Contents
02 — DNA structure Part 1 — Molecular foundations

02 — DNA structure

Before this: Ch 00 · Ch 01 · Time: ~40 min

What you'll be able to do

The core idea

DNA is a directed, charged, twisted, topologically constrained polymer. Each adjective does real work later:

The double helix is not a static ladder. It is a stressed, flexible, locally breathing object whose deformations are functionally load-bearing.


1. The monomer

A nucleotide has three parts: a nitrogenous base, a five-carbon sugar, and a phosphate.

The sugar is 2'-deoxyribose. Its carbons are numbered 1' to 5' — primed, so they never collide with the numbering of the base's ring atoms. Three positions matter:

PositionWhat is attachedWhy you care
1'the base, via an N-glycosidic bondBreak this and you get an abasic site — one of the commonest lesions (Ch 17)
3'a hydroxylThe . Polymerases extend here, and only here
5'a phosphateThe . Incoming nucleotides arrive as triphosphates
2'H in DNA, OH in RNAThat single oxygen is the difference between an archive and a message queue

The bases come in two sizes:

ClassMembersRingsSize
adenine (A), guanine (G)two fused rings, 9 atomslarge
cytosine (C), thymine (T), uracil (U, in RNA)one ring, 6 atomssmall

Thymine is uracil with a methyl group at position 5. That methyl is not decorative: it sticks into the as a hydrophobic marker proteins can feel, and it lets repair enzymes treat any uracil found in DNA as damage — because uracil in DNA is usually a deaminated cytosine (Ch 16).

2. The polymer: why direction is chemistry, not convention

Nucleotides link through phosphodiester bonds: one phosphate bridges the 3'-OH of one sugar and the 5'-OH of the next.

 5' end
  │
  O
  │
  P(=O)(O⁻)      ← phosphodiester: two ester bonds, one ionised oxygen.
  │                One negative charge per residue, at any cellular pH.
  O
  │
 5'┐
   │  2'-deoxyribose ──1'── N ── BASE
 3'┘
  │
  O
  │
  P(=O)(O⁻)
  │
  ⋮
  │
  OH             ← free 3'-hydroxyl: the only group a polymerase can extend
 3' end

The chain is therefore asymmetric. One end terminates in a phosphate on C5', the other in a hydroxyl on C3'. This is why sequence is written 5'→3' by universal convention, why FASTA records have an implied direction, and why a coordinate on the minus strand needs care (Ch 41).

The asymmetry has a mechanistic cause worth carrying. The 3'-OH attacks the α-phosphate of an incoming deoxynucleoside triphosphate, releasing pyrophosphate — so the energy for the bond arrives with the new nucleotide. If chains grew 3'→5' instead, the triphosphate would sit on the growing end, and any excision would destroy the chain's own energy source. Extension is 5'→3' because error correction is only affordable in that direction.

Two strands pair, and the pairing geometry forces them antiparallel:

5'-A  G  G  T  C  A  T  G-3'
   :  ‖  ‖  :  ‖  :  :  ‖       : = A·T, two hydrogen bonds
3'-T  C  C  A  G  T  A  C-5'     ‖ = G·C, three hydrogen bonds

3. Base pairing is a geometry problem

Two select the pairs, and neither is "A likes T".

Constant width. The C1'–C1' distance across a pair must stay near 10.8 Å for the backbone to run smoothly. + purine is too wide; + pyrimidine is too narrow. So every pair must be one of each.

Complementary hydrogen-bond patterns. Along the pairing edge each base presents donors (N–H) and acceptors (C=O, ring N). Of the four purine–pyrimidine combinations, only two produce donor opposite acceptor at every position:

A·C and G·T fail because they place donor against donor. They can pair if a base flips to a rare — the same molecule with one hydrogen sitting on a different atom, which converts a donor into an acceptor and so changes which base it fits. That flip is precisely how spontaneous mispairing happens and where a large share of the comes from (Ch 16).

4. What actually holds the helix together

Everyone learns "G·C has three hydrogen bonds, so GC-rich DNA is more stable". The conclusion is right and the reasoning is wrong.

