Immunofluorescent characterization of meiotic recombination in human males with variable spermatogenesis



Marieke de Vries, Department of Obstetrics and Gynaecology, Radboud University Nijmegen Medical Centre, P.O. Box 9101, 6500 HB Nijmegen, The Netherlands. E-mail:


Homologous recombination is the key to meiotic functioning. The basis of this process is provided by numerous SPO11-induced DNA double-strand breaks. Repair of these breaks occurs via the crossover (CO) and non-crossover (NCO) pathways. By means of immunofluorescence staining of Replication protein A (RPA) and MutL homolog 1 (MLH1) in combination with the DNA damage marker γH2AX, we studied transitional (CO and NCO) and late (CO) recombination nodules, respectively. Testicular samples were from non-obstructive azoospermic probands (testicular spermatozoa were found) and probands that had a history of normal fertility prior to a vasectomy. All probands were ICSI candidates. γH2AX foci mostly colocalized with delayed transitional nodules (RPA) for which variation was found among probands. Highest incidences of colocalization were found in patients. The level of MLH1 signal intensity was lower in probands who showed more frequent γH2AX RPA colocalization in late pachytene, suggesting communication between the CO and NCO pathways. Our results suggest the presence of a genetic risk pathway for children conceived from non-obstructive azoospermic probands and urge for follow-up studies investigating the level of recombination involved de novo mutations in these children.


Non-obstructive azoospermia (NOA) is the most severe form of male infertility. Several degrees have been described, ranging from Sertoli cell only syndrome and maturation arrest to focal (hypo)spermatogenesis (Levin, 1979). Spermatozoa are present in the testis in about 50% of cases (Vernaeve et al., 2006; Jarvi et al., 2010). From the introduction of ICSI (intra-cytoplasmic sperm injection) in 1992 (Palermo et al., 1992) on, these are used to fulfil a couple's child wish (TESE-ICSI) (Bonduelle et al., 1999). NOA can be either congenital or acquired (Irvine, 1998), but in most cases the origin of disturbed spermatogenesis is unknown. Because of the often encountered focal nature, a spermatogonial stem cell problem might be implicated. The histological characterization of testicular biopsies from NOA patients indicates a large variation in especially the quantitative aspects of spermatogenesis (Levin, 1979; McLachlan et al., 2007). In IVF laboratory practise, usually a small biopsy of the testis suffices to safeguard spermatozoa for ICSI. Here, we have taken advantage of this situation to study meiotic recombination at the cellular level in a series of patients, using the remnant biopsy material not needed for procreation.

Meiosis comprises about 1/3 of the duration of spermatogenesis (Heller & Clermont, 1964). During the leptotene, zygotene and pachytene stages of first meiotic prophase, crossing-over between the homologous chromosomes takes place. This ensures segregation, which leads to haploidization and also increases genetic diversity. For crossing-over, pairing and synapsis of the homologous chromosomes is essential and depends on the formation of numerous DNA double-strand breaks (DSBs) which occur at so-called recombination hotspots (Serrentino & Borde, 2012). Breaks are generated by SPO11, a topoisomerase-like transesterase protein, at the start of meiosis and become subsequently gradually repaired during the first meiotic prophase I stages (for recent reviews see (Handel & Schimenti, 2010; Inagaki et al., 2010)). The processing of DSBs creates single-strand 3’ overhangs, instrumental for the homology search, hence bivalent formation (Handel & Schimenti, 2010; Inagaki et al., 2010; Lichten & de Massy, 2011).

For repair of the SPO11-induced breaks, two routes exist, one leading to a crossover (CO) (exchange with flanking markers) and the other leading to a non-crossover (NCO) (exchange without flanking markers). In the case of a CO, a joined DNA molecule named a double Holliday Junction (dHJ) is assumed to be formed [for a review see (Schwartz & Heyer, 2011)]. In the case of a NCO, the break is preferentially repaired via synthesis-dependent strand annealing, only resulting into gene conversion [yeast (McMahill et al., 2007), an opinion gaining terrain (Bugreev et al., 2011)]. In mouse (Guillon et al., 2005) and human (Jeffreys & May, 2004) a molecular analysis of the recombination products of a hotspot for SPO11 DNA DSB induction showed CO and NCO to originate from the same initiating event, of which the processing tract is longer in case of a CO. Which repair route is taken, CO or NCO, can be influenced by single nucleotide polymorphisms present at the hotspot (Sarbajna et al., 2012).

Using fluorescence microscopy, repair can be followed from early zygotene on by the appearance of foci which, depending on their phase of development and protein composition, are called early, transitional, and late recombination nodules (Moens et al., 2007). The DNA repair enzymes RAD51 and DMC1 are representatives of early recombination nodules; RPA, MSH4/5 and BLM are representatives of transition nodules and MLH1/3 of late recombination nodules. These nodules initially are associated with the chromosomal cores (named axial elements) that, when the homology search is successful, participate in the synaptonemal complex (SC). This complex is often compared with a zipper [for recent reviews see (Costa & Cooke, 2007; Yang & Wang, 2009)]. In the normospermic human male, progression of recombination has been described by Oliver-Bonet and co-workers using immunofluorescence (IF) of a protein set involved in recombination nodules (Oliver-Bonet et al., 2005).

