Signatures of convergence in Neotropical cichlid fish

Convergent evolution of similar phenotypes suggests some predictability in the evolutionary trajectories of organisms, due to strong and repeated selective pressures, and/or developmental constraints. In adaptive radiations, particularly in cichlid fish radiations, convergent phenotypes are commonly found within and across geographical settings. Cichlids show major repeated axes of morphological diversification. Recurrent changes in body patterns reveal adaption to alternative habitats, and modifications of the trophic apparatus respond to the exploitation of different food resources. Here we compare morphologically and genetically two Neotropical cichlid assemblages, the Mexican desert cichlid, and the Nicaraguan Midas cichlid, with similar polymorphic body and trophic adaptations despite their independent evolution. We found a common morphological axis of differentiation in trophic structures in both cichlid radiations, but two different axes of differentiation in body shape, defining two alternative limnetic body patterns. Adaptation to limnetic habitats implied regulation of immune functions in the Midas cichlid, while morphogenesis and metabolic functions in the desert cichlid. Convergent phenotypic adaptions could be associated to divergent gene regulation.

Methods
Different morphotypes of Herichthys minckleyi and Amphilophus spp. were fished, measured, photographed in a standard position for morphometric analyses. The lower pharyngeal jaw (LPJ) was collected for morphometric analyses and spleen tissue was collected for gene expression analyses. We used the TPSdig2 pipeline to record landmarks of the body morphology and LPJ morphology. We analysed the landmark configuration with the R package geomorph and we used a linear discriminant analyses of principal components to evaluate the posterior probability for each individual to be reassigned to its original cluster (i.e. morphotype). We analysed meristic measurements of head and trunk proportions with an ANOVA in R.

We extracted RNA from spleen for paired-end sequencing on an Illumina HiSeq2000. We clipped adapters of demultiplexed reads with Trimmomatic, mapped them to the Amphilophus reference genome with STAR RNA-seq aligner and used StringTie to generate a consensus transcriptome. We quantified reads with featureCounts of the R package Rsubread and used the DESeq2 pipeline to infer differential expression. We extracted transcript sequences from the reference genome with AGAT, identified peptide-coding regions with Transdecoder and blasted them against the Swiss-Prot database from UniProtKB. We extracted Gene Ontology (GO) terms for each gene with Trinotate and used the R package topGO to identify enriched GO terms.