Genetics and Molecular Biology · Unit 3: DNA: copying, changing, repairing

DNA Repair and Recombination

DNA is damaged all the time, by its own chemistry, by light and by the polymerase that copies it. This lesson sorts the damage into kinds and matches each kind to the system that fixes it: direct reversal, base excision, nucleotide excision, mismatch repair, and the two ways of mending a broken chromosome. Afterwards you can calculate how much mismatch repair lowers the error rate of replication, explain how a cell tells the new strand from the old, and connect xeroderma pigmentosum, Lynch syndrome and BRCA1 or BRCA2 variants to the repair step each disrupts.

Watch first DNA Holds Its Information Twice · 3:08

Every day, in every one of your cells, some cytosines lose their amino group and turn into uracil, a base that pairs with adenine.

Water and body temperature are enough. If the uracil stayed, the next replication would put an A opposite it, and the round after that would leave an A·T pair where a G·C pair had been: a mutation made by chemistry alone. This lesson asks how cells find a few wrong bases among billions of right ones, and what happens to people who inherit a broken repair system.

Damage is constant: the lesions

A lesion is any change to DNA’s chemical structure, and it is not yet a mutation. It becomes a mutation, a heritable change in sequence, when replication copies it into a normal-looking base pair on both strands.

The lesions fall into a few kinds:

  • Spontaneous chemistry. Cytosine deaminates to uracil, and purine bases fall off their sugars, leaving an abasic (AP) site, a sugar with no base.
  • Oxidation and alkylation. Reactive oxygen produces 8-oxoguanine, which can pair with adenine. Alkylating agents add methyl groups; O6-methylguanine pairs with thymine.
  • Ultraviolet light. UV joins two neighboring pyrimidines on the same strand into a cyclobutane pyrimidine dimer or a 6-4 photoproduct. Both bend the helix and stall polymerases.
  • Strand breaks. Ionizing radiation and collapsed replication forks can break both strands at once.
  • Replication errors. The polymerase pairs normal bases wrongly or slips on repeats (see mutation).

One rule runs through almost all of repair: the double helix holds its information twice. If one strand is damaged, a cell can cut out the bad stretch and copy the partner strand, the same templating used in DNA replication. A break through both strands removes that option at one spot, which is why it needs separate machinery.

Reversing and cutting out: direct reversal, BER and NER

Direct reversal undoes the chemistry in place, without cutting the backbone. Photolyases use blue and near-UV light to split pyrimidine dimers back into separate bases. They work in bacteria, yeast and plants, but humans lack the photolyases that reverse UV dimers, so your skin cannot use this route. The second example is O6-methylguanine methyltransferase (MGMT). It moves the methyl group from the guanine to a cysteine in its own active site, which restores the base and inactivates the protein for good. One molecule repairs one lesion, so it is called a suicide enzyme.

Base excision repair (BER) handles most small, non-bulky base lesions that direct reversal does not: uracil, 8-oxoguanine, some alkylated bases, AP sites. It runs in steps:

  1. A DNA glycosylase specific for one kind of damaged base flips that base out of the helix and cuts the bond to its sugar, leaving an AP site.
  2. AP endonuclease cuts the backbone at the AP site.
  3. A polymerase (polymerase beta in human cells) inserts the correct nucleotide, using the partner strand.
  4. A ligase seals the nick.

In 1974 Tomas Lindahl reported an E. coli activity that released uracil from deaminated DNA and left the backbone uncut. In 1996 his group rebuilt human uracil repair from four purified proteins, which filled a one-nucleotide patch.

Nucleotide excision repair (NER) handles bulky lesions that distort the helix: UV dimers and 6-4 photoproducts, and adducts from drugs such as cisplatin. It senses the distortion, not the chemistry, so one machine copes with many lesions. It cuts the damaged strand on both sides and removes a short oligonucleotide. Aziz Sancar and Dean Rupp showed this dual incision in 1983 with the purified E. coli proteins UvrA, UvrB and UvrC. The E. coli fragment is about 12 to 13 nucleotides; in a human cell-free system the removed piece was 27 to 29 nucleotides long. A polymerase then fills the gap, and a ligase seals it.

