University of Groningen
Species selection and the spatial distribution of diversity
Herrera Alsina, Leonel
DOI:
10.33612/diss.99272986
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2019
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Herrera Alsina, L. (2019). Species selection and the spatial distribution of diversity. University of Groningen. https://doi.org/10.33612/diss.99272986
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
CHAPTER 3
Depth specialization decreases the rate of diversification in Lamprologini cichlids
Leonel Herrera-Alsina, Elodie Wilwert, Thijs Janzen, Martine E. Maanand Rampal S. Etienne
3
ABSTRACTThe high diversity of cichlid fish species, combined with their relatively young phylogenetic age, suggests that there are intrinsic characteristics of this clade that promote radiation. While one can compare cichlids to other taxa to identify these characteristics, one can also look for variation in characteristics within this clade that cause variation in diversification. A conspicuous characteristic is the water depth at which cichlid fish live, which varies between species. Many abiotic and abiotic factors, such as light spectrum, temperature, dietary resources and parasites, vary along the water column, generating depth-dependent selective regimes that may have macro-evolutionary consequences. Here, we present a new method in the family of state-dependent speciation and extinction models to establish whether 1) diversification rates depend on species depth ranges and 2) depth distributions change during speciation. We apply this inference framework to the Lamprologini clade (endemic to Lake Tanganyika, East Africa) to compare contrasting hypotheses that explicitly account for different models of trait evolution and modes of speciation. We do not find evidence for depth shifts during speciation. Instead, depth shifts occur along the branches of the tree. We do find an association between depth range and speciation rate: depth range generalists (i.e. species distributed along the entire water column) have higher rates of speciation than depth range specialists (i.e. species occupying either shallow or deep water). We show that transitions between shallow water and deep water primarily occur through a generalist phase, and that shallow-water specialization is a macro-evolutionary endpoint: it is unlikely to change to another state. To explain these findings, we hypothesize that specialization to a given depth range affects dispersal capacities, which could cause differences in speciation rates. Our study shows how the evolution of a trait can be tightly linked to speciation rates, but is independent of the speciation events themselves.
Keywords: Lake Tanganyika; species selection; trait inheritance; cladogenesis, anagenesis
INTRODUCTION
As for fish, they were numerous and often remarkable - writes Jules Verne (1870) as an
epitome to the vast diversity he encounters in his fictional voyage to the depths. Indeed, several striking fish radiations have been documented and among these, cichlid diversity in East African lakes has been the focus of extensive research. Cichlid species richness in East African lakes (over 500 species in lakes Malawi and Victoria, 250 in Lake Tanganyika; Genner et al. 2014) is higher than what is expected from clade age alone (McMahan et al. 2013); it is likely the outcome of elevated speciation rates or reduced extinction rates.
Lake depth has not only been suggested as an important predictor of cichlid species richness but it has also been associated with the extent of morphological variation (Recknagel et al. 2014; Wagner et al. 2014). In fact, depth segregation of species could account for the otherwise inexplicably high levels of sympatric speciation (Seehausen 2015). Although many species coexist within lakes, they inhabit different depth ranges and are exposed to abiotic and biotic conditions that differ across the depth gradient. For instance, freshwater fish from shallow and deep waters vary both in species of parasites and extents of infection (Karvonen et al. 2012, 2018) which suggests that independent processes of cichlid adaptation could take place at different depths. In addition, temperature varies with depth, which may affect diversification dynamics in at least two ways. First, marine fish inhabiting adjacent temperature-defined regions have been shown to display incipient species divergence (Teske et al. 2019) which suggests that adaptation to different temperatures could promote speciation (Keller and Seehausen 2012). Second, variation in temperature with depth could have major consequences for the rates of physiological processes in ectotherms (Porcelli et al. 2015). Hoekstra et al. (2013) found that warm temperatures accelerate the accumulation of genetic incompatibilities between populations of Drosophila. For aquatic habitats, this suggests that rates of species differentiation could be higher in the warm waters of shallow areas than in the colder waters of the deep parts of a lake. Further, algal communities vary with water depth in response to the changing light intensity and spectrum, promoting dietary specialization among fish at different water depths, as indicated by stomach contents (Hata and Ochi 2016); this may lead to resource partitioning and favor ecological speciation (Recknagel et al. 2014). The difference in light environment between shallow and deep water also drives adaptive evolution of fish vision (Yokoyama and Yokoyama 1996; Cornell 2013; Carleton et al. 2016); this may not only contribute to depth specialization, but also directly influence assortative mating, because visual perception affects mate recognition and assessment in many species (reviewed in Boughman 2002; Maan and Sefc 2013). Smith et al. (2011) report that the expression of pigments related to vision shows substantial variation within species and deep and shallow water; such
3
variation might cause differences in the tempo of species differentiation (via visualrecognition). Finally, specialization at a certain depth can constrain the ability to disperse, which affects range size and thereby probabilities of speciation and extinction. In summary, the aquatic environment harbours a multitude of depth-dependent selective pressures, indicating that the vertical distribution of fish could be of paramount importance in their evolution and diversification.
