Rad51 and Dmc1 Form Mixed Complexes Associated with Mouse Meiotic Chromosome Cores and Synaptonemal Complexes

PCR reactions were performed with the Pfu polymerase (Stratagene) according to the manufacturer's instructions. The nucleotide sequences of the DNA primers used are available upon request. The full-length hamster COR1(SCP3) cDNA was PCR-amplified and inserted in the EcoRI-BamHI sites of pGAD424 and pGBT9 as described previously (Tarsounas et al. 1997). The full-length mouse SYN1(SCP1) cDNA was excised from a pET28a construct (Tarsounas et al. 1999) with NcoI, ends were filled in with the Klenow fragment of E. coli DNA polymerase I and further digested with XhoI. This fragment was subcloned into the SmaI-SalI sites of pGAD424 and pGBT9 vectors.

The full-length mouse DMC1 cDNA was PCR-amplified from a pET3 construct (Habu et al. 1996) and subcloned into the EcoRI-BamHI sites of pGBT9 and pGAD424. The full-length mouse RAD51 cDNA was PCR-amplified from a pET3 construct (Habu et al. 1996) and subcloned into the SmaI-BamHI sites of pGBT9 and pGAD424. The fragments of mouse DMC1 and RAD51 proteins used in two-hybrid and in vitro protein–protein interaction analyses are shown in Fig. 1. The cDNA fragment encoding DMC1N was PCR-amplified and inserted into the EcoRI-SalI sites of pGBT9 and pGAD424. The cDNA fragment encoding DMC1C was obtained by removing the 5′ StuI-EcoRI fragment from the construct containing full-length DMC1 cDNA in pGAD424. After digestion with these two enzymes, the ends were blunted with the Klenow fragment of E. coli DNA polymerase I and religated, rendering the pGAD424 vector plus a cDNA fragment that encodes DMC1C. The cDNA fragment encoding RAD51N was PCR-amplified and subcloned into the SmaI-SalI sites of pGBT9 and pGAD424. The cDNA fragment encoding Rad51C was obtained by removing the 5′ EcoRI fragment from the construct of full-length RAD51 cDNA in pGAD424. After EcoRI digestion, the ends were religated, rendering the pGAD424 vector with a cDNA fragment that encodes RAD51C. All constructs were sequenced with MATCHMAKER5′ primer (Clontech) to verify the correct open reading frames.

The method for cotransformation of the GAL4 fusion constructs into the yeast strains HF7c and SFY526 (kindly provided by O. Kovalenko, Yale University, New Haven, CT) was described by Chen et al. 1998. The selective media required as well as the methods for determining β-galactosidase activity using the filter assay have been described previously (Tarsounas et al. 1997). Each one of the interactions reported was tested in at least three independent experiments.

For the in vitro protein–protein interaction experiments and polyclonal antibody generation we expressed the proteins involved as (His)₆ or HA fusions. The (His)₆-tagged proteins were generated in pET29b as COOH-terminal or in pET28a/pET3a as NH₂-terminal fusions. The COOH-terminal fusion of COR1(SCP3) with the (His)₆ tag was generated by PCR amplification of the coding region of hamster cDNA and cloning into the EcoRI-XhoI sites of pET29a. An NH₂-terminal (His)₆ fusion of the SYN1(SCP1)/SCP1 was generated in pET28a as previously described (Tarsounas et al. 1999) from a construct containing the full-length cDNA in pBluescript (provided by J. Sage, University of Nice). The pET3-derived plasmids used for expression of the DMC1 and RAD51 proteins (His)₆-tagged at the NH₂ terminus were described previously (Habu et al. 1996).

