Human RAD9A and RAD9B share 36% identity, while HUS1 and HUS1B are 48% identical (Dufault et al., 2003; Hang et al., 2002). Inspection of genomic sequences revealed that Rad9b genes are present in the syntenic genomic region of all placental species analyzed, whereas Rad9a was likely lost in a few species, including wallaby, tree shrew, and sloth (Figure 1A and B). Phylogenetic analysis suggested that the duplication event generating Rad9a and Rad9b occurred prior to the evolution of bony fish ancestors (Danio rerio), whereas the single-exon Hus1b gene likely arose after a retrocopy duplication event later in evolution in mammals. Ortholog matrix and evolutionary tree analyses of placental mammals further showed that Rad1 is highly conserved, with no identifiable paralog.

Human and mouse gene expression data indicate that the 9-1-1 paralogs are highly expressed in testes but not other tissues, hinting at a potential role for RAD9B and HUS1B in spermatogenesis (Figure 1—figure supplement 1A). To further define the cell type-specific expression patterns of the 9-1-1 subunits within the testes, we mined single-cell RNA-sequencing data from wild-type adult mouse testis (Grive et al., 2019), comparing relative expression in spermatogonia, spermatocytes, and Sertoli cells (Figure 1C, Figure 1—figure supplement 1B–D). Rad9b expression was highest in spermatocytes as compared to spermatogonia and Sertoli cells. Rad1 expression also was highest in spermatocytes, whereas Hus1b expression was similar in spermatocytes and spermatogonia. On the other hand, Rad9a and Hus1 relative gene expression was highest in spermatogonia. As expected, expression of Atr and meiotic-silencing genes Hormad1 and Hormad2 was significantly higher in spermatocytes than spermatogonia or Sertoli cells, whereas Etaa1 expression was relatively low in spermatocytes, consistent with prior reports that it has limited roles in meiosis (ElInati et al., 2017). Spermatogonia also displayed relatively high levels of Atr, along with both Topbp1 and Etaa1. Analysis of expression data from human testis showed that expression of the 9-1-1 paralogs RAD9B and HUS1B was highest in spermatocytes as compared to other testis cell types (Human Protein Atlas version 21.0 and Guo et al., 2018). Together, these results suggest that the 9-1-1/TOPBP1/ATR and ETAA1/ATR signaling axes are expressed in pre-meiotic spermatogonia and highlight potential roles for alternative 9-1-1 complexes in spermatocytes.

To further analyze the evolutionary relationships between 9-1-1 subunits, we performed evolutionary rate covariation (ERC) analysis, which assesses correlations in gene evolutionary history and can reveal functionally significant relationships (Clark et al., 2012; Wolfe and Clark, 2015). ERC analysis was performed between all of the 9-1-1 subunits in a pairwise fashion across 33 mammalian species. Significant ERC values were identified between the RAD1, HUS1, and RAD9B subunits, supporting the notion that alternative 9-1-1 complexes assemble in germ cells (Figure 2A). These findings are consistent with reports that RAD9B physically interacts with RAD1, HUS1, and HUS1B (Dufault et al., 2003), and similarly that HUS1B interacts with RAD1 (Hang et al., 2002), suggesting that the paralogs contribute to alternative 9-1-1 complexes that include RAD9B-RAD1-HUS1 (9B-1-1) and RAD9B-RAD1-HUS1B (9B-1-1B) (Figure 2B).

To determine how disrupting the subunit shared by all of the 9-1-1 complexes impacted meiosis, we created a Rad1 CKO model by combining a conditional Rad1 allele (Wit et al., 2011) with Stra8-Cre, which drives CRE expression in spermatogonia (Sadate-Ngatchou et al., 2008). A similar approach was previously used to create Hus1 CKO mice (Lyndaker et al., 2013a), also on the inbred 129Sv/Ev background, enabling direct comparison of results between the two models. Experimental Rad1 CKO mice carried one Rad1^flox allele, one Rad1-null allele, and Stra8-Cre (Rad1^-/fl; Cre⁺). Mice that carried a wild-type Rad1 allele (Rad1^+/fl; Cre⁺) or lacked Stra8-Cre (Rad1^-/fl; Cre^- or Rad1^+/fl; Cre^-) were used as littermate controls. Both Rad1 CKO and control mice were born at expected frequency.

