A Meiotic Chromosomal Core Consisting of Cohesin Complex Proteins Recruits DNA Recombination Proteins and Promotes Synapsis in the Absence of an Axial Element in Mammalian Meiotic Cells

Abstract

The behavior of meiotic chromosomes differs in several respects from that of their mitotic counterparts, resulting in the generation of genetically distinct haploid cells. This has been attributed in part to a meiosis-specific chromatin-associated protein structure, the synaptonemal complex. This complex consist of two parallel axial elements, each one associated with a pair of sister chromatids, and a transverse filament located between the synapsed homologous chromosomes. Recently, a different protein structure, the cohesin complex, was shown to be associated with meiotic chromosomes and to be required for chromosome segregation. To explore the functions of the two different protein structures, the synaptonemal complex and the cohesin complex, in mammalian male meiotic cells, we have analyzed how absence of the axial element affects early meiotic chromosome behavior. We find that the synaptonemal complex protein 3 (SCP3) is a main determinant of axial-element assembly and is required for attachment of this structure to meiotic chromosomes, whereas SCP2 helps shape the in vivo structure of the axial element. We also show that formation of a cohesin-containing chromosomal core in meiotic nuclei does not require SCP3 or SCP2. Our results also suggest that the cohesin core recruits recombination proteins and promotes synapsis between homologous chromosomes in the absence of an axial element. A model for early meiotic chromosome pairing and synapsis is proposed.

The eukaryotic cell cycle ensures that chromosomes are correctly replicated and symmetrically divided between daughter cells. Errors in the chromosomal segregation process can generate aneuploid cells, which are either not viable or contribute to cancer development, infertility, or other aspects of human disease. Two different strategies for cell division are active in eukaryotic organisms, mitosis and meiosis. Meiosis differs in several respects from mitosis; for example, meiotic cells undergo two cell divisions (M1 and M2) without an intervening DNA replication step, resulting in the generation of haploid cells. Furthermore, homologous chromosomes (each consisting of two sister chromatids) recombine and synapse in prophase I. The homologs are then separated at anaphase I, while the sister chromatids remain associated until the second meiotic division (33, 54).

How can the differences between mitotic and meiotic chromosomal behavior be explained? Our understanding of the mechanisms that regulate chromosome synapsis has increased tremendously over the past few years, and two different protein complexes have been shown to take part in these processes, the cohesin complex and the synaptonemal complex (SC) (25, 45). We now know that sister chromatids in mitotic cells remain associated by protein complexes called cohesins (14, 26), which consist of at least four different subunits (SMC1, SMC3, SCC1, and SCC3). SMC1 and SMC3 have been shown to bind DNA in vitro (2, 3). Cohesin complexes become attached to chromosomes in somatic cells in the G₁ phase and are deposited between sister chromatids during the S phase. The cohesin complexes act as a molecular glue between the two sister chromatids and create a bilateral symmetry which mimics the organization of the equally bilaterally organized mitotic spindles. The cohesin complex is lost from the chromosomes during mitosis in somatic cells, and as a result of the pulling forces applied on the chromosomes by the mitotic spindles, the two new cells each receive a copy of each chromosome. The cohesin complex has been shown to be required for chromosome pairing and segregation in yeast and for DNA recombination in meiotic cells (7, 8, 16, 23, 28, 47, 48).

In contrast to cohesin complexes, the SC is normally only found in meiotic prophase I cells between homologous chromosomes (33, 54). The SC was discovered more than 40 years ago, and its function has been intensely discussed since then (24). Ultrastructural analysis of the SC reveals a tripartite structure with two parallel lateral elements (LEs) and a central element. During the leptotene and zygotene stages of meiotic prophase I, the axial elements (AEs) (the LE is called AE prior to synapsis of the homologous chromosomes) form a proteinaceous core between the two sister chromatids of each chromosome. In a process called synapsis, the two AEs then connect along their entire lengths by fine fibers called the transverse filaments (TF), a process completed at the pachytene stage of meiotic prophase I (38).

While the SC is conserved at the ultrastructural level in most eukaryotic organisms, core components of this structure have as yet been characterized only in yeast and mammals. A meiosis-specific constituent of the TF called SCP1 (Syn1) in mammals and Zip1 in Saccharomyces cerevisiae has been analyzed in detail (11, 12, 21, 43). SCP1 and Zip1 both contain a long central coiled-coil motif surrounded by nonhelical ends. The TF has been postulated to consist of parallel dimers of SCP1 molecules, the C-termini of which are anchored in the LEs. SCP1 dimers that are attached to two opposing LEs are joined together by their N termini, a driving force in the zippering process that brings homologous chromosomes together as they synapse (12, 20, 39). In support of this, it has been shown that yeast cells lacking Zip1 maintain an intact AE but fail to synapse their meiotic chromosomes (42).

