Brca2–Pds5 complexes mobilize persistent meiotic recombination sites to the nuclear envelope

Meiosis is the process used by sexually reproducing organisms to generate haploid gametes. First, the parental chromosomes (homologs) are duplicated to form sister chromatids. The homologs then segregate during meiotic division I, and the sister chromatids are separated during division II. Prior to homolog segregation, a controlled number of DNA double-strand breaks (DSBs) are introduced at the entry into meiotic prophase I (Keeney, 2001). These DSBs are processed by the Mre11–Rad50–Nbs complex (MRN; Stracker and Petrini, 2011), which is required for homologous recombination. Homologous recombination uses the homolog as a template in an error-proof repair pathway (Kleckner, 1996; Handel and Schimenti, 2010). The homologs are aligned by the synaptonemal complex, which forms during prophase I (Carpenter, 1979a; Page and Hawley, 2001). Cohesins likely further stabilize homolog pairing, and most eukaryotes possess meiosis-specific cohesins (Dorsett and Ström, 2012). Many eukaryotes also have germline-specific lamins suggesting an involvement of the nuclear envelope in meiosis (Jahn et al., 2010; Link et al., 2013). The connections between homologous recombination, cohesins and lamins are currently mostly obscure.

Meiotic homologous recombination generates two repair products, noncrossovers and crossovers (Kleckner, 1996; Lichten, 2001). Noncrossovers result in the exchange of short stretches of DNA between the homologs near the DSBs and are the predominant outcome of homologous recombination. Crossovers lead to the reciprocal exchange of large portions of the chromosome arms between the homologs. Their formation is under precise control given that many organisms only generate an average of one crossover per chromosome arm (Allers and Lichten, 2001; Bishop and Zickler, 2004; Börner et al., 2004). Nevertheless, the majority of chromosome arms are expected to contain crossovers because they ensure proper chromosome segregation during meiotic division I (Kleckner, 1996).

Studies in fruit flies (Drosophila melanogaster) and other organisms have identified numerous genes with roles in homologous recombination (Lindsley and Sandler, 1977; Sekelsky et al., 2000). Among these are factors with a general role in homologous recombination including members of the Rad52 epistasis group such as the recombinase Rad51 (Ghabrial et al., 1998; Staeva-Vieira et al., 2003; Börner et al., 2004; McCaffrey et al., 2006). Rad51 is loaded onto the DNA by the Breast cancer, early onset, type 2 (Brca2) protein (Holloman, 2011). Two classes of genes have been shown to possess crossover-specific roles in Drosophila (Lindsley and Sandler, 1977). One class comprises the precondition genes, which encode factors with putative roles in the selection and/or protection of crossovers from resolution as noncrossovers (Blanton et al., 2005; Kohl et al., 2012). The second group consists of the exchange genes encoding factors that assemble into crossover-specific resolvase complexes (Joyce et al., 2009; Kohl et al., 2012).

Mounting evidence suggests that crossovers are repaired after noncrossovers in many organisms (Carpenter, 1987; Kleckner, 1996; Allers and Lichten, 2001; Börner et al., 2004; Yokoo et al., 2012). Electron microscopic and genetic studies in Drosophila have suggested that meiotic DSBs are incorporated into two types of recombination nodules with distinct morphologies and functions (Carpenter, 1975b; Carpenter, 1979a; Carpenter, 1979b; Carpenter, 1988). Early recombination nodules, which are small and ellipsoidal, appear early in pachytene, whereas late recombination nodules, which are large and spherical, appear later in pachytene. Numerous studies have suggested that late recombination nodules correspond to crossovers (von Wettstein et al., 1984; Carpenter, 1975b; Yokoo et al., 2012). Their maximal number is similar to genetically determined crossovers per genome and their numbers are reduced to the same extent that crossovers are in precondition-defective mutants. They also are sites of recombination-associated DNA synthesis (Carpenter, 1981). Further evidence for delayed crossover repair has been reported for worms, yeast and mammals, in which meiosis-specific DNA mismatch repair proteins accumulate during mid-pachytene at a subset of homologous recombination foci likely corresponding to crossovers (Svetlanov and Cohen, 2004; Oliver-Bonet et al., 2005; Yokoo et al., 2012). Although this temporal difference has been proposed to be due to a more complicated, and therefore slower, crossover repair (Börner et al., 2004), the delay could also serve to separate noncrossover and crossover repair for other reasons.

To test this, I studied the dynamics of meiotic homologous recombination in the female germ line of Drosophila using a fixation method that preserves MRN foci. The MRN complex accumulates at homologous recombination sites during early pachytene in pro-oocytes, but not in nurse cells. Although the majority of homologous recombination foci are repaired during early pachytene, up to five or six foci per pro-oocyte persist until mid-pachytene, when they move to the nuclear periphery. Their number corresponds to the one of late, large recombination nodules as well as crossovers (Carpenter, 1975b; Carpenter, 1979a).

The recombination mediator Brca2 is responsible for the peripheral localization of homologous recombination foci. Proteomic analyses of Brca2 complexes combined with other interaction studies have revealed that Brca2 associates with lamin and the chromosomal cohesion factor Pds5 (Dorsett and Ström, 2012). Pds5, which is enriched at the nuclear envelope, stabilizes the interaction between Brca2 and lamin and regulates the association of chromatin flanking persistent DSBs along the nuclear envelope. The loss of Brca2 or Pds5 disrupts the peripheral mobilization of persistent homologous recombination sites and causes a reduction in meiotic recombination frequency.

