Evolution of Interacting Proteins in the Mitochondrial Electron Transport System in a Marine Copepod

The extensive interaction between mitochondrial-encoded and nuclear-encoded subunits of electron transport system (ETS) enzymes in mitochondria is expected to lead to intergenomic coadaptation. Whether this coadaptation results from adaptation to the environment or from fixation of deleterious mtDNA mutations followed by compensatory nuclear gene evolution is unknown. The intertidal copepod Tigriopus californicus shows extreme divergence in mtDNA sequence and provides an excellent model system for study of intergenomic coadaptation. Here, we examine genes encoding subunits of complex III of the ETS, including the mtDNA-encoded cytochrome b (CYTB), the nuclear-encoded rieske iron-sulfur protein (RISP), and cytochrome c1 (CYC1). We compare levels of polymorphism within populations and divergence between populations in these genes to begin to untangle the selective forces that have shaped evolution in these genes. CYTB displays dramatic divergence between populations, but sequence analysis shows no evidence for positive selection driving this divergence. CYC1 and RISP have lower levels of sequence divergence between populations than CYTB, but, again, sequence analysis gives no evidence for positive selection acting on them. However, an examination of variation at cytochrome c (CYC), a nuclear-encoded protein that transfers electrons between complex III and complex IV provides evidence for selective divergence. Hence, it appears that rapid evolution in mitochondrial-encoded subunits is not always associated with rapid divergence in interacting subunits (CYC1 and RISP), but can be in some cases (CYC). Finally, a comparison of nuclear-encoded and mitochondrial-encoded genes from T. californicus suggests that substitution rates in the mitochondrial-encoded genes are dramatically increased relative to nuclear genes.


Introduction
The 13 proteins encoded by mitochondrial DNA (mtDNA) in animals are subunits of the electron transport system (ETS) and, together with some 65 nuclear-encoded subunits, comprise the oligomeric enzyme complexes of the ETS.Given the close association between nuclearencoded and mitochondrial-encoded proteins in these complexes and the importance of the ETS for energy production, coadaptation between nuclear-encoded and mitochondrial-encoded proteins has been strongly selected during the evolutionary history of these complexes.Two different scenarios of selection/drift could generate intergenomic coadaptation: (1) Selection pressure on ETS enzymes could be generated by the adaptation of a species (or population) to its environment (i.e., thermal environment for ectotherms) leading to changes in both nuclearencoded and mitochondrial-encoded subunits of ETS complexes.(2) High mutation rates in mtDNA, especially when coupled with fluctuating population sizes, may lead to fixation of deleterious amino acid substitutions and provide selection pressure for subsequent compensatory changes in nuclear genes, a two-step process that could drive evolution of coadaptation in ETS enzymes.
In mammals, several lines of evidence suggest that the interactions involving the nuclear-encoded soluble cytochrome c (CYC) and cytochrome c oxidase (complex IV) have evolved rapidly in higher primates: (1) In vitro studies with heterospecific combinations of CYC and complex IV enzymes across these groups show reduced enzyme activity (Osheroff et al. 1983).Additionally, analyses of human/orangutan cell lines containing portions of both species nuclear genomes and orangutan mtDNA show reduced complex IV enzyme activity consistent with significant mtDNA/nuclear coadaptation (Barrientos et al. 2000).( 2) Sequence analyses show that rates of amino acid substitution increase in simian primates in mtDNAencoded subunits of complex IV (COI and COII) and in seven other nuclear-encoded subunits in concert with increases in substitution in CYC (Adkins, Honeycutt, and Disotell 1996;Wu et al. 2000;Schmidt, Goodman, and Grossman 2002).( 3) At the structural level, rates of amino acid evolution are increased at interaction points of proteins for mtDNA-encoded proteins but not nuclearencoded proteins (Schmidt et al. 2001).( 4) Analyses of mtDNA-encoded cytochrome b (CYTB), nuclear-encoded cytochrome c 1 (CYC1), and rieske iron-sulfur protein (RISP) from ubiquinol-cytochrome c reductase (complex III) also show accelerated substitution in humans/simian primates correlated with the changes in complex IV and CYC and may argue for extensive coadaptation of these two ETS complexes and CYC in higher primates (Andrews, Jermiin, and Easteal 1998;Grossman et al. 2001).Given the role of CYC as a carrier of electrons between complex III and complex IV, this coadaptation has an obvious potential functional basis, although the evolutionary forces driving this divergence in higher primates are not known.
Because of high levels of interpopulation genetic divergence, the intertidal copepod Tigriopus californicus presents an excellent model system for the study of evolutionary coadaptation between nuclear and mitochondrial genes.Allopatric populations of T. californicus along the western coast of North America show extreme divergence in mtDNA genes, sometimes over relatively short physical distances.Two mtDNA-encoded subunits of complex IV, COI, and COII, display divergences above 20% for nucleotide substitutions and up to 15% for amino acid divergence (Burton and Lee 1994;Burton 1998;Burton, Rawson, and Edmands 1999).The nuclear-encoded protein CYC also shows substantial amino acid divergence between populations, particularly given the generally high level of conservation of this protein (Rawson, Brazeau, and Burton 2000).Despite high interpopulation divergence, within-population polymorphism is generally low; the resulting high F ST values suggest highly restricted gene flow.Clearly these populations have evolved independently for long periods of time and could be considered separate species under some species definitions (i.e., phylogenetic species concept); however, based on their reproductive compatibility, we will continue to refer to them as conspecific populations (also see Burton [1998]).
Several studies suggest that nuclear/mitochondrial coadaptation is evolving between genetically isolated T. californicus populations.Edmands and Burton (1999) used repeated backcrossing to study the effect of cytonuclear interactions on cytochrome c oxidase (complex IV) enzyme activity in hybrids.Reductions in complex IV activity in some interpopulation crosses were caused at least in part by nuclear/mitochondrial interactions.Rawson and Burton (2002) found direct evidence for functional coadaptation between CYC and complex IV.CYC derived from a San Diego (SD) population yielded significantly higher activity with SD population-derived complex IV than that derived from a Santa Cruz (SC) population, and reciprocally, SC population-derived CYC functioned best with SC population-derived complex IV.Willett and Burton (2001) found extensive differences in CYC genotypic viabilities in F 2 hybrids from controlled crosses between populations (estimated from deviations from Mendelian ratios).CYC viability differences were observed in reciprocal crosses between specific populations, suggesting cytoplasmic effects consistent with the inference of nuclear/mitochondrial interactions.
The combination of extensive divergence in mtDNAencoded genes in T. californicus and fitness/functional evidence for mitochondrial/nuclear coadaptation make this system a unique opportunity to examine molecular evolution of these interacting genes.In this paper, we will first examine the evolution of CYTB (the mtDNA-encoded subunit of complex III) and show that it is diverging rapidly between populations.Two nuclear-encoded subunits of complex III have functional domains containing redox centers, the rieske iron-sulfur protein (RISP), and cytochrome c 1 (CYC1).We have obtained nuclear DNA sequences encoding these proteins from T. californicus and examine polymorphism within populations and divergence between populations.We then compare these genes, which function in the ETS, with other nuclear genes that have been obtained from T. californicus.Two major conclusions can be drawn from these results: (1) Proteins closely interacting with mtDNA-encoded proteins in the ETS can show evidence for selection for high rates of evolution (CYC), but not all appear to be evolving rapidly (CYC1 and RISP).(2) The mtDNA in T. californicus appears to have a much higher rate of substitution between populations than has the average nuclear gene for both synonymous and nonsynonymous changes.