Immunoblotting of whole testis lysates from adult (12-week-old) Rad1 CKO mice confirmed significant reduction in RAD1 protein (Figure 2C). Reduced RAD1 expression was also observed in juvenile (postnatal day 14) Rad1 CKO testes (Figure 2—figure supplement 1). The residual RAD1 protein observed in Rad1 CKO mice may be attributed in part to somatic cells of the testis or pre-meiotic germ cells. However, additional results described below indicate that persistent RAD1 protein existed in some Rad1 CKO spermatocytes due to partial CRE recombinase efficacy or perdurance of RAD1 protein from pre-meiotic stages. Testes from Rad1 CKO males were one-third the size of control testes at 4 weeks of age, while body weight was not altered (Figure 2D). Hematoxylin and eosin (H&E) staining of testis sections from control and Rad1 CKO mice showed a reduction in tubule size and cellularity starting at 4 weeks in CKO mice, with the phenotype being much more severe in 12-week-old mice (Figure 2E). Similar to previous findings in Hus1 CKO males (Lyndaker et al., 2013a), histological analysis of Rad1 CKO mice revealed increased apoptosis of zygotene/pachytene-stage cells. In Rad1 CKO mice, round spermatids were infrequent but nevertheless observed in some histology sections from 4-week-old and 12-week-old mice, likely reflecting continued RAD1 expression in some meiocytes. Although severe germ cell loss in Rad1 CKO mice prevented precise staging of seminiferous tubules, 65.7% ± 2.7% of tubules in Rad1 CKO testes cross sections had fewer than 10 round spermatids, whereas no such tubules were identified in normal control testes (n = 3 mice per genotype; 50 tubules per mouse).

TUNEL staining confirmed significantly increased apoptosis in testes from Rad1 CKO mice starting at 4 weeks of age (Figure 2F and G). 4-week-old Rad1 CKO mice contained 2.4 ± 0.8 apoptotic nuclei per seminiferous tubule compared to 0.5 ± 0.4 in control mice. Apoptosis continued to be significantly elevated in 12-week-old Rad1 CKO mice (2.0 ± 0.7 positive nuclei per tubule) as compared to control mice (0.6 ± 0.4 positive nuclei per tubule) and was apparent in zygotene/pachytene-stage cells (Figure 2F, Figure 2—figure supplement 2A). To quantify the impact of Rad1 loss on germ cells, we stained testis sections for the germ cell-specific antigen TRA98 (Carmell et al., 2016). Tubules from control mice at 4 or 12 weeks of age contained an average of 220.2 ± 26.3 or 254.3 ± 45.5 TRA98-positive cells per tubule, respectively (Figure 2—figure supplement 2B and D). However, in the absence of RAD1, tubules contained only 74.9 ± 7.5 TRA98-positive cells in 4-week-old mice and 47.8 ± 8.3 in 12-week-old males.

Stra8-Cre expression occurs as cells are committing to undergo meiosis (Sadate-Ngatchou et al., 2008). We therefore anticipated that the apoptosis and germ cell loss observed in Rad1 CKO mice were due to meiotic defects. To address the possibility of pre-meiotic defects in Rad1 CKO mice, we assessed mice at 8 days postpartum (dpp), prior to meiotic entry. H&E staining, along with TUNEL and TRA98 staining of sections from both control and Rad1 CKO mice, showed no significant differences between genotypes at 8 dpp (Figure 2E–G, Figure 2—figure supplement 2B and D). To further confirm that RAD1 loss did not affect cells prior to meiotic entry, we stained sections for LIN28, a marker of spermatogonial stem cells (SSCs), which have not initiated meiosis (Aeckerle et al., 2012). As expected, no significant differences in LIN28 staining were observed between genotypes in testes from mice at 8 dpp or 4 weeks of age (Figure 2—figure supplement 2C and E), consistent with the notion that RAD1 targeting is specific to meiotic cells. However, 12-week-old Rad1 CKO mice had a significant decrease in LIN28-positive cells when compared to control mice. This later loss of LIN28-positive cells in Rad1 CKO mice can be attributed to large-scale germ cell loss, which could indirectly disrupt the environment required for proper SSC proliferation and survival.