It has been reported that cichlid species can switch from one depth to another (Moser et al. 2018) which could have consequences for the clade’s evolutionary dynamics. If this is the case, the nature of the relationship between depth range evolution and speciation could in principle be inferred from the phylogeny. There are two complementary scenarios for this relationship. First, depth shifts and speciation events may be associated. If (sub)populations colonize new depth ranges, differences between them could accumulate triggering speciation. In fact, because many factors change along the water column, it is likely that when lineages start to differ in depth preference, a major source of opportunity to speciate arises: it has been found that speciation was nonexistent or incipient in the absence of depth differentiation (Seehausen 2015). In this case, one expects sister species to differ in water depth range i.e. a phylogenetic tree should show that transitions across depth ranges take place not only in branches but also at nodes. However, speciation may not always occur: for instance if all populations of a given lineage experience the same change in depth range, ecological differentiation and reproductive isolation among populations will not occur and hence neither will speciation. In this scenario, speciation events are not accompanied by shifts in depth, and switches between depths happen only along the branches of the phylogenetic tree. Second, depth range itself may influence diversification rate. The specific environmental conditions in a certain depth range (parasitism, temperature, diet, visual environment) could spur or hamper diversification rates. In a phylogeny, evidence of this phenomenon would be found when per-lineage rates of speciation or extinction depend on what depth the lineage is at. Moreover, diversification rates would change over time, in line with the rate of transition between depths. Currently, it is unknown whether fish speciation tends to coincide with changes in depth range, and whether a species’ depth range affects its diversification rate.
Here we develop a likelihood-based inference framework to address these questions, and apply it to the cichlid tribe Lamprologini in Lake Tanganyika. Our framework belong to the family of SSE approaches (State-dependent Speciation and Extinction) that takes a phylogenetic tree and trait information to assess the interaction of an evolving trait and branching patterns. It extends the SecSSE model (Herrera-Alsina et al. 2018) by allowing trait changes during speciation, or alternatively, it extends the ClaSSE model (Goldberg and Igić 2012) by including a procedure that avoids elevated type I errors (i.e., we use a concealed-states framework as in HiSSE; Beaulieu and O’Meara 2016). We compare a
variety of models that explicitly account for contrasting hypotheses on the nature of trait change (i.e., depth range change) during speciation, the transition between depth ranges and its effect on diversification rates.
METHODS
Depth range and phylogeny of Lamprologini tribe
The Lamprologini form the most diverse lineage within Lake Tanganyika, with 84 described endemic species. All species are substrate spawners (with either maternal or biparental care) and show tremendous diversity in morphology and ecology, with most species having colonized shore habitats, while some inhabit open water habitats (Konings 2015). We compiled information on depth distribution of all Lake Tanganyika cichlid species based on field records (Poll 1956; Bailey and Stewart 1977; Kuwamura 1986; Nakai et al. 1990; Snoeks et al. 1994; Verburg and Bills 2007; Sturmbauer et al. 2010; Nagai et al. 2011; Muschick et al. 2012; Kullander et al. 2014a, 2014b; Hata et al. 2015; Janzen et al. 2017). We observed clusters at 0-8 and 8-30m depth, suggesting to use three categories: one for species occurring in shallow waters (0-8m; shallow-water specialists; n = 8 species), one for species occurring in deep waters (8-30; deep-water specialists; n = 23 species) and one for species occupying a broad depth range (0-30m; depth generalists; n = 43 species).
