Genome-Wide Control of the Distribution of Meiotic Recombination

Meiotic recombination events are not randomly distributed in the genome but occur in specific regions called recombination hotspots. Hotspots are predicted to be preferred sites for the initiation of meiotic recombination and their positions and activities are regulated by yet-unknown controls. The activity of the Psmb9 hotspot on mouse Chromosome 17 (Chr 17) varies according to genetic background. It is active in strains carrying a recombinant Chr 17 where the proximal third is derived from Mus musculus molossinus. We have identified the genetic locus required for Psmb9 activity, named Dsbc1 for Double-strand break control 1, and mapped this locus within a 6.7-Mb region on Chr 17. Based on cytological analysis of meiotic DNA double-strand breaks (DSB) and crossovers (COs), we show that Dsbc1 influences DSB and CO, not only at Psmb9, but in several other regions of Chr 17. We further show that CO distribution is also influenced by Dsbc1 on Chrs 15 and 18. Finally, we provide direct molecular evidence for the regulation in trans mediated by Dsbc1, by showing that it controls the CO activity at the Hlx1 hotspot on Chr 1. We thus propose that Dsbc1 encodes for a trans-acting factor involved in the specification of initiation sites of meiotic recombination genome wide in mice.


Introduction
During the first meiotic prophase, chromosomes undergo a series of unique events that lead to the formation of stable connections between homologous chromosomes (homologs), which are required for the reductional segregation at the first meiotic division.In most species, these connections result from the formation of at least one reciprocal recombination event or crossover (CO) per chromosome arm between chromatids from homologs, stabilized by the maintenance of cohesion between sister chromatids [1].In addition to this mechanical role, COs increase genetic diversity by reshuffling alleles in the genome and thus are thought to increase the efficiency of selection [2].
The main lines of the mechanism of CO formation have been uncovered in Saccharomyces cerevisiae and Schizosaccharomyces pombe and are conserved in other eukaryotes [3].Meiotic recombination is induced by the formation of localized DNA double-strand breaks (DSBs), catalyzed by the evolutionary conserved Spo11 protein at the leptotene stage, corresponding to the beginning of the meiotic prophase [4].These DSBs are repaired by homologous recombination preferentially with a chromatid from the homologous chromosome.This homologous repair leads to gene conversion with CO or to gene conversion without CO (noncrossover or NCO).One important implication of this mechanism is that the number and distribution of COs depend both on the number and distribution of DSBs and on the proportion of DSBs repaired towards COs or NCOs [5,6].
Direct analysis of DSB events in yeasts has shown that initiation occurs nonrandomly in the genome and preferentially in small regions known as recombination hotspots.In S. cerevisiae, DSBs occur almost exclusively in intergenic regions adjacent to transcription promoters [7,8].When tested, the transcription activity of the adjacent promoter was found not to be required, although binding of transcription factors can stimulate DSB formation [9].At several hotspots, DSBs were found to occur in accessible regions of the chromatin [10,11].In addition to these local constraints, another control appears to act at the level of chromosomal domains.In fact, an active initiation site loses its activity when inserted in a region with no or low DSB activity, suggesting a role for higher order chromosome structure [12].In S. pombe, the ade6-M26 initiation site is located next to the binding site of the Atf1/Pcr1 transcription factor, and DSB formation is enhanced upon binding of Atf1/Pcr1, which induces a reorganization and modification of the chromatin [13][14][15].Open chromatin configurations were also detected at several other DSB sites [16].Recently, the global analysis of DSB formation in S. pombe has shown that DSBs are clustered in small intervals separated by large regions with low or no DSB activity.DSBs do not necessarily occur next to transcription promoters; they appear to preferentially occur in large intergenic regions [17].Thus, in both yeasts, the parameters that define an initiation site are not defined by simple features of the primary DNA sequence; they are somehow related to a control of accessibility.This notion is indeed supported by experiments in S. cerevisiae that show that DSBs can be induced by targeting Spo11 to a Gal4 binding site through the expression of a Gal4-Spo11 fusion protein [18].In such situations, DSBs occur adjacent to the Gal4 binding site.Interestingly, constraints imposed by chromosomal domains are still observed in this case since Gal4-Spo11induced DSBs do not form in domains with low natural DSB activity [19].
