Light environment change induces differential expression of guppy opsins in a multi-generational evolution experiment

Abstract

Light environments critically impact species that rely on vision to survive and reproduce. Animal visual systems must accommodate changes in light that occur from minutes to years, yet the mechanistic basis of their response to spectral (color) changes is largely unknown. Here, we used a laboratory experiment where replicate guppy populations were kept under three different light environments for up to 8–12 generations to explore possible differences in the expression levels of nine guppy opsin genes. Previous evidence for opsin expression-light environment “tuning” has been either correlative or focused exclusively on the relationship between the light environment and opsin expression over one or two generations. In our multigeneration experiment, the relative expression levels of nine different guppy opsin genes responded differently to light environment changes: some did not respond, while others differed due to phenotypic plasticity. Moreover, for the LWS-1 opsin we found that, while we observed a wide range of plastic responses under different light conditions, common plastic responses (where the population replicates all followed the same trajectory) occurred only after multigenerational exposure to different light environments. Taken together this suggests that opsin expression plasticity plays an important role in light environment “tuning” in different light environments on different time scales, and, in turn, has important implications for both visual system function and evolution.

Evolution experiment, gene expression, light environment, opsin, Poecilia reticulata, visual system

Environments change over multiple time scales and organisms must accommodate these changes. Some environmental changes occur within an individual's lifetime, requiring physiological, developmental, and behavioral changes (phenotypic plasticity). Common responses to environmental change over periods longer than one generation can arise from both phenotypic plasticity and selection. Selection requires enough genetic variation to allow allele frequency changes and, over longer periods, the appearance of new alleles. Gene regulation should also respond to environmental changes and it may be beneficial for populations to have both a common plastic response in gene expression (to respond to shorter term environmental changes) as well as gene expression levels that can change over multiple generations (to respond to longer term environmental changes). The visual system is a good candidate to address this idea because visual inputs change at all time scales and vision mediates behavior critical for survival and reproduction.

Vision depends jointly on the interaction between environmental light, the reflection spectra of objects and the absorption spectrum of each photoreceptor in the viewer's eye. These interactions profoundly affect both reception and perception of visual stimuli (Endler 1990, 1993b, 1993a; Lythgoe 1979). Which cone photoreceptor properties are required for optimal visual performance differs among light environments and visual targets and the photoreceptors of many species appear to be “tuned” to work best in their most frequently used light environments (Lythgoe 1979). This yields the prediction that changed visual environments should be associated with phenotypically plastic and/or evolutionary (i.e., genetic) responses in the visual system. Furthermore we can hypothesize that, because light environments change at many different time scales over several orders of magnitude, the visual system should have components that can respond to short-term changes in the light environment that occur within an individual's lifetime as well as the potential to respond to longer term changes in the light environment after multiple generations of exposure (see references in Pigliucci et al. 2006).

The spectral sensitivity of the retina is primarily governed by the transmission spectrum of the eye's ocular media and the light absorption properties of the suite of photoreceptors present. For each photoreceptor, spectral sensitivity is described by the wavelength of peak absorbance (λ_max) and is largely determined by properties of the photoreceptor visual pigments (VPs), as well as any intracellular or ocular filters proximal to the light absorbing outer segment of the photoreceptor. The spectral sensitivity of a photoreceptor can change by altering the apoprotein (opsin), the chromophore (retinal), or both. The mechanisms for altering a retina's spectral sensitivity include, but are not limited to, a change in the chromophore used by the VP (Temple et al. 2006), the coexpression of different opsins in the same photoreceptor in specific proportions (Isayama et al. 2014), opsin amino acid changes that alter the function (tuning) of the VP (Carleton 2009) and temporal and spatial opsin gene expression changes in the retina (Dalton et al. 2014; Fuller and Claricoates 2011; Schremser and Williams 1995; Temple 2011). While the presence or absence of specific opsin genes in a species is a function of evolution, changes in opsin gene expression levels may result from phenotypic plasticity or from selection on both plastic and genetically polymorphic gene expression regulators within populations. Assuming opsin mRNA levels reflect protein abundance (Cheng and Novales Flamarique 2004; Fuller et al. 2004; Hagstrom et al. 1998; Schwanhausser et al. 2011), different opsin expression levels should change the ratio of visual pigments either within or among photoreceptor cells across the retina. In turn, these changes should impact vision (Calvert et al. 2001; Vorobyev et al. 2001). Hence differential opsin gene expression may enable the visual system to enhance efficiency of common visual tasks in different local environments.

Current evidence for evolution of opsin expression-environment tuning is based on the patterns of variation observed in nature (e.g., Bloch 2015; Carleton 2009; Sandkam et al. 2015). While this comparative approach has identified the light environment as a possible mechanism driving opsin expression changes, this empirical research is not conclusive because opsin variation is usually associated with correlated variation in several environmental factors. For example, opsin expression-light environment variation is typically associated with other variables such as, temperature, predator communities, seasonal water color differences, and food types and their availability.

To experimentally address whether opsin expression levels are directly influenced by environmental light, we used a laboratory evolution experiment where only the light environment varied. We exposed guppy (Poecilia reticulata) populations to different light environments over different time scales to investigate their effects on the expression levels of each opsin gene. In guppies, cone photoreceptors express nine opsin genes and what is currently known about these guppy opsin expression patterns is reviewed in Table S1. Furthermore, P. reticulata is one of the premier model organisms for the study of phenotypic variation, owing to rapid evolutionary responses to both predation and sexual selection (Houde 1997).

Initially, we used relative quantitative real-time PCR (qPCR) to measure population opsin expression levels for nine guppy opsins. We measured relative expression values separately for each opsin gene (calibrated normalized relative quantity [CNRQ] values, Hellemans et al. 2009) over proportional measures of opsin expression (Fuller and Claricoates 2011), because it was initially critical to examine whether the populations under different experimental light conditions differentially expressed each opsin gene. We did not aim to make inferences about color vision per se, but rather explicitly test which specific opsins are differentially expressed after exposure to different light environments. By initially examining the different light environment populations after 38 months (or approximately 8–12 overlapping generations) we can capture opsin expression differences caused by multiple nonexclusive mechanisms (i.e., phenotypic plasticity and selection), as would occur in a natural population responding to environmental change.

Subsequently, for guppy opsins that did undergo differentiation in their relative gene expression levels under the different light environments, we examined their gene expression levels over a 38-month multigenerational timeline and in a common garden experiment using digital PCR (dPCR). This allowed us to more accurately explore the relationship between opsin expression levels and time, as well as examine whether population opsin expression levels had undergone any longer term expression changes.

To our knowledge, this is the first time experimental evidence for opsin expression–environmental tuning has been presented in any species over a time frame longer than one to two generations. In turn, we suggest that our experimental results provide a validation of the first tenet of the sensory drive hypothesis (Endler 1992; Endler and Basolo 1998), that is, new environmental conditions during signal reception (e.g., ambient light) can directly or indirectly cause changes in sensory systems.

Materials and Methods

ANIMAL COLLECTION, CARE, AND TREATMENT

We collected the foundation of our experimental guppy population in September 2010 from Alligator Creek in Bowling Green Bay National Park (Queensland Permit WITK07655010), 30 km South of Townsville, Queensland, Australia (19^o26.79′ S 146^o58.65 E’), where they were introduced between 50 and 100 years ago (Brooks and Endler 2001). The light environments encountered by the Alligator Creek guppy populations are variable, but are mainly present in open canopy streams. Hence ancestral populations most probably experienced broad-spectrum light conditions.