The data reported here suggest that persistent homologous recombination sites are exposed to a developmentally controlled signal causing their mobilization to the nuclear envelope during mid-pachytene. Because it is likely that persistent homologous recombination foci correspond to late recombination nodules (i.e. nascent crossovers), the delay of their repair would be crucial for their transport to the nuclear lamina, which could serve as a crossover-specific repair compartment. Given that a general block of homologous recombination in Rad51 mutants leads to the accumulation of all homologous recombination sites at the nuclear envelope, a repair delay would be sufficient for this selection. Brca2 mediates the peripheral mobilization of persistent homologous recombination sites and directly anchors them to the nuclear lamina. Furthermore, Brca2–Pds5 complexes likely stimulate the attachment of chromatin regions flanking persistent repair foci to the nuclear envelope through mechanisms involving sister chromatid cohesion. This work also provides a first mechanistic explanation for the involvement of cohesins and lamins in meiotic recombination and identifies Brca2 as key mediator of meiotic recombination at the nuclear envelope.

In Drosophila females, meiotic homologous recombination is initiated in the germarium when cysts of 16 interconnected cells enter early pachytene (Spradling, 1993). About 20–25 DSBs are introduced into the genomes of all 16 cells; however, only an average of about 14–15 foci are detectable due to a steady state of DSB formation and repair during early pachytene (Mehrotra and McKim, 2006; Joyce and McKim, 2009). Each cyst has two pro-oocytes, which fully develop a synaptonemal complex that persists in both until completion of homologous recombination (Carpenter, 1975a; Page and Hawley, 2001). The two nurse cells connected to the pro-oocytes through three ring canals also transiently form a synaptonemal complex, which disappears before the end of early pachytene. During late pachytene, one of the pro-oocytes differentiates into a nurse cell and loses its synaptonemal complex (Page and Hawley, 2001; Mehrotra and McKim, 2006; Joyce and McKim, 2009). The cysts then remain in late pachytene until more advanced stages of oogenesis.

A temporal difference in homologous recombination foci turnover as suggested by the presence of late recombination nodules has not been reported in studies using antibodies against the phosphorylated histone variant H2A.X (also known as γH2Av in flies) as a DNA damage marker (Mehrotra and McKim, 2006). Given that γH2Av is introduced into extensive chromatin regions flanking DSBs and its removal depends on the Drosophila Tip60 chromatin remodeling complex, its clearance from nucleosomes could be temporally disconnected from homologous recombination (Sedelnikova et al., 2003; Kusch et al., 2004). It therefore was necessary to develop a direct marker for meiotic DSBs. Mammalian MRN complexes multimerize at homologous recombination sites to microscopically discernible foci and persist there in homologous recombination mutants (Stracker and Petrini, 2011), making the MRN complex a good candidate for such a marker. For these reasons, antibodies against Drosophila Mre11 were developed.

After the confirmation of the specificity of the antisera (supplementary material Fig. S1a–e), ovaries were labeled with anti-Mre11 antibodies. To identify cysts entering pachytene, the ovaries were co-stained with antibodies against the synaptonemal complex component C(3)G, which labels the nuclei of pro-oocytes (Page and Hawley, 2001). Additional weak and transient signals are present in two nurse cells near the pro-oocytes during early pachytene. In ovaries fixed with standard protocols, elevated Mre11 levels were found in the nuclei of early pachytene pro-oocytes; however, no focal staining was observed (supplementary material Fig. S1f). For these reasons, buffer conditions were established that allowed the visualization of Mre11-positive foci. Surprisingly, such foci were only detectable in the nuclei of pro-oocytes from early pachytene onwards, but not in the surrounding nurse cells. Colabeling experiments with antibodies against Nbs or Rad50 showed that Mre11 colocalized with the other MRN subunits in these foci (Fig. 1A; supplementary material Fig. S2a–d). Statistical image analyses confirmed that the overlap between Mre11 and Rad50 or Nbs was nearly complete (Pearson's overlap coefficients r = 0.999; Mander's overlap coefficients = 1 for both channels). Co-staining experiments with anti-γH2Av antibodies revealed a similar degree of overlap in early pachytene pro-oocytes, suggesting that the MRN foci correspond to meiotic DSBs (Fig. 1B; supplementary material Fig. S2e; Pearson's overlap coefficient r = 0.998). This is supported by the fact that their average number of 14.2±2.9 (mean±s.d.; n = 80) was consistent with previous studies (Mehrotra and McKim, 2006).

The experiments also revealed that some homologous recombination foci persisted longer in pro-oocytes, but not in nurse cells. In cysts with eight or fewer homologous recombination foci per pro-oocyte, nearly all nurse cells lacked γH2Av foci (Fig. 1A–C; supplementary material Fig. S2c–f; n = 40). Most of these persistent foci were in the nuclear periphery in these pro-oocytes, which was confirmed by colabeling experiments with anti-lamin antibodies (Fig. 1B,C; supplementary material Fig. S2e,f).

