Homologous Recombination

Homologous recombination is the process by which a broken or gapped DNA molecule is repaired using the sequence of a homologous, undamaged DNA duplex as a copying template. "Homologous" here means sequence-identical or nearly identical — typically the sister chromatid produced by replication, or, during meiosis, the homologous chromosome inherited from the other parent. Because HR uses a template rather than simply rejoining broken ends, it can regenerate exactly the sequence that was lost or damaged at the break site, making it fundamentally different from non-homologous end joining (NHEJ), which ligates ends together with little or no regard for sequence and often loses or adds nucleotides. Mechanistically, HR proceeds through a defined sequence of intermediates. A double-strand break is first resected into single-stranded 3′ overhangs. A recombinase protein — RecA in bacteria, Rad51 in eukaryotes — polymerizes onto this single-stranded DNA to form a helical nucleoprotein filament. This filament searches the genome for a homologous duplex and catalyzes strand invasion, in which the single strand base-pairs with its complement in the intact duplex while displacing the identical strand, forming a displacement loop, or D-loop. The invading 3′ end is then extended by DNA polymerase, using the intact strand as template, physically copying the missing sequence. Depending on the sub-pathway, this intermediate is either dissolved back into two duplexes without ever forming a stable crossed structure (synthesis-dependent strand annealing) or is converted into a double Holliday junction, a four-stranded structure in which the two duplexes are covalently interlinked, which must later be nicked apart (resolution) or unwound (dissolution) to separate the two DNA molecules. The two possible resolution geometries of a Holliday junction have different genetic consequences: one geometry regenerates the original arrangement of flanking markers (non-crossover), while the other exchanges the flanking arms between the two duplexes (crossover). This distinction is the molecular basis of genetic recombination frequencies measured in every cross since Mendel and Morgan, and it is why HR is simultaneously a DNA repair pathway and the engine of meiotic genetic exchange.

Step 1 — End resection. A double-strand break is first recognized by the MRN/MRX complex (Mre11–Rad50–Nbs1 in humans, Mre11–Rad50–Xrs2 in yeast) together with CtIP (Sae2 in yeast). MRN nicks the 5′-terminated strand a short distance from the break, and its Mre11 subunit then degrades that strand back toward the break end (3′→5′ exonuclease activity), while long-range resection nucleases — Exo1 (5′→3′ exonuclease) and Dna2 (helicase-nuclease, working with the BLM/Sgs1 helicase) — extend the resection tract outward. The net result is two long, 3′-tailed single-stranded overhangs, often several hundred to a few thousand nucleotides long, flanking the original break.

Step 2 — RPA loading, then Rad51 filament formation. The exposed single-stranded DNA is immediately coated by Replication Protein A (RPA), which prevents secondary structure and protects the DNA. RPA must then be displaced and replaced by the recombinase Rad51 (or bacterial RecA), a reaction that does not occur spontaneously because RPA binds single-stranded DNA more tightly than Rad51 does. This hand-off is catalyzed by mediator proteins — BRCA2 in humans (Rad52 in yeast, along with the Rad51 paralogs RAD51B/C/D, XRCC2/3) — which deliver Rad51 monomers directly onto RPA-coated DNA and nucleate filament assembly. The resulting Rad51–single-strand DNA nucleoprotein filament is a right-handed helical polymer that stretches and unwinds the DNA it coats, extending it roughly 1.5-fold relative to B-form DNA.

Step 3 — Homology search and strand invasion. The Rad51 filament conducts a search through the genome, sampling double-stranded DNA by transient, low-stability contacts until it finds a sequence complementary to the single strand it carries. This search happens far faster than random three-dimensional diffusion alone would allow, aided by chromatin organization that keeps the sister chromatid or homolog nearby. Upon finding homology, the filament catalyzes strand invasion: the single strand base-pairs with its complementary strand in the target duplex, displacing the identical (non-complementary) strand as a single-stranded loop. This structure — the D-loop — is the defining intermediate of homologous recombination, an actual physical joint molecule linking the broken and intact chromosomes.