In other organisms, DSBs have not been detected directly and the distribution of recombination is derived from mapping COs, cytologically, genetically or by population diversity analysis [20,21].In humans, CO distribution, which has been analyzed throughout the genome at high resolution, is not random and shows a specific clustering in small intervals (up to 1-2 kb long) separated by larger regions (50-100 kb) with no or low CO activity [22,23].Interestingly, a common sequence motif has been identified in a substantial fraction of human recombination hotspots [24].Moreover, when measured over several Mb-long domains, CO activity varies, some domains showing high (jungle) or low (desert) CO rates [25,26].In mice, low-resolution analysis of CO also suggests similar large-scale constraints [25,27].Several hotspots have been localized and shown to have similar properties to those described in the human genome, in particular with a tight clustering of COs in 1-2-kb intervals [21,28,29].A recent high-resolution CO map of mouse Chromosome 1 (Chr 1) has shown that this pattern of distribution of recombination is indeed a general feature in the genome and has led to the identification of several new hotspots in the mouse genome [30].
Interestingly, variations of CO activities between individuals, sexes, and populations in human, or strains and sexes in mice have provided additional clues to identify factors that contribute to hotspot localization and activity [31].The difference in CO activity between sexes reveals a DNA sequence-independent control of CO sites for which the mechanism, potentially related to epigenetic modifications, remains to be understood.Variations in genetic maps in different mouse hybrids have been observed [32] and could be the consequence of heterozygosity and/or of genetic factors.The contribution of genetic determinants on CO distribution and hotspot activity has been recently deduced from the analysis of interindividual variations in humans [33].Varia-tion of hotspot activity according to genotypes has also been shown at several sites in the mouse genome [29,[34][35][36].
In order to analyze the molecular basis of the control of hotspot activity, we have taken advantage of the unique properties of the Psmb9 hotspot located on Chr 17 in the mouse genome.Psmb9 hotspot activity is influenced by several factors such as sex and genotype [37].We have shown that both cisand trans-acting elements control the initiation activity of Psmb9.The regulation in trans was deduced from the observation that a specific haplotype on Chr 17 on one homolog could activate initiation at Psmb9 on the other homolog [38].In the present study, by generating new recombinant chromosomes, we have been able to locate the genetic element responsible for Psmb9 activity within 6.7 Mb on Chr 17.We show that this element actually acts by regulating the distribution of COs genome wide.Specific analyses on Chr 17 suggest that the regulation acts at the level of initiation.These results uncover a novel genetic control of the distribution of meiotic recombination events, predicted to influence where initiation occurs without altering the total CO activity along chromosomes.A similar control depending on a locus mapped in an overlapping region is described in the accompanying paper (Parvanov et al. [39]).

Results
The Psmb9 Hotspot Is Controlled by a Long Distance-Acting Element The activity of the Psmb9 hotspot has been shown to depend on the presence of the wm7 haplotype derived from an isolate of Mus musculus molossinus [36,40].Both Psmb9 and the wm7 haplotype are located on Chr 17 and the wm7 fragment required for Psmb9 activity extends from the centromere of Chr 17 to the center of the hotspot [37,38].The wm7 haplotype needs to be present in only one parental chromosome and leads to a 2,000-fold increase in CO activity at Psmb9 in hybrids in which the homologous chromosome can be from different genetic origins [41].The molecular analysis of CO and NCO events at Psmb9 in hybrids with or without a chromosome carrying the wm7 haplotype showed that the variation of CO frequencies were most likely due to variations in initiation frequencies, given the parallel variations of NCO events.Furthermore, these analyses revealed that the presence of the wm7 haplotype on one homolog was sufficient to induce initiation at Psmb9 on both homologs, indicative of a regulation acting in trans [38].
