Understanding the molecular basis of repeatedly evolved phenotypes can yield key insights into the evolutionary process. Quantifying gene flow between populations is especially important in interpreting mechanisms of repeated phenotypic evolution, and genomic analyses have revealed that admixture occurs more frequently between diverging lineages than previously thought. In this study, we resequenced 47 whole genomes of the Mexican tetra from three cave populations, two surface populations, and outgroup samples. We confirmed that cave populations are polyphyletic and two Astyanax mexicanus lineages are present in our dataset. The two lineages likely diverged much more recently than previous mitochondrial estimates of 5–7mya. Divergence of cave populations from their phylogenetically closest surface population likely occurred between ~161k - 191k generations ago. The favored demographic model for most population pairs accounts for divergence with secondary contact and heterogeneous gene flow across the genome, and we rigorously identified gene flow among all lineages sampled. Therefore, the evolution of cave-related traits occurred more rapidly than previously thought, and trogolomorphic traits are maintained despite gene flow with surface populations. The recency of these estimated divergence events suggests that selection may drive the evolution of cave-derived traits, as opposed to disuse and drift. Finally, we show that a key trogolomorphic phenotype QTL is enriched for genomic regions with low divergence between caves, suggesting that regions important for cave phenotypes may be transferred between caves via gene flow. Our study shows that gene flow must be considered in studies of independent, repeated trait evolution.

Albinism, the loss of melanin pigmentation, has evolved in a diverse variety of cave animals but the responsible evolutionary mechanisms are unknown. In Astyanax mexicanus, which has a pigmented surface dwelling form (surface fish) and several albino cave-dwelling forms (cavefish), albinism is caused by loss of function mutations in the oca2 gene, which operates during the first step of the melanin synthesis pathway. In addition to albinism, cavefish have evolved differences in behavior, including feeding and sleep, which are under the control of the catecholamine system. The catecholamine and melanin synthesis pathways diverge after beginning with the same substrate, L-tyrosine. Here we describe a novel relationship between the catecholamine and melanin synthesis pathways in Astyanax. Our results show significant increases in L-tyrosine, dopamine, and norepinephrine in pre-feeding larvae and adult brains of Pachón cavefish relative to surface fish. In addition, norepinephrine is elevated in cavefish adult kidneys, which contain the teleost homologs of catecholamine synthesizing adrenal cells. We further show that the oca2 gene is expressed during surface fish development but is downregulated in cavefish embryos. A key finding is that knockdown of oca2 expression in surface fish embryos delays the development of pigmented melanophores and simultaneously increases L-tyrosine and dopamine. We conclude that a potential evolutionary benefit of albinism in Astyanax cavefish may be to provide surplus L-tyrosine as a precursor for the elevated catecholamine synthesis pathway, which could be important for adaptation to the challenging cave environment.

35Understanding the molecular basis of repeated evolved phenotypes can yield key insights into 36 the evolutionary process. Quantifying the amount of gene flow between populations is especially 37 important in interpreting mechanisms of repeated phenotypic evolution, and genomic analyses 38have revealed that admixture is more common between diverging lineages than previously 39 thought. In this study, we resequenced and analyzed nearly 50 whole genomes of the Mexican 40 tetra from three blind cave populations, two surface populations, and outgroup samples. We 41confirmed that cave populations are polyphyletic and two Astyanax mexicanus lineages are 42 present in our dataset. The two lineages likely diverged ~257k generations ago, which, assuming 43 1 generation per year, is substantially younger than previous mitochondrial estimates of 5-7mya. 44