Journal of Cell Science partnership with Dryad

Journal of Cell Science makes data accessibility easy with Dryad

The choice in meiosis – defining the factors that influence crossover or non-crossover formation
Jillian L. Youds, Simon J. Boulton


Meiotic crossovers are essential for ensuring correct chromosome segregation as well as for creating new combinations of alleles for natural selection to take place. During meiosis, excess meiotic double-strand breaks (DSBs) are generated; a subset of these breaks are repaired to form crossovers, whereas the remainder are repaired as non-crossovers. What determines where meiotic DSBs are created and whether a crossover or non-crossover will be formed at any particular DSB remains largely unclear. Nevertheless, several recent papers have revealed important insights into the factors that control the decision between crossover and non-crossover formation in meiosis, including DNA elements that determine the positioning of meiotic DSBs, and the generation and processing of recombination intermediates. In this review, we focus on the factors that influence DSB positioning, the proteins required for the formation of recombination intermediates and how the processing of these structures generates either a crossover or non-crossover in various organisms. A discussion of crossover interference, assurance and homeostasis, which influence crossing over on a chromosome-wide and genome-wide scale – in addition to current models for the generation of interference – is also included. This Commentary aims to highlight recent advances in our understanding of the factors that promote or prevent meiotic crossing over.


Meiosis is the specialised reductive division that generates haploid cells. During this process, a single round of replication is followed by two rounds of chromosome segregation: in the first division (meiosis I), homologous chromosomes segregate, whereas in the second division, sister chromatids segregate (meiosis II). A key step in meiosis I is the recognition of homologous chromosomes, which then align and pair along the length of the chromosome. Once homologues have aligned, synapsis can proceed with the formation of the synaptonemal complex (SC), a protein structure that supports and maintains homologues in close juxtaposition and serves as a scaffold for crossover-promoting recombination factors. Meiotic crossing over involves the generation of meiotic double-strand breaks (DSBs), which are subsequently repaired either as crossovers or non-crossovers (Fig. 1). Meiotic recombination is not only necessary to create new allele combinations that generate genetic diversity, but is also essential in ensuring accurate chromosome segregation at the first meiotic division because the crossover acts as a tether between homologues, which ensures that each homologue will properly align at the metaphase plate and thereby correctly attach to the spindle. DSB repair occurs concurrently with SC formation and is required for normal synapsis in yeast and mice (Baudat et al., 2000; Roeder, 1997; Romanienko and Camerini-Otero, 2000), whereas in Caenorhabditis elegans and Drosophila melanogaster, homologue pairing and SC formation can occur independently of meiotic recombination (Colaiacovo et al., 2003; Dernburg et al., 1998; Liu et al., 2002; McKim et al., 1998).

The process of meiotic recombination is initiated when meiotic DSBs are created by the endonuclease SPO11, in conjunction with a number of additional proteins (Keeney and Neale, 2006). DSBs are then resected to generate 3′ single-strand DNA (ssDNA) overhangs that are initially bound by replication protein A (RPA), which is subsequently displaced by the recombinase radiation sensitive 51 (RAD51) and/or the meiosis-specific recombinase dosage suppressor of Mck1 (DMC1) to form nucleoprotein filaments. These filaments serve to find a complimentary sequence within a homologous chromosome, at which they instigate single-end strand invasions (Hunter and Kleckner, 2001) to generate so-called displacement loop (D loop) recombination intermediates (Fig. 1). If the second end of the original DSB binds with the homologous chromosome, a double Holliday junction is formed, which can be resolved to generate either a non-crossover or an interhomologue crossover, the latter of which is hereafter referred to simply as crossover (Bishop and Zickler, 2004; Schwacha and Kleckner, 1995). Double Holliday junctions can also be processed through dissolution, which results in a non-crossover (Fig. 1) (Wu and Hickson, 2003). Meiotic non-crossovers have also been proposed to form when strand invasion is transient, and when a limited amount of DNA synthesis occurs before the invaded strand dissociates and anneals to its partner strand, as in mitotic synthesis-dependent strand annealing (SDSA; see Box 1 and Fig. 1) (Allers and Lichten, 2001; Bishop and Zickler, 2004). During meiosis in budding yeast, non-crossover heteroduplex products are found to form with the same timing as double Holliday junctions, whereas crossovers occur later (Allers and Lichten, 2001), consistent with the idea that crossovers and non-crossovers are formed through distinct pathways. Analysis of ZMM mutants (Zip1, Zip2, Zip3, Zip4, Mer3, Msh4 and Msh5; further discussed below) in yeast indicates that the decision between crossover and non-crossover is made very early, i.e. at or prior to the establishment of a stable single-end invasion intermediate, when one of the two ends of the DSB invades its homologous chromatid (Bishop and Zickler, 2004; Borner et al., 2004; Hunter and Kleckner, 2001). Thus, the crossover–non-crossover decision is thought to occur around the time of strand exchange.

Fig. 1.

Model for meiotic crossover or non-crossover formation. Double strand breaks are generated and their 5′ ends are resected to generate a 3′ overhang. A strand invasion event then generates a single-end invasion D loop intermediate. If the second end of the original DSB also engages with the homologue, a double Holliday junction is formed (shown on the left). The double Holliday junction can be resolved to form either a crossover (interference-dependent) or a non-crossover. Alternatively, the junction can be dissolved by double Holliday junction dissolution to form a non-crossover. Instead of forming a double Holliday junction, the D loop can be dissociated and the invading strand can associate with the opposite end of the original break, as in synthesis-dependent strand annealing (SDSA), to form a non-crossover. Alternatively, the intermediate can be acted upon by enzymes such as Mus81 that can form interference-independent crossovers.

This Commentary will discuss the factors that contribute to crossover or non-crossover formation in meiosis, including the generation and positioning of meiotic DSBs, formation of the SC and generation of recombination intermediates. We will also discuss how interhomologue crossing over is promoted compared with intersister repair, how recombination intermediates are resolved into crossovers as well as how anti-recombinases prevent crossing over. Crossover interference, assurance and homeostasis will also be discussed (see text box), including a summary of the current models for how crossover interference is established.

