Meiotic interference among MLH1 foci requires neither an intact axial element structure nor full synapsis

During the prophase of the first meiotic division, homologous chromosomes (homologs) form stable pairs (bivalents), and non-sister chromatids of homologs exchange corresponding parts. Concomitantly, a zipper-like structure, the synaptonemal complex (SC), is formed between the homologs (Page and Hawley, 2004): the two sister chromatids of each chromosome form a common axis, the axial element (AE), and the AEs of homologs are connected along their length by transverse filaments, a process called synapsis, to form an SC. Along AEs and SCs, protein complexes occur that are involved in homologous chromosome pairing and crossover formation, and can be visualized as foci by immunofluorescence labeling (Ashley and Plug, 1998). One component of such protein complexes is the Escherichia coli MutL-homologous protein MLH1, which participates in a late step in crossover formation (Baker et al., 1996; Hunter and Borts, 1997). In the mouse, the number and spatial distribution of MLH1 foci along bivalents closely correspond with those of crossovers (Broman et al., 2002; Froenicke et al., 2002).

Crossovers are more evenly spaced along bivalents than is to be expected if they were placed randomly, a phenomenon named (positive) crossover interference. In early interference models, important roles were ascribed to SCs and/or synapsis (reviewed by Egel, 1995), but evidence against this is accumulating (reviewed by de Boer and Heyting, 2006). AEs, on the other hand, are still considered potentially important for crossover interference (Hillers, 2004; Kleckner and Zickler, 2003; Nabeshima et al., 2004).

We analyzed whether full synapsis and/or structurally intact AEs are required for crossover interference in the mouse, by studying cytological interference among MLH1 foci in Sycp3^–/– mice, which display structurally abnormal AEs and incomplete synapsis. The level of interference among MLH1 foci was the same in these mice as in the wild type, and we conclude therefore that crossover interference requires neither intact AEs nor full synapsis.

During meiotic prophase of Sycp3^–/– mice, the two sister chromatids of each chromosome still form a common axial structure, which contains cohesins, but lacks AE proteins SYCP3 and SYCP2 (Pelttari et al., 2001). This structure, further denoted as `cohesin axis', appears stretched and fragmented when visualized by labeling of cohesins in spread Sycp3^–/– spermatocytes or oocytes (Fig. 1) (Pelttari et al., 2001). Sycp3^–/– cohesin axes synapse incompletely and discontinuously (Fig. 1A,B) (Pelttari et al., 2001). Whereas Sycp3^–/– spermatocytes enter apoptosis presumably before MLH1 foci appear (Yuan et al., 2000), Sycp3^–/– oocytes assemble MLH1foci, and in some cases form chiasmata and complete meiosis (Yuan et al., 2002). The formation of MLH1 foci in Sycp3^–/– oocytes provided a unique opportunity to investigate at the cytological level the role of AEs and synapsis in interference, even in cells that cannot form chiasmata and complete meiosis.

We studied interference among MLH1 foci in oocytes at day 18 post coitus, when the proportion of MLH1-positive nuclei reached its maximum, both in Sycp3^–/– and wild-type mice (not shown). We focused on two long (1 and 2) and two short (18 and 19) chromosomes, which we identified by FISH (Fig. 1C). Because we inferred the strength of interference among MLH1 foci from the frequency distribution of interfocus distances (see below), it was essential that we could unambiguously recognize and locate every individual MLH1 focus on the cohesin axis. Therefore, we only analyzed bivalents of which: (1) both ends could be recognized; (2) at least one cohesin axis could be fully traced; and (3) all MLH1 foci could be clearly distinguished from background signals. Only 15% of the FISH-labeled bivalents in MLH1-positive oocytes fulfilled these criteria (see Fig. S1 in supplementary material). When measuring distances along cohesin axes, we excluded the gaps between the signals for meiotic cohesin REC8, for reasons explained in supplementary material Table S1 and Fig. S1Q,R.

