FANCM Limits Meiotic Crossovers

See allHide authors and affiliations

Science  22 Jun 2012:
Vol. 336, Issue 6088, pp. 1588-1590
DOI: 10.1126/science.1220381

No Crossing Over

To ensure the correct division of chromosome during the reduction division of meiosis, homologous chromosomes undergo double-strand breaks that—through crossing over and recombination—link the homologs together (and importantly introduce diversity into the genomes of gametes). But only a minority of these crossovers results in recombination—most are directed into non-crossover pathways. Lorenz et al. (p. 1585), working in the yeast Schizosaccharomyces pombe, and Crismani et al. (p. 1588), working in the higher plant Arabidopsis thaliana, looked for the factors that limit crossovers and promote non-crossover pathways. The homolog of the human Fanconi anemia complementation group M (FANCM) helicase protein was found to be a major meiotic anti-recombinase, which could drive meiotic recombination intermediates into the non-crossover pathway.


The number of meiotic crossovers (COs) is tightly regulated within a narrow range, despite a large excess of molecular precursors. The factors that limit COs remain largely unknown. Here, using a genetic screen in Arabidopsis thaliana, we identified the highly conserved FANCM helicase, which is required for genome stability in humans and yeasts, as a major factor limiting meiotic CO formation. The fancm mutant has a threefold-increased CO frequency as compared to the wild type. These extra COs arise not from the pathway that accounts for most of the COs in wild type, but from an alternate, normally minor pathway. Thus, FANCM is a key factor imposing an upper limit on the number of meiotic COs, and its manipulation holds much promise for plant breeding.

Crossovers (COs) are necessary for balanced chromosome segregation during meiosis. Meiotic recombination is initiated by the formation of double-strand breaks (DSBs), which are repaired to yield COs and noncrossovers (NCOs) (1). Two pathways to CO formation are known: Class I CO formation is subject to positive CO interference, whereby one CO reduces the probability of another nearby CO (2), and depends on a group of proteins called ZMMs in addition to MLH1 and MLH3; and class II COs, which are not subject to interference and depend on MUS81 and EME1/MMS4 (1). Mammals, budding yeast, and Arabidopsis thaliana have both classes of COs, but some organisms have only one or the other (2). Factors that limit meiotic CO number must exist because DSBs vastly outnumber COs [e.g., 230 versus ~10 in Arabidopsis Col-0 (3)].

To identify meiotic anti-CO factors, we designed a screen based on the idea that mutations that increase CO frequency will restore the fertility of CO-defective mutants. Arabidopsis zmm mutants, including zip4, lack meiotic COs, causing missegregation of homologs and thus reduced fertility leading to shorter fruit that can be visually discriminated from that of the wild type (3) (Fig. 1). Arabidopsis zip4 mutants still produce enough seeds, through random chromosome segregation, to perform a screen. zip4-2 seeds were mutated with ethylmethane sulfonate, and these developed into plants indistinguishable from zip4. However, among ~2000 families obtained by self-fertilization, eight lines segregated plants with increased fertility. The causal mutations were single locus, recessive, and fell into three complementation groups. The first complementation group contained five lines and was further studied here (table S1). Mapping and whole-genome sequencing of one suppressor [zip4(s)1] identified a missense mutation in the gene At1g35530 as the potential causal mutation (Fig. 2, fig. S2, and table S1). At1g35530 is the single Arabidopsis homolog of the human Fanconi anemia complementation group M (FANCM) (fig. S2). FANCM is involved in genome stability in various eukaryotes but has no known meiotic function (4, 5). Mutations in FANCM were found in the four other allelic suppressors, and the suppressor phenotype was recapitulated with a transferred DNA (T-DNA) insertion in FANCM (N620621). Three more allelic mutations in FANCM were identified in similar ongoing zmm screens (shoc1 and msh5) (Fig. 2 and fig. S2). The identified mutations resulted in amino acid changes, all in well-conserved residues of the helicase domain or in splicing sites (Fig. 2A and fig. S2). The single fancm mutants were indistinguishable from wild type in terms of growth and fertility [seeds per fruit: wild type = 66 ± 8 (n = 52), zip4 = 3 ± 2 (n = 65), fancm-1 zip4 = 44 ± 8 (n = 48), fancm-1 = 66 ± 7 (n = 47)].

Fig. 1

A screen for genes that limit COs. Wild-type plants have long fruit (A) and five bivalents at metaphase I (B), whereas zip4 mutants have short fruit (C) and predominantly univalents at metaphase I (D). Suppressors of zip4 were identified on the basis of increased fruit length (E) and confirmed to have increased bivalent formation (F). Scale bars: (A, C, E) 5 cm; (B, D, F) 5 μm.

