Pronounced heritable variation and limited phenotypic plasticity in visual pigments and opsin expression of threespine stickleback photoreceptors

Cone photoreceptors are specialized cells of the retina whose visual pigments capture light to begin the process of photopic vision (Fein and Szuts, 1982). Each visual pigment is composed of a protein (opsin) and a chromophore (a vitamin A derivative). There are four major classes of cone opsins (SWS1, SWS2, RH2 and MWS/LWS) and one rod opsin (RH1) (Yokoyama, 2000). When combined with the chromophore, these opsins form visual pigments with absorbance maxima (λ_max) in the ultraviolet (SWS1), short (SWS2), middle (RH2, MWS, RH1) or long (LWS) wavelength regions of the spectrum (Yokoyama, 2000; Isayama and Makino, 2012). Besides opsin type, visual pigment absorbance can be modulated by the choice of chromophore: coupling of an opsin to the vitamin A1 chromophore, 11-cis retinal, leads to a lower λ_max than combination with the vitamin A2 chromophore, 3-dehydro, 11-cis retinal (Isayama and Makino, 2012). In most vertebrates, the different visual pigments are distributed in morphologically distinct cone types, and comparison of neural outputs from these cells forms the basis for the perception of colour (Fein and Szuts, 1982). Thus, the molecular constituents of visual pigments are primary substrates for the genetic and environmental modulation of colour vision.

Many non-mammalian vertebrates modulate chromophore usage during life, and this often results from migrations to new photic habitats, as with fishes (Beatty, 1984; Novales Flamarique, 2005), or following metamorphic transitions, as with amphibians (Isayama and Makino, 2012). Modulation of opsin expression has also been documented during development (Wood and Partridge, 1993; Szél et al., 1994; Xiao and Hendrickson, 2000; Cheng and Novales Flamarique, 2004; Takechi and Kawamura, 2005; Spady et al., 2006; Hoke et al., 2006; Cheng et al., 2007; Applebury et al., 2007; Shand et al., 2008; Cottrill et al., 2009) and, to a lesser extent, following changes in habitat (Fuller et al., 2005; Shand et al., 2008; Cottrill et al., 2009; Hofmann et al., 2010; Fuller and Claricoates, 2011). Studies using fishes and rodents have demonstrated that opsin switching within individual cones is a major mechanism that re-structures the chromatic organization of the retina and the colour sensitivity of these animals during development and at early life stages (Szél et al., 1994; Cheng et al., 2006; Cheng et al., 2007; Cheng and Novales Flamarique, 2007; Novales Flamarique, 2013). By contrast, the mechanism(s) underlying, and functional significance of, environmentally induced changes in opsin expression are unknown. Despite the lack of this crucial information, such phenotypic plasticity is commonly assumed to be ecologically relevant (Fuller and Claricoates, 2011).

Studies that report phenotypic plasticity in opsin expression have relied on quantification of opsin transcripts from whole retinas using real-time PCR (Fuller et al., 2005; Shand et al., 2008; Cottrill et al., 2009; Hofmann et al., 2010; Fuller and Claricoates, 2011), a procedure that does not reveal the cellular patterns underlying opsin changes. Microspectrophotometry, the other technique employed in these studies, can only be used to support correlative trends between cone types and opsin expression as sample size is extremely small, in comparison with photoreceptor numbers in the retina (Fein and Szuts, 1982), and most studied animals have topographical variation in both opsin expression and photoreceptor types across the retina (see Hárosi and Novales Flamarique, 2012). In the above studies (Fuller et al., 2005; Shand et al., 2008; Hofmann et al., 2010; Fuller and Claricoates, 2011), differences in opsin expression are assumed to reflect changes in the abundance of the various cone types or differences in the size and/or visual pigment density of cone outer segments (Fuller and Claricoates, 2011), the ciliary structures whose lamellae house the visual pigments (Fein and Szuts, 1982). These studies further assume that the observed changes in opsin expression reflect fundamental alterations to the colour vision of the animal (Fuller and Claricoates, 2011).