Next, staining of surface spread spermatocyte nuclei was performed to test how localization of 9-1-1 subunits was affected by RAD1 loss. SYCP3, a component of the SC, was used to visualize the five substages of prophase I. Consistent with prior results (Freire et al., 1998; Lyndaker et al., 2013a), RAD1 localized in control meiocytes during leptonema as foci (209.3 ± 23.9 RAD1 foci), including on chromosome cores that were not yet synapsed, and during zygonema on both unsynapsed and synapsed chromosome cores (208.2 ± 9.2 RAD1 foci) (Figure 3A). In mid-pachynema, RAD1 was present on fully synapsed core axes of autosomes as well as along the X and Y chromosomes (120.8 ± 27.3 RAD1 foci). By late-pachynema, RAD1 was no longer present on autosomes but continued to be abundant along the X-Y cores. RAD1 localization was completely absent in 43% of spermatocytes from 12-week-old Rad1 CKO mice, whereas 100% of control cells showed proper RAD1 localization and abundance in zygotene- and pachytene-stage cells. The RAD1 localization observed in some Rad1 CKO meiocytes could be attributed to cells that failed to undergo CRE-mediated recombination or in which RAD1 levels were not yet fully depleted. Consistent with the latter possibility, pachytene-stage Rad1 CKO cells with detectable RAD1 focus formation had significantly fewer RAD1 foci than stage-matched control cells (106.0 ± 25.8 vs. 120.8 ± 27.3 RAD1 foci; p-value ≤ 0.0002; Figure 3—figure supplement 1A). The fact that Rad1 CKO cells were prone to apoptosis as described below would be expected to eliminate cells lacking RAD1, leaving RAD1-intact meiocytes enriched among the remaining cells. Additional functional analyses (see below) indicated that, overall, approximately 72% of Rad1 CKO cells had a meiotic defect (summarized in Figure 3—figure supplement 1B). We next evaluated how RAD1 disruption impacted RAD9A/B localization. In control samples, RAD9A and RAD9B localized to unsynapsed chromosomes as foci in leptotene-stage cells (145.5 ± 20.8 RAD9A foci; 211.5 ± 50.8 RAD9B foci) and to synapsed and unsynapsed chromosome cores in zygotene-stage cells (125.2 ± 25.6 RAD9A foci; 230.5 ± 40.5 RAD9B foci) (Figure 3B and C). By pachynema, RAD9A and RAD9B localized primarily as foci on autosomes and sex chromosome cores (107.3 ± 32.6 RAD9A foci; 125.5 ± 19.2 RAD9B foci). RAD9A and RAD9B foci were absent in 79 and 72% of Rad1 CKO meiotic spreads, respectively.

Rad1 CKO mice had no epididymal sperm (Table 1). To assess if Rad1 CKO mice were infertile, we bred control and Rad1 CKO mice with wild-type females. Control mice bred with wild-type females yielded 10 pregnancies and 66 viable pups, whereas Rad1 CKO mice had no viable offspring from 15 matings with wild-type females. Overall, these results indicate that RAD1 disruption severely compromised spermatogenesis and fertility. Moreover, the reduced testis weight and increased apoptosis in Rad1 CKO mice were more severe than those in mice with Hus1 or Rad9a loss (Lyndaker et al., 2013a; Vasileva et al., 2013), suggesting a broader role for RAD1 in meiocytes.

During meiosis, SC formation is critical for homologous chromosomes to pair and fully synapse (Zickler and Kleckner, 2015). Co-staining for the SC markers SYCP1 and SYCP3 revealed that 59.5% ± 4.3% of meiocytes from Rad1 CKO mice had whole chromosomes that remained unsynapsed and/or aberrant synapsis events involving multiple chromosomes, whereas 100% of meiocytes from control mice displayed normal homolog synapsis (Figure 3D, Figure 3—figure supplement 2A). RAD1 staining in meiocytes from 12-week-old Rad1 CKO mice revealed that all cells that lacked RAD1 displayed abnormal synapsis, with an average of only eight fully synapsed chromosomes per cell (Figure 3—figure supplement 2B). Cells with asynapsis that contained four or more synapsed homologous chromosomes were classified as pachytene-like cells. Unless otherwise noted, subsequent analyses described below focused on this population of RAD1-deficient meiocytes.

The γH2AX staining pattern was similar in Rad1 CKO and control spreads at leptonema and zygonema (Figure 3E). However, 97% of pachytene-like Rad1 CKO cells showed γH2AX present at asynaptic sites, with no clear presence of a sex body (n = 98 cells, three CKO mice). Interestingly, a subset of asynaptic regions in Rad1 CKO cells lacked detectable γH2AX staining (Figure 3E, white arrowheads; Figure 3—figure supplement 2C). Even in Rad1 CKO spermatocytes with apparently normal synapsis, 15.1% ± 11.5% of cells exhibited defects in γH2AX staining on the XY body, with partial or no coverage of γH2AX on the Y chromosome or expansion of the γH2AX domain to encompass an autosome (n = 205 cells, four CKO mice; Figure 3—figure supplement 2C). By contrast, such γH2AX staining defects were observed in only 3.2% ± 1.5% of pachytene-stage control cells. Together, these data suggest that RAD1 loss perturbed DNA damage signaling at asynaptic sites and the XY body.