Cytochrome b Sequencing
To study divergence in mitochondrial and nuclear encoded genes, copepods were collected from high intertidal rock pools at three sites in California, San Diego (SD, 328459N, 1178159W), Abalone Cove in Los Angeles County (AB, 338449N, 1188199W), and Santa Cruz (SC, 368579N, 1228039W).We obtained CYTB sequences from SD, SC, and AB copepods to test the extent of population divergence relative to other mitochondrial genes in T. californicus.Primers for SD population CYTB were designed from the sequence of the complete mtDNA molecule from SD (Rawson and Burton, unpublished data) and amplified the entire gene.PCR products were directly sequenced using Amersham Biosciences MegaBACE ET Terminators chemistry and run on a MegaBACE 500 sequencer following manufacturer's protocols.SD CYTB primers did not amplify the entire gene from either AB or SC populations (only smaller fragments were successfully amplified).We designed degenerate primers from comparisons of arthropod mtDNA sequences for both ND4 and COXII (genes flanking CYTB in T. californicus mtDNA) and within CYTB itself.Sequence obtained from this strategy allowed us to design population-specific primers that flanked CYTB and permitted us to amplify and sequence this gene from both SC and AB population copepods.Primer sequences and amplification protocols are available upon request from authors.DNA was extracted from single copepods using the proteinase-K cell-lysis method (Hoelzel and Green 1992).Direct sequencing of CYTB was performed from PCR products for multiple individuals from each population.