We reconstructed a new Lamprologini tree following the workflow of the most complete Lamprologini tree to date, which is a consensus tree based on the mitochondrial ND2 gene (Sturmbauer et al. 2010), but we added two newly described species (Lepidiolamprologus mimicus; Schelly et al. 2007) and Neolamprologus timidus (Kullander et al. 2014b). Using phyloGenerator (Pearse and Purvis 2013), we downloaded sequences from GenBank for five genes. For each species, sequences for at most four different individuals were downloaded. Genes were selected on the basis of species coverage (at least 25% of the 79 Lamprologini species for which molecular data is available). After selection, our full dataset consisted of two mitochondrial genes (NADH dehydrogenase subunit 2 (ND2) and cytochrome b (cytb)) and three nuclear genes (recombinase activating protein 1 (rag1), ribosomal protein S7 (rps7) and rod opsin 1 (RH1)). The references for gene sequences as well as the GenBank access numbers can be found in the Supplementary Material. Sequences were aligned using MAFFT (setting: --auto) (Katoh and Standley 2013), and subsequently, sequences were cleaned using trimAI (sites with more than 80% data missing were removed, e.g. setting –gt 0.2) (Capella-Gutiérrez et al. 2009). We partitioned the data into subsets with independent sequence evolution models, which is more suitable for a dataset which is expected to show incomplete lineage sorting or hybridization (Meyer et al. 2016). Substitution models were inferred jointly with the tree using the bModeltest package for BEAST 2 (Bouckaert and Drummond 2017).
3
Using *BEAST (Heled and Drummond 2010) within the BEAST 2 package (Bouckaert etal. 2014), we inferred the time-calibrated species tree. We used a fixed clock rate and applied two calibration points. Firstly, we calibrated the crown of the Lamprologini to be 4 million years old (log-normal prior, mean of 4 Myr 95% conf interval: [3, 5]), based on the results from Meyer et al. (2016). Secondly, we included two riverine Lamprologini species (L. congoensis and L. teugelsi), and calibrated the onset of their branching event at 1.7 Ma (offset 1.1, log normal distribution with mean 1.7, 95% conf interval [1.15, 3.47], “use originate = true”), following (Sturmbauer et al. 2010). We applied 1/X priors on the clock rates, and log-normal priors on the substitution rates. All other priors were left at their default setting. As tree model we used the birth-death model. Our BEAST configuration file (the Beauti xml) is given in the supplementary material. We ran 10 independent *BEAST MCMC chains, of 1750M trees each. Each chain was verified to have ESS values of at least 100 for all parameters. The first 10M trees were pruned from these chains as burn-in after which all ten chains were combined (we used the species tree, rather than the individual gene trees) into one large chain (of 17400M trees). Chains were subsequently thinned by taking each 5,000th tree. Using TreeAnnotator (from the BEAST 2 suite) we constructed a Maximum Clade Credibility tree, storing the mean heights. Contrasting models of trait evolution and diversification dynamics
We developed an extension of SecSSE (Herrera-Alsina et al. 2018) which belongs to the SSE family of macroevolutionary models (State-dependent Speciation and Extinction; Maddison et al. 2007) that allows the joint analysis of differential diversification rates between lineages, and their dependence on states of evolving traits. In a nutshell, in these models, the rates of speciation λi and extinction μi of a given lineage depend on its trait
state i and, over time, a lineage can switch from state i to another state j with rate qij. A
pruning algorithm is used to compute the likelihood of the data, i.e., a phylogeny and trait data of each extant species (trait states at the tips), given a model with its parameters λ, μ and qij (Maddison et al. 2007). High type I error rates have been reported for this family
of likelihood methods which has led to premature conclusions on trait-dependent diversification rates (Rabosky and Goldberg 2015; Rabosky and Huang 2016). However, recent studies have shown that the rate of type I errors decreases to acceptable limits when an appropriate “null” model is used for comparison (Beaulieu and O’Meara 2016; Herrera-Alsina et al. 2018). This is accomplished by contrasting the likelihood of a model in which diversification rates depend on the trait of interest (Examined-Trait-Dependent, ETD) with the likelihood of a model in which diversification rates depend on a trait that we are not analyzing (Concealed-Trait-Dependent, CTD). In the present study, the examined trait is the species’ depth category. In addition to ETD and CTD models, we also studied a Constant Rate (CR) model, in which rates are homogenous over time and across states. We defined seven different scenarios for trait evolution (i.e., the shift from one depth range to another), specifying what transitions are possible and what rates are allowed to be
different (Figure 1). For instance, in four scenarios of trait evolution, a shallow-water species cannot evolve directly into a deep-water species but must become a generalist species first, and vice versa (Constrained). In this case, all rates can be assumed the same (constrained one-rate), or different up and down the depth range (constrained two-rate) or all different (constrained four-two-rate).
