Divergence by depth in an oceanic fish

Introduction

Speciation phenomena in taxa diverging with gene flow, in the absence of obvious geographic barriers, remain a central focus in evolutionary biology (Bird et al., 2012; Smadja & Butlin, 2011; Fitzpatrick, Fordyce & Gavrilets, 2008). The classical model of allopatric speciation involves the evolution of reproductive isolation as a result of physical barriers that block gene flow in spatially separated populations; whereas populations diverging in sympatry lead to the formation of species from a single panmictic population which must exhibit strong divergent selection in order to overcome the homogenizing effects of gene flow (Gavrilets, 2003). Parapatric speciation represents an intermediate scenario of species formation whereby partial yet restricted contact zones exist between two populations with limited gene exchange (Gavrilets, 2003). Although exhaustive frameworks exist to identify and interpret speciation dynamics (Rettelbach et al., 2013; Bird et al., 2012; Smadja & Butlin, 2011), only recently has empirical attention been directed to the role of depth gradients in aquatic (Vonlanthen et al., 2009) and especially oceanic biota (Roy, Hurlbut & Ruzzante, 2012; Bird et al., 2011; Brokovich et al., 2010; Hyde et al., 2008).

The strong physical gradients across depth layers in the ocean pose strong selective pressures in aquatic organisms (Ingram, 2011; Somero, 1992), and one notable factor is the change in the light environment (Warrant & Locket, 2004), which affects vision. Visual sensitivity in marine vertebrates depends on the spectral tuning mechanism of the visual pigment (VP) (Yokoyama, 2000), which consists of an opsin protein (part of the largest family of G-protein-coupled receptors) bound to a light-sensitive chromophore. Differently-charged amino acid (AA) residues in the opsin will result in slightly different light absorbance by the photoreceptor cells (Yokoyama, 2002).

The percomorph marine family of rockfishes (Sebastidae) have played a central role in the understanding of depth-associated population divergence and speciation in the ocean (Ingram, 2011; Stefánsson et al., 2009a; Hyde et al., 2008; Alesandrini & Bernardi, 1999), and evidence exists that the rhodopsin gene may have evolved in response to different light environments in the main, ancient radiation of the genus, in the Pacific (Sivasundar & Palumbi, 2010). North-Atlantic Sebastes have a much more recent history (Hyde & Vetter, 2007), with the four recognised extant species having diversified during the Pleistocene (Bunke, Hanel & Trautner, 2012). In particular, the beaked redfish, Sebastes mentella, consists of two genetically distinguishable groups (Stefánsson et al., 2009a; Pampoulie & Daníelsdóttir, 2008): a widely-distributed shallow-pelagic (SP) form, found between 250 and 550 m depth and a more circumscribed deep-pelagic (DP) component, between 550 and 800 m. However, doubts remain as to the forces at play and the time scales associated with this divergence (Stefánsson et al., 2009b; Cadrin et al., 2010).

Here, we sought to investigate whether on-going processes of adaptation to different depth layers may leave a signature of disruptive selection in the rhodopsin gene in a recently diversifying Sebastes species. We also employed for the first time the mitochondrial DNA control region to reconstruct historical demography and to further elucidate the evolutionary relationships between ‘shallow’ and ‘deep’ pelagic beaked redfish.

Material and Methods

Data Analysis

Results and Discussion

We recovered 16 mtDNA haplotypes, defined by 15 total variable sites across the two groups. Haplotypes were almost completely segregated between depth layers, with eight haplotypes found in the deep, never recovered in the shallow area, and resembling a starburst pattern, three mutational steps away from the rest of the network (Fig. 1A). Of the 10 haplotypes found in the shallow, eight were exclusive of this habitat, and only two individuals collected in the deep were found to bear a ‘shallow-type’ sequence. The shallow-pelagic (SP) group exhibited much greater diversity ($\hat{h}$ = 0.887 ± 0.033, π = 0.00504 ± 0.00082) than the deep-pelagic (DP) group ($\hat{h}$ = 0.543 ± 0.119, π = 0.00238 ± 0.00071) (Table 1).

Partitioning of genetic variance between the populations showed highly significant and strong population structure (Table 2). The mismatch distributions of the two groups (Fig. 1B) were significantly different (K–S test: D₃₀₀ = 0.3, p ≪ 0.001), and confirmed what is visually apparent from the haplotype network: a scenario of more recent and pronounced demographic and spatial expansion in the deep-pelagic group compared to the shallow-pelagic (Table 3). Using Nei’s (Nei, 1987) formula for divergence time: T = D_a/2 µ, where 2µ represents a general mtDNA evolutionary rate, commonly assumed to be around 11% per million years for fish mtDNA control region (Patarnello, Volckaert & Castilho, 2007), we find that the “deep” and “shallow” lineages split over 44,000 years ago.

