Developmental effects of environmental light on male nuptial coloration in Lake Victoria cichlid fish

Fish rearing & maintenance

Offspring of wild caught P. pundamilia and P. nyererei, collected at Python Islands in the Mwanza Gulf of Lake Victoria (−2.6237, 32.8567 in 2010 & 2014), were reared in light conditions mimicking those in shallow and deep waters at Python Islands (as in: Maan et al., 2017; Wright et al., 2017). Lab-bred lines (hybrid and non-hybrid) were created opportunistically as reciprocal crosses, with 18 dams and 14 sires. Hybridization does occur with low frequency at Python Islands (Seehausen et al., 2008) and can be accomplished in the lab by housing females with heterospecific males. Fourteen F1 crosses (wild parents: 6 P. nye x P. nye; 4 P. pun x P. pun; 1 P. nye x P. pun; 3 P. pun x P. nye) and five F2 crosses (lab-bred parents: 1 P. nye x P. pun; 4 hybrid x hybrid) resulted in a test population of 58 males from 19 families (family details provided in Table S1). We included F2 fish due to low availability of F1 hybrids.

Pundamilia are maternal mouth brooders; fertilized eggs were removed from brooding females approximately six days after spawning (mean ± se: 6.3 ± 0.5 days post-fertilization; eggs hatch at about 5–6 dpf) and split evenly between light conditions. Upon reaching maturity, males displaying nuptial coloration were removed from family groups, PIT tagged (Passive Integrated Transponders, from Biomark, Idaho, USA, and Dorset Identification, Aalten, The Netherlands), and housed individually, separated by transparent, plastic dividers. All males were housed next to a randomly assigned male, with either one or two neighbour males (depending on location within the tank). Neighboring fish were the same for the duration of each sampling period (more details below). Fish were maintained at 25 ± 1 °C on a 12L: 12D light cycle and fed daily a mixture of commercial cichlid flakes and pellets and frozen food (artemia, krill, spirulina, black and red mosquito larvae). This study was conducted under the approval of the Institutional Animal Care and Use Committee of the University of Groningen (DEC 6205B; CCD 105002016464). The Tanzania Commission for Science and Technology (COSTECH) approved field permits for the collection of wild fish (2010-100-NA-2010-53 & 2013-253-NA-2014-177).

Experimental light conditions

Experimental light conditions were created to mimic the natural light environments of P. pundamilia and P. nyererei at Python Islands, Lake Victoria (described in greater detail: Maan et al., 2017; Wright et al., 2017). Species-specific light spectra were simulated in the laboratory (Fig. S1) by halogen light bulbs filtered with a green light filter (LEE #243; Lee Filters, Andover, UK). In the ‘shallow’ condition, mimicking P. pundamilia habitat, the spectrum was blue- supplemented with Paulmann 88090 compact fluorescent 15W bulbs. In the ‘deep condition’, mimicking P. nyererei habitat, short wavelength light was reduced by adding a yellow light filter (LEE #015). The light intensity differences between depth ranges in Lake Victoria are variable and can change rapidly depending on weather and sun angle (as much as 1,000-fold in sun vs. cloud cover); the mean (±se) light intensity in the deep environment (measured in 2010) was 34.15 ± 3.59% of that in the shallow environment (Fig. S1). Our experimental light conditions were designed to mimic in particular the spectral differences between depths and only partly recreated the intensity difference (the deep condition had a light intensity of ∼70% of that of the shallow condition).

Experiment 1: developmental colour plasticity

Males reared under experimental light conditions from birth were photographed repeatedly (three times each) in adulthood and assessed for body/fin coloration (details below). In total, we examined 29 pairs of brothers (mean age ± se at first sample: 689.9 ± 67 days; Pundamilia reach sexual maturity at ∼240 days), 29 from each light condition (2 × 10 P. pun, 2 × 9 P. nye, 2 × 10 hybrids, Table S1). Males were sampled from August–October 2016, with a mean (±se) of 13.25 ± 0.83 days between samples. Neighbour males (those housed next to test fish) were maintained for the duration of the sampling period.