The newly developed fixation buffer made it possible to lower the concentrations of antibodies against γH2Av by at least tenfold, which could be due to a more open chromatin structure at or near meiotic DSBs. Although this might also explain why MRN foci became detectable, these foci were only found in pro-oocytes, but not nurse cells. In mammals, the accumulation of MRN at homologous recombination sites is under control of a DNA damage amplification signaling pathway that involves H2Av phosphorylation and DNA repair factors of the Rad52 epistasis group (Stracker and Petrini, 2011). A similar feedback mechanism might be utilized in pro-oocytes in Drosophila. This was tested in H2Av mutants that express a truncated form of H2Av that lacks the DNA-damage-dependent phosphorylation site. These mutants show no γH2Av signals in germaria (Mehrotra and McKim, 2006). In germaria from these females, Rad50 or Mre11 accumulated in higher levels in pro-oocytes, but no foci were detectable (supplementary material Fig. S2h). Next, mutants for the spindle class genes, spnB, spnC (also known as mus301) and spnD were tested. All mutants show persistence of meiotic homologous recombination foci and an activated meiotic checkpoint (Ghabrial et al., 1998; Mehrotra and McKim, 2006). SpnB is homologous to mammalian XRCC3, whereas SpnD is related to Rad51C (Sekelsky et al., 2000). Both are required for MRN foci in mammals (Stracker and Petrini, 2011). The spnC homolog, HELQ, has unclear roles in MRN accumulation at homologous recombination sites, but interacts with regulators of this process. Germaria from mutant females for each gene showed persistence of γH2Av foci in both nurse cells and pro-oocytes in late mid and late pachytene cysts (supplementary material Fig. S2h–j). All showed elevated levels of Mre11 in pro-oocyte nuclei, but focus formation was affected. These results suggest that MRN focus formation is under control of a DNA damage signal amplification pathway similar to the one reported for mammals. This amplification pathway appears to be upregulated to sufficiently high levels for the detection of MRN foci only in pro-oocytes.

The number and turnover of persistent meiotic DSBs shown here are similar to those of late recombination nodules, which presumably correspond to nascent crossovers (Carpenter, 1979b). These parallels suggest that the MRN- and γH2Av-positive persistent foci correspond to late recombination nodules (although confirmatory immunoelectron microscopy studies would be necessary). Given that many studies refer to mid-pachytene to describe meiotic cells engaged in recombination (Carpenter, 1984; Börner et al., 2004; Svetlanov and Cohen, 2004; Joyce and McKim, 2009), this term will be used in the following to refer to cysts with mostly peripheral, late homologous recombination foci.

The peripheral accumulation of some meiotic DSBs implies that meiotic chromosomes redistribute during pachytene. This became apparent by the changed distribution of the synaptonemal complex. In early pachytene pro-oocytes with 14–15 foci, synaptonemal complex threads were randomly distributed throughout the nucleus. By contrast, cysts with five or fewer peripheral homologous recombination foci had large portions of the synaptonemal complex aligned along the nuclear envelope (Fig. 1A,C; supplementary material Fig. S2).

To assess when homologous recombination foci accumulated peripherally, the number of peripheral versus total foci was determined as a function of relative cyst position along the antero-posterior axis of the germarium. Given that cysts are produced in the tip of the germarium and migrate posteriorly, their relative position reflects, on average, their relative developmental age. Nevertheless, cysts are not produced at regular intervals, and not all developmental stages are always present in a given germarium (Carpenter, 2003). In addition, younger cysts can pass one another in the germarium. It therefore was necessary to determine the number of total versus peripheral Mre11 and γH2Av foci from 80 germaria (supplementary material Fig. S3). In early pachytene cysts with 14–15 foci, about two foci were at the nuclear envelope. Although their number remained constant in more posteriorly located early pachytene cysts, more advanced cysts with reduced γH2Av foci in the nurse cells had three or four peripheral foci in pro-oocytes. Older cysts lacking nearly all γH2Av foci in nurse cells had about five peripheral foci per pro-oocyte. These data suggest that the peripheral movement of meiotic chromosomes occurs while up to five or six DSBs are present in the pro-oocyte genomes. The overrepresentation of cysts with an average of five homologous recombination foci suggests that there is a transient delay in DSB repair during the early to mid-pachytene transition, which likely coincides with the peripheral mobilization of meiotic chromosomes. A previous study has also reported that the majority of cysts in comparable positions had about five or six γH2Av-positive foci, although no differentiation between nurse cells and pro-oocytes was made (Joyce and McKim, 2009).

To assess statistically whether the peripheral mobilization of persistent DSBs indeed occurs during advanced pachytene, the positions of all homologous recombination foci within three spheres of equal volume in nuclei from cysts in early pachytene (>10 foci), early mid-pachytene (5–8 foci), and late mid-pachytene (1–4 foci) were determined. Any deviation from the null hypothesis assuming a random distribution of 33% per sphere was tested for statistical significance using χ² analyses. The data revealed that 21% of homologous recombination foci form near the nuclear periphery at onset of early pachytene (n = 198; P<0.001; Fig. 1D). This lower number could be explained by the fact that meiotic DSBs do not form in the chromocenters, which distribute along the nuclear envelope in pro-oocytes (Mehrotra and McKim, 2006). In mid-pachytene pro-oocytes with five to eight foci, 69% were the nuclear periphery (n = 124; P<0.001), whereas more advanced pro-oocytes with one to four foci had 86% at the nuclear envelope (n = 44; P<0.001). In both sample groups, homologous recombination foci were almost completely excluded from the center of the nucleus. Combined with the predominantly peripheral distribution of the synaptonemal complex in pro-oocytes with five to six foci, these data suggest that the movement of meiotic chromosomes occurs during the early to mid-pachytene transition (Fig. 1A,C).