Control of meiotic DSB formation

Meiotic crossovers tend to form at specific sites in the genome of most organisms (Buard and de Massy, 2007; Mezard, 2006; Pryce and McFarlane, 2009); these are known as recombination hotspots. In mammals, hotspots are typically regions of 1–2 kb that occur every 50–100 kb within the genome (Jeffreys et al., 2001; McVean, 2010; Myers et al., 2008). Human recombination hotspots often occur relatively close to genes (within 50 kb), but are preferentially found outside of transcribed regions (Myers et al., 2005). At hotspots in eukaryotes, meiotic DSBs are generated by the conserved topoisomerase-like endonuclease Spo11 (Bergerat et al., 1997; Cao et al., 1990; Keeney et al., 1997). However, insights into what controls the position of break formation were unknown until recently.

Several studies have shown that Spo11 access to DNA is one of the determinants of DSB location, because DSBs are often found within open chromatin in yeast (Berchowitz et al., 2009; Ohta et al., 1994; Wu and Lichten, 1994). In C. elegans, the putative chromatin-modifying protein high incidence of males 17 (HIM-17) is required for successful DSB formation and for normal levels of dimethylation of lysine residue 9 of histone 3 (H3K9Me2) during meiosis, because mutants display a marked reduction in H3K9Me2 staining (Reddy and Villeneuve, 2004). Furthermore, DSB formation in him-17 non-null mutants can be enhanced by loss of LIN-35, the C. elegans retinoblastoma (Rb) protein homologue, which is a known component of chromatin modifying complexes (Reddy and Villeneuve, 2004). Collectively, these data suggest that the activity of HIM-17 alters the chromatin state, probably by opening the chromatin to allow access by SPO-11, and this governs DSB competence. More recently, components of the C. elegans condensin I complex have been implicated in regulating both the position and number of DSBs, because mutants show increased formation of meiotic DSBs and a distribution of crossovers that differs from the wild type (Mets and Meyer, 2009; Tsai et al., 2008). Here, it was demonstrated that disruption of the condensin I complex causes an increase in chromosomal axis length, which is independent of DSB formation but requires the axis-associated protein HIM-3 (Mets and Meyer, 2009). Thus, in C. elegans, it appears that the condensin I complex controls chromosome structure and the association of chromatin with the chromosome axis, which in turn might determine its accessibility by SPO-11. This and other evidence suggests that the chromosome axis has an important role in meiotic DSB formation. Certainly, proteins associated with the chromosome axis have also been shown to be important for DSB formation. For example, the C. elegans HIM-3 paralog – him three paralog 3 (HTP-3) – is a component of the meiotic chromosome axis that mediates DSB formation through its interaction with a complex required to generate DSBs (Goodyer et al., 2008). The conserved mouse protein meiosis-specific 4 (MEI4), which associates with the axis of meiotic chromosomes, was also shown to be required for DSB formation (Kumar et al., 2010). Thus, a functional relationship appears to exist between the chromosome axis and DSB formation and might have some bearing on where hotspots arise. For a detailed discussion and models of the interplay between the chromosome axis and DSB formation, readers are referred to the article by Kleckner (Kleckner, 2006).

If DSB hotspots are associated with sites of open chromatin, then the question arises what exactly determines the chromatin state at hotspots. Evidence suggests that some recombination hotspots in humans are associated with specific sequence motifs. Ten percent of the hotspots identified in a genome-wide survey of ~1.6 million single-nucleotide polymorphisms (SNPs) in three sample populations were associated with the 7-mer motif CCTCCCT (Myers et al., 2005). Using more detailed data from the human ‘haplotype map’ (HapMap) project, work by the same group identified a degenerate 13-mer sequence (CCNCCNTNNCCNC) that was present in one or more copies at 40% of all human hotspots (Myers et al., 2008). These data indicate that there is some correlation between DNA sequence and hotspot location. However, although humans and chimpanzees have more than 98% sequence identity, hotspot locations in these two species are not conserved (Ptak et al., 2005; Winckler et al., 2005). In addition, there is evidence that recombination rates vary both between ethnic groups (Evans and Cardon, 2005) and between individuals (Cheung et al., 2007; Coop et al., 2008). Thus, sequence alone is clearly not the only factor determining hotspot activity.

Recent studies indicate a role for specific epigenetic modifications at DSB sites. A recent study in C. elegans identified the new chromatin factor X-non-disjunction factor 1 (XND-1), which is required for DSB formation specifically on the X chromosome, as well as for the distribution of DSB formation (but not DSB number) on the autosomes (Wagner et al., 2010). Increased acetylation of histone H2A at lysine 5 (H2AK5Ac) that is likely to be mediated by the histone acetylation protein Myst family histone acetyltransferase-like (MYS-1; TIP60 in humans) in xnd-1 mutants was identified as a chromatin modification associated with the alteration in autosomal distribution of meiotic DSBs and reduced DSB formation on the X chromosome (Wagner et al., 2010). This study suggests that H2AK5Ac is a determinant of DSB distribution in C. elegans, although the influence of this modification on DSB formation in other organisms still needs to be shown. Trimethylation of lysine 4 of histone 3 (H3K4Me3) has been considered as a pre-existing marker for sites of DSB formation in S. cerevisiae because this histone modification is frequently found in regions close to DSB sites, independently of gene expression levels (Borde et al., 2009). The same study showed that deletion mutants of the SET-domain-containing 1 (set1) gene, which encodes the only H3K4 methyltransferase in S. cerevisiae, display a dramatic reduction in DSBs, and that those DSBs that form in the absence of Set1 are differentially localised (Borde et al., 2009). Specific histone modifications were also shown to be present at meiotic DSB

Box 1. Terms used to describe meiosis

Anti-recombinase: A protein that acts to prevent or inhibit recombination.

