Complete mitochondrial genomes of three Neotropical sleeper gobies: Eleotris amblyopsis, E. picta and Hemieleotris latifasciata (Gobiiformes: Eleotridae)

Abstract We report the first complete mitochondrial genomes of three species of eleotrid fishes from the Pacific and Atlantic watersheds of Panama: Eleotris amblyopsis, E. picta, and Hemieleotris latifasciata. The three species have similar mitochondrial genomes with identical gene order; however, there are differences in the length of control region, 16S rRNA, and in seven of the tRNAs. In addition, ATP8 is one codon shorter in E. picta than in E. amblyopsis or H. latifasciata. We infer a phylogeny for Gobiiformes based on all mitochondrial protein-coding genes, which supports the monophyly of Eleotridae but does not recover Neotropical members of Eleotris as a distinct clade.


Introduction
The freshwater fish fauna of Middle America contrasts to that of the Nearctic and South American regions in having a large number of salt tolerant species and marine taxa that have invaded the mainland, and are more or less permanent residents of freshwater streams (Myers 1966;Matamoros et al. 2015). This marine-derived and diadromous component comprises five families (Ariidae, Atherinidae, Gerridae, Eleotridae, and Gobiidae) that, despite representing an important fraction of the diversity of this Neotropical assemblage, is considerably understudied (McMahan et al. 2013;Galv an-Quesada et al. 2016); particularly relative to the other taxa in the region such as poeciliids and cichlids that have been the focus of many systematic and biogeographic studies (e.g. Martin and Bermingham 1998;Perdices et al. 2005;Alda et al. 2013;Marchio and Piller 2013;McMahan et al. 2017).
In this study, we present the first complete mitochondrial genomes of three fish species of the family Eleotridae that are distributed across Atlantic and Pacific watersheds primarily in the Neotropics (but can be found as far north as California and South Carolina). In addition, we present a mitogenomic phylogeny of the Gobiiformes, which illustrates the position of the Neotropical members of the eleotrids in relation to other species for which complete mitochondrial genomes are available.
Mitogenome sequences were obtained as a by-product of a hybrid target capture of ultraconserved elements (UCEs) (Faircloth et al. 2012). We trimmed our reads for low-quality bases and adapter contamination using the parallel wrapper around Trimmomatic (Bolger et al. 2014) in illumiprocessor (Faircloth 2013). Then, we mapped our reads to reference mitochondrial genomes from the evolutionary closest species available using Bowtie 2 (Langmead and Salzberg 2012) implemented in Geneious 10.1.3 (www.geneious.com; Kearse et al. 2012). Using Geneious 10.1.3, we then created consensus sequences and annotated them for each sample.
We aligned our mitogenomes with representatives from five of the seven families of Gobiiformes (sensu Agorreta et al. 2013), and extracted all 13 protein-coding genes for subsequent phylogenetic analysis. We partitioned the data by gene and by codon, and estimated the best arrangement of partitions under a GTR þ G nucleotide substitution model using PartitionFinder (Lanfear et al. 2012). We used these partitions in a Maximum Likelihood analysis in RAxML (Stamatakis 2014), that we run for 40 searches to find the best tree, and performed 500 bootstrap replicates to assess nodal support.
All protein-coding genes are of identical length across the three species, with the exception of ATP8 that is reduced by a single codon in E. picta. Non-coding regions, on the other hand, show more variation in length. For example, seven out of 22 tRNAs differ in size by a single base pair (bp): E. amblyopsis is one bp longer in tRNA-Trp, and one bp shorter in tRNA-Tyr; E. picta has an additional bp in tRNA-Pro and tRNA-Asp; H. latifasciata is one bp shorter in tRNA-Val, tRNA-Met, tRNA-Asn, and tRNA-Thr. 16S rRNA and control region differ drastically among the three species, ranging between 1687 and 1695 bp for the 16S rRNA, and 842 and 860 b for the control region. Within species, the two E. picta specimens exhibit 58 mutational differences throughout their mitochondrial genomes; 15 of these changes are found in non-coding regions, while the translation of eight codons differs as a result of these mutations.
The most common initiation codon, ATG, is found in all the mitochondrial protein-coding genes apart from COXI, that has GTG (Val), and ND6, that has TAC (Tyr), as start codons in all species. Five genes share the common termination codon TAA across all species. In ND1, E. amblyopsis and E. picta differ in having the termination codon TAG, while in E. picta remains TAA. For ND6, ATC is the stop codon in every species. COXII, COXIII, ND3, ND4, and CYTB coding regions have incomplete stop codons: T or TA. These incomplete stop codons are completed as TAA by the post-transcriptional polyadenylation of the corresponding mRNAs (Ojala et al. 1981).
Our phylogenetic reconstruction strongly supports the monophyly of the family Eleotridae. Hemieleotris latifasciata is the sister group to a clade containing all the species of the  Figure 1). We resist drawing any major biogeographic conclusion at this time, given the limited sampling, but do point out that major transoceanic relationships do exist in Gobiiformes (Chakrabarty et al. 2012).
Here, we present the first mitogenomes of three members of the Eleotridae from the Neotropics. These sequences will aid future studies into the evolutionary relationships of gobiiform fishes. This work will also aid our understanding of the diversity of Middle America freshwater fish assemblages as mitochondrial genomes of additional species and lineages become available.   Figure 1. Maximum likelihood (RAxML) phylogeny using all mitochondrial protein-coding genes from a selection of Gobiiformes species from which complete mitogenomes are available (GenBank accession numbers indicated), and Kurtus gulliveri, which was used as outgroup. Numbers next to nodes are support values obtained after 500 bootstrap replicates. New sequences obtained in this study are highlighted in bold.