Given that spermatocytes from Rad1 CKO mice exhibited significantly increased asynapsis, we next assessed meiotic progression in these cells by staining for the histone variant H1T and the recombination marker MLH1. First, we questioned whether RAD1-deficient cells were able to progress past mid-pachynema. Histone variant H1T is a marker of mid-pachynema and later stage wild-type spermatocytes (Cobb et al., 1999). Control cells demonstrate H1T staining as they progress into mid-pachynema (Figure 3—figure supplement 2D). However, H1T staining was absent in Rad1 CKO meiocytes with asynapsis, indicating that the cells failed to progress past mid-pachynema. By mid-pachynema, crossover sites are normally marked by MLH1 (Eaker et al., 2002). In contrast to the apparently normal MLH1 focus formation reported for Hus1 CKO cells lacking the canonical 9A-1-1 complex (Lyndaker et al., 2013a), MLH1 foci were not detected in any Rad1 CKO cells at the pachytene-like stage (Figure 3—figure supplement 2E), further suggesting that Rad1 CKO meiocytes with asynapsis fail to progress beyond early/mid-pachynema. Together, the observations of γH2AX abnormalities and SC defects in Rad1 CKO cells indicate important roles for 9-1-1 complexes in ensuring homologous chromosome synapsis and appropriate DDR signaling in response to asynapsis.

Localization of canonical 9A-1-1 subunit RAD9A to chromatin cores requires SPO11-induced DSBs (Lyndaker et al., 2013a), and testis-specific Hus1 or Rad9a CKO results in persistent meiotic DSBs with delayed repair kinetics (Lyndaker et al., 2013a; Vasileva et al., 2013). We therefore investigated how RAD1 loss impacts DSB repair. Following MRE11-RAD50-NBS1 (MRN)-mediated resection of SPO11-induced meiotic DSB, meiosis-specific with OB domains (MEIOB) and RPA localize to the ssDNA overhangs prior to RAD51 and DMC1 loading (Hinch et al., 2020; Luo et al., 2013; Shi et al., 2019). In control spermatocytes, RPA and MEIOB foci are abundant in early prophase I and diminish as DSBs are repaired (Figure 4A–D). RPA and RAD1 both formed foci on meiotic chromosome cores, but the extent of co-localization was modest, consistent with the notion that RPA coats single-stranded DNA, whereas 9-1-1 is loaded at recessed DNA ends (Figure 4—figure supplement 1A). Rad1 CKO testes had on average 50 fewer RPA foci than controls in leptotene-stage cells (194.4 ± 54.4 control; 145.8 ± 37.4 CKO; Figure 4A and B). Intriguingly, RPA foci in Rad1 CKO cells appeared larger than those in control cells (Figure 4A, Figure 4—figure supplement 1B). In the absence of RAD1, MEIOB focus formation on chromatin cores in leptotene-stage cells was also significantly decreased as compared to control cells (230.4 ± 45.4 control; 125.8 ± 36.6 CKO) (Figure 4C and D). In control spermatocytes, MEIOB and RPA levels on meiotic chromosome cores decreased as the cells progressed into pachynema (115.1 ± 26.6 control MEIOB; 52.3 ± 40.5 control RPA), whereas Rad1 CKO cells showed persistence of MEIOB and RPA staining (132.1 ± 38.0 CKO MEIOB; 100.0 ± 44.3 CKO RPA).

During prophase I in wild-type spermatocytes, RAD51 and DMC1 displace MEIOB and RPA from the ssDNA overhangs and drive the subsequent steps of homology search and strand invasion (Gray and Cohen, 2016; Hinch et al., 2020). The persistence of MEIOB and RPA foci in Rad1 CKO spermatocytes suggested that RAD1 loss might perturb RAD51 loading. On average, leptotene-stage cells from control mice contained 195.2 ± 29.0 RAD51 foci, whereas Rad1 CKO cells at the same stage had 115.7 ± 41.1 RAD51 foci (Figure 4E and F). RAD51 foci continued to be significantly lower in zygotene-stage Rad1 CKO meiocytes, which contained 77.1 ± 30.0 RAD51 foci per cells as compared to 137.5 ± 36.1 in controls. In control samples, RAD51 foci levels decreased as cells progressed from zygonema to pachynema, reflecting the successful repair of DSBs. However, Rad1 CKO spermatocytes retained relatively high levels of RAD51 foci in pachytene-like-stage cells (60.4 ± 29.5 RAD51 foci) as compared to control pachytene-stage meiocytes (11.5 ± 4.9 RAD51 foci) (Figure 4F). These results for RAD51 localization in Rad1 CKO spermatocytes differed from those in Hus1 CKO mice, where RAD51 appeared normal in early prophase and then was aberrantly retained at a small number of sites in pachytene-stage cells (Lyndaker et al., 2013a). Together, these results suggest that the 9-1-1 complexes are critical for DSB processing and repair during mammalian meiosis and that absence of RAD1, or to a lesser extent HUS1, leaves persistent unrepaired DSBs.

