From Dömel et al (2017):

2.1 Specimens and sampling sites
For the remainder of this study, we use the term Pallenopsispatagonica sensu lato (s.l.) when referring to the whole species complex including P. yepayekae, because it groups within clades morphologically originally identified as P. patagonica. Individuals of P. patagonica s.l. from the shelf of South America, Subantarctic islands as well as around the Antarctic continent were analyzed (Table 1, Figure 1). Chilean specimens were collected by divers during Huinay Fjordos expeditions (HF16, HF21, HF24 and HF26). Falkland samples (ZDLT1) were provided by Vladimir Laptikhovsky (Falkland Islands Fisheries Department, Stanley, Falkland Islands). Samples from the Southern Ocean were collected using different bottom trawls during several cruises on board the RRS James Clark Ross (British Antarctic Survey, Cambridge, UK) and the RV Polarstern (Alfred Wegener Institute Helmholtz Center for Polar and Marine Research, Bremerhaven, Germany). After collection, specimens were stored in ethanol (96%). Specimens were morphologically inspected and assigned to P. patagonica s.l. before being molecularly studied.

2.2 Molecular analyses
Muscle tissue was extracted from the tibia using sterile scalpel and forceps. DNA was isolated from the tissue using a modified salt precipitation protocol after Sunnucks & Hales (1996) (see Weiss & Leese 2016). Extracted DNA was eluted in 100 ul TE minimum buffer (0.1xTrishydrochloride buffer, pH 8.0, containing 0.1 mM EDTA). The amplification of the mitochondrial cytochrome c oxidase subunit I gene (COI) and a ribosomal gene region covering the 18S-ITS1-5.8S-ITS2-28S stretch (internal transcribed spacer, ITS) was carried out in 25 ul reactions containing 1x (2.5 ul) PCR buffer (5Prime), 0.2 mM dNTPs, 0.5 uM of each primer, 0.025 U/ul (0.125 ul) Hotmaster Taq (5Prime) and 1 ul template DNA, topped up to 25 ul with sterile water. A 658 bp long fragment of the COI was amplified using the common barcoding primer pair HCO2198 and LCO1490 (Folmer et al. 1994). The optimal temperature profile for the PCRs with these primers was an initial denaturation at 94 deg.C for 2 minutes, followed by 36 cycles of denaturation at 94 deg.C for 20 seconds, annealing at 46 deg.C for 30 seconds, extension at 65 deg.C for 60 seconds, and a final extension at 65 deg.C for 7 minutes. For ITS, an approximate 1000 bp long fragment was amplified using primers ITSRA2 and ITS2.2 (Arango & Brenneis 2013). PCR cycling program was initial denaturation at 94 deg.C for 3 minutes, followed by 35 cycles of denaturation at 94 deg.C for 30 seconds, annealing at 55 deg.C for 75 seconds, final extension at 65 deg.C for 5 minutes.

For sequencing, 10 U (0.5 ul) Exonuclease I (Thermo Scientific), 1.5 U (1 ul) FastAP Thermosensitive Alkaline Phosphatase (Thermo Scientific) and 9 ul PCR product per reaction were used. The purification mix was incubated for 25 minutes at 37 deg.C, followed by a denaturation step at 85 deg.C for 15 minutes. For sequencing at GATC Biotech AG (Cologne, Germany) 5 ul of purified PCR product was mixed with 5 ul of 5 pmol/ul primer. Forward and reverse primers were used to sequence both directions of the DNA strands.

For ITS sequences of samples reported by Harder et al. (2016) (herein labeled with GenBank numbers starting with KT98) DNA extraction was performed as stated in Harder et al. (2016). For ITS amplification the same primer pair as mentioned above was used. PCR mixture consisted of 1x PCR buffer, 0.75 U Taq DNA polymerase (5Prime), 2.5 mM Mg2+, 10 nmol of each dNTP, 1 ul of template DNA, 0.5 uM of each primer, and water to 25 ul. PCR cycling program was run, with an initial denaturation at 94 deg.C for 2 minutes, followed by 37 cycles of denaturation at 94 deg.C for 20 seconds, annealing at 55 deg.C for 30 seconds, extension at 65 deg.C for 80 seconds, with a final extension at 65 deg.C for 10 minutes. Successful amplification was confirmed by visualizing PCR products on a 1% agarose gel stained with ethidium bromide. Target PCR product was gel extracted and purified using a Qiagen QIAquick® Gel Extraction Kit according to the manufacturer’s recommendations. Bidirectional Sanger sequencing of amplicons was performed at High Throughput Genomics Center (Seattle, WA, USA).