Two meiosis-specific core components of the AE and LE, SCP2 and SCP3 (Cor1), have been identified in mammals (11, 19, 22, 27, 37). SCP2 and SCP3 first appear in leptotene-stage spermatocytes and disappear in late meiotic cells. Antibodies against the two proteins stain meiotic prophase I chromosomes in a continuous line from one end to the other. The SCP3 protein can, when expressed in cultured somatic cells, self-assemble into thick fibers that display similarities to the LE structure seen in vivo (53). We previously generated a mouse strain that is deficient in the SCP3 protein. We have shown that SCP3-deficient spermatocytes fail to form AEs and SC, as determined by silver staining. Furthermore, we found that mutant meiotic cells undergo apoptosis, resulting in male sterility (52). No homologs to SCP2 or SCP3 have been identified in nonmammalian organisms, but a protein associated with the AE and LE in S. cerevisiae, called Red1, has been identified (41). Red1 distributes along the meiotic chromosomes in a noncontiguous manner, arguing that it is not a core component of the AE. However, genetic analysis of RED1 function has shown that it is required for AE-LE formation, SC formation, and meiotic DNA recombination (32).

The molecular data available for cohesin complexes and for the AE suggest that the two protein complexes together contribute to the fidelity of meiotic chromosome segregation. Smc3 and a meiosis-specific cohesin subunit, Rec8 (related to Scc1), have been shown to be required for assembly of the AE and the SC and for correct chromosome segregation in yeast meiotic cells (7, 16, 45, 47). In addition, work with mammalian cells has shown that SMC1, SMC3, and a meiosis-specific variant of Scc3 (STAG3) are expressed in meiotic cells and that these proteins colocalize with the AE and the SC (13, 30). In vitro experiments have also indicated that SCP2 and SCP3 interact with cohesin complexes in mammalian meiotic cells (13). However, while the absence of cohesins has drastic effects on AE formation and chromosome pairing, it is not known how the expression and organization of cohesin complexes are affected by the absence of the AE. Furthermore, it is not known if the AE component SCP3 is required for recombination, alignment of intersister axes, and/or synapsis in early meiotic cells.

We have analyzed whether the absence of the AE protein SCP3 affects the expression and spatial distribution of cohesin complex subunits, DNA recombination proteins, and SC proteins in early meiotic cells. We find that SCP3 is required for assembly of a second AE protein, SCP2, on the meiotic chromosomes and that both SCP2 and SCP3 are likely required to form the AE structure observed in vivo. We show both wild-type and SCP3-deficient spermatocytes that in SMC1, SMC3, and STAG3 form fiberlike arrays of foci, likely corresponding to a chromosomal core that holds sister chromatids together. The transverse filament protein, SCP1, colocalizes with short regions of the AE-like cohesin core in SCP3-deficient spermatocytes, suggesting that synapsis takes place between homologous chromosomes even in the absence of an AE. This conclusion is reinforced by the finding that both early (DMC1) and late (Msh4) markers for DNA recombination colocalize with the AE-like cohesin core in SCP3-deficient spermatocytes and that Msh4 is found at regions of these structures where SCP1 is localized. Our data suggest, therefore, that the organization of cohesin complexes in meiotic cells is not substantially affected by the absence of the AE. Moreover, proteins involved in meiotic recombination organize themselves on the cohesin cores remaining in SCP3-deficient spermatocytes in a way similar to that seen on the AEs in wild-type spermatocytes.

MATERIALS AND METHODS

Immunocytology.

Male germ cells were taken from the adult testes of animals with different genotypes, and the spermatocytes were prepared using a “dry-down” technique (29). Briefly, testicular cells were fixed using 1% paraformaldehyde and 0.15% Triton X-100 and prepared for immunofluorescence microscopy using standard methods (52).

Primary antibodies.

The different sera were diluted as follows: CREST, 1:4,000; SCP1, 1:1,000 (21); SCP2, 1:600 (27); Dmc1, 1:500 (44); Smc1, 1:30 (13); Smc3, 1:30 (13); STAG3, 1:200 (30); and Msh4, 1:10 (36). The anti-FLAG antiserum (Sigma) was diluted 1:400, and the anti-Myc antiserum (Clontech) was diluted 1:500.

Plasmid constructs.

All sequences described here (except DMC1) were cloned into the eukaryotic expression vectors pCMX-pL1 and/or pCMX-pL1-FLAG. In addition to the cytomegalovirus promoter and a translational initiation site, the pCMX-pL1-FLAG vector contains an in-frame sequence for the FLAG epitope sequence (52). The full-length SCP2 sequence was amplified from rat testis cDNA (27), and the full-length RAD51 sequence was amplified from mouse testis cDNA, whereas the DMC1 sequence was amplified from a plasmid construct (44). The final constructs were DNA sequenced, and the expression of all constructs (except Dmc1-Myc) was tested by coupled in vitro transcription-translation using T7 polymerase and a rabbit reticulocyte system (Promega Biotec). The expressed proteins were detected by autoradiography or by Western blotting experiments using specific antibodies.

