GONZALO MACHADO-SCHIAFFINO, ANDREAS F. KAUTT*, JULIAN TORRES-DOWDALL*, LUKAS BAUMGARTEN, FREDERICO HENNING & AXEL MEYER
(*equal contribution)
Manuscript in preparation for submission
Summary
Sympatric speciation has been in the center of debate in evolutionary biology for decades and it is now generally accepted that when a trait under disruptive selection is also involved in assortative mating, the so called ‘magic traits’, divergence with gene flow is more likely to occur. Thus, it is critical to assess the role of disruptive selection and assortative mating during the early stages of divergence in order to predict whether the process will continue, be stalled, or even reverse. Here, we focus on hypertrophied lips as a trait potentially driving speciation in the Midas cichlids complex. We found a clear functional trade-off regarding feeding behavior between thick- and thin-lipped ecotypes, with thick-lipped fish performing better on non-evasive prey attached to “crevices” at the cost of poor performance on evasive preys. By using enclosures in the wild, we were able to validate our expectations that thick-lipped fish perform significantly better in a rocky than in a sandy habitat. Genome-wide differentiation between ecotypes was higher in the older source lakes (great lakes Nicaragua and Managua) than in the young crater lakes, where the differentiation between ecotypes seems to be related to the time of colonization, being subtle in L. Masaya (1,700 generations ago) and absent in the younger L. Apoyeque (<600 generations ago), although we found strong assortative mating. Overall, our results suggest that hypertrophied lips are very likely acting as a ‘magic trait’ promoting incipient sympatric speciation through divergent selection (ecological divergence in feeding performance) and non-random mating (assortative mating) in the young Nicaraguan crater lakes.
- 93 -
Introduction
The likelihood of speciation in the presence of gene flow has been controversially debated during the last decades (Mayr 1942; Coyne & Orr 2004; Gavrilets 2004; Bolnick &
Fitzpatrick 2007; Fitzpatrick et al. 2008; Nosil 2012; Feder et al. 2013). However, it is generally accepted that divergence in sympatry is more likely when the same trait that is under divergent selection is involved in assortative mating, also referred to as a ‘magic trait’
(Gavrilets 2004; Servedio et al. 2011). In this regard, it is fundamental to measure how strong key parameters such as divergent selection and assortative mating are at the early stage of sympatric speciation in order to predict whether the process will continue, be stalled, or even reverse (Papadopulos et al. 2011; Martin 2013). Crater lake cichlids exemplify some of the most convincing cases of sympatric speciation (Schliewen et al. 1994; Coyne & Orr 2004;
Barluenga et al. 2006; but see Martin et al. 2015) and represent an ideal system to assess the potential role of magic traits in the early stages of divergence. Despite the fact that ´magic traits´ might play a key role in speciation, convincing examples in nature are scarce (but see Servedio et al. 2011).
Midas cichlids (Amphilophus spp. species complex.) inhabit a group of small and isolated volcanic crater lakes located in Western Nicaragua which were colonized recently (less than 20,000 years ago) from the great lakes Managua and Nicaragua (Fig. 5.1A). This group of Neotropical cichlids is particularly variable in eco-morphological traits such as body shape (limnetic/benthic), pharyngeal jaws (papiliform/molariform), and hypertrophied lips (thick-/thin-lipped) that are the focus of this study (Barluenga et al. 2006;
Elmer et al. 2010b; Manousaki et al. 2013; Machado-Schiaffino et al. 2014). Interestingly, hypertrophied lips have evolved independently and repeatedly in African and Neotropical cichlids. Both Neotropical and African thick-lipped species seem to be associated with rocky habitats (Kohda & Tanida 1996). The fact that thick-lipped ecotypes have evolved in parallel suggests that this trait is an adaptation (Losos 2011), probably in response to selective pressures associated with foraging behaviour. It has been suggested that the feeding apparatus (hypertrophied lips in combination with thin and pointed heads) of thick-lipped species enhances the ability to forage in rocky substrates by facilitating the access to crustaceans and fish larvae that hide between rocks (Barlow & Munsey 1976; Konings 1998;
Arnegard & Snoeks 2001; Oliver & Arnegard 2010; Baumgarten et al. 2015). Although this has not yet been properly tested in Neotropical cichlids, it appears to be true for African cichlids (Baumgarten et al. 2015).
