Exon inclusion is the most prevalent type of alternative splicing during transition from mitosis to meiosis in germ cells
We have previously demonstrated that germ cells physiologically reduce their amount of MAX protein that constitutes the core of PRC1.6 with MGA to de-repress meiosis-related genes prior to meiotic onset21. In this study, we explored the possibility of an additional molecular mechanism that inactivates the function of PRC1.6 to ensure meiotic entry. Because the testis is known for its prevalence of alternative splicing similar to the brain24, 25, we pursued the possibility of involvement of alternative splicing in facilitating meiosis by deteriorating the function of PRC1.6. First, we examined which stages in spermatogenesis and which types of alternative splicing were prevalent in germ cells by inspecting publicly reported RNA sequence data. These analyses revealed that alternative splicing occurred most actively during meiotic onset (transition of spermatogonia to preleptotene spermatocytes) (Supplementary Figure 1A). In terms of the types of alternative splicing, the skipping exon (SE) type was the most prevalent, which accounted for more than 50% of all alternative splicing events among the five distinct splicing types. Furthermore, our analyses revealed that gain of a novel exon was approximately twice as frequent as loss of an exon among the SE-type alternative splicing events (Supplementary Figure 1B). Moreover, our data revealed that most transcripts that gained or lost an exon around meiotic onset maintained their forms by de novo synthesis and/or stabilization at least up to the round spermatid stage (Supplementary Figure 1C). We also classified alternative splicing events in neural progenitor cells and mesenchymal stem cells using publicly reported RNA sequence data. These analyses revealed that SE was also the most prevalent alternative splicing during differentiation of neural progenitor cells and mesenchymal stem cells (Supplementary Figure 1A). However, comparisons of genes that gained a new exon revealed that these three gene populations of germ cells, neural cells and mesenchymal stem cells barely overlapped (Supplementary Figure 1D), which indicated that at least genes subjected to this type of alternative splicing were selected distinctly in each cell type.
Identification of testis-specific Mga variant mRNA
Based on these data, we explored the possibility that the SE type of alternative splicing, particularly gain of a novel exon, was involved in the inactivation of PRC1.6 during the transition from mitosis to meiosis in germ cells. Because SpliceAI, a deep neural network that predicts mRNA splicing from a pre-mRNA sequence23, was developed recently, we used this technology to identify putative exon sequences within genes encoding a component of PRC1.6. First, we confirmed that SpliceAI identified all exons of genes encoding a component of PRC1.6 (Mga, Max, L3mbtl2, E2f6, Rnf2, and Pcgf6) (Fig. 1A and Supplementary Figure 2), which validated the accuracy of prediction by this deep learning program. More importantly, SpliceAI additionally predicted six genomic regions as putative exons (one each within E2f6 and Pcgf6 genes and two each in Mga and L3mbtl2 genes) (Fig. 1A and Supplementary Figure 2). Therefore, we performed RT-PCR analyses of RNAs from various tissues to determine the possibility that RNAs transcribed from these regions were incorporated as exon sequences into mature mRNA in certain tissues (Supplementary Figure 3). These analyses revealed that a putative exon located within the 18th intron of the Mga gene was specifically incorporated into RNA from the testis, but only marginal presence was evident in other tissue RNAs. We also confirmed testis-specific incorporation of the transcript from this region by quantitative PCR (qPCR) (Fig. 1B, upper panel). However, we found that the other five regions were not entirely incorporated into any examined tissue RNAs. According to the scores calculated by SpliceAI (0.695 and 0.761 for the splice acceptor and donor, respectively), this region was not predicted to be a constitutive exon (more than 0.9), but as an exon subjected to a substantial degree of alternative splicing regulation (usually between 0.1 and 0.9). Thus, our finding that this region was used for testis-specific alternative splicing was compatible with the prediction of SpliceAI. Therefore, we termed this sequence and Mga variant mRNA carrying this sequence as exon 19a and Mga splice variant (SV), respectively. With respect to canonical Mga mRNA, we confirmed broad expression in various cell types. We also noted its relatively higher expression in the testis compared with any other examined tissues (Fig. 1B, lower panel).
RNA in situ hybridization analyses of wildtype and variant Mga mRNAs in the testis and epididymis
To determine which portions of the testis expressed Mga SV, RNA in situ hybridization analyses were performed using testis and epididymis tissues of adult mice. To avoid the difficulty associated with the short length (145 bp) of the Mga SV-specific sequence, we employed the BaseScope in situ hybridization technique rather than the conventional RNA hybridization method26. These analyses clearly demonstrated that Mga SV-positive cells were not present in the interstitium portion, but restrictively present within seminiferous tubules, while canonical Mga-positive cells were detected in both portions (Fig. 2A). These analyses also revealed that cells positive for canonical Mga and those for Mga SV were abundantly and scarcely detected in the epididymis, respectively (Fig. 2B).
