Keywords

1 Introduction

Pigmentation varies dramatically across the animal kingdom—from crypsis to colorful ornamental displays—suggesting that coloration serves dynamic and adaptive functions. In many animals, these roles vary from mate choice selection (Protas and Patel 2008), defense from predation (Linnen et al. 2009), UV protection, structural support, and thermoregulation (Hubbard et al. 2010). Pigmentation traits have also served as a powerful approach for linking specific genes to phenotypic characters (Hoekstra 2006).

In this study, we investigated the naturally occurring pigmentation loss in the blind cavefish Astyanax mexicanus (Jeffery 2001; Borowsky 2008). This species harbors two distinct morphotypes: a pigmented surface-dwelling form that populates the rivers of NE Mexico, and several depigmented (or albino) cave-dwelling morphs that reside in the subterranean environment. This species demonstrates recurrent loss since Astyanax cave morphs have repeatedly colonized the cave, providing natural biological “replicates” (Gross 2012a).

The gene underlying the absence of melanin (albinism) was identified and confirmed by CRISPR mutagenesis in surface fish, as Oca2 in two independent cavefish lineages (Protas et al. 2005; Gross and Wilkens 2013; Ma et al. 2015; Klaassen et al. 2018). A second pigmentation phenotype, brown, is associated with the gene Mc1r in three cavefish populations (Gross et al. 2009). Although these studies have discovered the genetic basis for monogenic components of pigmentation loss, the genes contributing to complex pigmentation loss in cavefish have not been identified. A prior mapping study in Astyanax mexicanus did confirm that melanophore (pigment cell) numerical variation is indeed complex and linked to 18 QTL associated with pigment cell number, yet the identity of the genes underlying this trait still remains unknown (Protas et al. 2007).

The previous characterization of melanogenesis in other animals shows that melanophores are derived from a set of migratory cells that give rise to numerous cell types including cranial cartilage and bone, peripheral neurons, fat cells, and pigment-producing melanophores (Erickson and Perris 1993; Huang and Saint-Jeannet 2004). Due to the diversity of neural crest cell derivatives, it would be less likely to acquire mutations within genes of the neural crest pathway due to potentially lethal consequences (Jeffery 2009). Labeling experiments revealed normal neural crest migration during cavefish development (McCauley et al. 2004), and quantification of cell apoptosis after neural crest-derived precursor migration showed comparable numbers in both cavefish and surface (Jeffery 2006). These combined results suggest that evolutionary changes leading to pigment cell regression in cave morphs may be mediated by alterations late in melanogenesis (Jeffery 2009).

To identify pigmentation-related genes, we employed a second-generation linkage map (Carlson et al. 2015) inclusive of >3,000 genomic markers to perform high-resolution mapping of melanophore number diversity in a large cave × surface F2 pedigree (Fig. 1a–e). A quantitative trait locus (QTL) mapping study yielded numerous significant associations linked with 20 different regions of our linkage map. We then leveraged available genomics resources (McGaugh et al. 2014) to nominate candidate genes. Comparative genomics identified the critical syntenic region for each QTL in the Astyanax cavefish draft genome alongside conserved intervals in the distantly related zebrafish genome. We nominated candidate genes by screening the genes within these syntenic regions for gene ontology (GO) terms related to pigmentation. These analyses yielded two genes, Tyrp1b and Pmela, with well-characterized roles in melanin-based pigmentation in other animals. We further characterized the coding sequence and expression of candidate genes. Through these studies, we propose Tyrp1b and Pmela as genes that likely contribute to complex pigmentation in Astyanax cavefish. Identification of additional pigmentation genes provides a clearer picture of the mechanisms contributing to regressive evolution in Astyanax and informs our understanding of the broader principles governing trait loss in the natural world.

Fig. 1
figure 1

Surface × Pachón cave F2 sibling hybrids display a vast array of coloration. Hybrid offspring from a surface × Pachón cave cross reveal a varying degree of melanic-based pigmentation. The “dark” versus “light” appearance is associated with the number of melanophores (pigment cells) that an individual harbors. We observed levels of pigmentation that are darker (a) than normal surface fish (b). Some hybrid individuals showed dramatic reductions in pigmentation (c, d), while albino individuals (which still retain melanophores) produce no melanin, rendering melanophores invisible (e)