Experiment 2: colour plasticity in adulthood

Following experiment 1, a subset of fish (Table S2) was switched to the opposing light condition (mean age ± se when switched: 643.47 ± 50.61 days; sexual maturity is ∼240 days) and colour tracked for 100 days. Each fish was photographed 11 times over the 100-day period: 1, 2, 3, 4, 7, 10, 14, 18, 46, 73, 100 days after switching. We switched 24 males, 12 from each light condition (2 × 4 P. pun, 2 × 4 P. nye, 2 × 4 hybrid). As a control, we also tracked 18 males (nine from each light condition: 2 × 3 P. pun, 2 × 3 P. nye, 2 × 3 hybrid) that remained in their original rearing light, but were moved to different aquaria (thus, both experimental and control fish had new ‘neighbour’ males). All fish, control and experimental, were photographed at the same 11 time points (in addition to the three photographs from experiment 1). The experiment was conducted in two rounds: October 2016–January 2017 (24 fish moved: six experimental & six control from each light condition) and December 2016–March 2017 (18 fish moved: six experimental & three control from each light condition).

Photography

All males were photographed under standardized conditions with a Nikon D5000 camera and a Nikon AF-S NIKKOR18-200 mm ED VR II lens. Fish were removed from their housing tank and transferred to a glass cuvette, placed within a 62.5 cm × 62.5 cm domed photography tent (Kaiser Light Tent Dome-Studio). This tent ensured equal illumination for all photos provided by an external flash (Nikon Speedlight SB-600) set outside of the tent. To ensure consistency of colour extracted from digital images (Stevens et al., 2007), all photos contained a grey and white standard attached to the front of the cuvette (Kodak colour separation guide), were taken with the same settings (ISO: 200; aperture: F9; exposure: 1/200; flash intensity: 1/8), and saved in RAW format.

Colour analysis

In Adobe Photoshop CS4, we adjusted the white balance and removed the background from each photo, keeping the entire fish (except the eye and pelvic fins). Each fish was then cropped into separate sections (body excluding fins, dorsal fin, caudal fin, anal fin, anal fin spots) and saved as individual images. Each section was analyzed for coloration using ImageJ (https://imagej.nih.gov/ij/), following the same procedure as detailed in Selz et al. (2016). We defined specific colours by their individual components of hue, saturation, and brightness to cover the entire hue range, resulting in a measure of the number of pixels that met the criteria for red, orange, yellow, green, blue, magenta, violet, and black for each section (colour parameter details provided in Table S3).

Brightness

We also measured the mean brightness of fish. Using Photoshop, we recorded the luminosity of ‘whole fish’ and ‘anal fin spot’ images, calculated from RGB values as: 0.3R + 0.59G + 0.11B (defined as brightness in: Bockstein, 1986). The weighting factors used by Photoshop (0.3, 0.59, 0.11) are based on human perception and should be similar to the trichromatic visual system of Pundamilia (Carleton et al., 2005). We measured the mean brightness of all fish used in experiment 1 and from three time points in experiment 2 (days 1, 10, 100).

Colour scores

Colour scores were defined as a percentage of coverage: the number of pixels in each colour category divided by the total number of pixels in the section. We used principal component analysis (PCA) on the correlation matrix of all eight colour scores to obtain composite variables of coloration (separate PCA was performed for each section—loading matrices in Table S4). In experiment 1, we examined PC1–PC4, as PC5 accounted for <10% of the variance in all analyses (mean cumulative variance = 82.5%; mean across all sections). For all analyses, we first assessed ‘whole fish’ images (minus eye and pelvic fins), followed by examination of each individual section (body, dorsal fin, caudal fin, anal fin, anal fin spots). Anal fin spots contained only red, orange, and yellow, thus PC’s were based on only those colour scores (and consequently, only PC1 & PC2 were used in analyses, 96.8% cumulative variance, Table S4).

In experiment 2, we first calculated baseline mean PC scores per fish using the repeated samples from experiment 1. At each time point after the switch, we then assessed deviation from the mean, calculated as: PC score—mean baseline PC score. Measuring the deviations from individual means allowed us to track the direction of colour change for each fish, independent of individual variation in baseline. Once again, PC scores were calculated for each body part independently and we used only PC1–PC4 (mean cumulative variance = 79.8%; loading matrices in Table S5).

