Divergent mating preferences and nuptial coloration in sibling species of cichlid fish
Sluijs, I. van der
Citation
Sluijs, I. van der. (2008, June 26). Divergent mating preferences and nuptial coloration in sibling species of cichlid fish. Department of Animal Ecology, Insitute of Biology Leiden (IBL), Leiden University. Retrieved from
https://hdl.handle.net/1887/12988
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/12988
CHAPTER 5
N� �������� ��� � ������� �����������
������� ������ ������ ���������� ���
���� ��������� ������ ����� �� � L���
V������� ������� ����
Inke van der Sluijs, Ole Seehausen, Tom J. M. Van Dooren &
Jacques J. M. van Alphen Manuscript
A�������
Sexual selection by female mating preference for male nuptial coloration has been suggested as driving force in the rapid speciation of Lake Victoria cichlid fish. This process could have been facilitated or accelerated by genetic associations between loci that determine female mate preferences and loci that determine the preferred traits in males. Female mating preference as well as male coloration are heritable traits, and probably determined by more than one gene. However, li�le is known about potential genetic associations between these traits. In turbid water, we found a population that is variable in male nuptial coloration, which ranges from blue to yellow to red. The males at the extreme ends of the phenotype distribution resemble a sibling species pair that inhabits clear water. In clear water, divergence took place into two reproductively isolated species; one with blue-grey coloured males: Pundamilia pundamilia; and the other with bright red males: Pundamilia nyererei. Pundamilia females are cryptically coloured. In a previous study, we found that females of the turbid water population are variable in mating preferences coinciding with the male phenotype distribution. For current study, these females were mated to pure blue males of a largely reproductively isolated population. Body coloration of the sires and male offspring was measured. Parents-offspring regression showed that coloration of male offspring was determined by the identity of the dams, which suggests that there is heritable variation in male colour genes between females. However, we found that mate preferences of the dams were not correlated with the colour of the male offspring. As expected, the identity of the sires did not affect the coloration of the male offspring, which confirms earlier findings that males of the blue species breed true. Variation in body coloration within the clutches was large; which also suggests that male coloration is determined by several genes. There is no strong physical linkage between mate preference and the sexual trait in this population.
82
I�����������
Sexual selection plays an important role in the process of speciation (Andersson 1994, Darwin 1871). Mating preferences, usually of females, can evolve through direct benefits or indirect benefits. Potential direct benefits for females include the access to resources (Moller & Jennions 2001, Saetre et al. 1995), whereas indirect benefits could entail ‘good genes’ selection for offspring with high fitness (Hamilton & Zuk 1982) or Fishers’ runaway selection generating selection for sexually a�ractive offspring (Fisher 1930, Lande 1981).
Theoretically, genes affecting male secondary sexual traits can become associated with genes affecting female mating preferences for the male trait through selective mating (Gavrilets et al. 2007, Higashi et al. 1999, Kirkpatrick 1982, Lande 1981, Payne & Krakauer 1997, Turner & Burrows 1995, van Doorn et al. 2004, Wu 1985). The process requires that mating preference and the preferred trait both have a heritable basis. Covariance between mating preference and the preferred trait can also be caused by pleiotropic effects of one gene or tight linkage between genes affecting the traits. Genetic correlations between the traits can facilitate speciation (Kronforst et al. 2006, Lande 1981) because it protects once evolved covariance against recombination due to random mating. For stable coexistence of diverging species, subsequent processes are o�en necessary, for example intersexual selection through mate choice and intrasexual selection through competition for mates (Lande et al.
2001, Mikami et al. 2004, Seehausen & Schluter 2004, van Doorn et al. 2004).