Gamma H2AX denotes the phosphorylation of the H2A variant H2AX at serine 139, a post-translational modification most often associated with DNA repair [for a recent review see (Xu & Price, 2011)]. During male meiotic prophase in mammals, the repair of SPO11-induced DSBs is not systematically visualized by the phosphorylation of H2AX in neighbouring chromatin. In the male mouse (Chicheportiche et al., 2007) as well as in the human male (Roig et al., 2004) typical γH2AX signals, originating from the SC and protruding along the chromatin [termed L-foci (Chicheportiche et al., 2007)], represent a small subset of DNA DSB repair events at the pachytene stage. In the one human male of proven fertility studied, L-foci were found to be able to colocalize with RPA but not with MLH1 (only small γH2AX signals on MLH1 foci in 10% of nuclei) (Roig et al., 2004). In the mouse, L-foci did not colocalize with MLH1 either (Chicheportiche et al., 2007). In the human female, a significant association of γH2AX with MLH1 foci, decreasing from 40% at early pachytene to below 7% at late pachytene was reported by Lenzi et al. (Lenzi et al., 2005). In the mouse germline, more intense γH2AX signals can be induced by ionizing irradiation [pachytene (Chicheportiche et al., 2007), pachytene and round spermatids (Ahmed et al., 2010)], demonstrating their DNA DSB dependence. Increased numbers of L-foci are also observed in mice knock-out for genes involved in the processing of homologous recombination repair intermediates such as Rad54 (Ahmed et al., 2010), Blm (Holloway et al., 2010), Btbd12 (Holloway et al., 2011), Mus81 (Holloway et al., 2008), Ercc1 and p53 (Paul et al., 2007). These results provide strong evidence for the fact that interfering with recombination leads to an increased chance for γH2AX L-foci to develop.

We have focussed our attention on meiotic recombination because of a number of interrelated observations. (i) In NOA patients, often a surplus of the early first meiotic prophase stages leptotene and zygotene is noted [(de Boer et al., 2004) and reviewed in (Martin, 2008)] indicating a problem with the homology search. (ii) in previous reports, an increased frequency of synaptic problems at pachytene was found in this category [reviewed in (Martin, 2008; Sun et al., 2007a)], as well as (iii) lower numbers of crossing-over (Gonsalves et al., 2004; Sun et al., 2007a) and (iv) increased levels of meiotic non-disjunction [reviewed by (Harton & Tempest, 2012; Martin, 2008)]. Pregnancy rates by TESE-ICSI are lower compared with ICSI using ejaculated spermatozoa or spermatozoa from the caput epididymis (Boitrelle et al., 2011), but are regularly achieved. This, in theory, offers an opportunity to estimate the range in recombination nodule variation compatible with procreation.

To obtain a first insight we have investigated meiotic recombination in NOA males by means of recombination nodule processing and colocalization of nodules with γH2AX L-foci at the pachytene stage. We wanted to study γH2AX (damage) marked recombination intermediates to gain insight into the incidence of problematic recombination events. Problems during DNA DSB repair potentially increase the risk of introducing mutations in spermatozoa and subsequently in the offspring (Woldringh et al., 2009).

Our results show a relation between delayed processing of recombination intermediates and γH2AX signalling. We identified probands with a higher incidence of delayed recombination intermediates. Our data also suggest cross-communication between the CO and NCO routes of meiotic DSB processing.

Materials and methods

Human testis material

Testis material was obtained from five men with NOA willing to conceive, who underwent a testicular biopsy for sperm retrieval (TESE: testicular sperm extraction), which was successful. Spermatozoa were used for oocyte fertilization via ICSI. The diagnosis of NOA was established after the repeated absence of spermatozoa in the ejaculate, an elevated FSH level (>15 IU/mL) and/or no indications for obstruction (see Table 1 for patient details). Testis material from obstructive azoospermia men of proven fertility with a history of a vasectomy and absence of spermatozoa in the caput epididymis, who wished to conceive again, was used as control material (Table 1). We realize that in these men, despite the many reassuring reports on testis histology, a proper control testicular environment is not present. In a pilot project remnant biopsy material was used from six patients (P6-11, Table S1). All probands were tested for karyotype abnormalities and AZF deletions and none were found. Biopsies were taken following the procedure of Silber (Silber, 2000). From all probands a drop of spermatogenic cell suspension was smeared on a microscope slide prior to sperm retrieval. Cells were Giemsa stained and pachytene spermatocytes, spermatozoa and Sertoli cells were counted. In Table S1, ratios between pachytene nuclei and mature elongated spermatids (spermatozoa) and between mature elongated spermatids and Sertoli cells are given. The latter ratio indicates the spermatogenic activity of the tissue sampled, whereas the former ratio the efficiency of the production of mature spermatids per meiotic cell. Although not strictly comparable with histological studies, our data are consistent with those published (Rowley & Heller, 1971; Zhengwei et al., 1998; McVicar et al., 2005). For control men 1 and 2 histology of the biopsy was also evaluated by the Johnsen score (Johnsen, 1970) and found to be normospermic (scores of 9.1 and 9.5, respectively, with tubule scores of only 9 and 10). Remnants of the testicular samples were available for research after successful sperm retrieval. All men signed an informed consent for participation in a project to evaluate TESE-ICSI treatment, which was approved by the Dutch Central Committee on Research Involving Human Subjects (CCMO – NL12408.000.06).