Full size image

2 Materials and Methods

2.1 Melanophore Scoring

To quantify melanophore number, we analyzed numerous regions of the body where pigment cell number varied within our F2 pedigree. These included seven regions: near the anal fin (MelAnalfinSquare), below the mid-lateral stripe (MelUnderStripe), dorsal square (MelDorsalSquare), area above the stripe (MelAboveStripe), above the eye (MelHeadSquare), full head (MelHead), and neighboring the anal fin (MelAnalfinTriangle). Regions were selected based on a consistent set of landmarks for reproducibility and were similar to the areas previously assayed (Protas et al. 2007). For automated counts, we employed ImageJ (v.1.6; National Institutes of Health, Bethesda, MD) by inverting the color in a selected area and then counting the lighter objects (e.g., pigment cells, now white) with a preset “noise tolerance.” The noise tolerance was set so only markings (melanophores in this case) above the preset light pixel intensity were counted. Each image was reviewed and any melanophores “missed” in the automatic quantification were manually added. When appropriate, we transformed the melanophore counts to log10 values to generate a normal distribution for association studies.

2.2 Quantitative Trait Locus (QTL) Association Mapping

All QTL analyses were performed using a previously published linkage map (Carlson et al. 2015). We employed the software program R/qtl (v.1.30; Broman et al. 2003) for all association analyses. We analyzed each trait using four mapping methods: marker regression (MR; Kearsey and Hyne 1994), expectation maximization (EM; Xu and Hu 2010), Haley-Knott (HK; Haley and Knott 1992), and nonparametric (NP; Kruglyak and Lander 1995) as described in Gross et al. (2014). Significant linkages were set at a LOD score threshold of ≥4.0—as used in other QTL studies (Protas et al. 2007; Gross et al. 2009). To confirm associations, permutation tests involving 1000 iterations were performed to identify statistically significant QTL (P < 0.05). Effect plots for associations were generated using the closest linked genetic marker. QTL regions were then anchored (~6–8 cM on each side of the top marker) using the NCBI BLAST Toolkit (v.2.28+) to the Astyanax genome (Ensembl build v.75; McGaugh et al. 2014). We also determined the syntenic interval in the zebrafish genome, with previously demonstrated synteny with Astyanax (Gross et al. 2008; O’Quin et al. 2013; Carlson et al. 2015). Visual representations of synteny between our linkage map and the genomes were created with Circos (v.0.64; Krzywinski et al. 2009; Fig. 3a, d).

2.3 Gene Ontology (GO) Term Analysis

To nominate prospective candidate genes, we interrogated all genes within the syntenic interval in the Astyanax draft genome for gene ontology (GO) terms. From these analyses, we collected the GO terms for all genes located in the Astyanax syntenic interval using BioMart (v.0.8; Kasprzyk 2011). This approach yielded hundreds of GO terms for each significant QTL. We then narrowed our search to terms with potential involvement in pigmentation, such as “pigment”, “pigmentation”, “melanin”, “eumelanin”, “phaeomelanin”, “melanophore”, “melanocyte”, “melanosome”, “xanthophore”, “iridophore”, “chromatophore”, and “carotene”. This approach enabled the selection of genes based on annotation information from known functions in other organisms.

2.4 RNA-Seq, Qualitative, and QPCR Expression Analyses

We evaluated genes located within predicted critical genomic regions for expression differences using RNA-seq. Total RNA was isolated from pools of 50 surface or cave individuals using the RNeasy Plus Mini Kit (Qiagen) at each of five developmental stages. These included 10 hours post-fertilization (hpf), 24, 36, and 72 hpf, and from three individuals during juvenilehood (~4 months). Library preparation (TruSeq v.2 kit) and sequencing (Illumina 2500 Hi-Seq) was performed in triplicate (10–72 hpf) or duplicate (juvenile) at the DNA Sequencing Core (Cincinnati Children’s Hospital and Medical Center). All samples were sequenced to a ~10 million read depth for 50-bp, single-end reads. Normalized gene expression was calculated using ArrayStar (DNAStar). All expressions were evaluated using comparative read counts between cave and surface fish with the RPKM normalization method (Mortazavi et al. 2008). Raw sequencing reads are deposited at the NCBI SRA (BioProject: PRJNA258661).

