Limited effects of inbreeding on breeding coloration

Abstract

Animal color signals may function as indicators of fighting ability when males compete for access to females. This allows opponents to settle aggressive interactions before they escalate into physical combat and injury. Thus, there may be strong directional selection on these traits, toward enhanced signal quality. This renders sexually selected traits particularly susceptible to inbreeding depression, due to relatively low ratios of additive genetic variance to dominance variance. We measured the effects of inbreeding on an intrasexually selected color signal (the badge) in a population of Swedish sand lizards (Lacerta agilis) using the Rhh software based on 17 to 21 microsatellites. Males of this sexually dichromatic species use the badge during aggressive interactions to display, and assess, fighting ability. We found negative effects of homozygosity on badge size, saturation, and brightness. However, no such effects were observed on color hue. Pairwise correlations between badge size, hue, and saturation were all statistically significant. Thus, the sand lizard “badge” is a multicomponent signal with variation explained by covariation in badge size, saturation, and color hue. Body mass corrected for skeletal size (body condition) positively predicted badge size and saturation, encouraging future research on the extent that sexual signals may convey information on multigene targets (i.e. “genic capture”).

badge, color, display, inbreeding, sand lizards

Introduction

Strongly fitness-related traits, such as life-history traits, are expected to be more affected by inbreeding depression than those less tightly linked to fitness, such as many morphological traits (Falconer 1981; Roff 1998; DeRose and Roff 1999). This is because life-history traits have been under strong directional selection that decreases their additive genetic variance relative to dominance variance (Roff 1997; Merilä and Sheldon 1999). Inbreeding depression is caused by an increase in homozygosity at a number of fitness-related loci through either overdominance or partial dominance (Charlesworth and Charlesworth 1987). Thus, inbreeding is expected to impact the expression of polygenic fitness traits that represent relatively broad targets for mutations to act upon (i.e. “genic capture”; Houle et al. 1996; Merilä and Sheldon 1999). These traits should therefore reflect genetic quality involving a number of loci across the genome, with inbred individuals displaying inferior trait values relative to more outbred individuals.

Access to females and reproductive opportunities are often considered the ultimate mechanisms driving the evolution of secondary sexual traits (Darwin 1871). Sexual selection theory posits that these traits can evolve via 2 principally different processes. Firstly, male–male competition resulting in intrasexual selection may be associated with the evolution of male armaments, such as weapons (Andersson 1994), or color badges (Olsson 1992a; Andersson 1994) that may be used to signal fighting ability (Barrette and Vandal 1990; Olsson 1994a; Stapley and Whiting 2006). Secondly, female choice, or intersexual selection, may reinforce the evolution of male ornaments (Berglund et al. 1996) via Fisherian runaway processes (Fisher 1930; Andersson 1994) or indicator mechanisms (Fisher 1915; Zahavi 1975; Andersson 1994; Morris et al. 2007). Irrespective of the evolutionary sequence of events, variation in the magnitude of sexually selected traits may affect individual relative fitness (Hamilton and Zuk 1982; Zuk et al. 1990; Olsson 1994a; Vega-Trejo et al. 2017).

During the last few decades, a large body of literature depicts the effects of inbreeding on traits that are directly associated with fitness, such as survival and reproductive output (Falconer 1981; Keller and Waller 2002), while studies on inbreeding effects on secondary sexual traits are less common, in spite of clear evidence of their effects on reproductive success (Hamilton and Zuk 1982; Zuk et al. 1990; Olsson 1994a; Vega-Trejo et al. 2017). Studies of inbreeding effects on sexually selected traits in guppies (Poecilia reticulata) established compromised male coloration (Van Oosterhout et al. 2003; Zajitschek and Brooks 2010), mating behavior (Van Oosterhout et al. 2003; Mariette et al. 2006), and male attractiveness (Zajitschek and Brooks 2010). In mosquitofish (Gambusia holbrooki), inbred males experienced lower reproductive success relative to more outbred males when competing for females (Marsh et al. 2017; Vega-Trejo et al. 2017). Inbreeding in song sparrows (Melospiza melodia) reduces song repertoire size (Reid et al. 2005; Reid 2007) and in zebra finches (Taeniopygia guttata) it depresses song rate, beak color, and attractiveness (Bolund et al. 2010). However, it is not always evident if inbreeding effects on sexually selected traits act directly on genes associated with the traits themselves or are mediated through condition (Rowe and Houle 1996; Van Oosterhout et al. 2003).

