# DNA Replication: Semiconservative Copying and Its Machinery

> Each daughter molecule keeps one old strand and builds one new strand. This lesson uses Meselson and Stahl's E. coli density-gradient experiment to distinguish three inheritance models, then follows how E. coli and eukaryotic forks build the new strands, check errors and finish the ends of linear chromosomes.

Course: Genetics and Molecular Biology (DNA: copying, changing, repairing). Level: core university (third year). About 18 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/dna-replication/

Builds on: DNA Is the Genetic Material

## What you'll learn

- Predict the CsCl density-gradient bands for conservative, semiconservative and dispersive replication over three generations, and name the observation in Meselson and Stahl's E. coli experiment that rules out each alternative.
- Describe the E. coli replication machinery (oriC, DnaA, DnaB, DnaG, Pol III holoenzyme, Pol I, ligase, gyrase) and its eukaryotic equivalents (many origins, CMG helicase, Pol α-primase, Pol ε and Pol δ), keeping each organism's names.
- Explain why one strand at a fork is made in pieces, using the rule that polymerases only extend a 3' end and the pulse-labelling evidence of Okazaki and colleagues.
- Explain how base selection and proofreading keep errors rare, why linear chromosomes have an end-replication problem, and how telomerase extends the ends.

**One generation after Matthew Meselson and Franklin Stahl switched heavy-nitrogen *E. coli* to ordinary nitrogen, the extracted DNA formed one band halfway between heavy and light. Why halfway?**

This lesson builds on [DNA as the genetic material](https://stemhop.com/lessons/dna-the-genetic-material/) and the dNTPs from [nucleotide metabolism](https://stemhop.com/lessons/nucleotide-metabolism/).

> **The one idea.** Each daughter molecule keeps one old strand and builds one new strand.
>
> *Scope:* chromosome replication. Meselson and Stahl showed in *E. coli* that each daughter inherits one parental subunit, not which enzymes copy it. Heating fitted those subunits being the two strands, but their experiment alone did not prove that.

## Three models, one experiment

The first film ran three models: **conservative replication** keeps the parental duplex whole, **semiconservative** gives each daughter one parental and one new strand, and **dispersive** mixes old and new segments through both strands.

Meselson and Stahl grew *E. coli* B for many generations with heavy nitrogen-15, then added a tenfold excess of ordinary nitrogen-14 ammonium chloride. **Equilibrium density-gradient centrifugation** in concentrated caesium chloride then separates heavy, hybrid and light DNA by buoyant density.

**Predict.** Start with one heavy-heavy duplex. Assume complete synchronous rounds, equal-length molecules, fully light new strands and no recycled heavy nitrogen. Under semiconservative replication, what molecules appear after a third round in light medium?

> **Check your prediction.** Each original heavy strand stays in one hybrid; the other six molecules are light: 25% hybrid and 75% light.

| Model | After generation 1 | After generation 2 | After generation 3 |
| --- | --- | --- | --- |
| Conservative | 1 heavy + 1 light (50% heavy) | 1 heavy + 3 light (25% heavy) | 1 heavy + 7 light (12.5% heavy, 87.5% light) |
| Semiconservative | 2 hybrid | 2 hybrid + 2 light | 2 hybrid + 6 light |
| Dispersive, even mixing | one band 1/2 of the way from light to heavy | one band 1/4 of the way | one band 1/8 of the way |

**Observed.** A single hybrid band after one generation rules out conservative copying, though not dispersive. Hybrid and light bands after two generations, in equal amounts, rule out uniform dispersive copying.

**Predict.** Does the average heavy fraction per generation differ among the three models?

> **Check your prediction.** No. In every model it is one-half, one-quarter, then one-eighth. The distribution of label among molecules, not its average, distinguishes the models.

*Interactive on the lesson page:* Density band predictor: what does each model predict?

[Autoradiography](https://stemhop.com/methods/isotope-tracers/) of bean (*Vicia faba*) chromosomes by J. Herbert Taylor and colleagues in 1957 was separate evidence.

Copying happens at a Y-shaped **replication fork**. Every DNA polymerase extends only a free 3’-OH, whose hydroxyl attacks the incoming dNTP and releases pyrophosphate (see [enzyme action](https://stemhop.com/lessons/enzyme-action-basics/)). The replicative DNA polymerases also cannot start a chain: each needs a **primer**, a short paired stretch with a 3’ end to extend. The templates are antiparallel, so how can one fork extend 3’ ends on both? The second film answers.

**Film transcript: Why One Strand Is Built in Pieces**

### Film transcript: Why One Strand Is Built in Pieces

Each daughter molecule keeps one old strand and builds one new strand. But the two old strands point opposite ways. How can one fork build both new partners? Keep the first film's colour map. Purple marks parental templates, and cyan marks new DNA. The parental duplex opens into a moving Y-shaped replication fork. Each purple strand must guide a cyan partner, but the two templates run in opposite directions. Every DNA polymerase adds a nucleotide only to a 3' end. New DNA therefore grows 5' to 3'. A copying polymerase also needs that end paired to its template. Replicative polymerases cannot start a chain. Primase must first supply a short paired primer.

