Expression of Arf Tumor Suppressor in Spermatogonia Facilitates Meiotic Progression in Male Germ Cells

The mammalian Cdkn2a (Ink4a-Arf) locus encodes two tumor suppressor proteins (p16Ink4a and p19Arf) that respectively enforce the anti-proliferative functions of the retinoblastoma protein (Rb) and the p53 transcription factor in response to oncogenic stress. Although p19Arf is not normally detected in tissues of young adult mice, a notable exception occurs in the male germ line, where Arf is expressed in spermatogonia, but not in meiotic spermatocytes arising from them. Unlike other contexts in which the induction of Arf potently inhibits cell proliferation, expression of p19Arf in spermatogonia does not interfere with mitotic cell division. Instead, inactivation of Arf triggers germ cell–autonomous, p53-dependent apoptosis of primary spermatocytes in late meiotic prophase, resulting in reduced sperm production. Arf deficiency also causes premature, elevated, and persistent accumulation of the phosphorylated histone variant H2AX, reduces numbers of chromosome-associated complexes of Rad51 and Dmc1 recombinases during meiotic prophase, and yields incompletely synapsed autosomes during pachynema. Inactivation of Ink4a increases the fraction of spermatogonia in S-phase and restores sperm numbers in Ink4a-Arf doubly deficient mice but does not abrogate γ-H2AX accumulation in spermatocytes or p53-dependent apoptosis resulting from Arf inactivation. Thus, as opposed to its canonical role as a tumor suppressor in inducing p53-dependent senescence or apoptosis, Arf expression in spermatogonia instead initiates a salutary feed-forward program that prevents p53-dependent apoptosis, contributing to the survival of meiotic male germ cells.


Introduction
The Cdkn2a-Cdkn2b gene cluster (also designated Ink4-Arf) encodes two polypeptide inhibitors (p16 Ink4a and p15 Ink4b ) of cyclin D-dependent kinases (Cdk4 and Cdk6), as well as a third protein (p19 Arf ) that antagonizes the Mdm2 ubiquitin E3 ligase to activate p53 [1]. Although the Ink4a and Ink4b genes likely arose through gene duplication, the structure of the Ink4-Arf gene cluster is highly unusual, as major portions of the p16 Ink4a and p19 Arf proteins are encoded by alternative reading frames of a shared exon [2]. Induction of p16 Ink4a and p15 Ink4b prevents the phosphorylation of the retinoblastoma protein (Rb), thereby maintaining Rb in its growth suppressive state and preventing entry into the DNA synthetic (S) phase of the cell division cycle. In contrast, p19 Arf expression elicits a p53-dependent transcription program that either enforces cell cycle arrest or triggers apoptosis, depending on cell type, physiologic setting, and collateral modulating signals [1]. The Ink4-Arf genes prevent cell proliferation by implementing Rb-and p53-dependent programs that enforce cellular senescence and inhibit tissue regeneration as animals age, but their intimate genetic linkage facilitates their coordinate repression in embryonic and adult tissue stem cells, thereby allowing self-renewal [3,4]. Deleterious growth-promoting stimuli conveyed by activated oncogenes induce Ink4-Arf gene expression and engage both p53 and Rb to counter untoward cellular proliferation. Not surprisingly, bi-allelic deletion of the Ink4-Arf gene cluster abrogates this form of tumor suppression and is one of the more frequent events in human cancer.
Despite its canonical role as an inducer of p53 in response to oncogene signaling, Arf also has p53-independent tumor suppressive activity. Deletion of Arf together with Mdm2 and p53 expands the spectrum and decreases the latency of cancers that spontaneously arise in mice lacking p53, p53 and Mdm2, or Arf alone [5]. Although highly basic p19 Arf (,20% arginine) has been reported to physically interact with more than 25 different proteins other than Mdm2, the role of p19 Arf , if any, in regulating the functions of these putative ''target'' proteins remains controversial [6]. Indeed, numerous reports that p19 Arf regulates such diverse processes as ribosomal biosynthesis, transcription, DNA repair, apoptosis and autophagy in a p53-independent manner have generally relied on experiments performed with cultured cells but have not been buttressed by more extensive in vivo analyses.