Step 4 — DNA synthesis from the invading end. The 3′ end of the invading strand, now paired within the D-loop, serves as a primer. DNA polymerase (Pol δ in eukaryotes, with PCNA clamp loading) extends this end, synthesizing new DNA using the intact strand of the donor duplex as template. This step is the actual moment of information recovery: whatever sequence was lost at the break is now being resynthesized from an undamaged copy.

Step 5 — Branch migration and second-end capture, or SDSA. Two sub-pathways diverge here.

• In synthesis-dependent strand annealing (SDSA), the newly extended invading strand is displaced from the D-loop (unwound by helicases such as Rtel1/BLM), and its extended 3′ end anneals directly with the resected 3′ overhang of the other broken end via complementary single-stranded sequence. Gaps are filled and ligated. No Holliday junction ever forms, and the outcome is always a non-crossover.
• In the double-strand-break repair (DSBR) pathway, the second broken end is instead captured directly by the D-loop, base-pairing with the displaced strand. Both ends are now engaged with the donor duplex, DNA synthesis and ligation seal the structure, and the result is a double Holliday junction (dHJ) — two four-way branch points linking the two duplexes.

Step 6 — Holliday junction resolution or dissolution. A double Holliday junction must be removed for the two chromosomes to separate. Two fundamentally different mechanisms exist:

• Resolution by structure-specific nucleases (the GEN1/Yen1 or MUS81–EME1/Mms4 and SLX1–SLX4 resolvases) that cleave each junction symmetrically. Depending on the plane of cutting relative to the original break, resolution yields either crossover or non-crossover products, each with roughly equal likelihood if cuts are made independently at each junction, though cellular regulation biases the outcome.
• Dissolution by the BLM–TOPOIIIα–RMI1/2 (BTR / "dissolvasome") complex, which migrates the two junctions of a double Holliday junction together and then decatenates the resulting hemicatenane using topoisomerase. Dissolution is topologically constrained to produce only non-crossover products — it can never generate a crossover, because no strand cutting is involved.

The choice between resolution and dissolution, and between the crossover and non-crossover resolution planes, is a major control point: in mitotic cells, non-crossover outcomes are strongly favored (crossovers between sister chromatids are usually genetically silent but crossovers with a homolog would cause loss of heterozygosity), whereas in meiotic recombination a regulated subset of events is specifically directed toward crossover formation to ensure at least one crossover (the "obligate chiasma") per chromosome pair.

Robin Holliday's genetic model (1964). Studying gene conversion in fungi (Ustilago maydis), Holliday noticed patterns of non-Mendelian segregation at meiosis — aberrant 3:1 or 5:3 ratios of alleles among the four (or eight) products of a single meiosis, instead of the expected 2:2 or 4:4 — that could not be explained by simple independent assortment. He proposed a physical model: two homologous duplexes exchange single strands to form a symmetric, four-stranded crossed structure (now called the Holliday junction), which could migrate along the DNA (branch migration) and be resolved in two different ways, one giving crossover and one giving non-crossover products, with heteroduplex DNA (mismatched base pairs from strand exchange) explaining gene conversion when mismatch repair corrected one strand to match the other. This was a purely genetic inference — no protein or biochemical intermediate was known — yet it correctly predicted the existence of a four-way DNA junction more than a decade before anyone saw one directly.

Electron microscopy visualization of the Holliday junction (Potter and Dressler, 1976; and later crystallography). Huntington Potter and David Dressler used electron microscopy on recombination intermediates isolated from E. coli cells carrying plasmid DNA (ColE1), and directly visualized cross-shaped, four-armed DNA structures — physical joint molecules connecting two circular plasmids exactly as Holliday's model predicted. Later, the atomic structure of a synthetic four-way junction was solved by X-ray crystallography (Nash, Lilley and colleagues), confirming the stacked-X geometry and showing how the junction can adopt an open or stacked conformation depending on cations present, and how branch migration proceeds by sequential base-pair rearrangement without net energy input (isoenergetic in the absence of sequence heterology).

