The ability of men to remain fertile throughout their lives depends upon establishment of a spermatogonial stem cell (SSC) pool from gonocyte progenitors, and thereafter balancing SSC renewal versus terminal differentiation. Here, we report that precise regulation of the cell cycle is crucial for this balance. Whereas cyclin-dependent kinase 2 (Cdk2) is not necessary for mouse viability or gametogenesis stages prior to meiotic prophase I, mice bearing a deregulated allele (Cdk2Y15S) are severely deficient in spermatogonial differentiation. This allele disrupts an inhibitory phosphorylation site (Tyr15) for the kinase WEE1. Remarkably, Cdk2Y15S/Y15S mice possess abnormal clusters of mitotically active SSC-like cells, but these are eventually removed by apoptosis after failing to differentiate properly. Analyses of lineage markers, germ cell proliferation over time, and single cell RNA-seq data revealed delayed and defective differentiation of gonocytes into SSCs. Biochemical and genetic data demonstrated that Cdk2Y15S is a gain-of-function allele causing elevated kinase activity, which underlies these differentiation defects. Our results demonstrate that precise regulation of CDK2 kinase activity in male germ cell development is crucial for the gonocyte-to-spermatogonia transition and long-term spermatogenic homeostasis.
Human and mouse males are capable of reproduction throughout much of their lives owing to a continuously regenerating pool of spermatogonial stem cells (SSCs) in adult testes. In mice, the progenitors of SSCs, called primordial germ cells (PGCs), arise as a group of ∼45 cells in the epiblast of 6- to 6.5-day-old embryos (Ginsburg et al., 1990). The PGCs migrate to the genital ridges by ∼embryonic day (E) 10.5, and proliferate rapidly as the gonads differentiate into either primitive ovaries or testes. These germ cells, now called gonocytes (also called ‘prospermatogonia’ in males), reach a population of ∼25,000 by E13.5. Then, female gonocytes (‘oocytes’) directly enter meiosis, whereas the male gonocytes largely cease proliferation for the remainder of gestation (Sasaki and Matsui, 2008; Tam and Snow, 1981).
About 1 day after birth at postnatal day (P) 1.5, the gonocytes resume proliferation (Nagano et al., 2000) and complete a process known as the gonocyte-to-spermatogonia transition (GST). The GST is not precisely delineated, as there is no clear-cut distinction between gonocytes/prospermatogonia and spermatogonia, although marker analysis and single cell sequencing indicate that this transition begins in late gestation (Law et al., 2019; Pui and Saga, 2017). During the GST, some gonocytes establish the permanent pool of SSCs that will seed waves of spermatogenesis throughout adult life, whereas others initiate a distinct prepubertal round of spermatogenesis involving several mitotic spermatogonial divisions, meiosis, and postmeiotic development (Bellve et al., 1977). This first round of spermatogonial differentiation is unique because it originates from neurogenin 3-expressing gonocytes rather than SSC populations (Yoshida et al., 2004, 2006).
During adult life, new coordinated waves of spermatogenesis derive from divisions of an individual Asingle (As) type SSC and its progeny to form Apaired (Apr), and then Aaligned (Aal) type SSCs. Aal SSCs divide without cytokinesis to form clonal chains (Aal4 to Aal32) linked via their cytoplasm. Further divisions result in A2, A3, A4, intermediate (In) and B type spermatogonia before ultimately differentiating into preleptotene spermatocytes that enter meiosis. The ‘stemness’ potential of these cells decreases as spermatogonia chain length increases (reviewed by de Rooij, 2017). The balance between SSC renewal and SSC differentiation is crucial, and failure to maintain this balance can cause either insufficient sperm production or exhaustion of the stem cell pool. Genetic or environmental events that lead to SSC exhaustion, or which compromise the generation or viability of SSCs or their progenitors, can cause Sertoli cell only syndrome (SCOS), a histological phenotype categorizing a subset of patients with non-obstructive azoospermia (NOA).
Normal proliferation of cells is dependent on cell cycle regulation. Key players in this process are cyclin-dependent kinases (CDKs) and their activating partner proteins, cyclins, several of which exist in mammals. CDK activity is controlled during the cell cycle in part by the association of CDKs with cyclins. In their activated state, these complexes propel cells through successive stages of the cell cycle, including entry into and through the S and M phases (Satyanarayana and Kaldis, 2009). The transition between active and inactive states is governed by both interaction with CDK-inhibitory proteins (Lim and Kaldis, 2013) and also phosphorylation or dephosphorylation events at key regulatory sites on CDKs (Cuijpers and Vertegaal, 2018; Morgan, 1995).
Though the activities of cyclins and CDKs have been studied predominantly in cultured somatic cells and single-celled eukaryotes, their roles in the germline have also been investigated (Martinerie et al., 2014; Wolgemuth and Roberts, 2010). Despite its broad expression in many cell types, Cdk2 is not essential for mouse viability, yet its disruption causes male and female infertility (Berthet et al., 2003; Ortega et al., 2003). Specifically, Cdk2−/− meiocytes arrest during the pachytene stage of meiotic prophase I. This arrest is triggered by defective attachment of telomeres to the nuclear envelope, resulting in failed or incomplete synapsis of homologous chromosomes. In turn, these defects prevent homologous recombination repair of meiotic double-strand breaks (Viera et al., 2009, 2015). CDK2 is expressed in spermatogonia (Johnston et al., 2008; Ravnik and Wolgemuth, 1999), but SSCs apparently remain functional because mutant males produce spermatocytes (albeit destined for meiotic arrest) into adulthood. These results suggest that, as in most somatic cells, CDK2 function is not essential in spermatogonia, but it may provide redundant function in those cells and non-canonical function(s) in meiocytes related to recombination in the latter (Berthet et al., 2003; Krasinska et al., 2008). Although a spermatogonia-specific deletion of Cdk1 has yet to be described, this kinase is required for metaphase I entry at the end of the first meiotic prophase (Clement et al., 2015). CDK1 likely acts in concert with the meiosis-specific cyclin A1, which is also required at the same stage (Liu et al., 1998). In contrast, conditional ablation of cyclin B1 (Ccnb1), a CDK1 binding partner, blocks proliferation of gonocytes and spermatogonia, but does not impact meiosis (Tang et al., 2017).
To identify possible infertility alleles in human populations, we modeled a missense variant (SNP rs3087335) altering the Tyr15 phosphorylation site of CDK2 in mice (Singh and Schimenti, 2015). Surprisingly, homozygotes for this allele (Cdk2Y15S) caused an SCOS-like phenotype. Additionally, Cdk2Y15S heterozygotes exhibited age-dependent testis histopathology and reduced sperm count, indicating that Cdk2Y15S is a gain-of-function, semidominant, allele (Singh and Schimenti, 2015). In vitro studies have shown that Tyr15 phosphorylation, typically catalyzed by the WEE1 kinase, negatively regulates CDK activity and, thus, cell cycle progression (Gu et al., 1992; Welburn et al., 2007). We speculated that the Cdk2Y15S allele was hyperactive by virtue of being refractory to negative regulation by WEE1 (Hughes et al., 2013; Zhao et al., 2012), thus driving excessive spermatogonial proliferation and/or differentiation over SSC regeneration and maintenance.
