De novo assembly and curation of the surface fish genome. We set out to generate a robust reference genome for the surface morph of A. mexicanus from a single lab-reared female, descended from wild caught individuals from known Mexico localities (Fig. 1a, Supplementary Figure 1). We sequenced and assembled the genome using Pacific Biosciences single molecule real time (SMRT) sequencing (~73x genome coverage) and the wtdbg2 assembler29 to an ungapped size of 1.29 Gb. Initial scaffolding of assembled contigs was accomplished with the aid of an A. mexicanus surface fish physical map (BioNano) followed by manual assignment of 70% of the assembly scaffolds to 25 total chromosomes using the existing A. mexicanus genetic linkage map markers30. The final genome assembly, Astyanax mexicanus 2.0, comprises a total of 2,415 scaffolds (including single contig scaffolds) with N50 contig and scaffold lengths of 1.7 Mb and 35 Mb, respectively, which is comparable to other similarly sequenced and assembled teleost fishes (Supplementary Table 1). The assembled regions (394 Mb) that we were unable to assign to chromosomes were mostly due to 3.36% (2,235 markers total) of the genetic linkage markers not aligning to the surface fish genome, single markers per contig where the orientation could not be properly assigned, or markers that mapped to multiple places in the genome and thus could not be uniquely mapped. The uniquely mapped markers exhibited few ordering discrepancies and significant synteny between the linkage map30 and the assembled scaffolds of the Astyanax mexicanus 2.0 genome, validating the order and orientation of a majority of the assembly (Fig. 1b; Supplementary Fig. 2). In Astyanax mexicanus 2.0, we assemble and identify 11% more total masked repeats that in the Astyanax mexicanus 1.0.2 assembly (Supplementary Table 1). Amongst assembled teleost genomes, A. mexicanus seems to be an intermediate with 41% masked interspersed repeats estimated using WindowMasker31, compared to Xiphophorus maculatus (27%) and Danio rerio (50%). Possible mapping bias across cave population sequences to the cave versus surface fish genome references was also investigated by mapping population level resequencing reads to both genomes32. We found the number of unmapped reads is greater for all populations aligned to the Astyanax mexicanus 1.0.2 reference compared to Astyanax mexicanus 2.0 (Fig. 1c). Also, the percentage of properly paired reads, that is, pairs where both ends align to the same scaffold, is greater for all resequenced cave populations aligned to the Astyanax mexicanus 2.0 reference (Supplementary Fig. 3) and significantly more non-primary alignments with greater variation were observed (Supplementary Fig. 4). Both metrics indicate that the Astyanax mexicanus 2.0 reference has more resolved sequence regions than the Astyanax mexicanus 1.0.2 reference. Future application of phased assembly approaches will likely resolve a significantly higher proportion of chromosomal sequences in the Astyanax mexicanus genome33.
Gene annotation. Two independent sets of protein-coding genes were generated using the NCBI34 and Ensembl35 automated pipelines with similar numbers of genes found by each: 25,293 and 26,698, respectively (Supplementary Table 2). Gene annotation was aided by the diversity of transcript data derived from whole adult fish, embryos, and 12 different tissues available from the NCBI short read archive. As a result, the total predicted protein-coding genes and transcripts (mRNA) were consistent with other annotated teleost species (Supplementary Table 2) and 1,665 new protein-coding genes were added compared to Astyanax mexicanus 1.0.2. Additionally, long non-coding gene (e.g. lncRNA) representation is significantly improved in the Astyanax mexicanus 2.0 reference compared to the Astyanax mexicanus 1.0.2 reference (5,314 vs 1,062), although targeted non-coding RNA sequencing will be required to achieve annotation comparable to zebrafish (Supplementary Table 2). We assessed completeness of gene annotation by applying BUSCO (benchmarking universal single-copy ortholog) scores, which measure gene completeness among vertebrate genes. We find 94.6% of genes are complete, 4% are missing and 4.2% are duplicated (Supplementary Table 3). In addition, using NCBIs transcript aligner Splign (C++toolkit) gene annotation metrics, same-species RefSeq or GeneBank transcripts show 98.4% coverage when aligned to Astyanax mexicanus 2.0. In total, our measures of gene representation in the Astyanax mexicanus 2.0 reference show a high-quality resource for the study of A. mexicanus gene function.
