Ku70 and non-homologous end joining protect testicular cells from DNA damage

Meiosis is the central process of spermatogenesis by which the genome is bisected to generate haploid gametes. Prior to genome haploidization, homologous chromosomes pair, recombine and separate from each other during the first of two successive divisions that lack an intervening DNA replication (reviewed by Gerton and Hawley, 2005). At the onset of first meiotic prophase, chromosomes receive Spo11-dependent DNA double-strand breaks (DSBs) that are repaired by homologous recombination (HR) to promote crossing over, which ensures homolog separation during the meiosis I division (reviewed by Bannister and Schimenti, 2004; Keeney, 2008). HR in meiosis is facilitated by downregulation of the error-prone non-homologous end joining (NHEJ) DSB repair pathway at the onset and during the early substages of first meiotic prophase I (Goedecke et al., 1999). The NHEJ DSB repair pathway dominates in G0 and G1 phases of the cell cycle when homologous chromatids are not available. NHEJ involves binding of the Ku70/80 heterodimer to broken DNA ends, which are then ligated back together or to other DSB ends in a mutagenic process (for review, see Lieber, 2010).

In mice, Ku70 (also known as Xrcc6) and Ku80 (also known as Xrcc5) disruption leads to deficiency in NHEJ DNA repair, early cellular senescence and compromised growth (Gu et al., 1997; Nussenzweig et al., 1997; Ouyang et al., 1997). In somatic cells, Ku70 is required for full proliferation capacity (Gu et al., 1997; Nussenzweig et al., 1996) and acts as a switch between the HR and NHEJ DSB repair pathways. NHEJ is downregulated in early meiotic cells in which homology-directed repair (HDR) dominates (Goedecke et al., 1999). Ku proteins are again expressed in late pachytene and diplotene spermatocytes, and they mediate repair proficiency in somatic cells of the testis, such as spermatogonia and Sertoli cells (Ahmed et al., 2007).

Besides the well-known function in DSB repair, end-binding Ku proteins are components of the ends of linear chromosomes (Bailey et al., 1999; Celli et al., 2006; Hsu et al., 1999), the telomeres that ensure genomic stability (for review, see Bertuch and Lundblad, 2006; de Lange, 2005). In the mammalian somatic cell nucleus, mammalian telomeres are distributed without an apparent order (Ludérus et al., 1996; Vourc'h et al., 1993), whereas their meiotic counterparts attach to the nuclear envelope early in meiosis (Scherthan et al., 1996) through an interaction with a SUN/KASH protein complex in the meiotic nuclear envelope (Ding et al., 2007; Hiraoka and Dernburg, 2009; Morimoto et al., 2012). Meiotic telomeres move to cluster near the centrosome during the early substages of first meiotic prophase (bouquet formation) (for a review, see Scherthan, 2006). In budding yeast, cells deficient for Ku (Hdf1/2) function display a disturbed three-dimensional telomere location in the vegetative nucleus (Laroche et al., 1998), which leads to altered telomere dynamics at the onset of first meiotic prophase (Scherthan and Trelles-Sticken, 2008). However, the consequences of Ku disruption for mammalian meiosis are largely unexplored. As Ku and the mammalian telomere-interacting protein RAP1 have been shown to inhibit HDR at telomeres (Celli et al., 2006; Sfeir et al., 2010), we made use of Ku70 Rap1 double knockout mice (Sfeir et al., 2010) to determine the combined and single contribution of Ku70 and Rap1 to meiotic telomere dynamics and fidelity of spermatogenesis. Furthermore, we studied the consequences of the absence of these two proteins for the telomere clustering that has been observed in somatic Sertoli cells (Scherthan et al., 2000).

Ku knockout mice have been reported to display overall reduced body and organ size and a reduced life span owing to compromised cellular proliferation capacity (Gu et al., 1997; Holcomb et al., 2007; Nussenzweig et al., 1996). Investigation of Ku70^−/− and Ku70^−/−Rap1^Δ/Δ spermatogenesis revealed, in accordance with the literature, a significantly reduced testis size in all Ku70^−/− and Ku70^−/−Rap1^Δ/Δ animals compared with heterozygous, Rap1^Δ/Δ and wild-type testes (Fig. 1A,B). Investigating spermatogenesis in more detail using DAPI or Hematoxylin and Eosin (H&E) staining revealed an overall normal tubule histology in wild-type and heterozygous testes (Fig. 1C,D), whereas sperm formation and release was compromised in Ku70^−/− and Ku70^−/−Rap1^Δ/Δ stage VI–VII testis tubules (Fig. 1E,F). In agreement, we noted almost empty epididymides in all Ku70^−/− mice (Fig. 1H,I), demonstrating reduced fidelity of spermatogenesis in the absence of Ku70. These observations indicate that Ku70 is epistatic to Rap1, and that Ku deficiency is responsible for reduced testis size and compromised spermatogenesis in the Ku70^−/−Rap1^Δ/Δ condition.