The physiological literature, however, casts doubts on many of these assumptions. First, photoreceptor visual pigment density is likely to be at an optimum for proper lamellar packing and functionality of the phototransduction cascade, with overexpression of opsin resulting in expansion of lamellar volume and disrupted kinetics of the light response (Wen et al., 2009). It is therefore unlikely that visual pigment density would increase as a result of experiencing new photic environments in nature. Second, changes in outer segment morphology have only been reported following long-term exposure to artificial light environments (e.g. monochromatic, very bright light) (Kröger et al., 1999; Williams et al., 1999) that have no equivalent in nature. In contrast, migrations to new environments do not induce changes in cone outer segment morphology (e.g. size, taper) in salmonid fishes, and we are not aware of reports to the contrary for other animals featured in the literature (Hárosi and Novales Flamarique, 2012). It thus seems that the reported phenotypic plasticity in opsin expression of fishes is probably based on changes in the relative numbers of spectral cone types. Such alterations in cell numbers can arise via the production and/or pruning of photoreceptors, as occurs in several species (Cornish et al., 2004; Cheng et al., 2006; Hoke et al., 2006; Chen et al., 2008; Hu et al., 2011), or as a switch in the opsin expressed within individual photoreceptors, as has been demonstrated in the rods of the European eel (Wood and Partridge, 1993) and in the cones of salmonid fishes and rodents (Szél et al., 1994; Cheng et al., 2006, Cheng and Novales Flamarique, 2007; Glaschke et al., 2011). The importance of opsin phenotypic plasticity in colour sensitivity, and visual function in general, would then depend on the number and types of cones affected and their retinal distributions. It is the topography of these cells, and their relative outputs, that are primary determinants of visual acuity and the colour contrast detection capabilities of the retina (Fein and Szuts, 1982).

We have chosen the threespine stickleback, Gasterosteus acuelatus, to study genetic (heritable variability) and environmental (phenotypic plasticity) changes in opsin expression and their ecological significance. Literature on stickleback visually guided behaviour dates back to the 1830s (Rowland, 1994) and this fish has long been a model of ecology and evolutionary research to explore the role of colour in sexual selection and speciation (Rowland, 1994; Boughman, 2001; Smith et al., 2004; Rowe et al., 2004; Albert et al., 2007; Boulcott and Braithwaite, 2007; Rick and Bakker, 2008). In addition to the well-established role of colour in stickleback behaviour, this species is found in waters that expand the range of photic regimes, from dystrophic lakes dominated by long (red) wavelength light (Reimchen, 1989; McDonald and Hawryshyn, 1995; McDonald et al., 1995) to mesotrophic lakes and coastal waters characterized by a full visual spectrum (ultraviolet to red) in surface waters and peak transmission in the middle (green) wavelengths (Novales Flamarique et al., 1992; Novales Flamarique and Hawryshyn, 1993; Novales Flamarique and Hawryshyn, 1997; McDonald and Hawryshyn, 1995; McDonald et al., 1995). Thus, comparison of endemic and displaced sticklebacks from various photic regimes can be used to assess heritable variation and phenotypic plasticity in opsin expression, both components assumed important to the visual ecology of this animal and to the evolutionary divergence between populations (Boughman, 2001; Smith et al., 2004).

In this study, we examined the topographical distribution of opsins, visual pigments and photoreceptor types in adult retinas of seven populations of sticklebacks living in photic habitats ranging from red-stained to clear waters. We then compared the retinas of individuals from dystrophic populations after 14 years and within 6 months of transfer to clear waters to evaluate phenotypic plasticity in opsin expression and its photoreceptor origin.

