DAZL mediates a broad translational program regulating expansion and differentiation of spermatogonial progenitors

Abstract
Introduction
Results
Discussion
Materials and methods
Appendix 1
Data availability
References
Article and author information
Metrics

Abstract

Fertility across metazoa requires the germline-specific DAZ family of RNA-binding proteins. Here we examine whether DAZL directly regulates progenitor spermatogonia using a conditional genetic mouse model and in vivo biochemical approaches combined with chemical synchronization of spermatogenesis. We find that the absence of Dazl impairs both expansion and differentiation of the spermatogonial progenitor population. In undifferentiated spermatogonia, DAZL binds the 3' UTRs of ~2,500 protein-coding genes. Some targets are known regulators of spermatogonial proliferation and differentiation while others are broadly expressed, dosage-sensitive factors that control transcription and RNA metabolism. DAZL binds 3' UTR sites conserved across vertebrates at a UGUU(U/A) motif. By assessing ribosome occupancy in undifferentiated spermatogonia, we find that DAZL increases translation of its targets. In total, DAZL orchestrates a broad translational program that amplifies protein levels of key spermatogonial and gene regulatory factors to promote the expansion and differentiation of progenitor spermatogonia.

Introduction

The germline-specific DAZ family of RNA-binding proteins functions in germ cell development across metazoa. This family is comprised of Y-linked DAZ and its autosomal homologs DAZL (DAZ-like) and BOULE, all of which contain a highly conserved RNA recognition motif (RRM) and at least one DAZ repeat. BOULE is widely conserved across metazoa from sea anemones through humans, while DAZL is limited to vertebrates, and DAZ is further limited to Old World monkeys and apes (Saxena et al., 1996; Xu et al., 2001). In humans, deletions encompassing the Y chromosome’s Azoospermia Factor C (AZFc) region, which contains all four copies of the DAZ gene, are among the most common known genetic causes of spermatogenic failure, accounting for 10% of cases of azoospermia (no sperm detected in semen) or severe oligozoospermia (abnormally low number of sperm detected in semen) in the absence of any physical obstruction (Ambulkar et al., 2015; Fu et al., 2012; Girardi et al., 1997; Mascarenhas et al., 2016; Nakahori et al., 1996; Reijo et al., 1995; Simoni et al., 1997; Vogt et al., 1996). However, the mechanistic basis for the DAZ family’s role in spermatogenesis remains poorly defined.

Dazl is expressed in the embryonic germ line as well as during adult oogenesis and spermatogenesis (Seligman and Page, 1998) and therefore likely functions at multiple stages of germline development. In mice, genetic loss of Dazl causes infertility in both sexes (Ruggiu et al., 1997). DAZL is first required during embryogenesis for germ cell determination (Gill et al., 2011; Lin and Page, 2005; Nicholls et al., 2019b). On mixed genetic backgrounds and the inbred 129 strain, a small number of Dazl-null germ cells survive into adulthood with defects in spermatogonial differentiation (A_al to A₁ transition) (Schrans-Stassen et al., 2001) and in meiosis I (Saunders et al., 2003). However, because these studies examined a complete knockout of Dazl, it is unclear whether these adult phenotypes are a primary defect due to loss of DAZL activity in the adult germ line or a secondary effect of the absence of DAZL during embryogenesis. To address this, one study conditionally deleted Dazl after its embryonic requirement and reported roles for DAZL in spermatogonial stem cell maintenance, meiosis I, and spermatid development (Li et al., 2019).

DAZL interacts with the 3' UTRs of factors that contribute to multiple stages of spermatogenesis, but it is not clear what transcripts DAZL targets in specific spermatogenic cell types because these targets have been identified in either cultured primordial germ cell-like cells (Chen et al., 2014), ovaries (Rosario et al., 2017), or whole testes containing many stages of spermatogenesis (Chen et al., 2014; Li et al., 2019; Zagore et al., 2018). DAZL functions as a translational enhancer within the oocyte (Collier et al., 2005; Reynolds et al., 2005; Reynolds et al., 2007; Sousa Martins et al., 2016). However, during spermatogenesis, and particularly in spermatogonia, there is conflicting evidence whether DAZL regulates translation (Li et al., 2019; Reynolds et al., 2005; Reynolds et al., 2007) or RNA stability (Zagore et al., 2018).

Here, we study mouse Dazl using genetic and biochemical tools to define a function for the DAZ family in spermatogenesis. Using a conditional mouse model, we find that DAZL promotes expansion and differentiation of progenitor spermatogonia. By combining chemical synchronization of spermatogenesis with iCLIP and translational profiling, we demonstrate that DAZL mediates a broad translational program spanning ~2,500 protein-coding genes, including spermatogonial and gene regulatory factors, in undifferentiated spermatogonia in vivo. We propose that these novel insights into how DAZL functions within undifferentiated spermatogonia illuminate DAZ’s role in human spermatogenesis.

Results

DAZL promotes expansion and differentiation of the spermatogonial progenitor population

To determine when DAZL could play a direct role in spermatogenesis, we defined DAZL protein expression in the postnatal male germ line. At birth, male germ cells are present as mitotically quiescent gonocytes (also known as prospermatogonia). At postnatal days (P) 0 and 4, we found that gonocytes robustly expressed Dazl, as marked by a Dazl:tdTomato reporter (Figure 1—figure supplement 1A).

Shortly after birth, gonocytes resume proliferation and mature into spermatogonia, including spermatogonial stem cells. These spermatogonia initiate spermatogenesis, which encompasses a series of transit-amplifying mitotic divisions, followed by meiosis and then the cellular differentiation of spermiogenesis to form spermatozoa. DAZL protein expression has been previously reported in multiple spermatogenic cell types, including spermatogonia, spermatocytes, and round spermatids (Li et al., 2019; Ruggiu et al., 1997; Xu et al., 2001). Here, we sought to clarify DAZL’s expression patterns via immunohistochemistry. DAZL protein was robustly expressed in type A, Intermediate, and type B spermatogonia as well as in spermatocytes in meiotic prophase I (Figure 1—figure supplement 1B–C). However, DAZL was not detected in secondary spermatocytes in meiotic prophase II, or in round or elongating spermatids, after the completion of meiosis. These data provide greater detail as to the spermatogenic cell types that express DAZL (Li et al., 2019; Ruggiu et al., 1997; Xu et al., 2001) and are at odds with a previous study reporting DAZL expression in round spermatids (Li et al., 2019), potentially reflecting differences in antibody specificity or sensitivity of detection.

Next, we investigated whether DAZL is required for spermatogonial development, independent of DAZL’s earlier requirement in germ cell determination (~E10.5-12.5). Using a conditional Dazl allele with germline-specific Cre recombinase Ddx4^Cre, which is active from ~E14.5 onward (Figure 1A), we generated Dazl conditional knockout mice (Dazl^2L/-; Ddx4^Cre^/+; referred to as Dazl cKO) alongside phenotypically wild-type control animals (Dazl^2L/+; Ddx4^Cre^/+). Germ cells persist in adult Dazl cKO testes, but these animals are infertile (Nicholls et al., 2019b). However, conditional deletion of Dazl in the adult was incomplete, and while some tubules lacked DAZL protein, other tubules continued to express DAZL (Nicholls et al., 2019b). Here, at P10, we found that while many tubules contained DAZL-negative germ cells, nearly 100% of Dazl cKO tubule cross sections contained at least one DAZL-positive cell (Figure 1—figure supplement 2A–B). However, at 6 months, only ~40% contained at least one DAZL-positive cell, compared with 100% of controls (Figure 1—figure supplement 2B). We therefore focused our analysis on aged animals, in which the conditional deletion of Dazl was more complete.

Figure 1 with 3 supplements see all

Download asset Open asset

DAZL promotes spermatogonial expansion and differentiation.

