Nonneutral Evolution and Differential Mutation Rate of Gender-Associated Mitochondrial DNA Lineages in the Marine Mussel Mytilus

Corresponding author: David O. F. Skibinski, School of Biological Sciences, University of Wales, Swansea SA2 8PP, UK, d.o.f.skibinski{at}swansea.ac.uk (E-mail).

Communicating editor: A. G. CLARK

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX 1 LITERATURE CITED

Mussels have two types of mitochondrial DNA (mtDNA). The M type is transmitted paternally, and the F type is transmitted maternally. To test hypotheses of the molecular evolution of both mtDNA genomes, 50 nucleotide sequences were obtained for 396 bp of the COIII gene of European populations of Mytilus edulis and the Atlantic and Mediterranean forms of M. galloprovincialis. Analysis based on the proportion of synonymous and nonsynonymous substitutions indicate that mtDNA is evolving in a non-neutral and complex fashion. Previous studies on American mussels demonstrated that the F genome experiences a higher purifying selection and that the M genome evolves faster. Here we show that these patterns also hold in European populations. However, in contrast to American populations, where an excess of replacement substitution between F and M lineages has been reported, a significant excess of replacement polymorphism within mtDNA lineages is observed in European populations of M. galloprovincialis. European populations also show an excess of replacement polymorphism within the F but not within the M genome with respect to American M. trossulus, as well as a consistent pattern of excess of rare variants in both F and M genomes. These results are consistent with a nearly neutral model of molecular evolution and a recent relaxation of selective constraints on European mtDNA. Levels of diversity are significantly higher for the M than F genome, and the M genome also accumulates synonymous and nonsynonymous substitutions at a higher rate, in contrast with earlier reports where no difference for the synonymous rate was observed. It is suggested that a subtle balance between relaxed selection and a higher mutation rate explains the faster evolutionary rate of the M lineage.

RECENT reports have challenged the axiom that mitochondrial DNA (mtDNA) evolves under a strictly neutral model of molecular evolution (reviewed in BALLARD and KREITMAN 1995 ; Rand AND KANN 1996 ). Most of these reports on mtDNA also show a variety of complex patterns of variation, suggesting that mtDNA evolution can be governed by many of the same selective forces found to operate on the nuclear genome, but also by factors that do not affect nuclear variation (BALLARD and KREITMAN 1995 ). Tests of neutrality, beyond simply giving evidence for or against selection, can provide insights into the other diverse processes that shape and maintain genetic variation in nature.

Mussels of the genus Mytilus have an unusual mode of mtDNA inheritance that provides a novel model for investigating mechanisms of mtDNA evolution. Mussels have two mtDNA genomes called F and M (FISHER and SKIBINSKI 1990 ). Females transmit the F genome to both daughters and sons, whereas males transmit the M genome to sons, thus the M genome is inherited paternally and the F genome is inherited maternally (SKIBINSKI et al. 1994A , SKIBINSKI et al. 1994B ; ZOUROS et al. 1994A , ZOUROS et al. 1994B ). This mode of inheritance accounts for the high levels of heteroplasmy for the M and F genomes in males (FISHER and SKIBINSKI 1990 ; HOEH et al. 1991 ), and the high sequence divergence (>20%) between these two lineages (SKIBINSKI et al. 1994B ), whose separation predates current Mytilus taxa (RAWSON and HILBISH 1995 ; STEWART et al. 1995 ; QUESADA et al. 1996 ). Thus, Mytilus mtDNA provides a potentially powerful model for detecting the influence of natural selection because analyses can be applied at several different levels of divergence, between and within the F and M genomes, and between and within the different forms or species of Mytilus.

mtDNA is also noteworthy for evolving faster in Mytilus than in other metazoans, a feature that has been attributed to a relaxed selective constraint associated with biparental inheritance (HOEH et al. 1996 ). Several studies have also suggested that the M genome is more polymorphic and evolves faster overall than the F genome (SKIBINSKI et al. 1994B ; RAWSON and HILBISH 1995 ; STEWART et al. 1995 ; QUESADA et al. 1996 ), though without statistical support. However, for American populations of Mytilus edulis and M. trossulus, STEWART et al. 1995 , STEWART et al. 1996 demonstrated a significantly higher replacement rate for the M than the F genome in the COIII gene and in a mtDNA segment with no assigned function. STEWART et al. 1996 also noted a higher replacement rate for the M than the F genome in nonconserved regions, as well as a significant excess of replacement substitutions between M and F lineages when applying the McDonald-Kreitman test. They interpreted these observations as evidence for non-neutral evolution of Mytilus mtDNA associated with relaxed selection on the M genome. Because males carry both F and M genomes in their somatic tissue, but females usually only have the F genome, the M genome should experience more relaxed selection overall. While the departure from neutrality for mtDNA in American Mytilus taxa is opposite in direction to that reported in other non-Mytilus taxa, where excesses of replacement polymorphisms within lineages have been usually reported (e.g. BALLARD and KREITMAN 1995 ; Rand AND KANN 1996 ), American mussels are also noteworthy for having strong nuclear/mtDNA incompatibilities leading to the breakdown of biparental inheritance in hybrids and to intrinsic barriers blocking mtDNA introgression between different species (ZOUROS et al. 1992 ; RAWSON et al. 1996 ; SAAVEDRA et al. 1996 ). The restricted mtDNA introgression between American Mytilus taxa is in clear contrast to the extensive levels of mtDNA introgression observed between European taxa (QUESADA et al. 1995B , QUESADA et al. 1998 ), raising the possibility of diverse evolutionary forces governing mtDNA polymorphisms.

