1962 Electrophoresis 1999, 20, 1962±1977

Christoph Heller Separation of double-stranded and single-stranded

Abteilung Lehrach,
Max-Planck-Institut für
molekulare Genetik,
Berlin-Dahlem, Germany
DNA in polymer solutions:
I. Mobility and separation mechanism
We have studied the separation of single-stranded and double-stranded DNA in a
matrix of entangled, linear poly-N,N-dimethylacrylamide. Our results give better insight
into the mechanisms involved during separations in polymer solutions. The depend-
ence of different parameters on DNA size, electric field, pore size and the polymer
chain length are evaluated and compared to theoretical predictions. Striking differ-
ences between experimental data and predicted scaling laws are found. Our data
should help to optimize DNA separation in capillary electrophoresis and to improve
existing models for DNA separation in porous matrices.

Keywords: Separation matrix / Polymer solution / DNA separation / Electrophoresis theory /

Capillary electrophoresis EL 3522

1 Introduction matrix, fundamental studies on DNA separation in this

For DNA separation and other applications, capillary elec-
trophoresis (CE) has many advantages over slab gel
electrophoresis, in particular high speed, high sensitivity
and the possibility for automation. It also offers full com-
patibility to existing biochemistry, which is not always the
case in alternative techniques, such as MALDI-MS.
Instead of cross-linked gels, so-called ªentangledº (ªsemi-
diluteº) polymer solutions are now widely in use as sepa-
ration matrix (see [1] for review). Separation in unentan-
gled (ªdiluteº) polymer solutions has also been described
[2], but is probably irrelevant for most molecular biology
applications, as the resolution is rather low. Therefore, in
our study, we concentrate on entangled polymer solu-
tions.
1.1 Polymers used in CE
Most of the polymers used in CE are modified polysac-
charides or synthetic polymers such as derivatives of
acrylamide. One of these is poly-N,N-dimethylacrylamide
(pDMA), which has ideal properties as a separation
matrix, due to its low to medium viscosity and its self-coat-
ing ability. Solutions of pDMA are successfully used for
DNA separation, both in capillary electrophoresis [3±6] as
well as in microchips [7]. To better characterize this
Correspondence: Dr. Christoph Heller, Max-Planck-Institut für
molekulare Genetik, Ihnestraûe 73, D-14195 Berlin, Germany
E-mail: heller@mpimg-berlin-dahlem.mpg.de
Fax: +49-30-8413-1380
Abbreviations: BRF, biased reptation with fluctuation; BRM,
biased reptation model; pDMA, poly-N,N-dimethylacrylamide;
TBE, Tris-boric acid-EDTA buffer
polymer have been undertaken recently [8]. There, we
found hints that the separation of ssDNA gives superior
resolution compared to the same DNA fragments in the
double-stranded form (in a matrix of the same pore size).
This is in agreement with an earlier study on linear poly-
acrylamide [9]. The present study is a continuation of our
earlier work and is aimed at describing this and other
effects in more detail and getting a better insight into the
underlying mechanisms.
The properties of entangled polymer solutions have been
reviewed in several articles (see, e.g. [8, 10]). One impor-
tant parameter is the ªpore sizeº or ªmesh sizeº of such a
network. To estimate this value, we can regard a polymer
chain segment between two entanglement points as an
independent subunit, which can undergo a so-called ªran-
dom walkº. The volume, which this chain segment can
take, is called a ªblobº and is described by the ªblob sizeº,
xb. Viovy and Duke [11] argued that this ªblob sizeº could
be used as a ªpore sizeº of the network. It can be calcu-
lated as follows [12]:
xb = 1.43 Rp (c/c*)±(1+a)/3a (c > c*) (1)
with c* being the overlap threshold, i.e., the concentration
c above which the polymers become entangled, and Rp
being the radius of gyration of the polymer. The parame-
ter a is the exponent of the empirical Mark-Houwink equa-
tion, which describes the relationship between the intrin-
sic viscosity [h] and the molecular mass of the
polymer:[h] % K Mva. Here, Mv is the ªviscosity average
molecular massº, which is different, but often close to the
weight average molecular mass, Mw [13]. The Mark-Hou-
wink coefficients K and a have been determined for many
polymer-solvent systems (see, e.g. [14]). For flexible mol-