Next, we investigated the expression and localization of the Ku protein in testis sections of wild-type and Ku70-deficient mice by immunofluorescence (IF), which revealed that Ku70 is abundant in the nucleus of Sertoli cells, type A spermatogonia, post mid-pachytene (late) spermatocytes I, secondary spermatocytes and round spermatids as well as in the cytoplasm of metaphases I and II cells (Fig. 2A; supplementary material Fig. S1). Wild-type and Rap1-deficient mouse testes showed the same Ku70 localization pattern (not shown), whereas Ku70^−/− testes lacked any detectable protein (Fig. 2B).

Rap1 knockout mice, by contrast, lacked the Rap1 telomere protein in testis cells, but spermatogenesis was indistinguishable from that observed in wild type (Scherthan et al., 2011; Sfeir et al., 2010).

Because Rap1 and Ku70 inhibit homology-directed DNA repair at somatic telomeres (Celli et al., 2006; Sfeir et al., 2010) and are required for normal meiotic telomere behavior in yeasts (Chikashige and Hiraoka, 2001; Scherthan and Trelles-Sticken, 2008), we investigated whether the simultaneous absence of Ku and Rap1 affects meiotic telomere behavior and attachment to the meiotic nuclear envelope (NE). Investigation of (TTAGGG)_n fluorescence in situ hybridization (FISH)-tagged telomeres (n>1000) in pachytene nuclei of testis tissue sections revealed normal telomere localization to the nuclear periphery (Fig. 2C). IF for SUN1 (Schmitt et al., 2007) together with telomere repeat FISH on Ku70^−/− and Ku70^−/−Rap1^Δ/Δ spermatocytes showed a wild-type SUN1 distribution at all telomeres in 3D-preserved pachytene nuclei from tissue sections (Fig. 2C). Collectively, these data suggest that meiotic telomeres do not require Rap1 and/or Ku70 for their interaction with SUN1 and attachment to the meiotic nuclear envelope.

At the onset of prophase I, mouse telomeres undergo a dynamic redistribution that involves brief clustering during early zygotene stage (bouquet topology) (Scherthan, 2001). Bouquet spermatocytes, which can be assayed by centromere/telomere FISH (Tel/Cen FISH) in testicular suspensions, are rarely (∼0.3% of spermatocytes) encountered in wild-type mouse meiosis (Liebe et al., 2006), but their significance is underpinned by their enrichment in ATM-deficient meiosis (Pandita et al., 1999). When we determined the frequency of bouquet spermatocytes by Tel/Cen FISH in testicular suspensions of control mice, we noted similar frequencies in Ku70-deficient spermatocytes: Ku70^−/− (0.38%, n = 1055), Rap1^Δ/+Ku70^−/− (0.45%, n = 1103) and Rap1^Δ/ΔKu70^−/− (0.4%, n = 1004). The difference between the various genetic backgrounds tested was not statistically significant (P>0.05; Fisher's test). Furthermore, bouquet frequencies were similar to those observed in the single Rap1 knockout and in wild type (Liebe et al., 2006; Scherthan et al., 2011).

In contrast, we noted a significant increase (P<0.01, Fisher's test) in the so-called mid-preleptotene spermatocytes in Ku-deficient genotypes, i.e. in Ku70^−/− (2.3%, n = 1055), in Ku70^−/−Rap1^Δ/+ (2.6%, n = 1103) and in Ku70^−/−Rap1^Δ/Δ (2.6%, n = 1004) spermatocytes, relative to the Rap1 knockout and the wild type, which displayed 0.8% mid-preleptotene spermatocytes (n = 1164). These data corroborate earlier studies that observed a significant increase of mid-preleptotene but not bouquet spermatocytes in the Ku80 knockout mouse (Liebe et al., 2006).