Fig. 2

FANCM mutations restore bivalent formation in zmm mutants. (A) The FANCM protein with identified mutations indicated as arrows (table S1). The open triangle indicates the position of the fancm-9 T-DNA insertion. (B) Average number of bivalents (blue) and pairs of univalents (red) per male meiocyte. Number of cells analyzed is indicated in parentheses.

In wild type, each pair of homologous chromosomes is invariably linked by a minimum of one chiasma, the cytological manifestation of meiotic COs, resulting in five bivalents at metaphase I. In zip4, msh5, or shoc1, only ~1.5 bivalents form, leaving the ~3.5 remaining pairs of homologs as univalents (Figs. 1 and 2B). In contrast, bivalent formation in all the suppressors increases substantially (Fig. 2B), ranging from an average of 4.5 bivalents per cell, to essentially indistinguishable from wild type. A large increase in bivalent formation was also observed in female meiosis in fancm-1 (fig. S3) and fancm-5 (not shown). In contrast, fancm-1 did not restore chiasma formation of spo11-1 or dmc1 (Fig. 2), showing that chiasmata arising in fancm do not depend on ZMMs but require DSBs and strand invasion catalyzed by SPO11-1 and DMC1, respectively.

We then labeled chromosomes using fluorescence in situ hybridization and examined the shape of metaphase I bivalents (fig. S4). This technique determines if there is zero or at least one chiasma on each chromosome arm. In zip4, a majority of chromosomes received no chiasma, whereas in wild type, fancm-1 zip4, and fancm-1, a majority of chromosomes received at least one chiasma on each arm (table S2). This confirms that chiasma frequency largely increased genome wide when fancm was mutated in zip4, to a level approaching the upper limit of chiasma number that can intrinsically be detected with this technique. Recombination was therefore measured in a series of genetic intervals by tetrad-based analysis of fluorescent-tagged lines (6, 7) (Fig. 3 and table S3). Recombination was decreased in all tested intervals by a factor of 2 to 3 in zip4 compared to wild type, consistent with previous findings (3). In fancm-1 zip4 the genetic distances were increased, by a factor of 1.9 to 3.1 compared to wild type (P < 10−5). This confirms that the FANCM mutation not only restores CO formation in the absence of ZIP4, but also boosts CO frequency far above that of wild type. In fancm-1 (with a functional ZIP4), recombination was increased even more than in fancm-1 zip4, by an average of 12% (P < 0.05 in three individual intervals among six). Thus, when comparing fancm-1 to wild type, the genetic distance was increased in all eight intervals tested, by a factor ranging from 2 to 3.6 (P < 10−8), which highlights the importance of FANCM in placing an upper limit on CO frequency.

Fig. 3

Meiotic recombination is amplified in the absence of FANCM. Genetic distances in eight intervals, measured by using fluorescent-tagged lines, were calculated with the Perkins equation (20) and are given in centiMorgans. I2a and I2b are adjacent intervals on chromosome 2, and so on for the other couples of intervals as described in (6) (fig. S5 and table S3).

The difference in CO frequency between fancm-1 and fancm-1 zip4 is rather limited compared to the difference between fancm-1 zip4 and wild type, suggesting that the frequency of ZMM COs is not increased by the absence of FANCM. This was further supported by the result that the number of MLH1 foci, a marker of ZMM-dependent chiasmata (8), was identical in wild type and fancm-1 [11.8 ± 1.6 (n = 38) and 11.5 ± 2.1 (n = 49), P = 0.5] (Fig. 4, A and B). Furthermore, the chiasmata observed in fancm-1 zip4 are not marked by MLH1 (Fig. 4C). Also, DMC1 foci number is unchanged in fancm-1 (222 ± 52, n = 16) compared to wild type (215 ± 37, n = 14, fig. S6) (3). We cannot exclude the possibility that mutating FANCM does not also increase DSBs, although the unchanged number of DMC1 foci does not support this.

Fig. 4

fancm extra COs are MUS81 dependent. (A to C) MLH1 immunolocalization at diakinesis. (A) Wild type. (B) fancm-1. (C) fancm-1 zip4. (D to I) Analysis of the fancm mus81 synthetic growth defect. (D) Wild type. (E) fancm mus81. (F) fancm mus81 rad51. (G) fancm. (H) mus81. (I) rad51. (J to L) Meiotic chromosome spreads. (J) fancm-1 zip4 mus81 with chromosome fragmentation at the metaphase I to anaphase I transition. (K and L) fancm-1 mus81 showing chromosome fragmentation at anaphase I and five bivalent-like structures at metaphase I, respectively. Scale bars: 10 μm.

