DNA Repair

A genome under continuous attack

The genome is not a stable artefact. Water hydrolyses the glycosidic bond between bases and the sugar-phosphate backbone, and a typical human cell loses about 10,000 purines a day to spontaneous depurination. Cellular metabolism produces reactive oxygen species that oxidise guanine to 8-oxoguanine — another roughly 10,000 lesions per cell per day. Cytosines spontaneously deaminate to uracil at a low but steady rate. UV-B from the sun crosslinks neighbouring pyrimidines into cyclobutane dimers; ionising radiation and replication-fork collapse generate double-strand breaks; alkylating chemicals add methyls and ethyls to bases at random. Even the cell's own DNA polymerases misincorporate a base every ~10⁵ insertions during S phase.

If none of this were repaired the cell would be lost in a few divisions. Instead, each lesion class is handled by a dedicated pathway whose enzymes recognise the chemistry of the damage, excise the affected stretch, copy a template if available, and ligate the strand back together. The pathways overlap and back each other up — most lesions can be handled by more than one route — but each has a primary substrate and a characteristic phenotype when it fails.

The six major pathways at a glance

Pathway	Substrate	Cell-cycle phase	Key proteins	Accuracy	Failure phenotype
Base excision repair (BER)	Oxidation, alkylation, deamination, depurination — small base damage	All phases	OGG1, UDG, APE1, Polβ, Lig3, XRCC1, PARP1	High	Mutator phenotype; MUTYH-associated polyposis
Nucleotide excision repair (NER)	UV photoproducts, bulky adducts (cisplatin, BPDE)	All phases (TC-NER on transcribed strand)	XPA-G, ERCC1, RPA, TFIIH, Polδ/ε	High	Xeroderma pigmentosum; Cockayne syndrome (TC-NER)
Mismatch repair (MMR)	Replication mismatches, small indels at slippery repeats	S phase, post-replication	MSH2, MSH6, MLH1, PMS2, EXO1, PCNA	Very high	Lynch syndrome; MSI-H tumours
Homologous recombination (HR)	Double-strand breaks, interstrand crosslinks, collapsed forks	S and G2 only (sister chromatid required)	MRN, BRCA1, BRCA2, RAD51, PALB2, RPA	Very high (template-based)	BRCA1/2 cancers; Fanconi anaemia
Non-homologous end joining (NHEJ)	Double-strand breaks, V(D)J recombination	All phases (dominant in G0/G1)	Ku70/80, DNA-PKcs, Artemis, XRCC4, Lig4	Error-prone (small indels)	SCID (V(D)J failure); chromosomal translocations
Translesion synthesis (TLS)	Lesions at the replication fork (UV dimers, abasic sites)	S phase	Polη, Polκ, Polι, Rev1, PCNA-Ub	Low (mutagenic but life-saving)	XP variant (Polη loss); UV-induced skin cancer

Two patterns are worth noting. First, only HR is strictly cell-cycle-restricted; everything else operates whenever needed. The reason is that HR uses the sister chromatid as a template, and sisters only exist after replication — in S and G2. Second, only TLS and NHEJ are explicitly error-prone. The cell tolerates that error-proneness because the alternative — leaving a stalled fork or a free DNA end overnight — is worse than absorbing a few mutations.

Base excision repair: the workhorse of small damage

Most of the damage a cell sees is small: a single base oxidised, deaminated, depurinated, or alkylated. BER handles it. A lesion-specific glycosylase (OGG1 for 8-oxoguanine, UDG for uracil, MUTYH for mismatched A:8oxoG, MPG for alkylation damage) flips the damaged base out of the helix and snips it off, leaving an abasic site. APE1 nicks the backbone next to the abasic site. DNA polymerase β fills the single-nucleotide gap. DNA ligase III, with its scaffold partner XRCC1, seals the nick. The whole process takes seconds and goes silently.

PARP1, an enzyme that decorates broken DNA with poly(ADP-ribose) chains, is the recruiting flag for BER at single-strand breaks. Inhibiting PARP traps it on damaged DNA and converts the unrepaired SSB into a double-strand break at the next replication fork — useful because BRCA-deficient tumours then cannot repair the resulting DSB by HR and die. This is the textbook synthetic lethality that olaparib and its peers exploit.

