Discussion
Using a genotyping by sequencing approach, we provide a unique and robust dataset supporting the structuration of M. festivus populations by evolution through vicariance events. Our holistic sampling design, combining 12 sites from a wide territory, 231 M. festivus samples and an optimized SNPs calling bioinformatic pipeline render remarkably different conclusions than what other similar studies obtained (Cooke et al. 2009; 2014; 2012a; b; Beheregaray et al. 2015). Using a wider sampling design and a combination of phylogeographic and environmental association models, we challenge the previous ecological speciation hypothesis (i.e., genetic structure is linked to diverging water types) and show strong evidence that water type has a low-structuring power on M. festivus populations. Our results support a much more important influence of vicariance events associated with the Amazon’s formation and isolation by unidirectional downstream water current on structuring populations in this clade.
Phylogeographic analysis
According to the Admixture posterior membership probability plots [Fig. S1], the cross-validation error values from ADMIXTURE [Fig. S2] and the BIC values from “find.cluster” [Fig. S3], four genetic populations were detected across our 12 sampling sites. The geographic structuring of these four genetic populations can be used to infer the phylogeographic history of M. festivus, its colonization potential and to infer on the evolutionary history of closely related clades.
The first genetic group detected is formed by five Rio Solimões sites: CAT, JAR, JAC, MAN and PIR. Despite the small river course distance separating them, it appears that an important genetic distance separates CAT, JAR, JAC and MAN from the sites sampled in the Rio Negro [Fig. 2]. This first genetic gap [Fig. 2], between the Rio Negro and Solimões, has already been assessed in a series of scientific papers based on similar sampling designs (Cooke et al. 2009; 2012a; b; 2014; Beheregaray et al. 2015). In these studies, this genetic gap was associated with ecological speciation, an evolutionary process caused by the presence of an ecotone like the black and white water confluence. Strikingly, fish sampled at CAT are more closely related to PIR than to CEM, located respectively at 631 and 71 km from CAT [Fig. 1]. Here, the genetic distances seem to be unrelated to the geographic distance, but strongly correlated to other variables such as water type differences and isolation by unidirectional downstream currents. While this result is in concordance with the ecological speciation hypothesis and the results from Cooke et al. (2009; 2014; 2012a; b), it ultimately does not lead us to the same evolutionary conclusions.
Indeed, we observed a second genetic gap at the confluence of the Negro and Branco Rivers [Fig. 2], forming our second genetic group composed of the sites BAR and NEG. This genetic divergence has not been recorded in previous Amazonian population genetic studies, demonstrating the advantage of using a larger sampling design. While it might be interpreted as evidence of the divergence between the black (NEG) and white (BRA) water types, this conjecture is refuted by the higher relatedness between BRA and ANA, respectively white and black water sites, than between NEG and ANA, both black water, in the Admixture results and the Pairwise linearized Fst heatmap [Fig. 2-3]. This is despite the drastic environmental shift separating ANA and BRA, and both sites being at a similar river course distance from ANA, respectively 215 and 225 km for NEG and BRA. To reiterate, this result supports that gene flow is stronger between sites with different water types and that water type is not an important migration barrier for M. festivus at these sites. These genetic patterns are similar when comparing BRA and NEG with CEM [Fig. 2 and 3]. These sites: BRA, ANA and CEM, are forming the third genetic population observed.
Here, the strong genetic divergence between NEG and BRA could be caused by the strong unidirectional downstream water current that prevents the migration of M. festivus from the Rio Branco to upstream of the Rio Negro and vice versa. This downstream biased gene flow could lead to a partial isolation of upstream M. festivus populations, which admix in downstream rivers. Ultimately, this should lead to a downstream increase in intraspecific genetic diversity at the confluence of large rivers where multiple populations that are upstream and isolated by current genetic populations meet (Paz-Vinas et al. 2015). Systematically, the sites located downstream of the main rivers (i.e., CEM, ANA, CAT and JAR) did have a higher mean heterozygosity rate than their upstream relatives (i.e., NEG, BRA and MAN) [Table 1]. In fact, the sites with the lowest heterozygosity were the ones located the most upstream (i.e., BRA, BAR, PIR and TEF). Similarly, freshwater fish diversity hotspots in the Amazon watershed are usually located at the crossing of large rivers near the meeting of large watercourses (Oberdorff et al. 2019). For instance, the highest total species richness is found at the confluence of the Negro and Solimões Rivers and other large river confluences show similar diversity patterns, for example the Tapajós-Amazon Rivers confluence and Branco-Negro Rivers (Oberdorff et al. 2019). This is supporting the hypothesis of evolution by allopatry, where neutral evolutionary processes in fish populations located in isolated rivers were followed by a reconnection of the waterways and a mix of the newly formed and reproductively isolated species.