To determine more precisely the position of the genetic element responsible for the high level of recombination at the Psmb9 hotspot, we generated recombinant chromosomes carrying shorter wm7 fragments by genetic crosses.The starting strain was B10.A(R209), named R209.In this strain, the wm7 haplotype extends from the centromeric region of Chr 17 to Psmb9.R209 was crossed with C57BL/10 (B10), and progeny carrying recombined Chr 17 were screened.The extent of the wm7 fragments in the progeny were determined by genotyping using microsatellite markers, and the derived strains carrying recombined Chr 17 were mated with either B10 or B10.A in order to generate hybrids in which sperm DNA could be assayed for COs and NCOs at Psmb9.Six new strains carrying different segments of the wm7 haplotype (R5, RJ3, R115, RB2, RJ2, and RK2) were thus generated (Figure 1).
Unexpectedly, in the hybrids B10xR115 and B10xRJ3, neither CO nor NCO products could be detected at Psmb9, indicating that in these hybrids, initiation activity had been lost on both homologs and that the sequences of wm7 surrounding the hotspot are not sufficient for inducing its activity.On the R115 chromosome, the wm7 fragment extends 10.2-14.1 Mb to the left of Psmb9.In contrast, B10xR5 hybrid had high CO and NCO frequencies at Psmb9 indicating that a remote element was required for Psmb9 activity.Indeed, in the RB2xB10.A hybrid in which the RB2 chromosome contains a wm7 fragment from the centromeric region up to 9.3-10.2Mb from Psmb9, high frequencies of COs and NCOs, similar to the ones of B10xR209, were detected, demonstrating that an element from the wm7 haplotype located more than 10.2 Mb away from the hotspot is necessary and sufficient for its recombinogenic activity.Two additional recombinant Chr 17, derived from RB2 (named RJ2 and RK2), and carrying shorter wm7 fragments, retain full Psmb9 recombinogenic activity.Therefore, the genetic element necessary and sufficient for the activation of the Psmb9 recombination hotspot is located in a 6.7-Mb region of the wm7 fragment between positions 10.1 and 16.8 Mb of Chr 17.

The Psmb9 Control Element Influences CO Activity in Several Other Regions on Chr 17
In order to analyze the mechanism involved in Psmb9 activity, we designed an approach to determine whether the effect of the wm7 haplotype was specific to Psmb9 or if it could affect other regions on Chr 17.In order to address this question, we monitored CO activity by the detection of MLH1 foci on Chr 17 synaptonemal complex (SC) by immunofluorescent in situ hybridization (immuno-FISH) (Figure 2A).MLH1 foci are detected from middle to late pachytene, localize to future sites of chiasmata [42], and have the same global distribution as COs obtained from the genetic map [43].Given the high CO frequency at Psmb9, which is 1.9 6 0.6% in sperm DNA from the RB2xB10.A hybrid (Figure 1), one predicts that about 3.8 6 1.2% of pachytene spermato-cytes should have a MLH1 focus colocalizing with Psmb9 in this hybrid.We thus measured the positions of MLH1 and Psmb9 signals in B10xB10.A and RB2xB10.A hybrids.The signals were considered to colocalize when the centers of at least one Psmb9 focus and of the MLH1 focus on Chr 17 were separated by less than 300 nm (Figure 2B-2D).RB2xB10.A hybrid had a higher MLH1/Psmb9 colocalization frequency compared to B10xB10.A (4.7 6 1.2% vs. 1.4 6 1.2%; Fisher exact test, p-value ¼ 0.021).This suggests that 3.3 6 1.5% colocalization is due to the activity of Psmb9, which is in good agreement with the predicted value based on CO frequency.This analysis also indicates that the frequency of colocalization by chance is around 1.4%.The number of detectable Psmb9 foci per nucleus varied, as expected from one to four, depending on the relative positions of the four chromatids.RB2xB10.A mice showed a lower proportion of nuclei with two, three, or four foci as compared to B10xB10.A mice (Figure S1), indicating that the higher MLH1/Psmb9 colocalization frequency observed in the RB2xB10.A hybrid compared to B10xB10.A could not be explained by a higher average number of Psmb9 foci per nucleus.Taken together, these results validate our immuno-FISH approach and provide further evidence that MLH1 foci are faithful markers of CO events at our level of resolution.