The re-localization of persisting homologous recombination foci to the nuclear envelope is likely to be mediated by homologous recombination factors bound to meiotic DSBs. Brca2 homologs have been proposed to regulate aspects of recombination beyond the loading of Rad51 onto DNA at homologous recombination sites (Holloman, 2011). The Drosophila Brca2 homolog is required for meiotic homologous recombination (Klovstad et al., 2008) and therefore could mediate the targeting of persistent homologous recombination sites to the nuclear envelope. To systematically identify interaction partners of fly Brca2 that could provide links to this process, two cell lines expressing either N- or C-terminally Flag- and HA-tagged Brca2 were generated. Brca2 and associated proteins were then tandem-affinity purified and characterized by proteomics as previously published (Kusch et al., 2004; Ardehali et al., 2011; Kusch et al., 2014). Silver stains revealed that similar polypeptide mixtures were isolated in both purifications (Fig. 2A). Liquid chromatography tandem mass spectrometry (LC-MS/MS) studies identified about 150 Brca2-specific peptides in both purifications (Fig. 2B). About twice as many peptides for the chromosomal cohesion regulator Pds5 (Dorsett and Ström, 2012) were found in both preparations. APRIN (also known as Pds5b), a paralog of human Pds5, has also been shown to interact with Brca2 to assist it in somatic recombination (Brough et al., 2012). All other known cohesin subunits, except for Wapl, were also found in lower abundance in the purifications. Furthermore, both purifications contained three nuclear pore complex factors that are oriented towards the nuclear interior (Chow et al., 2012). Studies in yeast have revealed that unrepaired DSBs transiently locate to nuclear pores (Oza and Peterson, 2010). About 29% of peripheral homologous recombination foci localized to pores (n = 38), suggesting that nuclear pores are also only transiently targeted by persisting homologous recombination foci in flies (supplementary material Fig. S2k).

Immunoblotting experiments with anti-Pds5 antibodies confirmed the proteomic data (Fig. 2C). An association of Pds5 with Brca2 was also observed in co-immunoprecipitation experiments with anti-Brca2 antibodies as well as in reciprocal experiments with anti-Pds5 antibodies (Fig. 2C,D). Anti-HA immunoprecipitations in cells stably expressing C-terminally HA- and Myc-tagged Pds5 (Pds5HM) further confirmed the interaction. In these experiments, Pds5 associated with substantial amounts of other cohesin subunits. Their low abundance in the Brca2 interactome supports the hypothesis that Brca2–Pds5 complexes associate rather dynamically with cohesin rings in a chromosomal context.

To determine the regions of Brca2 interacting with Pds5, pulldown studies with maltose-binding protein (MBP) fusion proteins containing Brca2 fragments were performed (Fig. 3A). Of particular interest was the region that is conserved among all Brca2 relatives that contains the BRC repeats. BRCs are required for the loading of Rad51 onto single-stranded DNA (Carreira et al., 2009). These studies revealed that fragments containing the last two BRCs from fly Brca2 interacted with Pds5 (Fig. 3B). Further dissections of the BRC-containing region showed that single repeats failed to associate with Pds5 (Fig. 3C). To test whether the last two BRCs were necessary for the interaction, conserved amino acids required for the interaction between human Brca2 and human Rad51 were exchanged with alanine residues (Carreira et al., 2009; supplementary material Fig. S4a). As shown in Fig. 3D, the mutation of either BRC abolished the interaction with Pds5. Given that BRC1 contained the same conserved residues, but a BRC1–BRC2 fusion failed to bind, clearly the length of the spacers between the BRCs is relevant for Pds5 binding. In fact, a BRC1–BRC2 fragment with a shortened spacer interacted with Pds5 (Fig. 3D). Conversely, BRC2–BRC3 with an extended second spacer lost its ability to bind Pds5 (Fig. 3D).

To determine the regions within Pds5 contacting Brca2, GST fusion proteins of Pds5 fragments were next tested for their ability to bind to MBP–BRC2-3 (Fig. 3E). The studies revealed that a leucine-zipper-like domain of Pds5 interacted with BRC2–BRC3 (Fig. 3F). Given that leucine zippers mediate protein homodimerization, Brca2 might assist Pds5–Pds5 interactions. To test this hypothesis, a cell line co-expressing Flag–HA–Pds5 and Pds5–HA–Myc was used. Anti-Flag antibody immunoprecipitations on nuclear extracts from these cells confirmed that Flag–HA–Pds5 interacted with both Pds5–HA–Myc and Brca2 (Fig. 3G). Upon the RNA interference (RNAi)-mediated knockdown of Brca2 (supplementary material Fig. S4b), the association between Flag- and Myc-tagged Pds5 was disrupted. To confirm that the BRC repeats of Brca2 were essential for the interaction with Pds5, a cell line expressing Flag-tagged Brca2 with mutated BRCs was generated (Fig. 3H). Anti-Flag immunoprecipitations revealed that the BRC-defective Brca2 failed to associate with Pds5. In summary, the data reveal that two BRCs of Brca2 are not only required for Brca2–Pds5 interactions, but also mediate Pds5–Pds5 dimerization.

In humans cells, Rad51 and other homologous recombination factors accumulate at the nuclear envelope about 2 hours after DNA damage induction (Mladenov et al., 2006). Given that human Brca2–APRIN complexes have roles in homologous recombination (Brough et al., 2012), it was possible that fly Brca2–Pds5 complexes would redistribute in S2 cells after induction of DSBs. Immunostaining experiments revealed that Pds5 was already enriched in the nuclear envelope of cells prior to X-ray irradiation, whereas Brca2 was distributed throughout the nucleus (Fig. 4A). In irradiated cells allowed to recover for 2 hours, Brca2 had accumulated near the nuclear envelope. Quantitative immunoblotting experiments on nuclear matrix preparations revealed that the levels of nuclear-envelope-associated Pds5 were elevated by about 1.75-fold upon irradiation, whereas Brca2 increased by 7-fold (Fig. 4B). The RNAi-mediated depletion of Pds5 caused a 66% reduction of these signals, indicating that Pds5 mediates the DNA-damage-dependent association of Brca2 with the nuclear matrix.

