The zinc-finger protein basonuclin 2 is required for proper mitotic arrest, prevention of premature meiotic initiation and meiotic progression in mouse male germ cells

In mammals, primordial germ cells remain sexually bipotential during their migration across the embryo; they become sexually differentiated upon reaching the fetal gonad. Sexual differentiation is correlated with the timing of entry into meiosis. In mice, female germ cells enter meiosis shortly after they reach the fetal gonad, whereas male germ cells become quiescent in the fetal gonad and initiate meiosis after birth, when spermatogonia start forming spermatocytes. Transplantation studies have shown that bipotential germ cells are intrinsically programmed to differentiate following the female pathway and to undergo early meiosis (McLaren and Southee, 1997). Prevention of meiosis in fetal testis is achieved through synthesis in male Sertoli cells of CYP26B1, an enzyme that degrades the retinoic acid necessary for meiosis induction (Bowles et al., 2006; Koubova et al., 2006). When the level of testicular CYP26B1 decreases, prevention of meiosis is taken over by NANOS2, a protein that is synthesized by fetal male germ cells in the absence of retinoic acid signaling (Suzuki and Saga, 2008).

During spermatogenesis, spermatogonia undergo meiosis, in the first division of which homologous chromosomes pair, recombine and synapse. Disturbance of any of these processes leads to massive apoptosis-mediated death of the spermatocytes at a specific check point in stage IV of the seminiferous epithelial cycle (Hamer et al., 2008).

Basonuclin 2 (BNC2) is an extremely conserved zinc-finger protein (Romano et al., 2004; Vanhoutteghem and Djian, 2004). It is orthologous to BNC1, a protein that is essential for germ cells (Mahoney et al., 1998; Tseng and Green, 1992; Yang et al., 1997; Zhang et al., 2012). Mice with a homozygous disruption of Bnc2 die shortly after birth from a cleft palate (Hervé et al., 2012; Vanhoutteghem et al., 2011,, 2009).

Because of its abundance in testis (Vanhoutteghem and Djian, 2004), we thought that BNC2 might have a function in male reproduction in addition to its function in palatal mesenchymal cells. Direct evidence for a function of BNC2 as a DNA-binding protein that regulates meiosis and mitosis in male germ cells is described below.

In this study, we have used the previously reported Ayu21:18 mouse line, the Bnc2 gene of which is disrupted by a gene-trap insertion. The disrupted allele is likely to be null because all major Bnc2 mRNA isoforms are absent from the homozygous mutant (Vanhoutteghem et al., 2009).

We have previously described, in human keratinocytes, an insoluble form of human BNC2, which colocalizes with the splicing factor SC35 in nuclear speckles and has a presumed function in nuclear RNA processing. The apparent molecular mass of insoluble BNC2 is 145 kDa (Vanhoutteghem and Djian, 2006). We also found that, in human keratinocytes, there exists a second BNC2 isoform with a molecular mass of 160 kDa and a homogenous distribution throughout the nucleus. The two human isoforms result from alternative splicing of the transcript.

As shown in Fig. 1A, when the testes of newborn mice were analyzed by western blotting with an antibody against amino acid residues 722-849 of BNC2 (herein referred to as BNC2^722-849), three protein bands in the range of 120-160 kDa were detected in Bnc2^+/+ and Bnc2^+/− testis, but not in Bnc2^−/− testis. These proteins were present in the nuclear extract and absent from the cytoplasm. The major protein ran at ∼160 kDa, and we thought it probable that it was the mouse equivalent of human soluble BNC2. The identity of the 160-kDa protein was confirmed by its staining with another antibody against BNC2 (supplementary material Fig. S1A). Testicular BNC2 was found in the soluble nuclear extract and in association with chromatin. No BNC2 was detected in either the insoluble nuclear fraction or the cytoplasm (Fig. 1B). We conclude that in mouse testis, BNC2 is a largely soluble nuclear protein with an apparent molecular mass of ∼160 kDa.

Both of the antibodies against BNC2^722-849 and amino acid residues 661-782 of BNC2 produced homogenous nuclear staining in newborn testis (supplementary material Fig. S1B-E). Such a diffuse nuclear distribution is in keeping with the presence of BNC2 in the soluble nuclear extract and its association with chromatin. This suggests that BNC2 is a DNA-binding protein in testis.

