Predicting genetic biodiversity in salamanders using geographic, climatic, and life history traits

genetic variation is distributed within species depending on which taxonomic groups are being examined, and at what spatial scale. In order to expand these approaches, we conducted an assessment of genetic lineages in roughly 100 described salamander species using the phylogatR database [38]. PhylogatR aggre- gates DNA sequence data from both GenBank and BOLD into sequence alignments, providing associated GBIF occurrence records (i.e., GPS coordinates) for each sequence. There are over 700 nominal species of salamanders belonging to nine families [39], most located in the north- ern hemisphere. While salamanders contain a wide variety of life history strategies and habi- tats, they are likely to have high levels of cryptic diversity due to their moisture requirements and similar body forms. However, their eco-evolutionary processes can vary from species to species and sometimes oppose our expectations [40–45]. We follow methods from Parsons et al. [35] and use molecular species delimitation methods to estimate the number of genetic lineages present in previously collected data that is both openly available and easily tractable. We utilize these delimitation results to identify species that are likely to harbor undescribed diversity. Species for which delimitation reveals multiple genetic lineages are classified as hid- den species. The individual genetic lineages that comprise these hidden species are referred to as hidden genetic lineages. We then use a random forest classification to determine whether

any variables pertaining to geography, the environment, or life history traits contribute to the presence of hidden genetic lineages within species. We also discuss some of the difficulties in using open-source data that are often messy and potentially taxonomically and geographically biased. Materials and methods Collection of genetic and geographic data We downloaded all available data from the phylogatR database (https://phylogatr.org/) using the search term ‘Caudata’ on 2/4/22. The uncleaned data represented four families, 93 different species, and 14 loci with a total of 3768 DNA sequences. To begin cleaning the data, we calcu- lated nucleotide diversity (pi) values for each locus in every species and found outliers by set- ting lower and upper bounds of 2.5% (0) and 97.5% (0.2193634) respectively. For each of the four outliers and two species with missing pi values, we opened the DNA sequence file in Mes- quite v3.7 [46] and removed any extremely short or non-overlapping sequences (S1 Data). Additionally, we discovered a typo for the species Batrachuperus karlschmidti causing there to be two different species folders for the same species. Both the sequence and occurrence files were merged for the species and the sequence files were realigned to correct the error. Two species complexes were present in the dataset, and these were kept named as downloaded: Tri- turus cristatus x dobrogicus macrosomus and Ambystoma laterale jeffersonianum complex. A review of the available loci indicated that two genes, COI and cytb, were the most well-repre-

sented in both total number of sequences and species coverage. Consequently, we opted to uti- lize these two genes for downstream analysis. Species alignments from the download for both the mitochondrial genes Cytochrome oxi- dase I (COI) and Cytochrome b (cytb) were merged for all salamander species and aligned using MAFFT v7.5 [47] with the default settings and including the–adjustdirection command to account for reverse complement sequences. We visually inspected alignment files for both genes and removed all short sequences, which we classified as those missing 50% or more of the second half of the sequence. Twenty-one sequences were removed from the COI alignment and 99 were removed from the cytb alignment, leaving totals of 768 and 908 sequences for COI and cytb, respectively. The sequences for seven species were completely removed from further analysis due to their short length (missing 50% or more of the second half of the sequences). In PLOS ONE Predicting hidden lineages in salamanders PLOS ONE | https://doi.org/10.1371/journal.pone.0310932 October 18, 2024 3 / 20