(A) Schematic of *Dazl^2L* (two *loxP*) and *Dazl^1L* (one *loxP*) alleles. Recombination of *Dazl^2L* allele via *Ddx4^Cre* allele yields *Dazl^1L* allele in germ cells. (B) Schematic of spermatogenesis from A_single (A_s) spermatogonial stem cells to progenitor spermatogonia to differentiating spermatogonia, with expression of spermatogonial markers. A_s spermatogonial stem cells self-renew while also giving rise to A_s progenitor spermatogonia, which differentiate without self-renewal. Subsequent divisions with incomplete cytokinesis produce chains of two (A_paired, A_pr) as well as chains of four, eight, and sixteen (A_aligned, A_al) progenitor spermatogonia. Progenitors form differentiating spermatogonia in response to retinoic acid. In addition, chains of A_pr and A_al progenitors can fragment to form A_s spermatogonial stem cells, particularly under conditions of stress (not shown). (C) Quantification of FOXC2, ZBTB16, and KIT-positive spermatogonia (green) in histological sections of control and *Dazl* cKO adult testes. SOX9 marks Sertoli cells (gray), DAZL is in red, and DAPI marks DNA (blue). Select tubules are outlined via dotted orange line. Cells enlarged within insets are highlighted by arrowheads. Insets show spermatogonia that are DAZL-positive and FOXC2-positive (control and mosaic cKO panels); DAZL-negative and FOXC2-positive (mosaic and DAZL-negative cKO panels); DAZL-positive and ZBTB16-positive (control and mosaic cKO panels); DAZL-negative and ZBTB16-positive (DAZL-negative cKO panel); DAZL-positive and KIT-positive (control and mosaic cKO panels); and DAZL-negative and KIT-negative (DAZL-negative cKO panel). Scale bar = 20 μm. Populations were quantified from 50 tubules per animal, with two animals per genotype. Violin plots display medians with interquartile ranges. Difference between normalized FOXC2, ZBTB16, and KIT-positive populations was statistically assessed via two-sided Mann-Whitney U test. Difference between normalized KIT-positive populations over normalized ZBTB16-positive populations was statistically assessed via two-sided odds ratio. *, p<0.05; ***, p<0.001; ****, p<0.0001.

Figure 1—source data 1 RNA-seq analysis of Pou5f1:EGFP-positive spermatogonia from control and Dazl cKO testes.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig1-data1-v2.xlsx
Download elife-56523-fig1-data1-v2.xlsx
Figure 1—source data 2 Quantification of spermatogonial subpopulations and tubule cross-section area in control and Dazl cKO testes.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig1-data2-v2.xlsx
Download elife-56523-fig1-data2-v2.xlsx

First, we asked whether spermatogonial stem cells and early progenitor (A_s and A_pr) spermatogonia, as identified by FOXC2 (Wei et al., 2018; Figure 1B), are maintained in the absence of DAZL at 6 months. Using immunofluorescent staining, we quantified this stem/progenitor population in seminiferous tubule cross sections (Figure 1C). To account for shrinkage of the seminiferous tubules and concomitant changes in cell distribution caused by disrupted spermatogenesis (Figure 1—figure supplement 2C), the FOXC2 population was normalized to the population of SOX9-positive Sertoli cells, which lack DAZL and do not proliferate in adult testes (Kluin et al., 1984; Vergouwen et al., 1991). To account for the incomplete conditional deletion, Dazl cKO tubules that lacked DAZL-positive germ cells (‘DAZL-negative’) and those that exhibited one or more DAZL-positive germ cell (‘mosaic’) were analyzed separately. In DAZL-negative tubules from Dazl cKO testes, the FOXC2-positive population of stem/progenitor spermatogonia was modestly larger than in controls (two-sided Mann-Whitney U test, p=0.019), but no such difference was observed between mosaic tubules from Dazl cKO testes and controls (two-sided Mann-Whitney U test, p=0.095). We conclude that, within our cKO model, loss of Dazl has minimal impact on stem/progenitor spermatogonia within the testes.

Next, we examined whether the broader population of undifferentiated spermatogonia (stem cells as well as early and late progenitors), as identified by ZBTB16 (also known as PLZF) (Figure 1B), is affected in the Dazl cKO. The median number of ZBTB16-positive cells was reduced by 40–60% in cKO tubules (two-sided Mann-Whitney U test, p=1.85×10⁻⁶ and p=1.57×10⁻¹⁰ for mosaic and DAZL-negative cKO tubules, respectively; Figure 1C). Given that the numbers of stem/progenitor spermatogonia are not reduced, we conclude that fewer late progenitor (A_al) spermatogonia form in the absence of Dazl.

We then asked if the progenitor spermatogonia that do form are able to differentiate by quantifying the KIT-positive population of differentiating spermatogonia (Figure 1B). The median number of KIT-positive cells was reduced by 70–90% in cKO tubules (two-sided Mann-Whitney U test, p=7.06×10⁻⁹ and p=1.33×10⁻¹⁰ for mosaic and DAZL-negative cKO tubules, respectively; Figure 1C). We calculated the total rate of spermatogonial differentiation as the ratio of the normalized number of KIT-positive cells over that of ZBTB16-positive cells. In control testes, the differentiation rate was ~65%, which is consistent with previously observed rates (reviewed in de Rooij, 2017; Tegelenbosch and de Rooij, 1993). By contrast, the rate of spermatogonial differentiation was reduced to 25–35% in cKO testes (two-sided odds ratio, p=1.06×10⁻⁴ and p=1.40×10⁻⁶ for mosaic and DAZL-negative cKO tubules, respectively; Figure 1C). Therefore, without Dazl, progenitor spermatogonia do not differentiate efficiently.

To confirm these results, we isolated late progenitor and early differentiating spermatogonia expressing the Pou5f1:EGFP reporter at 6–8 months via FACS (Figure 1—figure supplement 3A–B; Garcia and Hofmann, 2012; La et al., 2018b) and analyzed gene expression via RNA-seq (Figure 1—figure supplement 3C–D). Several genes associated with stem/progenitor spermatogonia exhibited increased expression in the Pou5f1:EGFP-positive spermatogonial population from the Dazl cKO compared with those from the control, while several genes expressed in progenitor spermatogonia or the broader population of undifferentiated spermatogonia exhibited reduced abundance (adjusted p<0.05; Figure 1—figure supplement 3D). However, Zbtb16 and Kit RNA levels appeared unaffected, presumably because RNA-seq analysis of the bulk Pou5f1:EGFP-positive population does not provide the single-cell resolution of our immunohistological quantification (Figure 1C). It is also possible that the reduced population of progenitor and differentiating spermatogonia in the Dazl cKO expressed higher levels of Zbtb16 and Kit to compensate for the loss of DAZL. Regardless, other markers in this population-level RNA-seq analysis corroborate our immunohistological analysis.

In sum, the loss of Dazl results in dramatically fewer differentiating spermatogonia for two reasons: progenitor spermatogonia fail to fully expand their population, and they do not differentiate efficiently.

DAZL broadly targets the transcriptome in progenitor spermatogonia

Given that DAZL promotes the expansion and differentiation of progenitor spermatogonia, we set out to identify DAZL’s targets in undifferentiated spermatogonia using iCLIP (individual nucleotide resolution in vivo UV crosslinking and immunoprecipitation; Huppertz et al., 2014). To obtain the large numbers of undifferentiated spermatogonia needed for this biochemical analysis, we developmentally synchronized spermatogenesis (Hogarth et al., 2013; Romer et al., 2018; Figure 2A). By chemically regulating the level of retinoic acid, which is required for spermatogonial differentiation (Endo et al., 2015; van Pelt and de Rooij, 1990), we synchronized spermatogonial development and greatly enriched for undifferentiated spermatogonia in the testes. The successful accumulation of undifferentiated spermatogonia, without contamination from later stages of spermatogenesis, was verified in a testis biopsy by both histological analysis and immunohistochemical staining for STRA8 protein (Figure 2—figure supplement 1A), which is expressed at spermatogonial differentiation (Endo et al., 2015). To further verify synchronization, we isolated the synchronized germline population from ROSA26^tdTomato/+; Ddx4^Cre/+ testes (synchronize, stage, and sort or ‘3S’ protocol). Ddx4^Cre activates the ROSA26^tdTomato reporter in the vast majority (~95%) of germ cells by P1 (Nicholls et al., 2019b), prior to the initiation of synchronization treatments. We analyzed the 3S undifferentiated spermatogonia by RNA-seq (Figure 2A, Figure 2—figure supplement 1B–D). We compared these data to RNA-seq datasets of spermatogonia from unsynchronized testes (Kubo et al., 2015; La et al., 2018b; Maezawa et al., 2017). As expected, our 3S undifferentiated spermatogonia were more similar to sorted undifferentiated spermatogonia than sorted differentiating spermatogonia and exhibited minimal expression of differentiation markers Kit and Stra8 (Figure 2—figure supplement 1E–F). These data demonstrate that chemical synchronization of spermatogenesis provides an enriched population of undifferentiated spermatogonia.