Having a high-quality reference genome provides many benefits in exploiting Astyanax mexicanus as a model species. Among the more important uses, from the standpoint of utilizing the system to understand evolutionary mechanisms, are in mapping and identifying genes responsible for phenotypic change, and in unveiling genomic structural variation that provides a substrate for adaptive selection. Thus, to demonstrate the advantages Astyanax mexicanus 2.0 brings to the field, we have explored its use in each of these settings; in particular examining the genetic underpinnings of albinism and reduced eye size, and then looking at the contribution of genomic deletion to variation within Astyanax mexicanus populations.
First with respect to identification of evolutionary important genes, we took advantage of a critical attribute of A. mexicanus, nearly unique among cave animals: the ability to interbreed cavefish and surface fish populations to generate fertile hybrids that can be intercrossed to perform QTL studies (for review see36,37). The improved contig length of the surface fish genome and our syntenic analysis that unites the physical genome to a previously published linkage map based on genotyping-by-sequencing (GBS) markers30, should greatly aid the correct identification of genetic changes linked to cave traits. To this end, we demonstrate the power of the Astyanax mexicanus 2.0 reference in gaining deeper insight into an already mapped trait (albinism) and in carrying out new QTL analysis of eye reduction.
QTL analysis of albinism. Albinism has previously been mapped in QTL studies of both the Pachòn and Molino populations20. This work was carried out before the existence of any reference genome for the species. However, these mapping studies showed that a single major QTL in each population colocalized with a candidate gene oca2. Oca2 loss of function mutations are known to cause albinism in other organisms, including humans, mice and zebrafish, and deletions causing functional inactivation of Oca2 were identified in the albino fish from both the Molino and Pachón caves20. These compelling data indicating that oca2 mutations are causal of albinism in cavefish were subsequently confirmed by CRISPR-mediated mutagenesis in surface fish27. To build on these results, we first performed two separate de novo QTL analyses of surface/Pachón F2 hybrids (group A and B) in the context of the new reference genome (Fig. 2a). Both studies identified a single QTL for albinism, with a LOD score of 47.31 at the peak marker on linkage group 3 (69% variance explained) in group A, and with LOD score of 22.56 at the peak marker on linkage group 21 (37.8% variance explained) in group B (Fig. 2c, d; 2g, h). At the peak QTL positions, F2 hybrids homozygous for the cave allele are albino (Fig. 2e, i). Mapping the markers associated with each of these linkage groups to the surface fish genome revealed that they are both located on surface fish chromosome 13 (Table 1, Fig. 2f, j). This demonstrates a significant improvement over mapping to the original cavefish genome: Linkage group 3 from group A almost completely corresponds to surface fish chromosome 13 in Astyanax mexicanus 2.0, whereas it is split up into many contigs in Astyanax mexicanus 1.0.2 (Fig. 2f, j, Supplementary Fig. 5a). The QTL identified on linkage group 21 in group B corresponds to a 2.8 Mb region on surface fish chromosome 13 as determined using the sequences flanking the 1.5-LOD support interval as input for Ensembl basic local alignment search tool (Table 1, Fig. 2J). In line with previous mapping studies, the gene oca2 is found within this region on surface fish chromosome 13. We further functionally verified that a change in the oca2 locus is responsible for albinism in the F2 hybrids, by crossing an albino surface/Pachón F2 hybrid with a genetically-engineered surface fish heterozygous for a deletion in oca2 exon 2127. We found that 46.5% of the offspring completely lacked pigment (n=40/86, Fig. 2b). This confirms that a change in the oca2 locus is responsible for loss of pigment in the mapping population.
Table 1. QTL mapping of albinism and eye size in two groups of surface/Pachón F2 hybrids. Table comparing QTL marker positions on linkage maps, Pachón genome scaffolds, and surface fish genome chromosomes (1 Astyanax mexicanus 1.0.2; 2 Astyanax mexicanus 2.0).
