James H.

White
Russell A. Lund
Department of Mathematics
University of California
at Los Angeles
Los Angeles, CA 90024-1555

Twist, Writhe, and Geometry of

a DNA Loop Containing Equally
Spaced Coplanar Bends

William R. Bauer
Department of Microbiology
School of Medicine
State University of New York
Stony Brook, NY 7 7 794-5222

The,fi,rmation of a topologicallj. closed DNA loop is important in tnany biological proccsscs.

including the regulation oftranscription. recombination, and replication. Modeling DNA u.c an
isotropic elastic rod, we use ,finite element analysis to show that the dcpendmcc o f t h e twist
{ATw) and the writhe (Wr) upon the linking numher deficit ( A l k ) is .rtrongl.v influenced h y
intrinsic bmdx We determine how the geometry of a DNA loop changes as a ,function r?f'the
number of'itnif~rrnlyspaced coplanar 20" bend.r, oriented s o as to open toward the center y f t h e
loop. We also calculate the geometry of DNA rods that are stnoothlj! bent to the same exlent.
The response qfboth ATw and Wr ( f a bent DNA to changes in ALk,fulls into one of three
categories, depending upon the number qfbends. For a single bend of 20". Wr increases tnonotonicallji w>ithALk and the change in ATw with distance i.s constant along the entire DNA
axis. For two to ten 20" bends, Wr passe~sjrstthrough a local maximum, then through a local
minimiim, and,finallji increases monotonically us ALk increases. For rlrvrn t o eighteen 20"
bends, Wr again vuriccs monotonicallv with ALk . For all nutnbcrs ofbends greater than t"o, the
ATw per unit length depends upon the distribiition ?f intrinsic bends, being constant between
any two adjoining bends but varying MYth their position relative to the cut location. Aceompun),ing these Alk-associated changes in Wr and ATw per unit length arc characteristic changes in
geometry that are spec(fc,foreach category. The results ofthere calculations raise the possibility
that intrinsic bends can serve as u contro1,factor in the biologica1,finctions associated with loop
j0rtnation in DNA. 0 I996 John Wiley & Sons, Inc.

INTRODUCTION
The control ofgene expression often depends upon
the bringing together of DNA sequences that are
separated by a considerable distance along the contour.'-4 In this process a segment ofthe DNA forms
a closed loop of several hundred to a few thousand
base pairs and has a linking number change ALk of

Received February 21, 1995; accepted June 5 , 1995.

Biopolyrners. Vol. 38, 235-250 ( 1996)
0 1996 John Wiley &Sons. Inc.

between zero and a small positive or negative

value. We are investigating the properties of such
closed loops by modeling DNA as an isotropic elastic rod and using finite element analysis ( F E A ) to
determine the conformation of minimum elastic
energy.
We previously used a continuum elastic model
to investigate four distinct initial configurations of

CCC 0006-3525/96/020235- I6

235

236

W'hitc, Lirnd, and Buiirr

the DNA.' These were a straight line axis; an axis

straight except for the presence of one 20" bend; an
axis containing two symmetrically placed, coplanar 20" bends; and an axis bent smoothly into a
circle or O-ring. We showed that the distribution of
ALk between twist ( ATw) and writhe ( Wr) varies
markedly among the four cases. For the initially
straight DNA, a considerable change in ALk is required before the axis begins to writhe. Up to this
point all distortion appears as changes in ATw. The
changes in ATM.are uniform along the axis at all
values of ALk. The presence of a single intrinsic
bend converts a portion of ALk immediately into
W r , but the remaining A T w is still distributed uniformly along the DNA axis. A second intrinsic
bend, coplanar and diametrically opposite the first,
produces an acceleration in the conversion of ALk
into W'r, at the expense of AT%,.The nature of the
axial twist distribution, in addition, is strikingly altered: AT113 is uniform in the regions between the
two bends but is discontinuous at the bend locations. Finally, in the O-ring case ALk is converted
rapidly into W'r, and ATw is sinusoidally distributed along the O-ring axis.
In the current investigation we use a beam elastic model and focus upon three additional effects of
uniformly spaced bent segments in determining
the response ofthe DNA structure to applied A L k .
First, we vary the number of bends in the loop
while keeping their initial orientations coplanar.
Second, we change the magnitude of the bend angles while keeping the bend distribution otherwise
unchanged. Third, instead of discrete bends, we investigate the properties of continuously curved
DNA. Changes in ALk are kept below those resulting in the intersection of distant chain segments,
thus avoiding assumptions regarding self-contact.
As before, we model the DNA as a linearly elastic
rod having the appropriate cross section and material moduli, namely the modulus of elasticity and
the shear modulus. The calculations are performed
primarily for a DNA of length 628 nm, or 1870
base pairs ( b p ) , about half the length of pBR322
DNA (436 1 b p ) . We also demonstrate that the
DNA length per se is not a factor that affects the
results.

THEORY AND MODELS

Several distinct stages are involved in modeling closed
DNA by beam element based FEA. First, multiple bends
are introduced, ranging in number from two to eighteen.