Wild adult threespine stickleback (G. acuelatus L.) were obtained from seven locations in British Columbia (Canada): Doogan’s Lake, Swan Lake, Mayer Lake, Mayer Pond, Drizzle Lake, Drizzle Pond and the Broughton Archipelago (coastal ocean) (Fig. 1). All specimens were in reproductive condition except for those originating from the Broughton Archipelago. Mayer Lake and Drizzle Lake are dystrophic lakes, while Mayer Pond and Drizzle Pond are small mesotrophic ponds. No sticklebacks existed in these ponds until 1993 and 1997, when two of us (C.B. and T.E.R.) transferred sticklebacks from Mayer Lake and Drizzle Lake to Mayer Pond and Drizzle Pond, respectively (Leaver and Reimchen, 2012). Swan Lake is a high productivity, eutrophic, system with large mats of green algae present at various times of the year, but especially during the spring and summer. Doogan’s Lake and the oceanic location are mesotrophic systems, characterized by low to medium productivity and clear waters. The dystrophic lakes in this study have no transmission below 500 nm at 1 m depth and no light at 4 m depth (McDonald et al., 1995). The mesotrophic lakes and ocean are characterized by a full light spectrum in surface waters (320–750 nm) and peak transmission in the range 560–565 nm with depth [see comparable systems in previous publications (Novales Flamarique et al., 1992; Novales Flamarique and Hawryshyn, 1993; McDonald and Hawryshyn, 1995; McDonald et al., 1995)]. We measured downwelling and sidewelling light at all locations at 40 cm depth (Fig. 1) using a USB-2000 spectroradiometer equipped with a 600 μm diameter input, 0.22 NA, liquid light guide and a cosine collector (Ocean Optics, Dunedin, FL, USA). The sidewelling light was the average of two scans in the sun and anti-sun directions. Scans were acquired between 11:00 h and 12:30 h in June, with the sun’s elevation at ∼60 deg. The 40 cm depth is close to the middle of the range (0.1–1 m) at which stickleback nests are found during the breeding season in these water bodies.

Sticklebacks caught in the various locations were either killed on-site, under light adaptation, by quick spinal bisection and decerebration and the retinas fixed for histochemical analyses, or transported live to the aquatic facility of Simon Fraser University (SFU) where they were maintained for up to 6 months. Fish transported to SFU were housed in large aquaria containing water from their locations of origin, or placed in clear water of the same temperature and pH as the source water. Illumination was a 12 h:12 h light:dark cycle provided by tungsten–halogen fluorescent tubes. Fish were killed upon arrival to SFU for visual pigment measurements and at 3 weeks and 6 months after arrival for histochemical analyses. Supplementary material Table S1 shows the numbers of fish used for the different types of experiments. All experimentation was approved by the Animal Care Committee of SFU, in compliance with guidelines set by the Canadian Council for Animal Care.

Opsin partial cDNAs were generated by RT-PCR amplification from total RNA isolated from homogenized retina of adult sticklebacks (RNAqueous-Midi; Ambion, Austin, TX, USA). The cDNAs were synthesized (Ready-To-Go RT-PCR beads, Amersham Biosciences, GE Healthcare, Saskatoon, SK, Canada) using the following primers: SWS1 opsin partial cDNA forward 5′-ATG GGG AAA CAC TTC CAC TT-3′, reverse 5′-TGG AAC AAC AGA TTC CCA TA-3′; SWS2 opsin partial cDNA forward 5′-TCA GCT GAC AAT CAC ATA AA-3′, reverse 5′-AGC CAT CTT TCA AAA GCT AT-3′; RH2 opsin partial cDNA forward 5′-TGG ACA CTT CAC CAC TAT CT-3′, reverse 5′-TGA GTA CCA GAG AAC TTG AA-3′; and LWS opsin partial cDNA forward 5′-TTG CAA GTG ACG AGA AAG GT-3′, reverse 5′-TCC AGG AGA ACA CTA TTC CA-3′. Reverse transcription was carried out at 42°C for 15 min. Cycling parameters for the subsequent PCR were: 95°C × 5 min, 32 cycles of 95°C × 30 s, 56°C × 30 s, and 72°C × 1 min, and 1 cycle of 72°C × 10 min. Each cDNA was gel purified and cloned into a pCRII-TOPO cloning vector (Invitrogen, Carlsbad, CA, USA) and sequenced by AmpliTaq Dye terminator cycle sequencing (UBC Sequencing Laboratory). The identities of the sequences were confirmed by comparing them to nucleotide sequence databases using the BLASTn program (Altschul et al., 1997). To make each cRNA probe, a PCR fragment containing the insert and an RNA promoter amplified from the pCRII-TOPO vector were used to generate digoxigenin- and fluorescein-labelled RNA sense and antisense riboprobes by in vitro transcription (RNA labelling kit, Roche Diagnostics, Laval, QC, Canada).

To compare RH2 and LWS opsin sequences between stickleback populations, full-length opsin cDNAs were generated as described above using the following primers: RH2 opsin partial cDNA forward 5′-TGG ACA CTT CAC CAC TAT CT-3′, reverse 5′-TTA AGA CAC AGA GGA CAC TT-3′; LWS opsin partial cDNA forward 5′-TTG CAA GTG ACG AGA AAG GT-3′, reverse 5′-TTA TGC AGG AGC CAC AGA-3′. The deduced amino acid sequences were compared using ClustalW2, a multiple sequence alignment program (Larkin et al., 2007).

