The Effect of Ionic Conditions on the Conformations

of Supercoiled DNA. II. Equilibrium Catenation

Valentin V. Rybenkov
1
, Alexander V. Vologodskii
2
and Nicholas R. Cozzarelli
1
*
1
Department of Molecular and
Cell Biology, University of
California, Berkeley, CA
94720, USA
2
Department of Chemistry
New York University
New York, NY 10003, USA
We studied the equilibrium formation of DNA catenanes to assess the
conformational properties of supercoiled DNA as a function of ionic con-
ditions and supercoiling density. Catenanes were formed by cyclizing lin-
ear DNA with long cohesive ends in the presence of supercoiled
molecules. The efciency of the catenation depends on the distance
between opposing segments of DNA in the interwound superhelix. The
fraction of cyclizing molecules that becomes topologically linked with the
supercoiled DNA is the product of the concentration of the supercoiled
DNA and a proportionality constant, B, that depends on conformations
of supercoiled DNA. In parallel with these experimental studies, we cal-
culated the values of B using Monte Carlo simulations of the equilibrium
distribution of DNA conformations. There were no adjustable parameters
in the calculations because all three parameters of the DNA model, bend-
ing and torsional elasticity of DNA and DNA effective diameter, specify-
ing intersegment interactions, were known from independent studies. We
found very good agreement between measured and simulated values of
B for all the ionic conditions and DNA superhelix densities studied; the
discrepancy was less than a factor of 2 over the 200-fold variation in B.
The value of B decreases nearly exponentially with increasing superheli-
city, this dependence being especially strong at low salt concentration.
The dependence of B on the concentration of NaCl, MgCl
2
, and spermi-
dine can be described with good accuracy in terms of changes of the
DNA effective diameter. We found no indication of superhelix collapse
under any ionic conditions studied. We discuss, in light of these results,
the biological importance of the effect of DNA supercoiling on the
unlinking of the products of DNA replication.
# 1997 Academic Press Limited
Keywords: catenanes; supercoiled DNA; DNA topology; DNA modeling;
chromosome unlinking *Corresponding author
Introduction
In the preceding paper (Rybenkov et al., 1997a) ,
we used the sedimentation of supercoiled DNA to
study its conformations. The sedimentation coef-
cient of supercoiled DNA depends primarily on
the overall size and shape of the molecule in sol-
ution and is rather insensitive to local features of
molecule conformations. The same is true for most
physical methods, such as measures of viscosity or
elastic light scattering which allow determination
of the radius of gyration. Although knowledge of
global conformational features such as the branch-
ing frequency of the superhelix is essential for our
understanding of DNA supercoiling, it is equally
important to know local conformational properties,
such as the average diameter of the superhelix and
the angular distribution between juxtaposed DNA
sites. The circular nature of supercoiled DNA pre-
sents a unique opportunity to study local confor-
mations. We can measure the formation of
catenanes, topological links, between supercoiled
DNA and cyclizing linear molecules as illustrated
in Figure 1. The efciency of catenation depends
on the space between the opposing segments of
DNA that form the interwound superhelix. We
measured catenation efciency to evaluate the in-
tersegment space.
Abbreviations used: OC, open circular; SC,
supercoiled; cryo EM, cryo electron microscope.
J. Mol. Biol. (1997) 267, 312323
00222836/97/12031212 $25.00/0/mb960877 # 1997 Academic Press Limited
DNA catenanes are well known in biology. They
were discovered in 1967 in Vinograd's laboratory
as naturally occurring linked dimers in the mito-
chondria of malignant cells (Clayton & Vinograd,
1967; Hudson & Vinograd, 1967). Catenanes were
subsequently shown to be intermediates in the
terminal stage of replication of circular DNA (Sun-
din & Varshavsky, 1980) that accumulate when a
type II topoisomerase is inhibited (Yanagida &
Sternglanz, 1990; Adams et al., 1992). They also re-
sult from recombination between directly repeated
sites (Wasserman & Cozzarelli, 1986) and maintain
the structure of the kinetoplast network of unicel-
lular parasites (Ryan et al., 1988; Chen et al., 1995).
The topology of catenated DNAs has illuminated
the mechanisms of the processes that generate
them (reviewed by Wasserman & Cozzarelli, 1986;
Stark & Boocock, 1995).
The rst study of equilibrium catenation of DNA
in solution was carried out in 1967 by Wang and
Schwartz, who measured the fraction of catenanes
formed between phage 186 and l DNA molecules
upon their cyclization (Wang & Schwartz, 1967). In
1975, Vologodskii, Lukashin and Frank-Kamenets-
kii developed a computational method for deter-
mining the probability of catenation based on
Monte Carlo simulations (Vologodskii et al., 1975).
They used the Alexander polynomial (Rolfsen,
1976) to describe the topology of each particular
conformation of two closed circular chains. Klenin
et al. (1988) later extended the treatment to include
the excluded volume of DNA. Although these
computations concerned only open circular mol-
ecules, we used in this study essentially the same
method to simulate the equilibrium catenation of
supercoiled molecules.
We compared the experimentally determined frac-
tions of catenanes with the results of computer
simulations. We tested thereby how well the local
properties of simulated conformations corre-
sponded to those of actual DNA. Conversely, this
allowed us to interpret the experimental results in
terms of DNA conformations. We found very good
agreement between the theoretical and experimen-
tal data over a broad range of superhelical density
and ionic conditions, including those promoting
DNA aggregation. We determined that the prob-
ability of catenation decreases nearly exponentially
with increasing superhelical density, s, and that
this dependence is especially strong at low ionic
strength. The presence of magnesium or spermi-
dine increases the probability of catenation; an ob-
servation consistent with the efcient screening of
DNA charge by polyvalent ions. Under no con-
ditions did we nd the sharp drop in the prob-
ability of catenation that would be expected for
collapsed superhelices. These results and those of
the accompanying paper demonstrate that we un-
derstand well many of the conformational proper-
ties of supercoiled DNA.
Results
Method of analysis
We calculated the probability, P, that a given open
circular DNA (OC) will be linked with supercoiled
molecules (SC) of concentration c (Figure 1). To cal-
culate P, we introduce the function p(R), the prob-
ability of a link for chains whose centers of mass
are separated by distance R. This probability must
be averaged over all possible conformations and
orientations of two chains. We used the Monte
Carlo method to calculate p(R). Typical distri-
butions of p(R) versus R for closed DNA with s = 0
and 0.05 are shown in Figure 2. From these de-
pendences, we can calculate P according to the
equation:
P =

