Introduction

Species of mytiloid (Pteriomorpha: Mytiloida; marine mussels) (Fisher and Skibinski 1990; Hoeh et al. 1991; Zouros et al. 1992, 1994a, b; Skibinski et al. 1994a, b), veneroid (Heterodonta; Veneroida; marine clams) (Passamonti and Scali 2001), and unionoid (Palaeoheterodonta: Unionoida; freshwater mussels) (Hoeh et al. 1991, 1996, 2002; Liu et al. 1996; Curole and Kocher 2002) bivalves possess two independently inherited mitochondrial genomes. Males and females inherit a mitochondrial genome from their mother (the F mitotype); in addition, males possess a mitochondrial genome that they inherit from their fathers and pass on to their sons (the M mitotype), This mode of mitochondrial genome inheritance is designated doubly-uniparental inheritance (DUI) (Zouros et al. 1994a) or sex-limited mitochondrial DNA inheritance (Skibinski et al. 1994a).

Within the Unionoida, divergence of the M and F mitotype lineages is estimated at ≈450 MYBP (million years before present), with a minimum divergence of 213 MYBP for the paralogous genomes (Curole and Kocher 2002, but see Graur and Martin 2004). This ancient divergence is supported by independent analysis of several unionoid taxa (Curole and Kocher 2002; Hoeh et al. 2002). In addition, the sequence obtained from somatic tissue of the marine bivalve Neotrigonia margaritacea (Trigonioida: Neotrigoniidae) clusters with unionoid F mitotype sequences, indicating that the divergence of the M and F unionoid mitotypes precedes the divergence of these marine and freshwater bivalves (Hoeh et al. 2002).

In the 12 species of unionoid bivalves examined, the pattern of DUI has maintained fidelity since this ancient divergence (Hoeh et al. 1996, 2002; Curole and Kocher 2002). In contrast, DUI has not maintained fidelity in the mytiloids; failure to inherit the M mitochondrial genome has been observed in pair crosses and natural populations (Zouros et al. 1994b; Rawson et al. 1996; Saavedra et al. 1997; Quesada et al. 1999; Wood et al. 2003). When a male fails to inherit a mitochondrial genome from his father, the genome he received from his mother is recruited to function as the M genome and divergence between the F genome and the newly recruited, or masculinized, M genome begins de novo (Hoeh et al. 1996, 1997). Masculinized genomes may become fixed, resulting in an evolutionary history where the masculinized mitotype is more closely related to the conspecific F mitotype than to con/interspecific M mitotypes. This evolutionary pattern allows the inference of gender-switching events by phylogenetic analysis (Hoeh et al. 1997). Phylogenetic analysis of mytiloid and unionoid mitotypes supports a hypothesis of gender-switching in mytiloids but not unionoids; gender-switching events have occurred in the evolutionary lineages of four species of my- tiloids, while the pattern of DUI has maintained fidelity in all species of unionoids examined (Hoeh et al. 1997, 2002; Curole and Kocher 2002). These differences in the frequency of gender-switching led Hoeh et al. (1996) to hypothesize taxon-specific differences in the fidelity of DUI, a hypothesis supported by Curole and Kocher (2002) and Hoeh et al. (2002).

The lack of gender-switching in the Unionoida coincides with a unique mitotype-specific protein- coding polymorphism (Curole and Kocher 2002). The cytochrome c oxidase II (MTCO2) locus of the M genome contains two regions: a region homologous with the F genome and an ≈185 codon extension, continuous in reading frame with the homologous region (Curole and Kocher 2002). The extension is hypothesized to be functional protein- coding DNA based on several lines of evidence, including patterns and rates of nucleotide evolution, lack of premature stop codons, relative conservation of length over several million years of evolution, and presence of a polyadenylated mRNA transcript with the complete extension (Curole and Kocher 2002). As this extension is only present in the male mitotype, its function is inferred to be sex-specific, occurring within the mitochondrion, the cell, the organism, or at any combination of these levels. Although function within the mitochondrion is the current null hypothesis, high levels of divergence have hindered similarity comparisons with other protein sequences, failing to offer a hypothesis of protein function.

