Abstract

Deep-sea clams of the family Vesicomyidae live in symbiosis with intracellular chemosynthetic bacteria. These symbionts are transmitted maternally (vertically) between host generations and should therefore show a pattern of genetic variation paralleling that of the cotransmitted host mitochondrion. However, instances of lateral (nonvertical) symbiont acquisition could still occur, thereby decoupling symbiont and mitochondrial phylogenies. Here, we provide the first evidence against strict maternal cotransmission of symbiont and mitochondrial genomes in vesicomyids. Analysis of Vesicomya sp. mt-II clams from hydrothermal vents on the Juan de Fuca Ridge (northeastern Pacific) revealed a symbiont phylotype (designated symBVII) highly divergent from previously described symbionts of the same host lineage. SymBVII-hosting clams occurred at low frequency (0.02) relative to individuals hosting the dominant symbiont phylotype. Phylogenetic analysis of 16S rRNA genes from a wide range of symbionts and free-living bacteria clustered symBVII within the monophyletic clade of vesicomyid symbionts. Further analysis of 3 symbiont loci (23S, dnaK, and soxA) across 11 vesicomyid taxa unambiguously placed symBVII as sister to the symbiont of a distantly related host lineage, Vesicomya sp. from the Mid-Atlantic Ridge (98.9% median nucleotide identity across protein-coding loci). Using likelihood and Bayesian model discrimination methods, we rejected the strict maternal cotransmission hypothesis by showing a significant decoupling of symbiont and host mitochondrial (COI and mt16S genes) phylogenies. Indeed, decoupling occurred even when symBVII was excluded from phylogenetic reconstructions, suggesting a history of host switching in this group. Together, the data indicate a history of lateral symbiont transfer in vesicomyids, with symBVII being the most conspicuous example. Interpreted alongside previous studies of the vesicomyid symbiosis, these results suggest a mixed mode of symbiont transmission characterized by predominantly vertical transmission punctuated with instances of lateral symbiont acquisition. Lateral acquisition may facilitate the exchange of genetic material (recombination) among divergent symbiont lineages, rendering the evolutionary history of vesicomyid symbiont genomes much more complex than previously thought.

Introduction

Symbioses between chemosynthetic bacteria and marine invertebrates represent extreme examples of prokaryote–eukaryote coevolution. These mutualisms dominate the fauna at deep-sea hydrothermal vents and cold seeps, where the invertebrate host facilitates access to substrates (sulfur or methane, oxygen, carbon dioxide) needed for the chemoautotrophic metabolism of the symbiont. In exchange, symbiont carbon fixation supports most, if not all, of the host's nutrition (Stewart et al. 2005). This tight metabolic coupling exerts on each partner a strong selective pressure for maintenance of the interaction. Such coevolutionary interactions might be expected to drive the parallel diversification of chemosynthetic symbionts and hosts (cospeciation), as has been shown in other prokaryote–eukaryote symbioses (Chen et al. 1999; Clark et al. 2000; Lo et al. 2003; Degnan et al. 2004; Wade 2007). However, many marine chemosynthetic symbioses do not show strong patterns of cospeciation (Krueger and Cavanaugh 1997; Di Meo et al. 2000; Nelson and Fisher 2000; Won et al. 2003; Suzuki et al. 2006; Vrijenhoek, Duhaime, and Jones 2007). Notably, symbioses between gamma proteobacteria and heterodont clams of the Vesicomyidae are the only members of this group that show a significant coupling of symbiont and host genetic variation (Peek, Feldman, et al. 1998). These symbioses are therefore useful models for studying bacteria–eukaryote coevolution. This study examines an interesting exception to the general cospeciation pattern in vesicomyids and discusses its implications for inferences of symbiont transmission mode in these taxa.

The Vesicomyidae comprises at least 50 species occurring ubiquitously at vents and seeps throughout the world's oceans (Goffredi et al. 2003; Kojima et al. 2004; Krylova and Sahling 2006). Fossil evidence suggests that this diverse group radiated between 50 and 100 MYA, and all extant species are thought to host sulfur-oxidizing bacterial endosymbionts (Kiel and Little 2006). Several lines of evidence show that both the clam and the symbiont have undergone extensive coadaptation to accommodate life in symbiosis. The host retains only a rudimentary gut, relying instead on symbiont autotrophy for nutrition (Kennish and Lutz 1992). The bacterial symbionts reside within specialized cells of the gills (Cavanaugh 1983; Fiala-Médioni and Métivier 1986) and appear to be obligately symbiotic, as they have never been cultured or detected in the environment. Further, genomic studies show that the single chromosome of vesicomyid symbionts is only approximately 1.1 Mb in size, suggesting that, like the bacterial symbionts of insects (e.g., Buchnera; van Ham et al. 2003), vesicomyid symbionts may be experiencing genome reduction as an adaptation to an obligate intracellular lifestyle (Kuwahara et al. 2007; Newton et al. 2007).

The coadapted symbiont–host interaction is presumably maintained across generations by maternal (vertical) transmission of the symbiont via the clam's eggs. Maternal transmission in vesicomyids was first suggested by electron micrographs showing bacteria-like structures in ovarian tissue (Endow and Ohta 1990) and then by molecular probing that localized symbionts to the follicle cells of the primary oocyte (Cary and Giovannoni 1993). Subsequently, maternal transmission has also been inferred using phylogenetic and population genetic studies. Under a strict maternal transmission hypothesis, the symbiont and host mitochondrial genomes necessarily pass through the same individuals and are completely genetically linked (Hurtado et al. 2003). Patterns in symbiont variation should therefore parallel those of the mitochondrion at both the species and the population level (Clark et al. 2000; Hurtado et al. 2003). Indeed, using maximum likelihood (ML)–based hypothesis testing, Peek et al. (1998) showed a general congruence between symbiont 16S rRNA and host mitochondrial gene phylogenies for 9 vesicomyid taxa, though minor discrepancies occurred at 2 of the 7 internal nodes. Similarly, for clams of the Vesicomya pacifica/lepta species complex (eastern Pacific), Goffredi et al. (2003) showed that symbiont and host genealogies exhibited parallel partitions, suggesting a tight coupling of symbiont and host mitochondrial genomes. Further, focusing on intraspecies and intrapopulation variation, Hurtado et al. (2003) demonstrated symbiont–mitochondrion coupling in a population of the vesicomyid Calyptogena magnifica that was polymorphic for both symbiont and mitochondrial markers. Together, these studies suggest a hypothesis of nearly complete maternal symbiont transmission in these taxa (Hurtado et al. 2003).

Events that disrupt strict maternal transmission and decouple symbiont and clam phylogenies might still be possible. These include the acquisition of new symbionts from the environment or the horizontal transfer of symbiont lineages between distinct clam hosts (host switching), perhaps via hybridization events. Here, we use the term “lateral” to specify any nonmaternal (i.e., nonvertical) transfer events, including horizontal transfer between hosts and environmental acquisition. A decoupling of symbiont and host phylogenies has been used as evidence of lateral transfer in other symbiotic associations, notably those between insects and maternally transmitted parasitic bacteria of the genus Wolbachia (Werren et al. 1995; Schilthuizen and Stouthamer 1997; Haine et al. 2005). For vesicomyids, detection of lateral transfer events would suggest that the specificity of the symbiont–host interaction is not absolute, raising the possibility that a single host lineage could accommodate functionally divergent symbiont lineages (or vice versa) and that vesicomyid clams might be at different stages in the process of completely internalizing their nutritional symbionts. Further, lateral symbiont transfer in these taxa might create opportunities for the exchange of genomic material (recombination) between diverse symbiont lineages (e.g., see Papke and Ward 2004). To date, however, lateral symbiont transfer in vesicomyids has not been shown.

This study shows a significant decoupling of vesicomyid symbiont and host mitochondrial phylogenies, providing the first evidence of lateral symbiont transfer events in the evolutionary history of these clams. Genetic analysis of hydrothermal vent clams from the northeastern Pacific revealed a symbiont phylotype highly divergent from other symbionts of the same host lineage. Multilocus analysis showed that the divergent symbiont clusters tightly with the symbiont from a distantly related host clam. Further, Bayesian analysis using a computationally robust model choice statistic, the Akaike information criterion Monte Carlo (AICM; Raftery et al. 2007), demonstrated incongruence between symbiont and host mitochondrial gene trees even in the absence of the divergent symbiont lineage, suggesting a history of lateral symbiont acquisition in this clam family.

Methods and Materials

Vesicomyid Clam Specimens

This study analyzed clams of the Vesicomya sp. mt-II lineage, as named by Goffredi et al. (2003) according to mitochondrial haplotype. Vesicomya sp. mt-II clams (n = 3) were collected from a hydrothermal vent site on the North Endeavor segment of the Juan de Fuca (JdF) Ridge in 1991. Following the discovery of a unique symbiont lineage in 1 of these clams, we screened an additional 115 clams collected from 6 JdF sites in 1995 and 1999 (table 1). We also obtained samples of gill tissue from 9 other vesicomyid species (table 1), either via direct collection or from collaborators. These samples represent species from vent or seep sites in the Pacific Ocean, the Gulf of Mexico, and the Atlantic Ocean. Clams were typically dissected immediately following collection or frozen whole until dissection and analysis at the home institution.