Table 1. Proband details
Patient/controlAgeDiagnosisFSH level (IU/mL)No. of ICSI cyclesPregnancy (at ICSI cycle no.)
  1. Clinical data from patients (P) and controls (C). In every ICSI cycle, several oocytes are fertilized and one or two best quality embryos are transferred to the uterus. Ongoing pregnancy is defined as a positive heartbeat by ultrasound at 12 weeks after embryo transfer. Fertility treatment was not started for control 1 at the time of submission.

P127NOA311Yes (1)
P229NOA32Yes (2)
P337NOA133Yes (3)
C248OA5.51Yes (1)

Surface spread preparations

Nucleus spreads were made as described by Peters et al. (1997) with minor modifications. Briefly, a suspension of spermatogenic cells was made by crushing the remnant seminiferous tubuli with two ribbed forceps in a drop of SIM [spermatocyte isolation medium (Heyting & Dietrich, 1991)]. Remaining tubular remnants were separated from the cell suspension by a quick spin (25 G). Supernatant was transferred to a clean tube, centrifuged for 7 min (159 g), and the pellet was resuspended in 1 mL SIM. An equal volume of a hypotonic solution (17 mm sodium citrate, 50 mm sucrose, 30 mm Tris–HCl pH 8.2) was added for 7 min. Cells were centrifuged again (7 min, 159 g) and resuspended in 100 mm sucrose (pH8.2) at a concentration of 10–15 × 106/mL. Two 5 μL drops were pipetted onto a PFA (1% PFA, 0.15% triton-X-100 pH 9.2) fluid coated microscope slide which was placed in a levelled humid box for ~75 min, rinsed twice in 0.08% photoflow (Kodak, Paris, France) and air dried. Slides were stored at −80 °C until use.

Immunofluorescence (IF)

Surface spread preparations were washed twice in PBS containing 0.05% triton-X-100 and blocked for 1 h at 37 °C (blocking buffer: 1% bovine serum albumin, 10% normal donkey serum in PBS containing 0.05% triton-X-100). Primary antibodies were diluted in blocking buffer, and slides were incubated for 20 min at 37 °C, followed by overnight incubation at 4 °C (when using the MLH1 antibody we incubated for two nights at 4 °C) extended by 20 min at 37 °C. Then slides were rinsed and washed once in PBS containing 0.05% triton-X-100 and afterwards rinsed and washed once in PBS. A second 30 min blocking step was applied in blocking buffer without triton-X-100, followed by a 2-h incubation with the secondary antibodies diluted in blocking buffer without triton-X-100. After rinsing and washing once in PBS, nuclei were stained with DAPI (0.3 μg/mL) and mounted with Vectashield (Vector, Burlingame, CA, USA).


To detect SYCP3 a rabbit polyclonal antibody (ab15092; Abcam, Cambridge, UK) was used at 1 : 200 and a goat polyclonal antibody (AF3750; R&D systems, Minneapolis, MN, USA) was used at 1 : 100. To detect γH2AX a rabbit polyclonal antibody (07-164; Upstate) was used at 1 : 250 and a mouse monoclonal antibody (05-636; Upstate, Lake Placid, NY, USA) was used at 1 : 1000. A rabbit polyclonal antibody against RPA provided by D. Schaarschmidt was used at 1 : 1000 (Treuner et al., 1999). A mouse monoclonal antibody against MLH1 (BD Pharmingen, San Jose, CA, USA) was used at 1 : 100. A rabbit polyclonal antibody against RAD51 was used at 1 : 200 and was provided by R. Kanaar. To detect telomeres a monoclonal antibody (TRF2, clone 4A794) was used at 1 : 100 dilution. Primary antibodies were detected using donkey anti-mouse and donkey anti-rabbit secondaries with a green or red fluorochrome at 1 : 500 dilution (Invitrogen Alexa 488; A21202, Alexa 594; A21207, Paisley, UK). The donkey anti-goat secondary with a blue fluorochrome (Invitrogen Alexa 350; A21081) was used at a 1 : 100 dilution.

