Hans‐Jürgen Bandelt
Dear Sir, In science, it is not surprising to have exciting claims later refuted by hard evidence – although it may even be harder to have them corrected in the same journal where they originally appeared (Don't challenge a winning paper: Bär, 2006). One of the main reasons why false claims could pass through peer reviewing and eventually appear in print is (i) limited knowledge about certain genetic markers in a newly emerging field at the time of submission or (ii) insufficient screening of the published data for comparison and (iii) inadvertent documentation and analysis of the novel data underlying such claims. It may then take some years before the true causes for a particular finding would come to light. This apparently applies to the remarkable claim made by Holyoake et al. (2001) that the mitochondrial DNA (mtDNA) mutations G9055A and G11719A would compromise the semen quality of those men who possess any of these mutations in their mtDNAs (leading to reduced sperm motility and/or low sperm count). This study won the 2nd Prize of the European Academy of Andrology in 2001 – a winning paper, thus. The results of this study were aptly challenged by Montiel-Sosa et al. (2002), who (i) pointed to the fact that site 11719 determines a basal European mtDNA haplogroup (nowadays referred to R0), (ii) hinted at the peculiar frequency of the 11719 polymorphism as reported by Holyoake et al. (2001), and (iii) argued that ethnic origins of samples in a mixed population would matter and that 'the control and low sperm motility group could be non-homogenous'. Indeed, no reference to haplogroups, constituting the major branches of the mtDNA phylogeny, was made in the study by Holyoake et al. (2001), although it was submitted at a time (September 2000) where the role of the two particular (G9055A and G11719A) and other mutations was already quite well understood. Holyoake et al. (2002) in their reply to Montiel-Sosa et al. (2002) emphasized that they were only 'interested in single nucleotide substitutions in relationship to poor semen quality'. This popular 'allelic' approach gleaned from the analysis of the nuclear genome, however, is patently unsuitable for a deeply hierarchical genetic system such as mtDNA. First, the allelic view ignores the tight link between different mutations in the mtDNA phylogeny (as between G9055A and G11719A). Clusters of mutations would show up together and could then not serve as independent indicators of a disease phenotype. Second, disregard for the mtDNA phylogeny leaves out the option of phylogenetic proof-reading of obtained mtDNA results in regard to potential errors (Bandelt et al., 2005a). Not unexpectedly, the detailed haplogroup analysis of Pereira et al. (2005) for a large Portuguese sample did not support the notion that men carrying an mtDNA from a specific haplogroup overall had an increased or a reduced risk of infertility. These authors also warned that the observation by Ruiz-Pesini et al. (2000) that 'haplogroups H and K are significantly more abundant in nonasthenozoospermic and asthenozoospermic populations, respectively' may have been caused by population stratification. As long as patients and controls are not controlled for ethnic background, geographical matrilineal ancestry and social stratum (which certainly matters in countries with a considerable record of recent immigration), mere correlation of a mutation/haplogroup with a disease phenotype may be spurious and thus cannot provide sufficient evidence for an association. The complete mtDNA data of Ingman et al. (2000) as well as the RFLP-based analysis of West Eurasian mtDNAs of Macaulay et al. (1999) would have provided the necessary background information for the SSCP analysis aimed at by Holyoake et al. (2001). Curiously, the Ingman et al. (2000) study was referred to in a later article (Gemmell & Sin, 2002) by the same senior author (F. Y. T. Sin), but apparently without reappraising the findings of his former paper. If the data by Ingman et al. (2000) had been inspected at the time, then it would have become clear that the frequencies of the mutations observed by Holyoake et al. (2001) are mutually inconsistent and thus cannot realistically correspond to real-world data. To see this, take a preview of the worldwide mtDNA phylogeny as seen from the early Ingman et al. (2000) data, but now enriched with the up-to-date information about the nesting of the corresponding haplogroups, as displayed in Figure 1 of Bandelt et al. (2006). Nucleotide A at site 8860 is shared by virtually all mtDNAs worldwide that are not closest relatives of the revised Cambridge reference sequence (rCRS; Andrews et al., 1999). In Table 3 of Holyoake et al. (2001) the frequencies are recorded as 2/102 and 1/59 instead of