Here, we report that the apparent SCOS phenotype in Cdk2Y15S/Y15S testes is not due to an absence of germ cells; rather, SSC-like cells are present and can divide, but their progeny fail to differentiate and subsequently are lost before entering meiosis. The germ cell defects are first detectable at P3; GST appears delayed or disrupted as determined by analyses of key markers and single cell (sc)RNA-seq data. We provide evidence that CDK2Y15S-expressing cells display altered kinase activity, and that this defect underlies the phenotypes observed in such cells. This study highlights the importance of precise regulation of CDK kinase activity in establishing and maintaining testis homeostasis.
Ablation of the Tyr15 inhibitory phosphorylation site in CDK2 disrupts gonocyte and spermatogonia differentiation
As summarized above, adult Cdk2Y15S/Y15S testes lacked evidence of spermatogenesis and were essentially devoid of cells positive for DDX4 (hereafter MVH, mouse vasa homolog), which is strongly expressed in gonocytes and all juvenile germ cells (Toyooka et al., 2000). Our working hypothesis was that most gonocytes differentiated in the initial spermatogenic wave, leaving the adults devoid of a renewable SSC pool. To test this hypothesis, we first quantified gonocytes in neonatal testes. The number of MVH+ cells in P0 Cdk2Y15S/Y15S testes was no different than in control littermates (Fig. 1A,B), indicating that the loss of germ cells occurred not during gestation (for example, during PGC expansion), but during postnatal development. Next, to test the prediction that all SSCs would be exhausted by adulthood, we performed immunohistochemical (IHC) analysis of mutant adult (P180) seminiferous tubule sections, which lack ongoing spermatogenesis. Remarkably, Cdk2Y15S/Y15S tubules contained ample numbers of cells positive for LIN28, which is expressed in a subset of Type As spermatogonia, and essentially all Type Apr through Aal spermatogonia (Chakraborty et al., 2014b) (Fig. 1C; Fig. S1B,C), demonstrating that although mutant testes had an SCOS-like appearance, there were indeed undifferentiated A-type spermatogonia present. However, they apparently were not proliferating or differentiating in a normal manner.
We next characterized maturation of germ cells using markers of gonocytes, SSCs, and progressively more differentiated spermatogonia. The transcription factor FOXO1 (forkhead box O1), which in testis is expressed only in gonocytes and undifferentiated spermatogonia (As-Aal) but not more differentiated germ cells, transits from the cytoplasm to the nucleus as gonocytes differentiate into spermatogonia postnatally (Goertz et al., 2011). At P0, mutant gonads retained cytoplasmic localization of FOXO1 (Fig. S1A), consistent with normal prenatal development of gonocytes. However, translocation of FOXO1 to the nucleus, which normally begins at P3 and is complete by P21 (Goertz et al., 2011), was delayed in the mutant. Unlike wild type (WT), in which FOXO1 was localized in the nuclei of all positive cells at P30, >40% of Cdk2Y15S/Y15S cells exhibited cytoplasmic FOXO1 at this time; this fraction declined further to ∼10% at P90 (Fig. 2; Fig. S1B). These results suggest that Cdk2Y15S/Y15S germ cells are delayed in the GST and, with the exception of the unique first round of spermatogenesis, are unable to differentiate properly even after making this transition.
Studies of other markers by testis IHC confirmed abnormalities in the Cdk2Y15S/Y15S germ cell pool. At P3 and P15, there were fewer cells positive for PLZF (zinc finger and BTB domain containing 16; formally ZBTB16) and FOXO1 (Fig. S1B,C), which are expressed in all undifferentiated spermatogonia (Goertz et al., 2011). The number of cells positive for LIN28, which is expressed in both undifferentiated (As-Aal) and differentiated (A1-A4) spermatogonia (Chakraborty et al., 2014a; Gaytan et al., 2013), was also lower in mutants at P3 (Fig. S1C). The LIN28+ cell shortfall disappeared at P15, possibly reflecting large numbers of differentiated spermatogonia resulting from the first wave of spermatogenesis, but in adulthood (P90) when the absence of ongoing spermatogenesis in mutants was manifested, LIN28+ cells were again lower in the mutant, as was PLZF (Fig. S1B,C).
We next performed a series of studies on seminiferous tubule whole mounts to determine if there were disruptions to the normal patterns of Type A spermatogonia subtypes (e.g. As versus Apr versus Aal). A marker for SSCs is GFRA1 [glial cell line-derived neurotrophic factor family receptor alpha 1, the cell surface receptor for glial cell line-derived neurotrophic factor (GDNF)], which mainly labels As (a subset thereof) and Apr spermatogonia. We found that almost all As and Apr PLZF+ SSCs from P90 Cdk2Y15S/Y15S mice also expressed GFRA1, similar to age-matched WT controls. Mutants also contained clusters of four to eight PLZF+LIN28+ cells at P90, although these clusters did not have the typical appearance of Aal chains (possibly due to disrupted overall tubule architecture; henceforth we will refer groups of more than two cells as ‘clusters’). Furthermore, mutants lacked the longer PLZF+LIN28+ chains (Aal16-32) that are typical in adult WT testes (Fig. 3A) (Buaas et al., 2004; Zheng et al., 2009). This result is consistent with the lower overall number of PLZF+ and LIN28+ spermatogonia in testes of young (P5) and old (P90) Cdk2Y15S homozygotes (Fig. S1C).
To explore the basis for the defects in spermatogonial distributions, we examined GFRA1+ progenitor spermatogonia in adolescent (P15, during the first round of spermatogenesis) versus adult (P90) tubules, which lack spermatogonial differentiation. Cdk2Y15S/Y15S mutants had about twice as many GFRA1+ cells at both ages compared with WT (Fig. 3A-C). Strikingly, mutant tubules contained large clusters of GFRA1+ spermatogonia (referred to as ‘GFRA1+ Acluster>6’) in P15 tubules, which were virtually absent in WT (Fig. 3A,D,E). Hypothesizing that these GFRA1+ Acluster>6 cells might be in an abnormal, delayed state of differentiation, we examined them for co-expression of LIN28. Interestingly, many of the P15 GFRA1+ cells in clusters were also LIN28+ (Fig. S2). Additionally, whereas mutants had ∼2-fold more As GFRA1+ cells than WT, most were also LIN28+ as in WT (Fig. 3F). The clusters of GFRA1+ cells disappeared by P90, yet mutant tubules had more GFRA1+ As and Apr cells compared with WT at this age (Fig. 3B-D). The combined data indicate that the Cdk2Y15S allele not only causes a delay in the GST, but also impacts the differentiation of progenitor SSCs beyond the Apr developmental state when GFRA1 should be downregulated in chains of LIN28+ Aal spermatogonia.
Regulation of CDK2 activity is crucial for balancing spermatogonia progenitor self-renewal versus differentiation
Whereas SSC differentiation or maintenance was not obviously impacted in Cdk2 null (Cdk2−/−) mice, which demonstrate continued rounds of meiotic entry (Berthet et al., 2003; Ortega et al., 2003) (see also last section of Results and Fig. 7D), Cdk2Y15S heterozygotes exhibited an age-related decrease in spermatogenesis. This difference in phenotypes suggests that Cdk2Y15S is a hypermorphic or a gain-of-function allele. Given the increase in GFRA1+ spermatogonia in the Cdk2Y15S/Y15S testes (Fig. 3), we hypothesized that Cdk2Y15S either drives abnormal proliferation, and/or it skews these cells towards self-renewal instead of differentiation.
