Dataset: Systematics of the color-polymorphic spider genus Cybaeolus, with comments on the phylogeny of the family Hahniidae (Araneae)

Phylogenetic analysis of the spiders of the genus Cybaeulus, with outgroups in the marronoid clade. Data from six DNA markers, analyzed with maximum likelihood and parsimony.

PHYLOGENETIC ANALYSIS

We obtained sequences from 26 samples of the three known species of Cybaeolus, and of five additional species of Hahniidae. To these, we added legacy sequences of Cybaeolus and of other genera of Hahniidae, as well as representatives of the remaining families in the marronoid clade. For the new sequences, the extraction and amplification of DNA was made in the Laboratory of Molecular Tools at Museo Argentino de Ciencias Naturales (MACN), from tissues preserved in absolute alcohol at -18ºC. We targeted the markers histone H3 (H3), cytochrome oxidase subunit I (CO1), 28S ribosomal RNA (28S) and 16S ribosomal RNA (16S), previously used to estimate relationships of marronoid spiders (Wheeler et al., 2017). Details of extraction, primers and PCR protocols are the same as in Magalhaes & Ramírez (2022). Sequencing was outsourced to Macrogen Inc., South Korea. The resulting chromatograms were analyzed individually to detect contaminated sequences or ambiguous portions. In addition to these sequences obtained in the laboratory, we combined our data with additional sequences from previous work (Wheeler et al., 2017; Rivera-Quiroz et al., 2020), using the markers mentioned above plus 12S ribosomal RNA (12S) and 18S ribosomal RNA (18S). For the CO1 marker, additional sequences obtained by the Arachnology Division at MACN and deposited in the BOLDSYSTEMS platform (https://www.boldsystems.org/) were also used. Sequences were aligned with MAFFT Online v.7.463 (Katoh & Standley, 2013), using the L-INS-I algorithm. See Table 1 for list of vouchers and sequence identifiers.

Maximum likelihood
For the maximum likelihood analyses we used the program IQ-TREE 2.2.0 (Minh et al., 2020), partitioning the data by marker, and selecting the best combination of partitions and evolution models by Bayesian information criterion (best fitting models were TPM2+I+G4 for H3, GTR+F+I+G4 for 18S, GTR+F+I+G4 for 16S and 12S together, GTR+F+I+G4 for CO1, and GTR+F+I+G4 for 28S). Since the relationships of outgroup taxa in the resulting trees were slightly different to that found in recent phylogenomic studies, we used the study of Gorneau et al. (2023) based on ultraconserved elements as a backbone topology to constrain our tree search, considering only the taxa in common with our analysis (see supplementary Fig. S1); this means that all the rest of the taxa are free to move anywhere during tree search. Support for groups (branches) was estimated by 1000 cycles of ultrafast bootstrapping. Ten independent runs were performed; of those, six converged into nearly identical log likelihood values (-57417.7725 to -57417.9604) and identical topologies; the tree with top-ranking log likelihood is presented in Results, after collapsing branches with bootstrap below 0.5. To estimate the support of an alternative topology with Cybaeolus as sister to the rest of the hahniids, we used TNT 1.6 (Goloboff & Morales, 2023) to modify the optimal tree placing Cybaeolus in such position, and asked for the frequency of the branch of interest (all hahniids except Cybaeolus) in the 1000 bootstrapped trees previously saved by IQTREE.
Ancestral character states for the arrangement of spinnerets (grouped; separated in a transversal line) were estimated by maximum likelihood on the optimal tree, using the R packages phytools and ape, under the models ER and ARD, and the best fitting model selected by the Akaike information criterion. 

Parsimony
For the parsimony analyses we used TNT 1.6. For the equal weights analysis, a heuristic search was made using a driven search with the default parameters of the “new technologies”, aiming for 10 independent hits to minimum length. The resulting trees were then submitted to an additional round of tree-bisection reconnection (TBR) branch swapping. These results were compared to a simpler search strategy of 300 random addition sequences, each followed by TBR, which produced 20 hits to minimal length. As both strategies reached the same trees with multiple independent hits, it is likely that the optimal trees were found. Finally, the strict consensus of all the optimal trees was obtained, and on this consensus the support values were calculated by means of 1000 bootstrap pseudoreplicates.