the expected frequency of (nearly) 100%! The other frequencies in that table do not fare much better. Virtually all mtDNAs not belonging to the basal West Eurasian haplogroup R0 (which encompasses the sister haplogroups R0a and HV) bear the characteristic mutation G11719A. Assuming the frequency for this mutation in 'normozoospermic' men reported by Holyoake et al. (2001), one would be forced to conclude that all 80 control mtDNAs are members of haplogroup R0 (Table 1). Such a population sample has never been observed anywhere in the world. Moreover, this zero frequency is at odds with the other mutation frequencies recorded, as there is evidence that the other two basal West Eurasian haplogroups, JT and U (including K), were observed in both, the 'normozoospermic' and the 'subnormozoospermic' men (Table 1). The two frequencies of G11719A, on their own, are thus far apart from any realistic value, indicating that the recognition of G11719A via SSCP analysis must have been strongly hampered by technical problems. There is, by the way, little evidence that Polynesian mtDNAs played a noticeable role, because the vast majority of them would fall into a specific branch of haplogroup B. This haplogroup would be pinpointed by the 9 base-pair (bp) deletion, which, however, was found at a meagre 2.8% in the total sample. Note that the 9-bp deletion is also seen sporadically on many haplogroup backgrounds other than B. Furthermore, potential Sub-Saharan African mtDNA mutations (possibly C7789A) or (South-)East Asian mtDNA mutations (possibly G7853A) are present (if at all) at similarly minor frequencies; thus, in particular, the suggestion of considerable African mtDNA ancestry (Montiel-Sosa et al., 2002) receives no support from the data. Therefore, the vast majority (perhaps >95%) of mtDNAs analysed by Holyoake et al. (2001) must have been of European descent. Table 4 of Holyoake et al. (2001), which provides the full haplotype information, indicates further problems. Mutations that would be expected to be linked with other mutations in view of the above inequalities are not just absent but show up in unexpected combinations, almost random-like, which would rather suggest sample mix-up. Moreover, the haplogroup K mutation G9055A should always entail the mutation G11719A specific to non-R0 lineages, but according to that table, both mutations are rather unlinked. In the present Table 2, some haplotypes from Table 4 of Holyoake et al. (2001) are listed that combine at least two mutations. In three cases, all from the 'subnormozoospermic' group, one observes potentially mosaic patterns (Table 2). The best explanation for the odd frequency spectrum as well as the incomplete and mosaic haplotypes is, in the first place, massive failure of the SSCP analysis carried out by Holyoake et al. (2001). Such a mis-analysis is not an infrequent phenomenon in medical genetics (see e.g. Bandelt et al., 2005b). But why would have 'subnormozoospermia' triumphed over 'normozoospermia' e.g. in the screening of G11719A? And why would the 11719 polymorphism, if deemed crucial for 'subnormozoospermia', have only been analysed in a meagre 33 instead of, say, 131 patients? And what about G9055A? A normal mtDNA sample of predominantly mixed European descent would hardly show so few (<1%) haplogroup K members (Behar et al., 2006). The consistent trend in identifying mutations at frequencies that are far too low, especially for the 'normozoospermic' group, could suggest that the SSCP analysis was biased and effectively carried out only for a minority of the samples. It then seems that the number of fully screened mtDNAs was smaller for the 'normozoospermic' group than for the 'subnormozoospermic' group. In summary, there is no solid evidence that primary mtDNA substitutions other than recognized pathogenic mutations, mainly in heteroplasmic state, which usually cause complex disease phenotypes, influence sperm motility in any way. Therefore, it would be premature to accept all hypotheses and interpretations put forward by Gemmell & Sin (2002) and St. John et al. (2005), who took the results of Ruiz-Pesini et al. (2000) and Holyoake et al. (2001) at face value. Before far-reaching implications are discussed, the underlying data should come under scrutiny first. It would certainly be desirable to re-study the mtDNA samples employed by Holyoake et al. (2001) by performing highest quality sequencing. In any case, there is an urgent need for a new start of mtDNA analysis of samples from men with idiopathic infertility (or subfertility) by complete mtDNA sequencing in order to provide a solid database to which any subsequent case studies could get compared and evaluated. The most recent work of Pereira et al. (2007) is a promising first step in this direction.