The fork moves right, exposing more purple template. Which cyan strand can grow with that fork while its polymerase extends a 3' end? On one template, the growing 3' end faces the fork. Polymerase follows the opening helix and lengthens one continuous leading strand. Its direction of synthesis stays 5' to 3'. On the other template, the usable 3' end faces away from the fork. Polymerase cannot reverse its chemistry. Primase starts near the fork, then polymerase extends back toward the previous piece. As more template opens, the cycle starts again. These pieces are Okazaki fragments. The pieces must become one strand. In E. coli, DNA polymerase I removes each RNA primer while replacing it with DNA.

DNA ligase seals the remaining nick. The finished lagging strand is continuous, although it was built discontinuously. Several proteins keep the fork moving. DnaB helicase unwinds the parental duplex. Single-strand binding protein holds the templates apart, and gyrase relieves overwinding ahead.

DnaG primase places primers. DNA polymerase III, held by its β clamp, performs most synthesis. Eukaryotes keep the same geometry with different parts and many origins. CMG is the helicase. Pol α-primase starts chains. The usual summary assigns most leading synthesis to Pol ε. It assigns most lagging synthesis to Pol δ on PCNA. Names change, but the constraint does not. Every fork needs a way to open DNA, start new chains, extend them, replace starter material, and seal the products. The text compares these jobs and the evidence behind them. Polymerases extend a 3' end, so one strand is built in pieces. Both strands still obey the same chemistry.

The enzyme never changes synthesis direction to solve the geometry. This constraint turns opposite templates into different building patterns.

From the film

- Polymerases extend a 3' end, so one strand is built in pieces.
- Both new strands grow 5' to 3', although they grow in opposite directions at a fork.
- Primase restarts the lagging strand as the fork exposes more template.
- Primer removal and ligase turn short fragments into one continuous strand.

## Testing the pieces

The second film named the **leading strand** and the **lagging strand**’s **Okazaki fragments**, not the evidence for them.

*Figure (AI-generated illustration).* Both new strands grow 5' to 3', so one is made continuously toward the fork and the other in short primed pieces growing away from it. Image: a Y-shaped replication fork opening to the right. The upper new strand is one long piece growing toward the fork; the lower new strand is three short primed Okazaki fragments growing away from the fork. An inset shows a primer being extended from its 3' end.

In 1968 Reiji Okazaki, Tsuneko Okazaki, Kiwako Sakabe, Kazunori Sugimoto and Akio Sugino tested for pieces. They grew *E. coli* B at 20°C to slow replication, gave radioactive-thymidine pulses as short as five seconds, denatured the DNA and sized the labelled chains on alkaline sucrose gradients. After a five-second pulse most label was in a slow component of about 11S; with longer labelling it moved into long DNA. They suggested pieces of about 1,000 to 2,000 nucleotides, leaving their direction and joining for further proof. Eukaryotic Okazaki fragments are shorter, about 100 to 200 nucleotides.

## The fork’s jobs, organism by organism

The table adds what the second film skipped: origin opening, helicase direction and primer removal.

| Job | *E. coli* (one origin, **oriC**, circular chromosome) | Eukaryotes (many origins, linear chromosomes) |
| --- | --- | --- |
| Open the origin | DnaA binds oriC and opens a short AT-rich stretch | In G1, inactive MCM2-7 rings load at origins |
| Unwind | DnaB ring (loaded by DnaC, ATP-driven) moves 5’ to 3’ on the lagging-strand template; SSB coats single strands; gyrase, a topoisomerase, relieves overwinding | In S phase, Cdc45 and GINS join MCM as the active **CMG helicase**, which moves 3’ to 5’ on the leading-strand template |
| Prime | DnaG primase makes short RNA primers | **Pol α-primase** makes RNA and extends it briefly with DNA |
| Extend | Pol III holoenzyme; its β sliding clamp keeps it on DNA for thousands of nucleotides | In the usual summary, **Pol ε** most leading-strand and **Pol δ** most lagging-strand synthesis, both on the PCNA clamp |
| Remove primers | Pol I removes each primer with its 5’ to 3’ exonuclease while filling the gap | Pol δ displaces primer material as a 5’ flap; FEN1 cuts short flaps, Dna2 helps with longer ones, RNase H has a smaller part |
| Seal | DNA ligase | DNA ligase I |

Arthur Kornberg’s laboratory purified Pol I first, but a 1969 mutant with very little Pol I polymerase activity still grew; Pol III is the replicative polymerase.