These are equally spaced, arranged so that the bend

planes coincide with the plane in which the relaxed molecule lies, and the curvature of each bend is directed toward the interior. These structures, which are initially
free of strain, are then closed. For an individual bend
angle of 20", the eighteen-bend model in already closed,
a construction that it shares with the O-ring model. Second, a location is chosen midway between two of the
bends, and the closed rod is cut perpendicular to its axis
at this site. Third, the two faces of this cut are rotated
relative to one another about the local tangent to the rod
axis and are then resealed. This process has the effect of
changing the linking number ofthe DNA fractionally by
the a m o u n t of rotation imposed. It is actually the DNA
winding number, which can take on fractional values.
that is changed by this process. The sign of ALk is negative for a counterclockwise rotation of the cut faces
and positive for a clockwise rotation. For a topological
domain that is anchored by a protein, the relative position of each backbone is fixed with respect t o its continuation across the interface, once the rotation is
complete. In this case the fractional change in winding,
a n invariant. is included in the value of Lk for the domain. Such a domain satisfies the usual relationship
between L k , AT", and M/'r.6
In response to the imposed change in boundary conditions, as represented by a change in ALk, the various
rod models undergo deformation. The problem is to calculate the new coordinates of each rod-i.e., its new
shape. This is accomplished by the methods of solid mechanics using the techniques of displacement-based
FEA.' T o accomplish this. the initial configurations are
first modeled by an assemblage of relatively short straight
beam elements, rigidly joined end to end along the DNA
rod. The longitudinal axes of adjacent elements are
aligned, except at locations of discrete bends. Continuously curved DNA is modeled by an assemblage of short
beam elements with a small angle between the axis of
each adjacent element. In all cases, deformation of the
DNA rod. produced by moving one end relative to the
other, occurs by distortion of the individual elements,
not by hinging motion ofthe joints.
Beam elements incorporate bending, twisting (torsion), and extensional distortions, specialized for the
case of a slender rod. As the DNA model is gradually
deformed, the fundamental displacement variables for
these elements are incremental translations (in three orthogonal directions) and incremental rotations (about
three orthogonal axes) at each end. Bending, torsional,
and extensional strain increments are obtained along individual element longitudinal axes via interpolation of
the end displacements. The mechanics formulation,
which involves nonlinear equations for large motions as
considered here. includes equilibrium requirements and
linear elastic material behavior.
Deformed configurations of the DNA rod models are
obtained by moving the ends of the rod according to a

Twist. Urithe, und Geometry ofu DNA Loop

prescribed path in an incremental fashion. These strain
increments. in terms of unknown displacement increments, are entered in linearized versions of the governing
mechanics. which are then solved for the displacement
variables. An iterative procedure is used to correct for
errors introduced by the linearization of the governing
equations, and the deformed shape is determined at the
end of each increment. The formulation employed is restricted to small rotations during each increment, but
there is no limit on the total accumulated rotation. This
restriction was avoided in our previous work based on
a continuum element f o r m ~ l a t i o nHowever,
.~
the beam
element formation has the advantage of requiring far
fewer displacement variables to describe the motion. At
each joint in the beam element model, six displacement
variables are needed (three translations and three
rotations). We have typically used 80-180 beam elements for the various configurations. Consequently, the
models have approximately 500-1 100 independent displacement variables that must be determined in each increment. In contrast, our previous models using continuum elements utilized approximately 3500 variables per
increment. This reduction in the number of unknowns
results in significant savings in computer time and allows
practical solution of these problems on a personal computer.
We use beam elements with circular cross sections.
Consequently, bending behavior is independent of orientation. The cross section has radius 1 nm, for which
the moment of inertia I is equal to 7r/4 nm.4 The elements have linear elastic material properties. The modulus ofelasticity E is estimated from the persistence length
X with the expression E = X k B T / I .where k B is the Boltzmann constant and T the temperature (298 K ) . Using a
DNA persistence length (in dilute saline) of 200 bp (67
nm), the value of E is 3.40 X l o 9 dyne ern-'. The shear
modulus G is obtained from the torsional rigidity cwith
the expression G = C / 2 / . Using the value of C = 2.5
X 10-ergcrn. t h e v a l u e o f G i s 1.59X IO9dynecm-*.
The FEA calculations were performed using the commercially available general purpose nonlinear finite element program COSMOS/M.* This code has been widely
used to solve a large number of problems in mechanical
engineering, and the program accuracy has been demonstrated by many comparisons with known solutions for
specific applications. Computations were performed
with a 486/33 personal computer. We found that calculations for particular initial configurations required from
approximately 1 50- 1200 increments, with the initially
circular shape requiring fewer relatively large increments
and the initially straight shape requiring more small increments. Total run times were typically 30 minutes.
The nonlinear kinematic effects associated with large
translations and large rotations were treated using the
updated Lagrangian f o r m ~ l a t i o nThe
. ~ ~incremental
~
solution strategy utilized prescribed displacements in the
context of a force control method with regular Newton-

237

Raphson iteration. An automatic scheme that adjusts

prescribed displacement increments was employed to
avoid excessive incremental rotations. Iterations were
terminated based o n a displacement convergence criterion.

RESULTS
Writhe of Molecules Containing n Equally
Spaced20Bends
Each of the closed configurations was subjected to
rotation of the cut faces (change in ALk) up to the
point at which the writhe begins to change rapidly-that is, the point at which the slope of the
curve measuring Wras a function of change in ALk
becomes almost vertical. We have chosen to display the results as a function of right-handed rotation at the cut or positive changes in ALk. The corresponding negative changes in ALk produce
changes in ATw and Wvof the same magnitude but
of opposite sign. The resulting geometric configurations differ only by a reflection. From the fundamental equation ALk = ATw Wr,6a change in
sign of ALk results in a corresponding change in
sign of both A T w and Wr. Figures 1 and 2 present
plots of W r as a function of A L k for each set of
initial conditions, including the initially straight
case and cases with n = 1.2. 3 . 6 bends in Figure I ;
and cases with n = 9, 10. 1 I , 12, 18 bends and the
initially circular case in Figure 2.
The initially straight configuration can tolerate
a relatively large amount of rotation prior to the
onset of measurable writhe, as we showed prev i o u ~ l y Strikingly,
.~
introduction of even a single
bend causes the immediate onset of Wr upon
change in ALk, again in agreement with our previous results. The greater the number of bends, the
larger Wr per unit change in ALk. However, the
slopes of the Wr vs ALk curves differ dramatically
depending on the total number of bends. The results can be divided into three cases. First, for the
rod containing a single 20 bend, Wr increases
monotonically as A L k is increased. Second, for two
to ten bends, Wr initially increases, then reaches a
local maximum at a change in A L k of just below
1.O. The slope subsequently turns negative, until
Wr reaches a local minimum at a change in ALk of
approximately 1.2. The locations of the local maximum and minimum change little as the number
of bends changes. As ALk increases still further, the
slope becomes positive again, and Wr changes