Fixing the copier’s mistakes: mismatch repair

A mismatch is two normal bases paired wrongly, such as G with T. Neither base is damaged, so which one is wrong? The answer is whichever is on the new strand, because the template was correct before the polymerase arrived.

In E. coli the new strand carries a mark of absence. The Dam methylase adds a methyl group to the adenine in every GATC sequence, but only some time after the fork passes. For that window, new DNA is hemimethylated: methylated on the template strand only. MutS binds the mismatch, MutL joins it, and together they activate MutH, which nicks the unmethylated strand at a GATC site. An exonuclease removes the new strand from the nick past the mismatch, and the polymerase and ligase fill in. In Paul Modrich’s purified system a GATC site about 1 kilobase away still directed repair.

That GATC signal belongs to E. coli and its relatives. It is unusual, not universal. Eukaryotes keep MutS and MutL relatives (MSH2 paired with MSH6, and MLH1 paired with PMS2 in humans) but have no MutH. They appear to read strand breaks instead: new DNA has free ends, at the tip of the growing strand and between Okazaki fragments. A purified human system could be directed by a nick on either side of a mismatch. Human mismatch repair also corrects small loops left when the polymerase slips on a repeat.

Worked problem. The numbers here are round and chosen for illustration, not measured for one polymerase. Assume a replicative polymerase inserts a wrong base about once per 10^5 nucleotides, proofreading removes 99% of those errors, mismatch repair removes 99.9% of what is left, and a human diploid cell copies about 6.4 x 10^9 base pairs per S phase.

Step 1. Turn each percentage into a fraction that survives. Proofreading lets 1% through, 10^-2. Mismatch repair lets 0.1% through, 10^-3.

Step 2. Multiply the stages, because each acts on what the last one missed.

final error rate = 10^-5 x 10^-2 x 10^-3 = 10^-10 per base pair

Step 3. Scale to the genome.

6.4 x 10^9 bp x 10^-10 = 0.64 errors per genome per replication

Step 4. Remove mismatch repair. The rate is now 10^-5 x 10^-2 = 10^-7 per base pair, and 6.4 x 10^9 x 10^-7 = 640 errors.

So with these numbers, losing mismatch repair raises the error rate a thousandfold, from under one error per division to hundreds.

Repair pathway chooser: match the damage to the fix⚙ original · interactive
Kind of damage
Cell-cycle phase (breaks only)
Route to follow (breaks only)

    Match the repair to the damage: one helix strand is damaged, so cut it out and copy the partner; a break through both strands leaves no partner, so join the ends or splice from the sister. This is a sketch of the logic, not a timeline.

    Broken chromosomes: end joining and homologous recombination

    A double-strand break cuts the chromosome in two. Picture a rope cut through: you can knot the ends, or splice in a matching length from an identical rope. Cells do both.

    Non-homologous end joining (NHEJ) knots the ends. The ring-shaped Ku protein binds each end, recruits the kinase DNA-PKcs, and nucleases such as Artemis and polymerases trim or fill ends that do not fit. XRCC4 with DNA ligase IV joins them. NHEJ uses no template, so if the ends had to be processed, a few bases may be gained or lost at the joint. Calling it error-prone overstates this: how accurate it is depends on the structure of the ends, and clean, compatible ends are often rejoined exactly. NHEJ works throughout the cell cycle.

    Homologous recombination (HR) splices from a matching rope. Nucleases resect the 5′ ends, leaving 3′ single-stranded tails. A recombinase coats each tail, RecA in E. coli and Rad51 in eukaryotes, and the filament searches for a matching duplex and invades it, forming a displacement loop (D-loop). The invading 3′ end primes DNA synthesis, copying the information lost at the break. The best template is the sister chromatid, the identical copy made during replication, so HR acts mainly in S and G2.

