Mettl3-/Mettl14-mediated mRNA N⁶-methyladenosine modulates murine spermatogenesis

Abstract

Spermatogenesis is a differentiation process during which diploid spermatogonial stem cells (SSCs) produce haploid spermatozoa. This highly specialized process is precisely controlled at the transcriptional, posttranscriptional, and translational levels. Here we report that N⁶-methyladenosine (m⁶A), an epitranscriptomic mark regulating gene expression, plays essential roles during spermatogenesis. We present comprehensive m⁶A mRNA methylomes of mouse spermatogenic cells from five developmental stages: undifferentiated spermatogonia, type A₁ spermatogonia, preleptotene spermatocytes, pachytene/diplotene spermatocytes, and round spermatids. Germ cell-specific inactivation of the m⁶A RNA methyltransferase Mettl3 or Mettl14 with V asa -Cre causes loss of m⁶A and depletion of SSCs. m⁶A depletion dysregulates translation of transcripts that are required for SSC proliferation/differentiation. Combined deletion of Mettl3 and Mettl14 in advanced germ cells with Stra8-GFPCre disrupts spermiogenesis, whereas mice with single deletion of either Mettl3 or Mettl14 in advanced germ cells show normal spermatogenesis. The spermatids from double-mutant mice exhibit impaired translation of haploid-specific genes that are essential for spermiogenesis. This study highlights crucial roles of mRNA m⁶A modification in germline development, potentially ensuring coordinated translation at different stages of spermatogenesis.

Introduction

Spermatogenesis is a complex developmental process that typically consists of three stages: mitosis, meiosis, and spermiogenesis^1,2. In mammals, SSCs either self-renew or undergo repeated mitotic divisions to generate A paired (A_pr) and A aligned (A_al) spermatogonia³. The A_al spermatogonia differentiate into A₁ spermatogonia and subsequently form type B spermatogonia through a series of mitotic divisions. The type B spermatogonia give rise to primary spermatocytes that enter the meiotic phase, in which reduced division and genetic recombination occur, resulting in the formation of haploid round spermatids. Next, during spermiogenesis, round spermatids undergo extensive morphological and biochemical transformations to differentiate into elongated mature spermatozoa. These processes are precisely controlled by complex regulatory programs at the transcriptional, posttranscriptional, and translational levels. These programs direct the proper expression of specific sets of genes at different developmental stages⁴. For example, transcription ceases in elongated spermatids due to chromatin compaction. All mRNAs required for the late stage of spermiogenesis are transcribed in spermatocytes and round spermatids and stored steadily until they are needed⁴. Posttranscriptional and translational controls of stored mRNAs, therefore, ensure timely synthesis of proteins essential for the transcriptionally silent spermatids⁴. Although posttranscriptional and translational regulation are important for the appropriate expression of specific genes in spermatogenesis, limited knowledge exists about how this regulation is achieved in male germ cells. Thus, mammalian spermatogenesis provides a powerful system for studying gene regulation at the posttranscriptional and translational levels.

Emerging evidence has shown that N⁶-methyladenosine (m⁶A), the most prevalent mammalian internal mRNA modification, is implicated in regulation of nearly every aspect of the mRNA life cycle, including pre-mRNA splicing, mRNA export, stability, and translation^5,6,7,8,9,10, and is thus crucial for various cellular, developmental, and disease processes such as heat shock response¹¹, DNA repair¹², circadian rhythm¹³, maternal to zygotic transition¹⁴, and tumorigenesis^15,16,17. In mammals, the m⁶A modification is catalyzed by a multicomponent methyltransferase complex that includes methyltransferase-like 3 (METTL3)¹⁸, METTL14¹⁹, and Wilms' tumor 1-associated protein (WTAP)²⁰, and can be demethylated into adenosine by two known demethylases: fat mass and obesity-associated factor (FTO)²¹, and AlkB homologue 5 (ALKBH5)²². Several studies have shown that depletion of Mettl3 and Mettl14 (or their homologs in other species) caused a block in embryonic stem cell self-renewal and differentiation^23,24, embryonic developmental defects, sex reversal^25,26, and impaired gametogenesis^22,27,28 in diverse organisms. Because m⁶A is a newly discovered mechanism to coordinate translation and turnover of eukaryotic transcripts, we decided to study whether m⁶A on mRNA may play critical roles to ensure proper regulation of genes in mammalian spermatogenesis at the posttranscriptional and translational levels.

Here we show that m⁶A is dynamically regulated and plays crucial roles to shape gene expression in SSC development and during spermatogenesis. We reveal that lack of m⁶A by germ cell-specific inactivation of Mettl3 or Mettl14 results in SSC depletion due to significant changes in translational efficiency (TE). Double deletion of Mettl3 and Mettl14 in advanced germ cells leads to impaired spermiogenesis due to altered TE of m⁶A-containing transcripts. This study thus reveals m⁶A-dependent translation as a previously undefined mechanism that modulates protein synthesis in SSCs and in spermatids, highlighting a crucial role of m⁶A on mRNA in translational regulation, particularly of transcription-ceasing cells and in mammalian development.

Results

Germ cell-specific knockout of Mettl3 or Mettl14 causes loss of m⁶A, resulting in depletion of SSCs