Identification and Sequencing of Nuclear-Encoded Proteins
We sought to characterize two nuclear-encoded genes from complex III of the ETS that contains the mtDNAencoded protein CYTB.The T. californicus nuclear genes encoding the proteins CYC1 and RISP were initially identified using a 59 RACE (rapid amplification of cDNA ends) procedure on mRNA isolated from SD copepods.The kit Generacer (Invitrogen, Carlsbad, Calif.) was used with degenerate primers designed to match conserved regions of these two proteins (from alignments of human, Drosophila melanogaster, Saccharomyces cerevisiae, and Caenorhabditis elegans) to amplify from cDNA.Primer used for CYC1 was CYC1.rev (59-GCNCCNCCNGGRAARTANGGR-TT-39) and primer used for RISP was RISP.rev (59-TAR-TGNSWNCCRTGRCANGGRCA-39) in conjunction with a Generacer kit primer.Resulting PCR products were gel purified, cloned, and sequenced.The complete mRNA was obtained with 39 RACE using T. californicus-specific primers designed from the new sequence.
Genomic DNA sequences were obtained from the SD population corresponding to the mRNA transcript for RISP and CYC1.PCR products were generated using speciesspecific primers and these products were sequenced directly.Polymorphism was inferred for sites when two clear peaks of equal intensity were observed in the chromatogram.Sequences for most of the coding region and intervening introns were obtained from multiple individuals from SD, SC, and AB individuals (some 59 end and 39 end codons were not sequenced).CYC sequences were published previously based on sequences obtained from clones (Rawson, Brazeau, and Burton 2000), and polymorphisms in these sequences were verified by examining directly sequenced products from same individuals as described above.Sequences were edited with the program Sequencer version 4.1 (Genecodes, Ann Arbor, Mich.), and alignments were made using ClustalX with some adjustments by hand in large indels in introns.DNAsp version 3.53 (Rozas and Rozas 1999) was used to calculate p values, M-K tests, and numbers of fixed silent and replacement differences.PAUP* version 4.0b10 (Swofford 2001) was used to calculate amino acid identities between proteins and for phylogenetic analyses.
We also sought to characterize polymorphism and divergence in nuclear-encoded genes that are not a part of the ETS and not likely to interact closely with mtDNAencoded proteins.Additional sequences of the gene glutamate dehydrogenase (GDH) were obtained from SD, SC, and AB population copepods.A complete genomic sequence of GDH from SD was recently obtained (Willett and Burton 2003) and a region of 1,320 bp, including a single intron of 71 bp, was sequenced from multiple individuals in each population.Sequences were obtained by direct sequencing from PCR products.Population samples from two other nuclear encoded proteins, Á 1pyrroline-5-carboxylase synthase (P5CS) and Á 1 -pyrroline-5-carboxylase reductase (P5CR), were previously published (Willett and Burton 2002).To verify heterozygous sites, sequencing traces were visually inspected for polymorphism as discussed above.Alignments of sequences are available as Supplementary Material online (http:// www.mbe.oupjournals.org)and representative sequences from each population have been submitted to GenBank with accession numbers AY344446 to AY344470.

Patterns of CYTB Polymorphism and Divergence
We have examined three levels of CYTB variation for T. californicus, between species, within species, and within populations.The T. californicus CYTB protein is predicted to be 376 amino acids in length.The initiation codon appears to be TAA (based on alignments with other species CYTBs), which is reported to be the initiation codon for CYTB and four other mtDNA genes in a congener T. japonicus (Machida et al. 2002).Not surprisingly, the most similar CYTB among diverse eukaryotes is that of T. japonicus at 77% amino acid identity (table 1).CYTB amino acid sequence is not very informative phylogenetically, because few relationships are robustly supported in phylogenetic trees constructed using parsimony analysis of amino acids of largely arthropod CYTB proteins.For example, the clade Crustacea is not recovered in these trees (fig.1).The well-supported Tigriopus clade has comparatively long branches, suggesting rapid amino acid evolution in this group.This impression is supported by relative rates comparisons with other arthropods: T. californicus is 2.03fold faster when compared with Daphnia pulex, 2.7-fold -Amino acid identities are calculated for the entire CYTB protein.For both CYC1 and RISP proteins mature proteins, not including mitochondrial signal peptide sequence, were compared (cleavage site for human protein was used to determine approximate N-terminal end of mature proteins).Drosophila melanogaster proteins are CYTB-P18935, two homologous gene products for CYC1-AAF50785/AAL90066, and gene product homologous to RISP-AAF51354.Anonopheles gambiae proteins are CYTB-P34844, gene product homologous to CYC1-EAA04119, and gene product homologous to RISP-EAA14787.Human proteins are CYTB-P00156, CYC1-P08574, and RISP-P47985.Saccharomyces cerevisiae proteins are CYTB-P00163, CYC1-CCBY1H, and RISP-P08067.Tigriopus japonicus CYTB protein is BAB97221.
FIG. 1.-Phylogram for CYTB proteins of diverse arthropods, including T. californicus.This tree was generated from a parsimony analysis of CYTB complete amino acid sequences and was one of three most-parsimonious trees.This phylogram of a single tree is shown, rather than the consensus, to show branch lengths indicating rapid evolution in the Tigriopus clade that is common to all trees.Bremmer support indices (above branch) and bootstrap values from 1,000 replicates (below branch) are given for branches found in the consensus tree.GenBank accession numbers are listed before species names.T. californicus is denoted as Tcal with population designation (SD, SCN, or AB).Crustaceans are indicated with an asterisk (*).Outgroups include human, Xenopus, and the earthworm (Lumbricus terrestris) CYTB.