The abnormal localization profiles for MEIOB, RPA, and RAD51 observed in Rad1-deficient spermatocytes raised the possibility that DSB formation was impaired. To determine whether the defects were related to DSB formation or the subsequent repair steps, we treated Rad1 CKO and control mice with 5 Gy ionizing radiation (IR), harvested testes 1 hr post treatment, and quantified RPA and RAD51 focus formation in leptotene- and zygotene-stage cells. Since exogenously induced DSBs are repaired via meiotic processes in early stages of prophase I (Enguita-Marruedo et al., 2019), this approach allowed us to test whether the alterations in DSB markers in Rad1 CKO cells were due to reduced DSB formation or a DSB repair defect. Upon DSB induction via irradiation, control mice showed increased RPA and RAD51 focus formation at early prophase I stages (Figure 4G and H, Figure 4—figure supplement 1C and D). By contrast, irradiation did not induce increased focus formation by RPA or RAD51 in Rad1 CKO spermatocytes. Together, these results highlight the importance of the 9-1-1 complexes for meiotic DSB repair.

Given that the canonical 9A-1-1 complex plays a central role in stimulating ATR kinase activity in somatic cells, we sought to determine the effect of RAD1 loss on the localization of ATR and its substrates in meiocytes. ATR localizes to unsynapsed regions at early stages of prophase I, and by pachynema it is sequestered mainly at the XY body where it initiates MSCI (Abe et al., 2020; Turner, 2015). Cells from Rad1 CKO mice with synapsis defects showed ATR localization only at a subset of asynaptic regions (Figure 5A).

TOPBP1 is required for ATR activation following replication stress (Blackford and Jackson, 2017) and interacts with ATR during meiosis to ensure that meiotic silencing is properly initiated (ElInati et al., 2017; Jeon et al., 2019). In control meiocytes, TOPBP1 was observed as discrete foci on unsynapsed chromosome cores throughout leptonema and zygonema (Figure 5B). At pachynema, TOPBP1 was found exclusively along the unsynapsed regions of the X and Y and present as a faint cloud on XY chromosome loops. By contrast, in pachytene-like stage Rad1 CKO cells, TOPBP1 localized to only a subset of asynaptic sites, failing to coat the entirety of unsynapsed chromosome cores, similar to the pattern observed for ATR. These findings suggest a role for the 9-1-1 complexes in promoting ATR and TOPBP1 localization to unsynapsed chromatin, although this is, at least in part, likely an indirect effect of the extensive asynapsis in Rad1 CKO cells as the available pool of silencing factors can be insufficient to localize to all asynaptic sites under such circumstances (Mahadevaiah et al., 2008).

The best characterized ATR substrate in somatic cells is the transducer kinase CHK1. CHK1 has been proposed to play a role in meiotic DSB repair and is suggested to aid progression through prophase I by removal of DDR proteins such as γH2AX from autosomes (Abe et al., 2018; Fedoriw et al., 2015; Pacheco et al., 2018). In wild-type cells, CHK1 phosphorylation (S317) occurs during leptonema and zygonema at unsynapsed chromosomes. During pachynema, pCHK1 (S317) is apparent on XY cores and as a cloud over the sex body (Figure 5C). Interestingly, in the Rad1 CKO mutant, pCHK1 was absent at all stages of prophase I. Reduced CHK1 phosphorylation at ATR target sites S317 and S345 in the absence of RAD1 was confirmed by whole testis immunoblotting (Figure 5—figure supplement 1A). By contrast, meiotic spreads from Hus1 CKO mice showed normal patterns of CHK1 (S317) phosphorylation (Figure 5D). That meiotic CHK1 phosphorylation is normal in the absence of HUS1 but disrupted by RAD1 loss suggests that alternative 9-1-1 complexes play an important role in activating the transducer kinase CHK1 during meiotic prophase I.