Experiment 1: developmental colour plasticity

Using linear mixed modeling (lmer function in the lme4 package, Bates et al., 2013) in R (v3.3.2; R Development Core Team, 2016), we tested PC’s for the influence (and interactions) of: rearing light (shallow vs. deep), species (P. pun, P. nye, or hybrid), and body size (standard length, SL). Random effects included fish identity, parental identity, aquaria number, and position within aquaria to account for: (1) repeated sampling, (2) shared parentage among fish (Table S1), (3) location of aquaria within the housing facility, (4) number of neighboring males (1 or 2). The optimal random effect structure of models was determined by AIC comparison (Sakamoto, Ishiguro & Kitagawa, 1986) and the significance of fixed effect parameters was determined by likelihood ratio tests (LRT) via the drop1 function. Minimum adequate statistical models (MAM) were selected using statistical significance (Crawley, 2002; Nakagawa & Cuthill, 2007). We then used the KRmodcomp—pbkrtest package (Halekoh & Højsgaard, 2014) to test the MAM against a model lacking the significant parameter(s), which allowed us to obtain the estimated effect size of fixed effect parameters under the Kenward–Roger (KR) approximation (Kenward & Roger, 1997; Kenward & Roger, 2009). In the case of more than two categories per fixed effect parameter (i.e., species), we used post hoc Tukey (glht—multcomp package: Hothorn, Bretz & Westfall, 2008) to obtain parameter estimates.

Anal fin spot number

Following Albertson et al. (2014), the number of anal fin spots was counted as the sum of complete (1.0 each) and incomplete (0.5 each) spots for each fish (incomplete fin spots occur along the perimeter of the anal fin, often becoming complete with age/growth). Total spot number was compared among species, rearing light, and SL using the glmer.nb function in lme4 (Bates et al., 2013). Random effects were the same as above and reduction to MAM followed the same procedure. As KRmodcomp is unavailable for glmer.nb, final parameter estimates are reported from LRT via the drop1 function (Ripley et al., 2015).

Experiment 2: colour plasticity in adulthood

Using lme in package nlme (Pinheiro et al., 2014), we tracked fish coloration change over time, testing the influence (and interactions) of: species, treatment (rearing environment + ‘switched’ environment) and date (of sampling). We used lme because it allows specification of the optimal autocorrelation structure, as autocorrelation is common in longitudinal data (Crawley, 2007; Zuur et al., 2009). Random effects were the same as above, but with an additional random slope/random intercept term for date and fish identity (∼date|fish identity) to account for variability in the nature of colour change over time between individual fish. For simplification to MAM, models were fit with maximum likelihood (ML) and selected for statistical significance (Crawley, 2002; Nakagawa & Cuthill, 2007) by LRT using drop1. Final models were refit with restricted maximum likelihood (REML) and fixed effect parameters of MAM reported from the anova function. As above, we used post hoc Tukey (Hothorn, Bretz & Westfall, 2008) to obtain estimates for more than two categories per parameter.

Light-independent, interspecific differences

Coloration

To estimate the overall ‘colourfulness’ of fish, we calculated the sum of all measured colour scores for each male (whole body). Species did not differ in colourfulness (P = 0.29), nor did they differ in colours not defined by our colour parameters (calculated as: 100—sum of all measured colours; P = 0.29).

There was a significant difference between species (F_2,55.00 = 13.40, P < 0.001, Fig. 1A) in whole fish PC1 (positive loading yellow/orange). Tukey post hoc revealed that P. nyererei scored significantly lower than P. pundamilia (Z =  − 5.39, P < 0.001) and hybrids (Z =  − 3.76, P < 0.001). P. pundamilia was highest but did not differ significantly from hybrids (P = 0.47). There were tendencies for differences among species for whole fish PC3 (F_2,12.33 = 3.81, P = 0.051, Fig. 1C) and PC4 (F_2,55.00 = 2.49, P = 0.09, Fig. 1D). PC3 loaded positively with red/orange, with P. nyererei scoring highest and differing significantly from P. pundamilia (Z = 2.58, P = 0.026), but not quite so from hybrids (Z = 2.08, P = 0.09). PC4 had a strong, positive association with violet and followed the same general pattern as PC3 (P. nyererei highest). There were no significant differences for whole fish PC2 (P = 0.55; positive association with green/blue, Fig. 1B). Species differences for each body area separately are presented in Fig. S2. We saw a slight difference in mean brightness (F_2,55.00 = 2.5, P = 0.08): P. nyererei was lowest, differing somewhat from P. pundamilia (Z = 2.3, P = 0.053), while other comparisons were non-significant (P > 0.18).