Figure 1. Seven different models for transitions in water depth (range). Arrows with the same color/dashing have the same transition rate. S = Shallow-water specialist; D = deep-water specialist; G = water-depth generalist.
When speciation happens, the parental trait state may or may not be inherited by the descendant species. In the case of a daughter species exhibiting a different depth range than the mother species, speciation and depth range shift take place simultaneously. The probabilities of such inheritance modes shed light on the relative importance of vertical isolation (i.e., depth segregation) for speciation. To explore this, we studied eight different modes of speciation. 1) Dual Inheritance: both daughter species inherit the parental state. 2) Single Inheritance: one daughter inherits the parental state and the other obtains a different trait state. This means that there is a shift in trait state during speciation (e.g., speciation in a shallow-water lineage would produce a shallow-water species and a generalist or deep-water specialist). 3) Dual Symmetric Transition: both daughter species have the same trait state but they are different from the trait state of the parental species. 4) Dual Asymmetric Transition: the daughter species are different from each other and different from the mother species. These 4 modes are the basic modes of how a lineage
different (Figure 1). For instance, in four scenarios of trait evolution, a shallow-water species cannot evolve directly into a deep-water species but must become a generalist species first, and vice versa (Constrained). In this case, all rates can be assumed the same (constrained one-rate), or different up and down the depth range (constrained two-rate) or all different (constrained four-two-rate).
Figure 1. Seven different models for transitions in water depth (range). Arrows with the same color/dashing have the same transition rate. S = Shallow-water specialist; D = deep-water specialist; G = water-depth generalist.
When speciation happens, the parental trait state may or may not be inherited by the descendant species. In the case of a daughter species exhibiting a different depth range than the mother species, speciation and depth range shift take place simultaneously. The probabilities of such inheritance modes shed light on the relative importance of vertical isolation (i.e., depth segregation) for speciation. To explore this, we studied eight different modes of speciation. 1) Dual Inheritance: both daughter species inherit the parental state. 2) Single Inheritance: one daughter inherits the parental state and the other obtains a different trait state. This means that there is a shift in trait state during speciation (e.g., speciation in a shallow-water lineage would produce a shallow-water species and a generalist or deep-water specialist). 3) Dual Symmetric Transition: both daughter species have the same trait state but they are different from the trait state of the parental species. 4) Dual Asymmetric Transition: the daughter species are different from each other and different from the mother species. These 4 modes are the basic modes of how a lineage
3
can undergo speciation (Figure 2), but they are not mutually exclusive and a combinationof them could be necessary to explain diversification patterns. For example, a lineage could speciate through mode 1 and also mode 2, the rates for these two events could be different to account for differential contributions of each process. Therefore we defined 4 additional modes, based on the most intuitive combinations. In mode 5, speciation can happen through Dual Inheritance or Single Inheritance (modes 1 and 2) which means that during speciation the trait state is perfectly inherited or shows a shift in state in only one of the daughter species. We also combined Dual Inheritance with either Dual Symmetric Transition (mode 6) or with Dual Asymmetric Transition (mode 7). In the case of mode 8, speciation can happen in any of the four basic modes (mode 8 includes modes 1, 2, 3 and 4).
Figure 2. Four different models for how a trait state can be passed to daughter species during a speciation event.