The cohesin subunit SMC3 has been implicated as another likely meiotic ATR substrate (Fukuda et al., 2012). Loss of meiosis-specific cohesins results in phenotypes similar to those in Rad1 CKO mice, including SC assembly defects, impaired synapsis, and DSB repair failure and synapsis defects (Challa et al., 2019; Eijpe et al., 2003; Ishiguro, 2019; Llano et al., 2012; Ward et al., 2016). Notably, phosphoproteomic analyses revealed that phosphorylation of cohesin complex (SMC3) and SC (SYCP1, SYCP2, and SYCP3) components during meiosis is RAD1- and ATR-dependent (Sims et al., 2021). Furthermore, correlated evolutionary relationships, as measured by ERC analysis, were observed between genes encoding 9-1-1 subunits and those encoding cohesin and SC proteins, including SMC1β, RAD21L1, SYCP2, and SYCE1 (Figure 5—figure supplement 1B, C). We therefore further interrogated the relationship between RAD1 and SMC3. RAD1 and pSMC3 (1083) co-localized at the XY in pachynema-stage wild-type spermatocytes (Figure 5E). In control meiocytes, SMC3 was observed on chromatin cores throughout prophase I and was phosphorylated specifically at unsynapsed chromatin cores during leptonema and zygonema, and at the unsynapsed regions of the XY in mid-pachynema (Figure 5F). Although total SMC3 loading was unaffected by RAD1 loss, Rad1 CKO spermatocytes showed reduced accumulation of phosphorylated SMC3 (pSMC3 S1083) at unsynapsed chromatin regions in pachytene-like cells as compared to mid-pachytene-stage control cells. Western blot analysis of whole testis lysates confirmed that SMC3 phosphorylation (pSMC3 S1083) was significantly reduced in testes from Rad1 CKO mice (Figure 5—figure supplement 1D and E). Unlike Rad1 CKO spermatocytes, Hus1 CKO cells had grossly normal pSMC3 (S1083) localization to the XY in pachytene-stage spermatocytes (Figure 5F). To determine if SMC3 (1083) phosphorylation was ATR dependent, we performed immunostaining of spermatocytes from wild-type C57BL/6 male mice treated with the ATR inhibitor AZ20 (ATRi). Similar to the effects of RAD1 loss, acute ATRi treatment caused a decrease in pSMC3 (S1083) at X and Y chromatin loops and cores despite the fact that SMC3 localization to chromosome cores appeared normal, suggesting a specific defect in SMC3 phosphorylation. Together, these results suggest that 9-1-1 complexes and ATR act in conjunction to regulate meiotic cohesin phosphorylation.

Given that ATR is a primary regulator of MSCI (Fedoriw et al., 2015; Pacheco et al., 2018; Turner et al., 2005; Widger et al., 2018), the defects in ATR signaling noted above prompted us to assess meiotic silencing in Rad1 CKO mice. We first looked upstream of ATR and examined whether the absence of RAD1 impacted localization of HORMA-domain proteins 1 and 2 (HORMAD1 and HORMAD2). The presence of HORMADs at unsynapsed chromatin is important for meiotic silencing, and HORMAD1 is required for ATR recruitment to unsynapsed sites (Fukuda et al., 2012; Kogo et al., 2012; Shin et al., 2010; Wojtasz et al., 2012). In control cells, HORMAD1 and HORMAD2 were observed during early prophase I at chromosomal regions that were not yet synapsed (Figure 6A and B). By mid-pachynema, the HORMADs localized strictly at the unsynapsed regions of the XY, similar to the localization of ATR. Notably, RAD1 loss did not alter HORMAD1 or HORMAD2 localization to unsynapsed regions. Furthermore, the HORMADs were observed to entirely coat unsynapsed chromosome regions in Rad1 CKO cells, in contrast to the failure of ATR and TOPBP1 to localize to all unsynapsed sites.

ATR phosphorylates HORMAD1 (S375) and HORMAD2 (S271) at asynaptic regions (Fukuda et al., 2012; Royo et al., 2013). In control cells, HORMAD2 (S271) phosphorylation was observed on the X and Y chromosome cores in mid-pachytene-stage cells as expected (Figure 6C). However, in pachytene-like Rad1 CKO cells, phosphorylated HORMAD2 was detected at only a subset of unsynapsed regions. That HORMAD2 localized properly in the absence of RAD1 but lacked phosphorylation at an ATR-regulated site further supports the notion that meiotic ATR signaling requires the 9-1-1 complexes. BRCA1 is another key meiotic silencing factor, and its localization is interdependent with that of ATR in a stage-specific manner, with conditional Atr deletion disrupting BRCA1 localization to XY axial elements (Mahadevaiah et al., 2008; Royo et al., 2013). BRCA1 failed to localize properly in Rad1 CKO cells with extensive asynapsis, coating only a subset of asynaptic regions much like what was observed for ATR and TOPBP1 in RAD1-deficient cells (Figure 6D).