Transfection and indirect immunofluorescence microscopy.

DNA constructs were introduced into cells by electroporation. Briefly, three confluent 10-cm-diameter petri dishes containing COS cells were electroporated at 450 V and 250 μF (Bio-Rad) with 15 μg each of the relevant DNA construct in a total volume of 900 μl of phosphate-buffered saline. The electroporation cuvette was placed on ice for 10 min (for the cells to recover) before the cells were resuspended in 25 ml of Dulbecco's modified Eagles medium (GIBCO) supplemented with 10% fetal bovine serum (Life Technology) and gentamicin (0.06 mg/ml; Life Technology). Approximately 12 ml was transferred to a fresh 10-cm-diameter petri dish and cultured for 24 h at 37°C in a humidified atmosphere containing 5% CO₂ and 95% air. The cells were trypsinized and transferred to glass slides by cytospinning (Shandon). The cells were fixed in ice-cold methanol-acetone (50:50) for 5 min and preincubated with 3% bovine serum albumin before addition of the first antibody. The secondary antibodies were a fluorescein isothiocyanate-conjugated swine anti-rabbit immunoglobulin G (IgG) (diluted 1:50: Dakopatts), a rhodamine-conjugated sheep anti-mouse IgG (diluted 1:100; Boehringer Mannheim), and a rhodamine-conjugated goat anti-human IgG (diluted 1:100; Sigma). The cells were also stained with 1 μg of DAPI (4′,6′-diamidino-2-phenylindole)/ml for 15 s. The slides were mounted in 78% glycerol mounting medium containing 1 mg of para-phenylene diamine/ml, examined in a Leica DMRXA microscope, and digitally imaged using a Hamamatsu C4880-40 charge-coupled device camera, the Openlab software package from Improvision, and Photoshop software from Adobe. The pictures were printed on a Kodak 8650 PS CMYK sublimation printer.

RESULTS

Male germ cells deficient in the SCP3 protein do not contain AEs or SCs, as determined by silver staining and electron microscopy studies (52). Surprisingly, despite the absence of SCP3, SCP1 consistently forms short fiberlike structures, reminiscent of what is seen in wild-type cells (Fig. 1). The ability of SCP1 molecules to form linear arrays suggests the existence of an underlying meiotic chromatin structure which SCP1 can adhere to and use as a scaffold in its polymerization process (52). We have now investigated the molecular nature of the chromatin structure that remains in meiotic cells deficient in SCP3.

FIG. 1.

Expression of SCP1 and SCP2 in wild-type and SCP3-deficient spermatocytes. Spermatocytes prepared from wild-type and knockout animals were fixed and analyzed by immunofluorescence microscopy. Antiserum against SCP1 or SCP2 was used together with antiserum against CREST to label the different SC subunits (green) and the centromeres (red) in wild-type and mutant spermatocytes, respectively. Fibers formed by SCP1 were observed in both wild-type and mutant cells and included axial gaps. The SCP2 antiserum labeled fibrillar AEs only in wild-type spermatocytes. In mutant cells, no structures stained by the SCP2 antiserum were detected. DNA was detected by DAPI staining (blue).

SCP2 requires SCP3 to be incorporated into the AE.

The SCP2 protein has been shown to colocalize temporally and spatially with SCP3 in rodent meiotic germ cells (37). This suggests the possibility that SCP2, in the absence of SCP3, could form an AE-like chromosomal core with which SCP1 can associate. To investigate this possibility, the distribution of SCP2 was analyzed in wild-type and SCP3-deficient meiotic cells. Cells at an early stage of meiosis (i.e., leptotene-zygotene) were selected for indirect immunofluorescence microscopy analysis. Such cells were distinguished based on several criteria, i.e., the nuclear distribution of heterochromatin (as determined by DAPI staining of DNA), the distribution of centromeres within the nuclei (as visualized by staining with CREST antiserum), and the number of centromeric dots (53). A typical leptotene-zygotene-stage cell has a few heterochromatic regions containing a clustered set of centromeres, where the number of centromeric dots varies between 30 and 40, depending on how far synapsis has progressed. While the anticipated fibrillar AEs were detected by the anti-SCP2 antiserum in wild-type spermatocytes (the thin fibers labeled by the anti-SCP2 antiserum represent asynapsed chromosomes, whereas the thicker fibers represent synapsed chromosomes, as determined by costaining with an anti-SCP1 antibody [not shown]); no such structures were seen in the SCP3-deficient meiotic cells (Fig. 1). A weak and sometimes punctuated nuclear SCP2 staining was observed in the mutant germ cells. This result shows that SCP2 cannot form an AE-like structure on its own but that SCP3 is required for the incorporation of SCP2 into the AE. From this finding, we also conclude that SCP2 cannot be responsible for organizing the SCP1 fibers seen in SCP3-deficient meiotic cells.