Mga splice variant is restrictively present around the meiotic stage of germ cells
To determine whether restrictive expression of Mga SV in cells within seminiferous tubules of the testis represented specific expression in germ cells, we prepared four distinct germ cell types, i.e., undifferentiated (Thy1+) and differentiated (c-Kit+) spermatogonia from testes of mouse pups at postnatal days 5–8 and spermatocytes and round spermatids from adult mouse testes. Then, qPCR analyses were conducted using total RNAs from these cell populations. The analyses revealed that the levels of canonical Mga mRNA were comparable among the four cell populations, whereas much higher expression of Mga SV was evident in spermatocytes and spermatids compared with Thy1+ and c-Kit+ spermatogonia in which meiotic initiation is blocked 27 (Fig. 3A). These results indicated that germ cells in seminiferous tubules of the testis generated Mga SV and the timing of generation was coincidental with the timing of transition from mitosis to meiosis in cell division. Next, we examined publicly reported RNA sequence data of this sequence in spermatogonia, round spermatids, and germ cells undergoing meiosis in the testis (preleptotene and pachytene spermatocytes). In accordance with our qPCR data (Fig. 3A), visualization of publicly reported RNA sequence data of this putative exon (exon 19a) by Sashimi plot revealed specific production of Mga SV in meiotic spermatocytes and round spermatids, but not in spermatogonia (Fig. 3B). Male germ cells initiate meiosis in the testis at the beginning of puberty, whereas female primordial germ cells (PGCs) undergo meiosis and proceed to the diplotene stage of meiotic prophase I in the gonads during the mid-gestation stage28, 29. Therefore, we inspected the publicly reported RNA sequence data to determine whether female PGC-specific onset of meiosis in the embryonic stage was also accompanied by the production of Mga SV. The analyses revealed that expression of Mga SV in female PGCs of the gonads became detectable from 11.5 to 16.5 dpc with peak expression at 15.5 dpc when meiotic germ cells are mostly at the pachytene stage, whereas no apparent peak for the production of Mga SV was detected in male germ cells during embryonic stage (Fig. 3C). Taken together, our data demonstrated an intimate link in the time scale between expression of Mga SV and meiosis in both male and female germ cells. Because female PGCs undergo meiosis at around 13 dpc, initiation of Mga SV production appeared to occur prior to meiotic onset. It is also noteworthy that exon 19a bears in-frame stop codon (Fig. 3D), indicating that the coding sequence located downstream of the stop codon of exon 19a sequence, including that for the bHLHZ domain, was nullified as depicted. Intriguingly, inspection of the publicly reported RNA sequence data revealed that insertion of a PTC in conjunction with gain of a new exon was approximately three times more frequent in the transition from spermatogonia to germ cells at the preleptotene stage compared with differentiation of neural progenitor cells (Fig. 3E). These results indicated that production of Mga SV with a PTC represented rather common alternative splicing of new exon inclusion in meiotic germ cells.
Inefficiency of NMD substantially contributes to accumulation of Mga SV transcripts in meiotic spermatocytes and spermatids
Notably, an in-frame terminating codon was present in exon 19a (Fig. 3D). In general, transcripts containing a PTC are subjected to rapid degradation by NMD30-32. However, meiotic and post-meiotic germ cells in the testis are rather defective for PTC-mediated NMD, but not long 3′-untranslated region (UTR)-mediated NMD33, 34. Therefore, we speculated that an inefficient background of PTC-mediated NMD substantially contributed to the specific presence of the Mga SV transcript in meiotic spermatocytes and round spermatids. To test this hypothesis, three germ cell populations [germline stem cells (GSCs), spermatocytes (SCs), and round spermatids (RSs)] and two types of non-germ cells [mouse embryonic fibroblasts (MEFs) and embryonic stem cells (ESCs)] were treated with cycloheximide (CHX) that inhibits NMD. Then, RNAs of these cells were used to quantify expression levels of Mga SV by semi-quantitative and quantitative PCRs after conversion to cDNAs (Fig. 4A and B). These analyses revealed that CHX treatment clearly augmented the amounts of Mga SV transcripts in ESCs, MEFs, and GSCs compared with untreated cells, which indicated that Mga SV mRNA produced in these cells was indeed subjected to degradation by NMD. However, no noticeable alterations in the amounts of Mga SV due to treatment with CHX were observed in spermatocytes and round spermatids, which was consistent with the notion of low PTC-mediated NMD activity in these cells. We also noted that the levels of Mga SV mRNAs in CHX-treated ESCs and MEFs were significantly lower than those in spermatocytes and round spermatids, whereas CHX-treated GSCs showed almost equivalent levels of Mga SV mRNAs as those in spermatocytes and spermatids. Therefore, these results indicated that specific accumulation of Mga SV mRNAs in spermatocytes and round spermatids represented the combined consequence of two independent phenomena, i.e., preferential production of Mga SV mRNA in germ cells including spermatogonia and attenuated NMD activity against PTC-containing mRNAs in meiotic germ cells.