To test this, we assayed proliferation of spermatogonia by pulse labeling with the DNA analog 5-ethynyl-2′-deoxyuridine (EdU). P2 males were injected with EdU, sacrificed 4 h later, then the testes were immunolabeled for PLZF. In both mutants and WT, >99% of PLZF+ cells were negative for EdU (not shown), consistent with this being the period before gonocytes exit mitotic arrest to begin establishing the spermatogonial pool (Yang and Oatley, 2014). However, in older mutant animals, we noticed severe proliferation abnormalities in GFRA1+ and FOXO1+ cells. At P15, there were ∼1.6-fold more proliferating (EdU+) GFRA1+ cells in mutant homozygotes than WT, and this disparity persisted through P90 (Fig. 4A,C). Furthermore, the fractions of replicating As and Apr GFRA1+ spermatogonia were higher in mutant than WT, and, most dramatically, the Acluster>6 GFRA1+ category was unique to the mutant (Fig. 4A,D). Similarly, P90 Cdk2Y15S/Y15S testes, which contained cells with both nuclear (predominantly) and cytoplasmic FOXO1 (nFOXO1+ and cFOXO1+, respectively; Fig. 4B), had more proliferating FOXO1+ cells in both categories (Fig. 4E), in aggregate representing ∼50% more in the mutant than WT (22.5% versus 15%, P=0.01). In contrast, P90 mutant testes had ∼40% fewer proliferating LIN28+ germ cells, which include As spermatogonia (Fig. 4F). Collectively, our results support the notion that normal homeostasis of the stem cell niche is disrupted in Cdk2Y15S/Y15S mice, causing elevated proliferation of gonocytes and SSCs without normal differentiation.
Despite the evidence for spermatogonial cycling in the absence of differentiation, the seminiferous tubules never (up to 20 months of age) became replete with undifferentiated cells, a condition that might lead to, or resemble, tumorigenesis. Therefore, we hypothesized that such hyperproliferating progenitor spermatogonia were eliminated by apoptosis. Indeed, there were nearly 6-fold more terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)+ seminiferous tubules containing 2- to 4-fold more apoptotic germ cells in mutants versus WT at P30 (Fig. S3A-C). Furthermore, double immunolabeling for FOXO1 and cleaved PARP revealed the presence of spermatogonia undergoing apoptosis in mutants (Fig. S3D). These combined results suggest that Cdk2Y15S/Y15S undifferentiated progenitor GFRA1+ spermatogonia enter S phase normally but are incapable of differentiating; instead, they appear to proliferate, accumulate, and eventually undergo apoptosis.
Single cell transcriptome analysis reveals defects in differentiation of Cdk2Y15S/Y15S gonocytes and SSCs
The germ cell population at birth has substantial functional and molecular heterogeneity (Culty, 2013). Gonocytes in WT P3 testes, which constitute about half of all germ cells (Ohmura et al., 2004), can undergo one of three immediate fates: (1) re-enter the cell cycle, (2) migrate towards the basement membrane of the seminiferous cords to become SSCs, or (3) differentiate into spermatogonia that seed the first wave of spermatogenesis. Gonocytes decrease from 98% to 30% of all germ cells during the first week of life (Ohmura et al., 2004), thus lying within the GST interval (although the timing may differ in Swiss outbred mice; Pui and Saga, 2017). Given that Cdk2Y15S mutants are born with a normal number of germ cells (Fig. 1A), and that the first apparent abnormality was a deficit of PLZF+/LIN28+/FOXO1+ spermatogonia at P3 (Fig. S1C), we hypothesized that the neonatal germ cell pool was unable to differentiate into SSCs.
To test this at a molecular level, we performed scRNA-seq on unsorted cells from WT and mutant P3 testes. Data were obtained from WT (5061 cells), Cdk2Y15S/+ (4958 cells) and Cdk2Y15S/Y15S (4403 cells) testes. There was a median of 2239 genes and 5751 mRNA molecules detected per cell. Somatic and germ cells were evident in the clustering analysis of 14,422 cells from testes across all genotypes. Following unbiased k-means clustering of cells based on gene expression differences, we identified five major cell clusters (Fig. 5A). The cell types within each cluster were identified by markers diagnostic of particular gonadal lineages (Fig. S4). The percentages and numbers of cells in each of the five clusters were as follows: MVH+ germ cells (2.06%, n=299); WT1+ Sertoli cells (46.88%, n=6781); CYP11a1+ Leydig cells (1.5%, n=217); MYH11+ myoid cells (22.15%, n=3204); and VCAM1+ peritubular/epithelial cells (27.9%, n=3963).
To characterize potential GST defects, we focused on the MVH+ germ cell population. Interestingly, this group of cells consisted apparently of two populations that differed dramatically (>2-fold) in the number of unique transcripts (UMIs) per cell (Fig. S5A). This broad bimodal distribution was not observed in somatic cell populations. We considered two potential explanations for this observation. One is that the population with higher UMIs represents cell doublets (Ziegenhain et al., 2017). However, two algorithms used to predict doublets from single cell expression data, DoubletFinder (Fig. S5B) (McGinnis et al., 2019) and DoubletDetection (data not shown) (https://github.com/JonathanShor/DoubletDetection), indicated that the cell barcodes with higher UMIs (>10,000) were no more likely to be doublets than those with lower UMIs. A second explanation is that neonatal germ cells are transcriptionally more active than somatic cells, as are embryonic germ and stem cells (Percharde et al., 2017). To distinguish between these hypotheses, we quantified total RNA from equal numbers of Oct4-GFP+ cells isolated by fluorescence-activated cell sorting from neonatal testes of transgenic mice (Szabó et al., 2002). OCT4 (POU5F1) is a pluripotency marker and its expression is restricted to undifferentiated, prepubertal gonocytes/prospermatogonia (Ohbo et al., 2003; Szabó et al., 2002). This revealed 7-fold, 8.3-fold and 2.4-fold higher amounts of RNA in GFP+ germ cells compared with lung, brain and GFP− testis cells, respectively (Fig. S5C,D), supporting the hypothesis that neonatal germ cells have more transcripts. We next performed knee-plot analysis to identify a UMI cut-off for reliably profiled cells, revealing that the barcodes with low UMI counts fell in a low-quality region (under ‘knee’), indicating that those cells had poor detection (Fig. S5E) (Macosko et al., 2015). Therefore, we removed barcodes with low (≤10,000) UMI counts. After quality control and data filtering using Seurat (Butler et al., 2018), we identified 10,451 confidently quantified transcripts from WT and Cdk2Y15S/Y15S germ cells. Because Cdk2Y15S/+ clusters closely overlapped with those from WT, we only used the latter in subsequent comparisons with homozygous mutants.
Because only a fraction of all transcripts are detected in scRNA-seq, the resulting expression matrices are sparse. We therefore employed single cell analysis via expression recovery (SAVER) (Huang et al., 2018), which uses information across all genes and cells in a dataset, to impute gene expression values. We identified five subgroups of MVH+ germ cells, designated A-E (Fig. 5B), using hierarchical clustering based on the SAVER-imputed expression of the most divergently expressed genes (n=734; see Materials and Methods) (Fig. S6). As depicted in Fig. 5B, cluster A was exclusive to the WT sample, whereas D and E were exclusive to the Cdk2Y15S/Y15S sample. Clusters B and C were preferentially enriched in WT and the Cdk2Y15S/Y15S samples, respectively.