    When both ends engage the template, the two duplexes are linked by two Holliday junctions, four-stranded crossings named for Robin Holliday, who proposed the structure in 1964. Szostak, Orr-Weaver, Rothstein and Stahl put the break first in their 1983 yeast model. Before the chromosomes separate, the junctions are either dissolved, giving no crossover, or cut by resolvases, giving a crossover or not depending on which strands are cut. The same machinery, started by programmed breaks, makes the meiotic crossovers of chromosomes and meiosis that linkage mapping measures. How a CRISPR cut is repaired is taken up in forward and reverse genetics.

    When repair fails: three human conditions

    Unrepaired damage also triggers checkpoints run by p53, described in stress signals. Three inherited conditions show what happens when a repair pathway itself is lost.

    Xeroderma pigmentosum (XP) is autosomal recessive: affected people have two nonworking alleles of one gene. Most forms disrupt NER, so UV dimers in skin cells go unrepaired, and the condition is associated with sunlight-induced skin cancers. Cell fusion sorted patients into several complementation groups, each a different NER gene. One form, the XP variant, has normal excision. Its cells cannot replicate efficiently past UV damage, and in 1999 two groups, Masutani and colleagues and Johnson, Prakash and colleagues, found mutations in the gene for DNA polymerase eta, a bypass polymerase, in XP variant cell lines.

    Lynch syndrome is a dominantly inherited predisposition to colorectal, endometrial and other cancers, with risks that vary by gene. Carriers are heterozygous for a germline variant in MLH1, MSH2, MSH6 or PMS2, or have a deletion at the 3′ end of EPCAM that silences the neighboring MSH2. One working allele is enough for repair, so their normal cells repair mismatches. A tumor arises in a cell lineage that loses the second allele and with it mismatch repair. The link was made in 1993, when Fishel and colleagues and Leach and colleagues each reported MSH2 mutations in such families. The two-hit logic belongs to cancer genetics.

    BRCA1 and BRCA2 were cloned by Miki and colleagues (1994) and Wooster and colleagues (1995). Both are needed for efficient HR: Brca1-deficient mouse stem cells repaired chromosome breaks poorly by HR, and BRCA2-mutant human and mouse cell lines showed 6-fold to more than 100-fold less homology-directed repair, depending on the line. Carriers inherit one nonworking allele in an autosomal dominant pattern of risk for breast and ovarian cancers, and tumors typically lose the remaining allele. HR-deficient cells are unusually sensitive to PARP inhibitors, an example of synthetic lethality.

    Design problem

    Fibroblasts from four unrelated XP patients, P1 to P4, each show low unscheduled DNA synthesis after UV. You fuse cells in pairs and score the fused cells after UV: P1+P2 normal, P1+P3 low, P1+P4 normal, P2+P3 normal, P2+P4 low, P3+P4 normal. How many genes are involved? What would you do next to identify them?

    One way to do it

    Use the logic of complementation and pathways: XP is recessive, so a fused cell repairs normally only if each partner supplies the working gene the other lacks.

    1. P1+P3 stays low, so P1 and P3 are defective in the same gene. P2+P4 stays low, so they share a gene.
    2. Every pair across those groups (P1+P2, P1+P4, P2+P3, P3+P4) is normal, so the two groups are different genes.
    3. Answer: two complementation groups, {P1, P3} and {P2, P4}, and so at least two genes.

    To identify them, transfect cells of each group with cloned candidate NER genes, one at a time, and look for the gene that restores UV-induced repair synthesis in one group but not the other. Then sequence that gene in the group’s patients, expecting two nonworking alleles in each. A patient sensitive to UV but with normal unscheduled DNA synthesis would point instead to the XP variant, a defect in bypass by polymerase eta.

    How we measure it

    Unscheduled DNA synthesis

    Cells outside S phase are irradiated with ultraviolet light and given labeled thymidine. Any label they take up comes from filling the gaps left by excision repair, so the amount measures how well the cells cut out and replace damage.