Although the Ink4-Arf locus is not detectably expressed under most normal physiologic conditions, eye and male germ cell development provide notable exceptions [7]. Arf is required for early postnatal regression of the hyaloid vasculature in the vitreous, so that Arf-null mice form a retrolenticular mass predominantly composed of pericytes; the abnormal accumulation of these cells disrupts the retina and lens and leads to blindness [8].
Arf inactivation also results in a significant reduction of sperm production through as yet poorly defined mechanisms, although young male mice remain fertile [9]. In contrast, Arf-null females have no discernable reproductive defects.
Spermatogenesis involves a stereotyped sequence of mitotic and meiotic divisions followed by sperm differentiation [10]. In mice, male germ cell progenitors (gonocytes) renew in the testis between days 1-7 postpartum (P1-P7) and generate spermatogonia that line the basement membranes of developing seminiferous tubules [11,12]. At P7-P10, spermatogonia divide to form preleptotene spermatocytes that detach from the basement membrane, are displaced toward the tubular lumen, and enter meiosis-I as primary leptotene spermatocytes. During the extended prophase of meiosis-I, homologous pairs of maternal and paternal chromosomes align to form synaptonemal complexes and exchange genetic information through homologous recombination [13]. Meiosis-I is completed by P18, and is followed rapidly by meiosis-II, and by spermiogenesis (P19-P35), after which the first mature spermatozoa enter the epididymis. As spermatogenesis continues throughout life, spermatogonia within mature seminiferous tubules remain localized on the peripheral tubular basement membrane, whereas spermatocytes, spermatids, and mature sperm are arranged in a sequential order from the periphery towards the lumen [10].
Intriguingly, p19 Arf is transiently expressed in mitotically dividing spermatogonia, but not in the meiotic cells that arise from them [9]. Here, we provide genetic evidence demonstrating that Arf expression initiates a germ cell autonomous program that protects meiotic spermatocytes from undergoing p53-dependent elimination. This physiologic function of p19 Arf directly contrasts with its role as a tumor suppressor in inducing p53.

Arf Is Expressed in Mitotically Dividing Spermatogonia
Lineage tracing experiments in the mouse previously revealed that all viable male germ cells are derived from spermatogonial progenitors in which transient Arf expression neither inhibits proliferation nor subsequent meiotic commitment [9]. Underscoring these findings, expression of p19 Arf in young adult mice is observed in all types of spermatogonia, but not in Sox9-expressing Sertoli cells on the tubular basement membrane or in DAPI-stained intratubular spermatocytes, spermatids, or sperm ( Figure 1A). The fact that p19 Arf is not detected in cells that have detached from the basement membrane implies that Arf expression is extinguished at or near the primary spermatocyte stage of germ cell differentiation. Consistent with this interpretation, the Arf protein does not co-localize with Dmc1 [9], a meiotic recombinase expressed in leptotene spermatocytes. In the mature testis, spermatogenesis occurs in waves along the length of the seminiferous tubules, so that cross sections capture tubules in which dividing spermatogonia are in synchronous phases of the cell cycle. When five month-old mice injected intraperitoneally with BrdU were sacrificed two hours later, dual immunofluorescence analysis revealed that many cells on the tubular basement membrane that had synthesized DNA also expressed p19 Arf ( Figure 1B). Similarly, at P12 when the number of mitotically cycling progenitors exceeds those of more differentiated germ cells, p19 Arf was coexpressed with cyclin D1, a G1 phase marker of proliferating spermatogonia [14] ( Figure 1C), and strikingly, was detected during all stages of mitosis ( Figure 1D, 1E). Therefore, in spermatogonia, p19 Arf is expressed throughout the cell division cycle without interfering with proliferation.
Arf Deficiency Compromises Sperm Production, But Is Compensated by Ink4a Inactivation Total body weights of age-matched wild-type, Arf-null, Ink4a-null, and Ink4a-Arf double-null mice are equivalent, but testis weights of Arf-null animals were reduced relative to those of wild-type controls (Figure 2A), and this was associated with a significant reduction in numbers of mature sperm by the time Arf-null mice were two months old ( Figure 2B). Nonetheless, young Arf-null males remain fertile, and despite the widespread use of independently derived Arfnull strains by us and others, there is no suggestion that young fertile males produce reduced litter sizes. Hence, defects in spermatogenesis were not previously appreciated.