Table 1

Identifications and Collection Sites of Vesicomyid Clams Used in This Study

Host Species n Year Location Habitat Latitude, Longitude Divea Depth (m) 
Vesicomya sp. mt-IIb 3c 1991 JdF Ridge, North Endeavor, Vent 47-57.4N, 129-05.9W A 2413 2,200 
 32 1995 JdF Vent 47-58.1N, 129-05.2W ATV 2,200 
 1999 JdF Vent 47-57.78N, 129-05.51W A 3454 2,182 
 1999 JdF Vent 47-57.66N, 129-05.62W A 3456 2,184 
 1999 JdF Vent 47-58.11N, 129-05.24W A 3457 2,175 
 34c 1999 JdF Vent 47-57.79N, 129-05.48W A 3459 2,190 
 32 1999 JdF Vent 47-57.73N, 129-05.55W A 3462 2,190 
Vesicomya sp. MAR 2001 Logatchev, Mid-Atlantic Ridge Vent 14-45.32N, 44-58.79W A 3668 3,028 
Calyptogena magnifica 2003 East Pacific Rise Vent 09-50.88N, 104-17.61W A 3951 2,507 
Calyptogena kilmeri 2004 Monterey Canyon Cold seep 36-46.53N, 122-5.21W 970 
Calyptogena okutaniiNA  Sagami Bay, Japan Cold seep 34-57N, 139-12E HP 305 1,157 
Calyptogena ponderosa 1992 Gulf of Mexico Cold seep 27-40.88N, 91-32.10W J 3276 720 
Ectenagena extenta 2004 Monterery Canyon Cold seep 36-40.92N, 122-7.21W 1,464 
Vesicomya gigas 2001 Mendocino Fracture Zone Cold seep 49-20.06N, 127-39.69W T 473 1,694 
Vesicomya sp. mt-Ib 1999 Monterey Bay Cold seep 36-47N, 122-5W 500–1,000 
Vesicomya sp. mt-IIIb 2000 Monterey Bay Cold seep 36-37.79N, 122-19.92W T 217 2,200 
Host Species n Year Location Habitat Latitude, Longitude Divea Depth (m) 
Vesicomya sp. mt-IIb 3c 1991 JdF Ridge, North Endeavor, Vent 47-57.4N, 129-05.9W A 2413 2,200 
 32 1995 JdF Vent 47-58.1N, 129-05.2W ATV 2,200 
 1999 JdF Vent 47-57.78N, 129-05.51W A 3454 2,182 
 1999 JdF Vent 47-57.66N, 129-05.62W A 3456 2,184 
 1999 JdF Vent 47-58.11N, 129-05.24W A 3457 2,175 
 34c 1999 JdF Vent 47-57.79N, 129-05.48W A 3459 2,190 
 32 1999 JdF Vent 47-57.73N, 129-05.55W A 3462 2,190 
Vesicomya sp. MAR 2001 Logatchev, Mid-Atlantic Ridge Vent 14-45.32N, 44-58.79W A 3668 3,028 
Calyptogena magnifica 2003 East Pacific Rise Vent 09-50.88N, 104-17.61W A 3951 2,507 
Calyptogena kilmeri 2004 Monterey Canyon Cold seep 36-46.53N, 122-5.21W 970 
Calyptogena okutaniiNA  Sagami Bay, Japan Cold seep 34-57N, 139-12E HP 305 1,157 
Calyptogena ponderosa 1992 Gulf of Mexico Cold seep 27-40.88N, 91-32.10W J 3276 720 
Ectenagena extenta 2004 Monterery Canyon Cold seep 36-40.92N, 122-7.21W 1,464 
Vesicomya gigas 2001 Mendocino Fracture Zone Cold seep 49-20.06N, 127-39.69W T 473 1,694 
Vesicomya sp. mt-Ib 1999 Monterey Bay Cold seep 36-47N, 122-5W 500–1,000 
Vesicomya sp. mt-IIIb 2000 Monterey Bay Cold seep 36-37.79N, 122-19.92W T 217 2,200 
a

Dive: A = DSV Alvin, ATV = ROV Advanced Tethered Vehicle (Dive 95-48-149, High Rise Expedition), HP = ROV Hyper Dolphin, J = DSV Johnson Sea Link, T = ROV Tiburon, V = ROV Ventana (no associated dive numbers).

b

Vesicomya sp. mt-I, -II, and -III are unnamed clam species that have been designated according to mitochondrial type by Goffredi et al. (2003).

c

Populations in which the divergent symbiont phylotype (symBVII) was discovered.

d

C. okutanii host and symbiont sequences obtained from GenBank; collection details correspond to those associated with the sequencing of the V. okutanii symbiont genome (Kuwahara et al. 2007).

DNA Extraction, Polymerase Chain Reaction, and Sequencing

Total DNA was extracted from the symbiont-containing gill tissue of each clam using the DNeasy tissue kit (Qiagen, Valencia, CA). Gill DNA extracts were used to polymerase chain reaction (PCR) amplify and sequence portions of 3 host and 4 symbiont loci using the primer sets listed in supplementary table 1 (Supplementary Material online). Host markers included portions of the mitochondrial cytochrome oxidase c subunit I gene (COI; 729 bp), mitochondrial 16S rRNA gene (mt16S; 441 bp), and first internal transcribed spacer (ITS) of the nuclear rRNA operon (nucITS; 1,190 bp). These host loci have been used previously to characterize vesicomyids collected from the JdF (Peek et al. 1998; Goffredi et al. 2003). Symbiont markers included portions of the 16S rRNA gene (16S; 1,303 bp), the 23S rRNA gene (23S; 1,725 bp), and 2 single-copy protein-coding genes, dnaK and soxA (735 and 594 bp). The soxA gene encodes an enzyme in the sulfur oxidation pathway of thiotrophic prokaryotes and is presumably critical to energy generation by vesicomyid symbionts (Friedrich et al. 2001; Newton et al. 2007). The dnaK gene encodes a molecular chaperone that aids the refolding of damaged proteins. This housekeeping gene is highly conserved and has been used extensively in bacterial phylogenetics (e.g., Gupta 1998). PCR primers for dnaK and soxA were developed based on conserved regions in the genomes of 2 vesicomyid symbionts: Candidatus Ruthia magnifica and Candidatus Vesicomyosocius okutanii (the symbionts of the clams Calyptogena magnifica and C. okutanii, respectively; Kuwahara et al. 2007; Newton et al. 2007). Following visualization via agarose gel electrophoresis, PCR products were purified using the QIAquick PCR purification kit (Qiagen). Purified products were directly sequenced in both the forward and the reverse directions using standard BigDye techniques. Sequences were assembled and edited manually in Sequencher v4.7 prior to phylogenetic analyses. GenBank accession numbers for all sequences obtained in this study are listed in supplementary table 2 (Supplementary Material online).

Phylogenetic Analyses

Multiple gene trees were constructed to describe the symbiont phylotypes in this study. Here, the symbiont phylotype previously characterized for Vesicomya sp. mt-II clams is designated symAVII, whereas the divergent symbiont phylotype described in this study is designated symBVII. The corresponding host haplotypes are clamAVII and clamBVII, respectively. The broad phylogenetic placement of symAVII and symBVII was first determined in an analysis of 16S rRNA gene sequences from diverse vesicomyid symbionts (n = 20), symbionts of other marine invertebrates (n = 17), and several of the most closely related free-living bacteria that appeared in the BlastN results list for the symBVII16S sequence (n = 11).

To obtain greater phylogenetic resolution, we then conducted a multilocus analysis of 4 symbiont genes (16S, 23S, dnaK, and soxA) from 9 additional vesicomyid taxa (table 1). The corresponding host mitochondrial gene tree for this taxon set was obtained in an analysis of concatenated COI and mt16S sequences, as in Peek et al. (1998). Calyptogena okutanii was excluded from the mitochondrial analysis, as we did not have mt16S data for this species. Additionally, to confirm the identity of clamAVII and clamBVII host clams, nucITS sequences from these taxa were analyzed relative to published sequences from clams belonging to the Vesicomya sp. mt-I, -II, and -III lineages (see Goffredi et al. 2003). Due to their hypervariability, nucITS sequences from the other, more distantly related taxa analyzed in this study could not be unambiguously aligned and were excluded from the analysis. The above analyses were rooted using BlastN to identify the free-living bacterium most closely related to the symbiont taxa. Sequences were aligned in ClustalW (Thompson et al. 1994) and then edited manually in MacClade 4.0 (Maddison DR and Maddison WP 2000).

We first conducted rooted phylogenetic analyses on each locus using ML, maximum parsimony (MP), and Bayesian methodologies. We report only Bayesian phylogenies, as ML and MP phylogenies were not significantly different from those reported (supplementary figs. 2 and 3, Supplementary Material online). A second set of unrooted analyses using the taxa specified in table 1 (excluding C. okutanii) was then implemented to test for topological coupling between symbiont and host mitochondrial gene trees (these analyses are described below in Tests for Symbiont–Host Coupling).

Bayesian phylogenetic analyses were conducted on each gene separately using MrBayes v3.1.2 (Huelsenbeck and Ronquist 2001). Prior to the analyses, we tested a set of nested models of sequence evolution that are restrictions of the general time reversible (GTR) model with rate variation among sites (Tavaré 1986; Yang 1994). The best-fit model of sequence evolution for each locus was evaluated based on the Akaike information criterion using standard procedures in PAUP v4.0b10 (Akaike 1974; Swofford 1998; Posada and Crandall 2001) and then chosen according to Akaike weights (supplementary tables 3 and 4, Supplementary Material online). Diffuse priors were assumed for all analyses. Under models with site-specific rates, rate parameters (Dirichlet(1,1,1) prior) were assigned to partitions of first, second, and third codon positions. Topology (diffuse prior, all topologies equally weighted), branch lengths (Exponential(10) prior), instantaneous rate matrices (Dirichlet(1,1,1,1,1,1) prior), and equilibrium base frequencies (Dirichlet(1,1,1,1) prior) were shared among the data partitions. Other loci required gamma-distributed rates (α; Uniform(0.05,50) prior) or a proportion of invariant sites (I; Uniform(0,1) prior; supplementary table 4, Supplementary Material online).

All Markov chain Monte Carlo (MCMC) analyses were conducted using Metropolis coupling with 20 or more parallel chains. Swap rates between adjacent chains were >20% in all cases. Preliminary MCMC runs were conducted for each analysis to determine appropriate MCMC algorithm tuning, and short runs using optimal tuning parameters were repeated a minimum of 5 times. Two long chains were iterated 5.0 × 106 times, and parameters were sampled every 1,000 iterations. Of these 5,000 samples, the first 1,000 were discarded as burn-in. All repetitions of the analyses converged on very similar parameter estimates, and the 2 long chains were combined for a total of 8,000 posterior samples. MCMC convergence was assessed using the CODA package in R (R Development Core Team 2007). Furthermore, as expected for runs that have converged, the medians of the posterior distributions of model parameters for each locus were all close to ML estimates computed using PAUP (supplementary table 4, Supplementary Material online).

Tests for Symbiont–Host Coupling

Under the prevailing hypothesis of symbiont transmission for vesicomyids, symbionts and host mitochondria are cotransmitted via the egg and should therefore exhibit parallel patterns of genetic divergence (topologies). Coupling between symbiont and mitochondrial topologies was assessed using Bayesian and likelihood methods applied to unrooted data sets containing the vesicomyid taxa in table 1 (excluding C. okutanii and its symbiont).

Bayesian analyses compared a coupling model, T1, in which both host and symbiont loci share the same topology, with a decoupling model, T2, in which host and symbiont loci were allowed 2 separate topologies (table 2). The analyses were run separately for each symbiont locus concatenated to the 2 host loci (e.g., dnaK–COImt16S), as well as for a data set containing all 4 symbiont loci concatenated to the 2 host loci (16S–23S–dnaK–soxACOI–mt16S). Single symbiont locus analyses were run both for data sets that contained the divergent symbiont symBVII and its corresponding host clamBVII, as well as for data sets that lacked these sequences. The 6-locus analysis was run only for a data set lacking symBVII and clamBVII. If the data lent greater statistical support to T2, then a decoupling of symbiont and host mitochondrial phylogenies was indicated.