- 94 -
Fig. 5.1 Thin- and thick-lipped populations in the Midas cichlid species complex. (A) Map showing the four lakes where thin- and thick-lipped Midas cichlids coexist. Admixture plot showing clear genetic clustering among the four lakes is presented below the map. (B, C, D, E) Distribution of normalized lip area in wild-caught individuals from the great lakes Managua (purple) and Nicaragua (green) and the crater lakes Apoyeque (red) and Masaya (blue).
Within the Midas cichlid species complex, thick- and thin-lipped ecotypes are present in both great lakes (formally described as A. labiatus and A. citrinellus, respectively), and in the crater lakes Masaya and Apoyeque (Fig. 5.1A). Interestingly, L. Masaya and L.
Apoyeque are among the oldest and youngest crater lakes (Kutterolf et al. 2007) inhabited by Midas cichlids in Nicaragua, respectively. Morphological differentiation, principally in the mouth region, between thick- and thin-lipped ecotypes has been shown (Elmer et al.
2010c; Manousaki et al. 2013). Moreover, diet differences between thick- and thin-lipped ecotypes were previously reported for both African (Colombo et al. 2013) and Neotropical (Manousaki et al. 2013) cichlids. In fact, incipient ecological sympatric speciation between thick- and thin-lipped ecotypes has been suggested in lake Apoyeque (Elmer et al. 2010c).
The young age of the Midas cichlid complex and the repeated colonization of the crater lakes presumably at different time points represents a unique opportunity to assess the role of ecologically relevant traits, such as hypertrophied lips, during the early and more advanced stages of speciation. By comparing the pattern of genome-wide differentiation, demographic history, the degree of assortative mating and the functional trade-off between thick- and thin-lipped ecotypes, we aim to disentangle how these factors promote or constrain sympatric speciation in the young Nicaraguan Midas cichlids complex.
- 95 -
Results
Bimodal distribution of hypertrophied lips
We found strong evidence for phenotypic bimodality of lip area both in the great lakes and in the young crater lakes (Fig. 5.1B). Although bimodality in Lake Masaya was not significant, probably due to the small sample size, it was clearly not unimodal. Furthermore, the level of phenotypic differentiation was more pronounced in the source lakes (Fig. 5.1B, 5.1C) than in the recently colonized crater lakes (Fig. 5.1D, 5.1E). This pattern is concordant with theoretical predictions of eco-morphological traits under disruptive selection during the early stages of divergence.
Disruptive selection acting on lips
We detected significant differences in feeding performance between wild caught thick- and thin-lipped ecotypes from Lake Apoyeque (Fig. 5.2). On the one hand, thick-lipped fish performed better than thin-lipped fish when food was attached (sessile) to an acrylic structure with a continuously decreasing angle (Fig. 5.2A), mimicking crevices in the natural habitat.
Despite the fact that thick- and thin-lipped fish are different both in terms of lips and head shape, feeding performance was significantly correlated only with hypertrophied lips (Fig.
S5.1). We validated these findings in the lab with thick- (A. labiatus) and thin-lipped (A.
citrinellus) fish from great lake Nicaragua (Fig. S5.2). On the other hand, wild-caught thin-lipped fish performed significantly better than thick-thin-lipped fish when we exposed both ecotypes to free swimming fry (Fig. S5.3). Strikingly, the success ratio (number of eaten prey items divided by the total number of attempts) was negatively correlated with the size of the lips (Fig. 5.2B).
To test if this trade-off in performance would translate to fitness differences in the wild we tested whether the performance of thick- and thin-lipped ecotypes is different in regard to a rocky or sandy habitat. To do so, we employed single-fish enclosures (n=48) in L. Apoyeque in the respective habitats and monitored the change in weight of the fish as a proxy for fitness. All the fish employed in the enclosures had similar starting weight (mean
= 25 g, SD=7.64). The change in weight (before and after seven weeks of the experiment) differed significantly between ecotypes and habitats (Fig. 5.2C). In general, most of the fish lost weight in the enclosures, however, thick-lipped fish placed in the rocky habitat gained significantly more weight than thick-lipped fish placed in the sandy habitat (Mann-Whitney test, p<0.05). Furthermore, there was a trend (Mann-Whitney test, p=0.14) towards thick-lipped fish losing less weight than thin-thick-lipped in the rocky habitat. Weight change was not correlated with the initial weight of each individual (R2=0.102, p>0.05).