Subunits of the cohesin complex form an AE-like structure in the absence of the AE proteins SCP3 and SCP2.

Components of the cohesin complex have been shown to be required for assembly of AEs in yeast (16, 47) and to colocalize with the AEs and SCs in murine meiotic cells (13, 30). We have studied whether the cohesin structures seen in wild-type meiotic cells remain intact in SCP3 mutant cells. We first determined the cellular distribution of SMC1, SMC3, and STAG3 in wild-type cells by using specific antisera against these three proteins (13, 30). As previously described, we found that all three proteins, SMC1, SMC3, and STAG3, are constituents of fiberlike nuclear structures in wild-type meiotic cells at the leptotene-zygotene stages (Fig. 2). These fiberlike structures are likely to correspond to a chromosomal core that holds sister chromatids together, since they colocalize with the AE of the SC (13, 30). The distributions of SMC1, SMC3, and STAG3 were then analyzed in SCP3-deficient meiotic cells. Interestingly, all three proteins were found to be part of fiberlike structures even in the absence of SCP3 (Fig. 2). While SMC3 in both wild-type and SCP3-deficient spermatocytes is part of shorter fibers with a beadlike appearance, SMC1 and STAG3 are part of long AE-like structures (which we call cohesin cores). The relative intactness of the structures labeled by antibodies against the three different cohesin subunits in SCP3-deficient spermatocytes clearly suggests that these structures are not dependent on SCP3 or SCP2 for their formation or maintenance.

FIG. 2.

Expression of the cohesin complex subunits, SMC1, SMC3, and STAG3, in wild-type and SCP3-deficient meiotic cells. Spermatocytes prepared from wild-type and knockout animals were fixed and analyzed by immunofluorescence microscopy. Antiserum against SMC1, SMC3, or STAG3 was used together with antiserum against CREST to label the cohesin core (green) and the centromeres (red) in wild-type and mutant spermatocytes, respectively. Interestingly, the distributions of all three proteins were essentially unaltered in the SCP3-deficient cells. SMC3 forms short fibers and beadlike structures, whereas SMC1 and STAG3 staining revealed long AE-like structures even in the absence of SCP3. DNA was detected by DAPI staining (blue).

SCP1 colocalizes with the cohesin cores in the absence of the AE proteins SCP3 and SCP2.

The existence of AE-like structures in SCP3-deficient meiotic cells that contain cohesin complexes could explain the organized arrays of SCP1 molecules seen in these cells, if SCP1 could associate with such structures. To determine the distribution of the cohesin complexes relative to SCP1, we used a double-labeling immunofluorescence microscopy approach by using antisera against SCP1 and STAG3 (Fig. 3). We could distinguish two types of STAG3 labeling patterns in wild-type zygotene-stage spermatocytes, one that overlapped with SCP1 and one that did not. This very likely represents synapsed (SCP1-positive) and unsynapsed (SCP1-negative) regions of the AE, respectively. Interestingly, analysis of SCP3-deficient spermatocytes gives the same results as those seen in wild-type cells, i.e., SCP1 labels short stretches of the STAG3-positive fiberlike structures seen in these cells (Fig. 3). The fibers labeled by both STAG3 and SCP1 antisera are thicker than the fibers labeled by only the STAG3 antiserum, supporting the idea that SCP1 antiserum labels synapsed cohesin cores. The same results were also seen in double-labeling experiments using antisera against either SCP1-SMC1 or SCP1-SMC3 (not shown). Together, this suggests that SCP1 is able to interact with the AE-like chromosomal core that remains in SCP3-deficient spermatocytes and to promote synapsis.

FIG. 3.

Cohesin complex subunits and SCP1 colocalize in SCP3-deficient spermatocytes. Immunofluorescence double labeling was performed on fixed wild-type and mutant zygotene-stage spermatocytes and analyzed by immunofluorescence microscopy. Antibodies against SCP1 (red) were used together with antibodies against the cohesin complex subunit STAG3 (green) in order to analyze their relative distributions in spermatocytes. SCP1 is always found to colocalize with the fibers labeled by the anti-STAG3 antiserum, although only sections of the long fiberlike cohesin core labeled by the anti-STAG3 antiserum are labeled by the anti-SCP1 antiserum. This suggests that whereas STAG3 runs along the entire chromatin core, SCP1 associates only with regions that are being synapsed. The arrows indicate thick fibers labeled by both the anti-STAG3 antiserum and the anti-SCP1 antiserum but which protrude into thinner fibers labeled only by the anti-STAG3 antiserum. DNA is stained with DAPI (blue).

Dmc1 colocalizes with the cohesin cores even in the absence of the AE proteins SCP3 and SCP2.