The rhodopsin gene was tested for signatures of positive selection using the Creevey-McInerney method rooting the tree with S. alutus as an outgroup (Fig. 2). Values for the four substitution variables, G-test, and p-values along each branch are presented in Table 4. Two branches (numbers 26 and 27) showed significance at the p = 0.05 level (Fig. 2). Branches 26 and 27 show significant RI to RV deviations from neutrality, due to non-synonymous substitutions (Table 3), suggesting that positive disruptive selection is acting on the rhodopsin gene for the clade and on the internal branch leading to the shallow group. We observed a fixed non-synonymous AA substitution within the transmembrane domain, which strongly discriminates the two groups inhabiting shallow and deep environments (Fig. 3). The shallow-pelagic group exhibits a GTC at position 119, which codes for Valine (L119V), while the deep-pelagic type displays an ATC, coding for Isoleucine (L119I). Amino acid changes located in the transmembrane helical regions have been known to be important for spectral tuning (Yokoyama et al., 2008), yet in-vitro experimental spectral analyses of vertebrate rhodopsins suggest that amino acid substitutions at site 119 have negligible effect on absorption spectra (Yokoyama et al., 2008). Nevertheless, substitutions within the transmembrane protein domain III (helix-III), such as site 119, have been shown to affect the decay rate of metarhodopsin II (“meta II”; Ou et al., 2011), which is an intermediate of rhodopsin that binds and activates transducin, the visual G-protein (Smith, 2010). Ou et al. (2011) have shown that an AA replacement L119C against the wild-type rhodopsin resulted in shorter meta II lifetimes, suggesting more responsive structural alterations at the helix’ G-protein binding site.

Neither spectral nor conformational analyses have so far been conducted on shallow-pelagic and deep-pelagic Sebastes mentella, but the AA variation observed here could underlie differential hydrophobic activity and photoisomerization sensitivity (Ou et al., 2011) that could hold some adaptive value, even without net change in wavelength of maximal absorption.

Sivasundar & Palumbi (2010) discovered a striking association between AA replacements along the rhodopsin gene and inferred depth preference in many North Pacific Sebastes. Interestingly, four North-Pacific Sebastes (S. chlorostictus, S. elongatus, S. aurora and S. melanostomus) typically associated with deeper waters were observed to exhibit the same AA replacement L119I as detected in the deep-sea S. mentella. Similarly, one Pacific species (S. diploproa) exhibits the AA replacement L119V, which is linked to a shift back to shallower waters, and mirrors the polymorphism in the shallow-pelagic S. mentella.

Although larger sample sizes will be required in the future to test these patterns more robustly, the implications of these findings are twofold, and have powerful resonance for both marine evolution and fisheries management. First, mitochondrial variation between ‘shallow’ and ‘deep’ S. mentella in the North Atlantic unveil a degree of historical divergence that previously employed genetic markers either failed to detect (Bunke, Hanel & Trautner, 2012) or could not reliably frame in a phylogeographic context (Stefánsson et al., 2009b).

The level of differentiation and haplotype sorting is such that evolutionary independence can be broadly upheld for these two habitat-segregated lineages, and re-analysis of microsatellite data confirm this picture (Fig. 4). In particular, the comparison of frequency-based indicators of substructure (F_ST for mtDNA, and $G_{ST}^{\prime }$ and D_est to account for the hypervariability of microsatellite loci) reveal values (Table 2) that match theoretical expectations under neutral divergence, taking into account the four-fold strength of genetic drift at mitochondrial markers. Interestingly, we also noticed two individuals with a ‘shallow-type’ mtDNA haplotype, which were caught in the deep layer. One of these two (DP29, Fig. 2) also screened at the rhodopsin locus, exhibits a sequence typical of the shallow layer, and its multilocus microsatellite genotype also falls with the shallow group (Fig. 4), which can be interpreted as the occurrence of individual movements along the water column during the life cycle (i.e., short-term “dives” into the deep, by shallow-dwelling fish). Another deep-caught individual (DP1) also exhibits a “shallow” haplotype, but a “deep-like” ATC rhodopsin sequence and an inconclusive multilocus microsatellite genotype (Fig. 4). Collectively, this likely reflects the occurrence of introgressive hybridisation between the two groups, as previously suggested by Pampoulie & Daníelsdóttir (2008).

Furthermore, the stark pattern of depth-associated divergence at the rhodopsin gene is perhaps even more surprising, were it not for the fact that comparable evolutionary genetic patterns have recently been credited with a key role in the diversification of the more ancient Pacific Sebastes group (Sivasundar & Palumbi, 2010). It has been hypothesized that fast-evolving markers will allow to determine recent speciation events for closely related Sebastes spp. (Alesandrini & Bernardi, 1999; Cadrin et al., 2010). The present data provide a snapshot of the evolutionary mechanisms that may be at play in the young, species-poor, Atlantic Sebastes lineage, during its initial phase of adaptive radiation, underpinned by positive selection at the rhodopsin gene.

Less than a decade ago, S. mentella was assumed to be panmictic in the North Atlantic, and the rapidly increasing fishery pressure on these stocks did not recognise any possible substructure until 2009 (Cadrin et al., 2010). These latest results dismiss the notion of panmixia in this oceanic species, and, perhaps more intriguingly, open the possibility that the two ‘shallow’ and ‘deep’ groups may represent two lineages experiencing adaptation towards divergent environmental conditions. In the near future, it should be experimentally evaluated whether the amino acid replacements at the 119 position actually produce detectable changes in retinal absorbance or structural responsiveness, and whether more powerful molecular comparisons covering a wider portion of the genome (e.g. SNP-based genome scans; transcriptomic approaches) will offer further insights into the role of depth as a diversifying agent in the ocean.