Dominant negative effect of carboxy-terminally truncated MGA on PRC1.6
To explore the possible function of MGA SV, we first examined the ability of the protein to interact with other PRC1.6 components by coimmunoprecipitation analyses (Fig. 5A). These analyses revealed no noticeable difference in the efficiency of the interaction with endogenous PCGF6, HP1g, and RING1B in HEK293FT cells between flag-tagged canonical MGA and its derivative, i.e., MGA SV, which were forcedly produced by transient transfection. Our data also demonstrated that both types of MGA proteins did not bind to SUZ12, a component of PRC2. In addition, we confirmed that canonical MGA, but not MGA SV, interacted with MAX efficiently as expected. Next, we examined the effect of forced production of MGA SV on interactions of PRC1.6 components with genomic DNA. To this end, we also used the HEK293FT cell line. The advantage of using the HEK293FT cell line was complete disruption of Mga loci to eliminate the contribution of endogenous MGA in these cells without affecting their viability. In fact, we generated Mga-KO HEK293FT cells using the CRISPR-Cas9 system (Supplementary Figure 4) and used them to examine the interaction of overexpressed canonical and anomalous MGAs with the promoters of PRC1.6 target genes (CCND2, CDIP, and CNTD1) whose repression is crucially dependent on the bHLHZ domain of MGA in these cells17. These analyses revealed that canonical MGA bound much more efficiently to all three gene promoters than MGA SV in Mga-null HEK293FT cells (Fig. 5B). At present, we do not know why the interaction scores obtained with MGA SV were not sufficiently low enough to judge as background. However, it is possible that these data represent binding of MGA SV to DNA using its T-box domain and/or other components of PRC1.6., i.e., E2F6/DP1 and L3MBTL2, which can lead to direct and indirect binding of the complex to DNA, respectively. Our analyses also revealed that binding of MAX, RING1B, and PCGF6 to these gene promoters was much less efficient with forced expression of Mga SV than that with forcedly produced canonical MGA (Fig. 5C), which further validated that MGA SV is defective in functioning as an active component of PRC1.6.
Next, we used mouse ESCs in which PRC1.6 functions as a blocker of ectopic meiosis17, 20-22. We transiently introduced expression vectors for canonical Mga and Mga SV individually and then examined alterations in the expression levels of meiosis-related genes. Our analyses revealed that some meiosis-related genes (Sycp3, Sycp1, Hormad1, Dazl, and Meiosin) indeed showed significant elevation in their expression levels after overexpression of Mga SV, but not canonical Mga. However, other PRC1.6 target genes (Tex12 and Tdrkh) were not activated, but further repressed by forced expression of Mga SV and canonical Mga, while the Rec8 gene, a meiosis-related gene that is not subjected to PRC1.6-dependent regulation, did not show appreciable alterations in its expression levels after overexpression of canonical Mga or Mga SV (Fig. 5D). Because MGA SV lacks the bHLHZ domain, but carries an intact T-box domain, these data were in accordance with our recent observation that repression of Tex12 and Tdrkh genes is dependent on the integrity of the T-box domain of MGA, while the bHLHZ domain of MGA is crucially involved in repressing the expression of Sycp3, Sycp1, Hormad1, Dazl, and Meiosin genes35. Next, we investigated physiological alterations in the expression levels of meiosis-related genes during the conversion of spermatogonia to preleptotene spermatocytes in male germ cells using publicly reported RNA sequence data. We found that meiosis-related genes whose repression was crucially dependent on the bHLHZ domain (Sycp3 and Sycp1) profoundly elevated their expression levels during this conversion, whereas those primarily subjected to T-box-dependent repression, such as Tdrkh and Tex12, were much less significantly activated (Supplementary Figure 5). These results suggested that de-repression of a subset of meiosis-related genes by forced expression of Mga SV in ESCs faithfully recapitulated the physiological activation profile of meiosis-related genes during meiotic onset.