o
0
p(R)4pR
2
c
N
A
M
dRY (1)
where the term:
4pR
2
dR c
N
A
M
is the probability of nding a supercoiled molecule
in the volume 4pR
2
dR, c is the weight concen-
tration of DNA, N
A
is Avogadro's number, and M
is the molecular weight of DNA. Equation (1) is
valid only if c is small enough that we can ignore
the formation of three or more linked molecules. It
also supposes that the concentration of open circu-
lar molecules is sufciently low that the probability
of links between them is negligible.
For comparison with experimental data, it is con-
venient to introduce the constant B:
B =
N
A
M

o
0
p(R)4pR
2
dR (2)
Figure 1. Probing conformational properties of super-
coiled DNA by catenation. The procedure for formation
of catenanes between supercoiled and cyclizing linear
molecules is diagrammed. The small circle and carat on
the end of linear DNA symbolize its cohesive ends. We
measure the fraction of cyclized DNA that is catenated;
i.e. the ratio of the upper path to both paths.
Catenation Probability in Supercoiled DNA 313
Now we can express P as:
P = Bc (3)
The value of B does not depend on DNA concen-
tration and reects only the properties of a particu-
lar circular DNA. It depends on DNA length,
superhelix density, and on ionic conditions, which
change the conformational properties of super-
coiled DNA. Therefore, we compared calculated
and measured values of B here. We used equation
(2) to compute the value of B and equation (3) to
measure it.
There is a great variety of different links (Rolfsen,
1976). However, our calculations showed that for
DNA lengths used in this study, the simplest link,
the singly linked catenane shown in Figure 1, com-
prised at least 90% of all catenanes formed. Exper-
imentally, at the low DNA concentrations used
(values of P were always less than 0.1), only the
singly linked catenane was observed. Thus only
this type of link was analyzed throughout the
study.
We studied the probability of catenation between
circular pAB4 DNA and phage P4 DNA. Linear P4
DNA was added in excess of pAB4 DNA in
solution and allowed to cyclize via its cohesive
ends. When the cyclization reaction was complete,
we measured the fraction of P4 DNA linked to
pAB4 DNA among all cyclized molecules, P, and
calculated the corresponding values of B from
equation (3).
Figure 3(a) shows a typical autoradiogram of a
Southern blot from an agarose gel after electro-
phoresis that resolved all topologically distinct
species of interest. The PhosphorImager scan of
lane 4 from the gel (Figure 3(b)) shows that the
amount of catenated DNA was readily detectable
and quantiable.
The role of monovalent ions
Figure 4 shows both the measured and simulated
values of B for catenane formation in solutions
with three different concentrations of NaCl (0.02,
0.2 and 3.0 M) as a function of the DNA superheli-
cal density, s. (Throughout this and the accompa-
nying paper, open symbols are used for
experimental data and closed symbols for calcu-
lated values.) The simulated and measured values
of B agree closely over the whole range of s and
NaCl concentrations studied, even though the
range of B values exceeds two orders of magni-
tude. At each salt concentration, the value of B de-
creases nearly exponentially with increasing
supercoiling. In a 0.2 M NaCl solution, B is 30
times lower at s = 0.06 than in open circular
DNA. Typical simulated conformations of OC:OC
and OC:SC catenanes shown in Figure 5 clearly il-
lustrate why the change in catenation probability is
so large: the dense conformations of supercoiled
DNA offer much fewer opportunities for catena-
tion. The fraction of catenanes in supercoiled DNA
is also strongly dependent on salt concentration.
At a s of 0.05, the value of B is 30 times lower in
a solution containing 0.02 M NaCl than in a sol-
ution containing 0.2 M NaCl, and 100-fold lower
than in a 3.0 M NaCl solution. We conclude that
conformations of supercoiled DNA change sub-
stantially in this range of sodium ion concen-
trations.
The nding that the probability of catenation de-
creases with decreasing NaCl concentration is
counterintuitive, but can be readily explained. The
probability of catenation depends both on the aver-
age distance between opposing segments of inter-
wound superhelix and on the effective diameter of
the double helix. Although the average superhelix
diameter increases as the salt concentration is re-
duced, the concomitant increase of the DNA effec-
tive diameter turns out to be more important.
Thus, the probability of catenation decreases with
the reduction of the salt concentration as illustrated
in Figure 6.
We determined whether catenation probability is
also sensitive to the overall size of the superhelix.
We studied by computer simulation the depen-
dence of B on the number of branches in super-
coiled DNA. As in the accompanying paper, we
separated simulated conformations of supercoiled
DNA (s = 0.05) according to the number of
Figure 2. The probability of catenation, p(R), of two cir-
cular DNAs as a function of distance, R, between their
centers of mass. Shown are the calculated catenation
probabilities between open circular DNA 10 kb in length
and open circular (*) or supercoiled (!), s = 0.05,
DNA 7 kb in length. The calculations were done for a
solution containing 0.2 M NaCl. There is an optimal R
for catenation rather than a simple fall off of the catena-
tion with distance between the chains. For each value of
R a fraction of conformations of two chains is forbidden
because of overlap of one chain with the other. Because
this fraction increases sharply as R diminishes, there is a
decrese of p(R) at values less than optimum.
314 Catenation Probability in Supercoiled DNA
superhelix branches. We found that the probability
of catenation (B) was the same, within the accuracy
of the calculations, for molecules with three, four,
ve and six branches, and so was independent of
the number of branches in the supercoiled DNA.
Because the global size of the supercoiled mol-
ecules depends on the number of branches, this re-
sult means that the value of B must depend
primarily on the local properties of the superhelix,
particularly on the superhelix diameter. This result
also suggests that the position of catenane crossing
is distributed along the superhelix. We directly
analyzed the location of the catenane crossing by
