Evidence for Selection at Multiple Allozyme Loci Across a Mussel Hybrid Zone

The frequency and strength with which selection shapes patterns of genetic variation is unknown. Whereas all loci should be roughly equally affected by demography and population history, selected loci may exhibit increased or decreased genetic differentiation relative to neutral loci (Cavalli-Sforza 1966 ). Thus, one way to test for selection on a specific category of loci is to compare geographic differentiation of this particular category relative to a category presumed to be neutrally evolving (McDonald 1994 ; McDonald, Verrelli, and Geyer 1996 ). A number of recent studies have followed this approach, in particular comparing patterns of genetic differentiation at allozyme to nonallozyme loci. Where discordances between allozymes and other nuclear markers have been reported, less genetic partitioning has been observed for allozymes, consistent with balancing selection reducing differences among geographical populations (e.g., Karl and Avise 1992 ; Pogson, Mesa, and Boutilier 1995 ; Latta and Mitton 1997 ). Here, we report the reverse situation; geographic partitioning is greater at allozymes relative to nonallozyme loci across a mussel hybrid zone. Northern Hemisphere populations of the mussels Mytilus trossulus and Mytilus edulis are distinguished by nearly fixed allozyme allele frequency differences and some morphometric differences (McDonald, Seed, and Koehn 1991 ). Mytilus edulis is found throughout the northern Atlantic, whereas M. trossulus is found in three disjunct regions: the Pacific, Atlantic North America, and the Baltic Sea. A northern European hybrid zone between the Atlantic M. edulis and the Baltic Sea M. trossulus has been well described, based on allozyme surveys. A relatively sharp transition in mussel allozyme allele frequencies (over 100 km) occurs at the mouth of the Baltic with edulis-like alleles predominating in the Atlantic populations and trossulus-like alleles predominating in the Baltic. This pattern is particularly distinct for Aap,Est-D,Gpi,Lap,Mpi, and Pgm (Theison 1978 ; Bulnheim and Gosling 1988 ; Varvio, Koehn, and Väinölä 1988 ; Väinölä and Hvilsom 1991 ), and these loci, with the exception of Lap, are diagnostic between M. edulis and M. trossulus throughout the Northern Hemisphere (Varvio, Koehn, and Väinölä 1988 ; McDonald, Seed, and Koehn 1991 ). Rapid clinal transitions in allele frequencies, in combination with significant linkage disequilibria (particularly involving Gpi,Lap, and Pgm), have supported the idea that this hybrid zone represents a situation of secondary contact between pure populations of M. edulis and M. trossulus (Väinölä and Hvilsom 1991 ). Recent mtDNA surveys, however, have found that all northern European Mytilus, including Baltic mussels, are fixed for M. edulis-type mtDNA such as might result from asymmetric mtDNA introgression into Baltic M. trossulus populations (Wenne and Skibinski 1995 ; Rawson and Hilbish 1998 ; Quesada, Wenne, and Skibinski 1999 ). Thus, mtDNA and allozymes portray conflicting information regarding the ancestry of Baltic mussels. Here, we ask whether Baltic mussels represent a case of asymmetric (M. edulis) mtDNA introgression into an otherwise pure population of M. trossulus, or alternatively, whether the mtDNA introgression was accompanied by multiple M. edulis loci introgressing into the Baltic M. trossulus population. We employed four nuclear DNA markers that are diagnostic between M. trossulus and M. edulis: Glu 5′ (Rawson, Joyner, and Hilbish 1996 ), ITS (Heath, Rawson, and Hilbish 1995 ), MAL-I (Rawson, Secor, and Hilbish 1996 ), and PLIIa (Heath, Rawson, and Hilbish 1995 ). Glu 5′ and PLIIa primers target protein coding regions (Glu 5′: polyphenolic adhesive protein; PLIIa: protamine-like sperm packaging protein), whereas ITS primers amplify the ITS-1,5.8S, and ITS-2 regions of rDNA, and MAL-I primers amplify the intron of a protein coding region of unknown function. Glu 5′ primers produce species-specific–sized PCR products, whereas ITS,MAL-I, and PLIIa PCR products are digested with restriction enzymes to yield species-specific DNA fragments. These four markers were scored from 29 Baltic mussels collected from Hånko, Finland, a pure M. trossulus population based on allozymes (Väinölä and Hvilsom 1991 ). In marked contrast to allozyme patterns, all of these markers show the majority of the Baltic alleles to be of M. edulis origin, with the estimated M. edulis allele frequencies ranging from 37%–75% (table 1 ). The high frequency of M. edulis alleles from Hånko refutes the idea that the Baltic mussels represent a pure M. trossulus population. Similarly, a survey of Glu 5′ (one of the four markers employed in the present study) from a Gdansk, Poland population found high frequencies of