Knock-in of a cDNA encoding Cre recombinase in place of the first Arf exon creates a functionally null Arf allele that expresses Cre in lieu of p19 Arf under the control of the Arf promoter. Crossing Arf Cre/+ females to homozygous males containing Arf alleles flanked by LoxP sites (''floxed'' alleles) specifically results in the inactivation of Arf function in the testis of compound heterozygous Arf Cre/Flox male offspring. Although penetrance of Cre expression is not complete, more than 90% of spermatogonia in the seminiferous tubules of P21 mice had no detectable anti-p19 Arf fluorescence signals [9]. Overall, while p19 Arf was detected in the testes of haplo-insufficient Arf Cre/+ mice, any residual levels of the protein in Arf Cre/Flox testes were too low to be detected by immunoblotting analysis (representative data illustrated in Figure 3), confirming significant Cre-mediated Arf deletion in this setting. We therefore used this ''targeted'' deletion approach to compare the Arf loss-of-function phenotypes of Arf Cre/Flox males with those of Arf 2/2 males. Analysis of testis weights revealed no differences between those of Arf Cre/Flox mice and wild-type controls ( Figure 2C). However, the sperm counts of Arf Cre/Flox animals were reduced to levels approaching those of Arf 2/2 males ( Figure 2D). Notably, the Arf-Cre or Arf-Flox alleles alone had no significant effects in limiting sperm production unless coexpressed in compound heterozygotes. Therefore, tissue-restricted effects of Arf inactivation independently recapitulated those seen in mice that completely lack Arf function.

Author Summary
The intimately linked Arf and Ink4a genes, encoded in part by overlapping reading frames within the Cdkn2a locus, are induced by oncogenic stress, activating the p53 and Rb tumor suppressors, respectively, to inhibit proliferation of incipient cancer cells. As such, expression of the p19 Arf and p16 Ink4a proteins is undetected in most normal mouse tissues. However, p19 Arf is physiologically expressed in mitotically dividing spermatogonia, the progenitor cells that differentiate to form meiotic spermatocytes in which Arf expression is extinguished. We show that, instead of provoking cell cycle arrest or death, Arf expression in spermatogonia facilitates survival of their meiotic progeny, ensuring production of normal numbers of mature sperm. When Arf is ablated, meiotic defects ensue, along with p53-dependent cell death of spermatocytes, indicating an unexpected role of p53 in monitoring meiotic progression. Surprisingly, it is the absence of p19 Arf rather than its induction that enforces p53 expression in this setting. Coinactivation of Ink4a compensates for Arf loss by fueling proliferation of spermatogonial progenitors, but does not correct meiotic defects triggered by Arf loss. Although the Arf and Ink4a tumor suppressors are expected to restrain cellular self-renewal, Arf plays an unexpected role in male germ cells by facilitating their proper meiotic progression. Hormone signaling networks are involved in the proper control of spermatogenesis. Key regulatory gonadotrophins include luteinizing hormone (LH) and follicle-stimulating hormone (FSH) secreted by the anterior pituitary gland, and testosterone produced by testicular interstitial Leydig cells. The considerable day-to-day and even hour-to-hour variation over a 30-fold range in plasma testosterone levels in age-matched mice of a single strain precluded accurate measurements of strain-specific differences, even in a relatively large sample size ( Figure 4) [15]. Importantly, however, no discernable defects in pituitary or Leydig cell development have been observed in Arf-null, Ink4a-null, or doubly-deficient mice, and no significant differences were observed in the ranges of serum FSH and LH among all genotypes examined ( Figure 4). These findings suggest that spermatogenesis defects in Arf-deficient mice are not a secondary consequence of hormonal imbalances.