Molecular Evolution of ETS Proteins 445
Downloaded from https://academic.oup.com/mbe/article-abstract/21/3/443/1079533 by guest on 22 November 2018 higher when compared with Penaeus monodon, and 2.9fold higher when compared with Triops cancriformis (all comparisons use D. melanogaster as an outgroup).In contrast, when T. californicus and T. japonicus are compared, they show equal rates of evolution with either D. pulex or D. melanogaster used as outgroups (T.californicus 1.02fold and 1.05-fold higher than T. japonicus, respectively).
CYTB is remarkably divergent between the AB, SC, and SD populations, with pairwise amino acid differences averaging 6.9% to 7.4% (AB to SD, and AB and SD to SC, respectively).This translates into 26 and 28 amino acid differences in the CYTB protein for these interpopulation comparisons.Divergence was even more extreme at silent sites, with uncorrected pairwise divergences between 65% and 70% (table 2).A Jukes-Cantor correction for multiple hits suggests that the average silent site has undergone at least one change in the evolution of CYTB between populations.D n /D s values are low for CYTB, reflecting the high number of silent substitutions that have occurred.This large amount of genetic divergence between T. californicus populations is not unexpected, given previous results for mtDNA genes in this species; levels of nucleotide divergence for nonsynonymous sites in CYTB fall between those seen for the mitochondrial-encoded COI and COII of complex IV for comparisons between the same populations (table 2).Comparisons of these three genes across members of the Drosophila melanogaster subgroup of Drosophila show the same relative ranks of amino acid conservation with CYTB showing an intermediate rate of amino acid evolution (Ballard 2000a).
In comparison with the remarkable levels of divergence in CYTB between populations, levels of poly-morphism within populations are relatively normal.Between 11 and 20 alleles for the complete CYTB gene were sequenced from the AB, SC, and SD populations, and levels of polymorphism were calculated within each population (table 3).Average pairwise divergence between alleles within a population was about 0.25% for each of the three populations.These p values for CYTB are similar to those reported from a range of animal studies, including armadillos, cod, sardines, and shrews whose p values ranged from 0.3% to 0.53% for a single geographical population (Grant, Clark, and Bowen 1998;Arnason et al. 2000;Frutos and Van Den Bussche 2002;Ratkiewicz et al. 2002).The frequency distribution of alleles differed between T. californicus populations as indicated by Fu and Li's D test (Fu and Li 1993) and Tajima's D statistic (Tajima 1989), which were generally positive in the AB and SC populations and negative in the SD population.A positive value indicates excess high frequency alleles, whereas negative values indicate excess low frequency alleles (in the SD population, there were seven haplotypes, six of which were only represented by one individual).Different population histories or selection on the mtDNA are potential explanations for the differences we observe in haplotype frequency distributions among these populations.
Evidence is accumulating that animal mtDNA polymorphism includes an overrepresentation of slightly deleterious mutations (Rand 2001); this phenomenon is seen in many data sets as excess of amino acid polymorphism and can be detected by McDonald/Kreitman (MK) tests (McDonald and Kreitman 1991;Rand and Kann 1996;Nachman et al. 1996;Weinreich and Rand 2000).Neutrality index (NI) values can be used to measure   4).When JC corrected values are used for divergence, the AB/SC comparison is significant in an MK test consistent with elevated replacement polymorphism.Therefore, we find no consistent evidence for positive selection driving amino acid substitution between copepod populations in CYTB and some evidence that polymorphism could contain excess slightly deleterious mutations analogous to that seen in the evolution of other animal mtDNAs.