In the present study, we assess the effects of inbreeding on the lateral green coloration—referred to as the “badge” in our previous work (Fig. 1)—in a Swedish population of sand lizards (Lacerta agilis). These badges have been verified to be an intrasexually selected trait that reflects fighting ability (Olsson 1994a). The badge is developed immediately after hibernation in late April–early May, which is followed by the mating season involving male–male interactions that can lead to severe injuries (Olsson 1992b,1993). Toward the end of the mating season (in July–August), the badge fades gradually, coinciding with a decrease in aggression (Olsson 1994a). Larger males tended to win contests more often than smaller ones (Olsson 1992b) and males with artificially enlarged badges initiated and won contests more often (Olsson 1994a). Both the area of the green badge and a visual estimate of color saturation (Munsell color score) correlated with body condition and body mass in this population (Olsson 1994a). Thus, badge area along with pigmentary or structural components of color that contribute to this estimate of saturation may reflect important aspects of fighting ability (Olsson 1994a). Interestingly, at relatively equal badge size, contest duration was significantly prolonged (Olsson 1994a). These findings suggest that the area and saturation of the badge allow male sand lizards to assess a rival’s fighting ability, thus affecting fighting behavior during aggressive interactions. Ultimately, fighting decisions based on relative badge size may allow males to avoid costly interactions (Olsson 1992b, 1993) and be under intrasexual selection. In another experiment, badge-enlarged males experienced an increase in mate acquisition (Anderholm et al. 2004), thus highlighting the potential fitness returns of having a relatively larger badge.

Fig. 1.

The badge is located on the lateral area of a male’s body, between the front and the hind leg. The images show 3 males with badges that differ in size, from small to large (top to bottom image; photography courtesy of Willow Lindsay).

Given the described relationships between badge area and pigment saturation as important predictors of components of fitness (Olsson 1994a), we analyzed covariation between these traits and individual-level heterozygosity. Because lizard coloration should be comprised of pigmentary and structural components that may signal different aspects of individual quality (Olsson et al. 2013), we considered color metrics that may indicate variation in potential pigments, such as melanin and carotenoids (measured in badge size, saturation and hue), and structural components captured in “brightness.” We then investigated skeletally controlled body mass (body condition) as a predictor of each color metric trait separately, and applied principal component analysis and pairwise phenotypic correlations to determine the multicomponent nature of the badge and signal expression. We used long-term field data (1998 to 2008) with many males captured in more than 1 field season. Specifically, we asked whether our estimates of color traits were phenotypically correlated with our genetic estimate of inbreeding (using microsatellites, see below). If so, color may convey “genetic quality.”

Methods

Morphological measurements and molecular genetics

Details of the protocols used for field (Olsson 1994a; Olsson et al. 1996) and molecular genetics methods (Olsson et al. 2011b) have been described elsewhere. Here, we present a brief account of these protocols. Male sand lizards were captured by a loop of fishing line or by hand at Asketunnan, 50 km South of Gothenburg, Sweden. The captures took place every spring between 1998 and 2008. The individuals were weighed to the nearest 0.01 g and measured snout to vent (SVL, snout-vent length) to the nearest 1.0 mm. They were then individually marked by claw/toe-clipping (at the claw-base), and by using a numbered cloth tape tag on the back (for short-term marking until next skin shedding). DNA was extracted from blood samples (≈50 µL obtained from vena angularis, in the corner of the mouth, and stored in 70% alcohol). Once measured for all variables, males were released at the place of capture with a number of individuals being recaptured and measured in subsequent years. Note that a number of recaptured males had also hatched in captivity, allowing us to calculate their age (see Statistical methods). The DNA was then used for genotyping at 17 to 21 microsatellite loci (Olsson et al. 2011b; for a summary table of microsatellite data and their information content, see electronic supplement). Standardized heterozygosity (SH) (Coltman et al. 1999) was calculated as a proxy for genome-wide heterozygosity, with an average of 19.6 microsatellite loci used per individual using the Rhh software (Alho et al. 2010). Body condition was estimated for all males, represented as a skeletal size-corrected mass estimate (the residual score from a regression model using SVL as predictor and the cube root of mass as a response variable).