Because alternative promoters and alternative splicing are prominent features of the human BNC2 gene, we searched mouse testis for alternative Bnc2 promoters and alternatively spliced Bnc2 transcripts. For promoter mapping, the 5′ end of the mRNA was amplified by 5′ rapid amplification of cDNA ends (RACE), resolved by using agarose gel electrophoresis and then sequenced. A single transcription start site in exon 1 was found (supplementary material Fig. S2A,B). Alternative exons were identified by using RT-PCR with primers in either exons 1 and 3, or 4 and 6. The unfractionated products were cloned and sequenced. Because a large number of clones were sequenced, we could obtain an estimate of the relative abundance of each Bnc2 mRNA isoform. We found a total of 11 exons, of which five were constitutive (exons 1, 3, 4, 5 and 6) and six were alternative (1a, 2, 2a, 2c, 5c_S, 5c_L and 5b) (supplementary material Fig. S2C). The 5′ alternative exons (1a, 2, 2a and 2c) were all in frame with the rest of the coding region and caused variations in the N-terminal part of the protein, whereas the 3′ alternative exons (5c_S, 5c_L and 5b) resulted in either premature termination and suppression of the last three zinc-fingers or disruption of finger 4 with conservation of fingers 5 and 6. As a result, testis contained a family of BNC2 proteins with a constant central sequence that included the three N-terminal zinc-fingers and the nuclear localization signal, but variable N-terminal regions and variable numbers of C-terminal zinc-fingers (Fig. 1C).

The most abundant RNA isoform was composed of exons 1, 2, 2a, 3-6 (Fig. 1D) and encoded a 1099-amino acid protein. All other isoforms were much less abundant. The RNA comprising exons 1, 2a and 3-6, which is fairly abundant in human keratinocytes (Vanhoutteghem and Djian, 2007) and encodes the speckle-associated BNC2, was not detected, thus explaining the absence of speckle-associated BNC2 in mouse testis. Lentiviral expression in HeLa cells of the 1, 2, 2a, 3-6 isoform fused to c-myc resulted in the production of soluble BNC2 that had an apparent molecular mass undistinguishable from that of the testicular protein (Fig. 1E).

To define the cell types expressing Bnc2 in fetal testis, paraffin-embedded sections of Bnc2^+/+ testes were stained by using immunohistochemistry and the antibody against BNC2^722-849 (subsequently used in all immunofluorescence and immunohistochemistry analyses). We first examined testes on embryonic day (E)14.5, when germ cells undergo male differentiation. As shown in Fig. 2A, abundant nuclear BNC2 was detected in prospermatogonia, identified by their central position within the testicular cords and by the presence of VASA (DDX4 – Mouse Genome Informatics) (Castrillon et al., 2000). Sertoli cells, which entirely form the tubular epithelium until postnatal day (P)6, did not possess detectable levels of BNC2. This was confirmed by double staining for BNC2 and anti-Müllerian hormone. Some interstitial cells possessed smaller amounts of BNC2 (Fig. 2B). To determine whether these interstitial cells were Leydig cells, we double-stained testis for BNC2 and for 3-β-hydroxysteroid dehydrogenase (Majdic et al., 1998). No BNC2 was detected in Leydig cells (Fig. 2C). On E18.5, BNC2 was found in the nuclei of prospermatogonia at a lower level than that on E14.5. There was also some protein in interstitial cells. Neither Sertoli nor Leydig cells possessed detectable BNC2 (Fig. 2E,F).