Image capture and analysis

Nuclei were captured by a Zeiss AxioCam MR camera on a Zeiss Axioplan fluorescence microscope using Axiovision 3.1 software (Carl Zeiss, Oberkochen, Germany) at an exposure time reflecting the microscopic image. For the SYCP3-RPA-γH2AX/SYCP3-MLH1-γH2AX series, 25 early/mid and 25 late pachytene nuclei were captured per patient/control man per immunostaining. Nuclei of these series were analysed by two observers who were blinded to proband details. Data were imported in SPSS to generate an orderly depiction and for statistical analysis. When importing, data of the two observers were randomized if differences occurred, which was infrequent. Pachytene substages were determined, based on SYCP3 staining and electron microscopy surface spread images as described before by de Boer et al. (Chandley et al., 1984; de Boer et al., 2004). The stages I, II and III were collectively labelled early and the stages IV and V as late pachytene spermatocytes. The RPA patterns were determined based on the occupancy of SCs with foci (Fig. S2). Nuclei were classified as having a ‘full’ RPA pattern when all SCs were fully or almost fully occupied by RPA foci as closed chains. When only short stretches of adjacent RPA foci were found on the SCs, nuclei were classified as ‘intermediate’. In nuclei classified as ‘isolated’, single RPA foci in varying numbers were found, among which there often were more intensely stained and enlarged ones. An ‘empty’ pattern code was given to nuclei without any clear RPA foci. To investigate alterations in speed of RPA decrease between XY and autosomal SCs, subjective ratios were determined: ‘More’ indicates nuclei that show relatively more, and ‘less’ indicates fewer RPA foci on XY axial elements than on autosomal SCs. If no difference between autosomal SCs and XY axial elements was found, nuclei were classified as ‘equal’ (see Fig. S2 for examples). Based on MLH1 foci intensity and coverage of SCs, nuclei were classified into five groups (Fig. 3a–e). Groups 1–3 contain nuclei that have foci on all to nearly all SCs, but in a declining intensity. To check the objectivity of the intensity scoring system we used Image J to determine the average foci intensity in a subset of nuclei of the MLH1 1–3 groups selected from all probands (Fig. S4). Foci displaying an intensity above a set threshold were measured. A detailed description of the applied image J procedure is provided in the supplementary methods. Group 4 contains nuclei that show foci on less than half the SCs that often are weak. In group 5 nuclei, no MLH1 foci could be discerned with certainty. As to the evaluation of signals for γH2AX, we have adopted the system introduced by Chicheportiche and co-workers, using the monoclonal anti-γH2AX antibody (Chicheportiche et al., 2007). Numbers of clearly discernable L(arge)-foci (protruding more than 2 times the size of an RPA focus from the SC into loop domain chromatin) were counted. S(mall)-foci go hardly beyond the width of the SC (Chicheportiche et al., 2007) (Fig. S1d) and have not been investigated in detail. To determine the extent of colocalization between RPA and MLH1, two SYCP3-RPA-MLH1 series of 50 pachytene nuclei each were captured from probands C1 and P3 (see Table 1).


γH2AX in human male meiosis

To first obtain a more detailed insight into the γH2AX staining patterns of human male meiotic prophase stages, we applied IF of γH2AX using the monoclonal antibody combined with SYCP3 to follow SC formation. Sixty-six leptotene nuclei and 82 zygotene nuclei from 6 men (P6-11, Table S1) were evaluated. In contrast with previously published results on γH2AX in human meiosis (Roig et al., 2004) (using a polyclonal antibody and observing strong overall γH2AX staining), we observed at the leptotene stage a very faint dotted γH2AX signal (Fig. 1a) with stronger spots/little domains coming up in later leptotene nuclei colocalizing with developing axial elements (Fig. 1b) at the telomeric regions (tested by IF with SYCP3 and TRF2, Fig. S1a) (Brown et al., 2005). At the early zygotene stage, increasingly bright focal γH2AX domains were observed, that in many cases originated from developing axial elements, approaching each other at the telomeric ends of homologues (Fig. 1c, Fig. S1b). At mid zygotene, signals of variable sizes and originating from SC components mostly remained focal (Fig. 1d,e). At late zygotene, signals regularly were at synaptic forks, but were not systematically present in a high intensity in asynapsed regions (Fig. 1e). In a small series of nuclei (n = 27), we checked for γH2AX signals adjacent to DNA DSB repair patches (early recombination nodules) marked by RAD51. These usually were found (Fig. S1c, d). Next to the lack of a systematic γH2AX response at asynaptic axial elements (Fig. 1f) we observed at pachytene (besides XY body staining) additional SC born γH2AX signals (S-foci and L-foci) as described before in the mouse (Chicheportiche et al., 2007) (see Materials and methods section image capture and analysis, Fig. 1g). These foci were not rare in this material and could indicate a disturbance in the progression of recombination nodule development. Therefore we first studied the kinetics of these nodules by probing with RPA and MLH1.

Figure 1.

γH2AX staining patterns during human meiotic prophase I. (a–g) In the merge panels the nucleus is indicated in blue (DAPI staining). Scale bar: 10 μm. (a) Early leptotene nucleus showing small stretches of axial elements. Almost no γH2AX signal was detected throughout the nucleus. (b) Mid–late leptotene nucleus showing a subset of developing axial elements colocalizing with intense γH2AX patches of varying but small size. (c) Example of an early zygotene nucleus with axial elements starting to synapse (arrow). γH2AX domains of increasing size were found to originate from axial elements mainly at telomeric regions. (d,e) Mid–late zygotene nuclei with few unsynapsed axial elements and focal γH2AX domains, some protruding along chromatin loops away from the SC (arrowhead). Signals were also found at synaptic forks (arrow). (f) Early pachytene nucleus with extensive γH2AX staining at the XY body (*) and several asynaptic sites (arrowhead) which are not systematically marked by γH2AX. (g) Example of a late pachytene nucleus with strong staining at the XY body (*) and two L-foci (arrows).

RPA in early-to-late pachytene spermatocytes

To investigate the progression of recombination nodules, we used RPA as a processing marker for early recombination events. We studied the RPA staining pattern (see Materials and methods section image capture and analysis and Fig. S2) during the pachytene substages to determine variation between probands. As described by Oliver-Bonet and co-workers, RPA peaks at late zygotene and decreases during pachytene (Oliver-Bonet et al., 2005). We observed this pattern in all probands. Figure 2a shows that in early pachytene spermatocytes most nuclei contain SCs that are fully packed with RPA nodules. In the late pachytene spermatocytes most nuclei contain SCs that only show a small number of RPA foci (Fig. 2b). Within early and late pachytene, males were heterogeneous in the speed of RPA nodule loss (Fig. 2a,b).