Our data establish that Rap1 and Ku70 are both dispensable for meiotic telomere attachment and clustering in mouse meiosis, whereas passage through the so-called mid-preleptotene stage appears to be prolonged in the absence of Ku70 and NHEJ.

To investigate whether the increased mid-preleptotene levels in Ku70^−/− testes indicate elevated dsDNA damage in pre-meiotic cells, we performed immunostaining for the DSB repair markers γH2AX (Rogakou et al., 1998) and 53BP1 (Huyen et al., 2004) in paraffin-embedded tissue sections of Ku70 single and double knockout testes. Surprisingly, we noted a significant elevation of the numbers of DSB-associated foci in B spermatogonia of Ku70^−/− and Ku70^−/−Rap1^Δ/Δ mice (Fig. 3A,B) relative to the frequency in B spermatogonia of Rap1^Δ/Δ, heterozygous and control testes (Fig. 3C). These findings indicate that rapidly dividing Ku-deficient B spermatogonia with compromised NHEJ are accumulating replication-dependent DSBs prior to their entry into prophase I.

We next investigated the progress of DSB occurrence and DNA repair in Ku- and Rap1-deficient spermatocytes. HR is the dominant repair pathway during much of prophase I and repairs endogenous DSBs that are formed by Spo11 in leptotene chromatin, resulting in crossovers between homologous chromosomes. Leptotene spermatocytes show extensive histone H2AX phosphorylation in their nuclei and H2AX phosphorylation regresses with the progress of HR repair to the sex body of pachytene spermatocytes (Barchi et al., 2005; Mahadevaiah et al., 2001). Delayed repair progression at some DSB sites can lead to a few large, synaptonemal complex-associated γ-H2AX foci during the late pachytene stage of prophase I (Ahmed et al., 2010; Chicheportiche et al., 2007). These foci possibly relate to delayed repair progression as shown by RPA and γH2AX colocalization in mouse and human late pachytene meiocytes (Ahmed et al., 2010; de Vries et al., 2013; Roig et al., 2004). To determine whether carryover of DNA damage from pre-meiotic S phase of B spermatogonia leads to altered DNA repair progression in late Ku70^−/− spermatocytes, we examined surface-spread and SYCP3-stained Ku70- and Ku70 Rap1 doubly deficient late pachytene and diplotene spermatocytes for the presence of large γH2AX foci (Fig. 4). It was found that γ-H2AX signals appeared normally during leptotene stage and regressed during zygotene and early pachytene stages to the XY body in all genotypes (data not shown). In late pachytene and diplotene spermatocytes, in which NHEJ repair is reactivated (Ahmed et al., 2010; Goedecke et al., 1999), we noted a moderate but significantly elevated number of spermatocytes displaying on average two γ-H2AX foci/late Ku70^−/−Rap1^Δ/Δ spermatocyte nucleus (n = 55) (Fig. 4A,B) and in Ku70-deficient diplotene spermatocytes (n = 43) (Fig. 4C,D), as opposed to the one large focus/spermatocyte seen on average in wild-type (n = 49) and Rap1^Δ/Δ (n = 66) late pachytene and diplotene spermatocytes.

To investigate this matter further, we co-stained Mre11 and γH2AX in testis tissue sections of all genotypes and determined the number of γH2AX foci in 3D image projections of late pachytene and diplotene spermatocytes of stage X tubuli (Fig. 4E,F). Comparison of the γH2AX foci frequency per diplotene spermatocyte from five Ku70^−/− mice (n = 869 cells) revealed 0.082 (±0.06, s.d.) γH2AX foci/nucleus, whereas there were only 0.061 (±0.03) foci/nucleus in control diplotene spermatocytes (n = 896; four mice), representing a highly significant difference (t-test; P<0.001). These data agree with our γ-H2AX focus analysis in surface-spread spermatocytes and suggest that some DSBs in prophase I might go unrepaired in absence of Ku70/NHEJ. The role of Ku70 in repairing Spo11-dependent DSBs will require further analysis of a Ku70 Spo11 double knockout mouse.

