Replication: Copying Three Billion Letters
Unzip the ladder, and each half specifies its own replacement. The proofreading is what makes it astonishing.
Builds on: 2.1 DNA: Four Letters, One Molecule
The mechanism is in the structure
Watson and Crick’s paper ends with one of the most restrained sentences in science: they had not failed to notice that the pairing they proposed immediately suggests a copying mechanism. It does. Unzip the two strands, and each one carries the full instructions for rebuilding its partner.
The process, in order:
- Helicase unwinds the helix and breaks the hydrogen bonds, opening a Y-shaped replication fork.
- DNA polymerase runs along each exposed strand, matching free nucleotides to the template — A opposite T, C opposite G.
- The result is two double helices, each made of one old strand and one new one. That is why replication is called semi-conservative.
The awkward compromise
Here is where being antiparallel bites. DNA polymerase can only build in one direction: 5′→3′. At a fork, one template runs the convenient way, so its new strand is built smoothly and continuously — the leading strand.
The other template runs the wrong way. Polymerase cannot reverse, so it does something distinctly inelegant: it waits for a stretch to open, builds a short fragment backwards relative to the fork, jumps ahead, and does it again. Those pieces (Okazaki fragments) are then stitched together by ligase. This is the lagging strand, and it is a genuine kludge — the sort of thing that makes far more sense as the outcome of evolution than of design.
How accurate is it?
DNA polymerase makes roughly one mistake per 100,000 bases — which sounds fine until you remember the genome is 3 billion bases, giving about 30,000 errors per copy. So the enzyme proofreads: it checks each base it has just added and excises wrong ones. A separate mismatch-repair system sweeps through afterwards for anything missed.
That is an error rate of about one in a billion — comparable to copying every book in a large library by hand and making a single typo. And the residual errors are not purely a defect: they are mutations, and without them there would be no variation for evolution to act on. Copy too sloppily and the organism fails; copy perfectly and the lineage cannot adapt.
Three models were on the table: conservative (old helix intact, new one built fresh), semi-conservative, and dispersive (fragments mixed). In 1958 Meselson and Stahl grew bacteria on heavy nitrogen, switched them to light nitrogen and spun the DNA in a density gradient. After one round every molecule was intermediate in density — killing the conservative model. After two rounds there were two bands, half intermediate and half light — killing the dispersive one. It is often called the most beautiful experiment in biology, and it fits on one page.
⚗️ Lab — The Replication Fork
Drag the fork open and watch both new strands appear.
- The top strand is built continuously; the bottom one comes in fragments. That is the antiparallel problem made visible.
- Hide the lagging strand to see how tidy replication would be if polymerase could work in both directions.
- Every finished helix is one old strand plus one new — semi-conservative.
Work it by hand — 0 / 2
No multiple choice here: compute the value and type it. Suffixes like 2.5k or 20m are understood; answers within ±2% count.