Fish were killed in the light-adapted state. The left eyeball was removed, the iris and lens discarded, and the remaining eyecup, which had been marked on the ventral and nasal retina, immersed in cryo-fixative (4% paraformaldehyde in 0.08 mol l^–1 phosphate buffer, pH 7.4). After a minimum 24 h fixation at 4°C, the retina was extracted and flattened on a grid while maintaining its natural orientation. This procedure ensured the analysis of similar pieces of retina between individuals. Each retina was then cut into eight pieces, and each piece embedded in OCT frozen blocks for in situ hybridization as per previous studies (Cheng et al., 2006; Cheng and Novales Flamarique, 2007). Sections (8 μm thick) obtained from the blocks were collected serially and deposited on alternating slides for simultaneous double-labelling experiments with riboprobes against the SWS1 and SWS2, and the RH2 and LWS opsin transcripts, respectively. Details of the in situ hybridization protocol have been published previously (Cheng et al., 2006). Digital images of sections were acquired with an E-600 Nikon fluorescence microscope equipped with DIC optics and a DXM-100 digital camera. The images were used to characterize cone types and to compute the percentage of cones that expressed each opsin mRNA. The densities of double and single cones were counted over a 17,926 μm² area using a grid system on the computer monitor, and the ratio of double to single cones was computed for each sector of the retina.

The eyecup from the right eyeball of the same fish used for in situ hybridization was immersed in primary fixative (2.5% glutaraldehyde, 1% paraformaldehyde in 0.08 mol l^–1 phosphate buffer, pH 7.4). After 24 h fixation at 4°C, the retina was extracted from the eyecup, rinsed in 0.08 mol l^–1 phosphate buffer, and pieces cut using the same procedure as for in situ hybridization. These pieces were processed for embedding in EPON resin blocks, which were cut radially in 75 nm steps, as detailed previously (Novales Flamarique, 2011), to reveal the overall morphology and the ultrastructure of cones. Electron micrographs were used to obtain average dimensions of cone outer segments and lamellae, and lamellar spacing.

Individual fish were dark adapted overnight. Following this adaptation period, the fish was killed, one eye was enucleated, and the retina was removed under infrared illumination. Small pieces of retina were teased apart and prepared for viewing with the dichroic microspectrophotometer (DMSP) as per previous studies (Novales Flamarique and Hárosi, 2000; Cheng et al., 2006; Cheng et al., 2007). The DMSP is a computer-controlled, wavelength-scanning, single-beam photometer that simultaneously records average and polarized transmitted light fluxes through microscopic samples (Novales Flamarique and Hárosi, 2000). The DMSP was equipped with ultrafluar (Zeiss) objectives: 32/0.4 for the condenser and 100/1.20 for the objective. With the aid of reference measurements recorded through cell-free areas, individual photoreceptor outer segments were illuminated sideways with a measuring beam of rectangular cross-section ∼2×0.6 μm. Absolute absorbance and linear dichroism spectra were computed in 2 nm increments from the obtained transmittance (each spectrum consisted of an average of eight scans). The solid spectra (fits) were derived from experimental data by Fourier filtering (Hárosi, 1987).

Spectral irradiance measurements from the various water bodies revealed pronounced differences in light transmission (Fig. 1). The mesotrophic systems had broad spectrum downwelling (300–750 nm) and sidewelling (350–750 nm) light, with the downwelling light being ∼4–10 times as intense as the sidewelling light (Fig. 1E,F). Mayer Pond, also a mesotrophic system, had similar broad spectra but the shape of the curve was reduced in the middle to long wavelengths (Fig. 1B). Mayer Lake had a downwelling spectrum shifted toward longer wavelengths, with little downwelling to no sidewelling light below 400 nm, and a ratio of downwelling to sidewelling light of ∼33 (Fig. 1A). This light environment is characteristic of dystrophic lakes and is due to the high quantities of dissolved organic matter present in the water. Swan Lake had the most distorted spectra (Fig. 1G), and the ratio of downwelling to sidewelling light was ∼4, similar to that of the ocean location (Fig. 1E). Despite the large quantities of algae present in Swan Lake, a broad spectrum of downwelling light (320–750 nm) penetrated surface waters (Fig. 1G).