Figure 2 with 2 supplements see all

Download asset Open asset

Identification of DAZL-bound transcripts in undifferentiated spermatogonia via iCLIP reveals DAZL’s regulation of spermatogonial factors.

(A) Schematic of the synchronization of spermatogenesis to obtain undifferentiated spermatogonia via the 2S method for iCLIP (green arrows) and via the 3S method for RNA-seq (purple arrows). WIN 18,446 was used to synchronize spermatogenesis by blocking spermatogonial differentiation, and samples were histologically staged to verify successful synchronization. For 3S samples, germ cells were sorted from synchronized testes. (B) Genomic distribution of DAZL iCLIP peaks identified in three biological replicates (TPM ≥1; FDR < 0.05). (C) Venn diagram showing overlap of DAZL iCLIP peaks in expressed 3' UTRs (TPM ≥1) among three biological replicates. Replicated peaks (i.e., present in at least two of three replicates) were identified. After merging replicated peaks that fell on consecutive nucleotides, 11,882 DAZL binding sites (present in at least two of three replicates; highlighted in blue) were identified. These binding sites correspond to 2,633 genes, which are designated as the DAZL-bound genes. (D) Enrichment of factors that regulate the development and differentiation of undifferentiated spermatogonia in DAZL targets compared with all genes expressed in undifferentiated spermatogonia (one-tailed hypergeometric test). Number of genes (n) in each group designated in parentheses in labels along x-axis.****, p<0.0001. (E) DAZL iCLIP, IgG iCLIP, and 3S RNA-seq gene tracks showing exemplary DAZL-bound genes that are required for spermatogonial proliferation and expansion (*Lin28a and Sox3*) or differentiation (*Kit*, *Foxo1*, *Rarg*, and *Sall4*). Each iCLIP track represents the crosslinked sites from the sum of unique reads from three biological replicates. The RNA-seq track represents the sum of two biological replicates. The scale of each gene track is marked on the left. 3' UTRs are in light blue.

Figure 2—source data 1 Genes that regulate development and differentiation in undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig2-data1-v2.xlsx
Download elife-56523-fig2-data1-v2.xlsx
Figure 2—source data 2 Replicated DAZL iCLIP peaks and transcript expression levels in 3S undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig2-data2-v2.xlsx
Download elife-56523-fig2-data2-v2.xlsx
Figure 2—source data 3 Comparison of 3S undifferentiated spermatogonia to previously published spermatogonial datasets.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig2-data3-v2.xlsx
Download elife-56523-fig2-data3-v2.xlsx

We carried out DAZL iCLIP on synchronized and staged (‘2S’) wild-type testes (Figure 2A; Figure 2—figure supplement 2A–C). Crosslinked nucleotides captured by iCLIP were used to identify peaks in three biological replicates (Figure 2B). Within each replicate, the vast majority of peaks (≥89%) were in the 3' UTR of mRNAs (Figure 2B–C, Figure 2—figure supplement 2D). We therefore defined DAZL’s binding sites as 3' UTR peaks present in at least two of the three replicates (Figure 2C).

DAZL binding sites corresponded to 2,633 genes (Figure 2C), representing 19.5% of all protein-coding genes expressed in undifferentiated spermatogonia. Based on the transcript abundance of these genes, 31.8% of all mRNA molecules in undifferentiated spermatogonia contained DAZL binding sites. Therefore, DAZL interacts with about one-third of the transcriptome in undifferentiated spermatogonia. However, DAZL is not unusual in its number of targets, as CLIP experiments have revealed that many RNA-binding proteins interact with a comparably large number of targets (Van Nostrand et al., 2020; Yamaji et al., 2017).

Given that Dazl promotes expansion and differentiation of progenitor spermatogonia, we sought mechanistic insights into how DAZL instructs spermatogonial development. To identify DAZL targets that are known spermatogonial regulators, we queried a set of genes previously annotated as regulating development and differentiation of undifferentiated spermatogonia (Figure 2—source data 1) (reviewed in Mecklenburg and Hermann, 2016). While these genes comprised only 0.4% of expressed protein-coding genes in undifferentiated spermatogonia (53 of 13,497 expressed genes), they represented 1.1% of DAZL targets (29 genes of 2,633 DAZL targets), representing an enrichment of spermatogonial factors (Figure 2D; one-tailed hypergeometric test; p=1.71×10⁻⁸). Two of these DAZL-targeted factors, Lin28a and Sox3, promote the formation and proliferation of progenitor spermatogonia (Chakraborty et al., 2014; McAninch et al., 2020) and may contribute to the diminished progenitor population observed in the Dazl cKO (Figure 2E). Additional factors, such as Foxo1, Kit, Rarg, and Sall4, promote spermatogonial differentiation (de Rooij et al., 1999; Gely-Pernot et al., 2012; Goertz et al., 2011; Hobbs et al., 2012; Yoshinaga et al., 1991), consistent with the reduced spermatogonial differentiation in the Dazl cKO (Figure 2E). Therefore, DAZL preferentially interacts with factors that regulate both expansion and differentiation of progenitor spermatogonia, consistent with its genetically demonstrated roles in these cells.

DAZL interacts with broadly expressed, dosage-sensitive regulators of fundamental cellular processes

Since known spermatogonial factors comprise just a minor fraction of DAZL’s targets (Figure 2D), we sought to characterize other factors with which DAZL interacts. Gene Ontology (GO) analysis of all DAZL targets revealed enrichment for regulators of general transcription and RNA splicing (Figure 3A). At the level of transcriptional regulation, DAZL binds to Ep300 (also known as p300), a histone acetyltransferase; Polr2d, a component of RNA polymerase II; and Taf4b, a germline-enriched subunit of the general transcription factor TFIID (Figure 3B). At the level of splicing regulation, DAZL binds to Celf1, an RNA-binding protein involved in alternative splicing; Sf1, a splicing factor required for spliceosome assembly; and Snrpb, a core spliceosomal protein (Figure 3B). Via such interactions, DAZL-mediated regulation has the potential to broadly shape the transcriptome of progenitor spermatogonia.

Figure 3

Download asset Open asset

DAZL targets are enriched for broadly expressed, dosage-sensitive regulators of basic cellular processes.

(A) All statistically enriched GO biological processes, excluding parental nodes, for 2,633 DAZL-targeted genes compared with all genes expressed in undifferentiated spermatogonia (TPM ≥1) using the PANTHER GO Slim annotation set. P values are from binomial tests with Bonferroni correction. (B) DAZL iCLIP, IgG iCLIP, and 3S RNA-seq gene tracks showing exemplary DAZL-bound genes that are involved in transcription and RNA splicing. Gene tracks as described in Figure 2E. (C) DAZL targets are depleted of testis-specific factors compared with all genes expressed in undifferentiated spermatogonia (one-tailed hypergeometric test). (D) DAZL targets are more broadly expressed than nontargets within 12 adult mouse tissues (one-sided Mann-Whitney U test). (E) DAZL targets are more likely to have a yeast ortholog (one-tailed hypergeometric test). (F) DAZL targets have a reduced ratio of nonsynonymous substitutions per nonsynonymous site to synonymous substitutions per synonymous site (dN/dS) in alignments with human orthologs (one-sided Mann-Whitney U test). (G) Human orthologs of DAZL’s targets exhibit increased intolerance for copy number variation (CNV) (one-sided Mann-Whitney U test). (H) The human orthologs of DAZL’s targets exhibit increased probability of haploinsufficiency (one-sided Mann-Whitney U test). (I) The human orthologs of DAZL’s targets exhibit increased intolerance for deletions (one-sided Mann-Whitney U test). (J) The human orthologs of DAZL’s targets exhibit increased intolerance for duplications (one-sided Mann-Whitney U test). (K) DAZL targets exhibit higher miRNA conservation scores (P_CT) (one-sided Mann-Whitney U test). Violin plots display medians with interquartile ranges; number of genes (n) in each group designated in parentheses in labels along x-axis. *, p<0.05; ****, p<0.0001.