The fourth and last genetic population is formed by two sites located in the upper Rio Solimões, near the Rio Téfé: SOL and TEF. The third genetic gap is located between SOL and PIR, a site located only 6 km upstream of SOL [Fig. 2]. According to the linearized Fst/(1-Fst) heatmap, this genetic population does not arbour an important genetic dissimilarity with the other Rio Solimões sites [Fig. 3]. Yet, the pairwise Fst Values between SOL and PIR are similar to Fst values with sites located more than 500 km downstream the Rio Solimões. Therefore, TEF (black water) seems to lead to higher gene flow into SOL (white water) than PIR (white water) does. While this result goes against the ecological speciation hypothesis, the dissimilarity between SOL and PIR could, again, be caused by isolation by strong downstream water currents. We selected the site SOL as a site located right between PIR and TEF, two sites characterized by drastically divergent environmental conditions [Table 1]. It is possible that the river architecture between TEF and SOL is more favourable to gene flow coming from TEF than from PIR for M. festivus. Either way, this again supports that water type is not an important migration barrier for M. festivus since the populations from different water types are more genetically alike to each other’s than the site from the same water type.
According to the MRM analysis results, most of the gene flow between the sampled sites happened between sites connected by downstream water flow, with higher amounts of gene flow happening between sites located close to each other, irrespectively of the water type at each site [Fig. S4]. In this sense, the water type at each site had a non-significant and low power at explaining the genetic distance between sites. On the contrary, the connectivity between sites was very significant and inversely correlated to the genetic distance between sites. In the same way, sites separated by a small pairwise river course distance were more genetically related. According to these results, the divergent water types do not significantly affect the migration rate between sites for M. festivus. Conversely, the downstream water flows connectivity and the distance between rivers seemed to play a much more important role at structuring M. festivus genetic populations.
According to our results based on neutral loci, the three genetic gaps detected are caused by a combination of isolation by strong unidirectional water current and past evolution by allopatry between watersheds that recently reconnected. While the diverging environmental conditions must certainly have at least some effects on fish evolution, due to differences in productivity, food availability, species assemblages, environmental pressures, physiological demands, etc., we argue that isolation by strong downstream water current and past isolation by geological processes played a much more important role in shaping the genetic structure of M. festivus populations.
Environmental Association Study
While we did not detect an important impact of the water types on the structure of M. festivus genetic populations using neutral loci, the effect of an ecologically driven change in specific allelic frequencies in presence of gene flow could still be detected using genotype to environment association models. We conducted a complete environment to genotype association study (EAS) aiming to identify SNPs strongly associated with selected environmental variables and directly to water type. We chose to use DOC concentration, chlorophyll a concentration, conductivity, silicate in suspension and dissolved aluminum concentration since other authors have previously employed these to characterize black and white waters (Junk et al. 2011, 2015). While a low pH has proven to be a major characteristic of the black water environment (Ríos-Villamizar et al. 2013), its strong covariation with other parameters required its exclusion from the analyses. The aforementioned parameters have a very good power at differentiating black and white water sites in our study [Fig. 4]. If water type is responsible for a strong ecological speciation in our system, we expect to detect a strong pattern of differentiation between sites of different water types at the 172 SNPs associated with water type and its associated physicochemical parameters [Fig. 5]. Also, samples should not cluster according to their watershed of origin.
In a PCA, the clustering of samples according to the SNPs associated to the environment [Fig. 6] is similar to the result obtained using the full dataset [Fig. S10], which is compatible with the scenario of neutral divergence in allopatry. Likewise, the dominant influence of neutral evolutionary forces, mutation and genetic drift in conditions of low gene flow between certain populations, contrasts with the low influence of directional evolutionary forces in the genetic structure of environmentally associated SNPs. When clustered by watershed, samples are very well differentiated in the PCA plot [Fig. 6B], while clustering the samples by their water type gives a much more admixed PCA plot [Fig. 6C]. This is strong evidence that the presence of divergent water types is not one of the main evolutionary factors and that neutral evolutionary processes have a much stronger impact on the differentiation of these populations. This is even though we corrected for the neutral genetic structure in the three EAS methods. Since we only sampled a fraction of M. festivus’ genome and did not have access to a reference genome to map for SNPs associated with genes of interest, this result does not rule out the possibility that other genes could be positively selected in a specific water type. Doing an analysis with a reference genome could lead to the discovery of key genes that affect the fitness of M. festivus individuals in each environment. When combined with our previous analyses, the EAS results provide very strong evidence that evolution by ecological speciation did not have an important influence on M. festivus population structure in Amazonia.
