It is estimated that a human DNA sequence differs from that of one's parents at about 100 nucleotide positions1. These sites generally represent germline mutations that have arisen during the production of gametes in the parental generation. A classical question in human genetics is how many of these mutations come from mothers and fathers, respectively2. Is the paternal and maternal contribution to genetic novelty or erosion equal, or is there a male bias in the mutation rate? As the question has widespread implications for evolutionary, medical and molecular genetics, it has generated considerable interest over the years, without a consensus as to the extent of male bias. Some recent studies have indicated that the male bias may be much less than previously thought3,4. In a recent issue of Nature, however, Kateryna Makova and Wen-Hsiung Li5 provide evidence that a significant proportion of human mutations is indeed of paternal origin.

Measuring mutation rates

How to measure the rate of mutation in modern humans? One approach is to identify large numbers of random point mutations, which are transmitted from one generation to another in random families, thereby allowing one to calculate sex-specific mutation rates. This is a formidable task, however, and is hardly feasible with available technology. Rather than blindly grope for these random mutations, an alternative approach is to combine information from the very rare cases in which new mutations affect the phenotype of the offspring.

However, studies of the parental origin of novel mutations, like those causing achondroplasia and Apert syndrome, may be severely biased by recurrent mutations at individual CpG sites where methylation levels differ between the sexes. The observation that some diseases are caused exclusively by paternally derived mutations seems unlikely to apply to the genome as a whole6. The identification of de novo mutations in hypermutable markers such as microsatellites by pedigree analysis may offer another alternative, and has suggested a male:female mutation rate ratio (αm) of 3–5 (refs 7,8). We are then considering replication slippage–induced mutations in repetitive DNA, however, rather than the point mutations that are most prevalent in unique DNA.

Given the great difficulty in identifying random mutations in pedigrees, the issue of sex-specific mutation rates is better approached by molecular evolutionary studies of sex chromosomes9. Whereas the X chromosome spends only one-third of the time in males, the Y chromosome is present in males only. For paralogous genes shared between X and Y (ZFX/ZFY, for example; Fig. 1a), the rate of presumably neutral substitutions in primate comparisons is higher on Y than on X. The observed rate differences correspond to an αm of approximately 4–6, which suggests a rather distinct male bias in hominoids10. However, these data were recently challenged in a study by Bohossian et al.3, who suggested that αm is only 1.7, using data from a large region transposed from X to Y after the human lineage split with chimpanzees (Fig. 1b). Although perhaps surprising, this was a seemingly solid observation, given that the amount of sequence data included exceeded that of earlier studies by an order of magnitude.

Figure 1: Chromosomal approaches used to address sex-specific mutation rates.
figure 1

a, Comparison of substitution rates in homologous (gametologous) genes shared between X and Y. b, Comparison of substitution rates in a region recently transposed from X to Y. c, Comparison of substitution rates in a region transposed from chromosome 3 to Y. Although the comparison in b allows one to study mutations in the human lineage after the split with chimpanzees, the effect of ancestral polymorphism on estimates of divergence may be significant. Makova and Li5 argue that the comparison depicted in c is advantageous compared with the one in a, as it involves homologous sequences of more recent ancestry and avoids the possible effects of a specific reduction in the X-chromosome mutation rate.

Full size image

On ancestral polymorphism

Makova and Li5 now argue that the low estimate of αm was flawed, owing to the effect of ancestral polymorphism. Bohossian et al.3 inferred the number of nucleotide substitutions having arisen in the X and Y chromosome lineages subsequent to transposition by comparison to chimpanzee X-chromosome outgroup sequence (note that as the X-to-Y transposition occurred after the human–chimpanzee split, chimpanzee has only the X-linked sequence). One possible scenario for the X-to-Y transposition event in the human lineage is that the particular X-chromosome allele being transposed already contained polymorphic sites that distinguished it from other X-chromosome sequences. This could imply that some of the mutations interpreted as being of Y-chromosome origin (sites where human X = chimpanzee X ≠ human Y) could actually have arisen on the X chromosome, and were polymorphic prior to transposition. The effect of this would be to underestimate αm.

