Why should andrologists take an interest in evolution? A simple answer is that evolution can be seen as fertility averaged over a long period of time, and thus may highlight subtle aspects of fertility that might not be obvious in a single generation. Fertility varies between individuals because of both biological and social factors. Among the biological factors are genetic ones; indeed mutations in some 10% of mouse genes can lead to infertility (Kile et al., 2003) and the proportion is likely to be similar in humans, although the identity of most of these genes is unknown. An evolutionary perspective provides expectations about them. If a gene is functionally important, we expect it to be evolutionarily conserved and thus present in related species. Conversely, if a candidate infertility gene is not conserved, we might question its importance for fertility, and will discuss an example of such a gene below. Over the shorter period of evolution within human populations, we might expect to see that variants which increase fertility increase in frequency in the population, whereas variants that reduce fertility decrease in frequency. These simple expectations are, however, affected by the complete linkage over most of the Y chromosome: selection acts on the whole haplotype. Thus, a disadvantageous mutation on an otherwise beneficial haplotype background can still increase in frequency and even reach fixation, whereas an advantageous mutation on a disadvantageous background might still be selected against.

Variants may directly influence fertility, but they can also be used indirectly by marking lineages: there is complete linkage over most of the length of the Y chromosome. Variants can conveniently be considered within three categories: (i) SNPs (single nucleotide polymorphisms or base substitutions; sometimes used to include binary markers of many types including small insertions and deletions), (ii) STRs (short tandem repeats or microsatellites; multiallelic) and (iii) structural variants, of which CNVs (copy number variants; binary or multiallelic) have received most attention. Genome databases such as Ensembl (http://www.ensembl.org/Homo_sapiens/index.html) now list very large numbers of Y-SNPs: 72 220 in January 2008. Many of these, however, may represent differences between a Y sequence and a related sequence elsewhere in the genome (paralogs), and less than 1% are SNPs that have been placed on a phylogenetic tree (Jobling & Tyler-Smith, 2003). Databases of Y-SNP phylogeny are available (e.g. http://www.snp-y.org/) and there are plans to construct databases summarizing geographical distributions. Y-STRs can be identified from the reference sequence and a comprehensive survey identified 475 potentially useful loci (Kayser et al., 2004); these are the Y markers of choice for forensic work and comprehensive databases containing over 52 000 haplotypes exist (Willuweit & Roewer, 2007), but they have been less used in infertility studies.

Most variants have no detectable effect on the phenotype or fitness, and so are considered neutral. The evolutionary fate of a new neutral mutation depends on chance (genetic drift) and is influenced by characteristics of the population such as its size, including whether it is expanding or contracting, and the amount of exchange (migration) with other populations: the fields of population and evolutionary genetics (Jobling et al., 2004). In contrast, beneficial variants will be positively selected and tend to increase in frequency, whereas harmful variants will be negatively selected and decrease in frequency. In the most extreme case, for example, a variant that leads to complete spermatogenic failure, the variant will not be transmitted and will be found in the population at the level determined by mutation.

If we want to know whether or not a genetic variant influences a phenotype of interest, e.g. sperm count, a standard approach is to measure the frequency of the variant in samples of individuals who differ in the phenotype. If we found, say, 40% A allele in the men with high sperm count and 70% in the men with low sperm count, we might want to conclude that the A allele marked a genetic background that led to low sperm count. But we should be very cautious before coming to this conclusion: the two samples might differ for other reasons, for example, if they come from different geographical regions. This is known as ‘population stratification’. It is the important characteristic that makes the Y chromosome so popular for evolutionary studies, noted above, but it also makes association studies involving the Y chromosome fraught with difficulty. The magnitude of this effect is illustrated by a paper published in 1999 which investigated the association between Y haplogroup and infertility in Italy (Previderéet al., 1999). In the raw data, haplogroup P was present at 42% in the controls but 24% in the infertile men, a statistically significant difference. But the infertile men were mostly sampled in Central Italy, whereas the controls were from several parts of Italy. When only Central Italians were considered, the frequencies were 27 and 26% respectively, a non-significant difference. This is a far greater degree of geographical differentiation than detected with ~10 000 autosomal SNPs (Bauchet et al., 2007), rendering ineffective one of the recommended methods of correcting for stratification, genomic control. How then can the careful investigator of Y-chromosomal associations produce reliable results? Precise geographical matching is essential and replication in an independent sample is also necessary.

While population stratification is generally a confounding influence for association studies, it may be possible to take advantage of it some circumstances. Although Y microdeletions are indisputably associated with spermatogenic failure, the roles of the individual genes lost remain unclear. The microdeletions all remove multiple genes, but it should still be possible to identify the key ones by searching for de novo point mutations in single genes. Yet such an approach has not been very successful, and a major reason for this is the repeated nature of the AZFb and AZFc deleted regions: the genes they contain are present in multiple copies and inactivation of a single copy may not lead to spermatogenic failure. The b2/b4 complete AZFc deletion removes nine genes including all members of the three families BPY2, DAZ and CDY1 (Fig. 2; Kuroda-Kawaguchi et al., 2001), almost always resulting in spermatogenic failure. In contrast, the g1/g3 (=b2/b3) partial deletion removes five genes leaving one BPY2, two DAZ and one CDY1 (Fernandes et al., 2004; Repping et al., 2004; Fig. 2), but is found in haplogroup N men with normal spermatogenesis who make up half the population of northern Europe (Zerjal et al., 1997, 2001). In this relatively simple genetic background, a survey of haplogroup N men with spermatogenic failure might reveal point mutations in the remaining single-copy genes.

Although the AZFa microdeletion has the simplest structure and is the best understood of the Y microdeletions, carrying just two genes, USP9Y and DDX3Y (formerly DBY), there has still been debate about which gene/s is/are responsible for spermatogenic failure. A complete deletion of USP9Y was associated with severe oligospermia (Brown et al., 1998) whereas a 4-bp deletion in a splice site leading to truncation of the protein was found in an azoospermic man (Sun et al., 1999), suggesting that this gene is required for spermatogenesis. In contrast, two different partial deletions were found in men with moderate oligoasthenoteratozoospermia from families where these Y chromosomes were successfully transmitted under natural conditions (Krausz et al., 2006). It is therefore interesting that the two chimpanzee Y chromosomes sequenced both carry inactive forms of USP9Y (Hughes et al., 2005; Kuroki et al., 2006; Tyler-Smith et al., 2006) and the four inactivating mutations in this gene are shared by bonobos (Perry et al., 2007) indicating fertility among apes for perhaps two million years in the absence of USP9Y. Such observations point to a role for DDX3Y in human spermatogenesis and suggest that further studies of this gene are warranted.