Microspectrophotometry revealed four cone visual pigment types located within double and single cones, and one rod visual pigment present in the rods (Fig. 2). Single cones had either an ultraviolet (UV) or a short wavelength (S) visual pigment, with mean λ_max in the range 365–379 nm and 434–460 nm, respectively (Table 1). Double cones had either a long wavelength (L) visual pigment in both double cone members (mean λ_max range: 566–618 nm), as was predominantly the case in fish from dystrophic lakes or those that had been transferred into clear waters, or an L visual pigment in one member and a middle wavelength (M) visual pigment in the other (mean λ_max range: 514–547 nm), as was always the case in fish from the remaining water bodies (Table 1). Rods contained a rhodopsin visual pigment with mean λ_max in the range 507–531 nm. The mean λ_max values measured for the Broughton Archipelago (ocean) population are consistent with λ_max values of vitamin A1-conjugated visual pigments in other fish species (see Hárosi, 1994). None of the outer segments examined showed multiple α-band absorbance peaks, which would indicate multiple visual pigments within a photoreceptor (see Cheng et al., 2006; Cheng et al., 2009).

We used semi-empirical models (Hárosi, 1994) to predict the λ_max shifts of visual pigments when the opsin remains the same but the chromophore is switched from vitamin A1 to vitamin A2. These calculations revealed the following visual pigment A1–A2 pairs, computed based on the λ_max of vitamin A1-based visual pigments measured: 365–382 nm (UV), 434–441 nm (S), 514–546 nm (M), 566–638 nm (L) and 507–535 (rod). With the exception of the S pigment, the absorbance range of all other visual pigments could be accounted for by variable use of vitamin A1 and A2 chromophores within individual photoreceptors. The lower and upper extremes in the visual pigment absorbance ranges originated from cones of non-reproductive and reproductive fish, respectively. These denote retinas based primarily on vitamin A1 or A2 chromophore, as occurs in other non-mammalian vertebrates at different life stages or seasonally (Beatty, 1984; Novales Flamarique, 2005; Hárosi, 1994; Isayama and Makino, 2012).

Although the proportions of the two chromophores varied between populations, and between photoreceptors within populations, our results indicate vitamin A2-dominated retinas in dystrophic systems and vitamin A1-dominated retinas in the clear waters of the Broughton Archipelago (ocean) (Table 1). Such variation in chromophore use is in line with that found in a related species, the nine-spined stickleback (Pungitus pungitus), where the absorbance of all visual pigments (except S) could be accounted for by chromophore shifts (Saarinen et al., 2012). It therefore appears that, in both stickleback species, chromophore use varies between photoreceptor types and multiple SWS2 opsins may be expressed within the single cone population.

Sequencing of the LWS and RH2 opsins, corresponding to the L and M visual pigments (Fig. 2), which showed the greatest ranges in λ_max (Table 1), revealed only three amino acid differences in the LWS opsins and nine amino acid differences in the RH2 opsins between populations from dystrophic and mesotrophic systems (Figs 3, 4). The differences among the LWS opsins involved three amino acid substitutions (Fig. 3), two located in the N-terminus and one in the fifth transmembrane helix. None of these changes were within proximity of the chromophore and, as such, should be functionally inconsequential (Smith, 2010). In addition, none of these alterations occurred at any of the five key sites that produce major shifts in visual pigment absorbance (Yokoyama et al., 2008). Comparison of RH2 opsin sequences also showed conservation of amino acids at the five key sites between stickleback populations (Fig. 4). Except for position 114, all the differences between sequences involved amino acids belonging to the same, weakly charged, group (Smith, 2010). At position 114, Swan Lake fish had a glutamate (E), a negatively charged amino acid, whereas Mayer Lake and Broughton Archipelago fish had a glycine, a non-polar residue (Fig. 4). Position 114, however, is located in extracellular loop 1 of the stickleback RH2 opsin, which is not in the vicinity of the retinal chromophore and should therefore not influence the absorbance of the visual pigment (Smith, 2010; Yokoyama et al., 2008). Overall, these results corroborated the conclusion reached from visual pigment data that all stickleback populations studied had the same opsins.