Unlike Arf-null males, those lacking functional Ink4a instead exhibit increased testis weights and produce higher numbers of sperm than wild type mice (Figure 2A and 2B). Cdk4, the major target of p16 Ink4a protein inhibition in the adult testis, is expressed at maximal levels at the earlier stages of spermatogenesis, where spermatogonia undergoing mitotic cell divisions predominate [16,17], and Cdk4 inactivation compromises male fertility [18,19]. We therefore quantified the in vivo incorporation of BrdU in spermatogonia of young adult wild-type, Arf-null, Ink4a-null, and Ink4a-Arf-null mice by counting stained cells that had entered S phase during a two-hour pulse. The S phase fractions of wildtype and Arf-null spermatogonia did not differ from each other ( Figure 5), implying that the failure of Arf-null mice to produce normal numbers of sperm reflects a loss of meiotic cells or their  Immunoblotting analysis was performed using an antibody against p19 Arf . Whole testis lysates were prepared from two month old mice of the indicated genotypes. Immunoblotting with an antibody to actin was used to control for equal protein loading per lane. doi:10.1371/journal.pgen.1002157.g003 immediate progeny rather than spermatogonia. In contrast, we observed a significant two-fold increase (p,0.0001, Student's ttest) in the relative number of S phase spermatogonia from both strains that lack Ink4a ( Figure 5). Looking only at testis weights and sperm counts in Ink4a-Arf double-null animals, Ink4a inactivation appears to compensate for loss of Arf function ( Figure 2A and 2B), presumably by fueling the production of a greater number of mitotic progenitors. Together, the consequences of these two independent loss-of-function effects rebalance testis size and sperm output in the doubly null strain. In this sense, these two ''tumor suppressor'' genes play opposing physiologic roles in male germ cell development.

Arf Inactivation Leads to p53-Dependent Apoptosis in Primary Spermatocytes
Testes from two month-old Arf-null mice exhibited a significant increase in the numbers of apoptotic (TUNEL-positive) cells when compared to age-matched wild-type controls ( Figure 6A and 6E). The vast majority of apoptotic cells are spermatocytes as judged by the topological relation of TUNEL-positive cells to the expression of the meiosis-specific strand-exchange protein Dmc1, which is expressed during early prophase-I ( Figure 6A). Notably, however, intratubular TUNEL-positive cells were not stained with antibodies to Dmc1, implying that Arf-null cells die during a later stage of germ cell development after Dmc1 expression is greatly diminished. To examine this issue further, we conducted TUNEL staining of meiotic chromosome spreads. Characteristic stages of prophase during meiosis-I can be marked by staining chromosomes with antibodies to synaptonemal complex proteins, such as the axial element component SYCP3 [20], and by several ancillary criteria (see Materials and Methods). Unlike pachytene cells from wild-type mice, those from the Arf-null strain exhibited considerable TUNEL staining ( Figure 6C) with a concomitant reduction in the fraction of Arf-null diplotene spermatocytes ( Figure 6F) that correlated with decreased sperm production ( Figure 2B). Inactivation of Ink4a alone did not trigger spermatocyte apoptosis nor limit apoptosis in the Arf-null background ( Figure 6E) reinforcing the conclusion that the two closely linked genes play fundamentally different roles within the male germline.
Arf 2/2 ; p53 2/2 doubly-deficient males are even more susceptible to spontaneous tumor development than mice lacking either Arf or p53 alone [5]. However, the young tumor-free males produce sufficient viable sperm to remain fertile. Inactivation of p53 restored testis weights and sperm production in Arf-null males (Figure 2A and 2B), prevented the apoptotic elimination of Arf-null germ cells ( Figure 6D and 6E), and restored the number of . Hormone levels in mice of different genotypes. Blood was collected in the afternoon from 15 two to four month-old male mice of each indicated genotype, and sera were analyzed for hormone content. P values calculated using a Student's t-test indicated no significant differences between samples taken from mice of different genotypes. Wide day-to-day and even hour-to-hour fluctuations in plasma testosterone levels yield largely unpredictable changes of as much as 30-fold between individual samples, as determined by radioimmunoassay [19]. Error bars indicate standard deviations from the mean. doi:10.1371/journal.pgen.1002157.g004 diplotene spermatocytes ( Figure 6F). Accordingly, higher levels of p53 were detected in whole testis lysates from Arf-null mice as compared to those in wild-type mice ( Figure 7). Thus, in direct contrast to the role of p19 Arf in triggering a p53 response following abnormal hyperproliferative stress in somatic cells, it is instead the absence of Arf expression in spermatogonial progenitors that impairs the fidelity of meiotic progression and ultimately leads to p53-dependent elimination of Arf-null primary spermatocytes.