- 96 -
The absence of significant differences in weight between thick- and thin-lipped fish in the sandy habitat might be explained by the fact that evasive prey was free to leave the enclosures, thereby depriving thin-lipped fish from their putative advantage.
Fig. 5.2 Functional trade-off between ecotypes. (A) Correlation between normalized lip area and feeding performance (total length of mosquito larvae removed) for wild-caught thick- and thin-lipped fish from crater lake Apoyeque exposed to mosquito larvae attached to an acrylic structure with continuously decreasing angle. (B) Correlation between normalized lip area and feeding success ratio (number of attempts needed to eat three preys) for wild caught thick- and thin-lipped fish from crater lake Apoyeque exposed to free swimming fry. (C) Differential performance (change in weight per day) between thick- and thin-lipped ecotypes in rocky and sandy habitats in crater Lake Apoyeque.
Assortative mating between ecotypes
Given the strong bimodality and fitness trade-off of this trait we surveyed whether ecotypes mate assortatively, as expected under a magic trait hypothesis. Due to its small size and the clarity of the water, we chose Lake Apoyeque for carrying out a census of pair composition during two consecutive breeding seasons (2013-2014). We found highly significant non-random mating based on ecotype in a total of 68 pairs in Lake Apoyeque (Fig. 5.3A),
- 97 -
suggesting restricted gene flow between the ecotypes. Strikingly, we found several assortative thin-lipped pairs (20 out of 30 pairs with information about habitat) in the rocky habitat suggesting no role of habitat isolation.
We also validated these results in a pooled mate choice experiment carried out in the lab with A. citrinellus and A. labiatus from L. Nicaragua. Out of 25 pairs, all were assortative (Fig. 5.3B). Interestingly, there was no correlation in the size of the lips among the partners when A. labiatus was analyzed separately, suggesting that the presence of lips, rather than its size, might play a role for a female (or male) to choose its mate (Fig. 5.3C).
Fig. 5.3 Assortative mating by ecotype. (A) Assortative mating in crater lake Apoyeque. Census of pairs during breeding season 2013-2014. Observed number of pairs (red/orange) and expected under random mating (grey) are shown. (B) Strong Assortative mating under laboratory conditions between A. citrinellus (thin-lipped) and A. labiatus (thick-lipped) from great lake Nicaragua. Observed number of pairs (green or light green) and expected under random mating (grey) are shown. (C) Correlation between normalized lip area of mating females and males under laboratory conditions.
- 98 -
Recent colonization and weak genome-wide genetic differentiation between ecotypes in the crater lakes.
Based on more than 16,740 polymorphic markers, we found a clear genome-wide genetic differentiation among lakes, where each lake population forms a distinct genetic cluster (Fig.
5.1A, Fig. S5.4). It was previously suggested, based on few microsatellite markers, that Lake Masaya was colonized from Lake Nicaragua (Barluenga & Meyer 2010); however, using our comprehensive set of markers we found that both crater lakes seem to share a more recent common ancestor with Lake Managua (Fig. 5.4A). Thus, we have the unique opportunity to study the evolution of two independent replicates that were recently colonized, probably at different times, from the same source population (Lake Managua).
Lake Apoyeque and Masaya were genetically distinct from the great lake Managua (Table S5.1; Fig. S5.4). Moreover, we found clear, albeit low, genetic differentiation between thick- (A. labiatus) and thin-lipped (A. citrinellus) species in the source lakes Managua (FST
=0.050, p<0.00001) and Nicaragua (FST=0.019, p<0.00001) (see Fig. 5.4B, 5.4C; Fig. S5.5, Fig. S5.6). These findings are in agreement with the strong assortative mating between A.
labiatus and A. citrinellus from Lake Nicaragua that we found in lab experiments (Fig. 5.3B).