In accordance with the correspondence between visual pigments and morphological cone types, single cones contained a SWS1 or SWS2 opsin (Fig. 5) and double cones contained either LWS opsin or RH2 and LWS opsins, one per double cone member (Figs 6, 7). At the ellipsoid level, single and double cones formed square mosaics in the ventral retina (Fig. 5G) and mostly row mosaics in the rest of the retina (Fig. 5D). As is the case for the majority of fish species that have been studied (see Hárosi and Novales Flamarique, 2012), the square mosaic unit of the threespine stickleback consisted of four double cones forming a square, or cruciform, arrangement and a variable number of single cones, from 1 to 5 (Fig. 5A–C,G–I). Single cones were predominantly central in the mosaic unit, at the hypothetical intersection of neighbouring double cone partitions (Novales Flamarique, 2001; Cheng et al., 2006), but a few corner cones could also be found in the central and centro-dorsal areas of the retina (Fig. 5A,B). In contrast to reports on the adults of other fish species (Cheng et al., 2006; Cheng and Novales Flamarique, 2007), corner cones expressed SWS2 opsin, whereas the neighbouring centre cones expressed predominantly SWS1 opsin (Fig. 5A–F). In the ventral retina, where corner cones were absent, the remaining (centre) cones expressed primarily SWS2 opsin, though a minority expressed SWS1 opsin (Fig. 5G–I). UV cones were substantially smaller than S cones (Table 2) and showed proportionately smaller areas of mRNA expression in the cone myoids (Fig. 5J–L). All stickleback populations examined had similar distributions of UV and S cones. The highest densities of UV cones occurred in the centro-dorsal and centro-temporal areas, and S cones were always more numerous than UV cones throughout the retina (Fig. 8).

Double cones from dystrophic lake populations were predominantly (>95%) L/L pairs, the rest were M/L pairs (Fig. 6A,D, Fig. 7A–J, Fig. 8A). Double cones from transplanted populations, whether to mesotrophic ponds 15–20 years ago (Fig. 6B,E, Fig. 7B) or to full-spectrum holding tanks and examined at 3 weeks and 6 months post-arrival (Fig. 7C), were also overwhelmingly (>93%) L/L double cones. The ∼2% increase in M cones over the entire retina following transplantation was nonetheless statistically significant (ANOVA, F=24.93, P<0.001), and post hoc comparison tests (Student–Newman–Keuls, Tukey, with α=0.05) revealed that M cone numbers were the same between transplanted populations, regardless of holding time, and different from the source (dystrophic lake) population. The highest increases in M cones occurred away from the central retina, primarily in the ventral and temporal regions of the retina (Fig. 6D,E, Fig. 7D,E, Fig. 8B,C), though numbers between sectors were not statistically different. The M cone population did not exhibit any discernible topographical pattern (Fig. 6A,B,D,E, Fig. 7A–J) but appeared randomly distributed throughout the retina, reminiscent of the smaller populations of UV cones, blue cones and ‘green’ rods in the retinas of amphibians (Chen et al., 2008). In contrast, double cones of fish from the remaining water bodies were M/L pairs, alternating in position within the mosaic unit (Fig. 6C,F, Fig. 7K,L, Fig. 8D). As in salmonid fishes (Beaudet et al., 1997; Cheng and Novales Flamarique, 2007), the highest cone densities occurred in the ventral and nasal periphery, which were the areas characterized by the lowest UV cone presence (Fig. 8).

Cone dimensions of fish from the various populations were statistically the same (Table 2). Ultrastructure investigations revealed that lamellar thickness and inter-lamellar spacing were also invariant between populations, with means of 14.5 nm and 9.1 nm, respectively. In particular, these means ± s.d. for the dystrophic lake populations were 14.4±0.95 nm and 9.3±1.8 nm (N=63 cones, 6 fish), and 14.7±0.89 nm and 9.2±1.6 nm (N=75 cones, 8 fish) for the transplanted populations. These numbers, which are in the range reported for other fishes (Borwein and Hollenberg, 1973; Collin et al., 1996), indicate no changes in outer segment morphology as a result of transplantation.