Figure 2.

RPA staining pattern on XY and autosomes in early and late pachytene spermatocytes. (a) The RPA patterns in early pachytene spermatocytes for the seven probands. For chi-squared analysis, intermediate, isolated and empty categories were pooled: d.f. 6, p < 0.01. (b) The RPA patterns in late pachytene spermatocytes for the seven probands. For chi-squared analysis, full, intermediate and isolated categories were pooled: d.f. 6, p < 0.05. (c) The subjective ratio between RPA foci on XY and on autosomal SCs at early pachytene was determined for the RPA patterns. Ratios were determined as more, equal or less RPA foci on XY compared with autosomes. Significance was determined by chi-squared analysis: d.f. 4, p < 0.001. (d) The ratio between RPA foci on XY and on autosomal SCs in late pachytenes was determined for the RPA patterns. Ratios were determined as more, equal or less RPA foci on XY compared with autosomes. Significance was determined by chi-squared analysis: d.f. 4, p < 0.001.

Next, we investigated the decrease in RPA foci on XY axial elements relative to autosomal SCs at early and late pachytene (see Fig. S2 for examples). As an initial analysis could not identify differences between probands at late pachytene, data of the 7 probands were pooled to provide a better overview. The percentages of nuclei showing an accumulation of RPA nodules on XY (‘more’) are remarkable in the ‘intermediate’ and ‘isolated’ RPA categories of both early and late pachytene (Fig. 2c,d), indicating a slower repair of DNA DSBs of XY compared with the autosomes. Nuclei showing fewer RPA foci at XY compared with autosomes were mainly detected in early pachytene displaying a ‘full’ RPA pattern (Fig. 2c). Concomitant with a further decrease of autosomal RPA foci in late pachytene nuclei, also foci on XY decreased, although with a large variation between nuclei (Fig. S3).

A subset of early pachytene nuclei from all probands with synapsed sex chromosomes was selected to study X (n = 51) and Y (n = 67) chromosome RPA occupancy in more detail. PAR1 was in most cases recognized as the only synapsed region, and showed clear RPA foci in contrast with PAR2 that when synapsed showed RPA foci, albeit less intense. In 20% of nuclei synapsis was found for both PARs. For the Y axial element an RPA signal was found at PAR1 in 66 nuclei. In 30% of nuclei the PAR1 synapsed stretch seemed to be extended between the X and Y axial elements, which might indicate at non-homologous synapsis. On the non-PAR Y axial element, RPA could be clearly visible as a single dot to a few or a row for up to about 70% of its length coming from PAR1 (57% of nuclei). Two nuclei showed a signal in subtelomeric Yq in the absence of synapsis at PAR2. For the X axial element, RPA foci varied from sparse to abundant and occasionally a faint signal colocalized with a segment. Foci were from small (hardly exceeding the width of the axial element) to very conspicuous, but never seemed enlarged as was found on the autosomal SCs (as will be described in the last section). Of the 51 early pachytene nuclei investigated, in 25.5% of cases, the whole X axial element was occupied, in 45% of cases the focal distribution resembled the ‘intermediate’ category for autosomal bivalents and in 29.5%, 6 or fewer distinct foci were counted. Figure S2 (a, b, and e) shows early pachytene XY axial elements occupied by RPA with a strong signal at PAR1 (Fig. S2a,e).

Transition of RPA to MLH1

During progression of pachytene, part of the RPA containing early/transitional recombination nodules metamorphose into late MLH1 containing recombination nodules, which can still show RPA (Oliver-Bonet et al., 2005). To confirm this observation with an emphasis on late pachytene we probed two probands for RPA and MLH1 (C1 and P3, Table 1). More overlap between RPA and MLH1 was observed in early pachytene (not quantified) compared with late pachytene. In late pachytene nuclei with a ‘full’ MLH1 pattern (Materials and methods section image capture and analysis) and ‘isolated’ RPA pattern, only few colocalizing foci could be detected (C1 mean 1.7, n = 10, P3 mean 0.8, n = 10). This indicates that most remnant RPA foci of normal size do not represent late recombination nodules as defined by crossing-over, i.e. the presence of MLH1 (see also Fig. 4f).

MLH1 in early-to-late pachytene spermatocytes

Figure 3a–e illustrates the classification of MLH1 patterns and intensities into five groups (see Materials and methods section image capture and analysis). We first determined the frequencies of nuclei containing foci on all to nearly all SCs (groups 1, 2, 3) and nuclei containing partly occupied SCs (group 4) or no foci at all (5) (Fig. 3f,g). No variation was detected among probands in early pachytene (Fig. 3f). However, at late pachytene differences were observed (Fig. 3g). Next, we subjectively determined the average intensity of group 1–3 MLH1 foci among probands, which is depicted in Fig. 3h. To check our subjective classification we determined the average MLH1 foci intensity of group 1–3 nuclei using image J as well (Fig. S4). Significant differences in intensity levels were found between the groups indicating the validity of our subjective scoring method. Both in early and late pachytene, variation among probands is present. Usually there was an increase in intensity from early-to-late pachytene, which also varied among probands. Numbers of MLH1 foci were checked to determine if our probands belonged to the group of NOA males with lower to extremely low numbers of MLH1 foci (Gonsalves et al., 2004; Sun et al., 2007b). First observations on the presence of MLH1 foci on SCs indicated that the nuclei well represent the normal population. Ten nuclei per proband were counted as, knowing the averages and standard deviations of MLH1 counts from the literature (Gonsalves et al., 2004; Hassold et al., 2004; Codina-Pascual et al., 2005; Sun et al., 2005), this enables a sufficient approximation of the mean (Fig 3i). Significant differences were found among them (Fig. 3i), with P1 and P2 showing low numbers relative to the normal range. The intensity of MLH1 foci in late pachytene was not related to the average number of foci as indicated by the Spearman rank correlation test (rs: 0.46, p > 0.05).

