RAD21L, a novel cohesin subunit implicated in linking homologous chromosomes in mammalian meiosis

Meiosis is a sexual reproduction process in which chromosome number is halved from the diploid in a parental cell to the haploid in daughter cells. To this end, two successive meiotic divisions occur after a single round of DNA replication. Meiotic chromosomes exhibit highly unique behaviors particularly in the first meiotic division (meiosis I). Homologous chromosomes pair, synapse, and recombine with their partners during prophase I. The chiasmata resulting from crossover recombination, in conjunction with arm cohesion between sister chromatids, make physical connections between homologous chromosomes, leading to the formation of bivalent chromosomes by metaphase I. At the onset of anaphase I, resolution of the inter-sister arm cohesion causes separation of homologous chromosomes while sister chromatids are kept attached at their centromere regions. The sister chromatids eventually separate from each other when the centromeric cohesion is resolved in meiosis II (Miyazaki and Orr-Weaver, 1994; Petronczki et al., 2003). To achieve this highly intricate series of chromosomal events, meiotically dividing cells are equipped with specialized components and regulatory mechanisms.

Meiotic recombination is initiated by DNA double-strand breaks (DSBs) created by the topoisomerase II–like protein SPO11 (Keeney, 2001). The DSBs are then processed in a series of recombination events, and eventually converted into two products, crossovers and noncrossovers (Page and Hawley, 2004; de Boer and Heyting, 2006). At first, the DSBs are resected to yield 3′ single-strand DNA ends, which in turn allow single-end invasion mediated by RAD51 and DMC1 (Schwacha and Kleckner, 1997; Hunter and Kleckner, 2001; Keeney, 2001). In mammalian meiosis, the early recombination intermediates containing RAD51 and DMC1 are gradually replaced first by middle intermediates containing MSH4 (a homologue of Escherichia coli MutS) and eventually by late intermediates containing MLH1 (a MutL homologue; Plug et al., 1998; Santucci-Darmanin et al., 2000). MLH1 is essential for crossover recombination, appearing only at the crossover sites but not at the noncrossover sites during pachytene (Anderson et al., 1999; Santucci-Darmanin et al., 2000; Guillon et al., 2005).

In many organisms, the pairing, synapsis, and recombination of homologous chromosomes occurs in parallel with the formation of some cytologically discernible structures (Page and Hawley, 2004). The so-called axial elements (AEs) assemble along each chromosome in the leptotene stage, and start to pair and synapse in zygotene, thereby initiating the assembly of the synaptonemal complex (SC). In pachytene, synapsis completes along their entire length of the lateral elements (LEs; an alternative name for the AE after synapsis). The resulting SC has a tripartite structure with two LEs connected by transverse filaments called the central element (CE), which then starts to disassemble in diplotene. In mammals, SYCP2 and SYCP3 are the main components of the AE/LE whereas SYCP1 constitutes the CE (Meuwissen et al., 1992; Dobson et al., 1994; Offenberg et al., 1998), and all these components are essential for proper progression of meiosis. Both SYCP2- and SYCP3-deficient male mice fail to form the AEs and are sterile due to massive apoptotic cell death at or before pachytene (Yuan et al., 2000; Yang et al., 2006). In SYCP1-deficient mice the AEs are formed and aligned normally, but neither synapse nor form MLH1 foci (de Vries et al., 2005).

A series of recent studies has demonstrated that multi-subunit protein complexes, known as cohesins, play key roles in many aspects of meiotic chromosome dynamics. The canonical cohesin complex, originally identified as a central player in sister chromatid cohesion in mitosis, is composed of four subunits: two structural maintenance of chromosomes (SMC) subunits (SMC1α and SMC3) and two non-SMC subunits (RAD21 [known as Scc1/Mcd1 in budding yeast] and SA/STAG [known as Scc3 in the yeast]; Hirano, 2006; Nasmyth and Haering, 2009). In mammals, mitotic cells possess two types of cohesin complexes with different SA subunits (i.e., SMC1α–SMC3–RAD21–SA1 and SMC1α–SMC3–RAD21–SA2; Losada et al., 2000). On the other hand, meiotic cells, such as spermatocytes and oocytes, express at least three additional subunits, SMC1β (an SMC1α paralog), REC8 (a RAD21 paralog), and STAG3 (an SA1/SA2 paralog). Many if not all of these cohesin subunits have been localized along the AE/LEs of the SC (Eijpe et al., 2000, 2003; Prieto et al., 2001, 2002, 2004; Revenkova et al., 2001; Lee et al., 2003; Xu et al., 2004). Among them, REC8 and SMC1β have been shown to be essential for proper assembly of the SC and crossover recombination (Revenkova et al., 2004; Xu et al., 2005). Expression of these meiosis-specific subunits would predict the occurrence of a wide variety of different cohesin complexes with different subunit compositions in meiotic cells. However, the biochemical nature of these putative complexes is not fully understood.