Postnatal distribution of BNC2 was investigated by indirect immunofluorescence staining. At birth, BNC2 was found to occur in prospermatogonia and interstitial cells (Fig. 3A) but was absent from both Sertoli (Fig. 3B) and Leydig cells (Fig. 3C). On P8, prospermatogonia had been replaced by spermatogonia, located in the basal layer of the tubular epithelium, whereas spermatocytes had started to form suprabasally. BNC2 was confined to some basal cells, identified as undifferentiated spermatogonia by the presence of the promyelocytic leukemia zinc-finger protein (PLZF; ZBTB16 – Mouse Genome Informatics), a marker of spermatogonial progenitor cells (Buaas et al., 2004; Costoya et al., 2004). BNC2 was absent from both differentiated spermatogonia and spermatocytes (Fig. 3D-F). On postnatal day 30, BNC2 was found in a few cells within the tubules and in some cells outside of the tubules. Within the tubules, BNC2-positive cells were invariably basal; they were identified as undifferentiated spermatogonia by double staining for BNC2 and PLZF; all tubular BNC2-containing cells possessed PLZF and vice versa. Differentiating spermatogonia, spermatocytes and spermatids lacked both BNC2 and PLZF (Fig. 3G-I). The BNC2-containing cells located outside of the tubules (cells stained red in Fig. 3G,I) probably corresponded to the nonsteroidogenic interstitial cells that had been identified at earlier ages. Therefore, in the germ cell lineage, BNC2 was restricted to the most primitive cells – prospermatogonia and undifferentiated spermatogonia.

In view of the abundance of BNC2 in prospermatogonia, we thought that the protein might have a function in these cells. The fact that BNC2 has been implicated in cell multiplication (Vanhoutteghem et al., 2009) prompted us to compare the mitotic status of the prospermatogonia of the Bnc2^−/− mice with that of their wild-type littermates. During the fetal period, prospermatogonia divide actively from around E12 and then gradually become quiescent between E14 and E16, depending on the mouse strain (Vergouwen et al., 1991). We analyzed fetal germ cell proliferation by using immunohistochemical staining of testicular sections for VASA and Ki-67 (MKI67 – Mouse Genome Informatics), a nuclear marker of multiplying cells (Gerdes et al., 1983). On E14.5, ∼40% of Bnc2^+/+ germ cells contained Ki-67, whereas almost all Bnc2^−/− germ cells expressed Ki-67 (Fig. 4A-C). Real-time quantitative PCR (RT-qPCR) analyses have indicated that Lefty and Nodal, two genes known to be expressed in mitotically active prospermatogonia (Souquet et al., 2012), are overexpressed in Bnc2^−/− testis at E14.5. Expression of neither Nanos2, a marker of male differentiation, nor Oct4 (Pou5f1 – Mouse Genome Informatics), a germ cell marker, was affected (supplementary material Fig. S3). On both E17.5 and E18.5, all prospermatogonia in both Bnc2^−/− and Bnc2^+/+ testes lacked Ki-67 (Fig. 4D-G). Therefore, we conclude that fetal prospermatogonia lacking BNC2 multiply excessively on E14.5 but do reach mitotic quiescence by E17.5.

We then examined newborn testes by staining paraffin-embedded sections with hematoxylin & eosin (HE). In Bnc2^+/+ newborns, prospermatogonia were evident at the center of the cords as large round cells with a spherical nucleus and dispersed homogenous chromatin; no mitotic prospermatogonia were seen (Fig. 4H). In Bnc2^−/− mice, many prospermatogonia were similar to those of wild-type littermates, but 20-30% showed condensed chromosomes and appeared engaged in either mitosis or meiosis (Fig. 4I). To determine whether prospermatogonia with condensed chromosomes were engaged in mitosis, we double-stained newborn testes for Ki-67 and VASA. In the wild type, no prospermatogonia stained positively for Ki-67; all cells containing Ki-67 were somatic cells, as they lacked VASA (Fig. 4J). In the Bnc2^−/− newborns, approximately one-third of the prospermatogonia contained Ki-67 and therefore were engaged in the mitotic cycle (Fig. 4K).

