Phylogenomic analysis of Lake Malawi cichlid fishes: Further evidence that the three-stage model of diversification does not fit

Highlights

• 1307 ultra-conserved element marker sequences in Malawi cichlids.

• The sand-dwelling species Cyathochromis obliquidens is a mbuna.

• Phylogeny inconsistent with three-stage model of Malawi cichlid diversification.

Abstract

Adaptive radiations could often occur in discrete stages. For instance, the species flock of ∼1000 species of Lake Malawi cichlid fishes might have only diverged once between rocky and sandy environments during the initial stage of their diversification. All further diversification within the rock-dwelling (mbuna) or sand-dwelling (utaka) cichlids would have occurred during a subsequent second stage of extensive trophic evolution that was followed by a third stage of sexual trait divergence. We provide an improved phylogenetic framework for Malawi cichlids to test this three-stage hypothesis based on newly reconstructed phylogenetic relationships among 32 taxonomically disparate Malawi cichlids species. Using several reconstruction methods and 1037 ultra-conserved element (UCE) markers, we recovered a molecular phylogeny that confidently resolved relationships among most of the Malawi lineages sampled when a bifurcating framework was enforced. These bifurcating reconstructions also indicated that the sand-dwelling species Cyathochromis obliquidens was well-nested within the primarily rock-dwelling radiation known as the mbuna. In contrast to predictions from the three-stage model of vertebrate diversification, the recovered phylogeny reveals an initial colonization of rocky reefs, followed by substantial diversification of rock-dwelling lineages, and then at least one instance of subsequent evolution back into sandy habitats. This repeated evolution into major habitat types provides further evidence that the three-stage model of Malawi cichlid diversification has numerous exceptions.

Introduction

The incredibly species rich radiation of cichlid fishes in Lake Malawi (∼1000 species) has been hypothesized to provide a model of ecological divergence that many radiations mirror: evolution along three predictable niche axes in temporally discrete stages (Danley and Kocher, 2001, Streelman, 2003). This three-stage model posits that adaptively radiating clades diverge predictably and sequentially: first along a habitat axis, then along a trophic axis, and finally along a signaling, or sexual trait, axis. The three-stage model has strongly influenced the study of adaptive radiation and has been used to characterize diversity in cichlid lineages ranging from Central America to the East African rift lakes (Salzburger, 2009, Martin and Genner, 2009, Hulsey et al., 2010, Parnell and Streelman, 2011, Kautt et al., 2012, López-Fernández et al., 2012, Hulsey et al., 2013a, Husemann et al., 2014, Muschick et al., 2014, Salzburger et al., 2014, Santos-Santos et al., 2015; Ivory et al., 2016, Malinsky and Salzburger, 2016). This hypothesis of niche evolution has also been invoked as a putative explanation for diversification in a large number of other disparate clades including plants, invertebrates, and other vertebrate groups (Ackerly et al., 2006, Cowman et al., 2009, Harmon et al., 2008, Gavrilets and Losos, 2009, Arnegard et al., 2010, Glor, 2010, Sallan and Friedman, 2012). However, despite its widespread usage as a model for adaptive radiation, no explicit phylogenetic test of the three-stage model has been made in the group where it was first formulated, the Lake Malawi cichlid fishes.

In both freshwater and marine environments, transitions onto reefs could be key to determining how a clade diversifies (Alfaro et al., 2007, Price et al., 2011, Hodge et al., 2012, Price et al., 2013, Tornabene et al., 2015), but specialization to this habitat could also represent an evolutionary dead end (Alroy, 2008, Kiessling and Simpson, 2011). Reef-dwelling lineages might only rarely or never produce species that evolve to colonize other habitats. For instance, the mbuna constitute a putatively monophyletic group of approximately 400 cichlid species that could only inhabit the rocky shores of Lake Malawi (Ribbink et al., 1983; Genner et al., 2004, Genner and Turner, 2005). The habitat complexity that characterizes the rocky shores of Lake Malawi provides rich opportunities for niche partitioning and trophic specialization, and the mbuna do dominate and extensively exploit this habitat (MacArthur and Levins, 1964, Schoener, 1974, Ribbink et al., 1983; but see Martin and Genner, 2009). But, these rocky reefs could also have provided a habitat niche that was impossible for the mbuna to escape once they colonized it. For instance, modifications for algivory that is the dominant mode of mbuna feeding or the evolution of specific locomotory abilities associated with navigating complex rocky environments might have led to ecological specialization that was extremely difficult to reverse (Schluter, 2000, Alfaro et al., 2007, Price et al., 2011, Rupp and Hulsey, 2014). Once they initially colonized the rocky reefs in Malawi, the mbuna clade may also have entered a stage in which sexual selection and trophic evolution exclusively drove their diversification (Streelman, 2003, Malinsky and Salzburger, 2016). Phylogenetic analyses could help to resolve whether major habitat shifts back to sandy habitats have occurred during the mbuna radiation.