238

White, Lzind, and B u i i ~ r

0 20

0.4

F,
A

0 05

1--

0.00
0.0

0.3

one bend
l w o bends

eleven bends

twelve bends

2 0.2
4

*A

A/

0.1

-c
-

0.5

1.0

1.5

-3

20

ALk [turns]

0.0
C 0

0.5

1 .o

1.5

ALk [turns]

A eleven bends

twelve bends

00

0 5

10

15

20

ALk [turns]
FIGURE 1 ( a ) A plot of writhe ( W r ) vs change in
linking number ( A L L ) .or number of turns of one of the
faces of the cut relative to the other, for molecules containing a few intnnsic bends Represented are initially
straight D N A and D N A containing one, two, three, and
six bends All bends are initially coplanar, 20" in angle,
open toward the center, and are equally spaced The ALk
range varies for each model, terminating at a change in
ALk at which the slope of the Wr vs ALk plot is nearly
vertical, indicating the imminent onset of a snap through
phenomenon in the actual structure ( b ) An expanded
plot of ( a ) in order to show more clearly the behavior of
Wr at low ALL. The points represent calculated values,
and the connecting lines are smooth curves generated by
the Stineman method.

quite rapidly as ALk is further increased. The curve

finally becomes nearly vertical, indicating that selfcontact is imminent. Third, for eleven to eighteen
bends, the slopes of the Wr vs ALk curves are al-

ALk [turns]
FIGURE 2 ( a ) A plot of Wrvs ALk for molecules containing a larger number of bends. Represented are initially straight D N A and D N A containing nine. ten,
eleven, twelve, and eighteen bends, all of 20", as well as
D N A that is initially circular. ( b ) A magnified scale of
( a ) . The points and lines were obtained as described in
the caption to Figure 1.

ways positive, and local maxima and minima are

absent. The vertical slope on the Wr curve is
reached at a much lower magnitude of A L k .

Changes in Geometry Associated with

Changes in Writhe
We next describe the differences in geometry between the latter two classes of molecules discussed
in the preceding section: those containing two to

FIGURE 3 A series of representations ofthe nine-bend D N A molecule subject to increasing

changes in ALk. The spatial molecule is shown in darkened outline. Each three-dimensional
plot also includes three projections of this molecule onto three mutually perpendicular planes.
The solid arrow indicates the cut location. ( a ) The initial strain-free molecule containing nine
bends is shown after closure, with Wr = 0. The changes in ALk and resulting writhing numbers
for the rest of the figures are as follows: ( b ) ALk = 0.650, Wr = 0.030; ( c ) ALk = 1.075, Wr
= 0.034; ( d ) ALk = 1.200, Wr = 0.010; ( e ) ALk = 1.265, Wr = 0.034: ( f ) ALk = 1.365, Wr
= 0.195. Each point shown lies at the intersection of two adjoining elements.

Twr.ct, Urithe. and Goometry ofa DNA Loop

-\

,,-

(a) closed, ALk=O, Wr=O

(b) ALk=0.650,

W~0.030

-.,,-.

fc) ALk=l.O75,

Wr=0.034

(d) ALk=l . Z O O , Wr=O.O 10

(e) ALk=l.265,

Wr=0.034

(f) ALk=l.365,

Wr=O.l95

239

240

White, Lund. and Bailer

ten bends, and those containing eleven to eighteen

bends. We choose a model containing nine bends
as representative of the first class, and one containing eleven bends as representative of the second
class. Figure 3 presents various representations of
the nine bend molecule subject to increasing ALk.
These illustrate changes in geometry that occur
upon progression through the local maximum and
minimum depicted in Figure 2. In addition to the
three-dimensional representation, shown in darkened outline, each figure contains three views
showing projections onto three mutually perpendicular planes. The projected writhe may be
readily deduced from the figure in each left panel,
using the observation that the more nearly horizontal segment is over, and the other is under. The
cut location, at which the change in ALk is introduced, is shown by the solid arrow.
The initial configuration. shown in Figure 3a, is
the nine-bend molecule that has been closed up
into a polygon with curved edges and has Wr = 0.
All bends lie in the plane of the closed molecule
and open toward the interior. The remaining views
in this figure depict the changing geometry of the
DNA rod that accompany the ALk progression of
Figure 2. Three parts of the figure (b, c, and e ) compare the geometry at three different values of ALk,
chosen so as to have approximately the same Wr
(0.03) but located at quite different parts of the
curve in Figure 2. Figure 3d displays the geometry
at the minimum in Figure 2, and Figure 3f shows
the geometry at a relatively large value of A L k , well
beyond the minimum.
The result o f a change in ALkof0.65, just before
the local maximum in Figure 2, is shown in Figure
3b. The bottom projection shows that only a slight
change in overall geometry occurs, the molecule
becoming somewhat more oblong, but the projections of the bends remain oriented toward the interior. The side projection on the right is oblong,
showing some initial out-of-plane motion. The
projection on the left also shows that the molecule
has moved out of plane, with a projected writhe of
1. The ends of the projection (ears) are rotated
in opposite directions, the left one up and the right
one down.