visual inspection of 100 simulated catenanes on
the computer monitor, and found that in 65% of
the cases the positions were distributed along the
superhelix and the rest were localized in the
apices. Similarly, electron microscopy studies of
catenanes comprised of two supercoiled DNA
molecules one half the length we used showed
that 37% of the links were in subapical position
(Levene et al., 1995). Therefore, the catenane links
show only limited preference for an apical lo-
cation.
Figure 3. Separation by gel electrophoresis of topologically different DNA species formed during cyclization of P4
DNA. The products of the cyclization reaction were resolved by agarose gel electrophoresis and detected by a
Southern blot with labelled P4 DNA. (a) Autoradiogram of a typical gel resolving unlinked P4 DNA molecules (OC)
from catenanes of open circular (OC:OC) or supercoiled (OC:SC) pAB4 DNA. The simplest knots (Kn3) and dimeric
unlinked open circles (2xOC) are also seen on the gel. Lane 1, nicked pAB4 DNA; lane 2, P4 DNA cyclized alone;
lane 3, P4 DNA cyclized in the presence of nicked pAB4 DNA; lane 4, P4 DNA cyclization in the presence of super-
coiled pAB4; lane 5, cyclization as in lane 4 and cut with EcoO109 restriction endonuclease which cleaves pAB4 DNA
but not P4 DNA. (b) Scan of lane 4 from (a).
Figure 4. Measured and simulated probabilities of cate-
nation as a function of supercoiling at several NaCl con-
centrations. The experimental values of B (open
symbols) are shown together with calculated results
(lled symbols) for NaCl concentrations of 0.02 M (!,
!), 0.2 M (*, *), and 3 M (^, ^). The solid lines are
freehand ts to the calculated data.
Catenation Probability in Supercoiled DNA 315
Effect of polyvalent ions
It is well known that many di- and trivalent cat-
ions affect DNA conformations much more
strongly than monovalent ions. Polyvalent ions are
present inside the cell and are required for the ac-
tivity of many enzymes that act on DNA. This
prompted us to study the inuence of Mg
2+
and
spermidine
3+
on the probability of catenation in
supercoiled DNA. Spermidine in the millimolar
range promotes DNA aggregation and conden-
sation (Gosule & Schellman, 1976; Bloomeld,
1991), and high concentrations of magnesium
change the overall electrostatic interaction between
DNA segments from repulsion to attraction (Shaw
& Wang, 1993). Under these ionic conditions, if
anywhere, one should expect the collapse of inter-
wound superhelix reported in the studies of super-
coiled DNA by cryoelectron microscopy (cryo EM)
(Adrian et al., 1990; Bednar et al., 1994).
We found instead that the value of B actually in-
creases with increasing magnesium concentration
at all superhelical densities studied (Figure 7(a)).
This increase occurred both in the presence of mag-
nesium ions alone and with MgCl
2
and NaCl to-
gether (Figure 7(b)). The same pattern was
observed with increasing spermidine concen-
tration. The probability of catenation increased
Figure 5. Stereo views of typical
simulated conformations of DNA
catenanes. Open circular DNA
10 kb in length is shown by the
blue line; the red line corresponds
to DNA 7 kb in length that is open
circular in (a) and supercoiled,
s = 0.05, in (b). The simulations
were done for a solution containing
0.2 M NaCl.
Figure 6. Diagram of sections of an interwound super-
helix at high and low concentrations of sodium ions.
Double-stranded DNA is shown in black and the area
of strong electrostatic repulsion, specied by the value
of the DNA effective diameter, is shown in gray. The
geometric and effective diameter of DNA are shown to
scale but with idealized regular conformations.
Although the axis-to-axis distance between opposing
DNA segments is nearly two times smaller in 0.2 M
NaCl (a) than in 0.01 M NaCl (b), the increase in DNA
effective diameter in (b) is much more important for
catenation. The ability of the superhelix to accommodate
another DNA strand from a different molecule, a prere-
quisite for catenane formation, is dramatically lower at
low salt.
316 Catenation Probability in Supercoiled DNA
with increasing spermidine concentration at all
superhelical densities studied, and the supercoiling
dependence was the greatest at the lowest polyca-
tion concentration (Figure 7(c)). At the highest
spermidine concentration tested, the trend contin-
ued even though a signicant fraction of multimo-
lecular catenanes indicative of DNA aggregation
were formed (data not shown).
Figures 4 and 7 show that the probability of cate-
nation depends strongly on ionic conditions. In the
DNA model used in the simulations, the depen-
dence of B on ionic conditions was specied com-
pletely by the DNA effective diameter, d. Because
of the very good agreement between simulated
and measured data for NaCl solutions, we con-
cluded that for these solutions the dependence is
Figure 7. The probability of the catenation between supercoiled and open circular DNA in solutions of polyvalent
ions. The values of B were measured experimentally for solutions containing 0.01 M (~), 0.03 M (&) and 0.1 M (^)
MgCl
2
alone in (a) and supplemented with 0.1 M NaCl in (b). The probability of catenation in the presence of 10 mM
sodium phosphate, 10 mM MgCl
2
and 0 (~), 1.5 (!), 3.0 (&) or 3.5 (^) mM spermidine is shown in (c). The broken
lines are freehand ts to the experimental data.
Catenation Probability in Supercoiled DNA 317
really due to the changes in electrostatic interaction
between DNA segments. We tested whether this
conclusion also held true for solutions containing
polyvalent ions which have more complex inter-
actions with DNA. In Figure 8, we plotted the va-
lues of B for supercoiled DNA (s of 0.05) for all
the ionic conditions used against the corresponding
values of d (Rybenkov et al., 1997b). All the data
points t a single smooth curve in close agreement
with the simulated values of the catenation prob-
ability. Note that the value of d varies greatly from
nearly 0 nm to 12 nm. These data clearly show that
this simple interpretation of the effect of ionic con-
ditions on the probability of catenation works sur-
prisingly well.
The data in Figure 8 also demonstrate that super-
coiled DNA does not collapse in solution under a
very wide range of ionic conditions. In the col-
lapsed form observed by cryo EM, there is no vis-