Clusters A and B express several well-characterized markers of undifferentiated germ cells, such as Gfra1 and Id4 (Fig. 5C). Based on the following distinctions between the two groups, we tentatively classified cluster A as SSCs (ASSC) and cluster B as gonocytes (BGono). Cluster A was relatively depleted in cell cycle factors such as Ccnb1 and Cenpf (Fig. 5C), consistent with the idea that ‘true’ SSCs would have a lower proliferative index. Furthermore, this cluster was enriched for Txnip (an inhibitor of glucose transport), indicating decreased metabolism consistent with lower proliferation. Finally, to validate the cluster identities, we combined previously published scRNA-seq datasets from flow-sorted OCT4+ (Liao et al., 2017 preprint) and ID4+ cells (Song et al., 2016), and identified 50 and 100 genes uniquely expressed in gonocytes and SSCs, respectively, which we then used as diagnostic markers for cell identities (Fig. S7A,B; Table S1). Gene set enrichment analysis (GSEA) revealed that highly expressed genes in the SSC gene sets were upregulated in cluster A compared with the reference cluster B, whereas highly expressed genes in gonocytes showed upregulation in cluster B (Fig. S7B). The combined data led us to conclude that cluster A consists of SSCs, and cluster B consists of gonocytes.
To better define the cellular defects in mutant germ cells present at P3, we compared key expression patterns of clusters A and D/E, which are the populations most specific to WT and Cdk2Y15S/Y15S, respectively. The following features were characteristic of D/E: (1) cell cycle signature genes and E2F targets were enriched (Figs 5C,F and 6F); (2) SSC genes were expressed at relatively low levels (Fig. 5C); and (3) the PGC/gonocyte marker NANOS3 and differentiating spermatogonia signatures (Stra8, Lmo1, Uchl1, Dmrt1, Sohlh1, Dnmt3b) were co-expressed and enriched (Fig. 5C,F). In summary, mutant germ cells express an unusual combination of cell cycle, differentiation and gonocyte signatures. Importantly, when we repeated these analyses using gene expression data without SAVER-based imputation, our results were consistent (Fig. S8).
We next performed pseudo-time analysis (Trapnell et al., 2014) on the germ cell cluster transcriptomes to explore the implications for developmental states and trajectories. In WT, this analysis supported a trajectory path bifurcating from cluster B (WTgono) to clusters A (WTSSCs) and C (differentiation-primed gonocytes, or WTdiff-Gono) (Fig. 5D), with the latter being defined by virtue of retaining both gonocyte and differentiation signatures (Fig. 5C,D). This trajectory path in WT is consistent with the known developmental progression in the germline. Interestingly, only ∼8% of mutant germ cells fell into cluster B, and these eventually differentiate into two directions: (1) a small subset towards cluster C (YSdiff-gono; ∼12%) and (2) ∼80% towards mutant-specific clusters D+E (Fig. 5D). It is worth noting that trajectory analyses using non-imputed data gave a different topology. Cluster C, instead of branching out as a separate cluster, appeared as a transient stage on a developmental trajectory differentiating from cluster B to E (Fig. S7C,D). Nevertheless, the analyses indicate a profound defect in differentiation of mutant germ cells, and are consistent with the idea that mutant gonocytes do not undergo a normal GST.
To gain additional perspective on developmental defects in the mutant, we performed RNA velocity analysis of the scRNA-seq data (Fig. 5E). This method examines the expression dynamics of unspliced (nascent) versus spliced (mature) versions of transcripts to predict the future developmental states of cells (La Manno et al., 2018). This analysis supports the conclusion that whereas most WT gonocytes will give rise to SSCs, the Cdk2Y15S/Y15S mutation causes nearly all mutant gonocytes (including those in clusters B and C; Fig. 5E) to transition to an abnormal gonocyte-like state exemplified by clusters D and E (summarized in Fig. 5F).
CDK2Y15S has altered kinase activity that impacts gonocyte fate
Although we hypothesized that CDK2Y15S is hypermorphic by virtue of lacking the target (Tyr15) of inhibitory phosphorylation (Singh and Schimenti, 2015), we considered the possibility that Ser15 could be phosphorylated by an unknown kinase to alter CDK2 activity. To test this, we expressed MYC-tagged WT (CDK2-TYR15), mutant (Ser15) and also Phe15 cDNAs in HEK293T cells, performed mass spectrometry (LC-MS/MS) analysis on immunoprecipitates (Fig. S8A), then analyzed the mass:charge (m/z) spectra for evidence of phosphorylation of these residues. Phosphorylation was detected only at the WT CDK2Y15 residue, indicating that serine at this position is not a phosphorylatable substrate, at least in cultured cells (Fig. S9B).
Next, we assayed the ability of CDK2 isolated from WT and mutant P10 spleens to phosphorylate a histone H1 substrate (testis was not used as a source because of cellularity differences between mutant and WT). Consistent with ablation of the Tyr15 inhibitory phosphorylation site and previous reports examining Cdk2T14AY15F activity in mouse tissues and mouse embryonic fibroblasts (MEFs) (Zhao et al., 2012), CDK2 immunoprecipitated from Cdk2Y15S/+ spleens displayed 1.5-fold more kinase activity than that of WT (Fig. S10B-D). Counterintuitively, material immunoprecipitated from Cdk2Y15S/Y15S spleens had >5-fold reduced kinase activity compared with WT. This may reflect the consequence of excessive CDK kinase activity, which can be toxic to cell cycle progression in a mechanism involving p21 (CDKN1A)-mediated inhibition of CDK2/cyclin (see Discussion) (Hughes et al., 2013; Szmyd et al., 2019; Zhao et al., 2012).
As an orthogonal assessment of CDK2 activity in germ cells, we compared expression levels of 97 key CDK2 activity signature genes in the cell clusters defined earlier (Table S2) (McCurdy et al., 2017). WTSSCs (cluster A) had much lower expression of CDK2 activity signature genes compared with all other clusters, including WT cells in cluster B and all Cdk2Y15S clusters (Fig. 6A). Furthermore, CDK2 kinase activity, as inferred by the median of normalized expression of CDK2 activity signature genes per cell across clusters A, B, D and E, was lowest in WTSSCs and highest in mutant-specific clusters D and E (Fig. 6B,C). GSEA analysis indicated upregulation of CDK2 activity signature genes in clusters B and E compared with cluster A, suggesting that (1) cluster E is more like cluster B with respect to CDK2 kinase activity, and (2) that cells in cluster E have a higher propensity to cycle than those in cluster A (Fig. 6C). Interestingly, in support of this observation, genes upregulated in cluster E compared with cluster A (q≤0.05; FC≥0.2) also exhibited enrichment of cell cycle-related gene ontology (GO) terms (Fig. 6D). Overall, these data indicate that removing a layer of negative regulation of CDK2 activity disrupts the normal differentiation of gonocytes and SSCs into downstream cell types.
