«Taxonomic Revision, Molecular Phylogeny and Zoogeography of the huntsman spider genus Eusparassus (Araneae: Sparassidae) Dissertation for attaining ...»
Despite being the tenth largest of 112 spider families, Sparassidae was never subject to any comprehensive phylogenetic systematic study. Neither monophyly of Sparassidae and the majority of respected subfamilies has been tested (except for Deleninae; Agnarsson and Raynor, 2013), nor was the position of Sparassidae resolved other than being a member of the family-rich RTA-clade. Most of the many sparassid genera require revision and evaluation of their systematic position within the family and subfamilies. The latter is also true for the recently revised genus Eusparassus (Moradmand, 2013) and its subfamily Eusparassinae. Yet, the revision of Eusparassus with its six proposed species-groups based on morphological characters provides a good example to evaluate the diagnostic characters used for classification within these taxonomic ranks.
Results: chapter 3.3: Molecular phylogeny
In summary we aim to provide a rigorous insight into the phylogeny of Sparassidae (with a main focus on Eusparassus and the Eusparassinae) and its relationships to other spider families, especially those within the RTA-clade. Furthermore, the classification within Sparassidae is
revisited, concerning all but the species level, and addressing the following main questions:
1. What is the systematic position of Sparassidae within the RTA-clade? Is the group ‘Laterigradae’ a valid taxonomic entity or is the laterigrade leg position the result of convergent evolution? Is the Dionycha, based on the single character of two-clawed tarsi, a monophyletic group?
2. Are the family Sparassidae and its currently accepted subfamilies monophyletic?
3. Are the genus Eusparassus and its proposed species-groups delineated by Moradmand (2013) monophyletic?
4. When did Sparassidae diverge from RTA-clade relatives? And when did the genus Eusparassus originate?
Tissue samples for this study were mainly obtained from the Arachnology section, Senckenberg Research Institute, Frankfurt am Main (SMF) and partially from other spider collections. One leg of each freshly collected spider was preserved in pure 96% EtOH and subsequently stored at minus 28 C. The specimens were preserved in 70% EtOH and given individual ‘SD’ numbers (voucher number) which were used as preliminary identification numbers. A list of all voucher specimens with respective SD numbers, their collection localities, deposition institute and the genetic markers investigated with Genbank accession numbers is given in Table 1.
To determine the placement of Sparassidae within the RTA-clade and to test the monophyly of the ‘Laterigradae’, representatives of 24 of 41 extant spider families now accepted in the RTA-clade were included. This sample of taxa was appropriate to test other relationships proposed for Sparassidae, e.g. if being a member of Dionycha or to revisit its original placement in the Clubionidae. Finally, a number of Orbiculariae, Eresoidea and Palpimanoidea were included to root this RTA-clade ingroup as suggested by previous studies (Agnarsson et al.,
Results: chapter 3.3: Molecular phylogeny2013a; Griswold et al., 2005; Miller et al., 2010; Spagna and Gillespie, 2008). Sequences for most RTA-clade (non-sparassid) and outgroup taxa were acquired from Genbank (see Table 2).
Monophyly of Sparassidae and the relationships of its major subfamilies were tested by including a comprehensive coverage of taxa from across the overall geographic distribution range of the family, and by including representatives of all larger subfamilies. Geographic sampling covered Africa, Madagascar, Eurasia, Australasia, North America and Central and South America. The representatives of the following historically proposed subfamilies were included in the analyses: Eusparassinae, Sparassinae, Heteropodinae, Palystinae, Polybetinae, Staianinae, Sparianthinae and Deleninae, as well as some presently unplaced genera (see Figure 1). Fresh tissues for the following subfamilies remained unavailable: Clastinae (New Guinea), Chrosiodermatinae (Madagascar) and Tibellomatinae (Venezuela); all are considered monotypic, with Tibellomatinae based on a single juvenile.
To test relationships within Eusparassus and the validity of the six Eusparassus speciesgroups proposed by Moradmand (2013) at least two representatives of the dufouri-, walckenaeri-, doriae- and tuckeri-group were included. Of the jaegeri-group only a single member was available, and no individual of the vestigator-group (two species in Central and East Africa and one in India) could be obtained. Additionally, Cercetius perezi was included in the analyses to test its relationships with Eusparassus group taxa.