Coloration

We photographed males using a flatbed scanner and calculated badge size as the proportion of green coloration vs. the remaining side of the body (Olsson and Madsen 2001). Although this method is cruder than spectrometry analysis, staying with this method from before spectrometry were in general use allowed us to use 1 consistent method throughout the entire decade-long study period. We then imported 1 high-resolution image per individual and year into Photoshop and used the Auto Color Tool to standardize lighting between images. We extracted average hue (degrees), saturation (percent), and brightness (percent) within the male badge using the eyedrop color sampler and standardized the location for color capture between individuals by placing the sampler ca. 150 pixels posterior and 50 pixels dorsal to the lowermost notch of the forelimb. Because sand lizard badge color is patchy, with spots of darker and lighter green pigmentation, we collected average color within a 101 by 101 pixel sample window.

Statistical methods

Statistical tests were performed using Proc Mixed, Proc Corr, and Proc Princomp in SAS (Statistical Analysis Systems) 9.4.

The expression of the badge in males is age dependent (Olsson 1994a, 1994b; Fig. 1). Therefore, age had to be controlled for in the analysis of heterozygosity effects (SH) on badge size and other color metrics. However, we did not have the exact hatching date for a number of males, since they were captured as adults in the wild. To avoid a substantial decrease in sample size, we therefore used SVL as a proxy for age (Olsson and Shine 1996; Fig. 2).

Fig. 2.

Relationship between age in years and SVL in mm. The dots represent individual observations and the line the mean value per age category.

The effect of SH was tested separately for badge size, saturation, hue, and brightness, using a GLMM approach with Gaussian error distributions and individual ID as a random effect, since many males were resampled in several years. The fit of each model to the data was assessed by examining its −2 log likelihood value before and after excluding the variable of interest. Model fit was compared between the 2 models using a log likelihood ratio test (LL ratio test) with 1 degree of freedom (Wang et al. 2011). The marginal R² of each full or reduced model was calculated to estimate the proportion of variance in the response variable explained by either the joint effect of all predictors or by SH alone (Nakagawa et al. 2017). The above models allow us to estimate the effects of SH on the badge while controlling for body condition. The full model included individual ID as a random effect and year, sampling day, SVL, and body condition as fixed effects.

Pairwise correlations between badge size, hue, saturation, and brightness were assessed using Spearman’s rank-order correlations, since data could not be transformed to normality. One observation was randomly selected for each male, for a total of 292 observations. We utilized a principal components analysis in addition to pairwise correlations in order to assess how color metric components of badge (badge size, hue, saturation, and brightness) were segregated.

Sample sizes are model specific, depending on maximum available trait information and are provided for each analysis in the result section and in the electronic supplement.

Results

Relationship between age and SVL

Age had a significant effect on SVL, with older males having longer snout to vent length (n = 526; mean = 58.74; SD = 20.10; F_1,456 = 1518.92, P < 0.0001; estimate = 8.56; 95% confidence limits 8.13 to 8.99; Fig. 2). The marginal R² for the model was 0.75. SVL increased with age, but growth rate decreased in older age classes, consistent with the observations made by Olsson and Shine (1996).