Inappropriate c-H2AX Accumulation in Arf-Null Germ Cells Histone H2AX is phosphorylated at serine-139 in response to DNA strand breaks caused by ionizing radiation [21], UV irradiation [22], replication stress [23,24], failure of nucleotide excision repair [25], and at the leptotene stage of meiosis prior to synaptonemal complex formation [26]. H2AX phosphorylation also occurs in spermatocytes in a DNA damage-independent manner during formation of the sex body, a heterochromatic subnuclear domain encompassing the nonhomologous parts of the X and Y chromosomes [27]. Remarkably, staining of testis sections and immunoblotting of whole testis lysates revealed a profound increase in global c-H2AX levels when Arf was inactivated ( Figure 8A and 8B). Again, inactivation of Ink4a neither recapitulated nor ameliorated this Arf-null defect ( Figure 8A and 8B). Germ cells at the periphery of Arf-null seminiferous tubules exhibited the greatest increase in c-H2AX staining, suggesting that more immature cells were the ones most affected ( Figure 8A). Microscopic quantification revealed that the number of c-H2AXpositive spermatogonia, as well as the number of c-H2AX foci per cell, were increased ,2-fold when Arf was inactivated (Figure 9), but the greatest increase in c-H2AX staining was observed in primary Arf-null spermatocytes ( Figure 8A and 8B; see below). The increased c-H2AX in meiotic cells is especially striking because these cells do not normally express p19 Arf when the gene is present (Figure 1).  Meiotic double-strand DNA breaks are induced by the Spo11 transesterase and its accessory factors, which are loaded onto chromatin during the final pre-meiotic S-phase [13]. Autosomal c-H2AX staining is normally observed during the leptotene and zygotene phases of meiosis-I, which are the early stages at which chromatids undergo DNA scission as a prelude to homologous recombination. In contrast, c-H2AX foci are not normally detected by early pachytene (except in the sex body) once homologous synapsis is complete ( Figure 8C, top panels) [28]. Chromosome spreads from meiotic primary spermatocytes from Arf-null males revealed that 150 of 382 individually enumerated pachytene cells (39%) exhibited persistent autosomal c-H2AX foci in addition to normal sex body staining, whereas very few such cells (6.7%) were detected at the diplotene stage ( Figure 8C, bottom panels). It could be that disappearance of c-H2AX is delayed until diplonema, or that cells with aberrantly elevated c-H2AX are preferentially eliminated. The latter interpretation is supported by the prophase I apoptosis and depletion of diplotene cells observed in Arf-null mice ( Figure 6). Therefore, in the absence of Arf, c-H2AX accumulates to higher levels than normal starting in the least mature spermatogonia, continuing into meiotic prophase I, and persisting past the time when it would normally disappear from autosomes. Although inactivation of p53 suppresses the increased apoptosis of Arf-null spermatocytes, c-H2AX persists in cells lacking both of these genes ( Figure 8A, lower right panel). In meiotic chromosome spreads from p53; Arf double-null mice, 20.5% (34 of 166) of diplotene spermatocytes display persistant autosomal c-H2AX immunostaining as compared to 6.7% (5 of 74) of singly Arf-null and less than 1% (1 of 149) of wildtype diplotene spermatocytes. These data underscore the fact that the accumulation of c-H2AX in Arf-null spermatocytes is p53independent, whereas the elimination of defective spermatocytes that retain c-H2AX inappropriately is p53-dependent.