ible space between opposing strands in the DNA
superhelix (Adrian et al., 1990; Bednar et al., 1994).
Therefore, the probability of catenation should be
very low for a collapsed DNA. In contrast, we ob-
served an increase in the probability of catenation
at all superhelical densities with increasing so-
dium, magnesium, or spermidine concentration.
Effect of DNA length
We showed above that catenation of supercoiled
DNA occurs along the whole length of the super-
helix and displays only a limited preference to the
superhelix ends. The average number of branches
in the superhelix is proportional to the DNA
length. This implies that for DNA longer than 3 kb,
the probability of catenation should be directly
proportional to the total axial length of the super-
helix. Because we express the coefcient B in units
of weight rather than molar concentration, its
value should be independent of the length of
supercoiled DNA. We found, indeed, that both
measured and simulated values of B in supercoiled
3 kb DNA were exactly the same as for 7 kb DNA
in our standard ionic conditions, 0.2 M NaCl
(Figure 9). The measured values of B were also in-
dependent of DNA length for conditions which
were previously reported to promote collapse of
the superhelix, 0.1 M MgCl
2
(Figure 9).
These data also argue against the possibility that
loops at the ends of a collapsed interwound super-
helix could provide for catenation of supercoiled
DNA. When DNA is observed by cryo EM to be
collapsed there is hardly any branching both in
3 kb and 7 kb DNA molecules (Adrian et al., 1990;
Bednar et al., 1994; N. R. C. and M. Adrian, unpub-
lished data). For collapsed superhelices with just
two ends, the value of B should be inversely pro-
Figure 8. The probability of catenation of supercoiled
DNA as a function of DNA effective diameter. The
simulated data (s = 0.05) are shown with lled circles
(*) and open symbols corresponding to the measured
values of B in solutions containing: NaCl (*), MgCl
2
(!), MgCl
2
plus 0.1 M NaCl (~), 10 mM sodium phos-
phate, 10 mM MgCl
2
plus spermidine (^). The continu-
ous and broken lines correspond to the best t through
the measured and calculated data, respectively. The
values of the DNA effective diameter were calculated
from the probability of DNA knotting during cyclization
of linear DNA (Rybenkov et al., 1993, 1997b).
Figure 9. The probability of catenation for 3 kb and 7 kb
DNA. Small open symbols correspond to the values of
B found for 3 kb supercoiled DNA in solutions of 0.2 M
NaCl (*) or 0.1 M MgCl
2
(^). The corresponding data
for 7 kb supercoiled DNA are shown with large sym-
bols. The calculated values of B were the same for 3 kb
and 7 kb DNA. Therefore, the continuous curve rep-
resents the best t through both sets of calculated values
of B in 0.2 M NaCl; the broken curve is a polynomial t
to the experimental data for 0.1 M MgCl
2
for both 3 kb
and 7 kb DNA.
318 Catenation Probability in Supercoiled DNA
portional to DNA length. Yet, our data clearly
show that the probability of catenanes with a par-
ticular supercoiled DNA is instead proportional to
the superhelix length. This is inconsistent with the
collapsed unbranched superhelix and is the result
expected if the catenation position is distributed
along the axis of superhelix.
Catenation of two supercoiled molecules
Because of the important biological implications of
the substantial effect of supercoiling on the prob-
ability of catenation, we wished to determine the
magnitude of the effect when both DNA molecules
are supercoiled. It is very difcult to do such
measurements experimentally, but we were able to
calculate the effect. The very good agreement be-
tween simulated and measured results for cate-
nanes in which only one of the molecules was
supercoiled encouraged the view that the com-
puted results would be valid in this case as well.
The computed values of B for two supercoiled
DNA molecules 7 kb in length are presented in
Figure 10. We nd once again a nearly exponential
decrease in B with supercoiling except that the de-
crease is 400-fold for a s of 0.06 compared to
open circular DNA rather than 30-fold when only
one or two rings is supercoiled.
Discussion
Conformational properties of supercoiled DNA
are described surprisingly well in terms of the
simple model of the double helix
We studiedthe conformations of supercoiledDNA
by parallel experimental measurements and theor-
etical calculations of the equilibrium level of cate-
nanes under various ionic conditions. We found
using computer simulation that the probability of
catenation with supercoiled DNA depends on the
average diameter and on the axial length of the
superhelix rather than on the overall shape of a
superhelix or on the number of branches in the
supercoiled molecule. In contrast, the sedimen-
tation coefcient of supercoiled DNA studied in
the accompanying paper is sensitive primarily to
the overall size and shape of the molecule rather
than to the local organization of the superhelix.
The former is determined by the number of
branches as well as by random bends of the super-
helical axis. Therefore, these two studies comp-
lement each other. In both cases we found a very
good agreement between experimental and simu-
lated data. We conclude that the DNA model used
in the simulations predicts conformations of the
supercoiled DNA extremely well. Many other
properties of the supercoiled DNA which are hard
to measure directly can be calculated using simu-
lation (Vologodskii et al., 1992; Gebe et al., 1995;
Klenin et al., 1995; Vologodskii & Cozzarelli, 1996).
The agreement is especially important since there
were no adjustable parameters in our DNA model.
The only parameters of the model, DNA persist-
ence length, torsional rigidity and DNA effective
diameter, have all been determined independently.
Although the accuracy with which the values of
these parameters are known differs, the degree of
uncertainty in the values does not signicantly af-
fect the simulated DNA conformations (Vologods-
kii et al., 1992). The DNA model used in the
simulations provides a reliable quantitative de-
scription of the properties of linear and open circu-
lar DNA (for reviews and refs see Hagerman, 1988;
Vologodskii, 1992). However, it was not clear until
now how well the model describes the confor-
mational properties of supercoiled DNA; a form