The foundation population consisted of about 300 adults and 50 juveniles but the effective population size was probably larger because female guppies store sperm, (Kobayashi and Iwamatsu 2002; López-Sepulcre et al. 2013) leading to a mean of 3.5 sires per brood in nature (Neff et al. 2008). Guppies from the foundation population were distributed into 12 tanks (each 3 m × 1.5 m × 0.5 m deep; approximately 2220 L), as this reflects a typical natural pool size. The distribution was accomplished by first raising the newly caught fish in one tank until the population was over a thousand. Next, equal numbers of males, females, and juveniles were distributed into two tanks. The two were allowed to expand again, and then ¼ of each tank was distributed among four tanks. The next stage divided four tanks into eight, and finally eight into all 12. This ensured that the genetic variation among all tanks was very similar at the start of the evolution experiment. Throughout the experiment, all tanks were maintained in a constant temperature room at 24 ± 1°C and were illuminated from above using four low flicker (200 Hz) daylight fluorescent tubes under a 12:12 hour light-dark cycle. Population density (∼1000 adults/tank) was also kept stable throughout the experiment. Guppies were fed a combination of brine shrimp napulii (Artemia salina; INVE Aquaculture Inc.), algae discs (Wardley), and flakes (Fish Breeders Choice) daily. All care and treatment of these guppies was in compliance with and approved by the Deakin University Animal Welfare Committee (A21-2010 and G01-2012).

AMBIENT LIGHT TREATMENTS

We exposed the twelve guppy populations to three different ambient light treatments (Figs. (1) and S1), each with four replicates. From September 2011, the overhead fluorescent lighting was filtered through a “Moss Green” filter (Roscolux, filter number 89), a “Lilac” filter (Roscolux, filter number 55), or a clear film filter (functioning as a neutral density filter control to equalize total irradiance with the two other treatments without changing the lab light spectrum). This yielded each of the three filter light environment treatments named Green-F89, Lilac-F55, and Clear-CF, respectively (Fig. S1). The light treatments were assigned to tank positions randomly within the laboratory. The replicate mesocosm populations were labeled as T01, T05, T06, and T09 for Green-F89; T04, T07, T08, and T12 for Lilac-F55; and T02, T03, T10, and T11 for Clear-CF (control).

Figure 1

Schematic diagram of the experimental design. The different light environments are F89 (Green-F89 filter), F55 (Lilac-F55 filter) CF (Clear-CF filter), and LL (lab light). LL is used for the developmental (DP) and adult plasticity (AP) experiments. Samples were collected from each population at 4, 19, 26, and 38 months. The 4-month time point is a foundation population sample (where females lived for 4 months in their new light environments). For the 19, 26, and 38 months’ time points an isolation step was used to ensure sampling from the most recent generation. For the 38-month samples, there were three different light treatment exposure regimes, EV, AP, and DP. Prior to the different treatment groups (blue box), all populations had been under their respective light filters for approximately 8–12 overlapping generations (time = 27 months). After an additional 11 months in the EV, AP, or DP light conditions, females were sampled from each EV, AP, or DP population (population total = 36). See Material and Methods for further details.

The irradiance spectra of these three filter light treatments, plus that of the lab lights (LL) over the small aquarium racks used in the plasticity experiments were recorded by an Ocean Optics USB2000+ spectrometer calibrated for photon flux (visually relevant units, μmol photons m⁻¹ s⁻¹ nm⁻¹) with a Li-Cor LI-1800–02 optical radiation calibrator (see Endler 1990 for details). We estimated relative cone stimulation (RCS) under each light environment (Cole and Endler 2015) using the methods in (Endler and Mielke 2005) but substituting the cone λ_max values for guppies (Table S1) (Archer et al. 1987; Archer and Lythgoe 1990; Kawamura et al. 2016; Watson et al. 2011) and guppy lens transmission data was provided by Dr. R. H. Douglas (pers. comm. 2000; see also Thorpe et al. 1993). The irradiance spectra for each light environment are shown in Figure S1 and the RCS estimates for each light environment are shown in both Figures (2) and S1. With respect to guppy cone captures, our experimental light environments were not radically different from any natural light environments experienced by guppies (Endler 1993b; Gamble et al. 2003). In addition, we also calculated absolute cone stimulation (ACS) among the different light environments, using both (i) the measured peak absorbance (λ_max) of their respective photoreceptor MSP values (Fig. S2; Table S1 and references therein) and (ii) Kawamura et al.’s (2016) measured peak absorbance (λ_max) values for each opsin visual pigment reconstituted in vitro (Tables S1 and S2).

Figure 2

Heat map of the relative cone stimulations under four different light conditions. The different light environments are Green-F89, Lilac-F55, Clear-CF (control), Lab light (LL; used for the developmental and adult plasticity experiments, AP and DP). Relative cone stimulations were estimated for each reported guppy cone type using guppy ocular media transmission and the guppy λ_max values shown on the right side of this heat map (reviewed in Kawamura et al. 2016). The opsins predicted to be present in each guppy cone type are listed on the far right side of the heat map (Kawamura et al. 2016). RCS, Relative cone stimulation. d, evidence for more transcripts present in the dorsal retina. v, evidence for more transcripts present in the ventral retina. Ser, a serine positioned at residue 180 in LWS-1. Ala, an alanine positioned at residue 180 in LWS-1. See Material and Methods and Table S1 for further details.

EXPERIMENTAL DESIGN OVERVIEW

To investigate how P. reticulata opsin expression levels change under different light environments in our multigeneration experimental populations, we used a two-step approach (Fig. (1)). Initially, we determined which of the nine cone opsin genes were differentially expressed under the different filtered light environments after 38 months (or approximately 8–12 overlapping generations) using relative quantitative real-time PCR (qPCR). We measured relative opsin expression levels in the 38-month filtered light time point to capture any opsin expression differences that occurred via multiple nonexclusive mechanisms (e.g., phenotypic plasticity and selection), as would occur in natural population responding to light environmental change. This 38-month filtered light experimental treatment group is herein referred to as EV. In addition, we also examined the relative expression levels of nine cone opsin genes in two common garden experiments, namely our adult plasticity (AP) and developmental plasticity (DP) experiments (see Fig. (1) for details). Both common garden treatment groups were established to examine which populations exhibit phenotypic plasticity for opsin expression after being moved to a novel light environment after 38 months in the laboratory evolution experiment. For the DP experiment, we used common garden lab light conditions to investigate the effect of light environment on developing guppies and to determine the possibility of any opsin expression differences among our light environments being heritable. For the AP experiment we used common garden lab light conditions to investigate the effect of light environment on opsin expression after sexual maturity. In summary our qPCR experiment examines nine opsin expression levels separately in three different treatment groups (EV, AP, and DP), established from our 12 laboratory evolutionary experimental populations (Fig. (1)).

Subsequently, for any of the nine guppy opsins that were differently expressed in our experimental EV populations, we then investigated how their expression levels had changed over multiple generations using digital PCR (dPCR). To identify how the expression of these candidate opsins changed over time, we sampled and examined the experimental filter populations at 4, 19, 26, and 38 months. We chose to use dPCR (rather than qPCR) as it allows us to more accurately explore the relationship between opsin expression levels across generational time. Given the additional sensitivity of dPCR, we also reexamined the possibility that candidate opsin expression levels differences were heritable after 38 months in the laboratory evolution experiment, by again comparing the EV, AP, and DP experimental treatments using dPCR. We hypothesize that if opsin expression levels remain different between light environment populations after exposure to lab light conditions, opsin expression differences may be heritable.

SAMPLING OF THE FILTERED LIGHT POPULATIONS

We sampled female guppies from each of the 12 populations (four replicates for each of light environment; Green-F89, Lilac-F55, and Clear-CF) in January 2012 (4 months), April 2013 (19 months), December 2013 (26 months), and November 2014 (EV, 38 months). The January 2012 sample is a representation of the foundation population with adults exposed to their respective light environment for four months prior to sampling. As we were aiming to stimulate the social structure in a natural population, our experimental design allowed individuals within each population to reproduce naturally at normal densities.

To ensure we sampled from the most recent generation at the April 2013, December 2013, and November 2014 time points, we separated fry from the main population using an isolating chamber (94 L). This chamber was located within each of the tanks/light environments and had fine mesh sides that allowed water to flow between the main populations and the isolated individuals. The chambers kept part of the most recent generation separated from the main population while allowing them to experience the same environmental parameters. All isolated fry were estimated to be less than three weeks old (<15 mm standard length). The isolated fish were allowed to grow to sexual maturity before sampling. (For the April 2013 time point fry were isolated in September 2012, for the December 2013 time point fry were isolated in April 2013 and for the November 2014 time point fry were isolated in December 2013.) Female guppies sexually mature at approximately between 10 and 13 weeks of age (Houde 1997).

