Damage is the rule, not the exception
You arrive here already knowing the bad news. The last two guides showed that a mutation is any change in the DNA sequence, and that the genome is under constant assault — replication slips that escape the polymerase, spontaneous chemistry like the loss of a base or the deamination of cytosine, ultraviolet light welding neighbouring bases into a thymine dimer, radiation snapping strands, and chemical mutagens that disguise one base as another. Tot it all up and a single human cell suffers tens of thousands of damage events *every day*. If even a fraction stuck, the instructions for building you would dissolve into noise within a generation.
And yet your genome is astonishingly stable. The resolution to that paradox is the subject of this guide: cells do not tolerate damage, they *repair* it, all day, every day, mostly without you ever knowing. Here is the single idea that organizes everything below. There is no one universal repair enzyme that fixes everything. Instead the cell keeps a toolbox, and each tool is matched to a particular *kind* of damage — a single wrong base, a chemically scarred base, a bulky lump that distorts the helix. The art of repair is first recognising what sort of damage you are looking at, then calling the right pathway. Let us open the toolbox and meet the tools one by one.
Mismatch repair: who do you trust?
Start with the most common error: a replication slip. The replicative polymerase is fast and careful, and its built-in proofreading catches most of its own mistakes. But roughly once in ten million bases a wrong base still slips through and gets paired in — an A sitting opposite a C, say, where the geometry is subtly off but not catastrophic. This is exactly the chemically-perfect-but-wrongly-paired case. Mismatch repair is the pathway that sweeps up behind the replication fork and catches these survivors.
But mismatch repair faces a deep puzzle, and seeing the puzzle is the whole point. Suppose the enzymes find an A paired with a C. One of those two bases is the original, correct one, and the other is the freshly mis-inserted mistake — but *which is which*? Both are ordinary bases; neither carries a label saying "I'm the error." If the cell guessed wrong half the time and corrected the *template* strand instead of the new one, repair would be no better than a coin flip — it would lock in the mistake as often as it fixed it. So the pathway needs a way to know which strand is brand new.
Here is where the semiconservative design from the last rung pays off beautifully. There is always one old strand you can trust — the template — and the trick is just to mark it. In *E. coli*, the old strand carries chemical tags called methyl groups on certain A's, added slowly some minutes after the strand is built. So right after copying there is a brief window where the parent strand is already methylated and the new strand is not yet. The repair machinery reads that difference: "the unmethylated strand is the new one, so the error must be on it." It then excises a patch of the new strand spanning the mismatch and lets the polymerase re-fill it against the trusted template. Human cells use a different, less fully understood signal — strand breaks and the proteins still sitting at the fork seem to flag which strand is new — but the logic is identical: mismatch repair always corrects the *new* strand, because the old one is the reference of record.
Excision repair: cut out the bad part, copy back from the good
The next two tools share one gorgeous trick, and it is worth naming before the details: in a double helix, the two strands are complementary, so if one strand is damaged, the *other* strand still holds a perfect copy of the lost information. That means the cell can afford to be ruthless — it can simply cut the damaged piece clean out and rebuild it by reading the intact partner as a template. "Excision" just means cutting out. The two excision pathways differ only in *how big a piece* they remove, which in turn depends on how big the damage is.
Base excision repair handles the small stuff: a single base that has been chemically altered but has not deformed the helix — a cytosine that lost its amino group and turned into uracil, a base nicked by oxidation, the everyday wear-and-tear of spontaneous damage. The pathway is almost surgical in its precision. A specialised enzyme called a glycosylase recognises one specific kind of wrong base and snips it off its sugar, leaving the backbone intact but with a gap where the base used to hang. Other enzymes then trim the bare spot, and the polymerase drops in the single correct base read off the opposite strand, after which a DNA ligase seals the final nick. One base out, one base in. There is a different glycosylase for each common kind of damaged base — yet another layer of the dedicated-tool theme, nested inside base excision repair itself.