To explore the roles of m⁶A in spermatogenesis, we first examined whether two m⁶A writers, METTL3 and METTL14, are expressed in mouse testes, and found that both proteins localize to the nucleus of male germ cells (Supplementary information, Figure S1A and S1B). We then generated a Mettl3-floxed line and a Mettl14-floxed line using a CRISPR/Cas9 system (Supplementary information, Figure S1C). Mettl3^f/ΔVasa-Cre (hereafter referred to as Mettl3-vKO) mice were obtained to specifically inactivate Mettl3 in male germ cells as early as embryonic day 15 (E15)²⁹ (Supplementary information, Figure S1C). Immunostaining confirmed the absence of METTL3 protein in the male germ cells (Supplementary information, Figure S2). Analysis of m⁶A levels with quantitative ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS) in purified mRNA from control and Mettl3-vKO THY1⁺ undifferentiated spermatogonia showed that Mettl3 deficiency significantly but incompletely decreased m⁶A levels by ∼ 70% (Figure 1A). Mettl3-vKO males are sterile. Histological analyses of the Mettl3-vKO testes showed clear defects in the development of SSCs (Figure 1B-1H, Supplementary information, Figure S3A-S3G). However, Mettl3-vKO testes contained normal numbers of gonocytes at birth. By 6 weeks after birth, histological staining and immunostaining showed that seminiferous epithelium in Mettl3-vKO testes were completely devoid of any germ cells, with only SOX9-positive Sertoli cells remaining (Figure 1C-1F). The number of undifferentiated spermatogonia (PLZF-positive cells) in Mettl3-vKO testes was similar to that in the controls up to P5 but significantly reduced by P7 (Supplementary information, Figure S3A-S3C). Moreover, there was little if any difference in apoptosis of PLZF-positive spermatogonia between control and Mettl3-vKO testes (Supplementary information, Figure S3D and S3E). Furthermore, we found that EdU incorporation was significantly increased in GFRα1-positive A_single (A_s) spermatogonia, the most primitive set of spermatogonia, from Mettl3-vKO testes compared to those from controls, indicating higher proliferation of SSCs after Mettl3 depletion (Supplementary information, Figure S3F and S3G). Consistent with this, A_s spermatogonia were lost in the Mettl3-vKO testes at the age of 4 weeks; however, A_aligned (A_al) spermatogonia, which are derived from A_s spermatogonia, remained, demonstrating that the exhaustion of SSC pool is possibly due to SSC excessive proliferation after Mettl3 deletion (Figure 1G and 1H). Consistent with the notion that m⁶A is a key marker to determine cell state as previously shown^14,23,30, loss of m⁶A upon Mettl3 deficiency results in the loss of SSCs, causing depletion of germ cells in the Mettl3-vKO mice.

Figure 1

Characterization of germ cell-specific Mettl3 mutants. (A) UPLC-MS/MS analysis of m⁶A percentage relative to adenosine in purified mRNA from the undifferentiated spermatogonia of controls and Mettl3-vKO mutants. Data are expressed as mean ± SD from two biological replicates. ^**P < 0.01, Student's t-test. (B) Gross morphology of representative testes from an adult control and age-matched Mettl3-vKO mutant. (C, D) H&E staining of control (C) and Mettl3-vKO (D) testes at age of 6 weeks. (E, F) Immunohistochemical staining for germ-cell marker GCNA (green), Sertoli cell marker SOX9 (red), and DAPI (blue) in sections of 6-week-old control (E) and Mettl3-vKO mutant (F) testes. (G, H) Whole-mount immunostaining of seminiferous tubules for GFRα1 (a marker for early stage of the undifferentiated spermatogonia, Green), PLZF (a marker for the undifferentiated spermatogonia, red), and DAPI (blue) in 4-week-old controls (G) and Mettl3-vKO mutants (H). White and yellow arrowheads indicate both GFRα1- and PLZF-positive representative A_s and A_p spermatogonia, respectively, in controls, whereas there are no A_s and even A_p spermatogonia in mutants. (Scale bar: 40 μm).

We also found that Mettl14^f/ΔVasa-Cre (hereafter referred to as Mettl14-vKO) males showed an equivalent decrease in m⁶A levels on mRNA from germ cells and exhibited very similar SSC defects as compared to the Mettl3-vKO mice (Figure 2A-2H), suggesting that ablation of a single methyltransferase is sufficient to elicit SSC phenotypes. To test whether Mettl3 and Mettl14 could compensate for each other in SSCs, we generated double-mutant mice. We found the homozygous mutants (Mettl3^fΔ/Mettl14^f/ΔVasa-Cre, hereafter referred to as Mettls-vKO) displayed similar phenotypes to the Mettl3 or Mettl14 single-mutants (Supplementary information, Figure S4), suggesting there is no obvious compensative effect between Mettl3 and Mettl14. Collectively, these findings reveal that Mettl3 and Mettl14 control similar physiological processes in SSCs, consistent with the discovery that formation of a methyltransferase complex is required for m⁶A deposition³¹.

Figure 2

Characterization of germ cell-specific Mettl14 mutants. (A) UPLC-MS/MS analysis of m⁶A percentage relative to adenosine in purified mRNA from the undifferentiated spermatogonia of controls and Mettl14-vKO mutants. Data are expressed as mean ± SD from two biological replicates. ^**P < 0.01, Student's t-test. (B) Gross morphology of representative testes from an adult control and age-matched Mettl14-vKO mutant. (C, D) H&E staining of control (C) and Mettl14-vKO (D) testes at age of 6 weeks. (E, F) Immunohistochemical staining for the undifferentiated spermatogonia marker PLZF (red) and DAPI (blue) in sections of 6-week-old control (E) and Mettl14-vKO mutant (F) testes. (G, H) Whole-mount immunostaining of seminiferous tubules for GFRα1 (green), PLZF (red), and DAPI (blue) in 4-week-old controls (G) and Mettl14-vKO mutants (H). White and yellow arrowheads indicate both GFRα1- and PLZF-positive representative A_s and A_p spermatogonia, respectively, in controls (G). There is no A_s, even A_p spermatogonia in Mettl14-vKO mutants (H). (Scale bars: 40 μm.)

Combination deletion of Mettl3 and Mettl14 leads to impaired spermiogenesis

We next asked whether m⁶A plays a role in meiosis and/or spermiogenesis. To test this, we first conditionally deleted either Mettl3 or Mettl14 in advanced germ cells with Stra8-GFPCre, respectively. Stra8-GFPCre induces recombination starting from type A₁ spermatogonia (before meiosis). We found that Mettl3^f/ΔStra8-GFPCre (hereafter referred to as Mettl3-sKO), and Mettl14^f/Δ Stra8-GFPCre (hereafter referred to as Mettl14-sKO) male mice are fertile. Histological and immunohistochemical analysis showed normal spermatogenesis occurring in both Mettl3-sKO and Mettl14-sKO mice (Supplementary information, Figure S5A-S5O). Immunostaining verified that either METTL3 or METTL14 protein was absent in the respective mutant advanced germ cells, revealing high KO efficiency in both lines (Supplementary information, Figure S6A and S6B). By UPLC-MS/MS, we observed a 55-65% decrease in m⁶A levels in Mettl3-sKO or Mettl14-sKO pachytene spermatocytes, and a 45% decrease in round spermatids relative to controls (Figure 3A and Supplementary information, Figure S7D).

