DNA Supercoiling and Topology

DNA topology is not a historical curiosity — it is load-bearing machinery in modern molecular medicine and biotechnology. Fluoroquinolone antibiotics (ciprofloxacin, levofloxacin) and the earlier quinolones work by trapping the DNA gyrase–DNA cleavage complex, converting the enzyme that normally relieves positive supercoiling ahead of replication forks into a DNA-damaging poison — a direct pharmacological exploitation of bacterial topological regulation, and one of the most widely prescribed antibiotic classes precisely because gyrase has no close human counterpart in its ATP-dependent negative-supercoiling activity. Human topoisomerase II is similarly targeted by anticancer drugs (etoposide, doxorubicin), which trap the human type II enzyme's cleavage complex in rapidly dividing tumor cells, again converting a topology-resolving enzyme into a lethal DNA-damaging agent when inhibited mid-mechanism. Plasmid topology is a routine quality-control readout in biotechnology and gene therapy manufacturing: supercoiled plasmid DNA (the fully negatively supercoiled, most compact topoisomer) transfects cells and is taken up by bacteria far more efficiently than nicked or linear forms of the identical sequence, so agarose gel electrophoresis separating topoisomers by their differing $Lk$-dependent compaction is a standard release assay for plasmid-based vaccines and gene therapy vectors (including several COVID-19 DNA vaccine candidates). In structural biology, cryo-electron microscopy and single-molecule magnetic tweezers experiments now directly visualize and manipulate individual supercoiled DNA molecules, applying controlled torque to measure the twist–writhe buckling transition in real time and to watch individual topoisomerase strand-passage events change $Lk$ by exactly $\pm 1$ or $\pm 2$ per catalytic cycle, depending on enzyme type — a direct single-molecule confirmation of the integer, discrete nature of topological linking number changes predicted by the theory above. In genome organization more broadly, the loop-domain architecture of eukaryotic chromatin (topologically associating domains, TADs) means each loop is effectively a topologically semi-constrained unit, and supercoiling generated by transcription within a TAD is now understood to influence long-range enhancer–promoter contacts, linking DNA topology directly to the three-dimensional regulation of gene expression, a frontier actively connected to the enzymatic control exercised by topoisomerases.

Ask the tutor: How does a type II topoisomerase's strand-passage mechanism actually change the linking number by exactly two, step by step?

Consider a segment of B-DNA drawn as two backbones spiraling around a common axis, exactly as in the double helix, but now imagine both ends of both strands joined so the molecule closes into a circle — no free ends anywhere. This is not an exotic case: bacterial chromosomes, most plasmids, many viral genomes (SV40, papillomaviruses), and mitochondrial DNA are literally covalently closed circles, and even linear eukaryotic chromatin is organized into large loops anchored at both ends to a protein scaffold, which are topologically closed for practical purposes over the timescale of a cell cycle. Once the ends are sealed, the two strands are linked like two rings of a chain — not merely wound around each other locally, but linked in the rigorous topological sense that you cannot pull them apart without cutting one. The number of times strand 2 passes through the "disk" bounded by strand 1 (with a sign convention for direction) is the linking number, $Lk$, and it is a topological invariant: continuous bending, twisting, and deformation of the DNA in space cannot change it, only strand breakage (nicking) can. Here is the crucial subtlety that the rest of this article exists to unpack: $Lk$ can be partitioned between two geometrically distinct quantities. Twist ($Tw$) counts the local helical winding of the two strands about each other along the contour — essentially how many times the base pairs spiral around the helix axis, the same quantity familiar from a relaxed linear helix. Writhe ($Wr$) counts the coiling and crossing of the helix axis itself in three-dimensional space — whether the axis stays roughly planar or loops over itself like a twisted telephone cord. Twist is local and helical; writhe is global and axial. Crucially, only their sum is topologically fixed. If you take relaxed circular DNA (all $Lk_0$ turns accounted for as pure twist, $Wr = 0$, lying flat like a ring) and reduce $Lk$ by breaking one strand, unwinding, and resealing, the DNA cannot simply reduce its twist and stay planar — energetically it prefers to keep twist near the natural value (about 10.5 bp per turn, the value stabilized by base stacking and the sugar-phosphate backbone geometry) and instead throws the deficit into negative writhe: the axis coils upon itself in an interwound, plectonemic superhelix. This is exactly the rubber-tube experiment. The molecule is now negatively supercoiled.