Given that the three nuclear envelope-associated proteins, lamin, lamin B receptor and Otefin, were identified in the purifications of N-tagged Brca2 (Fig. 2B), Brca2 could directly associate with the nuclear envelope. These proteins were underrepresented in preparations of C-tagged Brca2, suggesting that the C-terminal tagging disrupts this interaction. This was confirmed by co-immunoprecipitation experiments, in which lamin only interacted with N-tagged Brca2 (Fig. 4C). Furthermore, MBP pulldown experiments revealed that the C-terminus of Brca2 interacts with recombinant GST-lamin (Fig. 4D). To assess whether Pds5 is required for the lamin-Brca2 interaction, anti-Brca2 immunoprecipitations on nuclear extracts from Pds5 RNAi (Pds5i)-treated cells were performed (supplementary material Fig. S4b,c). As shown in Fig. 4E, the interaction was indeed diminished in cells with reduced Pds5 levels, suggesting that Pds5 stabilizes the association of Brca2 with the nuclear lamina.

To assess whether Brca2 also dynamically redistributes during pachytene, ovaries from wild-type females were colabeled with antibodies against Brca2 and Pds5 (Fig. 4F; supplementary material Fig. S4d). During early pachytene, Brca2 distributed throughout the nuclear interior of pro-oocytes, in which it also accumulated more heavily compared to nurse cells. In mid-pachytene cysts, the highest levels of Brca2 were found at the nuclear envelope of pro-oocytes. Co-immunoprecipitations experiments on ovary extracts confirmed that Brca2 and Pds5 indeed interacted in ovaries (supplementary material Fig. S4e). In summary, these results support the hypothesis that Brca2 and Pds5 are responsible for the redirection and association of persisting homologous recombination foci with the nuclear envelope during mid-pachytene. This idea was tested next.

In yeast Rad51 mutants, the block of homologous recombination causes the accumulation of substantial numbers of DSBs at the nuclear periphery (Oza and Peterson, 2010). To test whether this was also the case for Drosophila, germaria from Rad51^−/− females were investigated next (Staeva-Vieira et al., 2003). Late pachytene oocytes from these mutants had indeed 80% of homologous recombination foci at the nuclear envelope (Fig. 5A; n = 57; P<0.001, supplementary material Fig. S4f). By contrast, Brca2^−/− females only had 14% of the persisting DSBs at the nuclear envelope (Fig. 5B; n = 55; P<0.001, supplementary material Fig. S4g). Immunoblots on ovary extracts from Brca2 mutants confirmed that Pds5 protein levels were unchanged in these mutants (supplementary material Fig. S4g′). The data reveal that Brca2 is responsible for the localization of persistent homologous recombination foci to the nuclear envelope. Persisting DSBs accumulated into larger foci in both Rad51 and Brca2 mutants, making statistical analyses difficult. Similar observations were made for mammals and in yeast, where persistent DSBs have been suggested to concentrate at so-called ‘repair centers’ (Lisby and Rothstein, 2004).

Given that the majority of Pds5^−/− cysts showed mitotic defects and died prior to meiotic prophase I, making results from mutant germ line clones unreliable, Pds5 was knocked down in the germ line by the expression of a Pds5-specific short hairpin RNA (shRNA) construct (supplementary material Fig. S4h,h′). In the ovaries from these females, the number of homologous recombination foci and levels of Brca2 protein were unchanged. The number of large homologous recombination foci in mid-pachytene pro-oocytes was also normal, however, only 7% located to the nuclear envelope (Fig. 5C; n = 63; P<0.001). These homologous recombination foci disappeared with a delay, indicating that Pds5 functions in their repair at the nuclear envelope.

To assess whether Brca2 and Pds5 were required for crossovers, Pds5^−/+, Brca2^−/+ and Brca2/Pds5 transheterozygous females were scored for the recombination rates between four marker genes on the third chromosome. In pds5^−/+ females, the frequency was reduced by 30–35%, in Brca2^−/+ females by 23–30%, and in Brca2/Pds5 transheterozygotes by 44–48% (Fig. 5D). The partially additive effect might be due to the general role of Brca2 in homologous recombination, although exchange mutants can have dominant-negative effects on recombination rates (Carpenter and Sandler, 1974; Baker et al., 1976). Brca2^−/− pro-oocytes expressing Pds5i constructs showed the same phenotypes as Brca2^−/− mutants, suggesting that Brca2 and Pds5 function in the same pathways during homologous recombination. The rather uniform drop of recombination rates between all loci is observed in heterozygous exchange mutants, suggesting that Brca2 and Pds5 primarily function in homologous-recombination-related processes.

Here, I report that a limited number of homologous recombination foci persist only in pro-oocytes and become mobilized to the nuclear envelope, presumably for their repair during mid-pachytene. Their number is similar to the number of late recombination nodules as well as the number of crossovers per genome. Although unique cytological markers for the identification of crossovers or noncrossovers are currently not available, mounting evidence suggests that late recombination nodules correspond to crossovers (Carpenter, 1979a; Kleckner, 1996; Allers and Lichten, 2001; Börner et al., 2004; Svetlanov and Cohen, 2004; Yokoo et al., 2012). In this context, it is pleasing that most late recombination nodules are near the nuclear envelope (A. Carpenter, personal communication). Although further confirmatory studies would indeed be necessary, the rather consistent evidence suggests an intriguing model in which the nuclear envelope might function as a compartment for crossover regulation.

The delay of crossover resolution combined with the developmentally orchestrated peripheral mobilization of meiotic chromosomes during mid-pachytene might be a simple, yet effective way to regulate the noncrossover and crossover differentiation during meiosis (for a model, see Fig. 5E).