Chiasmata: The (cytologically) visible structure that is the physical manifestation of a crossover between homologous chromosomes in meiosis.

Crossover assurance: The mechanism that makes certain each homologous chromosome pair will receive at least one meiotic crossover.

Crossover homeostasis: A mechanism that ensures a constant number of meiotic crossovers, even under conditions where more or fewer meiotic DSBs are generated. Homeostasis maintains meiotic crossovers at the expense of non-crossovers. How homeostasis is maintained is not well understood but it might be governed by the same mechanism as interference.

Crossover interference: A mechanism that distributes meiotic DSBs and crossovers such that adjacent crossovers tend to occur further apart than expected by chance. The basis of crossover interference is not well understood.

D loop: A recombination intermediate structure wherein one end of the DSB has invaded into the homologous chromosome and been extended to form a stable structure. A D loop can be dissociated to allow non-crossover repair through SDSA, or can go on to form a double Holliday junction. See Fig. 1.

Double Holliday junction: A recombination intermediate structure that can be formed following D loop formation if the second end of the original DSB also associates with the homologous chromosome. This structure can be processed into either a crossover or non-crossover. See Fig. 1.

Haplotype map: A map generated to describe common patterns of human genetic variation using single nucleotide polymorphisms in various populations. Also known as the HapMap.

Interference-independent crossovers: Crossovers that do not exhibit interference. Crossovers in this class are formed by Mus81 catalyzed recombination intermediate cleavage in many organisms. Because they do not display interference, interference-independent crossovers can occur in close proximity. Interference-independent crossovers contrast with interference-dependent crossovers, which always display interference in terms of their positioning.

Interhomolog crossover: Also known as a meiotic crossover, a crossover that has occurred between homologous chromosomes during meiosis.

Intersister repair: Repair of damage or a meiotic DSB (as in this article) that uses the sister chromatid as a template rather than the homologous chromosome. Repair between sister chromatids can be through crossover or non-crossover pathways.

Synthesis-dependent strand annealing (SDSA): A pathway for non-crossover repair wherein a D loop recombination intermediate is formed and limited DNA synthesis occurs in the region of the break, using the homologous chromosome as a template. The D loop is then dissociated and the invading end of the DSB that has been extended anneals back with the other end of the original DSB. DNA synthesis and ligation seals the break.

ZMM proteins: A group of proteins that includes Zip1, Zip2, Zip3, Zip4, Mer3, Msh4, Msh5 and the recently identified Spo16 in S. cerevisiae; they are involved in both SC formation and meiotic crossing over.

initiation sites in mice with H3K4Me3 enriched at active DSB sites, whereas histone H4 hyperacetylation was found to occur after DSB formation (Buard et al., 2009). H3K4Me3 is present in Spo11−/− mice, which lack meiotic DSBs, and thus H3K4Me3 is a marker for sites of DSB initiation that might contribute to hotspot activity (Buard et al., 2009). HapMap-methylation-associated SNPs, which are markers of germline methylation, are positively correlated with regional meiotic recombination rates in humans (Sigurdsson et al., 2009), providing further evidence that epigenetic modifications influence hotspot activity in mammals.

It remains unclear exactly how sites of frequent meiotic recombination are identified and marked as hotspots. However, recent studies in mice have identified the gene PR-domain-containing 9 (Prdm9) (Baudat et al., 2010; Parvanov et al., 2010), which encodes for a protein with histone methyltransferase activity. Different human alleles of PRDM9 show altered activity of recombination hotspots (Baudat et al., 2010). Furthermore, in vitro, the protein product of the human PRDM9 A allele was shown to bind to the specific 13-mer motif previously associated with recombination hotspots, suggesting that PRDM9 marks DSB initiation sites by recognising this motif (Baudat et al., 2010). Thus, sequence motifs that are modified by certain epigenetic marks appear to designate some of the known mammalian recombination hotspots by either signalling to the recombination machinery or allowing SPO11 to access the DNA. However, it is not known how these signals integrate with factors that control the association of chromatin with the chromosome axis. Furthermore, the H3K4Me3 mark is highly enriched in promoter regions of actively transcribed genes (Bernstein et al., 2005; Schneider et al., 2004), but meiotic DSBs do not form in all promoter regions and, currently, there is no known direct relationship between DSB formation and transcriptional activity (Hunter, 2006; Kniewel and Keeney, 2009). Thus, this chromatin modification alone is not likely to account for all hotspot activity. As the roles of epigenetic modifications in meiotic DSB formation and crossover control remain poorly understood, they should be an important focus of future investigation.

Proteins that influence crossover outcomes

The factors influencing DSB position are not yet well-understood, but a number of proteins that function downstream of DSB creation, to either promote or prevent crossover formation, have been characterised. Although it is not known exactly how the decision is made to form a crossover or non-crossover, many meiotic proteins influence whether or not a crossover can take place. Here, these proteins will be classified as having either pro-crossover or anti-crossover activities, although at least one of these proteins can be considered to fit into both categories, depending on the context of the protein activity or the model system in which it has been examined. A non-exhaustive list of these genes is presented in Table 1.

Pro-crossover proteins

Setting the stage for crossover formation

A number of pro-crossover proteins have roles in pairing or SC formation that indirectly promote crossovers. SC formation involves the assembly of several structural elements, specifically a pair of lateral elements connected by transverse elements and a central element, which facilitates crossovers by continuously maintaining homologues in close juxtaposition (Costa and Cooke, 2007; de Carvalho and Colaiacovo, 2006). Among the proteins that form the SC in S. cerevisiae are the Zip proteins (part of the ZMM group of proteins). Zip1 is a structural protein that makes up the central regions of the SC (Sym et al., 1993). Zip2, Zip3 and Zip4 are also required for SC assembly, and have been implicated in ubiquitylation and/or sumoylation of proteins associated with SC formation (Agarwal and Roeder, 2000; Borner et al., 2004; Cheng et al., 2006; Chua and Roeder, 1998; Perry et al., 2005; Tsubouchi et al., 2006). A common theme among SC proteins is that, although the sequence homologues of lower-organism SC proteins are not easily detectible in higher organisms, structural features of the proteins are conserved (Costa and Cooke, 2007). In mammals, the SC is made up of multiple proteins including SC proteins 1, 2 and 3 (SYCP1, 2 and 3), SC central element proteins 1 and 2 (SYCE1 and SYCE2) and testis-expressed gene 12 (TEX12) (Costa and Cooke, 2007), but these will not be discussed further here. A detailed review of proteins that are involved in homologue pairing and synapsis, as well as recent insights into meiotic pairing centres in C. elegans that are beyond the scope of this review can be found elsewhere (Costa and Cooke, 2007; Ding et al., 2010; Lynn et al., 2007; Yang and Wang, 2009; Zetka, 2009; Zickler, 2006).