Additional Meiotic Prophase Defects in Arf-Null Primary Spermatocytes
DNA double-strand breaks induced early in prophase I by Spo11 serve as substrates for the strand exchange proteins Rad51 and meiosis-specific Dmc1, which are required for double strand break repair during homologous recombination. Foci of staining using antibodies to Dmc1 ( Figure 10A) and Rad51 ( Figure 10B) were readily observed in zygotene spermatocytes from wild-type mice (left panels) but were fewer and less prominent in their Arfnull counterparts (right panels). The number and average fluorescence intensities of foci in 100 zygotene cells of each genotype were determined using commercial imaging software. In wild-type zygotene spermatocytes, the frequency of Dmc1 and Rad51 foci peaked at 100-125 per cell ( Figure 10C and 10D, respectively, blue bars) and exhibited a broad distribution of relative intensities over a ,10-fold range ( Figure 10E and 10F, blue bars). In contrast, both the number and intensities of Dmc1/ Rad51 foci were significantly reduced in Arf-null cells (average number of foci 6 S.D.: 103646 Dmc1 foci in wild-type vs. 44634 in Arf 2/2 ; 108645 Rad51 foci in wild-type vs. 45633 in Arf 2/2 ; average relative intensities 6 S.D.: 1023964959 Dmc1 foci in wild-type vs. 661163005 in Arf 2/2 ; 1098065783 Rad51 foci in wild-type vs. 563763340 in Arf 2/2 N = 100, p,0.0001, Student's t-test; Figure 10C-10F). An accumulation of Arf-null spermatocytes in zygonema ( Figure 6F) suggests that there may be a delay at this stage before progression to pachytene.
Pachytene cells are normally characterized by well developed synaptonemal complexes that stretch the length of autosome axes and by knob-like accumulation of SYCP3 at telomeres ( Figure 11A). However, 34% of Arf-null cells exhibited defects in synapsis (quantified in Figure 11D), including forked terminal structures and interstitial bubbles on autosomes ( Figure 11B) and complete asynapsis of sex chromosomes (arrow, Figure 11C). In addition, interrupted regions of SYCP3 staining (denoted by arrowheads in Figure 11C) were more frequently observed in meiotic chromosome spreads from Arf-null cells versus those in wild-type cells (191 versus 53 such segments, respectively, in 300 pachytene cells of each genotype). Because synapsis was complete in the majority of Arf-null pachytene cells, we could not distinguish whether the observed defects arose from regions in which synaptonemal complexes did not form at all, or where complexes had formed but subsequently disassembled. Taken together, Arfdeficiency results in a series of abnormalities during prophase I that include reduced loading of the Rad51 and Dmc1 recombinases, defects in synapsis, elevated and persistent c-H2AX expression, and p53-dependent apoptosis, ultimately associated with diminished production of mature sperm.

Discussion
With few exceptions, the Arf tumor suppressor is not expressed in normal tissues of healthy mice but is induced by abnormally sustained and elevated thresholds of proliferative signals, activating a p53 response that opposes the deleterious effects of oncogene activation. Notably, p53 responds to a much wider range of Arfindependent signal transduction cascades triggered by many other forms of cellular stress, including acute DNA damage, to which the Arf promoter does not respond [6]. By converging on p53, these different signaling pathways inhibit cell cycle progression or trigger apoptosis, acting to suppress tumor formation.
We now document a physiological role of Arf in mouse male germ cell development that is distinct from its tumor suppressive functions in key respects. First, Arf is expressed in spermatogonia, but not in the primary spermatocytes that arise from them. Expression of p19 Arf neither arrests spermatogonial mitotic progression nor triggers their p53-dependent apoptosis. However, the absence of Arf expression in spermatogonia leads to p53dependent apoptosis of spermatocytes before they exit meiosis-I. The defect in spermatogenesis is germ cell autonomous and results in a significant reduction in sperm counts by the time Arf-null mice are two months old, although residual sperm production maintains fertility in young males. Thus, expression of Arf in mitotic progenitor cells enhances the survival of their meiotic progeny in which Arf expression is normally extinguished. These features indicate that Arf expression initiates a salutary, feed-forward program that facilitates meiotic progression. Indeed, although Arf and Ink4a are widely viewed to convey tumor suppressive functions that coordinate the activities of the p53 and Rb signaling ''pathways,'' inactivation of Arf and Ink4a in the testes leads to opposing outcomes. Disruption of Ink4a increases the mitotic activity of spermatogonial progenitors to enhance sperm output and, in this respect, compensates for Arf loss of function without eliminating the cellular defects that arise in the Arf-null setting. In short, loss of Ink4a increases the spermatogonial pool size, but without Arf expression, spermatocytes undergo increased apoptosis, returning the number of mature sperm to normal levels.