We validated expression profiles at the 72 hpf stage using qualitative and quantitative PCR analyses, described in Stahl and Gross (2017). Template cDNA from surface and cavefish RNA pools (n = 50 embryos each; RNeasy Plus Mini kit, Qiagen) was synthesized for both experiments using the Transcriptor RT kit (Roche). Quantitative PCR (qPCR) experiments were performed as described in Stahl and Gross (2017). All samples were analyzed in sextuplet and normalized expression values (Cq) and significant differences (two-tailed Student’s t-test) were determined using the CFX Manager software program (v.3.1; BioRad).

Genes residing within the syntenic region in the Astyanax genome were analyzed for sequence alterations. Sequencing reads derived from surface fish and Pachón cavefish (~280 million reads total) were aligned to the draft Astyanax genome (Ensembl.v.75; McGaugh et al. 2014) using default parameters for the program SeqMan NGen (DNASTAR) and evaluated for diverse sequence mutations (e.g., SNP, indels) segregating between surface and cave morphs.

2.5 Whole-Mount In Situ Hybridization

RNA probes for in situ hybridization were generated by PCR for the genes Tyrp1b (forward primer: 5′-GAAACAGCCCTCAGTTCGAG-3′, reverse primer: 5′-AGGTGGGCCAGATTGTGTAG-3′) and Pmela (forward primer: 5′-CTACTGATGCTGCCACTGGA-3′, reverse primer: 5′-AGAGCCGTAGCGGTAGATCA-3′). The resulting PCR products were cloned into the TOPO TA Dual Promoter cloning vector (Life Technologies) and confirmed by sequencing. Sense and antisense digoxygenin (DIG)-labeled RNA probes for Astyanax Tyrp1b and Pmela were transcribed with either SP6 or T7 RNA polymerase (Roche). Whole-mount in situ hybridizations were performed on embryos at stages <12 hours post-fertilization (hpf), 24, 36, and 72 hpf according to Ma et al. (2014). Embryos were washed in NTM, PBT, and TBST and fixed in 4% PFA/PBS for storage at 4 °C and visualized using Leica Microscope M205 FA (with LAS software v.3.8.0) montage imaging (Fig. 5a–d). The numbers of cells with positive expression for Pmela or Tyrp1b, respectively, were counted.

2.6 Functional Validation in Zebrafish with MO Knockdowns

The translational blocking morpholino oligonucleotides were targeted to the first 25 base pairs of the zebrafish ORF of Pmela (5′-GAGGAAGATGAGAGATGTCCACAT-3′) and Tyrp1b (5′-GCACTAAACACACACTCTTCCACAT-3′; Gene Tools, LLC.). Morpholinos were injected into one-cell stage zebrafish embryos in a 1 nl volume at a 0.2 mM concentration. Control individuals were administered a mock injection composed of phenol red and Danieaux’s solution (Wingert et al. 2004), or a control oligo (5′-CCTCTTACCTCAGTTACAATTTATA-3′; Gene Tools, LLC). Imaging was performed using a Leica Microscope M205 FA (with LAS software v.3.8.0) stereoscope (Fig. 5e–g). Phenotypic analysis was performed on embryos fixed at 5dpf in 4% PFA. Counts were performed on the yolk sac of the number of light and dark melanophores, and the number of clustered and isolated melanophores, both relative to total melanophore number on the yolk sac (Fig. 5h, i).

3 Results

3.1 Cavefish × Surface Fish Hybrid Individuals Demonstrate Diversity in Melanophore Numbers in Distinct Regions Spanning the Body

Direct observations reveal a wide array of pigmentation variation among F2 siblings (Fig. 1a–e), and substantial pigment cell variation has been reported in four different body regions (Protas et al. 2007). We quantified melanophore number in seven areas and found variation across our F2 population (n = 170). Counts ranged within each region (reported as minimum to max number melanophore distribution for a given region among the entire F2 pedigree): 1–99 melanophores in the square near the anal fin (MelAnalfinSquare), 1–113 pigment cells below the mid-lateral stripe (MelUnderStripe), 4–600 cells in the dorsal square (MelDorsalSquare), 1–2,762 melanophores in the area above the stripe (MelAboveStripe), 1–89 pigment cells above the eye (MelHeadSquare), 11–1,395 melanophores present in the full head region (MelHead), and 0–252 pigment cells by the anal fin (MelAnalfinTriangle). Numerical variation in melanophore numbers did not simply increase or decrease proportionately across all regions of the body, instead pigment cell numbers in some cases varied independently between different assayed regions.

