Collagen and Gelatin PDF

THE STRUCTURE OF COLLAGEN AND GELATIN
WILLIAM F. HARRINGTON
McCollum-Pratt Institute, The Johns Hopkinr University, Baltimore, Maryland
and
PETER H. VON HIPPEL

Department of Biochemistry, Dartmouth Medical School, Hanover, New Hampshire
I. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
11. Synthetic Polypeptides Related t o Collagen. . . . . . . . . . . . . . . . . . . 6
A. Amino Acid Structures ...... ............. . . 7
B. Simple Peptides of Gly roline, oxyproline.. . . . . . ‘3
C. Physical Chemical Properties of Synthetic Polypeptides. . . . . . . . . 11
111. The Composition and Amino Acid Sequence of Collagen and Gelatin. . . . 30
A. Amino Acid Composition.. . . . . . . . . . . . . . . . . . . . . ............... 30
B. Amino Acid Sequence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
IV. The Structure of Collagen.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
A . Structural Studies in the Solid State. . . . . . . . . . . . . . . . . . . . . . . . . . . 42
B. Structural Studies in Solution.. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 64
C. The Collagen + Gelatin Transition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
D. The Use of Proteolytic Enzymes in Structural Studies of Collagen. . , . 83
E. The Role of Water in the Collagen Structure. .....
V. The Structure of Gelatin. ............................
A. The Molecular Properties of Gelatin a t Room Temperature and Above. 96
B. The Molecular Properties of Gelatin a t Low T
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............. 127
I. INTRODUCTION
The properties of collagen are of interest from many points of view.
Since it constitutes the major protein component of skin, bone, tendon,
and all the other forms of connective tissue, an understanding of collagen
seems to the clinician to be a necessary (though not sufficient) prerequisitc
to a rational attack on the many and diverse connective tissue disorders
currently lumped together as “collagen diseases.” The mechanisms of the
biosynthesis and incorporation into collagen of the unusual amino acids
hydroxylysine and hydroxyproline offer a continuing challenge to the
“pathway” biochemist. The leather chemist strives for a bettcr under-
standing of the interaction of collagen with tanning agents, for such undcr-
standing lies directly on the road to improved leather products. Thc
2 HAREINGTON AND VON HIPPEL
comparative biologist is intrigued by the direct correlation between the

temperature a t which the collagen of a given organism undergoes thermal
“denaturation” and the temperature of the normal habitat of that organism.
And finally, collagen interests the physical biochemist, not only because of
its intrinsic importance, but also because the properties of collagen on the
molecular level truly belong in a class set apart from those of all other
protein molecules.
Clearly, it is impossible to consider, even briefly, all these facets of the
collagen problem. So while recognizing both the importance and the
intrinsic interest of other areas, limits of space, patience, and competence
all have served to confine this review primarily to the last of the points of
view mentioned above. And even this area-the molecular properties of
collagen-is so large and developing so rapidly that no pretence to com-
pleteness of coverage can be made. Thus it will be apparent that this review
represents an attempt to summarize the present position as seen by the
authors, for although we have tried to take a balanced view throughout,
some aspects of certain of the problems discussed are controversial and
might have been handled quite differently by others in terms of emphasis,
coverage, and perhaps interpretation. However, despite these difficulties,
we hope that we have succeeded in assembling a reasonably complete and
coherent account of the current situation.
Since “The Structure of Collagen Fibrils” was last reviewed in these
pages by Bear (1952), tremendous progress has been made in the field of
protein chemistry and structure analysis. And this progress has naturally
been accompanied by comparable progress in our understanding of the
collagen molecule. The magnitude of this development in the field of
collagen might best be indicated by summarizing some of the major dr-
velopments of the decade which has passed since Rear wrote his review in
the spring of 1951.
1. It had been clear for many years prior to 1951 that collagen fibers from
all organisms have in common a distinctive wide-angle X-ray diffraction
pattern, and thus presumably also a unique polypeptide chain configura-
tion. However the nature of this configuration was not known. Since that
time advances in our understanding of the stereochemistry of polypeptide
chains and the theory of fiber X-ray diffraction first made possible, and
then led to, a generally accepted structural interpretation of the wide-
angle X-ray pattern of collagen.
2. The unusual amino acid composition of collagen had also been
recognized for some time. One-third of the residues of all collagens seemed
to be glycine, while about one-fourth were proline and hydroxyproline.
However, the stereochemical consequences of the presence of these residues
has only become clear as a result of detailed studies of synthetic homo- and
STRUCTURE OF COLLAGEN AND GELATIN 3
copolymers of glycine and proline. Consideration of such synthetic poly-

peptides as simplified models of certain features of collagen and gelatin has
been extremely helpful in recent years, and constitutes the rationale for the
inclusion of a section dealing specifically with these synthetic polypeptides
in this review.
3. Physicochemical studies of soluble collagen, which had just been
initiated in 1951, have been pursued vigorously during the last 10 years
and have added immeasurably to our understanding of the collagen system.
Such studies, in turn, have made it possible to interpret the long-range
periodicities which had previously been observed in collagen fibers using
electron microscopy and small-angle X-ray diffraction, and have led to a
partial understanding of the molecular basis of the patterns observed in the
various “reconstituted” collagens.
4. The collagen gelatin transformation in solution has been recognized
as a reversible first-order phase transition, subject to the same physical
laws which govern the crystalline S amorphous phase transitions observed
in systems of linear polymers. The direct relationskip between the transition
in solution and the well-known thermal shrinkage phenomenon exhibited
by collagen fibers has also been established.
5 . It has become clear that gelatin, long considered to be a degradation
product of uncertain composition and properties, can be prepared from
collagen under conditions which result in the production of perfectly
reproducible protein systems. Thus gelatin itself, and the mechanism of the
process by which it undergoes spontaneous reversion to a collagen-like
structure on cooling, have been the object of considerable study over the
last few years.
Specifically, it is the recent developments in these areas which we propose
to discuss in this review, For detailed considerations of earlier work and for
coverage of certain peripheral topics the reader is referred to the reviews
by Bear (1952) and Kendrew (1954), to the informative monographs by
K. H. Gustavson entitled “The Chemistry and Reactivity of Collagen,”
and “The Chemistry of Tanning Processes” (Academic Press, 1956);and to
numerous recent symposia dealing with various aspects of the collagen
problem, including: “Nature and Structure of Collagen” (J. T. Randall,
ed., Academic Press, 1953a); “Fibrous Proteins and Their Biological
Significance” (SOC.Ezptl. Bid. 9, 127 (1955) ; “Connective Tissue” (R. E.
Tunbridge, ed., Blackwell, 1957); “Recent Advances in Gelatin and Glue
Research” (G. Stainsby, ed., Pergamon Press, 1958); “Calcification in
Biological Systems” (R. F. Sognnaes, ed., Am. Assoc. Advancement Sci.,
1960); and “Central Leather Research Institute Symposium on Collagen”
(N. Ramanathan, ed., Interscience Publishers, 1961).
Before plunging into the detailed discussion, it is important to consider
4 HARRINGTON AND VON HIPPEL
the problem of the identification of collagen and its distribution through

the “animal kingdom,” and simultaneously to establish an operational
definition of this class of proteins. These problems have been treated at
great 1engt)h by Bear (1952), and since recent work has only served to
subshntiate most of his conclusions we will confine ourselves to a brief
summary of current views.
Collagen fibers may often be recognized histologically on the basis of one
or more of the following bulk characteristics: they tend to swell markedly
when immersed in acid, alkali, or concentrated solutions of certain neutral
salts and nonelectrolytes; they are generally relatively inelastic,; they are
more resistant than most protein fibers to degradation by proteolytic en-
zymes, but (in contrast to all other proteins studied) are readily attacked
by the enzyme collagenase; they undergo thermal shrinkage to a fraction of
their original length a t a temperature which is characteristic of the collagen
from a given animal (but varies from one species to another) ; and they are
converted in large part to soluble gelatin by prolonged treatment at
temperatures above the thermal shrinkage level.
I n practice, however, these criteria are not always easy to apply, nor
are they infallible. Thus in recent years the presence of the characteristic
collagen wide-angle X-ray diffraction pattern has come to be accepted a s the
fundamental defining criterion for collagen. This pattern, (which will be
discussed and interpreted in detail in Section IV) is easily recognized by
the strong 2.86 A meridional spacing and by the -11 A hydration-sensitive
reflection on the equator (see Fig. 8). It has served to demonstrate the
presence of collagen in the tissues of almost all multicellular animals which
have been investigated, ranging from the primitive porifera and co-
elenterates, through the annelids and echinoderms, and u p to the verte-
brates (see Marks et al., 1949; Bear, 1952; Gross et al., 1956).
The following physical and chemical features also seem to be generally
characteristic of collagen (see Gross et al., 1958; Watson and Silvester,
1059; Piez and Gross, 1959; Piez and Likins, 1960): (1) a content of glycyl
residues close to one-third of the total number present; (2) a high content
of pyrrolidine-ring-containing residues (relative to the composition of most
other proteins) ; (3) substantial numbers of hydroxyprolyl residues; (4) few
or no cystyl, methionyl, valyl, phenylalanyl, tyrosyl, or histidyl residues,
and thus minimal absorption a t 280 mp; (5) hydroxylysyl residues; (6) a n
extensive meridional small-angle X-ray diffraction pattern; (7) fibrils
exhibiting periodically-banded (-640 A) structure in the electron micro-
scope; and (8) an infrared absorption band a t 3330 cm-’.
In the following sections these criteria will be considered in detail,
mostly in terms of certain vertebrate collagens which have been extensively
studied. However, the available evidence suggests that in main outline the
a
w
STRUCTURE OF COLLAGEN AND GELATIN
results obtained with these collagens apply equally well to most vertebrate
(and perhaps invertebrate) collagens. Very recently careful studies of
invertebrate collagens have been launched by several groups, and it has
become apparent that some of these collagens differ substantially from
vertebrate collagen in certain respects. For example, the -640 A macro-
period seen via electron microscopy or small-angle X-ray diffraction in
vertebrate fibers has not been detected in certain of the invertebrate ma-
terials examined. Also, while in vertebrate collagens the ratio of proline to
hydroxyproline is generally in the neighborhood of unity, in the inverte-
brates this ratio varies widely. Thus gelatin from earthworm (Lumbricus)
cuticle contains 13 prolyl and 165 hydroxyprolyl residues per 1000, while
gelatin from Ascuris cuticle contains 280 prolyl and only 24 hydroxyprolyl
residues per 1000 (Watson and Silvester, 1959). It is expected that careful
physicochemical studies of such invertebrate collagens will help to il-
luminate further the role of these residues in the collagen structure.
Finally, it must be mentioned that there exist certain classes of protciiis
which yield the collagen wide-angle X-ray diffraction pattern and thus
meet the minimum requirements for membership in the collagen group, but
which differ significantly from the better known collagens in other ways.
Despite previous controversy, all the following now seem securely estab-
lished as collagens.
(a) Reticulin: This substance is found closely associated with collagen in
the connective tissue. It is identified histologically through the presence of
branching networks (see Bear, 1952; Kramer and Little, 1953; Kendrew,
1954; Robb-Smith, 1957, 1958).
(b) Ichthyocol and elastoidin: These terms simply refer to collagens
prepared, respectively, from the swim bladders and skins of fish, and from
the fins of the shark. Certain of the former, particularly the collagen derived
from the swim bladder of the carp, have been extensively studied and are
discussed in detail in subsequent sections. Elastoidin has been examined
much less extensively (see particularly Bear, 1952; Gross and Dumsha,
1958).
(c) Vitrosin: A fibrous protein obtained from the vitreous humor of the
eye. It exhibits the collagen wide-angle X-ray pattern, -640 A periodicity
and a typical collagen amino acid composition (see Gross et al., 1955b).
(d) Spongin, gorgonin, and cornein : Fibrous proteins derived from
sponges, corals, and coelenterates (see especially Marks et al., 1949; Bear,
1952; Gross et al., 1956; Pie2 and Gross, 1959).
(e) The “secreted” collagens: These may be differentiated from other
collagens in being secreted by epithelial cells instead of being mesodermal
or mesogleal in origin. The secreted collagens show the typical wide-angle
X-ray pattern, but apparently no macro-period. Examples include : earth-
worm and Ascaris cuticle, the bivalve byssus threads, and the “ejected
filaments” of the sea cucumber (see Bear, 1952; Kendrew, 1954; Watson
and Silvester, 1959).
Despite earlier claims to the contrary, it now seems relatively well
established that elastin, the third major histological component of verte-
brate connective tissue (with collagen and reticulin) is not a member of the
collagen class of proteins (Bear, 1952; Kendrew, 1954).
11. SYNTHETIC POLYPEPTIDES RELATED TO COLLACSN
I n this section we propose to consider the effects of proline, hydroxy-

proline, and glycine residues on the molecular architecture of polypeptide
chains. The physical chemistry of synthetic polypeptides containing these
residues will be examined in detail because of the remarkable similarities
which exist between the helical structures generated from polymers of these
amino acids and the three-dimensional pattern of the polypeptide chains
of the collagen molecule. Moreover, the mechanism of the reversion of the
component chains of collagen to form the parent collagen molecule a t low
temperatures in vitro (and possibly in the connective tissue in vivo) is now
thought to be intimately related to the configurational changes exhibited
by these model substances in solution.
Pauling (1940), in considering the general problem of protein configura-
tion, first pointed out that the insertion of a proline residue into a poly-
peptide chain has marked effects on the chain direction. Indeed, he demon-
strated that proline residues, by virtue of their unusual stereochemistry ,
are ideally suited to serve as hinge points for chain folding in globular
proteins. More recently, it has also become apparent that the geometry of
the pyrrolidine ring prevents L-proline from being accommodated in a left-
handed a-helix. Furthermore, this residue can only fit into an undistorted
right-handed a-helix when it occurs as one of the first three residues a t the
N-terminal end of the chain. On the other hand, Lindley (1955) has showii
that a change in helical sense (from a left-handed to a right-handed a-
helix) can occur without distortion a t a proline residue. I n all cases the
direction of an a-helix is altered at a proline residue and may be completely
inverted if the atomic grouping about the imide linkage is disposed in the
cis-configuration (Edsall, 1954).
In addition to the stereochemical features imposed on a polypeptide
chain as a consequence of the insertion of a proline residue, the imino
nitrogen atom of the pyrrolidine ring is devoid of a proton and is conse-
quently unable to participate in the formation of inter- or intrachain
hydrogen bonds. This means that the regularity of any systematic,
hydrogen-bonded peptide structure will be interrupted in proline-contain-
ing regions. The effect of such internal breaks on the stability of the a-helix
has been considered by Schellman (1955) who demonstrated that usu-
ally a t least three hydrogen bonds would be ruptured at such a discon-

tinuity, although little configurational entropy should be acquired by the
chain since peptide units on either side of the break are maintained by
hydrogen bonds. On the other hand it would be expected that insertion of
an appreciable number of randomly distributed proline residues would
lead to destabilization of the helix. Some evidence which supports this view
comes from the work of Szent-Gyorgyi and Cohen (1957) who compared
the a-helical content of a number of the keratin-myosin-epidermin-fibrin-
ogen (KMEI') rlass proteins, as deduced from optical rotatory dispersion,
with their proline content. They found that proteins with less than 30
proline residues per 1000 were in general highly folded, and that as the
proline content increased above this level a polypeptide chain without
benefit of stabilizing cross-bridges seemed to approach a random-coil-type
configuration. At very high concentrations of proline a completely new type
of configurational pattern seemed to emerge. This configuration, termed
poly-L-proline 11, will be the subject of detailed consideration below.
Before embarking upon the discussion of the relatively complex archi-
tecture of the synthetic polymers and copolymers related to collagen and
gelatin, it is important to examine the structure of glycine, proline, hydroxy-
proline, and simple peptides composed of these residues. As indicated in the
introduction, the stereochemistry of these residues is crucial to all current
theories of collagen structure. Their importance may be judged, without
reference to a particular structure, when it is recognized that, together, they
comprise over 50 70of the amino acid content of most collagens.
A , A mino Acid Structures

1. Glycine
The crystal structure of glyoine has been redetermined recently by Marsh
(1 9.58) from a complete three-dimensional Fourier analysis involving 1867
reflections. In general, the bond lengths and bond angles deduced were
essentially identical to those reported earlier by Albrecht and Corey (1939)
except that the C-N bond distance is 1.474 A instead of 1.39 A. The
former distance is in good agreement with that normally observed in other
amino acids. Carbon atoms 1 and 2 and oxygen atoms 1 and 2 (C, , CZ,
0, , and 0 2 ) * are nearly coplanar (see Fig. 1) whereas the nitrogen atom is
0.44 A out of plane.
2. L-Praline
A complete X-ray diffraction study of this imino acid has been prevented
as a result of its strongly hygroscopic character. However, Mathieson and
* The subscript numbers with symbols here indicate position of atom in the mole-
cule.
Welsh (1952) have deduced structure parameters from a careful investiga-

tion of copper-DL-proline dihydrate. In Fig. 1, atoms CI and N, CZ, CS,
and Cg of the pyrrolidine ring are coplanar. C4 lies at a distance of about
0.60 A out of plane while the carboxyl group attached to atom Cz is disposed
on the opposite side of the plane with respect to C4 . Bond angles N-C6-C4
-
L Pro1i ne
L - Hydroxyproline
FIG.1. Interatomic bond lengths and bond angles for glycine (Marsh, 1958)
L-proline (Mathieson and Welsh, 1952), and L-hydroxyproline (Donahue and True-
blood, 1952).
and C2-C3-Ca have the unusually low values of 96" and 97", respectively.
These values may be in error since in L-hydroxyproline (Donahue and
Trueblood, 1952), in L-lcucyl-L-prolylglycine (Leung and Marsh, 1957),
and in poly-L-proline I1 (Cowan and McGavin, 1955; Sasisekharan, 1959a)
the corresponding angles vary between 103" and 108".
3. L-Hydroxyproline
The crystal structure of L-hydroxyproline was first reported by Zussman
(1951) and a complete three-dimensional Fourier analysis published a year
STRUCTURE O F COLLAGEN AND GELATIN 9
later by Donahue and Trueblood (1952). As in copper-DL-prolinedihydrate,

the five-membered ring is appreciably puckered. C 4 ,the carbon atom bear-
ing the hydroxyl group, is about 0.40 A from a plane defined b y the other
four atoms of the ring and lies on the opposite side of this plane from the
carboxyl group. The angle between the two planes C6-N-Cz-C3 and
ca-C4-c6 is 17". Neuberger's (1948) prediction that the carboxyl and
hydroxyl groups should be in the trans-configuration relative to the ring has
been completely confirmed by the structure of Donahue and Trueblood.
B. Simple Peptides of Glycine, Proline, and Hydroxyproline

1. Glycylglycine
Glycylglycine can exist in three different crystal forms, a,0, and y, which
were originally observed by Bernal (1931) growing side-by-side in the same
mother liquor. The dimensions of this simplest of all linear peptides were
first determined by Hughes and Moore (1949) from a two-dimensional
Patterson projection analysis of the @-form. I n this structure the heavy
atoms are all coplanar except for the terminal nitrogen which is 0.64 A out
of the molecular plane. More recently Hughes et al. (1954) have carried out
a three-dimensional analysis of a-glycylglycine involving more than
2000 X-ray reflections. Again the whole molecule is planar except for the
terminal nitrogen atom which lies 0.73 A below the plane. I n this structure
the peptide bond length is 1.32 A, which is closer to the value usually
observed for this bond (Corey and Pauling, 1953) than is the 1.29 A distance
reported for @-glycylglycine.Both forms of the structure are shown in Fig.
2.
2. L-Leucyl-L-prolylglycine
A study of the stereochemistry of this peptide is of great value in gaining
an understanding of the configurational effect of a proline residue on the
relative arrangement of neighboring amino acid residues. The crystal
structure was determined by Leung and Marsh (1957) from a thorough
three-dimensional diffraction analysis utilizing about 1700 reflections.
Atoms C4, C7, C, , Cs, Nz , and 0 4 of the leucyl-prolyl peptide group
are coplanar within 0.02 A, with bond lengths and angles close to those
reported for other peptides (Corey and Pauling, 1953) except for the bond
angle N2-Cg-4~ , which is about 5" larger than expected (see Fig. 3). Di-
mensions of the pyrrolidine ring are closely similar to those observed for
L-hydroxy-proline (Donahue and Trueblood, 1952) with atoms C 4 ,Cg , C7 ,
and N Pcoplanar, and Caout of plane by 0.37 A. The dihedral angle between
the plane of the leucyl-prolyl amide grouping and that of the proline ring is
approximately 7".
Several noteworthy features emerge from an examination of the structure
which are of special interest to the present review. Most prominent among
these is the finding that the
0
//
C-N
Bond ( i )
bond length is close to 1.34 A. This value is normal for a peptide bond,
demonstrating that the imide linkage between the carbonyl group (C,) and
the imino nitrogen (Nz) of the pyrrolidine ring possesses virtually the same
degree of double bond character as do ot,her peptide linkages. Moreover thc
a - Glycylglycine
,9-Glycylglycine
FIG.2. Interatomic bond lengths and bond angles for wglyclglycine (Hughee and
Moore, 1949) and 8-glycylglycine (Hughes, et al., 1951).
angle CK--N~-C, of 113" is close to the value accepted by Corey and

Pauling (1953) for the C,-NH angle in a peptide group. Another prominent
feature is the disposition of the major groups. It appears that the leucyl-
prolyl-glycine residues are maximally extended in the crystal structure, with
the nine main-chain atoms of the proline and glycine residues (Ci-C*,
N1, and 01-03) essentially coplanar. Each C,-N bond is virtually cis to
the C--0 bond of the same residue. At the position of the proline residue
the chain undergoes a twist of about 120"from the fully extended configura-
tion, a twist required by the geometry of the pyrrolidine ring.
3. Tosyl-L-pro1 yl-L-h ydroxyproline Monohydrate

Only a partial three-dimensional analysis of this peptide has been re-
ported (Beecham et al., 1958) although the known occurrence of proline-
STItUCTURE OF COLLAGEN AND GELATIN 11
hydroxyproliiie sequence in the chains of collagen (see Section 111)endow it

with special significance. The molecule appears to be composed of four
approximately planar groups: p-methyl-thiophenyl, prolyl, peptide +
pyrrolidine, and carbonyl. These are disposed at right angles to each other.
Evidence of the flexibility of the pyrrolidine ring come's from the position of
the carbon atom opposite the C,-N bond of the proline residue. In this
structure, unlike that of proline (Mathieson and Welsh, 1952) and hy-
1.51
ti
12 I
L- Leucyl - L-Prolylglycine
FIG.3. Interatomic bond lengths and bond angles for ~-leueyl-~-prolylglyci~~e
(Leung and Marsh, 1957).
droxyproline (Donahue and Trueblood, 1952) this carbon atom appears to

be on the same side of the ring plane as the peptide C=O. Another example
of the apparent flexibility of the pyrrolidine ring will be given in the dis-
cussion of the poly-L-proline I1 structure.
C . Physical Chemical Properties of Synthetic Polypeptides
1. Polyglycine II
a. Studies on Polyglycine II in the Solid State. In 1934, Meyer and Go
reported an X-ray diffraction study of various polyglycine preparations, in
which it was shown that this substance can exist in two distinct and re-
producible isomeric forms. When polyglycine is cast into films from dichloro-
acetic or trifluoroacetic acid, X-ray diffraction patterns exhibit a strong 4.4
A spot and a somewhat more intense 3.45 A reflection (form I). However, if
the polymer is precipitated from concentrated aqueous lithium bromide or
calcium chloride solutions, X-ray diffraction powder diagrams show a
strong 4.15 A reflection (form 11).
0 5A
FIG.4a. Structureof polyglycine I1 (from Crick andRich, 1955). A projection down

the screw axis, showing seven chains. Hydrogen bonds, shown as dashed lines, run in
a number of directions linking neighboring chains.
X-ray evidence suggests that polyglycine I exists in the solid state as an

extended @-structure (Astbury, et al., 1948; Astbury, 1949; Corey and
Pauling, 1953) the two reflections of 3.45 A and 4.4 A corresponding,
respectively, to the side chain and backbone spacing generally observed for
this structural pattern (Bamford et aE., 1956). Additional supporting evi-
dence comes from infrared absorption and dichroism studies of a partially
oriented specimen of polyglycine I (Elliott and Malcolm, 1956) as well as a
strong 1.16 A X-ray meridional reflection (Bamford et al., 1955) which is
also found in @-poly-L-alanine.
Assignment of the polyglycine I1 structure proved to be more dificult.
Neither the very inteuse 4.15 A rcflectioii nor the infrared absorption bands
observed at 1648 cm-' (C = 0 stretch) and 1558 cm-I appeared to fit any
known systematic configuration such as the a-helix or the 0-structures.
Bamford et al. (1956) were thus led to postulate the existence of a new but
unknown folded structure for this form. Readers of Nature had little time
FIG.4b. Structure of polyglycine I1 (from Crick and Rich, 1955). A projection of

the structure with the screw axis vertical. The chain on the right is nearer the reader
than that on the left. The planar peptide groups are edge-on a t the bottom of the fig-
ure with hydrogen bonds from these groups virtually perpendicular to the plane of
the paper.
to ponder over this anomaly before a solution was offered by Crick and
Rich (1955).
In the polyglycine I1 structure proposed by Crick and Rich, all of the
polyglycine chains are parallel and are packed in an hexagonal array.
Each chain has a threefold screw axis and is hydrogen bonded t o each of
its six neighbors as shown in Fig. 4a, which is a projection down the screw
-
axis. The hydrogen bonds are linear with an 0 * N distance of 2.76 A, which
is within the range of values found for simple compounds.
Figure 4b presents a view normal to the screw axis where it will be seen
14 IlARRINGTON AND VON HIPPEL
that the planar peptide group is inclined at about 35", with its plane
perpendicular to that of the paper. Movement along any chain to the
peptide group of the next residue involves a rotation of 120" about, the
fiber axis and a displacement in the fiber direction of 3.1 A. There are thus
three residues per complete turn of the helix with a repeat distance of 9.3 A.
Left-handed or right-handed helices are equally probable, since the residues
of polyglycine are devoid of an asymmetric carbon atom. Moreover, for a
given set of left-handed or right-handed polyglycine screws in the form I1
configuration, it appears possible to remove a chain from the structure and
to replace it running in the opposite direction (Crick and Rich, 1955).
Coordinates for the polyglycine I1 structure are given in Table I.
TABLEI
Atomic Coordinates for the Polyglycine IIa and Poly-L-proline I10 Helices
(A) (A) 4)O

-
Atom
Glycine Proline Glycine Proline Glycine Proline
Or-C 0 0.00 1.27 1.25 0 0.00

C' 1.16 1.16 0.28 0.27 -10.5 20"30'
0 1.32 1.33 1.17 1.18 -111.5 112"30'
N 1.93 1.94 1.01 1.01 75 -75'30'
Or41 3.1 3.12 1.27 1.25 120 - 120"00'
B-C1 3.44 2.G5 - 105"30'
r-C1 2.78 3.19 -81"OO'
S-C1 1.70 2.47 -64"30'
~~ ~~ ~
4 Crick and Rich (1955). Coordinates are for right-handed polyglycine I1 helix.
b Sasisekharan (1959a).
b. Studies on Polyglycine II in Solution. I n solution, polyglycine exhibits

rather interesting form I $ form I1 transitions, which were first observed
and reported by Meggy and Sikorski (1956). Whcii polyglycirie I is dis-
solved in saturated aqueous calcium chloride and precipitated by dilution
with water at 20"C, form I1 of the polymer results. This material appears
as thin hexagonal plates in the electron microscope. If the precipitation is
carried out a t 100°C, only form I is obtained and electron micrographs of
this material reveal flat parallelograms. At temperatures between 60" and
100°C precipitation gives a product with an X-ray diffraction pattern
intermediate between that of form I and form 11, while below 60°C only
form I1 is obtained.
2. Poly-L-proline
The successful synthesis of poly-L-proline by Rerger d al. (1954a) has
provided one of the most, interesting and important model compounds which
we have for an understanding of the chemistry of collagen and gelatin.

Earlier attempts to synthesize this substance were unsuccessful (Astbury
et al., 1948) possibly as a result of the difficulties involved in the preparation
of a highly purified N-carboxyproline anhydride, which is crucial to the
synthesis.
The early preparations of poly-L-proline (polymerized in bulk or in
dioxane as solvent) were of relatively low molecular weight with an average
degree of polymerization (DP) of 35 to 133 as determined by end-group
titration (Sela and Berger, 1955). These were quite water-soluble, in
contrast to the insolubility in water of poly-a-amino acids such as poly-
glycine, poly-L-alanine, poly-L-leucine, and poly-L-phenylalanine. With
respect to its solubility properties, low molecular weight poly-L-proline
resembles polysarcosine, in which the amide linkages are also devoid of a
hydrogen atom (Wessely and Sigmund, 1926; Waley and Watson, 1949;
Hanby et al., 1950). More recently Blout and Fasman (1958) have
succeeded in obtaining very high molecular weight polyproline samples
(DP = 567 to 928) via polymerization in acetonitrile. These polymers
proved to be insoluble in water,
Soon after the synthesis of poly-L-proline, Kurtz et al. (1958a) observed
striking changes in the optical rotatory properties when aqueous or organic
acid solutions of the polymers were allowed to stand at room temperature.
Precipitation of the polymer from thc reaction medium with ether, followed
by dissolution in water or acetic acid, resulted in a dextrorotatory solution
with [a]z5= $40". On standing, the rotation slowly changed t o a highly
levorotatory form with [a]E5= -420" (this value could be increased to
[a]E5= -5540" by heating). The rate of mutarotation is strongly dependent
on the solvent. In formic acid, mutarotation of form I ([a]E5= +40°) -+
form I1 ([a]E5= -540") is complete in less than 1 hr at 25"C, whereas in
acetic acid or water a t this temperature rotatory changes are observed over
a period of several days.
Poly-L-proline I1 in the solid state was obtained by dissolution of form I
in hot glacial acetic acid or inm-cresol a t 100" C, followed by ether precipita-
tion after cooling. This material is readily soluble in water, acetic, or formic
acids, giving a specific rotation, [a]E5= -540°, which is invariant with
time. The high molecular weight samples of poly-L-proline synthesized by
Blout and Fasman (1958) are also quite water-soluble after transformation
to form 11.
a. Studies on Poly-L-proline in the Solid State. (i) X-ray difraction analy-
sis of poly-L-proline 11. In 1955, a structure for poly-L-proline I1 in the
solid state was proposed by Cowan and McGavin, on the basis of X-ray
diffraction patterns obtained from powders and oriented films. An intriguing
feature of the proposed structure is the essential identity of the backbone
configuration with that of polyglycine 11. Thus it seems likely that there
exist a t least three distinct classes of polypeptide chain configurations:
the a-helix (Pauling and Corey, 1951a), the @-structures (Pauling and
Corey, 1951b), and the recently “discovered” poly-L-proline I1 or poly-
glycine II-type helix.
In the poly-L-proline I1 structure of Cowan and McGavin, the peptide
grouping is planar and disposed in the trans-configuration. Movement along
the chain from one residue to the next involves a rotation of - 120” about
the fiber axis and a translation along the axis of 3.12 A. The axial repeat is
therefore 9.36 A (see Fig. 5). Unlike the polyglycine I1 helix, the poly-L-
proline I1 helix must be left-handed, since the screw sense is uniquely
determined by the absolute configuration of groups around the a-carbon
atom (Bijvoet et al., 1951). A right-handed helix of the required dimensions
cannot be constructed with L-amino acids. This property is fundamental
to the behavior and chemistry of the chains of collagen as will be evident in
subsequent discussion.
Sasisekharan (1959a) has recently published a more detailed X-ray
diffraction study of powders and films of poly-L-proline 11, including an
analysis of optical diffraction data. This elegant study confirms in all
essential details the structure proposed by Cowan and McGavin.
The bond lengths and bond angles of the pyrrolidine ring appear to be in
good agreement with those deduced for L-hydroxyproline by Donahue and
Trueblood (1952) and for L-leucyl-L-prolylglycine by Leung and Marsh
(1957) with one interesting exception : in the poly-L-proline I1 structure all
of the heavy atoms of the proline ring are virtually coplanar. Otherwise, too
much strain is introduced in packing the chains into the required unit cell
dimensions. The pyrrolidine rings of copper-DL-proline dihydrate, L-hy-
droxyproline, and L-leucyl-L-prolylglycine are all appreciably puckered,
with atom Cd lying 0.40-0.60 A out of the molecular plane. It will be
recalled that some flexibility in the proline ring has also been suggested for
the proposed tosyl-L-prolyl-L-hydroxyprolinemonohydrate structure
(Beecham et al., 1958).
(ii) X-ray diffraction analysis of poly-L-proline I . The spatial geometry
of poly-L-proline I in the solid state has not been rigorously established up
to the present time, since no one has succeeded in obtaining oriented films
or fibers of this substance. Cowan and Burge (1958) have reported pre-
liminary X-ray and electron diffraction analyses of powder patterns which
are compatible with a right-handed helix consisting of cis-prolyl residues
with three residues per turn and a pitch of 6.30 A, Rich and Crick (1959)
have also attempted X-ray diffraction analyses of form I powders and have
suggested a similar right-handed helix of residues disposed in the cis-
configuration, with, however, three and one-eighth residues per turn and an
axial repeat of 5.85 .A.
(iii) Infrared spectra of poly-L-proline. Infrared spectra of poly-L-pro-

line I and poly-L-proline I1 have been reported by Berger et al. (1954b)
and Blout and Fasman (1958). Both polymers show several common ab-
FIG.5. Structure of poly-L-proline. The figure on the right is a wire model of a

left-handed helix having all peptide bonds in the trans-configuration. This is the
structure proposed by Cowan and McGavin (1955) for poly-L-proline I1 in the solid
state. Neighboring ring planes are essentially perpendicular t o each other. The
figure on the left is a right-handed helix with all peptide bonds in the cis-configura-
tion (see text).
sorption bands, notably at 2950 and 2860 cm-' (C-H stretching modes),
1650 cm-' (C = 0 stretch), and 1485 cm-' (possibly C-H vibrations of the
pyrrolidine ring). At frequencies lower than 1400 cn-', the absorption
spectra differ markedly. The spectrum of form I exhibits strong bands a t
18 RAILRINGTON A N D VON HIPPEL
980 and 1355 cm-’, whcreas thcse arc abscrit in form TI. One rather striking
feature is a w r y strong band at 3480 cm-I, which ocrurs in the spcctra of
hoth polymers (Rergcr et al., 1954a; Rlout and Ihsman, 1958). On thorough
drying of thr polymcr this hand disappcars and Rlout arid Icasman con-
c2ludt.d that it is duc to very strongly adsorhrd watcr. It is of intmcst in this
c.oiincc*tioiito tiotc that Rradhury ct al. (ln.58) havr obscrved a similar h i d
in cwllagcri at 3450 cm-l, arid have demonstrated that it arises cxclusively
from adsortwd water by following rhsngcs in infrared absorption as a
func~tionof humidity.
Polarized infrared spcctra of *form I arid form 11 reveal pcrperidicular
dichroism of thc (”=() stretcahing frcqucnry a t 1650 (m-’, suggesting that
thc varlioiiyl groups extcnd away from thc main chain axis (see Fig. 5).
From the dirhroic ratios of -1.7 (form 11) and ~ 1 . (form 4 I) it may tie in-
ferrrd that thc C=O groups of form I1 arc disposed more normal to the
t)ac*kbonerhain than arc thosc of form I. .4 similar dichroism of the C=O
band has l m i i rcportcd for oricntcd c~ollagenpreparations (Badger and
l’ullin, 1954; Sutherlaiid et al., 19.54) and oricmted c.old-rvaporated gclatin
filnis (Xmbroso and IClliott, 1‘351a).
0. Studics on I ’oly-i,-prolirLe in Solution. (i) The mutarotation of poly-L-
prolinp. Thc mcchariisni of the mutarotation of poly-L-prolinc in solution
has engagcd the attcntion of a numbrr of lahoratorics. 111thcir carly work
Kurtz ct al. (19.56) suggcsted that thc change in optiral rotation whirh
occurs during thc transition from form I form I1 might rcsult from a
--j
srries of cis -+ trans-trnrisformatioiis of thc prptide bond groupings in e w h

molcculf.. In latcr studics, Stcinherg (1958) and Stcinherg et al. (1958)
discovrrcd that poly-L-proline I1 cwuld he rcconverted to poly-L-proline I
on dilution of a11 acetic or formic acid solution of form I1 with n-propaiiol
or n-butanol. The revcrsihle nature of the form I ~ form I1 transition
seems to rule out a chcmical cahange, and suggests that the observed for-
ward and rrversc mutarotations rcflect configurational transit ioris along
iiidividual polymrr molerules. This view is supported by the finding that
both form 1 arid form I1 of thc polymcr yicld L-proline on hydrolysis.
Since poly-L-prolinr is cmmposrd elltirely of imino prptide linkages, the
rhain is drvoid of any systematic. set of peptidr hydrogen bonds and one
may question whethcr thc form I and form I1 structurcs repremit homo-
gericous, ordcrcd configurations in solution. In considcring the various
fac‘tors which cwuld act to stablize an ordrred configuration in solution, it
may lj(1 srcii from IGg. 5 that the Ca-K bond of thc backbone chain is a
cornponelit linkagc of t hc pyrrolidine ring and ronsrclucntly rotation ahout,
thib hotid is impossible. The
0
//
C-N
Bond (1)
imide bond has a length of 1.34 A both in L-leucyl-L-prolylglycine and in the

proposed structure for poly-L-proline I1 (Cowan and McGavin, 1955;
Sasisrkharan, 1959a). Thus in the solid state this linkage behaves as a
normal peptidr bond with an rnrrgy barrier to rotation rstimated a t about
21 kcal/molr (I'auling and Sherman, 1933). From an examination of space-
filling models it> is apparrnt that a drfinitr restriction to rotation also
exists a t t hr
Bond (ii)
bond of t,he peptide backbone. Assuming t)he prpt,ide grouping to be in the
trans-configuration, the neighboring pyrrolidine ring can assume two
rotational posit,ions:
(1) The carbonyl oxygrn is cis wit,h respect to thr hydrogen of the
C,-at,om. I n this spatial arrangement (which has been termed cis'), ro-
tation of the pyrrolidine ring about bond (ii) is limited to about 15".
(2) The carbonyl oxygen is on the opposite side of the peptide bond with
respect to the C,-hydrogen atom (trans'). This disposition allows the
pyrrolidine ring a freedom of oscillation of about 60" about bond (ii).
A similar set of restrictions is observed when the peptide grouping is cis.
These rest,rictions severely limit the spectrum of configurational patterns
available to a polymer of proline and, coupled with the fact that two
distinct struct,ures are found in solution, indicated that these restraints
play an important stabilizing role. However, it must not be assumed,
given the planarity of the peptide grouping, that steric restrictions alone
are sufficient for stabilization. Solvation phenomena appear to play a key
role, as will be demonstrated below.
When all of the peptide groups of poly-L-proline are in the trans-configura-
tion and the groups about the (ii)linkage are trans', the resulting structure
is the left-handed helix proposed by Cowan and McGavin for poly-L-
proline I1 in the solid state. A photograph of a wire model built according
to this arrangement is shown in Fig. 5. If all of the peptide bonds are
disposed in the cis-configuration and t,he (ii) linkages are trans', a right-
handed helix is generated (Fig. 5). In constructing an accurate model
of this helix it, was observed that the helix dimensions conformed very
closely to t,hose predicted by Cowan and Burge from the preliminary
X-ray diffraction analysis of poly-L-proline I, ie., the helix has three
residues per complete turn and a repeat distance of 6.30 A. It is also possible
to construct, a right-handed helix with three and one-eighth residues per
turn and a repeat of 5.85 A, as suggested by Crick and Rich, but in this case
examination of wire models reveal that t8hepeptide bond grouping must
be distorted significantly from coplanarity.
20 HAILRIKGTON .4ND VON HIPI'EL
Harrington m d Scla (1958) con(-luded from optical rotatory, sedinicnt a-

tion, and viscosity studies that in thc form I 4 form I1 transition, a
right-handcd hclix with cis-imide litikagcs was traiisformrd to a lcft-handed
hrlix with trans-imide honds. The most pcrsuasivc argumcrit for this
proposal conics from thc optical rotatory data which may bc uscd to
calrulatc thc cotifigurational contrihution to thc spcrific rotation of mch
hclix and thus to obtain an indication of helical seiis". Howevrr, hcfore such
a cdculation c w i be made it is necebsary to estimatc thc residuc rotation of
L-proline in a polypcptidc chain. This has been awomplishcd in thrw
indcpcndciit ways (Harritigton and Scla, 1958; Striiiberg ct al., l!)(iOa).
(1) T h c spwific*rotation of a serirs of glyc.inc-proline copolymers with
incwasiug Gly/I'ro ratioh was mcasured. T h e specific rotation, corrcctcd
for the wcight-fraction of prolinc, approachcd [a]i5= -2.50" at high
Gly/I'ro ratios.
( 2 ) In aqurous solutions chontaining high conccntratiotis of JliBr or
CaC12 , thc specific. rotation of poly-r,-proline I1 approachcs [a]'f = -2240".
'ITndcr thcsc (widitions the iiitririsic viscosity of the highly asymmctric
form I1 strurturc dccreascs markcdly to valurs in the range of the globular
protrins and it can hc assumcd that the configurational cwntribution to
rotation of form I1 has becn climinated.
( 3 ) The rcsidur rotation of L-prolinr ('an also bc estimated from thr
reported optical rotations of relativcly simple pcptidc dcrivativcs of proline
and glyciiic.
These mcthods yicld an averagr [a]i5 = -250" for the L-prolinc residur.
Icrorn this valiic arid the spcrific rotations of form I ([a]:: = +10") arid
form I1 ([a]i5= --51O0), the contribution of the right-handcd hclix to
rotation is closc t o +300" whcrcas that of the left-handcd hclix is about
- 290".
(ii) The mechanism of mutarotation. Dircrt evidence for thc mrchanism
responsible for mutarotation comcs from kinetic studics. Stcinbcrg rt al.
(196Oa) ohscrved that the rates of both thc forward arid revws" mutarota-
tiori rcactions arr markrdly ac*ccleratcd by small amourits of strong arid.
Thus 0.052 molc of ITC104 per mole pcptide bond rrduces thc half-life of
mutarotation of form I -+ form I1 in acrtic acid from 5-15 to -1 min. Similarly
addition of tracc amounts of H@1(I4(0.15 M ) accclcratcs the rate of revcrw
mutarotation (form 11 -+ form I) in water-propanol (1:9 v/v) from a half-
time of 444 min to 60 min. The small amount of acid required in these
experiments suggests that the mechanism involves proton binding at the
imide linkages of thr polymer. Other cvidencc supports this virw: when
poly-L-proline is prcripitated from arctic arid solutions on addition of
anhydrous pcrchloric acid, thc resulting prccipitatc contains, after thorough
washing arid drying, about 0.30 molc tI(>l04 per molc peptide bond.
Potentiometric titration of simple proline derivatives and of poly-0-acetyl-

hydroxyproline in acetic anhydride also demonstrate protonation of the
imide linkages with about 0.35 mole proton bound per mole imide bond
(Steinberg et al., 1960a).
Protonation of simple amides has been found to markedly decrease the
double bond character of the C-N amide linkage. Thus Berger, et al. (1959)
found that protonation of the C-N bond in N-methylacetamide and
N , N'-dimethylacetamide resulted in a depression of the double bond
character as measured by nuclear magnetic resonance. In neutral aqueous
solution the nuclear magnetic resonance spectrum shows the presence of
two N-methyl lines indicating the absence of free rotation about the C-N
linkage. In the presence of acid the doublet is replaced by a single absorp-
tion band with the onset of free rotation.
Protonation appears to give an equilibrium between the three species:
(1) (11) (111)

of which free rotation is possible about the amide linkage in species (111).
It is probable that a similar mechanism applies in polymers of proline,
+
allowing cis trans-isomerizations at the peptide linkages.
In the absence of steric effects the activation energy of the form I -+
form I1 transition should approximate the energy barrier to rotation about
the peptide linkage (21 kcal/mole) provided that mutarotation involves
the proposed cis + trans-isomerization. Downie and Randall (1959)
measured the rates of forward mutarotation of poly-L-proline I in acetic
acid at various temperatures and obtained an activation energy, AE* =
23 kcal/mole. The rate of the reaction was independent of concentration
over a sevenfold dilution of the polymer. That is, a t any stage of mutarota-
tion (as measured by [a]:) the velocity constant, lc, was found to be in-
dependent of concentration. On the other hand k decreased from 15 X
sec-l to 2.5 X loF6sec -1 during the course of mutarotation.
The kinetics of the optical rotatory changes of poly-L-proline in various
solvents and at various temperatures have also been studied by Steinberg
et al. (1960a). Inacetic acid the course of the forward mutarotation reaction
was found to be independent of concentration (over the range 0.25 to 2.0
gm/100 ml) but, as observed by Downie and Randall, the rate constant
depends on the degree of mutarotation. An activation enthalpy, AH* = 21
kcal/mole, was determined for both the forward mutarotation of form I in
acetic acid and the reverse mutarotation of form I1 in acetic acid-n-pro-
panol.
The type of mutarotation kinetics found experimentally is strongly

dependent on the solvent. Forward mutarotation in acetic acid yields a plot
of log d[a]/dt versus log ( [ a ] t- [a],J which is linear over 97% of the
reaction with a slope of 1.33. On the other hand the rate is virtually in-
variant over two-thirds of the reaction in a solvent of 30% water-70%
acetic acid. Reverse mutarotation in acetic acid-n-propanol appears to
follow first-order kinetics over 90 % of the transformation. From these
experiments and the fact that the enthalpies of form I and form I1 are
essentially identical (Steinberg et al., 1960a) it seems likely that solvation is
decisive not only in determining the kinetic pathway of mutarotation but’
also the structural pattern which is stabilized in a given solution.
If we assume that mutarotation involves the rotation of randomly spaced
proline residues about the backbone linkages, it can be seen that the rate
of this isomerization need not remain constant throughout the reaction.
Because of the geometry of the pyrrolidine ring, a cis 4 trans-isomerization
at the peptide linkage virtually inverts the chain direction, leading to
proximal interactions between neighboring rings and to steric interference
to rotation of these vicinal residues. The average environment of any
particular isomerizing residue may therefore depend on the degree of
mutarotation. Steinberg et al. (1960a) have developed a kinetic analysis of
this process assuming that each individual L-proline residue associated with
a cis- or trans-peptide bond contributes to the observed specific rotation
[a],according to the specific rotations of poly-L-proline I, [elI,and that of
poly-L-proline 11, [ a ] I IThus
.
where C c i aand Ctransare the concentrations of cis- and trans-peptide bonds,