Hydrogen bonds between bases are formed in water, and water is an excellent hydrogen-bond partner itself. Separating the strands does not destroy those bonds — it replaces them with bonds to water, at close to no net cost. What is genuinely lost on melting is stacking: the face-to-face van der Waals and electronic interaction between adjacent base pairs, and the burial of flat hydrophobic surface away from water.

Yakovchuk and colleagues (2006) measured the two contributions separately, using duplexes with single nicks and gaps to isolate stacking. Their decomposition, at 15 mM Na⁺:

ContributionA·TG·C
Base pairing (ΔG per pair)+0.6 kcal/mol — destabilising≈ −0.1 kcal/mol — negligible
Base stacking (ΔG per contact)≈ −1.0 kcal/mol≈ −1.5 kcal/mol

GC-rich DNA melts at a higher temperature not because of the third hydrogen bond but because G and C stack better.

Base pairing supplies the specificity. supplies the stability. They are different jobs, and crediting the hydrogen bonds with both is the most common structural misconception in molecular biology. Pairing tells the helix which partner; stacking tells it how hard to hold on.

5. The grooves, and reading sequence without opening the helix

Because the two glycosidic bonds emerge from the same side of each base pair rather than diametrically opposite, the backbones are unevenly spaced around the helix. B-DNA therefore has a (~22 Å wide, ~8.5 Å deep) and a (~12 Å wide, ~7.5 Å deep).

This is not cosmetic. The edges of the bases are exposed in both grooves, and a protein can hydrogen-bond to them without melting the duplex. But only one groove carries enough information. Writing D = donor, A = acceptor, M = methyl, H = non-polar hydrogen, reading across each pair:

                 MAJOR GROOVE
   A·T      A(N7)   D(N6H)  A(O4)   M(C5-methyl)
   T·A      M       A       D       A
   G·C      A(N7)   A(O6)   D(N4H)  H
   C·G      H       D       A       A
            → four distinct patterns. Fully readable.

                 MINOR GROOVE
   A·T      A(N3)   H       A(O2)
   T·A      A       H       A          ← identical to A·T
   G·C      A(N3)   D(N2H)  A(O2)
   C·G      A       D       A          ← identical to G·C
            → two patterns. AT vs GC only; orientation invisible.

The minor groove cannot tell A·T from T·A, or G·C from C·G. The major groove can tell all four apart. That single asymmetry is why almost every sequence-specific DNA-binding protein inserts an α-helix — a "recognition helix", conveniently about 12 Å across — into the major groove, and why minor-groove binders read shape (local width, propeller twist, AT-tract narrowing) rather than base identity. Both mechanisms recur constantly in Ch 22.

6. B, A and Z

B-DNA is the physiological form. It is not the only one.

B-DNAA-DNA
Handednessrightrightleft
bp per turn~10.5 in solution (10.0–10.4 in fibres)~1112 (6 dinucleotides)
Rise per bp3.4 Å2.3–2.6 Å~3.7 Å
Diameter~20 Å~26 Å~18 Å
Sugar puckerC2'-endoC3'-endoalternating
Base tilt~1°~19°~9°
Groovesmajor wide/deep, minor narrowmajor deep and narrow, minor shallow and widesingle narrow, deep groove (minor-like); flattened to a convex surface
Occurs whenphysiological hydrationdehydrated; any RNA duplex or RNA:DNA hybridalternating /, e.g. (CG)ₙ, under negative

Two rows are load-bearing rather than trivia. A-form is not exotic — it is what RNA has to be: the 2'-OH clashes with C2'-endo pucker, so every RNA duplex and RNA:DNA hybrid is A-form, which means every bubble contains an A-form segment. Z-form is a torsional relief valve: a left-handed segment absorbs negative supercoiling, so Z-DNA forms transiently behind a transcribing polymerase, where it is recognised by Zα-domain proteins including ADAR1 and ZBP1 — linking it to RNA editing and innate immune signalling.

Remember the 10.5 bp/turn figure. Every topology calculation uses it.

7. Melting: reading stability off the sequence

Heat DNA and the strands separate, sharply rather than gradually, because unstacking is cooperative: opening the first pair inside a helix costs stacking on both sides, while extending an existing bubble costs it on one. It is a one-dimensional nucleation-and-growth problem with the structure of an Ising model.

