TSPY genes, as implied by their name, are located on the Y chromosome. TSPY was in fact one of the first human Y-chromosomal genes to be identified, in 1987 when Arnemann et al. published the results of a survey in which they used 18 DNA fragments enriched for Y-chromosomal sequences to search for transcripts, and found evidence for an abundant testis-specific mRNA [7]. The next year, a study of Y-chromosomal repeated sequences identified a tandemly-repeated array, DYZ5, on Yp, consisting of multiple homogeneous 20.3 kb units [5], and it soon became apparent that each unit carried a copy of TSPY. Several TSPY-related sequences were present elsewhere on the chromosome [8], the most substantial being designated TSPY minor [5]. Fuller details of the genomic structure, showing the presence of one intact TSPY gene at TSPY minor, and a more accurate unit size for the major array of 20.4 kb, were revealed by the finished sequence of the euchromatic part of the Y chromosome [9]. Nevertheless, only the edges of the array could be assembled, and a gap corresponding to most of the array remains even within the current human genome reference assembly (GRCh37/hg19; http://genome.ucsc.edu/cgi-bin/hgGateway; see Figure 1).

There is substantial variation in the number and organization of human TSPY genes between individuals, and this can be understood as a consequence of the complex repeated structures of the regions in which they lie. As indicated above, the number of copies in the tandem array varies, most likely as a result of homologous but unequal exchange events between sister chromatids leading to expansion and contraction of the array. Copy numbers reported in population samples range from 27–40 (n = 17 [5]), 18–40 (n = 42 [10]), 18–48 (n = 93 [11]) or 23–64 (n = 47 [12]), revealing the presence of greater than three-fold variation. In addition, a ∼4 Mb section of Yp containing the TSPY genes can be found in either orientation and has apparently undergone ≥12 inversion events mediated by flanking IR3 repeats during the evolutionary history of extant Y chromosomes [12], around 100,000 years. In one of these orientations, recombination can occur between DNA including the genes AMELY, TBL1Y and PRKY as well as some TSPY copies. Deletion carriers show no overt phenotypic effects, and the deletion is present at a frequency of ∼2% in the Indian subcontinent [14,15].

Two factors limit comparisons between human TSPY genes and those of other apes. First, some reference sequences, such as that of the gorilla, have been derived from females and thus provide no information about Y-specific genes such as TSPY. Second, although early studies had revealed the likely presence of multiple Y-specific TSPY genes in other apes [7,16,17], the complexity of the TSPY genomic structure meant that finished sequence was necessary for detailed comparison. Fortunately, a finished Y sequence is available for the chimpanzee [18]. The title of the paper presenting this work was “Chimpanzee and human Y chromosomes are remarkably divergent in structure and gene content”. The authors indeed documented 41 differences in gene content between the two Y chromosomes, far in excess of findings on other chromosomes. An earlier genomewide comparison had, for example, identified just 134 gene increase and six decrease events on the human lineage [19]. But even more remarkably, 70% of the Y-chromosomal gene copy differences (involving 29 gene copies) were due to different numbers of TSPY genes. Although TSPY was the most highly repeated gene on the chimpanzee Y chromosome, with six copies (equal to RBMY), the uniquely high copy number of TSPY in humans is matched only by the uniquely large differences in copy number between humans and chimpanzees. A more detailed comparison of TSPY organization in the two species is provided below in Section 2.

Two broad classes of evolutionary explanation could be considered for the extensive differences in TSPY copy number between humans and chimpanzees. TSPY might be evolving neutrally in one or both species, in which case the difference between six and ∼35 copies would have no functional significance and would be a consequence of neutral genetic drift. The alternative is that TSPY retains an important functional role in both species, and the different copy numbers have biological significance. The conservation of multiple TSPY copies on both Y chromosomes provides some evidence for functional relevance [20], and studies in humans tend to support the hypothesis of a functional role. TSPY is expressed specifically in spermatogonia in the testis [21], and decreased copy number has been associated with impaired spermatogenesis [22], although not in all studies [23,24]. Evolutionary analyses can also address the issue of the likely functional importance of the TSPY genes, and are considered below in Section 2.

Human TSPY genes, as noted above, are located in two regions of the Y: a large tandem array of ∼35 TSPY copies, and a single separate gene. Five unprocessed pseudogenes are also annotated in the reference sequence ([9], Table 1). The chimpanzee Y reference sequence, in contrast, has six active genes divided among three clusters containing four, one and one respectively, but 21 annotated unprocessed pseudogenes ([18], Table 1). Thus while the numbers of intact genes differ six-fold, the total numbers of genes plus pseudogenes are more similar, particularly when the variation within humans is taken into account.

In order to investigate the relationships between the two human and three chimpanzee TSPY clusters, we compared them and their flanking sequences using the program DOTTER [25]. Single-copy sequences with high similarity produce a diagonal line in such an analysis; tandem repeats produce a series of diagonal lines offset by the size of the repeating unit. The results of this analysis are shown in Figure 1. We can draw four main conclusions. First, the three panels representing the three chimpanzee TSPY clusters are broadly similar, apart from the different orientation of the cluster at 4.3 Mb. Second, the similarity between all of the different gene copies in the two species extends far outside the coding region to the entire repeat unit. This is expected from the locations of each chimpanzee TSPY cluster within one of the “pink” amplicons [18]. Third, the chimpanzee TSPY genes are organized into tandem arrays, but these are much smaller than the major human array, consisting of around four repeat units, and all have substantial deletions disrupting their regular structures. This is consistent with the inactive nature of the majority of the chimpanzee TSPY copies. Fourth, there is similarity between the sequences flanking the human major array on both sides and the sequences flanking the chimpanzee arrays. The sequences flanking the minor human array are less closely related.