3.2 QTL Analysis Revealed 19 Genomic Regions Associated with Complex Melanophore Variation

We scored melanophore variation in seven regions and detected numerous QTL (Fig. 2a). We did not discover QTL for melanophores on the entire head (MelHead). This was surprising since we detected QTL for pigment cell counts in the head square above the eye (MelHeadSquare). Additionally using four different mapping methods, our results yielded multiple significant (independent) associations (n = 41 markers) with the linkage map for the following pigment cell number traits: MelAnalfinSquare (n = 1 QTL), MelUnderStripe (n = 13 QTL; Fig. 2d, e), MelDorsalSquare (n = 3 QTL; Fig. 2b, c), MelAboveStripe (n = 1 QTL), MelHeadSquare (n = 3 QTL), and MelAnalfinTriangle (n = 6 QTL). In some instances, the different mapping methods identified significant associations with same top marker (e.g., MR, EM, HK, and NP methods all noted the marker ASTYANAX_414 at single map location (LG 16, 22.2 cM for MelUnderStripe), and in other cases, the different approaches yielded multiple significant markers within a small positional window—for example, threemarkers for each MR, EM, and HK respective method for MelDorsalSquare on LG 20 (53–61 cM). Not surprisingly, some of these QTL co-localized near the same position in our linkage map. One “hotspot” in our map was on linkage group 1 from 75 to 93 cM, wherein associations with five traits (MelHeadSquare, MelUnderStripe, MelAnalfinTriangle, MelDorsalSquare, MelAnalfinSquare) yielded significant QTL (Fig. 2a). Moreover, three different regions assayed (MelDorsalSquare, MelAnalfinTriangle, MelAboveStripe) returned associations with linkage group (LG) 20 (53–61 cM), and two traits (MelAnalfinTriangle, MelUnderStripe) tended to co-localize together at several positions: LG 2 (29.8–33.6 cM), LG 7 (54.64–63.7 cM) and LG 26 (8.98–18 cM). In total, we discovered 19 distinct regions (QTLs) of the genome are associated with melanophore variation (Fig. 2a).

Fig. 2
figure 2

High-resolution mapping identified multiple QTL associations with melanophore number variation. Our quantitative trait locus (QTL) study detected numerous associations (n = 40) with pigment cell variation (a). Many of these co-localized to similar positions on our linkage map. With this approach, we were able to capture the dramatic variation of pigment cell number, including the range of melanophores detected in the dorsal square (b) and under the mid-lateral stripe (d). The frequency denotes the number of individuals harboring a given number of melanophores. The markers underlying these two QTL demonstrated a significant effect of phenotype, where individuals harboring two copies of the surface allele “S/S” demonstrated increased numbers of melanophores, and hybrids with two cavefish alleles “C/C” exhibited fewer pigment cells. This effect is evident in both the dorsal (c) and stripe (e) regions assayed

Full size image

3.3 Comparative Analyses Narrowed Melanophore QTL Positions to Critical Genomic Regions in Astyanax and Danio

To further search for candidate genes residing near our top markers, we analyzed the positions of each QTL in the Astyanax cavefish draft genome (~10,000 scaffolds) and the current zebrafish genome (Ensembl.Zv9). First, we queried marker sequences to the Astyanax genome (Ensembl.v.75) using the standard nucleotide BLAST algorithm to find the position of the top genomic marker, and the locations of the markers within the QTL interval (immediately adjacent ~6–8 cM to the top marker), to identify regions of synteny. Since these scaffolds varied in length (876–9,823,298 bp), the number of scaffolds reported back from our BLAST search ranged from 4 to 27 scaffolds for each association (data not shown).

We also capitalized on the well-annotated genome of teleost fish Danio rerio. Three comparative studies previously identified large syntenic regions shared between Astyanax and Danio, which diverged ~120 MYa (Gross et al. 2008; Gross 2012b; O’Quin et al. 2013; Carlson et al. 2015). Our genomic comparisons revealed several syntenic blocks based on the positions of BLAST hits in Danio, yielding regions of synteny on different zebrafish chromosomes. We sought to further support the syntenic block from our direct comparisons of marker sequences to zebrafish, since usually <20 markers yielded direct BLAST hits for each QTL. Accordingly, we collected every predicted gene (often 100+ genes) from the respective Astyanax scaffolds associated with each QTL and queried the full-length gene sequences to the zebrafish genome.