You detect it by absorbance. Stacked bases have coupled π systems that absorb less UV than free ones, so unstacking raises A₂₆₀ by up to about 40% — the hyperchromic effect. Note what that implies: the standard assay for duplex stability is literally a measurement of stacking.

T_m is the temperature at which half the strands are duplexed. For long DNA an empirical relation works:

T_m ≈ 81.5 + 16.6·log₁₀[Na⁺] + 0.41·(%GC) − 500/L

but it is blind to sequence order, and order matters. The model everyone actually uses — inside every primer-design tool — is nearest-neighbour thermodynamics, whose parameters are indexed by dinucleotide stack, not by base. There is no term in it for "a G·C pair". You sum ΔH° and ΔS° over the L−1 stacks plus two initiation terms, and the thermodynamics of Ch 01 §6 turns that sum into a temperature. For a non-self-complementary duplex A + B ⇌ AB the standard result is

                ΔH°
  T_m  =  ───────────────────
          ΔS° + R·ln(C_T/4)

with C_T the total strand concentration and R the gas constant; the C_T/4 is what the equilibrium condition becomes at T_m, where half the strands are paired. Take the formula as given. This book does not ship the nearest-neighbour parameter tables — they are published, SantaLucia's unified set being the standard — and the point of this section is not arithmetic but what the model implies.

Test it. 5'-GCGCATATGC-3' and 5'-GACTGACTGC-3' share a composition (3 G, 3 C, 2 A, 2 T), share G·C termini, and therefore share a hydrogen-bond count: 26 each. Any composition-based rule must return the same T_m for both. Summing SantaLucia's unified parameters gives ΔH° = −78.4 vs −75.4 kcal/mol and ΔS° = −213.5 vs −206.6 cal/(mol·K), so at 2 μM strand in 1 M Na⁺, T_m ≈ 50 °C vs 47 °C. Same bases, same bonds, different order, different stability.

8. Topology: DNA is a closed object

Take a circular double helix — a plasmid, a bacterial , or a eukaryotic chromatin loop anchored at both ends. Its linking number Lk, the number of times one strand passes through the closed loop formed by the other, is an integer and cannot change without breaking a strand.

  Lk  =  Tw  +  Wr

Tw (twist) counts how many times the strands wind around each other; Wr (writhe) counts how many times the axis crosses itself in space. Lk is invariant, so Tw and Wr trade. Unwind a region locally, decreasing Tw, and Wr must rise to compensate: the molecule coils around itself. It supercoils.

Relaxed B-DNA has Lk₀ = N/10.5 for N base pairs, and superhelical density is the normalised deviation:

  σ = (Lk − Lk₀) / Lk₀

In vivo σ runs about −0.05 to −0.07: cells hold their DNA underwound, as a deliberate energy store. Negative pre-pays part of the cost of separating strands, so open and origins fire more readily.

and replication both manufacture supercoils. A polymerase tracking a helix must rotate relative to it once every 10.5 bp, and it cannot — it is tethered by a growing RNA, by bound processing factors, and in bacteria by . So the DNA must rotate instead; and where the DNA is anchored too, the rotation has nowhere to go. Liu and Wang's twin-domain model:

graph LR A["<b>behind</b><br/>(−) supercoiled<br/><i>underwound</i><br/>strands open easily"] P["<b>RNA polymerase</b><br/>cannot spin:<br/>tethered by nascent RNA"] B["<b>ahead</b><br/>(+) supercoiled<br/><i>overwound</i><br/>torque resists unwinding"] A --- P --- B T1["<i>bacteria:</i> topo I / topo III (type IA)<br/>relax (−) only, no ATP<br/>ΔLk = ±1"] -.-> A T2["<i>bacteria:</i> gyrase<br/>removes (+) by introducing (−)<br/>ATP-dependent"] -.-> B T3["<i>eukaryotes:</i> topo I (type IB) and topo II<br/>relax <b>either sign</b>, so either domain"] -.-> A T3 -.-> B