We then tested whether the increased activity at Psmb9 could reflect a global increase of CO frequencies on Chr 17 by comparing the average number of MLH1 foci on Chr 17 in B10xB10.A and RB2xB10.A hybrids.These were found to be identical (Table 1).SC length was also identical in the two hybrids (Table 1).In order to identify potential differences in SC compaction, we determined the positions of three different bacterial artificial chromosome (BAC) probes along the SC of Chr 17.In both hybrids, all three BAC probes localized to the same positions (Figure S2, Table 1), indicating that chromosome compaction along the SC is not significantly different between B10xB10.A and RB2xB10.A hybrids.
In order to test whether the Psmb9 control element could and orange, SGR (haplotype wm7).The 95% confidence intervals for recombinant product frequencies are calculated as described in [38].NCOs were detected either at the BsrFI site or the site ''38,'' depending on which is heterozygous in each hybrid.Values for B10xB10.A and B10xR209 are from [38].doi:10.1371/journal.pbio.1000035.g001 affect CO frequencies elsewhere on Chr 17, we mapped MLH1 foci along SC of Chr 17 in B10xB10.A and RB2xB10.A hybrids (Figure 3A).Surprisingly, the global distribution of MLH1 foci on Chr 17 was different in the two hybrids (Figure 3B).We evaluated the differences in MLH1 focus distributions, using two independent statistical tests.In the first test, the data (MLH1 focus positions) were grouped in 20 intervals, each 5% of SC length (about 350 nm for this chromosome), corresponding to the resolution limit of the analysis.After pooling intervals when sample sizes were too low, chi-square tests were applied to compare these distributions (see Materials and Methods).In the second test, Kolmogorov-Smirnov, we compared directly the distributions of the focus positions along the SC.Both analyses revealed a statistically significant difference of MLH1 distributions between B10xB10.A and RB2xB10.A hybrids (Table 2), indicating that the Psmb9 control element affects CO distribution in several regions of Chr 17.In some intervals, the number of MLH1 foci was similar in both hybrids (for instance, intervals from 85% to 95% of SC length).In contrast, in other intervals, the RB2xB10.A hybrid displayed either a higher (region from 40% to 60% of SC length) or a lower density of MLH1 foci (region from 70% to 85% of SC length) as compared to B10xB10.A.At least two interpretations could be considered to account for these variations: either the unusually high recombination frequency at Psmb9 leads to compensatory changes elsewhere on the chromosome or the Psmb9 control element has independent effects in different regions of Chr 17.
In order to distinguish between these possibilities, we mapped MLH1 foci along Chr 17 in SGRxSGR mice.In fact, these mice contain the genetic element-activating Psmb9; however, the Psmb9 hotspot has low or no CO activity due to the presence of a repressive element located at the distal side of Psmb9 on the SGR chromosome [38].This was deduced from the analysis of B10xSGR and B10.AxSGR hybrids, and  leads to the prediction that the same control should operate in the SGRxSGR mice analyzed here.In SGRxSGR mice, using the Kolmogorov-Smirnov test, we found that the MLH1 distribution was significantly different from that of B10xB10.A, but not from RB2xB10.A (Figure 3C, Table 2).
In particular, the region 70%-85% showed a significant reduction of MLH1 foci in SGRxSGR compared to B10xB10.A (Fisher exact test, p ¼ 0.00037).These results suggest that the redistribution of CO events along Chr 17 observed in RB2xB10A is not a consequence of the high activity at Psmb9, is not specific to the hybrid RB2xB10.A, and is rather due to the presence of the wm7 regulatory element having multiple independent effects in several regions of Chr 17.
