Causes of mutation and DNA repair

In S7.1 you classified what a mutation does once it exists: silent, missense, nonsense, frameshift. This lesson asks the prior question. Where does the change come from in the first place, and given how many things attack DNA, why is your genome not already shredded?

Here is the shape of the answer, in terms you already trust. Copying 3 billion bases is a lossy operation, so replication ships with built-in error checking, like a checksum on every write plus an automatic retry when the check fails. On top of that the cell runs a set of background maintenance daemons, repair enzymes that patrol the DNA and patch corruption after it happens. Mutations are the corruption events that slip past both layers. And aging, we will see, is partly those daemons falling behind the damage.

Four ways DNA changes

Start with the sources, because each one has a mechanism you can reason about, not just a name.

Copy errors during replication. When DNA is duplicated, a polymerase walks the old strand and pairs a fresh base against each one using the complementarity rules from S3, A with T, G with C. But pairing is a physical, probabilistic fit, not a lookup in a table, so now and then the enzyme seats the wrong base. The raw error rate is roughly 1 wrong base per 100,000, which sounds tiny until you multiply by 3 billion bases per genome copy. Uncorrected, that would be tens of thousands of new errors every division. Copying is the single largest source of raw mistakes, purely because there is so much of it.

Spontaneous chemical damage. DNA is not inert. It sits in water, and water plus body heat slowly break it. One base, cytosine, spontaneously loses an amino group and turns into uracil (a base that does not belong in DNA) on the order of a hundred times per cell per day. Bases also fall off the backbone entirely thousands of times a day, leaving a gap. Nobody attacked the cell. This is ordinary chemistry grinding on a fragile molecule around the clock.

Radiation. Ultraviolet light in sunlight carries enough energy to fuse two adjacent bases into a bulky knot that kinks the helix and blocks copying. Higher-energy ionizing radiation, X-rays and gamma rays, can snap the sugar-phosphate backbone outright, sometimes cutting both strands at once. Radiation does not invent a new alphabet. It deposits energy that breaks bonds already there.

Reactive metabolic byproducts. This is the one to remember for later. Making energy from oxygen, the metabolism you met earlier, leaks a trickle of reactive oxygen species, hot fragments like the hydroxyl radical that grab electrons from whatever they touch, including DNA. A common result is an oxidized guanine that mispairs during the next copy. The unsettling part is the source. This damage comes from the inside, from your own power plants running normally, every second you are alive.

The repair crew, layer by layer

The cell fights back with overlapping systems, and the layering is the point. Each one catches what the previous one missed.

Proofreading (checksum plus retry). The replication polymerase carries its own 3 prime to 5 prime editing head. Right after it inserts a base, it senses a bad fit, backs up one step, snips out the wrong base, and tries again. This alone improves fidelity roughly a hundredfold. It is a genuine checksum-and-retry, running inline as the strand is written.

Mismatch repair (a second pass). A few wrong bases still slip through. So a separate crew sweeps the freshly made DNA, finds the spots where the two strands do not pair cleanly, cuts out the offending stretch on the new strand, and resynthesizes it against the old one. To do this it must tell which strand is the newly written one, the copy that could be wrong, versus the original template. That layer buys another hundred to thousandfold. Stack the three (raw, then proofreading, then mismatch repair) and the surviving error rate falls to about 1 in a billion.

Excision repair (single-strand patches). For chemical and radiation damage, specialized enzymes recognize a specific kind of bad base or a bulky lesion, flip it out, cut a short window around it, and refill from the intact partner strand. The undamaged strand is the source of truth. Because DNA is double and complementary, one good copy is almost always sitting right there to copy back from.

Double-strand break repair (the hard case). Now the dangerous one. If both strands are cut at the same place, there is no intact partner to copy from. The information is gone on both sides. The cell has two options and they trade off. One route waits until after replication and uses the identical sister copy of that chromosome as a template, which is accurate. The other just grabs the two loose ends and glues them back together, fast but error-prone, often losing or adding a few bases at the seam. Double-strand breaks are the scary lesions precisely because the redundancy that saves every other repair, a good strand to copy from, is exactly what a double cut destroys.

The daemon analogy and where it breaks

So the software picture is fair. Replication runs a checksum with retries (proofreading), background daemons scrub corruption (excision and mismatch repair), and a lifetime of the daemons losing ground to the damage rate is part of what aging looks like.

Now the failure edge, because an analogy without its limit is a bug. In software, a checksum plus retry can always re-fetch a clean copy from an authoritative source, and a daemon that restarts comes back exactly as strong as before. DNA repair has neither guarantee. When both strands are lost in a double-strand break, there is no authoritative source to re-request, only a best guess or a lossy splice. And the daemons themselves are encoded in the same DNA they protect, so damage to a repair gene degrades the very machinery doing the repairing, a feedback loop with no clean equivalent in a stateless retry. The analogy gets you to the door. It does not walk you through it.

Germline versus somatic: only one is inherited

There is one distinction here that beginners get wrong constantly, and it decides whether a mutation matters to one body or to a whole lineage.

That split also tells you why cancer is a somatic disease of accumulation. A single cell picks up, over years, a run of somatic mutations that break the controls on when it is allowed to divide, and its descendants outgrow their neighbors. It is the same mutation machinery from this lesson, aimed inward, playing out in one lineage of cells rather than being copied into a child.

Key terms

mutation

A permanent change in the DNA sequence that survives into the next copy, as opposed to transient damage.

DNA damage

A physical lesion in DNA (a wrong base, a broken bond, a fused pair) that becomes a mutation only if it is copied before repair or fixed incorrectly.

reactive oxygen species

Hot chemical fragments leaked by normal oxygen metabolism that attack DNA and other molecules from the inside.

proofreading

The replication polymerase's inline ability to sense a wrong base, remove it, and retry, like a checksum with automatic retry.

mismatch repair

A second-pass system that scans freshly copied DNA for mispaired bases the polymerase missed and rewrites the new strand.

excision repair

Systems that cut out a damaged base or bulky lesion and refill the gap using the intact complementary strand as template.

double-strand break

A cut through both strands at once, the hardest lesion to repair because no intact partner strand remains to copy from.

germline versus somatic

Germline mutations sit in egg or sperm cells and are inherited by offspring, while somatic mutations sit in body cells, drive cancer and aging, and are not inherited.

Why the mutation rate is tuned, not minimized

You might expect evolution to drive the error rate to zero. It does not, and the reason is a trade-off. Repair costs energy and enzymes, and each added layer of fidelity buys less than the one before it. Past a point, spending more to prevent the next rare error is not worth it. There is a second, subtler reason. Mutation is also the raw material of variation, the thing S7.3 shows selection acts on. A lineage with zero mutation cannot adapt to anything new. So the rate we measure is a settled compromise between the cost of errors and the cost of preventing them, held at a level low enough to keep genomes stable across generations but not so low that change becomes impossible. Zero was never the target.