It remains unclear whether the peripheral movement of meiotic chromosomes and persistent meiotic DSBs is under active regulation or is the consequence of the drastic thickening of the synaptonemal complex during later pachytene (Carpenter, 1975a; Carpenter, 1979a). Nevertheless, several lines of evidence suggest that meiotic chromosomes are transiently anchored at the nuclear envelope. Normally, the chromosomes of advanced late pachytene oocytes condense into the so-called karyosome in the center of the nucleus. By contrast, the karyosome from late pachytene oocytes of recombination-defective mutants remains aligned along the nuclear envelope (Ghabrial and Schüpbach, 1999; Staeva-Vieira et al., 2003). This distribution would also explain the observed accumulation of persistent homologous recombination foci in Rad51 mutants (Fig. 5A). The phenomenon was interpreted as a chromosomal redistribution in response to an activated meiotic checkpoint (Ghabrial and Schüpbach, 1999; Staeva-Vieira et al., 2003); however, a more suitable interpretation would be that the detachment of meiotic chromosomes is blocked by persistent meiotic DSBs. This finds support in studies on the Barrier of Autointegration Factor (BAF), which anchors meiotic chromosomes to the nuclear envelope (Lancaster et al., 2007). The phosphorylation of BAF releases the chromosomes from the nuclear envelope in late pachytene. The mutation of these phosphorylation sites causes the constitutive attachment of the chromosomes to the nuclear envelope in the absence of persisting meiotic DSBs. A similar phenotype has also been reported for females lacking NHK1, which phosphorylates BAF (Lancaster et al., 2007).

Interestingly, the karyosome attachment to the nuclear envelope was not observed in Brca2 or Pds5 mutants despite their meiotic defects (Barbosa et al., 2007; Klovstad et al., 2008). This is consistent with the chromosomal distribution defects reported here (Fig. 5B,C). The analyses also revealed that Pds5-defective oocytes exhibited karyosome condensation defects, indicating that Pds5 has a role in chromosomal compaction in oocytes.

Although it cannot be excluded that Brca2–Pds5 complexes have cohesion-independent roles in homologous recombination, it will be difficult to determine the exact roles of Brca2–Pds5 complexes owing to the complex and context-dependent roles of Pds5. My proteomic and interaction studies revealed that a substantial fraction of Pds5 in nuclear extracts associates with cohesion factors, whereas Pds5–Brca2 complexes might only do so when associated with homologous recombination sites (Fig. 2A,B). This speaks for the existence of at least two nuclear pools of Pds5 complexes. In this context, it is interesting that Pds5 has been shown to destabilize chromosomal cohesion when associated with Wapl, which was not detected in the Brca2 interactome (Shintomi and Hirano, 2009; Cunningham et al., 2012).

Given that Pds5 is most abundant at the nuclear envelope, it might assist Brca2-dependent anchorage of persistent homologous recombination sites at the periphery as suggested by its requirement for Brca2–lamin interactions (Fig. 4E). Proteomic analyses have shown that other cohesin subunits also associate with the nuclear envelope (Kallappagoudar et al., 2010), and it is possible that such interactions might stabilize interactions between the homologous chromatids. Considering that exchange between the chromatids requires their precise alignment over extensive regions around homologous recombination intermediates, it is tempting to speculate that the Brca2-mediated Pds5–Pds5 interactions might serve the simultaneous clamping of two cohesin rings around the homologs. Interestingly, DNA repair studies in mammals have provided evidence for a role of Brca2 in sister chromatid cohesion and/or alignment (Abaji et al., 2005; Brough et al., 2012).

Although lamins are known to have meiosis-specific roles, their general roles in nuclear envelope architecture make functional studies addressing their homologous recombination-specific roles rather difficult. Nevertheless, the loss of lamin in the female germ line of Drosophila causes egg patterning defects typically found in meiotic mutants with an activated meiotic checkpoint, which would support a role for lamin in homologous recombination (Ghabrial et al., 1998; Guillemin et al., 2001). Remarkably, the nuclear envelope of pachytene cysts undergoes substantial morphological changes. Early pachytene nuclei are irregular in shape and later become more spherical (Carpenter, 1975a). This might explain why lamin signals in early pachytene cysts appeared weaker and less defined compared to mid-pachytene cysts (supplementary material Fig. S2e,f). Pds5 and Brca2 also only had strong signals at the nuclear periphery during more advanced pachytene stages in pro-oocytes (Fig. 4F; supplementary material Fig. S4d).

This raises the fascinating possibility that the nuclear lamina functions as a crossover-specific subnuclear repair compartment that matures with pachytene progression. Recent studies have reported that a splice variant of the recombination factor Rad9 accumulates at the nuclear envelope of pro-oocytes in Drosophila, which would lend further support for an involvement of the nuclear lamina in recombinational repair (Kadir et al., 2012). Such a spatiotemporally controlled compartimentalization would greatly minimize the chance that noncrossovers could be accidentally repaired as crossovers, and might explain why noncrossovers are repaired before crossovers. The necessity for the spatiotemporal separation becomes evident in Rad51 mutants, in which all unrepaired homologous recombination foci are mobilized to the nuclear envelope (Fig. 5A). In yeast, persisting DSBs also localize to the nuclear envelope upon block of DNA repair (Kalocsay et al., 2009; Oza and Peterson, 2010). Although this peripheral movement has been suggested to be part of a DNA damage checkpoint, these studies used the homologous-recombination-based mating-type switching system, which involves crossover formation. It therefore cannot be excluded that homologous-recombination-based repair processes requiring crossover resolution are coupled to a temporally controlled peripheral mobilization signal. This is also supported by the accumulation of homologous recombination factors at the nuclear envelope of mammalian cells, whereas repair factors with roles in non-homologous end joining were not found there (Mladenov et al., 2006).