are the enzymes that change Lk. Type I cuts one strand and reseals it, and splits into two mechanisms that behave differently. Type IA — bacterial topo I, topo III — passes the intact strand through the break, so ΔLk = ±1 exactly, and it relaxes negative supercoils only. Type IB — human topo I, poxvirus topo — instead lets the helix swivel about the intact strand before resealing, so it relaxes both signs and changes Lk by one or many turns per catalytic event. Neither spends ATP; they are releasing stored torsional energy. (The exception is reverse gyrase, a type IA enzyme fused to a domain, which burns ATP to add positive supercoils in hyperthermophiles.) Type II cuts both strands, passes an entire duplex through the gap, and reseals — ΔLk = ±2, ATP-dependent, and again indifferent to sign. Bacterial DNA gyrase is the special case that runs uphill: it actively introduces negative supercoils.

9. Flexibility, and what happens when B-form fails

At 3.4 Å per base pair, 1 kb of B-DNA is 340 nm long — so the 3.1 Gb haploid human genome is about 1 m of molecule, and a nucleus holds ~2 m of it. But DNA is not a wire. Its is about 50 nm — roughly 150 bp — at physiological salt: below that scale it behaves as a stiff rod, above it as a flexible polymer executing a random walk. Two consequences:

Under torsional stress or with the right sequence, DNA leaves B-form entirely:

StructureSequence requirementNote
Hairpin / cruciforminverted repeatExtrudes under negative ; a recombination and deletion hotspot
G-quadruplex (G4)four runs of ≥3 GFour guanines make a planar tetrad via Hoogsteen bonds; tetrads stack with K⁺ between them
i-motifthe C-rich complementary strandHemiprotonated C·C⁺ pairs; favoured at slightly acidic pH
Triplex / H-DNApolypurine·polypyrimidine mirror repeatThird strand in the
        G ———— G
        │  K⁺  │      Hoogsteen bonds, not Watson–Crick.
        │      │      Four G's per tetrad; tetrads stack;
        G ———— G      a potassium ion sits between them.

G-quadruplexes are not a curiosity. G4-seq mapped 716,310 G4-forming sequences in the human genome (Chambers et al. 2015) — most of them not predicted by the then-standard motif search. They are enriched at , 5' UTRs, replication origins, and the telomeric TTAGGG repeat, they impede polymerases, and dedicated exist to unwind them.

10. How this was worked out, and what it immediately implied

Until 1950 the consensus was that DNA was too monotonous to carry information. Levene's tetranucleotide hypothesis held that DNA was a repeating ACGT unit; protein, with 20 letters, looked the better candidate — which is largely why Avery, MacLeod and McCarty's 1944 demonstration that the bacterial transforming principle was DNA met scepticism.

Erwin Chargaff dismantled it. Measuring base composition across species he found that composition varies between species — killing the tetranucleotide hypothesis outright — and that within any species A = T and G = C. He reported the ratios; he did not propose pairing.

Rosalind Franklin and Raymond Gosling produced Photo 51 at King's College London in May 1952, a fibre diffraction pattern of the hydrated B-form. It said three things at a glance: the X-shaped cross of reflections is the signature of a helix; the layer-line spacing gives a 34 Å repeat with a strong 3.4 Å meridional reflection, i.e. ten residues per turn; and the absent fourth layer line means the two strands are offset by about 3/8 of a turn — exactly what produces grooves of unequal width.

The last piece was tautomeric. Watson's early models used the enol forms of the bases and could not be made to work; Jerry Donohue told him the keto forms were correct, and with keto bases the donor/acceptor patterns fall out immediately as A·T and G·C.

graph TD C["<b>Chargaff 1950</b><br/>A = T, G = C;<br/>composition varies by species"] F["<b>Franklin & Gosling 1952</b><br/>Photo 51: helical cross,<br/>34 Å pitch, 3.4 Å rise,<br/>4th layer line absent"] D["<b>Donohue 1953</b><br/>bases are keto,<br/>not enol"] M["<b>antiparallel double helix</b><br/>A·T and G·C pairs,<br/>unequal grooves"] R["<b>each strand specifies the other</b><br/>→ a copying mechanism<br/>falls out of the structure"] C --> M F --> M D --> M M --> R

The three papers appeared together in Nature on 25 April 1953. Watson and Crick's closes by noting that their pairing scheme immediately suggests "a possible copying mechanism for the genetic material" — the rare case of a structure that hands you the mechanism for free. If A always faces T and G always faces C, each strand fully specifies its partner, so separating them yields two templates. Meselson and Stahl confirmed it in 1958 (Ch 04).