Despite its anonymous nature, the extension is the first observed functional difference between the M and F genomes (notwithstanding levels of divergence). Although Curole and Kocher (2002) estimated the minimum age of divergence of the gender-associated genomes at 213 MYBP, the minimum estimate for the age of the extension was half this because the paternal genome of Margaritifara hembeli (Unionoida: Margaritiferidae) could not be amplified. Additionally, the length of the extension is consistent across all taxa examined and in ten additional species for which sequences were not reported. Here, we undertake a study of the phylogenetic distribution and evolution of the extension in the Unionoida; particularly, we examined the age of the extension, the lability of the extension, and its pattern of evolution with respect to the homologous MTCO2 sequence.

Methods

Mussels were sexed based on either gonadal type or shell sexual dimorphism. Male mitotype sequences were obtained by PCR amplification of DNA extracted from male gonads. Female mitotype sequences were derived from female gonads or mantle tissue. The Lampsilis straminea M mitotype was amplified with the HCO2198 (Folmer et al. 1994) and the COII.2 (Curole and Kocher 2002) primers. PCR products were TA cloned and three independent clones were sequenced. The M. hembeli paternal mitotype was amplified using the COII.2/HCO2198 primer pair, isolated on an agarose gel and sequence obtained using the COII.2 and HCO2198 primers. The remaining paternal mitotypes were amplified using the male-specific primer UNIOMCOIR and the COII.2 primer (Curole 2004). In all cases, male-specific internal primers were subsequently designed and used to PCR amplify internal regions and to sequence the resulting PCR products. Female type sequence was amplified with the UNIOFCOIR primer (5′CCACCAATCATTATTG GCATTAC3′) and the COII.2 primer. These shorter amplicons were sequenced directly using the UNIOFCOIR and COII.2 primers. Reagent and cycling conditions are given in Curole and Kocher (2002) and Curole (2004). Products were sequenced using the Amersham cycle sequencing kit following the manufacturer’s protocols and visualized on an ABI 377 fluorescent sequencer. All sequences have been deposited in GenBank (accession nos. AY951915-AY951930).

Sequences for Katharina tunicata and Crassostrea gigas were obtained from GenBank (accession nos. U09810 and AF177226); sequences obtained in Curole and Kocher (2002) were also included in the analysis. Nucleotide sequences were translated into putative amino acid sequences and aligned using CLUSTALX (Thompson et al. 1997); gaps were then inserted into nucleotide sequences at the corresponding sites. This aligned nucleotide sequence data matrix was used in all analyses. Phylogenetic estimation of the M and F mitotypes was based on nucleotide sequences of the homologous region of MTCO2; all phylogenetic analyses were done with the PAUP* software program. Distances were calculated with the Tamura-Nei model with a gamma correction (α = 0.3, p invar = 0) to account for multiple substitutions. These parameters were identified as the best fit to the data using the MODELTEST program (Posada and Crandall 1998). The neighbor-joining (NJ) algorithm was used for tree construction. Maximum-parsimony (MP) analysis was done using a heuristic search with stepwise random addition (1000 replicates) and tree-bisection reconnection branch swapping. Maximum-likelihood (ML) tree estimation was performed using a Tamura-Nei model with and without the gamma correction using the above parameters and without the assumption of a molecular clock. The data set was also bootstrapped and subjected to NJ (10,000 replicates), MP (1000 replicates), and ML (100 replicates) analysis using the Tamura-Nei model with a gamma correction for both ML and NJ.

Estimation of rates of synonymous (Ks) and non-synonymous (Ka) substitutions, as well as intercodon variation in Ka/Ks rate ratios (ω), was done using the CODEML program of the PAML package. The extension and homologous regions, as defined in Curole and Kocher (2002), were analyzed separately. Pairwise estimates were calculated for five sets of taxa, such that there is no shared evolutionary history across comparisons (phylogenetically independent; Felsenstein 1985), and z-tests were used to determine if differences between taxa were significant (Berry and Lindgren 1996). Intercodon variation was estimated using the one-rate, neutral, selection and discrete models of CODEML (see Yang et al. 2000 for a description of the models). The two representatives of the Anodontinae, as well as Potamilus purpuratus were excluded from this analysis. Significance of differences in likelihood between models was determined using the likelihood ratio test (Yang et al. 2000).