Genetic variation in opsin expression between populations was restricted to the longer wavelength opsins, RH2 and LWS. Dystrophic lake populations had double cones that were overwhelmingly L/L pairs whereas fish endemic to clear water systems had M/L double cones, each member alternating in position within well-organized mosaics (Figs 6, 7, 8). There was a clear correspondence between the mean λ_max of double cones, weighted by their cone fractions, and the light environments that fish inhabited. Based on our downwelling irradiance measurements and those in the literature (Novales Flamarique et al., 1992; Novales Flamarique and Hawryshyn, 1993; McDonald and Hawryshyn, 1995; McDonald et al., 1995), we computed the depth at which the integrated light power transmission median (λP₅₀), a measure of the prevailing ambient quanta (McFarland and Munz, 1975), best approximated the weighted λ_max of double cones. Surprisingly, this depth (60 cm) was the same for both dystrophic and mesotrophic systems, and was within 5 cm of the mid-range for nest deposition, an activity that marks the time when adult sticklebacks are found close to the water surface. At 60 cm depth, the mean λP₅₀ was 610 and 565 nm for dystrophic and mesotrophic systems, respectively. These numbers are almost a match for the weighted λ_max of double cones from the corresponding fish populations, with a mean of 613 nm for fish originating from dystrophic systems and 562 nm for those from mesotrophic systems. Such close correspondence, based on the vitamin A2-conjugated opsins found in reproductive fish, indicates evolutionary selection for double cone visual pigments with maximum sensitivity to the ambient light spectrum.

There were two cases where the λP₅₀ at 60 cm depth and the weighted λ_max of double cones were distant from each other. These involved fish transplanted to clear water ponds (λP₅₀≈540 nm and λ_max≈603 nm) and those from Swan Lake (λP₅₀≈613 nm and λ_max≈578 nm). The lack of correspondence in these cases reveals the slow pace of evolution and major differences in population ecology between water bodies. Stickleback populations from the dystrophic and mesotrophic systems examined are genetically very stable, having evolved over thousands of years in relative isolation (Deagle et al., 1996). In contrast, the Swan Lake population undergoes frequent genetic drift as this small water body becomes periodically anoxic, resulting in the death of resident fish, and marine stickleback populations can readily invade as a result of the ocean’s proximity. Thus, the Swan Lake population, though living in an environment dominated by long wavelengths, has an opsin complement resembling that found in mesotrophic waters and probably reflects the latest oceanic invasion. Similarly, the relatively recent introduction (15–20 years ago) of fish from dystrophic lakes into mesotrophic ponds has failed to change the inherited pattern of opsin expression from the source populations. As several generations of fish have been produced in the new environments, it appears that maximizing the sensitivity of double cones to the light background of surface waters, though beneficial in the long term, is not crucial for immediate survival.

Intraspecific, heritable, variation in opsin expression has also been reported in killifishes and in African cichlids using real-time PCR and microspectrophotometry. In Lake Victoria, cichlid populations living at greater depths have increased expression of LWS opsin alleles that absorb at longer wavelengths compared with those from surface-dwelling populations (Carleton et al., 2005; Hofmann et al., 2010; Smith et al., 2011). There is also an apparent increase in L/L double cones, as opposed to L/M double cones, in cichlids inhabiting murkier waters (Carleton et al., 2005). In Lake Victoria, the differential expression of LWS alleles with depth is thought to drive speciation according to the premises of the ‘sensory bias’ hypothesis (Terai et al., 2006; Seehausen et al., 2008), though such a scheme does not hold for Lake Malawi populations (Smith et al., 2011). In killifishes, populations from ‘tea-stained’ swamps appear to have fewer UV and violet cones than populations from clear water springs (Fuller et al., 2003), and this result correlates with mRNA expression of these opsins (Fuller et al., 2004). The topographical organization of spectral cone types in the retinas of cichlids and killifishes is unknown, however, making functional interpretation difficult.

By comparison, our study shows the chromatic organization of the retina of sticklebacks from a wide range of photic regimes, revealing a difference in double cone opsin expression that is much greater than that estimated for cichlid and killifish cones based on microspectrophotometry data (Fuller et al., 2003; Carleton et al., 2005). This difference in double cone opsin expression is reflected in the spectral sensitivity of ganglion cells as measured by compound action potential recordings from the optic nerve (McDonald and Hawryshyn, 1995). Such extracellular potentials indicate a single sensitivity peak in the long wavelengths (600–620 nm) in sticklebacks from dystrophic lakes and two peaks, in the middle (540–560 nm) and long (600 nm) wavelengths, in fish from a mesotrophic lake (Quamichan Lake, BC, Canada) with similar spectral characteristics to the one studied here (Doogan’s Lake). There is therefore a dominant contribution of the L cone mechanism to the spectral sensitivity of sticklebacks from dystrophic lakes, a result that is in line with expectations from our findings.