SAMPLING OF THE DEVELOPMENTAL AND ADULT PLASTICITY EXPERIMENTS

For the DP experiment (Fig. (1)), after 27 months under the filtered light environments, approximately 25 fry from each of the 12 mesocosms (all three light environments, four replicates each) were placed in 5.2 L tanks (23.5 × 13.5 × 16.5 cm deep) under unfiltered lab light (LL) conditions. Individuals within each of these 12 tanks were then allowed to grow to sexual maturity under their new light environment. We sampled female guppies from each of the 12 populations after 11 months under the common garden treatments.

For the AP experiment (Fig. (1)), after 36 months under the filtered light environments, we removed 15 sexually mature females from each of the 12 mesocosms and placed them in 5.2 L tanks under LL light conditions. Females within each of these 12 tanks were exposed to LL conditions for 8 weeks prior to sampling. Experiments DP and AP differ from EV in that all fish were exposed to LL light within a single lifetime, testing for within-generation plasticity in response to a change from their previous Green-F89, Lilac-F55, or Clear-CF light environments to lab light conditions (LL).

GUPPY EYE SAMPLING AND PREPARATION

All sexually mature female guppies were humanely killed in 1 g/L tricaine methanesulphonate (MS-222; Sigma) between 13:00 and 14:00 (Australian EST). We chose sexually mature females because their vision is critical in mate-choice (reviewed in (Houde 1997). In addition, there is no current evidence for sexual dimorphism in adult opsin gene expression levels in natural populations (Sandkam et al. 2015). Sampling time was guided by cichlid and zebrafish studies indicating that the highest level of cone opsin expression coincides with afternoon in these diurnal fish (Halstenberg et al. 2005; Li et al. 2005).

Between October and November 2014, eyes from 10 females (approximate age = 11 months) from each of the 12 EV tanks as well as 12 from each of the DP and AP tanks were excised and pooled separately (10 individual female eyes per sample, except for T5-AP where n = 8 and T7-DP where n = 7). Eyes from 10 females (approximate age = 8–9 months) were also taken from two additional EV experimental time points for each of the 12 tanks in April 2013 (April 2013-EV) and in December 2013 (Dec 2013-EV). Each individual was exposed to MS-222 for approximately 1 minute and each dissection took on average 45 seconds.

The same protocol was also used to sample 10 females from each of the 12 foundation populations. These sexually mature foundation population individuals (age unknown) were taken at the beginning of the evolution (EV) experiment (January 2012) after having 4 months in their respective mesocosms. This was to ensure that these females contributed to the next generation and to allow the foundation female guppies to acclimate the evolutionary experimental conditions prior to sampling.

For all samples (n = 72) the order of left and right eyes extractions were alternated among individuals, to eliminate RNA degradation bias that might have occurred between death and eye immersion in RNAlater (Sigma). All samples were kept in RNAlater (Sigma) at 4°C for 24 hours to aid absorption and were then transferred to –20°C or –80°C until RNA extraction.

RNA EXTRACTION AND cDNA SYNTHESIS

Total RNA from each sample was extracted in TRI Reagent (Sigma) following the manufacturer's protocol and the quality and the quantity was assessed by agarose gel electrophoresis and by NanoDrop ND-1000 (Thermo Scientific) UV spectrophotometry, respectively. To remove genomic DNA, total RNA (1 μg) was treated with DNase I (Invitrogen) following the manufacturer's protocol. Complementary DNA (cDNA) was synthesized from 0.75 μg DNase-treated RNA using Superscript III Reverse Transcriptase (RT) (Invitrogen) following the manufacturer's protocol. At the same time, for the assessment of genomic DNA contamination, no-RT control samples were prepared from the 0.25 μg DNase-treated RNA for the all RNA samples and tested by quantitative real-time PCR (qPCR) using the P. reticulata Elongation Factor 1-alpha (PrEF-1a) primer set.

QUANTITATIVE REAL-TIME PCR

To analyze the transcriptional expression profiles of guppy opsins under different ambient lights (e.g. Green-F89, Lilac-F55, and Clear-CF), we performed Quantitative real-time PCR (qPCR). All qPCR reactions were performed using 1:20 cDNA template in a 25 mL reaction mix that comprised cDNA, SYBR^® Premix Ex Taq Master Mix (Takara) and 0.25 μM of each primer, on a Corbett Rotor-Gene 3000 (Corbett Research). Gene-specific primers were designed to amplify each of the nine cone opsins and two P. reticulata reference genes [Beta-Actin (ActinB), and Elongation Factor 1-alpha (EF-1a)] and are listed in Table S3. All primers sets were designed over exon-exon boundaries. To normalize the level of transcription for obtaining relative gene expression values, we used the geometric mean of two reference genes ActinB and EF-1a. These two reference genes have been previously demonstrated to maintain uniform transcription throughout development and under different light conditions in P. reticulata (Laver and Taylor 2011; Sandkam et al. 2015) and their stability across all our experimental sample sets was validated (and the other candidate reference genes were also tested) using the GeNorm algorithm (Vandesompele et al. 2002) using Biogazelle qBase+ software (version 3.0).

For qPCR we used the following thermoprofile: initial denaturation 95°C for 10 s, denaturation 95°C for 5 s and annealing 65.5°C for 20 s. Forty cycles were performed, with a gain of 10 on SYBR Green. Melt curve data acquisition was from 72°C to 95°C, rising by 1°C per step, with an initial hold for premelt conditioning of 45 s on the 1st step and a hold of 5 s for all subsequent steps. All samples were run in triplicate. All qPCR runs were completed using a sample maximization strategy (one gene of interest per ambient light condition in each qPCR run). In addition to the cDNAs of interest, a no-template (H₂O) control, a calibrator sample (created from equal amounts of all 12 EV cDNA samples) and a fivefold serial dilution series (created from the calibrator sample) were also included in each qPCR run and for each primer pair. The first three dilution series samples were also used as inter-run calibrators for the comparison of EV, DP, and AP runs for each gene. Primer pair specificity was verified by the presence of a single peak in melt curve analysis and the presence of a single band when running post qPCR products on an agarose gel. In addition, due to the high sequence similarity among the four LWS transcripts, all four LWS transcripts were cloned and sequenced and primer specificity was confirmed by cross-amplification reactions using plasmids that contained each LWS (data available upon request).

The efficiencies of each primer pair were calculated using the formula E = 10(-1/slope) using Biogazelle qBase+ software (version 3.0) (Table S3). Relative expression values were calculated as the ratio of gene expression between the gene of interest and the geometric mean of the two reference genes (normalization factor) relative to an average of all samples (Hellemans et al. 2009). To do this, normalized relative quantities were calculated using a sample specific normalization factor and these values were then calibrated using a run and gene-specific calibration factor (Hellemans et al. 2009). The calibrated normalized relative quantity (CNRQ) values and their standard errors were calculated using qBase+ software (version 3). We chose to measure relative opsin expression of each transcript using CNRQ values (Hellemans et al. 2009), (also referred to as relative expression values), over proportional measures of opsin expression (Fuller and Claricoates 2011) because we are interested in investigating the process by which differences in opsin transcript expression arises.