3.4 Tyrp1b and Pmela Are Two Candidate Genes Associated with Numerical Melanophore Diversity

Nearly all of the assayed pigmentation traits yielded multiple QTL associated with multiple distinct regions of the linkage map (Fig. 2a). We characterized the wide range of melanophore number variation in our pedigree, including the dorsal square (Fig. 2b) and inferior to the mid-lateral stripe (Fig. 2d). Our association studies revealed multiple loci for these traits, including the markers ASTYANAX_28 on LG 20 (LODMR = 6.01, p = 0.015) and TP33309 on LG 26 (LODMR = 5.08, p = 0.041), respectively (see red circles in Fig. 2a). Corresponding effect plots demonstrated a significant effect of genotype for each marker: homozygous surface alleles “S/S” harbor more melanophores and homozygous cave alleles “C/C” have fewer melanophores (Fig. 2c, e). The heterozygous genotype for ASTYANAX_28 presented an intermediate number of melanophores at this locus, whereas the heterozygous genotype at TP33309 has melanophore numbers similar to the homozygous cave genotype, perhaps indicating the dominance of the cave allele in hybrid individuals.

To pursue candidate genes, we collected the ontology for genes within the syntenic blocks. For the QTL association on LG 20 (marker: ASTYANAX_28), we identified 951 terms affiliated with 157 genes on 10 Astyanax scaffolds, and 868 terms associated with 193 genes from the syntenic interval anchoring to chromosome 1 in zebrafish. Within this dataset, we identified two GO terms related to pigmentation: “melanosome membrane” (GO:0033162) and “melanin biosynthetic process” (GO:0042438) of which both terms are associated with two genes Tyrp1b and Tyrp1a in Astyanax, and “pigmentation” (GO:0043473) with the gene Tyrp1b in the zebrafish syntenic block. We note that the Tyrp1a paralog resides on a different chromosome (Chr7) in Danio rerio, whereas Tyrp1b resides on chromosome 1 in zebrafish, (described above). Additionally, we discovered 1,136 GO terms for 409 genes associated with 12 scaffolds for “MelUnderStripe” on LG 26 (marker: TP33309), and 3,017 terms for the 654 genes within the critical block on chromosome 11 in Danio rerio. This analysis revealed three pigmentation GO terms in Astyanax including “melanosome transport” (GO:0032402) linked to Ippk and “developmental pigmentation” (GO:0048066) and “eye pigmentation” (GO:0048069) assigned to Pmela. In Danio, we identified “developmental pigmentation” (GO:0048066) and “eye pigmentation” (GO:0048069) associated with Pmela. By co-analyzing gene ontology and positional information from both organisms, we nominated candidate genes for each QTL: Tyrp1b (LG20) and Pmela (LG26). These two genes reside in close proximity to the top two QTL markers, ~7 MB and ~0.05 MB, respectively (Fig. 3a, d).

Fig. 3
figure 3

Integrative analyses reveal two candidate genes for melanophore variation. The evaluation of melanophore variation using QTL analyses identified two top candidate genes that may be associated with pigmentation loss in Astyanax mexicanus cavefish. Pigment cell variation under the mid-dorsal stripe region showed a significant association with marker ASTYANAX_28 on LG 20 (LODMR = 6.01, p = 0.015), which anchors to genomic regions in Astyanax and Danio inclusive of Tyrosinase-related protein 1b (Tyrp1b; a). Distances between the top marker and candidate gene are in boldface. A second melanophore trait in the superior dorsal region demonstrates a significant association with marker TP33309 on linkage group 26 (LODMR = 5.08, p = 0.041; d). Regions of synteny were identified between Astyanax and Danio rerio. The syntenic region was mined for any genes with potential roles in the pigmentation pathway using gene ontology (GO) terms. This led to the discovery of the gene Premelanosome protein a (Pmela). These two candidates, and any other pigment-related genes that exist within the syntenic region, were evaluated for prospective coding mutations and expression alterations using RNA-seq technologies (b–c, e–f). This combined approach has led to discovery of candidate genes that may contribute to loss of pigmentation in cave-dwelling fish

Full size image

Next, we interrogated the coding sequences by aligning surface- and cave-tagged RNA sequencing reads to the Astyanax draft genome. Tyrp1b, near a marker where multiple melanophore traits were mapped, harbored three mutations, including a G-to-A substitution in exon 2 at position 630 (Fig. 3b). This is predicted to be a synonymous change (F210F); however, it falls within a tyrosinase copper-binding domain (205–222 bp; Ensembl.v.75) which could affect interactions with tyrosinase—an enzyme catalyzing melanin biosynthesis (Oetting 2000). Synonymous mutations such as these can affect splicing, stability, structure, and protein folding (Hunt et al. 2009). For the gene Pmela, sequence analyses revealed numerous alterations, such as a G-to-A change at position 1738 in exon 7 (Fig. 3e). This non-synonymous mutation impacts the amino acid sequence, causing a change from a hydrophobic alanine residue to a hydrophilic threonine in cavefish (A580T). In addition, Pmela demonstrates several more mutations including two other non-synonymous changes, two silent SNPs, and three potential splice variants.

