The existence of hypervariable sites (sites that evolve at a rate much faster than average) in the noncoding human mtDNA control region has been well documented via various analyses of human mtDNA variation (Hasegawa et al. ¹⁹⁹³; Wakeley ¹⁹⁹³; Excoffier and Yang ¹⁹⁹⁹; Meyer et al. ¹⁹⁹⁹); however, it is not known why such hypervariability exists. In the absence of any evidence, it has been generally assumed that hypervariable sites represent mutational hotspots, but recently it was suggested that hypervariable sites may actually reflect ancient mutations that have been redistributed among mtDNA lineages via recombination (Eyre-Walker et al. ¹⁹⁹⁹; Hagelberg et al. ¹⁹⁹⁹). Although recent claims of recombination in human mtDNA (Awadalla et al. ¹⁹⁹⁹; Eyre-Walker et al. ¹⁹⁹⁹; Hagelberg et al. ¹⁹⁹⁹) have not been substantiated in some instances (Macaulay et al. ¹⁹⁹⁹; Hagelberg et al. ²⁰⁰⁰) and remain controversial (Jorde and Bamshad ²⁰⁰⁰; Kivisild and Villems ²⁰⁰⁰; Kumar et al. ²⁰⁰⁰; Parsons and Irwin ²⁰⁰⁰), the suggestion that hypervariable sites are not mutational hotspots deserves consideration.

One way to test this hypothesis is to examine the evolutionary rate for sites at which new mutations have been observed. If hypervariable sites are indeed mutational hotspots, then new mutations should occur preferentially at hypervariable sites. However, if hypervariable sites are not mutational hotspots but, instead, reflect recombination (or some other process), then new mutations would not be expected to occur preferentially at hypervariable sites. In this report, I analyze the evolutionary rate for sites at which new mutations are observed, to test the hypothesis that hypervariable sites are mutational hotspots.

The evolutionary rate for each site in the two hypervariable segments (HV1 and HV2; positions 16024–16383 and 57–372, respectively, numbered according to the reference sequence [Anderson et al. ¹⁹⁸¹]) of the human mtDNA control region has been estimated elsewhere (Meyer et al. ¹⁹⁹⁹), from patterns of variation in worldwide populations; the overall average evolutionary rate has been found to be .60 for HV1 alone and .54 for HV1 and HV2. Two large family studies reported a total of 12 germline mtDNA mutations for HV1 and HV2 (Parsons et al. ¹⁹⁹⁷; Sigurdardottir et al. ²⁰⁰⁰), and the average evolutionary rate for the sites at which these 12 mutations occurred is 2.82 (table 1), which is 5.2 times greater than the overall average evolutionary rate for HV1 and HV2. Data are also available for 12 somatic mtDNA mutations in HV1 and HV2 that were observed in tumors but not in control normal tissues from the same individuals (Fliss et al. ²⁰⁰⁰). The average evolutionary rate for the sites at which these 12 tumor mutations occurred is 2.86 (table 1), which is 5.3 times greater than the overall average evolutionary rate for HV1 and HV2. Finally, a number of studies have documented the existence of heteroplasmy (i.e., the existence of more than one mtDNA type) for sites in the mtDNA control region (Comas et al. ¹⁹⁹⁵; Bendall et al. ^{1996, 1997}; Wilson et al. ¹⁹⁹⁷; Hühne et al. ¹⁹⁹⁸; Sigurdardottir et al. ²⁰⁰⁰; Tully et al. ²⁰⁰⁰). In principle, the heteroplasmic state of an individual could have arisen either via transmission from the mother (i.e., through the germline) or via a new somatic mutation occurring within the individual; most of the studies that have reported heteroplasmy have not attempted to determine which mechanism might be responsible. Therefore, for this analysis, all heteroplasmic mutations were grouped together, resulting in data on 51 mutations (table 1) in HV1 alone. The average evolutionary rate for the sites at which these 51 heteroplasmic mutations were observed is 2.87 (table 1), which is 4.8 times greater than the overall evolutionary rate for HV1.

Table 1

Sites in HV1 and HV2 at Which Germline, Tumor, and Heteroplasmic Mutations Were Observed and Their Estimated Evolutionary Rate

		No. of Mutations
Sitea	Rateb	Germline	Tumor	Heteroplasmic
16092	.69	1	0	0
16093	2.98	2	0	13
16104	.47	0	0	2
16111	2.57	1	0	1
16124	.81	0	0	1
16129	4.78	0	0	6
16172	3.76	0	1	1
16183	3	0	1	0
16185	.61	0	0	1
16187	3.67	0	1	0
16189	4.78	0	1	3
16192	3.6	0	0	3
16205	0	0	0	1
16222	.41	0	0	1
16234	.85	0	0	1
16239	.79	0	0	2
16256	2.3	1	0	2
16257	.44	0	0	1
16262	.05	0	0	1
16265	2.18	0	1	0
16292	2.32	0	1	0
16293	3.9	0	0	3
16295	.72	0	0	1
16300	2.1	0	1	0
16301	.3	0	0	1
16309	4.03	0	0	2
16311	4.78	0	0	1
16355	2.16	0	0	1
16362	4.72	0	0	2
16380	0	0	1	0
75	0	0	1	0
94	0	1	0	0
114	.17	0	1	0
150	6.12	0	1	0
153	3.99	1	0	0
185	4.86	1	0	0
189	6.22	1	0	0
195	6.22	0	1	0
207	3.29	2	0	0
234	.69	1	0	0
Totalc		12 (2.82)	12 (2.86)	51 (2.87)