F2
Surface/
Pachón
|
Phenotype
|
Marker
|
LOD
|
Linkage group position (cM)
|
Pachón scaffold position1
|
Surface
chromosome position2
|
group A
|
albinism
|
m6036
|
45.03
|
3:22.6
|
KB882185: 1429320
|
13: 35611239
|
group A
|
albinism
|
m17241
|
47.31
|
3:22.6
|
KB882122: 1818424
|
13: 38374549
|
group A
|
albinism
|
m36842
|
45.16
|
3:22.8
|
KB882122: 2246330
|
13: 37945372
|
group B
|
albinism
|
r219177
|
20.67
|
22:22.5
|
KB882122: 1091706
|
13: 39098911
|
group B
|
albinism
|
r219695
|
22.56
|
22:24.6
|
KB882122: 2349092
|
13:37844883
|
group B
|
albinism
|
r220017
|
20.98
|
22:25.0
|
KB882122: 3270816
|
13: 41898580
|
group A
|
eye size
|
m23384
|
3.39
|
1:62.6
|
KB882121: 3061587
|
3: 23014157
|
group A
|
eye size
|
m15225
|
5.46
|
1:66.8
|
KB882287: 356685
|
3: 9301997
|
group A
|
eye size
|
m30438
|
3.89
|
1:67.9
|
KB872232: 7153
|
3: 8742127
|
group B
|
eye size
|
r230947
|
8.45
|
15:31.0
|
KB882132: 1682036
|
20: 1802669
|
group B
|
eye size
|
r231141
|
11.98
|
15:35.9
|
KB882132: 2321700
|
20: 1169105
|
New insights on albinism locus. While up until this point our reanalysis of albinism was largely confirmatory, an advantage of using the surface fish genome as a reference, compared to the previously available cavefish genome, is the ability to make comparisons between regions that are deleted in cavefish and as such are not available for sequence alignment. We utilized the A. mexicanus 2.0 reference genome to align sequencing data obtained from wild-caught fish from different surface and cave localities32 and analyzed the oca2 locus (Supplementary Fig. 6,7). We found that 9 out of 10 Pachón cavefish carry the deletion in exon 24 that was previously reported in laboratory-raised fish (Supplementary Fig. 6;20). None of the individuals from other populations for which sequence information was available carried the same deletion. Consistent with the laboratory strains, exon 21 of oca2 is absent in all Molino cavefish samples (n=9, Supplementary Fig. 7). None of the other sequenced population samples harbor the same deletion of exon 21, however we found smaller heterozygous or homozygous deletions in exon 21 in some wild samples of Pachón (n=5/9), Río Choy (n=1/9), and Tinaja (5/9) fish (Supplementary Fig. 7). In summary, we were able to detect deletions in oca2 that would have not been discovered with alignments to Astyanax mexicanus 1.0.2 since exon 24 is missing in Pachón and thus no reference- based alignments were produced in that region.
Surface fish genome reveals new candidate genes from prior QTL studies. A prominent feature of cavefish is the absence of external eyes. While a number of eye size QTL have been identified18,19, there is an incomplete understanding of the genetic basis of eye regression. We uncovered additional information by remapping previously published QTL18-24 using the Astyanax mexicanus 2.0 reference. A total of 1,060 out of 1,124 markers (94.3%) mapped successfully with BLAST and were included in our surface fish QTL database. There were 77 markers that did not map to the cavefish genome28 but did map to the surface fish genome, 52 markers that mapped to the cavefish but not the Astyanax mexicanus 2.0 reference, and 12 markers that did not map to either reference. The improved contiguity of the chromosome-level surface fish assembly allowed us to identify several additional candidate genes associated with QTL markers that were not previously identified in the more fragmented cavefish genome (Supplementary Data 1). For example, the markers Am205D and Am208E mapped to an approximately 1 Mb region of chromosome 6 of the surface fish genome (46,516,926 - 47,425,713 bp) but did not map to the Astyanax mexicanus 1.0.2 reference. This region is associated with feeding angle21, eye size, vibration attraction behavior (VAB), suborbital neuromasts23, and maxillary tooth number19. Multiple strong candidate genes involved in eye development or morphology are contained in this region including rhodopsin, mib2, and ubiad1, as well as GABA A receptor delta which is associated with a variety of behaviors and could conceivably be involved in VAB. Notably, the scaffold containing these four genes in Astyanax mexicanus 1.0.2 (KB871939.1) was not linked to this QTL, demonstrating the utility of increased contiguity of Astyanax mexicanus 2.0.