With respect to expression, Tyrp1b showed reduced expression in cave relative to surface beginning at 24 hpf (Fig. 3c). We observed expression differences of 7-fold down (36 hpf) and 5.5-fold down (72 hpf; Fig. 3c). Expression profiles of cave versus surface forms demonstrated significant differences at 10 hpf (p = 0.0472), 36 hpf (p = 0.0312), and 72 hpf (p = 0.00167). Pmela similarly revealed reduced expression at 24 hpf (Fig. 3f). Pmela demonstrated a 12-fold reduction in expression at 72 hpf. Levels of gene expression were significantly different between morphs at four stages assayed: 10 hpf (p = 0.00607), 24 hpf (p = 0.00117), 36 hpf (p = 0.0311), and 72 hpf (p = 0.000239). Using qualitative and quantitative PCR, we observed weaker bands in Pachón cavefish compared to the surface fish for Tyrp1b and Pmela (gel images not shown), and significantly reduced expression in Pachón cavefish for Tyrp1b (p = 0.005967; Fig. 4a, c) and Pmela (p < 0.000001; Fig. 4b, d).

Fig. 4
figure 4

Quantitative (qPCR) confirms reduced expression of Tyrp1b and Pmela in Pachón cavefish. To further validate reduced expression patterns from our RNA-seq expression profiling (a, b), we performed quantitative PCR on surface and Pachón cavefish at 72 hpf. These analyses confirmed dramatic reduction of expression in cave-dwelling morphotypes for both Tyrp1b (p = 0.005967) and Pmela (p < 0.000001)

Full size image

3.5 Tyrp1b and Pmela Demonstrate Distinct Melanophore-Specific Expression Patterns Between Cave and Surface Morphotypes

Both Pmela and Tyrp1b appear to influence both melanophore position and melanin density in A. mexicanus and D. rerio (Fig. 5). Analysis of the location of Pmela (Fig. 5a, b) and Tyrp1b (Fig. 5c, d) gene expression following in situ hybridization indicates a heterochronic delay in cavefish compared to surface fish. Early in development (<12 hpf), Pmela and Tyrp1b gene expressions are apparent in more cells in surface fish than in cavefish. Later in development, surface fish exhibits higher numbers of cells with positive Pmela expression posteriorly than cavefish.

Fig. 5
figure 5

Expression and functional analysis of Pmela and Tyrp1b candidate genes. In situ hybridization of Pmela (a, b) and Tyrp1b (c, d) of cave (b, d) and surface fish (a, c) embryos. At this stage, surface fish embryos have begun to develop melanin pigmentation across the yolk sac and putative eye regions, which appears to coincide with Tyrp1b and Pmela gene expression. In individuals stained for Pmela, we see a difference in spatial distribution of gene expression between cave and surface embryos. Morpholino knockdown of 5dpf D. rerio embryos (e–g). The control-injected (e) embryos display a “normative” melanophore distribution and density in comparison to Pmela (f) and Tyrp1b (g) knockdowns. f Filled arrowheads show clustering of melanophores in Pmela morphants. (g) Open arrowheads show a higher dispersal of melanophores in Tyrp1b morphants. A total number of melanophores on the yolk sac were counted to determine the ratio of pale (h) or clustered (i) melanophores. Pmela knockdowns exhibit a clustering of melanophores across the yolk sac (p < 0.05) and a lesser ratio of “pale” melanophores (p < 0.05) than control-injected embryos. (i) Tyrp1b knockdowns have a higher distribution of melanophores across the yolk sac (p < 0.05) and a higher number of “pale” melanophores (n.s.) than control. Scale bars at 500 µm, unless otherwise noted

