Species boundaries of Gulf of Mexico vestimentiferans (Polychaeta, Siboglinidae) inferred from mitochondrial genes

vestimentiferans (Polychaeta, Siboglinidae) inferred mitochondrial At least six morphospecies of vestimentiferan tubeworms are associated with cold seeps in the Gulf of Mexico (GOM). The physiology and ecology of the two best-studied species from depths above 1000 m in the upper Louisiana slope ( Lamellibrachia luymesi and Seepiophila jonesi ) are relatively well understood. The biology of one rare species from the upper slope (escarpiid sp. nov.) and three morphospecies found at greater depths in the GOM ( Lamellibrachia sp. 1, L. sp. 2, and Escarpia laminata ) are not as well understood. Here we address species distributions and boundaries of cold-seep tubeworms using phylogenetic hypotheses based on two mitochondrial genes. Fragments of the mitochondrial large ribosomal subunit rDNA (16S) and cytochrome oxidase subunit I (COI) genes were sequenced for 167 vestimentiferans collected from the GOM and analyzed in the context of other seep vestimentiferans for which sequence data were available. The analysis supported ﬁve monophyletic clades of vestimentiferans in the GOM. Intra-clade variation in both genes was very low, and there was no apparent correlation between the within-clade diversity and collection depth or location. Two of the morphospecies of Lamellibrachia from different depths in the GOM could not be distinguished by either mitochondrial gene. Similarly, E. laminata could not be distinguished from other described species of Escarpia from either the west coast of Africa or the eastern Paciﬁc using COI. We suggest that the mitochondrial COI and 16S genes have little utility as barcoding markers for seep vestimentiferan tubeworms. mitochondrial large ribosomal subunit rDNA gene and mitochondrial cytochrome oxidase 1 gene of


Introduction
For the better part of the last century, marine biologists assumed oceans were largely interconnected by currents that enabled larvae and propagules to reach distant shores and assure gene flow even over great distances. More recently, the use of molecular tools has challenged assumptions regarding population structure and speciation in the ocean and demonstrated that marine animals often have genetically distinct populations despite geographic proximity (Palumbi and Warner, 2003). Although sharp genetic breaks between close populations have been recorded throughout the ocean, most of what is known about speciation patterns and phylogeography has been inferred from shallow-water and coastal systems, which represent only about 15% of the aquatic environment. Thus, our knowledge of processes that lead to population divergence and speciation in the open ocean is relatively limited (Thornhill et al., 2008, and references therein;Zardus et al., 2006).
Vestimentiferan tubeworms, which include 10 genera in the polychaete family Siboglinidae (Halanych et al., 2001;Kojima et al., 2002;McMullin et al., 2003;Rouse, 2001), are abundant at deep-sea hydrothermal vents and cold seeps at depths ranging from 80 to 9345 m (Cordes et al., 2007b;Mironov, 2000;Miura et al., 2002). In the deep Gulf of Mexico, six morphospecies have been reported (Cordes et al., 2009). Two described species, Lamellibrachia luymesi (van der Land and Narrevang, 1975) and Seepiophila jonesi (Gardiner et al., 2001), are relatively well studied, and their ecology and physiology are well understood (Bergquist et al., 2002;Cordes et al., 2007a, b). They occur on the upper Louisiana slope at between 500 and 950 m depth and occasionally co-occur with a rare undescribed species, escarpiid sp. nov. The three other morphological species are found on the lower Louisiana slope at depths greater than about 950 m (Lamellibrachia sp. 1, L. sp. 2, and Escarpia laminata).
In this paper, we present phylogenetic hypotheses based on the mitochondrial large ribosomal subunit rDNA gene (16S) and mitochondrial cytochrome oxidase 1 gene (COI) of over 200 vestimentiferans (sequenced for either or both genes) including 180 individuals from the six morphospecies that occur in the Gulf of Mexico. Phylogenetic trees are used to examine the distribution of vestimentiferans in the Gulf of Mexico and their relations to other vestimentiferans around the world. We examined the concordance between the morphological and phylogenetic data to identify differences between the genealogical and morphological species analyzed. Finally, we compared between-and within-species 16S and COI genetic distances and show that these two mitochondrial genes have little utility as ''barcoding molecules'' for vestimentiferans.