that is characterized by frequent close approaches
between DNA segments. Because these close ap-
proaches are rare in linear and open circular DNA
molecules, the accurate description of intersegment
interaction is not so crucial for the prediction of
most of their properties. The agreement between
simulated and measured results obtained in this
and the accompanying paper shows that our un-
derstanding of the intersegment interaction over a
broad range of ionic conditions is, in fact, quite
good.
The DNAeffective diameter, which species inter-
segment interaction, was the only parameter
whose value was changed in the simulation of
different ionic conditions. Over a 200-fold change
in the value of B, the discrepancy between the cal-
culated and experimental data is less than twofold
(Figure 8). The agreement between simulated and
measured sedimentation rates was similar over a
very broad range of ionic conditions (see Figure 6
in the accompanying paper), implying that with re-
gard to the conformational properties of super-
coiled DNA, the effect of ionic conditions is well
Figure 10. Calculated equilibriumconstant of catenation
of two supercoiled DNAs. The values correspond to
7 kb molecules in a solution of 0.2 M NaCl.
Catenation Probability in Supercoiled DNA 319
approximated by the value of DNA effective diam-
eter alone. It does not depend on the type or va-
lence of the counterions. The different ions,
however, have to be present at very different con-
centrations to yield a given value of DNA effective
diameter (Rybenkov et al., 1997b); for example, an
effective diameter of 5 nm is achieved at 0.2 M
NaCl or at 10 mM MgCl
2
.
Since DNA is a rigid polymer molecule, the inter-
segment interactions are only minor determinants
for many properties of linear and open circular
DNA, for which the probability of intersegment
collisions is very low (Tanford, 1967). In such
cases, changes of DNA persistence length, a, with
ionic conditions may be more signicant. For the
range of ionic conditions used in the current work,
the value of a does not change by more than 20%
from 50 nm (Hagerman, 1988). Because we could
not specify these changes with accuracy, we con-
sidered the value of a to be a constant. Although
we neglected this dependence of a on ionic
strength, it could conceivably contribute to the 1.8-
fold discrepancy between the theoretical and ex-
perimental values of B for open circular DNA
found for solutions containing MgCl
2
(compare
Figures 4 and 7). The change of the DNA persist-
ence length can also change the values of the DNA
effective diameter as derived from knotting prob-
ability which in turn would slightly change the
slope of the experimental dependence of B for
supercoiled DNA on d (shown in Figure 8). How-
ever, these effects are rather small, particularly for
supercoiled DNA.
Our results showthat there is no collapse of the in-
terwound superhelix at ionic conditions even re-
motely close to physiological conditions. Although
the low sedimentation rate in 100 mM MgCl
2
or
3.5 mM spermidine obtained in the accompanying
paper might have been interpreted as indicating a
collapse at these extreme ionic conditions, the re-
sults of this paper argue persuasively against it. In-
deed, even for these extreme ionic conditions,
under which d is less than the DNA geometric di-
ameter, the probability of catenation follows the
same monotonic dependence on d (see Figure 8). In
contrast, if collapse of the interwound superhelix
occurred, the value of B would certainly be
abruptly reduced. This conclusion is consistent
with the previous studies (Rau et al., 1984; Ma &
Bloomeld, 1994) which found no indication of
DNA aggregation at any MgCl
2
concentration
tested.
Physiological implications
The nding that the probability of catenation is
strongly dependent on DNA superhelical density
reveals what is likely to be an important physio-
logical role of DNA supercoiling. In order for
chromosomes to segregate during cell division, the
linking number of daughter DNAs must be equal
to zero. Because the intracellular DNA concen-
tration is very high (Kellenberger, 1987; Bohrmann
et al., 1993), relaxed daughter molecules would in-
stead be expected to be highly linked at topological
equilibrium. For example, for a 7 kb plasmid with
an intracellular concentration of about 0.2 mg/ml,
the probability of catenation calculated from
equation (3) equals 0.2. However, when both
DNAs are supercoiled, the reduction in the equili-
brium catenation probability is a factor of 400
(Figure 10), thereby reducing the probability of ca-
tenation in our example to 0.0005. Although the
situation becomes more complicated for chromoso-
mal DNA, it is clear that supercoiling will contrib-
ute greatly to the unlinking of replicated
chromosomes as well (Ullsperger et al., 1995).
Summary of conformational properties of
supercoiled DNA
This and the accompanying paper represent the
culmination of our experimental and theoretical
analyses of the conformations of supercoiled DNA.
We summarize our conclusions as follows.
(1) At all ionic conditions studied, the superhelix
has the interwound form.
(2) The superhelix conformations are often
branched. The number of branches is proportional
to DNA length for molecules longer than 3 kb, and
at near physiological ionic conditions and at s
equal to 0.05, there is about one branch per
1.7 kb DNA. Branching frequency decreases with
increasing superhelicity and salt concentration (for
details see Vologodskii et al., 1992).
(3) The conformations of supercoiled DNAdepend
strongly on ionic conditions. Screening of the elec-
trostatic repulsion between DNA segments in the
tight interwound structure by counterions is re-
sponsible for this dependence. At all ionic con-
ditions tested, including NaCl, MgCl
2
and
spermidine solutions, the effect of ionic conditions
on supercoiled DNA can be specied to a very
good approximation in terms of the DNA effective
diameter.
(4) At nearly physiological ionic conditions, Wr/
Tw equals 3 for DNA 52.5 kb and is indepen-
dent of s. This ratio decreases with decreasing salt
concentration and becomes dependent on s. At
0.01 M NaCl, Wr/Tw decreases from 1.6 at
s = 0.01 to 0.8 at s = 0.06 (Vologodskii et al.,
1992; Vologodskii & Cozzarelli, 1994; Rybenkov
et al., 1997b).
(5) The average superhelix winding angle is about
60