To address specifically DSB repair events at meiotic telomeres in the mutants, we also determined the frequency of occurrence of large (L) telomeric γ-H2AX foci at synaptonemal complex (SC) ends in ≥25 surface-spread late spermatocytes (>920 telomeres per genotype tested). The average colocalization between telomeres and L γ-H2AX foci at SC ends per wild-type late pachytene cell was 0.19 (±0.4 s.d.) and 0.15 (±0.36) in Rap1^Δ/Δ; 0.33 (±0.48) in Ku70^−/− and 0.20 (±0.41) in Ku70^−/−Rap1^Δ/Δ spermatocytes, indicating insignificant differences (t-test) between the genotypes studied. These results suggest that the single or combined deletion of Ku70 and Rap1 has no effect on HR at meiotic telomeres and agrees with the observations made in the Rap1^Δ/Δ mouse (Scherthan et al., 2011).

To test further whether Ku70 deficiency has an impact on chromosome pairing and segregation at meiotic metaphases, we investigated SC formation and X and Y chromosome pairing in pachytene and metaphase I cells (Fig. 5). SC staining with SCP3 antibodies revealed no overt differences in SC formation among the genotypes studied (Fig. 4A–D; data not shown), suggesting normal homolog pairing. X and Y chromosomes share only a small site of homology, the pseudoautosomal region (PAR) (reviewed by Otto et al., 2011), which undergoes late pairing and forms an obligate crossover in nearly all meioses (Kauppi et al., 2011). If there were problems in pairing or recombination we would expect defects in XY segregation. Studying XY distribution in pachytene and MI cells by two color XY chromosome painting in tissue sections revealed a normal association between the X and Y chromosomes in late pachytene and diplotene spermatocytes (Fig. 5A): paired (touching) X and Y chromosome paint signals were observed in 94.1% of wild type (n = 85), in 94.2% of Ku70^−/− (n = 154) and in 94.7% of Ku70^−/−Rap1^Δ/Δ (n = 95) late spermatocytes, the differences being statistically insignificant (P = 1; Fisher's test).

At metaphase I of meiosis, X and Y chromosome-specific signals were touching each other in 95.6% of wild-type MI cells (n = 45), in 92.7% of Ku70^−/− MI cells (n = 41) and in 91.9% of Ku70^−/−Rap1^Δ/Δ MI cells (n = 49) (Fig. 5B), with the difference between wild type and all Ku70^−/− genotypes being insignificant (P = 0.67; Fisher's test). Nonetheless, the slightly more XY univalents in the Ku70^−/− condition align with our data showing increased frequency of DSBs in late prophase (diplotene) spermatocytes.

To investigate checkpoint activation in mutant testes, we stained tissue sections of Ku70^−/− and Ku70^−/−Rap1^Δ/Δ testes for apoptosis using the TUNEL assay (Fig. 6A–D). In the testes of Ku70^−/− and Ku70^−/−Rap1^Δ/Δ mice, the frequency of apoptotic cells in stage IV-VI tubules, which are subject to HR checkpoint function (Barchi et al., 2005; de Rooij and de Boer, 2003), was similar for each genotype (Fig. 6A,B,E). This suggests that ectopic DSBs carried over from pre-meiotic cell cycles are processed by HR repair in a manner that allows for stage IV checkpoint bypass.

However, apoptosis in stage XII-I tubules that contain meiotic metaphases (Fig. 6C,D) was significantly increased (Fig. 6F), indicating an increased demise of Ku-deficient meiotic metaphase cells. Investigation of 3D image projections of metaphase I cells in stage XII tubules for the presence of γH2AX foci (supplementary material Fig. S2) revealed an average of 0.06 γH2AX foci per Ku70^−/− meiotic metaphase (n = 100) and 0.036 foci per meiotic metaphase in the wild type (n = 137). Again, there is a slight (1.4-fold) but insignificant (P = 0.54; Fisher's test) increase of meiotic metaphases with γH2AX foci in the Ku70^−/− condition.