Results

Neighbor-joining analysis using the gamma parameters identified with MODELTEST produced a tree topology with the the Pyganodon grandis F mitotype basal to all other mitotypes and the M. hembeli F mitotype basal to reciprocally monophyletic clades of M and F mitotypes, i.e., (Pg F(Mh F(M mitotype, F mitotype))). In addition, within the M mitotype the anodontines are placed as the most derived clade. We reject this topology for several reasons. It is inconsistent with the accepted relationships between families (Heard and Guckert 1970; Graf 2000; Hoeh et al. 2001; Lydeard et al. 1996); in particular, the Margaritiferidae are accepted as basal to the Unionidae, and within the Unionidae the Anodontinae are considered the basal taxon. It is also in contrast to previous analyses, where the P. grandis (Hoeh et al. 2002) and M. hembeli (Curole and Kocher 2002) F mitotypes clustered with other unionoid F mitotypes, forming a monophyletic F mitotype clade. Lastly, and perhaps most significantly, bootstrap analysis fails to support this topology, instead strongly supporting a topology with reciprocally monophyletic M and F clades, as well as supporting M. hembeli as basal to the Unionidae in the M clade (Fig. 1); in the F clade, M. hembeli clusters with P. grandis. The bootstrap-supported topology is consistent with NJ analysis of the dataset without the gamma parameters (analysis not shown).

Figure 1
figure 1

Maximum likelihood phylogenetic tree of male and female mitotypes using MTCO2 homologous region nucleotide sequences. Neighbor-joining bootstrap values are listed above the branches, maximum-likelihood bootstrap values below the branches, and maximum-parsimony bootstrap values are in parentheses.

Full size image

Maximum-parsimony analysis produced three trees differing in the arrangement of the Glebula rotundata, Plectomerus dombeyanus, and Amblema plicata M mitotypes. In contrast to the NJ analysis, all trees indicated reciprocally monophyletic M and F mitotype clades, a result strongly supported by bootstrap analysis. In addition, the anodontine M mitotypes are placed as basal to all other unionid M mitotypes, although this is not significantly supported by bootstrap analysis (26%). Inconsistent with bootstrap analysis, the M. hembeli and P. grandis F mitotypes are placed together as the most derived clade among all the F mitotypes; rather, bootstrap analysis weakly supports the placement of these two F mitotypes outside of all other unionid F mitotypes (Fig. 1).

Maximum-likelihood analysis using the gamma parameters identified with MODELTEST produced three trees with the highest likelihood (lnl = −3590) differing in the placement of the G. rotundata M mitotype. Similar to the MP analysis, all trees indicated reciprocally monophyletic M and F mitotype clades, again supported by bootstrap analysis (Fig. 1). Similar to the NJ analysis, the anodontine M mitotypes are placed as the most derived clade; similar to the MP analysis the M. hembeli and P. grandis F mitotypes are the most derived among the F mitotypes. Bootstrap support for these internal relationships is weak, with little support for clades other than the M mitotype anodontines and the two Quadrula taxa. Thus, in Fig. 1 we have chosen to present the topology derived using maximum likelihood with the Tamura-Nei model without a gamma correction (lnl = −3602), as it is consistent with (1) accepted relationships, (2) previously published analyses, and (3) the NJ, MP, and ML bootstrap analyses.

Supporting monophyly of the gender-associated mitotypes, a sex-specific extension of the MTCO2 gene, including a stop codon immediately upstream of the MTCO1 start codon, was observed in all species listed in Table 1. The longest extension is found in Anodonta implicata, with a length of 192 codons (including the stop codon; Fig. 2). With the exception of P. purpuratus, all species possessed an extension similar in length (range 177–192); in particular, M. hembeli possessed an extension 177 codons in length. In contrast, the unionid species P. purpuratus exhibits a significantly shortened extension that is only 48 codons long, but does have a termination codon. This is 35 standard deviations shorter than the average of 185.2 ± 3.9 codons excluding P. purpuratus and 3 standard deviations shorter than the average length including P. purpuratus (171.9 ± 45.6). In both cases, it is longer than F mitotype sequences, for which only one species differs by a single codon (0 is 47.1 and 4.1 standard deviations below the mean for the two respective standard deviations).

Table 1 Taxa analyzed for COII region
Full size table
Fig. 2
figure 2

Alignment of sequences for the M mitotype MTCO2 locus. The homologous and extension regions are labeled above the alignment. Hyphens indicate an indel and periods indicate amino acid identity with the A. implicata sequence. This figure was created with TexShade (Beitz 2000).