Figure 3

Analysis of advanced germ cell-specific Mettl3 and Mettl14 double-mutants. (A) m⁶A LC-MS/MS quantification in spermatid of control and different knockout mice. The data show the mean ± SD of two biological replicates, ^**P < 0.01, Student's t-test. (B) Morphology of testes and epididymis, from controls and Mettls-sKO double mutants. (Scale bars: 40 μm). (C) Number, total, and progressive motility of caudal epididymal sperm from 2-month-old control and Mettls-sKO double-mutant mice. Error bars represent SD, ^**P < 0.01, Student's t-test (n = 5-6). (D) Fluorescence staining of caudal epididymal sperms from controls and Mettls-sKO mutants with fluorescence dye-labeled peanut lectin (PNA, red) for acrosome, MitoTracker Green FM (green) for mitochondria, and DAPI (blue), respectively.

We then conditionally inactivated both Mettl3 and Mettl14 in advanced germ cells with Stra8-GFPCre. Testes from adult Mettl3^f/Δ/Mettl14^f/ΔStra8-GFPCre (hereafter referred to as Mettls-sKO) double-mutants were significantly smaller than control littermate testes (Supplementary information, Figure S7A and S7B). Immunostaining demonstrated the absence of METTL3 and METTL14 proteins in the advanced germ cells of Mettls-sKO testes (Supplementary information, Figure S7C). The m⁶A levels in spermatids of Mettls-sKO double mutants were significantly reduced compared to those in Mettl3 and Mettl14 single-mutant mice (Figure 3A). Interestingly, the differences in m⁶A levels between Mettls-sKO double-mutant and single-mutant spermatocytes were not significant (Supplementary information, Figure S7D). Histological analysis showed that seminiferous tubules contained very few mature spermatozoa in mutants (Figure 3B). Consistent with this, decreased sperm levels were found in Mettls-sKO epididymis compared to those in controls (Figure 3B). Sperm counting further showed that the number of caudal epididymal sperm in the Mettls-sKO mice was only about 2% of that in control mice (Figure 3C). Computer-assisted sperm analysis (CASA) revealed that sperm motility was severely destroyed in the Mettls-sKO mice as compared to controls, indicating defects in sperm flagella (Figure 3C). More than 80% of the caudal epididymal sperm from Mettls-sKO mice displayed abnormal heads (Figure 3D; Supplementary information, Figure S7E and S7F). Altogether, these abnormalities resemble those found in human oligo-astheno-teratozoospermia (OAT) syndrome. Given that yeast METTL3 orthologue ime4 is shown to be critical for meiosis^28,32,33,34, we expected a meiotic defect in Mettls-sKO mice. Unexpectedly, close examination of Mettls-sKO seminiferous tubules revealed no detectable abnormalities in germ cells of subsequent stages up to step 12 of elongating spermatids (Supplementary information, Figure S8), suggesting that the Mettls-sKO mice are normal in meiosis during spermatogenesis. The number of elongated spermatids (step 13 afterwards) dramatically decreased in the Mettls-sKO mice, and the heads of the Mettls-sKO mutant sperm were abnormal (Supplementary information, Figure S8). Thus, mRNA m⁶A modification produced by Mettl3 and Mettl14 is essential for spermatid differentiation in late stages of spermiogenesis. Given that mice with a conditional mutation for either Mettl3 or Mettl14 show normal meiosis and spermiogenesis, these results indicate that these two enzymes could have different or partially overlapping functions in late spermatogenesis. This observation also suggests the presence of additional factors that may mediate mRNA methylation, as suggested in a recent study³⁵. It will be of interest to test this hypothesis in the future.

Dynamic regulation of m⁶A in SSC development and during spermatogenesis

To elucidate the role of m⁶A in different stages of spermatogenesis, we first performed quantitative UPLC-MS/MS to monitor m⁶A levels on mRNA from six developmental stages of mouse spermatogenic cells: Thy1⁺ undifferentiated spermatogonia (including SSCs/progenitor cells), type A₁ spermatogonia, preleptotene spermatocytes, leptotene/zygotene spermatocytes, pachytene/diplotene spermatocytes, and round spermatids. We found that m⁶A was present in all mRNA samples tested, with particularly high enrichment in pachytene/diplotene spermatocytes and round spermatids (Figure 4A). To further characterize the dynamic nature of m⁶A in male germ cells during spermatogenesis, we performed m⁶A affinity purification and sequencing (m⁶A-seq) on purified mRNA from five developmental stages of mouse spermatogenic cells: Thy1⁺ undifferentiated spermatogonia (including SSCs/progenitor cells), type A₁ spermatogonia, preleptotene spermatocytes, pachytene/diplotene spermatocytes, and round spermatids⁶. From the above five samples, we identified 23 031, 16 392, 18 479, 15 656, and 10 950 m⁶A peak sites within the transcripts of 12 659, 12 167, 11 363, 11 293, and 12 491 genes, respectively, of which 9 092, 7 320, 7 471, 6 717, and 5 617 contained m⁶A (Supplementary information, Figure S9A and Supplementary information, Table S1). Consistent with previous findings^6,7, we found that m⁶A was present predominantly on its consensus motif of DRACH (D = A,G,U; R = A,G; H = A,C,U) at all five stages (Supplementary information, Figure S9B). m⁶A was distributed throughout mRNA transcripts, with increased read density in the CDS and stop codon (Figure 4B). The overall distribution of m⁶A peaks shifted dynamically between the five cell types. Notably, m⁶A read density in the CDS of pachytene/diplotene spermatocytes and round spermatids was greater than that of undifferentiated spermatogonia, A1 spermatogonia, and preleptotene spermatogonia, reflecting the UPLC-MS/MS findings (Figure 4A and 4B).

Figure 4

Dynamic change of m⁶A during spermatogenesis. (A) m⁶A LC–MS/MS quantification in six different developmental stages of spermatogonial cell. Un.S, undifferentiated spermatogonia; A1, type A1 spermatogonia; Prel, preleptotene spermatocytes; L/Z, lepotene/zygotene spermatocytes; Pa, Pachyotene/diplotene spermatocytes; Spd, round spermatids. The data show the mean ± SD of two biological replicates, ^**P < 0.01, Student's t-test. (B) Metagene distribution of m⁶A read density measured by m⁶A-seq depicting the subtranscript distribution pattern of m⁶A sites within the transcriptome of five different stages of spermatogenic cells. (C) RNA expression of transcripts with “emerging” or “resolving” peaks compared to unmethylated transcripts in five different stages of spermatogenic cells. A, undifferentiated spermatogonia; B, type A1 spermatogonia; C, preleptotene spermatocytes; D, pachytene/diplotene spermatocytes; E, round spermatids. “Emerging” peaks, m⁶A peaks with greater enrichment (enrichment ratio > 2) upon differentiation from a previous stage; “Resolving” peaks, m⁶A peaks with less enrichment (enrichment ratio < 0.5) upon differentiation to the next stage.