In summary, the data presented here, combined with other studies, support the hypothesis that the peripheral movement of persisting meiotic repair foci is essential during meiosis and homologous recombination. Although meiotic recombination is widely considered to be a special form of homologous recombination, all factors identified here function in both meiotic and somatic recombination. This might include cohesins and lamins. For example, patients suffering from laminopathies like the Hutchison–Wilford progeria syndrome exhibit increased DNA damage sensitivity and high frequencies of ovarian cancer similar to carriers of mutations in Brca2 (Kudlow et al., 2007; Link et al., 2013). Given that meiotic homologous recombination can be relatively easily followed and manipulated in Drosophila germaria, future studies in flies might clarify whether the nuclear lamina indeed functions as a subnuclear compartment for homologous recombination.

For antibody productions, the following fragments were fused with a C-terminal His6 tag using a laboratory-made vector: Brca2 (amino acids 1–425); Mre11 (CG16928; amino acids 121–620). Antibodies were produced by Covance Inc. (Princeton, NJ, USA). The following N-terminal MBP fusions of Brca2 were generated using a laboratory-made vector: Brca2-I (amino acids 1–425); Brca2-II (amino acids 324–722); Brca2-III (amino acids 622–971); Brca2-IV (amino acids 567–781); B1B2 (amino acids 567–706); B2B3 (amino acids 672–781); B1 (amino acids 567–604); B2 (amino acids 672–702); B3 (amino acids 746–781); truncated S1 (deletion of amino acids 646–670); expanded S2 (addition of amino acids 646–670 at position 745). Point mutations of BRCs are shown in supplementary material Fig. S4a. To generate N-terminal GST fusions, the following fragments of Pds5 were cloned into a laboratory-made vector: Pds5-I (amino acids 1–319); Pds5-II (amino acids 301–613); Pds5-III (amino acids 606–904); Pds5-IV (amino acids 891–1218); Pds5-V (amino acids 1–178); Pds5-Z (amino acids 138–319). GST–lamin was generated by cloning the full-length open reading frame into the laboratory-made GST vector. All primer sequences are available upon request. All fusion proteins were purified using manufacturers' standard protocols. 750 nmol of bacterial fusion proteins or 175 nmol of Pds5HM were used in interaction studies.

Co-immunoprecipitation assays were performed on nuclear extracts as previously described (Kusch et al., 2004; Kusch et al., 2014). For western blots, the following antibodies were used: anti-Brca2 (guinea pig, 1∶4000; this study); anti-Pds5 (rabbit, 1∶2000); anti-Nipped-B (guinea pig, 1∶4000); anti-Stromalin (SA; rabbit, 1∶2000); anti-Rad21 (rabbit, 1∶2000; all cohesin-specific antibodies were gifts of Dale Dorsett, Saint Louis University, MO); anti-Myc (mouse 9E10, 1∶2000, DSHB, Iowa City, IA); anti-HA (mouse 12CA5; 1∶2000, Sigma, Saint Louis, MO); anti-lamin (mouse ADL67.10, 1∶1000, DSHB); anti-tubulin (mouse, 1∶2000, Sigma); anti-Flag M2 (1∶1000, Sigma); anti-GST (rabbit, 1∶1000; Rockland, Boyertown, PA) and anti-MBP (rabbit, 1∶2000; Rockland); anti-Rad50 (rabbit, 1∶1000, a gift from Mauro Gatti, La Sapienza, Rome, Italy; Ciapponi et al., 2004) and anti-Nbs [rabbit, 1∶2000; a gift from Yikang Rong, NIH, Bethesda, MD (Gao et al., 2009)] antibodies. Band density quantifications were performed using ImageJ (NIH) as previously published (Ardehali et al., 2011).

S2 cells and ovaries were fixed in 20 mM HEPES pH 7.1, 7.5–10 mM EGTA, 1.5–4 mM MgSO₄ with 2–4% electron microscopy-grade formaldehyde at room temperature for a total of 20 min. The ovaries were dissected in fixation buffer, which was required for the preservation of MRN foci and nuclear architecture. The usage of PBS, Buffer A, modified Robb's Buffer or PIPES-based buffers caused the loss of MRN foci. Immunostaining protocols have been published before (Kusch et al., 2004; Ardehali et al., 2011). The following antibodies were used: anti-Brca2 (guinea pig; 1∶400); anti-Pds5 (rabbit; 1∶400); anti-Mre11 (guinea pig; 1∶15,000, this study); anti-Rad50 (rabbit, 1∶1000); anti-Nbs (rabbit, 1∶2000); anti-C(3)G [mouse, guinea pig, 1∶2000; gifts from R. Scott Hawley, Stowers Institute, Kansas City, MO (Page and Hawley, 2001)]; anti-γH2Av (rabbit, 1∶20,000; Rockland); anti-lamin (mouse; 1∶500); and anti-GP210 (mouse AGP78.20, 1∶500; DSHB) antibodies. Microscopy was performed on a Deltavision II Deconvolution system (Applied Precision, Issaquah, WA). Image reconstructions were performed with Huygens Essential (SVI, Hilversum, The Netherlands) using the CMLE method. Maximum intensity projections (MIPs) of the reconstructed image stacks were generated with ImageJ.

Student's t-tests were performed as paired two-tailed tests assuming equal variance. The significance of signal overlaps was calculated on z-stacks of germaria using the JACoP plugin for ImageJ (Bolte and Cordelieres, 2006). A complete colocalization of two signals exists when Pearson's or Mander's colocalization coefficients are 1.