+
respectively, while Co = C c i s Ctrona,is the concentration of all peptide
bonds. It follows that
-c e_i s - [a1 - [a111

(2)
co [a11 - b 1 I I
As pointed out above, the spontaneous cis trans-transformation requires
proportionality between dC,i,/dt and Ccisa t any time. Hence,
but K is not a true reaction constant since we have seen that it varies
throughout the course of mutarotation. The “constant” K is therefore a
function of [a]or of the degree of conversion. The over-all mutarotation
reaction may thus be represented by the equations
---
dCcis - KCcis, where K = f (4)
at
or by
--
dCci8 -
- K’Ccis, where K’ = dt) (5)
at
The function K’ = &t) may be determined if the conversion can be
expressed in terms of a reaction of a given order, p. In this case
where k(Cci,/Co)fl-lstands for K’(t) and k is independent of time and defines

the initial rate of mutarotation. Integration within the time limits t = 0
to t (corresponding to Ccis = Co to CciJ yields
= 1
+
(%>”-I
(7)
1 k(P - l>t
or for the particular case under discussion
1
K’(t) =
1 + k ( P - 1)t
We have seen that the form I -+ form I1 interconversion may take place
along different pathways depending on the solvent system, with differing
over-all mutarotation kinetics. In acetic acid the apparent order of the
forward mutarotation reaction is 1.33 (p). It will be seen from Eq. (8) that
in this case K‘(t) should decrease with time, in agreement with the experi-
mental observations. When p= 1, the mutarotation proceeds with first-
order kinetics. This behavior is seen in the reverse mutarotation of poly-L-
proline I1 in acetic acid-propanol. When p < 1, K’(t) increases with time.
This situation obtains in the initial stage of the forward mutarotation in
acetic acid-water solvent which proceeds with zero-order kinetics.
(iii) Hydrodynamic properties. The mutarotation of poly-L-proline I
in propionic acid proceeds extremely slowly (A[a]z5 = -17” per 24 hr).
Thus the molecular weight of a given polymer sample can be determined as
form I and compared with that of the same sample after conversion to
form 11. Combined sedimentation and viscosity studies on one sample of
form I in propionic acid gave an estimated molecular weight of 19,000
After mutarotation to form I1 in acetic acid, the weight-average molecular
weight was 18,400, while end-group analysis gave a number-average
moleculrtr weight of 19,000. It thus appears that association or dissociation
processes are not occurring during mutarotation in the simple aliphatic
acids, and that the fundamental unit particle in these solutions consists of a
single polymer chain.
The reduced viscosity (qsp/c) of form I1 is always greater than that of
form I in a given solvent, consistent with the left-handed form I1 helix
having a more asymmetric structure in solution than that of the form I,
right-handed helix (see Fig. 5 ) . During mutarotation of form I 3 form I1
in acetic acid, the viscosity increases monotonically (Blout and Fasman,
1958; Downie and Randall, 1959; Steinberg et al., 1960a; Steinberg et al.,
1960b). However Harrington and Sela (1958) observed marked reduction in
I I I I I I I
1.2 - -
-
6
1.0
\
g 0.8
v
2 0.6
0.4
0.2
I I I A ! I I I
0
0 100 200 300 400 500 600
-[a]~"'
FIQ.6. Reduced viscosity (C = 1%) of poly-L-proline at 30°C as a function of
[a]:. 0-0, datatakenduring forward mutarotationinglacial acetic acid; A-A, data
taken during forward mutarotation in acetic acid-water (7:3 v/v); 0-0,data taken
during reverse mutarotation in acetic acid-n-propanol (1:9 v/v); A, value in
aqueous 12 M LiBr. (From Steinberg et al., 1960b. Reproduced with kind permission
of the American Chemical Society.)
viscosity during the early phase of the mutarotation in water as solvent.

In these experiments the viscosity passed through a minimum, increasing
during the later stages of the reaction. Steinberg et al. (1960a, b) reported a
similar depression in the viscosity versus time plots in studies of the form I
+ form I1 interconversion in a solvent of 30 % water-70 % acetic acid. It is
apparent that the order of the reaction [4/3in acetic acid, zero (initially)
in water and in water-acetic acid] is related to the pathway of intercon-
version which is reflected in the hydrodynamic properties. Figure 6 il-
lustrates this situation graphically. In various solvent systems the poly-L-
proline molecules may assume different average configurations as measured
by viscosity during mutarotation but still exhibit the same optical rotation.
I t may be questioned whether the restrictions to rotation about the
backbone discussed above are sufficient to lead to rodlike particles for the
form I and form I1 structures in solution. Fair agreement is found between
the axial ratio estimated from viscosity studies (assuming a rigid prolate
ellipsoid) and that expected from the coordinates of the left-handed
Cowan-McGavin helix for low molecular weight polymers (Harrington and
Sela, 1958). At molecular weights of 12,000 and above the axial ratio found
is consistently lower than that expected, this deviation increasing with
increasing molecular weight (Steinberg et aZ., 1960a).
A similar examination of form I polymers for two different molecular
weight samples (12,000 and 19,000) gave axial ratios of 29 and 37. As-
suming that the form I chain in solution is a right-handed helix with all
peptide bonds in the cis-configuration, the axial ratios calculated using the
Rich and Crick (1958) dimensions are 27 and 29, respectively, whereas 31
and 45 were obtained using the Cowan and Burge (1958) dimensions (see
Fig. 5). Flexibility in the form I chain becomes apparent at higher molecular
weight. A sample of molecular weight 52,000 (Blout and Fasman, 19.58)
gave an axial ratio from viscosity measurements of 44 whereas the expectled
value is 116 (Rich and Crick) or 137 (Cowan and Burge).
(iiii) The efect of neutral salts on the poly-L-proline 11 configuration.
As mentioned earlier, the optical rotation and viscosity of poly-L-proline I1
are strikingly altered in the presence of certain neutral salts. The specific
rotation of poly-L-proline I (DP = 200) approaches [ag5 = -240' when
form I1 polymer is transferred from water to a concentrated aqueous
lithium bromide solution, while the intrinsic viscosity falls from [ q ] = 0.54
dl/gm to [7] = 0.07 dl/gm. Similarly, in the presence of 4 M sodium
thiocyanate the specific rotation of poly-L-proline I1 (DP = 550) is de-
pressed to [a]E5= -290" while [q] decreased from 0.67 (in water) to 0.07 in
this salt medium.
The most likely explanation of these changes is that they result from
destruction of an asymmetric, homogeneous structure (Harrington and
Sela, 1958). In this connection it is of interest that a sample of PO~Y-DL-
proline (DP = 100) exhibited an intrinsic viscosity in water of only 0.05
dl/gm, indicating a lack of configurational asymmetry. This value remained
unchanged in the presence of lithium bromide (Steinberg el al., 1960a).
Although destruction of the structural integrity of the poly-L-proline I1
helix in the presence of lithium bromide, calcium chloride (Harrington and
Sela, 1958), and sodium thiocyanate (Blout and Fasman, 1958) must
involve rotation of the pyrrolidine rings about the backbone chain, present
evidence indicates that loss of rigidity occurs at the
0
//
ce-c
Bond (ii)
linkages rather than a t the peptide bonds. For one thing, the N-methyl
doublet observed in the nuclear magnetic resonance spectrum of N , N’-
dimethylacetamide remains unchanged when this substance is transferred
from water to concent)rated aqueous solutions of lithium bromide or sodium
thiocyanate (Steinberg et al., 1960a). Moreover, on massive dilution of a
concentrated aqueous lithium bromide solution of poly-L-proline I1 with
water a t low temperature (3°C-1 l°C), the resulting solution undergoes
an extremely rapid mutarotation, with a rate about lo3 faster than that
expected for the “normal,” acid-catalyzed reaction. The activation enthalpy
of this reaction (AH* = 21 kcal/mole) is about the same a s that calculated
for the acid catalyzed mutarotation, but the entropy of activation is
markedly different (AS* = 0.84 e.u. for the dilut,ion reaction and -12.5
e.u. for the acid-catalyzed reaction) demonstrating that fundamentally
different mechanisms are involved in the two processes. It seems probable,
therefore, that configurational changes in the three-dimensional architecture
induced by neutral salts result from rotations of the rings about the (ii)
linkages. Since the helical pattern of form I1 is eliminated in the presence
of neutral salts, steric restrictions between contiguous groups of the chain
cannot be the primary source of stabilization a t bonds (ii). It would appear
more likely that the role of the neutral salts is to modify the helix-solvent
interaction.
3. Poly-L-hydroxyproline
Poly-L-hydroxyproline was first synthesized (Katchalski et al., 19,iCi;
Kurtz et al., 1958a,b) through polymerization of O-acetyl-N-carboxyhy-
droxy-L-proline anhydride in pyridine followed by deacetylation in aqueous
ammonia. The polymer is quite water-soluble, insoluble in the simple
aliphatic acids, but does not exhibit the mutarotation properties of poly-L-
proline I, probably because of the prolonged treatment in alkaline solution
required for deacetylation. Aqueous solutions of poly-L-hydroxyproline
exhibit an = -400” which is invariant with time. The magnitude of the
specific levorotation suggests that the backbone chain of poly-L-hydroxy-
proline is arranged in a structural pattern in solution similar to that of
poly-L-proline 11. This proposal is supported by the optical rotatory be-
havior of the polymer in aqueous lithium bromide. Addition of the salt
lowers the specific rotation to [a]i6 = -168”, paralleling the drop in
levorotation observed in poly-L-proline I1 under these conditions. The
optical rotatory dispersion constant, A,, falls from 206 to 191 mp, which is
also comparable to that observed for poly-L-proline I1 (A, = 202 mp in
water; X, = 185 mp in 8 M LiBr).
Sasisekharan (195Yb) has recently proposed a structure for poly-L-hy-
droxyproline in the solid state on the basis of X-ray diffraction patterns

obtained from powders and oriented films. The polymer chains appear t o
be packed in a hexagonal lattice. Each polypeptide chain takes up essen-
tially the same left-handed configurational pattern as that of poly-L-pro-
line 11, but the fundamental unit cell requires three helices as in collagen.
Although mutarotation of poly-L-hydroxyproline has not been observed
because of the conditions used in the deacetylation step, the acetylated
derivative, poly-0-acetyl-L-hydroxyproline, undergoes marked optical
rotatory changes with time in solution (Kurtz et al., 1958a, b). Imme-
diately after dissolution in formic acid, this polymer shows a specific
rotation, [a]:5 = +25" (form I). On standing 6 hr a t room temperature the
solution becomes strongly levorotatory with [a]:5 = -175" (form 11).
Reversal of the mutarotation (form I1 -+ form I) could be effected by boil-
ing in N,N'-dimethylformamide, by diluting a solution of form I1 in
formic acid with 40 volumes of pyridine or by dissolution in acetic anhy-
dride (Steinberg et al., 1960a).
Addition of trace amounts of a strong acid (HClOh) accelerates both the
forward and reverse mutarotation of poly-O-acetyl-L-hydroxyproline
indicating, as in the case of poly-L-proline, that isomerization is taking
place at the peptide linkages. If relatively large amounts of perchloric acid
are used, the specific rotation of the polymer immediately changes to
values intermediatebetween those of form I and form 11, being uniquely
determined by the ratio of acid/peptide bond. Since the same final specific
rotation is achieved on addition of a fixed amount of perchloric acid to
form I or form 11, it appears that the various levels of represent true
equilibrium states and that these reflect an equilibrium between the cis-
and trans-peptide group configurations in each molecule, which are in turn
determined by the extent of protonation. From titration studies it appears
that the polymer exists in the form I1 configuration after protonation of
about one-third of the peptide linkages. This situation is similar to that
found on titration of poly-L-proline.
Downie and Randall (1961) have recently measured the mutarotation of
poly-0-acetyl-L-hydroxyprolinein formic acid and the reverse mutarota-
tion in N ,N'-dimethylformamide at various temperatures and polymer
concentrations. The activation energy of forward mutarotation was 22.4
kcal/mole and of reverse mutarotation, 23.8 kcal/mole, consistent with a
trans- cis-isomerization mechanism involving rotation about the peptide
bonds.
Mutarotation phenomena are also exhibited by another derivative of
poly -L-hydroxyproline, i.e., poly-0-p-tolylsulphonylhydroxyl- L-proline
(Kurtz et al., 1957, 1958a). On dissolution in acetic acid, this polymer gives
28 HARRINQTON AND VON HIPPEL
[a]:' = 0" (form I) which changes to a terminal value, [a]:' = -120", on

standing (form 11). Dilution of the acetic acid solvent with pyridine
(40:1 v/v) brings about a reversal in mutarotation.
4. Copolymers Containing L-PTO~~TM
Copolymers of L-proline and glycine have been prepared by polymerizing
mixtures of N-carboxy anhydrides of proline and glycine in various weight
ratios. Copolymers were obtained with molecular weights ranging from
950 to 2050 as judged by end-group titration, and in mole ratios proline:
glycine (PG,) from PGa.a to PGs (Kurtz et at?., 1958s; Steinberg et al.,
1960a). Only the copolymers with low glycine content (below PG2) were
TABLEI F
Optical Rotation of Copolymers of L-Proline and Ulycine
Number
Polymer Molar ratio average
of residues molecular Solvent blDc [dD.oorreotadd
weightb
Poly-L-prolineI1 - 14,300 Formic acid -640" -540"

PzG 1:0.66 950 Formic acid -316" -438"
PG 1:1.06 2000 Trifluoroacetic acid -271" -440"
PG2 1:2.2 1700 Trifluoroacetic acid -190" -435"
PG: 1:3.3 1400 Trifluoroacetic acid -136" -398"
PGs 1:4.9 1630 Trifluoroacetic acid -90" -348"
PGa 123.8 2060 Trifluoroacetic acid -41" -256"
0 From Kurtz et al. (1960).
b End group.
Corrected for index of refraction of solvent.
d Based on weight-fraction of proline.
found to be water-soluble. All of the copolymers were quite soluble in

trifluoroacetic acid and in concentrated aqueous lithium bromide solutions,
but only slightly soluble in glacial acetic acid.
Mutarotation of a copolymer with a molar ratio of 1:1 (PG) has been
observed very recently by Downie and Randall (1961). In glacial acetic
acid the initial specific rotation ([a]:') was -118" which changed to -257"
on standing at 25°C. Similar behavior was observed in aqueous solution.
The effect of glycine residues in diluting out the poly-L-proline II-type
configuration may be seen in Table 11, which presents the specific rotation
of a series of L-proline-glycine copolymers in trifluoroacetic acid. As the
glycine content increases, bhe helical contribution of the left-handed form
I1 structure is eliminated (last column) and a t the highest g1ycine:proline
mole ratios (PGs) the specific rotation approximates that of the proline
residue. The disappearance of the form I1 structure with increasing glycine

content is also apparent in Fig. 7,where the specific rotation of each of the
copolymers is measured in varying concentrations of lithium bromide. The
effect of this salt on specific rotation becomes progressively less as the
0 DP-50
OP.30
-0 -0-0 1:9
OO
1 4 8 12
L i B r CONCENTRATION ( M ) .
FIQ.7. Specific rotation of various L-proline-glycine copolymers as a function of

lithium bromide concentration. Insert, la]:’ versus LiBr concentration for two sam-
ples of poly-L-proline 11. (From Kurtz et al., 1960.)
number of contiguous proline residues are decreased, suggesting that

neighboring pr pro line residues are required to give the observed salt effect.
This conclusion is entirely consistent with the studies discussed above on
poly-L-proline 11.
These experiments lead us to consider another aspect of the configura-
tional properties of proline-containing polypeptide chains which can be
stated in the form of a question: How many contiguous proline residues
are required to achieve the complete helical contribution to optical rot,ation

characteristic of poly-cproline? To answer this question Yaron and
Herger (1961) have recently prepared a skeleton copolymer consisting of an
optically inactive poly-DL-lysine backbone bearing short polybenzyl-m-
aspartate side chains. The side chains were used as sites for attachment of
poly-L-proline segments of varying length. It was found that the specific
rotation of such copolymers undergoes a sharp change to values charac-
teristic of high molecular weight poly-L-proline at chain lengths between
4 and 6 proline residues. In this range an abrupt transition was also ob-
served in the kinetics of mutarotation of the copolymers. I n formic acid the
specific rotation (based on the weight fraction of proline) approached the
value expected for form I1 whereas examination before mutarotation
showed a similar transition to the specific rotation characteristic of form I.
Copolymers of L-proline and sarcosine have been prepared by Fasman
and Blout (1961). These were found to exist in two forms, analogous to
poly-L-proline. Form I, which is obtained directly from the copolymeriza-
tion mixture showed anomalous optical rotatory dispersion and relatively
low viscosity. On dissolving form I in 2-chloroethanol, form I1 is obtained
which exhibits normal optical rotatory dispersion and relatively high
viscosity. Fasman and Blout have suggested that the transition, form I -+
form 11, involves a conversion of the structure from cis- amide bonds to
trans-amide bonds.
111. THE COMPOSITION
AND AMINOACID SEQUENCE
OF
COLLAGEN AND GELATIN
A. Amino Acid Composition

I n recent years the use of ion-exchange chromatography has greatly
improved the accuracy and speed of amino acid analysis, and in the case of
collagen and gelatin has provided us with complete compositions from
nearly every class of vertebrates. A number of invertebrate collagens have
also been analyzed.
Chemically, collagen is characterized by an unusually high content of
glycine and the imino acids proline and hydroxyproline; the presence of
hydroxylysine; and notably small amounts of aromatic and sulfur-con-
taining amino acids. The unusual fine structure of the polypeptide chains of
collagen, mirrored in the wide-angle X-ray diffraction patterns, and the
characteristic gross structural features, which are recognized by small-
angle X-ray diffraction and electron optics, are both intimately related to
this unusual composition. Our purpose in the present section is to draw
attention to those aspects of the chemical studies which may help in throw-
ing light on the fine structure of the protein. It is clear that any detailed
proposal of the molecular architecture of collagen must be compatible with
the composition of a wide range of collagen species. Moreover, a n examina-

t,ion of the compositions should help to characterize the chemical features
which link members of the collagen class of proteins.
1. Vertebrate Collagens
Much of the recent work on the amino acid composition of this class has
been collected by Eastoe and Leach (1958). Table I11 summarizes data
from the studies of Neuman (1949), Tristram (1953), Eastoe (1955, 1957),
Leach (1957), and Piez and Gross (1960) and includes both collagens and
gelatins. From a careful comparison of the compositions of collagen and
their derived gelatins, Eastoe (1955) and Eastoe and Leach (1958) have
given convincing evidence that the amino acid composition of collagen is
faithfully reproduced in the derived gelatin. I n fact from the chemical
point of view, Eastoe and Leach (1958) consider the preparation of gelatin
to be a purification of the parent collagen.
An examination of Table I11 reveals that the amino acid compositions of
mammalian collagens are very similar over a wide spectrum of species.
Very close to one-third of the total residues are glycine, about one residue
in ten is hydroxyproline, and twelve in a hundred are proline. Tyrosine,
histidine, and the sulfur-containing amino acids are present at concentra-
tions of less than 1 %. Eastoe and Leach (1958) suggest that the tyrosine
associated with the parent collagen may be, in part, an impurity since the
concentration of this amino acid is reduced in the derived gelatins, and
further decreased when these are fractionated with ethanol. Although
hydroxylysine is also present a t very low concentration (3 to 12 residues/
1000) it cannot be removed by purification procedures and is now thought
to be, along with hydroxyproline, a requirement for admission of a protein
to the collagen class. These two hydroxyl residues seem to be unique to
collagen among animal proteins. It is of interest that the protein of the
nemocysts of Hydra and Physalia appears to contain as much as 20%
hydroxyproline (Lenhoff et al., 1957). Hydroxylysine has been detected in
the free state in embryonic muscle extracts (Gordon, 1948).
The fish collagens exhibit an appreciably wider range of composition
than the mammalian species, in keeping with the greater evolutionary
time scale which they span. For example, the Australian lungfish, which is
one of the four surviving species of the class Crossopterygii is thought to be
more closely related to land vertebrates than any living fish (Young, 1950).
It is therefore of considerable interest that the amino acid composition of
this species approaches that observed for the mammalian collagens.
Similarly, the cod (Actinopterygii) which is among the most recently
developed of the bony fish, exhibits the greatest divergence from the
typical mammalian composition (Eastoe, 1957).
TABLE 111
Amino A r i d Composition of Vertebrate Collagens and Gelatins0
I l l 1 w
r.G
C?T
ox - jwlm Carp- Cod- Pike-
Amino acid skin blad- skin skin skin
(C) der ( G ) ( C ) (GI
- -._pI I
(C)
_ _ ~ _ _ _ _
Alanine 99.6 12 109.7110.8 99.9106 110.7 110.5 114.6 114.0 125.0 98.0'128.0 119.0 118.9 126 120 107 114
Glycine 138 i20 314 326 327 327 324 326 331 324 315 301 311 333 337 325 317 345 328
Valine 27.1 20 21.2 21.9 25.1 22 25.4 20.6 19.8 15.4 20.2 21.9 21.3 21.9 18.0 18 19 19 18
Leucine
Isoleucine
Proline
39.9 25 27.9 23.7 25.4 25 26.0 24.8 23.8
- 11 12.3 9.6 12.7 10 11.1 11.0 10.9
122.3 38 118.8 130.4 120.0 117 126.4 128.2 129.5
20.1 25.7 28.8 25.2 23.9
11.4 11.7 14.0 12.2 19.4
127.9 119.4 109.7 126.0 113.4
17.7
11.4
102.2
21 25
10 12 11
23 20
116 124 102 129

9.2 E
Phenylalanine
Tyrosine
14.1 13 16.3 14.4 13.6 13 14.2 13.0 14.3
5.1 2.6 2.9 3.2 4 . 8 3.2 3.6 3.6 3.4
17.7 14.2 19.3 15.3 13.9
3.3 1.8 6.1 1.1 1.4
14.1
2.4
14 14 13
14
2.0 3.2 3.5 1.8
5
Serine 29.9 36 37.8 36.5 27.9 41 36.9 41.0 28.6 42.1 43.6 66.3 43.7 44.5 50.5 37 43 69 41 $
Threonine 17.9 18 19.7 17.1 20.6 20 18.5 24.0 19.1 22.0 17.9 26.4 26.1 25.8 29.2 29 27 25 25 4
Methionine 5.0 4.3 5.1 5.4 5.7 6.3 5.7 4.7 6.2 6.5 6.1 8.7 4.0 10.0 8.8 13 12 13 12
Ar gin i n e 46.0 50 49.0 48.2 49.9 49 49.0 50.1 44.8 49.5 49.9 49.2 51.0 50.3 52.4 53 53 51 45
Histidine 4.5 5.0 5.8 6.0 4.2 5.1 5.4 5.7 4.5 4.7 4.7 6.5 5.1 7.4 4.8 3.8 4.5 7.5 7.4
Lysine 28.6 27 26.2 26.2 35.5 29 21.6 25.9 19.0 25.3 27.6 29.1 24.2 24.3 21.8 26 27 25 22
Aspartic acid
Glutamic acid
44.0 45 49.8 46.8 49.2 47 48.4 46.3 48.1
71.7 72 75.8 72.0 76.4 74 72.3 69.6 74.1
45.5 48.0 54.9 48.6 42.6
72.8 62.4 77.9 78.9 65.8
47.5
70.5
47
71 74
47 52
75
54 $
Hydroxyproline 99.6 94 100.8 95.5102.4100 92.1 89.1 98.5 92.8102.0 77.5 73.1 78.5 82.0 81 73 53
81
70
g
P
Rydroxylysine 6.3 7.4 6.4 5.9 - 5.7 8.9 5.8 9.6 4.9 4.0 4.3 5.3 4.7 10.7 7.4 4.5 6.0 7.9
Amide groups 43.9 46 41.8 40.8 - 51 44.0 25.6 40.1 25.5 22.1 53.9 46.8 29.4 41.0 38 26 33 42
-__--__ ~ ~ _ _ _ _ _
Total N 18.6 - 18.3 18.3 17.9 - 17.9 18.6 17.8 18.3 16.2 18.0 18.2 18.2 18.5 - - - I
Mean residue 91.3 - 92.4 91.3 92.4 - 91.6 91.1 91.1 91.4 91.2 93.3 91.4 90.8 90.7
weight
Referencec (1) (6) (2) (2) (3) (7) (2) (2) (4) (4) (4) (4) (5) (5) (5)
a Residues of amino acids per lo00 total residues. * Abbreviations: C = collagen; G = gelatin 3 = extract.
~References:(1) Tristram (1953). (2) Eastoe (1955). (3) S e u m a n (1949). (4) Leach (1957). (5) Eastoe (1957).
[6) Piee and Gross (1960). (7) Gross and Pie2 (1960)
STRUCTURE OF COLLAGEN AND GELATIF; 33
In agreement with earlier work (Gustavson, 1955a), the total imino acid
content within the fish group is significantly lower than that of the mam-
malian collagens, while the hydroxyamino acids serine (Neuman, 1949)
and threonine (Beveridge and Lucas, 1944) and sometimes hydroxylysine
are enhanced, leaving the level of hydroxyl groups about the same in both
the fish and mammalian series. The methionine content of fish collagens is
increased over that of the mammalian collagens except in the lungfish,
while tyrosine and histidine remain at less than 1 % of the total amino acids
throughout the series. Although significant variations in composition are
found among the fish collagens, the glycine content remains essentially
invariant throughout all species a t close to one-third of the total amino
acid residues (Piez and Gross, 1960).
Since the early work of Grassmann and Schleich (1935) and Beek (1941),
it has been known that a large variety of sugars are associated with the
collagens and gelatins. In addition to the small amounts ( < 1 %) of glucose,
galactose, and mannose reported by Grassmann and Schleich, glucosamine
(Schneider, 1940), fucose (Glegg et al., 1953), ribose, arabinose, and
galactosamine (Gross et al., 1958) have been found. The amounts of
carbohydrate determined vary, depending on the species and the degree of
purification of the collagen, but generally add up to less than 2 % by weight.
Purification or gelatinization, followed by fractionation, usually reduces
the carbohydrate content significantly. On repeated precipitation of
soluble collagens, the glucosamine disappears and the hexose content drops
to about 0.7% (Wolf, 1956; Grassmann et al., 1957b). This concentration
can be further reduced through oxidation with sodium periodate, and after
a short oxidation step only 0.15-0.20% of the hexoses remains (Hormann
and Fries, 1958; Kuhn et al., 1959). Since this residue cannot be destroyed
by oxidation, Schneider (1940) and Grassmann el al., (1957a) have sug-
gested that it, may be chemically bound to collagen through 0-glycosidic
linkages.
2. Invertebrate Collagens
At the present time complete amino acid analyses have been reported on
nine of the invertebrate collagens: the cuticle of the segmented roundworm
Lumbricus (Singleton, 1957; Watson and Silvester, 1959) and nonsegmented
roundworm Ascaris lumbricoides (Watson, 1958), ejected Cuverian fila-
ments of Thyone (Watson and Silvester, 1959), two spongins from Sp.
graminea, the collagen of the float of Physalia, the body walls of Metridium
and Thyone (Piez and Gross, 1959), and body wall of the garden snail
Helix aspersa (Melnick, 1958; Williams, 1960). Table IV gives the quantita-
tive amino acid analyses reported for these species.
Possibly the most striking feature of Table IV is the contrast between
TABLE IV
Amino Acid Composition of Invertebrate Collagens and Gelatinso
e3
-
:'
!-P
Coelente- Mollusca
Echinoderm
Amino acid Body wall duverian
(H jwskdi) cw)
Cuticle (Ascaris) km) Float Body
(G)b (G) Body wall (G) Spongin A 'POngin
fibers (GI (GI (G)
Alanine 113 114 103 69 70 66 56 94 72.3

Glycine 306 308 324 286 311 307 315 323 321
Valine 30 21 17 13 34 26 29 24 21.5
Leucine 22 28 29 18 37 31 28 24 23.5
Iso1eu cine 13 13 15 14 23 22 24 17 12.1
Proline 109 81 13 280 63 83 78 73 104.1
Phenylalanine 8.9 11 5.2 7.3 12 11 9.3 10 9.9
Tyrosine 7.9 11 0 1.2 7.9 5.6 4.7 4.0 8.8
Serine 43 55 105 23 54 47 38 24 61.4
P
Threonine 35 51 52 18 39 33 43 27 27.7
Methionine 2.2 6.9 0 12 8.8 5.8 4.7 3.1 1.2 3
Arginine 54 46 21 42 57 54 47 43 50.9 4
0
Histidine 2.8 4.8 0 8.4 5.1 1.9 3.9 3.2 2.6 2
Lysine 7.5 12 15 37 27 27 9.0 24 8.3
Aspartic acid 62 70 56 76 80 83 92 97 66.8
Glutamic acid 110 77 81 101 94 104 95 86 99.1
Hydroxyproline 60 54 165 24 49 61 108 94 99.5
Hydroxylysine 11.o 4.7 0 0 25 30 12.0 24 8.2
Cystine 2.5 - - - 3.2 1.6 3.3 6.0 0.0
Amide groups 75 84 97 44 71 66 102 90 46
Glucosamine 2.4 - - - 4.0 2.5 10 1.6 .92

Galactosamine 0 - - - 0 3.2 7.7 0 2.5
0 R.asidues of amino acids per lo00 total residues. * Abbreviations: G = gelatin. c References: (1) Pies and Gross (1959).
(2) Watson and Silvester (1959). (3) Watson (1958). (4) Williams (1960).
STRUCTURE OF COLLAGEN A N D GELATIN 35
the remarkable invariance of the glycine content and the wide range in
composition seen for most of the other amino acids. As in the vertebrate
collagens, glycine makes up nearly one-third of all of the amino acids and
it seems clear that this level of glycine must have a fundamental relation
to the collagen structures. The total imino acid content varies from 112
to 304 residues per 1000, but a much wider variation is observed in the
individual imino acids, with proline varying from 13 to 280 residues per
1000 and hydroxyproline from 24 to 1G5. Gelatins from Physalia float,
Metridium body wall, and spongin B have a high content of hydroxylysine,
whereas earthworm and roundworm cuticle apparently are devoid of this
amino acid and the former collagen is also lacking in histidine, tyrosine, and
methionine. The invertebrate collagens generally have a larger proportion
of polar amino acids, smaller amounts of imino acids, and more total hy-
droxyamino acids than do the vertebrates, while the aromatic and sulfur-
containing residues are consistently low.
A large variety of sugars are also found associated with the invertebrate
collagens. Chromatographic analyses reveal the presence of glucose,
galactose, glucosamine, galactosamine, fucose, mannose, and in some cases
arabinose (Gross et al., 1958). In contrast to the vertebrate connective
tissues, purification and gelatinization appears to be much less effective in
removing polysaccharide material. Furthermore, the sugar content is
generally much higher than in the vertebrates, ranging from about 18 % in
spongin A to 5 % in Thyone corium (Gross and Piez, 1960). At the moment
there is insufficient evidence to permit a decision as to whether the sugars
form an integral part of the invertebrate collagens or are simply associated
with them physically.
All of the invertebrate collagens which have been examined have yielded
t,he typical collagen wide-angle X-ray diffraction pattern, and electron
micrographs of the fibers show, with the exception of Physalia float and
Metridium body wall, the characteristic 600-700 A axial period (Gross
et al., 1958).
B. Amino Acid Sequence
I. Peptides Derived from Acid and Basic Hydrolyzates
In recent years, the isolation and identification of a large number of di-,
tri-, and tetrapeptides from acid and basic hydrolyzates of collagen and
gelatin have been achieved. Numerous longer peptides have also been
separated from tryptic and collagenolytic digests, and the composition,
and in some cases the partial sequence of these peptides have been deter-
mined. We now have at hand enough information from these studies to
establish the broad outlines of the primary structural pattern predominating
in the polypeptide chains of collagen.
In Table V are collected the most important peptides isolated by Heyns

et al. (1951), Schroeder et al. (1953), 1954), and Kroner et al. (1953, 1955),
following acid hydrolysis of steer hide collagen and gelatin. The peptide
fractions were separated on ion exchange columns, and individual peptides
resolved on celite, followed by sequence studies carried out by the methods
of Sanger.
The widespread distribution of glycine among the different peptides of
Table V, coupled with the fact that this amino acid makes up one-third of
the total residues of collagen, suggests that glycine occurs at every third
residue. It is apparent that this arrangement is not absolute, however,
since some peptides with the sequence Gly.Gly have been identified. The
presence of contiguous, identical residues such as Gly.Gly and Ala.Ala and
the variation in the type of amino acid residue associated with the amino
group of glycine, rule out extensive regular repeating patterns such as
-Pro.Gly.X.Pro.Gly.X- suggested by Astbury (1940). This structure, as
well as one made up of the repeating sequence -Gly.X.Pro.Gly.X.X-
proposed by Bergmann (1935) and Bergmann and Niemann (1938) cannot
be correct in view of the wide variation in total imino acid content en-
countered in the vertebrate (fish) and particularly in the invertebrate
collagens (see Tables I11 and IV) . Nevertheless, examination of Table V re-
veals that the imino acids seem to be distributed through the isolated
peptides in a remarkably systematic manner. Thus, almost all of the imino
acids occur next to glycine, either as Gly.Pro or as Hypro.Gly. Schroeder
et al. (1954) have suggested that the isolation of such a large quantity of
Gly.Pro indicates a very marked lability toward acid hydrolysis of the
peptide bond associated with the carboxyl group of proline, while the high
level of Hypro.Gly is evidence of the lability of the peptide bond associated
with the imino group of hydroxyproline. Since proline and hydroxyproline
are present in about equal amounts in steer hide gelatin, it was inferred
that sequences of the type -Gly.Pro.Hypro.Gly- could be an important
pattern in the primary structure. The work of Kroner et al. (1955) has
supported this proposal with the identification of the tripep tide Gly.Pro.-
Hypro and isolation of the tetrapeptide Gly.(Hypro, Pro) .Gly from partial
acid hydrolyaates of steer hide collagen. From the discussion given in
Section 11, it is apparent that the presence of neighboring imino acid
residues has a profound effect on the chain geometry in these regions. Since
insertion of a single proline results in rotation of the chain by -120", two
contiguous residues would give the chain a left-handed twist of -240" and,
assuming the usual restriction to rotation of neighboring pyrrolidine rings,
would generate an incomplete element of the poly-L-proline I1 helix along
the chain. This feature may be of fundamental significance to the structure
and will be considered in more detail in later sections.
V.
TABLE
Peptides from Acid and Basic Hydrolyzatea of Collagen and Gelatin'
Neutral peptides Basic peptides Acidic peptides Peptides with add and Peptides containing proline and
basic amino acids hy droxyproline
__ -
Ala.Ala 4.6~ Ala.Arg Gly.Asp Asp.Arg
1.oc 1.20 Gly.Pro 61.80
Ala.Gly 13.0. Ala. (Arg,Gly) Gly .Glu 7.0. ASP.(Arg,GW 0.7~Gly .Pro .Gly 0.4d
Gly.Ala 9.oc Ala.Lys Ala.Asp 1.9c Glu.Arg 1 . 8 c Gly .Pro .Ala 3.5"
Val.Gly 4.lc Arg.Gly .Gly Val.Glu 0.5~ Gly .Arg.Gly 1 . o c Gly .Pro .Glu 0.7d
Ser.Gly 18.40 Lys.Gly Leu.Glu 0.4~ - Ser. Pro.Gly 0.5"
Ser.Ala 1.5O Ser.Arg Glu.Gly 4.5c 4.7 Ala.Pro .Gly 0.3d
Thr.Gly 17.40 Glu.Ala 6.6. Lys.Pro.Gly A
Thr.Ala 1.1" Glu .Asp.Gly 0.5d Pro. (Gly,Lys) 1.0=