Figure 3c shows the molecule after a change in

ALk of 1.075, between the local maximum and local minimum in Figure 2. The two side plane projections are much the same as in Figure 3b, except
for some flattening near the cut location in the right
plane. An important change occurs in the bottom
projection and can also be seen in the three-dimensional representation. The bends near the cut location have reversed orientation and turned so that
some bends are now oriented toward the exterior
The left projection still has a projected writhe of
1, and the orientations of the ears are unchanged.
Figure 3d shows the molecule near the local
minimum in Figure 2; this occurs at ALk of 1.2,
and Wr has decreased to 0.0 1. The bends near the
cut have fully turned to the exterior, i.e., rotated
approximately 180with respect to the plane of the
initial molecule. This rotation has had the effect of
flattening the molecule again, as can clearly be seen
in the left side projection. In the left projection, the
ears have considerably flattened.
As ALk is further increased, the molecule begins
to writhe again. Figure 3e shows the molecule after
a change in ALk of I .265. This point was chosen so
that Wr is the same as that in Figures 3b and c. Here
the bottom projection has elongated in a direction
perpendicular to that in Figure 3c. More significant
yet, the projection in the left-side panel has reversed the orientation of the ears of the figure eight
and now has a projected writhe of -1. Figure 3f
shows a larger Wr of 0.0195, at a larger ALk of
1.365, and indicates how this results in exaggerations of the phenomena in the three projections
seen in Figure 3e.
All the molecules with two to ten bends have the
same characteristics indicated in the preceding discussion of the nine-bend case. First, the Wr vs ALk
curves all contain both a local maximum and a local minimum. Second, in the region of ALk before
the local maximum. the pro-jection onto the left
plane has a projected Wr of 1 and in the ALk region after the local minimum it has a projected Wr
of - 1 . Third, before the local minimum, the ears of
the figure eight in this projection satisfy the property that the left ear points up and the right ear
points down; after the local minimum, the reverse

FIGURE 4 A series of representationsof the eleven-bend DNA molecule subject to changes

in A L k . The data are plotted as described in the caption of Figure 3. The changes in ALk and
accompanying changes in Wr are as follows: ( a ) ALk = 0, Wr = 0: ( b ) ALk = 0.500, Wr
= 0.026; ( c ) ALk = 0.750, Wr = 0.074; ( d ) ALk = 0.850, Wr = 0.108; ( e ) ALk = 0.970, Wr
= 0.207; ( f ) ALk = 0.972, Wr = 0.226.

TttYst, W'rithe. and Geometry o f a DNA Loop

_,'

r;
x
'

(a) closed, ALk=O,

(c) ALk=0.750,

Wr=O

W~0.074

(b) ALk=0.500,

Wr= 0.026

Wr=O. 108

(d) ALk=0.850,

\'

(e) ALk=0.970,

Wr=0.207

//

( f ) ALk=0.972,

Wr=0.226

241

242

White. Liind. and Bailer

is true. Fourth, in the bottom projection, the molecule is oblong in one direction before the local minimum and oblong in the perpendicular direction
after the local minimum.
The molecules containing more than ten bends
have quite different characteristics. This is illustrated in detail for the case of eleven intrinsic bends
in Figure 4. First, Wr passes through neither a local
maximum nor a local minimum as ALk increases.
Instead, the Wr vs ALk curve has only a positive
slope up to the point of snap-through. Second, the
projection onto the left plane always has a projected Wu of $1. Third, the left ear of the figure
eight in that projection is always up, and the right
ear is always down. Finally, the bottom projection
is always oblong in the same direction. In summary, the molecules with more than ten bends exhibit the same characteristics over the entire range
of ALk that the molecules with fewer than 1 1 bends
have in the region before the local minimum.

Conformation of Molecules initially Curved

into a Smooth Semicircle as ALk Changes
We next show that changes in the conformation of
the DNA with changes in ALk depend only upon
the total curvature and are independent of the
number of discrete bends. This can be seen by an
examination of a molecule in which the 180" initial
curvature is introduced in a smooth bend of 180"
( a semicircle) instead of in nine discrete 20" bends.
Figure 5 presents a series of views of such molecules, with changes in ALk comparable to those in
Figures 3 and 4. The semicircle is first closed into a
circular molecule, as shown in Figure 5a, followed
by the appropriate changes in ALk in Figure 5b-f.
Comparison between the semicircular model and
the nine-bend model, Figure 3 , reveals that the
shapes are very similar in geometry, and that they
have similar projections onto the three planes. The
discrete nine-bend model is especially important in
that it gives a clearer explanation of the flattening
out of the molecule at a change in ALk of approximately 1.2. In this case the rotation from interior

to exterior of the bends near the cut location is especially apparent.

Changes in the Twist increment

Distribution for Bent DNA as ALk Changes
One of the main advantages of FEA is its capability
to calculate the local twist increment d i r e ~ t l yTo
.~
explain this briefly: the unit twist of a circular rod is
given by the ratio M / 2 G I , where M is the torsional
moment applied at opposite ends of the rod. For a
straight rod, this corresponds to the change in angle
between radii, separated by unit length along the
axis, which originally lie in a plane. This change in
angle is caused by twisting behavior of the rod in
response to the applied torsional moment. The torsional moment per element is calculated at specific
solution steps. The moment is divided by 2GI to
obtain the unit twist ( r a d / n m ) ; the latter is then
normalized by 27r to obtain the rate of change of
twist per unit length, 6ATwl6s (with length along
the DNA contour s measured in nm). The total
change in twist may be calculated as a function of
ALk by integrating the rate of change of twist per
unit length over the axis length L and ATw =
( ~ A T M ' / ~ds.
s ) Numerical checks (not shown)
confirm that ATw calculated in this fashion, when
combined with calculated values of W r , satisfies
the global relationship ALk = A T w
W r . FEA
thus allows ATw to be computed independent of
the conservation condition, in contrast to stochastic methods, in which ATw is not usually obtained
independent of ALk and Wr.".'*In addition, FEA
permits direct calculation of the manner in which
ATn, is distributed.
Figure 6 presents a plot of 6ATw/6s as a function of s for n = 3. 9, and I 1. and for the initially
circular molecule. The cut faces are at s = 0 and s
= 628 nm. The 6ATu./6s distribution is uniform
between bends and shows discrete jumps at the
bends. Since all models are closed and the bends
are uniformly spaced, it is not surprising that the
plots are all symmetric about the center of the
DNA ( s = 314 nm). These particular boundary

sL

FIGURE 5 A series of three-dimensional plots of the semicircular DNA molecule at various

values of ALk. The values of ALk at which the shapes were sampled were chosen to be consistent with those in Figure 3, in order to facilitate direct comparison of the geometric changes
with the nine-bend model. The changes in ALk and accompanying changes in Wr are as follows: ( a ) ALk = 0, Wr = 0; (b) ALk = 0.650, Wr = 0.026; ( c ) ALk = 1.080. Wr = 0.029; ( d )
ALk = 1.200, Wr = 0.005; (e) ALk = 1.265, Wr = 0.030; ( f ) ALk = 1.377. Wr = 0.243.