 WILEY-VCH Verlag GmbH, 69451 Weinheim, 1999 0173-0835/99/1010-1962 $17.50+.50/0

Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA
ecules, the value of a varies between 0.5 (polymer in an
ideal solvent) and 0.8 for so-called ªswollenº chains.
The so-called Flory-Fox theory relates the radius of gyra-
tion to the intrinsic viscosity: [h] = F Rp3/M, with F being
the Flory constant. Following Thielking and Kulicke [15],
this theory, which has been developed for the unper-
turbed state, can also be applied to dissolved polymers in
good solvents by introducing expansion coefficients. This
modified Flory viscosity constant takes into account
excluded volume effects by: F = F0 (1±2.63 b + 2.86 b2)
with b = (2a±1)/3 and F0 being 2.5 ´ 1023 mol±1 for
uncharged polymers [13, 16]. For ideal chains (a = 0.5; no
excluded volume), b becomes zero and we regain F =
F0. Following Cottet et al. [16], Eq. (1) can be rewritten to:
xb % 1.43 (K/63/2F)±1/3a c±(a+1)/3a (1.5/2 p NA)(a+1)/3a
(c > c*) (2)
with xb in cm and c in g/mL and NA being the Avogadro
number. Note that Eq. (2) is the more general form of Eq.
(11.12) in [10] and Eq. (6) in [17], respectively, with a
slightly different prefactor. However, note that the equa-
tions mentioned above are based on scaling laws and the
prefactors cannot be given exactly. Equation (2) means
that the ªpore sizeº of an entangled polymer solution does
not depend on the degree of polymerization but only on
its concentration, c, and the nature of the polymer
(through K and a).
1.2 Mechanism of DNA movement
The mechanism of DNA movement in a porous matrix
(gels and entangled polymer solutions) under the influ-
ence of an electric field has been the subject of intensive
research and reviewed in many articles (e.g., [10, 18,
19]). In the earliest model, the so-called Ogston-model
[20, 21], the gel is assumed to act as a sieve with a distri-
bution of pore sizes and the separation is regarded as a
kind of filtration, driven by the electric field. This leads to a
mobility m proportional to exp(±conc.):
m/m0 = exp[±KR c] (3)
where m0 is the mobility in pure solvent (free solution mo-
bility), KR is the retardation coefficient (proportional to (Rg
+ r)2 with r being the radius of the gel fibers and Rg the
radius of gyration of the DNA molecule), and c is the gel
concentration. The Ogston model was originally devel-
oped for the movement of spherical objects in a random
array of cylindrical obstacles, where the pore size follows
Poisson statistics. In entangled polymer solutions, how-
ever, the ªporesº are probably more uniform. Second, it is
not clear if one can model a quickly rotating rigid rod-like
1963
short DNA molecule as a so-called ªequivalent sphereº.
Therefore, in the context of DNA separation in entangled
polymer solutions we can only talk about ªOgston-likeº
sieving mechanisms. Recently, the Ogston model has
come under radical reconsideration [22], and an improved
model is under development.
Obviously, sieving models are not valid for the situation
when DNA molecules are larger than the pores of the gel
(Rg > x, with x being the pore size of the separation
matrix), and a number of models based on the so-called
reptation concept were developed for this case. In such
models, the DNA chain is assumed to thread its way
through a ªtubeº defined by the fibers (for a rigid mesh) or
the ªblobsº (for a flexible network) surrounding it. One of
the models to which experimental data fit best, the
ªbiased reptation with fluctuationsº (BRF) model [23], is
an improvement of the original ªbiased reptation modelº
(BRM) [24]. Here, we follow the notation by Viovy [25],
who has recently further improved the BRF by taking into
account the electrohydrodynamic forces acting on the
whole section of DNA in a pore. This replaces the so-
called ªlocal forceº idea, which assumes that each seg-
ment of the chain (Kuhn segment) is submitted to an elec-
tric force.
Very briefly, in the BRF, two different cases are consid-
ered: first, when the pore size x of the matrix is larger than
the Kuhn length bD of DNA (xk = x/bD > 1) and second,
when the pore size is smaller (so-called ªtight gelº or ªper-
CE and CEC
sistent chainº situation, xk < 1). The Kuhn length (twice
the persistence length) is a natural parameter and a
measure for the flexibility of a polymer. Here we have
introduced a dimensionless parameter, xk, which is the
pore size normalized to the Kuhn length. In the first case,
which can also be regarded as the ªflexible chainº situa-
tion, the BRF describes the electrophoretic mobility of a
polyelectrolyte as [25]:
m/m0 = xk2 [(1/3Nk)2 + (2 ek/(5+2 a ek xk2))2]1/2 (4)
where a represents the ratio between the free mobility
and the limiting in-gel mobility of DNA in strong fields and
was introduced phenomenologically to account for the
saturation of mobility at strong electric fields. In experi-
ments and computer simulations it is found to be of the
order of 3 [25]. In Eq. (4), we again introduced two dimen-
sionless parameters, i.e., the DNA size (curvilinear
length, N) also normalized to the Kuhn length: Nk = N/bD
and the ªscaled electric fieldº ek = hbD2m0E/kT, with h
being the ªlocal viscosityº (viscosity of pure buffer without
polymer), E the electric field, k the Boltzman constant,
and T the absolute temperature. Thus, all dimensionless
parameters depend on the properties of the DNA (i.e., the
1964 C. Heller
Kuhn length) and not on the properties of the separation
matrix. We have chosen this notation, as the derived
equations then directly give us the pore size dependence.
Theoretical curves obtained with this equation [25] fit the
experimental data of dsDNA mobility in agarose gels [26].
Equation (4) gives the well-known behavior described by
all reptation models: For small molecules (below a critical
size, Nk*), the first term dominates and the mobility is
inversely proportional to the DNA size (ªreptation without
orientationº), but for large DNA and/or high electric fields,
the second term dominates, which means that the mobili-
ty becomes independent of size (plateau mobility) and
separation fails (ªreptation with orientationº). For the criti-
cal size, the model predicts, that Nk* = ek±1, i.e., independ-
ent of pore size. This prediction, which is supported by
experimental results in gels [26] is the major difference to
earlier models, which predict a xk±1 [23] and a xk±4 de-
pendence [27], respectively. Also, the revised model pre-
dicts a mobility scaling with xk2 in the reptation with orien-
tation regime, whereas earlier models predict an
exponent of 3 [23] and 6 [25], respectively.
In tight gels (xk < 1), reptation still occurs below a critical
size:
m/m0 ~ 1/3 Nk Nk < Nk* (5)
but for large DNA (Nk > Nk*), the existence of three differ-
ent regimes with the following plateau mobilities are pre-
dicted [23, 25]:
m/m0 ~ ek xk3/2 ek < xk3/2 (6a)
m/m0 ~ (ek xk6)2/5 xk±1 > ek > xk3/2 (6b)
and:
m/m0 ~ ek2 xk4 ek > xk±1 (6c)
For the critical size, the predictions are Nk* = ek±3/2, ek2/5
xk±12/5 and ek±2 xk±4, respectively. The last Eq. (6c) repre-
sents the regime when x 5 bD and length fluctuations are
frozen. In this case we recover the prediction of the origi-
nal BRM. The second regime (Eq. 6b) has been observed
by Mitnik et al. [28] for dsDNA in entangled hydroxypro-
pylcellulose solutions.
Polymer solutions differ from gels in that the physical
entanglements between the chains have a finite lifetime
and we have to take into account the dynamics of the
matrix itself. Viovy and Duke [11] have adopted the physi-
cal concept of ªconstraint releaseº (CR) to describe this
phenomenon. It is based on the fact that polymers them-
selves move by a random curvilinear diffusion (reptation).
Electrophoresis 1999, 20, 1962±1977
The contribution of constraint release to the mobility of
DNA is given by:
mCR/m0 % (xb/bD) (Rp/xb)±5 (7)
Assuming that the biased reptation and the constraint
release are independent processes, the net DNA mobility
is then given by the sum of both mobilities: mtot = mBRF +
mCR. Note that the constraint release mobility is depend-
ent on both polymer concentration and the chain length,
but independent of DNA size and applied electric field.
This means that for efficient separation to be achieved in
entangled polymer solution, the constraint release mobili-
ty should be smaller than the BRF component. For large
pores, this condition can be rewritten to: c/c* > (NkbD/
Rp)1/3 [11].
1.3 Network dynamics
The influence of network dynamics on size-based electro-
phoretic separation has been studied in more detail by
Cottet et al. [16]. Estimating the reptation time of the
matrix polymer, they found that a threshold value exists
below which the electrophoretic mobility of the analyte is
influenced by the polymer motion. Above this value (i.e.,
for long chains or at high concentrations), the polymer
reptation ceases to be a parameter influencing the sepa-
ration, and mobility vs. size plots will fall onto the same
curve. The approach of comparing the lifetime of obsta-
cles to the renewal time of the analyte is consistent with
the concept of constraint release and allows further
insight into the complex behavior of electrophoretically
driven polyelectrolytes in polymer solutions. It is also
possible that the polymer chains are actively dragged by
the moving DNA (ªnetwork ruptureº) and indeed such
effects have been observed in videomicroscopic studies,
even in semidilute solutions [28, 29]. They can occur dur-
ing a so-called U-conformation, when a DNA molecule
becomes hooked around a polymer chain. In contrast to
gels, often the whole conformation is then seen to move
downfield. This behavior has some similarity to the ªen-
tanglement couplingº process, which has been studied
both experimentally and theoretically in dilute solutions [2,
30]. However, to date there is no theoretical treatment of
the combination or reptative movement and polymer
dragging, which might describe the DNA movement in
entangled polymer solutions.
2 Material and methods
2.1 Polymers
pDMA is synthesized from the monomer, dimethylacryl-
amide, by radical polymerization. The chain length can be
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA
influenced by changing the temperature or by adding radi-
cal traps (ªchain transfer reagentsº), such as isopropanol.
The preparation follows the protocol by Chiari et al. [3] as
described before [8]: N,N-dimethylacrylamide (Polyscien-
ces, Warrington, PA, USA) was dissolved in water to a
final concentration of 10% w/w and isopropanol was
added in various amounts. After degassing (with vacuum
and ultrasound), 1 mL TEMED per mL solution and 1 mL/
mL of a 40% ammonium persulfate solution was added.
After careful mixture, the solution was allowed to react for
1 h at 25oC or 50oC, followed by incubation at 25oC over-
night. The reaction product was extensively dialyzed
against water using 25 000 MWCO dialysis membranes
(Spectrum, Los Angeles, CA, USA) and lyophilized.
2.2 Electrophoresis
For all separations the ABI 310 capillary electrophoresis
device (Perkin Elmer ± ABI, Foster City, CA, USA) was
used, equipped with a 47 cm long (36 cm to the detection
window) 375/100 mm OD/ID fused-silica capillary (Polymi-
cro Technologies, Phoenix, AZ, USA). 0.5 ´ TBE buffer
(44.5 mM Tris, 44.5 mM boric acid, 1.25 mM EDTA, pH 8.3,
condictivity 625 mS) was used as background electrolyte
for the polymer (which was only in the capillary) and as
electrode buffer. For denaturing conditions, 4 M urea (final
concentration) was added to the polymer and electro-
phoresis took place at 50oC; for nondenaturing conditions
the temperature was set to 25oC (and urea was omitted).
For the free mobility measurements, the capillary was
dynamically coated by flushing a 2% pDMA solution
through the capillary followed by rinsing with buffer (0.5 ´
TBE). This coating proved to be stable for at least ten
injections.
2.3 Samples
As DNA sample we used Sizer 50±500 (Pharmacia,
Uppsala, Sweden), Low Range DNA Standard (pBR322-
HinfI digest), 100 bp ruler and 500 bp ruler (Bio-Rad,
Hercules, CA, USA), all of them being labeled with fluo-
rescein. For denaturing conditions, 8 mL formamide
(Amresco, Solon, OH, USA) were added to 2 mL DNA,
heated to 90oC for 2 min and chilled on ice before injec-
tion. For nondenaturing conditions, water was added
instead of formamide, without any heating. For unknown
reasons, under denaturing conditions, double peaks were
observed for all fragments of the 100 bp and 500 bp ruler.
Comparative runs revealed that in all cases, the first peak
of the first five doublets in the 100 bp ruler comigrates
with the respective fragment of the 50 bp ruler. Therefore,
the faster one of each double peak was taken for subse-
quent analysis.
1965
3 Results
3.1 Characterization of the separation matrix
We have produced the following three pDMA prepara-
tions of different chain length, which have been character-
ized by size exclusion chromatography: (i) ªShort chainº
pDMA, synthesized in presence of 2% isopropanol at
50oC. Mv = 169 039, Mw = 216 774, polydispersity (Mw/
Mn): 4.39, [h] = 139.07 mL/g. (ii) ªMedium chainº pDMA,
synthesized in presence of 1% isopropanol at 25oC. Mv =
477 488, Mw = 603 846, polydispersity (Mw/Mn): 5.29, [h]
= 326.71 mL/g. (iii) ªLong chainº pDMA, synthesized with-
out isopropanol at 25oC. Mv = 816 574, Mw = 1 146 319,
polydispersity (Mw/Mn): 5.35, [h] = 424.06 mL/g. Note that
the same polymer preparations were characterized before
by viscosity measurements (Table 1 of [8]), with molecu-
lar masses of 490 kDa, 934 kDa and 1536 kDa, respec-
tively. We assume that the measurements with size
exclusion chromatography are more precise than the esti-
mations based on intrinsic viscosity alone and therefore
we shall use the values given above.
As the published Mark-Houwink coefficients for pDMA are
disputable [8], we determined them again by a least-
square fit (not shown) of [h] vs. Mv, using the values
above. We obtained K = 0.023 mL/g and a = 0.72 (r2 =
0.98), which means that he exponent a is between the
values given in the literature for 25oC (K = 0.023 mL/g, a
= 0.81) and for 40oC (K = 0.02 mL/g, a = 0.65) [14]. The
entanglement threshold of polymer solutions is inversely
proportional to the intrinsic viscosity [12]: c* % 1.5/[h].
Taking into account the relation between the viscosity
average molecular mass and the intrinsic viscosity (see
Section 1), we determined c* for all three polymer prepa-
rations by using the molecular masses shown above and
the newly determined Mark-Houwink parameters: c* %
(1.5/K)Mv±a (data not shown). As a second method, the
Table 1. Free mobility of ssDNA and dsDNA measured
in capillary electrophoresis with 0% pDMA at
different temperature with and without urea
Buffer ssDNa dsDNA
(mm/s)/(V/cm) (mm/s)/(V/cm)
0.5 ´ TBE at 25oC 3.95 4.37c)
0.5 ´ TBE at 50oC 5.25 5.89
0.5 ´ TBE + 4 M urea
at 25oC 3.58a) 4.01
0.5 ´ TBE + 4 M urea
at 50oC 4.64b) 5.24
a) For comparison, see [36]
b) In agreement with data from Fig. 3
c) In agreement with data from [8, 35]
Other conditions as in Fig. 1
1966 C. Heller Electrophoresis 1999, 20, 1962±1977