Figure 3.

MLH1 foci dynamics, intensity and numbers at late pachytene spermatocytes. (a–e) Examples of nuclei showing the three MLH1 staining intensity categories from high (1) to low (3) and few (4) or no foci present (5). Scale bar: 10 μm. (f) Percentage of nuclei belonging to staining category 1–3, 4 or 5 is shown for early pachytene spermatocytes of the seven probands. For chi-squared analysis, categories 4 and 5 were pooled: d.f. 6, not significant. (g) Percentage of nuclei belonging to staining category 1–3, 4 or 5 is shown for late pachytene spermatocytes of the seven probands. For chi-squared analysis, categories 4 and 5 were pooled: d.f. 6, p < 0.01. (h) Intensity of category 1–3 MLH1 foci was numerically expressed by giving nuclei of category 1 three points, nuclei of category 2 two points and nuclei of category 3 one point. The average intensity was determined for early and late pachytene nuclei. Chi-squared analysis was performed on the distribution between probands of nuclei over the staining categories (1–3). Early pachytene: d.f. 12 p < 0.01 and late pachytene: d.f. 12 p < 0.001. (i) For each proband, the number of MLH1 foci was counted in 10 nuclei of staining categories 1–3. Symbols indicate the mean number of MLH1 foci and error bars the range. Variation between probands was analysed by one-way anova, F 663 = 6.55, p < 0.001. Tukey analysis was performed to specifically identify differences between two probands as indicated by horizontal lines; *indicates a difference with p < 0.05, **with p < 0.01 and ***with p < 0.001.

γH2AX L-foci in combination with RPA or MLH1

To study the relationship between recombination nodules and γH2AX signalling, we investigated γH2AX L-foci (Fig. 4). Frequency distributions were studied for outliers influencing average numbers of associations (γH2AX+RPA/MLH1), which were not observed (Fig. S5). For each γH2AX L-signal we first determined association with RPA (Fig. 4d). Enlarged RPA foci (Fig. 4b) were specifically recorded. The contribution of L-foci associated with RPA (normal and enlarged) to the total number of L-foci was determined for each proband (Fig. 4d). Both in early and late pachytene, L-foci associated with normal-sized RPA make up the largest part, a pattern observed for each proband. The distribution of L-foci differed among probands (Fig. 4d). In all but one (P5), more L-foci were found in late pachytene compared with early pachytene (Fig. 4d), but variation was detected in the degree of this increase. Highest numbers of L-foci were found in P3.

Figure 4.

L-foci colocalization with RPA and MLH1 foci in late pachytene spermatocytes. (a–c) Scale bar: 10 μm. (a) Example of a pachytene nucleus showing L-foci associated with normal RPA foci (arrows), the dotted squares zoom in at an L-focus (upper and lower panel). Arrowheads show L-foci without RPA. The XY body is indicated by *. (b) Example of a pachytene nucleus showing an L-focus associated with an enlarged RPA focus. The dotted square zooms in at an L-focus associated with an enlarged RPA focus. The XY body is indicated by *. (c) Example of a pachytene nucleus showing several L-foci associated with MLH1 (arrows), the dotted square zooms in at an L-focus. Arrowheads show L-foci without MLH1. The XY body is indicated by *. (d) The average number of L-foci for all probands in early and late pachytene determined by the monoclonal γH2AX antibody. Variation between probands was analysed by the Kruskall–Wallis test: early pachytene L-foci with RPA, chi-squared 26.9, d.f. 6, p < 0.001; early pachytene L-foci with enlarged RPA, chi-squared 10.4, d.f. 6, p > 0.05; early pachytene L-foci without RPA, chi-squared 16.9, d.f. 6, p < 0.05; late pachytene L-foci with RPA, chi-squared 16.4, d.f. 6, p < 0.05; late pachytene L-foci with enlarged RPA, chi-squared 14.1, d.f. 6, p < 0.05; late pachytene L-foci without RPA, chi-squared 9.9, d.f. 6, p > 0.05. (e) Average numbers of residual RPA foci on the autosomes in late pachytenes with an ‘isolated’ RPA pattern were determined for the seven probands. The ratio of the two types of RPA events (with and without γH2AX) over probands was tested for homogeneity by chi-squared analysis: d.f. 6, p < 0.001. Variation between probands on total numbers of RPA foci was analysed by Kruskall–Wallis test: chi-squared 13.96, d.f. 6, p < 0.05. (f) The average number of L-foci with and without association with MLH1 was determined for all probands in early and late pachytene using the polyclonal γH2AX antibody. Variation between probands was analysed by the Kruskall–Wallis test: early pachytene L-foci with MLH1, chi-squared 18.8, d.f. 6, p < 0.01; early pachytene L-foci without MLH1, chi-squared 46.7, d.f. 6, p < 0.001; late pachytene L-foci with MLH1, chi-squared 14.2, d.f. 6, p < 0.05; late pachytene L-foci without MLH1, chi-squared 42.2, d.f. 6, p < 0.001.

