The genus Cyclura includes nine extant species and six subspecies of West Indian Rock Iguanas and is one of the most imperiled genera of squamate reptiles globally. An understanding of species diversity, evolutionary relationships, diversification, and historical biogeography in this group is crucial for implementing sound long-term conservation strategies. We collected DNA samples from 1–10 individuals per taxon from all Cyclura taxa (n = 70 ingroup individuals), focusing where possible on incorporating individuals from different populations of each species. We also collected 1–2 individuals from each of seven outgroup species of iguanas (Iguana delicatissima; five Ctenosaura species) and Anolis sagrei (n = 12 outgroup individuals). We used targeted genomic sequence capture to isolate and to sequence 1,872 loci comprising of 687,308 base pairs (bp) from each of the 82 individuals from across the nuclear genome. We extracted mitochondrial reads and assembled and annotated mitogenomes for all Cyclura taxa plus outgroup species. We present well-supported phylogenomic gene tree/species tree analyses for all extant species of Cyclura using ASTRAL-III, SVDQuartets, and starBEAST methods, and discuss the taxonomic, biogeographic, and conservation implications of these data. We find a most recent common ancestor of the genus 9.91 million years ago. The earliest divergence within Cyclura separates C. pinguis from a clade comprising all Cyclura except C. pinguis. Within the latter group, a clade comprising C. carinata from the southern Lucayan Islands and C. ricordii from Hispaniola is the sister taxon to a clade comprising the other Cyclura. Among the other Cyclura, the species C. cornuta and C. stejnegeri (from Hispaniola and Isla Mona) form the sister taxon to a clade of species from Jamaica (C. collei), Cuba and Cayman Islands (C. nubila), and the eastern (C. rileyi) and western (C. cychlura) Lucayan Islands. Cyclura cychlura and C. rileyi form a clade whose sister taxon is C. nubila. Cyclura collei is the sister taxon to these three species combined.

MTDNA- The mtDNA were collected as bycatch from UCE sequencing and the PHYLUCE pipeline. We gathered cleaned contigs generated from Velvet alignments, organized by individual, and mapped Cyclura contigs to a reference mitogenome of C. pinguis (GenBank ID LN835346; Gan et al., unpubl.) in Geneious 10.2.5 (Biomatters^® Auckland, New Zealand) using the Geneious mapper with medium sensitivity. For outgroups, we used the reference sequence Anolis carolinensis (EU747728) for our A. sagrei sample and Iguana iguana (AJ278511) for our Iguana delicatissima and Ctenosaura samples (Table 1). For each taxon, we mapped all available contigs from all available specimens, and then deleted the reference genome and extracted the consensus sequence. Thus, each consensus mitogenome sequence is chimeric, in that it is a consensus of nucleotides from multiple individuals of the same taxon to improve overall mitogenome coverage. We retained ambiguity codes in the consensus sequence, then aligned all consensus sequences together to create a mitogenome alignment for all our Cyclura taxa as well as outgroups. We then annotated these mitogenomes using the MITOS2 webserver (http://mitos2.bioinf.uni-leipzig.de/index.py; Bernt et al. 2013) and verified correct annotation via alignment with the reference mitogenomes. We subsequently manually verified the positions of start and stop codons and indels in protein-coding regions of all alignments.

UCE DATA- We then batched raw fastq reads into the Linux phyluce pipeline— a toolkit comprising conda-packaged Python scripts to facilitate UCE identification and extraction (Faircloth 2016). We used the Illumiprocessor Python script (Faircloth 2014), which calls Trimmomatic (Bolger et al. 2014) to trim adapters and low-quality sequences reads. We then assembled contigs from the cleaned reads using the Velvet algorithm (Zerbino and Birney 2008) with kmer = 51. We then extracted UCE loci from the contigs file by matching contigs to the UCE Tetrapods 5kv1 probe set (available at https://github.com/faircloth-lab/uce-probe-sets/blob/master/uce-5k-probe-set/uce-5k-probes.fasta). We aligned UCE loci using the MAFFT algorithm (Katoh and Standley 2013) with edge trimming, then removed prepended locus names from the alignments. We created a 50% complete data matrix, whereby we retained aligned UCE loci with at least 50% of the taxa represented for every retained locus. This supermatrix dataset consisted of 82 taxa, 1,872 loci, and 687,308 bp. More stringent matrices (i.e., 75%, 95%) were found in initial study to have significantly lower phylogenetic resolution and were therefore not used. We conducted all UCE bioinformatics on a Dell® (Dell Inc., Round Rock, TX) PowerEdge© Server (64c, 128gb RAM) running Ubuntu 16.04.5 as well as the CHILABOTHRUS Dell® PowerEdge© Server (16c, 128gb RAM) running Ubuntu 20.04, both at the University of North Carolina Asheville.

The mtDNA sequences are chimeric, each species' sequence is a combination of 1-8 individuals of that species sampled near each other that were sequenced using UCE baits.