Next, we asked whether the presence of m⁶A on a transcript affects its expression level. We identified “emerging” peaks that harbor greater enrichment in m⁶A upon differentiation from a previous stage, as well as “resolving” peaks depleted in m⁶A upon differentiation to the next stage. In both cases, methylated genes demonstrated greater expression compared to unmethylated genes; transcripts had greater expression levels when they developed “emerging” peaks, as well as when they contained “resolving” peaks (Figure 4C and Supplementary information, Figure S9C). Taken together, our results reveal that m⁶A RNA modification is conserved and dynamically regulated during male germ cell development, suggesting its critical roles in spermatogenesis.

Knockout of either Mettl3 or Mettl14 results in dysregulation of translation in SSCs

Gene ontology (GO) analyses of m⁶A methylome revealed that genes with their transcripts methylated fall into diverse functional groups (Supplementary information, Table S2). Notably, genes identified in each type of spermatogenic cells are highly enriched for their specific functions (Supplementary information, Table S3). We found that the majority of transcripts of genes that are reported to be required for SSC/progenitor cell proliferation and differentiation were methylated in the undifferentiated spermatogonia and type A₁ spermatogonia, including Dnmt3b³⁶, Foxo1³⁷, Id4³⁸, Kit³⁹, Rptor⁴⁰, Sohlh2⁴¹, Sox3^42,43, Stat3⁴⁴, Stra8^45,46, and Zbtb16 (Plzf)^47,48 (Supplementary information, Table S3). Combined with SSC depletion in Mettl3 or Mettl14 single mutants, these data suggest that mRNA m⁶A modification could control SSC development.

To gain more comprehensive insight into the mechanisms underlying SSC depletion, we conducted high-throughput RNA sequencing (RNA-seq) and ribosome profiling assays to analyze the transcriptome and translatome of THY1⁺ undifferentiated spermatogonia from 5-day-old control, Mettl3-vKO, and Mettl14-vKO mutant testes, before overt morphological defects. The mutants had mild effects on transcript level (Supplementary information, Figure S10A). However, we identified 2 991 (1 416 up and 1 575 down) and 2 716 (1 258 up and 1 458 down) genes that were subjected to dysregulations of the TE (normalized read count of ribosome-protected fragments (RPFs)/mRNA fragments; Fold change > 2) upon either Mettl3 or Mettl14 deficiency, respectively, whereas the mRNA levels of most of these genes remained relatively stable (Figure 5A and 5B, Supplementary information, Table S4). Notably, these translationally dysregulated genes were exclusively enriched in those with transcripts bearing at least one m⁶A peak (Figure 5C; Supplementary information, Table S4). Interestingly, Mettl3-vKO and Mettl14-vKO mutant SSCs/progenitor cells shared significant overlaps of the translationally up- or down-regulated genes (Figure 5D; Supplementary information, Figure S10B and Supplementary information, Table S4), further demonstrating that the two methyltransferases are functionally similar in SSCs/progenitor cells. Nevertheless, we cannot completely exclude the possibility that Mettl3 may directly promote TE through interaction with the translation initiation machinery as shown in a recent study¹⁷.

Figure 5

The mRNA translation dysregulations in the THY1⁺ SSC/progenitor cells from the Mettl3 and Mettl14 single-mutants. (A, B) Scatter plots showing the fold changes of the RPF and mRNA of the genes in the THY1⁺ SSC/progenitor cells upon Mettl3 (A) or Mettl14 (B) knockout. (C) Different proportions of the transcripts with m⁶A modification in the genes with or without differential TE in the THY1⁺ SSC/progenitor cells upon Mettl3 (left) or Mettl14 (right) knockout. The P-value of such difference was calculated with the Fisher's exact test. (D) Overlaps between the genes that are translationally up- (top) or down- (bottom) regulated upon Mettl3 and Mettl14 knockout. P < 0.05. (E) Heat maps showing the relative levels of the mRNA and RPF read counts of the selected genes, which are known to be involved in spermatogenesis, in the THY1⁺ SSC/progenitor cells from the WT, Mettl3-vKO, and Mettl14-vKO mice.

Importantly, the above methylated transcripts whose TE is significantly affected in mutant compared to control SSCs/progenitor cells include multiple genes essential for proliferation and differentiation (Figure 5E; Supplementary information, Figure S10C and S10D; Supplementary information, Table S3). Among these genes, Sohlh2 is known to control SSC/progenitor cell differentiation through repressing expression of genes implicated in SSC maintenance and inducing expression of genes involved in differentiation (Supplementary information, Figure S10D)⁴¹. Forced expression of Dnmt3b in the undifferentiated spermatogonia causes SSCs/progenitor cells to exit the undifferentiated status toward differentiation³⁶. Consistent with hyperproliferation of SSCs in both mutants, TE for DNA replication factors such as MCM family proteins was significantly increased in mutants. Conversely, TE for Id4, which is required for SSCs³⁸, was significantly decreased in both mutants (Supplementary information, Figure S10D). Taken together, translational dysregulation of SSC/progenitor cell proliferation and differentiation regulators caused by disruption of the methyltransferase complex could lead to SSC phenotypes observed in Mettl3 or Mettl14 single mutants.

m⁶A methyltransferases regulate TE of methylated mRNAs for spermiogenesis

To determine the roles of m⁶A in spermatocytes and spermatids, we first performed GO analysis on genes containing “emerging” and “resolving” m⁶A peaks in pachytene/diplotene spermatocytes and found that these genes are important for spermiogenesis, including cilium morphogenesis, cell projection organization, and CatSper complex formation, indicating that the methylated mRNAs transcribed in spermatocytes may be stored until they are required in elongating spermatid (Supplementary information, Table S5). Notably, most mRNAs that encode proteins previously reported to be essential for spermiogenesis were highly methylated in round spermatids and pachytene spermatocytes, respectively (Supplementary information, Table S3).