In this study, we assess this possibility by examining patterns of mtDNA evolution in European populations of M. edulis and in the recently reported Atlantic and Mediterranean forms of M. galloprovincialis (QUESADA 1993 ). For comparative purposes, we examined the COIII gene for a segment that covers the same region that was analyzed previously in American mussels (STEWART et al. 1995 , STEWART et al. 1996 ). We ask the following questions: Are patterns of mtDNA nucleotide variation compatible with the expectations of a strictly neutral model when a combination of neutrality tests are applied? If not, is the high excess of replacement substitutions between F and M lineages in American mussels also seen in European mussels? Is there variation between European taxa in the degree of departure from neutral expectations? Are an evolutionary rate and diversity higher for the M than for the F genome general features of Mytilus mtDNA, and, if so, what is the role of the differences in mutation rate vs. relaxed constraint?

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX 1 LITERATURE CITED

Sampling:
Mussels from the three Mytilus taxa were collected from five European localities that are known (GOSLING 1992 ; QUESADA 1993 ) to contain pure M. edulis (Redcar and Mk. Point, northeastern and southwestern Britain, respectively), Atlantic M. galloprovincialis (Vigo, northwestern Spain), and Mediterranean M. galloprovincialis (Peñiscola and Chioggia, eastern Spain and northeastern Italy, respectively). Mussels were sexed by microscopic examination of gonads. A total of 50 individuals (24 males and 26 females) were chosen at random from samples from these populations for sequence analysis.

DNA preparation and sequencing:
Template for PCR amplification was prepared from single mussels using a CTAB-based method of mtDNA extraction (FISHER and SKIBINSKI 1990 ). Extractions from females were used for PCR amplification of a 476-bp fragment of the F mtDNA COIII gene using the following Mytilus specific primers: forward FOR1 5'CCAAAC CCGTCATCTACTAG-3' and reverse REV1 (ZOUROS et al. 1994B ) 5'-ATGCTCTTCTTGAATATAAGCGTACC-3', which correspond to nucleotide positions 805–824 and 1326–1301, respectively, of segment 5 described in HOFFMANN et al. 1992 . DNA from males was used for PCR amplification of the M genome using the following set of M specific primers (SKIBINSKI et al. 1994B ): forward FOR2 5'-AAACCCTTCGTCCAC AAGG-3' and reverse REV2 5'-AGCCTTTTTGTCATCATT CTGT-3', homologous to nucleotide positions 806–824 and 143–164 of segments 5 and 6, respectively. This 1.5-kb fragment was purified from 1% agarose gels and cleaned with Spin-X columns (Costar, Cambridge, MA). The eluted DNA was subsequently used as a source for the amplification of the same fragment as amplified from females using the primers FOR2 and REV1 given above. PCR reactions and cycling conditions were as described in QUESADA et al. 1996 , except that a total of 25 cycles were used. Double-stranded F and M PCR products were cleaned with the Quiaquick kit (Quiagen, Chatsworth, CA) before sequencing. Both PCR strands were sequenced using the Thermo Sequenase cycle sequencing kit (Amersham, Arlington Heights, IL) on an ALF automated sequencer (Pharmacia, Piscataway, NJ). About 200–300 ng of PCR product and 10 pM of primer were used per sequencing reaction.

Sequence analysis:
Sequences were aligned using the Clus-talW (HIGGINS et al. 1996 ) computer program. Nucleotide sequences were translated into amino acid sequences using the Drosophila mtDNA genetic code, as described in HOFFMANN et al. 1992 . Nucleotide diversity was calculated for each taxon using the FU 1994 estimator of . This new estimator has a variance that is substantially smaller than that of any existing estimator by making full use of phylogenetic information (FU 1994 ). The proportion of synonymous (K_S) and nonsynonymous (K_A) substitutions per site were calculated for all pairwise combinations of sequences according to the method of COMERON 1995 . This method minimizes stochastic errors by separating the twofold degenerate sites into sites where only transitional and transversional substitutions are synonymous. The Poisson-corrected proportion of amino acid substitutions was also calculated for all pairwise comparisons of sequences.

The sequences were tested for departures from neutral expectations using several tests as implemented in the DnaSP package (ROZAS and ROZAS 1997 ). The McDonald and Kreitman test (MCDONALD and KREITMAN 1991 ) was used to test that the ratio of replacement to silent substitutions should be the same within and between lineages. For their analysis, STEWART et al. 1996 pooled sequences of different taxa within the F and within the M lineages. Here we chose not to pool our data in this way because such a procedure has the potential to obscure any differences that might exist among taxa. Probability values for contingency tables were calculated by the Monte Carlo method (ROFF and BENTZEN 1989 ) using 10,000 permutations per analysis. To test for departures from neutrality in relation to the frequency distribution of F and M polymorphisms within each Mytilus taxon, we used the tests of TAJIMA 1989 and FU and LI 1993 . This was done for nondegenerate and fourfold degenerate sites to determine whether patterns of variation among lineages are as predicted by neutral theory (GILLESPIE 1991 ).

Phylogenetic relationships among sequences were estimated using the neighbor-joining method (MEGA version 1.02; KUMAR et al. 1993 ). The analyses were repeated using the Fitch-Margoliash method (PHYLIP 3.1; FELSENSTEIN 1993 ). Pairwise distances between sequences were calculated using Kimura's two-parameter correction for multiple hits. Gaps in the sequences were removed in pairwise comparisons. The level of support for the resulting phylogenetic trees was determined using 1000 bootstrap replications for the neighbor-joining method and 100 replications for the Fitch-Margoliash method. Similar topologies were obtained using the two methods, thus for brevity, the neighbor-joining tree alone is presented.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX 1 LITERATURE CITED

Phylogenetic relationships:
A total of 50 sequences (24 M and 26 F haplotypes) were scored for the same segment of 396 bp of the COIII gene. Polymorphic sites are given in the Appendix 1. There are 172 polymorphic sites over the whole data set. For comparative purposes, 308 bp of these sequences were aligned against the homologous 308 bp of the seven published COIII sequences of the highly diverged American M. trossulus M and F haplotypes (STEWART et al. 1995 ). The full 396 bp of the COIII sequence was used in all those analyses not involving American M. trossulus.