The wm7 Regulatory Element Influences DSB Formation in Several Regions on Chr 17 The analysis of cis and trans regulations of Psmb9 led to the conclusion that the high recombination activity at Psmb9 is due to a regulation of the initiation rate [28,38].In order to test whether the wm7 regulatory element in RB2xB10.A modifies the frequency of initiation events along Chr 17 or only alters the choice of DSB repair towards COs or NCOs, we mapped cH2AX foci, the phosphorylated form of H2AX, along this chromosome.Among the markers that have been used to mark all meiotic DSB repair events, such as RAD51, DMC1, RPA, MSH4, and cH2AX, cH2AX gave the most reliable signals with our immuno-FISH protocol.cH2AX is detected as large chromosomal domains at leptotene and zygotene [44] and forms small foci on the SC at the beginning of pachytene [45,46].The numbers of cH2AX foci, being in excess as compared to the number of MLH1 foci, are predicted to mark both CO-and NCO-designated events.This analysis shows an average number of 2.76 6 0.17 and 2.67 6 0.18 cH2AX foci per chromosome in B10xB10.A and RB2xB10.A hybrids, respectively.Interestingly, in both hybrids, the overall cH2AX distributions were significantly different from that of MLH1, with a larger fraction of foci on the proximal half of the chromosome (Figures 3-4, Table 2).This observation is similar to that described from the  comparison of MSH4 or RPA and MLH1 foci and is thought to reflect specific CO regulations such as interference and crossover assurance [47].In addition, the overall cH2AX distributions were significantly different in B10xB10.A and RB2xB10.A hybrids as determined by chi-square analysis (Table 2).Furthermore, regression analysis showed a significant correlation (R 2 ¼ 0.528; p , 0.01) between the variations in MLH1 and cH2AX foci distributions in these two hybrids (Figure 5).These results strongly suggest that the wm7 regulatory element is not simply inducing a redistribution of CO events on Chr 17 by regulating pathways of DSB repair but is influencing the formation of meiotic DSBs in several regions of this chromosome.We therefore named this genetic element, Dsbc1 for Double-strand break control 1 and the two alleles Dsbc1 wm7 (present in the wm7 haplotype) and Dsbc1 b (B10 strain).

Dsbc1 Regulates Crossover Distribution on Other Chromosomes
As the presence of the wm7 fragment leads to a redistribution of DSB and CO events over long chromosomal distances, we addressed the question of whether this regulation was specific to Chr 17 by monitoring CO distribution on other chromosomes.We thus analyzed MLH1 foci along Chrs 2, 15, and 18 in B10xB10.A and RB2xB10.A hybrids.For all three chromosomes, the total CO activity estimated by the average MLH1 number was identical in the two hybrids (Table S1).Using both chi-square and Kolmogorov-Smirnov tests, MLH1 distributions were found to be significantly different in B10xB10.A and RB2xB10.A on Chrs 15 and 18, showing that Dsbc1 acts in trans on different chromosomes.We did not detect significant differences on Chr 2, however (Figure 6, Table 2).
Dsbc1 wm7 Activates Recombination at the Hlx1 Hotspot on Chr 1 In order to show at the molecular level and at high resolution that Dsbc1 regulates recombination activity on another chromosome, we tested the recombination activity at the Hlx1 hotspot located on Chr 1 [30] in hybrids carrying the wm7 haplotype on Chr 17. Parvanov et al. [39] have shown that the activity of this hotspot requires an allele from Mus musculus castaneus at a genetic locus located on Chr 17 and named Rcr1 (see Discussion).Using an allele-specific PCR assay, we found a high CO activity at Hlx1 in the presence of the Dsbc1 wm7 allele (for one type of recombinant molecule, f(CO) ¼ 0.1 6 0.07%) and no detectable CO in the presence of Dsbc1 b (f(CO) , 0.015%; Figure 7).The frequencies of CO were found similar in the presence of Dsbc1 wm7 and in hybrids containing a Chr 17 from M. m. castaneus.

Discussion
We report the identification of a novel regulatory element of the distribution of meiotic recombination events located between positions 10.