We next isolated pachytene spermatocytes and round spermatids from controls and Mettls-sKO mutants, and performed RNA-seq and ribosome profiling for transcriptome and translatome analysis. In line with the observed mild changes in mRNA levels of either Mettl3 or Mettl14 mutant SSCs/progenitor cells, combined deletion of both enzymes resulted in slight effects on mRNA levels in spermatocytes (n = 357 differentially expressed genes; P-value < 0.05) and round spermatids (n = 265 differentially expressed genes; P-value < 0.05) as compared to controls (Supplementary information, Figure S11A and S11B).

We then compared the gene TE between control and Mettls-sKO mutant spermatocytes, and round spermatids. We identified that 1 051 genes (556 up and 495 down) exhibited significant TE changes (P-value < 0.05 and TE change > 50%) in mutant spermatids compared to controls, while 1 287 genes (706 up and 581 down) showed significant TE changes in mutant spermatocytes (Figure 6A, Supplementary information, Figure S11C, Tables S6 and S7). Compared with genes that showed no change in TE, genes whose TE was significantly affected upon combined deletion of Mettl3 and Mettl14 showed significant enrichment of transcript methylation (Figure 6B, Supplementary information, Figure S11D, Tables S6 and S7).

Figure 6

The mRNA translation dysregulations in spermatids from the Mettl3 and Mettl14 double mutants. (A) Scatter plot showing the fold changes of the RPF and mRNA of the genes in spermatids from the Mettl3 and Mettl14 double-mutant mice. (B) Different proportions of the transcripts with m⁶A modification in the genes with or without differential TE in spermatids upon Mettl3 and Mettl14 double mutation. The P-value of such difference was calculated with the Fisher's exact test. (C) Heat maps showing the relative levels of the mRNA and RPF read counts of the selected genes, which are known to be involved in spermatogenesis (spermatids from the WT and Mettls-sKO mice).

Notably, in mutant spermatids, many genes with down-regulated TE are previously reported to be essential for spermiogenesis, including Brd7⁴⁹, Cstf2t⁵⁰, Jmjd1c⁵¹, Parp11⁵², Lmtk2⁵³, and Tdrd12⁵⁴ (Figure 6C). Among these genes, Cstf2t knockout (KO) resulted in defects resembling those found in OAT, as well as in Mettls-sKO mutants (Supplementary information, Figure S11E)⁵⁰. Thus, translational inhibition of the key methylated transcripts of genes for spermiogenesis upon combined deletion of Mettl3 and Mettl14 may result in the observed OAT phenotypes in Mettls-sKO mutants. Interestingly, in mutant spermatocytes, genes with TE up-regulated include many factors for DNA replication and repair, which are required for early stage of spermatocytes but unnecessary for pachytenes (Supplementary information, Table S7), suggesting that m⁶A may repress translation of overexpressed transcripts to prevent protein overproduction. Accordingly, in male germ cells, overexpression of Gfer, one of the genes with TE up-regulated, has been reported to cause male infertility⁵⁵. Taken together, these results support the idea that m⁶A modification could serve as a mark to promote translation for producing proteins essential for spermiogenesis, and inhibit translation for preventing deleterious consequences of overproducing proteins during late spermatogenesis.

Consistent with this hypothesis, we found that, according to the effect of m⁶A on translation, the methylated transcripts could be divided into translation stimulated and repressed groups in mutant spermatids and spermatocytes. For example, in mutant spermatids, methylated transcripts with TE down-regulated are enriched for microtubule-based process, whereas genes with up-regulated TE are overpresented in metabolism and chromatin modification (Supplementary information, Figure S12A and S12B). In mutant spermatocytes, methylated transcripts with TE down-regulated are enriched for cilium formation, whereas the ones with TE up-regulated feature DNA metabolism and chromatin organization (Supplementary information, Figure S13A and S13B).

Discussion

Despite the recent extensive interest, dynamic regulation and functional roles of m⁶A in mammals, particularly in mammalian development, remain largely uncharacterized. Here we present a comprehensive m⁶A methylome landscape of developing male germ cells, as well as a detailed in vivo characterization of m⁶A biogenesis and functions during mammalian spermatogenesis, providing new insights into spermatogenesis.

Our analyses of the m⁶A methylome of spermatogenic cells at different developmental stages uncover transcripts that harbor extensive cell developmental stage-dependent common or unique m⁶A RNA modification, and reveal the dynamic regulation of m⁶A sites and the positive correlation of methylation with developmental stage-specific transcripts during spermatogenesis. These results will provide clues for further functional studies of m⁶A RNA modification in germline development.

Our results reveal an essential role of m⁶A RNA modification in the maintenance of SSC homeostasis. Early germ cell-specific inactivation of either of the methyltransferases leads to loss of m⁶A and impairs SSC/progenitor cell quiescence, resulting in rapid exhaustion of the SSC pool and complete germ cell-loss phenotype. m⁶A profiling indicated that most key regulators of SSC/progenitor cells harbor extensive m⁶A RNA modification, including Plzf, Id4, Dnmt3b, and Sohlh2. Despite subtle changes in mRNA levels, loss of m⁶A upon either Mettl3 or Mettl14 deficiency caused translational dysregulation of those key regulators in SSC/progenitor cells. We propose that Mettl3 or Mettl14 maintain SSC/progenitor cell homeostasis through methylating transcripts of key regulators governing SSC proliferation and differentiation.

During the late stages of spermiogenesis, the transcripts inherited from spermatocytes and round spermatids could govern synthesis of proteins required for sperm development. Our analyses show heavy enrichment of m⁶A in spermatids and spermatocytes, as well as substantial enrichment of the methylated transcripts in genes essential for spermiogenesis, revealing that m⁶A-dependent RNA translation may act as a mechanism to control late stages of spermiogenesis. Consistent with this, conditional KO of both Mettl3 and Mettl14 caused translational downregulation of the key m⁶A-modified transcripts for spermiogenesis, resulting in the defects in late stages of spermiogenesis. Thus, our work supports the idea that m⁶A RNA modification could provide a mark to those transcripts, which in turn modulates their translation and storage during spermiogenesis.