Nucleotide excision repair: bulky lesions, two-step incision

NER handles lesions that distort the helix shape — UV-induced thymine dimers, cisplatin adducts, polycyclic aromatic hydrocarbon adducts. It removes a stretch of about 27-29 nucleotides containing the lesion, rather than a single base. The damage is recognised either globally (by XPC-RAD23B scanning the genome) or transcription-coupled (by stalled RNA polymerase II handing off to CSB and CSA, the proteins lost in Cockayne syndrome). The TFIIH complex then unwinds the helix, XPA verifies the lesion, RPA coats the single-stranded patch, and the endonucleases XPF-ERCC1 and XPG cut on either side of the damage. Polymerase δ or ε fills in, and ligase I seals.

Loss of any of the seven XP genes (XPA-XPG) causes xeroderma pigmentosum: extreme UV sensitivity, hundreds of skin cancers in childhood, and progressive neurodegeneration. The disease is the cleanest demonstration that you cannot live without a functional NER pathway in a sun-exposed body.

Mismatch repair: cleaning up after the polymerase

Replicative polymerases are accurate to about 1 error per 10⁵ bases. With a 6-billion-base diploid genome, that would mean tens of thousands of mismatches per S phase if MMR did not clean up afterwards. MutSα (MSH2-MSH6) recognises base-base mismatches and small loops; MutSβ (MSH2-MSH3) handles larger insertions/deletions. They recruit MutLα (MLH1-PMS2), an endonuclease, which nicks the newly synthesised (daughter) strand. EXO1 then resects from the nick to the mismatch, polymerase δ resynthesises with the parental strand as template, and ligase I seals.

How does the cell know which strand is the new one? In E. coli, by transient hemimethylation of GATC sites. In humans, by the nicks already present on the lagging strand and by PCNA orientation on the leading strand. Lose any of MSH2, MSH6, MLH1, or PMS2 and you get Lynch syndrome — accelerated colorectal, endometrial, gastric, and urothelial cancer with high microsatellite instability and an unusually strong response to PD-1 immune checkpoint inhibitors, because the high mutational load generates many neoantigens.

Double-strand breaks: HR vs NHEJ

A double-strand break is the most dangerous lesion a cell can face — both strands are broken, no template remains across the gap, and an unrepaired DSB causes either chromosome loss or fusion to another DSB. Two pathways compete: homologous recombination, which is accurate but requires a sister chromatid, and non-homologous end joining, which is fast but error-prone.

The choice is regulated at the resection step. Initially the MRN complex (MRE11-RAD50-NBS1) binds both ends and removes a few nucleotides. In S/G2, CDK-driven phosphorylation activates BRCA1 to displace 53BP1 and license long-range resection by CtIP and EXO1, exposing 3' single-stranded tails. RPA coats those tails; BRCA2 (with PALB2) loads the recombinase RAD51. The RAD51 filament invades the sister chromatid, copies across the break, and the resulting Holliday junction is resolved without information loss. In G0/G1, 53BP1 wins instead; resection is blocked, Ku70/80 binds the blunt ends, DNA-PKcs assembles the synaptic complex, Artemis trims overhangs, and XRCC4-Lig4 ligates. The joint usually carries a small indel — which is exactly why CRISPR-Cas9 knockouts work: a Cas9-induced DSB at the gene of interest is repaired by NHEJ, and the resulting frameshifts disrupt the reading frame.

Translesion synthesis: the controlled mutagenic rescue

Sometimes the replication fork meets a lesion — a UV dimer, an abasic site, a cisplatin adduct — that the high-fidelity polymerase δ or ε cannot accommodate. Stalling at that fork is dangerous: collapsed forks become double-strand breaks. The cell uses translesion synthesis polymerases — Polη, Polκ, Polι, Rev1 — that have larger, more accommodating active sites and can copy past damage. They are also dramatically more error-prone than replicative polymerases. The switch to a TLS polymerase is signalled by mono-ubiquitination of PCNA at K164, which the TLS polymerases recognise; once past the lesion, the replicative polymerase resumes.

The trade-off is mutation in exchange for survival. Polη is specifically good at reading across cyclobutane pyrimidine dimers correctly; lose it and you cannot copy past sun-induced lesions accurately, leading to xeroderma pigmentosum variant — a UV-sensitive skin-cancer syndrome with the NER pathway still intact.