Similarly, we found significant genome-wide differentiation between ecotypes in crater lake Masaya (FST=0.016, p<0.00001) (Fig. 5.4E). However, no genetic differentiation was found between ecotypes in the young crater lake Apoyeque (FST=0.002, p=0.827) (Fig. 5.4D). This lack of genome-wide differentiation is not unexpected in the very early stage of speciation without geographical isolation.
We further inferred the demographic history of the crater lake populations (Fig.
5.4F) by performing coalescent simulations and comparing the fit against the site frequency spectrum (SFS) (Excoffier et al. 2013). We tested 13 and 17 different models for L. Apoyeque and L. Masaya, respectively (Table S5.3). According to the best model for L. Apoyeque the source population experienced a population bottleneck about 1,870 (95% confidence interval: 1,480-2,520) generations ago and was reduced from approximately 19,760 (19,060-20,670) individuals to only ca. 1,490 (1,040-2,290) individuals and it has since been growing exponentially reaching a population size of about 460,270 (0-898,670) individuals at present.
We note that the confidence intervals around the current population sizes are very broad, probably because small differences in the estimated growth rates will lead to a high variance since growth is exponential. L. Apoyeque was colonized by a small founder population of only about 110 (50-200) individuals around 580 (430-770) generations ago, growing to a current size of 14,720 (1,480-32,990) individuals. Continuous migration between the lakes is not supported, but a single admixture event ca. 380 (290-470) generations ago in which the crater lake population received ca. 16% (8.3%-22.4%) of its gene pool from the source
- 99 -
population is strongly supported. In the best model for L. Masaya the estimates for the population dynamics in the source lake are similar to above with a bottleneck happening ca.
1,700 (1,490-2,000) generations ago resulting in a population crash from about 18,490 (17,760-19,260) to approximately 1,210 (970-1,560) individuals followed by growth to a current size of around 683,670 (0-1,451,160) individuals. L. Masaya was colonized ca. 1,700 (1,490-2,000) generations ago. Surprisingly, a population size change in L. Masaya is not supported, the small increase in likelihood is outweighed by the penalty of adding a growth parameter, and the population in L. Masaya seems thus to have remained relatively stable at ca. 8,610 (7,800-9,760) individuals throughout its history. To put this in perspective, the long-term effective population size of L. Apoyeque (i.e. assuming no change) is estimated to be only 1,620 individuals and thus much smaller than L. Masaya. Migration from L.
Masaya into the source population has happened with a probability of 6.06 x 10-5 (i.e.
approximately 6 out of 100,000 alleles) per generation. Continuous migration in the other direction is not supported, yet L. Masaya received ca. 21.0% (14.5%-29.2%) of its gene pool from the source population in an admixture event about 240 (120-400) generations ago.
While we consider the model above the best model, a different model in which L. Masaya was colonized before the bottleneck in the source lake received a higher support. Yet this model assumes biologically unrealistically high amounts of gene flow. According to this model L. Masaya would have been colonized much earlier, ca. 6,390 generations ago, but continuously received alleles with a probability of 1.91 x 10-4 and moreover 57.1% of its gene pool in an admixture event from the source population around 1,190 generations ago.
Distinguishing between more ancient divergence events and high amounts of gene flow and very recent divergence with little or no gene flow based on genetic data is challenging (Hey et al. 2015), but taking the geographic isolation of these crater lakes into account we think the scenario assuming such high amounts of genetic exchange is unlikely.
- 100 -
Fig. 5.4 Genetic relationship and differentiation between ecotypes. (A) Neighbor-joining tree showing that the fish from crater lakes Apoyeque (red) and Masaya (blue) share a more recent common ancestor with those from great lake Managua (purple). (B, C) First three main axes of genetic variation (principal components) between ecotypes in the great lakes Managua (purple/light purple) and Nicaragua (green/light green). (D, E) First three main axes of genetic variation (principal components) between ecotypes in the crater lakes Apoyeque (red/orange) and Masaya (blue/light blue). (F) Schematic illustrations of the most supported demographic models of crater lakes Apoyeque and Masaya.