Tttist, Writhe, und Gmmetr.v ofa DNA Loop

>

Wr=O

(b) ALk=0.650,

Wr=0.026

Wr=0.029

(d) ALk=l.200,

Wr=0.005

(a) closed, ALk=O,

(c) ALk=l.O80,

(e) ALk=l.265,

Wr=0.030

(f) ALk=l.377,

Wr=0.243

243

244

Whit(., Lirnd, und Ruiier

0.0020

E
<

0.0015

0.0010

s-.0.0005

3
+

G
4
o

0.0000

-0.0005

157

314

47 1

628

[nml

FIGURE 6 A plot of twist increment, 6ATitl/6s, vs

distance along the contour from the cut location for molecules containing 3, 9, and l l uniformly spaced bends,
and for the initially circular molecule, for a change in
ALk of approximately 0.5. The values of ALk, Wr, and
ATw for each of the curves are as follows: 3 bends, ALk
= 0.480. Ur = 0.004. ATiz, = 0.476; 9 bends, ALk
= 0.500. Wr = 0.016. ATM, = 0.484; 1 1 bends, ALk
= 0.500, Wr = 0.026. AT%,= 0.474; and initially circular.
3I.k = 0.493, Ur = 0.087. A n t = 0.405. For the three
configurations containing discrete bends, the symbols indicate the elements in either side of each bend. The horizontal lines represent the calculated results. The nearly
vertical lines were added to make clear the identification
of each curve. The plot for the initially circular model
shows the smoothly connected calculated results for all
elements.

conditions were chosen to have the same change in

ALk of approximately 0.5. In all cases the ~ A T H , /
6s distribution either varies sinusoidally with location, as in the case ofthe initially circular molecule,
or has the shape of a polygonal approximation to a
sinusoidal curve in the discrete cases. The flattest
curve is that of the three-bend molecule. The
difference between the maximum and minimum
values of 6ATi4>/6.s increases as the number of
bends increases.

Changes in the Twist Increment

Distribution for the Semicircular Model
as ALk Changes
Since the discrete bend cases are essentially polygonal approximations of the continuous models,
and since many of the salient points are easier to
recognize in the continuous case, we first deal with
the semicircular model. Figure 7a presents a plot of
6ATw/6.~vs distance from the cut location, for the

same series of ALk values that were used in Figure

5 . The sinusoidal shape of the plot is quite evident
in all cases. What is especially significant is how the
minimum and maximum 6ATw/6.s locations
change with changing A L k .
As ALk is initially increased to 0.65,6ATw/6s is
largest near the cut location and smallest at the
point diametrically opposite. This curve contains
exactly one local maximum and one local minimum. At the point at which Wr is nearly at its local
maximum (see Figure 9 ) , however, this property
of the twist increment distribution changes. As
ALk increases further to 1.08, the local minimum
(which is also the global minimum) increases in
magnitude, but its location along the D N A remains unaltered. In addition, a second local minimum appears at the cut location. The increase in
ALk has caused the local maximum to shift inward
from the cut location so that, by symmetry, there
are now two local maxima near the cut location. As
ALk is further increased to 1.2, at which point the
molecule is clearly flattened out (Figure 5 d ) , the
global minimum increases yet more and the minimum at the cut location is reduced. The two local
maxima move further away from the cut location.
An additional shift in the two local maxima can be
seen at ALk = 1.27. where the two local minima
now come closer together. Finally, at ALk = I .38,
at which point the Wr curve is almost vertical, the
global minimum occurs at the cut location. The
other minimum is still at the same point but larger.
and the two maxima have become even smaller.

Changes in Twist Increment Distribution

with ALk for the Nine-Bend Initial
Conformation
Figure 7b shows the twist increment distributions
6 A T ~ , / 6 for
s changes in ALk for the nine-bend
model that correspond to those for the continuous
model in Figure 7a. As expected, the shapes here
are polygonal approximations to the behavior
shown by the continuous model. We emphasize
again that for all models containing discrete bends
the twist increment is uniform between the bends
and has discontinuities at the bend locations.
We next extend these results to show how the
twist increment distribution changes within each
interbend segment as ALk is changed. The results
are presented in Figure 8. Here the initial configuration contains nine uniformly spaced bends, and
the cut location is positioned exactly halfway between bend 9 and bend 1. The segment joining

TMw, U'rithe, und Geomrtrj- of u DNA Loop

245

AL k
0 65

- 108

1.20

127

I30

157

314

47 I

0.000

628

s [nml
0.003

-g

0.002

\
7

Alk
065

*-

108

--cr

120

3
'

127

I2

+
3

0.001

137

L a

0
0

0.000

L
0

I57

314

47 1

628

FIGURE 7 A plot oftwist increment, 6ATbi1/6s, vs distance along the contour from the cut
location for the semicircular ( a ) and nine-bend ( b ) models. The twist increment is plotted for
various values of ALk as indicated. The twist increment at the center location in ( a ) increases
as ALk increases. The nine-bend model is a polygonal approximation of the semicircular
model. The curves are otherwise drawn as described in the caption of Figure 6.