DIGITAL PCR

In addition to qPCR, for some target transcripts we also measured changes via digital PCR (dPCR), as it gives us absolute quantification of gene expression, in addition to allowing us to detect smaller fold changes accurately (Hindson et al. 2013). Again, due to the high sequence similarity among the four guppy LWS transcripts, primer specificity was confirmed by cross-amplification dPCR reactions using plasmids that contained each LWS (data available upon request). Absolute quantification of target transcripts was determined by dPCR, using EvaGreen Supermix (Bio-Rad) and the QX 200™ Droplet Digital™ PCR system (Bio-Rad), according to manufacturer's instructions. Briefly, for each assay a 25 μl final volume PCR mix was prepared containing 12.5 μl EvaGreen Supermix (Bio-Rad), 2.5 μl of each forward and reverse primer (2.5 μM; see Table S3) and 2 μl of cDNA template (diluted 1:20) or nuclease-free water as a no template control (NTC). Subsequent droplet generation and PCR amplification were performed using 20 μl of PCR mix. The 20 μl dPCR reaction mix was added to the droplet generator cartridge together with 70 μl droplet generation oil (Bio-Rad). Droplets were generated using a QX200 droplet generator (Bio-Rad), followed by gentle transfer to a 96-well PCR plate. Using a C1000 Thermal cycler (Bio-Rad), the following thermo-profile was used for target amplification: 1 cycle at 95°C for 10 min, 40 cycles at 94°C for 30 s and 63°C for 1 min, and 1 cycle at 98°C for 10 min, all at a ramp rate of 2.5°C/s. After PCR amplification, plates were analyzed using a QX200 droplet reader (Bio-Rad). The absolute quantification of each transcript was calculated using the Quantasoft v1.7.4 (Bio-Rad) and is represented as copies of transcript per microliter of amplified PCR mixture. The normalized target/EF-1a ratios are also calculated and presented to minimize any biases that are due to differences in reverse transcription (Heredia et al. 2013; Sanders et al. 2013).

ANALYSIS OF qPCR EXPERIMENTS

Differences in opsin expression CNRQ values among populations for each opsin gene were tested with linear mixed models using the lme4 package (Bates et al. 2015) within R software (R Development Core Team 2011). Initially for each opsin gene we examined the effect of the light environment (i.e., light; Green-F89, Lilac-F55, Clear-CF) within each experimental treatment (i.e., experiment; EV, DP, AP) using the linear model CNRQ ∼ light.

We also examine whether the expression level of each opsin gene can be predicted by light environment (Green-F89, Lilac-F55, Clear-CF), experimental treatment (EV, AP, and DP) or their interactions. The original light environment (Green-F89, Lilac-F55, Clear-CF), and treatment of the experiment (EV, DP, AP) were modeled as fixed effects and the 12 individual replicate source (EV) populations (i.e., tanks) as random intercept effects. This resulted in the linear-mixed model CNRQ ∼ light*experiment+(1|tank). This linear-mixed model was run separately for each of the nine opsins.

To further explore the multiple interactions between experiment and light on opsin expression levels (CNRQ), we conducted post-hoc tests that examined the effect of our common garden experimental light regime within each light environment population (i.e., multiple pairwise comparisons of EV, AP, and DP within F89 populations, F55 populations and CF populations, respectively). In addition, we also conducted post-hoc tests that examined the effect of the different light environments after 38 months of population light exposure (e.g., multiple pairwise comparisons of F89-EV, F55-EV, CF-EV). To control for error rate the reported P values were adjusted for using the Tukey method for comparing a family of three estimates. All post-hoc tests were conducted using the lsmeans package (Lenth 2015) with Tukey adjustment in R software (R Development Core Team 2011).

ANALYSIS OF dPCR EXPERIMENTS

We used a Generalized Additive Model for Location Scale and Shape (GAMLSS) in R software using the gamlss package (Bates et al. 2015) to examine whether light, time, or their interaction were significant factors in predicting LWS-1 or LWS-3 dPCR normalized expression values. We chose to use GAMLSS, rather than lme4 or lm, because of the nonlinearity and heterogeneity in dPCR expression over time and within different light environments (Rigby and Stasinopoulos 2010). The light environment (Green-F89 and Clear-CF) and sampling time (4, 19, 26, and 38 months) were modeled as fixed effects and the full model is dPCR ratio ∼ light * time. Differences between LWS-1 and LWS-3 dPCR normalized expression values in Green-F89 and Clear-CF populations at each individual time point were also investigated using Mann–Whitney and Fligner–Killen tests, or one-way ANOVAs.

Using a linear-mixed model, we also examined differences in dPCR gene expression using the model dPCR ratio ∼ experiment +(1|tank). This was followed by a post-hoc test that examined the effect of our common garden experimental light regime within the Green-F89 light environment populations. Reported P values were adjusted for using the Tukey method for comparing a family of three estimates. All post-hoc tests were conducted using the lsmeans package (Lenth 2015) with Tukey adjustment in R software (R Development Core Team 2011).

Results

To test directly whether opsin transcript expression levels are influenced by environmental light, we examined fish from our filtered light populations, where only the light environment varied (Fig. (1)). The three different ambient light treatments (Green-F89, Lilac-F55, and Clear-CF) stimulate guppy cone photoreceptors in different ways (Figs. (2), S1 and S2) with each type of cone photoreceptor associated with the expression of a different opsin gene/s (reviewed in Table S1). The divergence in cone stimulation estimates between the different light environments (Fig. (2)) allows us to explore the direct effect of different light colors (spectral shape) on the expression of guppy opsin transcripts after different experimental treatments (EV, AP and DP; see Fig. (1) for details). While not perfect representations of the light experienced by natural populations (mostly due to the lack of 300–350 nm wavelengths), the filtered light environments stimulate guppy cones in similar ways to the light experienced by guppies in their natural habitat ranges (Endler 1991, 1993b). In addition to our absolute and relative cone stimulation estimates (Fig. S2 and Fig. (2), respectively) for different guppy cone types, we also calculated absolute cone stimulation estimates for the different light environments, based on the previously measured peak absorbance (λ_max) of each of the opsin visual pigments reconstituted in vitro (see Table S2 for details). These cone stimulation calculations allow us to explore the direct effect of different light environments on the expression of guppy opsin transcripts.

EV, DP, AND AP EXPERIMENT

The results from our qPCR experiments show that the relative expression levels of nine different guppy opsin genes responded differently to light environment changes (Figs. (3) and (4) and Table (1)). We report that some opsin expression levels did not differ after populations were exposed for 38 months (or approximately 8–12 overlapping generations) to the Green-F89, Lilac-F55, or Clear-CF light (EV experiment) (Fig. (3), Table S4 and S5). In contrast other opsins were differentially expressed after exposure in the EV experiment (Fig. (3) and Table S4). However, we found no differential opsin expression levels among populations within in either of the common garden experiments (AP and DP experiments; Fig. (3) and Table S4). Yet, we did find differential expression patterns when comparing the opsin expression levels among the EV, AP, and DP experiments within the Green-F89, Lilac-F55, or Clear-CF populations (Fig. (4) and Table S6).

Figure 3

Summary of trends in guppy opsin relative transcript expression changes among F89, F55, and CF populations in the three different experimental treatments. Using qPCR relative expression values we summarize the response of different guppy opsins after multiple generations under the F89, F55, or CF filtered light environments, to three different experimental treatments (EV, AP, and DP). The three categories of a comparative increase (red), no change (gray) and a comparative (blue) were based on the statistical significance of light environments reported in Table S5 and S6. CNRQ, calibrated normalized relative quantity. F89, Green-F89 filter populations; F55, Lilac-F55 filter populations; CF, Clear-CF filter populations. EV, evolution experiment (filtered light conditions); AP, adult plasticity experiment; DP, developmental plasticity experiment.

Figure 4

Summary of trends in guppy opsin relative transcript expression changes among EV, AP, and DP within the F89, F55 and CF populations. Using qPCR relative expression values we summarize the response of different guppy opsins to three different experimental treatments, EV, AP, and DP for each different light environment (F89, F55, and CF), respectively. The three categories of a comparative increase (red), no change (gray), and a comparative (blue) were based on the significance of experiment treatments reported in Table S5. CNRQ, calibrated normalized relative quantity. EV, evolution experiment (filtered light conditions); AP, adult plasticity experiment; DP, developmental plasticity experiment. F89, Green-F89 filter populations; F55, Lilac-F55 filter populations; CF, Clear-CF filter populations.

Table 1.