We then tested whether Bnc2^−/− prospermatogonia could also be engaged in meiosis. Testes were stained for synaptonemal complex protein 3 (SYCP3), a marker of early meiosis (Lammers et al., 1994). On E17.5, wild-type testis did not contain any SYCP3-positive cells (Fig. 5A), whereas in the Bnc2^−/− littermates, ∼10% of the prospermatogonia contained foci of SYCP3 (Fig. 5B). By P0, Bnc2^+/+ prospermatogonia still stained negatively for SYCP3 (Fig. 5C), whereas in the Bnc2^−/− testis, the proportion of SYCP3-positive prospermatogonia had increased to 25-30%. SYCP3 was mostly in the form of large aggregates (Fig. 5D) but sometimes in threads typical of zygotene (Fig. 5E). The presence of SYCP3 in Bnc2^−/− prospermatogonia was confirmed using another antibody against SYCP3. Because a nonfilamentous distribution of SYCP3 is typical of cells in the preleptotene and leptotene stages that have not yet formed synaptonemal complexes, we hypothesized that most Bnc2^−/− prospermatogonia that had entered meiosis had remained thereafter blocked in preleptotene or leptotene. The presence of the premeiotic-specific protein stimulated by retinoic acid 8 (STRA8) (Koubova et al., 2006; Mark et al., 2008; Oulad-Abdelghani et al., 1996) confirmed the engagement of Bnc2^−/− prospermatogonia in meiosis. STRA8 was not detectable in wild-type testis (Fig. 5F,G). In order to determine whether the prospermatogonia that had abnormally engaged in meiosis were undergoing apoptosis, we stained testes at E17.5, E18.5 and P0 for activated caspase 3, which is associated with apoptotic cell death (Nicholson et al., 1995). Almost no cleaved caspase 3-containing cells were detected in either the Bnc2^−/− mice or their wild-type littermates (supplementary material Fig. S4A,B). Double staining for SYCP3 and cleaved caspase 3 was performed on P0 testis; none of the numerous SYCP3-containing prospermatogonia stained positive for the caspase (supplementary material Fig. S4C,D).

We finally sought to determine whether the prospermatogonia engaged in the mitotic cycle and containing Ki-67 (Fig. 4K) were the same ones that aberrantly produced SYCP3. We triple-stained newborn Bnc2^−/− testes for VASA, SYCP3 and Ki-67; prospermatogonia contained either Ki-67 or SYCP3, but never the two proteins together (Fig. 5H,I).

To identify the genes of which the expression was affected by a lack of BNC2, we used massive parallel sequencing (RNA-Seq). To achieve this, we prepared total RNA from the testes of six Bnc2^−/− and four Bnc2^+/+ newborn mice. We first examined the ribosomal (r)RNA gene transcripts. BNC1 is known to be a transcription factor for RNA polymerase I (Iuchi and Green, 1999; Tian et al., 2001). The first pair of zinc-fingers of BNC2 is very similar to that of BNC1 and has been shown to bind to the rRNA gene promoter in vitro (Romano et al., 2004). We did not detect any decrease in the number of 18S and 28S RNA molecules in the Bnc2^−/− testes. On the contrary, we observed a slight increase (supplementary material Table S1). We conclude that BNC2 is not a transcription factor for RNA polymerase I.

A list of genes specifically expressed in each testicular cell type is given in the supplementary material Table S2. The expression of the genes encoding DNMT3L and albumin showed the greatest changes. Both genes are associated with male germ cell differentiation and showed over a tenfold decrease in expression. DNMT3L is a zinc-finger protein that resembles cytosine-5-methyltransferase 3, but lacks its catalytic domains. Dnmt3l is expressed from E14.5 until the perinatal period in prospermatogonia, but not in oocytes; it is required for retrotransposon inactivation (Aapola et al., 2000; Bourc'his and Bestor, 2004; Bourc'his et al., 2001; Shovlin et al., 2007). The albumin gene is specifically expressed in prospermatogonia of newborn testis and is not expressed in the ovary (Gelly et al., 1994; McLeod and Cooke, 1989). We also observed an increase in the premeiotic-specific mRNAs encoding STRA8, MSX1 and MSX2 (Le Bouffant et al., 2011), all of which are normally specific to female germ cells during fetal life.

The RNA-Seq results for Dnmt3l and Stra8 were confirmed by using RT-qPCR. On E18.5, the level of the methylase-like RNA was approximately fourfold lower in Bnc2^−/− testis than in that of wild type. This difference reached tenfold on P0. On E18.5 and P0, the levels of the Stra8 mRNA were ∼15- and 3.5-fold higher in Bnc2^−/− embryos than in wild-type embryos, respectively, (supplementary material Fig. S5). Therefore, BNC2 is necessary for proper male differentiation.

Even though the Bnc2 mutation is overwhelmingly lethal (Vanhoutteghem et al., 2009), we obtained in the course of this study six Bnc2^−/− males that survived the perinatal period because their palates closed. Because BNC2 is found in undifferentiated spermatogonia, we thought that it might have a function in spermatogenesis.