To investigate further the cause of an activated spindle assembly checkpoint (SAC) in the Ku70^−/− condition, we examined the morphology of apoptotic MI cells for the presence of univalent chromosomes and for signs of congression failure that might activate the SAC, as noted on mouse oocytes (Zhu et al., 2010). To this end, we obtained maximum projection images from three-dimensional image stacks of TUNEL-positive apoptotic MI cells and investigated them for their chromosome distribution by DAPI staining where possible. We observed univalents off the metaphase plate (Fig. 6D, yellow arrow) in 18.8% of apoptotic MI cells in the Ku70^−/− background and 16.3% of wild-type and Rap1-deficient meiotic metaphase cells, a non-significant difference (P = 0.86, Fisher's test). However, the Ku70^−/− apoptotic MI cells displayed 35.6% of cells (n = 76) with numerous non-aligned chromosomes (Fig. 6D, red arrows), whereas in the controls congression failure was seen in 22.1% of apoptotic MI cells (n = 153), a significant difference (P = 0.005). This suggests that the absence of Ku70 leads to the activation of the SAC in meiotic metaphase cells when chromosome congression is still in process. Ku deficiency might cause metaphase cell death as a result of unrepaired DSBs carried over from late prophase I and/or a defective metaphase structure (Higashiura et al., 1992). Further analysis will be required to understand this in detail.

Spermatogenesis is under the control of Sertoli cells, which direct meiotic differentiation and are non-proliferative in the adult testes (see Petersen and Soder, 2006). Because of their non-proliferative nature, they depend on NHEJ to repair DSBs. Our analysis of DNA damage surprisingly revealed large 53BP1-positive DNA damage foci in the conspicuous nucleus of Ku70^−/− and Rap1^Δ/Δ Ku70^−/− Sertoli cells (Fig. 7), with >14.7% of Ku^−/− Sertoli cells expressing on average 1.22 DNA damage foci (range 1–6) per 53BP1 foci-positive cell (Fig. 7; Table 1). Co-staining of 53BP1 with either γ-H2AX (Fig. 7C,D) or activated (phospho-Ser-1981) ATM revealed a colocalization of ATMp and its substrate γ-H2AX at 53BP1 foci in the Sertoli cell nucleus (Fig. 7E,F), demonstrating an ongoing DNA damage response (DDR). Because dsDNA damage foci were only occasionally observed in Sertoli cells of all other genotypes investigated (Fig. 7A; Table 1), we conclude that the absence of NHEJ leads to the presence of persistent DNA damage also in differentiated Sertoli cells.

Sertoli cell nuclei display clustering of proximal telomeres near conspicuous chromocenters of pericentric heterochromatin (Scherthan et al., 2000). Because Rap1 and Ku70 protect telomeres from unwanted HR (Sfeir et al., 2010), deprotection of telomeres can lead to DDR activation and 53BP1 accumulation at dysfunctional telomeres (Takai et al., 2003). Thus, we investigated whether the large DNA damage foci in Rap1^Δ/Δ Ku70^−/− and Ku70^−/− Sertoli cells represent deprotected telomeres. Co-staining for 53BP1 by IF and telomeres by FISH revealed that the large DNA damage foci do not generally colocalize with the smaller telomere signals in Sertoli cell nuclei (Fig. 7G,H), ruling out the possibility that DNA damage foci in NHEJ-deficient Sertoli cell nuclei represent deprotected telomeres.

Here we observed that Ku70 disruption leads to reduced testis size, compromised fidelity of spermatogenesis and DNA repair, compromised sperm production and elevated metaphase I apoptosis. In contrast, combined or single deficiency of Ku70 and RAP1 did not affect meiotic telomere attachment and dynamics leading to normal telomere clustering (bouquet formation) in any of the mutants analyzed. These observations match with normal telomere dynamics in Rap1^Δ/Δ spermatogenesis (Scherthan et al., 2011) and normal telomere clustering in the Ku80 knockout mouse (Liebe et al., 2006). The observed normal meiotic telomere dynamics in mice contrasts with the situation in single-celled eukaryotes, in which yRap1 and yKu70 deficiency deteriorates meiotic telomere behavior (Chikashige and Hiraoka, 2001; Scherthan and Trelles-Sticken, 2008). Hence, our observations in the mouse establish that the Ku protein complex and Rap1 are dispensable for meiotic telomere dynamics in mammalian meiosis.

As in the Ku80 knockout mouse (Liebe et al., 2006), Ku70^−/− spermatogenesis was found to exhibit a significantly elevated level of so-called mid-preleptotene spermatocytes relative to wild-type and Rap1-deficient meiosis. It had been reasoned that the increase in mid-preleptotene spermatocytes is related to carry over of replication-dependent DNA damage into prophase I (Liebe et al., 2006). Along this line, it has been shown that yKu is required for recovery from replication stress (Miyoshi et al., 2009). Here, we observed that Ku70-deficient mice exhibit increased levels of dsDNA damage in somatic testis cells, as indicated by 53BP1, γ-H2AX and pATM foci in B spermatogonia and Sertoli cells. B spermatogonia proliferate and enter first meiotic prophase after a lengthy S phase. Our findings thus highlight the importance of Ku/NHEJ for repairing replication-dependent DNA damage prior to the entry into prophase I.