Effect of individual ID on badge size and color traits

Badge size, hue, and saturation differed significantly among males (size: n_males = 394; n_obs = 702; estimate = 0.0060; SE = 0.00123; Z = 4.90; P < 0.0001. Saturation: n_males = 221; n_obs= 369 estimate = 10.290; SE = 3.165; Z = 3.25; P < 0.0006. Hue: n_males = 221; n_obs = 369; estimate = 38.216; SE = 11.3392; Z = 3.37; P < 0.0004). However, male ID only showed borderline significance for badge brightness (n_males = 221; n_obs= 369; estimate = 3.532; SE = 2.2604; Z = 1.56; P = 0.0591).

Effect of individual-level heterozygosity (SH) on badge size and color

Level of heterozygosity (SH) significantly and positively predicted badge size, saturation, and brightness, but not hue (Table 1). Sampling year and sampling day (within year) both predicted all color metrics, except for an effect of day number on brightness (P = 0.164). SVL had a significant effect on badge size, saturation and brightness but not on hue (Table 1). Finally, body condition only significantly affected badge size. The results of the analyses, including the marginal R² values, are presented in Table 1. Descriptive plots of the relationships between SH and color metrics can be found in Fig. 3a–d.

Table 1.

For each badge color metric, the table shows the sample size (n), the mean value, the SD, the estimate of the effect of standardized heterozygosity (SH) and its standard error (SE), the R² value for the model with SH as only predictor and for the model including all fixed effects, and the results of the LL ratio test (the value of the chi-square statistic and its P value).

	Badge size		Saturation		Hue		Brightness
N	702		369		369		369
Mean	0.580		38.873		108.90		45.407
SD	0.193		10.231		18.204		5.419
SH: estimate (SE)	0.063 (0.028)		3.20 (1.732)		−0.210 (3.118)		2.682 (1.340)
R  2 (SH)	0.00038		0.0080		≈0		0.013
R  2 (all fixed effects)	0.529		0.587		0.594		0.206
LL ratio test: ${\chi }_1^2$ (P value)	4.8 (0.014)		3.4 (0.032)		0 (0.5)		3.8 (0.025)
Year	F  10,616 = 13.60	P < 0.0001	F  5,334 = 19.50	P < 0.0001	F  5,318 = 42.35	P < 0.0001	F  5,338 = 4.09	P < 0.0013
Day	F  1,669 = 60.13	P < 0.0001	F  1,351 = 10.31	P =  0.0014	F  1,333 = 26.64	P < 0.0001	F  1,366 = 1.94	P = 0.1641
SVL: estimate (SE)	0.020 (0.0009)	P < 0.0001	0.460 (0.068)	P < 0.0001	−0.210 (0.120)	P = 0.0816	−0.055 (0.0546)	P = 0.306
Condition: estimate (SE)	0.325 (0.083)	P = 0.0001	7.862 (6.344)	P = 0.2161	−15.462 (11.133)	P = 0.1657	−0.184 (5.171)	P = 0.9715

Also, the results for all fixed effects, including year and date of sampling, SVL and body condition, are presented. The statistically significant P values are bolded.

Fig. 3.

Descriptive plots of the relationship between standardized heterozygosity (SH) and a) badge size (%), b) badge saturation, c) badge hue, and d) badge brightness. The number of observations is the same as in the analyses (702 for badge size and 369 for the rest).

Relationship between SH and body condition

Body condition differed significantly between years (n_males = 440, n_obs = 805 full model; F_10,714 = 13.10, P < 0.0001) and sampling days (F_1,772 = 11.72; P < 0.0001). Males differed significantly in body condition (n = 440; estimate = 0.001260; SE = 0.0002; Z = 6.17; P < 0.0001). SH had a positive and significant effect on body condition but with a trivial effect size (n_males = 440, n_obs = 805 full model; estimate = 0.283; SE 0.012 = LL ratio test: ${\chi }_1^2=5.3$⁠, P = 0.010; marginal R² = 0.0038).