In all, our data indicate that recombination nodules with RPA which stay longer are at a higher risk for γH2AX labelling. Therefore, we determined the number of remaining RPA foci in the late pachytene nuclei displaying an ‘isolated’ RPA pattern, as most late pachytene nuclei belong to this group (Fig. 4e). Probands showed variation in the ratio of RPA foci with γH2AX/without γH2AX. Most males had a higher number of RPA foci without a γH2AX mark, however two (P1 and P3) showed the opposite pattern. A high correlation was found between numbers of L-foci in late pachytene (Fig. 4d) and the ratio between RPA+γH2AX/total RPA (Fig. 4e) (Spearman rank correlation test rs: 0.96, p < 0.01).

To estimate the colocalization frequency between MLH1 and γH2AX, a polyclonal antibody for γH2AX was used. Compared with the RPA labelling experiments involving the monoclonal γH2AX antibody, total numbers of L-foci for early and late pachytene differed. Generally, levels were lower and also the ranking of probands was not the same. For all probands at both early and late pachytene, the fraction of L-foci associated with MLH1 was smaller than the fraction without MLH1, which is the opposite for RPA (compare Fig. 4d with f). A higher relative labelling of RPA by γH2AX (Fig. 4e) is related to less bright MLH1 CO foci in late pachytene (Fig. 3h) (rs: 0.714, p = 0.05).


In this study we have analysed the kinetics of transitional and late recombination nodules during first meiotic prophase in testicular biopsies from seven men, of which five were diagnosed with NOA and two had a history of normal fertility before a vasectomy. Transitional nodules were marked by RPA and late nodules by MLH1. We have identified γH2AX signals emanating from the SC and studied colocalization with the recombination nodules. Both at early and late pachytene, men were heterogeneous as to the loss of RPA marked transitional recombination nodules. The appearance of late recombination nodules marked by MLH1 was uniform at early pachytene. Men differed in both the absolute levels of γH2AX foci and in the fraction of RPA foci marked by γH2AX. There was a high correlation between these two parameters at late pachytene. Men also differed in the intensity of the MLH1 foci at CO sites. Our data suggest that a higher fraction of RPA marked by γH2AX is observed when MLH1 foci are less bright. The remnant spermatogenic cells and cellular associations that are available after sperm retrieval (TESE) for ICSI offer abundant material for this type of analysis. In our sample of seven men, four fathered a child via ICSI (Table 1). Our results warrant further study into male meiotic stability in human artificial reproduction.

We choose to study these phenomena in detail over seven probands to obtain a first insight into the homogeneity/heterogeneity of the recombination process in the human male. The number of nuclei studied per proband was assessed using the limited mouse data on the occurrence and induction of this phenomenon (Chicheportiche et al., 2007; Ahmed et al., 2010). In our analysis of L-foci associated with RPA and MLH1, statistical analysis determined significant differences in distribution of associated L-foci among probands (except for the number of L-foci associated with enlarged RPA in early pachytene) indicating a sufficient sample size for a first investigation of this phenomenon in the human.

We found the two Upstate γH2AX antibodies not to give congruent results on SC involved signalling. The number of signals detected with the polyclonal antibody was generally lower and signals less bright. Using the polyclonal one, male meiotic recombination involved γH2AX signalling was initially overlooked in the mouse (Mahadevaiah et al., 2001) and became only apparent with use of the monoclonal antibody (Chicheportiche et al., 2007). When comparing average numbers of L-foci in the mouse (Chicheportiche et al., 2007) with our results in human, both generated by the monoclonal antibody, we can conclude that on average γH2AX signalling in mice is slightly lower.

Both in early and late pachytene, we found γH2AX foci to largely associate with RPA. We found a decrease in colocalization between RPA and MLH1 from early-to-late pachytene and an increase in γH2AX L-signals associated with RPA foci. These γH2AX marked RPA foci seem to involve recombination intermediates that are resolved towards a NCO (gene conversion) event, but are delayed in processing. Based on a molecular analysis at the DNA level of recombination products, Guillon et al. (Guillon et al., 2005) noted that recombination, whether in the NCO or the CO direction, proceeds over mid to the end of late pachytene in the mouse.

As mentioned in the introduction, evidence for the presence of γH2AX L-foci at recombination intermediates has been obtained in several mouse genetic models (a.o. Rad54, Blm and Btbd12 knock-outs). These three models address aspects of resolving recombination involved joined DNA molecules (dHJs). RAD54 is involved in branch migration of a dHJ. Branch migration might be involved in determining the length of the recombined or gene conversion region. Furthermore, it might be involved in determining the choice between a CO or a NCO (Mazin et al., 2010). BLM, which colocalizes with RPA in transitional recombination nodules (Walpita et al., 1999; Moens et al., 2002), is generally known to downregulate crossing-over (Wang et al., 2011). BTBD12 is a target of the ATM kinase (for which H2AX is a substrate) and possibly has a function in regulating recombination intermediate repair in the direction of a second CO pathway (discussed below) (Holloway et al., 2011). The role of these molecules in preventing the risks of problematic homologous recombination intermediates is supported by interactions (in the mouse) within the MSH4/5, MLH1/3 canonical crossing-over route: Blocking this pathway by knock-out of MLH3 leads to increased IF expression of BLM (Holloway et al., 2010). When BTBD12 function is compromised, the number of MLH1 foci is increased (Holloway et al., 2011).