Figure 1. Dependence of the electrophoretic mobility of linear DNA fragments on size, separated in
pDMA of different molecular mass and at different concentrations. (A) dsDNA in short-chain pDMA;
(B) dsDNA in medium-chain pDMA; (C) dsDNA in long-chain pDMA; (D) ssDNA in short-chain pDMA;
(E) ssDNA in medium-chain pDMA and (F) ssDNA in long-chain pDMA. Conditions: 0.5 ´ TBE at
210 V/cm in a 47 cm long (36 cm to the detector) fused-silica (100 mm ID) capillary. (A)±(C) Separa-
tion at 25oC, (D)±(F) separation in 4 M urea and at 50oC. Zones I±IV (separated by dotted lines) corre-
spond to the different regimes explained in Section 3.2. The straight lines in (E) indicate the finding of
the critical size, N*; the straight line in (F) has a slope of ±1.
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA 1967

threshold was also determined by measuring the viscosity regime IV) and smaller ones (70±250 bp, regime II) show
of the polymer solution at different concentrations and less difference in mobility. Regime II is most expressed at
finding the point of departure from linearity on the double higher polymer concentration (5 and 10%), whereas at
logarithmic specific viscosity vs. concentration plot (not low concentration, there is a smooth transition from
shown) [31]. Also, by using the independently measured regime I to III. This is in full agreement with earlier data
intrinsic viscosity values, we have previously determined [8]; however, the border between regimes III and IV has
the corresponding c* values of the different pDMA prepa-
rations (see Table 1 in [8]). Taking together the three
independent data sets, we get c* = 0.013  0.003 g/mL,
c* = 0.007  0.002 g/mL and c* = 0.004  0.001 g/mL for
ªshort chainº, ªmedium chainº and ªlong chainº pDMA, re-
spectively.

Using Eq. (2) and taking into account the new Mark-Hou-
wink parameters, we obtain ªmesh sizesº of about 3, 5, 11
and 19 nm, for 10%, 5%, 2% and 1% pDMA solutions, re-
spectively, independent of the chain length. Note that the
Mark-Houwink parameters were determined for a temper-
ature of 25oC. Under denaturing conditions (i.e., at 50oC
and in presence of urea), we might therefore obtain
slightly different values for the overlap threshold and the
mesh size. However, in any case ± as pointed out in Sec-
tion 1 ± the above calculations are based on scaling laws
only and precise prefactors can not be given. In our setup
it is difficult to measure the electroosmotic flow (EOF), but
under similar conditions, the residual EOF in a 0.01%
pDMA (98 kDa) solution was reported to be 0.2 ´ 10±4
cm2/Vs [5]. For our experiments, using longer pDMA
chains and a hundred times higher polymer concentra-
tion, we can therefore safely assume that the residual
EOF in our separation matrix is negligible.