The DNA recombination protein Dmc1 is required for chromosomal pairing before homologous chromosomes become closely associated by synapsis (31, 40, 50). Dmc1 has also been shown to be part of early recombination nodules that colocalize with the AE of the SC at the leptotene-zygotene stage (4, 44). In vitro interaction data suggest that SCP3 recruits Dmc1 to meiotic chromosomes (44), indicating that it promotes meiotic recombination. We find that the anti-Dmc1 antiserum labels a large number of foci in wild-type zygotene-stage spermatocytes (Fig. 4). These foci are organized in arrays where most of the foci at this stage of meiosis do not overlap with the synapsed chromosomes labeled by SCP1 but overlap with the AE (44). Interestingly, arrays of Dmc1 foci are also seen in SCP3-deficient spermatocytes (Fig. 4). The relative intensity of the foci labeled by the anti-Dmc1 antibody is in general weaker in SCP3-deficient cells, and some mutant spermatocytes have very few Dmc1-positive foci (not shown). To establish whether the DMC foci colocalize with the cohesin cores, which we have shown remain in SCP3-deficient spermatocytes, such cells were labeled with both STAG3 and DMC1 antisera. We found that most Dmc1 foci indeed colocalized with the cohesin cores labeled by STAG3 in wild-type spermatocytes (Fig. 4). Strikingly, arrays of Dmc1 foci also colocalize with the cohesin cores (labeled by anti-STAG3 antiserum) in SCP3-deficient spermatocytes. These observations show that neither SCP3 nor the AE is required for recruitment of Dmc1 to meiotic chromosomes. Instead, the cohesin cores seem sufficient for recruiting recombination proteins such as Dmc1 to meiotic chromosomes.

FIG. 4.

Expression of the recombination proteins Dmc1 and Msh4 in SCP3-deficient cells. Spermatocytes prepared from wild-type and knockout animals were fixed and analyzed by immunofluorescence microscopy. Anti-Dmc1 antiserum labels arrays of foci in both wild-type and knockout animals, although the labeling in cells lacking SCP3 is weaker. Double immunofluorescence labeling with antisera against Dmc1 (red) and SCP1 (green) shows little colocalization in wild-type or mutant cells (consistent with the observation that Dmc1 and SCP1 associate with asynapsed and synapsed meiotic chromosomes, respectively). Association of Dmc1 to asynapsed meiotic chromosomes is verified by the colocalization of Dmc1 (red) and STAG3 (green), also seen in SCP3-deficient spermatocytes. Using antisera against Msh4 and SCP1, it is shown that Msh4 foci (red) preferentially colocalize with SCP1 (green) in both wild-type and mutant cells. The arrows indicate colocalization between Dmc1 and STAG3, as well as Msh4 and SCP1. DNA was detected by DAPI staining (blue).

Msh4 accumulates with SCP1 at regions where cohesin cores are synapsed.

The colocalization of the recombination protein Dmc1 with cohesin complexes in meiotic cells deficient in SCP3 suggests that meiotic recombination does not require an AE. To further support this idea, we monitored the distribution of the MutS homolog, Msh4, in wild-type and SCP3-deficient cells. Msh4 is a meiosis-specific protein required for recombination, chromosome synapsis, and chromosome segregation but for which no DNA repair activity has been detected (17, 34). Msh4 has been shown to take part in a recombination pathway in meiotic cells downstream of Dmc1 and to localize to the regions of the homologous chromosomes that are undergoing synapsis (36). Analysis of wild-type meiotic cells shows that Msh4 foci colocalize with SCP1 in wild-type spermatocytes (Fig. 4). Interestingly, analysis of SCP3-deficient spermatocytes gives the same result as that seen in wild-type cells, i.e., Msh4 localizes to regions of meiotic chromosomes that also are labeled by SCP1. These results suggest that Msh4, even in the absence of AE, can bind to synapsed regions of meiotic chromosomes. It also strongly suggests that the SCP1 structures observed in mutant spermatocytes represent synaptic events between homologous chromosomes.

SCP3, but not SCP2 or SCP1, forms large fibrillar structures when overexpressed.

SCP3 has previously been shown to form homotypic fibers that have similarities with AEs when it is overexpressed in cultured somatic cells (53). In contrast, SCP1 will form only small nuclear and cytoplasmic foci under the same conditions (51). To test whether SCP2 has the ability to give rise to fibers when overexpressed, COS cells were transfected with a cytomegalovirus-driven eukaryotic expression vector containing full-length SCP1, SCP2, or SCP3. Transfection of SCP3-FLAG into COS cells gave rise to very large FLAG-positive fibrillar protein structures (Fig. 5). In contrast, transfection of either SCP2-FLAG or SCP1-FLAG into COS cells did not give rise to similar protein structures. Instead, in the last two cases, a large number of small foci were found in the nucleus (SCP2) or in both the nucleus and the cytoplasm (SCP1). These foci most likely represent small aggregates of overexpressed proteins.

FIG. 5.

Expression patterns of overexpressed SC proteins in cell cultures. pCMX vectors, encoding full-length SCP1, SCP2, or SCP3, were transfected into COS cells and analyzed by immunofluorescence microscopy. Expression of SCP1 and SCP2 results in small foci, localized to both the nucleus and the cytoplasm or restricted to the nucleus, respectively. In contrast, expression of SCP3 results in large fibers found predominantly in the nucleus.