Although it has been shown theoretically that genetic correlations can arise between mating preferences and preferred traits, there are only a few empirical examples of this phenomenon. Correlated responses in male trait and female preference have been found in artificial selection experiments in the laboratory. Selection experiments on the red coloration of male sticklebacks gave a positively correlated response in female preference for this coloration, consistent with a genetic correlation between the traits (Bakker 1993). Likewise, selection on the length of the eye span, the secondary sexual trait of male stalk- eyed flies, gave a correlated response in female preference for eye span length (Wilkinson & Reillo 1994). Empirical evidence for population divergence facilitated by genetic associations between mating preference and secondary sexual trait, predicted by models, is scarce. One recent example is mating preference in two completely interfertile species of Heliconius bu�erflies which is genetically linked to the sexual trait (Kronforst et al. 2006). This, in combination with natural selection, may have lead to the divergence of two Heliconius species within the last 500,000 years (Kronforst et al. 2006).
The rapid speciation of haplochromine cichlid fish in African lakes, and in Lake Victoria in particular, has probably been driven by interactions of ecological and sexual selection (Albertson & Kocher 2006, Kocher 2004, Kornfield & Smith 2000, Seehausen 2006b, Seehausen et al. 1997).
The divergence of the cichlid species in Lake Victoria started less than 120,000 years ago (Genner et al. 2007) and possibly even occurred within the last 16,500 years (Beuning et al. 2002). This lake contains more than 500 haplochromine species (Genner et al. 2004). The divergence of Lake Malawi cichlids started between two and five million years ago in the lake and nowadays, the lake contains between 450 – 600 species (Genner et al. 2004, Genner et al. 2007).
Previous studies have suggested that sexual selection by female mate choice for male coloration played a role in the rapid speciation of haplochromines (Dominey 1984, Knight & Turner 2004, Maan et al. 2004, Seehausen & van Alphen 1998, Seehausen et al. 1997, Van Oppen et al. 1998). Males differ to a great extent in nuptial coloration (Genner & Turner 2005, Seehausen 1996). Females of many species that were tested mate actively assortatively, and their choice is o�en based on male nuptial coloration (Knight & Turner 2004, Seehausen
& van Alphen 1998, Stelkens et al. in press). Directional sexual selection by females on male nuptial coloration has been shown in two species of African cichlids, one from Lake Victoria (Maan et al. 2004) and one from Lake Malawi (Pauers et al. 2004).
Li�le is known about the genetic architecture of female mating preference and male nuptial coloration. Direct tests for physical linkage between female mating preferences and secondary sexual traits are complicated by sex- limited expression of the secondary sexual traits, in contrast with, for example, Heliconius bu�erflies (Kronforst et al. 2006). A genetic association between female preferences and male coloration could potentially contribute to rapid speciation in cichlid fish.
Species of the genus Pundamilia inhabit rocky shores and islands throughout Lake Victoria (Seehausen 1996). Seehausen (in press) described the occurrence of two sibling species of this genus in the southern part of the lake along a transect of increasing water turbidity. At Luanso Island at the southern end of the transect in the Mwanza gulf, where the water is very turbid, rocky shores are inhabited by a single phenotypically variable species in which male nuptial coloration ranges from blue to bright yellow to red, and most males have an intermediate phenotype (Seehausen 1997, van der Sluijs et al. 2008b).
Female mating preferences vary in this population, many females do not have any preference but some have significant preferences for blue males and very few have significant preferences for red males (van der Sluijs et al. 2008b). In contrast, the water is relatively clear at Makobe Island at the northern end of the transect (open lake), where two distinct sister species coexist and there are no intermediate phenotypes present (Dijkstra et al. 2007, Seehausen 1997). Here males of one species, Pundamilia pundamilia, are blue (Seehausen et al. 1998), and males of the other one, P. nyererei (Wi�e-Maas & Wi�e 1985), have a bright red dorsum and yellow flanks. Further, the species show significant neutral genetic and ecological differentiation (Seehausen in press).
84
Females of Makobe Island exert vision-mediated preferences for males of their own species (Seehausen 1997). At intermediate water transparency near Python Islands in the Mwanza Gulf, species are phenotypically strongly but genetically only weakly differentiated (Seehausen in press). Females also mate assortatively at this location (Seehausen & van Alphen 1998, Seehausen et al.
in review).