3.2 Mobility and separation regimes

We have undertaken a systematic study of the mobility of
double-stranded and single-stranded DNA in pDMA prep-
arations of different molecular masses and at different
polymer concentrations. Figure 1 shows the dependence
of the electrophoretic mobility on DNA size (dsDNA and
ssDNA) at an electric field strength of 210 V/cm and at
polymer concentrations of 1%, 2%, 5% and 10% in ªshort
chainº, ªmedium chainº and ªlong chainº pDMA (with ªlong
chainº pDMA, we did not perform experiments at 10%,
due to the rather high viscosity, > 10 000 mm2/s). In all
cases, we can observe the usual sigmoidal curve on the
double logarithmic mobility vs. size plot, in agreement
with separation of dsDNA in agarose gels [26, 32], ssDNA
in polyacrylamide gels [33] or dsDNA in entangled poly-
mer solutions [17, 28]. Figure 2. Plot of mobility vs. the inverse of DNA size for
separation in medium chain pDMA (data from Figs. 1B
For dsDNA (Fig. 1A±C), we can define the following and E). (A) dsDNA; separation at 25oC; (B) ssDNA; sepa-
regimes (according to the notation in our previous work ration in 4 M urea and at 50oC. Other conditions as in Fig.
[8]): medium-sized molecules (250±900 bp) are well re- 1. The zones I±IV (separated by dotted lines) correspond
solved (regime III), whereas larger ones (> 900 bp; to the different regimes explained in Section 3.2.
1968 C. Heller Electrophoresis 1999, 20, 1962±1977

been slightly shifted to larger DNA sizes. A possible

explanation for this might be that in earlier studies, inter-
calators were used as fluorescent markers instead of a
covalently bound dye. Such intercalators can change the
properties of DNA (charge, flexibility and length), there-
fore leading to slightly different results. Also, note that the
limits are difficult to define as smooth transitions occur be-
tween the different regimes. In this new series of experi-
ments we did not use small DNA; therefore in this case,
regime I (< 70 bp) can not be well observed. For better
illustration, we refer our readers to Fig. 7 of [8].

In our previous work [8], we could identify the different

regimes as follows: Regime I corresponds to an Ogston-
type sieving mechanism, whereas regime III and IV indi-
cate unoriented reptation and reptation with orientation,
respectively. Regime II (plateau regime on the left hand
side) is a newly discovered regime and presumably corre-
sponds to the situation when the radius of gyration of
DNA, Rg, is larger than the pore size (so that sieving is
impossible), but at the same time the DNA is too stiff to
effectively reptate through the separation matrix [8]. This
is the case when the DNA is shorter than one Kuhn
length, and can occur in the ªtight gelº situation (see Sec-
tion 4). Presumably, the DNA molecules assume a ªrod-
likeº conformation and size-based separation is sup-
pressed in this regime.
In the case of ssDNA (denaturing conditions, Fig. 1D±F),
the situation is more complex. We can observe again that
plateau mobility (regime IV) starts to occur above a cer-
tain size, but this critical size increases with decreasing
polymer concentration. At the left-hand side of the plot,
we cannot observe a plateau regime, but instead a direct
transition from zone I to III. Again, the point of transition
increases with increasing polymer concentration. As in
the case of dsDNA, the slope in regime III decreases with
increasing polymer concentration but a slope of ±1, as
would be demanded by the reptation model, is not
reached. However, considering the rather strong electric
field (210 V/cm) used here, we cannot expect to observe
ªtrueº, i.e. unbiased reptation, but ± in analogy to the
dsDNA separation ± we can again assume that in this
regime the separations is still based on a reptation mech-
anism. To verify this assumption, we have replotted ± as
an example ± a subset of the data of Figs. 1B and 1E as
linear mobility vs. inverse of size plot (Fig. 2A and B). If
reptation is the dominant mechanism, the mobility is pro-
portional to 1/size and the data points will fall on a straight
line. Indeed, in Fig. 2 we observe such a linear relation-
ship within the domain of regime III, both for ssDNA as
well as for dsDNA. In zone IV (Fig. 1), the mobility levels
off, which indicates that it corresponds to the reptation
with orientation regime. As observed for the separation of
Figure 3. Semilogarithmic plot of the electrophoretic mo-
bility vs. pDMA concentration (ªFerguson plotº) for three
different ssDNA fragments, each one selected from one
of the three regimes of Fig. 1E (medium-chain pDMA,
603 kDa). Conditions as in Fig. 1. The straight line repre-
sents a fit of the data for the 75 b fragment to the following
relation: Y = 4.60 exp[±0.122c]; r2 = 0.995.
dsDNA in different entangled polymer solutions [8, 17,
28], the mobility does not become totally independent of
size.
In order to verify if zone I corresponds to an Ogston-like
separation, we picked one representative fragment from
each zone of Fig. 1E and plotted the data in a semiloga-
rithmic mobility vs. concentration plot (so-called Fergu-
son-plot [34], Fig. 3). If we assume that the radius of gyra-
tion of the DNA is larger than the fiber radius, Eq. (3)
predicts that the Ogston-type separation should lead to a
straight line with negative slope. Indeed, we find that the
75 b fragment fulfills this condition, whereas the other
fragments do not. The straight line intercepts the y-axis at
a value of 4.6 ´ 10±4 cm2/Vs, which agrees well with the
experimental data of free solution mobility (see below and
Table 1). The same behaviour has been verified for the
separations in ªshort chainº and ªlong chainº pDMA (not
shown).
3.3 Mobility in free solution
In Fig. 1, we also observe that the mobility of the small
ssDNA molecules (i.e.; at the left-hand side of the plots,
Fig. 1D±F) is generally higher than the mobility of the cor-
responding dsDNA standards (Fig. 1A±C). This is in
agreement with the findings by van der Schans et al. [9]
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA 1969

rithmic mobility vs. size plot, indicating that a higher reso-

lution might be achieved when using denaturing condi-

Figure 4. Mobility difference per base or per base pair, in

((mm/s)/(V/cm))/b or ((mm/s)/(V/cm))/bp, respectively, for
ssDNA and dsDNA in 5% pDMA (1146 kDa). The curves
are a guide to the eye only. Data are taken from Figs. 1C
and 1F.