Dmc1 colocalizes with SCP3 but not with SCP1 or SCP2.

To test whether the cell transfection system that we were using allowed us to detect protein-protein interactions between heterologous proteins (i.e., by monitoring protein colocalization), we transfected COS cells with constructs expressing full-length Rad51 and Dmc1 proteins (Fig. 6), which have previously been shown to interact with each other (44). We found that cotransfection of Rad51 and Dmc1 leads to the expected formation of short fibrillar nuclear structures containing both proteins. Thus, based on this result, it should be possible to use this system to look at interactions between two different proteins that are cotransfected. We then analyzed whether Dmc1 could also interact with core components of the SC in this cell transfection assay (Fig. 6). We found that Dmc1 and SCP3 showed extensive colocalization after cotransfection of expression plasmids encoding these two proteins, supporting the idea that Dmc1 and SCP3 interact in vivo during meiosis (44). Little colocalization was observed between Dmc1 and SCP1 or Dmc1 and SCP2, indicating that Dmc1 preferentially interacts with SCP3. We could not, unfortunately, study the cellular distribution of cohesin complex subunits, since we have not been able to express these proteins in transfected cells.

FIG. 6.

The recombination protein Dmc1 interacts with Rad51 and SCP3 when coexpressed in COS cells. pCMX-Myc-DMC1 (green) was cotransfected into COS cells with either pCMX-FLAG-RAD51/SCP3/SCP2 or -SCP1 (red) and analyzed by immunofluorescence microscopy. Coexpression of Dmc1 and Rad51 leads to the formation of joint fibrillar structures, as seen in the merged picture (yellow). Dmc1 and SCP3 also show extensive colocalization, whereas expression of Dmc1 together with either SCP2 or SCP1 does not yield a colocalization pattern.

Coexpression of SCP2 and SCP3 generates a novel fibrillar protein structure.

The assembly of the SC is likely mediated by interactions among some of the core components of the SC. To ascertain whether SCP1, SCP2, and SCP3 could interact with each other in cells expressing these proteins, different combinations of expression vectors encoding the three proteins were transfected into COS cells (Fig. 7). A striking finding from these experiments was immediately apparent from cells that had been transfected with constructs for SCP2 and SCP3, the two known components of the AE. Coexpressed SCP2 and SCP3 proteins form joint large fibrillar structures in the nuclei of COS cells. Furthermore, the fibrillar protein structures seen in cells expressing both SCP2 and SCP3 are clearly distinct from the protein structures formed by each protein when expressed on its own (compare Fig. 5 and 7). Both the large fibrillar structures observed in cells expressing only SCP3 and the diffuse nuclear staining pattern seen in cells expressing only SCP2 are radically different from the shorter, stubby fibers formed in cells coexpressing the two proteins. In addition, some cells cotransfected with both SCP2 and SCP3 displayed a networklike structure, i.e., the short fibers appeared to be linked (not shown). These results strongly suggest that SCP3 and SCP2 interact in vivo and that coexpression of the two proteins is important for the final structure of the AEs formed in meiotic cells.

FIG. 7.

SCP2 and SCP3 form a novel filamentous structure when coexpressed. pCMX vectors with or without FLAG encoding full-length SCP1, SCP2, or SCP3 were transfected into COS cells and analyzed by immunofluorescence microscopy. The distributions of SCP3 (pCMX-FLAG-SCP3; red) and SCP2 (pCMX-SCP2; green) show a clear colocalization pattern. The two AE proteins form novel fibrillar protein structures not seen in cells expressing only one of the proteins (Fig. 5). Coexpression of SCP3 (pCMX-FLAG-SCP3; red) and SCP1 (pCMX-SCP1; green) or SCP1 (pCMX-FLAG-SCP1; red) and SCP2 (pCMX-SCP2; green) did not produce any clear colocalization patterns.

In cotransfection experiments involving SCP1 and SCP2 or SCP1 and SCP3, we were not able to observe any significant colocalization patterns or formation of novel structures (Fig. 7). The partial colocalization observed for coexpressed SCP1 and SCP2 is most likely a coincidence resulting from the granular expression pattern displayed by both proteins. We have also carried out a triple-transfection experiment including expression plasmids for SCP1, SCP2, and SCP3 and monitored the expression of all three proteins in COS cells. However, this did not result in colocalization between the structures formed by SCP3 or SCP2 and SCP1 (data not shown). Our data therefore suggest that SCP1 is unlikely to interact directly with SCP2 or SCP3 in vivo.

DISCUSSION

Historically, the AE and the TF of the SC have been proposed to be responsible for sister chromatid cohesion and synapsis (46). However, the importance of cohesin complexes in meiotic chromosome pairing clearly suggests that this process is more complex than initially anticipated (16, 47). We have now investigated how cohesin complex assembly and meiotic chromosome synapsis are affected in murine spermatocytes that lack a component of the AE.