When traditional phylogenetic approaches have been used to reconstruct relationships among Malawi cichlids, they have often produced trees with limited resolution. Recovering robust phylogenetic hypotheses for Malawi cichlids has proven to be challenging due both to the young age of the entire clade (∼2 million years) and the high potential for hybridization within this largely sympatric radiation (Kocher et al., 1995, Albertson et al., 1999, Hulsey et al., 2010, Mims et al., 2010, Brawand et al., 2014, Genner et al., 2015). Most previous inferences of Malawi cichlid phylogeny have been based primarily on mitochondrial DNA sequences (Kocher et al., 1995, Meyer et al., 1996, Moran and Kornfield, 1993, Hulsey et al., 2007, Hulsey et al., 2010, York et al., 2015). However, the rampant shared polymorphism and limited resolution provided by physically linked mitochondrial markers provides little confidence in the phylogenetic position of most lineages within the Malawi radiation (Moran and Kornfield, 1993, Meyer, 1994, Won et al., 2005, Hulsey et al., 2013b, Brawand et al., 2014). Advantageously, next-generation sequencing technologies offer enormous promise for resolving even the most intractable of phylogenetic problems. For instance, sequence capture of regions anchored by ultra-conserved elements (UCEs) offers an efficient means of generating massive genomic data sets capable of resolving phylogenetic relationships at both deep and shallow scales (Bejerano et al., 2004, Faircloth et al., 2014). UCEs have become increasingly popular as phylogenetic markers and have been used to reconstruct evolutionary trees for ancient clades as divergent as mammals, fishes, birds, turtles, and arthropods (Crawford et al., 2012, Faircloth et al., 2014, Faircloth et al., 2013, McCormack et al., 2012). However, one of the most compelling characteristics of UCEs for use in systematics is that the flanking regions increase in variable sites, and thereby phylogenetically informative changes, as the distance from the UCE center increases (Faircloth et al., 2012). This variation, in theory, should allow for better resolution of nodes across a range of evolutionary timescales potentially even including the short timeframe over which Malawi cichlids have diversified (McGee et al., 2016).

If a robust phylogeny of Malawi cichlids were reconstructed, there are several lineages that could violate the pattern of diversification predicted from a strict interpretation of the three stage model (Konings, 1991). Within what are generally considered utaka, or non-mbuna, there are a number of species that have likely invaded rocky reefs. The algivorous species Protomelas taeniolatus is trophically similar to many mbuna but is likely more closely related to groups that are largely sand-dwelling (Ribbink et al., 1983). Other non-mbuna species such as Cheilochromis euchilus that possess hypertrophied lips are also likely specialized to feed on invertebrates located in the rocky crevices of these reef habitats (Konings, 1991, Baumgarten et al., 2015, Henning et al., 2017). A number of piscivores such as Tyrannochromis nigriventer also commonly feed on fishes that exploit rocky reefs (Ribbink et al., 1983). Additionally, there are several lineages of mbuna that exploit non-rocky habitats and currently have unclear phylogenetic affinities. For instance, the species Maylandia livingstonii lives and breeds primarily in sandy habitats and putatively belongs to a genus that is otherwise largely confined to rocky reefs (Fryer and Iles, 1972, Ribbink et al., 1983, Konings, 1991). Likewise, the currently ambiguous phylogenetic position of the sand-dwelling species Cyathochromis obliquidens has consequences for our understanding of both habitat and trophic diversification in Malawi cichlids (Fig. 1). C. obliquidens has frequently been considered to be closely allied to the mbuna clade (Fryer and Iles, 1972, Ribbink et al., 1983, Hulsey et al., 2010). Yet, C. obliquidens appears to both scrape aufwuchs that coat leaves as well as take bites from plants in the genus Vallisneria (Fryer, 1959, Ribbink et al., 1983). These plants grow almost exclusively in Lake Malawi’s sandy habitats (Konings, 1991). If C. obliquidens were the sister group to the other mbuna, the ubiquitous habit of scraping algae in the mbuna (Fryer and Iles, 1972, Ribbink et al., 1983, Rupp and Hulsey, 2014) might have evolved from herbivory on vascular plants. Alternatively, trophic specialization in C. obliquidens could have involved a transition from the mbuna habit of scraping algae to sometimes eating vascular plant material. Also, if C. obliquidens were well nested within the mbuna radiation, this placement would suggest that following extensive diversification on the rocky outcrops of Lake Malawi, a member of the mbuna clade diverged to exploit sandy habitats. Alternatively, if C. obliquidens were more closely related to non-mbuna living in sandy habitats or was found to be the sister lineage to the 400 species of rock-dwelling mbuna, then it would still be possible that the radiation of mbuna once it evolved to exploit reefs might have remained confined to these rocky habitats (Streelman, 2003, Malinsky and Salzburger, 2016). Robust phylogenies of Malawi cichlids could be used to test for this extreme niche conservatism and to evaluate through a more complete understanding of Malawi cichlid relationships the strictest interpretation of the three-stage model.

To provide a comparative framework for Malawi cichlid evolution and to test a particular case of whether adaptive divergence in Lake Malawi cichlids is consistent with the three-stage model, we generated several nuclear sequence based phylogenetic hypotheses. First, sequences of 1037 UCE loci were generated for many representatives of previously delineated major lineages within Lake Malawi. Then, using several methods for tree searching, the robustness of relationships to different reconstruction methods was determined. Finally, we assessed the phylogenetic affinities of the sand-dwelling species C. obliquidens to evaluate its placement with respect to the larger rock-dwelling mbuna radiation to determine whether the patterns of habitat and trophic evolution in Malawi cichlids are strictly consistent with the three stage model.