The neighbor-joining tree (Figure 1) indicates a primary division of American and European sequences into highly diverged male (M lineage) and female (F lineage) types. This result is consistent with earlier reports in Mytilus (RAWSON and HILBISH 1995 ; STEWART et al. 1995 ). The main branch leading to the M lineage is longer, as are secondary branches, compared with corresponding branches in the F lineage. The Poisson-corrected number of amino acid substitutions per site between European sequences and the highly diverged American M. trossulus sequences were significantly different (z-test, P < 0.001) between the F (0.018 ± 0.015) and M (0.100 ± 0.015) lineages. These results correspond with those published previously for American taxa (RAWSON and HILBISH 1995 ; STEWART et al. 1995 , STEWART et al. 1996 ), showing a higher evolutionary rate for the M lineage. There is clear and substantial geographic structure evident in the variation between Atlantic and Mediterranean M. galloprovincialis (QUESADA et al. 1995A , QUESADA et al. 1995C ). In the present study, Mediterranean M. galloprovincialis exhibit at least two highly diverged groups of F and M haplotypes that are not linked to particular geographical sites. F and M haplotype assemblages are not always concordant with the taxonomic identification of the individuals in which they are found. Hybridization and mtDNA introgression among European Mytilus taxa can explain these patterns of variation, as discussed elsewhere (QUESADA et al. 1995B , QUESADA et al. 1996 ).

View larger version (19K): In this window In a new window Download PPT slide   Figure 1. Neighbor-joining tree of the 50 sequences from three European taxa, plus the seven homologous published sequences from American M. trossulus (STEWART et al. 1995 ). The tree is based on 308 nucleotides and is rooted between the male and female lines. Only bootstrap values >50 are presented. Sequences are indicated by an alphanumeric code indicating the taxa (ED, M. edulis; AG, Atlantic M. galloprovincialis; MG, Mediterranean M. galloprovincialis; TR, M. trossulus) followed by the individual and locality from which the sequence was obtained: M, Mk. Point; R, Redcar; V, Vigo; P, Peñiscola; C, Chioggia; A, North America.

Diversity and divergence for replacement and synonymous changes:
Estimates of nucleotide diversity for both F and M genomes, as measured by , are given in Table 1. Mediterranean M. galloprovincialis has the highest diversity value for both F and M genomes. Genetic diversity is higher for the M genome than for the F genome within all taxa, although the difference is never significant (z-test). However, a two-tailed Wilcoxon sign rank test applied for the five populations across the three taxa does give a significant (P < 0.05) result (Table 2). When synonymous and nonsynonymous substitutions are analyzed separately, nucleotide diversity is significantly higher (P < 0.05) for synonymous than for nonsynonymous substitutions (Table 2). Moreover, the M genome shows a genetic diversity significantly higher than the F genome for synonymous substitutions, but not for nonsynonymous substitutions. A three-way analysis of variance can be carried out on the data set of 20 values for nonsynonymous and synonymous substitutions given in Table 2. The two two-way interactions (population x type of substitution and genome x population) are nonsignificant when assessed against the three-way interaction. The interaction between genome and type of substitution is, however, significant when assessed against these other interactions pooled as error (F_1,12 = 7.836, P < 0.025). These observations suggest differences in the evolutionary factors operating on each genome. However, stochastic factors could also account for a higher diversity of the M genome because local populations from the same taxa cannot be considered evolutionarily independent. However, synonymous substitutions are intermingled with nonsynonymous substitutions within both the F and M mtDNA molecules. Thus, the two types of substitutions should not show different patterns of variation between genomes as a result of drift, the opposite of what is observed. In addition, the above result is consistent with earlier claims suggesting higher diversity for the M than the F genome in populations not closely related to those analyzed here, such as American M. edulis and American M. trossulus (STEWART et al. 1995 , STEWART et al. 1996 ). These earlier reports did not provide statistical support for higher diversity for the M genome, nor did they consider the role of drift as a potential explanation for the observed differences between genomes. Consideration of all data sources suggests strongly that a higher diversity for the M genome is a general feature of Mytilus mtDNA—contradictory observations have never been reported (SKIBINSKI et al. 1994B ; RAWSON and HILBISH 1995 ; STEWART et al. 1995 , STEWART et al. 1996 ; QUESADA et al. 1996 , QUESADA et al. 1998 ).

View this table: In this window In a new window   Table 2. FU 1994 estimates of diversity per population

Table 3 gives the number of synonymous (K_S) and nonsynonymous (K_A) substitutions observed between European taxa in comparisons for the F or for the M genome. Both estimates are higher for the M lineage in five of the six comparisons, but the difference between M and F genomes is not significant, for K_A or K_S, in any one of these comparisons. The K_A/K_S ratio was unusually high, e.g., compared with STEWART et al. 1996, in all three taxa, and F and M genomes displayed similar ratios within each of the taxa. This observation implies either that F and M mtDNA genomes are under a similar and dramatic relaxed selection, or that selective pressures of similar strength are operating on both genomes. When pooled European sequences are compared with the highly diverged American M. trossulus sequences, a remarkably different pattern is apparent (Table 3). First, both K_A and K_S are significantly (P < 0.001) greater for the M genome. Second, the K_A/K_S ratio is 3 times higher for the M lineage. Third, the K_A/K_S ratios are lower than those observed in European comparisons, 2 to 3 times lower for the M genome, but 11–19 times lower for the F genome. These contrasting results, when we compare closely related European sequences with highly diverged M. trossulus sequences, suggest that patterns of mtDNA variation are critically affected by the time scale considered. Thus, in the long term, the M genome accumulates more synonymous and nonsynonymous substitutions than the F genome, although the nonsynonymous substitutions are accumulated at a higher rate. This observation is consistent with previous studies on American Mytilus taxa, suggesting a higher evolutionary rate and relaxed constraint for the M genome (STEWART et al. 1996 ).