We find that the different subunits of the cohesin complex form fiberlike structures irrespective of the presence or absence of the AE proteins SCP3 and SCP2 (Fig. 2). The corelike structure with which the cohesin complex subunits associate most likely corresponds to the chromatin structure that holds sister chromatids together in meiotic cells (16, 47). It has been shown that cohesin complex subunits associate with chromosomes as they replicate prior to entry into meiosis (8, 47), while ultrastructural studies have identified filamentous structures associated with the sister chromatids that are distinct from the AEs (10, 35). Furthermore, immunofluorescence microscopy studies using antibodies against SC and SMC proteins have suggested that chromatin structures with which the cohesins associate can sometimes be mechanically separated from the AEs (13). Our data show that cohesin complex subunits can indeed form an independent chromosomal core in early meiotic cells which does not require an AE (based on the absence of this structure in SCP3-deficient silver-stained nuclei) or SCP3 or SCP2 for its formation or maintenance.

How is the AE formed and how does this structure attach to meiotic chromosomes? The absence of a visible AE in SCP3-deficient spermatocytes, as determined by silver staining (52) or by immunofluorescence microscopy using anti-SCP2 antibodies (Fig. 1), shows that SCP3 is a primary determinant of AE assembly. This is supported by the observation that SCP3 forms long homotypic filamentous structures that resemble the AEs in cells overexpressing this protein, whereas SCP2 does not (Fig. 5). Furthermore, the inability of SCP2 to associate with meiotic chromosomal cores in SCP3-deficient spermatocytes (Fig. 1) suggests that a second important function of SCP3 is to anchor the AEs to meiotic chromatin. It has been shown that SCP3 can interact with SMC1 and SMC3 in vitro, indicating a possible mechanism for the chromatin-anchoring process (13). What, then, is the function of the SCP2 protein? SCP2 and SCP3 have been shown to interact with each other in a yeast two-hybrid protein interaction assay (44), while coexpression of SCP3 and SCP2 leads to the formation of novel protein structures not seen in cells expressing only one of the proteins (Fig. 5 and 7). This suggests that one important function of SCP2 is to help shape the in vivo structure of the AE by interacting with SCP3.

Why are mammalian meiotic chromosomes associated with two separate yet superimposed fibrillar protein structures, the cohesin cores and the SC? The absence of an AE in mitotic cells suggests that this structure promotes meiosis-specific functions, such as synapsis and/or the promotion of meiotic recombination. Our data, however, do not support such an interpretation. We show that SCP1, a TF component and a marker for synapsis in wild-type spermatocytes (20, 22, 39), forms the same type of short fiberlike structures that overlap with the cohesin cores in both wild-type and SCP3-deficient zygotene spermatocytes (Fig. 3 and 4). This implies that SCP1 in SCP3-deficient spermatocytes binds to meiotic chromatin and that interactions between SCP1 molecules attached to the two homologous chromosomes mediate synapsis. Our data also show that the AE is not required for recruitment of SCP1 to chromatin. Smith and Roeder (41) have shown in similar studies of yeast that the TF protein Zip1 can localize to meiotic chromosomes in a red1 mutant that lacks AEs. How, then, is synapsis achieved in the absence of an AE? Evidence from yeast and plant meiotic cells indicates the existence of multiple fibrous connections between prealigned chromosomal cores prior to SC formation (44, 49, 54). These aligned homologs are further apart than synapsed ones. Zip1 has been shown to accumulate at such axial associations and to promote synapsis (1, 9). Therefore, based on this principle, our data would suggest that SCP1 assembles at sites for axial associations and promotes synapsis in the absence of an AE in mammalian cells. It is possible, however, that the AE found in wild-type spermatocytes promotes a more robust synaptic process once SCP1 has become attached to meiotic chromatin. Such a model could explain the axial gaps seen in the fibrillar structures formed by SCP1 in SCP3-deficient spermatocytes.

A second potential role for the AE could be to recruit recombination proteins to meiotic chromosomes in order to promote homologous recombination and crossovers. It has been shown that the DNA recombination proteins Rad51 and Dmc1 colocalize with the axial element and that they interact with SCP3 in vitro (Fig. 6), suggesting that SCP3 has an important role in organizing the chromosomal distribution of these two proteins (44). We find, however, that the foci formed by the Dmc1 protein remain associated with the cohesin cores in the absence of an AE (Fig. 4). This suggests that Dmc1 is recruited to the sister chromatids by other components of meiotic chromatin, possibly by cohesin complexes. In agreement with this, cohesin complexes have been suggested to promote recombination in yeast and in mammalian cells (45). It is possible that the AE strengthens the binding of Dmc1 to the sister chromatids, since we find that SCP3-deficient spermatocytes have less strongly stained Dmc1 foci.