and can be attributed to a lower viscosity due to the high-

er temperature (50oC vs. 25oC). Table 1 gives the free
solution mobilities (measured in 0% pDMA) for both
cases, i.e., ssDNA separated in denaturing conditions
(50oC and 4 M urea) and dsDNA in nondenaturing condi-
tions (25oC without urea). In order to obtain the influence
of urea and temperature alone, we also measured the
free solution mobilities under the ªwrongº conditions, i.e.,
ssDNA in nondenaturing conditions and dsDNA in dena-
turing conditions. The result for dsDNA agrees well with
previously evaluated data [8, 35], but ssDNA is about two
times faster than reported in the literature [36]. Note, how-
ever, that in our case, the buffer ionic strength was about
five times higher. Generally, we find that under the same
conditions, the mobility of ssDNA is about 0.9 times the
mobility of dsDNA. Additionally, the presence of urea
slows down the mobility by about the same factor. How-
ever, this is more than compensated for by a 1.3-fold
increase in mobility due to the higher temperature under
denaturing conditions. Figure 5. (A) Dependence of the electrophoretic mobility
of linear DNA fragments on the pore size, separated in
medium chain pDMA (603 kDa). Conditions as in Fig. 1.
3.4 Mobility difference The lines represent a fit of the data to the following rela-
tion: 500 b ssDNA: Y = ±1.23 + 3.3 log(pore size), (r2 =
However, more striking than the mobility difference for
0.997); 5000 b ssDNA: Y = ±0.64 + 2.0 log(pore size),
small molecules is the fact that the large single-stranded (r2 = 0.999); 500 bp dsDNA: Y = ±0.27 + 2.2 log(pore
molecules (i.e., those nearly assuming a plateau mobility, size), (r2 = 0.998); 5000 bp dsDNA: Y = ±0.354 + 1.98
at the right-hand side of the plots, regime IV) migrate con- log(pore size), (r2 = 0.997). (B) Same data as in (A), but
siderably slower than their nondenatured counterparts. presented in a double logarithmic plot. The straight line
This leads to steeper slopes in zone III of the double loga- has a slope of 1.5, the dashed line has a slope of 1.
1970 C. Heller
tions instead of nondenaturing conditions, due to larger
spacing between the peaks. This effect is better displayed
in Fig. 4: We have taken the difference in mobility (nor-
malized to the size) between neighboring fragments, plot-
ted vs. the size of the smaller fragment of each ªpairº for
the data obtained in 5% pDMA (long chain), as an exam-
ple. We clearly see that the mobility differences (and
therefore the peak distances on the electropherograms)
obtained under denaturing conditions are considerably
higher than under nondenaturing conditions (for frag-
ments < 400 b). Similar results are obtained for different
Figure 6. Dependence of the electrophoretic mobility of
linear DNA fragments on the polymer chain length at a
concentration of 1% and 5% pDMA. (A) dsDNA; separa-
tion at 25oC; (B) ssDNA; separation in 4 M urea and 50oC.
Other conditions as in Fig. 1.
Electrophoresis 1999, 20, 1962±1977
polymer concentrations and different chain lengths (not
shown). This striking difference and the differences
explained above led us to the assumption that a different
separation regime might be valid for ssDNA compared to
dsDNA. As ssDNA has a much smaller Kuhn length than
duplex DNA, it might be possible that here we might not
have the situation of a ªtight gelº. In the following, we fur-
ther investigate this possibility.
3.5 Influence of pore size
To gain more insight into the separation mechanism, we
have investigated the influence of the pore size on the
mobility. Again, we only present the data from the experi-
ments in medium sized pDMA, as we could not find signif-
icant differences in the data sets obtained in ªsmall chainº
or ªlong chainº pDMA. We picked the 500 b (and 500 bp,
respectively) fragment as a representative for DNA sep-
arated by nonoriented reptation, and the 5000 b (5000
bp) fragment as an example for an ªorientedª molecule.
Figure 5 shows a semilogarithmic plot of the dependence
of mobility on the pore size for dsDNA and ssDNA. The
data can be fitted to logarithmic functions (see legend).
Interestingly, the slopes for both dsDNA fragments are
similar, whereas they change considerably from ssDNA in
the unoriented regime to ssDNA in the oriented regime.
We have replotted the same data in a double logarithmic
mobility vs. pore size plot in order to verify if the data
could be described by a power law. Clearly, the data do
not fit a straight line; only for small pore sizes (concentra-
tions of 5 and 10%) will they approximately fall on a line
with a slope of 1.5 for dsDNA and 1 for ssDNA.
3.6 Influence of chain length
As pointed out in Section 1, the length of the polymer
chains might influence the network dynamics and there-
fore the separation. To illustrate this effect, we have
regrouped part of the data from Fig. 1 into new plots pre-
senting the DNA mobility in matrices of the same concen-
tration but of different chain length (Fig. 6). Under native
conditions (Fig. 6A), the respective curves obtained are
identical (within experimental error), both at low (1%) and
at high (5%) concentration. Following the argument by
Cottet et al. (see Fig. 2 of [16]), this indicates that even for
short chain pDMA, at 1% the ªthreshold valueº is already
reached, above which the reptation of the polymer ceases
to affect the separation. The situation is slightly different
under denaturing conditions (Fig. 6B): in 1% and in 5%
pDMA, the mobility of the larger fragments (zone III and
IV) slightly decreases with increasing chain length, indi-
cating that network dynamics may be influencing the sep-
aration.
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA 1971

Figure 7. (A) Dependence of the electrophoretic mobility on the size of linear ssDNA fragments sep-
arated in 2% pDMA (1146 kDa) at different electric fields. Other conditions as in Fig. 1. The straight
line is for illustration only and has a slope of ±1. (B) Dependence of the electrophoretic mobility on the
size of linear ssDNA fragments separated in 5% pDMA (1146 kDa) at different electric fields. Other
conditions as in Fig. 1. The straight line illustrates a slope of ±1. (C) Dependence of the electropho-
retic mobility of linear ssDNA fragments on the electric field in 2% pDMA (1146 kDa). Data are from
Fig. 5A. The straight line has a slope of 0.4. (D) Dependence of the electrophoretic mobility of linear
ssDNA fragments on the electric field in 5% pDMA (1146 kDa). Data are from Fig. 5B. The straight
lines (for illustration only) have a slope of 0.4 and 1, respectively.

3.7 Influence of electric field
Besides the polymer concentration and the pore size, the
electric field strength is one of the major factors influenc-
ing the mobility of DNA in a porous matrix. As most elec-
trophoresis models are strictly valid for zero or low electric
field only, it is important to measure the field dependence
of a separation method. The influence of the electric field
on the mobility of dsDNA has been studied before [8]; we
will therefore restrict ourselves to the mobility of ssDNA.
Figures 7A and B show again double logarithmic plots of
mobility vs. size under denaturing conditions, obtained in
2% long chain pDMA and 5% long chain pDMA, respec-
tively, but at different electric fields. Again, the result is
similar to that obtained for dsDNA in pDMA [8] and in
other matrices [28]. However, in the example shown here
1972 C. Heller
Figure 8. Dependence of the critical size, N* on the pore
size. Conditions as in Fig. 1. The straight line is horizon-
tal, the dashed line (for illustration only) has a slope of
0.4. (B) Dependence of the critical size, N*, of ssDNA on
the electric field strength in 5% pDMA (1146 kDa). Condi-
tions as in Fig. 7. The straight line (for illustration only)
has a slope of ±0.5.
(Fig. 7A), a slope of ±1 could not be reached in 2%
pDMA. However, at higher concentration (5% pDMA; Fig.
7B) and low electric field, we nearly approach ideal repta-
tion conditions in the size range between 200 and 700 b.
A subset of the data from Figs. 7A and B was replotted in
order to demonstrate the dependence of the mobility on
the electric field (Figs. 7C and D). Again, the results are in
agreement to those obtained under nondenaturing condi-
tions [8]: At low electric field and small DNA size, the mo-
bility remains constant (but dependent on the size),
whereas larger molecules show an electric field depend-
ence. It is difficult to find scaling laws over this range of
electric field strengths, but for large molecules in 2% long-
chain pDMA the mobilities fall on a line with a slope of
0.4. In 5% pDMA and at strong electric field, there is a
deviation to a steeper slope, close to unity.
Electrophoresis 1999, 20, 1962±1977
3.8 Critical size, N*
We have also evaluated the crossover point N*, which
marks the transition from reptation in Gaussian conforma-
tion to oriented conformations. This parameter is impor-
tant because it indicates the point where separation starts
to fail. In DNA sequencing, for example, this can limit ±
among other factors ± the number of bases that can be
read. N* can be estimated graphically by finding the inter-
section between the average straight line with a steep
slope formed by the data in regime III of the double loga-
rithmic mobility vs. size plot and the nearly horizontal line
connecting the data points from the plateau mobility (for
example, see data for 5% pDMA in Fig. 1E). Figure 8A
shows a double logarithmic plot of N* vs. pore size for
ssDNA in all three polymer preparations. With the excep-
tion of 1% short-chain pDMA (which is probably only a
weakly entangled polymer solution), the data can be fitted
to straight lines with a slope of 0.4. As already mentioned
above, for dsDNA the ªcritical sizeº N* is constant over
about one order of magnitude of pore sizes. The cross-
over point has also been determined in dependence on
the electric field. In Fig. 8B we present the data obtained
from separations in 5% long-chain pDMA (Fig. 7B): the
data satisfy a power law of ±0.5.
3.9 Comparison to theoretical predictions
For small molecules (regime I), the separation of ssDNA
and dsDNA can be explained by a sieving mechanism
(see Fig. 3, and Fig. 8 of [8]). However, for larger DNA
(regime III and IV) the situation is more difficult. As the
theoretical treatment of the BRF is rather complex, we
have compiled the major scaling predictions in Table 2
(again, we follow [25]). dsDNA has a persistence length
of about 45 nm [37, 38], which means a Kuhn length of
about 90 nm, corresponding to about 300 bp. Taking into
account the pore sizes calculated above for the pDMA
solutions used here, we were confident that, in this case,
we were working under ªtight gelº conditions. The fact that
we can observe a plateau zone at the left-hand side of the
mobility vs. size plot (regime II) supports this assumption
(see also Fig. 9 and Section 4). Therefore, we have
compared our dsDNA data with the predictions for persis-
tent chains (lower part of Table 2). On the other hand, the
persistence length of ssDNA was recently determined to
be about 4 nm (in low ionic strength, [39]), corresponding
to a Kuhn length of 8 nm or about 20 nucleotides. This is
in the same order of magnitude as the ªpore sizeº of a 5±
10% pDMA solution, but given the fact that the prefactors
in Eq. (2) cannot be given exactly and that the flexibility of
DNA is greater in higher ionic strength buffers (persis-
tence length of ssDNA in 0.5 ´ TBE is about 2 nm [36]),
we think that we do not have a ªtight gelº situation here.
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA 1973