Gly.Gly d Pro.Ser 1.oe
Leu.Ala -d 12.4 Pro.Thr 0.7~
Ala.Gly.Ala -d Hypro.Gly 35.6.
Ala.Ala.Gly --. Ala.Hypro.Gly 3.6d
__ Ala.Hypro 1 .Od
69.1 Leu.Hypro 1.7~
Ser.Hypro.Gly 1 .4d
Glu.Hypro 0.9d
Glu.Hypro.Gly 2.4d
Gly.Pro.Hypx-o 4.2d
Gly. (Hypr0,Pro.o).Gly 3.ld
I -
23.8
- -
a Taken from Grassmann (1955).
b In micromoles peptides per 250 mg of gelatin or collagen.
0 Peptides have been assembled from the work of Schroeder et al. (1953,1954).
d Peptides have been assembled from the work of Kroner et al. (1953,1955).
Peptides have been assembled from the work of Heyns et a2. (1951).
2. Peptides Derived from Enzymatic Hydrolyzates

a. Tryptic Digests. The acid and basic hydrolyzates of collagen and
gelatin have given much valuable information, but it is clear that the
limited size of the isolated peptides prevents a detailed elaboration of the
primary structure. What is obviously needed is sequence information on
much longer peptide segments and it may be expected that the use of
specific enzymes, a technique employed so successfully in the delineation of
the primary structures of insulin and ribonuclease, will again prove in-
dispensable. Although the task is a formidable one, Grassmann, Hannig,
and their co-workers (Grassmann et al., 1956, 1960; Grassmann, 1960)
have made substantial progress in the separation and characterization of
tryptic hydrolyzates of calfskin collagen. Following digestion of a large
mass of purified, heat-denatured soluble collagen (62 gm) with chymo-
trypsin-free trypsin, 51 homogeneous peptides with lengths varying between
3 and 131 residues were isolated through a combination of preparative
“continuous” curtain electrophoresis and column arid paper chromatog-
raphy. The 51 peptides represent about 55% of the total initial protein.
N-terminal and C-terminal end groups have been determined for 18 of the
peptides, establishing glycine in the N-terminal position for every peptide
and either lysine or arginine in the C-terminal position. Of the 51 peptides,
90% contain between 30 and 38% glycine, giving very strong support to
the contention that, with a few exceptions, this amino acid appears at
every third residue throughout the component chains of collagen. A few
peptides have been analyzed in which the glycine content was significantly
higher than this and partial sequence work revealed two glycine residues
neighboring each other or separated by a single amino acid (Grassmann,
1960). On the other hand, three peptides were isolated which were devoid of
glycine over 4 t o 5 residues.
Proline- and hydroxyproline-rich peptides were found, in general, to be
deficient in diamirio and dicarboxylic acids, while peptides deficient in
imino acids contain large quantities of polar residues. Following the early
suggestions of Bear (1952), Grassmann and his collaborators (Kuhn el al.,
1958a, b ; Grassmann et al., 1960) have particularly emphasized the possi-
bility that the periodic occurrence of these regions could be responsible
for the characteristic band structure observed in the electron-optical in-
vestigation of phosphotungstic acid-stained collagen fibrils (Hall et al.,
1942; Schmitt and Gross, 1948), with the apolar, imino-rich regions cor-
responding to the light “interband” segments and the polar, imino-poor
regions to the dark “bands.” In this connection, investigation of a peptide
consisting of 43 or 44 amino acids (Grassmann et al., 1956) revealed that
the six N-terminal residues contained no imino acids, a central fragment
of 33 amino acids was made up primarily of proline, hydroxyproline, and

glycine (10 Pro, 5 Hypro, 12 Gly, 2 Ala, 1-2 Ser, 2 Tyr, 1 Glu) while the
5 C-terminal amino acids included 3 dicarboxylic acids, but no imino
residues. Similarly, two separate peptides containing 22 and 21 residues
were completely devoid of either proline or hydroxyproline, but yielded
41 and 33 mole % polar amino acids, respectively.
When the minimum molecular weight of a number of the isolated pep-
tides estimated from quantitative end-group determination is compared to
that expected from amino acid analysis, a surprising result emerges (see
Table VI). Twelve of the 18 peptides examined exhibit minimum molecular
weights by amino acid analysis which are, within experimental error, three
times that expected from the N-terminal analysis, suggesting that these
peptides are triple chain structures. Quantitative investigation of the C-
terminal end of these peptides revealed both lysine and arginine as C-
terminal residues. Additional evidence in support of the three-chain
postulate was obtained from titration studies. Assuming a molecular
weight of 360,000 gm/mole for the original collagen molecule, a n average
chain length of 18 residues was deduced from titration of the carboxyl
groups liberated on enzymatic cleavage. On the other hand, a n average
length of 60 residues was estimated from quantitative amino acid analyses
of the separated peptides. It was also observed that addition of leucine
aminopeptidase to some of these isolated peptides resulted in the concurrent
liberation of several amino acids a t about the same rate. This result is also
consistent with the presence of multichain peptide fragments.
These studies furnish strong evidence for the existence of relatively stable
interchain linkages in collagen. Since the currently accepted view suggests
that the collagen molecule is composed of three polypeptide chains, and a t
least in certain collagens it seems established that these chains can be
largely separated from one another by gelatinization, elucidation of the
chemical nature and distribution of interchain cross-links will be of funda-
mental importance to our understanding of the structure. We shall defer a
detailed discussion of this question to Sections IV and V, but it is important
to note a t this point that the degree of cross-linking which has been ob-
served in soluble collagens seems to vary with the source of the material
studied. For example, results to be presented in Section I V suggest that
the individual polypeptide chains of soluble calfskin collagen are more
highly cross-linked than those derived from the collagen of the swim
bladder of the carp. Furthermore, recent studies relating the degree of
solubilization of collagen to the age of the connective tissue suggest that
the degree of cross-linking increases progressively with age. Additional
work is needed to elucidate the relative structural importance of cross-
TABLE VIa
Peptides Derived from Tryptie Digation of the Soluble Collagen of Calfskin
-
>
z & g z
u
-
* G
- -
& , .
g :
z
I
& F s "
- - _ _ - ~ ~ - -__
N-Terminal Gly Gly Gly Gly Gly Gly GlY Gly Gly Gly Gly Gly Gly Gly Gly Gly Gly Gly
C-Terminal Arg Arg Lys Arg 2 Lys 2 Lys 1 Lys Lys Arg 2 Lys Arg Arg Arg 1 Lys 2 Lys Lys Arg 2 Lys
1 Arg 1 Arg 2 Arg 1 Arg 2 Arg 1 Arg 1 Arg
Molecular weight l#)o 2300 - 3350 870 1700 1100 1200 2400 2300 4650 - 1600 3000 1700 - - 1420
$!! by end-group
determination
Molecular weight 1321 6528 5528 3468 2499 4915 3330 4630 6936 7722 10035 4354 5234 9013 5840 591: 368 4670
by amino acid
analysis
Mole %, glycine
Mole %, Pro.
HYPro
-
38 33
26
33
25
32
24
39
11
36
22
34
17
32
261
30
131
33
24 I 34
191
34
241
34
191
33
26
34
14
24
11
33
-
34
25
a Taken from the work of Grassman et al. (1960). Column headings give the specific peptide assignment of the authors for various
acid (S),neutral (N), and basic (B) peptides.
links between the individual polypeptide chains of a single collagen mole-

cule, and those between adjacent molecules in the connective tissue matrix.
b. Collagenolytic Digests. The isolation and purification of a n enzyme
from the culture medium of Clostridium histolyticum (Gallop et aE., 1957b;
Seifter et al., 1959; MacLennan et al., 1953; Mandl et al., 1953; DeBellis et
al., 1954; Schuytema and Kallio, 1956) have provided us with a highly
specific agent for the degradation of the polypeptide chains of collagen.
Although collagenase attacks collagen and gelatin readily, no other protein
so far tested has been found to act as substrate. This remarkable proteolytic
property has stimulated a number of laboratories to investigate the specific-
ity requirements using a variety of low molecular weight synthetic sub-
strates. It appears from these studies (Seifter et al., 1959; Michaels et al.,
19.58; Nagai and Noda, 1959; Heyns and Legler, 1959; Kazakova et al.,
1958; Grassmann et al., 1959; Nagai et al., 1960; Poroshin et al., 1960) that
enzymatic activity requires the general sequence -Pro.X.Gly.Pro- with
cleavage occurring between the X arid Gly residues. Both proline residues
can be replaced by hydroxyproline, but Gallop and Seifter (1961) report
activity is substantially depressed by such substitution for the second
proline group. Collagenase hydrolyzates of calf collagen and ichthyocol
(derived from carp swim bladder) are found to contain peptides in which
the C-terminal residue (X) is virtually any amino acid; however, the N -
terminal residue of these same peptides is always glycine with perhaps one
exception. Preliminary studies indicate that this peptide may have an
N-terminal alanine (Gallop and Seifter, 1961). A few exceptions to the
general rule have also been observed in synthetic substrates. Nagai et al.
(1960) have reported that Gly.Pro.Leu.Gly.Gly.Pro is cleaved between
leucine and glycine, while Heyns and Legler (1959) and Kagai et al. (1960)
have demonstrated carbobenzoxy-Ala.Gly.Pro.-NH2 to be split between
Ala and Gly. The rate of splitting these peptides is markedly lower than
that observed for the general sequence.
Most of the peptides liberated from collagen after digestion with col-
lagenase have an average peptide weight of approximately 500-600, but
a number of larger peptides are also released which are too massive to pass
through a dialysis membrane (von Hippel et al., 1960). Schrohenloher et al.
(1959) have succeeded in isolating and identifying two peptides which
occur in relatively large amount, following digestion of steer hide col-
lagens with a crude collagenase extract. The two tripeptides were
identified as Gly.Pro.Hypro and Gly.Pro.Ala. Together they accounted for
23 % of the alanine, 14 % of the glycine, 23 % of the hydroxyproline, and
40% of the proline of the original substrate. Essentially the same results
were obtained by Gallop and Seifter (1961) using a highly purified col-
lagenase.
42 HARRINGTOR' AND VON HIPPEL
IV. THESTRUCTURE
OF COLLAGEN
A . Structural Studies in the Solid State

I n this section the structure of collagen in the solid stat,e will be con-
sidered a t two levels of magnification. First we will describe in detail the
development, of current ideas on the configuration and packing of the
polypeptide chains in collagen, and the wide-angle X-ray diffraction and
infrared absorption data upon which these ideas depend. Then cert,ain
aspects of the larger-scale structure of the collagen fiber, as viewed by
elect.ron microscopy and small-angle X-ray diff ractioii, will be discussed
more briefly.
FIG.8. Wide-angle X-ray diffraction patterns of collagen from rat tail tendon:
( a ) unstretched; ( b ) stretched 8%. (From Randall, 1954 )
1. The Polypeptide Chain Configuration of Collagen

The main features of the collagen wide-angle X-ray diffraction pattrrn
have been known for over 30 years. In fact, as pointed out in Section I,
this distinctive pattern has generally been considered the best analytical
test for the presenw of collagen in a sample of tissue. The principal re-
flections charactrristic of this material may be seen in the typical wide-
angle X-ray diffraction photograph which appears in Fig. 8a. (For a de-
tailed description, see Bear, 1952; or Millionova and Andreeva, 1957a.)
The 2.86 A meridional arc and the approximately 11 A equatorial reflections
are particularly prominent. Painter arcs may also be seen on the meridian
at 9.Y, A and 4.0 A, while a diffuse halo appears at about 4 A near the
equator. The -11 A equatorial sparing is very sensitive to hydration,
shifting from about 10.4 A in completely dry tendon to an upper limit of
15 to 16 A as the moisture content of the fiber is progressively increased
(see Rear, 1952; Rougvie arid Bear, 1953). The wide-angle meridional
STRUCTURE O F COLLAGEN A N D GELATIN 43
spacings are not appreciably affected by hydration. &lore recently, Cowan

et al. (1953) showed that stretching the collagen fiber by about 10%
during the recording of the X-ray photograph results in a considerably
sharper pattern containing many more discrete reflections (Fig. 8b). These
“stretched” patterns reveal that the meridional arc a t 4.0 A actually
consists of two off-meridional spots, and that the diffuse equatorial reflec-
tion a t -4 A also may be resolved into two off-axial intensity maxima,
leaving only the 2.86 A arc truly on the meridian and the hydration-
sensitive 11 A reflection on the equator.
Despite the early availability of reasonably good X-ray diffraction
photographs (as fiber pictures go) the structure corresponding to this
pattern proved remarkably elusive and only in 19.55 were a pair of struc-
tures finally proposed, more or less simultaneously, by Rich and Crick
(1955) and by Cowan et al. (1955a), which have been generally accepted as
being substantially correct.
Most of the early proposals for the structure of collagen were based
primarily on an attempt to achieve two objectives: (1) to provide a reason-
able explanation of the strong 2.86 A meridional reflection; and (2) to
somehow fit the large number of imino acid residues into the structure.
Furthermore, it was generally felt that proposed structures should account
for the apparent inextensibility of collagen fibers.l
On the basis of these criteria, structures were proposed by Astbury
(1940), Huggins (1943), Ambrose and Elliott (1951b), and others. For a
discussion of these proposals we refer the reader to the original papers or to
reviews by Bear (1952) and Kendrew (1954), because in 1951-52 two
developments totally altered the situation, making all previous structures
obsolete. First, in 1951, Pauling and Corey and co-workers wrote their
classic papers specifying stereochemical criteria which any proposed poly-
peptide-containing structure must satisfy. Then, the following year,
Cochran et a2. (19.52) presented their calculations of the Fourier trans-
forms for helical structures, making possible direct examination of wide-
angle X-ray patterns for the presence of helical configuration. These ad-
vances made the problem both easier and more difficult; easier because a
discrete structure now appeared possible, and more difficult because the
number of criteria which an acceptable structure had to satisfy became
much more numerous.
The first consequence of these developments was the general realization
It should be pointed out that Schmitt, et al. (1942) have shown, under certain
conditions in the electron microscope, t h a t collagen fibrils can apparently extend
manyfold. This led Bear (1952) to suggest t h a t a successful model should allow for
such extensibility. However, under all other conditions collagen fibers have proved
essentially iriextensible beyond about 10% over rest-length, and in recent years this
requirement for the collagen structure seems to have been generally abandoned.
that all of the structures proposed up to that time must be in error, due to
more or less scrious violations of the Pauling-Corey criteria regarding bond
lengths arid angles and coplanarity of amide groupings. Also, examination
of the diffraction pattern on the basis of the results of Cochran et al. (1952)
quickly showrd that collagen must be wound into a helical configuration
(Cohen and Bear, 1953; Cowan et al., 1953).z
In the same series of papers in which they outlined the stereochemical
ptable polypeptide structures and presented detailed
descriptions of the a- and y-helices and the p-structures, Pauling and Corey
(1951~)also proposed a structure for collagen. They suggested a three-
chain structure with each polypeptide chain coiled into a helix, all three
helices having a common axis. In this model equivalent amino acid residues
occurred a t the same level in each chain, the a-carbon atoms of the three
residues a t a given level occupying the corners of an equilateral triangle
located perpendicular to the fiber axis. These triangles werc spaced regularly
along the axis, a rotation of 40" and a translation of approximately 2.86 A
mnverting one three-residue element into the next. The required repeat of
2.86 A was achieved by specifying a cis-trans-cis sequence of peptidc bonds
and a mrrcsponding G1y.Pro.X arrangement of amino acids, where X
could bc any residue other than proline or hydroxyproline. Interchain
peptide hydrogen bonds were made between two of the three residues in
each triad, with the bond direction perpendicular to the fiber axis. No
hydrogen bonding between triple-chain elements was proposed, permitting
these elements to move relative to one another as suggested by the hydra-
tion-sensitivity of the 11 A equatorial spacing.
This structure represented an improvement over most of the prcvious
attcmpts in that it (naturally) satisfied the Pauling-Corey criteria, was
helically wound, aiicl gave a true 2.86 A repeat along the fiber axis. How-
ever, in the light of subsequent experience it proved inadequate and had to
be abandoned. The following difficulties became apparent:
(1) Although the model accounted satisfactorily for the 2.86 A axial
spaeing and the behavior of the 11 A equatorial reflection, Randall et al.
(1953b) pointed out that the structure did not explain certain other features
of the X-ray pattern, nor was it quantitatively compatible with infrared
dichroism measurements.
(2) Thr model rcquired that two-thirds of the total peptide bonds assume
the cis-configuration. Yet careful infrared absorption measurements
(Badger arid Pullin, 1954) showed that collagen contains very few, if any,
cis-amide groupings. Furthermore, Corey and Pauling (1953) themselves
pointed out that the trans-configuration is probably more stable than the
2 An excellent general discussion of the characteristics of wide-angle patterns ob-
tained from helical diffractors is given by Stokes (1955).
STRUCTURE OF COLLAGEN rlND GELA4TIN 45
cis, and suggested that since the cis-configuration had oiily been established
unequivocally in a single, rather unusual case (in the cyclic dipeptide,
diketopiperazine) , the trans-form of the amide grouping was to be pre-
ferred and should be used in proposed polypeptide structures whenever
possible.
(3) The amino acid sequence work of Schroeder et al. (1953, 1954) and
Kroner et al. (1953, 1955) showed that the sequence Pro.Hypro is quite
common in collagen; yet this sequence could not be accommodated in the
Pauling-Corey structure.
I n 1952, Randall and co-workers proposed a quite different st,ruct>ure
which met most of the above objections. They placed all the peptide bonds
in the trans-configuration (achieving the required meridional spacing by
tilting the residues to obtain a projected 2.86 A axial separation between
a-carbon atoms) and were able to orient the N-H groups so as to satisfy
the then-available infrared data. Furthermore, their model accounted
satisfactorily for some of the wide-angle X-ray reflectZionswhich the
Pauling-Corey model could not explain. However, the model they proposed
was basically a two-dimensional sheet structure, and as such was not in
accord with the helical configuration suggested by the general shape of the
X-ray pattern.
Discarding all previous models, Bear and co-workers began a more
systematic approach to the problem of collagen structure (Rear, 1952;
Cohen and Bear, 1953; Bear, 1955) .3 Instead of proposing a specific stmc-
t,ure, Cohen and Bear began by laying down a set of conditions which any
successful model must satisfy. From an analysis of the wide-angle pat,tern
they deduced that, the collagen structure must be helical and consist of
seven roughly “equivalent, scattering groups” located along a discon-
tinuous helix making approximately two turns per 20 A rise along the fiber
axis. Subsequently, Bear (1955) found slightly better agreement with the
X-ray data by assuming ten scattering groups located along three turns of
such a “genetic” helix.4 Bear also noted that t.he posit,ional aspects of
the diffraction pattern could be satisfied by specifying 2-, 3-, or n-chain
3 A similar, systematic survey of various alternative helical structures for collagen
was undertaken a t about the same time by Cowan and co-workers (1953, 1955 a , b).
It should be noted t h a t the term “equivalent scattering group” does not nec-
essarily imply single amino acid residues. In fact, for collagen, density considerations
suggested approximately three residues (with an average residue weight of 93 g/mole)
per equivalent scattering group. Bear (1955) adapted the term “genetic” helix from
botanical usage, defining the genetic helix as the single helix with the smallest number
of turns per period which could be passed through all the equivalent scattering
groups. For a more detailed discussion of these points and the general use of “helix-
net” theory in the systematic derivation of polypeptide chain structures from X-ray
data, the reader is referred to Bear’s (1955) lucid presentation.
46 HAHHISGTON A N D VON HIPPEL
helical strurtures, a t thr same time increasing the axial period 2-, 3-, or
n-fold. However, additional ambiguities exist, so that, as Bear stated :
“While the application of transform theory has limited the heliral modcls
which may br ent,eriained for collagen, these restrictions are far from
capable of isolating n unique structure.” Kevertheless, he felt that a
systematic approach based on these principles and a knowledge of the
stereochemical properties of polypeptide chains, with ultimate testing of
detailed structures by optical diffraction methods, should be able to isolate
the correct structures. As it turned out, the apparently successful models
were initially derived by others, partly by analogy with the structures of
certain synthetic polypeptides (see below and Section 11), but Bear (1956)
was able to use the systematic elimination approach to confirm independ-
cntly the essential correctness of these structures.
As these ideas were developing, the elaboration of specific collagen
structures continued. Considerably earlier, Bear (1952) had tentatively put
forward, as a collagen prototype, a slightly modified version of the y-helix
described by Pauling et al. (1951). This structure, a single helix with a
rather shallow pitch, could a t least account for the apparent sevcralfold
extensibility which the electron micrographs of Srhmitt et al. (1942) seemed
to require. However, a single helix of this type seemed unsatisfactory in that
it could not accommodate imino acid residues, did not account for the
perpendicular dichroism of the N-H and C=O stretching frequencies, and
did not incorporate a unique explanation of the 2.86 A meridional reflection.
I n 1934 Huggins proposed a quite different single-stranded helical struc-
ture. The polypeptide chain in his model was coiled in a left-handed helix,
with ten residues per turn and a pitch of 9.5 A. It was built assuming
l’nuling-Corey bond anglrs and distances and planar amide groupings,
and included rectilinear, intrachain N-H. . . O peptide hydrogen bonds of
equal length at two out of three residues. However, this model was also
based on a three-residue repeating unit, with only one position able to
accommodate either proline or hydroxyproline without distortion, again
excluding the common Pro.Hypro sequence from the crystalline portion of
the structure. Other difficulties included the requirement that half of the
peptide bonds he in the cis-configuration; also the density calculated for
the structure seemed slightly low.
Crick (1954) proposed a two-stranded helix, somewhat reminiscent of the
Watson-Crick double-strand helix which had been so successful for deoxy-
ribonucleic acid. Thcb model consisted of two polyprptide chains wound
helically around a common axis and held together by interchain peptide
hydrogen bonds. The peptide bonds were all in the cis-configuration,
and the rcprating unit consisted of a pair of amino acid residucs, one
perpendicular arid the other parallel to the fiber axis. As a result, plaiics
STRUCTURE OF COLLA4GES AND GEL.4TIK 47
perpendicular to the fiber axis and passing through the a-carbon atoms
of pairs of residues a t each level were equally spaced a t a separation of
2.95 A, close to the 2.86 A spacing observed. However, as Crick himself
pointed out, this model incorporated some unusually close van der Waals’
contacts. Furthermore, it suffered from incompatibility with the measured
infrared dichroism (Sutherland, et al., 1954), was built entirely from the less
favored cis-amide groups, and showed some minor discrepancies when an
optical transform, constructed by Wyckoff and Chow (see Bear, 1955) was
compared with the experimental X-ray pattern.
FIG.9. Three-chain collagen structure proposed by Ramachandran and Kartha

(1954). Dotted lines represent hydrogen bonds. Only one pyrrolidine ring is shown.
The center line indicates the direction of the crystal axis. (From Ramachandran and
Kartha, 1954.)
Finally, in 1954, Ramachandran and Kartha proposed a structure which,

while incorrect in certain particulars, was along the right lines and with
suitable modification did lead to the present generally accepted models.
I n this structure the residues were arranged in three equivalent poly-
peptide chains, each wound in a left-handed helix with three residues per
turn and a pitch of 9.5 A. While also a three-chain model, this structure
differed from that of Pauling and Corey in that each chain coiled around
its own axis, rather than around a common axis. The three chains were held
together by hydrogen bonds as follows: in each turn of each helix two of the
three peptide nitrogens were hydrogen bonded to one of the carbonyl
oxygens of each of the other two chains (see Fig. 9). This model had a
48 HARltINGTON A N D V O S HIPPEL
number of positive features: all amide groups were built to the Pauling-
Corcy dimcrisions aiid in the trans-configuration, hydrogen bond lengths
were (.lose to the accepted values aiid the N-H arid C=O groups were
oricrited so as to agree, both qualitatively and quantitatively, with the
infrared dirhroism measurements of Sutherland et al. (1954) ; (see Rama-
ehandran, 19ri.5). However, it still left unresolved some rather formidable
difficulties. First, it was clear that an arrangement of several polypeptide
chains with axes parallel to the fiber axis and to one another required that
the apparently meridional 2.86 A arc actually arise from the superposition
of two, somewhat off-axial, intensity maxima. Using the older, rather
fuzzy X-ray patterns, this possibility could still be entertained, but with
the advent of the stretched-fiber wide-angle patterns of Cowan et al. (1953)
it became increasingly clear that the 2.86 A reflection was truly meridional,
and must be accounted for as such by a successful model. Also, the structure
as given incorporated hydrogen bonds with much larger angles between the
?;-H and the N . . .O directions than generally considered reasonable. And
furthermore, as in several previous models, stereochemistry required a
G1y.Pro.X rcpeating sequence (with X any residue other than proline or
hydroxyproline) ; thus again excluding the common sequence Gly.Pro.-
Hvpro from crystalline regions.
Ramachandran and Kartha soon recognized some of these difficulties,
and accordingly published a somewhat modified model in 1955. They found
that the first two objections caould be rrctified by simply twisting the
thrce straight he1icc.s of thcir previous model into a three-membered,
right-handed super-coil winding around the fiber axis. This modification
placed the 2.86 ,4 reflection hack on the meridian, and also brought the
peptide hydrogen bonds closer to linearity. A t the same time it did not,
introduce much distortion into the original minor helices siiicc the major
helix coiled only very g r a d ~ a l l yIn
. ~ this structure the super-helix repeated
itself after thirty residues (per chain), over a fiber axis repeat distance of
85.8 A. The chains were symmetrically disposed around the fiber axis; a
rotation of -108” :md a translation of 2.86 A bringing one chain into
coincidence with the next. However, this revised model still failed to ac-
commodate the Pro.Hypro sequence, and, as Rich and Crick (1955) soon
pointed out, also contained some uncomfortably short van der Wads’
contacts.
A few months after Ramachandran and Kartha’s model appeared, Rich
The terms “major” and “minor helix” are used here as defined by Crick (1953)
in his original presentation of the coiled-coil (or super-helix) concept. The minor
helix defines the turns of the residues of an individual chain around its own axis.
The major helix define6 the turns of the minor helix around an axis outside of itself:
e.g., the common axis running up the center of the group of three chains in the col-
lagen case.
and Crick (1955) presented a still further modified version of this basic:
structure. Starting from their previous work on polyglycine I1 (Crick and
Rich, 1955; discussed in detail in Section 11) they showed that a bundle of
three chains, selected from the polyglycine I1 lattice in either of two ways
and twisted into a right-handed coiled coil similar to that of Raniachandran
and Kartha, resulted in the generation of a pair of structures which were
stereochemically entirely satisfactory and which would accommodate the
Gly.Pro.Hypro sequence, These structures have been generally accepted as
rather close approximations to the structure of collagen in the diffracting
regions of the fiber (but see Ramachandran et al. below) and will be dis-
cussed in some detaiL6
The development of the two structures may be visualized in two steps,
as follows. First, in both structures, the individual polypeptide chains are
coiled into a helix with a threefold, left-handed screw axis; movement along
a single helix from one residue to the next requires a rotation of -120"
and a translation of 3.12 A. Thus each complete turn of an individual (or
minor) helix requires three residues and a 9.4 A rise along the fiber axis.
As pointed out above, this is basically the backbone proposed by Crick and
Rich for polyglycine I1 and for poly-L-proline I1 by Cowan and McGavin,
1955 (see Section I1 and below). Two such chains are shown lying side by
side in Figs. 1Oa and lob.
The over-all collagen structure may be developed by taking three such
chains and setting them parallel so that, viewed from above, the axes form
the vertices of an equilateral triangle with sides about 5 A in length. The
three chains are also arranged so that equivalent elements (e.g., the a-car-
bon of residue 1) are at the same level. At this stage, the three chains are
related by a threefold screw axis running up the center of the group.
Considering now the two chains with axes located in the plane of the page
(Figs. IOU and b), it is clear that the third chain may be added by placing
it either behind (Fig. 10c) or in front of the other two (Fig. 1Od). In either
case it may easily be seen that only one out of every three residues along a
single chain (residue 1 in the numbering system of Fig. 10) lies near the
middle of the three-chain structure, while the other two lie on the outside.
Therefore, in each of the two arrangements of chains described above, only
one relative orientation can be found in which every third backbone N-H
can make a stereochemically proper hydrogen bond with every third C=O
of a neighboring chain. Thus in both structures each three-residue repeating
element is hydrogen bonded to each of the other chains via one hydrogen
6 The description of collagen structures I and I1 which follow are adopted in part
from a lucid discussion of these models by Rich and Crick (1958). A complete account
of these structures, including the details of the derivation of the final models and
complete sets of atomic coordinates has recently been published (Rich and Crick,
1961).
50 H.4RTZINGTOS rlND VON HIPPEL
bond, contributing ail K-H to one bond aiid a C=O to the other. Vicwed
end-on, the interrhain hydrogen bonds form the sides of the equilatcral
triangle mentioned above.
Figures 1Oc and d illustrate these points and also show the essential
difference bctwecn the two collagen structures. Adding the third chain
a b C
d e
FIG. 10. ( a ) Two polypeptide backbones shown side-by-side; each helically wound,
with a left-handed threefold screw axis (vertical lines). The dotted lines between the
two chains represent hydrogen bonds. ( b ) Simplified version of Fig. IOU; only C,-
atoms are shown. (c) Same as Fig. 106, but with a third backbone added behind t h e
other two. This arrangement is related to collagen I. The residues are numbered as in
Table VII. ( d ) Same as Fig. l o b , but with a third backbone added in front of the other
two. This arrangement is related to collagen 11. ( e ) Showing the deformation of the
axis of the minor helices of Fig. 1Oc and d to give the compound collagen helix. Note
t h a t the axes of the three polypeptide chains now follow gradual right-handed hel-
ices rather than running straight. The broken line represents the common axis around
which the axes of the three chains wind. (From Rich and Crick, 1958.)
behind the other two leads to a structure related to collagen I (Fig. LOc).
Adding the third chain in front of the others leads to a structure related to
collagen I1 (Fig. IOd). Thus these structures differ basically only in the way
the chains are placed relative to one another.
The actual collagen structures may be derived from these imaginary
constructions by siniply twisting the latter slightly so that the axes of the
individual chains coil slowly about one another in a gradual right-handed
(major) helix (see 1;ig. 1Oe) instead of running straight and parallel. This
deforms the threefold screw axis which previously related the chains to
one another in such a way that a rotation of -108” and a translation of
2.86 A takes one from a residue on one chain to the corresponding residue
on the next. After three such operations one arrives back on the poly-
peptide chain from which one started, but three residues higher up (see
Fig. 1Oc and d ) . Thus, as in the Ramachandran and Kartha structure, one
complete turn of the super-helix contains thirty residues per chain and
requires an 85.8 A translation along the fiber axis. It should be emphasized
again that the original helices (here the left-handed polyglycine I1 or
poly- proli line I1 backbone) are only slightly distorted in being incorporated
into the major helix.?
Having assembled the backbone structures of collagens I and 11, specific
amino acid side chains may be added. I n particular, the incorporation of the
Gly.l’ro.Hypro sequrnce (which contributed to the downfall of so many
previous models) must be considered. Table V I I summarizes the possible
positions of various types of side chains in terms of the residue numbering
system used in Fig. 10.
The year 19.55 seems to have been a vintage year for collagen structures.
Only 3 weeks after the publication of the Rich-Crick models, a paper ap-
peared by Cowan, McGavin, and North in which they described es-
sentially the same structures, deduced independently and from a different
point of view. As mentioned above, some years earlier the King’s College
group (like the Massachusetts Institute of Technology group) had launched
a systematic attempt to fit various types of helical systems to the observed
wide-angle X-ray diffraction pattern of collagen. By 1953, they had reduced
the number of feasible alternatives to a few major types (see Cowan et al.,
195.ib; Cowan et al., 1 9 5 5 ~ )iiicluding a three-chain, coiled-coil structure
much like that proposed by Ramachandran and Kartha (195.5). Then Co-
wan and McGavin worked out the structure of poly-L-proline I1 (see
Section 11) and it soon became apparent, that an acceptable collageii struc-
ture could be built by twisting a group of three chains, each originally
folded in the poly-L-proline I1 configuration, into a super-coil. I n attempting
The similarity of these structures to t h a t proposed by Ramachandran and Kartha
(1955) is striking, and therefore it is worth pausing t o point out just where the dif-
ferences lie. Both models (considering the Rich-Crick btructures I and I1 together)
are similar in containing minor helices built on a threefold left-handed scIew axis of
essentially identical pitch, and both feature identical right-handed major helices.
The models differ basically only in the way the three chains are packed together;
Ramachandran and Kartha attempted t o form t wo systematic interchain peptide
hydrogen bonds per three-chain segment, while Rich and Crick were content t o
form only one such bond per three residues in order t o achieve a stereochemically
more satisfactory structure.
52 HAKKINGTOS AND VON HIPPEL
to build this stru(Ature, Cowan et al. also discovered that the sequcnce
Gly.Pro.Hypro could only be accommodated by limiting themselves to one
peptide hydrogen h i d pcr three-residue element, and that, on this basis,
two acceptable modcls could be constructed. These two structures are
TABLEVII
Possible Positions of Side-Chains i n Collagens I and IIa
Collagen I
Position Collagen I1
Undeformed Deformedb
1 Gly only Other residues may Gly only

be possible; Pro or
Hypro impossible
2 Any residue Any residue includ- Any residue includ-

including ing Pro or Hypro ing Pro or Hypro
P r o or Hypro
3 Gly only Any residue, includ- Any residue, includ-

ing Pro or Hypro, ing Pro or Hypro
exrept Valine
__ ~
Bonding of the Can make a hydro- Projects rrtdially

OH of Hypro gen bond t o the away from the
in position 3 neighboring chain structure and can-
within the group not make a hydro-
of three chains gen bond within
the group of three
chains
Position of the From the NH of residue From the NH of

interchain 1 t o the CO of residue residue 1 t o the
hydrogen 1 on a neighboring chain CO of residue 2 on
bonds a neighboring
chain
a From Rich and Crick (1958).
b Side-chains could be added much more easily t o structure I by permitting certain
small deformations of the backbone structure, so allowed side-chains for a deformed
structure I are also listed.
equivalent to structures I and 11 of Rich and Crick, but were called the
“antklockwise” and “clockwise” structures, respectively, by Cowan et al.
(corresponding to the direction of the NH. . .O peptide hydrogcri bonds,
viewed from the carboxyl end of the chains).
Shortly thereafter, Ramachandran (1956) reconsidered his original
proposals on the hahis of the work of Rich and Crick, and Cowan el al., and
also agreed that thta Ricbh-Crick structures were probably, in general,

correct.* And when Bear (1956) announced the elimination of the other
theoretically possible models via a complrtrly indrpeiident, systematic
model-building approach, the general validity of the Rich-Crick structures
seemed rather well rstablished.
Most of the recent wide-angle and infrared work on collagen has been
aimed at determining which of the two proposed structures is actually the
more nearly correct.$ This is difficult, since the two models actually differ
too slightly to be definitely distinguished on the basis of the available data.
In fact, the possibility that stretches of both could exist in the collagen
fiber should not be entirely dismissed. However, what evidence there is
seems to favor the collagen I1 structure. Ramachandran (1956), Burge
et al. (1(358),and Rich and Crick (1958) all prefer structure I1 because it is
stereochemically more satisfactory and does not require deformation to
admit the Gly.Pro.Hypro sequence. Bear (1956) obtained optical trans-
forms of both structuresand found structure I1 in slightly better agreement
with the X-ray pattern. The King’s College group carried out a detailed
comparison between the diffraction pattern calculated for collagen I1
and the observed wide-angle diffraction pattern, and found reasonably
good agreement (Burge et al., 1958; Bradbury et al., 1958). I’auling (1958)
built both structurrs using high precision, dural molecular models and
found the identity period in collagen I1 to be rxactly 28.6 A, as required
by the diffraction pattern. Collagen I was less satisfactory in this respect.
However, both models were compatible with the infrared dichroism data.
In 1958, Beer et al. re-examined the infrared dichroism situation and
remeasured the dichroic ratio for several of the peptide bands in collagen.
On the basis of studies on model compounds, they calculated the directions
of the transition moments for the principal bands in the peptide link and,
using the atomic coordinates for various proposed collagen models, the
inc.lination of these transition moment vectors to the fiber axis. Since
Beer (19Fi6) had shown that the dichroism of a partially oriented polymer
may be considered equivalent to that of a sample containing fully aligned
arid completely random portions, a “disorientation parameter” (f) charac-
teristic. of the degree of order of the sample could be calculated for each
band. The “spread” of the values obtained for this parameter, which is
characteristic of the &ampleonly and not of the bands used to determine it,
* Raniachandran called the two models (‘plus” and “minus,” corresponding t o
the Rich-Crick structures I and 11, respectively. For simplicity, in t h e remainder
of this review w e will use the Rich-Crick nomenclature.
6 Certain more indirect evidence bearing on this question comes from physico-
chemical and infrared studies on interchain hydrogen bonding involving the OH

group of hydroxyproline, since the orientation of these groups is markedly different
in collagens I and I1 (see Table VII). These studies will be considered below.
54 HAIZRINGTON . W l ) V O N HIPPEL
rcprcsrnts a quantitative niwsure of the compatability of the obsrrvcd

iiifrarcd dic*hroismand the proposed strurturr. Perfect agreement should,
of course, lead to identical values off for each band. Brrr and co-workcrs
used this inethod t o trst three proposed collagrn models: t,hose of I’auling
and Corey (1951c), Iiamachandraii and Kartha (IS.%), aiid Rich and Crick
(1955) (collagen 11). The Ramachandran-Kartha structure gavr the best
agreement of the three; the Rich-Crick structure was somewhat less satisfac-
tory. However, Beer et al. pointed out that the latter structure should wr-
tainly not h r rcjrctcd 011 this basis alone. The I’auling-Corry structure was
found definitely unarceptable, in accord with the X-ray rrsults.
The structurrs proposed by Rich and Crick and by Cowan, McGavin,
and North, havr not attained completely universal a( ptancr. Huggins
(18.57) stated that, in his opinion, there was little rhance that these models
could be even approximately correct, because: “. . . they conform neither
to the principle that, hydrogen honds are formed whenever possiblr nor
to the principle that like groups tend to be surrounded by closr neighbors
in a like mannrr. Also. . . there seems to be no way of explaining the
large exteiisioiis of the band pattern occasiorially observed in thr elecatron
mirroscopr pictures.” Huggins then presented a siiigle-chain hrliral model
somewhat similar to his previous proposal (Huggins, 1954) but, improved
in that it now cwitaincd only trans-amide groupings. As in his prr\‘lous ‘
model, thr polypeptidr chain was coiled into a left-handed hrlix with
thirty residues (ten equivalent scattering groups) per three turns, arid per
28.6 A risr along the fiber axis. In the 1957 modrl, each amino group of
the peptidc linkage was hydrogen bondrd to a carboxyl oxygen locaatrd
three residucxs further down thr chain, again prrmitting the formation of
two intrachain peptidc hydrogen bonds per thrrr-rrsidue unit. As before,
oiw could ol~jerito this model on the grounds that the stereochemistry
required thc repeating sequrncc GIy.Pro.X, with X neither prolinc. nor
hydroxyprolinc. IIowever, Huggins stated that these residues could he
placed in position X wit>hout,murh disruption of the structure.
IKIdiscussing this model, l’auling (1958) pointed out, that, he and Corey
had also proposed an identical singlc-chain structiirr, but had found on
attempting to huild a prrcisr model that the stable configuration actually
contained four uiiits per turn rathrr than the 8.33 suggested by the X-ray
data. Icor this reason arid hecausr of the difficulty in fitting the Gly.Pro.-
Hypro se(1uencr, I’auling felt that this model must be abandoned.
Paulirig (1 958) and cdleagues also used the precise model-building
approarh to produce rcrtain stereochemically feasible variants of the
collagrri I aiid 11 structures, in which one amide group in csch of the thrcc-
residue elements of collagen had lieen rotated by 180” about thc single
bonds coiinrcting it to the adjacent carbon atoms. Howevrr, the “reversed”
STRUCTURE O F COLLAGES AX D GELATIN 55
structure corresponding to collagen I1 gave the wrong identity period,

while reversed structure I gave the proper identity period but could riot
accommodate the Gly.I’ro.Hypro sequence. Thus, from all points of view,
the collagen I1 structure seemed to represent the best available model for
the configuration of the polypeptide chains in the crystalline portions of the
collagen fiber.
Very recently, several long papers dealing with the structure of fibrous
proteins and polypeptides have emanated from the laboratory of G. hi.
Ramacharidran (Ramachandran, 1961; Ramachandran et al., 1961; Laksh-
manan et al., 1961 and Sasisekharan, 1961).1° Specifically, some new ideas
on collagen structure are included which seem worthy of careful considera-
tion. Though final evaluation of the new data and suggestions contained in
these papers must await the passage of time and consideration by workers
involved directly in the field of fiber X-ray diffraction analysis, we will
attempt to examine some particularly pertinent portions of this very
comprehensive work.
On the basis of an extensive compilation of bond lengths, angles, and
unbonded contact distances derived from various organic Structures which
have heen measured accurately, Sahisckharan emphasized that : (a) pri-
mary valence bond lengths and angles may deviate appreciably from
standard values; (b) hydrogen bond lengths and angles differing markedly
from the “normal” values are often found (see also k’uller, 1959, for a
similar recent compendium of hydrogen bond dimensions) ; and (c) UH-
bonded contact distances somewhat shorter than the sum of van der
Waals’ radii can occur. Moreover, known structures do not always exhibit
complete amide-group coplanarity. Sasisekharan classified these various
parameters into groups designated “common,” “occ.asiona1,” arid “rare,”
depending upon the frequency with which they occur in known structures.
On the strength of this compilation, Ramacharidran et al. have questioned
the rigidity with which stereochemical parameters have been applied in
formulating and defining acceptable structures for proteins and polypep-
tides. Though individually the standard values doubtless represent the
energetically most favorable situation in each case, it is clear that com-
promises must often be made in individual param rs to achieve maximum
stability of the over-all structure. Therefore structures should not, a priori,
be discarded because they do not satisfy the ideal Pauling-Corey criteria
completely, especially if the suggested distortions fall within the ranges
observed in known structures and if additional, previously excluded,
modes of stabilization thus become available.
The present, generally accepted model for collageii (collagen 11) has
10 We are grateful t o Professor Ramachandran for sending us manuscripts of these
papers prior t o publication.
56 HAHRINGTOS AND VON HIPPEL
heen re-c.xamined iii the light of thcse ideas and some new X-ray studies
(wried out, by Lakshmanan ct at. Considering first the collagen I1 strurturc
(which they term the “single-bonded 11” structure, referring to the fact
that this structure forms only one peptidc hydrogen bond per three-residuc
unit) Ramacbhandran and co-workers point out that this structure is
stereochemically far from unique. By varying the various dimensional
parameters systematically through the common range, a wide variety of
triple-chain, single-bonded structures may be generated, with no one
structure particularly stereochemically preferable to another. Thus there
appears to be no stereochemical rcason why such a single-bonded structurc
should assume the particular helical parameters characteristic of collagen
11. In fact, Ramachandran et al. show that a satisfactory single-bonded
structure can be made without invoking a coiled coil, without requiring
that every third residue be glycine, and with projected axial residue re-
peats varying withiii rather wide limits.
On thr other hand, by permitting a few short van der Waals’ contacts,
a triple-chain struct lire with two interchain hydrogen bonds per t hrec-
residue chain element can be coiistructed. This structure, which Ramachan-
dran ct at. refer to as the standard double-bonded structure, is actually
only slightly different from that previously put forward by Ramachandran
and Kartha (195.5). All bond angles and distances, uiiboiided eontact dis-
tances, and hydrogen bond dimensions are held within the observed (though
riot always the common) ranges, and therefore Ramachandran el aE. propose
that this structure he considered satisfactory; the extra strain energy re-
sulting from slight dwiations from the “standard” values of tht. stereochem-
i d parameters should be more than compensated by the stabilizing in-
fluence of the second hydrogen bond. This structure actually fits the X-ray
evidence more closely than the single-bonded I1 structure (see Itamachan-
dran et al. 1961; Lakshmanan et al., 1961) and also provides better agree-
ment with the infrared dichroism results (Beer ct al., 1958). Moreover, the
double-bonded strwture is highly specific and can accommodate only helical
parameters close to those which actually occur in collagen, thus offering a
natural stereochemiral explanation for the specific coiled-coil structure oh-
served and for the requirement that every third residue be glycine.
As beforp, only oiie member of the three-residue repeating chain can be
an imino acid without deforming the structure, so again, ideally, thc se-
quence Gly.Pro.Hypro is excluded. However, only a slight (and stereochem-
ically permissible) adjustment need be made t o accomrnodatc this sequence.
Figure 1l a shows a projection down the axis of the proposed double-bonded
structure, whilc Fig. 110 shows how this structure may be modified to ac-
commodate the Gly.€’ro.Hypro sequence. It should be noted that if only
one of the three chains a t a given level contains this sequence, then only
FIG.11. Axial projection of three polypeptide chains cast into the “double-
bonded” structure proposed for collagen by Ramachandran et d.(1961). The dotted
lines represent hydrogen bonds. ( a ) The standard structure; ( b ) structure modified
to accommodate two imino acid residues per three-residue repeating unit. The un-
distorted position of the B chain is also shown (lightly) t o indicate the extent t o which
the various atoms of the chains have been moved relative t o the undistorted struc-
ture. (From Ramachandran et al., 1961.)
57
58 HARR1NC:TON AND VON HIPPEL
that chain need adopt thc modified configuration and five iriterchairi hydro-
gen bonds can still be formed per iiiiie residues.
As a (wiisequeiiw of these developments, we must close this sectioii on a
slightly less definitive note than might have been possible a few months
ago, stating only that there seems, at least, to be general agreement that
collagen is a triplc-chain coiled-coil structure stabilized by one or two inter-
chaiii pept ide hydrogen bonds per three-residue repeat>ingm i t .
2 . The Structure of the Collagen Fibril