The HA fusions were generated in E. coli from a modified pET29a vector generated in our laboratory that we named pET29a^HA. To introduce the HA tag into the pET29a vector we synthesized two complementary 45-bp oligonucleotides with the following DNA sequences: 5′-GGC CAT ATG TAC CCA TAC GAT GTT CCA GAT TAC GCT GGT ACC CGC-3′ and 5′-GCG GGT ACC AGC GTA ATC TGG AAC ATC GTA TGG GTA CAT ATG GCC-3′. After annealing and digestion with KpnI, the resulting product was subcloned into the pET29a (Novagen) expression vector that had been digested with NdeI, ends filled in with the Klenow fragment of E. coli DNA polymerase I and further digested with KpnI. With this procedure, we replaced the NH₂-terminal S-Tag™ (Novagen) normally present in the pET29 vector, with the HA epitope. The new tag can be removed by thrombin digestion (Fig. 2): 20 μg of soluble extract from bacteria induced to express HA-COR1(SCP3) was incubated for 15 min at room temperature with 0.25 U thrombin and thrombin buffer (20 mM Tris-HCl, pH 8.3, 150 mM NaCl, and 2.5 mM CaCl₂), or with thrombin buffer alone. The HA-tagged COR1(SCP3) was detected with an anti-HA antibody, while no major band is visible in the Coomassie staining of the same gel. Thrombin treatment abolishes the anti-HA staining (Fig. 2), indicating complete removal of the HA tag from the protein. All the constructs made in pET29a^HA rendered NH₂-terminal in-frame fusions of the desired proteins with the HA tag and lacked the (His)₆ tag, as determined by DNA sequencing.

The full-length hamster COR1(SCP3) cDNA was excised as an EcoRI-SalI fragment from the pGBT9 construct (Tarsounas et al. 1997) and subcloned into the EcoRI-XhoI sites of the pET29a^HA. The full coding region of mouse SYN1(SCP1) cDNA was PCR-amplified with Pfu polymerase (Stratagene) from a construct in pBluescript provided by J. Sage (University of Nice) and subcloned into the BamHI-XhoI sites of pET29a^HA. The full-length mouse DMC1 and RAD51 cDNAs were excised as EcoRI-SalI fragments from the corresponding pGAD424 constructs (described below) and subcloned into the EcoRI-XhoI sites of pET29a^HA.

To express the RAD51 and DMC1 NH₂-terminal fragments (DMC1N and RAD51N; Fig. 1) as HA fusions, the corresponding sequences were PCR-amplified from constructs of the full-length cDNAs in pGAD424 and subcloned into the EcoRI-XhoI sites of pET29a^HA. The cDNA fragments encoding RAD51C and DMC1C were excised with EcoRI and SalI from the corresponding constructs in pGAD424 and subcloned into the EcoRI-XhoI sites of pET29a^HA.

To test protein–protein interactions in vitro, proteins were expressed as (His)₆ or HA fusions. The (His)₆-tagged protein was expressed in E. coli strain BL21(DE3) upon induction with 1 mM IPTG. After lysis and sonication in sonication buffer (50 mM NaH₂PO₄, pH 8.0, 300 mM NaCl, 10 mM imidazole), the soluble fraction was incubated with Ni-NTA agarose (Qiagen) and the bound protein was washed three times in sonication buffer with 50–100 mM imidazole. The remaining unbound sites on the beads were blocked by incubation in storage buffer (Fuchs et al., 1997; 40 mM Hepes, pH 7.6, 150 mM NaCl, 10 mM EDTA, 1 mM DTT, 0.5 mM PMSF, and 0.2% Nonidet P-40) for 2 h at room temperature. The protein bound to the Ni-NTA agarose was stored in 100% glycerol at −20°C. The HA-tagged proteins were expressed in E. coli strain BL21(DE3) after induction with 0.5 mM IPTG. After lysis under nondenaturing conditions and sonication the soluble fraction was isolated by centrifugation at 50,000 g for 1 h and stored at −20°C.

The protein-binding reaction contained 1× reaction buffer (Kovalenko et al., 1998; 20 mM Tris-HCl, pH 7.2, 25 mM NaCl, and 2 mM MgCl₂), 0.05% Tween 20, 50 μg BSA, 20 μl slurry with (His)₆-tagged protein preequilibrated in reaction buffer (∼10 μg of protein) and 50 μg soluble fraction of the bacterial extract containing the HA-tagged protein. The proteins were incubated for 1 h at room temperature on a Nutator. The beads were sedimented, washed three times with 500 μl reaction buffer with 50 mM imidazole, and resuspended in 2× SDS sample buffer (Laemmli 1970) supplemented with 4 M urea. Samples were boiled for 5 min and separated on 10% or 12% SDS–polyacrylamide gels. The (His)₆-tagged protein was detected in the control reaction using an antibody specific for this protein. The presence of the HA-tagged protein in the eluted samples was detected with a rat monoclonal anti-HA antibody (Boehringer-Mannheim).