Table 2. Scaling predictions of the BRF for different parameters, according to [25], in comparison to experimental results
for ssDNA and desDNA in entangled pDMA solutions
DNA size Electric field Pore size
Predicted Observed Predicted Observed Predicted Observed
Flexible chain ssDNA Flexible chain ssDNA Flexible chain ssDNA

m/m0 unoriented Nk±1 N±1 a) Independent Independent xk2 x1.5 0.2 d)

m/m0 oriented Independent Nearly indep. ek E0.4±1 b) xk2 x1.5 0.2 d)
 0.4  0.1
N* ± ± ek±1 E±0.5 0.1 Independent x

Persist. chain dsDNA Persist. chain dsDNA Persist. chain dsDNA


m/m0 unoriented Nk±1 N±1 a) Independent Independent Independent x1.5 0.2 d)

m/m0 oriented Independent Nearly indep. ek; ek2/5; ek2 E0.15±0.4 b, c) xk3/2; xk12/5; xk4 x1 0.2 d)
N* ± ± ek±1; ek±2/5; ek±2 n.d. xk±3/2; xk±12/5; xk±4 Independent
a) For small electric fields and small pore size
b) For large DNA
c) See Fig. 9C of [8]
d) Only for high polymer concentration, but better fitted by a logarithmic function (see Fig. 4)

aration, i.e., a separation inversely proportional to the size

and a plateau mobility above a critical size (sigmoidal
shape of the double logarithmic mobility vs. size plot).
However, other predictions, like the dependence of mobil-
ity and critical size on pore size, do not agree with the
results. Concerning the electric field, the independence of
the mobility in the unoriented reptation regime is con-
firmed; however, in the oriented regime, the electric field
dependence is more complex. For the critical size under
denaturing conditions, we find a dependence on the
square root of the inverse electric field whereas the BRF
predicts N* proportional to E±1. Our results are in contrast
to findings from Dovichi©s group, who indeed reported a
dependence of N* nearly inversely proportional to the
electric field in 5% non-cross-linked polyacrylamide (see
Figs. 10 and 11 of [40]).