It is intriguing yet enigmatic that the previous electrophysiological study failed to find any spectral sensitivity functions corresponding to either the UV or S cone mechanisms (McDonald and Hawryshyn, 1995). In fact, even the presence of an M cone mechanism in Swan Lake sticklebacks could not be clearly discerned. The electrophysiological technique used relied on extracellular recordings from a subset of optic nerve fibres and this limitation, along with restricted illumination of the ventral retina (McDonald and Hawryshyn, 1995), could easily have resulted in the absence of detected signal from smaller cone populations. Other taxa, like amphibians, also have reduced populations of UV and S cones dispersed throughout the retina (accounting for ∼1–6% of all cones) (Chen et al., 2008), and their physiological presence was only revealed following action potential recordings from the entire retina (Koskelainen et al., 1994; Deutschlander and Phillips, 1995). In primates, S cones (expressing a SWS1 opsin) are present in reduced numbers (<10%) compared with other cone types (Roorda et al., 2001), yet play a role in non-foveal vision. In the case of adult sticklebacks, it is important to note that these fishes often feed in surface waters, where a full spectrum of light is present, providing the necessary wavelengths to stimulate UV cones.

Fish transplanted to clear waters from dystrophic lakes had the same ∼2% increase in M cones regardless of residence time in the mesotrophic ponds or in aquaria exposed to a full light spectrum. Such an increase could have resulted from apoptosis of L cones and their replacement by newly differentiated M cones, or by a switch in opsin expression from LWS to RH2 within individual cones. Although we did not perform experiments to distinguish between these two possibilities, it is worth noting that full differentiation of cones in goldfish takes approximately 3 weeks (Wu et al., 2001). Apoptosis followed by differentiation, two processes that are usually sequential (Candal et al., 2005), would take longer than our first sampling time (3 weeks) to complete. Thus, our data suggest opsin switching as the cellular mechanism behind the double cone phenotypic plasticity observed. In the retinas of salmonid fishes and rodents, cell labelling studies with proliferation and differentiation markers have shown that cone production in the main (non-peripheral) retina is absent after development (Julian et al., 1998; Candal et al., 2005; Cheng et al., 2009; Glaschke et al., 2011). Our histological analyses further indicated that cone density and morphology did not change following transplantations. This result, together with the similarity in absorbance curves obtained for the various visual pigments between populations, strongly suggests that opsin density remained the same following transplantations.

The timing of M cone density increase, occurring within 3 weeks of transfer to clear waters, is in line with reported differences in opsin mRNA expression between adult killifish following 1–3 days of rearing in new light environments (Fuller and Claricoates, 2011). It is also consistent with the timing of the opsin switch in the single cones of salmonid fishes after 3–5 days of exogenous treatment with thyroid hormone (Cheng et al., 2009; Gan and Novales Flamarique, 2010).

In contrast to the opsin switches that occur during development of salmonid fishes and rodents (Cheng et al., 2006; Cheng and Novales Flamarique, 2007; Applebury et al., 2007), the increase in M cone density reported here for sticklebacks does not appear to have any significance for the visual ecology of the animal. This is because the total photon catch of M cones in the retina would be altered by less than 1% (see Novales Flamarique, 2013). Similar changes in cone numbers following metamorphosis of amphibians are considered inconsequential (Chen et al., 2008). Our results are in line with those from cave-dwelling mollies, where differences in LWS opsin mRNA expression between populations living in different light environments show little to no phenotypic plasticity (Tobler et al., 2010). It remains to be determined whether the changes in opsin mRNA expression due to phenotypic plasticity reported for other fishes (Fuller et al., 2005; Shand et al., 2008; Hofmann et al., 2010; Fuller and Claricoates, 2011) are ecologically meaningful.

We thank Lisa Grebinsky for laboratory and field assistance, and Dr Simon Jones (Pacific Biological Station, Department of Fisheries and Oceans, Canada) for sticklebacks from the Broughton Archipelago.