The molecular basis for this developmental disruption may stem from alteration of two known functions of active CDK2: (1) phosphorylation of the RB1 transcriptional repressor to inactivate RB1, enabling timely induction of the E2F transcription factors (TFs) to drive transition to S phase (Morris et al., 2000); and (2) inhibition of cytoplasmic-to-nuclear localization of the FOXO1 TF (nFOXO1 is essential for SSC maintenance) (Huang et al., 2006; Goertz et al., 2011). If CDK2 is indeed controlling the gene regulatory network via E2F and FOXO1, then levels or activity of their downstream targets might be affected in Cdk2Y15S/Y15S cells. Indeed, GSEA analysis revealed that FOXO1 targets are upregulated and highest in cluster A versus cluster E, where expression is lowest [84 genes; normalized enrichment score (NES): 3.27; family-wise error rate (FWER), p<0.001] (Fig. 6E,G,H). In contrast, E2F target genes are upregulated in cluster E (Fig. 6F; 66 genes; NES: −2.7; FWER, p<0.001). These results indicate that FOXO1 activity is significantly greater in cluster A than E (Student's t-test, P=3.7×10−19) (Fig. 6H).
Phosphorylation states at Tyr15 and Thr160 residues control CDK2 activity in male germ cells
If disrupting the ability to negatively regulate CDK2 in adult GFRA1+ SSCs favors cell cycle progression over differentiation, then a compensatory alteration that dampens or eliminates CDK2 activity might counteract the aberrant phenotype of Cdk2Y15S mutant spermatogonia. To test this, we mutated threonine 160 to alanine (T160A) in the Cdk2Y15S allele. Phosphorylation of Thr160 is required for activation of CDK2 (Gu et al., 1992; Kaldis, 1999). Cdk2T160A is a ‘kinase dead’ allele that causes infertility in both sexes, albeit with less severe meiotic phenotypes than in Cdk2−/− mice (Chauhan et al., 2016). Mice homozygous for the doubly mutated Cdk2 allele (Cdk2Y15S,T160A, abbreviated as Cdk2YS-TA) exhibited three striking phenotypic differences compared with Cdk2Y15S. First, whereas Cdk2Y15S adult (P120) heterozygotes had small testes and a markedly reduced sperm count, as previously reported (Singh and Schimenti, 2015), Cdk2YS-TA heterozygotes were indistinguishable from WT in both respects (Fig. 7A-C). Second, as with null but not Cdk2Y15S/Y15S mice, Cdk2YS-TA homozygous females were sterile (n=3 in matings to WT). Third, although Cdk2YS-TA/YS-TA adult males were severely hypogonadal (Fig. 7A,B) and azoospermic, similar to Cdk2Y15S/Y15S, they exhibited differentiating spermatogonia and meiocytes, which were completely missing from age-matched Cdk2Y15S homozygotes, indicating that the T160A alteration rescued the spermatogonial differentiation defect (Fig. 7D). At these levels of analysis, the Cdk2YS-TA allele resembles the null phenotype (Viera et al., 2009; Chauhan et al., 2016). Collectively, our results imply that the failed spermatogonial differentiation phenotype caused by the Cdk2Y15S allele is a result of altered kinase activity from abolition of a WEE1 phosphorylation site. As a consequence of this defective negative regulation, this allele acts semidominantly.
The longevity of spermatogenesis in both mice and humans depends upon proper establishment and homeostasis of the SSC pool. A key event in male germline establishment is the seeding of prepubertal testes with a quiescent (G1-arrested) population of gonocytes/prospermatogonia. About 2-3 days after birth, these cells re-enter the cell cycle to expand and differentiate into SSCs that establish a permanent, renewable pool of cells that can initiate spermatogenesis throughout life. Once established, there must be a fine balance of SSC self-renewal versus differentiation to maintain homeostasis and fertility, but this is not well-understood and is the subject of intense research.
There is some debate as to the character of SSCs and their behavior with respect to cycling activity (Huckins, 1971b; Sharma et al., 2019). In some tissues, the ability to continuously produce differentiated cells depends upon proper maintenance of a relatively infrequently cycling population of stem cells, as in the case of the hematopoietic system. Overproliferation of hematopoietic stem cells (HSCs) caused by deregulated cell cycle control can lead to their exhaustion or transformation (Pietras et al., 2011). It has been hypothesized that there is a slow-cycling population of SSCs that maintains the germline (Huckins, 1971a; Sharma et al., 2019). However, there is also evidence for rapid turnover of SSCs (Klein et al., 2010) and for the ability of chains of differentiating spermatogonia (Apr and Aal) to fragment and de-differentiate to become As SSCs (Hara et al., 2014; Nakagawa et al., 2010). Regardless of various models proposed for the identity and behavior of ‘true’ SSCs (reviewed by de Rooij, 2017), proper regulation of the cell cycle is essential. This is underscored by our studies, which implicate CDK activity regulation as being crucial for the GST and SSC renewal versus differentiation. Indeed, it is well-recognized that, in general, cell cycle progression and differentiation occur in a mutually exclusive manner (Dalton, 2015). For example, a decrease in CDK activity stimulates differentiation of neural stem cells (Lim and Kaldis, 2012), whereas elevated CDK activity decreases differentiation (Lange et al., 2009) and stimulates expansion of neural stem cells in the adult mouse brain (Artegiani et al., 2011).
Phenotypes of certain mouse mutants have provided some insight into how regulation of cell cycle impacts spermatogonial maintenance and proliferation. Conditional germline knockout of the Rb (Rb1) tumor suppressor, a negative cell cycle regulator, abolished the ability of SSCs to self-renew, causing the entire germ cell pool to undergo a single round of spermatogenesis (Hu et al., 2013). Cell cycle progression in normal cells requires inactivation of Rb by CDK/cyclin-mediated phosphorylation (including by CDK2/cyclinE), thus allowing expression of genes regulated by E2F transcription factors (Rubin, 2013). Moreover, ablation of Plzf, which negatively regulates the cell cycle by both inhibiting key regulators (McConnell et al., 2003; Yeyati et al., 1999) and also the self-renewal signal of GDNF (Hobbs et al., 2010), causes a less severe phenotype than Rb deficiency. Plzf−/− males are infertile due to a defect in SSC maintenance that leads to progressive germ cell loss and SCOS (Buaas et al., 2004; Costoya et al., 2004). Thus, a cell cycle-centric interpretation of these phenotypes is that unrestrained cycling (as in Rb−/−) causes efficient differentiation of all SSCs, but a moderate loss of cell cycle control (as in Plzf−/−) increases the propensity of SSCs to differentiate rather than self-renew. In the context of this model, CDK2Y15S might have a lower impact on cell cycle control than Rb and Plzf mutants, possibly mediated by abnormal but partial inactivation of Rb. The result is abnormal SSC proliferation but only partial differentiation, ultimately leading to loss of the aberrant cells (clusters D and E) before meiotic entry. The defective differentiation may explain the observation that P90 Cdk2Y15S/Y15S mutants had increased proliferating GFRA1+ cells but fewer proliferating cells expressing LIN28, which normally marks a subset of As ‘true’ SSCs.
MEFs derived from mice in which both the Thr14 and Tyr15 negative phosphoregulatory sites were mutated exhibited accelerated entry into S phase (Zhao et al., 2012). Interestingly, though no data were presented, the only significant defect reported in these mice was male infertility due to an apparent absence of germ cells in homozygotes and also in heterozygotes in one strain background. In that regard, this model appears to resemble our Cdk2Y15S mouse. This might indicate that the Thr14 residue is redundant or not important for negative regulation of SSCs.