Linear-mixed model results for each cone opsin gene analyzed separately using the model CNRQ ∼ light*experiment + 1|Tank

	SWS1				SWS2a				SWS2b				Rh2-1
Fixed effects	Estimate	Std. Error	t value	P-value	Estimate	Std. Error	t value	P-value	Estimate	Std. Error	t value	P-value	Estimate	Std. Error	t value	P-value
Intercept	0.96591	0.107	9.022	<0.001	1.42202	0.149	9.524	<0.001	1.043231	0.079	13.12	<0.001	0.92661	0.102	8.998	<0.001
EV experiment	0.16640	0.145	1.146	0.26	−0.33979	0.211	−1.609	0.119	0.164786	0.109	1.510	0.148	0.73232	0.145	5.028	<0.001
AP experiment	0.14499	0.145	0.999	0.31	−0.42893	0.211	−2.031	0.052	−0.161237	0.109	−1.478	0.156	0.03445	0.145	0.237	0.815
F89 light	0.16013	0.151	1.058	0.30	−0.31308	0.211	−1.483	0.149	0.090747	0.112	0.807	0.426	−0.02803	0.145	−0.192	0.849
F55 light	−0.03726	0.151	−0.246	0.80	−0.27413	0.211	−1.298	0.205	−0.139070	0.112	−1.237	0.226	−0.21361	0.145	−1.467	0.154
F89 light: EV experiment	−0.24447	0.205	−1.191	0.249	0.23875	0.298	0.799	0.431	−0.285841	0.154	−1.852	0.080	−0.07623	0.205	−0.370	0.714
F55 light: EV experiment	−0.02446	0.205	−0.119	0.906	0.28229	0.298	0.945	0.359	−0.011816	0.154	−0.077	0.939	0.15624	0.205	0.759	0.455
F89 light: AP experiment	−0.23037	0.205	−1.122	0.277	0.48755	0.298	1.633	0.114	−0.003463	0.154	−0.022	0.982	−0.00206	0.205	−0.010	0.992
F55 light: AP experiment	−0.05781	0.205	−0.282	0.781	−0.02711	0.298	−0.091	0.928	0.144032	0.154	0.930	0.362	−0.09060	0.205	−0.440	0.664
Random effects	Variance	Std. Dev.			Variance	Std. Dev.			Variance	Std. Dev.			Variance	Std. Dev.
tank	0.0037	0.06086			0	0			0.0014	0.0383			0	0
residual	0.0421	0.20530			0.0891	0.2986			0.0238	0.1543			0.042	0.206
	Rh2-2				LWS-1				LWS-2				LWS-3
Fixed effects	Estimate	Std. Error	t value	P-value	Estimate	Std. Error	t value	P-value	Estimate	Std. Error	t value	P-value	Estimate	Std. Error	t value	P-value
Intercept	1.3003	0.115	11.26	<0.001	1.41136	0.115	12.20	<0.001	0.92617	0.104	8.858	<0.001	1.58273	0.190	8.310	<0.001
EV experiment	−0.3930	0.163	−2.407	0.023	−0.56564	0.163	−3.459	0.001	0.38724	0.147	2.619	0.014	−0.86279	0.269	−3.203	0.003
AP experiment	−0.4140	0.163	−2.536	0.017	−0.59072	0.163	−3.612	0.001	−0.07589	0.147	−0.513	0.612	−0.64732	0.269	−2.403	0.023
F89 light	−0.1240	0.163	−0.760	0.454	−0.21772	0.163	−1.331	0.194	0.10182	0.147	0.689	0.497	−0.39162	0.269	−1.454	0.157
F55 light	−0.3034	0.163	−1.859	0.073	−0.08917	0.163	−0.545	0.590	−0.11830	0.147	−0.800	0.430	0.07075	0.269	0.263	0.794
F89 light: EV experiment	0.7519	0.230	3.257	0.003	0.63787	0.231	2.758	0.010	−0.09446	0.209	−0.452	0.655	1.11872	0.380	2.937	0.006
F55 light: EV experiment	0.3498	0.230	1.515	0.141	0.02326	0.231	0.101	0.920	0.10579	0.209	0.506	0.617	−0.10172	0.380	−0.267	0.791
F89 light: AP experiment	0.1905	0.230	0.825	0.416	0.36588	0.231	1.582	0.125	0.02719	0.209	0.130	0.897	0.47455	0.380	1.246	0.223
F55 light: AP experiment	0.1606	0.230	0.696	0.492	0.07003	0.231	0.303	0.764	−0.09644	0.209	−0.461	0.648	−0.10908	0.380	−0.286	0.776
Random effects	Variance	Std. Dev.			Variance	Std. Dev.			Variance	Std. Dev.			Variance	Std. Dev.
tank	0	0			0	0			0	0			0	0
residual	0.053	0.230			0.053	0.231			0.043	0.209			0.145	0.380
	LWS-4
Fixed effects	Estimate	Std. Error	t value	P-value
Intercept	0.87079	0.271	3.202	0.003
EV experiment	1.27386	0.384	3.312	0.002§
AP experiment	0.01323	0.384	0.034	0.972
F89 light	0.16013	0.384	0.416	0.680
F55 light	0.08127	0.384	0.211	0.830
F89 light: EV experiment	−1.11594	0.543	−2.052	0.051
F55 light: EV experiment	−0.98999	0.543	−1.820	0.070
F89 light: AP experiment	−0.08476	0.543	−0.156	0.877
F55 light: AP experiment	−0.29628	0.543	−0.545	0.590
Random effects	Variance	Std. Dev.
tank	0	0
residual	0.295	0.543

The linear-mixed model is CNRQ ∼ light*experiment + 1|Tank. CNRQ, calibrated normalized relative quantity values. Light = F89, Green-F89 filter; F55, Lilac-F55 filter; CF, Clear-CF filter. Experiment = EV, Evolution experiment; AP, adult plasticity experiment; DP, developmental plasticity experiment. Lab light (LL) was used for the AP and DP experiments. Bold numbers indicate statistically significant predictors. Tank = each evolutionary experimental population (n = 12). § when T0₂ populations are removed from the model, the EV experiment is no longer a significant predictor of LWS-4 CNRQ values.

SHORT-WAVE SENSITIVE OPSIN EXPRESSION

For the guppy short-wave sensitive (SWS) opsins we found no significant effect of light, experiment, nor their interactions on the relative expression levels of SWS1, SWS2a, and SWS2b (Table (1) and Fig. (3)). Photoreceptors containing SWS1 opsins are not predicted to be differentially stimulated among the three different light environments (Figs. (2) and S2; Table S2).

In contrast to SWS1, the photoreceptors containing the SWS2 opsins are predicted to be differentially stimulated among the three different light environments (Figs. (2) and S2; Table S2). However, for the SWS2a and SWS2b opsins, we find no obvious relationship between the probability of cone stimulation and gene expression in our multiple generation evolutionary experiment.

Rh2 and long-wave sensitive opsin expression

For the Rh2 and long-wave sensitive (LWS) opsins, we have demonstrated that five out of the six opsin transcripts, all show unique expression patterns in response to the different light conditions found in our evolution (EV), adult plasticity (AP), and developmental plasticity (DP) experiments (Table (1), Fig. (4)). These five unique expression pattern responses are reported in further details below. For the remaining opsin, LWS-4, differential expression was only found in a single CF-EV population replicate (T02 in Fig. S3). Photoreceptors containing LWS-4 are predicted to be differentially stimulated among the three different light environments (Figs. (2) and S2; Table S2). Hence, LWS-4 expression is not influenced by environmental light in our experiments (Table (1)). Of note, LWS-4 expression has only been reported in the peripheral-dorsal retina of some individuals (Rennison et al. 2011), its role in guppy vision is currently unknown and it is the only Rh2 or LWS opsin that is found on a separate chromosome (e.g., Künstner et al. 2016; Watson et al. 2011). From our Rh2 and LWS opsins DP and AP plasticity tests, Rh2-1 expression is lower in both plasticity experimental fish (DP, AP) compared to the evolution fish (EV), irrespective of the ambient light environment in which they lived under (all P < 0.05, Fig. (4); Table S6). Rh2-1 is the only guppy opsin transcript whose relative expression levels correlate with light intensity (Fig. S4, r = 0.85, P = 5.42 × 10^–11).