Turning now to the larger-scale features of the collagen complex, we
must consider first what fraction of the polypeptide chains is actually cast
into the ordered configuration drscribed above. Since the crystalline por-
tions give rise to the discrete reflections of the widc-angle X-ray pattern,
while the amorphous (nonrcpeating) regions are responsible for the difiuse
background, one should in principle be able to calculate the fraction in
each form by comparing intensities in appropriate portions of the pattern.
I n practice this is difficult and one must turn to other methods t o invcsti-
gate this question ; specifically, in this case, to electron microscopy and
small-angle X-ray diffraction.
Vicwed microscopically, collagen appears to he laid down in \midles
madc up of parallel fibrous elements. These fibcrs seem smooth and undif-
ferentiated at thc resolution of the light microscope, but as the early c~lcc-
tron microwope work of Wolpers (1043, 1944) and especially of Srhmitt
and co-workers (1942, 1945), Schmitt and Gross (1948) showed, the indi-
vidual fibrils which make up the collagen fiber are characterized by a dis-
tinctive pattern of fine cross-striations. At high resolution, in suitably
stained prcparations, these striations are seen to consist of regularly re-
penting srts of “baiids” and “interbands” having fixrd rclative positions
and electroil densities. In native fibrils, the over-all pattern repeats s t an
average axial interval of about 640 A (close to 700 A in hydrated samplcs).
Figure 12 shows a typical elec%ronmicrograph of stained fihrils at a fairly
low magnification; the interband structure of the repcating unit may be
setn in Fig. 13, which shows a greatly enlarged (and schematized) view
of one ti40 A period of native collagen. It should be noted that this intra-
period band pattern is polarized; i.e., it is not symmetrical about a plane
passed through the centcr of the period normal to the fiber axis. This intra-
prriod finc structure is not always observed, but a regular alternation of
slightly enlarged, phosphotungstic acid-staining bands, mid smaller diam-
eter, nonstaining interbands is characteristic of almost all of the native
collagens which have been investigated.
At ahout the time that collagen was first examined with the electron
mirroscope, Rear (1942, 1944) and Kratky and Sekora (1943) succcedcd
STRUCTUHE O F COLLAGEN AND G ELA TI N 59
in resolving the small-angle X-ray pat tern characteristic of this material

(Clark et al., 1935; Wyckoff et al., 1935) and showed that all the layer lines
which appeared spaced along the meridian of the pattern (see Fig. 14)
represent higher orders of a fundamental axial repeat of about 640 A. Using
special cameras and finely collimated beams, as many as thirty diffraction
orders have heen observed in carefully prcpared specimens (Bear et al.,
1951; Kaesbcrg and Shurman, 1953; Tomlin and Worthington, 1956). Thus
FIG.12. Electron micrograph of air-dried, chromium-shadowed collagen fibrils

from adult human corium; magnification. X 2G,000. (From Gross and Schmitt, 1948.)
the large-period repeating structure which gives rise to these lincs is ob-
viously highly ordered. No evidence of regularity prependicular to the fiber
axis has been ohserved in small-angle X-ray diffraction studies.
Since a fundamental axial repeat of -640 A could be observed by means
of both electron microscopy and X-ray diffraction, it appeared that this
spacing was not ttrtifactual but represented a definite repeating structure
of some kind. Bear (19.52) suggested that the band-interband repeat seen
in electron micrographs might bc due to a regular alternation of groups of
amino acids, the band regions containing high concentrations of long-chain,
polar amino acids and the intcrbands containing mostly the smaller, non-
60 J-IARRINGTON A N D VON I-IIPPEL
polar residues. He proposed that the interhands form the portions of the
fibril which give rise to the wide-angle X-ray pattern, the polar amiiio acids
located in the bands being too bulky and highly charged to pack properly.
Thus the bands would he the most amorphous portions of the fihril, and
FIG.13. Schematic view of the periodic fine structure (phosphotungstic acid-

stained) of the repeating elements of various reconstituted collagens: native pat -
tern, fibrous long spacing (FLS) collagen and segment long spacing (SLS) collagen.
(From Schmitt e t al., 1955.)
therefore also more easily penetrat,ed by st'ains. Investjigations of t,he cffccts

of tension and hydration on the wide- and small-angle X-ray diffraction
patterns (Bear et al., 19.51; Cowan et al., 1955b; Bear arid Morgan, 1957)
and cert,ain of t'he amino acid sequence studies carried out by Grassmanri
and co-workers (see Section 111) lend considerable support, t o t>hishypot,he-
sis.
STRUCTURE OF COLLAGEN . 4 m GELATIS 61
More detailed correlation of intraperiod fine structure, as viewed by

small-angle X-ray diff raction and high resolution electron microscopy, has
been attempted by several groups (see Cowan et at., 1955b; Randall el al.,
1955; Burge and Randall, 1955; Bear and Morgan, 1957) without, however,
casting significantly more light on t h r molecular factors involved. As it
turned out, electron microscope studies of “reconstituted” collagen fibrils
have proved much more informative.
Much earlier, Xageotte (1927) and others had found that collagen could
FIG. 14. Small-angle X-ray diffraction pattern of kangaroo tail tendon. Layer
line indices are indicated. Note the “fanning” seen in b and c. This is indicative of a
certain amount of chain distortion which accompanies drying; ( a ) moist specimen;
( b ) after soaking in water for 2 months, then drying under tension; (c) a f k r brief
exposure t o water, then drying under tension. (From Bear et al., 1951.)
be dissolved in dilute acid solution, and that, this soluble collagen could
then be reprecipit,at’edin fibrous form by appropriate manipulation of pH,
salt concentration, et,c. (see below). Furthermore, X-ray (Wyckoff and
Corey, 1936) and e l e c h n microscopic (Schmit,t el al., 1942) examination
revealed that many of the regularit,ics characteristic of native collagen had
reappeared in the reprecipita,ted or reconstituted material. These findings
have been greatly extended by Randall and co-workers (Randall et al.,
1952; Jackson and Randall, 1953; Randall et al., 1953a, 1955) and espe-
cially, in a classic series of papers, by Schmitt., Gross, and Highberger
(Highberger et al., 1950,19.51; Gross et al., 1952; Schmit,t et al., 1953; Gross
et al., 1954).
Schmitt arid his rollaborators found that by judicious manipulation of
the solvent environment they could reprecipitate soluble collagen in a t
least five different fibrous modifications. Briefly, they found that>by adding
increasing quantities of salt to a dilute acetic acid solution of soluble col-
lagen, t,hey could produce reconstituted collagens which showed : (I) the
iiormal -640 A polarized banding at 1 % NaCl, (2) an abnormal, -210 A
periodicity at 2 76 NaCl, and (3) no periodicity at 5 76 NaCl (see Fig. 15).
FIG.15. Schematic two-dimensional view of various modes of aggregation of rol-

lagen molecules in vilro, illustrating the generittion of the various periodirities. The
stagger responsible for the -210 A and the -6-20 A periods is not necessarily all in
adjoining molecules, as shown here. (From Schmitt, 1956.)
Furthermore, if instead of salt they added small amounts of acid glycopro-

tein (or a variety of other agents; see Gross et al., 1952; Randall et al.,
1952) the collagen precipitated in the so-called “fibrous long spacaing”
(FLS) form, with a nonpolarized repeating period of about 2600 A. The
addition of certain other agents (in particular, adenosine triphosphate)
under slightly different conditions induced the precipitation of collagen as
shorter, polarized segments having a total length of about 2600 A; the so-
called “segment long spacing” (SLS) form (see Figs. 1.5 and 13). Further-
more, these forms were shown to he completely interconvcrtiblc.
Schmitt and co-workers interpreted these results in terms of a hypotheti-
cal, polarized c*ollagen molecule which they called “tropocollagen,” and

which they deduced should be about 2600 A in length and 15-20 A in
diameter. Such long filamentous collagen molecules could then produce the
observed patterns by aligning themselves in parallel :in a quart er-staggered
arrangement to form the native collagen period, antiparallel with the ends
in register to form the FLS pattern, and parallel with the ends in register
to produce the SLS segments.” They assumed that the specificity respon-
sible for these discrete interactions must be “built-in” to the molecules via
specific and polarized arrangements of amino acids along the polypeptide
chains.
An independent demonstration of the general correctness of the “tropo-
collagen concept” came from the work of Boedtker and Doty (1956; see
below) who showed by physicochemical means that soluble collagen in acid
solution exists as long, relatively rigid, rod-shaped molecules having
a length of 2900 A, a diameter of 15 A, and a molecular weight of about
350,000. Shortly thereafter, as a consequence of improvements in elec-
tron microscope technique, Hall (see Hall, 1956; Hall and Doty, 1958) was
able to obtain electron micrographs of individual collagen molecules.
These confirmed directly the dimensions measured by Boedtker and Doty,
and deduced (for the hypothetical tropocollagen monomer) b y Schmitt
and coworkers.
The potent combination of chemical and enzymatic methods and electron
microscopy has been used in further studies of the structure and interactions
of the collagen molecule. In particular, this approach has been useful in
mapping the distribution of different classes of amino acid residues along
the molecule. As mentioned above, Bear (1952) originally suggested that
the phosphotungstic acid-staining bands observed in the electron micro-
scope represent amorphous regions containing primarily the bulky polar
residues of collagen; Kuhn et al. (1957) showed that these regions probably
take up phosphotungstic acid by relatively specific chemical combination
with basic amino acid residues. Hodge and Schmitt (1960) have carried
this analysis further using the SLS-type aggregates, in which all the like
features of a series of collagen molecules are arranged in accurate transverse
register. Taking advantage of this characteristic, they have been able to
use SLS band patterns as “molecular fingerprints” to localize various spe-
cifically bound, elect ron-dense compounds, and to deduce the packing
mechanisms which apply in other types of ordered collagen aggregates.
They examined and compared SLS aggregates stained with: (a) phospho-
tungstic acid, which reacts with the positively-charged lysine and arginine
residues and (b) uranyl ions, which bind to aspartic and glutamic acid
11 For more detailed reviews of these studies see Schmitt et al. (1955), Gross (1956),
and Schmitt (1959).

64 HARIZIKGTON AICD VON HIPPEL
residues. These two stains yielded SLS band patterns which differed sub-
stantially in the rclative intensities of comparable bands, but resembled
one another closely in terms of the relative positions of the bands along the
collagen molecule. Thus Hodge and Schmitt concluded, in agreement with
Bear, that both the acidic and basic polar residues are located in narrow
clusters separated by the nonpolar residue-containing interbands. The
enzymatic results of Grassmann and colleagues also tend to bear out this
interpretation (see Section 111).
Recently, Xishigai et al. (1960) treated SLS aggregates with collagenase
directly on the electron microscope grid. They found that collagenase
selectively digested away the interband regions, leaving the bands relatively
intact. Sirice it has becn established (sce Section 111) that collageriase spe-
cifically catalyzes the hydrolysis of the peptide bond between residues X
and Gly in scquenccs such as -Z.Pro.X.Gly.Pro.Y-, these results again
demonstrated that scqucnces of this sort arc confined primarily to the inter-
band regions, and further idcntifies these regions as the crystalline portions
of the collagen fibril.
R. Structural Studies in Solution
1. The Isolation of Soluble Collagen
Zachariades (1000a) was the first to show that collagen could be solubilizcd
using fairly mild techniques when he found that apprcciahle fractions of the
tail tendons of rats could be dissolvcd in dilute solutions of formic, acetic,
oxalic, hydrochloric, hydrobromic, sulfuric, and othrr acids. However, thc
significance of this finding did not become apparent until 1927, when
Kageotte showed t,hat acid extracts of soluble collagen could be reconsti-
tuted into fibers which microscopically resembled native collagen. This
similarity has been amply borne out by more recent X-ray diffraction and
electron microscope studies (see above). Since 1927, many workers have
demonstrated that collagen from a variety of tissues can be brought into
solution using dilute acids (e.g., see Leplat, 1933; FaurB-Fremiet and Gar-
rault, 1937; Orekhovich et al., 1948; Gallop, 1955a). Most of the recent
physicochemical work on soluble collagen has been done on collagen solu-
bilized by one of several variants of the citrate extraction procedure of
Orekhovich et al. (1948). Basically, this procedure involves extraction of
minced connective tissue with a dilute citrate buffer a t p H 3.5 to 4.0, fol-
lowed by removal of the insoluble residue, and dialysis of the extract against
tap water or dilute salt solutions. The dissolved collagen is reprecipitated
as needlelike crystals, which may be harvested and purified by several
repetitions of this cycle.
Ovcr the last few years it has been shown by several groups that collagen
can also he dissolved in a variety of neutral salt solutions (e.g., see Gross
et al., 1955a; ,Jackson and Fessler, 19.55; Gallop et al., 1957a) and in mild
alkali (Harkness et al., 1954) ; we will use the generic term “soluble collagen”
to refer t o the products of all these mild extraction procedures. A consider-
able literature has developed which deals with the differences between col-
lagens (and gelatins) extracted from the connective tissue in various ways;
this will be reviewed in Section V. For present purposes we need only point
out that, to a first approximation, the collagen molecules extracted by these
various procedures are essentially identical in terms of size, shape, chain
configuration, arid most other chemical and physicochemical properties
(e.g., see Gross, 1956; Orekhovich et al., 1937; Mazourov and Orekhovich,
1939, 1960; Jackson and Bentley, 1960).12
2. Physicochemical Studies of Soluble Collagen
a. Size and Shape of the Collagen Molecule. Since the existence of discrete
units of soluble collagen, capable of reconstituting larger scale native struc-
tures, had been apparent since the work of Nageotte, it is somewhat sur-
prising that physicochemical studies of soluble collagen were not undertaken
for so many years. Indeed, the first work along these lines was carried out
only about 12 years ago by Bresler, et al. (1950), on collagen extracted
from rat skin by the citrate method of Orekhovich. Bresler and co-workers
reported a molecular weight of about 70,000, and a particle length of about
380 A, on the basis of sedimentation and diffusion measurements. However,
the collagen used in these studies was dissolved a t 40°C, which subsequent
studies have shown brings about the total thermal denaturation of collagen
and its complete conversion to parent gelatin.13This work proved to be the
first of a series of physicochemical investigations of various soluble collagens
by several groups, including: M’Ewen and Pratt (1953), Mathews et al.
(1954), Noda (1955), Gallop (1955a), Orekhovich and Shpikiter (19554,
and Peng and Tsao (1956). The results obtained by these and subsequent
investigators are summarized in Table VIII. (Data obtained a t relatively
high temperatures, which therefore probahly relate to thermally denatured
collagens, have been omitted.) These groups obtained widely divergent
values of molecular weight for soluble collagen (see Table VIII) but were
generally agreed that the native molecule must he considerably larger and
certainly much more asymmetric than Bresler et al. had suggested. Also,
it soon became obvious (see cspecially Mathews et al., 1954; Gallop, 1955a
and Boedtker and Doty, 1956) that mild heating, or treatment with urea,
KSCN, etc., resulted in the conversion of these large, asymmetric particles
However, it is clear t h a t some of the more subtle interaction properties do vary
between soluble collagens extracted in various ways, and even between fractions of an
extract of a single type (e.g., see Fessler, 1960).
‘SThe term “parent gelatin,” coined by Statchard et al. (1944) t o designate the
ideal, undegraded, precursor gelatin molecule is defined in Section V.
HARIZINGTON ANI) V O N HIPPEL
TABLEV I I I
Physicochemical Properties of Soluble Collagen
Re-
Parameter Values Collagen and solvent fer-
:nce
Intrinsic viscosity, 15 Rabbitskin (citrate, p H 3) It

hl (100 ml/gm) 15 R a t tail tendon (acid) C
13.2 Ichthyocol (citrate, p H 3.7) a
10.5 Ratskin (citrate, p H 3.7) e
11.5 Ichthyocol (citrate, p H 3.7) f
12-16 Ichthyocol (neutral salts) h
13.5 Calfskin (citrate, p H 3.7) 3
13.7 Ratskin (citrate, p H 3.7) 1
13.2 Perch swim bladder (citrate, p H 1
3.7)
13.2 Cod swim bladder (citrate, p H 1
3.7)
15-17 Ichthyocol (neutral CaCl2) m
15 Calfskin (acetic acid) n
12.8 Codskin (citrate buffer, p H 3.5) P
Partial specific vol- 0.705 Ichthyocol (citrate, p H 3.7) a
ume, 0 (ml/gm)
Sedimentation coef- 3.5 Rat tail tendon (acetic acid) C
ficient s ~ O . l (Sved-
U 2.85 Ichthyocol (citrate, p H 3.7) a
berg units) 3.15 Ratskin (citrate, p H 3.7) e
3.28 Calfskin (citrate, pH 3.7) 3
3.0 Calfskin (acetic acid) n
3.17 Codskin (citrate, p H 3.5) P
Translational diffu- 0.33-0.5 Rattail tendon (acetic acid) C
sion coefficient, 0.35-0.4 Ratskin (citrnte, pH 3.7) e
Oh., (cm*/sec)
( x 101)
Refractive index in- 0.189-0.197 Ratskin (citrate, p H 3) and r a t a
crement, d n / d c tail tendon (acid)
(ml/gm) 0.192 Ichthyocol (citrate, p H 3.7) a
Molecular weight,
M (gms/mole) :
Sedimentation- 710,000 Rattail tendon (acetic acid) C
diffusion 700,000 Ratskin (citrate, p H 3.7) e
Sedimentation- 510,000 R a t tail tendon (acetic acid) C
viscosity 250,000 Ichthyocol (citrate, p H 3.7) f
352,000 Calfskin (citrate, p H 3.7) 3
280,000 Codskin (citrate, p H 3.5) P
Osmotic pressure 310,000 Ichthyocol (citrate, p H 3.7) f
400,000 Ratskin (--) 9
-
STRUCTURE OF COLLAGEN AND GELATIhT 67
TABLEVIII-Continued
-
Re-
Parameter Values Collagen and solvent fer-
ence
Light scattering 3.G-10 X 10' Itatskin (citrate, p H 3) a

10-25 X 106 R a t tail tendon (acid) a
1.67 X lo6 Ichtbyocol (citrate, p H 3.7) d
345,000 Ichthyocol (citrate, p H 3.7) f
360,000 Calfskin (citrate, p H 3.7) j
0.4-0.5 X lo6 Ichthyocol (citrate, p H 3.7) rrL
1.4 x 108 Ichthyocol (neutral CaC12) m
Flow birefringenct 350,000 Ichthyocol (citrate, p H 3.7) f
and Viscosity __
L D
~
Molecular length
and diameter, L
and I) (ang-
stronis) :
Hydrodynamic 3700-5200 18-22 R a t tail tendon (acetic acid) C
6000 - Ratskin (citrate, p H 3.7) e
2900 12-13. Ichthyocol (citrate, p H 3.7) f
3500 - Calfskin (citrate, p H 3.7) j
3000 - Calfskin (acetic acid) n
2810 12.1 Codskin (citrate, p H 3.5) P
Light,-scattering 87-~60x 103 12 Ratskin (citrate, p H 3) a
88 X lo3 16 R a t tail tendon (acid) a
13.4 x 103 - Ichthyocol (citrate, p H 3.7) d
3100 13 Ichthyocol (citrate, p H 3.7) f
3100 - Calfskin (citrate, pH 3.7) j
Electron 2820 15 Ichthyocol k
microscopy 3000 15 Calfskin n
~~
(1M'Ewen and P r a t t (1953). a Gallop et al. (1957a).

Mathews et al. (1954) 2 Doty and Nishihara (1958).
c Noda (1955). k Hall and Doty (1958).
d Gallop (1955a). 1 Burge and Hynes (1959a).
e Orekhovich and Shpikiter (1955a). m von Hippel et al. (1960).

f Boedtker and Doty (1956). Rice (1960).
g Peng, and Tsao (1956). p Young and Lorimer (19GO).
to considerably smaller, more flexihlc, subunits (gelatin, see below and

Section V).
Most of the quantitative discrepancies and contradictions arising from
these studies were resolved by the work of Boedtker and Doty (1956) who
carried out a very careful study of soluble ichthyocol collagen in pH 3.7
68 HARRINGTON AND VOX HIPPEL
citrate buffcr (isolated and purifird from carp swim bladdcrs using a procc-
dure developcd by Gallop, 195.5~~). Bordtker and Doty obtained by several
combinations of methods (including light scattering, osmotic presmre,
sedimentation, viscometry, and flow birrfringence) a molecular weight
(weight average) of 345,000 for the ichthyocol monomer in acid solution.
The molecule was found to be rodlike in shape, with a length of about
3000 A and a diameter of 13.6 A. These dimensions have since been con-
firmed by direct electron microscopic observation of single collagen mole-
cules (Hall and Doty, 1958; Rice, 1960). Also, as pointed out above, these
findings were particularly important in that they were in good agreement
with the dimensions deduced by Schmitt arid co-workers for the hypotheti-
cal tropocollagen monomer, and thus served to establish a bridge between
the studies of collagen in solution and in the solid state.
A careful examination of Table VIII reveals that thc chief discrcpancy
bctween thr results of Boedtker and Dotly and those of the earlier workcrs
lies in the light mattering values; the intrinsic viscosities and scdirnentation
coefficients dcterminrd by various groups differ remarkably little. This led
Boedtker and Doty to investigate thc light-scattering problem very care-
fully. They found that the difficulty seemed to be due to the presence of
small amounts of a large contaminating material-presumably large,
loosely bonded aggrcgatcs of collagen molecules-which could only be
removed by exhaustive, high speed crntrifugation at vcry low protein con-
centrations. Such aggrcgates would hc cxpected to have a great influence
on light-scattrring dctermiriations of particle weight and size, while a fYect-
ing measurements of [q]arid S ~ Ohardly
, ~ a t all. This seems to account for thc
high values of the earlier light-scattering results. Subsequent work (Doty
and Nishihara, 1958; Burge and Hynes, 195%; Young arid Ilorimer, 1960;
Rice, 1960) suggests that soluble collagen monomers derived from a variety
of other tissues are generally similar to ichthyocol in over-a11 size arid shape,
though frce monomers may br found only in acid solution. I n this connec-
tion, von Hippcl et al. (1960) have shown that in neutral 0.5 Ill CaClz,
ichthyocol owurs predominantly as small aggregates (approximately te-
tramers) .
Recent iriterpretat ions of wide-angle X-ray diffraction patterns have led
t o a three-chain structure for collagen, a t least in the ordered interband
regions. This strongly suggested that a three-stranded structure should
exist in the collagen monomer as well. Several lines of evidence support
this view:
(1) Thc average mas-to-length ratio ( M / L ) obtained for ichthyocol by
light scattering is 100 avograms per angstrom (Boedtker and Doty, 1956).
This ratio is in reasonable agrermcnt with the value of 98 avograms per
angstrom rrquircd by thc X-ray data (Bear, 1952).
( 2 ) The equatmial reflections of the wide-angle pattern suggest a center-

to-cent,er separation of 10 to 15 A (depending on the degree of hydration)
for adjacent collagen molecules. This compares favorably with the value
of 12 to 15 A derived from studies in solution (Table VIII).
(3) Complete deiiat,uration of ichthyocol, either by heating or by treat-
ment with neutral salts such as KSCN (see below) brings about a n approxi-
mat’ely threefold decrease in weight-average molecular weight, suggesting
that the three strands postulated for the native collagen molecule separate
under these conditions (Boedtker and Doty, 1956).
(4) Correlation of the kinetics of proteolysis of ichthyocol by the enzyme
collagenase with changes in various physical parameters, indicate that the
native molecule must be multistranded (von Hippel et al., 1960).
However, other data suggest that picturing the collagen monomer as a
structure composed of three equivalent polypept,ide chains twisted together
as required by the X-ray results, may be an oversimplification. For ex-
ample :
(1) The structure must somehow accommodate the various “unusual”
covalcnt bonds (e.g., hydroxylamine-sensitive “ester-type” bonds, y-car-
boxylglutamyl and r-aminolysyl peptide linkages, see Section V) which
have been report,ed in collagen and gelatin and which may possibly be in-
volved in interchain cross-linking.
(2) The actual molecular weight of the parent gelatins produced from
collagens by mild heating is generally not exactly one-third that of the
collagen monomer. For ichthyocol, Boedtker and Doty (1956) obtained
a weight-average molecular weight of 138,000, resulting in a molecular
weight ratio of collagen to gelatin of approximately 2.5. Earlier, Gallop
(1955b) had obtained a molecular weight of about 70,000 for parent
gelatin, yielding a collagen-gelatin ratio of about 5. Boedtker and Doty
postulated, on the basis of their result’s, that the collagen monomer con-
tains three chains of unequal weight. They suggested several possible ar-
rangements of such chains, including a staggered arrangement with a
single dangling chain protruding beyond the rigid three-stranded por-
tion of the molecule at either end. This suggestion has been adopted by
Hodge and co-workers in some of their recent studies of rcconstit’uted col-
lagen.
An important observation by Orekhovich and Shpikiter (1955b) seems
to have led to a partial explanation of the discrepancy between the molecu-
lar weights obtained for parent gelatin by Gallop and by Boedtker and
Doty. Orekhovich and Shpikit’er found that, parent gelatin derived from
rat skin contained two components of differing molecular weight. These
components, termed a and p, were subsequently observed in calfskin and
ichthyocol gelatin solutions as well (Chun and Doty, 1958; Nishihara and
70 IIAHltISGTON AND V O S HIPPEL
Doty, 1958; Orekhovivh arid (Lo-workerb, 1960; Orekhovich and Shpikiter,

lY.58b). Iloty and co-workers also found that the heavier, p-component ,
could be split by alkaline treatment or prolonged heating into units having
the sedimentation characteristics of the a-component. This suggested that
the @-componentis held together by a very labile bond(s), and that the
molecular weights of -70,000 were probably obtained on samples of parent
gelatin in which this bond had already been ruptured. This finding may
also account for the progressive decrease in the intrinsic viscosity of gelatin
which accompanies long standing at moderate temperatures (Boedtker and
Doty, 1956). The bond responsible for these changes is probably of the
ester type; it is definitely not a (a-amino) peptide bond, since cleavage
results in no release of ninhydrin-positive material (von Hippel, unpub-
lished material).
A more detailed discussion of the components of parent gelatin will be
deferred to the following section. The point to be made here is simply this:
that while the generally accepted, three-strand interpretation of the col-
lagen molecule is probably correct, the details of how the molecule is as-
sembled from its subunits remain obscure.
Additional information on the structure of the collagen molecule has
been derived from studies on sonicated collagens. Nishihara and Doty
(1958) found, from scdimcntatiori and viscosity measurements, that sonic
irradiation of soluble calfskin collagen (at a frequency of 9 kilocycles per
second) resulted in the progressive fragmentation of the molecules into
shorter segments, which, however, seemed to retain the compound-helical
structure and rigidity characteristic of the native molecule. Time-depend-
ence studies of the rate of molecular weight decrease suggested that clcav-
age occurred by nonrandom scission at three particular, rather evenly
spaced regions of the molecule. Nshihara and Doty pointcd out that these
susceptible regions seem to be separated by approximately 700 A, and
thus may bear some relation to the regions responsible for the -G40 A
spacing observed in phosphotungstie acid-stained samples of nativc. col-
lagen. This prefereiitial cleavage of the collagen monomer into halves and
quarters has been confirmed by Hodge and Schmitt (1958) using electron
microscopy.
Hodge and Schmitt (19.58) carried these ideas further. They demon-
strated that sonication did not impede the lateral packing of the degraded
molecules into an SLS-type pattern. On the other hand, the ability to
form end-to-end aggregates of the native type disappeared rapidly, even
before the molecular lengths had been appreciably altered. This suggested
that the ends of the collagen molecule must be particularly susceptible to
sonic damage, and that such damage destroys the normal md-to-end in-
teraction specificity of the molecules. On the basis of these results arid the
collagen model with dangling terminal peptidc chains proposed by Boedt-

ker and Doty (1956), Hodge and Schmitt postulated that formation of
the native-type pattern might involve a specific type of coiling of these
terminal chains about one another to form a highly ordered structure, and
that these terminal chains must therefore be particularly susceptible to
destruction by sonic irradiation.
In order to test this hypothesis, Hodge et al. (1960) attempted to iso-
late and characterize these terminal chains by treating intact collagen
molecules with trypsin. They found, as had been previously reported by
Gallop et al. (1957a), that the native collagen molecule is essentially im-
pervious to tryptic attack. However, at t,he temperature of their experi-
ments (room temperature), some small amount of proteolysis did occur,
which seemed to have no effect on the monomer lengt,h, but, like sonica-
tion, essentially abolished the ability of treated samples to reconstitute
native-type fibrils. This suggested that t,rypsin also attacks, primarily, the
end regions of the molecule, perhaps cleaving the postulated terminal pep-
tide chains. An acidic, tyrosine-containing peptide, which may perhaps be
part of such a structure, has been isolated by curtain electrophoresis from
such tryptic digests.
Certain additional, more indirect information on the structure and
properties of the collagen molecule may be derived from an examination
of the kinetics of the formation of fibrils from collagen solutions. In the
studies t o be discussed, fibril formation was initiated either by: (1) dilut-
ing an acidic or neutral salt solution of soluble collagen with low ionic
strength buffer adjusted to pH -7, or (2) increasing the temperature of a
neutral salt solution of collagen to 37°C. The latter procedure initiates the
curious phenomenon called “heat precipit,ation” of collagen. In both meth-
ods the development of fibrils was followed by measuring optical turbidity;
also b0t.h produce fibrils which show the native -640 A spacing in the
electron microscope. Gross (1956) and Gross and Kirk (1958) found, using
the heat precipitation t,echnique, that, specific elect,rolytes have a marked
effect on the rate of fibril formation. Bensusan and Hoyt (1958) came to
similar conclusions using procedure (1) to initiate fibril formation. Their
results indicated that, electrostatic factors must be operat’ive, but since
different ions a t the same ionic st,rength decreased the rate of fibril forma-
tion to markedly different extents, the effect is certainly not just one of
ionic strength. At constant ionic strength anions exhibited a wider varia-
tion than the cations tested. Bensusan and Hoyt, examined the effect, of a
a number of cat,ions on the electrophoretic mobility of soluble collagen,
and found evidence of specific ion binding.
Bensusan and Scanu (1960), in an examination of the effect of iodination
of collagen on the rate of fibril formation, found t,hat the rate increased
72 HARRINGTON A N D VON HIPPEL
markedly following iodination and that the only chemical consequence of

iodination seemed to be the conversion of tyrosyl residues to the diiodo-
tyrosyl form. These results again implicated the trace amounts of tyrosine
found in collagen in the interaction process (see Hodge et al., 1960). As a
consequence of these studies, Bensusan and Scanu suggested that the t)yro-
syl residue participates in the irit.eraction process in its ionized form, and
that the increase in rate on conversion to the diiodo form is due to the
decrease in p K which accompanies iodination (for tyrosine, pK =lo; for
diiodotyrosine, pK ~ 7 ) They. concluded that the ionization of 8 to 10
tryosyl residues is involved in the activation step; this is in reasonable
agreement with their analyt,ical finding of -12 tyrosines per 360,000 molec-
ular weight collagen unit. Ionized lysyl residues also appear to participate
in the int,eraction process. The electrostatic basis of the interaction
is furt,her supporkd by t.he finding that the rat.e of fibril format.ion
increases with decreasing dielectric constant of the medium (Bensusan,
1960) as well as with decreasing ionic strength. Very recent studies by
Martin et al. (1961) on the reconstitution of collagen fibrils as a function
of pH have suggested that an uncharged imidazole moiety may be crucial
in the alignment and binding of adjacent collagen molecules int,o fibrils
characterized by the -640 A axial periodicity.
Combining t.urhidometric rate measurements with electron microscopic
observat,ions, Wood and Keech (1960) have made extensive correlations of
fibril dimensioiis wit'h the ionic strength, pH, etc., at] which fibril forma-
tion takes place. Their kinetic findings are qualitatively similar tjo those of
Bensusari arid Hoyt, (1958), in that fibril formation under most conditions
showed a lag pcriod, followed by a sigmoid growth phase. I n a theoretical
analysis of the kinetic data, Wood (1960a) showed that the process of fiber
format,ion could he divided into a nucleation and a growth phase, corre-
sponding respectively to the observed lag and sigmoid growth periods.
Wood (1960b) has also examined the effects of chondroitin sulfat,e and
other naturally occurring polyanions on t,he rate of fibril formation; some
accelerated fihril formation, while others seemed to inhibit the process.
Basically, the data available are in accord with the postulat,e that specifi-
cally locat,ed charged groups on one molecule are attracted to specific
groups on another, and that most of these groups are probably located in
the phosphotungstic acid-staining bands of the collagen fibril.
b. Optical Rotatory Properties. In addition to its unique amino acid com-
position and wide-angle X-ray diffraction pattern, collagen is also chsrac-
terized by unusual optical rotatory properties. X-ray studies have shown
that the individual polypeptide chains of collagen in the solid state are
coiled into the left-handed, threefold helix characteristic of poly-L-proline
I1 (only marginally distorted by the superimposed, right-handed coiled-
coil structure). Physieochemical studies have suggested that this type of

organization carries through into the collagen molecule. Since it is now
well established that the formation of helical structures from random coil
precursors is accompanied by a large and characteristic change in optical
rotation, it is clear that a structure such as the poly-L-proline I1 helix
should exhibit definite and distinctive rotatory properties. And indeed,
as we have seen, this is the case.
In the last few years, attention has focused primarily on theoretical and
cxperimental optical rotatory studies of the a-helix and the @-structuresin
proteins and polypeptides (Ee article in this volume by Urnes and Doty).
(For recent reviews, see: Schellman and Schellman, 1958,1961; Blout, 1960;
Yang, 1961.) Collagen, when considered at all, has been quickly put aside
as a rather unpleasant exception to a number of otherwise generally applica-
ble empirical generalizations. Yet the rotatory properties of the poly-L-
proline II-type helices of collagen are at least as striking and as characteris-
t,ic as those of the a-helix. A brief comparative consideration of the rotatory
parameters of the two systems makes this clear:
(1) The rotary dispersion of collagen, like that of most of t,he a-helical
proteins, is of the simple type (over the region 400-700 mp) and may be
fit,ted to a one-term Drude equation:
where [a]is the specific rotation, X is wavelength, and A and A, are con-
stants characteristic of the system.14
(2) In the denatured (or essentially random coil) form, both collagen (as
gelatin) and the a-helix-forming proteins typically exhibit specific rotations
close to the mean residue rotation of the component amino acids ([a], N
-90" to - 120') and values of A, of -205 to 215 mp.
(3) However, in the native form the situation is quite different for the
two species. The a-helical proteins generally exhibit a lower specific levo-
rotation ([a], N -20" to -50") and an increased A, (generally to values
greater than 230 mp) relative to the denatured form (see the recent and
very extensive compilations of Schellman and Schellman, 1961). Collagen,
on the other hand, shows a vastly increased specific levorotation ([a],'V
-400") and an essentially unchanged A, (see Cohen, 1955; Harrington,
1958; Burge and Hynes, 1959a, b).
The optical rotatory parameters for some carefully studied collagens
(and the gelatins derived from them by mild heating) are summarized in
14 However, a few proteins with an exceptionally high a-helix content (e.g., some
of the myosins) and most of the synthetic polypeptides in the a-helical form, exhibit
anomolous dispersion.
Table IX; (further consideration of the rotatory properties of gelatin will

be deferred to Section V).
Certain ot,hcr features of the optical rotatory propertks :we worthy of
comment,. l h t , as mentioned previously and as might be expect,cd on
X-ray grounds, the structural portion of the specific rotatioil of collagcn is
the sodium D-line the specific rotation of collagen is -

very close to that obtained with poly-L-proline 11. Thus (see Table IX) a t
-400", while the
residue rotation is about - 125", yielding a configurational c o n t r i b u t h
TABLEIX
Optical Rotatory Properties of Collagen and Gelatin
Ichthyocol -350" 205 -110" 205 a

Calfskin -415" - 135" - 0
Rat tail -289 ' 217 -118" 217 c
Ichthyocol -330" 204 - - c
Bovine -350" 220 -146" 217 C
Codskin - 109.7' - d
Cod swim bladder -397" -116.0" - d
Herringskin -408" -124.5' d
Eelskin -398" - 125.7" d
Perch swim bladder -400" -127.5' d
Ratskin -409" -135.0' - d
Ichthyocol - - -124.8' 213 e
Dogfish sharkskin -345" 218 - 122" - f
Cohen (1955).
b Doty and Nishihara (1958).
Harrington (1958).
Burge and Hynes (1959b).
* von Hippel (unpublished).
f Lewis and Piez (1961).
of --275". The [a],for poly-L-proline I1 is more negative (---silo")

but the residue rotation of L-proline is also more negative (--250") re-
sulting in a structural rotation, [a10 of about -290" for this material
(Harrington and Sela, 1958). Furthermore, A, for poly-L-prolinc I1 is
-205 mp; very cose to the value obtained with collagen.
These findings suggest, of course, t,hat the polypeptide chain configura-
tions of poly-L-proline I1 and collagen are very similar, but also (unless
one is willing t o postulate some rather unlikely cancellation effects) that
practically all of the collagen peptide chains in the native molecule are in
this configuration. In this connection, it is intcrcsting to note that the
specific rotation measured for a number of collagens is essentially constant,
while that of the corresponding gelatins varies much more (see Table IX,
data of Burge and Hynes). Since the gelatin values reflect primarily the
amino acid compositions of the different samples, this also suggests that
the specific rotation per residue of the collagen helix does not depend pri-
marily on thc nature of the residues, but is a property of the helix itself
(Burge and Hynes, 1959a).
C. The Collagen -+ Gelatin Transition