View this table: In this window In a new window   Table 3. Jukes-Cantor corrected estimates of synonymous (KS) and nonsynonymous (KA) substitutions (COMERON 1995 )

Tests of neutrality:
Table 4 summarizes the number of replacement and silent substitutions observed within and between F and M genomes for each European taxon. Because there has been some controversy about the proper method for counting substitutions in the McDonald-Kreitman test (GRAUR and LI 1991 ; WHITTAM and NEI 1991 ), differences between lineages were counted in three different ways. For method 1, a site that is fixed in one lineage and polymorphic in another is classified as a polymorphism (MCDONALD and KREITMAN 1991 ). For method 2, a site fixed in one lineage and polymorphic for different substitutions in another is counted as a polymorphism and one fixed difference. This method attempts to cope with situations in which the observed polymorphism results from a single individual, most likely reflecting a multiple mutation after a fixed difference between the two highly diverged (>20%) F and M lineages has occurred. For method 3, branches on the phylogenetic tree were used to count mutations that appear on within-lineage branches and between-lineages branches (NACHMAN et al. 1994 ). These three methods resulted in similar patterns of association, as shown by a log linear analysis for each species (P > 0.85, for the three-way interactions). For the three methods considered, F/M interlineage comparisons reveal that Atlantic and Mediterranean M. galloprovincialis display a consistent pattern of excess replacement polymorphisms within mtDNA lineages over that predicted under a strictly neutral model of molecular evolution. The departure is always statistically significant for Atlantic M. galloprovincialis, and significant or marginally significant for Mediterranean M. galloprovincialis, depending on the method considered. By contrast, M. edulis shows good agreement with neutral expectations. The ratio of polymorphic to fixed differences is up to two times higher for nonsynonymous than for synonymous substitutions for the two M. galloprovincialis taxa for all three methods.

Additional McDonald-Kreitman tests can be performed for each genome by comparing the two major branches separating American M. trossulus and European taxa (Figure 1), taking advantage of the prediction that the ratio of polymorphic to fixed differences should be the same in each partition of a neutral genealogy (BALLARD and KREITMAN 1994 ). In each of these further tests involving F/F or M/M intralineage comparisons, the three counting methods produced similar results, thus only those for method 1 are given (Table 5). For this analysis, a significant excess of replacement polymorphism is observed within the F genome (P = 0.002), and inspection of the data reveals that all these polymorphisms occur in European populations. By contrast, a nonsignificant result is obtained for the M genome (P = 0.296). This contrasting pattern between genomes receives additional support from a log linear analysis (P = 0.008 for the three-way interaction). A similar excess of replacement polymorphism within the European F genome (P = 0.010), but not within the European M genome (P = 0.867), is observed if European M. edulis sequences alone are compared with American M. trossulus sequences. This indicates that M. edulis contributes to the contrasting pattern between genomes in F/F and M/M intralineage comparisons, despite the conformity of M. edulis to neutral expectations in F/M interlineage comparisons. These results imply either a stronger selection for adaptive mutations or a more relaxed selection for the F genome in European taxa with respect to American populations. Divergence in the genetic background, following the isolation of Amer-ican and European populations, could account for such differences in evolutionary factors operating on each mtDNA lineage.

Results of the TAJIMA 1989 and FU and LI 1993 tests are given in Table 1. Consistent results are obtained with both methods. There is an indication of an excess of rare nucleotide polymorphisms within F and M lineages, as revealed by the consistently negative D estimates. However, the only significant (P < 0.05) result is for the F genome of M. edulis at fourfold degenerate sites. In all three Mytilus taxa, the F genome displayed more highly negative D values, consistent with higher levels of purifying selection for this genome. Several observations support the hypothesis that the significant D value in M. edulis almost certainly results from the pooling of sequences from geographical regions differing in haplotype frequencies. First, the two M. edulis populations included in this study (Redcar and Mk. Point) are heterogeneous (P < 0.001) for haplotype frequencies, as revealed by a geographical survey based on RsaI and DdeI restriction patterns (C. GALLAGHER, unpublished results). Second, when the two M. edulis samples are analyzed separately, D values are substantially less negative in all cases (D < -0.10) and are not significant (P > 0.10). Third, the pooled M. edulis sample shows higher excesses of rare polymorphisms at fourfold than zerofold sites, but the opposite occurs in each single sample, in agreement with the pattern observed in the two other European taxa (Table 1).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX 1 LITERATURE CITED

Excess of replacement polymorphisms within European mtDNA lineages:
Sequence data for Mytilus mtDNA show a significant excess of replacement poly-morphisms in F/M interlineage comparisons for M. galloprovincialis and for the European F genome in intralineage comparisons between American M. trossulus and European taxa. Nonsignificant results are obtained in other comparisons. These results provide evidence for both neutral and nonneutral evolution of Mytilus mtDNA, suggesting different evolutionary histories for each taxon.

In the recent work by STEWART et al. 1996 on American M. trossulus and American M. edulis, rejection of the null hypothesis was attributed to an excess of replacement substitutions between the F and M lineages. Our data show a departure from neutrality in the opposite direction. Several questions need to be addressed: (1) What factors are responsible for the observed excess of replacement polymorphisms within mtDNA lineages? (2) What accounts for the reversal of the pattern in American mussels? (3) What are the origins of the different patterns observed among European taxa and among genomes in relation to the extent of departure from neutral expectations?