Open in a separate window

^aFrom Comas et al. (¹⁹⁹⁵); Bendall et al. (^{1996, 1997}); Parsons et al. (¹⁹⁹⁷); Wilson et al. (¹⁹⁹⁷); Hühne et al. (¹⁹⁹⁸); Fliss et al. (²⁰⁰⁰); Sigurdardottir et al. (²⁰⁰⁰); Tully et al. (²⁰⁰⁰). Sites are numbered according to the published reference sequence (Anderson et al. ¹⁹⁸¹).

^bFrom Meyer et al. (¹⁹⁹⁹).

^cNumbers in parentheses are average rates.

The distribution of inferred evolutionary rates for all sites in HV1 and HV2 is shown in figure 1, along with the distributions for the sites at which germline, tumor, and heteroplasmic mtDNA mutations were reported. A nonparametric (Mann-Whitney) test indicates that the distribution of evolutionary rates for sites at which germline, tumor, and heteroplasmic mutations were reported are all significantly different from the distribution of evolutionary rates for all sites in HV1 and HV2 (P<.001 for all comparisons). Thus, germline, tumor, and heteroplasmic mtDNA mutations all occur preferentially at hypervariable sites in the human mtDNA control region.

Open in a separate window

Figure 1

Frequency distribution of inferred evolutionary rates (Meyer et al. ¹⁹⁹⁹) for all sites in HV1 and HV2 and for sites at which germline, tumor, and heteroplasmic mutations were observed.

The average evolutionary rates for sites at which germline, tumor, and heteroplasmic mutations were observed were remarkably similar (table 1). Moreover, the distribution of evolutionary rates (fig. 1) does not differ among germline, tumor, or heteroplasmic mutations (pairwise Mann-Whitney tests; P>.05 for all comparisons). This overall similarity among germline, tumor, and heteroplasmic mutations may indicate that the underlying mutational processes for mtDNA in germ cells and in somatic cells are similar. However, it is also possible that some (or even all) of the supposed germline mutations inferred from family studies are actually somatic mutations that occurred early during the prenatal development of the individual and drifted to homoplasy and that heteroplasmic mutations may reflect an intermediate stage in this process. The observed close correspondence between the evolutionary rates for somatic mutations (both tumor and heteroplasmic) and for germline mutations would then arise because the supposed germline mutations are in fact somatic mutations. If not all mutations observed in family studies are true germline mutations, then this may also account for the fact that evolutionary rates that have been determined from family studies are higher than those inferred from evolutionary studies (Parsons et al. ¹⁹⁹⁷; Jazin et al. ¹⁹⁹⁸; Sigurdardottir et al. ²⁰⁰⁰). It would be of interest to know whether individuals who exhibit what appear to be newly arisen, germline mtDNA mutations actually transmit the new mutation to their offspring.

In conclusion, both somatic and (supposed) germline mtDNA mutations occur preferentially at hypervariable sites in HV1 and HV2. This result is not consistent with the idea that such hypervariable sites may reflect ancient mutations that have spread to other mtDNA lineages via recombination (Eyre-Walker et al. ¹⁹⁹⁹; Hagelberg et al. ¹⁹⁹⁹). A possible counterargument would be that new mutations actually are rare recombination events between paternal and maternal mtDNA. For this to be true, however, an unusual recombination mechanism must be postulated, because observed mutant mtDNA types differ from the parental type almost exclusively by single mutations, whereas paternal and maternal mtDNA typically differ at several nucleotide positions in the control region. Thus, any supposed mtDNA recombination must instead operate more like gene conversion, involving only short stretches of mtDNA sequence, perhaps even single nucleotide positions. It is difficult to imagine that such a mechanism could be operating in both the germline and in somatic cells (especially tumor cells). Moreover, if recombination involves only single nucleotide positions, then there is no longer the expectation that linkage disequilibrium should be negatively correlated with the physical distance between pairs of sites, a position still held by some (Awadalla et al. ²⁰⁰⁰) despite criticisms concerning the measure of linkage disequilibrium used (Jorde and Bamshad ²⁰⁰⁰; Kumar et al. ²⁰⁰⁰).

Given these difficulties with a recombinational mechanism for new mutations, it does appear that hypervariable sites are hotspots for both germline and somatic mutations. The challenge now facing mitochondrial geneticists is to determine what features of these hypervariable sites and/or the mtDNA replication machinery cause them to be mutational hotspots.