Wc turn iiow t o a more detailed consideration of how the multistranded
collage11 molecule is held together and stabilized in the native state, by
examining the cverits which take place when this stabilization breaks
down during the collagen --+ gelatin transition.
1. Thermal Shrinkage Studies

The collagen -+ gelatin transition can be detected a t various lcvels of
structural organization. Thus the so-called thermal shrinkage phenomenon,
which has been recognized as a characteristic property of collagen for
many years (e.g., see Ewald, 1919) has recently been shown to be a macro-
scopic manifestation of the molecular collagen -+ gelatin transformation
(Garrett and Flory, 1956; Flory and Garrett, 1958). The thermal shrink-
age phenomenon may be observed by simply subjecting a bundle of eol-
lagen fibers to slow heating. At a specific temperature, T,, which differs
from one species of collagen to another (and may vary slightly with the
rate of hcatiiig; see Weir, 1949) these fiber bundles undergo a sharp con-
traction to less than one-third of their original length. After thermal
shrinkage, many of the characteristic properties of the native fiber are lost,
including the wide-angle X-ray diffraction pattern (Herzog and Gonell,
1925), the small-angle pattern (Bear, 1944; Wright and Wiederhorn, 1951),
arid the resistance to proteolysis by trypsin (Grassmann, 1936). The wide-
angle X-ray pattern can be at least partially restored by stretching the
shrunken fiber to its original length (Astbury and Atkin, 1933) but the
small angle pattern is not restored by such treatment (Bear, 1944). Macro-
scopically, thermal shrinkage is irreversible, though the initial introduction
of molecular cross-links between chains (e.g., by subjecting the fibers to
mild tanning) does result in a fiber which re-elongates spontaneously when
cooled and which retains the small-angle diffraction pattern (Bear, 1942,
1944). Studies of the crystallization of cross-linked polymers have shown
that such behavior is not uncommon, and indeed is what one might expect
on thermodynamic grounds (e.g., see Mandelkern, 1956).
The shrinkage temperature of collagen is very much affected by inter-
action with various small molecules, including: electrolytes and nonelec-
trolytes, acids and bases, tanning agents, etc., and much interesting ex-
perimental and theoretical work has been done in attempting to uiider-

stand these phenomena. Since the work in this area prior to 1956 has been
lucidly summarized by Gustavson (1956) we will not attempt to review
this material, but will only supplement Gustavson's account by consider-
ing certain more recent developments. In particular, two specific areas
will be discussed: (1) the correlation between shrinkage temperature and
imino acid content; and (2) the demonstration that thermal shrinkage and
TABLEX
Itelation between Imino Acid Content and Shrinkage and Denaturation
Temperatures jor Various Collagens
~
Pro- Hydroxy- Tqtal Kefer-

Collagen linea proline" imino
acidsa T~ - T ~ j ences
Calfskin 138 94 232 65°C 36°C 29°C d , e, f ,

Ratskin 130 93 223 36°C - 9
Perch swim bladder 118 81 199 55°C 31°C 24°C g, h
Pikeekin 129 70 199 55°C 27°C 28°C e , h, i
Ichthyocol 116 81 197 54°C 29°C 25°C d, e
Sharkskin 113 79 192 53°C 29°C 24°C g,j
Cod swim bladder 103 57 160 16°C - B
Codskin 102 53 155 40°C 16°C 24°C e, f, i
Dogfish s h a r h k i n 99 57 156 16°C - k
Residues of amino acid per 1000 total residues.
a
Shrinkage temperature of tissue strips.
b
Denaturation temperature in solution, measured by viscosity or optical rotatiori
0
at pH 3.7.
d Doty and Nishihara (1958).
Pier and Gross (1960).
Gustavson (1955b).
Burge and Hynes (1959a).
h Gustavson (1956).
Esipova (1957).
i Eastoe (1957).
k Lewis and Pies (1961).
the collagen 3 gelatin trarisformatioii are both manifestations of the

same molecular process. The latter finding, of course, permits us to carry
the pertinent aspects of the fiber shrinkage work over into the molecular
domain.
a. Shrinkage Temperature and Imino Acid Content. In the course of a
study of the thermal shrinkage behavior of a number of skin collagens
from different organisms, Gustavson (1953) discovered an interesting
correlation between shrinkage temperature and imino acid content. Specifi-
cally, he found that, T. seemed to vary directly with the hydroxyproline
content (see Table X). Therefore, assuming interchain hydrogen bonds to
be primarily responsible for stabilizing the collagen structure, and assum-

ing that these bonds rupture in the shrinkage process, Gustavson (1955b,
1956,1957) postulated that -OH. .O=C- hydrogen bonds, between the
hydroxyl group of hydroxyproline and the carbonyl oxygen of the pep-
tide bond, play a key role in the stabilization of collagen, and that it is
the rupture of these bonds which precipitates thermal shrinkage. This hy-
pothesis was strengthened by a large number of measurements carried out
by Takahashi (Takahashi and Tanaka, 1953; Takahashi and Gustavson,
1956) on the skin collagen of a variety of Japanese teleosts, where again
a good correlation between T , and hydroxyproline content was obtained.
Other evidence which also apparently favored this point of view has been
summarized by Gustavson (1956, 1957).*6
In arriving at this conclusion, Gustavson dismissed the apparently
equally good correlation between either proline or total imino acid content
and T , (see Table X) on the grounds that proline cannot participate as a
hydrogen donor in hydrogen bonding. Thus Gustavson felt that the pres-
ence of proline would, if anything, reduce rather than increase the stability
of collagen.
Somewhat later, on the basis of other evidence, it became apparent that
hydrogen bonding might not be the sole factor involved in stabilizing
collagen.
(1) Szent-Gyorgyi and Cohen (1957) suggested that the presence of
large amounts of proline (or hydroxyproline), which cannot be accommo-
dated in an a-helical structure, might predispose a polypeptide chain to
take up the alternative poly-L-proline II-type configuration.
(2) Harrington and Sela (1958) demonstrated that poly-L-proline exists
in solution as a single chain structure cast into the poly-L-proline I1 con-
figuration (see Section 11) indicating that this structure is stabilized by
restrictions to the rotation of the pyrrolidine ring about the peptide back-
bone. Harrington (1958) also suggested that the presence of contiguous
pyrrolidine ring-containing residues might affect the geometry of the poly-
peptide chain in proteins (particularly collagen) by similar mechanisms.
(3) In the course of an examination of the gelatin 4 collagen-fold tran-
sition, von Hippel and Harrington (1959) proposed that the poly-L-proline
II-type configuration develops along single gelatin chains as an intermedi-
ate step in the formation of the collagen-type structure. This suggested
quite specifically that restricted rotation about both the
lfi In particular, acetylation experiments seemed to lend strong support. Gustav-

son (1954) found that complete N-acetylation of bovine collagen does not affect the
thermal shrinkage temperature of the fibers. On the other hand, combined N - and
O-acetylation (blocking both amino and hydroxy groups) lowered T, by approxi-
mately 20°C.
0
4
c,-c
Bond (ii)
and the peptide bonds adjacent to pyrrolidine rings is involved in stabiliz-
ing the collagen-fold (see Section V).
(4)Burge and Hynes (1959a) compared the proline and hydroxyproline
content with the dilute solution denaturation temperature (see below) of
several collagens, and again found good correlation between proline, hy-
drxyproline, and/or total imino acid content and T , (see Table X),
t,hough the correlation with total imino acid content secmed the best of
the three.
(5) Piez and Gross (1960) reported very careful measurements of tho
amino acid composition of a great many collagens, and compared these
results with values of T , culled from the literature (see Tahle X). They
also found statistically significant correlations between T , and proline,
hydroxyproline, and t,ot,al imino acid content; the best correlation again
being that betwecn T , and total imino acid residues (see Piez, 1960). These
observations led Burge and Hynes, arid Piez and Gross to suggest that the
correlation between T 8 (and T,) and total pyrrolidine content is the sig-
nificant one. This conclusion, in conjunction with the considerations cited
above, suggests that, the stereochemical properties of pyrrolidine ring-
containing residues in the polypeptide chain environment, rather than
interchain hydrogen bonding, might be the key factor in stabilizing the
collagen structure.
b. Relation between the Thermal Shrinkage of Fibers and the Dilute Solu-
tion Transition. Both the macroscopic thermal shrinkage of bundles of
collagenous tissue, and t,he t>hermaldenaturation of the soluble collageu
derived from such t,issue, have been known and studied for many years.
However, unt,il fairly recently the connection between these two phenom-
ena has not been entirely clear. About *5 years ago the feeling became rela-
tively general that these two thermally induced alterations must both be
manifestatioiis of a hasically similar process, presumably involving thc
collapse of the rigid three-stranded collagen molecule. I n this view, fiber
shrinkage occurred at higher temperatures than the dilute solution transi-
t,ion because, in the solid state, both the stabilization provided by the
crystallization energy (the energy of interaction of the native collagen
molecules) and the intramolecu1a.r forces stabilizing the molecules them-
selves, must be overcome (Boedtker and Doty, 1956).
To investigate this idea, Esipova (1957) and Doty and Nishihara (1958)
examined the temperatme at which solutions of various types of collagen
were transformed to gelatin, and compared these measurements of T a
with thermal shrinkage temperatures obtained by Gustavson and others.
They found that in all classes for which data were available, the tempera-
ture differences ( T , - To)were essentially constant a t 27” f 3°C. This
strongly suggested that differences in thermal stability of the various col-
lagens depend on intramolecular processes rather than intermolecular
interactions. And this conclusion, in turn, seemed to make possible a clear
choice between collagen structures I and 11.
One of the major differences between structures I and I1 (see Table
VII) lies in the orientation of the hydroxyl group of hydroxyproline rela-
tive t,o the rest of the molecule. In structure I, this group can form hydro-
gen bonds wit,h reccptor groups within the same three-stranded collagen
molecule, while in structure I1 these OH groups are oriented away from
t8hemolecule mid can only part,icipate in int,ermolecular hydrogen bonding.
Thus, accepting Gustavson’s explanation for the correlation between hy-
droxyproline and T, , Esipova, and Doty and Nishihara, concluded that
their results provided substantial support for the collagen I structure since
differences in hydroxyproline content only seemed to affect the magnitude
of the intramolecular stabilization. On the other hand, the more recent
views on the role of the imino acids in stabilizing the collagen structure
undercut the basis of this argument (Burge and Hynes, 1959a).
In their treatment of this problem, Garrett and Flory (1956; Flory and
Garrett, 1958) accepted the point of view of some of the earlier workers
(e.g., Wohlisch, 1932; Kuntzel, 1937) that the thermal shrinkage process
could be treated as the “melting” of crystalline arrangements of polypep-
tide chains, and suggested, in agreement, with the view advanced by Boedt-
ker and Doty (1956), that the dilute solution transition represents this
melting process stripped of its intermolecular component. Flory and Gar-
rett made a very careful study of the melting of dried samples of bovine
achilles tendon and rat tail tendon as a function of collagen concentration,
utilizing measurements on samplcs with collagen volume fractions, v 2 , in
et>hyleiwglycol ranging from 0.84 to 0.0004. Some of their results, plotted
as melting temperature versus v 2 , are shown in Fig. 16. The measurements
at the higher concentrations were made using either dilatometry or direct,
observation with a polarizing microscope.1G At the lowest concentratioii
(point 10 in Fig. 16) viscometry was used to determine the melting tem-
perature. As Fig. 16 shows, the transition temperature varies monotoni-
cally with volume fraction over the entire range, and the experimental
pointjs fall almost exactly on the theoretical curve calculated by substitut-
ing the parameters characteristic of collage11 into the usual polymer melting
point, c1quat.ion. (See Flory, 1953; Mandelkerii, 1956; Flory, 1961.) These
l6 In the dilatometric measurements, the temperature of the latent volume change
due to fusion was measured, while melting evidenced itself in the polarizing micro-
scope as the disappearance of the depolarization due to the birefringence of collagen
fibers.
results, when combined with a demonstration of the reversibility of this

transition, proved conclusively that both thermal shrinkage in the solid
state and the collagen ---f gelatin transformation in dilute solution are mani-
festations of a first-order crystalline -+ amorphous phase transition. (Stud-
ies of the reverse process in dilute solution, the gelatin 4 collagen transi-
tion, are treated in detail in Section V.)
q.
FIG.16. Melting temperature plotted as a function of composition for collagen-
ethylene glycol mixtures: Solid curve is calculated on the basis of a normal polymer
melting relation (see Flory, 1953). (From Flory and Garrett, 1958. Reproduced with
kind permission of the American Chemical Society.)
2. Dilute Solution Studies of the Collagen -+ Gelatin Transition

As mentioned previously, a rather dramatic alteration in physicochcmi-
cal properties takes place over a narrow temperature interval when a solu-
tion of soluble collagen is subjected to gradual heating. This process may
be conveniently observed using viscometry or optical rotation, since both
the specific viscosity and the specific rotation change enormously as a
consequence of the collagen -+ gelatin transition. Thus the intrinsic vis-
cosity [q] falls from the 12-15 dl/gm characteristic of soluble collagen (see
Table VIII) to about 0.4 dl/gm., while the specific rotation [a],changes
from approximately -400" to --125" (Table IX). Figure 17 shows typi-
cal phase transition profiles for three different collagens monitored visco-
metrically, while Fig. 18 demonstrates that the changes observed visco-

metrically are equivalent to those measured using optical rotation.
Examination of Fig. 17 reveals that the temperature interval over which
the transition takes place has a finite breadth, and thus may be charac-
terized by various temperatures defined in different ways. T o , the tem-
perature quoted in Table X, is defined as the temperature of the midpoint
of the transition. If one assumes: (1) that the fractional change in intrinsic
viscosity (or specific rotation) is proportional to the fractional conversion
of collagen from a single completely native state to a single completely de-
natured state, and (2) that equilibrium between these two forms is at-
tained at every temperature, then the relation T o = AH/AS applies and
these thermodynamic parameters can be calculated from data such as
TEMPERATURE OF 30 MINUTE HEATING ("GI
FIG.17. The collagen -+ gelatin transition for various collagens, measured vis-
cometrically. (From Doty and Nishihara, 1958.)
that of Fig. 17. This has been done by Boedtker and Doty (1956) for
ichthyocol, and by Burge and Hynes (1959a) for various other collagens.
However, subsequent measurement of the rate at which equilibrium is
attained at temperatures in the transition region (Harrington and von
Hippel, 1961) showed that times up to 24 hr may be required to reach the
final value at a given point. Since Boedtker and Doty, and Burge and
Hynes constructed their transition curves by waiting only 30 min at each
temperature, complete equilibrium at intermediate temperatures may not
have been attained. This leads to artificially sharpened transitions and
elevated values of T , . This difficulty, coupled with the possibility that
condition (1) above may also not apply (see Flory and Weaver, 1960)
suggests that calculations of AH and A S made in this way should not be
too rigorously interpreted. These considerations also indicate that probably
tjhe more significant parameter for characterizing phase transitions of this
sort is TW,, defined as the temperature at which the most ordered seg-
ments of the crystalline structure melt (see voii Hippel and Harrington,
1960), although this temperature is often more difficult to measure accu-
rately.
The temperature at which a givcn soluble collagen undergoes the col-
lagen -+ gelatin conversion (defined either as T , or T,,J is a usoful param-
eter for identification and characterization purposes. However, the transi-
tion temperature is constant only when measured in relatively dilute salt
3
TIME (MINUTESt
FIG-.18. Comparison between the rate of the collagen -+ gelatin transition for
soluble calfskin collagen a t 35.9"C, measured as the fractionaI change in specific
viscosity (solid line) and specific rotation (dotted line). (From Doty and Nishihar:t,
1958.)
solutions. Concentrated solutions of neutral salts, such as LiBr and KCNS,

have a marked depressant effect on TD (e.g., see Bocdtker and Doty,
1956; Harrington, 1958). Competitive hydrogen-bonding agents, such as
concentrated solutions of urea and guanidine-HCl, also lower the transi-
tion temperature markedly (see von Hippel and Harrington, 1960); T n
also is lowered slightly a t pH values close to 1 (Burge and Hynes, 1959a).I7
The mechanisms whereby some of these diverse agents exert their effects
17 The extensive and presumably closely related studies on the effects of various
electrolytes, urea, and pH on the thermal shrinkage of collagen in the solid state have
been comprehensively reviewed by Gustavson (1956).
8TltUCTURE OF COLLAGEN AND GELATIN 83
on collagen arid gelatin will be discussed in Section V. Anticipating this

discussion, it appears that the neutral salts may exert their effects through
a modification of solvent-polypeptide chain interaction, though binding of
ions to polar groups may also play a role. Urea and guanidine-HCI doubt-
less rupture hydrogen bonds, perhaps between water and carbonyl oxy-
gen groups (von Hippel and Harrington, 1960; Harrington and von Hippel,
1961). Low pH probably lowers T D by increasing intramolecular electro-
static repulsion (Burge and Hynes, 1959a).
D. The Use of Proteolytic Enzymes in Structural Studies o j Collagen
l’rotcolytic enzymes have been used extensively in the study of protein
structure. First, of course, purified proteases have been employed to cata-
lyze the cleavage of specific peptide bonds in proteins in the processof
establishing amino acid sequences and distributions. Within the last few
years it has also become apparent that proteolytic enzymes can be utilized
to obtain a t least semiquantitative information about protein (and nucleic
acid) configuration (or secondary-tertiary structure) as well as about amino
acid sequence. This use of proteases can be divided into two general areas:
(1) Physicochemical examination of the changes in protein (substrate)
structure which accompany progressive proteolytic degradation: here the
enzyme is used to rupture specific bonds a t known rates, and the molecu-
lar consequences of this action are studied.
(2) Analysis of the kinetics of proteolysis of a protein substrate as a
measure of steric or configurational features in the vicinity of suceptible
pept.ide bonds: in t,his approach the enzyme molecule itself is used as a
“configurational probe” to examine local aspects of substrate structure.
Both approaches, if properly applied in favorable situations, seem capa-
able of providing molecular insight of a different sort from that obtained
via the usual physicochemical methods, and both will be discussed below
as they have been applied in the study of collagen.
1. Physicochemical Studies of Collagen during Proteolysis
This approach, monitoring the changes in a macromolecular substrate as
a function of the number of enzymatic breaks in the polypeptide (or poly-
nucleotide) backbone, was first employed by Thomas (1956) and Schu-
maker et al. (1956) to prove that the native deoxyribonucleic acid (DNA)
molecule is composed of two chains, and to measure the strength and
distribution of the interchain hydrogen bonds which hold the two chains
together. I n these studies, Thomas related changes in viscosity and light
scattering to the number of enzymatic breaks, while Schumaker et al.
focused primarily on the viscometric changes. More recently, Sinsheimer
(1959) has used a similar procedure to show that the DNA isolated from
the 6x174 virus is single-stranded. This technique has been applied to

soluble collagen by von Hippel et al. (1960), who followed the changes in
specific viscosity, light-scattering molecular weights and radii of gyration,
nondialyzable protein and optical rotation which accompany the degrada-
tion of native ichthyocol by the collagenase isolated from Clostridium his-
to1yticum.
Von Hippel et al. found by light scattering that neither the molecular
weight nor the radius of gyration of the ichthyocol molecule changed
markedly during collagenolytic degradation at 5°C. Moreover, though
an,the molecular weight per free N-terminal residue, eventuslly fell to
-500, very little of the protein became dialyzable if both the enzymatic
treatent and the dialysis were conducted at low temperature. (On the
other hand, most of the protein did become very rapidly dialyzable if
dialysis was conducted a t temperatures above T D.) These results are con-
sistent with the current view of collagen as a three-stranded molecule
held together by interchain hydrogen bonds, and permitted a crude esti-
mate of the number of intact interchain hydrogen bonds needed to hold a
cleaved peptide fragment to the rest of the molecule a t low temperatures.
Apparently less than ten are required.18 Concurrent optical rotatory stud-
ies showed that the molecule not only retains most of its integrity in terms
of mass, but also that the intact interchain hydrogen bonds largely main-
tain the helical configuration of collagen despite very extensive backbone
cleavage (see Fig. 19).
These results were contrasted with viscometric studies, which showed
that the specific viscosity of soluble ichthyocol collagen decreases linearly
and without an initial lag period when plotted as a function of p , thc
probability that a given susceptible peptide bond has been split. If one
assumes (see Schumaker et al.) that: (1) the substrate is a rigid n-stranded
polymer held together by relatively weak interchain links; (2) the enzyme
attacks susceptible bonds of single strands a t random; (3) changes in the
specific viscosity are detected only when the molecule is split into two
smaller n-stranded pieces by cleavage of all the chains at loci which arc
sufficiently close so that the intervening interchain links cannot hold the
chains together; then the following equation applies:
where (vSp,t/Tlsp .o) is the ratio of the specific viscosity at time t to that at
time zero, n is the number of strands in the molecule, p is the probability
l* It is interesting to note, in this connection, that Thomas, and Schumaker et al.
found that only a short sequence of intact interchain hydrogen bonds (of the order
of 3-5)seemed to be needed to prevent the dissociation of the two chains of DNA a t
room temperature.
of bond cleavage defined above, and K is an arbitrary constant. Under

these conditions, it is clear that qep can be a linear function of p only if
n = 1; that is, if the efficiency of an enzymatic break with respect to its
effect on qBpis a constant from the start of the reaction. Yet this result is
143 ‘C
20-
-
xt
xo
10 -
08 -
06 -
04 -
01I
20
0 40 60 80 100 I20 140
TIME (MIN )
FIG.19. Comparison of the fractional decrease in specific viscosity and specific
rotation as a function of time after adding collagenase to soluble ichthyocol collagen
at 14.3”C. (From von Hippel et al., 1960. Reproduced with kind permission of the
American Chemical Society.)
incompatible with the three-chain collagen molecule and the observed in-
sensitivity of the particle mass to enzymatic attack, unless assumption (3)
is invalid and each backbone cleavage increases the molecular flexibility
(and thus decreases qap) regardless of whether the molecule separates into
two parts or not. These findings suggest that the rigidity of the collagen
molecule depends upon the “intactness” of all three polypeptide chains,
and that the super-helix, because of its extremely large pitch, does not con-
tribute very much to the molecular rigidity. These results also clarify the
basis of the viscometric assay for collagenase introduced by Gallop et aE.
(1957b), in which log qap is plotted against time for soluble ichthyocol in
the presence of collagenase, and the slope of the resulting straight line
is related directly to the activity of the enzyme preparation.
The progressive decrease in the specific viscosity of a solution of ich-
thyocol subjected to collagenolytic attack is compared to the accompany-
ing fall in specific rotation in Fig. 19. This shows strikingly, as pointed out
above, that the increase in molecular flexibility due to enzymatic cleavage
is not accompanied by an equivalent destruction of the compound collagen
helix. It is interesting to contrast this plot with Fig. 18, which demon-
strates that the specific viscosit,y and the rotation fall together when the
rigid, three-stranded collagen molecule is converted to random-coil gelatin.
2. Proteolytic Enzymes as “Probes” of Collagen Structure
The use of proteolytic enzymes as “configurational probes” grows out, of
the very old observation that, a native protein is often partially or even
completely resistant to prot>eolysisby an enzyme which readily attacks
the denatured form (e.g., see Linderst>r@m-Lang, 1952). Since denaturation,
by definition, does not alter covalent bonding or amino acid sequence,
configurational (steric) factors must &her prevent the enzyme from
reaching the susceptible bond in the native molecule, or prevent it from
orienting over the susceptible bond in such a way as to successfully cata-
lyze cleavage. Such behavior was observed by Grassmann (1936), for solid
collagen fibers with respect to tryptic hydrolysis; trypsin would not attack
native collagen fibers but easily degraded thermally shrunken specimens.
The analogous observation in solution was made by Gallop et al. (1957a)
who showed that native soluble collagen is not attacked by trypsin, but
that after conversion to gelatin it is readily digested. These qualitative
findings suggested that analysis of the kinetics of proteolysis might yield
quantitative data on polypeptide chain configurations in certain favorable
cases.
Studies of the tryptic hydrolysis of myosin and of the collagenolyt,ic
degradation of collagen seemed to confirm this view, and led Harringt~on
et al. (1959) to propose the use of proteolytic enzymes as probes of the
secondary structure of fibrous proteins, specifically as a measure of the
crystalline and amorphous (in an X-ray diffraction sense) regions along
the polypeptide chains. Subsequent work has supported this crystalline-
amorphous interpretation for the myosin-trypsin study, though in the
collagen-gelatin-collagenase system the situation appears to be more com-
plex.
Mihalyi and Harrington (1959) found that the digestion of rabbit myo-
sin, followed by pH-stat methods, did not obey kinetics of an integral

order. However, further investigation revealed that the kinetics could be
fitted nicely by assuming that the over-all reaction consists of two in-
dependent apparent first-order reactions proceedings] a t markedly differ-
ing rates.lg An investigation of the kinetics of proteolysis of ichthyocol
collagen (von Hippel et aE., 1960) and cold ichthyocol gelatin (von Hippel
and Harrington, 1959) by collagenase revealed similar behavior; in each
case the kinetics could be reduced to the sum of two concurrent apparent
first-order reactions. A typical example of the type of data obtained with
thc collagen-collagrnase system, illustrating the analysis into two reac-
tioils, is presented in Fig. 20a, where the fraction of the total bonds cleaved
is plotted as a function of time. Extrapolation of each linear segment back
to the ordinate gives the fraction of the total bonds split in that reaction,
and the apparent first-order rate constant for each reaction can be derived
from the slopes of the corresponding lines. These findings support the view
that the collagenase-sensitive (or trypsin-sensitive, in the myosin-trypsin
system) peptide bonds of ichthyocol collagen and cold gelatin can be
divided into two general reaction classes in terms of their susceptibility to
enzymatic attack.
Reactions run at elevated trmperatures demonstrated that the differ-
ences between the two classes of suseeptiblr bonds are primarily configura-
tional rather than chemical, since heating myosin to 41°C resulted in the
transfer of some of the trypsin-sensitive peptide bonds from the slower to
the faster reaction class (Harrington et at., 1959). Even more dramatic
was the finding that heating collagen to temperatures above the collagen
-+ gelatin transition temperature (27°C for ichthyocol) resulted in all the
bonds being split in a single apparent first-order reaction (Fig. 20b) differ-
ing in terms of thermodynamic parameters from either of the reactions ob-
served with native collagen (see Table XI). This seemed to imply that the
differences between the susceptible bonds, at least in the collagen-collagcn-
ase-gelatin system, are entirely configurational in origin.
I n Fig. 21, the logarithms of the apparent first-order rate constants ob-
tained from a series of pH-stat runs on ichthyocol collagen a t various
temperatures are plotted against the reciprocal of the absolute tempera-
ture as an Arrhenius plot. Clearly the data fall onto three straight lines:
those numbered (2) and (3) for the two reactions a t temperatures below
T , , line (1) for the single reaction at temperatures above T D. The appar-
ent enthalpy ( A H * ) , free energy (AF*), and entropy (As*) of activation,
together with the fraction of the susceptible bonds split in each reaction,
are compiled in Table XI.
Similar results were obtained by Connell (19GO) in an examination of the kinetics
of tryptie digestion of cod myosin.
Examination of Table XI and Fig. 21 brings out several interesting

points:
(1) There is a sharp break in the Arrhenius plot a t the collagen -+ gela-
tin conversion temperature; this is also true in the reverse reaction (see
Section V).
COLLAGENASE ON COLLAGEN
pH 8 0
1955 C
'
-
pm-pt
$03
I
I TIME (MINI
Fro. 20a. Fraction of the susceptible bonds of soluble ichthyocol collagen cleaved
by collagenase, as a function of time at 19.5OC. (a), Experimental pointjs; (e),fast
reaction (calculated by subtracting the extrapolated slow reaction from the experi-
mental data. (From von Hippel et al., 1960.)
(2) Above T D, all the collagenase-susceptible bonds appear to be split

in a single apparent first-order reaction, with AH* and AS* (-15 kcal/
mole and -0 e.u., respectively) close to the values usually associated with
the proteolysis of denatured proteins and synthetic substrates.
(3) The corresponding parameters for the two reactions in collagen and
cold gelatin are much larger than those for the reaction above T , .
(4) Approximately the same distribution of susceptible bonds between
the fast and the slow reaction obtains at all temperatures in both native
collagen and cold gelatin.
Points (1) and (2) above suggest, in agreement with physicochemical
evidence, that above T, gelatin exists as an open, essentially random coil
structure with the susceptible peptide bonds optimally available to the
COLLAGENASE ON GELATIN
1.0
0.9
00
0.7
06
b
0.5
04
b-Pt
pal
0.3
0.2
I 1 I I L
01 I I
10 20 30 40 50 60 70
TIME (MIN.)
FIG.206. Fraction of susceptible bonds of ichthyocol gelatin cleaved by colls-
genase, as a function of time at 37.35"C. (From von Hippel and Harrington, 1959.)
enzyme. Below T D , the bonds in both collagen and gelatin (which under-
goes a reversion to the collagen-type structure, see Section V) are not as
available to the enzyme. Hence we may speculate that since collagenase is
a highly specific enzyme (catalyzing only the cleavage of the sequence
Z.Pro.X.Gly.Pr0.Y) t,hat the rate of catalysis may be strongly dependent
on the orientation of the two required penultimate pyrrolidine rings rela-
tive to the polypeptide chain. Above T o , rotation of the pyrrolidine rings
90 HAllHINGTON AND VON HIPPYL
about the polypeptide chain should be relatively unhindered, and the

most favorable orientation for catalysis easily attained. However below
T o , where the chains as a whole exist in the poly-L-proline 11-type con-
TABLEX I
TherrrLodynaniic Data, Collagenase on Ichlhyocol Collagen and Gelatin"
AH* AF*
Substrate %
' (T)
!?- Bonds
(%)
(i%$ (kcal/
mole) mole)
(kcal/
mole)
_- -
Collagen (fast reaction) 10-23 10 f 3 1-42 $41 $15 +90 f 5

Collagen (slow reaction) 10-23 84 f 4 +47 +46 $15 4-110 f6
Gelatin 28-37 100 +15 $14 $14 $1 f7
Gelatin (fast reaction) 10-25 18 f 4 +23 +22 +14 $30 f4
Gelatin (slow reaction) 10-25 82 f 4 $30 +29 +15 $51 f4
0 Data from von Hippel and Harrington (1959) arid von Hippel et al. (19GO).
Oo0t COLLAGENASE ON COLLAGEN
I
3.20 3.30 3 40 3.50 3 60
+(x1031
FIG.21. Arrhenius plots for the proteolysis of ichthyocol collagen arid gelati11
by collagenase. pH-stat data: (1) gelatin above T D ; (2) collagen below T D , fast
reaction; (3) collagen below T D , slow reaction. (From von Hippel et al., 1960.)
figuration, the rate of catalysis might depend strongly on the relative

orientation of the pyrrolidine rings and 011 the nature of residue X.
We may assume that the enzyme distributes the susceptible bonds into
the two observcd reaction classes 011 the basis of some setluence-determiiicd
configurational basis (see Section V). Then the uniformly higher AH*
(about +20 kcal/mole) and AS* (about +

60 e.u.) found for each class in
native collagen relative to cold gelatin, implies that these differences may
constitute a measure of the “tighter” folding of the polypeptide chains in
the collagen molecule superimposed on the effect of local configuration
differences.20
E. The Role of Water in the Collagen Structure

Several types of evidence suggest that water is intimately involved in
t,he struct>ureand stabilization of collagen. For example, Gustavson (1956),
in reviewing the thermal shrinkage phenomenon, emphasized that the sub-
stitution of other solvents for water markedly influences the shrinkage
temperature and other properties of the fiber. The molecular consequences
of the hydration of collagen have been most extensively and revealingly
st,udied by X-ray diffraction techniques.
It has been known for many years that the characteristic wide-angle
X-ray pattern of collagen or cold gelakin is largely destroyed b y dehydra-
t,ion, and at least partially restored by rewetting the material. This sug-
gest,ed that water might be involved in stabilizing the collagen structure,
and led Rougvie and Bear (1953) to undertake a quantitative study of the
wrious features of both the wide- and small-angle diffraction patterns of
colla,gen (kangaroo tail tendon) which are affected by moisture content.
Rougvie and Bear constructed a moisture sorption isotherm (Fig. 22)
and found that the primary sorption amounted to -13.0 gm of water per
100 gm of collagen when analyzed by the standard B.E.T. method (Brun-
auer et al., 1938). Correlating moisture sorpt,ion with X-ray changes, they
found that the hydration-sensitive equatorial spacing observed in t,he wide-
angle X-ray pattern shifted from -10.6 A for dry fibers to a maximum
value of 14.6 A in completely wet specimens, and that the small-angle
meridional macro-period increased from approximately 600 A to -670 A
with increasing hydration. Accompanying changes in line intensities in the
small-angle pattern also indicated that improved ordering accompanied
the increase in moisture content (see Fig. 14). Rougvie and Bear concluded
t,hat the total moisture sorption range could be divided into four successive
intervals (in order of increasing water content) :
(1) Hydration of the polar side chains of the residues located a t the dis-
ordered fibrillar band regions.
(2) Hydration of polar groups in the interbands, accompanied by some
* O Very recently, Orekhovich e t al. (1960) have also reported marked differences
between the rate a t which collagen is degraded by collagenase in the native and in the
denatured state. They feel t h a t their results support t h e “local configurational
changes involving proline residues” interpretation of these phenomena offered by
von Hippel and Harrington (1959).
92 HARRINOTON AND VON HIPPEL
lateral chain separation as evidenced by the increase in wide-angle equa-

torial spacing.
(3) Straightening of kinked chains and further lateral separation.
(4) The attainment of apparently complete lateral separation of the
chains (or chain bundles, see below) at both the bands and interbands.
Several years later, Borge et al. (1958) reconsidered some of the hydra-
tion data of Rougvie and Bear from the point of view of the stereochemis-
Relative humidlty (%)
FIQ.22. Moisture sorption isotherm (25°C) for kangaroo tail tendon. The locrt-
tions a, b, c, d, and e divide the data into the four successive intervals discussed in
the text. (From Rougvie and Bertr, 1953.)
try of the collagen I1 structure. They pointed out that since only one-third
of the carbonyl oxygens and peptide nitrogens of this structure are in-
volved in hydrogen bonding within the three-chain collagen unit, a con-
siderable number of polar groups are left unbonded. Specifically, they
noted that if one assigns one water molecule t o each of the unbonded pep-
tide carbonyl and amide nitrogen groups, plus one to each polar side
chain, a total of 19.7 gm of water per 100 gm of protein should be taken
up in the primary reaction, compared to the 13.0 gm measured by Rougvie
and Bear. However, they pointed out that this experimental value is
close to that calculated for one water molecule per polar side chain plus
one for every two unbonded carbonyl groups (13.3 gm). Examination of
the three-chain collagen I1 model (incorporating either the Gly.Pro.Hypro
or the G1y.Pro.X repeat) showed that water molecules could be systemati-
cally placed so as to form hydrogen bonds with two carbonyl oxygens si-
multaneously in at least two ways: (1) by bridging the carbonyl oxygen of
the hydroxyprolyl (or X) residue and that of the adjacent glycyl residue
on the Same chain; or (2) by bridging the glycyl carbonyl oxygen and that
of the adjacent hydroxyprolyl (or X) residue of the next chain in the clock-
wise direction (viewed from the 6-terminal end of the collagen I1 model).
Burge et al. felt that neither arrangement was entirely stereochemically
satisfactory, and therefore did not pursue these possibilities, but concen-
trated their attention on the effects of singly-bonded water on the collagen
wide-angle pattern.2l
To this end, Bradbury et al. (1958) calculated the diffraction pattern
expected from collagen I1 with water molecules singly bonded in every
possible systematic position along the chains. This amounted to approxi-
mately 25 gm of water per 100 gm of collagen. The patterns calculated on
this basis were compared to those obtained experimentally, and the agree-
ment seemed somewhat better than that between the experimental results
and the transforms calculated for the anhydrous collagen I1 structure,
though the agreement was not good enough to definitely exclude other
hydration arrangements. It did seem clear, however, that systematically
disposed water molecules constitute an important portion of the diffract-
ing structure.
An X-ray study of collagen hydration has also been carried out recently
by Esipova et al. (1958). These workers measured moisture sorption iso-
therms and attempted to relate the results to intensity changes in the
wide-angle diffraction pattern, in order to establish the position of the ab-
sorbed water with respect to the diffracting portions of the collagen fiber.
They also found that hydration increased the crystallinity of the collagen
fiber. More specifically they inferred that the oxygen atoms of the water
bound to the ordered portions of the structure seemed to lie very close to
the axis of the polypeptide chains (within -3 A) and to be arranged in a
semiregular fashion along the chains (-3 A apart). Moreover, the stoichi-
ometry seemed to suggest that nonhydrogen-bonded peptide carbonyl
groups were primarily involved. On the basis of these results and the stereo-
chemical findings of Burge et al., Esipova and co-workers suggested that
the crystalline portions of the collagen structure might be sta.bilized by
21 Recent comparative studies on the rate of formation of the collagen-fold in H20
and D20 seem to support the possibility of intrachain water bridges linking adjacent
carbonyl oxygens. These studies are discussed in Section V.
94 HAHBINGTON AND VON HIPPEL
doubly hydrogcri-bonded water bridges of the type :
giving rise to continuous chains of structurally incorporated water along

the fiber axis in the diffracting regions. Similarly hydrated structures have
been proposed for poly-(glycyl-L-proline) by Millionova and Andreeva
(1958).
Infrared measurements have provided additional information about
hydration arid hydrogen bonding in collagen. Specifically, Bradbury et al.
(1958) examined the rate of deuteration of films of native ratskin collagen,
and found three groups of labile protons differing markedly in ratc of
exchange. The most rapidly exchangeable group was attributed to partially
degraded portions of the specimen, the second group t o labile protons on
side chains and N-H groups not involved in N-H. . .O=C-- hydrogen-
bonding, and the vcry slowly exchanging protons were assigned to N-H
groups involved in interchain hydrogen bonding. A value of 1:1.3 was
determined for the ratio of the number of N-H groups involved in inter-
chain hydrogen bonding to those bonded directly to water, in reasonable
accord with expectations based on the collagen I1 model. All groups showed
essentially instantaneous deuteration in heat-denatured (gelatinized)
specimens.
Bradbury and co-workers also employed infrared techniques to obtain
a moisture sorption isotherm for ratskin collagen. Their results are very
similar to those obtained by ltougvie and Bear for kangaroo tail tendon
(Fig. 22) ; however, they attributed the lateral separation accompanying
increased hydration to the separation of adjacerit three-stranded collagen
units, rather than to the separation of individual chains as suggested by
Rougvie and Bear. Bradbury and co-workers also concluded that water
plays a major role in the stabilization of both the intermolecular and the
intramolecular collagen structure.
Another investigation of hydrated collagen has been conducted by F’raser
and MacRae (1959) who concluded on the basis of infrared dichroic mms-
urements that the bound water molecules are primarily singly bonded to
the -C=O groups which project radially outward from the collagen
molecules. The bound water molecules seemed to be preferentially oriented
normal to the fiber axis.
These studies all point to water as an important, indeed, an integral,
component of the collagen structure. It is expected that future work will
establish unequivocally the precise arrangement of the water molecules
within the collagen framework.
STRUCTURE O F COLLAGEN AND GELATIN
V. THE STRUCTURE
OF GELATIN
The dissociation of the polypeptide chains of collagen by thermal or

chemical proeessrs leads to products variously termed gelatin. In thr pres-
ent section we propose to consider certain of the physicochemical properties
of gelatin, looking toward a deeper understanding of the structure and
chemistry of collagen.
For many years gelatin has been utilized in a large number of industries,
and thus for economic reasons the procedures used for extraction have
been based on the attainment of maximum yield. Since it has been of great
interest, from the industrial point of view, to learn more about the prop-
erties of these gelatins, the majority of the published studies in this area
have been carried out on such samples. Because of the diverse and rather
rigorous nature of the extraction procedures it is not surprising that a
rather broad spectrum of molecular properties has been reported. On the
other hand, it will be apparent from the discussion in Section IV that even
the mildest gelatinization procedures may not lead to a system of com-
pletely uniform polypeptide chains. Evidence is accumulating which sug-
gests that the component chains of the intact, fundamental collagen unit
may not be identical, either with respect to size or amino acid sequence.
Ideally, then, before undertaking detailed studies, the subunit chains
should be dissociated by mild procedures and resolved either chromato-
graphically or by some equivalent analytical technique. Undoubtedly, this
approach will be used a great deal in future studies and we may confidently
expect fruitful results.
However, despite the expected heterogeneity, studies of gelatin systems
have made possible a much deeper insight into the molecular properties of
collagen and, indeed, have made substantial contributions to other areas
of polymer and protein chemistry.
Two processes are commonly used in the commercial preparation of
gelatin. In the first process, hide or demineralized bone is extracted over a
prolonged period of time in alkaline solution. During this steeping opera-
tion hydrolytic changes occur, leading to a release of collagenous material
which is subsequently gelatinized at neutral pH a t temperatures of 60"-
65°C. In the alkaline medium used for extraction, the amide groups of the
glutamine and asparagine residues are released, resulting in a gelatin with
an isoelectric point at about p H 4.9-5.0 (Bowes and Kenton, 1948; Ames,
1952; Kenchington and Ward, 1954). Additionally, a significant number of
peptide linkages are split in the alkali pretreatment, as demonstrated by
the appearance of N-terminal residues (Bowes and Moss, 1953; Deasy,
1958; Courts, 1954, 1958, 1960).
The second process involves soaking skin, bone, or tendon in dilute acid,
followed by extraction with warm water at an acid pH. In this instance the
gelatin is not deamidated (Ames, 1957; Kuntzel et al., 1958; Veis et al.,
1958; Courts, 1960), and an isoelectric point of about pH 9 is obtained
for the resulting product. Generally speaking, acid-extracted gelatin ex-
hibits fewer N-terminal residues per unit weight than does alkali-processed
gelatin, but it is clear from the work of Courts (1960) that some degrada-
tion of the gelatin chains occurs during the steeping operation.
Gelatins may also be derived from soluble collagen preparations by
treatment with urea, sodium thiocyanate, lithium bromide, or by heating
to temperatures above T D . Since these collagen molecules have a very
uniform size distribution, it is to be expected that the derived gelatiiis
would exhibit the most meaningful physicochemical propertips. From the
pure protein-chemical point of view, it is a pity that so many of the in-
vestigations of gelatin have been carried out using the relatively degraded
acid- or base-processed material.
A . The Molecular Properties of Gelatin at Rocm Temperature and Above