It appears unlikely that the observed patterns could be generated by nonequilibrium conditions, in spite of good evidence supporting recent variation in population size and mtDNA introgression among European taxa (QUESADA et al. 1995B , QUESADA et al. 1998 ). Such violation of equilibrium conditions might affect the validity of Tajima's and Fu and Li's tests, but not the McDonald-Kreitman test (MCDONALD and KREITMAN 1991 ; BROOKFIELD and SHARP 1994 ). On the other hand, mtDNA introgression should lead to significant excesses of polymorphism within European mtDNA lineages for both synonymous and nonsynonymous substitutions, not just for nonsynonymous substitutions, as observed here (Table 4 and Table 5). Similarly, mtDNA introgression should generate significant excesses of polymorphisms within F and M genomes in intralineage comparisons of European sequences with American M. trossulus, not just within the F genome, as observed (Table 5). F and M lineages exhibit very similar ratios of replacement to silent substitutions across the three European taxa (Table 3) consistent with similar levels of introgression across genomes (QUESADA et al. 1998 ). Conversely, the rejection of an equilibrium neutral model for M. galloprovincialis but not for M. edulis in F/M interlineage comparisons (Table 4) cannot be explained by introgression alone. Another possibility is that excess replacement polymorphism within European mtDNA lineages could be caused by some form of balancing selection. However, under this hypothesis, an excess of high-frequency polymorphism within European mtDNA lineages should be observed. The data are not consistent with this prediction because both Tajima's D and Fu and Li's D' show a consistent excess of low frequency polymorphisms.

A more likely explanation for excess replacement polymorphisms is that replacement variants are mildly deleterious and persist within mtDNA lineages as short-lived polymorphisms, but they do not persist long enough to become fixed between F and M lineages or between American and European populations. This explanation is supported by the dramatic decrease in F and M K_A/K_S ratios when the closely related European sequences are compared with the highly diverged American M. trossulus sequences (Table 3), suggesting that a substantial proportion of the replacement polymorphism found within European mtDNA lineages never become fixed. This nearly neutral argument has been used to explain a similar pattern of excess replacement polymorphism in the mtDNA of Drosophila melanogaster (KANEKO et al. 1993 ; Rand AND KANN 1996 ), mice (NACHMAN et al. 1994 ), and humans and African apes (NACHMAN et al. 1996 ). One prediction of the nearly neutral theory (OHTA 1992 ) is that because silent polymorphisms are more likely to be governed by neutral forces than replacement polymorphisms, there should be a lower variance in the frequency of replacement variants. Data for European taxa are consistent with this prediction (Table 6). Synonymous sites always show a higher variance in frequency than nonsynonymous sites, although this difference is only significant for the F genome. This result suggests higher levels of purifying selection for the F genome. This conclusion is reinforced by two further observations: (1) excess of low-frequency polymorphisms greater within the European F than the M genome revealed by Tajima's D and Fu and Li's D' estimates (Table 1) and (2) the lower K_A/K_S ratios for the F genome in American M. trossulus–European comparisons (Table 3). Other features of the data also fit the nearly neutral model. For both genomes, the nucleotide diversity is greater for synonymous than for nonsynonymous substitutions (Table 2), and excesses of rare substitutions are larger for zerofold than for fourfold degenerate sites in M. galloprovincialis (Table 1) and in individual M. edulis samples (see RESULTS).

Our data contrast with those of STEWART et al. 1996 , where an excess of replacement substitutions between F and M lineages in American mussels for the same COIII gene region was reported. Nevertheless, both could be explained under the nearly neutral model. In a nearly neutral scenario, the magnitude of negative selection coefficients acting on mtDNA vary according to the reciprocal of the effective population size (OHTA 1972 , OHTA 1992 ). Thus, repeated founder events or bottlenecks associated with glacial fluctuations could have led to a striking relaxation of selective constraints on slightly deleterious mutations in European mtDNA. This relaxation of constraints would have a more dramatic effect on the F lineage, given a higher level of purifying selection operating on this genome. This would explain the high and similar K_A/K_S ratios for both F and M lineages in European taxa and the significant excess of replacement polymorphisms within the European F genome with respect to American M. trossulus. In the long term, most slightly deleterious mutations would not persist, but the more relaxed selection for the M lineage would generate an excess of replacement substitutions between M and F lineages, as observed by STEWART et al. 1996 . In conclusion, the deviations from neutrality opposite in direction for European and American mussels might have more to do with the time scale of selection than with a difference in the nature of evolutionary forces operating on mtDNA.

Similarly, the different patterns among European taxa with respect to the extent of deviation from neutral expectations may be related to differences in effective population size and levels of constraint. mtDNA genetic diversity is negatively correlated with latitude in European Mytilus populations, suggesting larger effective population sizes and more deme-structured populations in southern taxa (QUESADA et al. 1995B ). Thus, replacement mutations that might be neutral or nearly neutral in small Northern taxa populations (M. edulis) could be deleterious in Southern taxa (Mediterranean M. galloprovincialis).

However, despite the apparent fit of the data to the nearly neutral model, it is difficult to exclude the alternative hypothesis of a recent relaxation of selection on mtDNA that has occurred, not through a reduction in effective population size, but through a reduction in the magnitude of selection coefficients acting on mtDNA. This would result in some previously deleterious mutations becoming neutral and remaining within mtDNA lineages as polymorphisms. Support for this explanation is provided by the increasing evidence of complex nuclear–mtDNA interactions that restrict the exchange of mtDNA between hybridizing American Mytilus taxa (RAWSON et al. 1996 ; SAAVEDRA et al. 1996 ), which contrasts clearly with the extensive levels of mtDNA introgression observed in European taxa (QUESADA et al. 1996b, QUESADA et al. 1998 ). It is difficult to account for the compatibility between a single mtDNA genome and very different nuclear backgrounds in highly introgressed European populations, unless such a relaxation of selection is operating. Assumptions that need to be made in relation to variation in effective population size (N_e) and selective coefficient (s) to make this explanation consistent with a nearly neutral model seem by contrast to be more unrealistic because a general criticism of such models is that they require a very narrow range of s and N_e to fit data (GILLESPIE 1991 , GILLESPIE 1994 ). Alternative models and combinations of models, both selective and neutral, e.g., GILLESPIE 1994, may better account for the data as a general explanation of similar excesses of mtDNA replacement polymorphism in taxa as divergent as Drosophila, humans, and mice.