Table 1.

A non-exhaustive list of genes with either pro- or anti-crossover activities

Generation of the recombination intermediate

In the last few years, considerable progress has been made towards defining the activities that process DSBs to generate 3′ ssDNA overhangs, which are the substrate for initiating homologous recombination. Studies of C. elegans completion of meiotic recombination-1 (com-1) and its homologue in Arabidopsis thaliana, com1 (homologues of yeast sae2, human CtIP), found a similar phenotype in these two species, in which mutants failed to load RAD51 onto meiotic DSBs, suggesting that COM1 acts in DSB resection (Penkner et al., 2007; Uanschou et al., 2007). More recently, data from budding yeast and mammalian cells have revealed that DSB resection is dependent on the cooperative action of multiple factors, including the Mre11–Rad50–Xrs2 (MRX) complex [Mre11–Rad50–Nbs1 (MRN) in mammals] SUMO activating enzyme 2 (Sae2; CtIP in mammals), small growth suppressor 1 [Sgs1; Bloom's syndrome mutated (BLM) in mammals], exonuclease 1 (Exo1; EXO1 in mammals) and DNA replication helicase 2 (Dna2; DNA2 in mammals) (Gravel et al., 2008; Mimitou and Symington, 2008; Zhu et al., 2008) (Fig. 2). These studies have been validated by the analysis of DSB end processing in vitro using the respective purified proteins (Nimonkar et al., 2008). This area has been intensively reviewed and discussed elsewhere and we refer the reader to several reviews on this topic (Bernstein and Rothstein, 2009; Mimitou and Symington, 2009). Once the DSB has been processed, Rad51 and/or the meiosis-specific recombinase Dmc1 bind the 3′ ssDNA to generate a nucleoprotein filament that carries out a homology search and strand invasion into the homologous chromosome. Both Dmc1 and Rad51 are required for efficient meiotic recombination in yeast (Schwacha and Kleckner, 1997) and mice (Pittman et al., 1998; Sharan et al., 2004; Yoshida et al., 1998), but Dmc1 orthologues have not been identified in C. elegans or D. melanogaster. Human DMC1 has robust D-loop-forming activity in vitro (Li et al., 1997; Sehorn et al., 2004), whereas in S. cerevisiae, Rad54 is required together with Rad51 for D loop formation (Sung et al., 2003). In vivo, the Dmc1 nucleoprotein filament is more adept at forming D loops with the homologous chromosome than the Rad51 nucleoprotein filament (Shinohara et al., 2003; Tsubouchi and Roeder, 2003). Co-ordination of RAD51 and DMC1 activities in mammalian meiosis might be achieved through the breast cancer susceptibility protein 2 (BRCA2), because BRCA2 binds to both proteins at distinct sites (Sharan et al., 1997; Thorslund et al., 2007), and is required for localisation of RAD51 and DMC1 at foci that are presumed to be meiotic DSBs in mice (Sharan et al., 2004). Similarly, C. elegans BRCA2 (BRC-2) is essential for meiotic DSB repair, in which it functions to promote RAD-51 filament nucleation and stabilisation, and in the stimulation of RAD-51 mediated strand exchange (Martin et al., 2005; Petalcorin et al., 2007; Petalcorin et al., 2006).

Mismatch repair defective 4 and 5 (Msh4 and Msh5, respectively) and meiotic recombination 3 (Mer3), i.e. other members of the ZMMs, also promote crossovers, probably by facilitating the stable formation of recombination intermediates. msh4 and msh5 mutants in S. cerevisiae have reduced inter-homologue crossing over (Hollingsworth et al., 1995; Ross-Macdonald and Roeder, 1994). Similarly, in C. elegans, MSH-4 and MSH-5 are essential for crossing over (Kelly et al., 2000). Msh4 and Msh5 are known to form a heterodimer, and one proposal is that this acts as a clamp to hold homologous chromosomes together, thereby stabilising the Holliday junction and facilitating crossing over (Snowden et al., 2004) (Fig. 2). In Msh4 and Msh5 knockout mice, chromosomes fail to correctly pair, crossovers are absent and, consequently, animals are sterile (de Vries et al., 1999; Edelmann et al., 1999; Kneitz et al., 2000). In mammals, unlike in S. cerevisiae, Msh4 and Msh5 are required for correct chromosome pairing during meiotic prophase, and both are clearly associated with recombination intermediates destined to form both crossovers and noncrossovers (Kneitz et al., 2000; Santucci-Darmanin et al., 2000). Thus, the mouse phenotype associated with loss of Msh4 or Msh5 does not only reflect of a loss of crossovers, but is also more severe than that of other mutants, such as loss of the mut L homologue 1 (Mlh1; discussed below) (Edelmann et al., 1996). Interestingly, msh4 and msh5 are not found in D. melanogaster and in S. pombe. The latter relies solely on the structure-specific endonuclease complex between mutagenesis sensitive 81 (Mus81) and essential meiotic endonuclease 1 (Eme1) for crossover formation (Osman et al., 2003), whereas in D. melanogaster the clamp function of Msh4 and Msh5 can potentially be replaced by alternative proteins or an alternative mechanism might be present (Blanton et al., 2005). In addition to Msh4 and Msh5, the Mer3 helicase promotes normal crossover frequencies in yeast (Nakagawa and Ogawa, 1999). Mer3 functions to stimulate heteroduplex extension by Rad51, thereby stabilising nascent D loop structures to promote capture of the second free DNA end, double Holliday junction formation and, ultimately, crossovers (Mazina et al., 2004). In Sordaria, Mer3, Msh4 and Mlh1 have recently been shown to have roles in homologous chromosome pairing (Storlazzi et al., 2010).