1. Size and Shape
a. Gelatins Derived from Insoluble Collagens. In 1944, Scatchard et al.
published a short report on the molecular properties of a soluble, com-
mercial, alkali-processed gelatin derived from bone collagen. In this study
it was assumed that collagen consists of long polypeptide chains, and that
in the preparation of gelatin the bonds along each chain are hydrolyzed
a t random a t about the same rate. A few bonds, equally spaced along the
chain, were assumed to be hydrolyzed much more rapidly. According to
this view, there should exist an ideal parent gelatin molecule, defined as the
chain segment between two of the easily hydrolyzable bonds. Coupling tho
number-average molecular weight a,, (osmotic pressure) and weight-aver-
age molecular weight, a,,, (sedimentation equilibrium)) and using the
Montroll-Simha theory (1940) to predict the size distribution to be expected
from random degradation, Scatchard et al. (1944) estimated the molecular
weight of the parent gelatin chain to be about 110,000. In a more recent,
very thorough study, Boedtker and Doty (1954) found (light scattering)
of a similarly prepared gelatin (Knox, P20) to be 97,000. Calculation also
revealed the root-mean-square end-to-end distance (P)l/* = 258 A. Since
synthetic polyisobutylene with the same number of chain atoms gives
( P 2 ) ” 2 = 231 A, Boedtker and Doty proposed that the mean configuration
of these gelatin chains approaches that of a typical random chain polymer.
Further work in this direction was reported by Williams et al. (1954), and
Williams (1958) who found an M,,, of 95,000 for the same type of gelatin
on the basis of sedimentation equilibrium data. From a combination of
molecular weight and sedimentation coefficient distributions, a plot of

~ 2 0 versus
, ~ M;" was found to be linear, and therefore consistent with
random chain behavior. Gouinlock et al. (1955), arrived a t similar conclu-
sions through an analysis of viscosity, sedimentation, and light-scattering
measurements of various fractionated, base-processed gelatins.
The solution behavior of acid-precursor gelatins deviates quite markedly
from the pattern established for their base-processed analogs (Veis and
Cohen, 1956, 1957; Veis et al., 1958). At the same molecular weight, acid-
precursor gelatins have a lower intrinsic viscosity than do alkali-precursor
gelatins. Moreover they exhibit only diffuse changes in light scattering or
in electrophoretic mobility through the isoelectric pH region, whereas
base-processed gelatins exhibit sharply defined changes over this range.
Although the two materials differ in that base-processed gelatin is devoid
of amide groups, simple removal of amide nitrogen does not bring the
properties into coincidence. Evidently the differences are more fundamental
than this. Veis and Cohen believe that acid-precursor gelatins are lateral
aggregates of chains held together by occasional cross-links, since this type
of compact architecture would be expected to reflect a lower intrinsic
viscosity and exhibit less sharply defined configurational changes with pH
and ionic strength than the random chain, alkali gelatins.
The possibility of chain branching in the molecular structure of gelatin
has been investigated by a number of laboratories, following the early
suggestions of Mosimann and Signer (1944) and Ames (1952). Courts
(1954) measured the a-amino groups of various unfractionated commercial
gelatins and estimated i@% to lie between 50,000 and 70,000, in good agrec-
ment with osmotic pressure determinations of Pouradier and Venet (1950).
But the later work of Courts and Stainsby (1958) and Courts (1959) indi-
cated that higher molecular weight ox-bone and calfskin gelatins, both
alkali- and acid-processed, might well be multichain structures. Thus Mw
(light scattering) and [q] (intrinsic viscosity) were virtually independent of
the number-average chain weight, M,z, as determined by free a-amino
groups. Moreover, Ma for low molecular weight,, acid-processed gelatins
was found to be larger than M w , indicating the presence of some chains
without terminal a-amino groups. Pouradier (1958) has reported a similar
lack of equivalence in M m and the measured carboxyl groups in a base-
processed calfskin gelatin.
The preceding evidence, suggesting that both acid-processed and base-
processed mammalian gelatins are multichain structures, is entirely con-
sistent with the findings of Grassmann discussed in Section 111 and thus it
is pertinent, a t this point, to consider the various types of interchain
covalent bonds which might be involved in forming cross-linkages in
mature, collagenous tissue.
98 HASRINGTON AND VON HIPPEL
Since collagen is essentially devoid of cystine or cystcine residues, somc

type of linkage other than the ubiquitous disulfide bridge must be iii-
voked. Bowes and Kenton (1948) found that both the Van Slyke amino
nitrogen and the number of titratable amino groups of ox-hide gelatin were
lower than expected on the basis of the total lysine content,s and postulatcd
the existence of peptide bonds involving t-amino groups. This type of
linkage has also been favored by Mechanic and Levy (1959), based upon
the isolat,ion of a small amount of the tripeptide N'-(glycyl-a-glutamyl)
lysine following mild acid hydrolysis of bovine achilles tendon. Levy et al.
(1960) have inferred from their studies on ichthyocol that as much as 40 % of
the lysine of this collagen may be involved in N'-lysine bridges, although the
recent report of Betheil and Gallop (1960) that 94% of tJhe total lysirie
groups in ichthyocol can be guanidinated seems to contradict this conclu-
sion.
Ester bonds, imide bonds, and bonds involving side-chain carboxyls
of aspartic and glutamic acid have also been proposed as potential branch-
ing sites by a number of investigators. Thus, the possibility of interchain
imide bonds of the R-CO-NH-COR' type has been considered by Ames
(1944). Gustavson (1955a) has postulated the existence of ester linkages in
gelatin on the basis of an analysis of the free and amidated carboxyl groups.
Evidence of a more direct nature has come from the work of Grassmann et al.
(1954) on the citrat,e-soluble collagens of calfskin and Konno and Altmaii
(1958) on rat muscle collagen. They find that a small number of covalent,
bonds, presumably ester linkages, are reductively cleaved by treatment wit>h
lithium borohydride. Gallop et al. (1959) and Bello (1960) have shown that,
various gelatins react with hydroxylaminc (pH 9, 40°C) to yield a product
containing protein-bound hydroxamate. Gallop and co-workers showed
that t,he molecular weight of ichthyocol, calfskin, and acid-processed gela-
tins was lowered to approximately 20,000 as a consequence of this treat-
ment, without breaking peptide bonds. They have suggested that the
hydroxylamine labile bonds are of the ester type and may serve to link
together polypeptide segments of similar molecular weight. Some of these
hydroxylamine-sensitive ester links may involve the tightly bound carbo-
hydrate moiet'ies discussed in Section I11 (Hormann, 1960).
Recently, the presence of y-carboxylglutamyl peptide linkages has also
been reported in gelatin. Gallop et al. (1960) esterified a basc-processed
commercial gelatin and after cwnvcrsion t20 t>hc protein hydroxamate
followed by dinit~rophenylstion,hcat'ed the system in alkali t,o promot,e
Lossen rearrangement. of the aspartyl and glutamyl groups. E'rom a n
analysis of the products of acid hydrolysis it was inferred that about 31 of
the initial glutamic acid residues were involved in y-glutamyl peptide
linkages in the esterified protein.
BTRUCTURE OF COLLAGEN AND GELATIX 99
It is probable that differences in maturity of extracted collagens account

for a part of the reported variation in the structure of gelatin molecules.
The available evidence favors the view that collagen extracted with cold
neutral salt solution (tropocollagen) is the most recently formed protein
(Highberger et al., 1951; Gross el al., 1955a;Jackson and Fessler, 1955); that
collagen extracted with dilute, acidic citrate buffers (procollagen) is a some-
what more advanced morphological form (Orekhovich, 19.52) ; while
insoluble collagen represents the most advanced biogenetic stage, which
has become sufficiently cross-linked to resist solution. In this view, the
degree of cross-linking increaes with the age of the tissue, and the more
drastic gelatinizing processes required to solubilize the morphologically
advanced collagen structures involve cleavage of cross-linking peptide bonds
as well as the splitting of a few bonds along the backbone chain.
Orekhovich and Shpikiter (1958a; Orekhovich, 1952) observed that C14-
labeled glycine was more rapidly incorporated into citrate-soluble collagen
than into insoluble collagen, leading them to suggest that the individual
procollagen molecules are the precursors of the mature, cross-linked
connective tissue lattice. However, in later work Harkness et al. (1954)
reported a very rapid incorporation of C'4-glycine into that fraction of
rabbitskin collagen which is extracted with mildly alkaline buffer. The
rate of incorporation in to citrate-extracted collagen was appreciably lower,
and that into insoluble collagen negligible. Jackson (1956, 1957, 1958)
observed a similar sequence in his study of the C14-glycine incorporation
into the carrageenin granulomata of guinea pigs.
The chronological pattern of development of the insoluble, connective
tissue lattice from the soluble collagen monomers is also evident from
studies of the solubility of various collagen extracts as a function of salt
Concentration. It has been clear for a number of years from the work of
Gross and his associates (Gross et al., 195Fja; Gross, 1958) that the soluble
collagens can be divided into classes on the basis of the salt concentration
used in their extraction. More recently, Jackson and Bentley (1960) have
demonstrated that the amount of collagen which can be extracted from
the induced carageenin granulomata of guinea pigs increases with increaisng
NaCl concentration. This finding, coupled with the observation that
the most extractable collagen fraction also incorporates C14-glycine most
rapidly, led them to propose that a continuous spectrum of molecular
aggregates, increasingly strongly cross-linked, exists in the native connec-
tive tissue. Although it has been assumed that cross-linking in soluble colla-
gens involves only secondary bonds (Jackson and Bentley 1960), it seems
likely that covalent cross-links may well account for the insolubility of the
more mature, collagenous tissue.
The presence of covalent cross-linkages in the neutral-salt and citrate-
100 HARRING'I'ON AND VON HfPPEL
buffer insoluble collagens is particularly emphasized in the recent studies

of Veis et al. (1960), who have examined the sedimentation patterns of
such materials after bringing them into solution a t a temperature of 60°C.
These solutions exhibit heterodisperse Ultracentrifuge patterns with a t
least three major components. The two faster sedimenting components,
which have apparent molecular weights of 4 and 12 X 106, are not reduced
in weight or amount by hydrogen-bond competing reagents or by acidic
or basic environment at temperatures which normally destroy gelatin
aggregates. Since the weight of the collagen monomer is thought to be
around 350,000, it seems clear that some type of covalent linkages exists
between the collagen monomers in the intact fiber structure.
Most of the evidence for cross-linking in gelatin comes from work on
mammalian material. Fish collagens are in general more readily soluble
in dilute salts, suggesting a lower order of covalent bridges between the
collagen monomers. This point should be kept in mind when comparing the
physicochemical behavior of various gelatins.
0. Gelatins Derived from Soluble Collagens. In Section IV we discussed
the denaturation of the asymmetric, rodlike collagen molecules, demon-
strating that the viscosity and optical levorotation undergo a precipitous
decline at a relatively low and characteristic transition temperature in the
presence of certain neutral salts or the hydrogen-bond competing reagents
urea and guanidine-HCl. Because of the mild conditions used in such
denaturation procedures, it is likely that only secondary linkages are
cleaved and we would expect the multichain collagen structure to be dis-
sociated, under controlled conditions, into its component polypeptide
chains; providing only that these chains are not held together by covalent,
interchain cross-linkages.
In Section IV we referred briefly to the sedimentation studies of Orek-
hovick and Shpikiter (1955a) on the parent gelatin derived from ratskin
procollagen, pointing out that a two-component sedimentation pattern is
observed when this material is examined in the ultracentrifuge. Qualita-
tively similar bimodal sedimentation patterns have also been reported for
the parent gelatins prepared from the soluble collagens of the swim bladder
of the carp (Chun and Doty, 1958), calfskin (Doty and Nishihara, 1958;
Pize, et al., 1960) and codskin (Doty and Nishihara, 1958). The relative
amounts of the two components appear to vary somewhat, depending on the
species, but in general the sedimentation coefficients are quite comparable
when measured in identical solvent systems (Table XII).
At present only preliminary physicochemical studies have been carried
out on the two components. Orekhovich and Shpikiter (1958b) were able
to effect a separation of the components from ratskin procollagen by am-
monium sulfate fractionation of urea solutions. The isolated fractions each
exhibited a single boundary in the ultracentrifuge and diffusion experiments

yielded values of D20,w= 3.7 X lo-’ cm2/sec and 2.7 X 10-7 cmz/sec.
When coupled with sedimentation coefficients (sZo , w ) molecular weights of
80,000 and 130,000 were calculated for the slower sedimenting (a) com-
ponent and the faster sedimenting (a)
component respectively.
The molecular weights of the two components of ichthyocol and calfskin
gelatin have been estimated from their respective sedimentation coefficients
TABLEXI1
Sedimentation Coeficients of the a-and @-Componentso j Vurious
Soluble Collagens
Collagen source Solvent
(1) Calfskin 0.15 M Citrate-1.2 3.85s 5.45s -1 b

M KSCN, pH 3.7
0.15 M Acetate, pH 3.24 S 4.41 S -1 C
4.8
(2) Ratskin Phosphate-1 M 4.0 S 5.75 S -1 d
KSCN, pH 8
(3) Carp swim 0.15 M Citrate, pH 3.31 S - >>I e
bladder 3.7
0.15 M Citrate, pH 3.15 S 4.3 S >1 f
3.7
(4) Dogfish 0.15 M Citrate, pH 3.42 S 5.03 S -1 g
sharkskin 3.75
Ratio of areas under a- and p-peaks [not corrected for Johnston-Ogston effect
(194G)1.
* Doty and Nishihara (1958).
0 Piez et al. (1960). Ratio obtained from amino acid data.
d Orekhovich and Shpikiter (1958a). Determined by comparison with known mix-

tures.
e Gallop (1955b).
1 Chun and Doty (1958).
Q Lewis and Piez (1961). Ratio is corrected for the Johnston-Ogston effect.
and the relation between szo,w and aW proposed by Williams, et al. (1954) for
a commercial gelatin. Molecular weights of 80,000and 160,000for ichthyocol
(Chun and Doty, 1958) and 120,000 and 230,000 for calfskin (Doty and
Nishihara, 1958) were obtained in this way. I n the studies of Doty and
co-workers, as well as those of Orekhovich and Shpikiter, aqueous KSCN
solutions have been employed as solvents in order to minimize association
reactions between the gelatin chains. This solvent system gives consistently
higher sedimentation coefficients for the a- and 0-components than do simple
acetate or citrate buffers (Piez et al., 1960). When the molecular weight,s of
102 HAHHINGTON AND YON HIPPEL
the calfskin components are estimated from the sedimentation coef€icients

in citratr buffer using the Williams-Saunders-Cicerelli relation, molecular
weights of 80,000 and 160,000 are obtained (Piez et al., 1960).
The amounts of the a- and @-componentsobtained are about equal in
the parent gelatins derived from the soluble collagens of ratskin, calf-
skin, and codskin, but in ichthyocol the slower sedimentating a-peak is
appreciably larger than the &peak. Moreover, as pointed out in Section IV,
the relative distribution of these components can be altered by raising the
pH or by prolongrd treatment at a temperature of about 40°C. Under these
ronditions, the 0-peak of all of these species disappears and the a-peak
iitcreasrs in area, suggesting that the @-component is composed of two
polypeptide units similar in size to the a-component, these being held to-
gether by a thermal- or alkali-labile bond (Doty and Nishihara, 1958).
Differences in the stability of this bond(s) in the different species could
account for differences in the a-@distribution.
It seems clear that a part of the heterogeneity with respect to size
reported for commercially preparrd gelatins must be attributed to both the
molecular weight and mass distribution of the a- and @-components.Thus,
assuming the weight of the a-component to be 80,000, that of the 0-com-
ponent to be 160,000, and that dissociation of the collagen molecule yields
two a-components and one 0-component as suggested by several authors
(Orekhovich and Shpikiter, 1958s; Boedtker and Doty, 1955; Piez et al.,
1960), the number average molecular weight, an, of the mixture is 107,000
and the weight average molecular weight, aw, is 120,000.
There seems little doubt that the amino acid compositions of the a- atid
@-components differ significantly. Orekhovich and Shpikiter (195%) hy-
drolyzed the a- and @-fractionsderived from ratskin and demonstratrd, using
paper chromatography, that the a-component was richer in hydroxyproline
than the 0-component. Similar differences in hydroxyproline content were
found by Chun and Doty (1958) for the components of ichthyocol following
ethanol-water fractionation. The most complete study of these components
to date is that of Piez et al. (1960) who separated the a- and @-components
of calfskin parent gelatin on carboxymethyl (CM) cellulose columns.
Subsequent amiiio acid analyses of each showed appreciable differences
between the two components: the 0-component contained more than twice
as much histidine, about half as much tyrosine, and 50% more hydroxy-
lysine, leucine, isoleucine, and valine than the a-component. Proline, hy-
droxyproline and certain other amino acids also differed by values ranging
between 2 and 10 %. The amino acid analysis of the undenatured, purified
calfskin collagen was found to be identical with the mean composition of
the separated componrnts, indicating, in agreement with the sedimentation
patterns, that equal weights of a and 0 are present in the collagen monomer.
STRUCTURE OF COLLAGEN AND GELATXN 103
Very recent studies on the soluble collagen of ratskin have clarified con-
siderably the relatioilship between the a- and @-compoiients. Orekhovich
et al. (1960) and I'iez et al. (1961) have rcported that neutral-salt-soluble
collagen gives prcdomiiiaritly the a-component on denaturation, with only
a small amount of @-compoiient.Chromatography (I'iez et al., 1961) of this
material at 40°C in the preseiice of 6 M urea on CM-cellulose columns re-
veals the presence of two components, denoted a l and a2, with differing
amino acid composition. On the other hand, sedimentation analyses of acid
extracted (3% acetic acid) ratskin collagen gave two peaks in the ultra-
centrifuge with sedimentation coefficients characteristic of the a- and P-
components while chromatography of this material (40°C or 6 M urea)
revealed four major components. Two of these, a1 and a2, appeared to be
the same as the salt-extracted material as judged by sedimentation and
chromatographic behavior. Of the remaining two components, one, denoted
01, had a composition equivalent to an equal mixture of a1 and a2, whereas
the composition of the other component, 02, was identlical to that of a l .
Thus the amino acid analysis, when taken in conjunction with the sedi-
mentation properties of these components, suggests that the a-component]
is a mixture of two polypeptide chains of about the same mass but differing
composition, while the @-componentconsists of a mixture of two different
types of cross-linked chain pairs, i.e., al-a1 and al-a2.
Essentially similar conclusions have been reached by Grassmann et al.
(1961) on the basis of an analysis of sedimentation patterns of denatured,
acid-extracted calfskin collagen. They propose that the formation of the
@-componentarises from cross-linking of two a components which may or
may not be identical in amino acid composition, while a third component
always present in small amount, the y component, represents a structure
in which three a-components are cross-linked together.
2. Optical Rotatory Properties
a. Effect of Chain Weight and Composition. At room temperature and
above, the optical rotatory characteristics of gelatin are not much in-
fluenced by variation in chain weight. Ferry and Eldridge (1949) demon-
strated the specific rotation of a series of ossein gelatins in the molecular
weight range (aw) 33 t o 72 X lo3 to be virtually invariant (at [a1646 =
- 165") above a temperature of 30°C. Similar results have been reported by
Saunders and Ward (1958a) for higher molecular weight, fractionated ox-
hide gelatins above 40°C. A comparison of the rotatory properties of a wide
variety of gelatins should consequently be essentially independent of any
chain degradation resulting from isolation procedures.
The influence of composition can best be evaluated from the careful study
of Rurge and Hynes (1959b) who measured the specific rotation of a num-
104 HARRINaTON A N D VON HTPPEL
ber of gelatins a t 40°C following the thermal disruption of the respective

collagens. It will be seen from Table XI11 that the specific rotation ([LY];')
for these gelatins varies between -109.7' and -135", the increase in
levorotatiori paralleling the increase in imino acid content. Assuming a
random configuration for the gelatin chain, we would expect the specific
rotation t o be obtained by summing over the specific residue rotations of
the complete ensemble of amino acids. The residue rotation, [Rlpro,of pro-
line was shown to be close to -250" in Section 11; the residue rotation of
glycine, [RIcl, = 0. The residue rotation of L-hydroxyproline is not as well
TABLE XI11
Estimation of the Configurational Contribution to Optical Rotation f o r Various Gelatins
at 40°C
HY- Gly- Refer-
Gelatin source Prolinea droxy- cinea ences
proline
___
Calfskin 138 94 320 - 140" -27"
Rntskin 130 93 351 - 135" -27"
Perch swim bladder 118 81 333 -127.5" -21"
Dogfish sharkskin 99.4 57.G 340 - 120" -19"
Carp swim bladder 116 81 325 - 124" - 16"
Cod swim bladder 103 57 333 -116" -14"
Codskin 99.4 53 345 -110" -10"
a Residues of amino acid per 1000 total residues.
6 Assuming mean residue weight is 91.
c Piez and Gross (1960).
d Doty and Nishihara (1958).
0 Harrington (1958).
I Burge and Hynes (1959a).

Burge and Hynes (1959b).
h Lewis and Piez (1961).
i von Hippel (unpiiblished).
established as that of L-proline. However, the specific rotation of poly-L-

hydroxyproline in concentrated aqueous LiBr is - 168" and, reasoning by
analogy with the studies on poly-L-proline I1 in this solvent, we would
expect the structural contribution of the chain to be eliminated. Correcting
this specific rotation for the residue weight of hydroxyproline gives
[R]Hypro= - 190". Heynes and Legler (1960) have recently estimated the
residue rotation of hydroxyproline from the specific rotation of the tri-
peptide carbobensoxy-Ala.Gly.Hypro-NHz at -222". In the calculation to
follow we will assume R,,,,, = -200". I n view of the absence or trivial
presence of other amino acids with unusual residue rotations, the mean
residue rotation of all other amino acids in gelatin may be taken as - 110"
(Schellman and Schellman, 1958) and we may proceed to an estimation of

the specific rotation based on the composition.
'The specific rotation of a long polypeptidc chain which is locked into an
asymmetric configuration may be expressed as :
where [Ri]is the intrinsic residue rotation of the ith amino acid residue;
MRW, the mean residue weight; n, the number of residues in the chain; and
o,configuration , the
[a] structural contribution of the chain.
Calculation of the expected specific rotations from Eq. (9) and the known
amino acid compositions of the gelatins in Table XIII, demonstrate that a
significant amount of left-handed configuration remains in the polypeptide
chain of each species, even at temperatures of 40°C. Harrington (1958)
attributed this residual optical asymmetry to the existence of elements of a
poly-L-proline II-type configuration imposed by the presence of proline-
proline sequences. When two imino acid residues occur contigfuously in a
polypeptide chain, the orientation of nine backbone bonds along the chain
is fixed in the left-handed poly-L-proline II-t,ype configuration assuming
restricted rotation about, the
0
//
c.-c
Bond (ii)
(ii) bonds (see von Hippel and Harrington, 1959). These poly-L-proline I1
"nuclei" may be of special significance in the mutarotation of gelatin at low
temperature (see below).
b. E$ect of Salts. It has been known for many years that a number of
neutral salts have a rather profound effect on the optical rotation of gelatin
(Stiasny et al., 1925; Katz and Wienhoven, 1933; Carpenter, 1938). In
their studies, Carpenter and Lovelace (Carpenter, 1927; Carpenter and
Kucera, 1931; Carpenter and Lovelace, 1935a, b, 1936,1937) examined the
effect of a wide spectrum of electrolytes at concentrations up to 4 M , and
observed that the specific rotation, [a]: of calfskin gelatin decreased from
- 134.5' to values ranging between - 90" and - loo", the magnitude of the
change depending on the nature of the salt. The capacity of ions t o change
the specific rotation followed a Hofmeister series, with lithium cations and
thiocyanate anions exhibitng the most striking effects. Under these condi-
tions of varying electrolyte environment, the optical rotatory dispersion of
gelatin follows a one-term Drude equation over the visible region of the
spectrum, with an unchanging Drude dispersion parameter, A, = 220 mp.
At high concentrations of aqueous lithium bromide (8-12 M ) , may
10G HARRINGTON AND VON HIPPEL
decrease to values as low as -35”, and careful dispersion measurement’s

reveal an accompanying decrease in A, ranging between 2 and 22 m p ,
depending on tbe gelat,in species (Harrington, 1958). Alt,hough thc dath
are not extensivc, the magnitude of these changes appears to parallel the
imino acid content of the gelatins measured, being greatest for calfskin and
least for ichthyocol gelatin.
Neutral salts have been shown to have striking effects on other poly-
peptide chains as well. The rotatory parameters of oxidized ribonucleasr
and clupein, which are known to exist as unfolded chains in aqueous solu-
tion, change to those characteristic of a-helical proteins in high conceiitra-
tions of LiBr (Harrington and Schellman, 1957). These observations led tJo
the suggestion that LiRr, and other similar salts with unusually high activ-
ity coefficients, attenuate the disruptive effect of water on hydrogen-bonded
peptide structures by lowering the water activity.
The findings of Harrington and Schellman have recently been reinforced
by the deuterium-exchange studies of Stracher (1960), who has shown tjhatj
oxidized ribonuclease assumes a partially-folded, hydrogen-bonded con-
figuration i; this medium. Stracher also found, however, that comparable
concentrations of NaBr seem to have approximately the same effect as LiBr
on the deuterium-exchange properties, even though NaBr exerts a relatively
small effect on the activity of water. A similar correspondence between thesc
two salts was also observed in the optical rot,atory experiments of Rigclow
and Geschwind (1961). It seems probable that these salts produce a pro-
found configurational change in the polypeptide chain, but the fundamental
mechanism involved may be somewhat more complicated than originally
appreciated by Harrington and Schellman.
There can be little doubt that a marked structural change in poly-
L-proline I1 occurs in the presence of high concentrations of LiBr, and
NaSCN, since the intrinsic viscosity of this substance is strikingly depressed
in the presence of these salts. This cannot be a simple “solvent effect,”
since the effect of aqueous lithium bromide on the optical rotation of simple
derivatives of proline is relatively small, amounting to a change of a t most
5 % in the absolute value of the specific rotation (Steinberg et al., 196Ott).
Similar minor changes in optical rotation were reported by Bigelow and
Geschwind (1961) for several other low molecular weight substances in the
presence of aqueous LiBr solutions. To the present authors, the most likely
explanation of the optical rotatory changes in gelatin, poly-L-proline 11,
ribonuclease, and clupein is that these reflect deep-seated changes in chain
configuration brought about through fundamentally similar mechanisms.
This is believed to arise from a modification in the solvent-chain interaction
which is mediated by the salts. In the case of poly-L-proline 11, we have
already given evidence t.hat LiBr and NaSCN “unlock” the bond (ii)
linkages, allowing free rotation about, t,hcse bonds. This process is thought
to arise from a modification of the solvation of the carbon yl-oxygen atom,

the solvent binding in the peptide region in the absence of salt and thus
preventing rotation of the pyrrolidine rings about the backbone. Clearly
LiBr and NaSCN may be modifying the peptide group solvation through
quite different mechanisms, since it is well known that the lithium and
thiocyanate ions have markedly different effects on the activity of water.
Similarly, the optical rotatory changes in gelatin in the presence of these
same salts may well rest on the attenuation or modification of the solvent
binding. In gelatin, we might expect the bond (ii) linkage of the proline-
hydroxyproline sequence to be unlocked, thus eliminating any residual
poly-L-proline II-like structure. The distribution of proline residues along
the gelatin peptide chain would prevent the formation of an extensive a-
helical configuration, analogous to that observed in the oxidized ribonucle-
ase and clupein in the presence of LiBr. This is also consistent with the
rotatory parameters. Although the specific levorotation of gelatin is
greatly reduced in concentrated LiBr, A, exhibits only a slight decrease,
similar t o that observed in poly-L-proline I1 in this solvent, but opposite in
sense to that observed for clupein and ribonuclease. Indeed, the very low
specific levorotation of gelatin in concentrated aqueous LiBr (about 60"
below the specific rotation expected for a random chain, see Table XIII)
suggests that the configuration is right-handed, in contradistinction to the
left-handed sense of the chain as it exists in the collagen structure.
It is also apparent from this discussion that the stability of the three-
stranded collagen molecule in aqueous solution would be drastically lowered
in the presence of a high concentration of these neutral salts, the integrity
of the individual, left-handed poly-L-proline II-type helices of the macro-
molecule being distorted and destroyed as the bond (ii) linkages between
contiguous pyrrolidine rings are unlocked. These regions, in addition to
bcing the seat of the left-handed structure, must confer considerable
stability on the helix.
3. Nonaqueous Solvent Systems
The bchavior of gelatin in mixed organic solvent systems h i s bccii rc-
ported by Veis and Anesey (1959) and Steinberg et al. (1960a). Veis aiid
Anesey have observed that the addition of dimethylformamide (DMF) to a
formic acid solution of pigskin gelatin lowered the specific rotation from
[alD= - 116" to [aID= -58" (in 5 % formic acid-95 % dimethylformamide).
Similarly, Steinberg et al. observed an immediate decrease in [aIDon dilution
of a formic acid (FA) solution of bovine gelatin with n-propanol (8:l
propanol-formic acid, v/v). No further changes in rotation were observed on
standing. From rotatory dispersion data, A, was found to decrease from 217
to 203 mp in the mixed solvent system.
The intrinsic viscosity of gelatin in formic acid decreases on addition of
108 IIARRINGTON AND VON HIPPEL
dimethylformamide, reaching a minimum value a t a ratio FA/DMF = 1.

Further addition of dimethylformamide leads to an increase in intrinsic
viscosity (Veis and Anesey) . Since light-scattering studies indicate that the
molecular weight of gelatin remains unchanged throughout these trans-
formations, it appears that both the rotation and the viscosity changes
reflect alterations in the configuration of individual gelatin molecules.
The solvent system employed by Veis and Anesey has been shown to lead
to random coil -+ a-helix transformations in synthetic polypeptides (Yang
and Doty, 1957) but the decrease in A, and the anomalous minimum in the
viscosity versus solvent plot suggest that in the case of gelatin another type
of structure may be forming. Veis and Aniiescy propose a right-handed helix
of the poly-L-prolinc I-type, generated through trans 3 cis-isomerizations
a t the proline-proline peptide bonds. Steiriberg et al. also believe that
the rotatory changes result from induced rotations of neighboring pyrro-
lidine rings about the backbone, but suggest that this occurs at bonds
(ii). It may well be that increased rotational freedom a t both linkages is
involved in generating the new structure.
B. The Molecular Properties of Gelatin at Low Temperature
When gelatin solutions are cooled to low temperature, striking transfor-
mations are seen in a number of physical properties. The viscosity increases
with time, and if the protein concentration is above a critical level the
system soon sets to an elastic gel. Dilute isoionic solutions do not exhibit the
gelling phenomenon, but the viscosity increases continuously over a pro-
longed period of time at low temperatue and light-scattering studies
demonstrate an association between gelatin chains, leading to molecular
weights of the order of 40 X loe (Boedtker and Doty, 1954). I n the presence
of salt or at pH values removed from the isoionic point, lower molecular
weight aggregates are formed. These transitions are accompanied by sub-
stantial changes in optical rotation, the specific rotation increasing from
[a],= - 120" to values approaching those characteristic of native collagen
[a],= -350" to -400"; (Smith, 1919; Ferry, 1948a; Ferry and Eldridge,
1949; Robinson, 1953; Cohen, 1955; von Hippel and Harrington, 1959;
Flory and Weaver, 1960). Careful measurements during the gelling process
have demonstrated that the volume also decreases slowly with time (Derk-
sen, 1935; Heymann, 1936; Neiman, 1952, 1954; Flory and Garrett, 1958).
The over-all sol-gel transformation is reversible; the rigidity, levorotation,
volume, and heat content changing in unison through a fairly well-defined
temperature range (Smith, 1919; Pleass, 1930; K a t a , 1933; Holleman, et al.,
1934; Derksen, 1935; Ferry, 1948b; Ferry and Eldridge, 1949).
1. X-ray Di$raction and Infrared Studies
Cold gelatin gels formed by evaporation exhibit a rather high degree of
order, with X-ray reflections at 2.8, 7.8, and 11.5 A, and in addition, an
amorphous halo a t 4.4A (Hermann et al., 1930; Katz et al., 1931; Katz and
Derksen, 1932). It will be remembered that strong 2.8 and 1 1.5 A reflections
are also observed in native collagen. These reflections disappear from the
gelatin gel on heating above the transition temperature, but slowly reappear
on cooling. Gels which havc been oriented by stretching display X-ray
diffraction patterns strongly resembling those of collagen, with charac-
teristic meridional reflections at 2.86 A and equatorial spacings of 10 to 16 A
(Hermann et al., 1930; Derksen, 1935). When concentrated gelatin gels are
stretched, followed by heating, they show a well-defined shrinkage tem-
perature with characteristics quite comparable to the thermal shrinkage of
collagen (Hirai, 1953). Cold evaporated gelatin gels which have been treated
with neutral salts also exhibit a sequence of X-ray diffraction patterns
paralleling those obtained for native collagen under similar environmental
conditions (Ramachandran, 1958).
Noteworthy structural similarities between cold gelatin and native
collagen may also be inferred from infrared studies. Robinson and Bott
(19,51) found the N-H stretching frequency of a film prepared by evapora-
tion of a hot (40°C) gelatin solution to be 3310 cm-I, whereas in a film
evaporated at room temperature this band was shifted to 3330 cm-I, the
value characteristic of collagen. Comparable results have been reported by
Bradbury et al. (1958) who observed the infrared spectrum of cold-cast
gelatin t o be intermediate in character between native and completely
denatured collagen. Furthermore, when cold-cast gelatin films are stretched,
the resulting infrared dichroic ratios at 3330 cm-I (N-H stretch) and 1650
cm-1 (C=O stretch) have the same sense as those characteristic of native
collagen, but no dichroism is detected in hot evaporated gelatin films; again
indicating the lack of structural order in the polypeptide chains under these
conditions.
Taken in conjunction, the physical data summarized above strongly
suggest that the crystalline structure of collagen is partially regenerated on
cooling gelatin. This view is supported by the recent electron-optical
studies of Veis and Cohen (1960) and Rice (1960), who have demonstrated
that many of the morphological features of native collagen return on cooling
gelatin under controlled conditions.
2. Optical Rotatory Properties
The striking change in levorotation seen on chilling warm gelatin solu-
tions must be judged one of the most interesting properties of this system.
At concentrations of the order of 0.5 % gelatin and above, the mutarotation
is generally accompanied by gelation and the apparent relationship between
mutarotation and the gelling phenomenon has consequently been in-
vestigated by a large number of workers (see Ferry, 1948a). Certainly one
of the most careful early studies in this area was that of Smith (1919) who
110 HARRINGTON AND VON HfPPEL
found that a t temperatures of 35°C and above and a t gelatin concentrations

between 1 and 7%, the specific rotation of aqueous solutions of ossein
gelatin = -120") was virtually independent of temperature and
protein concentration. At low temperatures (< 15"C), the specific levorota-
tion was seen to change over a prolonged period of time a t a rate inversely
proportional to the concentration of gelatin, approaching a nearly constant
specific rotation of [aID= -313". On raising the temperature stepwise be-
tween 15" and 35"C, the specific rotation leveled off a t progressively lower
values, with [all, exhibiting a slight concentration-dependence at any
temperature. Smith concluded from these studies that two forms of gelatin
exist; one stable above 35°C (sol form) and one stable below 15°C (gel
form), and that at temperatures intermediate between 35" and 15°C these
are in equilibrium. Similar differences between the properties of gelatin at
high and low temperature were revealed by an investigation of dried gelatin
gels. Gelatin dried at temperatures above 35°C showed a n [aID = -120",
whereas samples dried at temperatures below 15°C displayed a much higher
levorotation, approximating [a] = -750".
The early proposals of Smith have been supported and extended by the
work of Robinson (Robinson and Bott, 1951; Robinson, 1953) who found
[aIDof a cold evaporated gelatin film to be about -1000", whereas a hot,
evaporated film showed [aI0 = -128". From these experiments, and his
infrared absorption and dichroism studies on gelatin films, Robinson has
suggested that gelatin exists as single molecules in a configuration which
cannot form interchain hydrogen bonds a t temperatures greater than
35°C. At lower temperatures, Robinson assumed that a unique configuration
(the collagen-fold configuration) develops which can form interchain hy-
drogen bonds. On the basis of her extensive studies on the optical rotation
and optical rotatory dispersion properties of ichthyocol gelatin, Cohen
(1955) reached similar conclusions, attributing the increase in levorotation
([a]~6 =0 -110'; [a]r = -350") observed on cooling gelatin solutions
specifically to the development of a helical configuration along the poly-
peptide chain.
A perusal of the data of Smith discloses that the specific rotation of 24-
hr, low temperature gelatin is remarkably insensitive to protein concentra-
tion, in contradistinction to the direct relationship between mutarotation
and chain association which some workers have assumed in the past
(Kraemer and Fanselow, 1925; Katz, 1933). The independence of the two
processes is also indicated by the investigations of Ferry and Eldridge
(1949). On chilling ossein, alkali-processed gelatin solutions of various
concentrations (2-6 % protein), each solution attained virtually the same
specific rotation ( [ 0 1 ] 6 6 ~ ~= -375") after 24 hr. Moreover, the melting-out
profiles of all of the solutions were essentially coincident, with [a]:46declining
from -265" to - 120" over the temperature interval 15" to 30°C. In these
studies, the terminal specific rotation of degraded, low molecular weight
gelatins maintained at low temperature varied with the number-average
molecular weight; but the specific rotation of various mixtures of two dif-
ferent gelatin fractions(an = 17 X lo3 and a,, = 44 X lo3)was found to
be additive in weight Concentration. Ferry and Eldridge concluded that the
change in optical rotation accompanying gelation was due primarily to an
intramolecular process within individual gelatin molecules and suggestcd
the formation of intramolecular cross-linkages or, alternatively, some type
of chain rearrangement.
The invariance of the ultimate value of specific rotation of chilled gelati11
with concentration has also been demonstrated by the work of Cohen
(1955) on ichthyocol gelatin and of Flory and Weaver (1960) on rat-tail
tendon gelatin. Von Hippel and Harrington (1960); Harrington and von
Hippel (1961) have followed rotatory changes in chilled ichthyocol gelatin at
very short wavelengths (A = 265-313mp) and thus have been able to
measure terminal specific rotations at gelatin concentrations nearly two
orders of magnitude below those investigated by Smith, and Ferry and
Eldridge. Again [a],was shown to be nearly independent of concentration,
although the final reduced viscosity varied about sixfold over a comparable
concentration range. In substance, then, the mutarotation phenomenon
reflects the development of a specific type of structure along each gelatin
chain, the formation of which is independent of chain association, at least
in the initial stages. In view of the close correspondence between many of
the physical properties of cold gelatin and collagen, we assume that the
ordered structure regencrated along each chain is that of a poly-L-proline
II-type helix.
Although the mutarotation of dilute gelatin at low temperature (3°C)
is apparently complete in 24 hr, careful measurement reveals that the
specific levorotation continues to increase very slowly over a period of
many days-paralleling the gradual increase in Viscosity observed during
this interval. In fact mutarotation is apparently not complete even after 28
days (Harrington and von Hippel, 1961) and it seems clear that a specific
association between gelatin chains may lead to an increased ordering of the
individual poly-L-proline I1 helices and consequently an incremental
increase in levorotation. It will be remembered that destruction of the
highly crystalline collagen structure leads to a change in [alD of about
f284". The change in [ a ] ,on cooling dilute gelatin amounts to only about
5 0 4 6 % of this value (Flory and Weaver, 1960; Harrington and von
Hippel, 1961) consistent with that expected for a partially disordered strue-
ture. The additional development of the poly- proli line II-type structure
on association of chains can be inferred from Fig. 23. After 24 hr at 3°C the
112 HAHRINGTON AND VON HIPPEL
specific lcvorotation of ichthyocol gelatin (1.67 mg/ml) has recovered

about 64% of the ordered collagen structure as measured by optical rota-
tion. On standing 28 days a t this temperature the recovery is about 80%.
TEMP. (OC.)
FIG.23. Specific rotation of ichthyocol collagen and gelatin at 313 mp a s a fuiic-
tion of temperature. Collagen concentration = 1.14 mg/ml, gelatin concentration =
1.67 mg/ml. v, gelatin, after 24 hr at 3°C; A, gelatin after 6 days at 3°C; 0, gelatin
after 28 days a t 3°C; 0, native soluble collagen. (From Harrington and von Hippel,
1961.)
Another noteworthy feature of Fig. 23 is the greater breadth of the mehing-