We added the speciation mode functionality to the R package secsse (Herrera-Alsina et al. 2018), which now integrates a) the general framework of ClaSSE (Goldberg and Igić 2012) where cladogenetic and anagenetic processes are considered, b) the hidden/concealed trait states of HiSSE (Beaulieu and O’Meara 2016) which controls for low Type I error, and c) MuSSE (Fitzjohn 2012) where multi-state traits can be analyzed. Optimization of likelihood and robustness analysis
We considered simultaneously the dependence of speciation on trait states (i.e., CR, CTD and ETD), the evolution of the trait (seven different transition models) and the mode of speciation (eight modes of inheritance), leading to 168 different model combinations. We maximized the likelihood for each model to find the best-fitting parameters (i.e., rates of speciation, extinction and transition). We used three sets of initial parameter values in the
likelihood optimization to avoid getting trapped in local likelihood optima. For one starting point, λ and μ were the estimates from a regular birth-death model; for the other two starting points, these values were either doubled or halved. The best likelihood of these three starting points was regarded as the global optimum and used to compute AIC weights (while penalizing the number of free parameters) to find the best supported model among the 168 model combinations. We then took the five best supported models to further test whether 1) extinction rates are different across trait states, and 2) transition rates of examined and concealed states are different.
For the best model, we evaluated the potential rates of Type I and II errors when detecting evidence of trait-dependent speciation, by using the parameters that maximized the likelihood to simulate data sets that are structurally similar to the empirical data set (i.e., phylogenetic trees and trait states for extant species). The simulation procedure is similar to that described in Herrera-Alsina et al. (2018), with the extension that here we allow different modes of trait state inheritance during speciation events (as described above). We simulated 100 datasets under the CTD model (with parameters taken from the best performing CTD model), and then, for each dataset, we ran our maximum likelihood inference framework with CR, CTD and ETD models and compared them (using AIC) to count in how many cases ETD was incorrectly selected, which contributes to the Type I error rate. Similarly, we simulated datasets under the ETD model and fitted CR, CTD and ETD models and counted the cases were the generating model (ETD) was not selected, which reflects the Type II error. Because we used the consensus tree for our analysis, we explored whether an alternative phylogenetic reconstruction would lead to a different conclusion by rerunning the analysis for the seven best-supported models using 100 trees from the posterior distribution produced during the Bayesian reconstruction.
RESULTS
The topology of the Maximum Clade Credibility tree is largely consistent with previous findings (Sturmbauer et al. 2010) (Figure 3). Placement of Neolamprologus fasciatus as a close relative to N. wauthioni seems to re-iterate previously published evidence for introgressive hybridization (Koblmüller et al. 2007). For the two species not previously included in the Lamprologini phylogeny, Lepidiolamprologus mimicus was placed as a close relative to the other species within the genus Lepidiolamprologus. In contrast to previous findings (Kullander et al. 2014b), Neolamprologus timidus is not placed as a sister species to Neolamprologus furcifer, but rather associates with the closely related
N. falcicula. can undergo speciation (Figure 2), but they are not mutually exclusive and a combination
of them could be necessary to explain diversification patterns. For example, a lineage could speciate through mode 1 and also mode 2, the rates for these two events could be different to account for differential contributions of each process. Therefore we defined 4 additional modes, based on the most intuitive combinations. In mode 5, speciation can happen through Dual Inheritance or Single Inheritance (modes 1 and 2) which means that during speciation the trait state is perfectly inherited or shows a shift in state in only one of the daughter species. We also combined Dual Inheritance with either Dual Symmetric Transition (mode 6) or with Dual Asymmetric Transition (mode 7). In the case of mode 8, speciation can happen in any of the four basic modes (mode 8 includes modes 1, 2, 3 and 4).
Figure 2. Four different models for how a trait state can be passed to daughter species during a speciation event.