Collection of material
Vestimentiferans were collected in the deep Gulf of Mexico from 1 2s i teso ntwocr ui sesin2 006and 2007, using the DSV ALVIN and R.V. Atlantis in 2006 and ROV JASON II and the NOAA ship Ronald Brown in 2007 (see Fig. 1). Vestimentiferans were collected using either the Bushmaster Jr. collection device (for samples destined also for community ecology analyses, see Cordes et al., 2010)o rt h e submersible manipulators and placed directly into a collection box. Aboadship, all vestimentiferans were identified using morphological criteria, and subsamples of vestimentum tissue were frozen for subsequent analyses at the Pennsylvania State University. Additional frozen vestimentiferan tissue samples collected previously from shallower sites on the upper Louisiana slope using the DSV JOHNSON SEA LINK were also analyzed for this study (see Table 1 for a complete list of specimens).

DNA sequencing
DNA was extracted either by boiling a small amount of frozen tissue in 600 mL of 10% Chelex solution (Bio-Rad) or using a CTAB+ PVP method modified from Doyle and Doyle (1987), followed by a standard ethanol precipitation.
A 524 bp fragment of the mitochondrial 16S gene was amplified using primers 16Sar and 16Sbr (Kojima et al., 1995). A 689 bp fragment of the mitochondrial gene COI was amplified using the primers HCO and LCO (Folmer et al., 1994). Amplification was performed under the following PCR conditions: 94 1C( 1 m i n ) ; 50 1C (2 min); and 72 1C( 2 . 5m i n )f o r3 0c y c l e s .A l lP C Rr e a c t i o n s were performed using 0.5 ml of each primer, 2.5 mlo f1 0 X B u f f e r , 2 mlo f1 0mMd N T P s ,0 . 2mlo ft a q ,1 6 . 5mlo fw a t e r ,a n d3mlo f template. The PCR product was first purified with the ExoSap-it protocol (USB, Affimetrix) and then run on a 2% agarose gel stained with ethidium bromide to enable us to check the quantity and quality of the product. The purified PCR product was used as a template for double-stranded sequencing that was carried out at the Pennsylvania State University Sequencing Core Facility, University Park, Pennsylvania, using ABI 3730 sequencer machines.

Phylogenetic analysis
Sequences were first assembled and edited using Geneious Pro 4.0.4 (Biomatters Ltd.), and then aligned using ClustalX (Thompson et al., 2002). All alignments were confirmed and edited visually in MacClade 4.06 OS X (Maddison and Maddison, 2000) to insure that indel variation was aligned consistently among all sequenced genes.
Phylogenetic analyses of the aligned sequences were conducted using the maximum parsimony (MP) optimality criterion and neighbor joining (Saitou and Nei, 1987) (NJ) in PAUP n version 4.0b10 for Macintosh (Wilgenbusch and Swofford, 2003), and the maximum likelihood (ML) optimality criterion in GARLI v0.951.OsX-GUI (Zwickl, 2006) and PhylML (Guindon and Gascuel, 2003). The best-fit model used in PhyML and PAUP n was assessed using the akaike information criterion as implemented in modeltest (Posada, 2003;Posada and Crandall, 1998). The best-fit model was (HKY +I + G) for the COI dataset and (GTR+G) for the 16S dataset. Clade stability was assessed by ML bootstrap analysis (Felsenstein, 1985) in GARLI (100 bootstrap replicates) and NJ (1000 replicates) in PAUP n . The ML analyses in GARLI were performed using random starting trees and default termination conditions. Within-and between-species distances were estimated in MEGA 4 (Tamura et al., 2007).