M. edulis alleles (54%, Borsa et al. 1999 ). This observation of moderate to high frequencies of M. edulis nonallozyme loci points to extensive M. edulis introgression at nonallozyme nuclear loci into the Baltic mussel populations. Multilocus genotypes also show extensive M. edulis introgression and are consistent with many generations of hybridization. Multilocus genotypes of individuals for our Baltic sample (excluding the dominant PLIIa locus) found two (out of 29 total) pure M. edulis individuals and no pure M. trossulus individuals. Several (14) individuals had F2-type genotypes and backcross individuals were predominantly edulis-like (10 edulis-like, 3 trossulus-like). Thus, from the four loci examined here, many Baltic mussels carry both M. edulis and M. trossulus alleles, and edulis-like genotypes are at a higher proportion than trossulus-like genotypes. We tested for deviations from Hardy-Weinberg expectations for Glu 5′ and MAL-I, the two codominant single locus markers, and found a significant deficit of heterozygotes for Glu 5′ (tested following Guo and Thompson 1992 , P < 0.001) and no deviations from Hardy-Weinberg for MAL-I. The low frequencies of Glu 5′ hybrids may indicate selection against hybrids at Glu 5′ or a linked locus, although no deviations from Hardy-Weinberg were detected for Glu 5′ among mussels from Gdansk (Borsa et al. 1999 ). In addition to frequency-based estimates of M. edulis introgression at nuclear loci, we used a genealogical approach to verify that individual Baltic mussels carry both M. edulis- and M. trossulus-derived alleles. The ITS regions of rDNA were PCR amplified using the ITS primers described above (Heath, Rawson, and Hilbish 1995 ) with high fidelity polymerase (Expand Hi-Fidelity Polymerase, Roche). We sampled from a Baltic population (Hånko, Finland, n = 5), two pure M. edulis populations (Trondheim, Norway, n = 5; Wood's Hole, Mass., n = 2), a pure population of M. trossulus (Coupeville, Wash., n = 5), and four individuals from an M. trossulus-M. edulis hybrid zone in Canada identified by our markers as pure M. trossulus individuals (Wolfville, Nova Scotia, Canada, n = 3; North Harbor, Newfoundland, Canada, n = 1). PCR products were TA cloned (Invitrogen), and an average of 10 clones per individual was sequenced to give 449 basepairs of sequence, including ITS-1 and 5.8S regions. Because rDNA has multiple copies, much of the within-individual variation is likely caused by imperfect gene conversion and is larger than expected from simple polymerase error. A subset of representative sequences was aligned with ClustalX (Thompson et al. 1997 ), and the remainder was manually aligned to the Clustal alignment. Sequences have been deposited in GenBank (accession numbers AF440869–AF441078). Because there was considerable ambiguity in the alignment, phylogenies were estimated both including (fig. 1 ) and excluding (not shown) gapped regions. Kimura's (1980) model of substitution with a gamma rate distribution was identified by MODELTEST (Posada and Crandall 1998 ) as the most appropriate model for our ITS sequences. The ratio of transitions to transversions and the gamma parameter were estimated from a parsimony tree and used as the input values for a maximum likelihood search (under a Kimura 1980 with gamma model). A heuristic search with the tree-bisection–reconnection method in PAUP* (Swofford 1998 ) found 27 equally likely trees when gapped regions were included and one likely tree when gaps were excluded. All trees, both including and excluding gapped regions, found reciprocally monophyletic lineages for M. edulis and M. trossulus, with Baltic alleles falling unambiguously in both the M. edulis and M. trossulus clades. Thus, the ITS gene tree points to hybrid origins of the Baltic mussel population. Three Baltic individuals had one M. trossulus allele each, whereas all five had multiple M. edulis alleles (fig. 1 ). Hence in this sample of ITS from five Baltic individuals, there were two pure M. edulis individuals, three hybrids, and no pure M. trossulus. It is notable that even in the three hybrid individuals, we found many distinct M. edulis alleles (beyond what is expected from polymerase error) but only found one M. trossulus allele each in three hybrid individuals. Similarly, M. trossulus bands, for the majority of individuals scored as hybrids, were substantially less intense on ethidium bromide–stained agarose gels than M. edulis bands. Although the high proportion of M. edulis alleles could be explained by the preferential amplification of M. edulis alleles from the tandem arrays, it may truly reflect a small proportion of M. trossulus alleles in hybrid genomes. If Baltic mussels do contain fewer M. trossulusITS copies relative to M. edulis copies, this would be further evidence of long-term introgression