Previous studies suggested that the gene retinal homeobox gene 3 (rx3) lies within the QTL for outer plexiform layer of the eye24. Another QTL for eye size, size of the third suborbital bone, and body condition18,19, may also contain rx3 (low marker density and low power of older studies, result in a broad QTL critical region in this area). The increased contiguity of Astyanax mexicanus 2.0 revealed that rx3 is within the region encompassed by this QTL, whereas in Astyanax mexicanus 1.0.2 the marker for this QTL (Am55A) and rx3 were located on separate scaffolds; thus, we could not appreciate that this QTL and key gene for eye development were in relatively close genomic proximity. While no amino acid coding changes are apparent between cavefish and surface fish, expression of rx3 is reduced in Pachón cavefish relative to surface fish28. In zebrafish, rx3 is expressed in the eye field of the anterior neural plate during gastrulation and has an essential role for the fate specification between eye and telencephalon38,39. We have compared rx3 expression at the end of gastrulation and confirmed Pachón embryos have reduced expression domain size (Fig. 3a, b). The expression area is significantly smaller in Pachón embryos compared to stage-matched surface fish embryos. The expression of rx3 is restored in F1 hybrids between cavefish and surface fish, indicating a recessive inheritance in cavefish (Fig. 3a, b). To test for a putative role of rx3 in eye development in Astyanax mexicanus we used CRISPR/Cas9 to mutate this gene in surface fish (Fig. 3c), and assessed injected, CRISPant fish for eye phenotypes. Wild- type surface fish have large eyes (Fig. 3d). In contrast, externally visible eyes are completely absent in adult CRISPant surface fish (n=5, Fig.3d). This is consistent with work from other species, in which mutations in rx3 (fish) or Rx (mice) result in a complete lack of eyes40-42. Together these data suggest that the role of rx3 in eye development is conserved in A. mexicanus. Further, they support the hypothesis that regulatory changes in this gene may contribute to eye loss in cavefish through specification of a smaller eye field, and subsequently, production of a smaller eye.
In addition to the compiled database of older QTL studies, a number of genomic intervals associated with previously described locomotor activity difference between cavefish and surface fish25 were also re-screened using the surface-anchored locations of markers from the high-density linkage map30. Within this Pachón/surface QTL map we confirmed the presence of 20 previously reported candidate genes, and identified 96 additional genes with relevant GO terms, including rx3, further demonstrating the power and utility of this new genomic resource (Supplementary Table 4). The new candidates include additional opsins (opn7a, opn8a, opn8b, tmtopsa, and two putative green-sensitive opsins), as well as several genes contributing to circadian rhythmicity (id2b, nfil3-5, cipcb, clocka, and npas2). While analyses of expression data and sequence variation are necessary to determine which of these candidates exhibit meaningful differences between morphs, the presence of clocka and npas2 in these intervals is of particular note, as the original analysis conducted using Astyanax mexicanus 1.0.2 did not provide any evidence of a potential role for members of the core circadian clockwork in mediating observed differences in locomotor activity patterns between Pachón and surface fish15,25.
Genetic mapping with surface fish genome reveals new candidate genes for eye regression.
Finally, to gain new insight into the eye reduction phenotype, we used the Astyanax mexicanus 2.0 reference to de novo genetically map eye size in the two surface/Pachón F2 groups that we used to map albinism. In surface/Pachón F2 mapping population A (n=188) we identified multiple QTL for normalized eye perimeter that were spread across four linkage groups (Fig. 4a). The QTL on linkage group 1 is significant above a threshold of p < 0.01 (Fig. 4b). The surface allele at the peak marker appears to be dominant since eye size in the heterozygous state is similar to the homozygous state (Fig. 4c). Mapping linkage group 1 to the Astyanax mexicanus 1.0.2 genome assembly results in markers under the QTL peak spread across different contigs, whereas these map to chromosome 3 in the new surface fish assembly, emphasizing how the improved quality of the surface fish genome allows the identification of candidate genes throughout the QTL region (Fig. 4d, Supplementary Fig. 5b, Table 1). This region has been identified previously18,28 and contains genes such as shisa2a and shisa2b, as well as eya1. Analysis of the region between the markers with the highest LOD scores (3:9301997-9505868), however, revealed one gene that has not been linked to eye loss in A. mexicanus before, ENSAMXG00000005961 (Fig. 4d). Alignment of this novel gene sequence showed homology to orofacial cleft 1 (ofcc1), also called ojoplano (opo). In medaka, opo has been shown to be involved in eye development. When knocked out, opo affects the morphogenesis of several epithelial tissues, including impairment of optic cup folding which resulted in abnormal morphology of both the lens and neural retina in the embryos43. Sequence comparisons of this gene in Pachón cavefish and surface fish revealed several coding changes, however, none affect evolutionarily conserved residues. Future studies are needed to investigate a putative role of opo in the eye loss of Pachón cavefish.