Full size image

Consistent with previous observations, the extension exhibits a greater rate of evolution than the homologous region when all sites are considered (Table 2). This increase in rate is generally twofold to threefold across all sites, but is negligible for the Quadrula comparison. Phylogenetically independent pairwise comparisons of Ka rates of substitution indicate a 2.8–6.3-fold greater rate for the extension; several comparisons of Ka in Table 2 are significant when subjected to a pairwise z-test. In contrast, similar pairwise comparisons of Ks substitution rates indicate roughly equivalent rates of substitution, with the extension being 0.14–1.34 times the homologous region rate. Although the Quadrula comparison is nearly tenfold different, it is not significant in a pairwise z-test. Average Ka divergence between the Ambleminae and Anodontinae subfamilies (Table 1) also reflects this result; corrected divergence between the two subfamilies for the homologous region averaged 14%, whereas corrected divergence for the extension averaged 81.9%. Differences in Ka rates are also reflected in estimates of the Ka/Ks ratio (ω), where ω is 3.5–20-fold greater for the extension (Table 2).

Table 2 Genetic distances for sets of phylogenetically independent taxa
Full size table

Patterns of intercodon variation differed between the two regions. In both cases, the one-ratio model was rejected, indicating significant intercodon variation in ω (Table 3). Among the models allowing for intercodon variation, both the selection and neutral models were significantly worse than the discrete model for both regions (Table 3). In the homologous region, a large proportion of codons are classified as being highly conserved with a low ω ratio (<0.01); the remaining sites had a moderate ω ratio (<0.30; Table 3). Analysis of complete M mitotype MTCO2 homologous region sequences available in GenBank indicate that results from the subregion analyzed are indicative of the complete homologous region, although more sites were classified in the moderate rate category (≈30%, unpublished analysis). Reflecting the extension’s overall greater ω rate, only 16% of sites were classified into the low ω rate category (Table 3). Nearly half (58%) of sites were assigned to the moderate-rate category, with the remaining sites assigned to a high-rate category not observed for the homologous region (>0.50; Table 3).

Table 3 Intercodon variation in the homologous and extension regions of MTCO2
Full size table

Discussion

The presence of DUI in three bivalve subclasses (Pteriomorpha: Mytiloida, Palaeoheterodonta: Unionoida, Heterodonta:Veneroida) indicates that DUI is not an aberration limited to the Mytiloida, but that its origin is likely very ancient. The widespread distribution of DUI offers many opportunities for understanding the evolution and mechanisms of mitochondrial DNA inheritance (Zouros 2000). That DUI exhibits different patterns of evolution between the Mytiloida and Unionoida enhances these opportunities. We examined the evolutionary pattern of DUI in the Unionoida, with a particular interest in sampling for the gender-switching events common in mytiloids but unknown in unionoids.

Evolutionary History of the M and F Mitotypes

As with previous studies of unionoids (Hoeh et al. 1996, 2002; Curole and Kocher 2002), we failed to observe gender-switching events among the 12 taxa sampled. This is supported by bootstrapped phylogenetic analysis of MTCO2 sequences, as well as the presence of a sex-specific extension of the MTCO2 gene in all M mitotypes. The relatively lower bootstrap support for these data is likely due to the small length of sequence analyzed (compared with Hoeh et al. 1996, 2002) and the inclusion of third codon positions of these highly divergent sequences (in contrast to Curole and Kocher 2002). Although bootstrap support is weaker than for previous analyses, the presence of an M mitotype-specific extension for all taxa analyzed strongly supports reciprocal monophyly. Thus, we conclude that gender-switching has not occurred in these lineages. These results bring to 17 the total number of Unionoidan taxa analyzed without a single observed gender-switching event, in stark contrast to the mytiloids (Hoeh et al. 1997), reaffirming our conclusion that the fidelity of DUI differs among taxa (Curole and Kocher 2002).

In contrast to Mytilus edulis (David Lunt, personal communication) and Mytilus trossulus (Mizi et al. 2005), where the M and F types do not exhibit large insertion–deletion protein-coding polymorphisms, a large protein-coding insertion is present in the M mitotype of the Unionoida. Previously, we reported that the lack of masculinization in the Unionoida coincided with an ≈185 codon extension of the MTCO2 gene in the M type genome (Curole and Kocher 2002). One possible explanation is that the extension is the result of a tandem mtDNA duplication that is degrading over time (e.g., Campbell and Barker 1999; Inoue et al. 2003; Shao and Barker 2003); however, we reject this hypothesis as the extension exhibits hallmarks of mitochondrial protein-coding sequence in all 12 species. Rates of synonymous substitution are an order of magnitude greater than non-synonymous substitutions, there are no premature stop codons, there is relative conservation of length for 11 of the 12 species, and we previously isolated a MTCO2 polyadenylated mRNA transcript with the full-length extension (Curole and Kocher 2002). Although we cannot rule out the possibility that the extension may be degrading in a manner that maintains these hallmarks, a model of selection that results in this pattern is unknown.