We added the speciation mode functionality to the R package secsse (Herrera-Alsina et al. 2018), which now integrates a) the general framework of ClaSSE (Goldberg and Igić 2012) where cladogenetic and anagenetic processes are considered, b) the hidden/concealed trait states of HiSSE (Beaulieu and O’Meara 2016) which controls for low Type I error, and c) MuSSE (Fitzjohn 2012) where multi-state traits can be analyzed. Optimization of likelihood and robustness analysis
We considered simultaneously the dependence of speciation on trait states (i.e., CR, CTD and ETD), the evolution of the trait (seven different transition models) and the mode of speciation (eight modes of inheritance), leading to 168 different model combinations. We maximized the likelihood for each model to find the best-fitting parameters (i.e., rates of speciation, extinction and transition). We used three sets of initial parameter values in the
3
Figure 3. Maximum Clade Credibility tree (left) reconstructed with two mitochondrial and three nucleargenes. Error bars indicate the 95% posterior distribution of the branching times. Overall uncertainty in the tree is shown by the densitree representation (right).
Among the 168 models we compared (see Table 1 for the best 30 models), we found the highest support for an ETD model that 1) assumes a dual inheritance mode of speciation (i.e., the parental state is inherited during speciation) and 2) does not allow direct transitions between shallow- and deep-water states, 3) assumes that the remaining four transition rates are all different from one another (constrained model with four rates). The estimated speciation rates were λ = 0.256 for shallow-water species, λ = 0.295 for deep-water species and λ = 1.46 for generalist species. Because the speciation rates for shallow- and deep-water species were similar, we further tested whether a simpler model would perform better by setting these two rates as equal (i.e., a model with 2 speciation rates, one for specialists and one for generalists) and re-ran the entire set of 168 models for this setting. The best performing model then becomes an ETD model where the speciation rate is the same for the specialist states, the shift between specialist states is not possible (constrained model with four rates) and speciation happens through a dual inheritance mode (AIC weight = 0.269; Table 1,Table S1).
Figure 3. Maximum Clade Credibility tree (left) reconstructed with two mitochondrial and three nuclear genes. Error bars indicate the 95% posterior distribution of the branching times. Overall uncertainty in the tree is shown by the densitree representation (right).
Among the 168 models we compared (see Table 1 for the best 30 models), we found the highest support for an ETD model that 1) assumes a dual inheritance mode of speciation (i.e., the parental state is inherited during speciation) and 2) does not allow direct transitions between shallow- and deep-water states, 3) assumes that the remaining four transition rates are all different from one another (constrained model with four rates). The estimated speciation rates were λ = 0.256 for shallow-water species, λ = 0.295 for deep-water species and λ = 1.46 for generalist species. Because the speciation rates for shallow- and deep-water species were similar, we further tested whether a simpler model would perform better by setting these two rates as equal (i.e., a model with 2 speciation rates, one for specialists and one for generalists) and re-ran the entire set of 168 models for this setting. The best performing model then becomes an ETD model where the speciation rate is the same for the specialist states, the shift between specialist states is not possible (constrained model with four rates) and speciation happens through a dual inheritance mode (AIC weight = 0.269; Table 1,Table S1).
Support for ETD models was generally high, as all the ETD models put together (regardless of speciation mode and transition type) summed up to an AIC weight of 0.5702 (note that our set of models is balanced; for every ETD model there is a corresponding CTD model). Likewise, we found high general support for the dual inheritance mode of speciation: the AIC weights of all the models with this speciation mode summed up 0.542. Finally, the sum of AIC weights for all the models with constrained transitions (i.e., no direct transitions between the shallow- and deep-water states), with different rates to and from the generalist state, was 0.756. Models with trait-dependent extinction did not perform better than models with trait-trait-dependent speciation. Likewise, models where examined and concealed traits had different transition rates did not perform better than models where these rates were identical. The best model shows that transitions between the generalist state and the deep-water state are much more likely than between the generalist state and the shallow-water state: the transition from deep-water to generalist state and the reverse have a per-lineage rate of 1.7 and 1.2 respectively, which is around 10 times higher than the transition rate from generalist to shallow-water (0.12; Figure 4). Lineages are very unlikely to shift from a shallow-water state to a generalist state (< 0.0001).