MLH1 is a well-known marker for sites designated as crossovers in the mouse (Anderson et al., 1999). MLH1 is required for crossovers, as germ cells from mice that lack MLH1 do not display sufficient chiasmata and do not progress beyond the meiotic pachytene stage, and thus, Mlh1 null mice are infertile (Edelmann et al., 1996) (Fig. 3). Budding yeast mlh1 deletion mutants have reduced crossing over (Hunter and Borts, 1997). In vitro, human MSH4 interacts with MLH1, and in vivo the two proteins colocalise during early to mid-pachytene, when crossing over takes place (Santucci-Darmanin et al., 2000). One possible function of MLH1 is that it mediates release of the MSH4 and MSH5 clamp (Snowden et al., 2004), thereby facilitating crossover completion. mlh-1 has not been characterised in C. elegans; instead, Zip homologous protein 3 (ZHP-3), the C. elegans homologue of yeast Zip3, has been shown to be a marker for meiotic crossovers (Fig. 3). ZHP-3 has two roles in meiosis: it promotes crossover formation and also restructures chromosomes at diakinesis, the chromosome condensation stage, so that appropriate segregation can take place (Bhalla et al., 2008; Jantsch et al., 2004). Thus, several proteins are required to generate and maintain a stable recombination intermediate. In order for appropriate chromosome segregation to take place, the stable recombination intermediate must have engaged the homologous duplex and not the sister chromatid, ultimately resulting in formation of an interhomologue crossover.

Fig. 2.

Biochemical activities that promote crossover and/or non-crossover recombination. Shown are schematic representations of specific recombination intermediates that are subject to biochemical activities of meiotic enzymes (helicases, nucleases, DNA-binding proteins) that act to promote one of two repair outcomes: non-crossover or crossover (recombination). Arrows indicate the directionality of the helicase activity; arrows with scissors indicate the position of cleavage of the nuclease activity; yellow circles indicate the preferred substrate for DNA binding. The human BLM orthologues Sgs1 (S. cerevisiae), Rqh1 (S. pombe), HIM-6 (C. elegans) and Mus309 (D. melanogaster) are presumed to perform roles in both resection of DSBs and dissolution of double Holliday junctions. Yen1 is the S. cerevisiae orthologue of human GEN1. Slx4, HIM-18 and Mus312 are the S. cerevisiae, C. elegans and D. melanogaster orthologues of human BTBD12, respectively.

Promoting interhomologue crossing over versus intersister repair

Meiotic crossover formation can be controlled by genes that allow meiotic DSB repair from only certain templates, i.e. not all. The proteins that participate in the barrier to sister chromatid repair are pro-crossover factors because they promote DSB repair that uses the homologous chromosome, but not the sister chromatid (see text box). In this way, these factors allow crossovers between homologues but not between sister chromatids. For example, in budding yeast, interhomologue recombination is ensured by phosphorylation of the axial element protein homologue pairing 1 (Hop1) by the serine/theronine protein kinases meiotic checkpoint 1 (Mec1) and telomere length 1 (Tel1), the homologues of mammalian ataxia telangiectasia and Rad3-related protein (ATR) and ataxia telangiectasia mutated (ATM), respectively (Carballo et al., 2008). It has been proposed that Hop1 phosphorylation leads to dimerisation of the meiotic axial element meiotic kinase 1 (Mek1), which enables it to phosphorylate its target proteins that prevent the repair of DSBs using the sister chromatid (Niu et al., 2005). In addition, Mek1 inhibits Rad51 activity by attenuating the formation of the Rad51–Rad54 complex; with less active Rad51, the activity of the meiosis-specific strand exchange protein Dmc1 is favoured (Niu et al., 2009). Dmc1 is thought to be more efficient at promoting interhomologue recombination than Rad51 (Shinohara et al., 2003; Tsubouchi and Roeder, 2003), thus, the activity of Mek1 promotes interhomologue recombination over sister chromatid repair. In C. elegans – which do not have Dmc1 – germ cells, at the onset of meiotic prophase, switch into a specialised mode of DSB repair that is characterised by the requirement for RAD-50 in loading RAD-51 onto meiotic DSB ends, a process that is essential for interhomologue crossover formation (Hayashi et al., 2007). By mid- to late pachytene, a second developmentally programmed switch occurs; RAD-50 is no longer required for RAD-51 association with DSBs, and competence for interhomologue crossing over is lost (Hayashi et al., 2007). Repair of any remaining DSBs might then occur through the sister chromatid, which requires BRCA homologue 1 (BRC-1) that is dispensable for repair through the homologue (Adamo et al., 2008). The requirement for RAD-50 in RAD-51 loading at DSBs is partially dependent on a number of proteins that have roles in meiosis-specific chromosome axis structure (Hayashi et al., 2007). Thus, mechanisms exist to ensure that crossing over with the homologue will occur at the correct time, rather than with the sister chromatid. Once the stable interhomologue crossover intermediate has been generated, the activity of a resolution enzyme will ultimately complete the crossover.

Fig. 3.

Examples of crossover interference. (A) Examples of crossovers marked by staining of MLH1 (green) on mouse chromosomes. The SC is marked in red by staining of the SC protein 3 (SCP3). When multiple crossovers occur on the same chromosome, they tend to be spaced far apart, demonstrating crossover interference. (B) Complete crossover interference in C. elegans. The images in the top row show the single recombination foci per chromosome marked by staining of ZHP-3 (six chromosomes, hence six foci) at meiotic diplotene in a wild type animal. The rtel-1 mutant exhibits defective complete crossover interference, as illustrated by the increased number of ZHP-3 foci per meiotic nucleus (bottom images).