Counterintuitively, although we observed elevated CDK2 kinase activity in spleens from heterozygous mice, the activity was greatly reduced in homozygous spleens. However, CDK2 kinase over-activation has been previously shown to be deleterious to cell cycle progression (Hughes et al., 2013). We postulate that, at least in spleen, cell-autonomous regulation maintains CDK2 activity within an appropriate range. Mutation of both negative phosphoregulatory sites in CDK2 (Thr14 and Tyr15) in human Hct116 cell lines caused premature S-phase progression, DNA-damage accumulation, and genomic instability leading to S-phase arrest (Hughes et al., 2013). In these cells, degradation of cyclin E was found to be increased and this could be relieved by p21 depletion, suggesting the presence of cellular feedback loops to attempt to reduce the levels of CDK2 activity upon premature activation (Hughes et al., 2013). In consideration of the aforementioned data of others, and our data showing elevated replicative activity and apoptosis of GFRA1+ cells in Cdk2Y15S/Y15S testes, we propose a model (Fig. 8) in which CDK2-associated kinase activity must be tightly regulated during specific stages of the cell cycle in SSCs, otherwise those cells attempting to differentiate will eventually die from cell cycle dysregulation.
A caveat to the model is that it is based on cell cycle studies in homogeneous cell cultures, as opposed to germ cells developing in a complex milieu that provides a stem cell niche containing several somatic cell types (Hofmann, 2008; Oatley and Brinster, 2012). In particular, the Sertoli cells, which associate intimately with the germ cells, provide key instructive signals including retinoic acid to induce differentiation (Endo et al., 2015). Furthermore, along with endothelial cells, Sertoli cells produce GDNF (which binds GFRA1), which is essential for spermatogonial maintenance (Bhang et al., 2018; Meng et al., 2000). It is possible that the phenotypes of Cdk2Y15S mice may not be a manifestation of germ cell-intrinsic defects exclusively, but rather to defects in one or more somatic cell types in addition to those of germ cells. Disruption to individual components of the testis niche could impact the transcriptome and behavior of germ cells, and vice versa. Although there were differences between mutant and WT Sertoli cells transcriptomes, EdU labeling of P30 mice revealed no evidence for active DNA replication in mutant or WT SOX9+ Sertoli cells (data not shown), suggesting that CDK2Y15S affects the cell cycle of germ cells (Fig. 4) rather than Sertoli cells. Ultimately, these issues of potential niche defects could be addressed via germ cell transplantation or cell type-specific ablation experiments.
Although there are mutations that reduce the number of perinatal gonocytes (typically stemming from PGC defects) (Hamer and de Rooij, 2018), the Cdk2Y15S phenotype is unique in that the GST is defective. Nevertheless, studies of other mutants give insight into how perturbations to cell cycle regulation impact gonocytes. Cdk2Y15S/Y15S germ cells fail to relocalize FOXO1 from the cytoplasm to the nucleus, a hallmark of the GST that normally occurs between P3 and P21 (Goertz et al., 2011; Kang et al., 2016; Pui and Saga, 2017). This cytoplasmic localization occurs as a result of CDK2-dependent phosphorylation (inactivation) of FOXO1 (Huang et al., 2006), which is essential for SSC maintenance and differentiation (Goertz et al., 2011). As FOXO1 nuclear localization is disrupted in Cdk2Y15S mutants, we conclude that proper cell cycle regulation and/or CDK2 kinase activity is a pre-requisite for the GST, and thus lies upstream in the developmental evolution/progression. Mice lacking the transcription factor Glis3 partially resemble Cdk2Y15S mutants; they lack FOXO1 nuclear localization in neonatal gonads and fail to undergo a normal GST, as indicated by decreased expression of genes associated with the permanent pool of undifferentiated spermatogonia (Kang et al., 2016). However, it is unknown whether GLIS3 regulates these genes directly or indirectly, so its relationship to CDK2 in the developmental hierarchy is unclear.
Interestingly, the first round of spermatogenesis in Cdk2Y15S/Y15S mice occurs normally, although meiosis is still interrupted, as in null mice. This first wave of spermatogenesis is claimed to arise directly from a subset of gonocytes that do not express neurogenin 3 (Yoshida et al., 2006). We can conclude that differentiation of this subtype of gonocytes neither requires CDK2 nor is affected by its abnormal activity in Cdk2Y15S mutants. This underscores the caution needed when extrapolating data from cultured cells or from somatic cells (such as spleen) to germ cells, and even from one germ cell type to another.
These investigations into the Cdk2Y15S mouse stemmed from a project to investigate the genetic basis of human infertility, and involves modeling human SNPs in mice to determine possible pathogenicity (Singh and Schimenti, 2015). A lesson learned from this and other alleles is that very few predicted deleterious variants are nulls, and that extensive phenotyping is required to understand their impact. Generally speaking, unless this mutation is entirely unusual in its effects, it raises the possibility that mice and people that present with non-obstructive azoospermia, and which have histopathology superficially resembling SCOS, may in fact have an occult pool of SSCs that have lost the ability to differentiate, but might be stimulated to do so after appropriate intervention. Additionally, as an example of an autosomal semi-dominant male infertility allele, it emphasizes the importance of considering monoallelic alterations when attempting to identify genetic causes of an individual patient's infertility.
MATERIALS AND METHODS
Mouse strains and breeding
Mice used were on a mixed genetic background [FVB/NJ and B6(Cg)-Tyrc-2J/J]. Experiments with the animals were performed under a protocol (2004-0038) approved by Cornell's Animal Care and Use Committee. For female fertility tests, 8- to 10-week-old WT or mutant homozygote females were mated with age-matched B6(Cg)-Tyrc-2J/J males. The Cdk2tm1Kald, Cdk2T160A and Cdk2Y15S alleles have been described previously (Berthet et al., 2003; Chauhan et al., 2016; Singh and Schimenti, 2015).
Production of CRISPR/Cas9-edited mice
The Cdk2Y15S-T160A (formally Cdk2em2Jcs) and Cdk2−/− (formally Cdk2em3Jcs) alleles were generated using CRISPR/Cas9 genome editing, essentially as described previously (Singh et al., 2015; Varshney et al., 2015). sgRNAs and ssODNs are listed in Table S1. Briefly, in vitro transcription of sgRNA was performed using a MEGAshortscript T7 Transcription Kit (Ambion, AM1354), and ssODN were obtained from IDT. sgRNA, ssODN, Cas9 protein and mRNA (Addgene plasmid #44758) were co-microinjected into zygotes [F1 hybrids between strains FVB/NJ and B6(Cg)-Tyrc-2J/J] using the reagent concentrations listed in Table S1. Edited founders were identified either by subcloning followed by Sanger sequencing using the primers and annealing temperatures listed in Table S1. For generating the Cdk2Y15S-T160A allele, Cdk2Y15S/Y15S females were used as embryo donors; thus, the Cdk2Y15S alteration was not re-introduced by CRISPR. The Cdk2−/− allele contained a 2-nucleotide deletion in exon 1, leading to a premature termination codon and a predicted truncated CDK2 protein of 31 amino acids.