Most eukaryotic species possess two kleisin subunits, RAD21 and REC8. Here we report, for the first time, the presence of the third kleisin subunit, designated RAD21L, in mammals. The orthologues of RAD21L are found in vertebrates, but not in invertebrate species. RAD21L is specifically expressed in meiotic cells (spermatocytes and oocytes), and localizes along the AE/LEs of the SC from leptotene to mid pachytene. One of the most striking features of RAD21L is that, unlike any other cohesin subunit, it dissociates from the SC at mid pachytene and is not detectable in the subsequent stages of meiosis. We provide several lines of evidence suggesting that RAD21L may be involved in synapsis initiation and crossover recombination between homologous chromosomes. The current results allow us to propose a model in which the spatiotemporal regulation of a variety of cohesin complexes contributes to progression of the elaborate sequence of chromosomal events during meiotic prophase I.

In the course of BLAST searches using the mouse RAD21 sequence as a query, we noticed that vertebrate genomes encode a protein closely related, but not identical, to RAD21 of each species. In the current study, we refer to this novel gene product as RAD21L (RAD21-like protein). A single copy of the RAD21L gene was found in each of the vertebrate genomes sequenced so far. RAD21L was, however, not found in any of the invertebrate species including amphioxus, living proxies for the chordate ancestor (Putnam et al., 2008). A phylogenetic analysis of kleisin subunits of cohesins among vertebrates indicated that they were categorized into three subfamilies, namely, RAD21, RAD21L, and REC8, and that RAD21L was much closer to RAD21 than to REC8 (Fig. 1 A). An alignment of the amino acid sequences of mouse RAD21, RAD21L, and REC8 is shown in Fig. S1. RAD21L displayed 44% and 17% identity to RAD21 and REC8, respectively, in total length: conserved residues were clustered into two winged-helix domains located at their N and C termini (Fig. 1 B). All the information taken together, it is reasonable to speculate that the RAD21L gene has diverged from the RAD21 gene immediately after the appearance of vertebrates. It should be noted that, unlike RAD21L, REC8 is found in most eukaryotic organisms.

We cloned a cDNA for RAD21L from a mouse testis cDNA library by PCR (see Materials and methods). In a Northern blot analysis using mRNAs isolated from various mouse organs, virtually no signal for RAD21L was detectable (unpublished data). Under the same condition, RAD21 mRNA was detectable ubiquitously and REC8 mRNA was found only in testes, as described previously (Lee et al., 2002). To detect a supposedly low level of RAD21L mRNA, we then performed quantitative PCR (qPCR) using cDNAs prepared from liver, spleen, and testis (Fig. 2 A). Consistent with the results of the Northern analysis, RAD21 mRNA (along with GAPDH mRNA used as a positive control) was readily detectable in all three organs whereas REC8 mRNA was detected only in testis, but hardly in liver or spleen. Most importantly, this approach allowed us to detect RAD21L mRNA in testis, but not in the others. The cycle numbers and the efficiencies of amplification in the real-time PCR allowed us to roughly estimate the relative abundance of mRNAs for each gene (see Materials and methods). The level of RAD21L mRNA was estimated to be extremely low even in testis, being <1% of that of RAD21 mRNA. In contrast, the level of REC8 mRNA was roughly equivalent (>60%) to that of RAD21 mRNA in testis. These results reveal that the expression of RAD21L is testis specific, and may be limited in specific cell types present in the testis organ.

To analyze RAD21L at the protein level, rabbit and rat polyclonal antibodies were raised against its C-terminal recombinant fragment. Both antibodies recognized a polypeptide with an apparent mol wt of ∼95 kD on an immunoblot against a testis extract (Fig. 2 B). The size was substantially bigger than the predicted one (∼63 kD), a characteristic commonly observed among many kleisin subunits. We then performed an immunoblot analysis against extracts prepared from various mouse organs using a set of antibodies against the kleisin subunits and SMC3 (a core subunit of cohesins; Fig. 2 C). Although RAD21 and SMC3 were detected in all the organs examined, RAD21L and REC8 were found only in testis. RAD21L was also found in ovary as judged by immunoblot analysis after immunoprecipitation using an anti-RAD21L antibody (unpublished data). We therefore conclude that RAD21L is specifically expressed in both male and female gonads.

To determine which cell types in the gonads might be positive for RAD21L, frozen sections of the ovary and testis were immunofluorescently labeled with anti-RAD21L (Fig. 2 D). To identify meiotic cells in the gonads, an antibody against SYCP3 (a component of the AE/LE) was used simultaneously. We found that, in both gonads, RAD21L was detectable only in the cell population positive for SYCP3. Virtually no specific signal of RAD21L was found in somatic cells, spermatogonia (premeiotic cells), or spermatids (post-meiotic cells). Thus, RAD21L is specifically expressed in meiotic cells in both male and female. The big difference in the mRNA level between RAD21L and the other kleisin genes, as judged by the quantitative PCR, might be attributed not only to the exclusive expression of RAD21L in meiotic cells but also to the high abundance of RAD21 and REC8 mRNAs in spermatids (Lee et al., 2002).