Marginal R² for the effect of body condition on badge size and color

Body condition explained 0.94% of the variance in badge size, 6.43% of the variance in hue, 6.55% of the variance in saturation, and 0.078% of the variance in brightness.

Phenotypic correlation between badge size and color properties

The pairwise correlations between badge size, hue, and saturation were statistically significant, with Spearman correlation coefficients ranging between 0.13 and 0.38 (Table 2). Badge brightness did not significantly correlate with badge size, hue, or saturation (Table 2). The principal component analyses produced 2 components with eigenvalues above 1 (see electronic supplement). Collectively, these 2 components accounted for 67.3% of the variance in badge size, hue saturation, and brightness (41.6% for PC1 and 25.7% for PC2). Badge size, hue, and saturation showed relatively high loadings on PC1, whereas badge brightness had a relatively high loading on PC2 (for principal component analysis results see the electronic supplement).

Table 2.

The table shows the results of the pairwise correlations between badge color metrics.

	Size	Brightness	Hue	Saturation
Size	1.00000	0.06816	0.12809	0.37709
		0.2456	0.0286	<0.0001
Brightness		1.00000	−0.03895	−0.04595
			0.5073	0.4340
Hue			1.00000	0.38375
				<0.0001
Saturation				1.00000

Each cell includes the Spearman correlation coefficient (the top value) and the P value for the correlation (the bottom value). The statistically significant P values are bolded.

Discussion

Strong directional selection on sexually selected signals should erode trait variance and result in a loss of underlying genetic variance (Van Homrigh et al. 2007). However, variation in sexually selected trait and underlying genetic variance often remain high (Pomiankowski and Møller 1995). Several hypotheses have been proposed to resolve this “lek paradox” (Kirkpatrick and Ryan 1991; Pomiankowski and Møller 1995; Rowe and Houle 1996). Empirical support from wild populations lags behind theoretical explanations and the relative importance of environmental vs. genetic effects in determining sexually selected trait variance remains unclear (Qvarnström and Price 2001; Martinossi-Allibert et al. 2019; Howie et al. 2019). In our analyses, we assess to what extent a sexually selected trait—the “badge” and its color metrics—is affected by loss/maintenance of individual-level standardized heterozygosity (SH). Importantly, level of heterozygosity had a positive effect on badge size, saturation, and brightness, with more inbred individuals having lower trait values than relatively outbred ones. However, the explained variance by the models was relatively low for all badge color metrics. SH only explained a small fraction of the variance in badge size and saturation, 0.038% and 0.8%, respectively, while it explained 1.3% of the variance in brightness. In previous research involving the same population, we found a relatively strong effect of heterozygosity on hatching success (a life-history trait), with an increase in hatching odds of ca. 30% for an increase of 10% in SH (Bererhi et al. 2019), for eggs hatched under controlled laboratory conditions. Similarly, Olsson et al. (1996) observed that matings between siblings resulted in 18.5% increase in the probability of malformation in newly hatched juveniles. Thus, strong episodes of selection during early ontogeny may attenuate the effects of inbreeding during later life stages (Keller et al. 1994; Hemmings et al. 2012). Thus, a parsimonious explanation for a significant, but weak, effect of inbreeding on badge size and color, could be relatively weak directional selection that results in a low ratio of dominance to additive genetic variance (Roff 1997; Merilä and Sheldon 1999).

Individual-level heterozygosity may act directly on genes that are involved in trait expression, or indirectly through condition (Rowe and Houle 1996; Van Oosterhout et al. 2003). The latter seems less plausible in our study population, as SH only explained 0.8% of the variance in body condition (with, however, a significant P value of 0.030). This suggests that direct or epistatic (Lynch and Walsh 1998) genetic effects of heterozygosity is a more likely explanation for the limited SH-dependent variance in badge components. Based on our findings, we cannot exclude condition as a mediator of the effects of heterozygosity on badge expression.