For the remaining four Rh2 and LWS opsins, we found two different trends in our qPCR data. First, Rh2-2, LWS-1, and LWS-3 all show relative expression differences among light environments in our evolution (EV) experiment (Table (1), Fig. (3)). Second, LWS-1, LWS-2, and LWS-3 all exhibit plasticity in the DP and AP experiments with relative expression differences depending upon the light environment in which they evolved (Table (1), Fig. (4), and Table S6). We find that overall the Green-F89 light environment had different responses than both the Clear-CF and Lilac-F55 environments (Figs. (3) and (4)). These qPCR results lead us to conduct the two following dPCR experiments. First we examined how Rh2-2, LWS-1, and LWS-3 expression patterns change over multiple generations in our EV experiment (Fig. (5)). Second we examined the likelihood that any differences in opsin transcript expression levels found in our EV experiment were heritable (Fig. (6)).

Figure 5

Normalized absolute gene expression ratios for Rh2-2, LWS-1, and LWS-3 over multiple generations in each Green-F89 and Clear-CF population replicate. Panels A, C, and E show the four replicate light environment populations separately for Rh2-2, LWS-1, and LWS-3 and panels B, D, and F show their mean and standard deviations. The four time samples during the experiment were taken as follows: foundation (sexually mature females sampled in January 2012 after having 4 months in their respective mesocosms [i.e., adult plasticity only]); April 2013 (19 months), December 2013 (26 months) and November 2014 (EV, 38 months). For each time point four replicate populations were analyzed by digital PCR from the Green-F89 and Clear-CF experimental light conditions. Each sample consists of 7–10 individual eyes all from sexually mature females. Target gene expression was normalized to Elongation Factor 1-alpha (EF-1a) (see Materials and Methods for details). T-numbers refer to tank numbers. An asterisk illustrates where there is a statistically significant difference between the Green-F89 and Clear-CF populations at a given time point.

Figure 6

Normalized absolute gene expression ratios for LWS-1 among experimental treatments for each Green-F89 and Clear-CF population replicate. Using digital PCR we measure LWS-1 expression in three different experimental treatments EV, AP, and DP for each Green-F89 and Clear-CF replicate population. Each sample consists of 7–10 individual eyes all from sexually mature females. Target gene expression was normalized to Elongation Factor 1-alpha (EF-1a) (see Materials and Methods for details). EV, evolution experiment (filtered light conditions); AP, adult plasticity experiment; DP, developmental plasticity experiment. T-numbers refer to tank numbers.

Here, it is also important to note that in guppies LWS-1, LWS-2, and LWS-3 are linked in a tandem array (Watson et al. 2011) and evidence from other teleost fish suggests that all three LWS opsins may be coregulated (Tam et al. 2011; Tsujimura et al. 2010). In our qPCR data we report relative expression differences in LWS-1 and LWS-3 between the different light environments, but not in LWS-2 (Table (1) and Fig. (3)). Hence based on our results we did not examine LWS-2 expression further via dPCR.

EFFECT OF TIME ON THE DIFFERENTIALLY EXPRESSED OPSINS

To examine the role of plasticity in the differential expression of Rh2-2, LWS-1, and LWS-3 under different light environments, we examined their gene expression over six time points in each of the Green-F89 and Clear-CF populations using digital PCR (dPCR). We omitted Lilac-F55 populations from our dPCR examination because their qPCR (i.e., CNRQ) opsin expression levels were similar to the Clear-CF populations (Fig. (3) and Table S4). We did not use dPCR over the entire experiment due to financial constraints. Our qPCR and dPCR expression values are correlated (Fig. S5; all P > 0.05).

Rh2-2 expression

For Rh2-2 we report no consistent relationship between expression and light environment within any of the time points (Fig. (5)A and B). In addition, we found no linear relationship between Rh2-2 and LWS-1 expression levels (Fig. S6A) or Rh2-2 and LWS-3 expression levels (Fig. S6B). Between the April 2013 and December 2013 samples we also observed an inverse relationship between the direction, variance, and mean of the Rh2-2 and LWS-1 expression levels (see Fig. (5)B compared to Fig. (5)D).

LWS-1 and LWS-3 expression

In contrast to Rh2-2 expression, we report a relatively more consistent relationship between LWS-1 and LWS-3 expression levels and the two different light environments over multiple generations (Fig. (4)C–F). Guppies from the Green-F89 light populations typically express more LWS-1 and LWS-3 than guppies from the Clear-CF light populations (Fig. (5)D–F; Table S7 and Wilcoxon rank sum tests [n = 32]; LWS-1, P = 0.035; LWS-3, P = 0.0005). In addition LWS-1 and LWS-3 expression is strongly correlated (Fig. S6c; P = 1.65 × 10^–7). We also report that time has no effect on mean LWS-1 and LWS-3 opsin expression levels in our full GAMLSS model (Table S7). However, the variance of LWS-1 and LWS-3 expression does change across multiple generations (Fig. (5)D–F; Table S8).

In the foundation populations (4 months after the filter placement), we report that LWS-1 and LWS-3 expression variance differs between the Green-F89 and Clear-CF light environments (Fig. (5)C–F, Fligner–Killeen tests P < 0.05, Table S8). We find high variance in LWS-1 and LWS-3 expression among the Green-F89 populations (Table S8). This corresponds with adult guppies being shifted from the broad-spectrum lab light environment to the relatively narrow spectrum of the Green-F89 light environment (see Fig. S1). Conversely, we find low variance in LWS-1 and LWS-3 expression among the Clear-CF populations (Table S8). This corresponds with adult guppies being moved into a similar light environment (LL to Clear-CF; see Fig. S1).

In the following April 2013 (19 months) and Dec 2013 (26 months) EV populations, we found no statistically significant differences in LWS-1 or LWS-3 expression between the Green-F89 and Clear-CF populations in their variance (Fligner–Killeen test, all P > 0.05) or medians (Mann–Whitney test, all P > 0.05) (Fig. (5)C–F and Table S8). However, in the Green-F89 populations there was a decrease in mean LWS-1 expression at the Dec 2013 time point compared to the other three time points measured (Fig. (5)C and D; Table S9).

Approximately one year later, after LWS-1 expression levels converged in the Dec 2013 populations, we observed both a general increase in LWS-1 expression (Fig. (5)C and D; Table S9) and a statistically significant difference for LWS-1 expression between Green-F89 and Clear-CF populations in the Nov 2014 (EV) time point (One way ANOVA, F_{(1, 6)} = 11.23, P = 0.01). We found no statistically significant differences in LWS-3 expression between the Green-F89 and Clear-CF populations at the Nov 2014 (EV) time point (One way ANOVA F_{(1, 6)} = 4.41, P = 0.08).

To examine if there was any evidence for longer-term changes in LWS-1 gene expression levels between Green-F89 and Clear-CF populations, we compared gene expression among the latest evolution samples (EV-2014) and guppies that were raised under lab light conditions, either from adulthood (AP-2014) or early development (DP-2014) (Fig. (6), Table S10 and S11). We find that LWS-1 expression levels are different among the EV, AP, and DP experiments for the Green-F89 populations (Fig. (6), Table S10 and S11), indicating that the Green-F98 populations remains phenotypically plastic, even after guppy populations had lived under different light environments for 38 months, or approximately 8–12 overlapping generations.

The lack of differentiation between the Green-F89 and Clear-CF populations within the AP and DP experiments suggests no evolutionary divergence has occurred for LWS-1 gene expression in response to the different light environments in our evolution experiment (One way ANOVAs, P > 0.05).