To directly compare the expression timing and distribution of RAD21 and RAD21L during meiosis, we performed immunofluorescent labeling of testis sections with specific antibodies to RAD21 and RAD21L (Fig. 3 A). In the testis, spermatogenesis proceeds in a centripetal direction from the inner basement membrane (Fig. 3 A, a and e; green dashed lines) to the lumen of the seminiferous tubules. RAD21L was specifically detected only in some population of spermatocytes near the basement membrane (Fig. 3 A, f and h; white arrowheads). In contrast, the signal of RAD21 was prominent in somatic cells outside the seminiferous tubules or in spermatogonia, and was also detected in some spermatocytes localizing away from the basement membrane (Fig. 3 A, g and h; yellow arrowheads). Remarkably, however, the RAD21 signal was hardly detectable in the RAD21L-positive spermatocytes. We also performed immunofluorescent colabeling of RAD21 and RAD21L on nuclear spreads of testicular cells (Fig. 3 B). Again, most of the RAD21L-positive spermatocyte nuclei were not labeled with RAD21, and vice versa (Fig. 3 B, a–d). In a small population of pachytene-like nuclei, however, faint signals for both RAD21 and RAD21L were detectable (Fig. 3 B, e–h). Interestingly, the two signals hardly overlapped with each other and apparently alternated along the SC (Fig. 3 B, f′–h′), most likely corresponding to the transition stage from RAD21L to RAD21.

These results revealed that RAD21L largely replaces RAD21 during early stages of prophase I, and that, conversely, RAD21 reappears in later stages of prophase I to replace RAD21L. Two previous studies had reported conflicting results regarding the temporal expression of RAD21 in mouse spermatocytes: one study reported a transient appearance of RAD21 at a late stage of prophase I (Prieto et al., 2002), whereas the other reported its persistent presence throughout prophase I (Xu et al., 2004). The former study was more consistent with our observation reported here.

To examine when RAD21L starts to appear in meiotic cells, nuclear spreads of testicular cells were immunofluorescently colabeled with antibodies against RAD21L and PCNA, a marker for S phase (Bravo and Macdonald-Bravo, 1987). We found that the nuclei of some spermatocytes were simultaneously positive for RAD21L and PCNA (Fig. S2), suggesting that RAD21L is expressed from premeiotic S phase, as had been observed for REC8 (Eijpe et al., 2003).

To determine the exact stages in which RAD21L appears in meiotic chromosomes, nuclear spreads of spermatocytes were colabeled with antibodies against RAD21L and SYCP3. In the leptotene stage when the AEs started forming, RAD21L was already present and localized along the AEs (Fig. 4, a–d). In the zygotene stage when synapsis initiated, RAD21L was detected as dot- or thread-like signals along the AE/LEs of the SC being assembled (Fig. 4, e–h). Notably, the signals of RAD21L appeared enriched on synaptic sites where two AEs made a contact to initiate synapsis (Fig. 4, h′ and h′′; arrowheads). In pachytene when the synapsis was apparently completed, the RAD21L signals distributed evenly along the entire length of the SC in some nuclei (Fig. 4, i–l), and were hardly detectable in others (Fig. 4, m–p). The RAD21L signals were undetectable in later meiotic stages, from diplotene (Fig. 4, q–t) through metaphase I (not depicted). Taken together with the results from testis sections (Fig. 3), these data strongly suggested that RAD21L disappears from the SC in the middle of the pachytene stage (i.e., mid pachytene). In fact, this conclusion was supported by a colabeling experiment with antibodies against MSH4 and MLH1 (see below, Fig. S3 B, and Fig. 6 B), whose signals had been used to assess sub-stages within pachytene (Plug et al., 1998; Santucci-Darmanin et al., 2000). It was somewhat surprising for us to find that RAD21L disappeared from the meiotic chromosomes at mid pachytene when the SC was still intact as judged by SYCP3 labeling. This behavior in early prophase I is absolutely unique to RAD21L, being in striking contrast to the behaviors of known cohesin subunits.

Previous studies had shown that REC8 is localized along AE/LEs of the SC from leptotene to diplotene (Eijpe et al., 2003; Lee et al., 2003). To directly compare the distribution of RAD21L with that of REC8, nuclear spreads of spermatocytes were colabeled with antibodies against the two meiotic kleisins. Both RAD21L and REC8 signals first became detectable as arrays of dot-like structures in leptotene (Fig. 5, a–d), and were gradually lined up into more continuous structures from zygotene (Fig. 5, e–h) through early pachytene (Fig. 5, i–l). In diplotene, the RAD21L signal was no longer detectable whereas REC8 was retained on the desynapsed LEs (Fig. 5, m–p). Very importantly, although RAD21L and REC8 largely coexisted along the AE/LEs from leptotene to early pachytene, their detailed localizations were distinct from each other. Indeed, magnified images revealed that the signals of RAD21L and REC8 did not overlap and apparently alternate along the AE/LEs at zygotene (Fig. 5, f′–h′) and early pachytene (Fig. 5, j′–l′). These results provide the first line of evidence for the existence of alternating sub-chromosomal domains containing distinct kleisin subunits along meiotic chromosomes.