and does not depend on s or on ionic con-

ditions. Therefore, the length of the superhelix is
always about 40% of the DNA length, but the aver-
age superhelix diameter decreases rapidly with s
(Vologodskii & Cozzarelli, 1994). At s = 0.05,
the average superhelix diameter (i.e. the distance
between the axes of the opposing DNA segments
in the regular superhelix) increases from 10 nm at
0.2 M NaCl, to 17 nm at 0.01 M NaCl (Table 2 of
accompanying paper). At nearly physiological
ionic conditions and s value about 0.05, the
320 Catenation Probability in Supercoiled DNA
superhelix is extremely long and narrow: for a 7 kb
molecule, the axial ratio is about 100.
Materials and Methods
DNA
The 7.0 kb plasmid, pAB4 (Wasserman et al., 1988), and
the 2.95 kb plasmid, pUCd2, were puried by the Triton
lysis method (Ausubel et al., 1989). pUCd2 was derived
from pUC18. Samples with different superhelical den-
sities were prepared as described (Rybenkov et al.,
1997a) and their superhelical density measured by band
counting (Keller, 1975). The 10.0 kb bacteriophage
P4del22 DNA was puried as described (Rybenkov et al.,
1993). Singly nicked circular DNA was obtained by lim-
ited digestion with DNase I in the presence of ethidium
bromide (Barzilai, 1973).
DNA cyclization
P4 DNA with 19 bp cohesive ends was linearized by
heating for 15 minutes at 65

C. It was then rapidly

cooled to 0

C, and diluted to 1 mg/ml into a mixture of

supercoiled DNA at 25 to 500 mg/ml, depending on salt
conditions and superhelical density. Solutions contained
the indicated amounts of NaCl, MgCl
2
, and spermi-
dine 3HCl as well as 0.5 mM sodium phosphate (pH 7.5)
as a buffer and 0.2 mM EDTA. The cyclization reaction
was carried out at room temperature for 4 to 20 hours
depending on salt conditions, after which the reaction
was terminated by addition of an oligonucleotide comp-
lementary to one of the cohesive ends (Rybenkov et al.,
1993). After an additional one hour incubation, the mix-
ture was analyzed by gel electrophoresis through a 1.0%
(w/v) agarose gel containing 0.1% (w/v) SDS and TAE
buffer (Sundin & Varshavsky, 1980) at 1.2 V/cm for 84
hours. The DNA was detected by Southern blot hybrid-
ization with
32
P-labelled P4 DNA as a probe and quanti-
ed using a PhosphorImager (Molecular Dynamics).
During cyclization some fraction of the supercoiled DNA
was nicked. Since the fraction of catenanes is pro-
portional to the concentration of circular DNA, this
could lead to an underestimate of the measured prob-
ability. Therefore, we determined the nicked fraction pre-
sent at the end of the cyclization reaction using agarose
gel electrophoresis and probing with pAB4 DNA, and
corrected for it.
Another correction took into account the nite width of
the topoisomer distribution of our DNA samples (see the
accompanying paper). For each salt condition, the
measured dependencies of B on s (Figures 4 and 7) were
tted with a function B = B
0
exp(bs), the correction fac-
tor was calculated as:
B(s))
B(s))
= exp

1
2
b
2
(ds)
2
)

Y
where (ds)
2
) is the variance of the topoisomer distri-
bution. We corrected all measured values of B for super-
coiled DNA by a corresponding factor. The value of this
factor was always less than 1.3.
To avoid formation of catenanes composed of several
molecules, we used relatively low concentration of both
DNAs so that the probability of catenation did not ex-
ceed 0.1. We also tested the validity of equation (3) by
repeating the measurements with one half the concen-
tration of supercoiled DNA and obtained in each case
the same value of B within 5%. However, because of the
uncertainty in the baseline position on a gel scan
(Figure 3(b)) the error in the measured probability of ca-
tenanes could be up to threefold higher.
At the ionic conditions tested, the cyclization reaction
was at least ve orders of magnitude slower than the
conformational relaxation of 10 kb DNA (Perkins et al.,
1994). Therefore, in our experiments DNA cyclization re-
sulted in an equilibrium distribution of topological
species. We also tested directly the premise that the frac-
tions of catenanes did not depend on the cyclization rate
by comparing the fractions of catenanes produced upon
cyclization at 23