Divergence Between T. californicus Populations in Nuclear Genes
Given the extensive amino acid divergence seen in CYTB between populations, it is interesting to ask whether proteins that interact closely with CYTB also show large amino acid divergences between populations.Two nuclear genes that encode protein subunits of complex III and contain the other redox active sites in this complex are CYC1 and RISP.We isolated the genes encoding these proteins from T. californicus and found that inferred amino acid sequences show high similarity to CYC1 and RISP proteins of other species (table 1).For most comparisons between the taxa in table 1, the RISP protein appears to be the most highly conserved, with CYC1 being more conserved than CYTB.Similar to the results seen for CYTB, phylogenetic analyses on CYC1 and RISP amino acid alignments do not produce robust and well-resolved phylogenies (data not shown).Rates of evolution between these taxa are more difficult to compare than for CYTB with relative rates comparisons because there are very few sequences of these proteins in arthropods currently available.Comparing the rate for T. californicus to D. melanogaster CYC1 (AF50785), with human as the outgroup, suggests the rate of amino acid evolution is 0.93-fold slower in T. californicus lineage.The same set of comparisons for RISP suggests T. californicus lineage is evolving 1.21-fold faster.Finally, a comparison of soluble CYC from T. californicus and the prawn (Macrobrachium malcolmsoni) with a D. melanogaster outgroup suggests T. californicus CYC is evolving slightly faster (1.2-fold higher).
We obtained sequences of CYC1 and RISP from three T. californicus populations, SD, AB, and SC.The intron/ exon structure of these genes was determined by comparing mRNA and genomic sequence for SD, and it is largely conserved between populations (fig.2).A potential exception to this is intron 1 of CYC1 in the AB and SC populations, where the 39 intron-ending AG in SD has been changed to GG, and the next closest potential splice AG is 5 bp closer to the predicted start codon (this predicted boundary has not been verified by mRNA sequencing from SC and AB copepods).The four small introns in the two genes vary little in size between populations, whereas the two larger introns vary considerably (fig.2).Based on the current annotations of the A. gambiea and D. melanogaster CYC1 and RISP genes, there appears to be little conservation in intron position among these two Diptera and T. californicus.However, unlike the other introns, the first intron in RISP does occur in approximately the same position in all three taxa (in the unalignable putative mitochondrial signaling peptide).
The complete coding region for CYC1 and most of that of RISP were sequenced from individual T. californicus copepods from each of the SD, SC, and AB populations (sequenced regions indicated in figure 2).Divergence in these genes between populations, although substantial, was considerably lower than that of the mtDNA genes (table 5).The nuclear-encoded CYC has extensive interactions with both complex III and IV and was previously sequenced from these populations (Rawson, Brazeau, and Burton 2000).Divergences between these populations for CYC are also presented in table 5.For these three nuclear-encoded ETS genes, the average rate of nonsynonymous substitution compared with the average for three mtDNA genes was 5.8fold lower for the SD/SC comparison, sixfold lower for the AB/SC comparison, and 9.7-fold lower for the AB/SD comparison.Insertion and deletion variation was common in noncoding regions of these three genes between these populations but absent within coding regions.Nucleotide divergence between populations was also higher in general  for these noncoding regions than for synonymous sites in the coding regions (2.2-fold higher for SC/SD, 1.5-fold higher for AB/SC, and 1.4-fold higher for AB/SD).
Although not as rapidly evolving as mtDNA genes, are the rates of CYC1, RISP, and CYC nonsynonymous substitution elevated above typical nuclear genes that do not interact with mitochondrial-encoded proteins in T. californicus?To address this question, nuclear divergences for such ''typical'' nuclear genes are compiled in table 6.In addition to divergences from three previously published genes (P5CS and P5CR [Willett and Burton 2002] and H1 [Burton and Lee 1994;Burton 1998]), sequences were also obtained from glutamate dehydrogenase (GDH) of just over 1,300 bp, including parts of exons 1 and 2 and the 71-bp intron 1 for the AB and SC populations (SD sequence obtained previously [Willett and Burton 2003]).The divergence in these genes at synonymous sites is roughly equal to that of CYC, CYC1, and RISP with these three genes evolving 1.14-fold faster for SD/SC, 1.16-fold faster for AB/SC, and 1.83-fold faster for AB/SD.These results contrast sharply with rates of synonymous substitution for the average of these seven nuclear genes compared with the average rate of synonymous substitution in the three mtDNA genes (using the JC-corrected values in table 2): 38.2-fold lower for SD/SC, 34.6-fold lower for AB/SC, and 26.7-fold lower for AB/SD.To compare rates of nonsynonymous substitution normalized to regional mutation rate differences in CYC1, CYC, and RISP with the ''typical'' genes, we can look at D n /D s ratios for each set of genes.On average, the four ''typical'' genes are evolving at a 2.7-fold lower rate for the SD/SC comparison, a 3.9-fold lower rate for the AB/SC comparison, and a 1.3-fold lower rate for the AB/SD comparison.On the surface, it appears that the proteins interacting with mitochondrial encoded genes are evolving more rapidly; however, these averages mask differences between the genes.CYC and H1 stand out with high D n /D s ratios in their respective sets of proteins and bias the averages.Based upon comparisons with D. melanogaster amino acid sequences, CYC is a more conserved protein than CYC1 or RISP (75% identity), so its elevated rate of evolution is unexpected and could indicate positive selection in the T. californicus lineage.With regard to the elevated rate of evolution observed for histone H1, it is important to note that H1 is the linker histone and is not as conserved as other core histones.Although T. californicus and D. melanogaster amino acid identity (63%) appears to  be reasonably high, the H1 proteins align poorly and the value excludes numerous gaps and potentially high homoplasy, as H1 has a large bias toward lysine, proline, and alanine residues (61% of T. californicus H1).Hence, H1 could evolve faster than other nuclear genes because of lower constraint on the H1 gene product, but this is not likely to explain evolution at CYC.