To test whether the disappearance of RAD21L from meiotic chromosomes at mid pachytene might correlate with other chromosomal events, we performed colabeling experiments. We initially focused on the dynamics of histone H2AX phosphorylated at S139 (γ-H2AX). In addition to two strong signals of γ-H2AX that had been reported before, a hitherto unrecognized signal on the SC was noticed as described below. A strong and chromosome-wide signal of γ-H2AX first appeared in leptotene and zygotene (Fig. 6 A, a–d and e–h). This signal, which is thought to be dependent mostly on ATM in response to DSBs created by SPO11 (Mahadevaiah et al., 2001; Barchi et al., 2005; Bellani et al., 2005), largely disappeared in early pachytene, yet a sub-fraction was left along the proximity of the SC (Fig. 6, i–l). At the same time, the second population of strong γ-H2AX signals started to accumulate onto the XY sex chromosomes in early pachytene, as had been reported previously (Chicheportiche et al., 2007; Blanco-Rodríguez, 2009). This is an indicator of the formation of XY body, and is thought to be dependent on ATR (Turner et al., 2004). It was noteworthy that, although the γ-H2AX signal on XY body stayed robust throughout pachytene, the signal on the SC was weakened gradually around mid pachytene and lost almost completely by late pachytene (Fig. 6, m–p and q–t). The disappearance of γ-H2AX from the SC occurred highly coincidentally with that of RAD21L, raising the possibility that RAD21L might contribute, directly or indirectly, to processing of DSBs and/or completion of their repair.

To test this idea further, we next examined the temporal relationship between RAD21L signals and RAD51 foci, a marker for early recombination intermediates, or MSH4 foci, a marker for middle recombination intermediates. We found that RAD51 foci appeared in leptotene and largely disappeared in early pachytene, as had been reported previously (Plug et al., 1998; Tarsounas et al., 1999). The disappearance of RAD51 foci from meiotic chromosomes precedes the disappearance of RAD21L signals (Fig. S3 A). On the other hand, MSH4 foci were detected at late zygotene (unpublished data) and early pachytene on the SC, but not at late pachytene (Fig. S3 B), as had been reported previously (Santucci-Darmanin et al., 2000). The disappearance of RAD21L signals nicely correlated with that of MSH4 foci.

Finally, we looked at MLH1 foci, a marker for crossovers (Anderson et al., 1999). In early pachytene when RAD21L signals localized along the SC, MLH1 foci were rarely observed (Fig. 6 B, a–d). A few MLH foci started to appear in mid pachytene when the RAD21L signals got gradually reduced (Fig. 6 B, e–h and h′). The number of MLH1 foci increased in late pachytene when RAD21L was no longer detectable (Fig. 6 B, i–l). Thus, the disappearance of RAD21L from pachytene chromosomes correlates with the formation of crossovers.

Despite an extensive series of cytological studies, there have been only a few biochemical studies concerning multiple cohesin complexes in mammalian meiosis (Lee et al., 2003; Revenkova et al., 2004; Suja and Barbero, 2009). As a consequence, we currently have very limited information about how many different combinations of subunits might create how many different types of cohesin complexes in meiotic cells. Our current discovery of RAD21L, a novel, putative subunit of cohesins, provides an additional layer of complexity to this problem. To clarify this important problem, we performed systematic coimmunoprecipitation assays with the antibodies to RAD21L as well as all known cohesin subunits, using an extract from testes.

In the first set of experiments, a testis extract was subjected to immunoprecipitation with antibodies against the three kleisin subunits, and the precipitates were analyzed by immunoblotting (Fig. 7 A). Only a single kleisin subunit was detected in each of the immunoprecipitates, indicating that different kleisin subunits do not coexist in any cohesin complexes. SMC1α was coprecipitated with RAD21 but with neither RAD21L nor REC8, whereas SMC1β (and SMC3) was coprecipitated with all three kleisins. The amounts of the SMC subunits coprecipitated with RAD21L were small in all cases, most likely reflecting the very limited amount of RAD21L present in the testis extract (Fig. 2; also see Fig. S4 A). SA1 and SA2 were coprecipitated only with RAD21, whereas STAG3, the STAG/SA subunit whose expression is limited to meiosis, was coprecipitated with all three kleisins. Interestingly, only the fraction of STAG3-coprecipitated RAD21L displayed a mobility shift (Fig. 7 A, arrow), which turned out to be dependent on phosphorylation (Fig. 7 D). In the next set of experiments, antibodies against three different SMC subunits were used for immunoprecipitation (Fig. 7 B). Anti-SMC3 antibody precipitated all subunits tested here (except SA1), and SMC1α and SMC1β were never coprecipitated with each other, indicating that either SMC1α–SMC3 or SMC1β–SMC3 represents the core subunit of all cohesin complexes. RAD21 and RAD21L associated with either SMC1α or SMC1β, whereas REC8 associated with SMC1β only. Thus, the failure to detect SMC1α in the RAD21L immunoprecipitates likely reflects the limited amount of RAD21L present in the extract as described above. Moreover, SA2 was coprecipitated with SMC1α, whereas the vast majority STAG3 was coprecipitated with SMC1β. In the final set of experiments, antibodies against the SA–STAG subunits were used for immunoprecipitation (Fig. 7 C). Like kleisin subunits, each of the SA–STAG subunits formed distinctive cohesin complexes. We found that RAD21 associated primarily with SA1 or SA2 ,whereas RAD21L and REC8 associated preferentially with STAG3.

