The Replication Fork and Its Machinery

The replication fork machinery is not a historical curiosity confined to textbook diagrams — it is under continuous chemical and pharmacological attack and exploitation across medicine and biotechnology. Fluoroquinolone antibiotics (ciprofloxacin, levofloxacin) trap bacterial DNA gyrase and topoisomerase IV in a cleaved-DNA complex, converting an essential, transient enzyme intermediate into a fork-blocking, genotoxic lesion — directly exploiting the topological-relief mechanism described here. Nucleoside reverse transcriptase inhibitors and many anticancer nucleoside analogs (e.g., cytarabine, gemcitabine) are incorporated by replicative or reverse-transcriptase polymerases at the fork and act as chain terminators, exploiting the same 5′→3′ polymerization chemistry that also underlies leading- and lagging-strand extension. PCNA (the eukaryotic sliding clamp) is a major hub for coordinating replication with translesion synthesis, mismatch repair, and cell-cycle checkpoint signaling — small molecules that disrupt PCNA's protein-protein interaction surface are in active development as anticancer agents that would selectively sensitize replicating tumor cells. In biotechnology, rolling-circle amplification and isothermal amplification methods (e.g., LAMP, used in rapid diagnostic testing including for SARS-CoV-2) directly harness strand-displacing polymerases and helicase-like or nicking-enzyme-based strand separation, mechanistically descended from replisome biology. Structural studies of the CMG helicase and replisome by cryo-EM (a field advancing rapidly through the 2010s–2020s) have resolved atomic-level detail of how helicase, polymerase, and clamp physically dock together — informing both basic cell-cycle biology and the design of replication-stress-inducing chemotherapeutics that selectively kill fast-dividing cancer cells whose forks are already under intrinsic replication stress.

Ask the tutor: How does the replisome coordinate the timing of Okazaki fragment termination with primer removal so that gaps never accumulate faster than ligase can seal them?

Picture the replication fork as a Y: the stem is parental duplex DNA, and the two arms are the newly separated template strands, each now paired with a nascent daughter strand. The fork itself is not a static structure — it is the leading edge of a translocating protein machine, the replisome, that physically moves along the DNA, converting one duplex into two. At the very point of strand separation sits the helicase. In bacteria this is DnaB, a homohexamer that encircles the lagging-strand template as a ring and translocates 5′→3′ along it, using ATP hydrolysis at each of six subunit interfaces to ratchet forward and sterically exclude the leading strand — physically wedging the duplex apart rather than "melting" base pairs by direct chemical attack. In eukaryotes the analogous ring is the CMG complex (Cdc45–MCM2-7–GINS), a hexameric MCM ATPase core that instead encircles the leading-strand template and translocates 3′→5′ on it. Either way, the outcome is the same: an actively pumped, ATP-driven separation of two H-bonded strands, running measurably faster than passive thermal fraying would allow. Immediately behind the helicase, the freshly separated single strands are dangerously exposed — vulnerable to nucleases, to forming intrastrand hairpins that would stall polymerase, and simply to reannealing with each other (which is thermodynamically favorable, since base pairing releases free energy). Single-strand binding protein (SSB in bacteria, tetrameric; RPA in eukaryotes, heterotrimeric) coats this ssDNA cooperatively — binding of one SSB tetramer increases the affinity of the next for adjacent sites — forming a continuous protein coat that keeps the template extended and accessible, yet must still be displaceable by primase and polymerase as they act. Ahead of the helicase, a topological crisis brews. DNA is a double helix with roughly one turn every 10.5 bp; unwinding it at the fork does not make that twist vanish, it pushes it ahead as compensatory positive supercoiling in the still-duplex DNA in front of the fork (or, in a topologically closed loop, converts it into positive writhe). Left unrelieved, this strain would rise until the fork could no longer be pushed forward at all. Topoisomerases — DNA gyrase (a type II topoisomerase, using ATP to pass one duplex through a transient double-strand break in another) and topoisomerase I (type I, nicking and resealing a single strand to allow controlled rotation) — continuously discharge this superhelical tension. Behind the helicase, the two template strands must each be copied by DNA polymerase, but here the antiparallel geometry becomes decisive. Take the fork moving left to right. The template strand running 3′→5′ (left to right, matching fork movement) permits the polymerase to synthesize its complementary daughter strand continuously in the 5′→3′ direction, moving in the same direction as the fork — this is the leading strand, needing only a single priming event at the origin. The other template strand runs 5′→3′ in the fork's direction of travel; since polymerase can only extend 5′→3′, synthesis on this strand must run away from the fork, opposite to fork movement, and can only proceed in short stretches before it must restart closer to the fork. This is the lagging strand, synthesized in short pieces called Okazaki fragments (roughly 1000–2000 nt in bacteria, 100–200 nt in eukaryotes), each independently primed. Each Okazaki fragment begins with a short RNA primer (about 10–12 nt) laid down by primase (DnaG in bacteria; the Pol α-primase complex in eukaryotes) — a specialized RNA polymerase capable of de novo synthesis without a pre-existing 3′-OH, something DNA polymerase itself cannot do. DNA polymerase III (bacteria) or Pol δ/ε (eukaryotes) then extends this RNA primer with DNA. Crucially, this extension is tethered to the template by a sliding clamp — β-clamp in bacteria (a homodimer forming a closed ring) or PCNA in eukaryotes (a homotrimer) — loaded around the primer-template junction by a dedicated clamp loader (the γ-complex/DnaX clamp loader in bacteria; RFC in eukaryotes) in an ATP-dependent reaction. The clamp encircles the DNA like a topologically closed ring, sliding freely along the duplex while physically preventing the polymerase from diffusing away — converting an intrinsically low-processivity enzyme (able to add only a handful of nucleotides before dissociating) into one that can synthesize tens of thousands of nucleotides without releasing the template. Because leading and lagging strand synthesis happen simultaneously at the same fork, physically tethered together via multiple polymerase copies bound to a single helicase-primase core (the trombone model, in which the lagging-strand template loops out so its polymerase, still moving 5′→3′, can nonetheless travel in the same net direction as the fork), the entire ensemble — helicase, primase, two or more polymerase-clamp units, SSB, topoisomerase — moves as one coordinated unit: the replisome. Once an Okazaki fragment is extended until it bumps into the RNA primer of the previous fragment, a maturation pathway removes the RNA primer (RNase H trims most of it; the 5′→3′ exonuclease/flap-endonuclease activity of Pol I in bacteria, or FEN1 in eukaryotes, removes the last ribonucleotide and any strand-displaced flap), fills the resulting single-nucleotide gap with DNA polymerase, and finally DNA ligase — using ATP or NAD⁺ as its energy source — seals the remaining nick by forming the final phosphodiester bond, leaving one continuous daughter strand.