Resolution of the recombination intermediate

Crossover formation is controlled directly by enzymes that act on recombination intermediates and resolve these either as crossovers or non-crossovers. Mus81–Eme1 can resolve intermediates into crossovers (Fig. 2), but also has multiple biochemical abilities, including a preference for acting on structures that include D loops, nicked Holliday junctions, replication forks with the lagging strand at the junction point, and 3′ flap structures (Bastin-Shanower et al., 2003; Fricke et al., 2005; Osman et al., 2003). Yeast Mus81–Eme1 is also reported to have robust cleavage activity on intact Holliday junctions, although they are not its preferred substrate in vitro (Gaskell et al., 2007). Mus81 is responsible for generating interference-independent crossovers in S. cerevisiae (Argueso et al., 2004; de los Santos et al., 2003) and in mouse (Holloway et al., 2008), and also generates the majority of crossovers in S. pombe, which has only interference-independent crossovers (Boddy et al., 2001; Osman et al., 2003; Smith et al., 2003). In C. elegans, MUS-81 is also responsible for a subset of crossovers that occur when meiotic non-crossover repair through SDSA is blocked by mutation in the anti-recombinase regulator of telomere length 1 (RTEL-1) (Youds et al., 2010).

A number of DNA processing enzymes have recently been identified in various organisms that possess the biochemical activity to resolve Holliday junctions – specifically, the ability to cleave Holliday junctions symmetrically. These resolvases have the potential to generate either crossover or non-crossover products, and other factors, perhaps the MSH4-MSH5 complex, might influence this decision. The human resolvase XPG-like endonuclease 1 (GEN1), and its S. cerevisiae orthologue Yen1, have been independently identified through their ability to symmetrically cleave Holliday junctions (Ip et al., 2008) (Fig. 2). Yen1 has functional overlap with Mus81 in budding yeast (Blanco et al., 2010) and human GEN1 was found to rescue the meiotic phenotype of mus81 S. pombe mutants (Lorenz et al., 2009), indicating conserved, functionally similar roles for Mus81 and Yen1 or GEN1 in yeast and humans, respectively. However, we have not observed a meiotic phenotype for gen-1 mutants in C. elegans (J.L.Y. and S.J.B., unpublished data) and, together with the recently demonstrated role for gen-1 in DNA damage signalling and repair in the nematode (Bailly et al., 2010), this indicates that additional enzyme(s) act as the meiotic resolvase in this organism.

In addition to GEN1, concurrent works from several groups described roles for the orthologous proteins Drosophila Mus312, C. elegans HIM-18, and mammalian synthetic lethal of unknown function 4 (SLX4) in resolving Holliday junctions (Andersen et al., 2009; Fekairi et al., 2009; Munoz et al., 2009; Saito et al., 2009; Svendsen et al., 2009) (Fig. 2). In mus312 mutants of D. melanogaster, meiotic crossing over is reduced by ~95%, consistent with the hypothesis that Mus312 is the key protein required for Holliday junction resolution in this organism (Andersen et al., 2009; Yildiz et al., 2002). Gene expression patterns and knockdown studies suggest that mammalian SLX4 has a meiotic function in humans that is conserved with that of D. melanogaster Mus312 (Andersen et al., 2009). SLX4 interacts with SLX1, and the SLX4–SLX1 complex has a robust Holliday junction cleavage activity in vitro, which probably accounts for its role in meiosis (Fekairi et al., 2009; Munoz et al., 2009; Svendsen et al., 2009). SLX4 also binds the structure-specific complex between the endonucleases Xeroderma Pigmentosum complementation group F and excision repair cross-complementing 1 (XPF–ERCC1) complex and MUS81–EME1, among other proteins. Depletion of SLX4 also causes sensitivity to a number of DNA damage agents, indicating that it serves as a platform for different structure-specific endonucleases that are most likely to act both in meiosis and in DNA repair (Fekairi et al., 2009; Munoz et al., 2009; Svendsen et al., 2009). Similarly, the C. elegans SLX4 orthologue HIM-18 is required for processing meiotic recombination intermediates, because him-18 mutants display phenotypes that are consistent with reduced meiotic crossover formation (Saito et al., 2009). Moreover, HIM-18 physically interacts with the structure-specific endonucleases SLX-1 and XPF-1, and might serve as a scaffold that binds certain nucleases to allow cleavage of Holliday junction intermediates in various cellular contexts (Saito et al., 2009). For further details on recent advances in the area of resolvases, the reader is directed elsewhere (Mimitou and Symington, 2009; Svendsen and Harper, 2010).

Anti-crossover proteins

Several proteins are known to inhibit crossover formation by promoting different pathways for meiotic DSB repair, and these involve alternative processing of recombination intermediates. BLM, together with topoisomerase TOPOIIIα, has double Holliday junction dissolution activity in vitro, which resolves these junctions without crossover formation and, thereby, explains the increased degree of sister chromatid exchange observed in BLM-deficient cells (Wu and Hickson, 2003) (Figs 1 and 2). The budding yeast BLM homologue Sgs1 suppresses the formation of multi-chromatid joint molecules during meiosis and, thereby, prevents aberrant crossing over (Oh et al., 2007). These anti-crossover activities of Sgs1 are normally antagonised by the ZMM proteins (Jessop et al., 2006). Conversely, the fission yeast BLM homologue Rec Q helicase 1 (Rqh1) might extend hybrid DNA and, thereby, bias the recombination outcome toward crossover formation (Cromie et al., 2008). In C. elegans mutants for the BLM orthologue him-6, meiotic crossovers are decreased compared with those in wild type (Wicky et al., 2004). Similarly, meiotic recombination is reduced to about half of the wild type frequency in D. melanogaster with mutations in the BLM orthologue mus309 (McVey et al., 2007). Therefore, BLM might have context-dependent effects or differing roles in different organisms and, thus, could be classified as having both pro- and anti-crossover activities. It is also possible that the pro-crossover function of BLM reflects a role in meiotic DSB resection (Fig. 2).