The defects in ATR signaling observed in Rad1 CKO mice suggested that disruption of 9-1-1 complexes might impair meiotic silencing. To test this possibility, we first examined the localization of RNA Pol II, the exclusion of which from the XY body is an indicator of MSCI (Figure 6E). In control cells, RNA Pol II was excluded from the sex body in the vast majority of cells, with only 8.8% ± 2.4% of cells showing any RNA Pol II at the XY chromosomes. In Rad1 CKO meiocytes with extensive asynapsis, RNA Pol II was diffusively distributed, but the sex chromosomes cannot be distinguished in such cases. We next assessed RNA Pol II in Rad1 CKO cells with apparently normal synapsis and observed that 27.8% ± 19.1% of such cells failed to fully exclude RNA Pol II from the XY body. This contrasted with our previous analysis of Hus1 CKO spermatocytes with disruption of the canonical 9A-1-1 complex, in which RNA Pol II was properly excluded from the sex body (Lyndaker et al., 2013a). Finally, we directly evaluated meiotic silencing via RNA fluorescent in situ hybridization (FISH) for the X-chromosome gene Scml2 that should be silenced in early pachynema-stage cells (Royo et al., 2010). Because autosome asynapsis antagonizes MSCI (Mahadevaiah et al., 2008), the analysis of Scml2 expression focused on cells with normal homolog synapsis and excluded those with asynapsis. Inappropriate Scml2 expression was detected in 7.1% ± 0.6% of early pachytene-stage control cells, but 28.9% ± 3.2% of Rad1 CKO cells (p<0.0001; Figure 6F), indicating that meiotic silencing was disrupted by RAD1 loss. This quantification underestimates the extent of the silencing defect upon RAD1 loss since some cells with apparently normally synapsis in Rad1 CKO mice retain normal RAD1 expression. Nevertheless, approximately 30% of Rad1 CKO cells with apparently normal synapsis show both defective exclusion of RNA Pol II from the XY body as well as inappropriate expression of an X-linked gene. Together with the clear evidence for defective ATR signaling upon RAD1 loss, these results demonstrate the importance of the canonical and alternative 9-1-1 complexes in promoting ATR-mediated MSCI.

Rad1 CKO and control mice in the 129Sv/Ev background were generated by crossing Rad1^flox/flox mice with Rad1^+/+, Stra8-Cre⁺ mice to generate Rad1^+/fl, Stra8-Cre⁺ (Rad1^+/-, Stra8-Cre⁺) mice. Stra8-Cre mice containing one null Rad1 allele (Rad1^+/-, Stra8-Cre⁺) were crossed with Rad1^flox/flox mice to generate experimental germ-cell specific Rad1 CKO mice (Rad1^-/fl, Stra8-Cre⁺) and control mice (Rad1^+/fl, Stra8-Cre⁺; Rad1^+/fl, Stra8-Cre^-; Rad1^-/fl, Stra8-Cre^-). Rad1 flox mice feature a conditional Rad1 allele containing a K185R mutation that does not affect RAD1 function (Wit et al., 2011). Hus1 CKO mice were used as previously reported (Lyndaker et al., 2013a). All mice used for this study were handled following federal and institutional guidelines under a protocol approved by the Institutional Animal Care and Use Committee (IACUC) at Cornell University. The Key resources table lists the genotyping primers used for mice in this study.

For fertility testing, 8- to 12-week-old Rad1^-/fl, Stra8-Cre+ and control males were singly housed with wild-type FVB females, where copulatory plugs were monitored daily. Once a plugged female was detected, the female was removed to a separate cage and monitored for pregnancy. Viable pups were counted on the first day of life.

Both caudal epididymides from 12-week-old mice were minced with fine forceps in 37°C in a Petri dish containing 1× phosphate buffered saline (PBS) and fixed in 10% neutral-buffered formalin (1:25 dilution). Sperm were counted using a hemacytometer and analyzed statistically using a Student’s t-test between control and Rad1 CKO mice.

For irradiation, control and Rad1 CKO mice were placed in a ¹³⁷Cesium-sealed source irradiator (J.L. Shepherd and Associates) with a rotating turntable and irradiated with 5 Gy IR. Testes were harvested for meiotic spreads 1 hr post radiation. For in vivo ATR inhibition, wild-type B6 mice were treated via oral gavage with 50 mg/kg AZ20 (Selleck Chemicals, S7050) reconstituted in 10% DMSO (Sigma), 40% propylene glycol (Sigma), and 50% water, and collected 4 hr later.

Whole testis lysates from Rad1 CKO and control mice were prepared in RIPA buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl) supplemented with aprotinin, leupeptin, sodium orthovanadate, and phenyl-methylsulfonyl fluoride. Cell lysates were resolved by SDS-PAGE and immunoblotted using standard procedures. Bands were visualized on a VersaDoc MP 5000 Model (Bio-Rad) using a 1:1 ratio of WesternBright ECL Luminol/enhancer solution to WesternBright Peroxide Chemiluminescent peroxide solution (Advansta). Antibody information is provided in the Key resources table.

Testes were harvested from mice aged to 8 dpp, 4 weeks or 12 weeks of age. Testes were then fixed overnight in either Bouin’s (RICCS Chemical) for H&E staining or 10% neutral-buffered formalin (Fisher) for LIN28 (RRID:AB_1310410), TRA98 (RRID:AB_1056334), and TUNEL staining. Fixed testes were embedded in paraffin wax and sectioned at 5 µm. Immunofluorescence staining was used to detect LIN28 using rabbit polyclonal anti-LIN28 antibody (RRID:AB_1310410). Immunohistochemistry staining was used to detect TRA98 using rat monoclonal anti-TRA98 antibody (RRID:AB_1056334). TUNEL assay was performed using the Apoptag kit (EMD Millipore) as per the manufacturer’s instructions. LIN28, TRA98, and TUNEL data were quantified in ImageJ by counting the number of positive cells per tubule for 50 tubules of each genotype for each age group. Differences between controls and Rad1 CKOs were analyzed using Welch’s unpaired t-test in GraphPad (RRID:SCR_002798). Staging of spermatocytes in stained sections was performed as described by others (Ahmed and de Rooij, 2009; Meistrich and Hess, 2013).