Mice with conditional deletion of either Mettl3 or Mettl14 exhibit complete loss of SSCs, but show normal meiosis and spermatogenesis, indicating these enzymes could have different or partially overlapping functions in late stages of spermatogenesis. Yeast METTL3 orthologue ime4 were shown to be essential for meiosis^28,32,33,34. In contrast to these studies, our results argue against an indispensable role for Mettl3 and Mettl14 in mouse meiosis because ablation of both genes in advanced germ cells only disrupts spermiogenesis without affecting meiosis. However, substantial enrichment of m⁶A within meiosis-associated transcripts and dynamic changes of the methylated sites on these transcripts in different developmental stages of meiotic cells support the idea that m⁶A RNA modification could regulate mouse meiosis, raising the question of whether more methyltransferases in addition to Mettl3 and Mettl14 contribute to the m⁶A deposition on transcripts during mammalian meiosis. Accordingly, combined KO of both Mettl3 and Mettl14 did not significantly reduce m⁶A levels in spermatocytes compared to single mutants. Thus, it is reasonable to speculate that diverse methyltransferases are present in cells in order to offer multiple ways to decorate transcripts with m⁶A responsible for key biological functions. It would be interesting to further investigate whether a similar mechanism operates in other systems.

Collectively, our data indicate mRNA m⁶A modification as a critical regulator to control the timely translation of groups of transcripts to coordinate proper production of proteins, which is essential for mammalian spermatogenesis. In the absence of either methyltransferase in early male germ cells, loss of m⁶A leads to dysregulated translation of SSC/progenitor cell proliferation and differentiation factors, causing SSC depletion. Furthermore, Mettl3/Mettl14 double mutants, but not single mutants, showed disrupted spermatid differentiation due to dysregulated translation of key factors for spermiogenesis, despite normal meiosis, which may suggest potential functional redundancy of these methyltransferase components in the late stages of the spermatogenesis program. Moreover, in addition to a role of m⁶A in mRNA stability, our analyses indicate that m⁶A on mRNA could provide an identity to transcripts for their coordinated translation, thus directing the proper expression of stage-specific genes during late spermatogenesis. We postulate that the precise effect of m⁶A on translation might depend on its recognition by different reading mechanisms. It will be of interest to explore this hypothesis in the future.

Materials and Methods

Mice

The conditional mutant alleles for Mettl3 and Mettl14, and the Stra8-GFPCre knockin mouse line were generated by the CRISPR/Cas9 technology. To generate a Mettl3-floxed line, in which exon 4 of the Mettl3 allele is flanked by loxP sites, two independent guide RNAs targeting Mettl3 introns 3 and 4 were designed. The donor vector containing exon 4 flanked by two loxP sites and two homology arms was used as a template. Two founder mice containing floxed exon 4 of Mettl3 on the same allele were obtained. Using similar procedure, we obtained two founder mice containing floxed exon 2 of Mettl14 on the same allele. Resulting founder male mice were mated to WT C57BL/6J (B6) female mice to obtain heterozygous Mettl3-floxed and Mettl14-floxed mice, respectively. Progeny were screened by PCR for germ line transmission of the targeted alleles. For conditional deletion of Mettl3 or Mettl14 in advanced germ cells, we established a Stra8-GFPCre knockin mouse line. A cDNA encoding the GFPCre fusion protein was inserted into the last coding exon of Stra8, and a 2A peptide sequence was included to link Stra8 and GFPCre to allow expression of both genes. The Stra8-GFPCre lines were generated by Shanghai Biomodel Organism Co., Ltd. All mice described above were maintained on the C57BL/6J (B6) background. Mettl3- and Mettl14-floxed mice (Mettl3^flox/flox and Mettl14^flox/flox) were then bred with germ cell-specific expressed Cre mice including Vasa-Cre mouse line (Jackson Laboratory, Bar Harbor, Maine, USA) and Stra8-GFPCre mouse line for excising the loxP-flanked exon 4 and exon 2 to generate germ cell-specific Mettl3 and Mettl14KO mice, respectively. Germ cell-specific Mettl3 and Mettl14 double KO mice were obtained by crossing Mettl3^flox/floxMettl14^flox/flox with Mettl3^flox/+Mettl14^flox/+ or Mettl3^flox/+Mettl14^flox/flox carried germ cell-specific expressed Cre mice. All of the primers for PCR genotyping were listed in Supplementary information, Table S8. All animal experiments were conducted in accordance with the guidelines in the Animal Care and Use Committee at Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Science.

Isolation of spermatogenic cells

We used male mice in B6 background for preparations of different types of spermatogenic cells from undifferentiated spermatogonia, type A₁ spermatogonia, preleptotene spermatocyte, pachytene/diplotene spermatocyte, and round spermatid. Isolation of the undifferentiated spermatogonia, THY1⁺ spermatogonia, from 5-7-day-old mice was carried out using magnetic-activated cell sorting (MACS) as described previously⁵⁶. Stra8-GFPCre mice were used for synchronous spermatogenesis to generate type A₁ spermatogonia and preleptotene spermatocytes, respectively. Spermatogenesis was synchronized as previously described with modifications⁵⁷. Briefly, 2-dpp Stra8-GFPCre mice were pipette fed with 100 μg/g body weight WIN 18 446 (Sigma), suspended in 1% gum tragacanth, for 7 consecutive days. At Day 8 of WIN 18 446 treatments, these animals received an i.p. injection of RA (Sigma; 30 μg/g body weight) in dimethyl sulfoxide (DMSO), and were then left to recover for 24 h. Type A₁ spermatogonia were then collected based on their GFP fluorescence label using fluorescence-activated cell sorting (FACS) (Becton Dickinson). Spermatocytes and round spermatid from adult mice were collected based on Ho33342/PI staining using FACS as previously described⁵⁸.