Homologous recombination during meiosis exchanges genetic information between maternally and paternally derived chromosomes and also guides proper segregation of chromosome pairs to maintain correct chromosome numbers in gametes [13]. During meiosis, in contrast to mitotically diving cells, homologous chromosomes are favored over sister chromatids as templates for recombinational DNA repair. Double-strand DNA breaks are formed by the topoisomerase-II-related transesterase Spo11. This process activates the Atm kinase and leads to phosphorylation of the H2AX histone variant near sites of strand breakage during early prophase I. Binding of the RecA family strand exchange proteins, Rad51 and meiosis-specific Dmc1, to Spo11-induced DNA ends generates filaments that search for and invade homologous duplex DNA molecules, leading to pairing of homologous chromosomes. Loading of Rad51 and Dmc1 is normally reversed by early pachytene when chromosomes are fully synapsed, after which c-H2AX foci are no longer detected.
In the Arf-null setting, a modest but significant increase in c-H2AX staining was first detected in the least mature spermatogonia, and primary spermatocytes displayed accentuated signals that persisted inappropriately into the pachytene stage. Arf-null cells also formed fewer Dmc1/Rad51 foci at zygotene and exhibited focal regions of asynapsis at pachytene. Aberrant Arfnull spermatocytes underwent apoptosis at pachytene, resulting in the emergence of fewer diplotene cells and a significant reduction in sperm output. Importantly, Arf 2/2 ; p53 2/2 double-null pachytene cells also exhibited persistent c-H2AX staining, but these cells escaped elimination. Thus, apoptosis was p53dependent, but aberrant c-H2AX accumulation was not.
Although the underlying mechanisms remain unknown, we consider here two plausible interpretations of this apoptotic arrest. First, it may be that reduced Rad51/Dmc1 focus formation and persistent c-H2AX staining in Arf-null male germ cells connote a defect in DNA repair that then activates p53 through Arfindependent but Atm/Atr-dependent signaling pathways. In this scenario, Spo11-induced DSBs would form at normal levels but Rad51/Dmc1 loading would be impaired such that some DNA damage would persist into pachytene. This might conceivably involve the p53-independent ability of p19 Arf to promote the sumoylation of numerous target proteins by inhibiting the SUMO2/3 protease Senp3 [29][30][31]. SUMO2/3 accumulates at sites of DNA damage in mammalian cells [32,33], and various aspects of DNA repair are regulated by the SUMO conjugation pathway [34]. There is fragmentary evidence that absence of p19 Arf compromises nucleotide excision repair in cultured cells [35,36] raising the possibility that Arf may play an as yet undefined role in promoting homologous recombination. All meiotic mutants that cannot properly synapse homologous chromosomes arrest during pachytene [37], and accompanying defects in sex body formation and failure to properly silence transcription of the sex chromosomes during prophase is itself sufficient to eliminate pachytene cells [27,38]. However, spermatocytes can also undergo apoptosis in direct response to unrepaired Spo11-induced breaks even if sex body formation is normal [27,39]. Where tested, spermatocyte apoptosis in meiotic mutants with chromosome synapsis errors has been found to be p53-independent [40][41][42]. Moreover, Spo11-dependent activated phospho-p53 can be transiently detected from leptonema and zygonema in wild-type male mice, and in Drosophila, p53 activity is prolonged in cells defective for meiotic repair [43]. Thus, it remains a formal possibility that meiotic recombination defects can trigger p53dependent apoptosis.
A second, alternative interpretation rests on the idea that the earlier and less profound accumulation of c-H2AX in Arf-null spermatogonia might be a symptom of an underlying defect affecting chromatin structure or Atm/Atr signaling. The appearance of c-H2AX reflects chromatin modifications that flank sites of DNA damage rather than strand breaks themselves, so the kinetics of c-H2AX formation and dissolution do not necessarily coincide with the appearance and repair of DNA damage [44,45]. Moreover, aberrant Atm/Atr signaling is itself sufficient to activate p53, whether triggered by DNA breaks or not [46]. Thus, it may be that Arf deficiency causes inappropriate Atm/Atr signaling that provokes p53-dependent apoptosis in a DNA damage-independent manner. In this view, the observed meiotic prophase defects in Arf-null spermatocytes may possibly be a separate downstream consequence of this earlier anomaly, and may not be the cause of apoptosis. Regardless of which interpretation is correct, it is important to note that our findings provide strong evidence that p53-dependent monitoring promotes proper meiotic maturation, in addition to the previously documented p53-independent pathway(s). Whatever the underlying mechanisms, the role of Arf in male germ cell development contrasts with the general paradigm of p19 Arf acting as an activator of p53. Instead, it is the absence of Arf in spermatogonia that consequently leads to p53dependent apoptosis of spermatocytes. Studies reported herein did not unnecessarily duplicate previous research, and were undertaken only because suitable non-animal models were unavailable. The number of animals used was consistent with good statistical design. Anesthesia, analgesia and tranquilization were used to relieve pain and distress in accordance with the IACUC recommendations.