Sertoli cells also displayed significantly increased DNA damage. They nurse and direct spermatogenic cells (see Petersen and Soder, 2006) and proliferate before puberty after which they arrest (Hayrabedyan et al., 2012). Because they are in cell cycle stage G0/1, Sertoli cells rely on NHEJ for DNA repair. The presence of DNA damage-responsive signaling molecules, such as ATMp and 53BP1, at the large nuclear DNA damage foci of Sertoli nuclei indicates an ongoing DNA damage response. Deprotection of telomeres leads to a DDR involving ATM activation and 53BP1 accumulation at telomeres (Takai et al., 2003). However, Ku70^−/− Sertoli cell telomeres did not colocalize with the DNA damage foci, suggesting that the absence of Ku70 or Rap1 does not lead to deprotected telomeres in this cell type. Thus, the formation of large persistent dsDNA damage foci in the absence of NHEJ is likely to relate to the clustering of replication-dependent DNA damage after prepuberal Sertoli cell proliferation. Clustering of DSB-harboring chromatin domains has, for instance, been noted in α-irradiated somatic cells (Aten et al., 2004). Furthermore, it is possible that the action of reactive oxygen species might have induced accumulation of DNA damage in this cell type. However, because other NHEJ-deficient testicular cells did not display such large DNA damage foci, we consider it likely that these damage foci result from replication stress during the prepuberal Sertoli proliferation that remains irreparable in absence of NHEJ, the major repair pathway of cell cycle-arrested Sertoli cells (Ahmed et al., 2009; Ahmed et al., 2007).

In all, it appears that NHEJ is required for DNA repair in somatic cells of the testis, where Ku70 colocalizes with the Mre11 protein and acts as a switch between NHEJ and HR repair pathways. In spermatocytes, Ku is downregulated during the early stages of meiotic prophase, i.e. in leptotene to mid-pachytene spermatocytes, a time frame during which meiotic DSBs are repaired by HR (Goedecke et al., 1999). In Ku70^−/− mice, we noted the absence of an activated meiotic (recombination) checkpoint in stage IV tubules, normal synapsis and a regression of the Spo11- and DSB-dependent γH2AX signal to the XY body of pachytene and diplotene spermatocytes. In contrast to the Ku70^−/− testis, SCID mice, which carry an inactivating mutation in the DNA-PK catalytic subunit and are compromised in NHEJ, display numerous apoptotic spermatocytes at tubule stage IV, indicating activation of the recombination-dependent meiotic checkpoint (de Rooij and de Boer, 2003; Hamer et al., 2003). The absence of elevated stage IV pachytene apoptosis in our Ku70^−/− mice indicates that HR processes the DSBs carried over from pre-meiotic cell cycles to a state that allows stage IV checkpoint bypass. This distinguishes them from SCID mice, which display reduced DNAPKcs activity and elevated stage IV apoptosis (Hamer et al., 2003).

However, there was a significant elevation of large γH2AX foci in diplotene Ku^−/− spermatocytes, indicating some unrepaired DSBs in these late prophase cells. The presence of γH2AX foci in low frequency meiotic metaphase cells indicates that HR repairs most, but not all, of the DSBs that are present during prophase I. The observed 1.4-fold increase of meiotic metaphases with γH2AX focus signals (possibly marking unrepaired or not completely repaired DSBs) might explain the ∼1.6-fold increase in meiotic metaphase apoptosis in Ku70^−/− stage XII tubules noted by TUNEL staining. A cross talk between the DNA damage response and the spindle assembly checkpoint has been noted previously (Choi and Lee, 2008; Maringele and Lydall, 2002; Yang et al., 2012) and might be instrumental for this situation.