The mouse DMC1 and RAD51 full-length proteins were overexpressed in E. coli using the pET3 expression system (Habu et al. 1996). The (His)₆-tagged proteins were purified on a Ni-NTA agarose column (Qiagen) and injected into mice and rabbits, respectively. The preimmune sera gave no staining when tested on meiotic prophase chromosomes. The polyclonal anti-DMC1 serum was depleted of anti-RAD51 cross-reactive components by adsorption on Sepharose-bound (Pharmacia) mouse RAD51 protein. The depleted antiserum was used at a 1:200 dilution in blocking buffer (5% skim milk powder, 0.05% vol/vol Triton X-100 in PBS) for Western blotting and 1:2 dilution in antibody dilution buffer (10% vol/vol goat serum, 3% BSA wt/vol, 0.05% vol/vol Triton X-100 in PBS) for immunocytochemistry. Similarly, the rabbit anti-RAD51 polyclonal serum was immunodepleted of components recognizing DMC1 epitopes by adsorption on a DMC1-bound Ni-NTA agarose column (Qiagen). This procedure abolished the cross-reactivity with the DMC1 protein initially observed on Western blots. As an additional specificity test, the two polyclonal sera were each reacted separately with either the mouse DMC1 or RAD51 proteins. The immune complexes were pelleted by centrifugation, and the supernatant used in immunofluorescent staining of spermatocyte nuclei.

COR1(SCP3) was overexpressed in an E. coli expression system and purified using a (His)₆-Tag. The full-length protein was injected into mice to generate a polyclonal serum. COR1(SCP3) localization on meiotic chromosomes was described previously (Dobson et al. 1994; Tarsounas et al. 1999).

The localization of RAD51 and DMC1 antigens was determined on spread rat spermatocyte nuclei prepared as described in Tarsounas et al. 1999. Human CREST sera were used to identify the centromeres. The fluorochromes were visualized with a Polyvar epifluorescent microscope (200 W mercury-vapor light source) and recorded on 400ASA black/white or color slides. For EM, nuclei were attached to a 50-nm-thick plastic carrier film and treated for 5 min at room temperature with DNaseI at a final concentration of 1 μg/ml in MEM (GIBCO BRL). The surface-spread nuclei were incubated with primary antibody for 2 h at 37°C and with secondary antibody (goat anti–mouse conjugated with 5 nm gold, goat anti–rabbit conjugated with 10 nm gold and goat anti–human conjugated with 15 nm gold, 1/50 dilution) for 1 h at 37°C. The immunogold-labeled preparations were postfixed in 4% OsO₄ before observing under the EM. The electron microscope images were recorded at 8,500× magnification on 35-mm film.

Testis preparation was performed as described by Tarsounas et al. 1999. Separation of the spermatocyte fractions by elutriation was performed following the procedure described by Grabske et al. 1975 and Meistrich 1977, with modifications by Heyting et al. 1985. 100 ml were collected from each fraction at increasing flow rates from 15 to 30 ml/min (5-ml/min increase at every step). The composition of these fractions in cells at various meiotic stages was assayed by immunofluorescent staining for the COR1(SCP3), SYN1(SCP1), and centromeric marker proteins. The composition varies from early to late prophase I stages (i.e., fraction I contains 45% leptotene cells, 45% zygotene cells, and 10% pachytene cells; fraction IV contains 85% pachytene cells and 15% leptotene/zygotene cells; fraction VI contains 80% diplotene cells and 20% pachytene cells). The number of spematogonia observed was too small to be taken into consideration. The cells in each fraction were centrifuged and resuspended in sample buffer (4 M urea, 10% SDS, 0.25 M Tris-HCl, pH 6.8, 20% glycerol, 0.015% bromophenol blue, 1 mM EDTA, and 10% vol/vol β-mercaptoethanol) to a concentration of 10⁶ cells/ml. For Western blot analysis, protein corresponding to equal numbers of cells were loaded in each lane.