Table 2

Topology Tests Based on Vesicomyid Hosta and Symbiont DNA Sequence Data

  AICMb BFc 
Comparison Modeld Thin ρ1 forumlamax forumla AICM MCSE wi Log[f(X|Mi)] Log(BF) Odds Ratio P 2log(BF) 
w/ sym/clamBVII              
COI versus mt16S T2 0.01 −3,380 66 6,825 3.6 0.99 −3,393     
T1 −0.03 −3,397 40 6,834 2.2 0.01 −3,408 15 3.3 × 1006 0.00 30 
16S versus host T2 0.03 −5,837 76 11,750 4.1 1.00 −5,853     
T1 −0.02 −5,909 50 11,868 2.7 0.00 −5,919 66 4.6 × 1028 0.00 132 
23S versus host T2 0.02 −6,658 70 13,385 3.8 1.00 −6,679     
T1 −0.003 −6,761 51 13,572 2.8 0.00 −6,775 96 4.9 × 1041 0.00 192 
dnaK versus host T2 0.04 −5,429 64 10,922 3.5 1.00 −5,444     
T1 0.06 −5,659 46 11,363 2.5 0.00 −5,668 224 1.9 × 1097 0.00 448 
soxA versus host T2 0.006 −5,362 69 10,792 3.7 1.00 −5,376     
T1 −0.003 −5,547 49 11,143 2.7 0.00 −5,557 181 4.0 × 1078 0.00 362 
w/o sym/clamBVII              
COI versus mt16S T2 0.03 −3,361 57 6,779 3.8 1.00 −3,374     
T1 −0.01 −3,380 36 6,796 2.5 0.00 −3,391 17 2.4 × 1007 0.00 34 
16S versus host T2 −0.01 −5,815 66 11,696 3.6 1.00 −5,832     
T1 0.01 −5,842 46 11,730 2.5 0.00 −5,853 21 1.3 × 1009 0.00 42 
23S versus host T2 −0.01 −6,607 65 13,279 3.5 1.00 −6,620     
T1 0.004 −6,630 46 13,305 2.5 0.00 −6,638 18 6.6 × 1007 0.00 36 
dnaK versus host T2 0.03 −5,361 64 10,786 3.5 1.00 −5,377     
T1 −0.01 −5,407 47 10,861 2.6 0.00 −5,420 43 4.7 × 1018 0.00 86 
soxA versus host T2 0.01 −5,320 62 10,702 3.4 1.00 −5,332     
T1 0.02 −5,354 45 10,753 2.5 0.00 −5,367 35 1.6 × 1015 0.00 70 
All sym versus hoste T2 −0.003 −12,979 96 26,053 4.6 1.00 −12,997     
T1 0.03 −13,034 74 26,143 3.6 0.00 −13,058 61 3.1 × 1026 0.00 122 
  AICMb BFc 
Comparison Modeld Thin ρ1 forumlamax forumla AICM MCSE wi Log[f(X|Mi)] Log(BF) Odds Ratio P 2log(BF) 
w/ sym/clamBVII              
COI versus mt16S T2 0.01 −3,380 66 6,825 3.6 0.99 −3,393     
T1 −0.03 −3,397 40 6,834 2.2 0.01 −3,408 15 3.3 × 1006 0.00 30 
16S versus host T2 0.03 −5,837 76 11,750 4.1 1.00 −5,853     
T1 −0.02 −5,909 50 11,868 2.7 0.00 −5,919 66 4.6 × 1028 0.00 132 
23S versus host T2 0.02 −6,658 70 13,385 3.8 1.00 −6,679     
T1 −0.003 −6,761 51 13,572 2.8 0.00 −6,775 96 4.9 × 1041 0.00 192 
dnaK versus host T2 0.04 −5,429 64 10,922 3.5 1.00 −5,444     
T1 0.06 −5,659 46 11,363 2.5 0.00 −5,668 224 1.9 × 1097 0.00 448 
soxA versus host T2 0.006 −5,362 69 10,792 3.7 1.00 −5,376     
T1 −0.003 −5,547 49 11,143 2.7 0.00 −5,557 181 4.0 × 1078 0.00 362 
w/o sym/clamBVII              
COI versus mt16S T2 0.03 −3,361 57 6,779 3.8 1.00 −3,374     
T1 −0.01 −3,380 36 6,796 2.5 0.00 −3,391 17 2.4 × 1007 0.00 34 
16S versus host T2 −0.01 −5,815 66 11,696 3.6 1.00 −5,832     
T1 0.01 −5,842 46 11,730 2.5 0.00 −5,853 21 1.3 × 1009 0.00 42 
23S versus host T2 −0.01 −6,607 65 13,279 3.5 1.00 −6,620     
T1 0.004 −6,630 46 13,305 2.5 0.00 −6,638 18 6.6 × 1007 0.00 36 
dnaK versus host T2 0.03 −5,361 64 10,786 3.5 1.00 −5,377     
T1 −0.01 −5,407 47 10,861 2.6 0.00 −5,420 43 4.7 × 1018 0.00 86 
soxA versus host T2 0.01 −5,320 62 10,702 3.4 1.00 −5,332     
T1 0.02 −5,354 45 10,753 2.5 0.00 −5,367 35 1.6 × 1015 0.00 70 
All sym versus hoste T2 −0.003 −12,979 96 26,053 4.6 1.00 −12,997     
T1 0.03 −13,034 74 26,143 3.6 0.00 −13,058 61 3.1 × 1026 0.00 122 
a

Host loci: COI and mt16S concatenated (1170 bp).

b

AICM: thin = thinning interval; ρ1 = first-order autocorrelation of likelihood samples after thinning; forumlamax= estimate of the maximum likelihood of the data derived from posterior simulation; forumla = estimate of the effective number of parameters; AICM = 2forumlamax − 2forumla; wi = Akaike weights; see Tests for Symbiont–Host Coupling section of Methods and Materials for full details.

c

BF: log[f(X|Mi)] = marginal log-likelihood of the model; log(BF) = difference in support of the models in log-likelihood units; odds ratio = relative probability of the 2 models; P = probability of model; 2log(BF) = values >10 are considered very strong evidence against model T1 (Kass and Raftery 1995).

d

Model: T1 = coupling model that constrains both data sets in the comparison to a single topology; T2 = decoupling model that allows 2 separate topologies for the data sets being compared.

e

Symbiont loci (16S, 23S, dnaK, soxA) concatenated versus host loci (COI, mt16S).

We used MrBayes to compare the marginal likelihood of the data under models T1 and T2. Selection of best-fit nucleotide substitution models for each data partition (locus) was as described above and also involved a second level of assessment to avoid overparameterization and to aid convergence during MCMC runs (see supplementary table 3, Supplementary Material online). Under both T1 and T2, each data partition was assigned its own variable rate parameter (Dirichlet(1,1,1,1) prior). MCMC runs were conducted using Metropolis coupling as in the rooted single-gene phylogenetic analyses (above), with the following exception. For the 6-locus analysis, 5 chains (rather than 2) were iterated 5.0 × 106 times applying the same thinning and burn-in as above and then combined for a total of 20,000 posterior samples.

Statistical support for models T1 and T2 was assessed using Bayes factors (BF) computed from the harmonic mean of the likelihood scores over the MCMC run (Kass and Raftery 1995). However, the harmonic mean estimator is known to be computationally unstable (Newton and Raftery 1994; Nylander et al. 2004; Raftery et al. 2007), and some researchers have cautioned against its use in phylogenetic analyses (Lartillot and Philippe 2006). Due to this instability, we also applied an alternative model discrimination measure, the AICM (Raftery et al. 2007), using the program MrAICM (Young CR, unpublished data).

AICM is computed from the marginal posterior distribution of log-likelihood scores for the data and is defined as 

graphic
where forumlamax is an estimate of the ML of the data derived from posterior simulation and forumla is an estimate of the effective number of parameters (i.e., model complexity), also derived from posterior simulation. Posterior simulation–based estimates of these quantities are forumlamax = forumla+ s2 and forumla= 2s2, where forumla and s2 are the sample mean and variance of the log-likelihood scores, respectively (Raftery et al. 2007). Therefore, AICM is 
graphic
Raftery et al. (2007) point out that AICM is equivalent to the deviance information criterion as defined in Gelman et al. (2004). The Monte Carlo standard error (MCSE), given B independent MCMC draws, is 
graphic
(Raftery et al. 2007). Subsampling the MCMC data sets every 2–4 simulation replications, depending on the model and the data set, reduced autocorrelation enough so that our samples from the marginal posterior distribution of log-likelihood scores were independent.

We measure the relative support of competing models by computing the Akaike weights, wi, for each model i (Burnham and Anderson 2002). The best model in a set R of competing models is the one with the maximum AICM, AICMmax. The difference between the best model and other models that we wish to compare is Δ AICMi = AICMmax − AICMi, so that the Akaike weights for model i are defined as 

graphic
Symbiont–host congruence was further assessed using methods implemented in the program CONSEL (Shimodaira and Hasegawa 2001). Briefly, the unrooted ML tree for a given symbiont gene was generated via a heuristic search in PAUP using 100 random addition replicates, tree bisection-reconnection branch swapping, and best-fit substitution models chosen as described above (see supplementary table 3, Supplementary Material online, for best-fit models). The sitewise log-likelihoods for this tree were compared with those generated when the same sequence data were constrained to the ML topology of the host data (COI + mt16S). For the concatenated host data, ML analyses were run under a GTR model with variable rates assigned to each codon position of the COI gene and to the mt16S gene. Using sitewise log-likelihoods for both constrained and unconstrained topologies, CONSEL was implemented to calculate P values according to the approximately unbiased (AU) test using multiscale bootstrapping (Shimodaira 2002), as well as posterior probabilities (PPs) under the Bayesian information criterion (BIC) approximation (Schwarz 1978). P values correspond to the probability of obtaining a more extreme test statistic (the difference in log-likelihoods between constrained and unconstrained topologies) by chance under the null hypothesis of no difference between topologies. These analyses were also run reciprocally, that is, host data constrained to the ML topology for each symbiont gene.

Results

Sequence analysis of Vesicomya sp. mt-II clams from the JdF Ridge revealed a symbiont phylotype (symBVII) that was highly divergent from that previously described for this host lineage (symAVII). This discovery prompted the hypothesis that symBVII represented a decoupling of symbiont–host genetic variation and was acquired laterally (nonvertically) by the Vesicomya sp. mt-II host lineage. To test these hypotheses, we conducted a multilocus analysis that characterized the phylogenetic placement of symBVII and applied statistical tests that confirm symbiont–host decoupling.

Host Clam Phylogeny

Sequencing of 3 host loci (COI, mt16S, and nucITS) unambiguously identified the 3 host clams collected in 1991 from the JdF as members of the Vesicomya sp. mt-II lineage of the V. pacifica/lepta complex (see naming in Goffredi et al. 2003). Two individuals are identical across both mitochondrial loci and differ from the third individual at only 3 bp of the COI sequence (0.41%; supplementary table 5, Supplementary Material online). The third individual, which harbors the divergent symbiont lineage symBVII (see below), was identical across both mitochondrial loci to a clam species previously identified as Calyptogena pacifica (Vrijenhoek et al. 1994) and later reclassified as Vesicomya sp. mt-II (see GenBank AF008295 and AF035732 in Peek et al. (1998) and AY143320 in Goffredi et al. (2003)). We refer to the host haplotype of the first 2 individuals as clamAVII and that of the third individual as clamBVII. Bayesian analysis of the mitochondrial data (COI and mt16S concatenated) confirmed that clamAVII and clamBVII clustered in a monophyletic clade with the clam lineages Vesicomya sp. mt-I and -III from the eastern Pacific (PP = 0.97; fig. 1A). The relationship of nucITS sequences from clamAVII and clamBVII to those from Vesicomya sp. mt-I–, mt-II–, and mt-III–type clams characterized by Goffredi et al. (2003) is shown in figure 1B. ClamAVII and clamBVII were of the same host lineage for this nuclear gene and were identical to a previously sequenced Vesicomya sp. mt-II clam.