The results described above allow us to conclude that the testis extracts used in the current study contain at least four different meiotic cohesin complexes in addition to the canonical mitotic complexes (Fig. 7 E). In brief, RAD21 and RAD21L can form complexes with both SMC1α and SMC1β, whereas REC8 associates with only SMC1β. Two meiotic kleisin subunits, RAD21L and REC8, associate preferentially with STAG3, whereas RAD21 associates with SA1, SA2, and STAG3. Intriguingly, at least under the current conditions tested, phosphorylation of STAG3 is prominent only when it is complexed with RAD21L.

In the current study, we report a novel cohesin subunit, referred to as RAD21L, which localizes along the AE/LEs of the SC exclusively from leptotene to mid pachytene in mammalian meiosis. Combined with the expression profiles of known cohesin subunits (Eijpe et al., 2000, 2003; Prieto et al., 2001, 2002, 2004; Revenkova et al., 2001; Lee et al., 2003; Xu et al., 2004), we propose the following diagram in which spatiotemporal coordination of multiple cohesin complexes supports the elaborate sequence of chromosomal events during meiotic prophase I (Fig. 8). According to this scenario, RAD21-containing cohesin (cohesin^RAD21) is largely replaced by cohesins containing RAD21L and REC8 (cohesin^RAD21L and cohesin^REC8, respectively) during premeiotic S phase. Cohesin^RAD21L exists as two distinct types: one containing SMC1α and the other SMC1β, whereas cohesin^REC8 is most likely to represent a single form containing SMC1β (Fig. 7 E). Although cohesin^RAD21L apparently colocalizes with cohesin^REC8 along the AE/LEs by mid pachytene, their detailed distributions do not overlap with each other, implicating the existence of alternating sub-chromosomal domains along the SC. Notably, cohesin^RAD21L disappears from the SC after mid pachytene, and is replaced with cohesin^RAD21, whereas cohesin^REC8 stays on meiotic chromosomes. At this and later stages in prophase I, the majority of cohesin^RAD21 is likely to contain SMC1β rather than SMC1α (Revenkova et al., 2001, 2004).

A previous study using SYCP3-deficient (SYCP3^−/−) mice indicated that a cohesin axis is formed even in the absence of the AE (Pelttari et al., 2001). This observation led the authors to propose that the cohesin axis is preformed along each chromosome and acts as the organizing framework for subsequent AE/LE anchorage and formation. On the other hand, in REC8^−/− or SMC1β^−/− mice, the AEs are assembled partially (Revenkova et al., 2004; Xu et al., 2005). Because thread-like signals of RAD21L are observed before the appearance of linear signals of SYCP3 (unpublished data), as has been reported for REC8 (Eijpe et al., 2003; Lee et al., 2003), it is reasonable to speculate that RAD21L might act in concert with REC8 to help assemble the AEs on their own axes.

In early zygotene, SYCP1 is recruited to between the two AEs of homologous chromosomes and participates in the assembly of the SC (Meuwissen et al., 1992). What factor(s) might recruit SYCP1 to the AEs and to help initiate synapsis is currently unknown. We suggest that RAD21L would be an excellent candidate for such a factor for the following two reasons. First, RAD21L foci appear enriched at connection sites between two AEs undergoing synapsis at zygotene (Fig. 4, h′ and h′′). Second, our preliminary results show that RAD21L, but neither RAD21 nor REC8, is coimmunoprecipitated with a fraction of SYCP1 (Fig. S4 B). If this hypothesis were correct, then the disappearance of RAD21L at mid pachytene would be an essential prerequisite for starting desynapsis (and to preventing resynapsis) at diplotene.

It is important to note that, unlike RAD21L, all other meiotic cohesin subunits characterized so far (SMC1β, REC8, and STAG3) persist at the inter-sister chromatid regions at least until the onset of anaphase I when homologous chromosomes start to separate. It is therefore reasonable to assume that they possess the canonical function of cohesins (i.e., holding sister chromatids together) at least in part. Indeed, precocious separation of sister chromatids is observed in SMC1β^−/− oocytes or oocytes depleted of SGO2, a protector of centromeric REC8 (Revenkova et al., 2004; Hodges et al., 2005; Lee et al., 2008). Interestingly, in REC8^−/− mice, synapsis occurs between sister chromatids instead of between homologous chromosomes (Xu et al., 2005), implicating that REC8-mediated cohesion between sister chromatids is a prerequisite for proper recognition of homologous chromosomes to initiate their synapsis. The unique behavior of RAD21L reported in the current study indicates that, if RAD21L had a function in holding sister chromatids, it would be limited until mid pachytene. We therefore suggest the alternative (yet not mutually exclusive) possibility that RAD21L uses early recombination intermediates to recognize homologous chromosomes and then holds a pair of nonsister duplexes from homologous chromosomes (Fig. 8, inset). This situation could in turn provide a structural base for recruitment of SYCP1, thereby coupling homologous chromosome pairing to synapsis initiation.