Figure 3—source data 1 GO analysis of DAZL targets.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig3-data1-v2.xlsx
Download elife-56523-fig3-data1-v2.xlsx
Figure 3—source data 2 Testis-specific genes and expression breadth for genes expressed in undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig3-data2-v2.xlsx
Download elife-56523-fig3-data2-v2.xlsx
Figure 3—source data 3 Mouse:yeast orthology status for genes expressed in undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig3-data3-v2.xlsx
Download elife-56523-fig3-data3-v2.xlsx
Figure 3—source data 4 dN/dS ratios of mouse:human orthologs for genes expressed in undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig3-data4-v2.xlsx
Download elife-56523-fig3-data4-v2.xlsx
Figure 3—source data 5 Haploinsufficiency scores, ExAC data, and mean miRNA P_CT scores for genes expressed in undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig3-data5-v2.xlsx
Download elife-56523-fig3-data5-v2.xlsx

While GO analysis highlighted that DAZL interacts with many factors that regulate fundamental cellular processes, it failed to reveal any enrichment for germline-specific processes. We therefore tested whether DAZL targets tend not to be germline-specific factors. Using RNA-seq data from 12 male mouse tissues (Naqvi et al., 2019), we identified testis-specific genes (Figure 3—source data 2) and confirmed that Dazl expression is testis-specific, as previously described (Nicholls et al., 2019b). We found that testis-specific factors comprised 5.3% of the genes expressed in undifferentiated spermatogonia but only 2.7% of DAZL targets, representing a depletion of testis-specific factors among DAZL targets (one-tailed hypergeometric test; p=1.78×10⁻¹³; Figure 3C). Given this, we assessed whether DAZL’s targets were broadly expressed. Across 12 mouse tissues, DAZL targets exhibited greater expression breadth than non-target genes expressed in undifferentiated spermatogonia (one-sided Mann-Whitney U test, p<2.2×10⁻¹⁶; Figure 3D). Therefore, while DAZL itself is a germ cell-specific factor, its targets are biased towards broadly expressed genes that are not unique to germ cells. The broad expression of DAZL targets is consistent with their regulation of fundamental cellular processes.

To further test whether factors bound by DAZL regulate fundamental cellular processes, we examined the conservation of DAZL targets. If these targets are critical to fundamental cellular functions, then they should be preferentially conserved across the 900–1,400 million years of evolution that separate mice from yeast (Berbee et al., 2017). This conservation would exist independent of DAZL family members, which only evolved in multicellular organisms and are absent from Saccharomyces cerevisiae. While 23.8% of mouse genes expressed in undifferentiated spermatogonia had orthologs in Saccharomyces cerevisiae, 30.2% of DAZL targets had an ortholog, representing an enrichment in yeast orthologs among DAZL’s targets (one-tailed hypergeometric test; p=2.09×10⁻¹⁷; Figure 3E). Therefore, DAZL preferentially interacts with factors conserved between mouse and yeast, indicating that these factors support basic cellular functions.

Given that DAZL preferentially interacts with conserved factors, we hypothesized that DAZL targets are subject to stronger purifying selection, which limits changes to the amino acid sequence and consequently causes an excess of synonymous substitutions relative to nonsynonymous ones. In mouse:human orthologs, DAZL targets exhibited a reduced ratio of nonsynonymous to synonymous substitution rates (dN/dS) compared with nontargets (one-sided Mann-Whitney U test, p<2.2×10⁻¹⁶; Figure 3F). Therefore, DAZL targets are under strong purifying selection, consistent with their roles in fundamental cellular processes.

DAZL targets include conserved components of transcriptional and splicing complexes. To maintain stoichiometry, dosage of proteins within these complexes is strictly regulated (Veitia and Birchler, 2009; Veitia and Potier, 2015). Therefore, we asked whether DAZL targets are sensitive to dosage decreases and increases. First, as a metric of general dosage sensitivity, we assessed human copy number variation (CNV) intolerance scores (Lek et al., 2016). The human orthologs of genes targeted by DAZL were more intolerant of CNV than nontargets expressed in undifferentiated spermatogonia (one-sided Mann-Whitney U test, p=7.51×10⁻⁵; Figure 3G). Next, we examined whether DAZL targets are sensitive to dosage decreases. Compared with nontargets, human orthologs of DAZL targets had higher probabilities of displaying haploinsufficiency (one-sided Mann-Whitney U test, p<2.2×10⁻¹⁶; Figure 3H; Huang et al., 2010) and were more intolerant of deletions (one-sided Mann-Whitney U test, p=5.6×10⁻¹⁶; Figure 3I; Lek et al., 2016). We then examined whether DAZL targets are similarly sensitive to dosage increases. Human orthologs of genes targeted by DAZL exhibited a greater intolerance for duplications than nontargets (one-sided Mann-Whitney U test, p=0.045; Figure 3J; Lek et al., 2016). We also examined conserved miRNA targeting, as genes that are sensitive to dosage increases exhibit conserved targeting by miRNAs, which modulate gene dosage by lowering the levels of their mRNA targets (Bartel, 2009; Naqvi et al., 2018). Indeed, DAZL targets had higher probabilities of conserved targeting (P_CT scores) (Agarwal et al., 2015; Friedman et al., 2009), and were consequently more sensitive to dosage increases, than nontargets (one-sided Mann-Whitney U test, p<2.2×10⁻¹⁶; Figure 3K). Therefore, DAZL targets are sensitive to both increases and decreases in dosage.

In summary, within undifferentiated spermatogonia, DAZL binds transcripts that facilitate proliferation and differentiation of progenitor spermatogonia, consistent with our genetic findings in the conditional model. However, DAZL also binds a broad set of conserved, dosage-sensitive transcripts that control the fundamental cellular processes of transcription and splicing, which may also affect the efficiency of proliferation and differentiation in spermatogonia.

DAZL binds a UGUU(U/A) motif

To biochemically characterize DAZL’s binding preferences, we analyzed DAZL’s binding sites for de novo motif discovery. The top motif from two independent tools contained GUU (Figure 4A), consistent with previously identified motifs (Chen et al., 2011; Jenkins et al., 2011; Li et al., 2019; Maegawa et al., 2002; Reynolds et al., 2005; Zagore et al., 2018). We then asked whether DAZL-bound GUU motifs are flanked by a specific sequence. By comparing GUU motifs from DAZL binding sites to unbound GUUs from the same 3' UTRs, we expanded the DAZL-bound motif to UGUU(U/A) (Figure 4B). Beyond this five-nucleotide motif, there was a bias for U’s 5' to the motif as well as U’s or A’s 3' to the motif. Many RNA binding proteins exhibit similarly broad flanking sequence preferences (Dominguez et al., 2018).

Figure 4 with 2 supplements see all

Download asset Open asset

DAZL binds a UGUU(U/A) motif within 3' UTRs.

(A) De novo motif discovery from replicated DAZL peaks in 3' UTR exons. Motif analyses were carried out with HOMER and MEME tools using crosslinked peaks ± 10 nucleotides, with all expressed 3' UTRs (TPM ≥1) as background. The top three ranked motifs identified via HOMER are shown. One statistically significant motif was identified via MEME. (B) GUU-centered motif analysis of replicated peaks in 3' UTRs via kpLogo. For each crosslinked peak ±10 nucleotides, the closest GUU was identified, and all sequences were aligned along the GUU. Background was a subset of unbound GUUs randomly selected sequences from the full-length 3' UTRs that contain DAZL peaks. As P values are extremely small (<1×10⁻³⁰⁸), residues are scaled by test statistics. (C) Position of all UGUU(U/A), GUU, and UUU motifs relative to crosslinked nucleotides from replicated peaks in 3' UTRs. 0 represents the crosslinked nucleotide. Enrichment was identified relative to randomly selected sequences from the full-length 3' UTRs that contain DAZL peaks. (D) Conservation of DAZL binding sites across vertebrates based on phyloP and phastCons scores. DAZL-bound nucleotides identified via iCLIP were compared with unbound nucleotides from the same 3' UTRs (two-sided Mann-Whitney U test). (E) DAZL’s 3' UTR binding site in *Celf1* is conserved among vertebrates. Blue shading highlights nucleotides that reflect the consensus. Bold designates DAZL’s UGUU(U/A) motif. Asterisk marks crosslinked nucleotide in DAZL iCLIP data. Sequence shown is absent from coelacanth. (F) Frequency of DAZL binding sites per DAZL-bound transcript. The majority of DAZL targets have more than one DAZL binding site (those targets to the right of the vertical dashed line). (G) Distance between adjacent DAZL binding sites in 1,649 DAZL-bound transcripts with more than one DAZL binding site. (H) Relative position of DAZL binding sites along the 3' UTR. The start and end of the 3' UTR were designated as 0 and 1, respectively. DAZL binding sites are enriched at the end and, to a lesser extent, at the start, relative to randomly selected sites in the same 3' UTRs (dashed line) (two-sided Kolmogorov-Smirnov test). (I) Absolute position of DAZL binding sites along the 3' UTR. DAZL binding sites exhibit a sharp accumulation 20–100 nucleotides from the end of the 3' UTR and a broader accumulation 100–240 nucleotides from the start of the 3' UTR relative to randomly selected positions within the same 3' UTRs (dashed line) (two-sided Kolmogorov-Smirnov tests). ****, p<0.0001.