Figure 4. Estimates of rates of speciation (λ), extinction (μ) and transition across states (q) for the best
supported model (see Table 1). Our analysis provides four main insights: 1) there is strong evidence that the trait state affects the diversification process, 2) both specialists (shallow- and deep-water species) have similar speciation rates, 3) during speciation, both daughter species inherit the parental state and 4) specialist species can only switch to another specialist state in two steps (via the generalist state). We show the rates of speciation for generalist (A) and specialist species (B) as well as the distribution of frequencies of the estimates for 100 simulated trees under the best supported model. The switch from one state to another takes place between speciation events (i.e., along the branch of a phylogeny) with four different rates (C). All states have the same rate of extinction (μ). S = Shallow-water specialist; D = deep-water specialist; G = water-depth generalist.
With the simulation-inference procedure, we found that when trait-dependent diversification is the generating model (i.e., ETD), our analysis identified ETD as the best model in 96% of the simulations, Indicating that our method has high power. However,
Support for ETD models was generally high, as all the ETD models put together (regardless of speciation mode and transition type) summed up to an AIC weight of 0.5702 (note that our set of models is balanced; for every ETD model there is a corresponding CTD model). Likewise, we found high general support for the dual inheritance mode of speciation: the AIC weights of all the models with this speciation mode summed up 0.542. Finally, the sum of AIC weights for all the models with constrained transitions (i.e., no direct transitions between the shallow- and deep-water states), with different rates to and from the generalist state, was 0.756. Models with trait-dependent extinction did not perform better than models with trait-trait-dependent speciation. Likewise, models where examined and concealed traits had different transition rates did not perform better than models where these rates were identical. The best model shows that transitions between the generalist state and the deep-water state are much more likely than between the generalist state and the shallow-water state: the transition from deep-water to generalist state and the reverse have a per-lineage rate of 1.7 and 1.2 respectively, which is around 10 times higher than the transition rate from generalist to shallow-water (0.12; Figure 4). Lineages are very unlikely to shift from a shallow-water state to a generalist state (< 0.0001).
Figure 4. Estimates of rates of speciation (λ), extinction (μ) and transition across states (q) for the best
supported model (see Table 1). Our analysis provides four main insights: 1) there is strong evidence that the trait state affects the diversification process, 2) both specialists (shallow- and deep-water species) have similar speciation rates, 3) during speciation, both daughter species inherit the parental state and 4) specialist species can only switch to another specialist state in two steps (via the generalist state). We show the rates of speciation for generalist (A) and specialist species (B) as well as the distribution of frequencies of the estimates for 100 simulated trees under the best supported model. The switch from one state to another takes place between speciation events (i.e., along the branch of a phylogeny) with four different rates (C). All states have the same rate of extinction (μ). S = Shallow-water specialist; D = deep-water specialist; G = water-depth generalist.
With the simulation-inference procedure, we found that when trait-dependent diversification is the generating model (i.e., ETD), our analysis identified ETD as the best model in 96% of the simulations, Indicating that our method has high power. However,
3
ETD was incorrectly selected in 21% of datasets simulated under the best CTD model(AIC weight = 0.016). The extinction rate which was estimated for this CTD model and used for simulation was quite high (λ = 0.338, 1.453 and μ = 0.161), which may be partly responsible for this high type I error rate (see Discussion).
We found that the best supported model for the consensus tree was ranked first or second in 92 out of 100 of trees taken from the Bayesian posterior (Figure 5).
Figure 5. AIC weights comparing seven strongly supported models of diversification across 100 trees taken
from the posterior distribution. The rightmost bar shows the support for the models using the consensus tree. Shades of green represent different assumptions of Examined trait-dependent model of diversification (ETD) whereas shades of gray show models where speciation rate is the same over the time and across species (CR).
We note that our optimization procedure found higher likelihoods and AIC weights (for CTD models that assume different modes of speciation across states: one state speciates with dual inheritance and dual asymmetric transitions whereas the other states speciate through dual inheritance only) than the ones reported above, but the parameter estimates were biologically completely unrealistic (speciation rates of over 8,000), so we dismissed these.