The C. elegans anti-recombinase RTEL-1 promotes non-crossover repair both in meiosis and mitosis (Barber et al., 2008; Youds et al., 2010). Analysis of the meiotic phenotypes of rtel-1 mutants revealed that these animals had multiple crossovers per chromosome, whereas wild type animals usually have only a single crossover per chromosome (Youds et al., 2010). rtel-1 mutants also have an increased number of recombination foci, as marked by ZHP-3 in meiotic nuclei, indicating excess crossovers. RTEL-1 has the ability to disrupt D loop recombination intermediates in vitro (Barber et al., 2008; Youds et al., 2010) (Fig. 2). Thus, RTEL-1 is likely to promote non-crossovers by functioning in meiotic SDSA, where it probably dissociates the D loop intermediate to facilitate the association of the invading DNA strand back with the original duplex DNA, thereby inhibiting a stable strand invasion event. In addition to its meiotic functions, RTEL1 acts in interstrand crosslink repair in C. elegans and human cells, and is required for maintenance of telomere length in mice (Barber et al., 2008; Ding et al., 2004). Thus, RTEL1 is believed to have important roles as an anti-recombinase in multiple cellular scenarios. BLM and RTEL1 are the only examples of meiotic anti-crossover proteins in the current literature. However, it has been reported that S. cerevisiae mutator phenotype 1 (Mph1) has biochemical abilities similar to those of RTEL1 (Prakash et al., 2009) (Fig. 2); therefore, future experiments should address whether Mph1, like RTEL1, has a role in meiotic SDSA.

Possible mechanisms for crossover interference

The proteins discussed thus far all make an individual contribution to the formation of either a crossover or non-crossover. These roles must be carefully regulated and work in concert for successful completion of either crossover or non-crossover repair. However, the factors described above are not the only factors that control the decision between crossover and non-crossover; it is also influenced by crossover interference, assurance and homeostasis on a chromosome-wide and genome-wide scale.

It is well known that crossovers are distributed non-randomly along chromosomes. This phenomenon exists because of two underlying factors: first, DSBs are not formed randomly and, second, crossover choice is governed by crossover interference. Crossover interference dictates that adjacent crossovers on the same chromosome tend to occur at sites that are further apart than expected if they were randomly positioned. Crossover assurance makes certain that all chromosomes will obtain at least one crossover, known as the obligate crossover. The presence of at least one crossover between homologous chromosomes serves as a physical linkage between homologues so that they can each attach to the correct spindle and, subsequently, segregate accurately in meiosis I. In budding yeast, crossover interference and assurance are separately regulated by the ZMM proteins (Shinohara et al., 2008). Mutations in the ZMM genes result in defective SC formation together with reduced crossovers, but non-crossovers are unaffected (Borner et al., 2004). Different subcomplexes formed by ZMM proteins might have different functions. For instance, Mer3 and the Msh4–Msh5 heterodimer participate in promoting the differentiation of crossovers, which establishes crossover interference; whereas the newly identified ZMM protein Spo16 acts in complex with Zip4 to efficiently implement crossovers, which is important for crossover assurance (Shinohara et al., 2008). However, observations by de Boer et al. (de Boer et al., 2007) suggest that, in mice, a separate mechanism for crossover assurance, in addition to that of interference, is not necessary (de Boer et al., 2007).

C. elegans meiosis presents an extreme case for the control of crossover formation. Here, meiotic crossover interference is referred to as ‘complete’ because each chromosome has only a single crossover, indicating that the number of crossovers per chromosome is tightly regulated (Hillers and Villeneuve, 2003; Wood, 1988); <1% of oocyte meioses have a chromosome that lacks a crossover (Dernburg et al., 1998) and double crossovers are very rare (Lim et al., 2008). Studies of this system have shown that complete crossover interference in C. elegans can be altered by either increasing the number of meiotic DSBs, for example, by mutating the condensin dumpy 28 (dpy-28), or by changing the balance between crossover and non-crossover repair of meiotic DSBs (by mutation in rtel-1). It is, therefore, possible that crossover interference is regulated on at least two levels in C. elegans: by the number of DSBs generated and by the pathway through which they are repaired.

Related to crossover assurance and interference is crossover homeostasis, a mechanism that regulates the balance between crossovers and non-crossovers. In S. cerevisiae, experiments have shown that crossover homeostasis exists. Specifically, a set level of crossovers is always maintained at the expense of non-crossover pathways, such as SDSA, in situations where fewer meiotic DSBs are generated (Chen et al., 2008; Martini et al., 2006). In mutants with reduced Spo11 activity, which have reduced numbers of meiotic breaks, crossover homeostasis is maintained by shifting the ratio between crossovers and non-crossovers towards fewer non-crossovers (Martini et al., 2006). Homeostasis also exists in C. elegans when additional breaks are generated by ionising radiation (IR); when excess DSBs are induced by IR treatment in wild type animals, little or no increase in recombination is observed (Youds et al., 2010). However, in rtel-1 mutants treated with IR, a large, dose-dependent increase in recombination is observed, indicating that rtel-1 mutants cannot maintain homeostasis in the presence of additional meiotic DSBs (Youds et al., 2010). Homeostasis and crossover interference are potentially regulated by a common mechanism because mutants defective in interference, such as budding yeast zip2 and zip4 mutants and C. elegans rtel-1 mutants, also show compromised homeostasis (Chen et al., 2008; Joshi et al., 2009; Youds et al., 2010).

Several models have been proposed over the years to explain how crossover interference is propagated (Fig. 4). One early hypothesis in the field is the polymerisation model, in which the completion of a crossover initiates polymerisation of an inhibitor of recombination along the chromosome (King and Mortimer, 1990). More recently, the stress model has been proposed, which posits that mechanical stress along the chromosome results in axis buckling that places one recombination intermediate into a position where it commits to becoming a crossover and, subsequently, tension is released in the area near the crossover so that no other DSBs nearby will form crossovers (Borner et al., 2004; Kleckner et al., 2004) (Fig. 4).