It seems warranted to ask “Why are inbreeding effects not stronger on color traits”? A full discussion of the validity of multilocus microsatellite markers from a perspective of inbreeding analysis is outside the scope of this study, but suffice to say that results have varied across taxa and that power to detect heterozygosity-fitness correlations is low when 10 or fewer markers are typed (Slate and Pemberton 2002). Other work has shown that sample sizes in excess of 600 individuals are necessary to detect reasonable effect sizes with maintained power (Coltman and Slate 2003). Our work used 17 to 21 microsatellites, depending on analysis. Even if we appreciate the higher resolution offered in large single nucleotide polymorphism analysis or whole-genome sequencing (e.g. Howard et al. 2017; Baes et al. 2019), 21 microsatellites are still a substantial set of markers in a natural population and our sample sizes are in the upper tail end of published field studies with this level of resolution. Our microsatellites were initially selected for paternity analysis and therefore chosen based on their above average variability (electronic supplement). Thus, from this perspective inbreeding could even be underestimated in our work (with markers being more variable than the true genome-wide average).

In our study population, only larger/older males seem to be under significant sexual selection for badge size (Olsson 1994a), i.e. males that have completed somatic growth. Younger/smaller males instead invest more into growth and less into producing integumental color traits, as illustrated by the significant effect of SVL on badge size and saturation (Olsson 1994a, 1994b). Furthermore, our experimental enlargement of badges in small males resulted in a 4-fold increase in female mate acquisition (with badges that proportionally matched badges of the largest males; Anderholm et al. 2004). Importantly, alternative strategies could result in weak selection on badge size and/or color in younger males. This implies that the reproductive success of younger males might be affected by other strategies than resource-holding displays, such as sneaking. Finally, the number of loci underlying the expression of badge size and color may limit the magnitude of the effects of inbreeding (Houle et al. 1996; Merilä and Sheldon 1999). Thus, knowledge of the genetic architecture of these traits, combined with estimates of additive genetic variance and more detailed age-specific measures of selection, may improve our understanding of the impact of inbreeding on the expression of the badge in the Asketunnan population.

Association between brightness and fitness is yet to be experimentally investigated in our sand lizard study population. Depending on habitat-specific light conditions, a bright color signal should increase contrast and detectability (Leal and Fleishman 2004; Stuart-Fox et al. 2007). The detrimental effects of a decrease in brightness could be accentuated if it concomitantly causes a decrease in detectability involving another reflectance peak than green color, such as ultraviolet (UV) wave lengths. The sand lizard badge reflects UV (De Lanuza and Font 2007; Olsson et al. 2011a), which may function as a signal amplifier (Hasson 1989; Lappin et al. 2006), and affects mate acquisition in experiments involving a treatment with a UV blocker in the wild (Olsson et al. 2011a).

Sampling year, and days within years, significantly affected the development of badge color at the beginning of the mating season (no data are included in the analyses during the time of year that badges fade). Individual ID affected all badge characteristics but brightness. Thus, genetic differences among males seem less likely to explain differences in brightness than the fixed effects included in our models (year, day, SVL, and condition).

To conclude, we have shown that the sand lizard badge reveals information on an individual’s level of inbreeding, using microsatellite-estimated heterozygosity, although effect sizes for these estimates are low. The correlations between badge size, hue, and saturation may indicate their joint evolution as part of a multicomponent trait, while brightness seem to make up a separate entity (and perhaps convey different underlying information). Future research would benefit from contrasting genetic variation in the genes regulating developmental pathways of integumental coloration and to what extent variance in those genes are more or less upheld than in the genome overall.

Funding

This long-term study was supported with a series of grants from the Swedish Science Council (VR) and the Australian Research Council. The latest grant in this series was 2015-04835.

Acknowledgments

We thank the numerous students and field workers for invaluable contributions over the many years of the study, and the comments by an anonymous reviewer, which improved the quality of the manuscript.

Data availability

All data for this MS will be made available on Dryad upon acceptance for publication. The paper has now been updated with the correct DOI, https://doi.org/10.5061/dryad.dncjsxm4w.