Phylogenetic analyses
For COI, P.patagonica s.l. sequences from Weis et al. (2014) (n=34 including five downloaded from NCBI) and Harder et al. (2016) (n=26) were added to the final data set. ITS sequences of specimens of both previous studies were generated and included into the ITS alignment, too. For both gene regions, sequences were edited with Geneious v. 8.1.3 (http://www.geneious.com, Kearse et al. 2012) and aligned in Geneious using MAFFT v. 1.3.3 Multiple Sequence Alignment (Katoh & Standley 2013) with default parameters as implemented in Geneious, with a gap opening penalty of 1.53 and offset value of 0.123. For COI, sequences were translated into amino acids using the invertebrate mitochondrial genetic code (translation table 5) to verify that all codons could be translated without stop codons. For ITS, a version of the alignment where ambiguously aligned regions were removed was produced with Gblocks v. 0.91b (Castresana 2000) using less stringent parameters (smaller blocks, gaps in final alignment allowed, less strict flanking positions) as has been done in Dietz et al. 2015b. For analyses when only unique copies were needed, sequences were collapsed into unique sequences (‘haplotypes’ for COI data) with the online tool FaBox v. 1.41 (Villesen 2007). For both, nuclear and mitochondrial data sets a maximum-likelihood (ML) analysis was performed with RAxML v. 8.2.4 (Stamatakis 2014) using the GTRCAT model of sequence evolution and branch support was assessed with 10000 rapid bootstrap replicates. In addition, for the mitochondrial data set a resolved ultrametric gene tree was calculated using BEAST v. 1.8.3 (Drummond et al. 2012) with the model specified by jModelTest v. 2.1.10 (Guindon & Gascuel 2003; Darriba et al. 2012). An XML file was created with BEAUti v. 1.8.3 (Drummond et al. 2012) with the following settings: HKY+G+I as substitution models and 80 x 10^6 as length of MCMC chain sampling every 1000th tree. Convergence of the likelihood and appropriate effective sampling size (ESS > 200) of parameter estimates were checked using TRACER v. 1.6 (Rambaut et al. 2014), and a consensus tree was calculated using TREEANNOTATOR v. 1.8.3 of the BEAST package. Furthermore, uncorrected pairwise distance matrixes were created using MEGA v. 7 (Tamura et al. 2011).

2.4 Species delimitation methods
For species delimitation analysis of the COI data set ABGD (Automatic Barcode Gap Discovery; Puillandre et al. 2012) was run to test for presence of distinct clades. As no clear barcode gap was found in the pairwise distance data, ABGD results varied drastically depending on single sequences and run parameters. Hence, for results presented default settings but Kimura-2-parameter (K2P) for distance were used. The same settings were applied to the ITS alignment (including and excluding ambiguously aligned regions). Due to the smaller data set for the ITS alignment and the fact that informative alignment gaps cannot easily be interpreted as additional character in the tree-based delimitation methods, further species delimitation methods were only applied to the COI data set. The final mitochondrial COI ML tree was used to perform a Bayesian Poisson Tree Processes (bPTP) analysis using the web server (http://species.h-its.org/ptp; Zhang et al. 2013). Furthermore, a Generalized Mixed Yule Coalescent (GMYC) analysis based on the resolved ultrametric gene tree was conducted at the web server (http://species.h-its.org/gmyc; Fujisawa & Barraclough 2013) using the single-threshold method only (see Fujisawa & Barraclough 2013).

2.5 Molecular clock analysis
A calibrated molecular clock rate for sea spiders has not been reported in previous studies. However, in order to infer possible divergence date ranges for the different clades we applied a widely adopted COI molecular clock rate reported for insects: 1.15% permyr and lineage (Brower 1994). BEAST v. 1.8.2 was used to estimate divergence times using an HKY+I+G evolution model as well as an uncorrelated local clock model. Analyses were run for 10 x 10^6 generations sampling every 1000th tree. Convergence of parameter estimates and ESS control and subsequent steps were done as described above.