C and 37

C. The rates of cyclization,

and thus the time to reach equilibrium at these two tem-
peratures, differ more than threefold (Wang & Davidson,
1966). We found, however, no difference in the measured
probability of catenation (data not shown).
Monte Carlo simulation of catenation probability
We used equation (2) to calculate the values of B for
given conditions. This equation reduces the problem to
calculating the equilibrium probability, p(R), of the link-
ing of two circular chains whose centers of masses are
separated by the distance R. To obtain p(R), we used the
following steps. First, we constructed the equilibrium
sets of conformations of closed chains using the Metro-
polis procedure (Rybenkov et al., 1997a). One set corre-
sponded to supercoiled or open circular DNA 7 kb in
length; the other set corresponded to open circles 10 kb
in length. Then we placed all possible pairs of chain con-
formations from the two sets at a distance R between
their centers of mass. During this placement, the seg-
ments of one chain could freely cross the segments of the
other chain. Any conformation in which the chain seg-
ments overlapped was forbidden. We then determined
the topological state of the allowed chains by calculating
the value of the Alexander polynomial, (s, t), at s = 1
and t = 1, which has different values for the simplest
links and for unlinked curves (Rolfsen, 1976). Since this
analysis of chain topology was proposed 20 years ago
(Vologodskii et al., 1975), several methodical improve-
ments have been introduced in the algorithms for the
polynomial calculation (reviewed by Frank-Kamenetskii
& Vologodskii, 1981; Michels & Wiegel, 1986). For each
pair of chain conformations, we tested their topological
states for 144 mutual orientations obtained by all poss-
ible 90