We first examined 9-day-old males. At this age, the germ cell population comprises spermatogonia and spermatocytes at the initial stages of meiosis (Bellvé et al., 1977; Goetz et al., 1984). Undifferentiated spermatogonia were identified by double staining for BNC2 and PLZF, and spermatocytes by the presence of SYCP3. In 9-day-old wild-type mice, numerous PLZF-containing undifferentiated spermatogonia were visible in the basal layer of the tubules (Fig. 6A-C). The number of these spermatogonia was not markedly different in the Bnc2^−/− littermates (Fig. 6D-F). In contrast to the similarity of the PLZF staining, the SYCP3 staining differed between Bnc2^−/− and wild-type testis. In the latter, only ∼12% of the tubules had started producing spermatocytes, in which SYCP3 had mostly formed threads typical of the zygotene stage (Scherthan et al., 1996) (Fig. 6G). In Bnc2^−/− testes, over 80% of the tubules contained vast numbers of SYCP3-positive cells, most of which showed large aggregates of SYCP3; no zygotene spermatocytes were observed (Fig. 6H). These aggregates were similar to those seen in Bnc2^−/− prospermatogonia (see Fig. 5B,D). We thought it probable that cells with large aggregates had accumulated SYCP3 that could not disperse in synaptonemal complexes because the cells had not progressed to zygotene. These spermatocytes did not contain activated caspase 3 and therefore were not apoptotic (supplementary material Fig. S4E,F). We conclude that in the 9-day-old Bnc2^−/− mouse, the entry of spermatogonia into meiosis is greatly accelerated, but the cells are then blocked or delayed at the leptotene-zygotene transition.

We then compared the Bnc2^−/− and Bnc2^+/− testes of 1-month-old mice. Staining with HE showed that all Bnc2^+/− tubules contained basal spermatogonia, several layers of spermatocytes and groups of spermatids towards the lumen (Fig. 7A). By contrast, the Bnc2^−/− tubules could be divided into type I tubules that contained only basal cells on most of their circumference (Fig. 7B), and type II tubules that were stratified and sometimes contained typical pachytene spermatocytes, often located in the center of the tubule instead of in their normal peripheral location (Fig. 7C,D). No spermatids were seen in any of the Bnc2^−/− tubules. Type I tubules represented 46% of total tubes (n=100).

In order to clarify the nature of the cells present in Bnc2^−/− tubules, we first characterized undifferentiated spermatogonia by using PLZF staining. Heterozygous testes contained an average of 1.3 PLZF-positive spermatogonia per tubule (Fig. 7E); a similar number was found in Bnc2^−/− tubules (Fig. 7F). We then identified mitotic and meiotic cells by double staining for Ki-67 and SYCP3. In the Bnc2^+/− testis, tubules generally possessed a small number of basal cells that were positive for Ki-67 and a suprabasal layer of meiotic cells with filamentous SYCP3 (Fig. 7G). In the Bnc2^−/− testis, tubules were smaller and could be divided into two types, which were identified by using HE staining. Type I tubules contained only basal cells, virtually all of which had nuclear Ki-67 and no SYCP3 and therefore were dividing. The situation was reversed for type II Bnc2^−/− tubules, which contained very few Ki-67-positive cells, but numerous basal and suprabasal SYCP3-positive cells (Fig. 7H).

The stage of meiosis of Bnc2^+/− and Bnc2^−/− cells was determined by double staining for SYCP3 and phosphorylated histone H2AX (γ-H2AX). Phosphorylation of H2AX is required for the formation of γ-H2AX foci at meiotic double-strand breaks (Mahadevaiah et al., 2001). γ-H2AX produces a weak diffuse nuclear staining in type A spermatogonia (Hamer et al., 2003), forms multiple foci in leptotene and zygotene spermatocytes and concentrates in a unique XY body in pachytene spermatocytes. Bnc2^+/− testis contained numerous pachytene spermatocytes, located suprabasally; in some tubules, leptotene and zygotene spermatocytes were detected in the basal layer (Fig. 7I). In type I tubules of Bnc2^−/− testes, most of the cells showed weak diffuse nuclear γ-H2AX staining and no SYCP3 staining, and therefore were type A spermatogonia (Fig. 7J). The majority of these spermatogonia were differentiating because they contained c-KIT (Fig. 7K), a marker of differentiating spermatogonia (Yoshinaga et al., 1991). The differentiating Bnc2^−/− spermatogonia were nearly all engaged in the mitotic cycle as they contained both nuclear Ki-67 and histone H3 phosphorylated on serine residue 10, a protein that is strictly associated with the chromosomal condensation that occurs during mitosis and meiosis (Goto et al., 1999; Gurley et al., 1978) (Fig. 7L). Bnc2^−/− type II tubules showed intense staining of SYCP3 and γ-H2AX over several suprabasal layers. Both SYCP3 and γ-H2AX had a distribution similar to that of leptotene-zygotene spermatocytes, except that the staining was much more intense (compare Bnc2^−/− spermatocytes in Fig. 7M,N with normal leptotene-zygotene spermatocytes indicated by the arrowhead in Fig. 7I). We presumed that these cells were abnormal spermatocytes blocked in leptotene-zygotene (Liebe et al., 2006; Mahadevaiah et al., 2001). In other tubules, the cells in the luminal region showed filamentous SYCP3 and XY bodies typical of pachytene spermatocytes (Fig. 7N,O).