2.2 DNA amplification, sequencing and pre-analysis data handling
Extraction of genomic DNA used the CTAB method after Wallace (1987; for details see Bayer and Schönhofer, 2013). The nuclear 28S rRNA gene (28S), Histone H3 (H3) and the mitochondrial cytochrome c oxidase subunit I (COI) and 16S rRNA gene (16S) were amplified using primers and PCR conditions specified in Table 3. These genes have previously shown to be useful for inferring phylogenetic relationships in other groups of spiders (Arnedo et al., 2004;
Spagna and Gillespie, 2008; Arnedo et al., 2009; Crews et al., 2010; Miller et al., 2010, Dimitrov et al., 2012; Bayer and Schönhofer, 2013; Dimitrov et al., 2013; Wood et al., 2013). Fragments were purified using the QIAquick PCR purification kit (Qiagen) and sequencing was performed with the BigDye Terminator Cycle Sequencing Kit v3.1, sequencing fragments from both directions, using primers as mentioned. The majority of previously listed procedures were realised by the Scientific Research and Development GmbH, Bad Homburg, Germany (SRD).
Results: chapter 3.3: Molecular phylogenySequence contigs were assembled and edited using DNA Baser v3.5.1 (Heracle BioSoft, http://www.DnaBaser.com) and Sequencher v4.1.4 (Gene Codes Corporation, Ann Arbor, MI).
Sequences were initially aligned using MEGA v4.1b (Tamura et al., 2007) checking for correct amino acid translation in COI and H3. To incorporate structural information, 28S and 16S were further aligned with MAFFT v7 (Katoh and Standley, 2013) using the Q-INS-i strategy as recommended by Katoh and Toh (2008). In rare cases, sequences of 28S were manually readjusted or sequences removed from the alignment. In some specimens, paralogous copies of H3 and 28S were observed, and excluded from the analyses. Models of sequence evolution were evaluated for each gene using jModeltest v0.1.1 (Posada, 2008) under three substitution schemes (JC, HKY and GTR) on a fixed BIONJ tree, allowing for unequal base frequencies and amongsite rate variation and based on the Akaike information criterion. COI and H3 were further partitioned into single codon positions using Mesquite v2.75 (Maddison and Maddison, 2011) and models evaluated for each partition as defined above.
2.3. Alignment strategies and phylogenetic analyses
Aligning deeply divergent taxa introduced many gaps and regions of alignment uncertainty to the 28S and 16S partitions. We thus assumed the composition of taxa would influence our analyses, and investigated changes caused by alternate taxa combinations, thereby also testing the
robustness of our data. We investigated taxon combinations as follows:
1) the full data set including all ingroup and outgroup taxa (All-taxa set);
2) all outgroups, but Sparassidae reduced to a set of eight most divergent taxa (see Fig. 2) as predetermined by other analyses (All-outgroups-8-Sparassidae set);
3) Sparassidae reduced to eight taxa as previously defined, all members of the RTA-clade, but other outgroups reduced to Araneus diadematus Clerck, 1757 to root this ensemble (RTA-out-8Sparassidae set);
4) including only Sparassidae and accepting Thelcticopis Karsch, 1884 as outgroup, as predetermined by other analyses (Full-ingroup set);
5) as previously defined but reducing terminal taxa in groups with many representatives, e.g.
including only two Eusparassus species (Full-ingroup-2-Eusparassus set).
These taxon combinations allowed evaluation of all initial hypotheses using more than one alignment version. For example the position of Sparassidae within the RTA-clade was estimated in data sets 1, 2 and 3, and the ingroup phylogeny of Sparassidae and the position of Eusparassinae in data sets 1, 4 and 5. For all combinations gene-partitions of 28S and 16S were re-aligned using the described strategies, while this proved unnecessary for the coding genes. For these rRNA genes we also investigated additional alignment versions of the specified five data sets by removing areas of alignment uncertainty using Gblocks (Castresana, 2000; available at http://www.phylogeny.fr/) with least stringent settings. Models were re-evaluated for all partitions of different taxon-combinations or alignment versions (Table 4), and applied to the subsequent analyses. Concatenation and alignment conversion was handled with Mesquite v2.75, while alignment manipulation and curation used MEGA v4.1b (Tamura et al., 2007). For a comparison of the length of all partitions in all data sets see Table 4.