    Cell-fusion complementation

    Cells from two patients with the same recessive repair defect are fused into one cell with both nuclei. If repair returns, each patient supplies the gene product the other lacks, so their mutations are in different genes; if not, they are in the same gene.

    Reconstitution from purified proteins

    A repair reaction is rebuilt in a test tube from separately purified proteins and a DNA substrate with a defined lesion. It shows which proteins are sufficient for the reaction and lets each step be read separately, though it does not show how the reaction runs on chromatin in a living cell.

    Site-specific break reporter

    The yeast endonuclease I-SceI cuts a recognition site placed once in a chromosome, next to a broken reporter gene. Repair of the break by homologous recombination restores a working reporter, so the fraction of reporter-positive cells measures homology-directed repair at that one site.

    Finish line

    Test yourself

    8 questions

    Practice and MCAT-style, with explained answers.

    Next · Lesson 16

    Transcription: From DNA to RNA

    Which strand of a gene does RNA polymerase read, and which way does it go?

    Continue →

    ← Lesson 14 · Mutation: Where Variation Comes From

    E. coli DNA photolyase — light-driven direct reversal, absent in humans
    E. coli DNA photolyase — light-driven direct reversal, absent in humans ● LIVE · RCSB PDB 1DNP
    open on RCSB PDB ↗
    Loading live structure 1DNP from RCSB PDB…

    Find the FAD cofactor buried in the protein and the groove on the surface where a pyrimidine dimer is thought to be flipped in to meet it. This is a bacterial enzyme; placental mammals do not have one.

    Drag to rotate · scroll or pinch to zoom · right-drag to pan
    Human uracil-DNA glycosylase bound to DNA — the first step of base excision repair
    Human uracil-DNA glycosylase bound to DNA — the first step of base excision repair ● LIVE · RCSB PDB 1SSP
    open on RCSB PDB ↗
    Loading live structure 1SSP from RCSB PDB…

    Look for the nucleotide flipped out of the double helix into the enzyme's pocket. The glycosylase reads one base at a time this way and cuts it from its sugar.

    Drag to rotate · scroll or pinch to zoom · right-drag to pan
    E. coli MutS bound to a G·T mismatch — the mismatch sensor
    E. coli MutS bound to a G·T mismatch — the mismatch sensor ● LIVE · RCSB PDB 1E3M
    open on RCSB PDB ↗
    Loading live structure 1E3M from RCSB PDB…

    Two MutS subunits clamp the DNA, which is sharply kinked at the mismatch. Only one of the two subunits contacts the mismatched base directly.

    Drag to rotate · scroll or pinch to zoom · right-drag to pan
    Human Ku70/Ku80 bound to a DNA end — the first protein of end joining
    Human Ku70/Ku80 bound to a DNA end — the first protein of end joining ● LIVE · RCSB PDB 1JEY
    open on RCSB PDB ↗
    Loading live structure 1JEY from RCSB PDB…

    The two Ku subunits form a ring that the broken DNA end threads through. A ring can slide onto an end but grips no particular sequence, which suits a protein that must bind any break.

    Drag to rotate · scroll or pinch to zoom · right-drag to pan
    James Cleaver

    James Cleaver

    Showed in 1968 that skin fibroblasts from people with xeroderma pigmentosum make far less repair synthesis after ultraviolet light than normal cells, tying a human hereditary disease to a defect in DNA repair.

    Xeroderma pigmentosum was known to physicians as an inherited condition of extreme sensitivity to sunlight, with skin cancers appearing early. What was wrong inside the cells was not known. Cleaver, a radiation biologist, grew fibroblasts from patients and from unaffected people, exposed them to ultraviolet light and measured repair replication: the short stretches of new DNA a cell makes to fill the gaps left when damage is cut out, separate from the bulk copying of S phase. Normal cells made plenty of it. The patients' cells made little. He reported this in Nature in 1968. A year later, in PNAS, he showed that the same cells could still repair breaks in the DNA chain while failing to repair damage to the bases, so the defect was specific. Because the disease is recessive, cells from different patients could now be fused and tested against each other. That test sorted patients into groups, and the groups led to the genes of nucleotide excision repair.