Figure 9. Diagram of the different regimes existing in the

electrophoretic separation of DNA in polymer solutions.
The dotted line indicates the border between (A) the ªper-
sistent chainº situation and (B) the ªflexible chainº situa-
tion. The dashed line indicates the predicted critical size
(for better clarity, we have only indicated one of the three
different possibilities existing in ªtight gelsº). The arrows
indicate the situation for ssDNA and dsDNA in polymer
solutions of different concentration. For explanation of the
different regimes, see Section 3.2.
Also, we cannot observe a zone II as exists in dsDNA
separations and therefore we have compared our ssDNA
data to the predictions for flexible chains (upper part of
Table 2).
As can be seen from Table 2, the BRF can only correctly
predict the major observation in electrophoretic DNA sep-
4 Discussion
In ªclassicalº gel electrophoresis, dsDNA and ssDNA are
usually separated in different matrices. The pore size of
the gel is adapted to the flexibility of the DNA, i.e., we use
agarose with large pores to separate the rather stiff
dsDNA and cross-linked polyacrylamide (small pores) for
the more flexible ssDNA. In both cases, we can assume
that the average pore size is larger than the Kuhn length
of the DNA and we have a ªflexible chainº situation. In
capillary electrophoresis, often the same matrix is used
for both types of DNA. In most cases, we shall then have
two different situations: Native DNA with a large Kuhn
length has to thread through rather small pores (ªtight
gelº), whereas for denatured and ssDNA, the ªflexible
chainº situation remains.
1974 C. Heller
4.1 ssDNA
This leads to different regimes for the mobility in depend-
ence on size, which is explained in Fig. 9: For ssDNA, the
Kuhn length is smaller than the pore size (except maybe
for very high polymer concentrations), i.e., bD < x, or xk >
1. For short molecules (i.e., smaller than the pore size, Rg
< x), the DNA undergoes separation by a sieving mech-
anism (regime I). With increasing size, the point will come
where the radius of gyration is bigger than the pore size
and the DNA will start to reptate (regime III, Rg > x > bD).
Naturally, this first crossover point shifts to larger DNA
with increasing pore size, which indeed can be observed
in Figs. 1D±F. The radius of gyration of a polymer can be
described by the so-called Kratky-Porod formula and has
been estimated by Viovy et al. [10] for both ssDNA and
dsDNA (see Fig. 4 of [10]). For ssDNA, a size of about 40
b, 100 b and 400 b would give a radius of gyration of 3, 5
and 11 nm, respectively corresponding to the estimated
pore sizes of the 10%, 5% and 2% pDMA solutions (see
above). The crossover points from regime I to regime III
for ssDNA (see dotted lines in Fig. 1D±F), do not fully
match these values, but come rather close. By further
increasing the size, the DNA will become oriented and we
reach regime IV (Rg 4 x > bD). According to the reptation
model, here the mobility should become independent of
size; however, in entangled polymer solutions we often
observe a slight size dependence in this regime. This
means that ± in contrast to gels ± the separation does not
totally fail for large DNA.
4.2 dsDNA
For small dsDNA, we have Rg < x < bD and the DNA is
sieved (regime I). As in ssDNA, with increasing size the
DNA will become larger than the pore size; however, as
the molecules are rather stiff, their radius of gyration can
be smaller than their Kuhn length. This leads to the spe-
cial situation of bD > Rg > x (regime II), where the mole-
cules are too large to be separated by sieving, but too stiff
to effectively reptate. Presumably, they adopt a rod-like
conformation and travel through the matrix at the same
speed. The crossover from regime I to regime II should
also be dependent on the polymer concentration, which
does not fully agree with our observations. However, the
crossover point is difficult to determine and the fact that
for large pore sizes, zone II becomes small (Fig. 9),
explains that for low polymer concentration, we have a
smoother transition from zone I to zone II (Fig. 1A±C).
The crossover from regime II to III (Rg > bD > x), however,
is given by the Kuhn length and the observed transition
point at about 250 bp agrees well with the estimated Kuhn
length of 300 bp. As in ssDNA, by further increasing the
size, the molecules become oriented during electrophore-
Electrophoresis 1999, 20, 1962±1977
sis, and we reach regime IV (Rg 4 bD > x), again with a
possible, but less than good separation.
4.3 Separation mechanism
As outlined above, we have attributed the separation in
zone I to a sieving mechanism, due to the fact that the
analytes are smaller than the estimated pore size of the
matrix. The ªclassicalº sieving theory applied in electro-
phoresis is the so-called Ogston model and, in fact, the
electrophoretic behavior of short DNA in our experiments
can be explained using this model (see Fig. 8 of [8] and
Fig. 3 in this study). However, this is somewhat surpris-
ing, because under our conditions the main assumptions
of the Ogston model are not valid: the Ogston model was
originally developed for the movement of spherical
objects in a random, fixed array of fibers, where the pore
size follows Poisson statistics. In entangled polymer solu-
tions, the ªporesº are probably more uniform and the
obstacles themselves might move (ªconstraint releaseº).
Second, a short DNA molecule is rod-like and it is ques-
tionable to what extent it can be regarded as a so-called
ªequivalent sphereº. Therefore, in the context of electro-
phoretic separation of short DNA molecules in entangled
polymer solutions we can only talk about ªOgston-likeº
sieving mechanisms and more theoretical work might be
necessary to clarify this point.
The separation of larger DNA molecules, i.e., larger than
the pore size, is classically explained reptation models. In
Table 2, we have compiled the predictions of the latest
reptation model, the biased BRF, taking into account elec-
trohydrodynamic forces. The BRF correctly predicts the
major observation in electrophoretic DNA separation, i.e.
the inversely proportional dependence of the mobility on
size (see Fig. 2) and the plateau mobility for large DNA
and high electric field. However, the model dramatically
fails when we look at the predicted dependencies on the
pore size. Neither for ssDNA nor for dsDNA could the pre-
dicted dependence of the mobility on the pore size be
verified. Even more important, the dependence of the crit-
ical size, N*, is predicted to be independent of pore size
for flexible chains, but increases with pore size in our
experiments. In tight gels, it should decrease with pore
size (in all three predicted cases; see Eq. 6A±C), but it
was found to be constant within experimental error. In the
original version of the BRF [23] and in the BRM [27], N*
was even predicted to scale with the pore size with a neg-
ative exponent, i.e. ±1 and ±4, respectively.
This wrong prediction of the pore size dependence seems
to be a major weakness of reptation models and only the
latest improvements have brought it closer to reality. For
instance, in gel electrophoresis, N* has been found to be
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA
nearly independent of pore size for dsDNA in agarose
gels (see Fig. 3 of [26]) and also for ssDNA in polyacryl-
amide gels (see Fig. 2 in [33]), which can now be
explained by the most recent version of the BRF [25].
However, for DNA separation in polymer solutions, it still
cannot account for the pore size dependence. The wrong
theoretical scaling of N* with pore size has led to the
wrong assumption that the read length in DNA sequenc-
ing could be increased with increasing gel concentration,
in contrast to the practical findings (e.g., [41]). Based on
the findings of the BRF model, for example, Slater et al.
[42] have predicted that the read length in gel electropho-
retic DNA sequencing could be increased by increasing
simultaneously the electric field and the gel concentration.
If we apply their rule (Eq. 25 of [42]) to our experimental
data in polymer solutions, we cannot confirm this predic-
tion. In 2% pDMA at 105 V/cm, we find a higher critical
size (about 1100 bases) than in 160 V/cm (1.6 ´ the elec-
tric field) and 5% pDMA (1.62 = 2.5 ´ the polymer concen-
tration).
The reason why the severe failure of reptation models
concerning the pore size dependence has been over-
looked might be due to the fact that, in gels, the effective
pore size is difficult to estimate. In fact, the relation be-
tween the pore size (and the pore size distribution) and
the concentration of agarose gels and polyacrylamide
gels, was the subject of intense discussion (see e.g.,
Table 1 in [43]). In entangled polymer solutions, we can
assume that the pore size is more uniform and that it can
correctly be described by the blob size. Still, in entangled
polymer solutions, the assignement of the pore size to a
solution with known concentration depends heavily on the
correct estimation of Mark-Houwink parameter a. Accord-
ing to Eq. (2), the pore size scales with c±(a+1)/3a. In our
case, this exponent is ±0.79. However, even if this expo-
nent were not correct, it should only vary between ±0.75
(for a = 0.8; swollen chain) and ±1 (a = 0.5; ideal solvent).
This could not explain the striking differences between
the predicted and the observed pore size dependence of
the mobility and the critical size.
The observed difference could be due to other factors.
For instance, it is not certain if the blob size corresponds
to the effective pore size of the network. In other words,
the DNA could ªseeº pores which are different from the
pores derived by the theory of polymer solutions. Also, as
the applied electric fields are rather high, the possibility of
ªhernia formationº [44] is increased. ªHerniasº or ªloopsº
are a sidewise escape of the DNA from the ªtubeº, and
reptation models become invalid. However, they should
only occur in flexible chains. Most models were originally
developed for low electric field strengths. To gain more
insight into the pore size dependence of mobility and criti-
1975
cal size, more experiments at different field strengths are