Fig. 1.—

Bayesian phylogenies for host loci of vesicomyid clams. PPs of taxon bipartitions are displayed if >0.50. Host trees based on (A) the mitochondrial COI and 16S (mt16S) genes (concatenated) from clams sequenced in this study, where taxa are named according to top blast hits (>99% nucleotide identity) and (B) the nuclear ITS region (nucITS) from host clams clamAVII and clamBVII sequenced in this study and clams of the Vesicomya sp. mt-I, -II, and -III lineages that were sequenced and named in Goffredi et al. (2003). Host clam haplotypes clamAVII and clamBVII are shaded and belong to the Vesicomya sp. mt-II lineage.

Fig. 1.—

Bayesian phylogenies for host loci of vesicomyid clams. PPs of taxon bipartitions are displayed if >0.50. Host trees based on (A) the mitochondrial COI and 16S (mt16S) genes (concatenated) from clams sequenced in this study, where taxa are named according to top blast hits (>99% nucleotide identity) and (B) the nuclear ITS region (nucITS) from host clams clamAVII and clamBVII sequenced in this study and clams of the Vesicomya sp. mt-I, -II, and -III lineages that were sequenced and named in Goffredi et al. (2003). Host clam haplotypes clamAVII and clamBVII are shaded and belong to the Vesicomya sp. mt-II lineage.

Symbiont Phylogeny

Symbiont genetic diversity in the 1991 JdF clams was initially assessed by direct sequencing of the bacterial 16S rRNA gene (16S; 1,303 bp). Genomic analyses indicate that the rRNA operon occurs in a single copy in vesicomyid symbionts (Kuwahara et al. 2007; Newton et al. 2007), and the chromatograms in our study showed no evidence of a mixed DNA template during direct sequencing. The symbiont 16S sequence from hosts bearing haplotype clamAVII belonged to the Vesicomya sp. mt-II symbiont lineage, differing from the published sequence (see Peek et al. 1998) at only 1 nucleotide (0.08% divergence); we designate this symbiont phylotype symAVII. In contrast, the 16S of the individual bearing haplotype clamBVII was highly divergent, differing from symAVII at 14 sites (1.1%; supplementary table 5, Supplementary Material online); we designate this divergent symbiont phylotype symBVII. The level of 16S divergence of symBVII fell within the range shown for symbionts of distinct vesicomyid host species and contrasted with prior results showing zero symbiont 16S variation among hosts of the same species (Peek et al. 1998). A BlastN search of the symBVII16S sequence returned a top hit not to the Vesicomya sp. mt-II symbiont (symAVII) but to the symbiont of a vesicomyid (Calyptogena phaseoliformis) from the Japan Trench in the western Pacific (99% identity).

Bayesian phylogenetic analysis of 16S sequences from a wide range of chemosynthetic symbionts and free-living bacteria showed that the 16S of symBVII fell in the monophyletic clade of vesicomyid symbionts with PP = 1.00 (fig. 2). Within this clade, the placement of the symBVII16S sequence is poorly resolved, as this sequence appears to have accumulated few changes relative to that of other vesicomyid symbionts (fig. 3). The 16S analysis also failed to resolve (PP < 0.5) relationships among several of the other symbiont taxa, particularly those within the clade containing Vesicomya sp. mt-I, -II, -III and C. ponderosa symbionts (figs. 2 and 3).

Fig. 2.—

Bayesian phylogeny for bacterial 16S rRNA genes from chemosynthetic vesicomyid clam symbionts, symbionts of other marine invertebrates, and closely related free-living bacteria and environmental clones (as identified by BlastN of the symBVII16S sequence). PPs of taxon bipartitions are displayed if >0.50. Symbionts are identified according to host taxon. Those sequences not obtained in this study are identified by GenBank accession numbers. Symbiont phylotypes symAVII and symBVII from Vesicomya sp. mt-II clams are shaded.

Fig. 2.—

Bayesian phylogeny for bacterial 16S rRNA genes from chemosynthetic vesicomyid clam symbionts, symbionts of other marine invertebrates, and closely related free-living bacteria and environmental clones (as identified by BlastN of the symBVII16S sequence). PPs of taxon bipartitions are displayed if >0.50. Symbionts are identified according to host taxon. Those sequences not obtained in this study are identified by GenBank accession numbers. Symbiont phylotypes symAVII and symBVII from Vesicomya sp. mt-II clams are shaded.

Fig. 3.—

Rooted Bayesian phylogenies for symbiont 16S rRNA, 23S rRNA, dnaK, and soxA genes from 11 vesicomyid clam taxa. PPs of taxon bipartitions are displayed if >0.50. Symbiont phylotypes symAVII and symBVII from Vesicomya sp. mt-II clams are shaded. Substitution rates in protein-coding loci are approximately 10× rates in rRNA genes (note differences in scale bars). Symbionts are identified according to host clam. Outgroups were chosen for each locus using BlastN to identify the most closely related free-living bacterium. Relative to 16S sequences, sequences of bacterial 23S, dnaK, and soxA genes are not well represented in GenBank. As such, choice of outgroup for these genes was constrained by availability, with the outgroup coming from whole genomes or large environmental contigs and generally falling outside of the extended 16S phylogeny shown in figure 2.

Fig. 3.—

Rooted Bayesian phylogenies for symbiont 16S rRNA, 23S rRNA, dnaK, and soxA genes from 11 vesicomyid clam taxa. PPs of taxon bipartitions are displayed if >0.50. Symbiont phylotypes symAVII and symBVII from Vesicomya sp. mt-II clams are shaded. Substitution rates in protein-coding loci are approximately 10× rates in rRNA genes (note differences in scale bars). Symbionts are identified according to host clam. Outgroups were chosen for each locus using BlastN to identify the most closely related free-living bacterium. Relative to 16S sequences, sequences of bacterial 23S, dnaK, and soxA genes are not well represented in GenBank. As such, choice of outgroup for these genes was constrained by availability, with the outgroup coming from whole genomes or large environmental contigs and generally falling outside of the extended 16S phylogeny shown in figure 2.

The unresolved placement of symBVII in the 16S tree prompted analyses of 3 additional symbiont genes, 23S, dnaK, and soxA (fig. 3). In contrast to that of the 16S, analysis of the 23S rRNA gene, which occurs in the same operon as the 16S gene, yielded a highly resolved vesicomyid symbiont phylogeny, with high PPs at most nodes. This analysis unambiguously placed symBVII as sister to the symbiont of an unnamed clam from the Logatchev vent site on the Mid-Atlantic Ridge, Vesicomya sp. MAR (PP = 1.00; also see Peek et al. (2000) for a characterization of the host COI of Vesicomya sp. MAR). Analysis of the housekeeping chaperone gene dnaK and the sulfur oxidation gene soxA confirmed the coupling of symBVII and the Vesicomya sp. MAR symbiont, with PPs of 1.00 and 0.97, respectively (fig. 3). Nucleotide identities between symBVII and Vesicomya sp. MAR symbionts for dnaK and soxA were 98.8% and 99.0%, respectively (supplementary table 5, Supplementary Material online). In contrast, the median pairwise nucleotide identity across all the symbionts in figure 3 was 90.2% for soxA and 86.5% for dnaK.

The discovery of symBVII led us to examine the frequency at which this phylotype occurs in other individuals. To do so, we sequenced the symbiont 16S from 115 additional clams collected from the JdF Ridge in 1995 and 1999 (table 1). The symBVII phylotype was detected in a single clam from the 1999 collection (Alvin dive 3459). Additional sequencing of 23S, dnaK, and soxA confirmed that the symbiont from this clam was identical to the divergent symBVII phylotype originally detected in the 1991 clam. Similarly, the host loci (COI, mt16S, nucITS) from this individual were identical to those of clamBVII (i.e., correspond to a Vesicomya sp. mt-II clam). The 16S sequences from the remaining 114 clams collected in 1995 and 1999 were identical to the 16S of symAVII, the phylotype previously characterized for the Vesicomya sp. mt-II host lineage. The frequency of symBVII-hosting clams in the sampled populations was therefore 2 in 118 (0.02).

Upon first finding symBVII in the original 1991 JdF clam, we took several measures to rule out the possibility that this sequence represents a contaminant (i.e., a free-living bacterium associated with the gill). These methods are detailed in Supplementary Material online. However, the question of contamination was effectively nullified by finding symBVII in a second clam from a distinct population (JdF 1999), as well as by our multilocus characterization that unambiguously placed symBVII as sister to another vesicomyid symbiont.

Tests for Symbiont–Host Coupling

The phylogenetic placement of symBVII with the symbiont of a distantly related host clam (Vesicomya sp. MAR) indicates a clear decoupling of symbiont and host mitochondrial data. To statistically test for symbiont–host coupling, the symbiont and host mitochondrial data were concatenated and a coupling model in which symbiont and host loci share the same topology, T1, was compared with a decoupling model in which symbiont and host loci are allowed to have separate topologies, T2. As shown by BF and Akaike weights (wi) computed using AICM, analyses of all 4 of the symbiont genes significantly rejected a coupling of symbiont and host topologies (table 2). Indeed, for each test, the decoupling model T2 was >1,000 times more likely than the coupled topology model (see Akaike weights in table 2). BIC and likelihood-based AU tests additionally confirmed this decoupling, showing for each symbiont gene a reciprocal rejection of symbiont and host mitochondrial topologies (table 3). Curiously, the Bayesian methods (BF, AICM, and BIC) also suggested that the topologies of the 2 host mitochondrial loci, COI and mt16S, may be decoupled (table 3 and supplementary table 6, Supplementary Material online). However, this pattern was not well supported by the AU tests, which failed to reciprocally reject incongruence between these loci (table 3).