In Sycp3^–/– oocytes, MLH1 foci were located in regions of synapsis (identified by the presence of TF-protein SYCP1; Fig. 1A) or at sites where cohesin axes converged without detectable SYCP1 (Fig. 1A,B,D-F) (cf. Higgins et al., 2005), whereas they were lacking from the paracentromeric region on all four analyzed chromosomes, in both Sycp3^–/– and wild-type mice (cf. de Boer et al., 2006; Froenicke et al., 2002). In the wild type, MLH1 foci were distributed rather uniformly along the remainder of the chromosomes, and this pattern was maintained on Sycp3^–/– chromosomes 1 and 2. For Sycp3^–/– chromosomes 18 and 19, the numbers of analyzed MLH1 foci were too small to decide whether they were distributed similarly along the bivalents as in the wild type (Fig. 1I).

The average number of MLH1 foci per oocyte nucleus was lower in Sycp3^–/– than in wild-type mice (Table 1), which fits with the decreased number of chiasmata in Sycp3^–/– oocytes (Yuan et al., 2002). Nevertheless, three of the four analyzed chromosomes had similar numbers of MLH1 foci in Sycp3^–/– and wild-type oocytes (Table 1); only on Sycp3^–/– chromosome 19 was the number of MLH1 foci reduced. Accordingly, Yuan et al. (Yuan et al., 2002) reported a differential effect of the SYCP3 disruption on recombination on different chromosomes. Among the chromosomes analyzed by these authors chromosome 12 displayed reduced recombination in a Sycp3^–/– background. Both chromosome 19 and chromosome 12 carry a nucleolus-organizing region (Dev et al., 1977). Possibly these chromosomes pair later than average, and in a Sycp3^–/– background part of them might pair too late to form MLH1 foci.

As is explained in detail elsewhere (Broman et al., 2002; de Boer et al., 2006; McPeek and Speed, 1995; Stam, 1979), the strength of interference among foci can be estimated by fitting the frequency distribution of interfocus distances to the gamma distribution. This yields an estimate of the so-called interference parameter of the gamma model, here denoted by ν, which is a measure of the strength of interference: if the best fit is obtained for ν=1, there is no interference, and the higher ν is above 1, the stronger (positive) interference is.

Since chromosomes 18 and 19 rarely have more than one MLH1 focus (Table 1), we could only estimate the strength of interference from the interfocus distances on chromosomes 1 and 2. As judged by the P values in Table 2, the interfocus distances on these chromosomes fitted reasonably to the gamma distribution. In wild-type oocytes, the estimates of ν (v̂ values in Table 2) for chromosomes 1 and 2 were high, 8.9 and 11.7 respectively, indicating strong interference among MLH1 foci on these chromosomes. For Sycp3^–/– oocytes, we obtained similarly high v̂ values (Table 2), indicating a similar high interference level among MLH1 foci in Sycp3^–/– oocytes as in the wild type. Intact AEs are thus not required for wild-type interference levels.

In Sycp3^–/– oocytes, the cohesin axes between two MLH1 foci were often not (fully) synapsed (Fig. 1A-H). To analyze whether synapsis had any influence on interference, we compared interference among MLH1 foci that were separated by fully synapsed cohesin axes with interference among foci that were separated by stretches of unsynapsed or incompletely synapsed cohesin axes. For both situations, we obtained similar high v̂ values (Table 2), indicating that full synapsis is not required for wild-type interference levels among MLH1 foci.

The ν estimates (Table 2) are based on the observed interfocus distances, and thus only on bivalents with two or more foci. Interference levels can also be estimated from the distributions of the numbers of foci per bivalent. Although this approach is generally less valuable than the approach based on interfocus distances (see Broman and Weber, 2000), it has the advantage that bivalents with fewer than two foci can be included in the analysis. Therefore, we determined for each of the four analyzed chromosomes, in Sycp3^–/– and wild-type oocytes, the proportions of bivalents with 0, 1, 2, 3 and 4 foci. For chromosome 19, we found a significant difference between Sycp3^–/– and the wild type, as was to be expected, because chromosome 19 bivalents had on average fewer MLH1 foci in Sycp3^–/– than in wild-type oocytes (Table 1). For chromosomes 1, 2 and 18, there was no significant difference between Sycp3^–/– and wild-type oocytes (Table 3).