Current evidence indicates that the establishment of crossover interference is linked to chromosome axis structure. In yeast, the conserved ATPase pachytene checkpoint 2 (Pch2) regulates meiotic axis morphogenesis by controlling the overall levels and localisation patterns of the structural protein Hop1 and the association of Zip3 with meiotic chromosome axes (Joshi et al., 2009), but it is not required for crossover formation (Joshi et al., 2009; Zanders and Alani, 2009). A proposed model for Pch2 function is that it reorganises chromosome axes into long-range, one-crossover modules, in which only a single crossover is assured and additional crossovers are prevented (Joshi et al., 2009). The authors suggest that multiple single-crossover modules exist along the length of each homologous chromosome pair, and that tension release – as proposed by the stress model (Borner et al., 2004; Kleckner et al., 2004) – prevents multiple crossovers from occurring within any one module (Joshi et al., 2009). Therefore, higher-order chromosome structure appears to have an important role in crossover interference, at least in yeast.

Data from mouse studies have also provided insights into the establishment of crossover interference. It has been reported that MLH1 foci, which mark crossovers in meiotic pachytene in mice, exhibit strong interference (de Boer et al., 2006). The same group showed that, prior to pachytene, late zygotene MSH4 foci and RPA foci – which mark interhomologue interactions that could be resolved into either crossovers or non-crossovers – also exhibit substantial interference (de Boer et al., 2006). These data are consistent with the idea that interference occurs in at least two consecutive steps. Furthermore, mice that lack the axial element protein SYCP3 show structurally abnormal axial elements and incomplete chromosome synapsis, but have similar levels of crossover interference when compared with wild type animals, indicating that axial element structure and full synapsis are not required for the establishment of normal interference (de Boer et al., 2007). However, correlations between SC length and interference in mammals have been reported by other groups (Lynn et al., 2002; Petkov et al., 2007), so further research will be required to clarify the exact relationship between the SC, meiotic chromosome structure and crossover interference.

Fig. 4.

Illustration of two proposed mechanisms of crossover interference. In the polymerisation model (left), the formation of a crossover (blue X) leads to the localisation of a recombination inhibitor (orange oval) at the crossover and its polymerisation along the vicinity, preventing other potential sites (blue circle) nearby from forming crossovers, thereby allowing non-crossover repair (green region of DNA) instead. In the mechanical stress model (right), compression stress (red arrows) along the chromosome axis leads to buckling of the axis at one of the potential crossover sites. Buckling at the site changes the configuration of the recombination intermediate, committing it to become a crossover. Crossover formation (blue X) at this site allows local stress release, preventing other crossovers from forming close by. Similar to the polymerisation model, other potential recombination sites nearby are repaired as non-crossovers (green region of DNA).

One further possibility is that epigenetic signals are also involved in marking the crossover site and preventing other crossovers from occurring nearby. The above mentioned study that showed that HapMap methylation-associated SNPs are positively correlated with regions of meiotic recombination (Sigurdsson et al., 2009) suggests that either methylation is a marker for DSB sites or, alternatively, its occurrence after crossover formation inhibits further crossovers, which serve as the signal for interference. Further studies will be essential to understand the relationship between epigenetic marks, chromosome axis morphogenesis and the crossover–non-crossover decision.

Future perspectives

Although recent studies have enhanced our understanding of the factors that control meiotic crossing over, many questions remain. Further experiments are needed to clearly define how the chromosome axis and its association with DNA control the location of meiotic DSBs. Understanding the timing of epigenetic marks will also be important in clarifying how these DNA modifications control DSB initiation. For example, one question is whether there are different modifications at crossover and non-crossover sites. Prdm9 has recently been identified as a histone modifying protein that might mark DSB initiation sites (Baudat et al., 2010), but are there other proteins that act similarly? Answers to these and other questions will help to establish the involvement of epigenetic signals in designating DSB and/or crossover sites.

The activities of a number of pro-crossover proteins are relatively well understood. However, only a few meiotic non-crossover proteins have been identified. Given that, in most eukaryotes, far more meiotic DSBs occur compared with the number of crossovers that will eventually be formed, meiotic non-crossover proteins have an extremely important role in preventing excess crossovers. Thus, additional anti-crossover proteins probably exist but need to be identified. Synthetic lethal screens in the genetic backgrounds of anti-crossover mutants, such as rtel-1 or BLM, might identify additional genes with overlapping functions. Furthermore, any proteins that can act on D loop recombination intermediates should be tested for roles in meiotic crossover control. In addition, other modes of regulating the promotion of non-crossovers should also be considered, aside from a direct activity on recombination intermediates. For instance, are there proteins that promote DSB repair by using the sister chromatid rather than the homologous chromosome? Or is the activity of certain proteins that aid in the formation or resolution of stable recombination intermediates temporally inhibited to favour non-crossovers?

Finally, recent years have seen strides toward a greater understanding of crossover interference, assurance and homeostasis, but many questions remain regarding these processes. Further details are needed to clarify how a higher-order chromosome structure contributes to crossover interference and whether a role for the chromosome axis in interference is conserved in higher organisms. If a common mechanism controls one or several of these processes, how does this involve the chromosome axis and other factors? Mutants that display phenotypes in which interference and homeostasis defects are uncoupled would be highly valuable to address these questions. Whatever the answers to these questions, future investigations into the factors that control the crossover-non-crossover decision promise to be intriguing.


We acknowledge Carrie Adelman for images of mouse MLH1 foci, and thank Jennifer Svendsen and Carrie Adelman for comments on the manuscript. Research in the lab of S.J.B. is funded by Cancer Research UK. S.J.B. is a Royal Society Wolfson Research Merit Award holder.


View Abstract