It is noteworthy that Ku70 is highly expressed in the cytoplasm of meiotic metaphase I and II cells and in mitotic metaphase cells where it also localizes to the surface of condensed metaphase chromosomes (Higashiura et al., 1992; Koike et al., 1999). Furthermore, Ku70 seems to play a role in apoptosis regulation, possibly through its interaction with BAX (Amsel et al., 2008; Harrington et al., 1992) and by influencing the cell survival transcription network (Fell and Schild-Poulter, 2012). In all, absence of Ku in metaphase cells could lead to apoptosis by pleiotropic effects influencing DNA damage response, SAC pathways and survival networks. How this is regulated requires further investigation.

In summary, our data indicate that Ku/NHEJ contributes to the repair of replication-dependent and exogenous DNA damage in somatic testis cells, and might also be required for the cleanup repair of some persistent Spo11-dependent DSBs in late meiotic prophase cells and for a normal metaphase structure. By these mechanisms, Ku ensures the fidelity of spermatogenesis.

Rap1 knockout mice were produced by a conditional knockout strategy of exon2 (Rap1^Δex2/Δex2) (Sfeir et al., 2010) and denoted Rap1^Δ/Δ (Scherthan et al., 2011; Sfeir et al., 2010). Rap1^Δex2/Δex2 targeted mice were maintained in a C57BL/6J background. Rap1^Δ/Δ and Ku70 knockout mice (Gu et al., 1997) were bred and crossed to give double knockout mice as described elsewhere (Sfeir et al., 2010).

Testes were recovered from mice immediately after sacrifice by cervical dislocation. Usually one testis was paraffin embedded according to standard procedures and the other was frozen in isopentane (Sigma) until spreading or suspension preparation. Testicular preparations, surface spreading, immunostaining as well as telomere and centromere fluorescence in situ hybridization (FISH) for bouquet stage investigation and chromosome painting were carried out as described in detail elsewhere (Liebe et al., 2006; Scherthan, 2009). Differentially labeled X and Y mouse chromosome paint probes were obtained from MetaSystems (Altlussheim, Germany) and hybridized to tissue sections as described (Scherthan et al., 1996).

TUNEL analysis was performed to detect apoptotic nuclei in 5-µm paraffin-embedded sections (Ahmed et al., 2007) according to the manufacturer's protocol (In Situ Cell Death Detection Kit, POD; Roche Diagnostics GmbH, Mannheim, Germany).

The following affinity-purified antibodies were used in the immunostaining experiments: rabbit anti-SYCP3 (1∶250, Abcam, UK), rabbit anti-γ-H2AX (1∶500, Millipore, Schwalbach), goat anti-Ku70 (sc-1487, 1∶10, Santa Cruz, Heidelberg). Guinea pig anti-SUN1 was a kind gift of M. Alsheimer, University of Würzburg, Germany. Antisera were diluted in PBS, 0.1% Tween 20, 0.2% BSA, 0.1% gelatin (PBTG). All antibodies were tested in individual staining reactions for their specificity and performance. Controls without primary antibodies were all negative (not shown).

Immunostaining was performed as described earlier (Scherthan et al., 2011). Preparations were mounted in antifade solution (Vectashield, Vector Labs) containing 0.5 µg/ml DAPI (Sigma) to reveal nuclear DNA. In mouse testicular suspension preparations, the meiotic prophase stage-specific distribution of SYCP3 proteins and/or DAPI-bright heterochromatin clusters (Liebe et al., 2006; Scherthan et al., 1996) was utilized to identify spermatocytes at various stages of prophase I.

Microscopy was carried out using a Zeiss Axioplan 2 epifluorescence microscope (Carl Zeiss, Oberkochen) equipped with single band pass filters for excitation of green, red and blue fluorescence (Chroma Technologies, Bellows Falls) and 10×, 40×, 63× and 100× plan-Neofluoar lenses. Three-dimensional evaluation of immunostained nuclei was performed in some experiments by carefully focusing through the nuclei using a 63× lens. Digital black and white images were recorded with a cooled CCD camera and merged with RGB images using the ISIS fluorescence image analysis system (MetaSystems). In some cases, a MetaFer image analysis system was used to automatically record image stacks and to obtain maximum projection images from tissue sections. In this case, image collections were then scanned interactively for scoring.

We thank Titia de Lange, Rockefeller University, and H.H. Ropers MPI-MG, Berlin, for continuous support; and M. Alsheimer, University of Würzburg, Germany, for providing SUN1 antiserum. H.S. acknowledges technical support from G. Schrock and C. Nowak in the late phase of the project.