Table 3

Reciprocal Tests for Congruence between ML Topologies of Symbiont and Host Loci

Comparisona  Diff-lnLb AUc BICd 
Locus 1 Locus 2 Ho = 2 Ho = 1 Ho = 2 Ho = 1 Ho = 2 Ho = 1 
w/ sym/clamBVII        
COI mt16S 36.2 4.2 0.001 0.243 2 × 10−16 0.014 
COI + mt16S 16S 151.5 47.7 3 × 10−82 0.003 2 × 10−66 2 × 10−21 
 23S 176.6 64.8 2 × 10−79 6 × 10−68 2 × 10−77 8 × 10−29 
 dnaK 180.2 218.9 7 × 10−114 8 × 10−57 5 × 10−79 8 × 10−96 
 soxA 230.7 126.4 7 × 10−40 8 × 10−05 7 × 10−101 1 × 10−55 
w/o sym/clamBVII        
COI mt16S 15.3 4.2 0.047 0.190 2 × 10−07 0.011 
COI + mt16S 16S 13.2 16.9 0.257 0.042 2 × 10−06 4 × 10−08 
 23S 16.9 10.5 0.084 0.042 5 × 10−08 3 × 10−05 
 dnaK 54.7 68.8 4 × 10−07 4 × 10−06 2 × 10−24 1 × 10−30 
 soxA 59.1 23.1 0.002 0.004 2 × 10−26 9 × 10−11 
Comparisona  Diff-lnLb AUc BICd 
Locus 1 Locus 2 Ho = 2 Ho = 1 Ho = 2 Ho = 1 Ho = 2 Ho = 1 
w/ sym/clamBVII        
COI mt16S 36.2 4.2 0.001 0.243 2 × 10−16 0.014 
COI + mt16S 16S 151.5 47.7 3 × 10−82 0.003 2 × 10−66 2 × 10−21 
 23S 176.6 64.8 2 × 10−79 6 × 10−68 2 × 10−77 8 × 10−29 
 dnaK 180.2 218.9 7 × 10−114 8 × 10−57 5 × 10−79 8 × 10−96 
 soxA 230.7 126.4 7 × 10−40 8 × 10−05 7 × 10−101 1 × 10−55 
w/o sym/clamBVII        
COI mt16S 15.3 4.2 0.047 0.190 2 × 10−07 0.011 
COI + mt16S 16S 13.2 16.9 0.257 0.042 2 × 10−06 4 × 10−08 
 23S 16.9 10.5 0.084 0.042 5 × 10−08 3 × 10−05 
 dnaK 54.7 68.8 4 × 10−07 4 × 10−06 2 × 10−24 1 × 10−30 
 soxA 59.1 23.1 0.002 0.004 2 × 10−26 9 × 10−11 
a

First column for each test is for locus 1 data constrained by the ML topology for locus 2 data (Ho = 2); second column for each test is for locus 2 data constrained by the ML topology for locus 1 (Ho = 1); bold indicates instances of reciprocal rejection of topological congruence (P < 0.05).

b

Diff-lnL = difference in log-likelihood between constrained and unconstrained topologies.

c

AU: P value of the AU test calculated from the multiscale bootstrap; P value = probability of obtaining a more extreme test statistic (diff-lnL) by chance under the null hypothesis of no difference between topologies.

d

BIC: PP calculated by the BIC approximation.

To determine whether symBVII provided the only information in the data rejecting symbiont–host coupling, we repeated the analysis excluding clamBVII and symBVII. Even without these sequences, the decoupling model T2 was >1,000 times more likely than the coupling model for each symbiont gene tested, as assessed via BF and AICM (table 2). Similarly, BIC PPs showed a reciprocal rejection of symbiont and host topologies in the absence of clamBVII and symBVII (table 3). The AU tests also provided evidence of decoupling, though congruence was reciprocally rejected for only 2 of the 4 symbiont loci (dnaK and soxA). Clearly, taxa other than symBVII contributed to the decoupling of symbiont and host phylogenies, as demonstrated by the unrooted symbiont and host mitochondrial topologies shown in figure 4. Notably, the mitochondrial data provided strong support (PP > 0.98) for the sister grouping of {C. ponderosa with C. magnifica} as well as {V. sp. mt-II and -III}, whereas the symbiont data grouped the symbionts of {C. ponderosa, V. sp. mt-I, -II, and -III} and {V. sp. mt-I and -III} (PP = 1.0 and 0.88, respectively; fig. 4). Together, the data showed that the symbiont phylogeny is incompatible with that of the host, a clear violation of the pattern predicted if symbionts and host mitochondria have undergone strict maternal cotransmission throughout the evolutionary history of this group.

Fig. 4.—

Unrooted Bayesian phylogenies for the analysis of 4 concatenated symbiont loci (16S, 23S, dnaK, and soxA) and 2 concatenated host mitochondrial loci (COI and mt16S) from 9 vesicomyid clam taxa. The divergent symbiont symBVII and its corresponding host haplotype (clamBVII) are excluded. Symbiont phylotype symAVII and host haplotype clamAVII (from a Vesicomya sp. mt-II clam) are shaded. PPs of taxon bipartitions for both the symbiont and the mitochondrial data are shown on each topology to illustrate incongruities between the data sets. Symbionts are identified according to host clam.

Fig. 4.—

Unrooted Bayesian phylogenies for the analysis of 4 concatenated symbiont loci (16S, 23S, dnaK, and soxA) and 2 concatenated host mitochondrial loci (COI and mt16S) from 9 vesicomyid clam taxa. The divergent symbiont symBVII and its corresponding host haplotype (clamBVII) are excluded. Symbiont phylotype symAVII and host haplotype clamAVII (from a Vesicomya sp. mt-II clam) are shaded. PPs of taxon bipartitions for both the symbiont and the mitochondrial data are shown on each topology to illustrate incongruities between the data sets. Symbionts are identified according to host clam.

Discussion

Symbiont–Host Decoupling Suggests Lateral Symbiont Acquisition

The results of this study suggest that symbionts, or mitochondria, have been transferred laterally (nonvertically) among hosts during the evolutionary history of vesicomyids. This study, interpreted alongside previous studies of the vesicomyid symbioses (Peek et al. 1998, Hurtado et al. 2003), suggests a mixed mode of symbiont transmission characterized by predominantly vertical transmission punctuated with instances of lateral symbiont acquisition. This conclusion is supported by two primary lines of evidence that reject the pattern of genetic coupling predicted if symbionts and mitochondria are strictly maternally cotransmitted (see Moran et al. 1993; Peek et al. 1998; Clark et al. 2000).

First, this study provides the first report of 2 similar symbiont lineages, symBVII and the Vesicomya sp. MAR symbiont, occurring in distantly related hosts. SymBVII co-occurs with the previously characterized symbiont phylotype, symAVII (Peek et al. 1998; Goffredi et al. 2003), in clams of the Vesicomya sp. mt-II host lineage from the JdF Ridge. However, symBVII is highly divergent from symAVII across rRNA and protein-coding genes, clustering instead with the symbiont of a distinct host lineage, Vesicomya sp. MAR from the Logatchev vent field on the Mid-Atlantic Ridge (fig. 3). Given the geographical isolation of the JdF (northeastern Pacific) from the Logatchev site (equatorial Atlantic), the grouping of symBVII and the MAR symbiont might seem curious. However, clustering of Atlantic and Pacific species has been shown repeatedly in vesicomyids (Peek et al. 1997, 1998, 2000; Kojima et al. 2004). Prior studies show the Vesicomya sp. MAR (Logatchev) host falling in a clade with clams from the northern Pacific, including unnamed species from the Aleutian Trench, C. kakoi from the Nankai Trough off the coast of Japan, and C. phaseoliformis from the Japan Trench (Peek et al. 2000; Kojima et al. 2004). Indeed, a BlastN search of the sym16S sequence of symBVII returns a top hit to the symbiont of C. phaseoliformis. It therefore seems likely that symBVII shares a most recent common ancestor with the symbiont of an as-yet unidentified Pacific clam, which may cluster within the {V. sp. MAR, C. phaseoliformis} clade. The current study was constrained to the available samples; broader taxonomic sampling may reveal the source clam or population from which symBVII originated. Nonetheless, the clustering of symBVII and MAR shown here represents a clear uncoupling of symbiont and host phylogenies, suggesting lateral transfer of symBVII into the Vesicomya sp. mt-II host lineage.

Second, Bayesian model discrimination methods (AICM, BF, BIC) show that even in the absence of the divergent symbiont lineage symBVII, the symbiont phylogeny is incompatible with that of the host (tables 2 and 3 and fig. 4). This decoupling is further supported for 2 symbiont genes by the likelihood-based AU test (table 3). These results are inconsistent with prior studies. Using a different set of taxa, Peek et al. (1998) showed that the clam mitochondrial topology was not significantly worse than the symbiont 16S topology at explaining symbiont diversification. This pattern was taken as strong evidence of cospeciation and of nearly complete vertical symbiont transmission in vesicomyids, though the authors hypothesize that minor discrepancies at 2 of the 7 internal nodes in their topologies might be explained by rare lateral symbiont transfer events. Subsequently, Hurtado et al. (2003) examined 80 individuals from populations of the clam C. magnifica and found no evidence of a decoupling of distinct symbiont and mitochondrial variants, as might be expected if occasional lateral symbiont transfer occurs between hosts. Together, these studies have led to a general acceptance of the hypothesis that vesicomyid symbionts undergo complete, or nearly complete, maternal transmission (e.g., see review by Papke and Ward 2004). However, detecting symbiont–host coupling likely depends on the taxa included in the study and on the resolution provided by the sequence data (Page 2003). For instance, our study included Calyptogena ponderosa, a taxon for which symbiont–host codivergence had not been previously assessed and whose phylogenetic placement in this study appears inconsistent between symbiont and host gene trees (fig. 4). Further, our study utilized a relatively large amount of symbiont sequence data (∼5 kb per taxon), yielding resolution beyond that provided by traditional analyses of the bacterial 16S rRNA gene.

The evidence outlined above argues against the strict maternal cotransmission of vesicomyid symbionts and host mitochondria. Rather, it appears that vesicomyid symbionts, though predominantly vertically transmitted (Peek et al. 1998; Hurtado et al. 2003), also experience occasional instances of lateral symbiont acquisition, with symBVII being the most obvious example. This mixed transmission strategy may be similar to that observed in associations between arthropods and the bacterium Wolbachia. As in this study, lateral transfer of cytoplasmically inherited Wolbachia has been inferred from the detection of highly similar Wolbachia strains in divergent host species, strong incongruence between symbiont and host phylogenies, and evidence of genomic exchange (recombination) between symbiont lineages (Werren et al. 1995; Schilthuizen and Stouthamer 1997; Haine et al. 2005; Baldo et al. 2006). For vesicomyids, the strong phylogenetic signal for symbiont–host codiversification shown in prior studies suggests that, relative to Wolbachia, lateral transfer events in vesicomyids may be relatively rare. Clearly, studies aimed at determining the modes and rates of lateral symbiont acquisition for this group are warranted. It is possible that different host lineages (or conversely, symbiont lineages) within the Vesicomyidae are differentially capable of lateral transmission. We now propose several hypothetical mechanisms by which lateral symbiont transfer could occur in vesicomyids (fig. 5). These include acquisition via 1) host hybridization, 2) horizontal transfer, and 3) environmental acquisition. We examine evidence for and against each of these hypotheses below, focusing explicitly on the acquisition event by which symBVII entered the Vesicomya sp. mt-II host lineage.