out profile of chilled gelatin compared t o that of collagen, indicating a
lower degree of order and perhaps near neighbor cooperation in the gelatin
structure (Schellman, 1955, 1958; Zimm and Bragg, 1958, 1959; Gibbs arid
DiMarzio, 1958). Yet the threshold temperature, T , ,is identical for the two
systems, revealing the fundamental similarity of their basic molecular
architecture. The significance of this relationship has been particularly

emphasized by Flory and Garrett (1958; Garrett and Flory, 1956).
3 . Kinetics of Mutarotation
When concentrated gelatin solutions (> 1 %) are held a t low temperature,
the levorotation increases rapidly over a period of several hours, after which
the rate declines sharply to a comparatively low value. Smith (1919) found
the [a],versus time profile to correspond closely to the behavior expected
for a second-order reaction , the over-all rate varying inversely wit,h protein
concentration.
Qualitatively similar kinetic patterns are observed in gel rigidity (Ferry,
194813). During the initial stages of gelation the rigidity increases very
rapidly, this phase leading into a prolonged interval in which the rigidity
changes a t a greatly reduced rate. Although Smith interpreted his mutarota-
tion kinetics in terms of a second-order reaction, the over-all time-de-
pendent change, both in optical rotation and rigidity, could rcsult from
combination of a fast process followed or accompanied by a slow process.
That this is indeed the case may be seen from the results of two recent,
studies on the mutarotation of gelatin at low protein concentrations. Flory
and Weaver (1960) found the rate of mutarotation of rat tail tendon gelatin
to be independent of concentration (in the range 0.5 to 4 mg/ml) at tem-
peratures between 5°C and 23°C. Von Hippel and Harrington (1960)
reported a similar invariance in rate of mutarotation of ichthyocol gelatin
over the range 0.1 to 1.7 mg/ml (at 3.9”C). It was demonstrated in the
latter study that the over-all process consists of two phases (Harrington and
von Hippel, 1961) of which the primary phase is independent of protein
concentration and is completed in about 15 hr at 3.9”C1while the secondary
phase is, as we have seen, a concentration-dependent association between
gelatin chains and leads, in this concentration range, to an incremental
increase in rotation which takes a very long time to go to completion. At
high concentrations the chain association reaction would be expected to be
markedly increased in rate, leading to a compression of the [a],versus time
curve which would mimic a second-order process.
Flory and Weaver (1960) have advanced a mechanism for the reversion
kinetics, in which they postulate the formation of an intermediate (in-
volving the intramolecular rearrangement of a single gelatin molecule) as
the rate-determining step. The intermediate, which may be a helix of the
poly-L-proline 11-type, is converted rapidly to a three-chain compound helix
a t a rate “sufficiently rapid to have no effect on the over-all rate.” The
scheme may be represented as follows:
ki’ h’
C - I L %(H)
‘
b h
where C, I, and H represent t,he random coil, single-chain intermediate, and

nat,ive, three-chain collagen helix, respectively. The concentration of
intermediat,e is assumed to be very small compared to the concentration of
C, so that t,he over-all process should be first-order (with rate, R' = Icl'
[C]). The second step of the process consists of the lateral aggregation of the
intermediate species and is thought t o he both comparatively rapid and
easily reversible.
A crucial question with respect to the Flory-Weaver scheme is t,he rat,e of
the second step, i.e., the lateral association between chains t,o form t,he
native three-stranded compound helix. This process should he strongly
concentration-dependent, a change in concentration of tenfold leading tjo a
change of a hundred- to a thousandfold in rate. In fact we have seen that, at,
concentrations between 1 and 7 % (Smith, 1918), t,he over-all rate of
mutarotation is strongly influenced by concentration. On the other hand,
the lack of a concent,rat,ion-dependenceat, the very low prot,eiri concent,ra-
tions used by Flory and Weavcr (1960) and Harringtm and von Hippel
(1961) suggests that this st3ep,t'he lateral associat>ionhet,weeii chains, has
been reduced in rate to a comparatively low value. Further evidence in
support of this supposition comes from the viscosity studies at, low con-
centration. At concentrations of less than 1.5 mg/ml, the primary st,ep of
mutarotation is completed in 24 hr at 4°C. During this time interval the
reduced viscosity, qsPic has increased from -0.30 to -0.65 dl/gm. In con-
t,rast to this latter value, the reduced viscosiy of the three-chain compound
helix is about 14 dl/gm (Table VII). Clearly the molecular structure present
a t the end of the primary phase cannot be the three-stranded compound
helix of collagen.
These considerations lead us to an alternative mechanism, which had
been proposed earlier by the authors of this review (von Hippel and Har-
rington, 1959, 1960; Harrington and von Hippel, 1961). We assume t,hat
the first-order kinetics of the primary phase reflects the development, of a
configuration of the poly-L-proline II-type along each chain, and that this is
a stable configurat,ion at low temperature. The chain may exist either as a
single helix or as helical segments separated by random chain elements.
Helical segment,s of neighboring chains are able to pack laterally, giving
(locally) the highly ordered interchain hydrogen-bonded structure found in
native collagen. Lateral association between chains allows a further ordering
of the noncrystalline regions of each chain.
The difficulty with this proposal is the requirement for intramolecular
stabilization of the helical elements. As we have seen, the left-handed helix
of poly-L-proline I1 is stabilized in aqueous solut,ion a s a result of the
restraints imposed a t the peptide linkage and the
0
Bond (ii)
bonds. I n the gelatin chain, however, free rotation is theoretically possible

about every third backbone linkage in the non-imino acid residues. It is
highly unlikely that a systematic set of peptide hydrogen bonds, acting in a
- 1900
-1700
-1500
n
;--I300
23
-1100
- 900 -
- 700
0 10 20 x) 40 50
TEMPERATURE PC)
FIG.24. The specific rotation of ichthyocol gelatin a t 313 mp as a function of tem-
perature. Protein concentration = 1.67 mg/ml. Solvent: 0 , 0.5 M CaClz in DzO;
0 , 0.5 M CaClz in HzO. Samples on solid curves were held a t 3°C for 24 h r after
quenching from 45"C, samples on dotted lines were held for G days after quenching.
(From von Hippel and Harrington, 1960.)
inaiiiicr analogous to that of tJhea-hclix, could stabilize this stmcture. This

hydrogen-bonded arrangement is ruled out both by the lack of peptide
hydrogen donors in the imino acid residues and by the steep pitch of the
poly-L-proline 11-type helix. Nevertheless, some type of hydrogen bonding
is indicated, since no observable time-dependent changes in optical rotation
are observed on cooling aqueous gelatin solutions in the presence of high
concentratioiis of the hydrogen-bond-competing reagents urea or guanidine-
HC1. Moreover, when the melting-out profiles of cold (24 hr a t 3°C) dilute
ichthyocol gelatin in DzO are comparcd to those in HZO (Fig. 24), the mid-
116 IIARRINGTON AND VON HIPPEL
point of the transition is found to be elevated about 3.7”C (Harrington and

von Hippel, 1961).
A similar temperature differential was found by Hermans and Scheraga
(1959) in comparing the “melting” temperature of the hydrogen-bonded
secondary-tertiary structure of ribonuclease in these two solvents. The
explanation for the elevation in temperature advanced by these authors is
that the -0. . . H peptide hydrogen bond in HzO is less stable than the
corresponding 0 . ..D peptide hydrogen bond in DzO. This suggestion
is supported by X-ray studies on simple crystals, which have demonstrated
an appreciable shift in hydrogen bond length on replacement of hydrogen by
deuterium (Gallaghcr, 1959). The change in bond length, which is thought
to be primarily due to the difference in zero point vibration energies of the
two isotopes, is rclatcd to the type of hydrogenbond and may occur either as
u coritractiori or elongation of the bond, depending on the original H-bond
length.
On the basis of these arguments, we are led to the possibility that u
systematic set of hydrogen bonds involving water molecules may develop
along the geltttin chain at low temperature. We have in mind specifically
the formation of doubly hydrogen-bonded water bridges between adjacent
carbonyl oxygen atoms, these forming a cooperative set along the chain and
serving to stabilize the poly-L-proline II-type configuration. Evidence
for the involvement of water as an integral part of the structure of collagen
has been presented in Section IV. We may recall here that the work of
Burge et al. (1958), Esipova et al. (1958), and Bradbury et al. (1958), has
provided evidence that water molecules will fit into the geometry of the
chains of collagen, forming hydrogen-bonded bridges between adjacent
carbonyl groups without severe strain. We should also note the extra-
ordinarily strong affinity for water exhibited by imino residues, as shown
by the infrared studies of Blout and Fasman on poly-L-proline (1958).
Although the rate of mutarotation of cold dilute gelatin is independent
of protein concentration, kinetic analysis reveals an exponential dependence
of d[a]/dt on the concentration of chain elements in the unfolded form.
Thus van’t Hoff plots of log (d [ a ] / d t )versus log ([& -[aI1),
derived from
the general equation:
are linear over about 80 % of the intramolecular phase of the reaction, with
n = 2.2 (f0.15) independent of protein concentration or temperature (see
Table XIV).
The apparent negative temperature dependence of mutarotation is
another striking aspect of the kinetics which is of signal importance in
understanding the mechanism of the gelatin -+ collagen-fold transition

(Smith, 1919; Flory and Weaver, 1960; von Hippel and Harrington, 1960).
Flory and Weaver found the rate of mutarotation of rat-tail tendon gelatin
to increase over a hundred-fold when the temperature was decreased from
23°C to 5°C. It was also observed that a logarithmic plot of the half-time of
mutarotation against the reciprocal of absolute temperature (log [tx]
versus 1/T) was nonlinear over this temperature range, the magnitude
of the (negative) activation energy' of mutarotation increasing with tem-
perature. A large apparent negative activation energy (- - 30 kcal/mole)
TABLE XIV
Kinetics of Optical Rotatory Changes During the Gelatin -+
Collagen-Fold Transition
Gelatin SolventD Concentration Tern erature Order of Reaction6

(mg/ml) PC, (4
I. Ichthyocol H2 0 1.33 3.70 1.9
Ha 0 1.33 5.20 2.1
H2 0 1.33 8.00 2.4
H2 0 1.33 11.40 2.3
H2 0 1.28 8.00 2.3
H 20 0.64 8.00 2.0
H20 0.32 8.00 2.1
H20 0.16 8.00 2.4
DzO 1.57 3.70 2.2
D20 1.57 5.05 2.1
D20 1.57 8.00 2.2
D2 0 1.57 11.35 2.2
11. Calfskin H2 0 1 .?O 3.7 2.2
D2 0 2.0 3.7 1.9
Over-all average: 2 . 2 f 0.15
All solutions 0.5 M in CaClz .
* Slope ofvan't Hoff plot.
over the temperature interval 3°C to 12"C, has also been reported for
ichthyocol gelatin (Harrington and von Hippel, 1961). The magnitude of
the negative activation energy and its variation with temperature suggest
a phase transition, similar to a crystallization process developing from
preformed or rapidly formed nuclei (Flory and Weaver, 1960; Beckrr and
Doring, 1935; Flory, 1949; Flory and McIntyre, 1955; Price, 1959; Laurit-
Zen and Hoffman, 1959).
Flory and Weaver have shown how their mechanism (see above) could
lead to a negative temperature coefficient for the reversion kinetics. Since
the intermediate, I, consists of an ordered configuration, its entropy should
be appreciably lower than that of C, the random coil. The over-all (C -+ H)
enthalpy change is negative (approximately 1.4 kcal per peptide unit), and
Flory and Weaver assume that part of the enthalpy should be lost in the first
step of the process, the formation of I. This result, coupled with the as-
sumption of an unstable intermediate of ordered configuration, leads t o a
negative heat of activation and a large positive free energy of activation.
Crystal growth is assumed to arise from nuclei consisting of helical seg-
ments constructed t,hrough the joining of three primary helix intermediates,
I, of a critical length of n residues.
The apparent intramolecular nature of mutarotation has prompted
the present authors tjo examine the kinetics of t’he process in terms of a
one-dimensional crystallization along individual gelatin chains (Harrington
and von Hippel, 1961). According to this view, the crystal nuclei would be
small segments of the poly-L-proline II-type helix which are residual above
the transition temperahre, T,,, , or, alternatively, which form very rapidly
on lowering the temperature below T, . If it is assumed that crystal growth
is propagated a t constant velocity from these nuclei, the apparent order of
the kinetics (n = 2.2 f 0.15) can be easily derived. The analysis suggests
that the nuclei are grouped in clusters along the gelatin chain.
The overriding experimental consideration which led both Flory and
Weaver (1960) and the present authors (von Hippel and Harrington, 1959,
1960; Harrington and von Hippel, 1961) to postulate a single-chain inter-
mediate in their mechanisms for collagen reformatlion from cold gelat,in,
was the observation that the mutarotation process appeared to be in-
dependent of concentration, in marked contrast to properties attributed to
interchain interactions as measured by viscosity and light scattering. How-
ever, the evidence already discussed in Section I11 and the present section
strongly suggests the presence of interchain covalent cross-linkages, a t
least in some collagens and gelatins. This raises the possibility that interac-
tjions between cross-linked chains might, nevertheless, be involved in the
generation of the collagen-fold. Such a process would still, of course, be
intramolecular and so could not be ruled out on the basis of the observed
concentration independence of tJhe primary step. These arguments sug-
gested that an investigation of “de-esterified” gelatins might be helpful.
As pointed out in Section 111, Gallop et aE. (1959) showed that &w of
both ichthyocol and calfskin gelatins is reduced to -20,000 on treat,ment
with aqueous hydroxylamine (pH 9, 40°C). Such treatment breaks ester-
type linkages, but has been shown to have no effect on peptide bonds. Thus
for studies of the mutarotation process de-esterified gelatin would have two
advantages : (1) the probability that cross-linked multichain units are
involved in the process should decrease with decreasing moelcular weight,
and (2) the bonds which are broken by the hydroxylamine treatment
might well be the postulated cross-linkages.
A recent examination of the mutarotation phenomenon in de-esterified

gelatin (von Hippel and Wong, 1961) has shown that this process is es-
sentially unchanged from that observed using untreated gelatins, ex-
hibiting similar large negative temperature coefficients and apparent
energies of activation, the same order of the reaction from van't Hoff plots
(n = 2.0 f0.2; compare with Table XIV), and a similar response to altera-
tions in the ionic environment. Moreover, tjhe number-average molecular
weight (estimated from protein-bound hydroxamate) and the weight-av-
erage molecular weight (by short-column equilibrium ultracentrifugation)
are closely similar, as expected for material consisting primarily of single
polypeptide chains of roughly comparable length.
TABLE XV
Rate of Formation of the Collagen-Fold i n Different Gelatins, Following
Quenching to 3.7%'
Total imino Initial rate

acid content Transition dedmin) [d:l8
Gelatin
(residues/ :ERp::tj (final)
1000) In H ~ O In D ~ O In Hzo
Dogfish sharkskin 158a 11 0.39 0.92 1940

Ichthyocol 197* 18 4.8 13.2 1900
Calfskin 232b 23" 13.1 27.7 1972
a Piez, personal communication.
* Piez and Gross, 1980.
Smith, 1919.
Another revealing feature of the mutarotation process is the cf!fect of the

solvent environment on the rate. Harrington and von Hippel (1961) have
shown that the initial rate of mutarotation of gelatin in DzO is markedly
accelerated over that in HzO. At 3.7"C, for example, the initial rate of
mutarotation of ichthyocol gelatin in DzO is over 2.5 times that in HzO.
Similltr ratios have been observed with dogfish shark and calfskin gelatin
(see Table XV). These findings may be explained on the basis of the large,
negative temperature cxfficient of the mutarotation process and the
finding that the melting temperature in DzO is elevated by approximately
3.7"C above that in HzO (see Fig. 24). Since measurement of the rate of
helix formation a t a fixed low temperature involves a larger temperature
differential between the experimental temperature and T, in the DzO
solvent, the rate in DzO would be expected to be greater than that in HzO.
In fact, if the difference in T , in the two solvent systems and the apparent
energy of activation of the process are taken into account, the rate in DzO
becomes essentially identical to that in HzO.
Other methods of tampering with the solvent environment lead to

comparable results. In the study discussed above, all the experiments were
carried out in an identical ionic environment, namely neutral 0.5 M CaC12.
Recently von Hippel and Wong (1961) have examined the effect of vary-
ing the ionic milieu of the gelatin chains, and find that the initial rate of
mutarotation is a very sensitive function of both ion concentration and ionic
composition. For example, the initial rate is approximately tenfold larger in
0.025 M CaClz than it is in 0.5 A4 CaClz (see Fig. 25), and the melting profile
for both ichthyocol and calfskin collagen in 0.025 M CaClz is found to be
Temp. = 4. C
Cone. = 1.6 m*lm~
0 CoClt (PH 7 )
A C a C I e 4 Glyclne (PH 25)
V C o C l o + Olycin. (pH 10.5)
ionlo Strcngth
FIG.25. Initial rate of rnuLarotation of ichthyocol gelatin cooled to 4°C at zcro
time, as a function of ionic strength (CaClZ). (From von Hippel and Wong, 1961.)
shifted upward several degrees when compared to the profiles obtained in

0.5 M CaCl2.
The effect of a variety of cations and anions on the rate of mutarotatioti
and the terminal rotations obtained on standing a t low temperature have
also been investigated (von Hippel and Wong, 1961). In agreement with
the findings of Carpenter and Lovelace, these studies have shown, as might
be expected, that the effect is not simply one of ionic strength, but varies
enormously from one ionic species to anot,her. Thus, while small increments
of (dciuni, barium, lithium, thiocyanate, bromide, etc., ions have large
effects 011 the initial rates, other ions, particularly tetramethylammonium
and acetate, have almost no effect at all. It seems reasonable to assume,
in line with the discussion given earlier, that ions exert their effects by
modifying the solvation of the gelatin polypeptide chains.
The proline and hydroxyproline residues are fundamental in the processes
which have been described above. It is not simply that these residues “fit
into” a left-handed helix as suggested by some authors. In our view, it is
more reasonable to suppose that their geometry and rigidity, detailed in
Section 11, establish and direct the left-handed configuration which is
formed a t low temperature. The optical rotatory evidence suggests that
elements of the poly-L-prolirie 11-type structure remain in the gelatin chain
above the transition temperature, T,. It seems likely that this residual
structure is due to the presence of a significant number of contiguous proline
residues in the chain, a situation recurring frequently via the sequence
Gly.Pro.Hypro.22 These regions may act as crystal nuclei for the growth of
the poly-L-proline I1 helices. We may imagine, for definiteness, that these
segments “lock-in” when the temperature is lowered below T,. As sug-
gested earlier this may occur through the formation of water bridges be-
tween adjoining carbonyl oxygen atoms, the peptide segments neighboring
the nuclei slowly crystallizing into the preordained poly-L-proline I1 helix
through this type of hydrogen bond mechanism. It is possible that nuclea-
tion may also occur at the G1y.Pro.X sequences in the chain, although this
situation seems less likely because of the increased number of bonds al-
lowing free rotation.
Additional supporting evidence for the type of nucleation proposed above
comes from studies in which the enzyme collagenase was used to probe the
configurational changes occurring during the gelatin -+ collagen-fold
transition. As pointed out in Section 111, the requirement for activity of
this enzyme is the peptidc sequence -1’ro.X.Gly.Pro-, with cleavage taking
place between X and Gly. At temperatures above T,, von Hippel arid
Harrington (1959) found all peptide linkages in ichthyocol gelatin cleaved
by the enzyme to belong to a single class in that the reaction obeyed simple
first-order kinetics over several half-lives. When the gelatin solution is
cooled below the transition temperature, the kinetics become complex and
analysis suggests that the potentially cleavable bonds are distributed among
two classes with about 20% of the total undergoing cleavage at a rate
nearly tenfold that of the remaining bonds. Moreover, the complex kinetics
attain their final form within 1 hr after lowering the gelatin temperature
(i.e., within the time required to complete an enzymatic reaction) in keeping
with the proposal that the enzyme is sensing primarily the nucleation
process.
In view of the requirement for the peptide sequence Pro.X.Gly.Pr0.Y
22 Heyns and Legler (1960) have recently found t h a t pyrrolidine residues separated
by one or two non-imino residues can still generate a left-handed structural contri-
bution t o optical rotation in single peptides. Thus the residue rotation of proline in
the tripeptide carbobenzoxy-Gly.Gly.Pro-NHa is -217” whereas t h a t estimated for
carbobeneoxy-Pro.Ala.Gly.Pro-NHais -272” and that for csrbobenzoxy-Pro.Ala.-
Gly.Hypro is -308”.
by the enzyme, it seems reasonable to assume that the change in the form of
the enzyme kinetics below T,,, results from a rapid, temperature-dependent
alteration in the chain configuration in this region. The types of peptidc
linkages cleaved can be divided into two general classes: those which result
in a terminal G1y.Pro.Y sequence (amounting to about 80% of the total
peptides formed) and the remaining 20% which have Gly.Pro.Hypro as
the terminal sequence. The correspondence between the ratio of bonds
cleaved in each reaction and that expected for these two classes of peptides
indicates that below T , the sequence Pro.Hypro.Gly.Pro is cleaved much
more readily than the sequence Pro.X.Gly.Pro.Y, and that this difference
in rate is related to the nucleation or “lock-in” phenomenon about the
Pro.Hypro elements discussed above. We should note that the number of
these nuclei would be relatively small compared to the total residues in a
chain, so that significant changes in optical rotation would not be expected
during the nucleation step. It is possible that some change occurs but that
it is overshadowed by the mutarotation taking place during the time
required for completion of an enzymatic reaction.
In summary, tlhe following three-step mechanism for the reformation of
the collagen-type structure in dilute gelatin solution, following cooling to
temperatures below T,, has been proposed (von Hippel and Harrington,
1959; von Hippcl and Harrington, 19GO; Harrington and von Hippel, 1961).
1. An initial configurational change takes place in the pyrrolidine-rich
portions of the gelatin chains, which “niicleat,es” the poly-L-prolinc II-type
helix. This st,ep goes to completion rapidly arid is detected by the change
from simple to complex collagenase kinetics.
2. The poly-L-proline I1 configurat,ion propagates outward from these
nuclei along single gelatin chains. This process is responsible for the more
rapid, concentration-independent portion of the mutarotation phenomenon.
3. The formation of the unique collagen-fold type structure along
individual chains makes possible lateral chain association, which may be
monitored hy the relatively slow changes in viscosity and light, scattering
accompanying this step.
4. Properties of Gelatin Gels
The physical properties of gelatin gels and their dependence on protein
concentration, temperature, molecular weight, pH, and added reagents
have been thoroughly reviewed by Ferry (1948a). We shall refer only
briefly to supplementary work.
Considerable insight into the mechanism of gelation is afforded by a study
of the influence of chain weight on gel behavior. Below a chain weight of
about 60,000, the rigidity of gelatin gels (shearing stress/shearing strain)
which have been matured at low temperatures, increases markedly with
increasing molecular weight (Ferry and Eldridge, 1949; Saunders and

Ward, 1958a). Above this critical size the rigidity appears to be essentially
independent of molecular weight (Saunders and Ward, 1958b) and, in fact,
may decrease somewhat (Stainsby, et al., 1954; Pouradier et al., 1954;
Saunders and Ward, 1958b). Similarly, although Ferry observed the
specific levorotation of mature gelatin gels to increase progressively over
the molecular weight range 33-72 X lo3, it appears from the work of
Saunders and Ward (1958) that [a], is unaffected by chain weight above
this range. Melting behavior reveals, correspondingly, that the melting
point is elevated appreciably with increasing chain weight below Mw =
70,000, whereas the relative change in melting temperature with molecular
weight is distinctly reduced at higher Bwvalues (Ferry and Eldridge, 1949;
Saunders and Ward, 1958a). Moreover, the melting point is substantially
unchanged by variations in concentration a t high molecular weight
(>70,000), but rises sharply with increasing concentration a t low chain
weights (Ferry and Eldridge, 1949).
Qualitatively, the mechanical behavior of gelatin gels resembles that of
rubber. The damping of elastic response is small and there is a rapid
response t o stress, analogous to that observed for rubberlike materials.
On the molecular level, these properties are characteristic of network gels
in which the interaction between chains occurs a t specific loci distributed at,
intervals along each chain. In polymeric systems where chain-chain inter-
action can occur at every residue along the molecule, the resulting gel
exhibits appreciable elastic damping, response to stress is slow and, more-
over, gelation generally requires a careful balance between precipitation
and solution. A number of recent studies on gelatin gels have supported and
amplified the rubberlike model (Hirai, 1953; Saunders and Ward, 195813;
Hastewell and Roscoe, 1956; Jopling, 1956) and are in general compatible
with Ferry’s early postulate (1948a) that the loci available for cross-linking
on each gelatin molecule are both specific in nature and limited in number.
At low molecular weight it might be supposed that the number of these
interaction sites would be too small t o form a stable network. When the
chain length reaches a critical size, a relatively rigid lattice can be gen-
erated and the physical properties of the gel state should be less dependent
on the length of each polypeptide chain.
a. Dependence of Melting Points on Added Reagents. The melting points of
gelatin gels exhibit only a small dependence on pH (Ferry) 1948a; Bello et al.,
1956) but demonstrate a very marked dependence on the presence of cer-
tain salts, acids, and nonpolar substances. The effectiveness of salts in
lowering the melting point has been found to be additive with respect to
their ions, with highly hydrated cations such as Li+, Ca++, and Mg++ and
large nonpolar or large, highly polarizable anions (Gordon and Ferry, 1946;
I24 HARRINGTON AND VON HIPPEL
Ferry, 1948a) bringing about the most dramatic effects. Thus sodium chlo-
ride, a t a concentration of 1 M , lowers the melting point of gelatin about
2.4"C, whereas lit~hium-2-hydroxy-3,5-diiodobenzoate (diiodosalicylate),
TABLEXVI
Melting Points of 6% Gelatin Gels Containing Salts"
Salt Concentration Melting point

(moles/liter) ("C)
None 30.4
Sodium fluoride 1.0 34.5
Sodium methanesulfonate 1 .o 31.5
Sodium chloride 1 .o 28.0
Sodium bromide 1. 0 22.8
Sodium nitrate 1 .o 22.3
Sodium thiocyanate 1 .o 16.0
Sodium iodide 1.0 10.0
Sodium benzenesulfonate 1.0 15.5
Sodium salicylate 1 .o No gel
Sodium trimethylacetate 1. 0 26.5
Sodium chloroacetate 1 .o 26.6
Sodium dichloroacetate 1 .o 19.6
Sodium trichloroocetate 1.0 12.7
Sodium dibromoacetate 1.0 16.3
Sodium tribromoacetate 1.0 No gel
Sodium diiodoacetate 1.0 12.3
Sodium trifluoroacetate 1.0 24.1
Sodium acetylt ryptophan 0.75 13.7
Sodium maleate 0. 5 33.7
Sodium succinate 0.25 32.3
Sodium fumarate 0.5 31.5
Sodium acetylenedicarboxyltte 0. 5 19.4
Lithium chloride 1.0 25.4
Lithium iodide 0.25 26.7
Lithium salicylate 0.25 20.5
Lithium diiodosalicylate 0.23 No gel
Calcium chloride 1 .o 16.1
Magnesium chloride 1 .o 22.9
From Bello et n l . (1956). Reproduced with kind permission of the American
Chemical Society.
atl an cquivalent concentration, lowers the melting point by 120°C (Rello

et aE., 1956). Table XVI, taken from the work of Bello, Ricse, and Vinograd,
lists the melting points of gelatin gels containing a wide variety of salts. I n
general the order of effectiveness of cations and anions follows a Hofmeister
series (I<at,z and Wienhoven, 1933; Carpenter, 1938; Ruchner and Mey-
link, 1943).
High concentrations of the competitive hydrogen-bonding reagents urea

and guanidinium chloride also show strong depressant effects on the melting
temperature, with guanidinium ion being about twice as effective as urea.
On the other hand, little change in melting point is observed in the presence
of guanidinium sulfate, the elevating effect of the sulfate anion counter-
acting the depressing effect of the guanidinium ion.
I n a comparative study of the gelation of various gelatins which had been
modified chemically so as to mask polar groups, Bello et al. (1956) have
shown that the striking alterations in melting point in the presence of
electrolytes are essentially unaffected by these changes, and therefore can-
not depend on ion binding at the polar groups. They suggest binding at the
peptide linkages as t hc most likely explanation.
1). DPpendence of Optical Rotation on Added .Reagents. In general the
effect of various electrolytes on the optical rotation of gelatin gels closely
parallels their effect on the melting point. Most salts lower the specific
levorotation appreciably, with a corresponding loss in gel rigidity. The
detailed study of Carpenter and Lovelace (Carpenter, 1927; Carpenter and
Kucera, 1931; Carpenter and Lovelace, 1935a,b, 1936, 1937) remains the
classic work in this field and demonstrates that the change in optical
rotation of gelatin gels (at 0.5"C) follows a Hofmeister series. Comparing
the salts of potassium, Carpenter and Lovelace found the following order in
the effectiveness of the anions: CNS- > I- > c103- > NO,- > Br- >
C1- > CHICOO- > CH3CH200- > HCOO-. The similarity of this series
with that laid out by Bello, Riese, and Vinograd (see Table XVI) for the
melting point of the gels clearly indicates the close relationship between the
optical rotatory and melting phenomena. This view is supported by
iiivestigations on numerous other reagents; those substances which have a
strong effect on rigidity and melting point show a qualitatively similar
response in optical rotatory properties (Katz and Wienhoven, 1933; Ferry,
1948a; Bello et al., 1956).
c. The Mechanism of Gelation. A number of workers have investigated the
possibility that the polar side-chain groups plily a significant role in the
gelation phenomenon. Grabar and Morel (1950) reported inhibition of
gelation when approximately 45 % of the guaiiidino groups of arginine arc
destroyed by alkaline hypobromite, and suggested that these groups are
responsible for gelation. Later work by Davis (1957) supported this view
but indicated that gelation was only inhibited when the amount of hy-
pobromite is sufficient to destroy virtually all of the guanidino groups. The
interpretation of these experiments has recently been contested b y Bello
(1959) who finds that hypobromite reduces the gelatin chain weight ap-
preciably, suggesting that rigidity is depressed through inability to form a
stable gel network. Moreover, chemically modified gelatins in which thc
polar groups are either masked or altered exhibit only minor changes in
gelling properties. Gelatins with amino and hydroxyl groups acetylated,
carboxyl groups esterified, and guanidino groups either nitrated or in-
creased in number have essentially the same melting points, rigidities, and
solution viscosities as unmodified gelatins (Bello et al., 1956; Kenchington,
1958).
Since the side-chain groupings do not appear to play a fundamental
role in the gelation process, the mechanism of gelation must involve, pri-
marily, groups along the polypeptide backbone. The probable involvement
of peptide groupings is also indicated by work on the gelling properties of
biuret-gelatin complexes (Bello and Vinograd, 1958). At peptide to copper
ratios as high as 17, gelation was completely suppressed a t O'C, but could
be restored by destroying the biuret complex a t low pH. As a result of thew
studies Bello and Vinograd have proposed that blocking of the peptide
groups may interfere with the formation of necessary cross-linkages at thesc
sites or, alternatively, that the copper complex may interferc with the
formation of a required intramolecular configuration.
Recently it has been found that formation of the biuret complex also
suppresses mutarotation almost completely, while acidification (and coii-
sequent destruction of the complex) reactivates the mutarotation process
(von Hippel and Wong, 1961). It was inferred from these results that
chelation of the cupric ion across the peptide bond inhibits the develop-
ment of the poly-L-proline 11-type helix by restraining adjacent residues in
sterically unfavorable configurations.
Many of the physical properties of gelatin gels are consistent with thc
assumption that the formation and structural integritiy of the gel network
depends on the establishment of a specific configurational pattern along
segments of the individual gelatin chains. This point of view was firstj
proposed by Hermann and Gerngross (1932). As a result of their X-ray
diffraction studies of gelatin gels they suggested that the nctwork is formed
through crystallization of certain regions of the gelatin chain, these seg-
ments being associated 1:tterally to form rrystalline bundles of chains, or
xystallites. Ferry (19484 latcr argued that the chains might first associate
in pairs rather than bundles, with further association of paired chains leading
to the formation of crystallites. He estimated that as few as 5 or 6 loci per
gelatin chain would be sufficient to form a rigid network. The proposal that
gelation proceeds through the formation of crystallites is also supported by
the light-scattering results of Boedtker and Doty (1954) and by the well-
known decrease in volume and evolution of heat characteristic of crystal-
lization processes (Flory and Garrett, 1958; Flory and Weaver, 1960).
In the present review, as indicated earlier, we have assumed that the
association loci of the gelatin chains are segments of the chain arranged in
STRUCTURE OF COLLAGEN AND OEILATTN 187
the poly-L-proline I1 helix configuration. The somewhat broad melting-

out profile of the gelatin gel would result from the lack of continuity in order
along the chain. The close relationship between the melting point of the
gel and the temperature at which the high levorotation is attenuated is
understandable, since both phenomena depend on the integrity of the
intrachain helices. This close relationship is also demonstrated by the
analogous effect of salts on the melting point and on optical rotation.
Destruction of the poly-L-proline I1 helices through the action of salts, by
way of a mechanism such as that proposed earlier, would destroy the loci
necessary for maintenance of the gel network. Since the association between
gelatin chains leads to partial restoration of the collagen structure, it will
be evident that close packing of the chain segments allows the type of
interchain hydrogen bonding found in collagen. I n fact, if the proper condi-
tions of temperature and concentration are employed, the gross macro-
molecular properties of collagen can also be regenerated.
Veis and Cohen (1960) found that cooling a fractionated 7 % bull hide
corium gelatin at 4°C for 20 hr, followed by a short heating step a t 40”C,
resulted in the precipitation of fibrils with the cross-striated pattern
typical of native collagen. The fibril width was variable, but electron
micrographs revealed typical -640A spacings along the longitudinal axis.
Rice (1960) has recently utilized the technique of Hall and Doty (1958) for
viewing individual molecules in the electron microscope, and has obtained
rods on spraying cold gelatin which have morphological properties similar
to those of native collagen. Calfskin collagen solutions a t concentrations
between 0.2 and 0.4 % were heated 19 min a t 70°C and sprayed a t elevated
temperatures onto freshly cleaved mica surfaces. After shadowing with
platinium, electron micrographs of these preparations showed only globules.
When the heated solutions were stored a t 5°C for several hours before
spraying, regenerated rods were observed. Although these varied somewhat
in length, many showed the 2800-3000 A length and 15 A diameter char-
xteristic of the native collagen molecule. Typical SLS-type fibrous aggre-
gates were also obtained using the “renatured” collagen.
ACKNOWLEDGMENTS
Preparation of this article was supported by research grants (A-4349) and (A-3412)
from the National Institutes of Health, United States Public Health Service One of
us (P. H. von Hippel) is indebted t o the Public Health Service for a Senior Research
Fellowship (SF-360). Thanks are also due Miss S. Himmelfarb and Mrs. M. Backer
for help in t h e final preparation of the manuscript, and t o Mr. H. Walton for construc-
tion of the wire models of poly-L-proline.
REFERENCES
hlbrecht, G., and Corey, R. B. (1939). J . Am. Chem. Soc. 61, 1087.
Ambrose, E. J., and Elliott, A. (1951a). Proc. Roy Soc. A208, 75.
Ambrose, E. J., and Elliott, A. (1951b). Proc. Roy. Sac. A206, 206.
Ames, W.M. (1944). J. SOC.Chem. Znd. (London) 63, 200.
Ames, W.M. (1952). J. Sci, Food Agr. 3, 47, 579.
Ames, W. M. (1957). J . Sci, Food Agr. 8, 169.
Astbury, W.T. (1940). J . Intern. Soc. Leather Trades’ Chemists 24,69.
Astbury, W . T. (1949). Nature 163, 722.
Astbury, W.T., and Atkin, W. R. (1933). Nature 132, 348.
Astbury, W.T., Dalgliesh, C. E., Darmon, S. E., and Sutherland, G. B. B. M. (1948).
Nature 162, 596.
Badger, R. M., and Pullin. A. D. E. (1954). J . Chem. Phys. 22, 1142.
Bamford, C. H., Brown, L., Cant, E. M., Elliott, A,, Hanby, W. E., and Malcolm,
B. R. (1955). Nature 176, 396.
Bamford, C. H., Elliott, A . , and Hanby, W. E. (1956). “Synthetic Polypeptides,” p.
283. Academic Press, New York.
Bear, R. S. (1942). J . Am. Chem. SOC.64, 727.
Bear, R. S. (1944). J. Am. Chem. SOC.66, 1297.
Bear, R. S. (1952). Advances in Protein Chem. 7, 69.
Bear, R. S. (1955). In “Fibrous Proteins and Their Biological Significance.” Sym-
posia SOC.Exptl. Biol. 9, 97.
Bear, R. S. (1956). J . Biophys. Biochem. Cytol. 2, 363.
Bear, R . S., and Morgan, R. S. (1957). In “Connective Tissue,” Council Intern. Org.
Med. Sci. Symposium (R. E. Tunbridge, ed.), p. 321. Blackwell, Oxford, England.
Bear, R. S., Bolduan, 0,E. A., and Salo, T.P. (1951). J . Am. Leather Chemists’ Assoc.
46, 107.
Becker, R.,and Doring, W. (1935). Ann. Physik 24, 719.
Beecham, A. F., Fridrichsons, J., and Mathieson, A. McL. (1958). J. Am. Chent. Soc.
80,4739.
Beek, J. (1941). J. Am. Leather Chemists’ Assoc. 36, 696.
Beer, M. (1956). Proc. Ray. SOC.A236, 136.
Beer, M.,Sutherland, G. B. B. M., Tanner, K. N., and Wood, D. L. (1958). Proc. Roy.
Soc. M 4 9 , 147.
Bello, J. (1959). Trans. Faraday SOC.66, 2130.
Bello, J. (1960). Nature 186, 241.
Bello, J., and Vinograd, J. R. (1958). Nature 181,273.
Bello, J. Riese, H. C. A., and Vinograd, J. R. (1956). J. Phys. Chem. 60. 1299.
Bensusan, H. B. (1960). J. Am. Chem. SOC.82, 4995.
Bensusan, H. B., and Hoyt, B. L. (1958). J. Am. Chem. SOC.80, 719.
Bensusan, H. B., and Scanu, A. (1960). J . Am. Chem. SOC.82, 4990.
Berger, A., Kurtz, J., and Katchalski, E. (1954a). Bull Research Council Israel 4A, 113.
Berger, A., Kurtz, J., and Katchalski, E. (1954b). J. Am. Chem. SOC.76, 5552.
Berger, A., Lowenstein, A., and Meiboom, S. (1959). J . Am. Chem. SOC.81, 62.
Bergmann, M. (1935). J. Biol. Chem. 110, 471.
Bergmann, M., and Niemann, C. (1938). Ann. Rev. Biochem. 7, 99.
Bernal, J . D. (1931). 2. Krist. 78, 363.
Betheil, J. S., and Gallop, P. M. (1960). Biochim. et Biophys. Acta 46, 598.
Beveridge, J. M. R., and Lucas, C. C. (1944). J. Biol. Chem. 166, 547.
Bigelow, C. C., and Geschwind, I. I. (1961). Compt. rend. trau. lab. Carlsberg 32,
89.
Bijvoet, J. M., Peerdeman, A. F., and Van Bommel, A. J. (1951). Nature 168, 271.
Blout, E. R. (1960). In “Optical Rotatory Dispersion” (C. Djerassi, ed.), p. 238.