    ✓ verbatim Cleaver, Defective repair replication of DNA in xeroderma pigmentosum, Nature 218:652 (1968); Cleaver, Proc. Natl. Acad. Sci. USA 63:428 (1969).
    Tomas Lindahl

    Tomas Lindahl

    Found in E. coli an enzyme that removes uracil from DNA by cutting the bond between base and sugar, the first step of base excision repair, and later helped define the human methyltransferase and the minimal human base excision pathway.

    Lindahl's starting point was chemistry. Experiments in his laboratory and others' showed that DNA in water is not inert: it slowly loses purine bases, and its cytosines lose their amino groups and become uracil. If that happened inside cells, something had to undo it. In 1974 he reported a partly purified activity from E. coli extracts that released uracil from DNA in which cytosines had been deaminated. It cut the bond between the base and the sugar and left the phosphodiester backbone whole, so the product was DNA with a missing base, an abasic site. That was a new way to repair DNA: remove one wrong base, not a stretch of strand. His group went on to study a human enzyme that takes methyl groups off O6-methylguanine and is used up by the reaction, and in 1996 reported the human base excision steps rebuilt from four purified proteins. He shared the 2015 Nobel Prize in Chemistry with Paul Modrich and Aziz Sancar for mechanistic studies of DNA repair.

    ✓ verbatim Lindahl, Proc. Natl. Acad. Sci. USA 71:3649 (1974); Nobel Prize in Chemistry 2015, nobelprize.org.
    Paul Modrich

    Paul Modrich

    Rebuilt E. coli methyl-directed mismatch repair from purified proteins in 1989, and went on to reconstitute human mismatch repair, showing that a strand break can direct which strand is corrected.

    In the 1970s geneticists had proposed that E. coli tells a newly made strand from its template by methylation: the Dam enzyme adds methyl groups to the adenines of GATC sequences, and it does so a little after the fork passes, so new DNA is briefly unmethylated. Modrich's laboratory at Duke University set out to test the idea with pure parts. In 1989 Robert Lahue, Karin Au and Modrich reported that a set of purified E. coli proteins, including MutS, MutL and MutH, corrected mismatches in a test DNA and chose the strand by its methylation, even when the only GATC signal lay about 1 kilobase from the mispair. The group then turned to human cells. In 1995 Guo-Min Li and Modrich restored mismatch repair to extracts of a colorectal tumor cell line by adding a fraction from HeLa cells, and in 2005 a purified human system was shown to be directed by a nick on either side of the mismatch. The work gave a biochemical basis for the mismatch-repair defects found in hereditary colon cancer.

    ✓ verbatim Lahue, Au & Modrich, Science 245:160 (1989); Li & Modrich, Proc. Natl. Acad. Sci. USA 92:1950 (1995); Nobel Prize in Chemistry 2015, nobelprize.org.
    Robin Holliday

    Robin Holliday

    Proposed in 1964 that recombining chromosomes exchange single DNA strands, forming hybrid DNA and a four-stranded crossed junction, now called the Holliday junction.

    Many fungi, including yeasts and Neurospora, keep all four products of a single meiosis together, so a geneticist can count alleles directly. For a heterozygous locus the count should be two of each. Sometimes it is three to one, as if one allele had been converted into the other. Holliday's 1964 paper in Genetical Research offered a physical model for this gene conversion. He proposed that two homologous chromatids swap single strands, so that a stretch of each duplex has one strand from each parent. Where the alleles differ, that hybrid DNA holds a mismatch, and correcting it can turn one allele into the other. The strands cross at a junction that can be cut in two ways, one giving a crossover of the flanking markers and the other not. In 1983 Szostak, Orr-Weaver, Rothstein and Stahl revised the model to start from a double-strand break and to contain two such junctions. The junction itself was later seen in DNA molecules and carries his name.