The signal of γ-H2AX, a marker for DSBs, initially appears chromatin-widely from leptotene through zygotene, and largely disappears by early pachytene. At this stage, a sub-fraction of the γ-H2AX signal remains on autosomes only at the proximity of RAD21L localizing the SC (Fig. 6 A, c, g, and k). Remarkably, this particular sub-fraction of γ-H2AX and RAD21L disappear highly coincidentally around mid pachytene, and MLH1 foci appear at the same time. Thus, RAD21L apparently persists along the SC until processing of all DSBs is complete. This observation implies that RAD21L may be involved in the processing of recombination intermediates occurring along the SC until mid pachytene. In this context, it is important to ask why RAD21 and RAD21L might appear on chromosomes in a mutually exclusive manner in meiotic prophase I. A recent study shows that RAD21 facilitates recombination between sister chromatids, whereas it suppresses recombination between homologous chromosomes at the same time (Covo et al., 2010). It is tempting to speculate that replacement of RAD21 with RAD21L in early prophase I would help reverse this situation, thereby creating an environment in which recombination between homologous chromosomes becomes preferable. The proposed function of RAD21L to hold two nonsister duplexes from homologous chromosomes (Fig. 8, inset) provides a mechanistic explanation for how RAD21L might promote meiotic recombination. However, we cannot rule out the possibility that RAD21L might be involved in other events occurring during pachytene, such as pachytene checkpoint responses (Roeder and Bailis, 2000), apoptosis of cells defective in meiotic sex chromosome inactivation (Turner, 2007), and deposition of histone H1t (Drabent et al., 1996).

Perhaps one of the most important observations made in the current study is that RAD21L disappears from the SC at mid pachytene. No such behavior has been reported for the canonical cohesin subunits or the canonical SC components (Suja and Barbero, 2009). Moreover, BRCA1 and ATR are localized along unsynapsed AEs (Turner et al., 2004), whereas HORMAD1 and HORMAD2, recently identified HORMA domain–containing proteins, preferentially associate with unsynapsed and desynapsed regions of the AE/LEs (Wojtasz et al., 2009; Fukuda et al., 2010). Thus, the apparent loss of RAD21L at mid pachytene from the fully synapsed chromosomes is exceptionally unique (except the γ-H2AX signal described above).

Then what mechanism might promote the dissociation of RAD21L from the SC at mid pachytene? In mitotic prophase, phosphorylation of the STAG–SA subunits (SA1 and SA2) is known to contribute to the release of cohesins from chromosome arms (Losada et al., 2002; Sumara et al., 2002; Hauf et al., 2005). In the current study, we find that STAG3 associates with the three different kleisin subunits (RAD21, RAD21L, and REC8) in meiosis, but only the fraction of STAG3 associating with RAD21L is phosphorylated. It is therefore possible that phosphorylation of STAG3 contributes to the release of cohesin^RAD21L at mid pachytene. However, the current data do not exclude the possibility that STAG3 phosphorylation might regulate other unique functions attributed to cohesin^RAD21L on the SC. Another candidate factor that could contribute to the release of cohesin^RAD21L is WAPL, a protein that plays a direct role in releasing cohesins from chromosome arms in mitotic prophase (Gandhi et al., 2006; Kueng et al., 2006; Shintomi and Hirano, 2009). Interestingly, mouse WAPL has been localized along the SC in pachytene spermatocytes (Kuroda et al., 2005), making the possibility highly plausible that WAPL also functions as a key regulator of meiotic cohesins, including cohesin^RAD21L. Future studies should explore this possibility.

In Drosophila, male meiosis involves neither SC formation nor meiotic recombination. Although the SC is formed in female meiosis, homologous synapsis does not depend on DSBs or recombination (Page and Hawley, 2004). Curiously, no putative orthologue of REC8 has been identified in Drosophila. Instead, C(2)M, a protein that has been reported to have a very limited similarity to kleisins, is expressed specifically in meiosis (Manheim and McKim, 2003; Schleiffer et al., 2003), and may function together with SMC3 (Heidmann et al., 2004). C(2)M is localized in the central region adjacent to the LEs, and is required for both the assembly of the CE component C(3)G and proper formation of crossovers (Manheim and McKim, 2003; Anderson et al., 2005). It is very interesting to note that C(2)M disappears from the SC in late pachytene (Manheim and McKim, 2003; Heidmann et al., 2004). This particular feature and the proposed function of C(2)M are highly reminiscent of those of mouse RAD21L reported in the current study. We therefore suggest that the “functional homologue” of Drosophila C(2)M in vertebrates could indeed be RAD21L rather than REC8. REC8-equivalent functions could be assigned to other chromosomal proteins in Drosophila, such as ORD and SOLO, although neither of them displays any similarity to cohesin subunits at the primary structure level (Webber et al., 2004; Yan et al., 2010).