To determine the chromosomal distribution of RAD51 and DMC1, we generated antibodies against the full-length mouse proteins. As the overall degree of identity at the amino acid level between the two proteins is 54%, each antibody displayed immunoreactivity with the homologous protein on a Western blot (e.g., the anti-DMC1 antibody recognized the RAD51 protein; see Fig. 3 A). To eliminate the cross-reactivity of the anti-DMC1 antibody with the RAD51 protein, we produced a Sepharose matrix with covalently bound RAD51. The polyclonal anti-DMC1 serum was adsorbed on this matrix twice; the immunodepleted antibody reacted with DMC1 but showed no reactivity with the RAD51 protein (Fig. 3 B). Similarly, the anti-RAD51 antibody was immunoadsorbed twice on a matrix containing the mouse DMC1 protein covalently attached to it. This treatment rendered a specific anti-RAD51 antibody (Fig. 3 D). Equal amounts of proteins purified on Ni-agarose column were loaded in each lane, as shown in the Coomassie staining of the gel (Fig. 3 C). The purified antibodies were further used in immunostaining preparations for light and EM, and Western blotting of the spermatocyte fractions.

As an additional method for determining the specificity of the antibodies, we reacted each of them with the RAD51 and DMC1 proteins separately, discarded the immune complexes by centrifugation, and used the depleted sera in immunofluorescence staining of rat meiotic chromosomes (Fig. 4). Normal numbers and position of the RAD51 foci on the chromosome cores were observed when the anti-RAD51 antibody was reacted with the mouse DMC1 protein (Fig. 4A and A′), and when the anti-DMC1 antibody was reacted with the mouse RAD51 protein (Fig. 4C and Fig. C′). Fig. 4A and Fig. C, encompass several nuclei of early prophase I spermatocytes, where the RAD51 and DMC1 proteins are abundant. In Fig. 4A′ and C′, a section of the A and C figures is enlarged. When the anti-RAD51 antibody was reacted with the RAD51 protein, the ability of this antibody to stain discrete foci on the chromosomes was completely blocked (Fig. 4 B). Similarly, reacting the anti-DMC1 antibody with DMC1 protein abolished the staining by this antibody (Fig. 4 D). These experiments established that each purified antibody recognized only the protein against which it was generated.

Surface-spread mouse meiotic chromosomes were reacted with the purified rabbit anti-RAD51 and mouse anti-DMC1 antibodies, each visualized with EM of immunogold particles of different sizes (10 and 5 nm, respectively) attached to the secondary antibodies. The 15-nm gold particles identify the centromere stained with a human CREST serum. The majority of foci observed at various prophase I stages contained a mixture of 5- and 10-nm gold particles (Fig. 5 B′). We counted 150 of these EM foci present in leptotene, zygotene, and pachytene cells, and 11% of them contained only one size of grains. This is consistent with the null class estimate predicted from the Poisson distribution of the results (data not shown). It seems reasonable to conclude that the foci contain both sizes of grains, and the presence of foci with a single grain size is due to the random binding of the antibody to the available sites. This distribution pattern of the gold grains indicates that the two proteins colocalize on the chromosomal cores before they synapse and in the confines of the SC, suggesting that they might act in a concerted manner during meiosis. Interestingly, at leptotene, when the cores are in the process of alignment, the gold grains are physically associated with the cores (Fig. 5 A). Later on, at zygotene, the grain distribution coincides extensively with interstitial connections between the homologues (Fig. 5 B), cytologically similar to the axial associations reported in yeast to be the synapsis initiation sites (Rockmill et al. 1995). In lily and mice, these sites most likely identify the early recombination nodules (Anderson et al. 1997; Moens et al. 1997). At pachytene (Fig. 5 C), the grains label electron-dense structures localized in the central region of the SC.

Because the foci observed with EM contained both RAD51 and DMC1 recombination proteins, we refer to them as recombination complexes. We attempted to show that at the level of fluorescence microscopy the numbers and relative positions of DMC1 and RAD51 foci are also identical. The distribution of RAD51 foci during successive stages of the meiotic prophase I in mammalian oocytes and spermatocytes has been reported previously (Moens et al. 1997; Barlow et al. 1997a). Using double staining of the DMC1 and RAD51 proteins (Fig. 6) and superimposing the two colors as described in Gasior et al. 1998, it can be seen that RAD51 and DMC1 foci coincide. A correspondence of 95% was recorded for 10 prophase nuclei. This is in agreement with the colocalization results of EM where 89% of the foci are mixed. As the distribution of COR1(SCP3) during meiosis is well characterized (Dobson et al. 1994; Lammers et al. 1994), we used a mouse anti-COR1(SCP3) antibody (red) to identify the prophase I stages and sub-stages. This antibody was added in trace amounts to merely mark the cores, without interfering with the staining of the anti-DMC1 antibody also generated in a mouse. In Fig. 6 A the leptotene nucleus was identified based on the fact that the cores are just beginning to form, while in the adjacent zygotene nucleus, the cores are extensive (Fig. 6 C). The coincident staining with anti-RAD51 and anti-DMC1 antibody shows that the foci are associated with the cores (Fig. 6C and Fig. F) and that the number of foci per nucleus is high at leptotene/zygotene (250–300 foci per nucleus) and low at pachytene (∼100 foci or less per nucleus).