Biochemical reconstitution of strand invasion — RecA in vitro (Charles Radding and colleagues, late 1970s–1980s). Radding's laboratory purified RecA protein from E. coli and showed, in a purified biochemical system with single-stranded DNA, a homologous double-stranded DNA circle, and ATP, that RecA alone catalyzes formation of a D-loop-like joint molecule and promotes strand exchange over kilobase distances. This was the direct biochemical demonstration that a single protein, polymerized on DNA, can perform the entire homology search and strand invasion reaction — turning Holliday's genetic inference into a defined, reconstitutable enzymology. Matthew Meselson and Charles Radding's 1975 model further integrated single-strand invasion, D-loop formation, and mismatch correction into a unified molecular scheme that anticipated much of the modern pathway.

Szostak, Orr-Weaver, Rothstein and Stahl's double-strand-break repair model (1983). Using yeast (Saccharomyces cerevisiae) transformation experiments, Jack Szostak, Terry Orr-Weaver, Rodney Rothstein, and Franklin Stahl showed that a plasmid deliberately cut with a restriction enzyme to create a double-strand break, rather than being poorly repaired, was recombined into the yeast genome far more efficiently and precisely than an uncut plasmid — the break stimulated recombination rather than preventing it. This was the decisive result against the earlier "single-strand break" models of recombination (in which recombination was thought to initiate from a nick, not a full break) and established the double-strand-break repair (DSBR) model essentially as described above, with resection, strand invasion, D-loop formation, synthesis, and double Holliday junction resolution — the model still taught today, later refined by Wolf-Dietrich Heyer, Stephen Kowalczykowski, and others who added the SDSA branch and biochemically reconstituted human BRCA2-mediated Rad51 loading in vitro (Jasin, Sung, West laboratories, 2000s).

Physical monitoring of recombination intermediates in yeast meiosis (Neil Hunter and Nancy Kleckner, 2001). Using Southern blotting of DNA from synchronized meiotic yeast cultures, Hunter and Kleckner physically detected joint molecules with the branched, X-shaped mobility expected of double Holliday junctions appearing and disappearing at the predicted times during meiotic prophase, directly confirming that dHJs are genuine in vivo intermediates on the pathway to meiotic crossovers, not just a theoretical construct.

Example 1 — A meiotic tetrad and gene conversion. In a cross between two strains of Neurospora crassa differing at a single locus (allele A versus allele a), a particular ascus (the sac containing the eight ordered products of one meiosis, following the two post-meiotic mitoses) is dissected and typed. Normal, Mendelian segregation predicts a 4:4 ratio of A:*a spores. Suppose instead the ascus shows a 6:2 ratio (six A* spores, two a spores). Reasoning: a 6:2 ratio cannot arise from simple independent segregation of two alleles at a diploid locus — it requires that, at some point, one strand of DNA carrying allele a was physically converted to carry the A sequence, without a reciprocal reciprocal exchange occurring. This is exactly the signature of heteroduplex DNA formed during strand invasion: when the invading strand (say, carrying A) pairs with the donor duplex (carrying a) at the locus, a mismatch is created in the heteroduplex region. If the mismatch repair machinery in the cell corrects that mismatch using the A strand as the correction template before the subsequent mitotic divisions, then the originally a-bearing duplex "becomes" A, converting what should have been a 4:4 tetrad into 6:2 (or, if correction goes the other way, 2:6). This is gene conversion, and it was precisely this class of observation — non-Mendelian ratios in ordered fungal tetrads — that led Holliday to propose the heteroduplex-forming, four-stranded junction model in the first place. Note also that whether the flanking markers on either side of the converted locus show a crossover or non-crossover configuration is an independent readout of which Holliday junction resolution plane was used, and in real data roughly half of gene conversion events are associated with crossing-over of flanking markers, consistent with the two resolution geometries being used with similar frequency in meiosis.