Histological and immunohistochemical analysis

Testes were fixed in Bouin's buffer or 4% paraformaldehyde (PFA), embedded in paraffin and sectioned. Sections were deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E). For immunofluorescence analysis, sections were boiled in 10 mM sodium citrate buffer (pH 6.0) for 15 min, brought to room temperature, washed in PBS with 0.1% Triton X-100. The sections were then blocked with blocking buffer (10% donkey serum and 0.1% Triton X-100 in PBS) for 60 min at room temperature, and later incubated with the primary antibodies in blocking buffer overnight at 4 °C. The following primary antibodies were used in this study: rabbit anti-METTL3(1:200; Abcam), rabbit anti-METTL14(1:200; Sigma), rabbit anti-PLZF (1:100; Santa Cruz Biotechnology), goat anti-LIN28a (1:200; R&D Systems), goat anti-GFRα1 (1:50; R&D Systems), rabbit anti-SOX9 (1:200; Millipore), rabbit anti-STRA8 (1:200; gift from Michael Griswold, Washington State University, Pullman, WA), mouse anti-γH2AX (1:200; Millipore), rat anti-GCNA IgM (1:50; kindly provided by Dr G Enders, University of Kansas, Kansas City, KS), rabbit anti-VASA (1:100; Abcam), rabbit anti-DMC1 (1:100; Santa Cruz Biotechnology), and FITC-conjugated peanut agglutinin (1:500; Sigma). On the following day, slides were washed four times for 15 min in PBS with 0.1% Triton X-100, and Alexa Fluor 488- and Alexa Fluor 594-conjugated donkey secondary antibody (Jackson ImmunoResearch Laboratories) were then added at a 1:500 dilution. After 60 min at room temperature, the sections were washed in PBS, rinsed quickly in pure ethanol, mounted in Prolong Gold Antifade medium with DAPI (Molecular Probes), and then analyzed by fluorescence microscopy (Olympus) or confocal microscopy (Olympus). Apoptotic cells were detected using an In Situ Cell Death Detection Kit, Fluorescein (Roche Applied Science) according to the manufacturer's instructions.

Whole-mount immunohistochemistry

Mouse testes were removed from the tunica albuginea, and untangled seminiferous tubules were fixed in 4% PFA with 0.5 mM CaCl₂, and PBS on ice for 4 h. The seminiferous tubules were washed in PBS with 0.2% NP40 (Sigma) for 20 min, and dehydrated through a graded methanol series (25, 50, 75, and 100%) in PBS containing 0.1% Tween 20 (Sigma) (PBST) on ice for 1 h each. After rehydration in PBST for 5 min twice, the tubules were blocked in blocking buffer (1% bovine serum albumin (BSA) and 4% donkey serum in PBST) for 1 h, and incubated with primary antibodies against GFRA1 (1:50), LIN28A (1:200), and PLZF (1:100) in blocking buffer at 4 °C overnight. After washed in PBST, the tubules were incubated with Alexa Fluor 488- and Alexa Fluor 594-conjugated donkey secondary antibody (Jackson ImmunoResearch) for 2 h at room temperature. The tubules were then washed in PBST, mounted, and observed using confocal microscopy (Olympus).

For EdU labeling, as described previously, mice were i.p. injected with EdU (Invitrogen) (50 μg/g body weight) in PBS⁵⁹. The mice were killed 2 h later and testes were removed and treated either for sections or whole-mount staining. The samples were immunostained with primary antibody first, and then detected for the EdU incorporation by Click-It EdU Alexa Fluor 594 Imaging Kit according to the manufacturer's protocol (Invitrogen).

CASA

Cauda epididymides from control and Mettls-sKO double mutants were minced at 37 °C in 1 ml of Dulbecco Modified Eagle Medium (Invitrogen) supplemented with 3% BSA. After 5 min at 37 °C, tissue was removed, and aliquots of sperm suspension were diluted with fresh medium to adjust an approximate concentration to 6 million/ml, and CASA was performed using the HTM-IVOS system (Version 10.8; Hamilton-Thorne Research). At least 1 000 spermatozoa and 10 fields were assessed for each specimen (n = 5-6 independent experiments), and the percentages of motile and progressively motile spermatozoa were determined.

RNA isolation

Total RNA isolation for UPLC-MS/MS analysis: total RNA was isolated with Trizol reagent (Invitrogen). mRNA was extracted using GenElute mRNA miniprep (Sigma-Aldrich) followed by further removal of contaminated rRNA by using RiboMinus Transcriptome Isolation Kit (Invitrogen) according to the manufacturer's instructions. mRNA concentration was measured by Qubit.

mRNA-seq

Total RNA was isolated from THY1⁺KIT⁻ spermatogonia, spermatocytes, and spermatids from control and germ-cell mutant mice using Trizol reagent (Invitrogen). RNA purification and libraries of cDNA were constructed by the Omics Core of CAS-MPG Partner Institute for Computational Biology at Shanghai Institutes for Biological Sciences using the TrueSeq Stranded Total RNA Library Prep Kit (Illumina) following manufacturer's instructions. Libraries were sequenced using single reads (100 nt) on an Illumina HiSeq 2000. Sequencing reads were mapped to the ENSEMBL Mouse Reference Genome (GRCm38 release 87) using STAR (version 2.5.1) with the following parameters: --alignEndsType EndToEnd--outFilterMismatchNmax1--outFilterMultimapNmax 5--outSAMtype BAM SortedByCoordinate. Read assignment and counting were achieved using HTSeq-count (version 0.7.2) in intersection-strict mode⁶⁰.

UPLC-MS/MS analysis of m⁶A levels

About 50-100 ng of purified mRNA was digested by nuclease P1 (1U; Sigma) in 20 μl of buffer containing 10 mM of NH₄Ac (pH 5.3) at 42 °C for 4 h. About 100 mM NH₄HCO₃ and alkaline phosphatase (0.5 U) were then added to the reaction for another incubation at 37 °C for 4 h. The digested sample was centrifuged at 4 °C, 13 000 rpm for 20 min and the supernatant was injected into UPLC-MS/MS. The nucleosides were separated by UPLC (SHIMADZU) equipped with ZORBAX SB-Aq column (Agilent), and detected with Triple Quad 5500 (AB SCIEX) in positive ion multiple reaction-monitoring (MRM) mode. Quantitation of modifications was based on nucleoside-to-base ion mass transitions: m/z268.0-136.0 for A, and m/z282.0-150.1 for m⁶A. Pure nucleosides were used to generate standard curves, from which the concentrations of A and m⁶A in the sample were calculated. The level of m⁶A was then calculated as a percentage of total unmodified A.

m⁶A-seq

m⁶A-seq was performed as previously described⁶. Briefly, spermatogonial cells were isolated as described above. PolyA mRNA was enriched using GenElute mRNA Miniprep Kit following the manufacturer's protocols. mRNA was sonicated to ∼100 nt, mixed with 2.5 mg affinity-purified anti-m⁶A polyclonal antibody (Synaptic Systems) in IP buffer (150 mM NaCl, 0.1% NP-40, and 10 mM Tris-HCl, pH 7.4), and incubated for 2 h at 4 °C. The antibody-RNA complex was isolated by incubation with protein A beads (Invitrogen) at 4 °C for 2 h. The beads were washed three times and eluted competitively with an m⁶A monophosphate solution. RNA in the eluate was isolated using RNA Clean and Concentrator (Zymo Research) and used for library preparation with TruSeq stranded mRNA sample preparation kit (Illumina).