Polymorphism Within T. californicus Populations for Nuclear Genes
We examined polymorphism within T. californicus populations for these nuclear genes to both contrast levels of polymorphism in nuclear and mtDNA genes and to use in MK tests for evidence of selection.Nonsynonymous polymorphisms are extremely rare in six nuclear proteins, with only two segregating sites of this type observed (table 7), but more common in H1 (see table 8).In contrast, synonymous sites and noncoding regions show moderate levels of polymorphism in all three populations.The p syn and p tot-sil values vary considerably between genes with a high of 1.98% for CYC in SC and a nonzero low of 0.045% for RISP in AB.Average values of p syn and p tot-sil for nuclear-encoded genes are higher in the SC population than the AB or SD populations; this pattern did not hold for the mtDNA gene CYTB (table 3).With equal mutation rates, the neutral expectation is a fourfold reduction in polymorphism for mtDNA-encoded genes in comparison with nuclear-encoded genes.For the genes we have surveyed, the difference in average levels of polymor-phism (less than a fourfold decrease for mtDNA-encoded genes) suggest that synonymous sites in mtDNA have a higher neutral mutation rate (SD is 10.5-fold higher, AB is 7.2-fold higher, and SC is twofold higher).
Polymorphism within T. californicus populations and divergence between populations were compared for the six nuclear genes in table 7. We tested for differences from neutral expectations using the MK test.For five of the six genes (excluding CYC), no significant deviations were found from neutral expectations (data not shown).In contrast, for CYC comparisons involving the SC population, there are deviations consistent with selection for amino acid replacement (statistically significant for the comparison of AB and SC) (table 8).Histone H1 data also showed some evidence for rapid evolution based upon D n /D s ratios for the SD/SC comparison.In an MK test, however, the excess is in the opposite direction, excess nonsynonymous polymorphism within populations rather than divergence between populations.These results suggest that positive selection has acted on CYC, leading to rapid amino acid evolution, whereas lower functional constraint could explain the pattern of evolution at H1.