In the surface/Pachón F2 mapping population B (n=219) we identified a single QTL for normalized left eye diameter on linkage group 13 with a LOD score of 11.98 at the peak marker that explains 24.3% of the variance in this trait (Fig. 4e, f). F2 hybrids homozygous for the cave allele at this position have the smallest eye size, and the heterozygous state is intermediate (Fig. 4g). The values for left and right eye diameter mapped to the same region and, notably, we obtain the same peak when including eyeless fish in the map and coding eye phenotype as a binary trait (i.e. eyed, eye- less). The QTL on linkage group 13 corresponds to a 633 KB region on surface fish chromosome 13 as determined using the sequences flanking the 1.5-LOD support interval as input for Ensembl basic local alignment search tool (Fig. 4h, Table 1). There are 22 genes in this region (Supplementary Table 5). Of note, none of the previously mapped eye size QTL in Pachón cavefish map to the same region44. A promising candidate gene in this interval is dusp26. Morpholino knock-down of dusp26 in zebrafish results in small eyes with defective retina development and a less developed lens during embryogenesis45. We used whole genome sequencing data32 to compare the dusp26 coding region between surface, Tinaja, Pachón and Molino and found no coding changes. However, previously published embryonic transcriptome data indicates that expression of dusp26 is reduced in Pachón cavefish at 36 hpf (p<0.05) and 72 hpf (p<0.01) (Supplementary Fig. 8)46. These data suggest a potential role for dusp26 in eye degeneration in cavefish, however, we cannot exclude critical contributions of other genes or genomic regions in the identified interval.
Structural variation in cave populations. Another important advantage of having a robust reference genome is that it allows one to interrogate the genomic structural variation at a population level. Knowledge of population-specific A. mexicanus structural sequence variation is lacking.
Therefore, we aligned the population samples from Herman et al.32 against Astyanax mexicanus 2.0 to ascertain the comparative state of deletions. We used the SV callers Manta47 and LUMPY48 to count the numbers of deletions present in each sample compared to Astyanax mexicanus 2.0 (Supplementary Fig. 9). While LUMPY tended to call a larger number of short deletions and Manta a smaller number of long deletions, there was high correlation (R2=0.78) between the number of calls each made per sample (Supplementary Fig. 10), so we used the intersection of deletions called by both callers for further analysis. We then classified these deletions based on their effect (i.e. deletions of coding, intronic, regulatory, or intergenic sequence) (Supplementary Fig. 11). Among the cavefish populations measured for deletion events, 412 genes contained deletions with an allele frequency >5%. We found that the Molino population has the fewest heterozygous deletions, while Río Choy surface fish have the least homozygous deletions, mirroring the heterozygosity of single nucleotide polymorphisms32 (Supplementary Fig. 11). In addition, the Tinaja population showed the most individual variability of either allelic state (standard deviation of 427). Pachón and Tinaja contained the highest number of protein-coding genes altered by a deletion in at least one haplotype, while Río Choy had the least (Fig. 5a). In two examples, per3 and ephx2, we find the deletions that presumably altered protein-coding gene function varied in population representation, number of bases affected, and haplotype state for each (Fig. 5b, c). Of the 412 genes that contained deletions, 109 have assigned gene ontology in cavefish (Supplementary Table 6). We tested these 109 genes for canonical pathway enrichment using WebGestalt49 and found genes significantly enriched (p<0.05) for AMPK and MAPK signaling, as well as metabolic and circadian clock function (Table 2). In addition, some genes were linked to diseases consistent with cavefish phenotypes including: ephx2, which is linked to familial hypercholesterolemia, or hnf4a, which is linked with non-insulin dependent diabetes mellitus, based on the gene ontology of OMIM and DisGeNET50. Notably, these enrichment analyses recovered the deletion known to be the main cause of pigment degeneration in cavefish, oca227. Further validation of these disease phenotype or pathway inferences is warranted.
Table 2. Protein-coding genes altered by deletion events and their enrichment among canonical pathways and disease. Red text denotes significant tests for disease enrichment (see methods).