The replisome's behavior can be captured in a small number of quantitative relationships that connect fork speed, Okazaki fragment number, energetic cost, and topological strain.

Fork rate and Okazaki fragment count. If the fork moves at rate $v$ (nucleotides per second) and each Okazaki fragment has average length $L$, the number of priming events per unit time on the lagging strand is:

where:

• $v$ = fork translocation rate (nt/s)
• $L$ = average Okazaki fragment length (nt)
• $N_{\text{primed/s}}$ = number of new primers laid down per second at that fork

Reading: a faster fork or shorter fragments both demand more frequent priming events — this sets the required turnover rate of primase and the clamp loader.

Topological constraint on unwinding. The linking number of a closed or constrained DNA segment is invariant except by topoisomerase action, and decomposes as:

where:

• $Lk$ = linking number (integer, the number of times one strand winds around the other in a closed topological domain)
• $Tw$ = twist (helical turns of the duplex)
• $Wr$ = writhe (supercoiling of the duplex axis in three dimensions)

Reading: unwinding at the fork locally reduces $Tw$ in the region being melted, which (with $Lk$ fixed) must be compensated by increased $Tw$ or $Wr$ ahead of the fork — this is precisely the positive supercoiling that gyrase/topo II must relieve.

Free energy cost of primer synthesis and removal (futile cycling). Each Okazaki fragment requires laying down ~10 ribonucleotides (cost: pyrophosphate release, $\Delta G \approx -2$ high-energy phosphoanhydride bonds equivalent per NTP incorporated) that are later degraded and resynthesized as DNA. The net ATP-equivalent cost per fragment is approximately:

where:

• $n_{\text{primer}} \approx 10$ (ribonucleotides in the primer, discarded)
• $n_{\text{DNA}} \approx L$ (deoxynucleotides incorporated to complete the fragment, including the gap-filling step replacing the excised primer)
• $\Delta G_{\text{PPi hydrolysis}} \approx -7.3$ kcal/mol per phosphoanhydride bond driving the reaction essentially to completion (via subsequent pyrophosphatase hydrolysis of PPi)

Reading: RNA primer synthesis is not "free" — every ribonucleotide laid down and later excised still cost one NTP hydrolysis, so lagging strand synthesis is intrinsically more expensive per base pair than leading strand synthesis, though the difference is a small fraction of total replication cost.

Example 1 — Foundational: counting priming events per second at one E. coli fork. E. coli replication forks move at approximately $v = 1000$ nt/s, and Okazaki fragments average $L = 1500$ nt. How many new RNA primers must primase synthesize per second at a single fork, and how does this compare to the primer's own synthesis time? Using $N_{\text{primed/s}} = v/L$:

So on average, primase initiates a new primer roughly every 1.5 seconds at each fork. Primase itself synthesizes an ~10–12 nt RNA primer at a much faster intrinsic rate than replicative polymerase extension (DnaG synthesizes primers rapidly, within a fraction of a second once engaged), so priming is not rate-limiting — the fork's overall speed is set by helicase translocation and polymerase extension, not by how fast primase can act. This matches observation: E. coli has two forks per chromosome (bidirectional from oriC), each needing roughly 1533 total primers over the ~2300 s (38 min) required to replicate the full 2.3×10⁶ bp per fork at 1000 nt/s — consistent with $1533/2300 \approx 0.667$ primers/s, exactly recovering the rate computed directly.