Discussion
Protein-coding mtDNA genes appear to be evolving at a more rapid rate than protein-coding nuclear genes in T. californicus.Synonymous sites are evolving approximately 25-fold to 40-fold faster for sequence comparisons between populations for mtDNA genes (COI, COII, and All genes are completely linked in animal mtDNA, so either of these processes would reduce polymorphism for the entire molecule at all sites.Nonsynonymous rates are also higher in mtDNA than in the seven nuclear genes (7.9-fold faster for SD/SC, 9.7-fold faster for AB/SC, and 13.8-fold faster for AB/SD).Accelerated evolution of mtDNA relative to nuclear DNA is of course not unique to T. californicus; this pattern has also been seen in mammals and birds (Brown, George, and Wilson 1979;Arctander et al. 1995).However, such increases in mtDNA rates of substitution are not universal.Both Drosophila and sea urchins (genus Strongylocentrotus) appear to have roughly equal rates of evolution of mtDNA and single-copy nuclear DNA (Powell et al. 1986;Vawter and Brown 1986).The rate of evolution in plant mtDNA is lower than that of nuclear DNA (Palmer and Herbon 1988).Few other studies have explicitly examined the average level of divergence at synonymous and nonsynonymous sites for comparisons of mtDNA and nuclear genes.Pesole et al. (1999) found 22-fold higher synonymous substitution in complete mtDNA from humans and chimps in comparison with divergence in 20 nuclear genes from both species.Nonsynonymous sites had a threefold higher rate of substitution in mtDNA for comparisons of the same proteins.In an examination of 13 complete animal mtDNA sequences, Lynch and Jarrell (1993) reported that rates of amino acid substitution may be somewhat lower in mitochondrial-encoded proteins than in nuclear-encoded proteins.Comparisons of COI, COII, and CYTB with nuclear genes for D. melanogaster and D. simulans show less than a two-fold increase in synonymous rates but greater than a three-fold decrease in nonsynonymous rates for the mtDNA-encoded genes (Ballard 2000b;Betancourt, Presgraves, and Swanson 2002;C. S. Willett, unpublished results).Both synonymous and nonsynonymous rates of substitution appear to be substantially higher in T. californicus, suggesting a much higher mutation rate for mtDNA than nuclear.Noncoding sequence in large introns is likely to be the least constrained sequence we have analyzed, and comparisons of divergences of noncoding and synonymous sites for the three genes with large introns analyzed (CYC1, RISP, and CYC) show a 1.5-fold to twofold higher rate for noncoding sites, suggesting some constraint on synonymous sites in coding regions of nuclear genes.For comparison of mutation rates between nuclearencoded and mitochondrial-encoded genes, this constraint could inflate the difference if mitochondrial-encoded genes have less constrained synonymous sites (but only by about twofold).
Given the central importance of the mitochondrialencoded subunits in the ETS, a high mutation rate of mtDNA in comparison with nuclear DNA in T. californicus can only explain a rapid rate of amino acid replacement if the fixed substitutions behave neutrally.An alternate explanation for increased amino acid divergence in mtDNA-encoded proteins could be environmental selection on the ETS.Is there evidence for selection for amino acid replacement acting on CYTB?A negative Tajima's D statistic or Fu and Li's D test could indicate recovery from a recent selective sweep.Polymorphism for CYTB has nonsignificant negative values for both of these measures in the SD population.Both SC and AB populations have positive Fu and Li's D test values and nonsignificant Tajima's D statistic values.These tests then provide no strong evidence for recent selective sweeps on mtDNA.MK tests can detect repeated fixations driven by selection, but comparisons of polymorphism and divergence across these three populations for CYTB do not show a pattern consistent with positive selection.In fact, for comparisons involving SC, an opposite deviation consistent with excess replacement polymorphism is seen.Therefore, examination a Numbers do not match those in table 7 because some polymorphic sites were found in noncoding regions with alignment gaps between populations and were excluded.
b Polymorphism in H1 is taken from sequences from AB, SD, and SC populations previously obtained (Burton and Lee 1994;Burton 1998).
of polymorphism and divergence in T. californicus populations lends little support for selection acting on CYTB to fix amino acid changes; instead, there is some evidence for excess slightly deleterious polymorphism, a pattern common in animal mtDNA.
If amino acid substitution in CYTB has not been driven by positive selection, could fixed differences be deleterious?Because of its ephemeral high intertidal habitat, T. californicus experiences dramatic fluctuations in population size (Burton 1997), which could facilitate the stochastic fixation of deleterious mutations.Extinction and recolonization in subdivided populations has been shown to increase the probability of fixation of deleterious mutations (Cherry 2003).Given the close physical and functional contacts of the mitochondrial-encoded and nuclear-encoded subunits of the ETS, compensatory amino acid change (via selection on nuclear-encoded subunits) is a potential consequence of accumulated deleterious substitutions.For two key subunits of complex III, RISP and CYC1, we find little evidence for selection for increased amino acid replacement in MK tests on these two proteins.This contrasts with apparently pervasive coadaptation occurring in primate complex IV proteins where amino acid substitution appears to be accelerated in many subunits of the complex (Adkins, Honeycutt, and Disotell 1996;Andrews, Jermiin, and Easteal 1998;Wu et al. 2000;Schmidt, Goodman, and Grossman 2002).Our finding shows that substantial change in CYTB can occur without numerous compensatory changes in RISP or CYC1.This could imply that the majority of the amino acid changes observed in CYTB between T. californicus populations are functionally neutral, or alternatively, most compensatory change occurs within the same subunit (for example within CYTB).
Mapping the positions of the amino acid changes from T. californicus populations in RISP and CYC1 on vertebrate complex III structures could also provide functional insights.Both differences from the SD sequence for CYC1 are found in the a19 helix, which is thought to be near the interaction site with soluble CYC (Iwata et al. 1998;Zhang et al. 1998) although they are not the specific interacting residues in yeast (Lange and Hunte 2002).They include adjacent sites 68 and 69 (based on the vertebrate CYC1 mature protein positions) and are Val to Ile in SC/AB, and Met to Ile in SC.In RISP, residue 24 (based on vertebrate RISP mature protein position) is in the matrix portion of the protein and is a Lys to Asn change in AB.At position 50 in the transmembrane helix, there is a change from Gly to Ala in SC/AB.None of these residues have been implicated in a specific intermolecular interaction in yeast or vertebrate complex III.
Sequence comparisons provide no evidence for selection acting on RISP or CYC1 in the evolutionary divergence of these two proteins between T. californicus populations; however, there is evidence for selection acting on the soluble CYC protein that transports electrons between complex III and complex IV.This evidence comes from MK tests that suggest an excess of amino acid fixation for comparisons involving the SC population (table 8).Divergences between T. californicus populations for CYC appear high based on the high level of conservation of CYC in animals, but a relative rate comparison with the CYC of prawn does not suggest a dramatic increase in the rate of CYC in the Tigriopus lineage in general (1.2-fold higher).CYTB in contrast shows a twofold to threefold increase along the Tigriopus lineage, suggesting that rates of amino acid substitution in CYC and in CYTB may not be evolving in concert over long periods of time.A more relevant comparison for CYC evolution may be with COII of complex IV rather than CYTB with which CYC has no direct interactions.Like CYTB, COII is highly divergent between T. californicus populations, and hydrophobic portions of this protein (involved in CYC binding) are unexpectedly divergent (Burton, Rawson, and Edmands 1999;Rawson and Burton unpublished results).
Functional and fitness data also suggest selective divergence of CYC between genetically isolated T. californicus populations.Rawson and Burton (2002) showed that the enzymatic activity of complex IV differed when using SD population-derived CYC versus SC population-derived CYC as substrate.Enzymatic activity was consistently higher when complex IV and CYC were derived from the same natural population.Large temperature/enzyme interaction effects were also detected.Other studies show that fitnesses (as determined by deviations from expected Mendelian ratios) of different CYC genotypes vary dramatically in F 2 hybrids of AB, SC, and SD populations (Willett and Burton 2001).For AB/SC hybrids, dramatic influences of the rearing environment (including a different temperature regime) were observed; selection on CYC genotype was totally eliminated by a change from one environmental regime to another (Willett and Burton, 2003).The prevalence of environmental influences on CYC fitness and function suggest that local adaptation to temperature environment could be a selective force on CYC and influence its evolution.A model for molecular evolution in this system could involve a high mutation rate on mtDNA, leading to frequent fixation of neutral or slightly deleterious mtDNA mutations perhaps compensated by intramolecular changes.Coadaptation between mtDNA-encoded and nuclear-encoded proteins may then be driven by occasional bouts of adaptation to environmental conditions, leading to the evolution of specific components of the ETS complexes.