bends 9 and 1 is denoted the cut segment. Because

of the symmetry of the initial configuration, the
twist increment distribution should also behave
symmetrically with respect to the cut location.
Thus, the uniform 6ATw-/6sbetween bends 1 and
2 should be the same as that between bends 8 and
9, etc. This symmetry is indeed borne out by the
calculations. as demonstrated in Figure 7b. For
this reason, in Figure 8 we denote the segment
between bends 1 and 2 ( 8 and 9 ) the second segment; that between 2 and 3 ( 7 and 8 ) the third
segment: that between 3 and 4 ( 6 and 7 ) the
fourth segment; and that between 4 and 5 ( 5 and
6 ) the fifth segment.
U p to a change in ALk of approximately 1 .O,
a point just beyond the local maximum in Wr,
6ATw~/6.s is greatest in the cut segment and

monotonically decreases in neighboring segments, reaching a minimum in segment five. As

the Wr vs ALk curve descends from its local maximum to the local minimum, there is a change in
this hierarchy. The second segment now has the
largest twist increment, the cut segment being
second in magnitude. As ALk is changed further,
and Wr passes through its local minimum and
begins to increase rapidly, 6ATw/6s in the third
segment now becomes largest, followed by that
in the fourth segment, etc. Numerical problems
have made it difficult to investigate the further
development of this pattern. However, we speculate from the accessible data that this change in hierarchy continues, until one after another a later
segment has the larger value of twist increment until the fifth segment is largest. It will be particularly

246

White, Lzind. and Bazier

0.003

- 0.20

- 0.15

- 0.10

- 0.05

00

02

04

06

ALk

08

12

10

1 4

- 0.00

[turns]

FIGURE 8 A plot o f t h e twist increment ofeach segment between successive bends. 6ATw/
fis, as measured from the cut location, as a function of A L k , for the nine-bend model. Because
of the symmetry of the molecule. 6ATw/6.s is divided symmetrically along the molecule as
measured in either direction from the cut location. Thus, if the bends are counted from 1 to 9,
the cut segment is located between 9 and 1, the second segment is between 1 and 2 ( 8 and 9 ) .
the third segment is located between 2 and 3 ( 7 and 8 ) , the fourth segment is located between
3 and 4 ( 6 and 7). and the fifth segment is located between 4 and 5 ( 5 and 6 ) . The solid lines
are smooth curves generated by the Stineman method. The I@" plot taken from Figure 2 is
shown for reference.

interesting to discover if this phenomena is periodic as ALk increases still further and supercoils
are formed, or if this is only a characteristic of molecules with small changes in A L k .

Changes in Writhe with ALk Are

Independent of DNA Length
In all preceding calculations, the DNA length was
taken to be 628 nm ( 1870 bp 1. In order to ascertain

Comparison of Writhe for DNA with

Smooth Curvature and Discrete Bends

0.25

As we showed above, the conformation as a func-

0.20 -

tion of changes in ALk depends only upon the

amount of initial uniform curvature rather than
upon the presence of discrete uniformly spaced
bends. The next two figures establish that Wr similarly depends only upon the total curvature. In
Figure 9 two discrete and continuous models are
compared. The W r of the nine-bend and semicircular 180" continuous models, and of the elevenbend and 220" continuous models, are plotted as a
function of changes in A L k . The two plots show
remarkable agreement. Equally in agreement are
the two plots in Figure 10 of Wr vs ALk for molecules with 240" of intrinsic curvature introduced in
two different ways: by twelve 20" bends or by six
40" bends. This again demonstrates that the total
curvature, rather than the size of the individual
bends, determines the response to changes in A L k .

rn
9 Bends

A
0.15

I I Bends

r'

Semicircular

220 Degrees

0.10 #
0.05

V.""

0.0

0.2

0.4

0.6

ALk

0.8

1.0

1.2

1.4

[turns]

FIGURE 9 A plot of Wr vs ALk for two discrete and

continuous models with the same intrinsic curvature: the
nine-bend and 180"semicircular model, and the elevenbend and 220" smooth model. The smooth curves connect data points for the nine-bend and eleven-bend
models.

247

T ~ a \ t ,U'rrtkc, and Gcometrji of a DNA Loop

0.20

0.30

INITIALLY STRAIGHT
0.15
6 40 Degree Bends

0.20

$ 0.10

$
0.10

0.05

- /
n nn
".""

0.0

0.2

0.4

0.6

0.8

1 .o

0.00
0.0

0.5

1 .o

1.5

2.0

0.6

ALk [turns]

ALk [turns]
FIGURE 10 A plot of U'r vs ALk for two 240" discrete
bend models, one with 12 X 20" bends. the other with 6"
X 40" bends. The plots show that. if the total curvature
due to the equally spaced bends is the same, Wv is the
same for a given A L k . The smooth curves connect the
data points for the 12" X 20" bends case.

--

0.4

Initially Circular

Langth [ml

157
628

the effects of the absolute DNA axial length, we determined the variation of Wr with ALk for DNAs
of three different lengths, using three models: initially straight, initially circular, and initially semicircular. These effectively bracket the range of behavior that we have observed in the calculations.
We calculated the Wr plots for DNA one-fourth
the length of the standard, or 157 nm (467 bp),
and four times the length of the standard, or 25 12
nm (7476 bp). The results are shown for these
three initial configurations in Figure 1 1. It is clear
that, in all cases, the results show no significant dependence upon the absolute DNA length. We also
determined that the same result is obtained if discrete bends are used instead of the continuous
curves (calculations not shown).

0.0
0.0

2513

0.2

0.4

ALk [turns]

0.3 I

Initially Semicircular
?

DISCUSSION
The present study is part of a continuing investigation of the influence of DNA initial configuration
upon the geometric response of a closed DNA loop
to changes in ALk. In particular, we have examined the consequences of the presence of varying
numbers of fixed angle, equally spaced, in-plane
bends upon both the writhe and the twist increment with increasing ALk. More generally, we
show that discrete bends of this type can be replaced by smooth curvature, provided that the total
extent of bending remains unchanged. We deal

ALk [turns]
FIGURE 1 I A plot of Wr vs ALk for ( a ) initially
straight, ( b ) initially circular, and ( c ) initially semicircular DNA of varying lengths. All three plots show
the results ofcalculations for molecules oflengths 157
n m (467 b p ) , 628 n m ( 1870 b p ) , a n d 2513 n m (7480
b p ) . In all cases, the smooth curves connect the data
points for the 628 n m case.