Females of both species are cryptically coloured and difficult to distinguish. Interspecific female mating preference is heritable, and probably determined by few genes (Haesler & Seehausen 2005). Interspecific differences in male nuptial coloration are also heritable (Seehausen 1997, Seehausen et al.
1997). F1 hybrid males between P. pundamilia and P. nyererei are intermediate in coloration and F2 hybrids segregate for colour (Seehausen in press, van der Sluijs et al. in press). This suggests that male coloration is not determined by a single gene, but not by very many either. These heterospecific crosses yielded fully viable and fertile hybrid offspring (van der Sluijs et al. 2008a). To test whether speciation in this system was aided by physical linkage between female mating preference and male nuptial coloration, we used females of the naturally variable population at Luanso Island.
M�������� ��� �������
Experimental procedure
Female preferences of the colour polymorphic population from Luanso Island was tested in a previous study (van der Sluijs et al. 2008b). Female mating preferences were tested for 30 females (see Figure 1 for preferences of the females used in current study). A�er determining their mating preferences, each female was allowed to mate with a blue (P. pundamilia) male from Python Islands. We reared the male offspring and measured their coloration to test the null hypothesis of no association between the mating preference of the dam and the colour of its male offspring. We used sires of one species only, to standardize the effect of sire on the coloration of the offspring. We chose the blue species as sire because previous observations suggested that blue coloration was recessive to red (OS, pers. obs.).
FishFish were collected from Luanso Island and Python Islands in February 2003 and August / September 2005 and shipped to the University of Leiden, the Netherlands. Females were allowed to mate with blue males which resulted in 21 clutches from 12 different females and seven different males.
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
1 2 3 4 5 6 7 8 9 10 11 12
Female
Preference
Figure 1 Mating preferences of 12 females of a population that is variable in male coloration. A positive score is a preference for males of the red species (black bars), and a negative score is a preference for males of the blue species (dashed bars).
The open bars are females without a preference, not significantly different from zero.
The order of the females is the same in all figures, thus in decreasing preference (from red to blue).
Photography
Male offspring was photographed a�er reaching sexual maturity, at the age of 1 to 1.5 years. Four tanks of 100 x 20 x 20 cm (l x w x h) each, were divided into five compartments by transparent Plexiglas dividers, perforated for water exchange. Males were placed in the compartments one week before the photographs were taken to allow them to become territorial and express nuptial coloration. A Kodak Colour strip (No. Q13, Eastman Kodak, Rochester, NY) was a�ached to the front glass to calibrate the photos in Photoshop 6.0 (Adobe Systems Inc.). In total, 187 male offspring and seven sires were photographed with a Sony digital camera (DSC-F707).
Colour analysis
Colour was measured from the calibrated digital photos of all males. Three different colour measurements were taken, two quantitative measures and one categorical. Yellow and red coloured pixels and the total body area in pixels, excluding fins and eyes, were measured with SigmaScan Pro 4.0 (Systat Inc.).
The hue windows were 0-26 plus 232-255 for red, and 27-45 for yellow.
86
The saturation window was 40-97% (Maan et al. 2004). Red and yellow coloration were expressed as the percentage red and yellow pixels of the total body area.
For the categorical measurement, males were assigned visually to one of five different colour classes, based on their total body coloration (van der Sluijs et al. 2008b), herea�er referred to as colour score. Colour score was used because it was not possible to measure the amount of blue coloration quantitatively, and males were assigned to a certain colour score based on the amount and the location of the colour. A male was assigned to colour score 0 when it was completely blue / grey without any yellow or red on the body.
When a male had yellow on the flanks but no red, it had colour score 1. When there was additionally some red present on the flanks along the upper lateral line, this was classified as colour score 2. Colour score 3 males had yellow flanks with some red on the dorsum and colour score 4 males had yellow flanks and a completely red dorsum.
We tested whether the three colour measurements were correlated within the individual male offspring with Spearman correlations (SPSS 15.0, Inc., Chicago, IL).