In contrast to the 11 other species examined, the P. purpuratus M mitotype possesses an extension that is significantly shorter than the 185.2 codon average. The remaining 11 species have extensions that are very similar in length despite large divergences in sequence composition. This stark difference leads us to speculate that the remaining P. purpuratus extension sequence may no longer be functional. Testing this hypothesis by functional assay is likely difficult, particularly considering that the function of the extension is unknown and may be best done by identifying the extension elsewhere in the genome. Thus, sequencing of the P. purpuratus M type mitochondrial genome is the next step in identifying whether the extension has been completely lost from this mitochondrial genome.

The presence of the extension in M. hembeli indicates that this protein-coding sequence evolved prior to the divergence of the Margaritiferidae, which is estimated to be the basal unionoid family (Heard and Guckert 1970; Graf 2000; Hoeh et al. 2001; but see Graf and ÓFoighil 2000; Hoeh et al. 2001, 1998), and the Unionidae (213 MYBP; Smith 1976, 2001). This doubles the estimated age of the extension and supports our previous observations that the extension is ancient and functionally important. Based on evolutionary relationships of the Unionoida, it is not surprising that an extension has also been found in the related family Hyriidae (order Unionoida; J. Walker, personal communication). Whether the age of this extension precedes the invasion of freshwater by the Unionoida is also of great interest. The marine order Trigonioida is the sister clade to the Unionoida (Hoeh et al. 1998; Graf and ÓFoighil 2000), and the placement of the N. margaritacea MTCO1 sequence with the unionoid F mitotypes suggests that N. margaritacea may possess DUI (Hoeh et al. 2002); if this is the case, examining the N. margaritacea M mitotype would be an important test of the phylogenetic distribution and age of the extension.

Molecular Evolution of the Extension

Additional pairwise comparisons between taxa support the previous observation that the extension is evolving more rapidly than the homologous region of MTCO2 (Curole and Kocher 2002). The extension appears to be evolving about twice as fast as the homologous region when averaged over all sites. As differences in Ka rates are significant for all pairwise comparisons but Ks estimates are not significantly different, we conclude that the extension is under relaxed selection relative to the homologous region. This conclusion should be tempered by the observation that synonymous divergence is saturated for all comparisons. Levels of non-synonymous divergence at the extension are extraordinarily high; with the exception of the Quadrula taxa, non-synonymous divergence at the extension exceeds divergence levels across all sites for the homologous region. This is reflected in the ω rate, which is much greater for the extension as compared with the homologous region. Estimates of ω and the proportion of sites in each category for the homologous region are similar to estimates for the mammalian mitochondrial genome (Yang et al. 2000), with a large proportion of sites classified as having a low ω rate and a smaller proportion having a moderate ω rate (Yang et al. 2000). A third class of sites, under positive selection, identified in the mammalian mitochondrial genome, was not estimated from this data set. Thus, the M mitotype MTCO2 homologous sequence is evolving in a manner similar to mitochondrial protein-coding genes of other metazoans.

In contrast, only a small number of sites in the extension are under strong constraining selection, while one-quarter of sites experience one non-synonymous substitution per site for every two synonymous substitutions per site. These sites are clearly under relatively low constraint but are not evolving neutrally as the neutral model was rejected. Addition of the two anodontine taxa to the analysis does not substantially affect the proportion of sites in each class and marginally increases the ω rates for each class (analysis not shown). Thus, only 16% (≈30) of the ≈185 codons in the extension, in contrast to 84.5% of the homologous region, are subject to strong stabilizing selection, while 26% (≈46) of the codons are evolving nearly free from constraint. The selection model did detect a small number of sites with ω > 1; however, this model was significantly worse than the discrete model. The ability to detect sites under positive selection is limited because of the large levels of divergence between taxa, which is well over 100% at synonymous sites. Intraspecific comparisons of the extension would be ideal to test the hypothesis that there are no sites under positive selection in the extension.