Meiotic spreads were prepared from 8- to 12-week-old mice as previously described (Kolas et al., 2005). Briefly, tubules from mice were incubated on ice in hypotonic extraction buffer for 1 hr. Tubules were then minced into single-cell suspension in 100mM sucrose, and cells were spread on slides coated with 1% PFA with 0.15% TritionX-100 and incubated in a humidifying chamber for 4 hr or overnight. For immunostaining, slides were blocked using 10% goat serum and 3% BSA, followed by incubation overnight with primary antibody (listed in the Key resources table) at room temperature in a humidifying chamber. Secondary antibodies were incubated at 37°C for 2 hr in the dark, and slides were then coverslipped using anti-fade mounting medium (2.3% DABCO, 20 mM Tris pH 8.0, 8 µg DAPI in 90% glycerol). Meiotic chromosomal spreads were imaged with an AxioCam MRM using a Zeiss Imager Z1 microscope (Carl Zeiss, Inc) and processed with ZEN software (version 2.0.0.0; Carl Zeiss, Inc). Quantification of meiotic spreads was performed using Fiji for ImageJ. Statistical analysis was performed using Welch’s unpaired t-test using GraphPad Prism9 (RRID:SCR_002798).

RNA-FISH was carried out with digoxigenin-labeled probe using BAC DNA, Scml2: RP24-204O18 (CHORI), and immunofluorescence using rabbit HORMAD2 antibody (gift from A. Toth) as previously described (Mahadevaiah et al., 2009). Images of RNA-FISH with immunofluorescence were captured using a Deltavision Microscopy System with a ×100/1.35 NA Olympus UPlanApo oil immersion objective.

Human 9-1-1 subunit sequences were used to obtain their respective orthologs from Ensemble 101 (2020) and/or NCBI Gene from 33 representative mammalian species. Orthologs found in Ensemble having a ≥ 50% of both target and query sequence identity and a pairwise whole-genome alignment score of ≥50 were considered to have high confidence. Orthologs that did not meet those criteria were considered to have low confidence. Sequences only found in the NCBI Gene database were considered as high confidence if they were found to be syntenic. Synteny was determined based on whether the gene had at least one shared neighbor gene upstream or downstream that also was conserved. Species divergence across time was obtained from TimeTree website (http://timetree.org/).

Protein sequences of 9-1-1 orthologs were obtained using NCBI HomoloGene. Multiple alignment of protein sequences was done using Clustal Omega (1.2.2) implemented in Geneious Prime (2020.0.5). A substitution model was tested using ProtTest (v. 3.4.2). The selected substitution model with specific improvements was JTT + I + G + F (Jones–Taylor–Thornton;+ I: invariable sites; + G: rate heterogeneity among sites; + F: observed amino acid frequencies). Improvements were included to take account of any evolutionary limitations due to conservation of protein structure and function. A nonrooted phylogenetic tree was made using Maximum Likelihood interference (four gamma distributed rate) (Nguyen et al., 2015) and implemented with iTOL (itol.embl.de) (Letunic and Bork, 2019). Branch distance represents substitution rate, and branch support was performed with 1000 ultrafast bootstrap replicates. Nodes below 70% branch support were collapsed.

ERC calculations were completed using the ERC web tool at https://csb.pitt.edu/erc_analysis/ (Wolfe and Clark, 2015). Group analysis was performed to examine ERC values between all gene pairs indicated in Figure 2A and Figure 5—figure supplement 1B and C using UCSC gene sequences from 33 mammalian species as described in Priedigkeit et al., 2015. For Figure 5—figure supplement 2A and B, the protein set list for Gene Ontology subontology Meiosis I (GO:0007127) was obtained from AmiGO 2 (v2.5.13). ERC values were calculated against each of the 9-1-1 subunits using the ERC analysis website. Using R (v4.0.3), ERC values were depicted as a heatmap and a network plotted using the packages pheatmap (v1.0.12) and qgraph (v1.6.5), respectively. A cutoff of ERC value of 0.4 was used to determine significant comparisons. The Fruchterman and Reingold algorithm was used to generate a forced-directed layout to help determine clusters of highly connected nodes, and after 500 iterations the distance between nodes shows absolute edge weight (ERC values) between nodes.