Material and methods
Leucothoid amphipods for this study were collected from marine sponges, ascidians, and coral rubble from throughout the Ryukyu Archipelago, Japan, as reported in White and Reimer (2012a, b, c). These collections included detailed host association and location data (GPS coordinates, depth etc.), as such associations were previously unknown for most of the described species in the Leucothoidae. Specimens, station data, and GenBank accession numbers are listed in Table S1.
Genomic DNA was extracted from the urosomes of representative individual ethanol-preserved animals after morphological examination. In total, DNA from one to six specimens each of 24 morphologically determined Leucothoidae species was extracted. Voucher specimens are deposited in The University of the Ryukyus Museum (Fujukan), with the prefix RUMF for museum numbers. Extractions were performed using a guanidine extraction protocol (Sinniger et al., 2010). Partial 18S rDNA sequences were amplified (697-851 base pairs including the V4 hypervariable region) using polymerase chain reaction (PCR) (Saiki et al., 1988). PCR protocols were used following Spears et al. (2005). The following peracarid-specific primers were used for 18S rDNA: 329 (5’-TAATGATCCTTCCGCAGGTT-3’) and Hi- (5’-GTGCATGGCCGTTC TTAG-TTG-3’) (Spears et al., 2005). PCR products were purified using shrimp alkaline phosphatase (SAP) and gel purified with a Qiagen gel extraction kit if non-specific PCR products were apparent after gel electrophoresis.
Sequencing was done by Fasmac (Yokohama, Japan). Sequences were edited, combined, and initially aligned in Bioedit 220.127.116.11 (Hall, 1999) and the alignments were further edited by eye using SeaView 4.3.3 (Guoy et al., 2010) and Se-Al v.2.01 (Rambaut, 2002). Three different alignments were constructed in order to examine the influence of hypervariable regions on resulting phylogenetic trees, as some studies suggest deleting these regions (e.g. Kim and Abele, 1990; Spears et al., 1992; Carmean et al., 1992; Pashley et al., 1993; Campbell et al., 1994; Friedrich and Tautz, 1995; Kim et al., 1996; Chalwatzis et al., 1996; Giribet et al., 1996; Friedrich and Tautz, 1997; Spears and Abele, 2000) while others propose retaining them (Crease and Taylor, 1998; Hwang et al., 2000; White, 2011b). The first alignment contained all variable regions including the V4 hypervariable region (‘long’ alignment). The second alignment retained all variable regions except the V4 hypervariable region (‘medium’). The final alignment retained only conserved regions (‘short’). All three generated alignments contained sequences for 40 taxa (including two outgroup taxa). Representatives of the amphipod family Liljeborgiidae Stebbing, 1899 were chosen as outgroup taxa based on the loss of molar structure and rudimentary charpochelation of gnathopod 1 (more primitive than the fully carpochelate gnathopod 1 in Leucothoidae) in the Liljeborgiidae (Barnard, 1974). The long alignment dataset contained 1205 sites, the medium alignment 751 sites, and the short alignment 426 sites. All three alignments are available from the corresponding author and at the homepage http://web.me.com/miseryukyu/.
All three alignments resulted in trees with identical topologies based on preliminary maximum likelihood (ML) and neighbor-joining (NJ) analyses, with the short and medium trees showing generally lower bootstrap support values at most nodes. For these reasons, we therefore utilized the long alignment for all subsequent analyses. Uncorrected measures of percent-sequence divergence were obtained using Mesquite V2.74 (Maddison and Maddison, 2011). The levels of divergence used were 0-0.023 between different populations of a species, 0.090-0.295 between species within a genus, 0.260-0.349 between different genera within the Leucothoidae, and greater than 0.306 between leucothoid genera and the outgroup, Liljeborgiidae (after White, 2011b).
Twenty one ingroup taxa were included in the analyses due to lack of specimens or difficulty of obtaining molecular data from the other four species collected in the Ryukyus. PhyML (Guindon and Gascuel, 2003) was used for ML analyses using an input tree generated by BIONJ (Neighbor Joining) with the Generalized Time Reversible (GTR) model, incorporating invariable sites and a discrete gamma distribution with eight substitution rate categories. Base frequencies were estimated from the dataset. Support for the maximum likelihood tree was measured using the bootstrap method with 1,000 replicates in PhyML.
NJ analysis was accomplished in SeaView 4.3.3 using Jukes Cantor (JC) distance methods. Support was measured using the bootstrap method with 1,000 replicates in SeaView 4.3.3.
Bayesian phylogenetic inference analysis (B) was accomplished using MrBayes 3.1.2 (Ronquist and Huelsenbeck, 2003) under GTR + I + Γ. One cold and three heated Markov chains Monte Carlo (MCMC) with default-chain temperatures were run for 5 million generations, sampling log-likelihoods. Trees at 1000- generation intervals were saved (5000 InLs and trees saved during MCMC). The likelihood plot suggested that MCMC reached a stationary phase after the first 1,000,000 generations, and thus the remaining 4,000 trees were used to obtain clade probabilities and branch-length estimates.