For the first time, we analysed 89 strictly selected patients with severe oligozoospermia, 37 with azoospermia due to SCOS and 100 with normal spermatogenesis as controls by array-CGH. Although interpretation of the results would have been more straightforward if only one type of microarray had been used and while knowing that amount of DNA available as well as funding would not permit us to repeat the analyses on the first set of samples, we favoured switching to the higher resolution array to detect smaller CNVs in almost half of our subjects, which in the eand appeared to be beneficial. As result, we report several genes and genomic regions on autosomes and more pronouncedly on the sex-chromosomes that might either be risk factors (also found in controls) or causative by themselves (not found in controls) for spermatogenic failure.
We hypothesised that an increased number or specific distribution of CNVs could result in defective recombination, meiotic and thereby spermatogenic failure. Structural chromosomal aberrations are found more frequently in men with oligo- and azoospermia with an emphasis on autosomal translocations in the former (3–4% compared to 0.5–1.5% in controls) and sex-chromosomal aneuploidy in the latter (13–16% compared to 0.5–1%) 
. The causal relation between chromosomal rearrangements and impaired sperm production has been suggested to be a structural effect related to alterations in the process of chromosome synapsis during meiosis 
, but whether submicroscopic chromosomal rearrangements (CNVs) can result in meiotic recombination defects is not known. By comparing the number of all CNVs, duplications and deletions separately and amount of DNA change, gain and loss no significant differences were found between the groups analysed. The differences in chromosomal distribution of CNVs were attributed to single, recurring CNVs (see below). Also particularly large (>1 Mb) variants were not found more frequently in oligozoospermia or SCOS. Therefore, only even larger variants of several megabases, microscopically detectable upon conventional karyotyping, might impair chromosome synapsis and meiosis. Whether a size threshold for chromosomal aberrations having such a “direct”, gene-independent impact on spermatogenesis exists cannot be concluded from the presented data. It seems, however, that structural variation below the detection limit of routine karyotyping should not be regarded, per se
, as an obligate cause of spermatogenic failure. Considering that translocation carriers may have normal spermatogenesis 
, the link between (large) structural chromosomal variation and spermatogenesis remains to be elucidated.
In principle, CNVs may result in altered gene transcription/protein function through different mechanisms: they might encompass dosage-sensitive genes, a deletion may demask a recessive mutation on the homologous chromosome, genes overlapped by structural variation may be disrupted directly or a CNV can exert position effects 
. By comparing CNVs between the controls and the two patient groups, 11 and 4 CNVs specific for severe oligozoospermia and SCOS, respectively, were identified in more than one patient each (). In addition, 12 and 14 private, sex-chromosomal CNVs were listed () because these may have a functional consequence of the naturally haploid genes contained therein. Also private, autosomal CNVs currently found in one patient each may contribute to the infertility phenotype, but their significance may only be evaluated by much larger studies. According to our moderately conservative approach, six CNVs were found in highly significantly different frequencies in patients with SCOS but also in controls and may be new risk factors associated with male infertility. All of these CNVs remained significant also after Bonferroni-correction. Interestingly, only one CNV in 12p13.31 without known genes was associated with severe oligozoospermia. The 81.9 Kb deletion was found with increasing frequency from controls (4%) to oligozoospermic men (12%) to men with SCOS (35%) as the most severe phenotype. However, the common deletion polymorphism in UGT2B17
was found in decreasing frequency while a duplication of the same locus was found in increasing frequency. These glucoronidase variants determine urinary excretion of testosterone 
, but have not been studied in male infertility.
Genes may be prioritised according to a known function or indirectly through their expression in the testis. Therefore, the 4 autosomal and 8 X-chromosomal genes presented in are the distilled outcome of the current study with respect to identification of new, promising candidate genes. Most of these were shown to be expressed in Sertoli or germ cells, but their functions are mostly unknown. EPHA3
were mentioned because the Anks proteins were found to modulate degradation of EphA receptors in mice 
and therefore the three genes might represent members of a functional unity. CNVs altering one of these genes were found in 6 men (7%) with severe oligozoospermia and 5 (14%) with SCOS.
In two recent studies, our group analysed patients with POF and XY gonadal dysgenesis by array-CGH 
. Interestingly, the putative causal genes identified were enriched for genes also implied in spermatogenesis. When comparing results of these and the current study, CNVs in the PLEC
loci were found not only in men with SCOS, but also in either patients with POF or XY gonadal dysgenesis (personal communication). These cover 5 of 11 genes identified in the current study which might hint to a common genetic origin of loss of spermatogonia in the male and loss of oogonia in the female resulting in SCOS, XY gonadal dysgenesis and POF, respectively.
With sperm counts ranging from zero to hundreds of millions per ejaculate, sperm output may be viewed as a quantitative trait and male infertility is sometimes postulated as a polygenic disease 
. Some evidence for this hypothesis may be gained from the significant negative correlation between sperm counts and number of deletions described in our 78 normozoospermic men analysed by 244A array. Especially as hundreds of genes are implicated in spermatogenesis, it is plausible that more deletions also more often involve spermatogenesis-relevant genes. Therefore, a man bearing more deletions may have a less efficient spermatogenesis and therefore lower sperm output. We could, however, not detect a comparable correlation in the, albeit at most half as large, groups of normozoospermic men analysed by 400K arrays or in oligozoospermic patients (irrespective of array used).
While on the one hand the selection of our control group from the patient clientele avoids population stratification, on the other hand the conclusions drawn primarily remain limited to the phenotype of spermatogenic failure and cannot be readily extended to fertility. For this purpose, a group of proven fertile men (fathers) would be needed as additional controls. However, at least one-fifth of our normozoospermic controls had fathered a child before then presenting with either secondary infertility or infertility in a new relationship. Contrariwise, the usually utilised Database of Genomic Variants (DGV) of ‘healthy’ controls is not amenable to be used with respect to the phenotype of spermatogenic impairment, as the fertility status (let alone spermatogenesis) is unknown. Thus, as a larger group of proven fertile men was neither available to us nor has - to our knowledge - been analysed anywhere else yet, our control group may well be used to study spermatogenesis. The CNV data of the 100 controls is provided as supplement as well as accessible through dbVar and will be valuable for other studies on genetics of male infertility.
In contrast to many candidate-gene approaches, only one recently published study analysed 80 men with normozoospermia, 52 with OAT, and 40 with non-obstructive azoospermia (histology not mentioned) using whole genome SNP arrays 
. Twenty-one SNPs were found significantly associated with azoo- or oligozoospermia, but only four could be replicated in a larger follow-up study 
. We applied a more stringent patient selection especially for the group of azoospermia as the known histology of SCOS leads to one of the most homogeneous phenotypes possible in male infertility. Accordingly, we could identify a larger number of putative causal CNVs/genes in SCOS than in oligozoospermia. All the recurring, patient-specific and private, sex-chromosomal CNVs as well as the CNVs associated with SCOS described herein are candidates deserving further detailed analyses in larger patient and control groups as well as other populations. Especially with respect to the sex-chromosomal CNVs, analyses of trios would be helpful to determine their relevance. It should be considered, however, that involving the parents of infertile patients is very difficult in comparison to other diseases.
In conclusion, by the first CNV study in male infertility, we provide evidence that CNVs contribute to the complex origin of male infertility and present a number of candidate genes possibly causing or being risk factors for spermatogenic failure.