To monitor the presence of the two recombinases at progressive stages of meiotic prophase, we have isolated fractions of testicular cells by centrifugal elutriation (Grabske et al. 1975). In general, this method allows cell fractionation by size. As the size of spermatocytes increases with their progression through meiosis, it is possible to separate prophase I cells in fractions enriched for cells at specific meiotic stages (Tarsounas et al. 1999). A sample of each fraction was fixed in 2% paraformaldehyde and used to determine the composition in cells at various meiotic stages, using immunofluorescence staining with anti-SYN1(SCP1), anti-COR1(SCP3), and anti-centromeric antibodies. The prophase I spermatocytes increased from 90% leptotene/zygotene cells in fraction I to 80% diplotene cells in fraction VI (data not shown). To demonstrate that the protein extracts prepared from these fractions give an accurate representation of the stages identified with immunofluorescence staining, we demonstrate the distribution of marker proteins COR1(SCP3) and SYN1(SCP1) in these fractions (Fig. 7). In accordance with its distribution on the meiotic chromosomes from early prophase to metaphase I, COR1(SCP3) was detected in all fractions as a constantly present 30-kD polypeptide. A higher molecular mass form of the protein was also detected, which changed from 35 kD in early prophase I to 33 kD in the late stages. Previously, COR1(SCP3) was being referred to as the 30/33-kD polypeptide, because of its constant detection on Western blots of testicular cells as a doublet of these sizes. The results presented here indicate that the COR1(SCP3) protein (estimated M_r 28,000) most likely corresponds to the 30-kD polypeptide, while the slower migrating band (35 or 33 kD) is possibly the result of posttranslational modifications such as phosphorylation (Lammers et al. 1995), which vary with the meiotic stage. SYN1(SCP1) was used as a second marker for the protein composition of the testicular fractions collected. This protein is known to be present on the chromosomes during zygotene when synapsis initiates, and at pachytene in synapsed chromosomes. The protein disappears at the end of pachytene/onset of diplotene when the homologues separate (Meuwissen et al. 1992; Dobson et al. 1994). Consistent with its presence in the meiotic chromosomes, the SYN1(SCP1) protein was only detected in fractions II–IV, and not in the fractions where the initial and final prophase I stages were predominant. The presence of RAD51 and DMC1 (Fig. 7) was only detected in the first three fractions containing cells in early prophase I stages, in accordance with the immunocytological observation of abundant RAD51/DMC1 foci on the chromosomal cores of early stages of prophase I.

We used the yeast two-hybrid system to test possible protein–protein interactions involving mouse RAD51 and DMC1 (Table A). Homotypic interactions were detected for both mouse proteins, as shown previously for the yeast and human homologues (Donovan et al. 1994; Kovalenko et al. 1996, Kovalenko et al. 1997; Dresser et al. 1997). We tested heterotypic interactions with each protein expressed either as a fusion to the GAL4 activation domain or to the GAL4 DNA-binding domain. Compared with their ability to self-interact, the heterotypic interactions between RAD51 and DMC1 were weaker in the sense that the reporter gene expression (HIS3 and lacZ) was delayed. In the negative controls no expression was detected under identical conditions.

The interaction between RAD51 and DMC1 was also tested using an in vitro assay in which the RAD51 protein was expressed in E. coli as a (His)₆-fusion and bound to Ni-NTA agarose column. The HA-tagged DMC1 from a soluble E. coli extract was incubated with the RAD51 matrix, precipitated, and detected with an anti-HA antibody (Fig. 8 A). These data indicate that the formation of the RAD51/DMC1 complexes can be mediated by direct interactions between these two proteins. As a positive control for this experiment, precipitation of HA-DMC1 with a (His)₆-DMC1 fusion confirms the homotypic interaction detected in the in vivo two-hybrid assay.