Positive supercoiling is the mirror image: over-twisting a closed circle ($\Delta Lk > 0$) also cannot be absorbed as pure extra twist without energetic cost, so the excess again partitions into writhe — but of the opposite handedness, producing right-handed superhelical crossings rather than the left-handed crossings of negative supercoiling. Positive supercoils resist strand separation (they tighten the helix), which is exactly the topological obstacle that a replication fork or transcribing RNA polymerase generates ahead of itself as it unwinds the duplex — the "twin supercoiled domain" problem. Biologically, nearly all genomic DNA in bacteria, and large domains of eukaryotic chromatin, is maintained negatively supercoiled relative to relaxed B-DNA, at a superhelical density $\sigma \approx -0.05$ to $-0.07$. This underwinding is not accidental: it is actively imposed by DNA gyrase (a type II topoisomerase that pumps negative supercoils in using ATP) and partially constrained by nucleosome wrapping and other DNA-binding proteins. The stored torsional free energy lowers the energetic cost of the strand separation required for replication origin firing, promoter melting by RNA polymerase, and recombination — supercoiling is, in effect, a pre-paid energetic deposit toward opening the double helix, precisely the process whose thermodynamics (base-pair opening, stacking energy) is treated in the physical chemistry of helix stability.

The entire quantitative structure of DNA topology rests on three relations. First, the definition of the linking number for a relaxed circle:

where: $N$ is the number of base pairs in the closed duplex, $h$ is the helical repeat (base pairs per turn, $h \approx 10.5$ for relaxed B-DNA in solution), and $Lk_0$ is the linking number of the fully relaxed, torsionally unstressed circle (rounded to the nearest integer, since $Lk$ must be an integer). Second, the White–Fuller partition theorem, the load-bearing identity of the whole subject:

where: $Lk$ is the (integer, topologically fixed) linking number, $Tw$ is the twist — the total helical winding of the two strands about the local axis, integrated along the contour — and $Wr$ is the writhe — the coiling of the helix axis in three-dimensional space, a purely geometric quantity that need not be an integer and can change continuously without breaking any bond. Third, the definitions used to quantify biological supercoiling:

where: $\Delta Lk$ is the linking number difference (negative for underwound, positive for overwound DNA), and $\sigma$ (superhelical density) is $\Delta Lk$ normalized by the relaxed linking number, making it comparable across plasmids of different sizes. Typical bacterial values are $\sigma \approx -0.06$. Finally, the elastic (harmonic) free energy of supercoiling, an empirical but well-calibrated relation from torsional elasticity theory:

where: $R$ is the gas constant, $T$ absolute temperature, $N$ the number of base pairs, and the numerical prefactor (about $1050\, RT$, often written using a torsional rigidity constant $K \approx 1100$–$3000\ RT$ depending on convention and salt conditions) is calibrated from measurements of supercoiling free energy in real plasmids; this article uses the widely cited Depew–Wang/Bauer value of $\sim 10\, RT$ per superhelical turn scaled appropriately, giving the $1050\,RT/N$ form per $(\Delta Lk)^2$. The key qualitative content — free energy rising quadratically with the square of the linking difference and falling inversely with molecule size — is robust across conventions. A reading of these four equations together: $Lk_0$ tells you where "zero" is; $Lk$ is the fixed number the cell (or the topoisomerase) has actually set; $\Delta Lk$ and $\sigma$ measure the departure from zero in absolute and normalized units; and $\Delta G_{sc}$ tells you the energetic price (or subsidy) of that departure — a price that grows quadratically, so doubling the supercoiling deficit quadruples the stored strain energy, and a price that is diluted across a longer molecule, so a small plasmid is far more torsionally stressed for the same $\sigma$ than a whole bacterial chromosome would be per unit length.

Example 1 — Foundational: linking number and superhelical density of a small plasmid. A circular plasmid has $N = 4000$ bp. Take the relaxed helical repeat as $h = 10.5$ bp/turn (the standard solution value for B-DNA). Step 1: compute the relaxed linking number.