Full size image

3.6 Functional Analyses of Tyrp1b and Pmela Reveal Altered Melanophore Dispersion and Structure in Morphants

We utilized the closely related (120 MYa diverged), and laboratory-established fish species Danio rerio, to carry out morpholino experiments (Fig. 5e–g). Pmela knockdown via MO injection resulted in clustering of melanophores across the yolk sac (Fig. 5f, i), coincident with a lesser number of pale melanophores (Fig. 5h). In contrast to Pmela morphants, Tyrp1b morphants have a higher number of pale cells compared to control individuals (Fig. 5h). This corresponds to the wider dispersal of melanophores (Fig. 5g) and a higher number of “pale” melanophores on the yolk sac of Tyrp1b morphants. Both Pmela and Tyrp1b appear to influence melanophore darkness and position, based upon phenotypes of D. rerio morphants.

4 Discussion

4.1 Tyrp1b and Pmela Contribute to Melanophore Number Variation in Cavefish

Cave-dwelling Astyanax has evolved diverse phenotypes upon colonization of the cave, including a reduction in pigmentation and eyes, and character expansion of taste and touch sensation (Montgomery et al. 2001; Gross 2012a; Kowalko et al. 2013). Here, we utilized an integrative approach to identify two genes—Tyrp1b and Pmela—that play a role in complex melanophore number diversity in Astyanax. In Astyanax, deeply conserved sets of pigmentation-related processes may be governed by Tyrp1b and Pmela. Tyrp1 serves as a stabilizing protein for tyrosinase which functions in melanin production, and when the Tyrp1b protein is absent, tyrosinase rapidly degrades (Müller et al. 1988; Kobayashi and Hearing 2007). In mice, two distinct Tyrp1 alleles in transgenic mice yield degenerative pigmentation: one causing brown coat coloration or nearly white fur relative to the normally black mice (b allele), and another form that could induce albinism (c allele; Kwon et al. 1989). Similarly, variants of Tyrp1 are the cause of “chocolate” and “cinnamon” coat colors in the domestic cat (Lyons et al. 2005; Schmidt-Küntzel et al. 2005). Numerous Tyrp alleles have also been identified in humans as a cause of oculocutaneous albinism type III, which are often caused by small deletions or single base pair changes impacting the amino acid sequence (Forshew et al. 2005; Rooryck et al. 2006; Chiang et al. 2008; Kenny et al. 2012).

While most mutations in Tyrp1 are the result of non-synonymous changes, we discovered a synonymous alteration in the coding sequence (F210F; Fig. 3b). Some studies of Tyrp1b also include synonymous mutations (e.g., six silent SNPs in the domestic cat) for which the precise impact has not yet been explained (Lyons et al. 2005). Our RNA-seq and qPCR studies indicated significantly reduced Tyrp1b expression in Pachón cavefish compared to the surface-dwelling form. The reduced expression in cavefish may contribute to instability of tyrosinase and may be responsible, in part, for the reduced numbers of melanophores observed in Pachón cave morphotypes.

The gene Pmela which is also known as “Silv” or “Pmel17” is a similarly well-described gene associated with pigmentation (Theos et al. 2005). This gene is responsible for recessive dilution of coat color that deteriorates with age in inbred mouse strains (Dunn and Thigpen 1930). Variants of Pmela cause dilution or hypotrichosis of coat color in the domestic yak (Zhang et al. 2014), cattle (Jolly et al. 2011; Schmutz and Dreger 2013), and the “silver” phenotypes in horse (Brunberg et al. 2006) and zebrafish (Schonthaler et al. 2005).

We discovered three non-synonymous changes to the Pmela coding sequence, including G580A in cavefish, which causes an alanine-to-threonine substitution. This is intriguing since Pmela undergoes posttranslational modification and processing imperative for proper functioning (Theos et al. 2013). The change from a hydrophobic alanine (nonpolar side chains) to a hydrophilic threonine (polar side chains) may alter the Pmela protein. Ala-to-Thr residue changes in diverse proteins have been shown to induce self-aggregation into amyloids (i.e., an insoluble β-pleated sheet formed by the alteration in secondary structure). Although some amyloids are native to Pmel, additional changes in this structure-sensitive protein may impact normal functioning (Fowler et al. 2005).