If the extra COs produced in the absence of FANCM arise through a non-ZMM pathway, as suggested by the above data, they should be insensitive to interference (2, 9). We measured interference between adjacent intervals (tables S4 and S5) using the interference ratio (6, 7) (table S4), which indicated strong interference in four pairs of intervals in wild type (P < 3 × 10−4). In contrast, in fancm and fancm zip4, these ratios were not different from 1 (no interference), but were different from the wild-type ratios (P < 9.10−3), indicating an absence of interference (table S4). Interference was also measured within each interval by calculating the nonparental ditype ratio (10), which confirmed the absence of detectable interference in fancm zip4 (table S5).

The extra COs arising in fancm are independent of ZIP4, not marked by MLH1, and not sensitive to interference. These characteristics are reminiscent of MUS81-dependent COs, which are a minor fraction in wild type (11, 12). Therefore, we tested whether the extra COs were MUS81-dependent, by introducing the mus81 mutation into fancm zip4. Notably, fancm mus81 or fancm mus81 zip4 mutants have a strong growth defect. The fancm mus81 growth defect was suppressed by mutating RAD51 (Fig. 4, D to I). This suggests that MUS81 and FANCM have redundant roles in homologous recombination during somatic DNA repair.

Despite the growth defect, meiocytes could be obtained from fancm zip4 mus81 and fancm mus81 plants. Although meiosis appears to be normal in the double fancm zip4 or single mus81 (Fig. 1F and fig. S7), the triple mutation leads to a meiotic catastrophe, showing an absence of bivalents at metaphase I and chromosome fragmentation at anaphase I (Fig. 4J), further arguing that the extra COs appearing in the fancm mutant are MUS81 dependent.

In the fancm mus81 double mutant, with a functional ZIP4, the chromosome fragmentation defect was still present at anaphase I (Fig. 4K), suggesting that the ZMM pathway cannot repair recombination intermediates that accumulate in the absence of both MUS81 and FANCM. However, in fancm mus81, unlike in fancm zip4 mus81, structures resembling bivalents were seen at metaphase I (Fig. 4L), before fragmentation appeared at anaphase I, suggesting the formation of chiasmata. In addition, a wild-type number of MLH1 foci (8) were observed at diplotene (9.9 ± 2.1, n = 33) and diakinesis (9.6 ± 1.8, n = 14) (fig. S8) in fancm mus81, suggesting that the formation of ZMM-dependent COs is indifferent to the defective MUS81 FANCM pathways.

In mitotic cells, budding and fission yeast orthologs of the FANCM helicase—Mph1 and Fml1, respectively—have been shown to unwind displacement loops (D loops) in NCO pathways and to be somatic CO suppressors (13, 14). The biochemical properties of FANCM helicases are likely similar at mitosis and meiosis. We suggest that FANCM processes meiotic DSB repair intermediates, possibly D loops, driving them toward NCO resolution (or sister chromatid events) (fig. S1). In the absence of FANCM, MUS81 repairs these intermediates as interference-insensitive COs, whereas ZMMs cannot process these intermediates as COs. This implies that recombination intermediates on which FANCM can act are distinct from the intermediates that ZMMs converts to COs, supporting an early irreversible channeling of intermediates toward the ZMM or MUS81 pathways (9). To date, only two other meiotic anti-CO factors have been described (1): R-TEL in Caenorhabditis elegans (15) and Sgs1 in budding yeast (16, 17). Thus, anti-CO factors identified so far are helicases, raising the possibility that different helicases are used in various organisms to prevent excessive COs. Additionally, our findings show that, although most eukaryotes have only one to three COs per chromosome on average, CO number can be largely increased without obvious negative phenotypic effects, suggesting that COs are naturally constrained below their possible maximum. This finding supports the idea that crossover frequency is maintained by natural selection at a specific equilibrium between the long-term advantages and costs of recombination (18). Finally, the hyperrecombination provoked by fancm mutation could be of great interest for crop improvement, which relies on the production of new allele combinations through meiotic recombination whose frequency is a limiting factor (19).

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Tables S1 to S5

References (2130)

References and Notes

  1. Acknowledgments: R.M. and J.L.S. thank the EU-FP7 program (Meiosys-KBBE-2009-222883), and G.P.C. thanks the NSF (MCB-1121563) for financial support. We thank V. Borde, M. Grelon, C. Mézard, F. Nogué, A. Demuyt, and O. Loudet for critical reading of the manuscript and helpful discussions. A provisional patent application based on the work has been filed by INRA.
View Abstract

Navigate This Article