Fig. 5.—

Hypothetical mechanisms of lateral (nonvertical) symbiont acquisition in vesicomyid clam symbioses. (A) Host hybridization hypothesis. Symbionts (s) and/or mitochondria (m) are transmitted paternally via the sperm during hybridization between species 1 and 2. Paternal transmission may be occasional (leakage) or persistent via DUI, though neither has not been shown for vesicomyids. Potential outcomes of hybridization include the occurrence of multiple symbiont or mitochondrial lineages in the same host (heteroplasmy), or displacement, in which the paternal symbiont or mitochondrial lineage replaces the maternal lineage. (B) Horizontal transfer hypothesis. Symbionts are transmitted between contemporary species without hybridization. Potential transfer mechanisms include direct contact between symbiont-associated eggs, between eggs and host tissue (e.g., gill), and between host tissue and symbionts that have been released into the water column. (C) Environmental acquisition hypothesis. Symbionts are acquired directly from a stable free-living population of cells. Symbiont acquisition may involve displacement of the existing maternally transferred symbiont lineage, as multiple phylotypes have not been documented within the same host.

Fig. 5.—

Hypothetical mechanisms of lateral (nonvertical) symbiont acquisition in vesicomyid clam symbioses. (A) Host hybridization hypothesis. Symbionts (s) and/or mitochondria (m) are transmitted paternally via the sperm during hybridization between species 1 and 2. Paternal transmission may be occasional (leakage) or persistent via DUI, though neither has not been shown for vesicomyids. Potential outcomes of hybridization include the occurrence of multiple symbiont or mitochondrial lineages in the same host (heteroplasmy), or displacement, in which the paternal symbiont or mitochondrial lineage replaces the maternal lineage. (B) Horizontal transfer hypothesis. Symbionts are transmitted between contemporary species without hybridization. Potential transfer mechanisms include direct contact between symbiont-associated eggs, between eggs and host tissue (e.g., gill), and between host tissue and symbionts that have been released into the water column. (C) Environmental acquisition hypothesis. Symbionts are acquired directly from a stable free-living population of cells. Symbiont acquisition may involve displacement of the existing maternally transferred symbiont lineage, as multiple phylotypes have not been documented within the same host.

Potential Mechanisms of Lateral Symbiont Acquisition

Host Hybridization

SymBVII could have originated from a hybridization event between Vesicomya sp. mt-II and an as-yet unidentified clam. This hypothesis would necessarily involve paternal transfer of either symbionts or mitochondria via the sperm (fig. 5A). Such an event presumably would have been followed by displacement or loss of the existing symbiont or mitochondrial genome of the egg, as our sequence traces (and the results of prior studies; Kim et al. 1995; Peek et al. 1998; Goffredi et al. 2003) provide no evidence of multiple symbiont or mitochondrial phylotypes in the same host individual, that is, heteroplasmy. Several lines of evidence argue against a hybridization hypothesis. First, paternal transfer of either mitochondria or symbionts would be necessary to explain the incongruence between mitochondrial and symbiont genealogies observed for the vesicomyids in our study. Paternal symbiont transfer has not been reported in these clams, though this hypothesis has not been rigorously tested. Similarly, paternal mitochondrial transfer, via either occasional leakage (e.g., Kvist et al. 2003; Sherengul et al. 2006) or doubly uniparental inheritance (DUI, Zouros et al. 1994), has not been reported in vesicomyids. However, DUI has been identified in the venerid clam Tapes philippinarum and may be an ancestral feature of the Bivalvia (Passamonti and Scali 2001), suggesting the need for further study of this mitochondrial transmission mode in other clams. Second, sequences from 2 nuclear loci, nucITS (fig. 1) and histone (Hist2h3c1, 300 bp; data not shown), show no variation between host clams clamAVII and clamBVII, suggesting that clamBVII is not a hybrid.

Nevertheless, 2 factors suggest that the hybrid acquisition hypothesis deserves further attention. First, analysis of only 2 nuclear loci might not identify hybrids that have backcrossed for a sufficient number of generations. Indeed, introgression between Vesicomya sp. mt-II and an as-yet-unknown species might occur at a rate that would be sufficient for occasional horizontal transmission of symbionts but would not create strong disequilibrium between nuclear and cytoplasmic genes within the host lineage. Low rates of genetic introgression have been demonstrated for Drosophila pseudoobscura and Drosophila persimilis (Hey and Nielsen 2004), subspecies of chimpanzees (Won and Hey 2005), cichlid fishes (Hey et al. 2004), Heliconius butterflies (Bull et al. 2006), and possibly hydrothermal vent limpets (Johnson et al. 2006). Second, our results show topological incongruence between the 2 host mitochondrial loci used in the analysis (COI and mt16S; tables 2 and 3). Similarly, a low-level of COI–mt16S incongruence was demonstrated previously for some vesicomyid taxa (Peek et al. 2000). Such incongruence is suggestive of recombination, though other factors, including inadequate substitution models, may contribute to this pattern (Peek et al. 2000). If low levels of introgression occur between Vesicomya sp. mt-II and another species, then multilocus screening of host genes should also show introgression between the species. Clearly, analysis of additional nuclear loci, as well as broader taxonomic sampling of vesicomyid hosts and symbionts, is necessary to verify or rule out the hybridization hypothesis. However, the data currently available suggest that symBVII did not originate via hybridization.

Horizontal Transfer

SymBVII could have originated via horizontal host-to-host transfer that did not involve hybridization (fig. 5B). Horizontal transfer of cytoplasmically inherited endosymbionts has been observed repeatedly in associations between arthropods and the parasitic bacterium Wolbachia (e.g., Haine et al. 2005). The mechanism of Wolbachia transfer in nature is poorly understood but may involve shared interactions with common food substrates, predators, or parasites (Huigens et al. 2004; Sintupachee et al. 2006). In vesicomyid clams, a horizontal transfer hypothesis would involve both symbiont release from the donor clam and acquisition by the recipient species (Vesicomya sp. mt-II; fig. 5B). As in the host hybridization hypothesis, symbiont uptake would likely be followed by displacement of the existing symbiont population. Symbiont release could occur via attachment to the egg, as vesicomyid symbionts have been shown to occur at high density in the nutritive follicle cells surrounding the primary oocyte of the clam (Cary and Giovannoni 1993). Alternatively, symbionts could be released directly into the environment, perhaps from a dead or moribund clam, though there is no evidence to suggest this mechanism. Acquisition by the recipient clam would then involve contact with either egg-associated or freely dispersing symbionts, followed by symbiont internalization into the tissue of the new host. One could imagine this process occurring at the gamete stage if eggs from distinct species come into physical contact, thereby enabling symbiont exchange. Several aspects of this hypothesis, including the possibility of egg-to-egg transfer, may be experimentally testable and require further attention.

Environmental Acquisition

SymBVII could have been acquired directly from the environment from a free-living pool of symbionts (fig. 5C). Environmental acquisition has been suggested in other vent chemosynthetic symbioses, including mytilid mussels and vestimentiferan tube worms (Won et al. 2003; DeChaine et al. 2006; Nussbaumer et al. 2006; Vrijenhoek et al. 2007). Environmental acquisition in these taxa has been inferred from molecular evidence showing multiple symbiont phylotypes within a single host individual, as would be expected if symbionts are acquired from a diverse external population. Other studies of these taxa have used experimental or cytological data to directly show symbiont internalization or acquisition by the host (Kadar et al. 2005; Nussbaumer et al. 2006). However, similar evidence has not been reported for vesicomyids, and vesicomyid symbionts have not been found free living in the environment, though intensive searches for these bacteria in a free-living form have not been conducted. Further, accelerated substitution rates in vesicomyid symbionts are consistent with a maternal transmission strategy that inhibits gene exchange with a free-living bacterial pool (Peek, Vrijenhoek, and Gaut 1998). Recent genome sequencing of 2 vesicomyid symbionts suggests that the symbiont genome (∼1.1 Mb) has experienced some level of size reduction (Kuwahara et al. 2007; Newton et al. 2007). Genome reduction is well documented in other obligately endosymbiotic bacteria, including the gut bacteria of insects (size range: ∼0.42–0.65 Mb; Pérez-Brocal et al. 2006; Gómez-Valero et al. 2007). Although larger than that of other endosymbionts, the reduced genome of vesicomyid symbionts nonetheless may indicate adaptation to the intracellular environment. Such adaptation may inhibit growth outside the host, though this hypothesis requires further testing. Nevertheless, the available data argue against long-term persistence of a free-living vesicomyid symbiont pool.

Implications of Lateral Symbiont Acquisition in Vesicomyids

This study provides compelling evidence that symBVII was acquired laterally (i.e., nonmaternally). However, additional analyses, as well as broader taxonomic sampling, are necessary to address the mechanisms by which this could have occurred. Of the above hypotheses, environmental acquisition may be the most difficult to test experimentally, as definitive proof of this hypothesis would require recurring detection of vesicomyid symbionts in their free-living form. Hypotheses regarding acquisition via host hybridization or horizontal host-to-host transfer may warrant particular attention, as aspects of these hypotheses (e.g., paternal symbiont or mitochondrial transfer, egg-to-egg transfer) can be tested experimentally. Additional studies should also examine the temporal scale and frequency of lateral transfer. The presence of symBVII in Vesicomya sp. mt-II most likely resulted from a single lateral transfer event, followed by maintenance of symBVII at low frequency via vertical transmission. This is supported by the fact that the two clams hosting symBVII have exactly the same mitochondrial and symbiont haplotypes, suggesting a tight coupling of symbiont and mitochondrial genomes. Nonetheless, we cannot exclude the possibility that symBVII symbionts are being actively transferred into the Vesicomya sp. mt-II host lineage at low frequency, and the occurrence of this symbiont phylotype in two clams represents 2 lateral transfer events. Additional population-level sampling of symbiont and host mitochondrial and nuclear variation may resolve these 2 scenarios. Our results should encourage such studies, as lateral symbiont acquisition has important implications for understanding the evolution of vesicomyid symbioses.

Lateral symbiont acquisition may disrupt coadapted symbiont–host gene interactions. For most vesicomyids, vertical transmission has likely maintained a tight association between a clam and its symbiont over millions of years of evolution. Tight coevolution of this sort may impose severe constraints on the specificity of the symbiont–host interaction. For example, Edmands and Burton (1999) show that in hybrid lineages of Tigriopus copepods, introgression of the nuclear genotype from 1 population onto the cytoplasmic (mitochondrial) genotype of a divergent population causes reduced fitness and loss of function of major mitochondrial enzyme complexes. Hybridization presumably disrupts the strong positive epistatic interactions that normally occur between coadapted mitochondrial and nuclear gene products in these complexes (Ellison and Burton 2006). Similar molecular or physiological coadaptations may exist between the vesicomyid symbiont genome and the mitochondrial or nuclear genome of its host, particularly given the extent to which the clam host relies on its symbiont for nutrition (Fiala-Médioni et al. 1993). Our results are therefore notable, as they indicate that at least 1 host lineage (Vesicomya sp. mt-II) retains the ability to acquire divergent symbionts. This observation indicates that the specificity of the symbiont–host interaction is not absolute and suggests the possibility that a core set of genes may be conserved across symbiont lineages, thereby allowing the infection of hosts with nonspecific symbiont lineages. Indeed, recent analyses reveal high levels of synteny and gene content conservation in the genomes of the distantly related symbionts of C. okutannii and C. magnifica (Kuwahara et al. 2007; Newton et al. 2007; Newton ILG, Cavanaugh CM, unpublished data), perhaps indicating that genome reduction occurred prior to large-scale diversification in these symbioses. Additional experimental and genomic studies should focus on the extent to which genome structure and content impact the coadapted symbiont–clam interaction.