We were puzzled by the existence of two kinds of tubules in the mutant testis, and we questioned whether, in type I tubules, spermatogonia had been unable to form spermatocytes or whether they had formed spermatocytes that subsequently underwent apoptotic death. To investigate apoptosis, TdT-mediated dUTP-biotin end labeling (TUNEL) was used. A low level of apoptotic death was observed in heterozygous testis (approximately one cell per five tubules; see Fig. 8A). By contrast, ∼15% of the mutant littermate tubules contained many apoptotic cells (Fig. 8B) that accumulated in the suprabasal layers (Fig. 8C). Staining for activated caspase 3 confirmed the results of the TUNEL assay (Fig. 8D,E). We interpret the findings as follows: during spermatogenesis in Bnc2^−/− mice, spermatocytes enter meiosis but undergo apoptotic death sometime during meiotic prophase. Death of the spermatocytes during stages I-V of the cycle of the seminiferous epithelium (Oakberg, 1956) results in type I tubules, which contain only spermatogonia until stage VI, when new preleptotene spermatocytes are formed that enter leptotene in stage VIII (Fig. 7M) and pachytene at the end of stage XII (Fig. 7N), thus generating type II tubules. The situation observed here is analogous to that of the SYCP3 mutant, in which tubules containing only Sertoli cells and spermatogonia form through massive apoptotic death during meiosis prophase (Yuan et al., 2000).

Finally, we examined 2-month-old Bnc2^−/− mice. In the wild-type testis, all tubules contained densely packed germ cells at all stages of differentiation (Fig. 9A). Undifferentiated spermatogonia were detected by the presence of PLZF (Fig. 9B). In the Bnc2^−/− testis, some seminiferous tubules lacked germ cells altogether, whereas others retained a homogenous population of cells that appeared to be germ cells. Neither spermatocytes nor spermatids were observed (Fig. 9C). No PLZF-positive cells were detected among the 300 Bnc2^−/− tubules that we examined (Fig. 9D). The presumed germ cells of Bnc2^−/− mice did not look like tumor cells, and how these cells could have persisted in the absence of PLZF-positive spermatogonia in 2-month-old mice is unclear. Younger Bnc2^−/− mice still retain spermatogonial stem cells, and the presumed Bnc2^−/− germ cells may have been produced from these earlier spermatogonial stem cells before their pool had been exhausted.

Because BNC2 was found in nonsteroidogenic interstitial cells (Fig. 3C), we examined these cells in the Bnc2^–/– mice. We detected no abnormalities of interstitial cells in either Bnc2^−/− fetuses or newborns. We also examined Sertoli cells by using both immunostaining of anti-Müllerian hormone in paraffin-embedded sections and RT-qPCR analysis of the Sertoli cell markers Cyp26b1, Fgf9 and Sox9. We detected no overt defect in Bnc2^−/− Sertoli cells on either E14.5 (Fig. 10A-C) or E18.5 (Fig. 10D-F). We found no abnormalities in the number or the morphology of Leydig cells that we identified by histological staining for 3-β-hydroxysteroid dehydrogenase. RT-qPCR analysis of Leydig cell markers detected a decrease in insulin-like 3 mRNA on E14.5 and inhibin βA mRNA on E18.5. The other Leydig cell-specific mRNAs tested (Cyp11a1, Star and Notch) were unaffected (Fig. 10G-L).