Apart from the canonical MSH4/5, MLH1/3 route of crossing-over, a minor second CO pathway in mouse makes use of the highly conserved MUS81 endonuclease pathway (Holloway et al., 2008). The fact that this pathway is free of crossing-over interference and when blocked, leads to upregulation of the MSH4/5, MLH1/3 CO pathway (Holloway et al., 2008), points towards a role for MUS81 in the regulation of CO vs. NCO also. MUS81 is a conserved protein and is expressed in the human testis ( Human chiasma counts (Laurie & Hulten, 1985) are very similar to MLH1 counts (Hassold et al., 2004; Martin, 2008). However, both show high variation within and between males, making it difficult to determine whether there would be a difference between chiasma and MLH1 foci counts on a per cell basis, offering a possible contribution of MUS81 to the CO pathway in the human.

Probands were not heterogeneous in acquiring MLH1 at early pachytene. We found, however, MLH1 foci to be variable between patients as to intensity of the signals observed in early and in late pachytene. Also two patients (P1, P2) had mean MLH1 counts that were at the lower range of average values, based on men with normal or near-normal spermatogenesis [46.2–54.5 (Hassold et al., 2004), 42.5–55.0 (Sun et al., 2005), 42.6–50.4 (Gonsalves et al., 2004)] consistent with reports on lower counts in NOA [reviewed in (Martin, 2008)]. The standard deviations of MLH1 foci numbers, determined in 10 nuclei were comparable with those reported in the literature for normal men (Gonsalves et al., 2004; Hassold et al., 2004).

When studying expression of MSH4 and MSH5 per spermatocyte by RT PCR, Terribas and co-workers (Terribas et al., 2010) noted that RNA levels of both MutS homologues were reduced in males with compromised spermatogenesis. Hence, variation in expression of MLH1 might well be an explanation of the observed variation in foci intensity. Lower MLH1 foci intensities at late pachytene were found to correlate with a higher occupation rate of RPA foci with γH2AX. This indicates that substantial γH2AX signalling in the RPA marked NCO pathway influences the CO pathway (or vice versa) leading to a lowered MLH1 capacity. Alternatively, suboptimal DNA repair could influence both the CO and NCO pathway.

From the biopsy samples used here, P3 stands out for showing a coherent pattern: a delay in loss of RPA foci, a delay in the acquisition of MLH1 foci, an initially very low intensity of these foci, a high number of late pachytene γH2AX-positive RPA foci and a very high occupancy of late RPA foci with γH2AX. Also, with the polyclonal antibody, this man had high levels of colocalization between MLH1 and γH2AX. Interestingly, at the third ICSI attempt a pregnancy was established (Table 1). Furthermore, three men (C1, C2 & P4) had near-normal rates of meiotic prophase stages per spermatozoa in the wet preparations and 4 (P1-3,5) clearly had a paucity of spermatozoa (Table S1). The probands with normal spermatogenic efficiency ratios (C1, C2 & P4) tended to have more intensive MLH1 foci, consistent with results described by Terribas et al. (2010).

It is not yet possible to link γH2AX signalling of mainly NCO directed recombination intermediates with known mutation patterns in the human population. The suggestion here is that this type of signalling can be more often found in men with a high pachytene/sperm ratio (Table S1). A link between Copy Number Variation (mainly copy loss) after ICSI and delayed synapsis was recently suggested (Woldringh et al., 2009). In a random search for human genomic structural variation by deep sequencing, gene conversion (resulting from mainly a NCO) stands out as one of the major causative mechanisms next to non-allelic homologous recombination (NAHR) and Line 1 (L1) transposition (Kidd et al., 2010). Non-allelic gene conversion is known to play a role in human disease (Chen et al., 2007).

In summary, we have demonstrated heterogeneity between probands in processing of presumably delayed NCO recombination intermediates as these were variably marked by the DNA damage sensing pathway marker γH2AX. Highest numbers of damage marked recombination intermediates were found among NOA patients with a low yield of spermatozoa per spermatocyte. The relevant clinical question is whether the progeny of these patients is more at risk for mutations introduced at meiosis or would cell selection during spermiogenesis prevent transmission?


We would like to thank the following colleagues for providing antibodies: W.M. Baarends, Department of Reproduction and Development, Erasmus MC, Rotterdam, The Netherlands; D. Schaarschmidt, (formerly) Department of Biology, University of Konstanz, Konstanz, Germany; R. Kanaar, Department of Cell Biology and Genetics, Erasmus MC, Rotterdam, The Netherlands and H. Scherthan, Institute for Radiobiology, München, Germany. We would like to thank K. D'Hauwers, Department of Urology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands for taking the testicular biopsies and Cindy Dieteren from the department of Cell Biology, Nijmegen Centre for Molecular Life Sciences, Nijmegen, The Netherlands for her kind help with the image J analysis.