Differential evolutionary rates of male and female mtDNA lineages:
This study demonstrates a higher diversity for the M than the F genome, and that the M lineage evolves faster than the F lineage in European populations of M. edulis and M. galloprovincialis. These results are consistent with earlier reports in Mytilus using both RFLP (SKIBINSKI et al. 1994B ; QUESADA et al. 1996 , QUESADA et al. 1998 ) and sequence data (RAWSON and HILBISH 1995 ; STEWART et al. 1995 , STEWART et al. 1996 ), and thus support the universality of this difference between genomes. What factors might be responsible? Our data suggest a combination of relaxed constraint and an increased mutation rate for the M genome.

In the earlier work of STEWART et al. 1996 , a rate of replacement substitution significantly higher for the M than the F genome, but no significant difference for the synonymous rate, was reported. It was suggested that the higher rate for the M genome could thus only be the result of relaxed constraint for replacement substitutions; a role for differences in mutation rate was disregarded. In the present study, substantial differences in both the synonymous rate and, to a greater extent, the nonsynonymous rate are observed between genomes in American–European comparisons (Table 3). This result is clearly supportive of a higher mutation rate for the M genome, although it can also be explained under a nearly neutral model if the effective population size for the M genome is smaller, as suggested by STEWART et al. 1996 . However, levels of diversity lower for the M than the F genome are expected under the assumption of a smaller effective population size for the M genome (KIMURA 1983 ), but the opposite is observed (Table 2).

If similar levels of constraint were operating on both genomes, then differences in the synonymous but not in the nonsynonymous rate would reflect primarily differences in mutation rate. The striking and similar relaxation of constraints for both F and M genomes in European taxa, leading to no substantial differences of constraint between genomes, provides a test of this prediction. First, nucleotide diversity is significantly higher for the M than the F genome for synonymous substitutions, but not for nonsynonymous substitutions (Table 2). Second, the ratios of nonsynonymous to synonymous substitutions within the F and M lineages are very similar in European taxa (Table 3), but this ratio was always smaller for the M genome because of an increased synonymous rate. Third, no differences are observed between the F and M genomes in the number of sites showing replacement substitutions, but both M. edulis and Mediterranean M. galloprovincialis show a significant (Monte Carlo test, P < 0.01) increase in the number of sites with synonymous substitutions for the M genome (calculated from the data of Table 6).

Several alternative hypotheses fail to account for the observed differences in diversity and divergence between genomes. Genome-specific population bottlenecks in the recent history of the species could reduce variability of the F genome, but they should increase sequence divergence, as diversity becomes converted to divergence. However, this pattern is the opposite of that observed in the data. Hitchhiking on the F genome associated with a selective sweep could cause the observed reduction in diversity, as reported for Drosophila mtDNA (BALLARD and KREITMAN 1994 ; Rand et al. 1994 ). However, the hitchhiking effect predicts no particular change in the rate of substitution of unselected mutants (BIRKY and WALSH 1988 ; KAPLAN et al. 1989 ) and, hence, cannot explain the lower divergence for the F genome. Another possibility is that hybridization and introgression between European Mytilus taxa (GOSLING 1992 ; QUESADA et al. 1995B ) have caused higher polymorphism for the M than the F genome because of more relaxed taxonomic boundaries for the M genome. However, this should lead to lower sequence divergence among taxa for the M genome, the opposite of that observed.

In conclusion, the results presented here suggest that relaxed selection coupled with a higher mutation rate enhance each other's effect to generate a higher evolutionary rate for the M lineage. A higher level of oxidative damage in sperm mtDNA (see SKIBINSKI et al. 1994B ; QUESADA et al. 1996 ) might cause the reported higher mutation rate for the M genome, as might an increase in the number of replications in the male germ line of mussels (STEWART et al. 1995 ; RAWSON and HILBISH 1995 ). The increase in mutation rate for the M genome might be small, but the accumulated effect could be very dramatic on lineages that are evolving independently over a long period (RAWSON and HILBISH 1995 ; STEWART et al. 1995 ; QUESADA et al. 1996 ). In fact, the underlying differences among genomes in K_A and K_S are statistically significant only in comparisons including the highly diverged M. trossulsus sequences. How general is the conclusion of a higher mutation rate for the M genome remains to be seen. It is possible that the contrary results of STEWART et al. 1996 could result from a difference in mutation rate stemming from different American and European thermal habits, e.g., RAND 1994. However, it is also possible that divergence of genetic background after the isolation between American and European populations, or differences of statistical power between both studies, given the higher number of sequences analyzed here, could explain these contrasting results.

In summary, many factors are affecting the levels of mtDNA polymorphism in Mytilus, from rates of molecular evolution to population history. Depending on the particular interaction of these factors, it appears that levels of mtDNA polymorphism within and between species deviate from a neutral equilibrium model, and that natural selection is in part responsible for the overall patterns of nucleotide variation. The data presented here suggest that the balance between mutation and selection may be very subtle, and that different histories of adaptive and neutral evolution can lead to complex differences between populations. However, the excesses of replacement polymorphisms reported here in mussels are similar to those observed in species as divergent as mice, Drosophila, and humans, raising the possibility of a general mechanism governing mtDNA evolution.