Lateral symbiont transmission may also create opportunities for the exchange of genomic material (recombination) between divergent symbiont strains (Baldo et al. 2006). Indeed, the intracellular insect parasite Wolbachia, which undergoes maternal transmission as well as occasional instances of lateral transmission (e.g., Haine et al. 2005), exhibits extensive recombination (Baldo et al. 2006). The detection of recent recombination in vesicomyid symbionts would suggest that certain genes are not functionally constrained by high levels of symbiont–host coadaptation and may be exchanged freely among divergent taxa. Further, recombination could impact genome evolution by attenuating the effects of Muller's ratchet observed in strictly vertically transmitted symbionts (Moran 1996). The impact of recombination, if it occurs, would depend on the rate of gene exchange among lineages. This rate might be low, as contact between divergent strains would likely be limited to instances of lateral symbiont transfer, which are presumably rare. Nonetheless, this study provides evidence that lateral symbiont transmission does occur in vesicomyids, indicating that the coevolutionary history of these clams and their internal symbionts is far less static and far more complex than previously thought.

Supplementary Material

Supplementary figures 1–3, supplementary tables 1–6 and supplementary material are available at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals.org/).

We thank Bob Vrijenhoek, Scott Edwards, and 2 anonymous reviewers for their generous and constructive input during the preparation of this manuscript. We also thank Pete Girguis, Zoe McCuddin, Chuck Fisher, and Bob Vrijenhoek for generously providing us with vesicomyid samples. This work was supported by National Science Foundation grants EF-0412205 and OCE-0453901 awarded to C.M.C. and by the Genetics and Genomics Training Program at Harvard University.

References

Akaike
H
A new look at the statistical model identification
IEEE Trans Automat Control
 , 
1974
, vol. 
19
 (pg. 
716
-
723
)
Baldo
L
Bordenstein
S
Wernegreen
JJ
Werren
JH
Widespread recombination throughout Wolbachia genomes
Mol Biol Evol
 , 
2006
, vol. 
23
 (pg. 
437
-
449
)
Bull
V
Beltran
M
Jiggins
CD
McMillan
WO
Bermingham
E
Mallet
J
Polyphyly and gene flow between non-sibling Heliconius species
BMC Biol
 , 
2006
, vol. 
4
 (pg. 
1
-
17
)
Burnham
KP
Anderson
DR
Model selection and multi model inference. A practical information-theoretic approach
 , 
2002
New York
Springer
Cary
SC
Giovannoni
SJ
Transovarial inheritance of endosymbiotic bacteria in clams inhabiting deep-sea hydrothermal vents and cold seeps
Proc Natl Acad Sci USA
 , 
1993
, vol. 
90
 (pg. 
5695
-
5699
)
Cavanaugh
CM
Chemoautotrophic bacteria in marine invertebrates from sulfide-rich habitats
Nature
 , 
1983
, vol. 
302
 (pg. 
58
-
61
)
Chen
X
Li
S
Aksoy
S
Concordant evolution of a symbiont with its host insect species: molecular phylogeny of genus Glossina and its bacteriome-associated endosymbiont, Wigglesworthia glossinidia
J Mol Evol
 , 
1999
, vol. 
48
 (pg. 
49
-
58
)
Clark
MA
Moran
NA
Baumann
PB
Wernegreen
JJ
Cospeciation between bacterial endosymbionts (Buchnera) and a recent radiation of aphids (Uroleucon) and pitfalls of testing for phylogenetic congruence
Evolution
 , 
2000
, vol. 
54
 (pg. 
517
-
525
)
DeChaine
EG
Bates
AE
Shank
TM
Cavanaugh
CM
Off-axis symbiosis found: characterization and biogeography of bacterial symbionts of Bathymodiolus mussels from Lost City hydrothermal vents
Environ Microbiol
 , 
2006
, vol. 
8
 (pg. 
1902
-
1912
)
Degnan
PH
Lazarus
AB
Brock
CD
Wernegreen
JJ
Host-symbiont stability and fast evolutionary rates in an ant-bacterium association: cospeciation of Camponotus species and their endosymbionts, Candidatus Blochmannia
Syst Biol
 , 
2004
, vol. 
53
 (pg. 
95
-
110
)
Di Meo
CA
Wilbur
AE
Holben
WE
Feldman
RA
Vrijenhoek
RC
Cary
SC
Genetic variation among endosymbionts of widely distributed vestimentiferan tubeworms
Appl Environ Microbiol
 , 
2000
, vol. 
66
 (pg. 
651
-
658
)
Edmands
S
Burton
RS
Cytochrome c oxidase activity in interpopulation hybrids of a marine copepod: a test for nuclear-nuclear or nuclear-cytoplasmic coadaptation
Evolution
 , 
1999
, vol. 
53
 (pg. 
1972
-
1978
)
Ellison
CK
Burton
RS
Disruption of mitochondrial function in interpopulation hybrids of Tigriopus californicus
Evolution
 , 
2006
, vol. 
60
 (pg. 
1382
-
1391
)
Endow
K
Ohta
S
Occurrence of bacteria in the primary oocytes of vesicomyid clam Calyptogena soyoae
Mar Ecol Prog Ser
 , 
1990
, vol. 
64
 (pg. 
309
-
311
)
Fiala-Médioni
A
Métivier
C
Ultrastructure of the gill of the hydrothermal vent bivalve Calyptogena magnifica, with a discussion of its nutrition
Mar Biol
 , 
1986
, vol. 
90
 (pg. 
215
-
222
)
Fiala-Médioni
A
Boulegue
J
Ohta
S
Felbeck
H
Mariotti
A
Source of energy sustaining the Calyptogena populations from deep trenches in subduction zones off Japan
Deep-Sea Res
 , 
1993
, vol. 
40
 