The Psmb9 Recombination Hotspot Is Regulated by a Distant Locus, Dsbc1
The Psmb9 hotspot was known to be active in hybrids in which one parent carries the wm7 haplotype from M. m. molossinus, the region of the wm7 haplotype required for this activity extending from the centromere to the center of the hotspot.By genetic analysis, we show that the region necessary and sufficient for Psmb9 activity is located between positions 10.1 and 16.8 Mb of Chr 17, at least 17.5 Mb away from Psmb9, indicating a long-distance control.Interestingly, the Psmb9 hotspot was also found to be active in hybrids in which one parent carries the cas3 MHC haplotype derived from M. m. castaneus [48].This haplotype extends from the MHC to at least position 18.6 Mb towards the centromere (unpublished data) and therefore potentially overlaps with the regulatory element from the wm7 haplotype.It is important to note that in most hybrids tested, Dsbc1 is heterozygous, indicating a dominant or semidominant effect of the Dsbc1 wm7 allele over the Dsbc1 b allele from the B10 strain.Heterozygosity is not required, however, since we have shown high Psmb9 activity in hybrids homozygous at Dsbc1 wm7 [38].

Dsbc1 Controls Initiation of Meiotic Recombination
Based on measurements of CO and NCO frequencies, we infer that the enhanced recombination activity at Psmb9 results from an enhancement of initiation frequency.The only alternative hypothesis would be to assume that B10 strain has a high initiation frequency at Psmb9, but that all DSBs are repaired using the sister chromatid, an unlikely hypothesis given the repression of sister-chromatid recombination during meiosis.Given that the presence of Dsbc1 wm7 on one homolog is sufficient to promote initiation on both [38], this locus acts, not only at long distance, but also in trans.A complementary approach to follow initiation events during meiosis is through the detection of cH2AX foci which mark meiotic DSBs [44].Using this approach, we have shown that the distribution of cH2AX on Chr 17 is significantly different in B10xB10.A and RB2xB10.A hybrids, therefore in the absence or presence of the Dsbc1 wm7 allele.Although indirect, this observation is in agreement with the molecular analysis at Psmb9, and thus suggests that the Dsbc1 locus controls the initiation of meiotic recombination in multiple genomic regions on Chr 17.Through the analysis of MLH1 focus distribution on Chr 17, we detect significant differences between the hybrids tested and found that the variations of cH2AX distribution are strongly correlated with those of MLH1.We therefore consider that these cH2AX variations are not likely due to modifications of the pathway leading to H2AX Ser129 phosphorylation that would be independent from DSB formation.We thus propose that Dsbc1 regulates initiation activity and, therefore, CO rates.We do not

Dsbc1 Acts Genome Wide
In order to test whether the Dsbc1 locus acts on other chromosomes, we determined by immuno-FISH the localization of MLH1 foci on Chrs 2, 15, and 18.In all cases, the total CO activity as measured by the average number of MLH1 foci per chromosome was not changed.However, excepted for Chr 2, we found significant differences in the distribution of MLH1 foci along Chrs 15 and 18, depending on the presence or absence of the Dsbc1 wm7 allele.These differences are validated both by chi-square analysis after subdividing the data in intervals and by a comparison of the focus position distributions by the Kolmogorov-Smirnov test.In some regions, CO is increased, in others decreased, or not affected.The detection of these effects in single intervals is obviously limited by our assay in two respects: The spatial resolution and the statistical significance when sample size is small.The spatial resolution of MLH1 foci is close to 300-400 nm, corresponding to about 5 Mb or 1/20th of the chromosome length (Chrs 15, 17, and 18 are of similar lengths).If Dsbc1 controls initiation as proposed above, it might do so independently at single sites or over domains, or both, and the net result over 5 Mb intervals is the sum of these effects.In the case of Chr 2, the number of foci measured per unit of physical distance was lower than on other chromosomes, potentially weakening the statistical significance of the measures.Pooling data to normalize for focus number per interval did not reveal significant differences, possibly due to decrease of spatial resolution.Within these constrains, it is not possible to determine which fraction of the hotspots are influenced by Dsbc1.