In the RNA-Seq analysis, markers specific for Sertoli cells showed almost no difference between Bnc2^−/− and Bnc2^+/+ newborn mice. Myoid cell markers were also relatively unaffected, except for actin α2. Leydig cell-specific mRNAs tended to be increased in the Bnc2^–/– mice. The mRNA for the androgen receptor was not significantly affected by the absence of BNC2 (supplementary material Table S2). Therefore, lack of BNC2 appeared to affect predominantly, if not exclusively, the germ cell lineage.

We finally asked whether Bnc2 was expressed in female germ cells. We first studied postmigratory germ cells at the time of their sexual determination. Male and female primordial germ cells were purified on E12.5 and E13.5 by using magnetic-activated cell sorting and stage-specific embryonic antigen 1 (FUT4 – Mouse Genome Informatics) (Solter and Knowles, 1978). Microarray analysis showed that the Bnc2 mRNA was much more abundant in male than in female germ cells. The difference in Bnc2 expression was comparable to that of Nanos2 (Fig. 11A), a marker of male primordial germ cells (Tsuda et al., 2003).

We then examined fetal ovaries by immunostaining for BNC2 and VASA. On E14.5, BNC2 was undetectable in ovarian germ cells (Fig. 11B), whereas it was abundant in male germ cells (Fig. 11C). On P0, double staining for BNC2 and SYCP3 showed that oocytes did not contain detectable levels of BNC2 and that entry into meiosis had not been affected by lack of the protein (Fig. 11D-H).

The mouse Bnc2 transcript is subject to alternative splicing. As a result, the mouse, like the human (Vanhoutteghem and Djian, 2007), possesses a major BNC2 isoform with six zinc-fingers, and less abundant isoforms with different N-terminal sequences and variable numbers of zinc-fingers. Alternative-splicing events conserved between human and mouse are likely to be of primary biological importance (Yeo et al., 2005). Therefore, both the more and the less abundant mouse BNC2 isoforms that are shared with the human are likely to be functionally important. The pleiomorphism of the BNC2 effect might be explained by the existence of multiple isoforms of the protein, as well as by the presence of multiple zinc-finger pairs that each potentially binds to a different target sequence. In mouse testis, BNC2 is found in the soluble nuclear extract and associated with chromatin. Such a distribution certainly suggests that BNC2 binds to DNA through its zinc-fingers, but whether BNC2 is a transcription factor remains to be determined.

In testis, BNC2 is synthesized by germ cells and nonsteroidogenic interstitial cells. This raises the question of whether BNC2 acts on germ cells directly or through interstitial cells. We did not find any evidence that the number or morphology of interstitial cells was markedly affected by a lack of BNC2. Two somatic cell types that could affect germ cells are Leydig and Sertoli cells, but BNC2 was undetectable in both. Neither of the two main sertolian regulators of germ cell commitment, Fgf9 and Cyp26b1, was affected by the lack of BNC2, and RNA-Seq analysis did not detect meaningful alterations in the number of Leydig or Sertoli cell-specific transcripts in mice that lacked BNC2. The very strong expression of Bnc2 in prospermatogonia, particularly at the time of sexual differentiation on E14.5, and the precise regulation of its expression during spermatogenesis further support the notion of a cell-autonomous phenotype.

The nature of the nonsteroidogenic interstitial cells that synthesize BNC2 is not clear. Nonsteroidogenic interstitial cells include macrophages, myoid cells and mesenchymal cells (Skinner, 1991). Because BNC2 has been found in several kinds of mesenchymal cells, it is possible that BNC2 also resides in mesenchymal cells in the testis, where it might have a function in cell multiplication, as it does in the developing palate and urethra (Bhoj et al., 2011; Vanhoutteghem et al., 2009).