Pt 1
(pg. 
1241
-
1258
)
Friedrich
CG
Rother
D
Bardischewsky
F
Quentmeier
A
Fischer
J
Oxidation of reduced inorganic sulfur compounds by bacteria: emergence of a common mechanism
Appl Environ Microbiol
 , 
2001
, vol. 
67
 (pg. 
2873
-
2882
)
Gelman
A
Carlin
JB
Stern
HS
Rubin
DB
Bayesian data analysis
 , 
2004
2nd ed
London
Chapman and Hall
Gómez-Valero
L
Silva
FJ
Simon
JC
Latorre
A
Genome reduction of the aphid endosymbiont Buchnera aphidicola in a recent evolutionary time scale
Gene
 , 
2007
, vol. 
389
 (pg. 
87
-
95
)
Goffredi
SK
Hurtado
LA
Hallam
S
Vrijenhoek
RC
Evolutionary relationships of deep-sea vent and seep clams (Mollusca: vesicomyidae) of the ‘pacifica/lepta’ species complex
Mar Biol
 , 
2003
, vol. 
142
 (pg. 
311
-
320
)
Gupta
R
Protein phylogenies and signature sequences: a reappraisal of evolutionary relationships among Archaebacteria, Eubacteria, and Eukaryotes
Microbiol Mol Biol Rev
 , 
1998
, vol. 
62
 (pg. 
1435
-
1491
)
Haine
ER
Pickup
NJ
Cook
JM
Horizontal transmission of Wolbachia in a Drosophila community
Ecol Entomol
 , 
2005
, vol. 
30
 (pg. 
464
-
472
)
Hey
J
Nielsen
R
Multilocus methods for estimating population sizes, migration rates and divergence time, with applications to the divergence of Drosophila pseudoobscura and D. persimilis
Genetics
 , 
2004
, vol. 
167
 (pg. 
747
-
760
)
Hey
J
Won
YJ
Sivasundar
A
Nielsen
R
Markert
JA
Using nuclear haplotypes with microsatellites to study gene flow between recently separated Cichlid species
Mol Ecol
 , 
2004
, vol. 
13
 (pg. 
909
-
919
)
Huelsenbeck
JP
Ronquist
F
MrBayes: Bayesian inference of phylogenetic trees
Bioinformatics
 , 
2001
, vol. 
17
 (pg. 
754
-
755
)
Huigens
ME
Almeida
RP
de Poons
PAH
Luck
RF
Stouthamer
R
Natural interspecific and intraspecific horizontal transfer of parthenogenesis-inducing Wolbachia in Trichogramma wasps
Proc R Soc Lond B Biol Sci
 , 
2004
, vol. 
261
 (pg. 
509
-
515
)
Hurtado
LA
Mateos
M
Lutz
RA
Vrijenhoek
RC
Coupling of bacterial endosymbiont and host mitochondrial genomes in the hydrothermal vent clam Calyptogena magnifica
Appl Environ Microbiol
 , 
2003
, vol. 
69
 (pg. 
2058
-
2064
)
Johnson
SB
Young
CR
Jones
WJ
Waren
A
Vrijenhoek
RC
Migration, isolation, and speciation of hydrothermal vent limpets (Gastropoda; Lepetodrilidae) across the Blanco transform fault
Biol Bull
 , 
2006
, vol. 
210
 (pg. 
140
-
157
)
Kadar
E
Bettencourt
R
Costa
V
Santos
RS
Lobo-Da-Cunha
A
Dando
P
Experimentally induced endosymbiont loss and re-acquirement in the hydrothermal vent bivalve Bathymodiolus azoricus
J Exp Mar Biol Ecol
 , 
2005
, vol. 
318
 (pg. 
99
-
110
)
Kass
RE
Raftery
AE
Bayes factors
J Am Stat Assoc
 , 
1995
, vol. 
90
 (pg. 
773
-
795
)
Kennish
MJ
Lutz
RA
The hydrothermal vent clam, Calyptogena magnifica (Boss and Turner, 1980): a review of existing literature
Rev Aquat Sci
 , 
1992
, vol. 
6
 (pg. 
29
-
66
)
Kiel
S
Little
CTS
Cold-seep mollusks are older than the general marine mollusk fauna
Science
 , 
2006
, vol. 
313
 (pg. 
1429
-
1431
)
Kim
YW
Yasuda
M
Yamagishi
A
Oshima
T
Ohta
S
Characterization of the endosymbiont of a deep-sea bivalve, Calyptogena soyae
Appl Environ Microbiol
 , 
1995
, vol. 
61
 (pg. 
823
-
827
)
Kojima
S
Fujikura
K
Okutani
T
Multiple trans-Pacific migrations of deep-sea vent/seep-endemic bivalves in the family Vesicomyidae
Mol Phylogenet Evol
 , 
2004
, vol. 
32
 (pg. 
396
-
406
)
Krueger
DM
Cavanaugh
CM
Phylogenetic diversity of bacterial symbionts of Solemya hosts based on comparative sequence analysis of 16S rRNA genes
Appl Environ Microbiol
 , 
1997
, vol. 
63
 (pg. 
91
-
98
)
Krylova
EM
Sahling
H
Recent bivalve molluscs of the genus Calyptogena (Vesicomyidae)
J Molluscan Stud
 , 
2006
, vol. 
72
 (pg. 
359
-
395
)
Kuwahara
H
Yoshida
T
Takaki
Y
, et al.  . 
(10 co-authors)
Reduced genome of the thioautotrophic intracellular symbiont in a deep-sea clam, Calyptogena okutanii
Curr Biol
 , 
2007
, vol. 
17
 (pg. 
881
-
886
)
Kvist
L
Martens
J
Nazarenko
AA
Orell
M
Paternal leakage of mitochondrial DNA in the great tit (Parus major)
Mol Biol Evol
 , 
2003
, vol. 
20
 (pg. 
243
-
247
)
Lartillot
N
Philippe
H
Computing Bayes factors using thermodynamic integration
Syst Biol
 , 
2006
, vol. 
55
 (pg. 
195
-
207
)
Lo
N
Bandi
C
Watanabe
H
Nalepa
C
Beninati
T
Evidence for cocladogenesis between diverse Dictyopteran lineages and their intracellular endosymbionts
Mol Biol Evol
 , 
2003
, vol. 
20
 (pg. 
907
-
913
)
Maddison
DR
Maddison
WP
MacClade 4: analysis of phylogeny and character evolution
 , 
2000
Sunderland (MA)
Sinauer Associates
Moran
NA
Accelerated evolution and Muller's ratchet in endosymbiotic bacteria
Proc Natl Acad Sci USA
 , 
1996
, vol. 
93
 (pg. 
2873
-
2878
)
Moran
NA
Munson
MA
Baumann
P
Ishikawa
H
A molecular clock in endosymbiotic bacteria is calibrated using the insect hosts
Proc R Soc Lond B Biol Sci
 , 
1993
, vol. 
253
 (pg. 
167
-
171
)
Nelson
K
Fisher
C
Absence of cospeciation in deep-sea vestimentiferan tube worms and their bacterial endosymbionts
Symbiosis
 , 
2000
, vol. 
28
 (pg. 
1
-
15
)
Newton
MA
Raftery
AE
Approximate Bayesian inference by the weighted likelihood bootstrap
J R Stat Soc B
 , 
1994
, vol. 
56
 (pg. 
3
-
48
)
Newton
ILG
Woyke
T
Auchtung
TA
, et al.  . 
(16 co-authors)
The Calyptogena magnifica chemoautotrophic symbiont genome
Science
 , 
2007
, vol. 
315
 (pg. 
998
-
1000
)
Nussbaumer
AD
Fisher
CR
Bright
M
Horizontal endosymbiont transmission in hydrothermal vent tubeworms
Nature
 , 
2006
, vol. 
441
 (pg. 
345
-
348
)
Nylander
JAA
Ronquist
F
Huelsenbeck
JP
Nieves-Aldrey
JL
Bayesian phylogenetic analysis of combined data
Syst Biol
 , 
2004
, vol. 
53
 (pg. 
47
-
67
)
Page
RDM
Tangled trees: phylogeny, cospeciation, and coevolution
 , 
2003
Chicago (IL)
University of Chicago Press
Papke
RT
Ward
DM
The importance of physical isolation to microbial diversification
FEMS Microbiol Ecol
 , 
2004
, vol. 
48
 (pg. 
293
-
303
)
Passamonti
M
Scali
V
Gender-associated mitochondrial DNA heteroplasmy in the venerid clam Tapes philippinarum (Mollusca Bivalvia)
Curr Genet
 , 
2001
, vol. 
39
 (pg. 
117
-
124
)
Peek
AS
Gaut
BS
Feldman
RA
Barry
JP
Kochevar
RE
Lutz
RA
Vrijenhoek
RC
Neutral and nonneutral mitochondrial genetic variation in deep-sea clams from the family Vesicomyidae
J Mol Evol
 , 
2000
, vol. 
50
 (pg. 
141
-
153
)
Peek
AS
Gustafson
RG
Lutz
RA
Vrijenhoek
RC
Evolutionary relationships of deep-sea hydrothermal vent and cold-water seep clams (Bivalvia: vesicomyidae): results from the mitochondrial cytochrome oxidase subunit I
Mar Biol
 , 
1997
, vol. 
130
 (pg. 
151
-
161
)
Peek
AS
Feldman
RA
Lutz
RA
Vrijenhoek
RC
Cospeciation of chemoautotrophic bacteria and deep-sea clams
Proc Natl Acad Sci USA
 , 
1998
, vol. 
95
 (pg. 
9962
-
9966
)
Peek
AS
Vrijenhoek
RC
Gaut
BS
Accelerated evolutionary rate in sulfur-oxidizing endosymbiotic bacteria associated with the mode of symbiont transmission
Mol Biol Evol
 , 
1998
, vol. 
15
 (pg. 
1514
-
1523
)
Pérez-Brocal
V
Gil
R
Ramos
S
Lamelas
A
Postigo
M
Michelena
JM
Silva
FJ
Moya
A
Latorre
A
A small microbial genome: the end of a long symbiotic relationship?
Science
 , 
2006
, vol. 
314
 (pg. 
312
-
313
)
Posada
D
Crandall
KA
Selecting the best fit model of nucleotide substitution
Syst Biol
 , 
2001
, vol. 
50
 (pg. 
580
-
601
)
R Development Core Team
R: a language and environment for statistical computing
 , 
2007
Vienna (Austria)
R Foundation for Statistical Computing
Raftery
AE
Newton
MA
Satagopan
JM
Krivitsky
PN
Bernardo
JM
Bayarri
MJ
Berger
JO
Dawid
AP
Heckerman
D
Smith
AFM
West
M
Estimating the integrated likelihood distribution via posterior simulation using the harmonic mean identity
Bayesian statistics 8
 , 
2007
Oxford (UK)
Clarendon Press
(pg. 
1
-
45
)
Schilthuizen
M
Stouthamer
R
Horizontal transmission of parthenogenesis-inducing microbes in Trichogramma wasps
Proc R Soc Lond B Biol Sci
 , 
1997
, vol. 
264
 (pg. 
361
-
366
)
Schwarz
G
Estimating the dimension of a model
Ann Stat
 , 
1978
, vol. 
6
 (pg. 
461
-
464
)
Sherengul
W
Kondo
R
Matsuura
ET
Analysis of paternal transmission of mitochondrial DNA in Drosophila
Genes Genet Syst
 , 
2006
, vol. 
81
 (pg. 
399
-
404
)
Shimodaira
H
An approximately unbiased test of phylogenetic tree selection
Syst Biol
 , 
2002
, vol. 
51
 (pg. 
492
-
508
)
Shimodaira
H
Hasegawa
M
CONSEL: for assessing the confidence of phylogenetic tree selection
Bioinformatics
 , 
2001
, vol. 
17
 (pg. 
1246
-
1247
)
Sintupachee
S
Milne
JR
Poonchaisri
S
Baimai
V
Kittayapong
P
Closely related Wolbachia strains within the pumpkin arthropod community and the potential for horizontal transmission via the plant
Microb Ecol
 , 
2006
, vol. 
51
 (pg. 
294
-
301
)
Stewart
FJ
Newton
ILG
Cavanaugh
CM
Chemosynthetic endosymbioses: adaptations to oxic-anoxic interfaces
Trends Microbiol
 , 
2005
, vol. 
13
 (pg. 
439
-
448
)
Suzuki
Y
Kojima
S
Watanabe
H
, et al.  . 
(10 co-authors)
Single host and symbiont lineages of hydrothermal vent gastropods Ifremeria nautilei (Provannidae): biogeography and evolution
Mar Ecol Prog Ser
 , 
2006
, vol. 
315
 (pg. 
167
-
175
)
Swofford
DL
Paup*. Phylogenetic analysis using parsimony (*and other methods)
 , 
1998
Sunderland (MA)
Sinauer
Tavaré
S
Some probabilistic and statistical problems on the analysis of DNA sequences
Lect Math Life Sci
 , 
1986
, vol. 
17
 (pg. 
57
-
86
)
Thompson
JD
Higgins
DG
Gibson
TJ
ClustalW: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice
Nucleic Acids Res
 , 
1994
, vol. 
22
 (pg. 
4673
-
4680
)
van Ham
RCHJ
Kamerbeek
J
Palacios
C
, et al.  . 
(16 co-authors)
Reductive genome evolution in Buchnera aphidicola
Proc Natl Acad Sci USA
 , 
2003
, vol. 
100
 (pg. 
581
-
586
)
Vrijenhoek
RC
Schutz
SJ
Gustafson
RG
Lutz
RA
Cryptic species of deep-sea clams (Mollusca, Bivalvia, Vesicomyidae) in hydrothermal vent and cold-seep environments
Deep Sea Res II
 , 
1994
, vol. 
41
 (pg. 
1171
-
1189
)
Vrijenhoek
RC
Duhaime
M
Jones
WJ
Subtype variation among bacterial endosymbionts of tubeworms (Annelida: siboglinidae) from the Gulf of California
Biol Bull
 , 
2007
, vol. 
212
 (pg. 
180
-
184
)
Werren
JH
Zhang
W
Guo
LR
Evolution and phylogeny of Wolbachia: reproductive parasites of arthropods
Proc R Soc Lond B Biol Sci
 , 
1995
, vol. 
261
 (pg. 
55
-
71
)
Wade
MJ
The co-evolutionary genetics of ecological communities
Nat Rev Genet
 , 
2007
, vol. 
8
 (pg. 
185
-
195
)
Won
YJ
Hey
J
Divergence population genetics of chimpanzees
Mol Biol Evol
 , 
2005
, vol. 
22
 (pg. 
297
-
307
)
Won
YJ
Hallam
SJ
O'Mullan
GD
Pan
IL
Buck
KR
Vrijenhoek
RC
Environmental acquisition of thiotrophic endosymbionts by deep-sea mussels of the genus Bathymodiolus
Appl Environ Microbiol
 , 
2003
, vol. 
69
 (pg. 
6785
-
6792
)
Yang
Z
Maximum likelihood phylogenetic estimation from DNA sequences with variable rates over sites: approximate methods
J Mol Evol
 , 
1994
, vol. 
39
 (pg. 
306
-
314
)
Zouros
E
Ball
AO
Saavedra
C
Freeman
KR
An unusual type of mitochondrial DNA inheritance in the blue mussel Mytilus
Proc Natl Acad Sci USA
 , 
1994
, vol. 
91
 (pg. 
7463
-
7467
)

Author notes

1
F.J.S. and C.R.Y. contributed equally to this work.
Edward Holmes, Associate Editor