Supplementary Material
Amino acid alignments for RISP, CYC1, and CYTB for comparisons of the T. californicus proteins with homologs in other species are available online at the journal's website (http://www.mbe.oupjournals.org).Sequences for genes are available in GenBank with accession numbers AY344446 to AY344470.DNA sequence alignments in nexus format for all genes are also available online.

a
FIG. 2.-Gene structure for CYC1 and RISP from T. californicus.Sizes of introns and exons are indicated below figure (introns vary in size between populations as indicated).Region sequenced for all three populations is indicated by lines above alignment and also the distance to the nearest gene feature.Locations of start and stop codons are predicted from sequence and alignments with other homologs.Asterisk (*) indicates intron 1 end appears to have shifted 5 bp closer to start codon in SC and AB (from 6 bp away in SD).

Table 1
Amino Acid Identities for CYTB, CYC1, and RISP Proteins from SD T. californicus NOTE.

Table 2
Genetic Divergence in T. californicus CYTB in Comparison with COI and COII Site type refers to synonymous or nonsynonymous category using Nei/Gojobori method as implemented by the program DNAsp.D n /D s value is ratio of nonsynonymous divergence divided by synonymous divergence (using JC-corrected values).aUncorrected distance is given for this comparison because value is too high to use for JC correction.

Table 3
Polymorphism in CYTB Gene in T. californicus Populations (Rand and Kann 1996)xpected polymorphism and divergence ratios, with NI values greater than 1 consistent with negative selection and NI values less than 1 consistent with positive selection(Rand and Kann 1996).The levels of divergence at synonymous sites add uncertainty to calculations of NI for comparisons between T. californicus populations.The Ni values are greater than 1 for both uncorrected and JC corrected comparisons involving the SC population (table

Table 4
Neutrality Index Values for CYTB in T. californicus Populations

Table 5
Genetic Divergence Between T. californicus Populations for Nuclear-Encoded Proteins Interacting Directly with Mitochondrial-Encoded Proteins Site type refers to synonymous, total silent, which is synonymous plus noncoding, or nonsynonymous category using Nei/Gojobori method as implemented by the program DNAsp.D n /D s value is ratio of nonsynonymous divergence divided by synonymous divergence (in synonymous row) or divided by total silent divergence (in total silent row).

Table 6
Genetic Divergence Between T. californicus Population for Non-ETS Nuclear-Encoded Proteins

Table 7
Polymorphism Within Populations for Six T. californicus Nuclear Genes a Number of alleles sequenced for each locus, nonsynonymous segregating sites, synonymous segregating sites, and the total number of noncoding and synonymous segregating sites (total silent).bAverage number of pairwise differences for nonsynonymous sites (p nonsyn ), synonymous sites (p syn ), and total synonymous and noncoding sites (p tot-sil.).All values calculated using program DNAsp.c SD average.d SC average.e Only a single allele of P5CS was obtained from the AB population.f AB average.

Table 8
McDonald/Kreitman Tests for CYC and H1 Genes from T. californicus Populations Kreitman tests were performed using polymorphism in both species and fixed divergence between species for nonsynonymous and noncoding and synonymous sites (total silent.).Fisher's exact test was used to calculate the probability that observed changes differ from the neutral expectation.