**Predict.** The *E. coli* K-12 chromosome is about 4.6 million base pairs. Two forks each move at about 1,000 nucleotides per second. How long does one round take, and how can cells divide about every 20 minutes in rich medium?

> **Check your prediction.** Each fork copies half: (4.6 × 106 / 2) / 1,000 = 2,300 seconds, about 38 minutes. Fork speed rises only modestly with growth rate, far too little to close the gap, so new rounds begin at oriC before earlier rounds finish, and each division still receives a finished chromosome.

## Three layers of accuracy

Polymerases first **select** bases: the active site closes well around a correct pair and poorly around a mispair, leaving roughly one error per 104 to 105 nucleotides. They then **proofread**: a 3’ to 5’ exonuclease removes a mispaired 3’ end. In *E. coli* Pol III that activity is in the ε subunit. Proofreading improves accuracy about 100-fold, and up to 1,000-fold depending on polymerase and mispair. Not every polymerase proofreads: eukaryotic Pol α lacks a functional proofreading exonuclease. **Mismatch repair** then fixes errors after the fork ([DNA repair and recombination](https://stemhop.com/lessons/dna-repair-and-recombination/)). Together they bring the *E. coli* rate to roughly one error per 109 to 1010 base pairs per round.

**Predict.** At that rate, about how often does copying a 4.6-million-base-pair chromosome add an error?

> **Check your prediction.** About one copy in 200 to 2,000. These are orders of magnitude, not fixed constants; assay and sequence matter.

## Linear chromosomes need finishing work

**Predict.** When an RNA primer is removed, why can a circular chromosome always fill the gap, while the gap left by the last primer at a linear tip may not be filled?

> **Check your prediction.** A circle has DNA behind every gap, supplying a 3’ end for extension. At a linear tip, no upstream 3’ end may remain. This is the classical **end-replication problem**.

By the mid-1990s it was clear that real telomeres end in a single-stranded 3’ overhang of the G-rich strand, which leading-strand synthesis cannot regenerate and telomerase makes up; the lagging strand has its own finishing problem (Going deeper).

*Figure (AI-generated illustration).* Removing a terminal primer can leave no 3' end to extend; real telomeres also require overhang restoration and C-strand fill-in. Image: a linear chromosome with a telomere at each end, and two close-ups of a lagging-strand tip. In this classical model, with the last RNA primer in place the new strand reaches the end; once it is removed, the new strand stops short of the template end.

**Telomeres** are repeated sequences at chromosome ends: TTGGGG in the ciliate *Tetrahymena* and TTAGGG in humans. In 1985 Carol Greider and Elizabeth Blackburn found that *Tetrahymena* extracts added TTGGGG repeats to synthetic telomeric primers; heat and protease destroyed the activity. In 1989 they reported that the enzyme’s RNA contains CAACCCCAA, suggested as the template.

**Telomerase** is a reverse transcriptase (it makes DNA from an RNA template) that carries its own RNA and extends the G-rich 3’ strand. The C strand is filled in separately; in engineered human cells that required the CST complex with Pol α-primase. Most human somatic cells make little telomerase, so their telomeres shorten as they divide; germline cells and most cancer cells keep it active. Blackburn, Greider and Jack Szostak shared the 2009 Nobel Prize in Physiology or Medicine.

### Design problem

*E. coli* are given a 5-second pulse of tritiated thymidine, and their DNA is denatured and sized. Right after the pulse, most label is in pieces of about 1,000 to 2,000 nucleotides. After a 60-second chase in unlabelled thymidine, the label is in long DNA. In a ligase-defective mutant, the short pieces persist after the chase. What do you conclude, and what control would you add?

> **One way to do it.** **Conclusion.** At least some new DNA is made as short pieces later joined into long strands. The chase shows the same labelled molecules moving from short to long, a precursor becoming a product, and the mutant shows that joining needs ligase.
>
> **Controls.** Long-labelled DNA handled the same way should stay long (not extraction breakage); repair mutants should still make the pieces (not repair patches). Hybridizing pulse-labelled pieces to separated strands of a region with known fork direction tests the semidiscontinuous prediction: pieces mostly from the lagging strand. Early data could not exclude pieces on both strands, and later *E. coli* work by Tye and colleagues in 1977 suggested that excising misincorporated uracil also makes short fragments, so a uracil-excision-deficient strain gives a cleaner test.

## Going deeper

In a replisome rebuilt from purified budding-yeast proteins, Pol δ helped establish leading strands before handing off to Pol ε, so the usual division is dynamic.

Work published in 2024 added a second problem on the lagging strand: in budding-yeast replisomes rebuilt on linear DNA, no primers were laid on the overhang, and lagging-strand synthesis stopped in a zone more than 26 nucleotides from the template end. CST with Pol α-primase supplies the C-strand fill-in. This refines, rather than overturns, the classical last-primer model.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/dna-replication/