Next, we wished to determine which region of the mouse RAD51 and DMC1 proteins was essential for homotypic and heterotypic interactions. We investigated whether the NH₂-terminal domains (Fig. 1), which show the lowest identity between the two proteins, could have distinct abilities in establishing protein–protein interactions. In a two-hybrid assay, we show that the DMC1 COOH-terminal fragment (DMC1C; M_r 31,000) containing amino acids 63–340 interacts with the full-length protein (340 amino acids; M_r 37,800), while the NH₂-terminal fragment (DMC1N; M_r 7,200) containing amino acids 1–62 does not. This indicates that the COOH-terminal region, or a portion of it, is required for the DMC1 homotypic interaction to occur. Similarly, the COOH-terminal RAD51 fragment (RAD51C; M_r 23,000), or a portion of it, mediates RAD51 self-interacting capacity. In addition, the COOH-terminal region of RAD51 is able to interact weakly with the full-length DMC1 in a two-hybrid assay and conversely, the COOH terminus of DMC1 establishes weak interactions with the full-length RAD51 (Table A). These data suggest that the COOH-terminal regions of the two proteins may be involved in mediating heterotypic RAD51/DMC1 interactions as well.

The homotypic interactions observed in the two-hybrid assays were confirmed by the in vitro binding experiments. Fig. 8 B shows that only DMC1C, and not DMC1N, interacts with the full-length DMC1 protein. The truncated derivatives of RAD51 act similarly in the presence of the full-length RAD51 protein. These data suggest that the NH₂- and COOH-terminal regions of the two proteins behave similarly in mediating self-interactions.

To gain further insight into the relationship between the SC and the RAD51/DMC1 recombinases, we tested possible interactions between the SC proteins and the components of recombination complexes. We performed in vivo two-hybrid assays with DMC1, RAD51, and each of the SC components, SYN1(SCP1) and COR1(SCP3), the results of which are presented in Table B. COR1(SCP3) is a component of the chromosomal cores from early prophase I when the cores start to assemble, to pachytene when the cores synapse and form the lateral domains of the SC, and until late diplotene when the homologues separate from each other (Dobson et al. 1994; Lammers et al. 1994; Tarsounas et al. 1999). Both RAD51 and DMC1 interact with COR1(SCP3) in the in vivo two-hybrid system and in the in vitro coprecipitation assays. This is consistent with the presence of the recombination complexes in association with the cores from leptotene to pachytene, as detected with electron and light microscopy (Fig. 5 and Fig. 6). SYN1 (SCP1) is pachytene-specific, possibly identifying the protein that establishes and maintains synapsis between homologues (Meuwissen et al. 1992; Dobson et al. 1994; Tarsounas et al. 1999). RAD51 interacts with this synaptic protein, while DMC1 does not (Fig. 8 B).

The third component of the SC tested for interaction with the RAD51/DMC1 recombinases was SCP2. This protein localizes in the cores/lateral domains of the SC and has a temporal and spatial distribution very similar to that of COR1(SCP3) (Offenberg et al. 1998). In a two-hybrid screen of a hamster testis cDNA library for proteins interacting with COR1(SCP3) (Tarsounas et al. 1997), we have isolated a 50-kD polypeptide from the COOH-terminal region of the SCP2 protein (M_r 190,000). We tested this fragment in a two-hybrid assay for its ability to participate in direct interactions with the DMC1/RAD51 proteins. No interaction was detected, suggesting that the COOH-terminal SCP2 region which mediates the interaction between the two core components (COR1(SCP3) and SCP2) is not required for the binding of the recombination complexes to the cores. This finding does not exclude the possibility that another region of the large SCP2 protein may be functional in this respect.

A major difficulty in studying the distribution of RAD51 and DMC1 in mammalian meiotic chromosomes is that sera raised against the full-length proteins recognize both proteins in immunoblots and presumably in cytological preparations. A polyclonal serum raised against a 15–amino acid NH₂-terminal peptide of RAD51, while showing a high specificity for RAD51 on Western blots, does not render any signal on meiotic chromosome spreads (data not shown), possibly because the NH₂ terminus of this protein is not available to the antibody. To differentiate between these two proteins, we have used sera that were depleted of the cross-reactive antibodies. The high specificity of the resulting antibodies was apparent in Western blot analyses of bacterially expressed proteins (Fig. 3) and in spermatocyte immunostaining in which each antibody was prereacted with the antigen or with the homologous protein (Fig. 4). Data presented here, therefore, accurately reflect the distribution of the two proteins on the chromosomes.