Example 2 — Applied: how much ssDNA must SSB coat at a moving fork, and how many tetramers are needed? Suppose a fork moves at $v = 1000$ nt/s and the lagging-strand template is exposed as single-stranded DNA for an average "window" of one full Okazaki fragment cycle before being converted to duplex — i.e., the exposed ssDNA stretch is on the order of $L = 1500$ nt at steady state (from the point of unwinding to the point where the most recent Okazaki fragment's primer has just been laid and extension is beginning). One SSB tetramer in E. coli occludes about 65 nucleotides of ssDNA per tetramer (measured binding site size, "SSB65" mode). How many SSB tetramers coat this stretch at any instant?

This is a steady-state occupancy, not a cumulative count — as the fork advances and each Okazaki fragment is primed and extended, the same ~23 tetramers' worth of ssDNA is continually displaced by the extending polymerase and replaced by newly unwound template ahead. The number scales linearly with exposed ssDNA length, so if fragment length were instead 200 nt (as in eukaryotic replication with RPA), roughly $200/30 \approx 7$ RPA heterotrimers (RPA's occluded site size, ~30 nt) would coat the equivalent stretch — a smaller number consistent with eukaryotic forks generating shorter transient ssDNA windows.

Example 3 — Exam/competition level: topological strain and the ATP cost of unrestrained unwinding without gyrase. Consider a circular bacterial chromosome-like domain of $N = 10{,}000$ bp, initially fully relaxed ($Lk_0 = N/10.5 = 952.4$, using 10.5 bp/turn). A replication fork unwinds $\Delta = 500$ bp of this domain (converting that stretch locally from duplex twist into single strands ahead of the fork's own template usage — i.e., this 500 bp is removed from the "twistable" pool), with no topoisomerase present to relieve strain, and gyrase is fully inhibited (e.g., by a fluoroquinolone). (a) Compute the writhe that must accumulate in the remaining, still-double-stranded 9,500 bp to conserve linking number. Since the fork is unwinding a closed domain, $Lk$ is conserved at its pre-unwinding value $Lk_0 = 952.4$ (assume $Lk$ was already fixed at this value before the fork began, i.e. the domain was topologically sealed at both ends, as in a plasmid or a chromosomal loop anchored by proteins). After unwinding 500 bp, the twist in the unwound region drops essentially to zero (single strands, no helical twist to speak of over that stretch), while the intrinsic twist that would "belong" to that 500 bp under B-form geometry is:

The remaining 9,500 bp, if it stayed fully relaxed B-form, would have twist:

Conservation of $Lk = Tw + Wr = 952.4$ requires, since the unwound region no longer contributes 47.6 turns of twist:

If the remaining duplex DNA retains its natural B-form twist ($904.8$ turns) rather than absorbing any of the deficit as altered twist, all of the "missing" 47.6 turns must appear as compensatory writhe:

(b) If DNA gyrase, once reactivated, removes this strain by introducing negative supercoils at a rate of 1 negative supercoil (one $Lk$ decrement) per ATP hydrolyzed (gyrase's actual mechanistic stoichiometry is approximately 1 ATP hydrolyzed per catalytic cycle introducing 2 negative supercoils via strand passage, but we use 1:1 here as a clean illustrative rate — see below for the corrected stoichiometry), how many ATP are required to fully relax this domain back to $Wr = 0$? Using the illustrative 1:1 rate:

Correcting to gyrase's actual stoichiometry (each catalytic cycle of strand passage introduces 2 negative superhelical turns per 2 ATP hydrolyzed, i.e. 1 turn per ATP is in fact close to the literature value of ~1–2 supercoils introduced per ATP hydrolyzed depending on conditions), the estimate of order-40–50 ATP molecules per fork-domain-sized relaxation event stands as the right order of magnitude — small compared to the thousands of ATP-equivalents spent on the polymerization itself, confirming that topological relief is energetically cheap relative to synthesis, even though it is mechanistically indispensable (without it, the fork physically cannot advance further once superhelical density exceeds a critical threshold, typically $\sigma \approx -0.05$ to $+0.05$ before torque stalls unwinding).

Limiting case check: if the fork unwinds only $\Delta \to 0$ bp (negligible), $Wr \to 0$ and no topoisomerase action is needed — consistent with the fact that only actively progressing forks generate meaningful topological strain, and a stalled fork (zero net unwinding) requires no ongoing gyrase turnover to maintain, only to resume.

The mechanistic understanding of the replication fork emerged from a lineage of biochemists working through the 1950s–1980s who insisted on reconstituting replication with purified components rather than relying solely on genetics.