As a direct molecular demonstration that Dsbc1 regulates, not only the Psmb9 hotspot, but also hotspots on other chromosomes, we tested the Hlx1 hotspot on Chr 1.We found that indeed the presence of Dsbc1 wm7 leads to a high CO frequency at Hlx1.This result not only demonstrates that Dsbc1 acts on other chromosomes, but also that it might correspond to the same locus as one identified by Parvanov et al. [39].In fact, the work by Parvanov et al. reports the identification of a locus, named Rcr1, whose position overlaps with the one of Dsbc1 and which influences initiation activity at Hlx1.More specifically, the Rcr1 allele from the Cast/Eij strain (M.m. castaneus) leads to a high CO frequency at Hlx1 as well as variations of CO rates at several hotspots in the neighboring region of Chr 1. Genetically, although the mouse strains used in both approaches were different, it is interesting to note that M. m. molossinus is a hybrid between M. m. domesticus and M. m. castaneus.It is therefore possible that the Dsbc1 wm7 allele is derived from M. m. castaneus.Alternatively, the two loci Rcr1 and Dsbc1 might be distinct.

How Does Dsbc1 Act?
Our analysis reveals that the strains from B10 background and from M. m. molossinus wm7 haplotype have two different alleles at the Dsbc1 locus, Dsbc1 b and Dsbc1 wm7 .Although the size of the genomic fragment from wm7 origin is large and differs in several hybrids, the consistency of the effects leads us to assume that these are under the control of this single locus.The presence of one Dsbc1 wm7 allele is sufficient to induce hotspot activity at some sites (Psmb9 and Hlx1, for instance).This allele thus behaves as a dominant trait specifying initiation at new locations.However, given that the total amount of recombination is constant (estimated by cH2AX and MLH1 foci), one expects the activity at other hotspots to decrease in the presence of Dsbc1 wm7 (heterozygous or homozygous).This could be due either to a regulation of the frequency of initiation events such as the one observed in S. cerevisiae in which DSBs can be suppressed by insertion of strong DSB sites nearby [19,49] or to a competition between Dsbc1 wm7 and Dsbc1 b alleles.These observations raise several questions: Is the Dsbc1 wm7 allele an inducer of new initiation sites?Does the Dsbc1 b play a role in initiation site specification?If so, which fractions of sites are controlled by the Dsbc1 locus?At the molecular level, Dsbc1 is expected to be a diffusible factor, RNA, or protein, that could act either on the recombination machinery involved in initiation or at the level of the substrate.One could, for instance, envision that Dsbc1 controls directly or indirectly some aspects of chromatin accessibility that are known to be important for initiation site specification as shown by studies in yeast [50].We have indeed obtained evidence that the structure of the chromatin at Psmb9 differs in the context of Dsbc1 wm7 or Dsbc1 b (J.Buard, P. Barthe `s, C. Grey, and B. de Massy, personal correspondence).rec-3, have been found to influence meiotic recombination in a region-specific way and were proposed to modify patterns of initiation [51].Recently, in Caenorhabditis elegans, mutations in a subunit of a condensing complex, DPY-28, were found to alter CO distribution by influencing DSB activity.Differently from the effect mediated by Dsbc1, dpy-28 mutants show an overall increase of CO frequencies [52].However, such observations indicate that modifications in structural components of chromosomes can affect recombination initiation, and Dsbc1 could therefore interact directly or indirectly with these components.
Genetic control of the frequency of meiotic recombination has been suggested from analysis of several hotspots in the mouse genome.For instance, a very active hotspot was identified in the Eb gene, for which activity was suppressed in the presence of the p haplotype [53,54].Although the Eb hotspot is active in many different genetic contexts, its activity is highly variable in different hybrids [29,54].The activity of the Eb hotspot was suggested to depend on distal elements or to be influenced by genetic background [34].Among other less-characterized hotspots some are haplotypespecific such as the Pb hotspot, active in the presence of the cas4 haplotype derived from M. m. castaneus [55,56], whereas others, such as G7c, have detectable activity in a variety of hybrids [57].Whether these variations depend on local or distal controls is not known.In humans, the regulation of CO activity by remote control has been proposed from the analysis of the MSTM1a hotspot showing variable CO frequencies in men with identical haplotypes in the 100-kb region around the hotspot [58].In addition, genome-wide mapping has shown a heritable variation in hotspot activity between individuals [33]; it would be interesting to know whether specific loci are responsible for such effects and whether any locates to the region of the human genome synthenic to that of Dsbc1.