Figure 4—source data 1 Characterization of DAZL binding within 3' UTR.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig4-data1-v2.xlsx
Download elife-56523-fig4-data1-v2.xlsx

As iCLIP captured nucleotides that directly interact with DAZL, we verified that the UGUU(U/A) motif was overrepresented at crosslinked nucleotides, relative to the GUU and UUU identified by the de novo motif analysis (Figure 4C). To verify that all five nucleotides of the UGUU(U/A) motif contribute to DAZL’s binding preferences, we examined truncations of the motif (i.e., UGUU and GUU(U/A)) and confirmed that they were less enriched than the full-length motif (Figure 4—figure supplement 1A).

The UGUU(U/A) motif is consistent with the UGUU and UUUGUUUU motifs identified by previous studies (Chen et al., 2014; Li et al., 2019). Other studies had characterized DAZL’s motif as GUUG (Zagore et al., 2018), GUUC (Maegawa et al., 2002; Reynolds et al., 2005), or UUU(C/G)UUU (Chen et al., 2011). All of these motifs share features with, but are distinct from, the UGUU(U/A) motif identified here. To test whether previously identified motifs similarly capture DAZL’s binding preferences, we examined their enrichment at crosslinked nucleotides. We found that each motif exhibited less enrichment than UGUU(U/A) at crosslinked nucleotides (Figure 4—figure supplement 1A) and at DAZL binding sites (Figure 4—figure supplement 1B). In addition, UGUU(U/A) was more enriched than these other motifs in our reanalysis of an independent DAZL iCLIP dataset derived from P6 testes (Figure 4—figure supplement 1C; Zagore et al., 2018). Finally, we found that the UGUU(U/A) motif was enriched at the small number of DAZL iCLIP peaks identified outside of 3' UTRs (Figure 4—figure supplement 1D). In total, UGUU(U/A) robustly captures DAZL’s binding preferences.

DAZL’s 3' UTR binding sites are conserved among vertebrates

Mouse DAZL, human DAZL, and human DAZ likely have similar binding preferences, as their RNA-binding domains are highly conserved (Jenkins et al., 2011; Saxena et al., 1996). To explore whether DAZL’s binding sites in mice may be bound by DAZ family members during spermatogenesis in humans as well as in other vertebrates, we asked whether mouse DAZL binding sites are conserved among vertebrates. First, we analyzed phyloP scores, which quantify the conservation of each individual nucleotide (Figure 4D). We found that DAZL-crosslinked nucleotides were significantly more conserved than other 3' UTR nucleotides (two-sided Mann-Whitney U test, p<2.2×10⁻¹⁶). Next, we analyzed phastCons conservation scores, a per-nucleotide score that reflects the conservation of each nucleotide as well as its neighbors (Figure 4D). DAZL-crosslinked nucleotides were located within more conserved regions of the 3' UTR than noncrosslinked nucleotides (two-sided Mann-Whitney U test, p<2.2×10⁻¹⁶). These highly conserved sites include one of DAZL’s 3' UTR binding sites in Celf1 (Figure 4E), as well as binding sites in Lin28a and Ep300 (Figure 4—figure supplement 1E–F).

Next, we characterized DAZL’s binding behavior along the 3' UTR. The majority of DAZL targets had at least two binding sites (Figure 4F). When ranked according to the number of DAZL binding sites, the top 5% of DAZL targets had ≥15 sites and included spermatogonial factors Foxo1, Lin28a, Sall4, and Sox3 (Figure 2E) as well as splicing factor Celf1 (Figure 3B). Indeed, based on Gene Set Enrichment Analysis (GSEA), undifferentiated spermatogonial factors were overrepresented among DAZL targets with many binding sites (p<0.001; Figure 4—figure supplement 2A). Enriched GO categories associated with transcription and RNA splicing also showed some overrepresentation among DAZL targets, but they did not meet thresholds for statistical significance. The number of binding sites weakly correlated with transcript abundance (Figure 4—figure supplement 2B), which suggested that these data underestimate the actual number of binding sites for lowly expressed transcripts. The number of binding sites also weakly correlated with 3' UTR length (Figure 4—figure supplement 2C), indicating that the frequency of DAZL binding on a gene is not solely driven by its 3' UTR length. There was a similarly weak correlation between the number of binding sites and the number of UGUU(U/A) motifs in the 3' UTR (Figure 4—figure supplement 2D). In part, this limited correlation is likely due to DAZL having flanking sequence preferences beyond the 5-nt motif (Figure 4B). For DAZL targets with multiple binding sites, we examined the distance between adjacent binding sites, and found that binding sites are positioned close together, with the median distance between peaks being 28 nt (Figure 4G).

We then examined the distribution of DAZL binding sites along the 3' UTR. While DAZL binding sites were located throughout the 3' UTR, they were strongly enriched at the start and end of the 3' UTR (two-sided Kolmogorov-Smirnov test, p<0.0001; Figure 4H–I, Figure 4—figure supplement 2E–F). DAZL’s preference for the end, but not the start, of the 3' UTR was previously observed (Zagore et al., 2018). Given that the prior analysis used DAZL CLIP data from whole testes, the accumulation of binding sites at the start of the 3' UTR in our data may be a unique feature of DAZL binding in undifferentiated spermatogonia. The accumulation of binding sites at the end of the 3' UTR, near the poly(A) tail, is consistent with DAZL interacting with poly(A)-binding protein to regulate its targets (Collier et al., 2005).

In summary, DAZL interacts with the start and end of the 3' UTR at sites that are conserved among vertebrates. Therefore, DAZL’s targets in mouse are likely regulated by DAZ family proteins during spermatogenesis across vertebrates, including human.

DAZL enhances translation of its targets

DAZL functions as a translational enhancer within the oocyte (Collier et al., 2005; Sousa Martins et al., 2016), but its molecular function during spermatogenesis remains poorly defined. We therefore asked whether DAZL similarly enhances the translation of its targets in undifferentiated spermatogonia. To isolate actively translating ribosomes specifically from germ cells, we used the RiboTag allele, which expresses HA-tagged RPL22 after Cre-mediated excision and allows for isolation of polysomes via anti-HA immunoprecipitation (IP) (Figure 5A–B; Sanz et al., 2009). To recombine the RiboTag allele specifically in germ cells, we used the Ddx4^Cre allele. We performed RiboTag IP-seq in Rpl22^HA^/+; Ddx4^Cre^/+ testes chemically synchronized via the 2S method for undifferentiated spermatogonia, along with controls for nonspecific binding to the antibody (RiboTag IP-seq in Rpl22^+/⁺; Ddx4^Cre^/+2S testes) and input RNA-seq from all IP-seq samples to control for differences in mRNA abundances among samples. We also used the RNA-seq data from 3S undifferentiated spermatogonia to control for mRNA levels in germ cells.

Figure 5 with 1 supplement see all

Download asset Open asset

DAZL enhances translation of its targets in undifferentiated spermatogonia.