Anal fin spots

Anal fin spot coloration did not differ among species (PC1: P = 0.25; PC2: P = 0.15) but the number of anal fin spots differed significantly (df = 2, LRT = 8.50, P = 0.014; Fig. 2). P. nyererei had significantly more spots than P. pundamilia (Z = 2.85, P = 0.017), while hybrids were intermediate and did not differ from either parental species (P > 0.18). A statistical trend indicated that anal fin spot brightness also varied between species (F_2,55.00 = 2.56, P = 0.08): P. nyererei had the brightest spots, differing slightly from hybrids (Z = 2.17, P = 0.07) but not from P. Pundamilia (P = 0.81). The total surface area (P = 0.10) or the size of the largest anal fin spot did not differ among species (P = 0.19).

Experiment 1: developmental colour plasticity

Increased green in deep light

We predicted that deep-reared fish would increase long-wavelength reflecting coloration, which would imply lower PC1 scores and higher PC3/PC4 scores. However, this was not the case (PC1 & PC3/PC4 scores did not differ between rearing environments, P > 0.59 for all). Instead, we found that, independent of species, deep-reared fish had significantly higher PC2 scores (F_1,40.07 = 9.08, P = 0.004, Fig. 3A), which could be attributed to body PC2 (F_1,40.12 = 4.89, P = 0.03, Fig. 3B) and, to a lesser extent, caudal fin PC2 (F_1,30.93 = 3.18, P = 0.083, Fig. 3C). The strongest positive PC2 loadings were with green/blue (body PC2 also loaded positively with red/magenta; caudal fin PC2 with red/violet). We also found a non-significant trend for deep-reared fish to have lower PC4 body scores (F_1,56.00 = 3.77, P = 0.057; PC4 loaded negatively for green/black), again indicating increased green colour in deep light. Separate analyses of each colour category confirmed this pattern; only green differed between rearing conditions (F_1,40.11 = 11.36, P = 0.001, Fig. 4A). This difference was species-independent, observed in P. pundamilia, P. nyererei, and hybrids (see Fig. S3). Increased green in deep-reared fish did not correspond to higher brightness (P = 0.43). For species-specific coloration in each light environment, see Fig. S4.

Experiment 2: colour plasticity in adulthood

Little effect of treatment

As seen in Fig. 5, fish that were moved between light conditions did not display consistent changes in coloration compared to baseline or to controls. For whole fish PC3 and for body PC4, we found significant three-way interactions between treatment, species, and date (F_6,408 = 2.33, P = 0.031 and F_6,408 = 3.34, P = 0.003) but treatment did not cause consistent changes in coloration (Figs. S5 and S6). Treatment had no effect on mean fish brightness (P = 0.41) or anal fin spot coloration/brightness (P > 0.26). We found a significant effect of ‘date’ in nearly all analyses (Table S6), indicating that both experimental and control fish continued to change colour over the 100-day sampling period. Thus, the lack of treatment effect was not due to fish colour being inflexible in adulthood.

Limited colour plasticity

P. pundamilia and P. nyererei are naturally depth segregated and occupy different light environments in Lake Victoria (Maan et al., 2006; Seehausen et al., 2008; Castillo Cajas et al., 2012). Given the close proximity of the two habitats (overlapping at some locations), selection might favour some level of flexibility in colour expression to cope with different signaling environments. Previous studies in other fish species have shown light-induced plasticity in coloration (Fuller & Travis, 2004; Lewandowski & Boughman, 2008; Hornsby et al., 2013). Contrary to our predictions, we found that deep-reared fish did not express more long-wavelength reflecting coloration. Instead, deep-reared fish were greener. Our light manipulations did not affect the male colours that most clearly differentiate the two species (blue/red). Additionally, males switched between light environments as adults showed little colour change. We propose two explanations for the lack of plasticity in nuptial coloration.

The first is that nuptial coloration in Pundamilia is under strong genetic control. Previous work has shown that species-specific coloration in Pundamilia is heritable (Magalhaes et al., 2009) and likely controlled by a small number of loci (Magalhaes & Seehausen, 2010). Common garden experiments by Magalhaes et al. (2009) found higher plasticity in morphological traits than in male colour scores but fish were reared under standard aquarium lighting (for light spectra comparison, see Fig. S1). We observed light-induced plasticity in green, which, unlike other male colours, is not subject to strong divergent selection by female choice (as demonstrated by: Selz et al., 2014) and perhaps less rigidly controlled (see below).