Data analysis
The general idea we used to structure our analysis is that the correlation between the genotypic value of one trait and the phenotypic value of another trait is equal to the genetic correlation. For traits determined in a polygenic manner and no sex-specificity, the expected genotypic value of an offspring is half that of the dam, plus half the genotypic value of the sire (Lynch & Walsh 1997). Since male nuptial coloration is not expressed in females, we can estimate genotypic values of dams only from the phenotypic data of their offspring. On the other hand, phenotypic data on both sires and male offspring can be used to estimate or predict genotypic values of sires.
To determine the genotypic effects of dam and sire on coloration of male offspring, we fi�ed linear mixed-effect models in R so�ware (version 2.4.0, Ihaka & Gentleman 1996) and used the lme4 library (version 0.9975-10, Bates
& Sarkar 2006) in order to fit the mixed models. Offspring male coloration was the explanatory variable, and dam and sire were random effect variables. The heritability of the traits is equal to the variance of the random effect divided by the total variance.
Since there is no so�ware available that estimates the genetic correlation between a normally distributed trait and a binomial trait, and in which several random effects can contribute to phenotypic variation at the same time, we tested the genetic correlation separately from fi�ing genetic and phenotypic variances. To determine whether the mate preference of the dams was linked to the coloration of their male offspring, we calculated Spearman rank correlations between the estimated preferences of the dams (female mating preference was
estimated with a binomial generalized linear model, see for methods and results van der Sluijs et al. 2008b) and the estimated coloration of the male offspring based on the minimum adequate mixed-effect model. We used Spearman rank correlations because the female preference parameters we used as preference phenotypes are not normally distributed.
We observed significant shrinkage in the predicted random effects for dams with small numbers of offspring. Since this affected the ranking of colour traits substantially, we also fi�ed linear models with least-squares fixed effects instead of random effects. Linkage between the estimated mating preference of the dams and the estimates of the fixed effects of the coloration of the male offspring was also tested by Spearman rank correlation tests.
If no significant effect of the sire was found on the coloration of the offspring, we used bivariate mixed-effect models using SAS proc nlmixed (Gueorguieva 2001), assuming conditional independence between the two traits (female mating preferences and male offspring coloration), given the random variation between offspring from different dams. That allows us to estimate the genetic correlation between female mating preferences and male offspring coloration along with the other parameters in a single fit. We used adaptive gaussian quadrature (AGQ) to approximate the likelihood of a model, with ten quadrature points. Since we can only fit a single random grouping level in this approach, there were no corrections made for variation in response between different mate choice trials.
We investigated separate models for each of the following traits:
percentage yellow, percentage red and colour score. Per trait, we combined data on parents and offspring in a single model, such that we used the maximum available data to estimate genotypic variances and predict or estimate genotypic values. In order to discount the genotypic value of a parent in the offspring correctly, we fi�ed (random) regressions without intercept and a weight covariate w. For example, in order to estimate the genotypic variance between sires in a mixed model, the covariate w of the random regression was equal to one for observations on a sire and one half in the offspring. By stratifying the random regression over sires, we could then estimate the genotypic variance between sires based on data from two generations. For the estimation of random colour variation between dams, only data on offspring could be used, which all received weight w = 0.5. When fi�ing fixed effect models, we fi�ed models with pairwise interactions between the covariate w and dam and sire effects. The maximum model contained dam and sire as groups and separate regressions for each effect. Conservative likelihood-ratio tests or F tests were used to assess the significance of random effects and to simplify the model if warranted. From the minimum adequate models selected, random effects were extracted or parameter estimates inspected. The bivariate models we fi�ed using SAS only included data on the offspring, since variation between sires was not considered there.
88
R������
Individual colour measurements
Within individuals, the percentage of red coloration was significantly positively correlated with the percentage of yellow coloration (Spearman correlation = 0.521, p < 0.001, n = 187). Further, the percentage of red coloration was also positively correlated with the categorical colour score (Spearman correlation = 0.261, p < 0.001, n = 187). In contrast, the percentage yellow was not correlated with the colour score (Spearman correlation = -0.003, p = 0.966, n = 187).