(A) Schematic of synchronization of spermatogenesis by WIN 18,446 to enrich for undifferentiated spermatogonia for RiboTag IP and sequencing experiments. (B) Schematic of the RiboTag allele. The *Rpl22* locus carries a floxed exon 4, which is expressed in the absence of recombination, followed by an engineered exon four that encodes an HA tag before the stop codon. Recombination via the *Ddx4-Cre* allele removes the floxed exon four and allows for expression of the exon four that encodes the HA tag specifically in germ cells. The germ cells’ ribosomes can then be immunoprecipitated via the HA tag. (C) Translational efficiency of DAZL targets compared with all nontargets from undifferentiated spermatogonia (two-sided Mann-Whitney U test). (D) Translational efficiency of DAZL targets compared with nontarget subset datasets that were sampled to match transcript abundance (TPM) and CDS length in undifferentiated spermatogonia (two-sided Mann-Whitney U test). (E) Adjusted translational efficiency in undifferentiated spermatogonia (two-sided Mann-Whitney U test), calculated from a multiple log-linear regression model that included transcript abundance (TPM), CDS length, 3' UTR length, codon usage (Codon Adaptiveness Index, CAI), and DAZL binding as variables. Adjusted translational efficiency was calculated by subtracting the contributions (as calculated by the model) of transcript abundance, CDS length, 3' UTR length, and codon usage from biochemically measured translational efficiency values. ****, p<0.0001.

Figure 5—source data 1 Translational efficiencies for genes expressed in undifferentiated spermatogonia.: https://cdn.elifesciences.org/articles/56523/elife-56523-fig5-data1-v2.xlsx
Download elife-56523-fig5-data1-v2.xlsx

We measured translational efficiencies (i.e., number of ribosomes bound per transcript molecule, calculated as the enrichment in RiboTag IP-seq relative to control samples Baser et al., 2019) for 11,996 protein-coding genes expressed in wild-type undifferentiated spermatogonia. To verify our translational efficiencies, we examined replication-dependent histones, whose translation is coupled to S phase of the cell cycle (Marzluff et al., 2008). We would expect these histones to be poorly translated because our population of undifferentiated spermatogonia is not enriched for S phase. Indeed, we confirmed that replication-dependent histones were less efficiently translated than the average transcript (two-sided Mann-Whitney U test, p=5.62×10⁻¹⁰; Figure 5—figure supplement 1A). With respect to translational efficiency, Kit fell within the bottom 10% of all genes, consistent with its translational repression in progenitor spermatogonia in the absence of retinoic acid (Busada et al., 2015).

We observed that DAZL targets were more efficiently translated than nontargets in undifferentiated spermatogonia (two-sided Mann-Whitney U test, p<2.2×10⁻¹⁶; Figure 5C). Next, we examined variables known to correlate with translational efficiency. Compared with nontargets, DAZL targets exhibited longer 3' UTRs and less optimized codon usage (Codon Adaptation Index, CAI) (two-sided Kolmogorov-Smirnov test, p<2.2×10⁻¹⁶ for each variable; Figure 5—figure supplement 1B). However, these variables were associated with lower translational efficiencies (Figure 5—figure supplement 1C), and therefore did not contribute to the higher translational efficiencies observed in DAZL targets. DAZL targets also exhibited higher transcript abundances (TPMs) and longer coding regions than nontargets (two-sided Kolmogorov-Smirnov test, p<2.2×10⁻¹⁶ for each variable; Figure 5—figure supplement 1B). As these variables positively correlated with translational efficiency (Figure 5—figure supplement 1C), they were confounding variables in DAZL targets’ higher translational efficiencies.

To control for differences in transcript abundance and coding region length, we sampled nontargets to obtain a subset that matched DAZL targets for each individual variable. We then compared translational efficiencies, and found that DAZL targets exhibited enhanced translational efficiencies compared with nontargets matched for abundance and coding region length (two-sided Mann-Whitney U test, p=1.22×10⁻¹⁴ and p<2.2×10⁻¹⁶, respectively; Figure 5D). When comparing the median DAZL target to the nontarget matched for transcript abundance, which was the variable most correlated with translational efficiency (Figure 5—figure supplement 1C), DAZL binding increased translational efficiency by 0.43 ribosomes per transcript (Figure 5D).

As a second statistical approach, we tested whether DAZL affects its targets’ translation using a multiple log-linear model of translational efficiency that included DAZL binding and other covariates: transcript abundance, CDS length, 3' UTR length, and codon usage. Inclusion of DAZL binding as a variable in the model significantly improved the fit to the translational efficiency data (likelihood-ratio test, p=2.97×10⁻⁴³). We then used this model to calculate adjusted translational efficiencies that control for transcript abundance, CDS length, 3' UTR length, and codon usage. By comparing the adjusted translational efficiencies of DAZL targets and nontargets, we found that DAZL targets were more efficiently translated (two-sided Mann-Whitney U test, p<2.2×10⁻¹⁶; Figure 5E). DAZL binding increased the adjusted translational efficiency for the median transcript by 0.39 ribosomes per transcript (Figure 5E), similar to the effect size estimated by our first statistical approach (Figure 5D).

Therefore, based on in vivo profiling, DAZL amplifies translation in undifferentiated spermatogonia. This DAZL-bound program encompasses 20% of expressed genes in undifferentiated spermatogonia, particularly factors not specific to the testes that regulate broad cellular process such as transcription and splicing, and is required for the expansion and differentiation of progenitor spermatogonia.

Discussion

Prior to the studies reported here, DAZL’s direct functions in the spermatogonial progenitor population were not known. We demonstrate that Dazl promotes expansion and differentiation of spermatogonial progenitors, independent of its embryonic requirement for germ cell determination (Figure 6). Via biochemical analyses in undifferentiated spermatogonia in vivo, we further demonstrate that DAZL accomplishes these functions by enhancing translation of thousands of genes, including recognized spermatogonial factors that promote proliferation or differentiation of undifferentiated spermatogonia. More generally, DAZL interacts with broadly expressed, dosage-sensitive regulators of transcription and RNA metabolism, secondarily transforming the cell’s transcriptome to facilitate proliferation and differentiation of spermatogonia.

Figure 6

Download asset Open asset

Model of DAZL’s regulation of a broad translation program in undifferentiated spermatogonia.

(A) DAZL promotes robust translation of a broad set of transcripts, ranging from those required for spermatogonial proliferation and differentiation to those that regulate fundamental cellular processes like transcription and RNA splicing. (B) In the absence of DAZL, transcripts normally bound by DAZL are translated less efficiently. (C) DAZL’s broad translational program promotes the expansion of late progenitor (A_al) undifferentiated spermatogonia and subsequent spermatogonial differentiation (‘+DAZL’ arrows). In the absence of DAZL (‘-DAZL’ arrows), both expansion and differentiation occur at reduced rates.

Our biochemical characterization of DAZL targets in undifferentiated spermatogonia in vivo reveals a surprising breadth to the DAZL-regulated program, which encompasses ~30% of transcripts in this spermatogenic cell type. This breadth was not previously appreciated because prior studies identified DAZL targets in whole testes containing mixed stages of spermatogenesis (Li et al., 2019; Zagore et al., 2018). At the same time, this breadth suggests a role for DAZL in the regulation of global translational levels, which balances stem cell renewal with progenitor formation in other transit-amplifying populations (Blanco et al., 2016; Signer et al., 2014; reviewed in Tahmasebi et al., 2019; Zismanov et al., 2016). Specifically, hematopoietic, hair follicle, and muscle stem cells exhibit lower rates of protein synthesis compared with the progenitor cells to which they give rise (Blanco et al., 2016; Signer et al., 2014; Zismanov et al., 2016), and increased translation promotes the differentiation of stem cells into progenitors (Zismanov et al., 2016). Conversely, reduced translational levels limit progenitor formation in favor of self-renewal (Blanco et al., 2016). The latter phenotype is analogous to the constrained spermatogonial expansion and differentiation observed in the absence of Dazl. Therefore, DAZL promotes progenitor formation and function via its broad translational regulation. As DAZL promotes translation initiation, the rate-limiting step of protein synthesis during which the ribosome is assembled onto mRNA, this regulation can occur without changes in expression of the translational machinery (Collier et al., 2005). Increased translation of specific DAZL targets further impacts spermatogonial progenitors. DAZL-mediated expansion of the progenitor population likely requires DAZL’s targeting of, for example, Lin28a (Figures 2E and 6A), which enhances proliferation and formation of late progenitor (A_al) spermatogonia, at least in part, by blocking the biogenesis of the anti-proliferative miRNA let-7g (Chakraborty et al., 2014; Johnson et al., 2007; Newman et al., 2008; Viswanathan et al., 2008). Spermatogonial differentiation, which occurs in response to retinoic acid, is likely facilitated by DAZL’s targeting of, for example, Rarg (Figures 2E and 6A), a retinoic acid receptor that transduces retinoid signaling (Gely-Pernot et al., 2012; Van Pelt and De Rooij, 1990). Prior CLIP studies using whole testes identified many spermatogonial factors as DAZL targets (Li et al., 2019; Zagore et al., 2018). However, by characterizing DAZL:RNA interactions specifically in undifferentiated spermatogonia, we provide greater insights into DAZL’s binding behavior by identifying novel mRNA targets (i.e., Rarg and Ep300) and additional DAZL binding sites at previously reported targets (i.e., Lin28a and Celf1).