Discussion

After 38 months in our multigeneration, large-scale evolutionary experiment (EV), the relative expression levels of nine different guppy opsin genes responded differently to light environment changes. For the SWS opsins, we report no differential relative expression among the light environments. In contrast, the environmental light environment regime induced large plastic effects on the relative expression of Rh2 and LWS opsins. Below we discuss each of our major experimental findings in the context of their potential roles in “tuning” guppy vision to different light environments. We then outline some new considerations for future opsin expression studies and discuss the potential evolutionary consequences of opsin phenotypic plasticity in the context of sensory drive. These results are relevant to a broad range of taxa because opsins are the universal photoreceptor molecules of all visual systems in the animal kingdom. We then close with a brief and general discussion of the theoretical implications of common plastic response requiring multiple generations under new environment conditions.

SWS OPSIN EXPRESSION AND THE LIGHT ENVIRONMENT

We found no differential relative expression in SWS opsins (SWS1, SWS1a, and SWS2b) among light environments (Fig. (3)). The SWS1 result is consistent with the essentially invariant estimated relative cone stimulation (RCS) in the three light environments (Fig. (2)). SWS2a expression levels also showed no relative expression differences. This may be due to SWS2a transcripts being relatively rare in qPCR studies (this study; Ehlman et al. 2015; Laver and Taylor 2011; Sandkam et al. 2015). SWS2b expression levels were also not affected by light environments, despite their different RCS estimates (Fig. (2)) in conjunction with their relatively high expression levels (this study; Ehlman et al. 2015; Laver and Taylor 2011; Sandkam et al. 2015). This was unexpected given that relative expression differences have been reported for SWS2b among natural guppy populations from different local environments (Sandkam et al. 2015).

There are several possible reasons for the disparity in SWS2b expression between our experiment and natural populations. First, as opsin transcripts are expressed in a circadian rhythm (Li et al. 2005; Sandkam et al. 2015) and all our eye samples were consistently taken in the afternoon (between 13:00 and 14:00 AEST), we may not have captured gene expression differences that may appear at other times of day, although lights were constant in our experiment. Second, our light conditions are a subset of that in the wild (i.e., more similar to forest shade and small gaps natural light environments; Endler 1993b), with relatively little light below 400 nm, resulting in only a subset of naturally occurring SWS2b expression patterns. Additional factors, such as population foraging preferences and male coloration patterns, could also affect SWS opsin transcript expression levels in guppies, as has been suggested for cichlids (Hofmann et al. 2009) and birds (Bloch 2015).

In summary, SWS opsin relative expression levels did not respond to light environment treatments in our experiment, but this may reflect the lower relative availability of shorter wavelength light in our experiments compared to natural light environments.

Rh2 OPSIN EXPRESSION AND THE LIGHT ENVIRONMENT

For the Rh2 opsins, we report their differential relative expression under different light environment treatments (Fig. (3)) in conjunction with the predicted differential stimulation of their photoreceptors under the different light environments (Figs. (2) and S1).

For Rh2-1, although we do not find differences among the three EV light environments, we do find that Rh2-1 expression is lower in both the lab light experimental populations (DP, AP) compared to the evolution fish populations (EV), irrespective of the ambient light environment in which they evolved (Fig. (4)). Moving guppies from the Green-F89 and Lilac-F55 to the lab light (LL) in the plasticity experiments corresponds to changes in light color (spectral shape) and a reduction in light intensity, whereas moving from the Clear-CF to Lab light-LL corresponds only to reduction in light intensity (see Fig. (2) compared to Fig. S2; The reduction in light intensity is due to the greater distance between the ceiling fluorescent lamps and the DP and AP tanks compared to the EV tanks). This suggests that Rh2-1 responds to light intensity rather than color (Fig. S4). Of note, Rh2-1 expression levels also decreased when guppies were reared in turbid water developmental treatments in contrast to clear water developmental treatments (Ehlman et al. 2015). These findings are consistent with the expression of Rh2 opsin subfamilies in double cones (Hofmann and Carleton 2009) and the potential role of double cones in luminance vision (Lythgoe 1979). Taken together, we suggest that differential Rh2-1 expression levels play an important role in optimizing guppy vision under different light intensities.

For Rh2-2, we find relative qPCR expression differences among the different light environments after 38 months in our multigeneration, large-scale evolutionary experiment (EV) (Figs. (3) and (4)). However, further analysis of Rh2-2 via digital PCR revealed no consistent relationship between its expression and the light environment over multiple generations (Fig. (5)A and B). This suggests that Rh2-2 gene expression plasticity is not simply a direct consequence of the light environment.

One possible explanation for the differences in Rh2-2 expression over time is the potential functional roles Rh2-2 may play in intraretinal variability and spectral sensitivity. More specifically, unlike all the other Rh2 and LWS opsins transcripts expressed in the double cones, Rh2-2 expression is spatially uniform throughout the retina and is putatively assigned to the shortest (a.k.a. accessory) member of a double cone (reviewed in Laver and Taylor 2011). The shortest member of a double cone is typically paired with a longer wavelength sensitive cone (e.g., Rh2-1 in the ventral retina, or a LWS class opsin in the central or dorsal retina) (Rennison et al. 2011). As such, phenotypically plastic changes in Rh2-2 expression levels may allow both the dorsal and the ventral regions of the guppy retina to divergently specialize with respect to wavelength discrimination and/or sensitivity under different light environments over different time scales. Alternatively, differences in Rh2-2 expression levels over time may be caused by other factors such as genetic drift. In either case, our experimental results are consistent with the differential Rh2-2 expression levels found among natural guppy populations from different local environments (Sandkam et al. 2015).

LWS EXPRESSION AND THE LIGHT ENVIRONMENT

We find relative expression differences for both LWS-1 and LWS-3 among the different light environments after 38 months (or approximately 8–10 overlapping generations) in our multigeneration, large-scale evolutionary experiment (EV) (Fig. (3)). Further analysis via digital PCR revealed that guppies from Green-F89 populations have higher LWS-1 expression levels than guppies from Clear-CF populations at the 38-month time point (Fig. (4)C and D), whereas LWS-3 expression levels were not statistically different (Fig. (4)E and F). As the light environment is the only introduced variable in our experiments, we infer that the light environment directly influences LWS-1 expression levels. Moreover, we find that a common plastic response in LWS-1 expression (i.e., a narrowing of the phenotypic range) is only present after multiple generations under the Green-F89 light environment. To our knowledge this is the first reported experimental evidence for differences in opsin expression between one or two generations compared to multiple generational exposure to the same spectral light environment.

POSSIBLE VISUAL CONSEQUENCES OF MULTIGENERATION DIFFERENTIATION IN LWS-1 EXPRESSION LEVELS

Given that we find an increase in LWS-1 expression levels under the Green-F89 light environment, Green-F89 populations should have a higher sensitivity to peak wavelengths at 560 nm (LWS-1_Ala) or 572 nm (LWS-1_Ser), depending on which allelic variant is expressed (Kawamura et al. 2016). Based on our absolute cone stimulation estimate comparisons (Table S2) we predict that LWS-1_Ala and LWS-1_Ser containing photoreceptors should be less stimulated in Green-F89 than in Clear-CF light conditions. Therefore, we propose that in given environmental conditions of low photon flux (of appropriate wavelengths), LWS-1 opsin expression may be upregulated to increase the sensitivity of the photoreceptor cell. In turn, we hypothesize that the increased sensitivity of LWS-1 containing photoreceptors may enhance the efficiency of common visual tasks in the Green-F89 light environment (i.e., “tune” guppy vision).

Given the positive correlation between LWS-1 and LWS-3 expression levels (Fig. S6C), we suggest that LWS-1 and -3 expression ratios may also be important in enhancing visual system efficiency under different light environments. Changes in the ratio of LWS-1 and LWS-3 expression levels may influence a guppy spectral sensitivity (λ_max) across different spatial scales. For example, if LWS-1 and LWS-3 are expressed together in a single photoreceptor cell, changes in their ratio may change the photoreceptors spectral sensitivity (λ_max). This is congruent with the “coexpression” model, where the large spectral sensitivity (λ_max) variance is reported for the “Green” to “Yellow” cone classes in guppies (Kawamura et al. 2016).