    ✓ verbatim Holliday, A mechanism for gene conversion in fungi, Genetical Research 5:282 (1964); Szostak, Orr-Weaver, Rothstein & Stahl, Cell 33:25 (1983).
    Test yourself
    Question 1 of 8 Score 0 / 0

    Which enzyme repairs O6-methylguanine by moving the methyl group onto one of its own cysteines, and is inactivated by doing so?
    Sources & Integrity

    Every structure, slide, simulation, and clip on this page is listed below with the source it came from. None of the 3D structures is drawn by hand or by AI: each one loads by accession ID from RCSB PDB or PubChem. This panel is generated from the lesson's own data file, so what it lists is what the page loads.

    Molecular structures live data

    • E. coli DNA photolyase — light-driven direct reversal, absent in humans — RCSB PDB 1DNP · coordinates CC0 · Escherichia coli DNA photolyase, X-ray crystal structure (Park, Kim, Sancar & Deisenhofer, Science, 1995)
    • Human uracil-DNA glycosylase bound to DNA — the first step of base excision repair — RCSB PDB 1SSP · coordinates CC0 · Human uracil-DNA glycosylase bound to uracil-containing DNA (Parikh et al., EMBO Journal, 1998)
    • E. coli MutS bound to a G·T mismatch — the mismatch sensor — RCSB PDB 1E3M · coordinates CC0 · Escherichia coli MutS with a G·T mismatched DNA (Lamers et al., Nature, 2000)
    • Human Ku70/Ku80 bound to a DNA end — the first protein of end joining — RCSB PDB 1JEY · coordinates CC0 · Human Ku heterodimer with DNA (Walker, Corpina & Goldberg, Nature, 2001)

    Simulations

    • Repair pathway chooser: match the damage to the fix · Original interactive — Biochemistrypedia (Deppmann Lab).

    Scientist profiles & quotes

    • James Cleaver — Cleaver, Defective repair replication of DNA in xeroderma pigmentosum, Nature 218:652 (1968); Cleaver, Proc. Natl. Acad. Sci. USA 63:428 (1969).
    • Tomas Lindahl — Lindahl, Proc. Natl. Acad. Sci. USA 71:3649 (1974); Nobel Prize in Chemistry 2015, nobelprize.org.
    • Paul Modrich — Lahue, Au & Modrich, Science 245:160 (1989); Li & Modrich, Proc. Natl. Acad. Sci. USA 92:1950 (1995); Nobel Prize in Chemistry 2015, nobelprize.org.
    • Robin Holliday — Holliday, A mechanism for gene conversion in fungi, Genetical Research 5:282 (1964); Szostak, Orr-Weaver, Rothstein & Stahl, Cell 33:25 (1983).

    Quotes are verbatim from the cited source. Profile text is original to this site and was checked against the sources cited.

    Animated explainer

    DNA Holds Its Information Twice · narration: an AI narrator voice (Gemini, Charon) reading narration written from this lesson's text · animation: AI-generated in code (Manim) by Claude, a schematic with no drawn molecular structures, plots computed from their equations · captions from the narration script.

    Images & figures

    This lesson carries no raster figures. Every structure is rendered live from RCSB PDB or PubChem, and every plot is drawn from its equation in the browser.

    integrity Machines check the mechanics. The build confirms that every local file exists, that every question's answer key points at a real choice, and that any AI-generated scientific figure names the person who checked it; a weekly job re-resolves every accession ID against its database. Questions are original; MCAT-style items are written for this site and are not real AAMC questions. Banned sources (the Tymoczko textbook, publisher decks, copyrighted figures) are not used.