rotations of each molecule around the X, Y, and

Z directions p(R) was calculated as the number of linked
conformations divided by the total number of tested con-
formations of the two chains including both allowed and
forbidden conformations. We usually tested 625 different
pairs of chain conformations for each distance R. This
provided a statistical error of computations of B of less
than 15%.
Acknowledgements
This work was supported by NIH grant GM31657 to N.
R. C. and NIH grant GM54215 to A. V. V.
References
Adams, D. E., Shekhtman, E. M., Zechiedrich, E. L.,
Schmid, M. B. & Cozzarelli, N. R. (1992). The role
of topoisomerase IV in partitioning bacterial repli-
Catenation Probability in Supercoiled DNA 321
cons and the structure of catenated intermediates in
DNA replication. Cell, 71, 277288.
Adrian, M., ten Heggeler-Bordier, B., Wahli, W., Stasiak,
A. Z., Stasiak, A. & Dubochet, J. (1990). Direct
visualization of supercoiled DNA molecules in
solution. EMBO J. 9, 45514554.
Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D.,
Seidman, J. G., Smith, J. A. & Struhl, K. (1989). Cur-
rent Protocols in Molecular Biology, Wiley-Inter-
science, New York.
Barzilai, R. (1973). SV40 DNA: quantitative conversion
of closed circular to open circular form by an ethi-
dium bromide-restricted endonuclease. J. Mol. Biol.
74, 739742.
Bednar, J., Furrer, P., Stasiak, A., Dubochet, J., Egelman,
E. H. & Bates, A. D. (1994). The twist, writhe and
overall shape of supercoiled DNA change during
counterion-induced transition from a loosely to a
tightly interwound superhelix. J. Mol. Biol. 235,
825847.
Bloomeld, V. A. (1991). Condensation of DNA by
multivalent cations: considerations on mechanism.
Biopolymers, 31, 14711481.
Bohrmann, B., Haider, M. & Kellenberger, E. (1993).
Concentration evaluation of chromatin in unstained
resin-embedded sections by means of low-dose
ratio-contrast imaging in STEM. Ultramicroscopy, 49,
235251.
Chen, J., Rauch, C. A., White, J. H., Englund, P. T. &
Cozzarelli, N. R. (1995). The topology of the kineto-
plast DNA network. Cell, 80, 6169.
Clayton, D. A. & Vinograd, J. (1967). Circular dimer and
catenate forms of mitochondiral DNA in human
leukaemic leucocytes. Nature, 216, 652656.
Frank-Kamenetskii, M. D. & Vologodskii, A. V. (1981).
Topological aspects of the physics of polymers: the
theory and its biophysical applications. Sov. Phys.-
Usp. 24, 679696.
Gebe, J. A., Allison, S. A., Clendenning, J. B. & Schurr, J.
M. (1995). Monte Carlo simulations of supercoiling
free energies for unknotted and trefoil knotted
DNAs. Biophys. J. 68, 619633.
Gosule, L. C. & Schellman, J. A. (1976). Compact form
of DNA induced by spermidine. Nature, 259, 333
335.
Hagerman, P. J. (1988). Flexibility of DNA. Annu. Rev.
Biophys. Biophys. Chem. 17, 265286.
Hudson, B. & Vinograd, J. (1967). Catenated circular
DNA molecules in HeLa cell mitochondria. Nature,
216, 647652.
Kellenberger, E. (1987). The compactness of cellular plas-
mas; in particular, chromatin compactness in re-
lation to function. Trends Biochem. Sci. 12, 105107.
Keller, W. (1975). Determination of the number of super-
helical turns in simian virus 40 DNA by gel
electrophoresis. Proc. Natl Acad. Sci. USA, 72, 4876
4880.
Klenin, K. V., Vologodskii, A. V., Anshelevich, V. V.,
Dykhne, A. M. & Frank-Kamenetskii, M. D. (1988).
Effect of excluded volume on topological properties
of circular DNA. J. Biomol. Struct. Dynam. 5, 1173
1185.
Klenin, K. V., Frank-Kamenetskii, M. D. & Langowski, J.
(1995). Modulation of intramolecular interactions in
superhelical DNA by curved sequences-a Monte
Carlo simulation study. Biophys. J. 68, 8188.
Levene, S. D., Donahue, C., Boles, T. C. & Cozzarelli, N.
R. (1995). Analysis of the structure of dimeric DNA
catenanes by electron microscopy. Biophys. J. 69,
10361045.
Ma, C. & Bloomeld, V. A. (1994). Condensation of
supercoiled DNA induced by MnCl
2
. Biophys. J. 67,
16781681.
Michels, J. P. J. & Wiegel, F. W. (1986). On the topology
of a polymer ring. Proc. Roy. Soc. London ser. A. 403,
269284.
Perkins, T. T., Quake, S. R., Smith, D. E. & Chu, S.
(1994). Relaxation of a single DNA molecule
observed by optical microscopy. Science, 264, 822
826.
Rau, D. C., Lee, B. & Parsegian, V. A. (1984). Measure-
ment of the repulsive force between polyelectrolyte
molecules in ionic solution: Hydration forces
between parallel DNA double helices. Proc. Natl
Acad. Sci. USA, 81, 26212625.
Rolfsen, D. (1976). Knots and Links, Publish or Perish,
Inc., Berkeley, CA.
Ryan, K. A., Shapiro, T. A., Rauch, C. A. & Englund, P.
T. (1988). Replication of kinetoplast DNA in
trypanosomes. Annu. Rev. Microbiol. 42, 339358.
Rybenkov, V. V., Cozzarelli, N. R. & Vologodskii, A. V.
(1993). Probability of DNA knotting and the effec-
tive diameter of the DNA double helix. Proc. Natl
Acad. Sci. USA, 90, 53075311.
Rybenkov, V. V., Vologodskii, A. V. & Cozzarelli, N. R.
(1997a). The effect of ionic conditions on the confor-
mations of supercoiled DNA. I. Sedimentation
analysis. J. Mol. Biol. 267, 299311.
Rybenkov, V. V., Vologodskii, A. V. & Cozzarelli, N. R.
(1997b). The effect of ionic conditions on DNA heli-
cal repeat, effective diameter, and free energy of
supercoiling. Nucl. Acids Res. In the press.
Shaw, S. Y. & Wang, J. C. (1993). Knotting of a DNA
chain during ring closure. Science, 260, 533536.
Stark, W. M. & Boocock, M. R. (1995). Topological selec-
tivity in site-specic recombination. In Mobile Gen-
etic Elements (Sherratt, D. J., ed.), pp. 101129, ARL
Press, Oxford Univ. Press, Oxford.
Sundin, O. & Varshavsky, A. (1980). Terminal stages of
SV40 DNA replication proceed via multiply inter-
twined catenated dimers. Cell, 21, 103114.
Tanford, C. (1967). Physical Chemistry of Macromolecules,
Wiley, New York.
Ullsperger, C. J., Vologodskii, A. V. & Cozzarelli, N. R.
(1995). Unlinking of DNA by topoisomerases
during DNA replication. Nucl. Acids Mol. Biol. 9,
115142.
Vologodskii, A. V. (1992). Topology and Physics of Circular
DNA, CRC Press, Boca Roton.
Vologodskii, A. V. & Cozzarelli, N. R. (1994). Confor-
mational and thermodynamic properties of super-
coiled DNA. Annu. Rev. Biophys. Biomol. Struct. 23,
609643.
Vologodskii, A. V. & Cozzarelli, N. R. (1996). The effect
of supercoiling on the juxtaposition and relative
orientation of DNA sites. Biophys. J. 70, 25482556.
Vologodskii, A. V., Lukashin, A. V. & Frank-
Kamenetskii, M. D. (1975). Topological interaction
between polymer chains. Sov. Phys. JETP, 40, 932
936.
Vologodskii, A. V., Levene, S. D., Klenin, K. V., Frank-
Kamenetskii, M. D. & Cozzarelli, N. R. (1992). Con-
formational and thermodynamic properties of
supercoiled DNA. J. Mol. Biol. 227, 12241243.
Wang, J. C. & Davidson, N. (1966). Thermodynamic and
kinetic studies on the interconversion between the
322 Catenation Probability in Supercoiled DNA
linear and circular forms of phage lambda DNA.
J. Mol. Biol. 15, 111123.
Wang, J. C. & Schwartz, H. (1967). Noncomplementarity
in base sequences between the cohesive ends of
coliphages 186 and lambda and the formation of
interlocked rings between the two DNAs. Biopoly-
mers, 5, 953966.
Wasserman, S. A. & Cozzarelli, N. R. (1986). Biochemi-
cal topology: applications to DNA recombination
and replication. Science, 232, 951960.
Wasserman, S. A., White, J. H. & Cozzarelli, N. R.
(1988). The helical repeat of double-stranded DNA
varies as a function of catenation and supercoiling.
Nature, 334, 448450.
Yanagida, M. & Sternglanz, R. (1990). Genetics of DNA
topoisomerases. In DNA Topology and its Biological
Rffects (Cozzarelli, N. R. & Wang, J. C., eds),
pp. 299320, Cold Spring Harbor Laboratory Press,
Cold Spring Harbor, New York.
Edited by D. E. Draper
(Received 15 August 1996; received in revised form 12 December 1996; accepted 12 December 1996)
Catenation Probability in Supercoiled DNA 323