Testes histology and immunohistochemistry
For histological analyses, testes were fixed for 24 h at room temperature (RT) in Bouin's solution, paraffin-embedded, sectioned at 7 μm, and then stained with Hematoxylin & Eosin. For IHC, testes were fixed in 4% paraformaldehyde for ∼24 h, paraffin-embedded, sectioned at 7 μm, and deparaffinized. Antigen retrieval for different antibodies was performed as indicated in Table S4. Sections were blocked in PBS containing 5% goat serum for 1 h at RT, followed by incubation with primary antibodies for 12 h at 4°C and detection by the secondary antibody (see Table S4 for antibody details).
Whole-mount seminiferous tubule staining
Seminiferous tubule whole mounts were prepared as described previously (Savitt et al., 2012), with minor modifications. Briefly, seminiferous tubules were manually isolated in PBS and interstitial tissue was washed thoroughly followed by fixation in 2% paraformaldehyde for 2 h. After extensive washing, tubules were processed for detection of progenitor spermatogonial markers and EdU labeling. To detect ZBTB16, FOXO1, cPARP1 and LIN28, tubules were blocked in PBSS buffer (5% goat serum, 0.5% Triton X-100 in 1× PBS) for 2 h at RT following primary antibody incubation for 12-18 h at 4°C in the same buffer. Tubules were washed at room temperature twice for 15 min, five times for 1 h in PBSS, and then incubated overnight at 4°C with secondary antibodies. To detect GFRA1, the blocking and all incubations and washing steps were carried out exclusively in PBS-AT (1% bovine serum albumin, 0.1% Triton X-100 in 1× PBS) buffer. Briefly, isolated tubules were blocked for 2 h and incubated in goat anti-GFRA1 antibody diluted in PBS-AT buffer. This was followed by five subsequent washes (30 min each) and detection by secondary antibodies. When performing co-immunolabeling of GFRA1 with other spermatogonial markers, ZBTB16, FOXO1, cPARP1 and LIN28 were always detected prior to GFRA1.
For EdU incorporation, mice were injected intraperitoneally with 100 mg/kg EdU (PY7562, Berry & Associates) before harvesting and processing testes 4 h later for whole-mount staining, or fixed in 4% paraformaldehyde for IHC. Following antibody staining, freshly prepared Click reaction cocktail [10 mM (+)-sodium-L-ascorbate, 0.1 mM 6-caboxyfluorescein-TEG azide (FF 6110, Berry & Associates) or Alexa Fluor 594 Azide (A10270, Invitrogen) and 2 mM copper (II) sulfate in water] was added into whole-mount tubules or testes cross-sections and removed quickly after 45 s. Excess EdU was removed by extensive washing (2×1 h and 2×12 h at 4°C) using PBS-AT buffer. Following washing, tubules were mounted in Vectashield (Vector Laboratories). Click reaction was always performed following antibody detection. In each animal, we counted all As, Apr and Aal GFRA1+ cells positive or negative for EdU seminiferous tubules at least >8 mm in length.
Slide preparations were scanned, and tiled images of testes cross-sections or seminiferous tubule whole mounts were acquired using a laser scanning confocal microscope (u880, Carl Zeiss) using Plan Apo 40× water immersion objective (1.1 NA) and Zen black software. Initial laser power adjustment was performed to avoid saturation of signal: argon laser, 488 nm; blue-diode, 405 nm; DPSS laser, 561 nm; HeNe, 633. Additional images of testes cross-sections were obtained through an Olympus BX51 microscope with objectives 100×/1.35 NA infinity/0.17 or objective or 10×/0.3 NA, respectively, using Olympus Cell Sens software (Olympus). Following identical background adjustments for all images, cropping, color and contrast adjustments were made with Adobe Photoshop CC 2017.
In vitro kinase assay
In vitro kinase assays were performed as described elsewhere (Lim et al., 2017). Briefly, 1 µg of anti CDK2 antibody was added to 0.5 µg of protein lysate and rotated overnight (∼16 h) at 4°C. Twelve microliters of protein A beads was added to each sample and left to mix for a further 2 h. Prior to their use, these beads were washed twice previously using 1 ml of EBN buffer (80 mM β-glycerophosphate, 15 mM MgCl2, 20 mM EGTA, adjusted to pH 7.3 with KOH; followed by adding 150 mM NaCl, 0.5% NP40 and 1× protease inhibitor cocktail). After binding, the beads were spun down and washed with EBN buffer three times. Pellets were re-suspended in 14 µl EBN buffer (containing protease inhibitors: cOmplete, Mini, 11836153001; Roche) followed by adding 6 µl of histone H1 (260 ng/μl) and 9 μl of kinase assay buffer (15 mM EGTA, 25 mM NaF, 250 mM sodium β-glycerophosphate, 5 mM DTT, 20 mM MgCl2, 21 µM ATP). Following 20 min incubation at room temperature, 1 µl of ATP [γ-32P] containing a specific radioactivity of 5 µCi (Perkin Elmer, BLU502A) was added to each sample and left for 30 min at 30°C. Then 6× SDS sample buffer [2 M β-mercaptoethanol, 0.375 M Tris (pH 6.8), 12% SDS, 60% glycerol, 0.6 M DTT, 0.06% Bromophenol Blue] was added to a final concentration of 1× and each sample was boiled at 95°C for 5 min. Next, 12 µl of each sample were analyzed for Coomassie-stained histone H1 protein bands and phosphosignal on a phosphor screen cassette for 6-24 h. Phosphosignal was quantified using a FLA7000 phosphorimager (Fujifilm).
Protein was extracted using EBN buffer (see above) with protease inhibitors. Samples were boiled at 95°C for 5 min and were separated by SDS/PAGE (12% acrylamide). Separated proteins were electrotransferred to 0.2 µM nitrocellulose (Bio-Rad, 1620112) or PVDF membrane (Immobilon-P membrane, IPVH00010, EMD Millipore), and then blocked in 5% nonfat milk for 45 min at RT. Membrane was washed 3×10 min in TBST (0.14 M NaCl, 15 mM KCl, 25 mM Tris Base, 1% Tween20) at room temperature followed by ∼16 h incubation with primary antibody, washing, and ∼1 h incubation with HRP-conjugated secondary antibody (as stated in Table S4) for 45 min at room temperature. Signal was detected using Luminata Classico Western HRP substrate (WBLUC0100; EMD Millipore). Densiometric analysis of western blot bands was performed using Fujifilm Multi Gauge software Ver. 3.1.
Site-directed mutagenesis for obtaining CDK2-Y15F and CDK2-Y15S cDNA was accomplished using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent, 210518) using the primer sets listed in Table S3. All mutations were confirmed by Sanger sequencing.
CDK2 overexpression and immunoprecipitation
The (Myc-DDK-tagged)-CDK2 cDNA was obtained from Origene (RC200494). Y15F and Y15F mutant cDNA were generated as described above and transfected into HEK293T cells (ATCC; CRL-3216) using TransIT-LT1 Transfection Reagent (Mirus, MIR2305) following the manufacturer's instructions. Lysates were prepared from harvested cells in lysis buffer (50 mM Tris pH7.5, 150 mM NaCl, 0.5% Triton X-100, 5 mM EDTA) containing phosphatase (Thermo Fisher Scientific, 78428) and protease inhibitors (Roche, 04693159001). 400 µg of each lysate was immunoprecipitated using the c-Myc-Tag IP/Co-IP Kit (Pierce, 23620). Proteins were separated on 12% polyacrylamide gel by SDS-PAGE and CDK2 protein was visualized by western blot analysis.