Like female Drosophila, Caenorhabditis elegans does not require DSBs for alignment or synapsis of homologous chromosomes (Bhalla and Dernburg, 2008). In C. elegans, three kleisin subunits (REC-8, COH-3, and COH-4) apparently cooperate to support cohesion-related functions by meiosis I, although REC-8 alone is sufficient for sister centromere cohesion in meiosis II (Severson et al., 2009). Therefore, either one or both COH-3 and COH-4 could play a role similar to that of Drosophila C(2)M and vertebrate RAD21L. Given their highly divergent primary sequences, however, neither C(2)M nor COH-3 or COH-4 is to be placed at the position of the “RAD21L orthologues.” It is most likely that they might have acquired related if not identical functions through convergent evolution. Whatever the precise relationship might be, our finding of RAD21L in vertebrates shed new light on the mechanism by which intricate combinations of cohesins and noncohesin components have evolved to support the fundamental program that ultimately creates physical connections between homologous chromosomes in meiosis.

The present study illuminates the molecular mechanism by which concerted actions of multiple forms of cohesins contribute to the progression of a series of meiotic events, e.g., pairing, synapsis, and recombination of homologous chromosomes and repair of DSBs. In vertebrates, RAD21L might have evolved as a conductor that orchestrates a series of complicated events to proceed in a timely and organized fashion. To test the proposed functions of RAD21L, we are now embarking on the generation of RAD21L knockout mice.

All animal experiment protocols used in the present study have been approved by the Director of RIKEN Wako Institute, following a review by the Wako Animal Experiment Committee.

Single-stranded cDNAs were generated from total RNA extracted from testes of an adult C57BL/6 mouse with Superscript III reverse transcription (Invitrogen). The resultant cDNAs were used as a template for PCR to amplify the cDNA sequence encoding a hypothetical protein (NCBI Protein database accession no. NM_001114677), referred to here as RAD21L. The following primers were designed and used for RT-PCR: 5′-ATGTTCTACACTCATGTGCT-3′ and 5′-TCACATCTTATAGAACATTG-3′. The PCR products were cloned into pGEM-T-Easy vector (Promega) and then subjected to sequence analysis. The nucleotide sequence data will appear in the GenBank/EMBL/DDBJ databases with the accession no. AB574185.

Total RNAs from liver, spleen, and testes of 7–10-wk-old C57BL/6 mice (Agilent Technologies) were treated with TURBO DNase (Applied Biosystems) to remove contaminating genomic DNAs. Using the RNAs as a template, single-stranded cDNAs were generated and used for real-time PCR. Quantitative analysis of gene expression was performed using a Fast Real-Time PCR System (model 7900HT; Applied Biosystems) in 30 µl of total reaction volume containing 100 ng of cDNAs with 250 nM TaqMan probes and 900 nM forward and reverse primers. The probes and primers used for the analysis are listed in Table S1. As an internal control, gene expression of GAPDH was also monitored using its probe and primers (4352339E; Life Technologies). Standard curves for each gene were drawn by plotting the Ct (threshold cycles) values (Y axis) against the log of cDNA amount (X axis) through conduction of real-time PCR using a serial dilution of a known amount of testis cDNA. The standard curves and accompanying amplification efficiencies in PCR for each gene are used for rough estimation of the relative amount of each gene transcript in a testis.

To produce specific antisera, C-terminal fragments of mouse RAD21 (290–635 aa), mouse RAD21L (185–552 aa), and mouse REC8 (250–591 aa) were expressed in E. coli, purified, and used as antigens to immunize rabbits or rats. The anti-RAD21L and anti-REC8 antisera were passed through a column conjugated with the recombinant fragment of RAD21 to remove fractions potentially cross-reacting with RAD21. Likewise, the anti-RAD21 antiserum was passed through a column conjugated with recombinant RAD21L. Then, the flow-through fractions were loaded onto columns conjugated with their own antigens, and specific antibodies bound to the columns were purified. Rabbits were also immunized with a synthetic peptide containing the C-terminal 18 amino acids of mouse SMC1β to produce polyclonal antisera. Other antibodies used in this study were described elsewhere or purchased from companies: anti-hSMC1 antibody (Gandhi et al., 2006); anti-XSMC3, anti-XSA1, anti-XSA2 antibodies (Losada et al., 1998); anti-SYCP3 antibody (Lee et al., 2003); anti-SYCP1 antibody (ab15087) and anti-MSH4 antibody (ab58666) from Abcam; anti-H2AX phosphorylated (Ser139) antibody (613401) from BioLegend; and anti-RAD51 antibody (sc-8349), anti-STAG3 antibody (sc-20341), anti-PCNA antibody (sc-56), and anti-MLH1 antibody (sc-133203) from Santa Cruz Biotechnology, Inc.

Various organs were collected from 3-mo-old mice and minced with a surgical blade. 50 mg of each sample was homogenized with a pestle, sonicated, and boiled for 3 min in 500 µl of 1x SDS sample buffer (62.5 mM Tris-HCl, pH 6.8, 1% SDS, 10% glycerol, 50 mM DTT, and 0.02% bromophenol blue). After centrifugation at 9,100 g for 10 min at 4°C, the supernatant was collected and used for immunoblot analysis.