Example 2 — A clinical genetics/oncology calculation: BRCA2 loss and synthetic lethality. A patient carries a germline heterozygous loss-of-function mutation in BRCA2 (one functional allele, one null allele) and develops a breast tumor. Sequencing the tumor shows loss of the remaining wild-type BRCA2 allele (loss of heterozygosity), so the tumor cells are fully BRCA2-deficient, while the patient's normal tissues remain heterozygous (one functional copy, which is sufficient for normal HR). Consequence for HR: BRCA2 is required to load Rad51 onto RPA-coated single-stranded DNA at resected double-strand breaks. Without BRCA2, Rad51 filaments cannot form efficiently, D-loop formation fails, and double-strand breaks that would normally be repaired by HR are instead shunted to error-prone pathways (NHEJ, or single-strand annealing when repeated sequences flank the break), producing chromosomal deletions, translocations, and genomic instability — this instability is what drives tumorigenesis in the first place. Therapeutic logic (synthetic lethality): PARP1 normally helps repair single-strand nicks and stalled replication forks; when PARP1 is inhibited (e.g., olaparib), unrepaired single-strand lesions collapse into double-strand breaks during replication. In a normal, BRCA2-proficient cell, these breaks are efficiently repaired by HR, so PARP inhibition alone is well tolerated. In the BRCA2-null tumor cell, however, these breaks cannot be repaired by HR at all, and the cell dies. Quantitatively, clinical trials of PARP inhibitor maintenance therapy in BRCA-mutated ovarian cancer (e.g., the SOLO2 trial) showed median progression-free survival of roughly 19.1 months (olaparib) versus roughly 5.5 months (placebo), a difference directly attributable to this HR-pathway-specific synthetic lethality, only exploitable because the tumor — unlike the patient's normal cells — has completely lost HR capacity.

Robin Holliday (1932–2014) was a British molecular biologist who, while working on genetic recombination in the fungus Ustilago maydis at the John Innes Institute in Norwich in the early 1960s, confronted a puzzle that pure Mendelian genetics could not solve: tetrads and octads from single meioses sometimes showed allele ratios like 3:1 or 6:2 instead of the expected 2:2 or 4:4. Rather than treating these as noise, Holliday built a physical, structural model of the DNA molecules themselves — proposing that homologous chromosomes exchange single strands to form a symmetric four-way joint that could slide along the DNA and be cut in two different ways. Published in 1964, years before anyone had biochemical tools to test it directly, the model was a triumph of reasoning from genetic pattern to molecular structure, and it correctly anticipated intermediates that electron microscopy and crystallography would later visualize directly. Holliday also made major contributions to theories of epigenetic inheritance and cellular aging later in his career, but it is the four-way DNA junction that bears his name in every molecular biology classroom.

Homologous recombination underlies CRISPR-based precision genome editing: after Cas9 introduces a targeted double-strand break, cells can be induced to repair it via HR using a co-delivered donor DNA template, allowing researchers to introduce specific point mutations, tags, or gene corrections rather than the small insertions/deletions produced by NHEJ — the entire strategy of "knock-in" editing depends on hijacking the cell's own HR machinery. In oncology, tumors with HR deficiency (BRCA1/2 mutations, or a broader "BRCAness" phenotype from other HR gene defects) are now routinely identified by genomic scarring signatures and treated with PARP inhibitors, exploiting synthetic lethality; HR status is a standard biomarker guiding treatment selection in ovarian, breast, prostate, and pancreatic cancers. In agriculture and plant breeding, understanding meiotic HR (crossover number, placement, and the enzymes controlling crossover interference) allows breeders and genome engineers to manipulate recombination rates, either increasing crossovers to break unwanted linkage between desirable and undesirable traits, or suppressing recombination to preserve favorable haplotype combinations in hybrid seed production. In conservation genetics and population biology, recombination rate estimates from HR-generated crossover patterns are used to reconstruct effective population sizes, detect selection, and map disease- or trait-associated loci through linkage disequilibrium, since the density of historical recombination events determines how finely genetic markers can localize a causal variant. In basic research on genome stability, HR defects are used to generate synthetic lethal genetic screens and to model human diseases of genomic instability (Fanconi anemia, ataxia-telangiectasia-like disorders), since many Fanconi anemia proteins act to promote HR-mediated repair of interstrand DNA crosslinks at stalled replication forks.

Ask the tutor: How does the choice between HR and NHEJ get made at a double-strand break, and which proteins act as the molecular switch?