248

R'hite, Ltind. and Bauer

Table I

Summary of FEA Writhe Results for Closed DNA Loops

Number of
20" Bends

Range of ALk
(Figures 1 and 2 )

0 to zz 1.8
> 1.8

All
Low
Intermediate

2-10

110

High
All

only with ALk changes below the threshold that

produces intra-axial ( self-)contact.
The FEA calculations establish that closed DNA
loops undergo highly significant changes in writhe,
geometry, and twist increment as A L k is increased.
The nature of these changes depends upon the
number of bends or, more generally, upon the total
bend angle. The principal results with regard to Wr
are summarized in Table I. For more than a single
20" bend, two distinct behaviors are evident. For
two to ten 20" bends, Wrpasses first through a local
maximum, then a local minimum, and finally enters a region of monotone increase. If the number
of 20" bends exceeds ten, Wr always increases
monotonically with A L k . Perhaps more significant
from a functional point of view, these two regions
exhibit quite different changes in geometry as a
function of A L k . As A L k increases, the bends, originally opening toward the interior of the loop, rotate away from the loop center. This occurs first for
bends nearest the cut location. For ten or fewer 20"
bends, those bends located nearest the cut rotate
approximately 180" as Wr approaches the local
minimum (the overall shape is relatively planar).
This behavior occurs at physiologically relevant
values of both the number of bends and of A L k
and might well affect the binding of proteins that
recognize bent DNA. In contrast, for more than ten
20" bends the corresponding rotation of bends always occurs with substantial Wr. These results
raise the possibility that intrinsic bends of varying
total angle, in combination with changes in ALk,
may provide a metabolic control function.
The FEA calculations also demonstrate that the
distribution of twist increment, as ALk is changed,
is strongly influenced by the presence of intrinsic
bends in a closed DNA loop. The local twist increment changes as one proceeds in either direction

Change in Wr with ALk

~ 7 r -0
Rapid increase with ALk
Monotonic increase with ALk
Increases to local maximum
Decreases from local maximum
to local minimum
Rapid increase with ALk
Monotonic increase with ALk

from the original cut, the location of which is preserved in protein mediated loop formation. The
twist increment distributions depend upon ALk .
For example, if the total bend angle is 180", the
twist increment is largest in the vicinity of the original cut and diminishes progressively (and
symmetrically) as one progresses away from the
original cut, for ALk up to approximately 1 .O. The
dependence of the twist increment upon location
might affect the binding of reagents whose association constants are sensitive to the local twist, such
as intercalating drugs and certain DNA binding
proteins.13
We emphasize that the behavior of DNA predicted by FEA is restricted to that associated with
its character as an elastic rod. In particular, neither
the theory nor the calculations deal with the entirely separate question of the conformational distribution due to Brownian motion. Our results
therefore refer to the average (equilibrium) state of
the DNA, apart from thermal fluctuations.
The restriction to a static DNA model has the
great advantage that we are able to address directly
the behavior of a DNA formed under conditions
such that the cut location is effectively preserved by
combined protein-protein and protein-DNA interactions. When a loop is formed connecting axially distant sites of a DNA, the closure is mediated
by the binding of proteins attached at each site. The
protein-protein junction marks the closure of the
topological domain and is the equivalent of the cut
location that is characteristic of a naked DNA loop.
The creation of a topological domain by this mechanism means ( 1 )that the history ofthe cut position
is not lost due to thermal motion and ( 2 ) that fractional values of ALk are meaningful and allowed.
Since our calculations take no explicit account of
the presence of protein, the quantitative predic-

Twist. W'rithe, und Geometry qfu DNA Loop

tions are probably directly applicable only to real

DNA loops larger than about 250 b ~ . ~
The situation of a free closed circular DNA in
solution is clearly more complicated, and our results provide only partial information in this case.
In particular, thermal agitation of DNA free in solution is expected to abolish the memory of the location of the cut. Likewise, the twist increments,
6ATw/6s, will be observable only as a population
average, from which most of the detailed information concerning the changes in 6ATw/6s with location and extent of bending cannot be abstracted.
The predictions of the effects of bends on the Wr
will, however, still lead to changes in the expected
population average ( W r ) as a function of ( A L k ) .
This subject will be treated more fully elsewhere.
Protein-stabilized closed DNA loops are of common occurrence in at least three vital biological
processes: regulation of transcription, recombination, and DNA replication. Examples of all three
are well documented in procaryotic systems and,
to a more restricted extent, in eucaryotic systems as
well.4 Loop sizes vary considerably, from only a
few base pairs to as long as 4 kilobase pairs (kbp).
A partial list of transcription regulatory elements/
loop sizes includes X repressor, 50-850 bp 14; AraC,
280 bp15; lac repressor, 65-535 bp"; dcoR repressor, 600 bpI7 and 1-5 kbp'*; and N R I , normally
100- 130 bp, but as long as 700-950 bp." Examples
of recombination-associated loops include bacteriophage X integration, 50- 150 bp2' and site-specific inversion in Salmonella, 100-4000 bp.2' Finally, replication specific loops of 1200-2000 bp
occur in the replication of certain plasmids.22
Bends that are located within the appropriate
DNA segment also appear to play a regulatory role
in the biological function of looped DNA. These
bends are usually induced by protein binding. For
example, CAP bends the DNA loop associated
with AraC repression, thereby modulating the repression.2' Likewise, internal bending by IHF at a
location between the N R , binding sites and the
promoter is required for initiation of glnHp2 tran~cription.'~
As a further example, the DNA loop
stabilized by the Int protein of X is facilitated by the
DNA bending induced by IHF.25
We have not yet attempted to extend FEA to
deal directly with the elastic effects of a DNA loop
with an internally attached protein. However, at
least in some cases, the function of the protein appears to be limited to producing the bend, and the
protein itself is dispensable. For example, in the
N R I system, IHF is no longer required if the loop