	ACKNOWLEDGMENTS

We thank the Natural Environmental Research Council of the UK for financial support. We are grateful to A. QUESADA and F. RODRÍGUEZ for help in the collection of mussels, and to K. A. NAISH for introducing us to automatic sequencing. H.Q. was supported by a postdoctoral fellowship from the Ministerio de Educación y Ciencia (Spain). Sequences have been deposited in GenBank, accession numbers AF063251, AF063252, AF063253, AF063254, AF063255, AF063256, AF063257, AF063258, AF063259, AF063260, AF063261, AF063262, AF063263, AF063264, AF063265, AF063266, AF063267, AF063268, AF063269, AF063270, AF063271, AF063272, AF063273, AF063274, AF063275, AF063276, AF063277, AF063278, AF063279, AF063280, AF063281, AF063282, AF063283, AF063284, AF063285, AF063286, AF063287, AF063288, AF063289, AF063290, AF063291, AF063292, AF063293, AF063294, AF063295, AF063296, AF063297, AF063298, AF063299, AF063300.

Manuscript received November 11, 1997; Accepted for publication March 11, 1998.

	APPENDIX 1

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX 1 LITERATURE CITED

View larger version (187K): In this window In a new window Download PPT slide   Figure 2. Variable nucleotides from F and M COIII mtDNA in three Mytilus taxa R, replacement; S, synonymous; SM, synonymous within M lineage. F sequences appear in regular type; M sequences appear in bold type.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX 1 LITERATURE CITED

BALLARD, J. W. O. and M. KREITMAN, 1994  Unravelling selection in the mitochondrial genome of Drosophila.. Genetics 138:757-772[Abstract].

BALLARD, J. W. O. and M. KREITMAN, 1995  Is mitochondrial DNA a strictly neutral marker? TREE 10:485-488.

BIRKY, C. W., JR. and J. B. WALSH, 1988  Effects of linkage on rates of molecular evolution. Proc. Natl. Acad. Sci. USA 85:6414-6418[Abstract/Free Full Text].

BROOKFIELD, J. F. Y. and P. M. SHARP, 1994  Neutralism and selectionism face up to DNA data. Trends Genet. 10:109-111[Medline].

COMERON, J. M., 1995  A method for estimating the numbers of synonymous and nonsynonymous substitutions per site. J. Mol. Evol. 41:1152-1159[Medline].

FELSENSTEIN, J., 1993 PHYLIP: phylogeny inference package. Version 3.1. University of Washington, Seattle.

FISHER, C. and D. O. F. SKIBINSKI, 1990  Sex-biased mitochondrial DNA heteroplasmy in the marine mussel Mytilus.. Proc. R. Soc. Lond. B 242:149-156.

FU, Y. X., 1994  A phylogenetic estimator of effective population size or mutation rate. Genetics 136:685-692[Abstract].

FU, Y. X. and W. H. LI, 1993  Statistical tests of neutrality of mutations. Genetics 133:693-709[Abstract].

GILLESPIE, J. H., 1991 The Causes of Molecular Evolution. Oxford University Press, Oxford.

GILLESPIE, J. H., 1994  Substitution processes in molecular evolution. II. Exchangeable models from population genetics. Evolution 48:1101-1113.

GOSLING, E. M., 1992 Systematics and geographical distribution of Mytilus, pp. 1–20 in The Mussel Mytilus: Ecology, Physiology, Genetics and Culture, edited by E. M. GOSLING. Elsevier, Amsterdam.

GRAUR, D. and W. H. LI, 1991  Neutral mutation hypothesis test. Nature 354:114-115.

HIGGINS, D. G., J. D. THOMPSON, and T. J. GIBSON, 1996  Using Clustal for multiple sequence alignments. Methods Enzymol. 266:383-402[Medline].

HOEH, W. R., K. H. BLAKLEY, and W. M. BROWN, 1991  Heteroplasmy suggests limited biparental inheritance of Mytilus mitochondrial DNA. Science 251:1488-1498[Abstract/Free Full Text].

HOEH, W. R., D. T. STEWART, B. W. SUTHERLAND, and E. ZOUROS, 1996  Cytochrome c oxidase sequence comparisons suggest an unusually high rate of mitochondrial DNA evolution in Mytilus (Mollusca: Bivalvia). Mol. Biol. Evol. 13:418-421[Medline].

HOFFMANN, R. J., J. L. BOORE, and W. M. BROWN, 1992  A novel mitochondrial genome organization for the blue mussel Mytilus edulis.. Genetics 131:397-412[Abstract].

KANEKO, M., Y. SATTA, E. T. MATSURA, and S. CHIGUSA, 1993  Evolution of the mitochondrial ATPase 6 gene in Drosophila: unusually high level of polymorphism in D. melanogaster.. Genet. Res. 61:195-204[Medline].

KAPLAN, N. L., R. R. HUDSON, and C. H. LANGLEY, 1989  The "hitchhiking effect" revisited. Genetics 123:887-899[Abstract/Free Full Text].

KIMURA, M., 1983 The Neutral Theory of Molecular Evolution, Cambridge University Press, Cambridge.

KUMAR, K., K. TAMURA and M. NEI, 1993 MEGA: Molecular Evolutionary Genetics Analysis. Version 1.02. Pennsylvania State University, University Park, PA.

MCDONALD, J. H. and M. KREITMAN, 1991  Adaptive protein evolution at the Adh locus in Drosophila.. Nature 351:652-654[Medline].

NACHMAN, M. W., S. N. BOYER, and C. F. AQUADRO, 1994  Nonneutral evolution at the mitochondrial NADH dehydrogenase subunit 3 gene in mice. Proc. Natl. Acad. Sci. USA 91:6364-6368[Abstract/Free Full Text].