Maximum likelihood analyses (ML) were performed using raxmlGUI v0.95 (Silvestro and Michalak, 2011). Node support was assessed from 1000 non-parametric bootstrap (bt) pseudoreplications (Felsenstein, 1985) as implemented in raxmlGUI (according to Stamatakis et al., 2008). Bayesian inference (BI) topologies were assessed using MrBayes v3.2.1 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003). Analyses were run for 5,000,000 to 20,000,000 generations, where the standard deviation of split frequencies had dropped below 0.01, and convergence diagnostics were satisfied for all sampled parameters with few exceptions (ESS 200, PSRF = 1.00; Ronquist et al., 2011). The first 25% of trees were discarded as burnin, with remaining trees used to reconstruct a 50% majority rule consensus tree. Split frequencies were interpreted as posterior probabilities (pp) of clades. We considered branches receiving 1.00 pp and 87 bt as highly supported (Zander, 2004) and branches receiving 0.94 pp and 69 bt as sufficiently supported.
2.4. Divergence time estimation
Approximate taxon divergence times were estimated from the concatenated data using BEAST v1.7.2 (Drummond et al., 2012), applied to two selected data sets: All-outgroups-8-Sparassidae set and Full-ingroup set with MAFFT alignments. Due to computational limitations, the data were partitioned by genes only, with models set to HKY and invariable sites and gamma as suggested by jModeltest. Nodes were not constrained to be monophyletic and no tree model
Results: chapter 3.3: Molecular phylogeny
operators were removed, allowing BEAST to re-estimate the topology. Analyses were run until ESS values exceeded 200 for all priors, which was after 50,000,000 generations, and checked for convergence using Tracer v1.5. We also investigated changes in the speed of substitutions per lineage using an uncalibrated random local clock model as implemented in BEAST (Drummond and Suchard, 2010). For final analyses in BEAST a Yule model of speciation and a relaxed lognormal clock were set. Clock rates were unlinked and estimated for individual partitions.
For the BEAST analysis only well supported nodes of the All-outgroups-8-Sparassidae set were calibrated using fossil data (see Fig 4). Calibrations (parameters given in parentheses) were set as suggested by Wood et al. (2013) treating them as lognormal distributions with a hard lower bound, based on the minimum age of the fossil, and a soft 95% upper bound to allow
significantly older ages:
1) Oecobiidae-Hersiliidae: Penney et al. (2012) state comparable divergence times between Eresidae, Oecobiidae and Hersiliidae, with the common ancestor of Oecobiidae and Hersiliidae known from Lebanese amber (125-135 MA; Penney and Selden, 2011) (parameters: mean 2.3, stdev 1, offset 123 = 5%: 124.9 MA, median: 133 MA, 95%: 174 MA; calibration according to Wood et al., 2013),
2) Salticidae-Philodromidae: with the origin of Salticidae still debated, we used the compelling salticid fossil record as not predating Eocene Baltic amber (Penney and Selden, 2011;
parameters: mean 2.3, stdev 1, offset 43 = 5%: 44.9 MA, median: 53 MA, 95%: 94.7 MA),
3) Dictynidae-other families: According to the presence of Dictynidae in Burmese amber (100 MA; Penney and Selden 2011), the divergence with other families is set to slightly predate this
deposit (parameters: mean 2.89, stdev 1, offset 96.5 = 5%: 100 MA, median: 114.5 MA, 95%:
189.7 MA; Penney and Selden, 2011; Wood et al., 2013),
4) Lycosidae-Pisauridae: As for the previous family, Pisauridae fossils are present in Burmese amber (100 MA; Penney and Selden 2011). We did not follow Wood et al., (2013), assuming much younger divergence between Lycosidae and Gnaphosidae, as this family relation was not found supported and Gnaphosidae fossils already date back to Baltic amber (Penney and Selden 2011; parameters: mean 2.89, stdev 1, offset 96.5 = 5%: 100 MA, median: 114.5 MA, 95%:189.7 MA).
To constrain the root of the “Full-ingroup set” we applied the divergence age of Thelcticopis from all other Sparassidae as based on the analysis of the “All-outgroups-8
Results: chapter 3.3: Molecular phylogeny
Sparassidae set” (normal distribution; parameters: initial value/mean 163.0, stdev 20 = 2.5%: 130 MA, median: 163 MA, 97.5%: 202 MA). For this set we further calibrated the earliest divergence within Eusparassus (mean 1.7, stdev 1.51, offset 43.5 = 5%: 43.96 MA, median: 48.97 MA, 95%: 109.1 MA) using the minimum age based on the amber fossil E. crassipes (Dunlop et al.,
2011) and the maximum age based on the biogeographic evidence (assuming Eusparassus originated after the breakup of Gondwana; Briggs, 1995; Sanmartín and Ronquist, 2004).