Phenotypic Characterization of Mouse Testes and Sperm Count Analysis
Caudal epididymides were harvested before dissection of the testes. For each male mouse, two cauda were minced into 1 ml of Dulbecco's modified Eagle's medium (DMEM) containing 25 mM HEPES buffer (pH 7.5) and 4 mg/ml bovine serum albumin and incubated at 37uC for 20 minutes. Suspensions of sperm were fixed at a 1:25 dilution in 10% formalin and counted on a hemocytometer. All sperm counts were performed between 1:00-3:00 PM. Dissected testes were weighed in pairs.

Analysis and Staging of Meiotic Spreads
Testes were decapsulated and minced in 5 ml of DMEM per testis and transferred to a 15 ml Falcon tube. After further dissociation of the tubules by pipeting up and down, large pieces were allowed to settle to the bottom of the tube by gravity for 10 minutes on ice. One ml of the supernatant, containing a suspension of spermatocytes, was transferred to a 1.5 ml Eppendorf tube and centrifuged for five minutes at 58006 g. The pellet was resuspended in 40 ml of a 0.1 M sucrose solution, and 20 ml of spermatocyte suspension was applied evenly to a slide containing a thin layer of 1% paraformaldehyde (pH 9.2) containing 0.1% Triton X-100. Slides were allowed to dry for two hours at room temperature in a closed humidity chamber before rinsing in Photoflo (Kodak 1464510, diluted 1:250 in doubly distilled H 2 O) and air dried at room temperature. For immunofluorescence, slides were incubated in PTBG (0.2% bovine serum albumin, 0.2% gelatin, 0.05% Tween 20 in PBS) for 10 minutes with shaking. Primary antibodies were diluted in PTBG, applied to the slide, and covered with parafilm before incubation overnight at 4uC in a humidity chamber. Antibodies were directed to SYCP3 (Santa Cruz G-3, 1:500) to mark the synaptonemal axial element [20], to c-H2AX (Cell Signaling 2577, 1:500) to identify sex body formation and sites of DNA damage, and to Rad51 (Calbiochem Ab-1, 1:500) and Dmc1 (Santa Cruz H-100, 1:750) to demonstrate formation of complexes required for DNA strand exchange during homologous recombination. Slides were washed three times in PTBG at room temperature for 3 minutes with shaking. Secondary antibodies, also diluted in PTBG, were applied to slides which were covered with parafilm and incubated at 37uC for one hour in a humidity chamber. Slides were washed three times in PTBG for 3 minute intervals in the dark with shaking and mounted with Vectashield (Vector Labs) containing DAPI. Surface spread spermatocytes were visualized by a Marianas spinning-disc confocal microscope, and images were assembled and analyzed using Slidebook 5.0 SDC software (Intelligent Imaging Innovations, Denver CO). Meiotic spreads from three adult mice (age three months) were analyzed. One hundred spermatocytes were scored each from mouse.
Distinct staining patterns allow for classification of each stage of meiotic prophase [51,52]. Leptotene cells were categorized by short stretches of axial elements accompanied by intense c-H2AX staining throughout the nucleus and the absence of a distinct sex body. Zygotene cells also display intense c-H2AX staining throughout the nucleus and lack a sex body, but can be distinguished by longer stretches of SYCP3 staining, some of which are synapsed. Pachytene cells have fully formed and synapsed axes that appear as thick, continuous SYCP3-stained threads, while displaying intense c-H2AX staining only in the sex body. Dmc1 and Rad51 foci are normally present at leptotene and zygotene, and largely disappear by pachytene. Diplotene cells have c-H2AX localized only to the sex body, but fully formed axes are desynapsing and chiasmata are visible.