Protein identification by nanoLC-ESI-MS/MS analysis
In-gel trypsin digestion of immunoprecipitated CDK2 protein was performed as described previously (Yang et al., 2007). The tryptic digests were subjected to nanoLC-ESI-MS/MS analysis on an Orbitrap Fusion Tribrid (Thermo Fisher Scientific) mass spectrometer equipped with a nanospray Flex Ion Source, and a Dionex UltiMate3000RSLCnano system (Thermo Fisher Scientific) following a protocol described previously (Thomas et al., 2017; Yang et al., 2018). The column was re-equilibrated with 0.1% formic acid for 23 min prior to the next run. For data-dependent acquisition (DDA) analysis, the instrument was operated using ‘FT mass analyzer’ in MS scan to select precursor ions followed by 3 s ‘Top Speed’ data-dependent CID ion trap MS/MS scans at 1.6 m/z quadrupole isolation for precursor peptides with multiple charged ions above a threshold ion count of 10,000 and normalized collision energy of 30%. MS survey scans at a resolving power of 120,000 (fwhm at m/z 200), for the mass range of m/z 375-1575. Dynamic exclusion parameters were set at 30 s of exclusion duration with ±10 ppm exclusion mass width. All data were acquired using Xcalibur 3.0 operation software (Thermo Fisher Scientific).
Mass spectrometry data analysis
The DDA raw files for CID MS/MS were subjected to database searches using Proteome Discoverer (PD) 2.2 software (Thermo Fisher Scientific) with the Sequest HT algorithm. All three raw MS files for three samples were used for the database search. The PD 2.2 processing workflow containing an additional node of Minora Feature Detector for precursor ion-based quantification was used for protein and post-translational modifications identification. The database search was conducted against a mouse database containing ∼20,153 entries downloaded from NCBI database on 12 January, 2018, plus some common contaminants (246 entries). Two missed trypsin cleavage sites were allowed. The peptide precursor tolerance was set to 10 ppm and fragment ion tolerance was set to 0.6 Da. Variable modification of methionine oxidation, deamidation of asparagines/glutamine, phosphorylation of serine, threonine and tyrosine, and fixed modification of cysteine carbamidomethylation, were set for the database search. Only high-confidence peptides defined by Sequest HT with a 1% FDR by Percolator were considered for the peptide identification. The final protein IDs contained protein groups that were filtered with at least two peptides per protein.
Sperm counting was performed as described previously (Singh and Schimenti, 2015).
scRNA seq sample preparation
At P3.5 testes were decapsulated in HBSS (Mediatech) and digested with 0.642 ml of Trypsin/EDTA (0.25%, Invitrogen) and 0.071 ml DNase I (1 mg/ml in HBSS, Sigma-Aldrich) at 37°C for 3-5 min. Digestion was stopped by adding 0.5 ml of media (HBSS+10% FBS), the cells were filtered through a 70-µm cell strainer and re-suspended in HBSS with 0.04% FBS. Single cell 3′ RNA-seq sequencing libraries for Illumina were constructed using a 10x Genomics Chromium instrument, using the Chromium Single Cell 3′ Reagent kits (v2) following the manufacturer's protocols. The final libraries were quantified by digital PCR and sequenced on Illumina sequencers, using an Illumina NextSeq500/550 75 bp kit (26 bp+8 bp index read+58 bp). The target number of cells captured from the input, single cell suspension was 8700. The data were demultiplexed and aligned to the reference genome using the 10x Genomics Cellranger software (v2.2) and visualized using the 10x Genomics Loupe Cell Browser software (v2.0) packages.
scRNA-seq data analysis
The raw scRNA-seq data were analyzed using cell ranger from the 10x platform to generate a matrix of raw read counts, which was further analyzed in R using Seurat (Butler et al., 2018; Satija et al., 2015) and Monocle (Qiu et al., 2017). A cluster of germ cells from WT and Cdk2Y15S/Y15S were separated from the rest of the testis cells based on Ddx4 gene expression. We then isolated cell barcodes that had more than 10,000 UMIs per cell (see Results). Raw read counts for these selected germ cells were further filtered to keep only those genes that were detected in at least 10% of the cells. This resulted in an expression matrix of 10,451 genes across 141 cells (WT=69; Cdk2Y15S/Y15S=72). Because of the high rate of drop-outs in mRNA-capture in scRNA-seq approaches, the expression values for mid- and low-expressed genes are often unreliable due to missing information. Hence, we recovered expression for each gene using SAVER (Huang et al., 2018), an approach that estimates gene expression from UMI-based scRNAseq data using information across all genes and cells. This estimated gene expression was used for further analysis. The most variable genes across all cells were identified using principal component analysis (PCA). Specifically, PCA was performed using log-transformed counts-per-million (CPM) expression values. The top 300 genes with highest absolute loading for PC1, top 300 genes for PC2 and top 300 genes for PC3 were chosen, which resulted in 744 unique genes. Furthermore, this list of variable genes was filtered to remove mitochondrial genes (n=10). The 141 germ cells were clustered into five groups using the ‘Ward.D2’ method based on expression estimates for the 734 highly variable genes. The pseudotime and trajectory analyses were performed based on the expression of the 734 most variable genes using Monocle R code. RNA velocity analysis (La Manno et al., 2018) was performed using Velocyto, available at velocyto.org . GSEA (Subramanian et al., 2005) was performed on the web server software.broadinstitute.org/gsea/index.jsp.
P-values were calculated from unpaired Student's t-tests for all IHC and whole-mount experiments.
The authors would like to thank R. Munroe and C. Abratte of Cornell's transgenic facility for generating the Cdk2T160A allele alone and in cis to the Cdk2Y15S change. Authors also thank the Proteomics Facility of Cornell University for providing the mass spectrometry data and NIH SIG 1S10 OD017992-01 grant support for the Orbitrap Fusion mass spectrometer. Confocal imaging data was acquired in the Cornell BRC-Imaging Facility using the NIH-funded (S10OD018516) Zeiss LSM880 confocal/multiphoton microscope (u880).
Conceptualization: P.S., D.P., J.C.S.; Methodology: N.P., P.K., A.G.; Formal analysis: P.S., R.K.P., N.P., J.K.G.; Investigation: P.S., N.P., D.P.; Writing - original draft: P.S., J.K.G., J.C.S.; Writing - review & editing: R.K.P., J.C.S., P.K., A.G.; Supervision: P.K., A.G., J.C.S.; Project administration: J.C.S.; Funding acquisition: J.C.S.
This research was funded by National Institutes of Health grants (R01 HD082568 to J.C.S. and P50 HD076210 to A.G. and D.P.), the Biomedical Research Council, the Agency for Science, Technology and Research (A*STAR; to P.K.), a Singapore International Graduate Award (SINGA; to N.P.), a Biomedical Research Council – Joint Council Office grant (1231AFG031 to P.K.), the National Medical Research Council Singapore, NMRC (NMRC/CBRG/0091/2015 to P.K.), a National Research Foundation Singapore grant (NRF2016-CRP001-103 to P.K.) and contract CO29155 from the New York State Stem Cell Science Program (NYSTEM). Deposited in PMC for release after 12 months.
The 10x single cell RNAseq data have been deposited in Gene Expression Omnibus with accession GSE130554.
The authors declare no competing or financial interests.