Results
The complete COI dataset includes 146 sequences (Table 1)o f the six Gulf of Mexico (GOM) cold-seep morphospecies, the available GenBank sequences of E. southwardae, E. spicata, and assorted Lamellibrachia species from around the world. Sequences from the hydrothermal vent-dwelling genera Riftia, Oasisia, Tevnia, and Arcovestia were used as outgroups. We restricted our analyses to   Table 1 for the complete list of samples), 127 of which were from the Gulf of Mexico. Sequences from the vent-dwelling genera Tevnia and Ridgeia were used as outgroups. The aligned 16S dataset consisted of 524 bp, of which 433 were invariant sites, 72 were phylogenetically informative, and 19 were autapomorphies. MP, ML, and NJ analyses produced congruent trees, and the GARLI ML phylogeny is presented in Fig. 2 A and B and 3A and B.
Both 16S and COI phylogenies identify five distinct monophyletic clades of vestimentiferans in the Gulf of Mexico. Four of the clades represent single morphospecies, S. jonesi, E. laminata, Lamellibrachia sp. 2, and escarpiid sp. nov., from the upper slope. However, the fifth clade includes both Lamellibrachia sp. 1 from the collections in the deeper GOM and L. luymesi from the upper Louisiana slope sites. They were, therefore, considered a single species when within-and between-species distances for the 16S and COI datasets were estimated. Additionally, COI sequences of E. laminata did not differ from those of E. spicata and