Since $Lk$ must be an integer, round to $Lk_0 = 381$ (this rounding is itself a real physical statement: a truly relaxed closed circle of exactly 4000 bp cannot have a non-integer linking number, so the last fractional turn is accommodated as a small residual twist strain even in the "relaxed" reference state — a subtlety usually neglected at this level of precision). Step 2: suppose topoisomerase I has relaxed the plasmid to $Lk = 375$ in vivo — i.e., six fewer links than the relaxed value, consistent with typical bacterial underwinding. Compute $\Delta Lk$ and $\sigma$.

This $\sigma$ is on the low end of physiological (typical values are $-0.05$ to $-0.07$); it illustrates the arithmetic cleanly with modest, easy-to-follow numbers before the next example uses a more realistic value. Step 3: partition into twist and writhe under the simplifying (small-$|\sigma|$) assumption that essentially all of the deficit remains as $Tw$ change with negligible $Wr$ (appropriate here since $|\sigma|$ is small):

This is the "underwound but still nearly planar" regime described in Step 2 of the second derivation.

Limiting-case check: if $\Delta Lk \to 0$ (fully relaxed), $\sigma \to 0$ and the free energy expression $\Delta G_{sc} = 1050\,RT(\Delta Lk)^2/N \to 0$, correctly recovering the unstressed, relaxed circle with no stored torsional energy — as it must, since a relaxed closed circle is by definition the reference (zero-strain) state.

Example 2 — Applied: free energy of physiological supercoiling in a bacterial plasmid, and comparison to a genomic domain. Take a plasmid with $N = 4000$ bp and physiological superhelical density $\sigma = -0.06$ (a realistic in vivo value). Step 1: compute $Lk_0$ and $\Delta Lk$.

Step 2: compute the elastic supercoiling free energy using $\Delta G_{sc} = 1050\,RT(\Delta Lk)^2/N$.

At $T = 310\ \text{K}$ (body temperature, $RT \approx 2.58\ \text{kJ/mol}$), this is

per mole of plasmid molecules — a substantial stored free energy, entirely available (in principle) to subsidize strand separation events such as origin melting, distributed over the whole 4000 bp molecule. Step 3: compare to a topologically isolated domain ten times larger, $N' = 40{,}000$ bp, held at the same superhelical density $\sigma = -0.06$ (same fractional underwinding).

Note that $\Delta G_{sc}'/\Delta G_{sc} = 1376.6/138.9 \approx 9.91 \approx 10$: at fixed $\sigma$, the total stored free energy scales linearly with $N$ (since $\Delta Lk \propto N$ at fixed $\sigma$, so $(\Delta Lk)^2/N \propto N$), even though the free energy per base pair is identical between the two domains. This clarifies the earlier qualitative claim: a smaller domain forced to the same absolute $|\Delta Lk|$ (not the same $\sigma$) is far more strained per base pair — the scaling direction depends on which quantity ($\sigma$ or $\Delta Lk$) is held fixed, and both limits are biologically relevant (topoisomerases regulate toward a target $\sigma$, while a stalled polymerase generates a roughly fixed $\Delta Lk$ regardless of domain size).

Limiting-case check: as $N \to \infty$ at fixed $\Delta Lk$ (a fixed absolute linking number perturbation spread over an ever-larger domain), $\Delta G_{sc} \to 0$ — correctly capturing the intuition that the same local topological "insult" produces vanishing torsional stress if the domain into which it can relax is unboundedly large, exactly as diluting a fixed twist deficit over more base pairs reduces the strain per base pair to zero.

Example 3 — Exam/competition level: resolving Tw and Wr from an experimental linking number distribution. A population of a 3000 bp plasmid is nicked, relaxed by topoisomerase I under conditions where the effective helical repeat is $h = 10.4$ bp/turn (slightly different from the idealized 10.5, as commonly observed under differing ionic conditions), then resealed. Gel electrophoresis of the resulting topoisomer distribution (which runs as a ladder of bands differing by exactly one $Lk$ unit, since more compact, faster-migrating topoisomers correspond to more negatively or positively supercoiled forms) shows a Gaussian-like distribution of $Lk$ values centered at $Lk = 288$, with adjacent bands populated according to a Boltzmann weighting from the quadratic free energy penalty. Step 1: find the reference relaxed linking number for this molecule.