A second potential explanation for our results is that our light manipulations mimicked natural spectral variation, but only partially reproduced variation in light intensity. The difference in light intensity between the deep and shallow habitat in Lake Victoria is variable, but the deep habitat is consistently darker. While spectral differences have repeatedly been shown to correlate with numerous Pundamilia characteristics, light intensity may also play a role. Future studies could examine this by manipulating light intensity independent of spectral composition.

Finally, our sample size was modest (9–10 and 4 individuals per group in experiment 1 and 2, respectively). However, while increasing the sample size might increase statistical support for some of our results, we found relatively small effect sizes and these would likely not be affected. Thus, we expect that larger sample sizes would not change the main conclusions presented here.

Increased green in deep

Males reared in the deep light environment were significantly greener than shallow-reared fish. Our light manipulations differed primarily in short-wavelength availability and green wavelengths were abundant in both conditions (Fig. 4B). If plasticity in the species-specific male colours (blue/red) is limited, then increased green reflectance in darker conditions might be an alternative solution to increase visibility. To test this, we measured mean brightness of fish reared in both conditions. We found no difference in brightness between rearing environments, nor did brightness change when fish were switched in adulthood. This would suggest that differences in green colour do not contribute to increased visibility. However, these results are based on measurements of RGB values from photographs and may not properly capture contrast and perception in a specific light environment. Green covers a relatively small proportion of the fish (∼6% in deep light) and is not concentrated in a specific area of the body (unlike the red dorsum of P. nyererei, for example), making reflectance spectrometry difficult. Moreover, changes in green coloration coincided with non-significant changes in multiple other colours (see Fig. 4A), all of which may contribute to detectability.

Contrary to our prediction, we found that deep-reared fish expressed higher total amounts of shorter-wavelength colours, while longer-wavelength colours did not differ. These findings resemble those of Hornsby et al. (2013), who reared Nile tilapia in an environment lacking short-wavelength light and found higher expression of short-wavelength colours. The authors suggested that this response might be adaptive as it increases the contrast against the short-wavelength-poor background (Lythgoe, 1968; Hornsby et al., 2013). Possibly, this response represents a common strategy in cichlid fish.

Anal fin spots

Haplochromine cichlids possess carotenoid-dependent, yellow-orange, circular spots on their anal fins (Goldschmidt, 1991; Tobler, 2006). While the adaptive significance of these spots is debated (Maan & Sefc, 2013), previous studies have documented environment-contingent spot coloration in a number of species (Goldschmidt, 1991; Castillo Cajas et al., 2012; Theis et al., 2017). We examined the coloration, brightness, number and size of the anal fin spots and found that none of these measures were influenced by our light manipulations. However, we did find species differences: P. nyererei had the highest number of anal fin spots and the spots were generally brighter. Given that P. nyererei naturally occurs in the deep, short-wavelength poor habitat, this follows the general patterns presented by Goldschmidt (1991) and Theis et al. (2017); the exception being that P. nyererei in our study did not exhibit larger fin spots. The absence of colour differences in the anal fin spots of P. pundamilia and P. nyererei is consistent with earlier results based on reflectance spectrometry of wild fish (Castillo Cajas et al., 2012).

Implications for species isolation

P. pundamilia and P. nyererei differ in nuptial coloration and colour has been shown to co-vary with light conditions (Maan, Seehausen & Van Alphen, 2010; Castillo Cajas et al., 2012). Females display divergent preferences for conspecific male colour (Seehausen & Van Alphen, 1998; Haesler & Seehausen, 2005; Stelkens et al., 2008; Selz et al., 2014) and these preferences are key to species isolation (Selz et al., 2014). Differences in visual system characteristics (Carleton et al., 2005; Maan et al., 2006; Seehausen et al., 2008) correspond to differences in light environments, male coloration, and female preferences, suggesting a role for divergent sensory drive in speciation (Maan & Seehausen, 2010). Recently, we have shown that the same light manipulations that we used here significantly influenced female mate preference, potentially interfering with reproductive isolation (shallow-reared females preferred blue males, while deep-reared females favoured red males (Wright et al., 2017)). Plasticity in male colour expression could weaken the linkage disequilibrium between colour and preference even further. However, we find little evidence for such plasticity here, suggesting that blue and red are likely under strong genetic control. This may preserve reproductive isolation between populations inhabiting adjacent visual environments. In contrast, the plastic response in green coloration may aid in overall detectability of males, without interfering with species-assortative mating decisions that rely on interaction at closer range.