Investigators recently reported a different conditional deletion of Dazl in which undifferentiated spermatogonia are gradually lost and then depleted altogether (Li et al., 2019). We did not observe such a defect in stem cell maintenance in our cKO model (Figure 1C), likely because DAZL protein expression persisted in at least a subset of spermatogonial stem cells. The potential fragmentation of DAZL-positive A_pr and A_al spermatogonia into A_s stem cells, which may occur under conditions of stress (Hara et al., 2014; Nakagawa et al., 2007; Nakagawa et al., 2010; Zhang et al., 2016), may have also contributed to the maintenance of a stem cell pool in our model. Using different conditional Dazl and Cre alleles, Li et al. more efficiently deleted Dazl in spermatogonial stem cells and thereby revealed a role for DAZL in this cell type. This study complements our own and highlights that DAZL functions at multiple points along spermatogonial development.

Because we characterized translation and DAZL’s targets in undifferentiated spermatogonia from a synchronized first round of spermatogenesis, some of our results may represent the unique regulation that occurs within the first round, which originates from prospermatogonia, as opposed to later cycles of spermatogenesis, which arise from spermatogonial stem cells derived from prospermatogonia (Hermann et al., 2015; Law et al., 2019; Yoshida et al., 2006). Our data may also reflect the larger spermatogonial stem cell pool that is established in the neonatal testis during synchronization (Agrimson et al., 2017). Despite these caveats, much of the regulation identified here is likely relevant to spermatogonial function in the adult, particularly as many spermatogonial factors targeted by DAZL, such as Lin28a and Rarg, contribute to adult spermatogenesis (Chakraborty et al., 2014; Gely-Pernot et al., 2012). In addition, we genetically demonstrated DAZL’s requirement in progenitor spermatogonia in adult animals, thereby supporting a role for DAZL beyond the first round of spermatogenesis.

Our in vivo findings are, in part, inconsistent with the in vitro conclusions from Zagore et al., 2018, who proposed that DAZL regulates transcript stability, but not translation, in spermatogonia based on work in mouse GC-1 cells (derived from immortalized Type B differentiating spermatogonia). This cell line is likely not an appropriate model for in vivo spermatogonia, as RNA-seq data (Zagore et al., 2018) shows low abundance of germ cell-specific markers such as Dazl and Ddx4, suggesting that these cells do not faithfully reflect the biology of germ cells. Our analyses of spermatogonia in vivo demonstrate that, for the vast majority of targets, DAZL increases translational efficiency, though we were unable to assess how the loss of DAZL translationally impacts specific genes because our conditional model’s high rate of mosaicism (Figure 1—figure supplement 2B) obscures the effect of DAZL binding on translation. While we could not formally rule out the possibility that DAZL also regulates mRNA stability, our translational profiling of undifferentiated spermatogonia in vivo more accurately captures DAZL’s molecular activity in spermatogonia.

Our refinement of DAZL’s motif to UGUU(U/A) (Figure 4A–C) expands upon previous characterizations of DAZL’s binding preferences, including a UGUU motif identified via CLIP (Chen et al., 2014; Jenkins et al., 2011; Li et al., 2019). However, our motif contrasts with the GUUG motif identified by another CLIP study (Zagore et al., 2018). This latter motif originated from CLIP-based computational analyses that fail to fully capture a protein’s binding preferences (Sugimoto et al., 2012). By contrast, our computational analysis, which relies on iCLIP’s ability to capture DAZL:RNA crosslinked sites with single-nucleotide resolution, provides a more comprehensive characterization of binding preferences (Sugimoto et al., 2012). In DAZL iCLIP data from this study and others (Zagore et al., 2018), UGUU(U/A) was more enriched at crosslinked sites than GUUG (Figure 4—figure supplement 1A-C) and thus is more representative of DAZL’s in vivo binding preferences.

Our mechanistic characterization of DAZL function in mouse spermatogonia likely captures the activity of human DAZ, whose deletion is implicated as among the most common known genetic causes of spermatogenic failure. Like mouse DAZL, human DAZ is expressed in spermatogonia (Menke et al., 1997; Nickkholgh et al., 2015; Szczerba et al., 2006). At the molecular level, the two proteins’ highly conserved RNA-binding domains recognize similar motifs - U(G/U)UUU by human DAZ compared with UGUU(U/A) by mouse DAZL (Figure 4A–B; Dominguez et al., 2018; Jenkins et al., 2011; Saxena et al., 1996). As mouse DAZL’s 3' UTR binding sites show strict conservation among vertebrates (Figure 4D–E), human DAZ likely elicits similarly broad translational regulation by binding human orthologs of DAZL’s murine targets. At the phenotypic level, the majority of men with reduced DAZ dosage resulting from an AZFc deletion form spermatozoa at a decreased rate (Girardi et al., 1997; Hopps et al., 2003; Kim et al., 2012; Nakahori et al., 1996; Reijo et al., 1995; Simoni et al., 1997; Vogt et al., 1996; Zhang et al., 2013). This reduced spermatogenesis is consistent with the decreased spermatogonial expansion and differentiation observed in the Dazl cKO (Figure 6) and consequently points to a spermatogonial function for human DAZ. While a previous study reported that AZFc deletions did not affect maintenance and differentiation of human spermatogonia in vitro, culture conditions probably obviated these cells’ requirement for DAZ, as wild-type spermatogonia stopped expressing DAZ protein after long term culture (Nickkholgh et al., 2015). Therefore, human DAZ likely promotes spermatogonial progenitor expansion and differentiation in vivo, and reduction of DAZ dosage may disrupt these developmental transitions, resulting in reduced spermatogenesis and male sterility.

In conclusion, DAZL promotes expansion and differentiation of progenitor spermatogonia by enhancing the translation of thousands of genes, including spermatogonial factors and dosage-sensitive regulators of transcription and RNA metabolism. This broad translational program in progenitor spermatogonia reflects the function of human DAZ, whose deletion is associated with one of the most common known genetic causes of spermatogenic failure.

Materials and methods

Animals

Mice carrying the fluorescent reporter B6D2-Dazl^{em1(tdTomato)Huyc} (Dazl:tdTomato; Nicholls et al., 2019b) were back-crossed for four generations to a C57BL/6N (B6N) background. Mice carrying the Cre recombinase allele Ddx4^{tm1.1(cre/mOrange)Dcp} (Ddx4^Cre; Hu et al., 2013) were backcrossed to B6N for at least nine generations. The conditional Dazl allele B6N-Dazl^em1Dcp (Dazl^2L; Nicholls et al., 2019b) was generated via a CRISPR/Cas9-mediated strategy on the B6N background. Experimental animals were backcrossed to B6N for an additional two to three generations. Two Dazl-null alleles were used in this study: 129P2-Dazl^tm1Hjc (Dazl^-; Ruggiu et al., 1997), backcrossed to B6N for over 30 generations; and B6N-Dazl^em1.1Dcp (Dazl^1L; Nicholls et al., 2019b) which was generated by crossing animals carrying the Dazl^2L allele with Ddx4^Cre/+ animals. These two null alleles contain deletions of largely the same exons (Nicholls et al., 2019b; Ruggiu et al., 1997). In addition, when homozygosed, both alleles produce germ cell loss on B6N animals and cause teratoma formation in 129S4/SvJae animals (Nicholls et al., 2019b; Ruggiu et al., 1997). Mice carrying the fluorescent reporter Tg(Pou5f1-EGFP)2Mnn (Pou5f1:EGFP; Szabó et al., 2002), which is a multi-copy transgene array near chromosome 9’s telomere (Nicholls et al., 2019a), were maintained on a B6N background. Mice carrying the fluorescent Cre reporter Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze} (ROSA26^tdTomato; Madisen et al., 2010) were backcrossed to B6N for at least 10 generations. Mice carrying the RiboTag allele B6N.129-Rpl22^tm1.1Psam/J (Rpl22^HA or RiboTag allele; Sanz et al., 2009) were maintained as homozygotes (Rpl22^HA/HA). For wild-type mice, B6N mice were used. All wild-type B6N mice for experiments and backcrossing were obtained from Taconic Biosciences.