Distribution of vestimentiferan species in the Gulf of Mexico and relation to other seep species
Vestimentiferans have been collected from both hydrothermal vent and cold-seep sites. The vent and seep species fall into two different clades. However, it should be noted that ''seep species'' are sometimes found in sedimented hydrothermal vent areas with low levels of diffuse flow, and that ''cold-seep'' fluids may have temperatures elevated over background (Black et al., 1998;Kojima et al., 1997;MacDonald et al., 2000;Joye et al., 2005); so this separation really reflects more aspects of their habitat than temperature alone. Vestimentiferans found at cold seeps worldwide can be further divided into two clades. One clade includes at least five named and three unnamed species in the genus Lamellibrachia. The other clade includes three named species in the genus Escarpia, S. jonesi, Paraescarpia echinospica, and a rarely collected species (escarpiid sp. nov.) from the shallow GOM. Although Arcovestia seems basal to the Lamellibrachia clade (Fig. 2B), this position is not well supported.
Three species in the escarpiid clade of seep vestimentiferans are found in the GOM: S. jonesi has been collected from numerous sites, 2A Fig. 2. COI maximum likelihood (ML) tree. Outgroups are shown in italics, and bootstrap support above 50% (NJ 1000 replicates) is indicated below each node. All new sequences are preceded by a number, followed by the abbreviation for the seep site or lease block from which they were collected. VK ¼ Viosca Knoll, BH¼ Bush Hill, and BP ¼Brine Pool. Those sites, together with GB425, GC234, and GC354, are from the upper Louisiana slope of the GOM ( o 800 m depth). All other lease blocks are on the lower slope (Fig. 1).  Table 3 Between-and within-species (in bold, on diagonal) p distances for the COI gene.
[ ranging in depth from 500 to 950 m; escarpiid sp. nov. from two sites ranging in depth from 600 to 640 m, where it co-occurs with S. jonesi (although it has been reported also from GC234 at 525 m; see Cordes et al., 2003); and E. laminata from 950 to 3200 m depth. S. jonesi and E. laminata co-occurred at only one site, GB647, at a depth of 950 m. The undescribed escarpiid differs morphologically from S. jonesi, as it lacks the curl of the ventral vestimental fold that is a defining character of the genus Seepiophila (Gardiner et al., 2001). Additionally, the obturacular process of the undescribed escarpiid forms a spike, whereas it is flat in S. jonesi and barely protrudes from the top of the obturaculum. Both the COI and 16S phylogenetic trees distinguish these three species and place them within the escarpiid clade of seep vestimentiferans (Figs. 2 and 3). Both the 16S tree and the 16S p distance matrix suggest E. laminata is more closely related to S. jonesi (between-species uncorrected p ¼2%) than to the undescribed escarpiid (between-species uncorrected p ¼3.50%). However, the COI tree groups the undescribed escarpiid with the described Escarpia spp. The bootstrap value based on COI data supporting this clade is low (61%), and the grouping observed for the 16S dataset has a bootstrap below 50%. Neither tree allows us to state clearly whether this new escapiid is more closely related to Escarpia, Paraescarpia,o rSeepiophila.
As previously noted by other authors, COI does not separate Escarpia southwardae, E. spicata, and E. laminata, respectively, from cold seeps on the west coast of Africa in the eastern Atlantic, Guaymas Basin, off the coast of Mexico, and the GOM (Black et al., 1998). Also, there is very little to no intra-clade diversity within this group (Table 3). This result may indicate that those three nominal species represent a single genealogical species with a surprisingly wide geographic distribution and variable morphology. However, this assumption would require a high level of gene flow between quite distant localities, especially since the closing of the Isthmus of Panama 3.5 million years ago, followed the closing of the deep sea exchange 10 million years ago (Burton et al., 1997). This level of genetic exchange over these distances seems quite unlikely, considering what is known about larval development times for vestimentiferans (Marsh et al. 2001, Young et al., 1996. Although the life span of Escarpia larvae has not been determined, the larval life span of the vent species Riftia pachyptila is estimated at about three weeks (Marsh et al., 2001) and the larval life span of the seep vestimentiferan L. luymesi is estimated to be about one month (Young et al., 1996). Tyler and Young (1999) estimate that the maximal dispersal distances for these species are on the order of 60 km per generation, which is unlikely to support the level of genetic mixing necessary to maintain genetic homogeneity among the three described species of Escarpia from such widely separated geographic locations. It is possible, however, that undiscovered seeps around South America could connect all of these species.
The lack of fixed COI differences within Escarpia spicata, E. laminata, and E. southwardae could also be due to different rates of evolution of the COI gene in different taxa. COI has been used for higher level phylogenetic reconstructions in other groups of annelids (Halanych and Janosik, 2006) and has been adopted as an appropriate gene for the ''barcode of life'' for animals in general by the barcode of life initiative (BOLI; http://www.dnabarcodes. org/). However, the fact that COI fails to identify morphologically distinct populations of Escarpia from such widely separated areas implies that in this clade the mutation rate may be considerably slower than in other lineages. Slower rates of evolution in the mitochondrial DNA have been recognized in some other groups, such as the cnidarian class Anthozoa, where this phenomenon has been linked to an especially efficient repair system of their mitochondrial DNA (France and Hoover, 2002;Pont-Kingdon et al., 1998); however, no evidence of a similar system has been found in vestimentiferan mitochondrial DNA. Seep vestimentiferans can also be extremely long-lived (Bergquist et al., 2000;Cordes et al., 2007a), which may contribute to a slower rate of change of mitochondrial DNA (see for example Nabholz et al. (2008) for a consideration of longevity effects on mitochondrial rates of evolution in vertebrates).
In the COI dataset, the Lamellibrachia clade is divided into eight distinct groups that represent presumptive species, including five basal species (L. juni, L. barhami, L. satsuma, L. sp. Japan, and L.s p . West Pacific), all of which are from the Pacific Ocean and four of which are from the western Pacific. This observation is consistent with the hypothesis that the genus Lamellibrachia originated in the Pacific, likely the western Pacific, and subsequently radiated to the eastern Pacific, the Atlantic, and the GOM.
Three morphological species of Lamellibrachia were identified in collections from the GOM: L. luymesi, from the upper slope at between about 400 m and 800 m;L.sp. 1, from 950 to 2320 m; and L. sp. 2, from 1175 to 2320 m. L. luymesi and L. s p .1h a v eas i m i l a r number of sheath lamellae, but the deep-water L. sp. 1 generally has more gill lamellae, ranging between 21 and 27 in the 28 individuals examined, whereas the shallow-water L. luymesi has between 15 and 22 gill lamellae in the 20 individuals examined for the species description. The morphological character that allowed rapid identification of animals aboardship was the relatively short and fat vestimentum of L. sp. 1. The ratio of length to width of the vestimentum of L.sp.1rangesfrom2.4to4.7andfrom6.2to16.4in L. luymesi. L. sp. 2 has a similar number of sheath and gill lamellae as L. sp. 1, and the vestimentum length to width ratio tends to be shorter (1.9 to 3). The most distinct field character for L. sp. 2 is the lack of a ventral vestimental fold, which is present on L. sp. 1.
Despite morphological characters that distinguish the three GOM Lamellibrachia presumptive species, only either the COI or the 16S phylogenetic trees resolved two of them. Specifically, both genes failed to separate L. luymesi from the shallow GOM and L. sp. 1 from the deeper GOM sites. This lack of genetic differences between individuals that span such a wide depth range is unusual (Chase et al., 1998;Zardus et al., 2006) and surprising, given the morphological differences. Both 16S and COI genes consistently identify Lamellibrachia sp. 2 as a separate clade, sister to the L. luymesi/L. sp. 1 clade.
There were no apparent geographic distributional patterns that were independent of depth for the seep vestimentiferans in the Gulf of Mexico. The common species present on the upper Louisiana slope (L. luymesi and S. jonesi) have been found at both the eastern-most and western-most sites where we have collected vestimentiferans. E. laminata from the lower slope ranges from the Alaminos Canyon sites, our most westerly collection sites for this study, to the Florida Escarpment in the eastern GOM (Cordes et al., 2009;McMullin et al., 2003). Both of the Lamellibrachia spp. found at the deeper sites occurred over the entire E-W range of sites within their depth range (from the Alaminos Canyon sites in the west to AT340 in the east).