In yeast, Rad51 and Dmc1 proteins form discrete foci along the meiotic chromosome cores with variable degrees of colocalization (Bishop 1994; Dresser et al. 1997). We report here the results of a similar study performed on mammalian meiotic chromosomes. The larger size of these chromosomes, as well as the spreading techniques available for their fixation allow a more detailed analysis of the proteins associated with the chromosomal cores. The RAD51 and DMC1 foci detected by immunofluorescence staining of rat spermatocyte nuclei with antibodies against each of the two proteins overlap in 95% of the cases. This observation does not necessarily imply that the two proteins colocalize on the chromosomes (Freire et al. 1998). Thus, we attempted to define the relative position of the RAD51 and DMC1 foci using EM with immunogold visualization. RAD51 was visualized with 10-nm gold grains and DMC1 with 5-nm gold grains. The two types of particles occurred in mixed groups in 89% of the cases, while only 11% of the groups were composed of single-size particles. According to the Poisson analysis of our numeric data, the latter focus type is expected to occur by chance alone if the number of labeled proteins is small relative to the total number of protein molecules per focus. Similarly, in immunofluorescence staining of meiotic nuclei a small percentage of foci (∼5%) were composed of only one or the other type of protein. We conclude that RAD51 and DMC1 colocalize in mixed foci on the mouse meiotic chromosomes.

The number of RAD51/DMC1 complexes is most abundant at early meiotic stages and decreases thereafter, becoming undetectable at late pachytene/early diplotene. Consistent with this is the detection of the two proteins in spermatocyte fractions containing predominantly early meiotic stages (Fig. 7). The downregulation of DMC1/RAD51 during meiotic prophase I is correlated with the progression of recombination. The DMC1/RAD51 complexes are most abundant at the leptotene/zygotene stages, during which homologous recombination is initiated, and less numerous at pachytene, when maturation of the recombination intermediates is postulated to occur. This is consistent with a role for these proteins in the early steps of meiotic recombination, presumably in homology recognition and initiation of strand exchange reactions as previously proposed (Baumann et al. 1996).

We addressed the question whether the RAD51/DMC1 mixed foci observed with immunocytological methods are formed by direct interactions between the mouse RAD51 and DMC1 proteins. The interaction detected between RAD51 and DMC1 in an in vivo two-hybrid assay was weak, as estimated from the delayed expression of the HIS3 and lacZ reporter genes. One possibility is that the interaction between RAD51 and DMC1 is indirect and requires an adaptor protein. A strong candidate for such an adaptor protein is the yeast Tid1 (Dresser et al. 1997), a Rad54 homologue which interacts with both Rad51 and Dmc1 in a two-hybrid screen. It is possible that the yeast Tid1 protein can mediate the interaction between the two mouse homologues when they are expressed in the same yeast cell in the two-hybrid assay. Its recruitment, however, may delay the establishment of an interaction between the two proteins and, consequently, the expression of the reporter genes. The second possibility is that each of the two proteins (RAD51 and DMC1) form homotypic interactions very easily, especially in the presence of the DNA substrates as may be the case in vivo, and, therefore, little protein is left available for the heterotypic interaction to occur (Dresser et al. 1997). Using biochemical assays, we showed that RAD51 and DMC1 can be coprecipitated in vitro (Fig. 8). This suggests that the mixed RAD51/DMC1 foci detected immunocytologically may assemble through direct interactions between the two proteins.

Determining the regions of RAD51 and DMC1 involved in homotypic interactions will help understand the assembly of the recombination complexes on the chromosomes. We used truncated derivatives of the two proteins in two-hybrid and in vitro binding experiments, and showed that a version of RAD51 or DMC1 bearing a deletion at the NH₂ terminus still allows establishment of heterotypic and homotypic interactions.

Biochemical studies have shown that RAD51 assembles nucleoprotein filaments with single- and double-stranded DNA (Benson et al. 1994; Baumann and West 1997) and DMC1 may have similar properties (Li et al. 1997). Our EM data indicating colocalization of these proteins on the meiotic chromosomes and their ability to establish direct heterotypic protein–protein interactions suggest the possibility that RAD51/DMC1 form mixed filaments associated with DNA at meiosis. Moreover, we scored 150 EM foci containing >1,500 gold grains and found that 46% of the grains correspond to RAD51 and 54% to DMC1 (data not shown). Therefore, the possibility that these filaments have a constant stoichiometry of RAD51 to DMC1 molecules is not excluded.