Step 2: since the observed topoisomer distribution is centered at the integer closest to $Lk_0$ (as expected: topoisomerase I, a strand-passage enzyme with no energy input in the relaxation direction, drives the population toward the Boltzmann-weighted center near $Lk_0$, and the nearest achievable integer is 288), this confirms $Lk_0 \approx 288.46$, consistent with the observed center at $Lk = 288$ (the residual $0.46$ turn simply cannot be achieved by an integer-valued $Lk$, so the "relaxed" ensemble is actually centered slightly off zero strain, at $\Delta Lk \approx 288 - 288.46 = -0.46$, an effect only visible because we deliberately used a non-round $h$). Step 3: now suppose a second sample of the same plasmid is extracted directly from actively growing bacteria and found by gel electrophoresis to run as the topoisomer band $Lk = 271$. Compute $\Delta Lk$ and $\sigma$ relative to the true relaxed reference (using the rounded, physically achievable $Lk_0 = 288$, since that is the actual relaxed-state topoisomer population's center).

This value, $\sigma \approx -0.059$, sits squarely in the physiological range ($-0.05$ to $-0.07$), a check that the numbers used are biologically sensible rather than arbitrary. Step 4: estimate how many of these 17 linking number units are likely stored as writhe rather than twist, using the empirical rule of thumb (well-supported by cryo-EM and light-scattering studies of native plasmids) that at physiological $\sigma$, roughly 70% of the deficit appears as writhe once the plectonemic regime is reached, with the remaining 30% as residual twist strain.

Check: $Tw + Wr \approx -5 + (-12) = -17 = \Delta Lk$. ✓. This means the observed 17-unit linking deficit manifests physically as roughly 12 plectonemic superhelical crossings of the axis (visible, in principle, by electron microscopy as interwound loops) plus a smaller residual 5-unit reduction in the local helical winding, exactly the qualitative picture developed in the second derivation's discussion of the two structural regimes.

Limiting-case check: if the plasmid were instead a linear fragment of the same 3000 bp with free ends (not topologically closed), then $Lk$ would not be defined (there is no fixed linking number for a linear duplex with free rotation at the ends), $Tw$ could relax freely along its length, $Wr$ would fluctuate thermally about zero with no net constraint, and none of the machinery above would apply — this is precisely why closure (a covalently closed circle, or a domain anchored at both ends) is the essential precondition for everything in this article, a limiting case worth stating explicitly since it is the most common student error.

The topology of closed DNA was worked out through an unusual convergence of pure mathematics and molecular biology in the 1960s–70s. James H. White, a mathematician, proved the general theorem in 1969 relating the linking number of two closed curves to the integral of twist plus a global geometric writhe term, extending earlier work by the Romanian mathematician Gheorghe Călugăreanu (1959–61) on the linking of ribbon boundary curves — pure differential geometry, developed with no biological application in mind. F. Brock Fuller, another mathematician working at Bell Labs and then Caltech, independently derived and popularized the same relation in a form directly usable by molecular biologists, coining much of the terminology ("writhing number") still used today and showing how it connects to the physical coiling of elastic rods and closed ribbons — the Călugăreanu–White–Fuller theorem is now the standard name for the combined result. The experimental side was driven largely by Jerome Vinograd, who in the mid-1960s first isolated and characterized covalently closed circular DNA (from polyoma virus) as topologically distinct from nicked circles, showing by ultracentrifugation that superhelical and relaxed circular DNA sediment differently — the physical basis for later gel electrophoresis assays of supercoiling. James C. Wang, working at Berkeley, discovered DNA topoisomerase I in E. coli in 1971, the first enzyme shown to relax supercoiled DNA by transient strand breakage, directly connecting the abstract topology to an enzymatic mechanism, and later Martin Gellert's group discovered DNA gyrase (1976), the type II topoisomerase that actively introduces negative supercoils using ATP hydrolysis — establishing that cells do not merely tolerate supercoiling but actively regulate it as a control layer.