Pmela has been described as a melanocyte-specific type 1 transmembrane encoded protein enriched in melanosomes (i.e., the pigment-producing organelles within the melanocytes/melanophores; reviewed in Theos et al. 2005). The gene Pmela plays a critical role in the premelanosome “fibril” ultrastructure, and Pmel-targeted antibodies are evident in fibrous, stage II melanosomes as Pmel is thought to polymerize fibrillar arrays that ultimately form the backbone of eumelanosomes (Spanakis et al. 1992; Solano et al. 2000; Raposo et al. 2001; Berson et al. 2001; Raposo and Marks 2002). Moreover, high levels of Pmela expression in non-agouti (solid dark) mice are necessary to construct the fibrils associated with the shape of eumelanosomes (Theos et al. 2005).

We detected the substantial reduction in Pmela expression across early development in Pachón cavefish relative to surface fish. Reduced Pmela expression could impede fibril formation, leading to the aberrant melanophore morphology observed in our knockdown studies. Other reports similarly describe changes in melanosome shape but modest effects on overall body pigmentation in mice (Hellström et al. 2011). Pmela mutants often have reduced or scattered pigment granules in individual hairs (Dunn and Thigpen 1930). Furthermore, the “merle” phenotype in domestic dogs (e.g., Australian shepherds) is caused by codominance of the merle (M) and non-merle (m) Pmel alleles in heterozygous individuals that collectively yield patches of dark and light-colored fur due to spatially random protein instability (Clark et al. 2006; Schmutz and Berryere 2007). These degenerative pigmentation phenotypes, combined with the structural changes and severely reduced expression in colorless cavefish, suggest that Tyrp1b and Pmela contribute to the complex trait of melanophore variation in Astyanax mexicanus.

4.2 Analysis of Complex Pigmentation Informs the Genetic Basis for Regressive Evolution in Cavefish

Three hypotheses seeking to explain regressive evolution include natural selection, neutral mutation/genetic drift, and pleiotropy (Culver 1982). A prior QTL study yielded 18 loci associated with numerical melanophore variation. In certain cases, the homozygous cave genotype was associated with increased numbers of melanophores (Protas et al. 2007). Our study revealed a similar result—genotypic effect plots varied in their polarity for members our experimental F2 pedigree. This may provide evidence for neutral mutation/genetic drift since it would be unlikely that depigmented cavefish (if under selection) would harbor alleles that increase pigment cell number (Protas et al. 2007). However, one potential “selective” benefit for loss of pigmentation in cave-dwelling morphs may be activation of the catecholamine pathway to encourage foraging behavior in the nutrient-poor subterranean environment (Bilandžija et al. 2013). Although the precise evolutionary pressures governing regression remain unknown, further studies will reveal the genetic mechanisms accompanying trait loss in the wild.

4.3 Astyanax mexicanus Enables Investigation of Degenerative Pigmentation Disorders

Degenerative pigmentation can be found widely around the globe, including clinical diseases impacting humans such as albinism, vitiligo, and skin melanoma (Oetting et al. 1996; Agarwal 1998; Hocker and Tsao 2007). Often the etiologies for these disorders remain unknown. Animals that naturally demonstrate degenerative traits, such as the blind Mexican cavefish, can help understand pigmentation losses since they have recurrently evolved extreme pigmentation changes as a consequence of the extreme cave environment (Jeffery 2005; Gross 2012a). Since the functions of many pigmentation genes are shared broadly across taxa, we can make progress toward understanding the comprehensive role of novel genes in the regulation of animal pigmentation, including albinism (Protas et al. 2005) and brown (Valverde et al. 1995; Flanagan et al. 2000; Rees 2003).

Here, we suggest a role for Tyrp1b and Pmela in melanophore variation in cavefish, and these genes are also critical for the normal production of melanin in humans. For instance, in humans, Tyrp1 is responsible for oculocutaneous albinism type III due to coding sequence alterations. In cavefish, Tyrp1 has been repeatedly mutated, including alleles associated with different geographic regions (Manga et al. 1997; Forshew et al. 2005; Rooryck et al. 2006). However, more common Tyrp1 alleles contribute to normal hair, skin, and iris variation in humans (Frudakis et al. 2003; Sulem et al. 2008; Han et al. 2008; Liu and Fisher 2010; Eriksson et al. 2010). Our knowledge of Pmela variants in humans is more limited, and some of the disorders present in other animals (e.g., double merle dogs) also cause hearing and ocular deficiencies due to pigmentation losses in the ears and eyes (Clark et al. 2006). Evaluation of pigmentation genes in cavefish will continue to identify vulnerable genes, shared broadly across taxa that may improve our knowledge of human, pigmentation-related diseases.