Implications for Hotspot Evolution
If a single locus can modify initiation activity in such a dramatic way as we have uncovered, it obviously has consequences, not only on genetic maps as mentioned above, but also on their evolution.In particular, it provides an answer to the hotspot paradox.In fact, due to the directionality of mismatch repair during DSB repair in meiosis, the initiating chromatid is the recipient of genetic information.Consequently, in a population, if an allele appears that lowers initiation frequency in cis, and is located in the region subject to gene conversion, its frequency will tend to increase, and it will eventually be fixed.This phenomenon has been called the hotspot paradox because it is expected to lead to loss of hotspot activity in the genome and to strongly limit the appearance of new hotspots [59].In fact, we have already reported that at the Psmb9 hotspot, a suppressive mutation, as observed in the SGR strain, might have spread rapidly through the population due to the disparity of gene conversion and the very high efficiency of the Psmb9 site in this context [38].The forces that maintain hotspot activity remain unknown.If the activity of a hotspot can be influenced by elements outside the region that is subject to gene conversion, alleles that lower initiation frequency might not necessarily be driven to fixation, and new hotspots could appear independently from mutations in the region of the hotspot itself [60].We propose that elements such as Dsbc1 that regulate in trans the distribution of recombination sites can counteract the loss of activity at some sites.If true, this suggests that the Dsbc1 locus itself might be under specific selection to maintain the recombination activity at an appropriate level in the genome.

Figure 1 .
Figure 1.CO and NCO Events at Psmb9 Are Activated by a Long Distance-Acting Element from the wm7 Haplotype The origin of the Chr 17 fragments in the different recombined lines is represented: black, B10 (B10 or B6 for R115) (haplotype b); blue, A (haplotype a);and orange, SGR (haplotype wm7).The 95% confidence intervals for recombinant product frequencies are calculated as described in[38].NCOs were detected either at the BsrFI site or the site ''38,'' depending on which is heterozygous in each hybrid.Values for B10xB10.A and B10xR209 are from[38].doi:10.1371/journal.pbio.1000035.g001

Figure 4 .
Figure 4. Different Distributions of cH2AX Foci on Chr 17 in B10xB10.A and RB2xB10.A (A) Chr 17 of a pachytene spermatocyte labeled with SYCP3 (blue), cH2AX (red), Chr 17 probe (mix of BACs RP23-10B20, and RP24-67L15; green), and stained with DAPI (white).(B) Percentage of cH2AX foci per 5% SC intervals on Chr 17 in B10xB10.A and RB2xB10.A hybrids (n ¼ 549 and 513 foci, respectively).doi:10.1371/journal.pbio.1000035.g004 1 and 16.8 Mb on mouse Chr 17. Comparative analysis between several hybrids suggests that a single locus, named Dsbc1, is regulating the distribution of crossing over in several regions of Chrs 15, 17, and 18.Two different alleles of Dsbc1 appear to be present in B10 and mice derived from M. m. molossinus.Direct analysis of recombinant molecules shows that the M. m. molossinus allele of Dsbc1, Dsbc1 wm7 , stimulates the CO activity of two recombination hotspots, Psmb9 on Chr 17 and Hlx1 on Chr 1.We also demonstrate that Dsbc1 controls the distribution of cH2AX foci along Chr 17.We thus propose that the Dsbc1 locus controls the distribution of initiation of meiotic recombination genome wide in mice.

Table 2 .
Statistical Analysis of MLH1 and cH2AX Foci Distributions