NACHMAN, M. W., W. M. BROWN, M. STONEKING, and C. F. AQUADRO, 1996  Nonneutral mitochondrial DNA variation in humans and chimpanzees. Genetics 142:953-963[Abstract].

OHTA, T., 1972  Population size and rate of evolution. J. Mol. Evol. 1:305-314[Medline].

OHTA, T., 1992  The nearly neutral theory of molecular evolution. Annu. Rev. Ecol. Syst. 23:263-286.

QUESADA, H., 1993 Estructura genética de poblaciones naturales del mejillón Mytilus galloprovincialis Lmk. Servicio de Publicaciones de la Universidad de Santiago de Compostela, Santiago de Compostela, Spain.

QUESADA, H., C. M. BEYNON, and D. O. F. SKIBINSKI, 1995a  A mitochondrial DNA discontinuity in the mussel Mytilus galloprovincialis Lmk: pleistocene variance biogeography and secondary intergradation. Mol. Biol. Evol. 12:521-524[Medline].

QUESADA, H., R. WENNE, and D. O. F. SKIBINSKI, 1995b  Differential introgression of mitochondrial DNA across species boundaries within the marine mussel genus Mytilus.. Proc. R. Soc. Lond. B 262:51-56.

QUESADA, H., C. ZAPATA, and G. ALVAREZ, 1995c  A multilocus allozyme discontinuity in the mussel Mytilus galloprovincialis: the interaction of ecological and life-history factors. Mar. Ecol. Prog. Ser. 116:99-115.

QUESADA, H., D. A. G. SKIBINSKI, and D. O. F. SKIBINSKI, 1996  Sex-biased heteroplasmy and mitochondrial DNA inheritance in the mussel Mytilus galloprovincialis Lmk. Curr. Genet. 29:423-426[Medline].

QUESADA, H., C. GALLAGHER, D. A. G. SKIBINSKI, and D. O. F. SKIBINSKI, 1998  Patterns of polymorphism and gene flow of gender-associated mitochondrial DNA lineages in European mussel populations. Mol. Ecol. in press.

RAND, D. M., 1994  Thermal habit, metabolic rate and the evolution of mitochondrial DNA. TREE 9:125-131.

RAND, D. M., M. DORFSMAN, and L. M. KANN, 1994  Neutral and nonneutral evolution of Drospholia mitochondrial DNA. Genetics 138:741-756[Abstract].

RAND, D. M. and L. M. KANN, 1996  Excess amino acid polymorphism in mitochondrial DNA: contrast among genes from Drosophila, mice and humans. Mol. Biol. Evol. 13:735-748[Abstract].

RAWSON, P. D. and T. J. HILBISH, 1995  Evolutionary relationships among the male and female mitochondrial DNA lineages in the Mytilus edulis species complex. Mol. Biol. Evol. 12:893-901[Abstract].

RAWSON, P. D., C. L. SECOR, and T. J. HILBISH, 1996  The effects of natural hybridization on the regulation of doubly uniparental mtDNA inheritance in blue mussels (Mytilus spp.). Genetics 144:241-248[Abstract].

ROFF, D. A. and P. BENTZEN, 1989  The statistical analysis of mitochondrial DNA polymorphisms: X² and the problem of small samples. Mol. Biol. Evol. 6:539-545[Abstract].

ROZAS, J. and R. ROZAS, 1997  DnaSP version 2.0: a novel software package for extensive molecular population genetics analysis. Comput. Appl. Biosci. 13:307-311.

SAAVEDRA, C., D. T. STEWART, R. R. STANWOOD, and E. ZOUROS, 1996  Species-specific segregation of gender-associated mitochondrial DNA types in an area where two mussel species (Mytilus edulis and Mytilus trossulus) hybridize. Genetics 143:1359-1367[Abstract].

SKIBINSKI, D. O. F., C. GALLAGHER, and C. M. BEYNON, 1994a  Mitochondrial DNA inheritance. Nature 368:817-818[Medline].

SKIBINSKI, D. O. F., C. GALLAGHER, and C. M. BEYNON, 1994b  Sex-limited mitochondrial DNA transmission in the marine mussel Mytilus edulis.. Genetics 138:801-809[Abstract].

STEWART, D. T., C. SAAVEDRA, R. R. STANWOOD, A. O. BALL, and E. ZOUROS, 1995  Male and female mitochondrial DNA lineages in the blue mussel (Mytilus edulis) species group. Mol. Biol. Evol. 12:735-747[Abstract].

STEWART, D. T., E. R. KENCHINGTON, R. K. SINGH, and E. ZOUROS, 1996  Degree of selective constraint as an explanation of the different rates of evolution of gender-specific mitochondrial DNA lineages in the mussel Mytilus.. Genetics 143:1349-1357[Abstract].

TAJIMA, F., 1989  Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123:585-595[Abstract/Free Full Text].

WHITTAM, T. S. and M. NEI, 1991  Neutral mutation hypothesis test. Nature 354:115-116.

ZOUROS, E., K. R. FREEMAN, A. O. BALL, and G. H. POGSON, 1992  Direct evidence for extensive paternal mitochondrial DNA inheritance in the marine mussel Mytilus.. Nature 359:412-414[Medline].

ZOUROS, E., A. O. BALL, C. SAAVEDRA, and K. R. FREEMAN, 1994a  Mitochondrial DNA inheritance. Nature 368:817-818.

ZOUROS, E., A. O. BALL, C. SAAVEDRA, and K. R. FREEMAN, 1994b  An unusual type of mitochondrial DNA inheritance in the blue mussel Mytilus.. Proc. Natl. Acad. Sci. USA 91:7463-7467[Abstract/Free Full Text].