Using computer simulation, and assuming that the MLH1 foci are distributed uniformly along the bivalents (which is roughly correct, see Fig. 1I) and that the gamma model applies for the spacing of MLH1 foci, we then estimated for each of the four analyzed chromosomes, in the wild type and Sycp3^–/–, the expected numbers of bivalents with 0, 1, 2, 3 and 4 foci, based on the corrected v̂ values (corr. ν in Table 2), and the observed average numbers of foci per bivalent (Table 1). For the long chromosomes, we found generally a good fit of the observed with the expected numbers (Table 3). This supports the use of the gamma model and the ν estimates for these chromosomes, and thus confirms the occurrence of strong interference among MLH1 foci on Sycp3^–/– chromosomes 1 and 2.

For chromosomes 18 and 19, we could not estimate ν from the interfocus distances, because these chromosomes rarely have more than one focus. For these chromosomes we performed a series of simulations based on the observed average numbers of foci per bivalent and increasing integer values of ν. For ν=1 (no interference), the expected frequencies of bivalents with 0 or more than 1 focus were much higher than those observed, both for chromosome 18 and 19, and in both Sycp3^–/– and wild-type oocytes (Table 3), which indicates positive interference among MLH1 foci on these chromosomes. For wild-type chromosomes 18 and 19, the observed numbers of bivalents with 0, 1, 2, 3 or 4 foci fitted only with the expected numbers for high ν values (ν⩾16 for chromosome 18, ν⩾33 for chromosome 19). This suggests that in wildtype, interference on chromosomes 18 and 19 is as strong as or stronger than interference on chromosomes 1 and 2. Broman et al. (Broman et al., 2002) inferred similar high ν values for the short chromosomes of wild-type female mice from genetic intercrossover distances. For Sycp3^–/– chromosome 19, only few MLH1 foci were available for analysis, and the observed distribution of the numbers of foci per bivalent was therefore compatible with a broad range of ν values (Table 3). However, it was not compatible with ν=1 (Table 3), implying that the lower average number of foci per bivalent of Sycp3^–/– chromosome 19 is not accompanied by total loss of interference. For Sycp3^–/– chromosome 18, the observed distribution of the numbers of foci per bivalent fitted with the expected distribution for ν⩾8, indicating strong interference on this chromosome. Altogether, the observed distributions of the numbers of foci per bivalent are consistent with wild-type interference levels on all four analyzed Sycp3^–/– chromosomes.

For all analyzed chromosomes, except Sycp3^–/– chromosome 19, the frequency of bivalents without MLH1 foci was low. Because bivalents without crossovers are generally rare, a mechanism ensuring at least one `obligate' crossover is sometimes assumed (cf. Jones and Franklin, 2006). However, for the analyzed wild-type and Sycp3^–/– chromosomes it is not necessary to assume such a mechanism besides interference itself, because we found in all cases ν values for which the observed numbers of bivalents with 0, 1, 2, 3, and 4 foci fitted with the expected numbers. However, a mechanism ensuring at least one MLH1 focus per bivalent cannot be excluded.

Because the reconstruction of bivalents in Sycp3^–/– oocytes was particularly laborious (see Fig. S1 in supplementary material) (Wang and Höög, 2006), we could analyze only a limited number of bivalents. Nevertheless, the results clearly show that neither full synapsis nor intact axial elements are required for wild-type levels of interference in mouse Sycp3^–/– oocytes. Furthermore, the SYCP3 protein and the proper localization of SYCP2 are not essential for interference, which fits with a preliminary analysis of the distribution of MLH1 foci among selected SCs with (almost) full synapsis in Sycp3^–/– oocytes (Wang and Höög, 2006).