POSSIBLE MECHANISMS OF LWS-1 DIFFERENTIATION OVER MULTIPLE GENERATIONS

To address whether LWS-1 expression population differences are phenotypically plastic or whether they potentially evolve under the different light conditions, we examined LWS-1 expression over multiple generations in both the Green-F89 and Clear-CF light environments. We report both evidence for an immediate difference in the plastic response of LWS-1 expression levels under the two different light conditions, in conjunction with evidence that the phenotypic range of LWS-1 expression narrows over multiple generations for populations under the Green-F89 light conditions. However, our common garden experiments demonstrate that the evolution of LWS-1 expression has not (yet) occurred in our evolutionary experimental populations after 38 months.

The Green-F89 light environment is very different compared to either of the other experimental light environments (Clear-CF and Lilac-F55), as well as when compared to the preferred light conditions of the native Alligator Creek populations (Cole and Endler 2016). This may have resulted in much stronger physiological stress and stronger selection than in the other two experimental light environments. It has been suggested that, in response to stressful environments, plastic responses can result in increased variance around the phenotypic mean due to the expression of cryptic genetic variation, facilitating evolution if some of the new variants have higher fitness (reviewed in Ghalambor et al. 2007). It is therefore noteworthy that a common plastic response in LWS-1 expression (i.e. a narrowing of the phenotypic range) was only present after multiple generations under the Green-F89 light environment in our experiments. We therefore suggest that any plastic-induced LWS-1 expression variance among Green-F89 populations may provide the opportunity for LWS-1 expression to evolve over longer time periods via the release of cryptic genetic variation(reviewed in Ghalambor et al. 2007). Yet our common garden experiments demonstrate that the evolution of LWS-1 expression has not yet occurred (i.e., LWS-1 expression has not become genetically assimilated in our populations after 38 months). Instead our findings raise the interesting possibility that the genes regulating the plastic response of LWS-1 may evolve via selection in response to different light environments.

More generally, the finding that a common plastic opsin response requires multiple generations in a new environment is relevant to other species that rely on vision because it implies that new light environments can create the conditions for the rapid adaptation of opsins via the release of cryptic genetic variation that selection can then act upon. It also suggests the possibility of evolution of the plastic mechanisms themselves.

FUNCTIONAL UTILITY OF LWS-1 PLASTICITY

In summary, we suggest that LWS-1 plasticity is an effective mechanism that enhances the efficiency of common visual tasks in different spectral light environments (i.e., “tune” guppy vision). The idea that opsin expression plasticity plays a critical role in light environmental-tuning is reasonable given that (i) guppy populations (and almost all other species of vertebrates and invertebrates) are frequently exposed to variable light environments in nature, (ii) light environments produce constant and reliable cues (iii) different light environments should favor different opsin expression profiles and (iv) no single opsin phenotypic profile is likely to exhibit superior fitness across all environments (e.g., Hofmann et al. 2009).

It is also important to note here that over longer evolutionary time periods, directional selection for different LWS-1 expression levels may be feasible given that LWS-1 expression is variable among different populations (this study; see also Sandkam et al. 2015), and hence there is a potential genetic component to opsin expression (Carleton 2009; Fuller et al. 2005; Fuller and Claricoates 2011; Hofmann and Carleton 2009) and that differential reproduction among different “LWS-1 expressing” females appears reasonable (i.e., sensory drive hypothesis; Endler 1992; Endler and Basolo 1998). In addition, our experimental results are also consistent with the LWS opsin changes reported among natural guppy populations from different local environments (for details see Sandkam et al. 2015; Tezuka et al. 2014).

FUTURE CONSIDERATIONS FOR OPSIN EXPRESSION STUDIES

The multigenerational opsin experimental results presented here provide us with a framework to further explore how the spectral light environment may affect opsin expression levels in animal species that rely on vision for survival and reproduction. We suggest that future experiments and discussions around opsin expression-light environmental tuning should take the following three points into consideration. First, different opsin genes’ expression levels can respond differently to changes in the light environment, allowing for some independent flexibility in response to particular parts of the visible spectrum. Second, the temporal scale of the light environment change is an important factor in determining in the type of opsin phenotypic plasticity response, and may affect both its phenotypic variance and mean. In other words, opsins expression levels may respond to fine-scale or coarse-scale temporal changes in the light environment, or both. Third, common plastic opsin responses may require multiple generations under new light environments. When taken together, these considerations should advance our understanding of how visual sensory systems can respond to different spectral and temporal changes in the light environment and thus are relevant to any animal that relies on vision to survive and reproduce under changing light conditions.

CONCLUSIONS

In summary, we provide novel, direct experimental evidence that LWS-1 expression levels show plasticity and change as a result of different environmental light conditions over multiple generations. Previously, the evidence for evolution of opsin expression-light environment tuning was either correlative or focused exclusively on the relationship between the light environment and opsin expression over one or two generations. Over 8–12 overlapping generations, our results suggest that the common plastic responses of LWS-1 enable the visual system to function efficiently under different light environments.

Moreover, the plasticity in guppy opsin expression levels may provide both the mechanism to adapt quickly to short-term changes in the light environment and provide the additional variation upon which selection can operate. Once a phenotype has higher fitness under a particular light environment this may translate to genetic variation ultimately driving the evolution of the visual system over many generations (Pigliucci et al. 2006). However, the genetic changes involved in the evolution of the visual system in response to different light environments still needs to be determined experimentally. While further work is needed to understand the consequences of guppy LWS-1 opsin expression variation on opsin protein expression, neural processing, visual perception, and visually based behaviors (as well as whether genetic expression levels confer to genetic polymorphism) we believe our results provide a validation of the first tenet of the sensory drive hypothesis (Endler 1992; Endler and Basolo 1998) that is, new environmental conditions during signal reception (e.g., ambient light) can directly or indirectly cause changes in sensory systems. Critically, if different LWS-1 opsin expression levels do confer a change in an animal's visual sensitivity, this could in turn drive the diversification of visual signals and preferences and ultimately lead to speciation (Boughman 2001; Endler 1992; Endler and Basolo 1998; Seehausen et al. 2008).

Beyond sensory systems, our research also supports the more general idea that plastic responses may evolve in response to environmental change (Grenier et al. 2016; Pigliucci et al. 2006; Scheiner 1993). Currently understanding how a trait's plastic responses might evolve is a significant and complex challenge in animals, especially because of the lack of knowledge around the genetic basis of trait plasticity. Future studies that examine how genetic (and potentially epigenetic) mechanisms regulating plasticity function and evolve under different environmental conditions should further our understanding of adaptive evolution. Indeed, the opportunity for selection to influence the genetic variation for plasticity in ecologically relevant traits may be an under-appreciated mechanism by which populations can evolve in response to changing environments. Whether early adaptive evolution in response to novel environments commonly includes divergent changes in plastic responses also remains to be seen.

AUTHOR CONTRIBUTIONS

J.A.E and A.M.K conceived the study and developed and designed the experiments; A.M.K gathered and conducted all experiments with contributions from L.G.F and G.L.C.; Data analysis was performed by A.M.K and J.A.E.; A.M.K and J.A.E interpreted the data and wrote the article.

ACKNOWLEDGMENTS

We thank P. Singh for assistance with fish husbandry; R. Douglas for providing the guppy cornea and lens transmission spectrum; Deakin University for financial support; and Daniel Osorio, Beata Ujvari, Ellis Loew, and Lee-Ann Rollins and the anonymous reviewers for excellent comments on the manuscript. Guppies (Poecilia reticulata) were collected in 2010 with a Queensland Wildlife Parks permit (WITK07655010) and Deakin University ethics approval (A21-2010 and G01-2012). This work was supported by Australian Research Council grants DP110101421 and DP150102817 to J.A.E. The authors declare no conflict of interest.

DATA ARCHIVING

All relevant data for our analyses has been archived at Dryad https://doi.org/10.5061/dryad.0d19278.

Associate Editor: E. Tibbetts

Handling Editor: Mohamed A.F. Noor