Parents – offspring regression
As expected, the percentage of yellow and red coloration, as well as the colour scores of male offspring did not differ significantly among sires. Dam’s identity however, had a significant effect on all three colour measurements of the offspring in the mixed models (see for test values and significance Table 1), as well as in the linear models (Table 2). In the mixed model, dam’s identity explained 26% of the variation in yellow coloration of the male offspring.
Likewise, dams explained 37% of the variation in red coloration of the male offspring. Finally, the colour score was significantly affected by dam’s identity too, explaining 30% of the variation among male offspring.
Offspring coloration, expressed in percentage of yellow coloration, red coloration, and colour score is plo�ed per dam in Figure 2-4 respectively. The dams are ranked along the x-axis by decreasing mating preference, thus from red-preferring on the le� to blue-preferring females on the right.
Table 1 Results of parents-offspring regression models for three measures of male coloration (percentage red, yellow, and colour score) with random effects of dams and sires. χ2 test values, degrees of freedom (df) and p-values are given for the random effects of the dams and sires. Spearman correlations were calculated between estimates of the offspring coloration and estimates of the mate preference of the dams.
Colour measure Sire Dam Significant Spearman Rho p male offspring X2 df p X2 df p effects in correlation
model S
Yellow (%) 1.050 1 0.306 3.060 1 0.080 Dam 262 0.084 0.792 Red (%) 0 1 1 9.687 1 0.002 Dam 290 -0.014 0.974 Score 0 1 1 4.821 1 0.028 Dam 270 0.006 0.860
Correlation preference and coloration
None of the genotypic colour values was correlated with the estimated preference of the dam (Spearman correlations in Table 1 and 2, bivariate mixed- effect models in SAS in Table 3). In the bivariate mixed-effect models, dam’s identity explained 21% of the variation in yellow coloration, 8% of the variation in red coloration, and 9% of the variation in colour score.
Table 2 Results of parents-offspring regression models for three measures of male coloration (percentage red, yellow, and colour score) with fixed effects of dams and sires. F-test values, degrees of freedom (df) and p-values are given for the fixed effects of the dams and sires. Spearman correlations were calculated between estimates of the offspring coloration and mate preference of the dams.
Colour measure Sire Dam Significant Spearman Rho p male offspring F df p F df p effects in correlation
model S
Yellow (%) 1.711 6 0.121 1.836 11 0.051 Dam 250 0.126 0.692 Red (%) 0.899 6 0.497 2.432 11 0.007 Dam 262 0.084 0.792 Score 0.947 5 0.452 2.546 10 0.007 Dam 284 0.007 0.983
1 2 3 4 5 6 7 8 9 10 11 12
0102030405060
Female
Yellow coloration (%)
Figure 2 Boxplot of yellow coloration (%) of the male offspring of the 12 different dams.
90
1 2 3 4 5 6 7 8 9 10 11 12
051015
Female
Red coloration (%)
Figure 3 Boxplot of red coloration (%) of the male offspring of the 12 dams.
1 2 3 4 5 6 7 8 9 10 11 12
0.00.51.01.52.02.53.0
Female
Colour score
Figure 4 Boxplot of the colour scores in categories of the male offspring of the 12 dams. Male colour score categories range from 0 (= blue) to 4 (= red).
Table 3 Results of parents-offspring regression bivariate mixed-effect models for male coloration and the heritabilities of male coloration.