For immunoprecipitation analysis, whole extracts from testes of 8–10-wk-old mice were prepared as follows. Five pairs of testis were removed, decapsulated surgically, and homogenized with a Dounce homogenizer in 10 ml of extraction buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 1% Na deoxycholate, 0.1% SDS, 50 mM NaF, 1 mM Na₃VO₄, and 5 mM 2-mercaptoethanol) supplemented with a Protease Inhibitor Cocktail tablet (Roche). The homogenized solution was sonicated and centrifuged at 200,000 g for 30 min at 4°C, and the supernatant was collected and used as testis extracts.

For immunoprecipitations, testis extracts were incubated with one of the antibodies against cohesin subunits for 1.5 h at 4°C with rotor agitation. After the addition of rProtein A Sepharose (GE Healthcare), the extracts were further incubated for 1 h; the beads were then washed five times with the extraction buffer, and the immunoprecipitates were analyzed by SDS-PAGE followed by immunoblotting. The signals were detected using either Immobilon Western Chemilluminescent HRP Substrate (Millipore) or SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific). For protein phosphatase treatment (Fig. 7 D), immunoprecipitates obtained with anti-RAD21L antibody were washed five times with the extraction buffer and twice with 1x λ-protein phosphatase (λ-PPase) buffer, then incubated with 20 U/µl λ-PPase (New England Biolabs, Inc.) in the buffer for 30 min at 30°C. As a control, the immunoprecipitates were incubated without λ-PPase in the buffer. Then the immunoprecipitates were subjected to immunoblot analysis with anti-STAG3 and anti-RAD21L antibodies.

Testes or ovaries were frozen in O.C.T. Compound Embedding medium (Sakura) and kept at −80°C until use. Cryostat sections of 10-µm thickness were prepared at −20°C, air dried on MAS-GP–coated slides (Matsunami), and fixed in 1% paraformaldehyde in PBS for 10 min. After washing with PBS three times, the sections were incubated with primary antibodies at appropriate dilutions in 1% BSA in PBS at 4°C overnight. After washing three times with PBS, the sections were incubated with secondary antibodies (Alexa 488– or 568–conjugated anti–mouse, anti–rat, or anti–rabbit IgG antibodies; Invitrogen) at 1:500 dilution at room temperature for 1 h. DNA was counterstained with Hoechst 34580 (Invitrogen).

Meiotic nuclear spreads were prepared according to the methods described previously (Lee et al., 2003). In brief, testicular cells were put on poly-l-lysine–coated coverslips. The cells on the coverslips were placed in 85 mM NaCl for 3 min, and fixed in 1% paraformaldehyde solution (pH 8.2 adjusted with 10 mM sodium borate) for 15 min. The coverslips were rinsed three times for 1 min each with 4% (vol/vol) Photo-Flo (Kodak) in distilled water (pH 8.0) and air dried overnight. The nuclear spreads on coverslips were placed in PBS for 10 min, in 1 µg/ml DNase I (Sigma-Aldrich) in PBS for 20 min, in a detergent solution (5 mM EDTA, 0.25% gelatin, and 0.05% Triton X-100 in PBS) for 10 min, in PBS for 10 min, and finally in a blocking buffer (3% BSA and 0.05% Triton X-100 in PBS) for 30 min. The samples were incubated with the primary antibodies at appropriate dilutions in the blocking buffer at 4°C overnight, and washed with PBS for 10 min, with the detergent solution for 10 min, and with PBS for 10 min. The coverslips were then incubated with the secondary antibodies at appropriate dilutions in the blocking buffer. DNA was counterstained with Hoechst 34580. The samples were mounted with VECTASHIELD Mounting Medium (Vector Laboratories) and observed under a confocal microscope (model LSM710; Carl Zeiss, Inc.) with a Plan Apochromat 20x/0.8 or 100x/1.46 oil DIC objective lens. All images were obtained by scanning a single section of samples using ZEN2008 software (Carl Zeiss, Inc.) and imported into Photoshop (Adobe). Then gamma adjustment was performed for each of the RGB channels. For double labeling of RAD21L and REC8 (Fig. 5), chromosome spreads were treated simultaneously with a mixture of the two primary antibodies. We also tested sequential treatments with the two antibodies and obtained exactly the same result, eliminating the possibility that they may compete with each other for a common epitope.

Fig. S1 shows the alignment of amino acid sequences between mouse RAD21, RAD21L, and REC8. Fig. S2 shows that RAD21L appears in premeiotic S phase. Fig. S3 shows the relative timing of the disappearance of RAD21L and RAD51 or MSH4 from the synaptonemal complex. Fig. S4 shows that the relative amount of RAD21L is extremely low compared with other kleisin subunits in testis and that RAD21L is coimmunoprecipitated with a fraction of SYCP1. Table S1 shows probes and primers used for quantitative real-time PCR.

We thank A. Matsuura for her contribution to production of antibodies. We are also grateful to the Support Units for Bio-material Analysis and for Animal Experiment, RIKEN BSI Resources Center, for help with sequence and real-time RT-PCR analyses and for production of rat polyclonal antibodies. We thank all the members of our laboratory for critically reading the manuscript.

This work was supported by Grant-in-Aids for Young Scientists (B) (20780203) to J.Lee and for Specially Promoted Research (20002010) to T. Hirano.