BNC2 appears to have functions in both prospermatogonia and spermatogonia, and these functions are all related to mitosis and meiosis. In contrast to Bnc2^−/− spermatocytes, Bnc2^−/− meiotic prospermatogonia do not show evidence of apoptotic cell death, possibly because prospermatogonia are not part of a cycling seminiferous epithelium. Apoptosis of abnormal meiotic cells is thought to result from a stage IV paracrine signal that is derived from Sertoli cells. Such a signal is unlikely to exist in the non-cycling epithelium of fetal and neonatal testis. Many defects in either DNA repair or meiotic pairing cause apoptosis at a checkpoint in stage IV of the epithelial cycle (Hamer et al., 2008). This raises the question of a possible function of BNC2 in meiotic pairing, recombination or synapsis. Such a function would have to be indirect because BNC2 is absent from meiotic cells.

Testes were frozen in liquid nitrogen and homogenized on ice in a tissue grinder using ten strokes. Cytoplasmic and nuclear extracts were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Kit (Pierce). We used the Subcellular Protein Fractionation Kit for Tissues (Pierce) for subcellular fractionation. The residual pellet was thoroughly mixed with Laemmli buffer (Laemmli, 1970). The protein concentration was determined using a Bio-Rad protein assay. Proteins were then resolved by electrophoresis and electroblotted onto nitrocellulose membranes. The membranes were blocked for 1 h at room temperature, incubated for 1 h in the presence of the indicated antibody and then overnight at 4°C. Membranes were washed and incubated for 1 h at room temperature with a horseradish peroxidase-conjugated secondary antibody. The membranes were then washed and the blots developed by using chemiluminescence (SuperSignal West Femto Maximum Sensitivity Substrate, Pierce). Further detail is given in the supplementary material methods and antibodies are listed in supplementary material Table S4.

For RNA preparation, testes were pulverized under liquid nitrogen. RNA was extracted using the Tripure isolation reagent (Roche). Reverse transcriptase (RT)-PCR was performed using the superscript first-strand synthesis system for RT-PCR (Life Technologies). First strand synthesis was initiated from total RNA (5 μg) with either oligo(dT) or a Bnc2-specific primer (supplementary material Table S3). The cDNAs were then subjected to 30 cycles of amplification (95°C for 1 min, 57°C for 1 min and 72°C for 1 min) using various combinations of primers (supplementary material Table S3). Sequencing was performed by Eurofins MWG Operon. RT-PCR was also used to generate the lentiviral expression vectors, detailed information is given in the supplementary material methods.

Gonads were frozen in RLT buffer (Qiagen), and total RNA was extracted using the RNeasy Mini-Kit (Qiagen). RNA (200 ng) was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) and amplified by using RT-qPCR as previously described (Souquet et al., 2012). Either β-actin or Vasa mRNA were included as an endogenous reporter. Results are presented as a fraction of the maximum. The final concentration of all primers was 400 nM. Samples were run in duplicate.

Total RNA was prepared as described under RT-PCR. Sequencing was carried out by Eurofins MWG Operon.

For HE staining, samples were either embedded in paraffin or frozen in optimal cutting temperature (OCT) compound. For paraffin embedding, samples were fixed in 4% paraformaldehyde, washed in PBS, embedded in paraffin and sectioned at 7 µm. For frozen sections, samples were successively incubated in 4% paraformaldehyde for 1 h, PBS containing 15% sucrose for 3 h and PBS containing 30% sucrose overnight, and snap-frozen in OCT (Miles).

Indirect immunofluorescence was performed as previously described (Vanhoutteghem and Djian, 2007) with some modifications, described in the supplementary material methods. Briefly, frozen sections (6 μm) were fixed in 4% formaldehyde for 5 min on ice and permeabilized using 0.2% Triton X-100, or fixed in acetone without permeabilization. For primary polyclonal antibodies, 5% BSA in PBS was used for blocking. For mouse monoclonal antibodies, MOM (Vector Laboratories) was used for blocking. Incubations with primary antibodies were performed overnight at 4°C, except for the anti-γ-H2AX antibody, which was incubated for 2 h at room temperature. Secondary antibodies were incubated for 1 h at room temperature.

Immunohistochemical analyses were performed as previously described (Messiaen et al., 2013). Antibodies are listed in supplementary material Table S4.

In situ analysis of DNA fragmentation was performed using the ApopTag Fluorescein In Situ Apoptosis Detection Kit (Chemicon International), according to the manufacturer's recommendations.

Acquisition of images by using microscopy is described in the supplementary material methods.

We are very grateful to one of the reviewers for insights into the interpretation of the different types of tubules found in the adult Bnc2^−/− mice.