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Gene (Mus musculus)	deleted in azoospermia-like (Dazl)	Mouse Genome Informatics	MGI: 1342328
Strain, strain background (Mus musculus)	B6	Taconic	TAC: B6-F TAC: B6-M	C57BL/6NTac
Genetic reagent (Mus musculus)	Dazl:tdTomato	Nicholls et al., 2019b		B6D2-Dazl^{em1(tdTomato)Huyc}
Genetic reagent (Mus musculus)	Ddx4^Cre	Hu et al., 2013	MGI:5554579	Ddx4^{tm1.1(cre/mOrange)Dcp}
Genetic reagent (Mus musculus)	Dazl^2L	Nicholls et al., 2019b		B6N-Dazl^em1Dcp
Genetic reagent (Mus musculus)	Dazl^-	Ruggiu et al., 1997	RRID:IMSR_JAX:023802	129P2-Dazl^tm1Hjc
Genetic reagent (Mus musculus)	Dazl^1L	Nicholls et al., 2019b		B6N-Dazl^em1.1Dcp
Genetic reagent (Mus musculus)	Pou5f1:EGFP	Szabó et al., 2002	RRID:IMSR_JAX:004654	Tg(Pou5f1-EGFP)2Mnn
Genetic reagent (Mus musculus)	ROSA26^tdTomato	Madisen et al., 2010	RRID:IMSR_JAX:007909	Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}
Genetic reagent (Mus musculus)	Rpl22^HA	Sanz et al., 2009	RRID:IMSR_JAX:011029	RiboTag allele; B6N.129-Rpl22^tm1.1Psam/J
Antibody	anti-DAZL (rabbit polyclonal)	Abcam	Abcam ab34139; RRID:AB_731849	IHC 1:200; iCLIP 10 μg antibody per 100 μl Dynabeads and 500 μl lysate
Antibody	anti-DAZL (mouse monoclonal IgG₁)	BioRad	BioRad MCA2336; RRID:AB_2292585	IF 1:100
Antibody	anti-FOXC2 (sheep polyclonal)	R&D Systems	R&D Systems AF6989; RRID:AB_10973139	IF 1:250
Antibody	anti-HA (mouse monoclonal IgG₁)	BioLegend	BioLegend #901513; previously Covance #MMS-101R; RRID:AB_2565335	IP 2.5 μl antibody per 400 μl lysate and 200 μl Dynabeads
Antibody	anti-KIT (goat polyclonal)	R&D Systems	R&D Systems AF1356; RRID:AB_354750	IF 1:250
Antibody	anti-mCherry (goat polyclonal)	SICGEN	SICGEN AB0040-200; RRID:AB_2333092	IF 1:300
Antibody	anti-SOX9 (rabbit polyclona)	EMD Millipore	EMD Millipore AB5535; RRID:AB_2239761	IF 1:300 for postnatal testis sections; 1:250 for adult testis sections
Antibody	anti-STRA8 (rabbit polyclonal)	Abcam	Abcam ab49405; RRID:AB_945677	IHC 1:500
Antibody	anti-ZBTB16 (goat polyclonal)	R&D Systems	R&D Systems AF2944; RRID:AB_2218943	IF 1:250
Antibody	IgG control, anti-rabbit polyclonal	Santa Cruz Biotechnologies	Santa Cruz Biotechnologies sc-2027; RRID:AB_737197	iCLIP 10 μg antibody per 100 μl Dynabeads and 500 μl lysate
Commercial assay or kit	ImmPACT DAB Peroxidase Substrate	Vector Laboratories	Vector Laboratories SK-4105
Commercial assay or kit	ImmPRESS HRP anti-Rabbit Detection Kit	Vector Laboratories	Vector Laboratories MP-7401–50
Commercial assay or kit	TruSeq Stranded mRNA kit	Illumina	Illumina 20020594
Commercial assay or kit	SMART-Seq v4 Ultra Low Input RNA Kit	Takara Bio	Takara Bio 634888
Commercial assay or kit	SMARTer Stranded Total RNA-Seq Kit v2 – Pico Input	Takara Bio	Takara Bio 634411
Chemical compound, drug	N,N′-Octamethylenebis(2,2-dichloroacetamide) [Win18,446]	Santa Cruz Biotechnology	Santa Cruz Biotechnologies sc-295819	Used at 0.1 mg/gram body weight
Software, algorithm	ASPeak v.2.0.0	Kucukural et al., 2013	RRID:SCR_000380	https://sourceforge.net/projects/as-peak
Software, algorithm	CellProfiler v3.1.8	Kamentsky et al., 2011	RRID:SCR_007358	https://cellprofiler.org
Software, algorithm	coRdon v1.2.0			https://github.com/BioinfoHR/coRdon
Software, algorithm	Cutadapt v.1.8	Martin, 2011	RRID:SCR_011841	https://cutadapt.readthedocs.io/en/stable/
Software, algorithm	DESeq2 v1.26.0	Love et al., 2014	RRID:SCR_015687	http://bioconductor.org/packages/release/bioc/html/DESeq2.html
Software, algorithm	FASTX-Toolkit v.0.0.14		RRID:SCR_005534	http://hannonlab.cshl.edu/fastx_toolkit/index.html
Software, algorithm	HOMER v4.9.1	Heinz et al., 2010	RRID:SCR_010881	http://homer.ucsd.edu/homer/
Software, algorithm	GSEA v4.0.3	Mootha et al., 2003; Subramanian et al., 2005	RRID:SCR_003199	https://www.gsea-msigdb.org/gsea/index.jsp
Software, algorithm	kallisto v0.44.0	Bray et al., 2016	RRID:SCR_016582	https://pachterlab.github.io/kallisto/
Software, algorithm	kpLogo v1.0	Wu and Bartel, 2017		http://kplogo.wi.mit.edu
Software, algorithm	MEME Suite v4.11.2	Bailey and Elkan, 1994; McLeay and Bailey, 2010	RRID:SCR_001783	http://meme-suite.org/
Software, algorithm	MetaPlotR	Olarerin-George and Jaffrey, 2017		https://github.com/olarerin/metaPlotR
Software, algorithm	PANTHER v.13.1	Mi et al., 2017	RRID:SCR_004869	http://www.pantherdb.org
Software, algorithm	STAR v.2.5.4b	Dobin et al., 2013	RRID:SCR_015899	https://github.com/alexdobin/STAR

Share this article

Cite this article

DAZL promotes spermatogonial expansion and differentiation.

Figure 1—source data 1

Figure 1—source data 2

Identification of DAZL-bound transcripts in undifferentiated spermatogonia via iCLIP reveals DAZL’s regulation of spermatogonial factors.

Figure 2—source data 1

Figure 2—source data 2

Figure 2—source data 3

DAZL targets are enriched for broadly expressed, dosage-sensitive regulators of basic cellular processes.

Figure 3—source data 1

Figure 3—source data 2

Figure 3—source data 3

Figure 3—source data 4

Figure 3—source data 5

DAZL binds a UGUU(U/A) motif within 3' UTRs.

Figure 4—source data 1

DAZL enhances translation of its targets in undifferentiated spermatogonia.

Figure 5—source data 1

Model of DAZL’s regulation of a broad translation program in undifferentiated spermatogonia.

Author details

Maria M Mikedis

Contribution

Competing interests

Yuting Fan

Contribution

Competing interests

Peter K Nicholls

Contribution

Competing interests

Tsutomu Endo

Contribution

Competing interests

Emily K Jackson

Contribution

Competing interests

Sarah A Cobb

Contribution

Competing interests

Dirk G de Rooij

Contribution

Competing interests

David C Page

Contribution

For correspondence

Competing interests

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

Categories and tags

Research organism

Further reading