Franklin's unpublished data reached Watson and Crick without her knowledge or consent. She died in 1958, aged 37; the 1962 Nobel went to Watson, Crick and Wilkins, and the prize is limited to three recipients. The statute formally barring posthumous awards came later, in 1974 — but she had died four years before the announcement, and even the pre-1974 exception reached only those who died after being nominated in the same year.


Common misconceptions

What people thinkWhat's actually true
G·C pairs are stronger because they have three hydrogen bondsIn water those bonds are nearly free — the strands simply hydrogen-bond to water instead. GC-rich DNA melts higher because G and C stack better. By one careful decomposition, A·T pairing is net destabilising
DNA is a rigid ladder ~150 bp. Beyond that it is a floppy polymer. It bends, writhes, and breathes open locally — all functionally required
A protein must open the helix to read the sequenceThe exposes a distinct donor/acceptor/methyl pattern for each of the four base-pair orientations. Sequence-specific readout happens on intact duplex
The just carries less informationIt carries categorically less: it cannot distinguish A·T from T·A, or G·C from C·G. Minor-groove binders read shape, not identity
5'→3' is a bookkeeping conventionIt is a chemical asymmetry with a mechanistic cause: the growing chain must carry the so that does not destroy the reaction's energy source
You can unwind DNA locally without consequenceIn a topologically closed domain Lk is invariant. Unwinding here forces overwinding there. Every helix-opening process manufactures supercoils
Negative means "wound too tightly"The opposite: underwound relative to relaxed B-form. Cells maintain σ ≈ −0.05 to −0.07 because it stores that helps open and origins
A-DNA and are laboratory artefactsA-form is the obligatory conformation of every RNA duplex and RNA:DNA hybrid. Z-DNA forms behind transcribing polymerases and is read by Zα-domain proteins
Watson and Crick discovered DNAThey determined its structure. DNA was isolated in 1869 and shown to be the genetic material in 1944 and 1952

Worked example: how much supercoiling does one gene generate?

Take II transcribing a 10 kb gene at 2 kb/min.

1. Convert to turns per second. 2 kb/min = 2,000/60 ≈ 33 bp/s. At 10.5 bp per turn:

  33 bp/s ÷ 10.5 bp/turn  =  3.2 turns/s

2. Each turn is two supercoils. Lk is conserved, so every turn the polymerase fails to make around the helix appears as +1 ahead and −1 behind. The enzyme is generating roughly 3 positive supercoils per second downstream and 3 negative per second upstream.

3. Total over the gene. 10,000 bp ÷ 10.5 = 952 turns, taking 300 s. So ~950 supercoils of each sign, per polymerase, per pass.

4. How much can the DNA absorb? Suppose the gene sits in a chromatin loop of 100 kb anchored at both ends (Ch 50). Its relaxed linking number is:

  Lk₀ = 100,000 / 10.5 = 9,524

At the physiological σ ≈ −0.06, the domain's resting ΔLk is:

  ΔLk = σ · Lk₀ = −0.06 × 9,524 ≈ −571

So the entire torsional budget of a 100 kb domain is roughly ±570 turns.

5. Conclude. One polymerase generates ~950 turns of each sign in five minutes, well past what the domain can hold; torque would stall it in under three. And that is one polymerase — an active gene carries many, and build positive supercoils faster still. therefore run continuously, everywhere, at high throughput. They are not occasional maintenance; they are a mandatory part of every and replication event.

The clinical corollary is immediate. Topoisomerases work through a transient covalent enzyme–DNA intermediate, so a drug that traps that intermediate turns an ordinary catalytic cycle into a permanent DNA break. That is how fluoroquinolones (ciprofloxacin, against bacterial gyrase and topo IV), etoposide and doxorubicin (human topo II), and irinotecan and topotecan (human topo I) kill cells. Three drug classes downstream of one conserved quantity.

Connections