Where DNA repair lands in the clinic

• BRCA1/2 and PARP inhibitors. Carriers of pathogenic BRCA1 or BRCA2 variants face 50-80% lifetime breast cancer risk and 15-40% ovarian cancer risk. Their tumours are HR-deficient. Olaparib (FDA 2014), rucaparib, niraparib, and talazoparib exploit synthetic lethality with HR loss and have transformed maintenance therapy for BRCA-mutant ovarian, breast, prostate, and pancreatic cancers.
• Lynch syndrome (MMR loss). Carriers face 50-80% lifetime colorectal cancer risk. Their tumours are MSI-high and respond extraordinarily well to pembrolizumab and other checkpoint inhibitors — so well that pembrolizumab has been approved as first-line therapy for any MSI-H solid tumour regardless of tissue of origin.
• Xeroderma pigmentosum (NER loss). Children develop hundreds of skin cancers; affected families avoid sunlight entirely. The condition demonstrates that the daily UV burden on human skin requires functional NER for survival past childhood.
• Fanconi anaemia. Loss of any of 22+ FA proteins (FANCA, FANCC, BRCA2/FANCD1) impairs interstrand crosslink repair. Patients show progressive bone-marrow failure, congenital anomalies, and 500-fold increased leukaemia risk.
• Radiation and chemotherapy. Most cytotoxic chemotherapy works by overwhelming repair pathways: cisplatin generates intrastrand crosslinks (NER substrate), doxorubicin traps topoisomerase II producing DSBs, gemcitabine inhibits ribonucleotide reductase. Tumours selected for repair-pathway loss become hypersensitive to the matching cytotoxic — an active area of personalised oncology.

Variants of the basic logic

• Direct reversal. Some lesions are reversed without excision: MGMT removes a methyl group from O6-methylguanine in a single step (and is itself destroyed in the process). Bacterial photolyase reverses UV dimers using blue light energy.
• Microhomology-mediated end joining (MMEJ / alt-NHEJ). A backup DSB pathway that uses 5-25 bp of microhomology to align broken ends. Always produces deletions; particularly active when classical NHEJ is impaired and is one route by which BRCA-deficient tumours generate the genomic scars that PARP inhibitors detect.
• Single-strand annealing (SSA). When a DSB occurs between repeated sequences, the cell can resect both ends, anneal the repeats, and ligate — collapsing the intervening sequence into one. Fast but always deletes information.
• Interstrand crosslink repair. A coordinated dance between the FA pathway, NER endonucleases, TLS, and HR. The two strands cannot be unzipped, so the repair must be done within a stalled replication fork.
• Mitochondrial repair. Mitochondria run their own BER (with mitochondrial-targeted versions of OGG1, UDG, Polγ) but no NER, no MMR, no HR. The asymmetry is one reason mitochondrial DNA accumulates mutations 10-20× faster than nuclear DNA.

Pitfalls and easy misreadings

• "DNA repair prevents all mutations." It cannot. Repair operates after damage and cannot restore information that was identical on both strands when the lesion arrived (a missed mismatch in MMR, a wrong base copied by TLS). The cell trades fidelity for survival routinely.
• "NHEJ is the bad pathway." NHEJ is essential — it is the only DSB pathway available in G0/G1, it is what allows V(D)J recombination to generate antibody and TCR diversity, and it is the cell's default fallback. It is error-prone but it is not optional.
• "BRCA1 and BRCA2 do the same thing." Both are HR proteins but they act at different steps. BRCA1 promotes resection (and antagonises 53BP1); BRCA2 loads RAD51 onto resected ssDNA. BRCA2 mutations also cause Fanconi anaemia subtype D1.
• "More repair is always better." Hyper-active TLS or NHEJ over HR drives mutagenesis and translocations. Overexpression of POLQ (microhomology-mediated end joining) is associated with worse cancer prognosis.
• "Checkpoint failure equals cancer immediately." A mutator phenotype takes many divisions to convert a single cell into a tumour. Lynch syndrome carriers, despite a 100-1000-fold elevation in mutation rate, develop cancers at 30-50 years, not in childhood — most pre-cancer cells either die or are eliminated by immune surveillance long before they form a tumour.