The colocalization of the recombination protein Dmc1 with cohesin complexes in meiotic cells deficient in the SCP3 protein suggests that an AE is not required for initiation of meiotic recombination. Similar ideas have been proposed by Rockmill et al. (32), who have shown that AEs are not absolutely required for meiotic recombination, since red1 mutants fail to form AEs but still exhibit residual levels of crossing over. To further support these ideas, we monitored the distribution of the DNA recombination protein, Msh4, in wild-type and SCP3-deficient cells. Msh4 has been shown to take part in a meiotic recombination pathway downstream of Dmc1 and to localize to regions of homologous chromosomes in mammalian cells that are undergoing synapsis (17, 36). We find Msh4 predominantly at regions of meiotic chromosomes that are labeled by SCP1 in wild-type and mutant spermatocytes (Fig. 4). This also suggests that in the absence of an AE Msh4 can associate with synapsed meiotic chromosomes and that the SCP1 structures observed in mutant spermatocytes represent synaptic events. We conclude from the above-mentioned experiments that synapsis, as well as recruitment of recombination proteins, is not dependent on an AE.

Based on data presented here and elsewhere (cited below), we propose a model for early meiotic chromosome pairing and synapsis (Fig. 8). In this model, cohesin complex subunits initially assemble between sister chromatids and ensure that they remain bound to each other until the second meiotic division (7, 16, 47, 48). Recombination proteins, such as Dmc1 and Rad51 (as well as proteins such as Zip2 and Zip3, as yet identified only in budding yeast cells), are then recruited to the cohesin cores and promote the formation of axial associations between intersister axes (1, 44). Some of these sites are likely to represent early recombination nodules (4, 6). Binding of TF proteins, such as Zip1 or SCP1, to axial association sites at the meiotic chromosomes initiates synapsis between the two homologs (1). Msh4 binds to synapsed regions and, together with other recombination proteins, gives rise to late recombination nodules and crossovers (36). As shown in Fig. 8, we find that recruitment of neither recombination proteins nor TF proteins is dependent on AE functions in SCP3-deficient spermatocytes. Rather, the cohesin core appears to be sufficient for these functions.

FIG. 8.

Model for initiation of synapsis in mammalian meiotic cells. Schematic model showing the successive stages of early meiotic prophase I cells in wild-type and SCP3-deficient spermatocytes. At the leptotene-zygotene stage (middle) in wild-type meiotic cells, two superimposed chromatin structures, the AE and the cohesin core, are attached to the two homologous chromosomes. Recombination proteins, such as Dmc1, are recruited to the two intersister axes. The formation of axial associations between the two intersister axes then leads to a parallel alignment of the two homologs during the zygotene stage. SCP1 and Msh4 are recruited to axial association sites and mediate synapsis and crossovers between homologs. Synapsis is completed at the pachytene stage (right). The left panel shows that recombination proteins, such as Dmc1 and Msh4, as well as TF proteins, such as SCP1, are recruited to the cohesin cores in the absence of an AE. This suggests that the AE is not required for initiation of DNA recombination or synapsis and that these functions are instead provided by proteins that are part of the cohesin cores.

It has been argued that meiosis represents a modification of mitosis (15). To convert mitotic cells into meiotic cells would require several changes in chromosomal behavior that can be achieved either by functional modifications of an already existing chromatin-associated structure (the cohesin cores) or by the creation of a new structure (the AE or SC). In support of the first model, it has been shown that meiosis-specific subunits of the cohesin complex exist and that these variants are essential for meiosis (7, 16, 47). Synapsis is not observed in fission yeast meiotic cells, and no SC, no TF, and only a fragmented AE-like structure are observed in these cells (5, 18). Fission yeast meiotic cells might, therefore, have undertaken a minimal set of changes to accomplish a meiotic function, relying to a large extent on modified cohesin cores. In this sense, the zygotene-stage cells seen in SCP3-deficient spermatocytes are similar to fission yeast meiotic cells, as they also lack an AE and an SC. Despite this, however, our results suggest that the remaining cohesin cores in SCP3-deficient spermatocytes are sufficient for recruiting recombination proteins and TF proteins. Assuming that the cohesin cores in meiotic cells were initially modified to promote meiotic functions (as is perhaps seen in today's fission yeast meiotic cells), these functions appear to be retained in mammalian meiotic cells despite the parallel existence of AEs in these cells. Meiosis-specific cohesin complex subunits have also been found in mammalian cells (28, 30), strengthening the argument that the cohesin complex has an important function in these cells. The challenge now remains to reveal a particular function for the AE of the SC and to elucidate why the AEs have been superimposed on the cohesin cores in mammalian cells. Based on our experimental data, putative functions of the AE include stabilization of recombination protein complexes bound to meiotic chromosomes, completion of synapsis, and chromosomal segregation. One or more of these functions are essential in early meiotic cells, as the absence of the SCP3 protein leads to apoptosis at the zygotene stage of meiosis (52).