We show here that interactions between RAD51/DMC1 complexes and SCs are established by direct protein–protein interactions between their components. These are detected in an in vivo two-hybrid system and in vitro binding analyses using bacterially expressed proteins. Coimmunoprecipitation of these proteins from testis extracts is technically unattainable due to the high insolubility of the SC components in these extracts (Tarsounas, M., P. Moens, and R.E. Pearlman, unpublished data). The interactions detected between SC components and the two recombinases support the cytological observations from immunofluorescence and EM, where the RAD51/DMC1 complexes are detected in association with the chromosomal cores at the very early prophase I stages or with the SCs at pachytene (Barlow et al. 1997a; Moens et al. 1997; this study). Based on these data, it is possible that the SC components may provide the structural frame that stabilizes complexes formed of recombination proteins and DNA (Yuan et al. 1998).

Several lines of evidence support this hypothesis. In mammals, SC formation is obstructed in the absence of Atm (Barlow et al. 1996; Keegan et al. 1996; Xu et al. 1996). This structural defect may cause the mislocalization of RAD51/DMC1 complexes detected in the Atm−/− spermatocytes (Barlow et al. 1997b, Barlow et al. 1998). In yeast, the core-associated Red1 protein with a meiotic distribution similar to that of COR1/SCP3 (Smith and Roeder 1997) is necessary for recombination (Rockmill and Roeder 1990; Storlazzi et al. 1996) and specifically for normal levels of interhomologue joint molecule formation (Schwacha and Kleckner 1997).

Using genetic analyses and EM of spread yeast chromosomes, Rockmill et al. 1995 have shown that Rad51 and Dmc1 proteins are required to establish the axial associations between homologues, which represent the synapsis initiation sites. Here we show that in rat chromosome spreads the RAD51/DMC1 complexes coincide with cytologically similar axial associations. The gold grains may detect a nucleoprotein filament formed at the site of a DSB after resection of the ends as predicted (Szostak et al. 1983; Bishop 1994). This nucleoprotein filament bridges the two homologues and possibly plays a role in homology recognition as well. By these criteria, a RAD51/DMC1-coated ssDNA visualized in association with single cores at leptotene/early zygotene (Fig. 5 A) may recognize homologous DNA sequences on another chromosome and establish the axial associations visualized in Fig. 5 B, as previously suggested (reviewed by Baumann and West 1998). This hypothesis is also supported by the recent observation that RAD51 foci are associated with ssDNA in meiotic cells (Raderschall et al. 1999).

At pachytene, the gold grains appear in the central region of the SC (Fig. 5 C). The recombination complexes may be anchored there by a direct interaction of RAD51 with the synaptic protein SYN1(SCP1) which also localizes in the central region of the SC (Meuwissen et al. 1992; Dobson et al. 1994). We detected such a direct interaction between RAD51 and SYN1(SCP1) in the two-hybrid and in vitro binding analyses. The significance of this is unclear at present. It is possible, however, that the few recombination complexes present at pachytene are anchored in the central region of the SC through a direct interaction with SYN1(SCP1).

In addition to the RAD51/DMC1 recombinase, other proteins involved in DNA repair (e.g., MLH1, BRCA1, BRCA2), as well as sensors of DNA damage (e.g., HRAD1, ATR) have been detected in association with the meiotic chromosome cores and/or the SCs (Baker et al. 1996; Keegan et al. 1996; Scully et al. 1997; Barlow et al. 1998; Chen et al. 1998; Freire et al. 1998; Moens et al., 1998). This suggests a concerted action of all these proteins in the repair of meiotic DSBs and maintenance of genomic integrity in the germ line. Determining the molecular details of this process requires further experimentation.

We are grateful to Julien Sage (University of Nice) who provided the full-length SYN1/SCP1 cDNA, Oleg Kovalenko (Yale University) for the SFY526 yeast strain, Barbara Spyropoulos for generating the antibodies used here, Bryan McNeil (U. of Toronto) for plasmids, and Roberto Botelho, Gagan Gupta, and Anita Samardzik (all from York University) for technical help.

This work was supported by Natural Sciences and Engineering Council of Canada grants to P.B. Moens and R.E. Pearlman. M. Tarsounas was supported in part by an Ontario Graduate Scholarship.