Colour measure X2 p Genetic variance Environmental Heritability
male offspring colour variance colour
Yellow (%) 0.1 0.75 42.6 163.6 0.21
Red (%) 1.6 0.21 0.27 3.10 0.08
Score 0.8 0.37 0.04 0.43 0.09
D���������
Physical linkage between genes affecting mating preferences and sexual traits can enhance the likelihood of speciation (Kronforst et al. 2006, Lande 1981) because it protects once evolved covariance against recombination due to random mating. We studied a population of Lake Victoria cichlid fish that contains considerable variation in female mating preferences and in male nuptial coloration, the extremes of which resemble differences between sympatric species, with red and blue males, at other sites (Seehausen 1997, Seehausen in press, van der Sluijs et al. 2008b). Females with known preference or known absence thereof were mated with blue males from a closely related species.
Body coloration of male offspring was analysed for potential correlations with the colour preference of its dams. As expected, the identity of the sire had no effect on male offspring coloration. This is consistent with previous data that suggest that the blue species that we used breeds true (Seehausen 1997, Seehausen et al. 1997). We found that dam’s identity significantly affected the coloration of the offspring, suggesting heritable variation in male colour genes between females. However, there was no detectable association with mating preference of the dams for either blue or red males.
Thus, we found no evidence for any genetic correlation between female mating preference and male nuptial coloration. However, this does not rule out the existence of weak linkage. Haesler & Seehausen (2005) have shown that female mating preferences for red versus blue males in this system are likely codetermined by several genes. The variation in body coloration within our clutches was large, which suggests that this trait is determined by several genes too. When female mating preference and male coloration are both determined by several genes it may be difficult to detect potential linkage of some of these genes. However, our data do rule out strong physical linkage between mating preference and the preferred trait that would be protected against recombination and aid speciation as have been observed in Heliconius bu�erflies (Kronforst et al. 2006). The implication is that the sympatric build-up or maintenance of preference / trait covariance in the Pundamilia cichlid system requires strong divergent selection.
92
Speciation by sexual selection models predict that genes affecting mating preferences and genes affecting sexual traits can become associated relatively easily (Higashi et al. 1999, Kirkpatrick 1982, Lande 1981, Payne & Krakauer 1997, Turner & Burrows 1995, Wu 1985). Positive correlations between mating preferences and sexual traits have been found in artificial selection experiments in guppies, sticklebacks, and stalk-eyed flies (Bakker 1993, Houde 1994, Houde
& Endler 1990, Wilkinson & Reillo 1994). Genetic covariance between female and male components of mate recognition can coevolve rapidly under sexual selection in Drosophila serrata and D. birchii hybrids (Blows 1999). In these studies, speciation by sexual selection through mate choice has been suggested but it is unknown whether speciation was facilitated by genetic linkage of the traits. The divergence of two species of Heliconius bu�erflies was suggested to be facilitated by physical linkage between mate preference and the preferred trait (Kronforst et al. 2006).
We do not know for certain whether the population at Luanso Island was once differentiated into a blue and a red species that coalesced due to recently increased rates of hybridization or, alternatively, if this population had never been differentiated. Either way, our results imply that random mating and recombination in this turbid water site have been sufficient to prevent any build-up of, or remove, previously existing genetic correlations between female mating preference and male nuptial coloration. The absence of detectable association also implies that the persistence of the variation in preference and trait in this population is not explained simply by immigration of individuals from clear water sites. In other words, the rare red-preferring females at Luanso are not migrants from red populations further north in the Mwanza Gulf.
A���������������
We thank the Tanzanian Fisheries Research Institute for hosting and use of the facilities (Prof. Philip Bwathondi; Egid Katunzi) and the Tanzania Commission for Science and Technology for the research permits. Mhoja Kayeba and Mohamed Haluna provided assistance in the field. Peter Dijkstra, Kees Ho�er, Charlo�e Lindeyer, Martine Maan, John Mrosso, and Machteld Verzijden are greatly acknowledged for collaboration in the field and discussions. Martine Maan is also acknowledged for comments on an earlier version of this manuscript. This research was supported by the Netherlands Science Foundation (NWO-ALW 810.64.011), and research grants from Leiden University Fund, and the Schure- Beijerinck-Popping Foundation. This study was approved by the Animal Ethics Screening Commi�ee (UDEC 03061).