For subnuclear distribution analyses of homologous recombination foci, the average diameter of pro-oocyte nuclei was determined using the Deltavision measurement tools (3.5±0.2 µm; ±s.d.; n = 50). Three equivoluminal spheres spanning 0–220 nm (outer), 220–534 nm (middle) and 534–1750 nm (inner) from the nuclear lamina were calculated. The position of homologous recombination foci relative to the nuclear periphery was determined from single planes of 0.1 µm z-stacks. For χ² analyses, the actual positions of foci were compared to a random distribution of 33% of foci per sphere. Ambiguous stages between early and mid-pachytene were excluded. These cysts had less than ten homologous recombination foci in both nurse cells and pro-oocytes.

The full-length open reading frames of Brca2 (Brough et al., 2008; Klovstad et al., 2008), and Mre11 (Gao et al., 2009) were amplified from 1 µg of total RNA. For the generation of S2 cell lines, error-free clones were cloned into FLAG–HA- or HA–FLAG expression vectors (Kusch et al., 2004; Kusch et al., 2014). The open reading frame of Pds5 was amplified from a cDNA clone (a gift from Dale Dorsett), then cloned into the HA–Myc expression vector. The latter was generated by replacing the HA–FLAG cassette in the published vector by PCR. The mutations of the BRCs in Brca2 were performed by site directed mutagenesis of three conserved residues in the BRCs (supplementary material Fig. S4a). The generation of stable transgenic Drosophila S2 cell lines and tandem-affinity purifications were described before (Ardehali et al., 2011; Kusch et al., 2014). The affinity-purified proteins were separated in a 5–30% glycerol gradient and immobilized in a 14% Tris-Tricine gel. Single gel slices were excised for mass spectrometric analyses. For the purification of full-length Pds5–HA–Myc for interaction studies, HA beads were washed in nuclear extraction buffer containing 1 M NaCl and 1% Triton X-100. The purity of the samples was assessed by silver stain and immunoblotting analyses. The proteomic analyses were performed using the Global Proteome Machine (Craig et al., 2004; Ardehali et al., 2011; Chalkley, 2013). A search of mass spectrometry data was performed against the latest releases of databases for the Drome.fasta (Uniprot.org) and common contaminant proteins (cRAP.fasta, GPM common Repository of Adventious Proteins). The search parameters are available upon request. The accuracy of the data was further assessed by comparison to negative controls from cells carrying empty vectors and other datasets from unrelated proteins (Kusch et al., 2004; Ardehali et al., 2011). Irradiation experiments were conducted as described previously (Kusch et al., 2004). Nuclear matrix fractions were prepared as published (Mladenov et al., 2006).

For loss-of-function studies, females carrying mutations in spnA/Rad51 (SpnA¹; Staeva-Vieira et al., 2003), spnB (spnB^BU; Ghabrial et al., 1998), spnC/mus301 (mus301^D1; McCaffrey et al., 2006), spnD (spnD¹⁵⁰; Ghabrial et al., 1998), or Brca2 (Brca2^KO, Brca2^56E; gifts from Gertrud Schüpbach, Princeton University, Princeton, NJ; Klovstad et al., 2008) were used. H2Av⁸¹⁰; H2Av^ΔCT flies have been previously described (Clarkson et al., 1999). For Pds5-RNAi in the germ line, females carrying P{w^+mC = otu-GAL4::VP16.R}1, w*; P{ w^+mC = GAL4-nos.NGT}40; P{w_+mC = GAL4::VP16-nos.UTR}CG6325^MVD1; y¹ sc* v¹; P{y^+t7.7 v^+t1.8 = TRiP.GL00479}attP2 were analyzed. For the determination of recombination frequencies, st¹ Sb^sbd-1 e^s ro¹ ca¹/+, pds5^E3/+; st¹ Sb^sbd-1 e^s ro¹ ca¹/+, Brca2^KO/+; st¹ Sb^sbd-1 e^s ro¹ ca¹/+, or pds5^E3/Brca2^KO; st¹ Sb^sbd-1 e^s ro¹ ca¹/+ females were crossed to ru¹ h¹ th¹ st¹ cu¹ sr¹ e^s ca¹ males. The resulting progeny were scored for recombination between st, Sb, e and ca. Flies were raised on standard cornmeal-molasses-agar medium food at 25°C.

RNAi in tissue culture was performed as previously published (Kusch et al., 2004; Ardehali et al., 2011). RNAi treatments were conducted for 72 h. The following primers were used (T7 promoter sequences are omitted): Brca2i-fwd, 5′-GCGATAGCATGCAATCGGTTCCCAC-3′; Brca2i-rev, 5′-GCCATTAGAGCACTGCATTGGTTGCC-3′; Pds5i-fwd, 5′-CCACTCTCTTGTCTCGTATGTTCTCC-3′; Pds5i-rev, 5′-GCTCTTCCGGAGCTAGTTTATAGGG-3′; Mre11-fwd, 5′-CGGTAGCCCAACACACCGGTCATCC-3′; Mre11-rev, 5′-CACCTTCAAGCCCTGTGTGTCAAGC-3′.

I thank Dale Dorsett, Maurizio Gatti, R. Scott Hawley, Kim McKim, Yikang Rong, Gertrud Schüpbach, the Bloomington Stock Center (Bloomington, IN, USA), and the Developmental Studies Hybridoma Bank for fly stocks and reagents; Camtu Nguyen and Chris Cordona for technical assistance; and Adelaide T. C. Carpenter and David T. Denhardt for insightful comments on the manuscript. Special thanks to Adelaide Carpenter for sharing unpublished results and invaluable discussions.