An extension this length for MTCO2 is unknown across a wide diversity of metazoans, including vertebrates, arthropods, nematodes, and molluscs, where variation in length of the 3′ end of MTCO2 is less than 10 codons. The extension may represent one of the remaining mitochondrially encoded genes. The ATP synthase 8 (MTATP8) gene is the only locus not identified in the Inversidens japonenesis male mitochondrial genome available in GenBank Accession no. AB055624; this is consistent with the presumed absence of this gene in the M. edulis female genome (Hoffman et al. 1992). Although sequence similarity may be limited by the rapid rate of evolution of MTATP8 (Boore and Brown 1994), the length of the extension is three times the 53 codon average (Drosophila melanogaster, Caenhorhabditis elegans, and K. tunicata) for MTATP8 and hydrophobicity plots of the extension are not consistent with K. tunicata and D. melanogaster plots (Boore and Brown 1994). The extension may also represent one of the ten nuclear encoded cytochrome c oxidase subunits, but again levels of Ka divergence are inconsistent with this hypothesis. Corrected divergence levels (1st and 2nd positions) between Homo sapiens and Bos taurus (90–98 MYBP fossil divergence) for cytochrome c oxidase loci do not exceed 20% (Saccone et al. 2000). In contrast, corrected divergence levels between the Ambleminae and Anodontinae (≈100 MYBP fossil divergence) for the extension are four times this, although divergence for the homologous region is comparable (20% for 1st and 2nd positions between Homo and Bos MTCO2 vs. 14% Ka divergence for Ambleminae-Anodontinae homologous MTCO2). Thus, we conclude that the extension is not one of the standard 13 mitochondially encoded proteins but is novel mitochondrial protein-coding sequence (Curole and Kocher 2002) and we speculate that the extension may not be part of the cytochrome c oxidase complex.

Evolution of DUI

Hypotheses as to the cause of the taxon-specific difference in fidelity can be divided into two ideological groupings: breakdown of DUI in mytiloids is actively promoted (e.g., hybridization, see Hoeh et al. 2002) or inhibition of gender-switching events (or their fixation) in the Unionoidea (e.g., via sequence divergence, see Hoeh et al. 2002; functional protein-coding differences, see Curole and Kocher 2002). The first hypothesis postulates that gender-switching occurs and can become fixed in unionids given the proper environment. Breakdown of DUI is elevated for hybridization events (Fisher and Skibinski 1990; Quesada et al. 1995; Rawson et al. 1996; Wenne and Skibinski 1996; Quesada et al. 1999; Wood et al. 2003); however, breakdown of DUI is not limited to hybridization events and appears to be male dependent (Zouros et al. 1994b; Saavedra et al. 1997). Supporting the hybridization hypothesis is the failure to observe gender-switching events in Mytilus californianus, which is not known to hybridize (B. Ort, personal communication). Regardless, as suggested by Hoeh et al. (2002), examination of unionid hybrid zones is an essential test of this hypothesis.

The second group of hypotheses are derived from the novel (at least in metazoans) situation of a paternally inherited mitochondrial genome; rather than the inability of natural selection to act upon mitochondria in the sperm environment (Frank and Hurst 1996; Zeh 2004), sperm-derived mitochondrial haplotypes are under increased selection, i.e., selection can now act upon sperm mitochondria to maximize the fitness of sperm (Zeh 2004). In particular, Hoeh et al. (2002) suggested that the extensive divergence between the M and F mitotypes may have resulted in the reification of nuclear-cytoplasmic interactions, and individuals experiencing role-reversal events would have reduced fitness. It is clear that the extensive divergence between M and F mitotypes has resulted in substantial coding differences between the two genomes (28–34% for unionoideans as compared with 18–21% for mytiloids at MTCO1, Hoeh et al. 1996) and there is limited evidence that role-reversal events may lead to reduced fitness in M. edulis × M. gal-loprovincialis hybrids (Wood et al. 2003). In contrast, reduced fitness of hybrid males lacking an M mitotype does not serve as a complete block to masculinization in the mytiloids. Curole and Kocher (2002) outlined several hypotheses for how the extension may serve as a block to masculinization. Unfortunately, these hypotheses are dependent upon identifying a specific function for the extension, a rather formidable task. However, it is tempting to speculate that the extension is a result of the new selective environment and provides a fitness increase to sperm that may inhibit masculinization. In this respect, identification of taxa without a functional extension will provide an opportunity to test this hypothesis.