The SC and synapsis play a major role in early interference models (Egel, 1995; Fung et al., 2004; Roeder, 1997). However, we found no difference in interference level across synapsed or (partially) unsynapsed stretches, which argues against a role of synapsis in interference, or an effect of a delay in synapsis on the strength of interference (Chua and Roeder, 1997; Conrad et al., 1997). Thus, although we cannot formally exclude the fact that the Sycp3^–/– cohesin axes were briefly fully synapsed when interference was imposed, our results add to a growing list of diverse observations arguing against a role for synapsis in interference (reviewed by de Boer and Heyting, 2006).

Support for a role of AEs in interference comes from the Caenorhabditis elegans him-3(me80) mutant. HIM-3 is thought to be a structural AE protein (Zetka et al., 1999); the C. elegans him-3(me80) mutant incorporates reduced amounts of HIM-3 in AEs, and shows reduced levels of crossover interference (Nabeshima et al., 2004). This might suggest that an intact AE structure is essential for wild-type interference levels. However, the mouse Sycp3^–/– phenotype argues against this: despite grossly abnormal AEs, Sycp3^–/– oocytes show wild-type levels of interference among MLH1 foci. Also, preliminary evidence in Tetrahymena suggests crossover interference in the absence of both SCs and AEs (Loidl and Scherthan, 2004). It is possible that the C. elegans HIM-3 protein has roles in interference other than a structural one, or there is a species-specific difference in the role of the intact AE structure in interference.

If neither intact SCs nor intact AEs function in interference, which structure (if any) does? One recent interference model dealing with this question is the stress model (Kleckner et al., 2004), according to which, mechanical stress is generated by expansion of chromatin against some chromatin-constraining structure. Mechanical stress would promote covalent changes in recombination intermediates that lead to crossover formation, and if such a change would occur, it would trigger local stress relief around the involved recombination complex. The stress relief would spread along the bivalent and decrease the probability of the formation of additional crossovers nearby, thus bringing about crossover interference. In this model, the relevant chromatin-constraining structure is essential for generating interference (by withstanding expanding chromatin), and might also be important for conducting the interference signal along the bivalent. The mouse Sycp3^–/– phenotype shows that the intact AE structure is required neither for generating interference nor for guiding the interference signal. In the context of the stress model this is not surprising, because Kleckner et al. (Kleckner et al., 2004) suggested that expansional stress and stress relief might function in the coordination and spacing of events not only in meiosis, but also in the mitotic cycle, when AEs are not formed. One candidate structure that might be involved in generating and guiding interference is the chromosome scaffold.

Unexpectedly, some MLH1 foci in Sycp3^–/– oocytes localized to converged cohesin axes without detectable SYCP1 (Fig. 1A-F). This appears to contradict the dependence of MLH1 focus formation on SYCP1 (de Vries et al., 2005). Possibly, SYCP1 colocalized with MLH1 in Sycp3^–/– oocytes at earlier time points than our measurements, and/or minute amounts of SYCP1 colocalized with MLH1 foci in Sycp3^–/– mice. That would be consistent with the proposed role of yeast TF protein Zip1 in crossover formation, besides its role in synapsis (Storlazzi et al., 1996).

All antibodies used have been described (de Vries et al., 2005). Oocytes of wild-type and Sycp3^–/– mice were isolated, spread and incubated for immunofluorescence and FISH as described (de Boer et al., 2006; de Vries et al., 2005). We measured SCs and cohesin axes and processed the images as described (de Boer et al., 2006).

Estimation of the interference parameter ν and correction of the ν estimates (Table 2) for the limited range of observable interfocus distances have been described (de Boer et al., 2006); the correction results in slightly lower estimates of ν. We determined the distribution of the numbers of foci per bivalent expected for integer values of ν using a macro in Excel (Microsoft), which is available on request from the corresponding author.

We thank N. Vischer (University of Amsterdam) for the Object Image program, F. Lhuissier (Wageningen University) for adapting the program, the animal facilities at Wageningen University for expert technical support, and H. Offenberg and F. Lhuissier for useful comments. The Swedish Research Council financially supported C.H.