    Like I'm five

    Your DNA is a very long instruction book, and it gets smudged and torn every day. Your cells have repair crews. One kind wipes off a smudge, another cuts out a bad letter and writes it back in by looking at the matching side of the page, and another tapes torn pages back together. When a crew is missing, mistakes pile up, and some people are born missing one.
    High school

    DNA gets damaged by its own chemistry, by sunlight and by copying mistakes. Because the double helix has two matching strands, most repair cuts out the damaged piece of one strand and rebuilds it using the other strand as a guide. Different machines handle different damage: one removes single wrong bases, one cuts out bulky sunlight damage, and one fixes copying mistakes by recognizing which strand is new. When both strands break, the cell either joins the ends directly or copies the missing information from the duplicate chromosome made during replication. People who inherit broken repair genes, as in xeroderma pigmentosum or Lynch syndrome, have a higher risk of certain cancers.
    College

    Repair pathways are matched to lesions. Direct reversal undoes a lesion in place: photolyase (absent in placental mammals) splits pyrimidine dimers with light, and O6-methylguanine methyltransferase moves a methyl group to its own cysteine in a single-use reaction. Base excision repair uses a glycosylase to remove a damaged base, leaving an AP site that AP endonuclease, polymerase beta and a ligase fill, often with one nucleotide. Nucleotide excision repair recognizes helix distortion, cuts on both sides of the lesion and removes an oligonucleotide (about 12 to 13 nucleotides in E. coli, about 27 to 29 in a human cell-free system). Mismatch repair corrects replication errors on the new strand, identified by hemimethylated GATC in E. coli and by strand breaks in eukaryotes. Double-strand breaks are joined without a template by non-homologous end joining throughout the cycle, or repaired by homologous recombination from the sister chromatid in S and G2 via resection, Rad51 strand invasion and Holliday junctions.
    Graduate

    Fidelity is multiplicative: base selection, exonucleolytic proofreading and mismatch repair each lower the error rate by orders of magnitude, so loss of one stage raises the mutation rate by roughly its own factor, about a thousandfold for mismatch repair in round textbook numbers. Strand discrimination differs by lineage. E. coli MutH reads hemimethylated GATC; eukaryotic MutS-alpha and MutL-alpha use pre-existing nicks, and MutL-alpha's latent endonuclease, activated in a PCNA-dependent manner, can create entry points on the strand that carries them. At double-strand breaks, the decisive step is resection: 3′ single-strand tails commit a break toward Rad51-dependent homologous recombination, whereas end protection favors Ku-dependent joining. Work since 2018 named end-protection factors such as shieldin, whose loss restored homologous recombination in BRCA1-deficient experimental systems. Labels such as error-free and error-prone fit outcomes better than pathways, since canonical end joining is often accurate at compatible ends, and recombination can produce crossovers and loss of heterozygosity.

    Open question: how a eukaryotic cell marks the new leading strand for mismatch repair in vivo, where pre-existing nicks may be few, is still debated; candidate signals include ribonucleotides left in new DNA and the orientation in which PCNA is loaded.
    An enzyme that works once

    Most enzymes are catalysts: they come out of a reaction unchanged and go on to the next substrate. The human methyltransferase that repairs O6-methylguanine does not. Harris, Karran and Lindahl reported in 1983 that the enzyme from human lymphoid cells transfers the methyl group to one of its own cysteines and is inactivated by it. Repair-proficient cell lines held roughly 10,000 to 25,000 molecules per cell, so a cell's capacity to repair this lesion is a fixed count, and a large enough dose of an alkylating agent can use it up.

    — Harris, Karran & Lindahl, Cancer Research 43:3247 (1983).
    A junction that outlived its model

    Holliday proposed his crossed-strand junction in 1964 to explain odd allele counts in fungal meioses. His model began with single-strand nicks. In 1983 Szostak, Orr-Weaver, Rothstein and Stahl proposed that recombination in yeast starts instead from a break through both strands, which is widened to a gap and repaired from the homologous chromosome, producing two Holliday junctions. The junction survived the change of starting point, and the break-first version became the standard model for how recombination and break repair share one mechanism.

    — Holliday, Genetical Research 5:282 (1964); Szostak, Orr-Weaver, Rothstein & Stahl, Cell 33:25 (1983).

     ↑ Top