Data analyses were performed as previously described^8,14. Briefly, after removing adapters, sequencing reads were aligned to the reference genome (mm10) using TopHat (v2.0.14)⁶¹. The longest isoform was used if multiple isoforms were detected. Aligned reads were extended to 100 nt (average fragment size) and converted from genome-based coordinates to isoform-based coordinates to eliminate interference from introns in peak calling. To call m⁶A peaks, the longest isoform of each human gene was scanned using a 100 nt sliding window with 10 nt steps. To reduce bias from potentially inaccurate gene structure annotation and the arbitrary usage of the longest isoform, windows with read counts < 1/20 of the top window in both m⁶A IP and input sample were excluded. For each gene, the read count in each window was normalized by the median count of all windows of that gene. A negative binomial model was used to identify the differential windows between IP and input samples by using the edgeR⁶². The window was called positive if FDR < 1% and log2 (enrichment score) ≥ 1. Overlapping positive windows were merged. The following four numbers were calculated to obtain the enrichment score of each peak (or window): (a) read count of the IP sample in the current peak/window, (b) median read count of the IP sample in all 100 nt windows on the current mRNA, (c) read count of the input sample in the current peak/window, and (d) median read count of the input sample in all 100 nt windows on the current mRNA. The enrichment score of each window was calculated as (a × d)/(b × c). When comparing the m⁶A profiles between different samples to determine “emerging” and “resolving” peaks, a peak (m⁶A IP/input > 2) was considered “enriched” if its enrichment ratio was > 2.

Ribosome profiling and total RNA sequencing in parallel

Cells were washed with prechilled PBS (with 100 μg/ml cycloheximide, Sigma-Aldrich) and then harvested by centrifugation at 4 °C (2 000 rpm, 5 min). Cell pellet of each sample was resolved in prechilled lysis buffer (20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM MgCl₂, 0.1% NP-40, 1% Triton X-100, 1 mM DTT, 25 U/ml of RNase-free DNAse I, and 100 μg/ml cycloheximide). After 10 min of incubation on ice with periodic agitation, the lysate was centrifuged for 10 min (20 000 × g at 4 °C), and the supernatant was collected.

The libraries for paralleled ribosome profiling and total RNA-seq were prepared as previously described^63,64. Briefly, the total RNA was isolated from 50 μl of the cell lysate using the Acid Phenol: chloroform extraction method and heat fragmented at 95 °C for 25 min. The ribosome protected RNA fragments (RPF) were obtained from 100 μl of the cell lysate treated with 60 U/A260 of RNase I, and further purified with Sephacryl S400 columns (GE Healthcare). Finally, the RPFs with the length of 25-35 nt were selected on a 15% Urea-Polyacrylamide gel. Next, the same procedure was followed to prepare the sequencing libraries from both the total RNA fragments and the RPFs. This includes 3′ adaptor ligation, reverse transcription, circularization of the cDNA, and PCR amplification^63,64. The libraries were then sequenced on the Illumina HiSeq 2 500 system with 50 cycles of single-end reading.

Ribosome profiling data analysis

The pre-processing procedure of the ribosome profiling and the paralleled RNA-seq data has been described previously⁶⁵. Specifically, the cutadapt program⁶⁶ was used to trim the 3′ adaptor in the raw reads of both mRNA and RPF. Low-quality reads with Phred quality scores lower than 25 (> 75% of bases) were removed using the fastx quality filter (http://hannonlab.cshl.edu/fastx_toolkit/). Next, sequencing reads originating from rRNAs were identified and discarded by aligning the reads to rRNA sequences of the particular species using Bowtie (version 1.1.2) with no mismatch allowed. The remaining reads were then mapped to the mouse genome (GRCm38 (Ensembl release 87)) and spliced transcripts using STAR (version 2.5.1) with the following parameters: --alignEndsType EndToEnd --outFilterMismatchNmax 1 --outFilterMultimapNmax 5 --outSAMtype BAM SortedByCoordinate. To control the noise from multiple alignments, reads mapped to multiple genomic positions were discarded.

For each gene, mRNA expression was estimated by the RNA-seq reads, which were counted using HTSeq-count (version 0.7.2)⁶⁰ in intersection-strict mode. For quantification of RPF, multiple filters were implemented on raw reads to reduce the technical noise of ribosome profiling and extract the reads originating from ribosome-binding and translating sequences in coding regions. First, RPF reads with length between 25 and 35 nt were deemed high quality and most likely to be from ribosome occupation in mammalian cells^67,68. Second, to reduce noise due to multiple alignments, only the reads uniquely mapped to the coding regions were counted as RPFs. Third, due to the potential accumulation of ribosomes around the starts and ends of coding regions^68,69, reads aligned to the first 15 and last 5 codons were excluded for the counting of RPFs. Finally, The Xtail package⁶⁵ was used to identify the genes subjected to differential translation. For the various gene sets selected by the different analysis, GO enrichment analysis was conducted using the tool Metascape. The GO terms with P-values < 0.001 were selected and imported into REVIGO, which visualizes the terms as nodes in a network⁷⁰. Each GO term was color-coded according to the P-value (−log10). The size of each node is proportional to the number of genes belong to the GO term, whereas the link between different terms represent the number of shared genes.

Author Contributions

M-HT conceived the project, and with CH and XY, designed the project and data analysis. M-HT, CH, and XY wrote the manuscript with contributions from all authors. ZL conducted phenotype analysis, and isolated spermatogonial cells and RNAs for all experiments in the project. PJH and ZL conducted m⁶A-seq and analysis. XX, JF, QZ, and WS conducted ribosome profiling experiments and data analysis. K-JZ, TZ, YZ assisted in cell isolation and phenotype analysis. XZ and GJ carried out m⁶A-LS/MS analysis. YZ conducted CASA assay.

Competing Financial Interests

The authors declare no competing financial interests.