Discussion
We found differences in terms of morphology and genome-wide differentiation between thick- and thin-lipped fish in the Nicaraguan Midas cichlid species complex. Moreover, we found a clear functional trade-off with respect to feeding behavior for the very first time between thick- and thin-lipped ecotypes, where thick-lipped fish perform better feeding on non-evasive prey attached to narrow angles (“crevices”) but need significantly more attacks to successfully feed on free swimming fry. We later validated in situ these findings using
- 101 -
enclosures in the crater lake Apoyeque, where thick-lipped fish performance was significantly better in the rocky than in the sandy habitat. Overall, this is the first time that a clear association between phenotype and feeding performance has been found for hypertrophied lips in cichlids both in the field and under laboratory conditions. Moreover, our results suggest that this trait might also be involved in assortative mating, pointing out that hypertrophied lips could be one of the few convincing examples of magic traits.
Disruptive selection acting on lips
This study provides evidence that hypertrophied lips, and not a correlated trait, is subject to divergent selection. In this regard, a clear binomial distribution for this trait was found in all the lakes studied, the effect being more pronounced in the older lakes (great lakes Managua and Nicaragua). Moreover, a clear functional trade-off with respect to feeding performance was found between thick- and thin-lipped fish where the former ones are more successful feeding on small crevices (characteristic of rocky habitat) and the later ones feeding on free swimming fish (sandy/open-water habitat) (Fig. 5.2). Thus, it is expected that hybrids between thick- and thin-lipped ecotypes, which are known to exhibit an intermediate phenotype (Machado-Schiaffino et al. 2014), would perform worse on either resource than either of the more extreme phenotypes (Rueffler et al. 2006) and thus have a lower fitness (Rundle & Whitlock 2001).
One of the most accepted hypotheses about the function of hypertrophied lips is that it increases suction power during foraging by sealing cracks and grooves in rocky substrate (Barlow & Munsey 1976; Konings 1998; Oliver & Arnegard 2010; Baumgarten et al. 2015).
Evidence for his hypothesis has recently be presented in African cichlids (Baumgarten et al.
2015). Here, we found that the benefits of having hypertrophied lips have associated costs in the open-water (Fig. 5.2). In concordance, clear diet differences were previously reported for thick- and thin-lipped fish in Midas cichlids (Colombo et al. 2013; Manousaki et al. 2013) suggesting that thin-lipped fish tend to predate more on evasive prey (e.g. fish) than thick-lipped ones (e.g. hard-shelled invertebrates). At a proximal level, the kinetics of the buccal protrusion during prey capture might differ among these ecotypes. Hypertrophied lips might be constraining the size and shape of the mouth in thick-lipped fish and create more turbulence in the water, thereby decreasing the feeding success (Skorczewski et al. 2012) on free swimming fish (Fig. 5.2B). Preliminary analyses furthermore suggest that A. labiatus and A. citrinellus differ significantly in mouth shape, the former one having a more oval mouth shape determined by a longer gape height and ascending, and descending process (Fig. S5.7).
Thus, it is expected that a more planar and circular mouth shape, such as in A. citrinellus, increases feeding performance by maximizing the capacity of suction feeders to exert
- 102 -
hydrodynamic forces on the prey (Skorczewski et al. 2012). Further studies about the kinetics of the buccal protrusion are needed to corroborate these findings.
Field transplant experiments have been proved to be a successful way to test the role of natural selection in divergent environments (Soria-Carrasco et al. 2014). To our knowledge, this is the first time that transplant experiments (enclosures) have been carried out in Neotropical cichlids to test the performance of different ecotypes in different habitats.
Strikingly, a clear trend where thick-lipped fish perform better in the rocky habitat was observed (Fig. 5.2C).
Reproductive isolation: assortative mating
The evolution of non-random mating is crucial for the success of speciation driven by selection for local adaptation in the presence of gene flow (Schluter 2000). In this regard, non-random mating based on coloration has been observed in Midas cichlids (Elmer et al.
2009) suggesting that assortative mating based on different cues is possible and might be one of the mechanisms driving divergence in this young system.
Both ecotypes are present in sandy and rocky habitat during the breeding season in Lake Apoyeque and several thin-lipped assortative pairs were observed in both habitats.
Moreover, our laboratory experiments allow us to confirm that the complete assortative
Moreover, our laboratory experiments allow us to confirm that the complete assortative