necessary. It is possible that extrapolations to low or zero
electric field could then reveal and asymptotic trend to dif-
ferent scaling laws.
For flexible chains, there is also a discrepancy between
the prediction and the observation concerning the electric
field dependence of the mobility in the oriented regime
and of the critical size, N*. In both cases, the scaling is
less strong than predicted. Only in highly entangled poly-
mer solutions and at rather strong electric fields do we
approach a linear scaling for the mobility of oriented
ssDNA on the electric field (Fig. 7D). Similarly, for dsDNA,
a slope of 0.4 could only be found for strong electric fields
(see Fig. 9C of [8]). Again, what has been said above
about the possible sources of the discrepancies remains
valid. Moreover, it is difficult to determine precise scalings
over a range of field intensities of only one order of magni-
tude.
4.4 Separation matrix
When working with polymer solutions, we also have to
take into account the flexible nature of the network itself.
The polymer chains are not linked to each other but can
move themselves. This leads to a renewal of the obsta-
cles and can be described by the concept of ªconstraint
releaseº [11]. The constraint release mobility adds to the
electrophoretic mobility and increases with shorter chain
length and lower polymer concentration (see Eq. 7). How-
ever, in our experiments with dsDNA in entangled pDMA
solutions we could not observe an effect on the separa-
tion with increasing chain length (Fig. 6A), indicating that
the threshold value, above which the reptation of the poly-
mer ceases to affect the separation, is already reached.
Therefore, we assume that constraint release does not
play a major role under these conditions.
However, when using the same matrix for the separation
of ssDNA, a slight effect on the electrophoretic mobility
can be observed with increasing chain length. This could
be explained as follows: The more flexible ssDNA has a
longer renewal time (biased reptation time) than dsDNA.
The reptation time of the matrix polymers, however,
remains the same. According to Cottet et al. [16], the net-
work dynamics has an influence on the separation when
the reptation time of the polymer is smaller than the
renewal time of the analyte, which could be the case here.
Another explanation could be given using the concept of
constraint release: The Kuhn length bD for ssDNA is much
smaller than for dsDNA, so that the first term in Eq. (7)
becomes larger. This in turn means a larger contribution
of constraint release to the total mobility than compared
to the case under nondenaturing conditions.
1976 C. Heller
Note, however, that the polymer chains can also be
actively dragged by the moving DNA, a process which ±
in contrast to constraint release ± should be dependent
on the DNA size and the electric field. The more flexible
ssDNA molecules probably get hooked around matrix
fibers more frequently than do the rather stiff duplex DNA
chains. Thus, a network rupture would be more likely.
Such effects could also explain the fact that even large
molecules are still separated and their mobility does not
fully level off. So far, no model exists that takes into
account such ªdraggingº effects in entangled polymer sol-
utions. We feel that the electrophoretic separation of DNA
in entangled polymer solutions is not fully understood and
more experimental and theoretical work would be needed
to reveal these phenomena.
Under denaturing conditions, the mobility of large DNA is
lower than under native conditions. Consequently, DNA
up to the critical size is better separated and we can say
that the system has a better selectivity under denaturing
conditions than under nondenaturing conditions (Fig. 4).
Therefore, if high resolution is necessary (i.e., one or two
base resolution) the DNA should be separated under
denaturing conditions. This is especially interesting in
applications such as genotyping and fingerprinting and, in
fact, such conditions are successfully applied to different
biological applications (e.g., [6]). We also found that for
ssDNA in pDMA solutions, the critical size, i.e., the point
beyond which separation fails, is increased with decreas-
ing polymer concentration. This effect could probably be
used for longer read lengths in DNA sequencing. How-
ever, both a higher resolution and longer read length can
only be achieved if the peak width does not increase
simultaneously with the selectivity or the critical size. The
dependence of the peak width on different experimental
conditions is the subject of a further study [45].
4.5 Conclusion
In this study, we have used pDMA as the separation
matrix. The results obtained for dsDNA are very similar to
those obtained in other matrices such as hydroxypropyl-
cellulose [28] or dextran [17]. Therefore, we can assume
that the findings for ssDNA might also be valid in other
entangled polymer matrices. We preferred pDMA
because of its unique properties such as its self-coating
ability and its low viscosity [5, 8]. Very recently, pDMA
has also become commercially available, but only as a
100 kDa preparation, which is considerably shorter than
the preparations used here. The work presented here
should help to optimize DNA separation in capillary elec-
trophoresis and to improve existing models for DNA sepa-
ration in porous matrices.
Electrophoresis 1999, 20, 1962±1977
I would like to thank R. Reinhard for providing the capil-
lary electrophoresis instrument and R. Steinkamp, V.
Egelhofer and J. Schacherl for their help with the charac-
terization of the polymer solutions and with the DNA sepa-
rations. J.-L. Viovy is thanked for providing his manuscript
prior to publication and for helpful discussions and one of
the referees is thanked for very helpful comments. I am
very grateful to M. Millequant, Labo PCM, ESPCI, Paris
(France) for performing the size exclusion chromatogra-
phy analysis. This work was supported by a grant from
the German Ministry of Research (BMBF), FKZ 0311040
and by a grant from the Commission of the European
Union, CT 97-2627.
Received February 1, 1999
5 References
[1] Heller, C., in: Heller, C. (Ed.), Analysis of Nucleic Acids by
Capillary Electrophoresis, Vieweg Verlagsgesellschaft,
Wiesbaden 1997, pp. 3±23.
[2] Barron, A. E., Blanch, H. W., Soane, D. S., Electrophoresis
1994, 15, 597±615.
[3] Chiari, M., Riva, S., Gelain, A., Vitale, A., Turati, E., J. Chro-
matogr. A 1997, 781, 347±355.
[4] Rosenblum, B. B., Oaks, F., Menchen, S., Johnson, B.,
Nucleic Acids Res. 1997, 25, 3925±3929.
[5] Madabhushi, R. S., Electrophoresis 1998, 19, 224±230.
[6] Wenz, H. M., Robertson, J. M., Menchen, S. M., Oaks, F.,
Demorest, D. M., Scheibler, D., Rosenblum, B. B., Wike, C.,
Gilbert, D. A., Efcavitch, J. W., Genome Res. 1998, 8,
69±80.
[7] Waters, L. C., Jacobson, S. C., Kroutchinina, N., Khandur-
ina, J., Foote, R. S., Ramsey, J. M., Anal. Chem. 1998, 70,
5172±5176.
[8] Heller, C., Electrophoresis 1998, 19, 3114±3127.
[9] van der Schans, M. J., Kuypers, A. W. H. M., Kloosterman,
A. D., Janssen, H. J. T., Everaerts, F. M., J. Chromatogr. A
1997, 772, 255±264.
[10] Viovy, J.-L., Heller, C., in: Righetti, P. G. (Ed.), Capillary
Electrophoresis: An Analytical Tool in Biotechnology Series,
CRC Press, Boca Raton 1996, pp. 477±508.
[11] Viovy, J.-L., Duke, T., Electrophoresis 1993, 14, 322±329.
[12] Broseta, D., Leibler, L., Lapp, A., Strazielle, C., Europhys.
Lett. 1986, 2, 733±737.
[13] Cowie, J. M. G., Polymers: Chemistry and Physics of Mod-
ern Materials, Blackie Academic, London 1991.
[14] Kurata, M., Tsunashima, Y., in: Brandrup, J., Immergut, E.
H. (Eds.), Polymer Handbook, John Wiley and Sons, New
York 1989, pp. VII/1±VII/46.
[15] Thielking, H., Kulicke, M. W., Anal. Chem. 1996, 68,
1169±1173.
[16] Cottet, H., Gareil, P., Viovy, J.-L., Electrophoresis 1998, 19,
2151±2162.
[17] Heller, C., Electrophoresis 1998, 19, 1691±1698.
[18] Slater, G. W., Mayer, P., Drouin, G., Methods Enzymol
1996, 270, 272±295.
Electrophoresis 1999, 20, 1962±1977 Separation of double-stranded and single-stranded DNA
[19] Slater, G. W., in: Heller, C. (Ed.), Analysis of Nucleic Acids
by Capillary Electrophoresis, Vieweg, Wiesbaden 1997, pp.
24±66.
[20] Ogston, A. G., Trans. Faraday Soc. 1958, 54, 1754±1757.
[21] Morris, C. J. O. R., Morris, P., Biochem. J. 1971, 124,
517±528.
[22] Slater, G. W., Guo, H. L., Electrophoresis 1995, 16, 11±15.
[23] Duke, T., Viovy, J.-L., Semenov, A. N., Biopolymers 1994,
34, 239±247.
[24] Slater, G. W., Noolandi, J., Biopolymers 1986, 25, 431±454.
[25] Viovy, J.-L., Rev. Mod. Phys. 1999, in press.
[26] Heller, C., Duke, T., Viovy, J.-L., Biopolymers 1994, 34,
249±259.
[27] Slater, G. W., Rousseau, J., Noolandi, J., Turmel, C.,
Lalande, M., Biopolymers 1988, 27, 509±524.
[28] Mitnik, L., SalomØ, L., Viovy, J. L., Heller, C., J. Chromatogr.
A 1995, 710, 309±321.
[29] Carlsson, C., Larsson, A., Jonsson, M., in: Heller, C. (Ed.),
Analysis of Nucleic Acids by Capillary Electrophoresis,
Vieweg, Wiesbaden 1997, pp. 67±89.
[30] Hubert, S. J., Slater, G. W., Viovy, J.-L., Macromolecules
1995, 29, 1006±1009.
[31] Grossman, P. D., Soane, D. S., J. Chromatogr. 1991, 559,
257±266.
1977
[32] Heller, C., Viovy, J.-L., Biopolymers 1995, 35, 485±492.
[33] Heller, C., Beck, S., Nucleic Acids Res. 1992, 20,
2447±2452.
[34] Ferguson, K. A., Metabolism 1964, 13, 985±1002.
[35] Stellwagen, N. C., Gelfi, C., Righetti, P. G., Biopolymers
1997, 42, 687±703.
[36] Pluen, A., Tinland, B., Sturm, J., Weill, G., Electrophoresis
1998, 19, 1548±1559.
[37] Hagerman, P. J., Ann. Rev. Biophys. Biophys. Chem. 1988,
17, 265±268.
[38] Hagerman, P. J., Ramadevi, V. A., J. Mol. Biol. 1990, 212,
351±362.
[39] Tinland, B., Pluen, A., Sturm, J., Weill, G., Macromolecules
1997, 30, 5763±5765.
[40] Yan, J. Y., Best, N., Zhang, J. Z., Ren, H., Jiang, R., Hou,
J., Dovichi, N. J., Electrophoresis 1996, 17, 1037±1045.
[41] Ansorge, W., Barker, R., J. Biochem. Biophys. Methods
1984, 9, 33±47.
[42] Slater, G. W., Mayer, P., Drouin, G., Electrophoresis 1993,
14, 961±966.
[43] Stellwagen, N. C., Electrophoresis 1997, 18, 34±44.
[44] Deutsch, J. M., Science 1988, 240, 922±924.
[45] Heller, C., Electrophoresis 1999, 20, 1978±1986.