249

is formed from a segment of a DNA that was supercoiled, thereby already in a bent configuration.26In X recombination, the CAP protein from
Eschcrichiu coli can substitute for IHF when the
CAP site is substituted for the IHF site,27suggesting that it is the bend itself, not some other aspect
of IHF binding, that is important. Even more significant, an appropriate intrinsic bend, formed by
six repeats of a tract of A6 so phased as to produce a
120" bend, can also substitute for the IHF induced
bend.27.2*These observations argue strongly that
our FEA predictions, based on the effects of protein-free intrinsic bends, have direct relevance to
understanding the in vivo situation.
The magnitudes of the bend angles involved in
regulation of naturally occurring DNA loops often
lie within the range ofcomplex Wr response to ALk
that are revealed by FEA analysis. That is, many
are near to but below 10 X 20" bends or 200". Thus,
the CAP-induced bend angle has been estimated to
lie in the range 100"- I 30.29.30
The Fis bending angle is 90",29 the combined Fos and Jin bending angle can range up to 1 1 I ', 3 1 and the IHF bending
angle is in excess of 140.29
" No cases of proteinmediated DNA loops have yet been reported in
which examples of bends greater than 220" occur.
Our results suggest that the role of bend formation
might be more extensive than merely to facilitate
loop formation, and it is possible that the complex
behavior of Wr and 6ATw/6s play a natural role in
the modulation of the biological response to loop
formation. In particular, both the local torsional
distortion and the three-dimensional shape of the
loop are determined, in part, by the total extent of
bending.
This research was supported in part by grant GM37525
from the National Institutes of Health.

REFERENCES
I . Wang, J . C. & Giaever. G. N. ( 1988) Science 240,
300-304.
2. Schleif, R. ( 1988) Scicnccp 24, 127- 128.
3. Lobell, R. B. & Schleif, R. F. ( 1990) Science 250,
528-532.
4. Hochschild, A. ( 1990) in DNA T o p o l o g ~und
~ Its Biological Eflcts, Cozzarelli, N . R. & Wang, J. C.,
Eds., Cold Spring Harbor Laboratory Press, Cold
Spring Harbor. NY, pp. 107-138.
5. Bauer, W. R., Lund, R. A. & White, J. H. (1993)
Proc. Nutl. Acud. Sci. USA 90, 833-837.
6. White, J. H. ( 1969) A m . J . Muth. 91,693-728.

250

U'hite. Liind, and Buirer

7. Hagerman, P. J. ( 1988)Ann. Rev. Biophys. Biophys.

Chem. 17, 265-286.
8. COSMOS/M Finite Element Sjstem ( 1992) Structural Research and Analysis Corp., Santa Monica,
CA.
9. Bathe, K.-J. (1982) Finite Element Procedures in
Enginwring Anulj~si.s, Prentice Hall, Englewood
Cliffs, NJ.
10. White, J. H. ( 1992) Proc. Symp. Appl. Math. 45, 1737.
1 1 . Schlick, T. &Olson, W. K. ( 1992) J . Mol. Bid. 223,
1089-1 119.
12. Vologodskii, A. V., Levene, S. D., Klenin, K. V.,
Frank-Kamenetskii, M. & Cozzarelli, N. R. (1992)
J . Mol. Biol. 227, 1224- 1243.
13. Bauer. W. R. & Gallo, R. ( 1989) in Chromosomes:
Eiikuryotic, Prokuryotic. and Virul, Adolph, K. W.,
Ed., CRC Press, Boca Raton. FL. pp. 87- 126.
14. Hochschild, A. & Ptashne, M. ( 1986) Ce1144,68 I 687.
15. Dunn, T. M.. Hahn, S., Ogden, S. & Schleif, R. F.
( 1984) Proc. Nutl. Acad. Sci. US.4 81, 5017-5020.
16. Kramer, H., Amouyal, M., Nordheim, A. & MullerHill, B. ( 1988) EMBO J . 7,547-556.
17. Dandanell, G . ( 1992) Niicleic Acid.7 Res. 20, 54075412.
18. Dandanell, G., Valentin-Hansen, P., Larsen, J. E. L.
& Hammer, K. ( 1987) Nuriire325, 823-826.

19. Ninfa, A. J., Reitzer, L. J. & Magasanik, B. (1987)

C~jll50, 1039- 1046.
20. Weisberg, R. A. & Landy, A. ( 1983) in Lamhdu I I ,
Hendrix, R. W.. Ed., Cold Spring Harbor Labordtory, Cold Spring Harbor, NY, pp. 2 1 1.
21. Johnson, R. C. &Simon, M. I. (1985) Cell41,781791.
22. Mukherjee, S., Erickson, H. & Bastia, D. (1988)
Proc. Nutl. Acud. Sci. USA 85,6287-629 I .
23. Lobell, R. B. & Schleif, R. F. ( 1991) J. Mol. Biol.
218,45-54.
24. Claverie-Martin, F. & Magasanik, B. ( 199 1 ) Proc.
Nutl. Acud. SLY. lJSA88, 1631-1635.
25. Moitoso de Vargas, L., Kim. S. & Landy, A. ( 1989)
Science 244, 1457- 146 1 .
26. Claverie-Martin, F. & Magasanik, B. ( 1992) J . Mol.
Biol. 227, 996- 1008.
27. Goodman. S. D. & Nash, H. A. ( 1989) Nutzire341,
25 1-254.
28. Goodman, S. D., Nicholson, S. C. & Nash, H. A.
( 1993) Prctc. Null. Acad. Sci. USA 89, 1 19 10- I I9 14.
29. Thompson, J. F. & Landy. A. ( 1988) Niicleic A d s
RCS.16, 9687-9708.
30. Zinkel, S. S. & Crothers, D. M. ( 1990) Biopolymrrs
29,29-38.
3 1. Kerppola, T. K. & Curran, T. ( I99 I ) Science 254,
12 10- 12 14.
32. Robertson, C. A. & Nash, H. A. (1988) J. Biol.
C'hen?. 263,3554-3557.