Makova and Li5 also make a more general point about ancient polymorphism and estimates of αm. Ancestral polymorphism can have a significant effect on the average divergence between two species—or two sequences—if they are closely related (Fig. 2). In other words, the time to the most recent common ancestor may significantly exceed the time to speciation. If ancestral polymorphism is low or effectively absent, however, the times to the most recent common ancestor and speciation would be more similar. This dichotomy may apply to X-and Y-chromosome sequences. The Y chromosome is typically very much lower in nucleotide diversity than the X11, probably because the smaller effective population size and the absence of recombination allows selection to wipe out polymorphism. Therefore, when closely related species are compared, ancient nucleotide diversity may contribute to the estimates of divergence in X, but not in Y. The combined effect would be to lower the estimates of αm, relative to that when more distantly related species are compared and where ancient nucleotide diversity has less influence on divergence.

Figure 2: The effect of ancestral polymorphism (width of branches) on the possible times to the most recent common ancestor in relation to time to speciation.
figure 2

The left panel depicts a scenario with significant ancient nucleotide diversity, implying that the time to the most recent common ancestor may be longer than the time to speciation. The right panel shows a situation with limited ancestral polymorphism, where the time to the most recent common ancestor is about the same as the time to speciation.

Full size image

DAZ to the rescue

To avoid these problems, and the potential pitfall of a suggested chromosome-specific reduction in the X-chromosome mutation-rate12, Makova and Li5 studied the DAZ (deleted in azoospermia) locus that was transposed from chromosome 3 to Y after the split between New World and Old World monkeys (Fig. 1c). In phylogenetic trees based on sequences from a number of primate species, they noted that internal branch lengths were significantly longer in trees constructed using Y sequences than in trees based on autosomal sequences. In contrast, external branches generally showed less pronounced differences in length between the two trees. This is consistent with divergence of autosomal, but not Y, sequences being affected by ancestral polymorphism.

By using the formula Y/A = 2αm /(1 + αm) on summed internal branch lengths of human–bonobo–gorilla–siamang–gibbon trees, they estimated αm to be 5.2. Clearly, this estimate is based on mutations that arose during several million years of evolution and does not necessarily reflect the mutation processes in contemporary human populations. However, there are no obvious reasons to think that human αm would be lower than in other primates. If anything, we should rather expect human αm to be higher owing to a longer generation time and an older age at reproduction than our ancestors and the apes.

The new αm estimate of 5.2 carries a large 95% confidence interval (2.44 to ∞), but it should be noted that it is similar to that obtained in previous comparisons of homologous X–Y genes10. If we accept a significant male bias in human mutation rate, it would suggest that a large proportion of germline mutations derive from replication-associated processes (replication errors, for example), a conclusion that can be drawn from the fact that the number of germline cell divisions in spermatogenesis vastly exceeds that in oogenesis13. The relative importance of replication-independent mutagenic factors on germline mutation would in this case be minor. As spermatogonial germ (stem) cells are continuously mitotically active in adult men, another consequence of these results is that we should expect the male mutation rate to increase with age14.

Is the new study without caveats? Although the results seem convincing, it might be premature to consider the case entirely closed. One potentially confounding factor relates to a recent observation that there is significant heterogeneity in the local mutation rate15. One could argue that the use of homologous sequences on different chromosomes would imply that the only cause of a difference in their mutation rate is the varying length of time spent in the male and female germlines. However, if the local mutation rate varies over the genomic landscape in a way that it is primarily determined by the regional genomic context rather than the primary sequence context, then homologous sequences may not necessarily have similar intrinsic mutation rates. To judge whether this is the case, we need to learn more about the determinants of local mutation rate variation.