Gene Pathway or Disease Association
|
Genes
|
Database
|
P value
|
AMPK signaling
|
hnf4a;lipe;ppp2r5a
|
KEGG
|
0.01
|
Wnt signaling
|
csnk2b;hltf;map3k7cl;pcdh9;ppp2r5a
|
Panther
|
0.01
|
MAPK signaling
|
gadd45a;mapkapk3
|
Panther
|
0.01
|
Metabolism
|
cmpk1;cmpk2;ephx2;mtmr6;ndufa4l2;pig p;prps1;sgms1;sms;uxs1
|
KEGG
|
0.02
|
Circadian clock
|
per3
|
Panther
|
0.05
|
Hypercholesterolemia
|
ephx2
|
OMIM
|
0.005
|
Diabetes mellitus
|
nnf4a
|
OMIM
|
0.02
|
Albinism
|
pp3b1;oca2
|
DisGeNET
|
0.002
|
Sialorrhea
|
ikzf1;prps1;vps13a
|
DisGeNET
|
0.001
|
Coding mutations affecting hypocretin signaling. We next used the surface fish genome to compare amino acid composition of key genes hypothesized to be involved in cave-specific adaptations. The wake-promoting hypothalamic neuropeptide Hypocretin is a critical regulator of sleep in animals ranging from zebrafish to mammals51-53. We previously found that expression of hypocretin was elevated in Pachón cavefish compared to surface fish and pharmacological inhibitors of Hypocretin signaling restore sleep to cavefish, suggesting enhanced Hypocretin signaling underlies the evolution of sleep loss in cavefish9. In teleost fish, Hypocretin signals through a single receptor, the Hypocretin Receptor 2 (Hcrtr2). We compared the sequence of hcrtr2 in surface fish and Pachón cavefish and identified two missense mutations that result in protein coding changes, S293V and E295K (Fig. 6a). To examine whether these were specific to Pachón cavefish or shared in other cavefish populations we examined the hcrtr2 coding sequence in Tinaja and Molino cavefish. The mutation affecting amino acid 295 is shared between Pachón and Tinaja populations (Fig. 6a). Further, we identified a six base pair deletion that results in the loss of two amino acids (amino acid 140 and 141; SV) in Molino cavefish. The presence of these variants was validated by PCR and subsequent Sanger sequencing on DNA from individuals from laboratory populations of these fish. The E295K variant in Tinaja and Pachón as well as the two amino acid deletions in Molino affect evolutionarily conserved amino acids suggesting a potential impact on protein function (Fig. 6b). We performed an in-silico analysis to test whether the identified cavefish variants in hcrtr2 are predicted to affect the protein structure and stability of Hcrtr2. We used iStable54 to test for potential destabilization effects of the substitution mutations found in Tinaja (E295K) and Pachón (S293V and E295K). iStable predicted a destabilization effect of E295K on the protein structure with a confidence score of 0.842. However, we found a stabilization effect of the S293V mutation in Pachón using iStable with a confidence score of 0.772. We repeated this analysis using MUpro55 and obtained similar results. These findings raise the possibility that the evolved changes differentially affect the function of the same receptor. Structural changes in proteins upon amino acid deletion are difficult to predict with common tools such as iStable and MUpro. To analyze whether the deletion of S140 and V141 in the Molino hcrtr2 could potentially influence the structural integrity of Hcrtr2, we performed a different analysis. We used the SWISS-MODEL protein structure prediction tool56 to identify potential differences in the protein structure between surface and Molino Hcrtr2. We modeled the surface and Molino Hcrtr2 protein using crystal data from the human HCRTR2 (5wqc;57). We then used the VMD visualization software to overlay the surface fish and Molino predicted structure. This analysis indicates that the deletion of S140 and V141 disrupts the structural integrity of an alpha helical structure in the transmembrane region of Hcrtr2 that could potentially affect the stability of this receptor (Fig. 6c). We also tested the Tinaja and Pachón hcrtr2 sequences using a similar approach and found only minor differences between the respective cavefish Hcrtr2 structure when overlaid with the surface fish structure (Fig. 6c). To confirm these potential structural changes in cavefish Hcrtr2 advanced in-situ and in-vivo protein analysis need to be performed in future studies, however, the identification of coding mutations in three different populations of cavefish supports the notion that hypocretin signaling is under selection at the level of receptor. This is in line with population sequencing data32, which identified hcrtr2 in the top 5% of FST outliers between surface fish (Rascon) and Pachón cavefish.