McGraw Hill, New York.
Blout, E. R., and Fasman, G. D. (1958). In “Recent Advances in Gelatin and Glue
Research” (G. Stainsby, ed.), p. 122. Pergamon Prese, London.
Boedtker, H., and Doty, P. (1954). J . Phys. Chem. 68, 968.
Boedtker, H., and Doty, P. (1955). J . A m . Chem. SOC.77, 248.
Boedtker, H., and Doty, P. (1956). J . A m . Chem. SOC.78, 4267.
Bowes, J . H., and Kenton, R. H. (1948). Biochem. J . 43, 358.
Bowes, J. H., and Moss, J. A. (1953). Biochem. J . 66, 735.
Bradbury, E. M., Burge, R. E., Randall, J. T., and Wilkinson, G. R . (1958). Discus-
sions Faraday SOC.26, 173.
Bresler, S. E., Finogenov, P. A., and Frenkel, S. Y. (1950). Doklady Akad. Nauk
S.S.S.R. 72, 555.
Brunauer, S., Emmett, P. H., and Teller, E. (1938). J . A m . Chem. SOC.60, 309.
Buchner, E. H., and Meylink, B. (1943). Rec. trav. chim. 62, 337.
Burge, R. E., and Hynes, R. D. (1959a). J . Mol. Biol. 1, 155.
Burge, R. E., and Hynes, R. D. (1959b). Nature 184, 1562.
Burge, R. E., and Randall, J. T. (1955). Proc. Roy. SOC.A233, 1 .
Burge, R. E., Cowan, P. M., and McGavin, S. (1958). In “Recent Advances in Gelatin
and Glue Research” (G. Stainsby, ed.), p. 25. Pergamon Press, London.
Carpenter, D. C. (1927). J . Phys. Chem. 31, 1873.
Carpenter, D. C. (1938). Cold Spring Harbor Symposia Quant. Biol. 6,244.
Carpenter, D. C., and Kucera, J . J. (1931). J . Phys. Chem. 36, 2619.
Carpenter, D. C., and Lovelace, F. E. (1935a). J . Am. Chem. SOC.67, 2342.
Carpenter, D. C., and Lovelace, F. E. (1935b). J . A m . Chem. SOC.67, 2337.
Carpenter, D. C., and Lovelace, F. E. (1936). J. Am. Chem. SOC. 68, 2438.
Carpenter, D. C., and Lovelace, F. E. (1937). J . A m . Chem. SOC.69, 2213.
Carpenter, D. C., and Lovelace, F. E. (1938). J . A m . Chem. SOC.60,2289.
Chun, E. H. L., and Doty, P. (1958). I n “Recent Advances in Gelatin and Glue Re-
search” (G. Stainsby, ed.), p. 261. Pergamon Press, London.
Clark, G. L., Parker, E. A., Schaad, J. A., and Warren, W. J. (1935). J . Am. Chem.
Soc. 67, 1509.
Cochran, W., Crick, F. H. C., and Vand, V. (1952). Acta Cryst. 6, 581.
Cohen, C. (1955). J. Biophys. Biochem. Cytol. 1 , 203.
Cohen, C., and Bear, R. S. (1953). J . A m . Chem. SOC.76, 2783.
Connell, J. J. (1960). Biochem. J . 76, 530.
Corey, R . B., and Pauling, L. (1953). Proc. Roy. SOC.B141, 10.
Courts, A. (1954). Biochem. J. 68, 70, 74.
Courts, A. (1958). In “Recent Advances in Gelatin and Glue Research” (G. Stainsby,
ed.), p. 145. Pergamon Press, London.
Courts, A. (1959). Biochem. J . 73, 596.
Courts, A. (1960). Biochem. J . 74, 238.
Courts, A., and Stainsby, G. (1958). In “Recent Advances in Gelatin and Glue Re-
Cowan, P. M., and Brirge, R. T. (1958). Reported in the Gelatin and Glue Research
Discussions. See Downie, A . R., and Randall, A. A . , In “Central Leather Re-
search Institute Symposium on Collagen” (N. Ramanathan, ed.). Interscience,
New York. In press.
Cowan, P. M., and McGavin, S. (1955). Nature 170, 501.
Cowan, P. M., North, A. C. T., and Randall, J. T. (1953). In “Nature and Structure
of Collagen” (J. T. Randall, ed.), p. 241. Butterworth, London.
Cowan, P. M. McGavin, S., a n d North, A. C. T. (1955a). Nature 176, 1062.

Cowan, P. M., North, A. C. T., and Randall, J . T. (3955b). “Fibrous Proteins and
Their Biological Significance.” Symposia SOC. Exptl. Biol. 9, 115.
Crick, F. H. C. (1953). Acta Cryst. 6, 685, 689.
Crick, F. H. C. (1954). J . Chem. Phys. 22, 347.
Crick, F. H. C., and Rich, A. (1955). Nature 176, 780.
Davis, P. (1957). Trans. Faraday SOC.63, 1390.
Deasy, C. (1958). J . A m . Leather Chemists’ Assoe. 63, 203.
DeBellis, R., Msndl, I., MacLennan, J. D., and Howes, E. L. (1954). Nature 174,
1191.
Derksen, J. C. (1935). Dissertation, Amsterdam, Holland.
Donohue, J., and Trueblood, K . N. (1952). Acta Cryst. 6, 414.
Doty, P., and Nishihara, T. (1958). In “Recent Advances in Gelatin and Glue Re-
Downie, A. R., and Randall, A. A. (1959). Trans. Faraday SOC.66, 2132.
Downie, A. R., and Randall, A. A. (1961) In “Central Leather Research Institute
Symposium on Collagen” (N. Ramanathan, ed.). Interscience, New York. I n
press.
Eastoe, J. E. (1955). Biochem. J . 61, 589.
Eastoe, J . E. (1957). Biochem. J . 66, 363.
Eastoe, J. E., and Leach, A. A. (1958). In “Recent Advances in Gelatin and Glue
Research” (G. Stainsby, ed.), p. 173. Pergamon Press, London.
Edsall, J. T. (1954). J . Polynzer Sci. 12, 253.
Elliott, A., and Malcolm, B. R . (1956). Trans. Faraday SOC.62, 528.
Esipova, N. G. (1957). Biofizika 2, 455.
Esipova, N. G., Andreeva, N. S., and Gatovskaia, T. V. (1058). Biofziku 3, 529.
Ewald, A. Z. (1919). 2.physiol. Chem. Hoppe-Seyler’s 106, 115, 135.
Fasman, G. D., and Blout, E. R. (1961). (Personal communication.)
FaurB-Fremiet, M. E., and Garrault, A. (1937). Arch. Anat. Microbial. 33,81.
Ferry, J. D. (1948a). Advances i n Protein Chem. 4, 1.
Ferry, J. D., (1948b). J . A m . Chem. Soc. 70, 2244.
Ferry, J. D., and Eldridge, J. E. (1949). J . Phys. & Colloid Cheni. 63, 184.
Fessler, J. H. (1960). Biochem. J . 76, 452, 463.
Flory, P. J. (1949). J . Chem. Phys. 17, 223.
Flory, P. J. (1953). “Principles of Polymer Chemistry.” Cornell Univ. Press, Ithaca,
New York.
Flory, P. J. (1961). J . Polymer Sci. 49,105.
Flory, P. J., and Garrett, R. R . (1958). J . Am. Chem. Soc. 80, 4836.
Flory, P. J., and McIntyre, A. D. (1955). J . Polymer Sci. 18, 592.
Flory, P. J., and Weaver, E. S. (1960). J . Am. Chem. Soc. 82, 4518.
Fraser, R. D. B., and MacRae, T. P. (1959). Nature 183, 179.
Fuller, W. (1959). J . Phys. Chem. 63, 1705.
Gallagher, K. J. (1959). In “Hydrogen Bonding” (D. Hadsi, ed.), p. 45. Pergamon
Press, New York.
Gallop, P. M. (1955a). Arch. Biochem. Biophys. 64, 486.
Gallop, P. M. (1955b). Arch. Biochem. Biophys. 64, 501.
Gallop, P. M., and Seifter, S. (1961). In “Central Leather Research Institute Sym-
posium on Collagen (N. Ramanathan, ed.). Interscience, New York. In press.
Gallop, P. M., Seifter, S., and Meilman, E. (1957a). J . Biophys. Biochem. Cy t ol . 3,
646.
Grtllop, P. M., Seifter, S., and Meilman, E. (1957b). J . Biol. Chem. 227, 891.
Gallop, P. M., Seifter, S., and Meilman, E. (1959). Nature 183, 1659.
Gallop, P.M., Seifter, S., Lukin, M., and Meilman, E. (1960). J . Biol. Chem. 236,
2619.
Garrett, R. R., and Flory, P. J . (1956). Nature 177, 176.
Gibbs, J. H., and DiMarzio, E. A. (1958). J . Chem. Phys. 98, 1247.
Glegg, E., Eidinger, D., and Leblond, C. P. (1953). Science 118, 614.
Gordon, A. H. (1948). Nature 169, 778.
Gordon, R. S., and Ferry, J. D. (1946). Federation Proc. 6, 136.
Gouinlock, E. V., Flory, P. J., and Scheraga, H. A. (1955). J . Polymer Sci. 16, 383.
Grabar, P., and Morel, J. (1950). Bull. S O C . chim. biol. 32, 643.
Grassmann, W. (1936). Kolloid - 2. 77, 205.
Grassmann, W. (1955). Leder 10, 241.
Grassmann, W. (1960). Svensk Kem. Tidskr. 72, 275.
Grassmann, W., and Schleich, H. (1935). Biochem. 2.227, 320.
Grassmann, W . , Endres, H., and Steber, A. (1954). 2. Naturforsch. 9B,513.
Grassmann, W., Hannig, K . , Endres, H., and Riedel, A. (1956). 2. physiol. Chem.
Hoppe-Seyler’s 306, 123.
Grassmann, W., Hormann, H., and Hafter, R . (1957a). 2. physiol. Chem. Hoppe-
Seyler’s 307, 87.
Grassmann, W., Hofmann, U. Kuhn, K., Hormann, H., Endres, H., and Wolf, K.
(1957b). In “Connective Tissue.” Council Intern. Org Med. Sci. Symposium
(R. E. Tunbridge, ed.), p. 157. Blackwell, Oxford.
Grassmann, W., Hormann, H., Nordwig, A., and Wunsch, E. (1959). Z . physiol. Chem.
Hoppe-Seyler’s 316, 287.
Grassmann, W., Hannig, K., and Schleyer, M. (1960). 2. physiol. Chem. Hoppe-Seyler’s
322, 71.
Grassmann, W . , Hannig, K., and Engel, J. (1961). 2. physiol. Chem. Hoppe-Seyler’s
324, 284.
Gross, J. (1956). J . Biophys. Biochem. Cytol. 2 , 261.
Gross, J. (1958). J . Exptl. Med. 107, 265.
Gross, J., and Dumsha, B. (1958). Biochim. et Biophys. Acta 28, 268.
Gross, J., and Kirk, D. (1958). J . Biol. Chem. 233, 355.
Gross, J., and Piez, K. A. (1960). I n “Calcification in Biological Systems” (R. F.
Sognnaes, ed.), p. 395. Am. Assoc. Advancement Sci., Washington, D. C.
Gross, J., and Schmitt, F. 0. (1948). J . Exptl. Med. 88, 555.
Gross, J., Highberger, J. H., and Schmitt, F. 0. (1952). Proc. SOC.Exptl. Biol. Med.
80, 462.
Gross, J., Highberger, J. H., and Schmitt, F. 0. (1954). Proc. Natl. Acad. Sci. U . S .
40, 679.
Gross, J., Highberger, J. H., and Schmitt, F. 0. (1955a). Proc. Natl. Acad. Sci. U.S.
41, 1.
Gross, J., Matoltsy, A. G., and Cohen, C. (1955b). J . Biophys. Biochem. Cytol. 1, 215.
Gross, J., Sokal, Z., and Rougvie, M. A. (1956). J . Histochem and Cytochem. 4, 227.
Gross, J., Dumsha, B., and Glazer, N. (1958). Biochim. et Biophys. Acta 30, 293.
Gustavson, K . H. (1953). Svensk Kem. Tidskr. 66, 70.
Gustavson, K . H. (1954). Acta Chem. Scand. 8 , 1300.
Gustavson, K. H. (1955a). Svensk Kem. Tidskr. 67, 116.
Gustavson, K. H. (1955b). Nature 176, 70.
Gustavson, K. H. (1956). “The Chemistry and Reactivity of Collagen.” Academic
Press, New York.
Gustavson, K.H. (1957). In “Connective Tissue.” Council Intern. Org. Med. Sci.
Symposium (R. E. Tunbridge, ed.), p. 185. Blackwell, Oxford, England.
Hall, C. E. (1956).Proc. Natl. Acad. Sci. U.S. 42, 801.
Hall, C. E., and Doty, P. (1958).J . Am. Chem. SOC.80, 1269.
Hall, C. E.,Jakus, M. A., and Schmitt, F. 0. (1942).J . Am. Chem. SOC.64, 1234.
Hanby, W. E., Waley, 5. G., and Watson, J. (1950).J . Chem. SOC.p. 3009.
Harkness, R. D., Marko, A. M., Muir, H. M., and Neuberger, A. (1954).Biochem. J .
66,658.
Harrington, w. F. (1958). Nature 181, 997.
Harrington, W.F.,and Schellman, J. A. (1957).Compt. rend. trau. lab. Carlsberg SO,
167.
Harrington, W. F., and Sela, M. (1958). Biochim. et Biophys. Acta 27, 24.
Harrington, W. F.,and von Hippel, P.H. (1961).Arch. Biochem. Biophys. 92, 100.
Harrington, W. F., von Hippel, P. H., and Mihalyi, E. (1959). Biochim. et Biophys.
Acta 32, 303.
Hastewell, L. J., and Roscoe, R. (1956). Brit. J . Appl. Phys. 7, 441.
Hermann, K.,and Gerngross, 0. (1932).Kautschuck 8, 181.
Hermann, K.,Gerngross, O., and Abitz, W. (1930). 2. physik. Chem. (Leipzig) B10,
371.
Hermans, J., Jr., and Scheraga, H. A. (1959).Biochim. el Biophys. A d a 36, 534.
Heraog, R. O.,and Gonell, H. W. (1925).Naturwissenschaften 12, 1153.
Heymann, E. (1936). Trans. Faraday SOC.32, 462.
Heyns, K.,and Legler, G. (1959).Z.physiol. Chem. Hoppe-Seyler’s 316,288.
Heyns, K.,and Legler, G. (1960).Z.physiol. Chem. Hoppe-Seyler’s 321, 184.
Heyns, K.,Anders, G., and Becker, E. (1951).2. physiol. Chem. Hoppe-Seyler’s 287,
120.
Highberger, J. H., Gross, J., and Schmitt, F. 0. (1950).J . Am. Chem. SOC.72,3321.
Highberger, J. H., Gross, J., and Schmitt, F. 0. (1951).Proc. Natl. Acad. Sci. U S .
37, 286.
Hirai, N. (1953).J . Chem. SOC.Japan, Pure Chem. Sect. 74,347,441,443,539,593,810.
Hodge, A. J., and Schmitt, F. 0. (1958). Proc. Null. Acad. Sci. U.S. 44, 418.
Hodge, A. J., and Schmitt, F. 0. (1960).Proc. Natl. Acad. Sci. U.S. 46, 186.
Hodge, A.J., Highberger, J. H., Deffner, G. G. J., and Schmitt, F. 0. (1960).Proc.
Natl. Acad. Sci. U.S. 46, 197.
Holleman, L. W. J., Bungenberg de Jong, H. G., and Modderman, R.S. T. (1934).
Kolloid-Beihe. 40,211.
Hormann, H. (1960).Leder 11, 173.
Hormann, H., and Fries, G. (1958). 2. physiol. Chem. Hoppe-Seyler’s 311, 19.
Huggins, M.L. (1943).Chem. Revs. 32, 195.
Huggins, M. L. (1954).J . Am. Chem. SOC.78, 4045.
Huggins, M. L. (1957).Proc. Natl. Acad. Sci. U.S. 43, 209.
Hughes, E. W.,and Moore, W.J. (1949). J . Am. Chem. SOC.71, 2G18.
Hughes, E. W., Biswas, A. B., and Wilson, J. N . , cited by Pauling, L., and Corey,
R.B. (1954).Forlschr. Chem. org. Naturstofe 11, 190.
Jackson, D. S. (1956).In “Connective Tissue.” Council Intern. Org. Med. SOC.Sym-
posium (R.E.Tunbridge, ed.), p. 62. Blackwell, Oxford, England.
Jackson, D. S. (1957). Biochem. J . 86, 277.
Jackson, D. S. (1958). In “Recent Advances in Gelatin and Glue Research” (G.
Stainsby, ed.), p. 50. Pergamon Press, London.
Jackson, D. S., and Bentley, J. P. (1960).J . Biophys. Biochem. Cylol. 7.37.
Jackson, D. S.,and Fessler, J. H. (1955).Nature 176, 69.
Jackson, S. F., and Randall, J. T. (1953). In “Nature and Structure of Collagen”
(J. T. Randall, ed.), p. 181. Butterworth, London.
Johnston, J. P., and Ogston, A. G. (1946). Trans. Faraday SOC.43, 789.
Jopling, D. W. (1956). Bull. Brit. SOC.Rheol. 46, 2.
Kaesberg, P., and Shurman, M. M. (1953). Biochim. e t Biophys. Aeta 11, 1.
Katchalski, E.,Kurtz, J., Fasman, G., and Berger, A. (1956). Bull. Research Council
Israel Sect. 6A, 264.
Katz, J. R. (1933). Trans. Faraday SOC.29, 279.
Katz, J. R., and Derksen, J. C. (1932). Rec. trav. chim. 61, 513.
Katz, J. R., and Wienhoven, J. F. (1933). Rec. trav. chim. 63, 36.
Katz, J. R., Derksen, J. C., and Bon, W. F. (1931). Rec. trav. chim. 60,725.
Kazakova, 0. V., Orekhovich, V. N., and Shpikiter, V. 0. (1958). Doklady Akad.
Nauk S.S.S.R. 122, 657.
Keiichington, A. W. (1958). Biochem. J . 68, 458.
Kenchington, A. W., and Ward, A. G. (1954). Biochem. J . 68, 202.
Kendrew, J. C. (1954). In “The Proteins” (H. Neurath and K. Bailey, eds.), Vol. 11,
Part B, p. 845. Academic Press, New York.
Konno, K., and Altman, K. I. (1958). Nature 181, 994.
Kraemer, E. O., and Fanselow, J. R. (1925). J . Phys. Chem. 29, 1167.
Kramer, H., and Little, K . (1953). In “Nature and Structure of Collagen” (J. T.
Randall, ed.), p. 33. Academic Press, New York.
Kratky, O., and Sekora, A. (1943). J . makromol. Chem. 1, 113.
Kroner, T. D., Tabroff, W., and McGarr, J. J. (1953). J . A m . Chem. SOC.76,4084.
Kroner, T. D., Tabroff, W., and McGarr, J. J. (1955). J . A m . Chem. SOC.77, 3356.
Kuhn, K., Grassmann, W., and Hofmann, U. (1957). Naturwissenschaften 44, 538.
Kuhn, K., Grassmann, W., and Hofmann, U. (1958a). 2. Naturforsch. 13B. 154.
Kuhn, K., Hofmann, U., Grassmann, W., and Gebhardt, E. (1958b). Naturwissen-
schaften 46, 521.
Kuhn, K., Grassmann, W., and Hofmann, U. (1959). 2. Naturforsch. 14B, 436.
Kuntzel, A. (1937). In “Stiansny Festschrift” (K. H. Gustavson, ed.), p. 191. Roether,
Darmstadt.
Kuntzel, A., Cars, N., and Heidemann, E. (1958). In “Recent Advances in Gelatin
Kurtz, J., Berger, A,, and Katchalski, E. (1956). Nature 178, 1066.
Kurtz, J., Berger, A., and Katchalski, E. (1957). Bull. Research CounciE Israel Sect.
6A, 319.
Kurtz, J., Berger, A., and Katchalski, E. (1958a). I n “Recent Advances in Gelatin
Kurtz, J., Fasman, G., Berger, A., and Katchalski, E. (1958b). J . Am. Chem. Soc. 80,
393.
Kurtz, J., Harrington, W. F., Sela, M., Berger, A,, and Katchalski, E. (1960). Unpub-
lished.
Lakshmanan, B. R., Ramakrishnan, C., Sasisekharan, V., and Thathachari, Y. T.
(1961). I n “Central Leather Research Institute Symposium on Collagen (N.
Ramanathan, ed.). Interscience, New York. In press.
Lauritzen, J. I., and Hoffman, J. D. (1959). J . Chem. Phys. 31, 1680.
Leach, A. A. (1957). Biochem. J . 67, 83.
Lenhoff, H. M., Kline, E. S., and Hurley, R. (1957). Biochim. et Biophys. Acta 26,204.
Leplat, G. (1933). Compt. rend. S O C . bid. 112, 1256.
Leung, Y. C . , and Marsh, R. E. (1957). Acta Cryst. 11, 17.
Levy, M.,Fishman, L., and Cabrera, G. (1960). Federation Proc. 19, 343.
Lewis, M. S., and Pier, K. A. (1961). Federation Proc. 20, 380.

LinderstrZm-Lang, K. U. (1952).In “Proteins and Enzymes” (Lane Memorial Lec .
tures) p. 53. Stanford Univ. Press. Stanford, California.
Lindley, H. (1955). Biochim. et Biophys. Acta 18, 194.
MacLennan, J. D., Mandl, I., and Howes, E. L. (1953). J. Clin. Invest. 32, 1317.
Mandelkern, L. (1956). Chem. Revs. 66, 903.
Mandl, I., MacLennan, J. D., and Howes, E. L. (1953). J. Clin. Invest. 32, 1329.
Marks, M. H., Bear, R. S., and Blake, C. H. (1949). J. Ezptl. 2001.111, 55.
Marsh, R. E. (1958). Acta Cryst. 11, 654.
Martin, G. R., Mergenhagen, S. E., and Scott, D. B. (1961). Biochim. et Biophys.
Acta 49, 245.
Mathews, M. B., Kulonen, E., and Dorfman, A. (1954).,Arch. Biochem. Biophys. 62,
247.
Mathieson, A. McL., and Welsh, H. K. (1952). Acta Cryst. 6, 599.
Mazourov, V. I., and Orekhovich, V. N. (1959). Biokhimiya 24, 33.
Mazourov, V. I., and Orekhovich, V. N. (1960). Biokhimiya 26, 814.
Mechanic, G., and Levy, M. (1959). J . Am. Chem. SOC.81, 1889.
M e g a , A. B., and Sikorski, J. (1956). Nature 177, 326.
Melnick, S. C. (1958). Nature 181, 1483.
M’Ewen, M. B., and Pratt, M. I. (1953). In “Nature and Structure of Collagen” (J.
T. Randall, ed.), p. 158. Academic Press, New York.
Meyer, K. H., and Go, Y. (1934). Helu. Chim. Acta 17, 1488.
Michaels, S., Gallop, P. M., Seifter, S., and Meilman, E. (1958). Biochim. e t Biophys.
Acta 29, 450.
Mihalyi, E., and Harrington, W. F. (1959). Biochim. et Biophys. Acta 36, 447.
Millionova, M. I., and Andreeva, N. S. (1957a). Biojizika 2, 45.
Millionova, M. I., and Andreeva, N. S. (1957b). Biojizika 2, 292.
Millionova, M. I., and Andreeva, N. S. (1958). Biofizika 3, 259.
Montroll, E . W., and Simha, R. (1940). J . Chem. Phys. 8, 721.
Mosimann, H., and Signer, R. (1944). “Svedberg Memorial Volume,” p. 464. Alm-
quist and Wilksells, Uppsala.
Nagai, Y., and Noda, H. (1959). Biochim. el Biophys. Acta 34, 298.
Nagai, Y., Sakakibara, S., Noda, H., and Akabori, S. (1960). Biochim. e t Biophys.
Acta 37, 567.
Nageotte, J. (1927). Compt. rend. soc. biol. 96, 172; Compt. rend. acad. sci. 184, 115.
Neiman, R. E. (1952). Colloid J. (U.S.S.R.) (Engl. Transl.) 14, 170.
Neiman, R . E. (1954). Colloid J . (U.S.S.R.) (Engl. Transl.) 16, 280.
Neuberger, A. (1948). Advances in Protein Chem. 4, 297.
Neuman, R . E. (1949). Arch. Biochem. 24, 289.
Nishigai, M., Nagai, Y., and Noda, H. (1960). J . Biochem. (Yokyo) 48, 153.
Nishihara, T., and Doty, P. (1958). Proc. Natl. Acad. Sci. U.S. 44, 411.
Noda, H. (1955). Biochim. et Biophys. Acta 17, 92.
Orekhovich, V. N. (1952). “Procollagens, Their Chemical Structure, Properties and
Biological Role.” Moscow.
Orekhovich, V. N., and Shpikiter, V. 0. (1955a). Biokhimiya 20, 438.
Orekhovich, V. N., and Shpikiter, V. 0. (1955b). Doklady Akad. Nauk S.S.S.R. 101,
529.
Orekhovich, V. N., and Shpikiter, V. 0. (1958a). Science 127, 1371.
Orekhovich, V. N., and Shpikiter, V. 0. (1958b). In “Advances in Gelatin and Glue
Orekhovich, V. N., Tustanovskii, A. A., Orekhovich, K . D., and Plotnikova, N. E.

(1948). Biokhimiya 13, 55.
Orekhovich, V. N., Paulikhina, L. V., and Shpikiter, V. 0. (1957). Biokhimiya 22,
210.
Orekhovich, V. N., Shpikiter, V. O., Mazourov, V. I., and Kounina, 0. V. (1960).
Bull. S O C . chim. biol. 42, 505.
Pauling, L. (1940). J . Am. Chem. SOC.62, 2643.
Pauling, L. (1958). I n “Recent Advances in Gelatin and Glue Research” (G. Stainsby,
ed.), p. 11. Pergamon Press, London.
Pauling, L., and Corey, R. B. (1951a). Proc. Natl. Acad. Sci. U.S. 37, 236.
Pauling, L.,and Corey, R . B. (1951b). Proc. Null. Acad. Sci. U.S. 37, 251.
Pauling, L., and Corey, R. B. (1951~).Proc. Natl. Acad. Sci. U.S. 37, 272.
Pauling, L., and Sherman, J. (1933). J. Chem. Phys. 1, 606.
Pauling, L., Corey, R. B., and Branson, H. R. (1951). Proc. Natl Acad. Sri. U.S. 37,
205.
Peng, Chia-Mu, and Tsao, Tien-Chin (1956). Sci. Sinica (Peking) 6 , 691.
Piel;, K. A. (1960). J. A m . Chem. Soc. 82, 247.
Piez, K. A . , and Gross, J. (1959). Biochim. et Biophys. Acta 34, 24.
Piez, K . A., and Gross, J. (1960). J. Biol. Chem. 236, 995.
Piez, K. A., and Likins, R. C. (1960). In “Calcification in Biological Systems” (R. F.
Sognnaes, ed.), p. 411. Am. Assoc. Advancement. Sci., Washington, D.C.
Pies, K. A., Weiss, E., and Lewis, M. S. (1960). J. Biol. Chem. 236, 1987.
Piez, K. A., Lewis, M. S., Martin, G. R., and Gross, J. (1961). Biochim. et Biophgs.
Acta (In press.)
Pleass, W. B. (1930). Proc. Roy. SOC.Al26, 406.
Poroshin, K. T., Kozarenko, T. D., Shibnev, V. A., and Debabov, V. G. (1960). Izvest.
Khim. Nauk p. 550.
Pouradier, J. (1958). In “Recent Advances in Gelatin and Glue Research”
(G. Stainsby, ed.), p. 265. Pergamon Press, London.
Pouradier, J., and Venet, A. M. (1950). J. chim. phys. 47, 11.
Pouradier, J., Venet, A. M., and Trigny, L. (1954). Compt. rend. congr. intern. ehim.
ind. 273, Congr., Brussels. 709.
Price, F. P. (1959). J. Chem. Phys. 31, 1679.
Ramachandran, G. N. (1955). J . Chem. Phys. 23, 600.
Ramachandran, G. N. (1956). Nature 177, 710.
Ramachandran, G. N. (1958). In “Recent Advances in Gelatin and Glue Research”
Ramachandran, G. N. (1961). I n “Central Leather Research Institute Symposium
on Collagen” (N. Ramanathan, ed.). Interscience, New York. In press.
Ramachandran, G. N., and Kartha, G. (1954). Nature 174, 269.
Ramachandran, G. N., and Kartha, G. (1955). Nature 176, 593.
Ramachandran, G. N., Sasisekharan, V., and Thathachari, Y. T. (1961). In “Central
Leather Research Institute Symposium on Collagen” (N. Ramanathan, ed.). In-
terscience, New York. In press.
Randall, J. T. (1954). J . SOC.Leather Trades‘ Chemists 38, 362.
Randall, J. T., Fraser, R. D. B., Jackson, S. F., Martin, A. V. W., and North, A. C. T.
(1952). Nature 169, 1029.
Randall, J. T., Brown, G. L., Jackson, S. F., Kelly, F. C., North, A. C. T., Seeds,
W. E., and Wilkinson, G . R. (1953a). In “Nature and Structure of Collagen”
(J. T. Randall, ed.), p. 213. Butterworth, London.
Randall, J. T., Fraser, R. D. B., and North, A. C. T. (195313). Proc. Roy SOC.B141,
62.
Randall, J. T., Booth, F., Burge, R. E., Jackson, S. P., and Kelly, F. C. (1955). In
“Fibrous Proteins and Their Biological Significance.” Symposia SOC.ExptZ.
B i d . 9, 127.
Rice, R. V. (1960). Proc. Natl. Acad. Sci. U.S. 46, 1187.
Rich, A. Personal communication.
Rich, A., and Crick, F. H. C. (1955). Nature 176, 915.
Rich, A., and Crick, F. H. C. (1958). In “Recent Advances in Gelatin and Clue Re-
Rich, A., and Crick, F. H. C. (1959). Personal communication from A. Rich.
Rich, A , , and Crick, F. H. C. (1961). J . Mol. Biol. 3, 71.
Robb-Smith, A. H. T. (1957). I n “Connective Tissue.” Council Intern. Org. Med. Sci.
Symposium (R. E. Tunbridge, ed.), p. 177. Blackwell, Oxford, England.
Robb-Smith, A. H. T. (1958). I n “Recent Advances in Gelatin and Glue Research,”
Robinson, C. (1953). I n “Nature and Structure of Collagen,” (J. T. Randall, ed.), p.
96. Academic Press, New York.
Robinson, C., and Bott, M. J. (1951). Nature 168, 325.
Rougvie, M. A., and Bear, R . S. (1953). J . A m . Leather Chemists Assoc. 48, 735.
Sasisekharan, V. (1959a). Actu Cryst. 12, 897.
Sasisekharan, V. (1959b). Actu Cryst. 12, 903.
Sasisekharsn, V. (1961). In “Central Leather Research Institute Symposium on Col-
lagen” (N. Ramanathan, ed.). Interscience, New York. I n press.
Saunders, P. R., and Ward, A. G. (195%). In “Recent Advances in Gelatin and Glue
Saunders, P. R., and Ward, A. C. (195813). “Rheology of Elastomers,” p. 45. Perga-
mon Press, New York.
Scatchard, G., Oncley, J. L., Williams, J. W . , and Brown, A. (1944). J . A m . Chem.
Sac. 66, 1980.
Schellman, C., and Schellman, J. A. (1958). Compt. rend. trav. lab. Carlsberg SO, 463.
Schellman, J. A. (1955). Compt. rend. trav. Zab. CarZsberg 29, 230.
Schellman, J. A. (1958). J . Phys. Chem. 62, 1485.
Schellman, J. A , , and Schellman, C. (1961). J . Polymer Sci. 49, 129.
Schmitt, F. 0. (1956). Proc. Am. Phil. Sac. 100, 476.
Schmitt, F. 0. (1959). Revs. Modern Phys. 31, 349.
Schmitt, F. O., and Gross, J. (1948). J . A m . Leather Chemists’ Assoc. 43, G58.
Schmitt, F.O., Hall, C. E., and Jakus, M. A. (1942). J . CeZZular Comp. Physiol. 20,
11.
Schmitt, F. O., Hall, C. E., and Jakus, M. A. (1945). J . Appl. Phys. 16, 263.
Schmitt, F. 0.)Gross, J., and Highberger, J. H. (1953). Proc. Natl. Acad. Sci. U.S.
59, 459.
Schmitt, F. O., Gross, J., and Highberger, J. H. (1955). In “Fibrous Proteins and
Their Biological Significance.” Symposia SOC.Exptl. Biol. 9, 148.
Schneider, F. (1940). Collegium, p. 97.
Schroeder, W. A,, Honnen, L., and Green, F. C. (1953). Proc. N a t l . Acad. Sci. U.S.
39, 23.
Schroeder, W. A., Kay, L. M., LeGette, J., Honnen, L . , and Green, F. C. (1954).
J . Am. Chem. SOC.76,3556.
Schrohenloher, R. E., Ogle, J. I)., and Logan, M. A . (1959). J . Rial. Chem. 234, 58.
Schumaker, V. N., Richard, E. G., and Schachman, H. K . (1956). J . A m . Chem. Soc.

78, 4230.
Schuytema, E. C., and Kallio, R. E. (1956). Bacteriol. Proc. 63.
Seifter, S., Gallop, P. M., Klein, L., and Meilman, E. (1959). J . Biol. Chem. 234,285.
Sela, M., and Berger, A. (1955). J. Am. Chem. SOC.77, 1893.
Singleton, L. (1957). Biochim. et Biophys. Acta 24, 67.
Sinsheimer, R. L. (1959). J . Mol. Biol. 1, 43.
Smith, C. R. (1919). J . Am. Chem. SOC.41, 135.
Stainsby, G., Saunders, P. R., and Ward, A. G. (1954). J . Polymer Sci. 12, 325.
Steinberg, I. Z. (1958). Bull. Research Council Israel 74, 97.
Steinberg, I. Z . , Berger, A,, and Katchalski, E. (1958). Biochim. el Biophys. Acta 28,
647.
Steinberg, I. Z., Harrington, W. F., Berger, A., Sela, M., and Katchalski, E. (1960a).
J . A m . Chem. SOC.82, 5263.
Steinberg, I. Z., Sela, M., Harrington, W. F., Berger, A., and Katchalski, E. (1960b).
Bull. Research Council Israel 9A, 130.
Stiasny, E., das Gupta, S. R., and Tresser, P. (1925). Collegium p. 23.
Stokes, A. R. (1955). Progr. i n Biophys. and Biophys. Chem. 6, 140.
Stracher, A. (1960). Compt. rend. trav. lab. Carlsberg 31, 468.
Sutherland, G. B. B. M., Tanner, K. N., and Wood, D. L. (1954). J . Chem. Phys. 22,
1021.
Seent-Gyorgyi, A. G., and Cohen, C. (1957). Science 126, 697.
Tukahushi, T., und Gustavson, K. H. (1956). In “The Chemistry and Reactivity of
Collagens” (K. H. Gustavson, ed.), p. 225. Academic Press, New York.
Takahashi, T., and Tanaka, T. (1953). Bull. Japan. SOC.Sci. Fisheries 19, 603.
Thomas, C. A., Jr. (1956). J . Am. Chem. SOC.78, 1861.
Tomlin, S. G., and Worthington, C. R . (1956). Proc. Roy. SOC.A236, 189.
Tristram, G. R. (1953). I n “The Proteins” (H. Neurath and K. Bailey eds.), Val. I,
Part A, p. 181. Academic Press, New York.
Veis, A., and Anesey, J. (1959). J . Phys. Chem. 63, 1720.
Veis, A., and Cohen, J . (1956). J . Am. Chem. SOC.78, 6238.
Veis, A., and Cohen, J. (1957). J . Polymer Sci. 26, 113.
Veis, A., and Cohen, J. (1960). Nature 186, 720.
Veis, A., Anesey, J., and Cohen, J. (1958). In “Recent Advances in Gelatin and G ~ I W
Veis, A., Anesey, J,, and Cohen, J. (1960). J . Am. Leather (‘hemists’ Assac. 66, 548.
von Hippel, P. H. (1960). Unpublished.
von Hippel, P. H., and Harrington, W. F. (1959). Biochirn. e t Biophys. Acla 36, 427.
von Hippel, P. H., and Harrington, W. F. (1960). In “Protein Structure and Func-
tion.” Brookhaven Symposia i n B i d . 13, 213.
von Hippel, P. H., and Wong, K. Y. (1961). Unpublished data.
von Hippel, P. H., Gallop, P. M . , Seifter, S., and Cunningham, R . S. (1960). J. A m .
Chem. Soc. 82, 2774.
Waley, S . G., and Watson, J. (1949). Proc. Roy. SOC.AlQB, 499.
Watson, M. R. (1958). Biochem. J . 68, 416.
Watson, M. R . , and Silvester, N. R. (1959). Biochem. J . 71, 578.
Weir, C. E. (1949). J . Am. Leather Chemists Assoc. 44, 108.
WesseIy, F., and Sigmund, F. (1926). 2.physikal. Chem. (Leipzig) 169, 102.
Williams, A. P. (19GO). Biochem. J . 74, 304.
Williams, J. W. (1958). In “Recent Advances in Gelatin and Glue Research” (G.

Stainsby, ed.), p. 106. Pergamon Press, London.
Williams, J. W., Saunders, W. M., and Cicirelli, J. S. (1954). J . Phys. Chem. 68, 774.
Wohlisch, E. (1932). Biochem. 2.247,329.
Wolf, K. (1956). Dissertation, Munich, Germany.
Wolpers, C. (1943). Klin. Wochschr. 22, 624.
Wolpers, C. (1944). Klin. Wochschr. as, 169.
Wood, G. C. (1960a). Bioehem. J . 76, 598.
Wood, G. C. (1960b). Biochem. J . 76, 606.
Wood, G. C., and Keech, M. K. (1960). Biochem. J . 76, 588.
Wright, B. A., and Wiederhorn, N. W. (1961). J. Polymer Sci. 7 , 105.
Wyckoff, R. W. G., and Corey, R. B. (1936). Proc. SOC.Ezptl. Biol. M e d . 34, 285.
Wyckoff, R. W. G., Corey, R. B., and Biscoe, J. (1935). Science 82, 175.
Yang, J. T.(1961). Tetrahedron 13, 143.
Yang, J. T., and Doty, P. (1957). J. Am. Chem. Soc. 79, 761.
Yaron, A., and Berger, A. (1961). Bull Research Council Israel 10A, 46.
Young, E. G., and Lorimer, J. W. (1960). Arch. Biochem. Biophp. 88, 373.
Young, J. Z. (1950). “The Life of Vertebrates.” Clarendon Press, Oxford.
Zachariades, P. A. (1900). Compt. rend. soc. biol. 62, 182, 251, 1127.
Zimm, B. H., and Bragg, J. K. (1968). J . Chem. Phys. 28, 1246.
Zimrn, B. H., and Bragg, J. K. (1969). J . Chem. Phys. S1, 526.
Zussman, J . (1961). Acta Cryst. 4, 72, 493.

Collagen and Gelatin PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Collagen and Gelatin PDF

Uploaded by

Copyright:

Available Formats

THE STRUCTURE OF COLLAGEN AND GELATIN

PETER H. VON HIPPEL

comparative biologist is intrigued by the direct correlation between the

copolymers of glycine and proline. Consideration of such synthetic poly-

the problem of the identification of collagen and its distribution through

I n this section we propose to consider the effects of proline, hydroxy-

ally a t least three hydrogen bonds would be ruptured at such a discon-

A , A mino Acid Structures

Welsh (1952) have deduced structure parameters from a careful investiga-

later by Donahue and Trueblood (1952). As in copper-DL-prolinedihydrate,

B. Simple Peptides of Glycine, Proline, and Hydroxyproline

angle CK--N~-C, of 113" is close to the value accepted by Corey and

3. Tosyl-L-pro1 yl-L-h ydroxyproline Monohydrate

hydroxyproliiie sequence in the chains of collagen (see Section 111)endow it

droxyproline (Donahue and Trueblood, 1952) this carbon atom appears to

FIG.4a. Structureof polyglycine I1 (from Crick andRich, 1955). A projection down

X-ray evidence suggests that polyglycine I exists in the solid state as an

FIG.4b. Structure of polyglycine I1 (from Crick and Rich, 1955). A projection of

(A) (A) 4)O

Or-C 0 0.00 1.27 1.25 0 0.00

b. Studies on Polyglycine II in Solution. I n solution, polyglycine exhibits

we have for an understanding of the chemistry of collagen and gelatin.

(iii) Infrared spectra of poly-L-proline. Infrared spectra of poly-L-pro-

FIG.5. Structure of poly-L-proline. The figure on the right is a wire model of a

srries of cis -+ trans-trnrisformatioiis of thc prptide bond groupings in e w h

imide bond has a length of 1.34 A both in L-leucyl-L-prolylglycine and in the

Harrington m d Scla (1958) con(-luded from optical rotatory, sedinicnt a-

Potentiometric titration of simple proline derivatives and of poly-0-acetyl-

(1) (11) (111)

The type of mutarotation kinetics found experimentally is strongly

where C c i aand Ctransare the concentrations of cis- and trans-peptide bonds,

-c e_i s - [a1 - [a111

where k(Cci,/Co)fl-lstands for K’(t) and k is independent of time and defines

viscosity during the early phase of the mutarotation in water as solvent.

droxyproline in the solid state on the basis of X-ray diffraction patterns

[a]:' = 0" (form I) which changes to a terminal value, [a]:' = -120", on

Poly-L-prolineI1 - 14,300 Formic acid -640" -540"

found to be water-soluble. All of the copolymers were quite soluble in

residue. The disappearance of the form I1 structure with increasing glycine

FIQ.7. Specific rotation of various L-proline-glycine copolymers as a function of

number of contiguous proline residues are decreased, suggesting that

are required to achieve the complete helical contribution to optical rot,ation

A. Amino Acid Composition

the composition of a wide range of collagen species. Moreover, a n examina-

116 124 102 129

Alanine 113 114 103 69 70 66 56 94 72.3

Glucosamine 2.4 - - - 4.0 2.5 10 1.6 .92

In Table V are collected the most important peptides isolated by Heyns

Thr.Ala 1.1" Glu .Asp.Gly 0.5d Pro. (Gly,Lys) 1.0=

2. Peptides Derived from Enzymatic Hydrolyzates

of 33 amino acids was made up primarily of proline, hydroxyproline, and

links between the individual polypeptide chains of a single collagen mole-

A . Structural Studies in the Solid State

1. The Polypeptide Chain Configuration of Collagen

spacings are not appreciably affected by hydration. &lore recently, Cowan

FIG.9. Three-chain collagen structure proposed by Ramachandran and Kartha

Finally, in 1954, Ramachandran and Kartha proposed a structure which,

1 Gly only Other residues may Gly only

2 Any residue Any residue includ- Any residue includ-

3 Gly only Any residue, includ- Any residue, includ-

Bonding of the Can make a hydro- Projects rrtdially

Position of the From the NH of residue From the NH of