by M. edulis alleles. In summary, mtDNA and all nonallozyme nuclear markers examined in the Baltic mussels show that there has been extensive introgression of the M. edulis genome into the Baltic population of M. trossulus. In contrast, there has been no apparent evidence for extensive M. edulis introgression among previously published allozyme surveys (e.g., Theison 1978 ; Bulnheim and Gosling 1988 ; Varvio, Koehn, and Väinölä 1988 ; Väinölä and Hvilsom 1991 ; Wenne and Skibinski 1995 ) where mussels were collected from 1976 (Theison 1978 ) to 1991 (Wenne and Skibinski 1995 ). There is no reason to suspect that allozyme frequencies at Hånko or any other Baltic location have changed in the last decade, although we cannot currently exclude this possibility. If the allozyme allele frequencies given in table 1 are representative of current allele frequencies at Hånko, then the probability that all four nuclear DNA loci have higher M. edulis allele frequencies than the five diagnostic allozyme loci is less than 0.008 (based on 126 possible combinations of 4 items from 9 total loci). Taken at face value, the strong discordance between allozyme and nonallozyme markers can only be explained by selection acting on some loci. Although M. edulis mtDNA and other (nonallozyme) nuclear loci could, in principle, have some selective advantage over native M. trossulus loci, it is more likely that multiple M. trossulus allozyme loci, which are involved in metabolic functions, have been selectively maintained among the Baltic mussels. Cohesion of the M. trossulus allozyme phenotype may be caused by either the coadaptation of M. trossulus genes to each other or by independent local adaptation. Salinity in the Baltic is very low, and strong allozyme differentiation between Atlantic and Baltic populations has been observed for several osmoconformers (such as Mytilus) but not for osmoregulators (reviewed by Bulnheim and Gosling 1988 ; Väinölä and Hvilsom 1991 ). Thus, it seems probable that the Baltic environmental conditions affect the performance of mussel allozyme alleles and may act to keep M. trossulus-like alleles in high frequencies among the Baltic mussel populations. Whatever the cause of selection on allozyme loci, the discordant patterns of allozyme and nonallozyme introgression across the Baltic mussel hybrid zone presents a qualitatively different pattern than other examples of discordance between allozymes and other nuclear markers. In American oysters (Karl and Avise 1992 ), Atlantic cod (Pogson, Mesa, and Boutilier 1995 ), and limber pine (Latta and Mitton 1997 ), allozymes show less differentiation among geographically distant populations than nonallozyme markers, consistent with balancing selection maintaining similar allozyme allele compositions among populations. Here, allozymes are more differentiated between the Baltic M. trossulus and Atlantic M. edulis populations than nonallozyme loci. Although the patterns are qualitatively different, these studies point to the same conclusion: selection may frequently shape patterns of genetic variation at multiple allozyme loci. David Rand, Reviewing Editor Keywords: allozyme asymmetric introgression hybridization Mytilus edulisMytilus trossulus selection Address for correspondence and reprints: C. Riginos, Department of Biology, P.O. Box 90338, Duke University, Durham, North Carolina 27707. riginos@duke.edu . Table 1 Frequencies of edulis-type Alleles in Baltic Mussels for Loci Diagnostic Between M. trossulus and M. edulis Table 1 Frequencies of edulis-type Alleles in Baltic Mussels for Loci Diagnostic Between M. trossulus and M. edulis Fig. 1.—Maximum likelihood tree from ITS clones. This is one of 27 equally likely trees under a Kimura (1980) model of substitution with a gamma rate distribution and gapped sites included. Clones from each mussel are labeled by geographic origin of the mussel (FI: Finland; MA: Massachusetts; NF: Newfoundland; NS: Nova Scotia; NWY: Norway; and WA: Washington State) and also assigned a unique number for that location. Where multiple identical clones are included from the same mussel, the number of identical clones is given in parentheses. For example, MA1 (7) refers to seven identical clones from MA individual one. Mussels from MA and NWY are pure Mytilus edulis, and mussels from NF, NS, and WA are pure M. trossulus. The two major clades corresponding to M. edulis and M. trossulus alleles are well supported with 89% bootstrap support using parsimony (above branch; 1,000 replicates) and 100% consistency (below branch) among all 27 maximum likelihood trees. Filled circles indicate alleles from Finnish mussels We thank two anonymous reviewers for their comments and suggestions. This research was funded by the Duke University program in Molecular Evolution and Comparative Genomics and NSF.