Refuting Previous Assumptions of Strong Ecological Speciation
The recently accepted assumption that ecological speciation was a major Teleostean evolution driver in the tropics seems to have been derived from observations of a site effect in papers with similar experimental designs (Cooke et al. 2009; 2012; 2012a; 2015). Our results support that, in these papers, the structuring effect caused by the water type was potentially confounded with the effect of a strong unidirectional water flow at the sampled sites. Furthermore, adding new sites to these past studies could lead them to different conclusions. For instance, if we consider only the sites that were used by Cooke et al. (2009; 2012; 2012a; 2015) in our results, it leads us to very similar results and conclusions, and this is even though this series of previous studies focused on different species that are genetically distant from M. festivus. For these papers, Cooke et al. sampled the genetic gap happening at the confluence of the Rio Negro and Solimões. Our genetic data behaves identically to theirs at these sites and we detected an extensive genetic distance between the Rio Negro and Solimões populations. It is only after adding the data from sampling sites at two other black and white water confluences that we detected evidence that adaptative divergence to specific water types was not the main driver of population structure in M. festivus. There seems to be a real possibility that adding new sites to Cooke’s studies would lead to results like ours. The application of a wide sampling design and a deep sequencing approach have previously resolved fine-scale phylogeographic patterns in other teleostean species (Fairweather et al. 2018; Fang et al. 2018).
A Demographic Scenario Based on the Amazon’s Geological History
Our study detected three main genetic gaps at the confluences of: (1) the Negro and Solimões Rivers, (2) Branco and Negro Rivers, and (3) Lago Tefé and Solimões Rivers. What do these three genetic gaps have in common? First, they all represent a confluence of black and white water rivers. Second, gene flow is always going downstream, which could mean the water current is the primary driver of gene flow for M. festivus. Considering its physiology and sedentary behaviour (Pires et al. 2015), strong current velocity represents an important environmental gene flow barrier for M. festivus. Additionally, there have been anecdotes of fishers seeing M. festivus floating in high current while imitating a dead leaf (Pires et al. 2015), which could explain why there are sources of downstream gene flow over very long distances. Lastly, the three confluences result from the formation of the modern Amazon during the late Pliocene, between 2.5 Ma and 700 Ka (Campbell et al. 2006; Ribas et al. 2012).
The formation of the modern Amazon is relatively recent and has shown to be one of the main causes of the Amazonian Animalia terrestrial diversity boom (Albert et al. 2018; Araújo‐Silva et al. 2017; Lynch Alfaro et al. 2015). Numerous authors support the Pleistocene refugia hypothesis, assuming that Amazonia diversity is partly due to geographic isolation caused by geologic processes and salt-water incursion events (Farias and Hrbek 2008, Bragança and Costa 2018, da Rocha and Kaefer 2019). Amazonian cichlids are much older, dating from the separation from their African sister clade, and have evolved mostly in riverine ecosystems (Concheiro Pérez et al. 2007). For this reason, M. festivus, a ubiquitous Cichlid species, is a good model to understand the impact of the formation of the modern Amazon on fish evolution. The modern transcontinental Amazon got established 2.5 Mya, when the west and east territories were divided in two (Campbell et al. 2006). The water drainage led to sediment accumulation, creating thousands of small lakes and rivers. The Negro and Branco Rivers were formed approximately between 1.0 to 0.7 Mya and Tefé River approximately between 2.0 to 1.0 Mya (Ribas et al. 2012). M. festivus was present well before the formation of these rivers and probably colonized the new downstream environments as they formed. As previously mentioned, M. festivus has a very good colonization potential of new environments and show strong proof of downstream migration. In this sense, the founding population of M. festivus probably came from the Andes, in the west, and admixed in multiple downstream rivers, irrespective of their water type. Likewise, the genetic gaps detected at the confluences of black and white water rivers are probably related to the formation of the rivers (geologic history) rather than the confounding effect of the water type. As a result, the strong genetic divergence between the M. festivus populations of the Negro and Solimões rivers probably stems from geological processes that led to a neutral evolution by allopatry.
M. festivus low upstream migration potential (Pires et al. 2015) must have slowed the admixture process after the reconnection of the waterways as observed in the present watershed structures. Considering this and the similarity between our results and those from Cooke et al. (2009; 2012; 2012a; 2015), multiple Amazonian fish species could have evolved in these same vicariant conditions, potentially leading to the genesis of many new species, who could have dispersed after the connection of the waterways. In order to test whether some population divergence times are linked to important geological events, it would be interesting to do a demographic analysis with a molecular clock estimate based on M. festivus mitochondrial DNA from sites positioned close to past major geologic processes.
Our study aimed to investigate the support for the two main speciation hypotheses (ecological and allopatric) used to explain the evolution of M. festivus within the Amazon basin. We showed strong evidence that the divergent physicochemical characteristics between black and white water have a weak structuring power on M. festivus populations in the Amazonian watershed. Furthermore, our results challenge the recently suggested ecological speciation hypothesis explaining fish diversification in Amazonia. Unlike previous studies focusing on a single confluence between black and white water, our extensive sampling design comprising 12 populations of M. festivus detected a genetic structure congruent with isolation by unidirectional downstream water current, past geologic events, and waterways connectivity shift. While the Brazilian Amazon supports one of the richest fish faunas on Earth, our comprehension of the evolutionary processes which shaped its biodiversity is still lacking. Understanding the origin of such richness would help us protect its diversity. Our study not only constitutes a step forward in understanding these important processes but also provides a conceptual framework that should benefit the sampling designs of future investigations on this matter.