To examine the phylogenetic relationships between the five TSPY clusters in more detail, we identified a flanking region with high sequence similarity (within the red circles in Figure 1), aligned the five sequences using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/) [26] and corrected the alignment manually. The alignment spanned 6277 bp and came from two nearby segments (human 6,147,115–6,148,562 and 6,150,328–6,155,088; 9,400,312–9,401,761 and 9,403,547–9,408,316; chimpanzee 4,275,601–4,277,071 and 4,269,058–4,273,824; 9,990,626–9,992,099 and 9,993,868–9,998,636; 13,631,075–13,632,544 and 13,634,296–13,639,063). After excluding length variations in mononucleotide runs, variable positions consisting of base substitutions or insertions/deletions were identified and, if immediately adjacent, assigned to a single mutational event. In this way, 515 mutational events were identified, 514 of which were consistent with a simple parsimony tree. This unrooted tree is shown in Figure 2.

From these two analyses, we can see that the human minor array is distinct from all the other arrays because most of its flanking regions do not align with them (Figure 1), and even where there is good alignment, it is the most divergent sequence (Figure 2). In contrast, the human major array and its flanking sequences show strong sequence similarity to the three chimpanzee arrays (Figure 1). The phylogenetic reconstruction, which in this part of the tree can be rooted using the human minor array as an outgroup, shows that the chimpanzee arrays were all derived from a shared common ancestor on the chimpanzee lineage after the human-chimpanzee split, and that the arrays at 4.3 and 9.9 Mb are the most closely related. Since homologous clusters of multiple TSPY genes are found in both species, the common ancestor is likely to have carried such a cluster, and if this is the case, the multiple TSPY pseudogenes in the chimpanzee arrays are likely to have been generated independently from functional genes on the chimpanzee lineage. Since the large number of structural differences between the human and chimpanzee Y chromosomes, and the extensive structural polymorphism even within humans, suggest that large numbers of rearrangements have occurred during the descent from this ancestral structure, the intermediate steps cannot be reconstructed in more detail using the current data. The lack of an outgroup, such as a finished gorilla Y chromosome sequence, also limits the deductions we can make about the ancestral structure, although cytogenetic analyses show one signal of intermediate intensity in gorillas [16]. Nevertheless, a simple model consistent with the data available would propose a moderate-sized ancestral tandem array of TSPY genes in the human-chimpanzee ancestor, with expansion on the human lineage contrasted with array duplications and pseudogenization on the chimpanzee lineage.

Here we return to the question of whether the rapid evolution of the TSPY genes is more likely to represent relaxation of constraint or selection for different functional configurations in humans and chimpanzees. If the TSPY genes were functionally unimportant in either species, we would expect to see an increase in the rate of nucleotide substitutions in the coding region, which might approach the rate in predominantly neutral regions of the genome such as introns. Hughes and colleagues [18] tabulated a comparison of the extent of divergence of the ampliconic gene sequences, divided into coding and intronic sequences (Table 2).

Table 2 shows that introns have diverged by 1.4 to 2.9%, similar to the overall average figure 1.7% for the X-degenerate regions of the Y chromosomes. In contrast, the TSPY coding region shows 0.8% divergence, substantially lower. These conclusions are consistent with an earlier study that examined part of exon 1, exon 2 and intron 1, and estimated 1.9% divergence for the intron-exon segment as a whole [27]. TSPY shows the lowest coding divergence of any of the ampliconic genes, but the highest intronic divergence. It therefore seems unlikely that ampliconic organization or a reduced mutation rate could account for the low coding divergence: functional constraint provides the best explanation.

To investigate this coding divergence in more detail, we compared the rates of synonymous substitutions, which do not change the coding sequence and are usually neutral, with those of non-synonymous, which alter the coding sequence, using the dN/dS statistic implemented in DNaSP (http://www.ub.edu/dnasp/). For these comparisons, we chose the two alternative CCDS annotations of TSPY as the human sequences (CCDS48204, CCDS48205), the chimpanzee sequence (ENSPTRT00000055849 with manual annotation according to information from the Page lab and new alignment id), and used the TSPY sequence from a new world monkey, the marmoset, as an outgroup (ENSCJAG00000034791). The results are shown in Table 3.

A dN/dS value below 1 indicates purifying selection, a value of 1 neutrality, and a value above 1 positive selection. The marmoset provides a relatively distant outgroup, and this has the advantage that many differences between the marmoset-great ape sequences have accumulated. The dN/dS values between marmoset and either human or chimpanzee are similar, and less than 1, showing that selection has acted to reduce the number of amino acid changes: purifying selection is seen here, as is expected to predominate in protein-coding regions. This supports the idea that these TSPY sequences are functional, and have been for most of their evolutionary history. The human-chimpanzee value, in contrast, is positive for CCDS 48204 and undetermined for CCDS 48205 since there are no synonymous differences in this comparison. If the number of synonymous differences were conservatively set to 1 instead of zero, the dN/dS value would be 2.11. These positive values could indicate positive selection for change in the amino acid sequence of TSPY since the human-chimpanzee split. However, this interpretation needs to be considered cautiously: if there had been two synonymous differences in each case, dN/dS would be close to 1, consistent with neutrality. However, the low number of human-chimpanzee differences in the TSPY coding region, compared with introns and other genes (Table 2) is not consistent with neutrality. In all, the analyses of the TSPY coding region support the idea of purifying selection on the TSPY sequence for most of its history, and reveal likely positive selection in the last few million years.