Nucleotide excision repair handles the big, bulky stuff: lesions that physically distort the helix, with the classic example being the thymine dimer that ultraviolet light creates by fusing two neighbouring T's into a kinked, locked-together lump. A base glycosylase is no good here — the damage is not a single recognisable base, it is a gross distortion of the backbone's shape. So this pathway recognises the *bend* rather than the chemistry. Once it senses a kink, it cuts the damaged strand on both sides of the lesion and lifts out a whole short stretch — in humans, a patch of roughly two to three dozen nucleotides — dimer and all. The polymerase then re-synthesises the gap from the undamaged strand, and ligase seals it. Because it reads shape, not a specific molecule, nucleotide excision repair is the cell's general-purpose tool for almost any large, helix-warping adduct, whatever chemical happened to cause it.
BASE EXCISION REPAIR (one altered base; helix not distorted) 5'-...A T C[U]G A...-3' U = damaged base (was C, deaminated) 3'-...T A G G C T...-5' glycosylase snips out the single bad base -> gap -> polymerase fills 1 base -> ligase seals NUCLEOTIDE EXCISION REPAIR (bulky lesion; helix kinked) 5'-...A T[T=T]G C A...-3' T=T = thymine dimer (UV-fused), strand bent 3'-...T A A A C G T...-5' cut on BOTH sides -> lift out a ~24-30 nt patch -> polymerase re-fills the gap -> ligase seals In both: the intact opposite strand is the template. Cut out the bad, copy back the good.
Direct reversal: just undo it
The fourth tool is the simplest of all, and the rarest. Sometimes a chemical change to DNA can be reversed directly — the very bond that the damage created is simply broken again, restoring the original base with no cutting and no copying. The textbook example is photolyase, an enzyme found in many bacteria, plants, and animals (though, tellingly, not in placental mammals like us) that uses the energy of visible light to pry a thymine dimer back apart into two separate, normal T's. No strand is opened, no patch removed; the lesion is just *undone*, in a single chemical step.
Direct reversal is fast and error-free — there is no gap to fill, so there is no chance of filling it wrong — but its reach is narrow. An enzyme that simply undoes one bond can only handle the one specific lesion that bond defines; you cannot "un-make" a base that has been broken into pieces, or a missing base, by reversal. So direct reversal exists only for a handful of damages, and it is precisely the cell's choice of when *not* to use it that illustrates our theme. Humans, lacking photolyase, must instead send thymine dimers to nucleotide excision repair — the slower, costlier cut-and-patch route. The cell does not use one favourite tool everywhere; it picks the tool that fits the lesion in front of it, and which tools it even owns varies from species to species.
When the toolkit breaks: repair and disease
The cleanest proof that these pathways matter is what happens when one of them is broken from birth. Xeroderma pigmentosum is an inherited condition in which the genes for nucleotide excision repair are defective. Sunlight still creates thymine dimers in the skin exactly as it does in everyone — but these patients cannot cut them out. The dimers pile up, force error-prone copying, and seed mutations at a furious rate, so affected children develop extreme sun sensitivity and skin cancers thousands of times more often than usual, often before the age of ten. The lesson is stark: the damage is normal and universal; only the missing tool turns it into catastrophe. The repair system is not a luxury, it is what stands between ordinary sunlight and runaway mutation.
The same story plays out for mismatch repair. People who inherit a broken copy of a mismatch-repair gene have Lynch syndrome (hereditary nonpolyposis colorectal cancer), one of the most common forms of inherited cancer predisposition. Without proper mismatch repair, replication slips accumulate, especially in short repeated sequences that the polymerase tends to stutter over — a measurable signature called microsatellite instability. The connection is general and worth holding onto: most cancers are at heart diseases of accumulated mutation, so anything that lets mutations build up unchecked — a missing repair pathway — raises cancer risk. Repair failure does not cause cancer directly; it removes a guardrail, and then ordinary, ongoing damage does the rest.
Two honest qualifications before the next guide. First, this toolkit is not the whole story: we have deliberately left out the most dangerous lesion of all, a break clean across *both* strands, which has no undamaged template to copy from and demands its own dramatic machinery — that is exactly the subject of the next guide. Second, beware the tempting but false idea that repair is about *perfection*. It is not. As the mutation guide stressed, most surviving changes are neutral, variation is the raw material of evolution, and a cell that erased every change would freeze its own lineage out of the future. Repair keeps the error rate low enough to preserve the message, while leaving it high enough for life to keep exploring. The toolkit's job is not zero mutations; it is the right, survivable amount.