Within-species diversity of the GOM vestimentiferans
Tables 2 and 3 report within-and between-species p distance calculated for the GOM genetic species. In most cases, withinspecies diversity for both 16S and in the COI genes is strikingly low, a finding that is in contrast to previous studies on deep-sea mollusks and echinoderms, where large amounts of genetic variation were observed over small distances (Chase et al., 1998;Howell et al., 2004;Quattro et al., 2001). However, largescale studies indicate that low within-species genetic variation may be typical of deep-sea organisms (Bisol et al., 1984) and even suggest that it may decrease with increase in depth (France and Kocher, 1996). Genetic variation has been suggested to be an important feature of the genome of an organism that allows it to adapt to a changing environment (Powers et al., 1991). Organisms that live in the deep sea may experience a long-term stable environment, resulting in low levels of within-species genetic diversity. Alternatively, low within-species genetic diversity may be the result of fewer replication errors, more efficient repair in the germ line, or repeated population bottlenecks.
E. laminata, E. spicata, and E. southwardae clade and L. luymesi sp. 1 and L. sp. 2 have a moderate degree of intra-specific diversity (Figs. 2 and 3). However, as with all of the GOM vestimentiferans analyzed, none of the within-species clades grouped by specific geographic locations or depth. A similar pattern was found in the seep mussel Bathymodiolus childressi, which, based on markers ranging from microsatellites to mitochondrial genes, has a panmictic population in the GOM ranging across 550 km east to west and from 540 to 2200 m depth (Carney et al., 2006;Cordes et al., 2007b). In contrast, genetic breaks and barriers that restrict gene flow were identified in both hydrothermal vent vestimentiferans and mussels along the East Pacific Rise (EPR). Specifically, Won et al. (2003) used COI sequences to identify two highly divergent clades on the EPR on the two sides of the Easter Island Microplate. Similarly, Hurtado et al. (2004) used COI sequences to identify several geographic breaks and barriers that restrict gene flow in three genera of annelids along the EPR, including two species of vestimentiferan (Riftia pachypitla and Tevnia jerichonana).

Summary
In this study, our primary goals were to identify and characterize the distributions of vestimentiferans at seep sites covering a wide geographic and depth range in the Gulf of Mexico and to investigate their relationship to other seep vestimentiferan species, using phylogenetic analysis of mitochondrial gene sequences. Although the genetic analyses confirmed identification of most of the morphological species during collections, we also identified an unexpected discrepancy between the morphospecies identified during the collections and genealogical species identified using the mitochondrial genes COI and 16S. Using morphological characters, we identified two new species of Lamellibrachia (spp. 1 and 2). However, neither COI nor 16S distinguished the deeper occurring morphospecies L. sp. 1 from L. luymesi, the common Lamellibrachia species on the upper Louisiana slope. Our molecular genetic analyses confirm the presence of three vestimentiferan species within the escarpiid clade in the Gulf of Mexico. However, since COI also does not differentiate between E. laminata found in the Gulf of Mexico and the other described Escarpia species off the coast of Africa or in the eastern Pacific Ocean, we suggest that COI or 16S genes may not reliably distinguish closely related species of long-lived seep vestimentiferans. We are currently evaluating the usefulness of several nuclear genes to clarify the relationships among the named species of Escarpia and the Lamellibrachia species in the Gulf of Mexico.