DISCUSSION
In this study we explored the evolution of depth range and depth range specialization in cichlid fish, and their relationships with speciation and extinction rates. We found that shifts in water depth range occur frequently and that depth specialization strongly varies with rates of speciation: generalist species have higher rates of speciation than species that are specialized to either deep or shallow water. We also found that speciation events do not coincide with shifts in water depth range; instead depth preferences change along
the lifetime of lineages. A potential explanation for the absence of water depth changes during speciation this is that the duration of a lineage transition to a different depth may be short making depth segregation brief, thereby minimizing the chances of reproductive isolation and species divergence. Information on how fast species change depth preferences is not well documented, but it could be as fast as within 50 generations in the case of the ancestor of Astatotilapia in Lake Chala (Moser et al. 2018). The absence of water depth changes during speciation seems to suggest that speciation in sympatry, i.e. at the same depth indicating that speciation occurs on ecological, /morphological or behavioral axes. However, allopatric speciation is still possible if populations at the same water depth but different locations in the lake diverge.
We found that depth range generalists have around six times higher rates of speciation than depth range specialists and this is robust to different assumptions of the models. This could be explained by differential capacities of horizontal range expansion.
Lamprologini species are mostly substrate-bound and are not likely to venture into open
waters which poses an important barrier to dispersal. For shallow-water rock specialists for instance, the expansion of range within a lake is limited to dispersal along the rocky shore, whereas generalist species can use the benthic zone of the lake to move anywhere in the lake. One can imagine that a shallow water species could expand its range by colonizing a distant shore and, in order to do this, it has to travel all the lake’s perimeter. This, however, could not easily be accomplished by some species as river mounds or other stretches of unsuitable habitat along shores are actual barriers for dispersal (Markert et al. 1999; Steenberge et al. 2018). On the other hand, generalist species can move vertically to reach a zone where lake’s perimeter is smaller and progress horizontally until reaching the opposite shore and to move towards the surface; this increases the chances of successful dispersal and therefore range expansion. Thus species that are adapted to a wide range of depths may not be restricted in their geographical range which may lead to increased chance for speciation by encountering new habitats or being exposed to different abiotic conditions (Gaston 1998). Elevated speciation rates in depth range generalists may reflect a more general phenomenon that ecological generalists are more likely to successfully colonise new environments (Marvier et al. 2004). This advantage in dispersal might especially be important for cichlids as they lack a highly mobile larval stage (Ribbink et al. 1983; Van Oppen et al. 1997).
The robustness analysis showed that the evidence for trait-dependent diversification is strong because we were able to correctly detect it in 96% of our simulations where it was present. When simulations were performed under CTD, ETD was falsely detected in 21% of the simulations, which is a relatively high type I error rate. One possible explanation is that extinction under the CTD model was estimated to be rather high (μ = 0.161) making simulated trees contain a relatively large number of missing lineages (i.e., extinct species) which leads to poor performance in likelihood computations. For instance, Herrera-Alsina
3
et al. (2018) reported that with zero extinction, SecSSE has a Type I error rate of 11%but for an extinction rate of 0.1, the Type I error increases to 18%. ETD models are found as the most likely scenario when the analysis is repeated on slightly different phylogenetic reconstructions (i.e., trees from the posterior distribution) which adds support for our conclusions on depth-dependent speciation in Lamprologini and robustness of our phylogeny. Moreover, analyses of these alternative trees confirm our conclusion that speciation events are independent from shifts in depth range.
We found that there is an important asymmetry in rates of change across states. The shift from a deep-water specialist state to the generalist state is much more frequent than the shift from a shallow-water specialist state to the generalist state. This implies that once a lineage attains the shallow-water specialist state, it is very unlikely to switch to a different state, suggesting a macro-evolutionary endpoint. This does not lead to an accumulation of shallow-water species because the speciation rates in this state are relatively low, and the transition rate to this state is not very high. Transitions from and to a deep-water specialist state are more frequent. Although the process of adaptation to a different depth zone is multifactorial, we point at two possible candidates to explain why transitions to shallow water are relatively rare. First, adaptation to a new thermal environment requires several physiological changes (Porcelli et al. 2015), and specifically, the adaptation to the warm temperature of shallow water as well as the daily fluctuations in temperature could be particularly challenging. For instance, in a gradient of thermal tolerances, phenotypic plasticity in ectotherms at the upper limit of the gradient (high temperature) is much smaller than at the lower limit (Araujo et al. 2013). The struggle of species to develop adaptations that allow them to thrive (only) in warmer areas, could therefore be responsible for the low transition rates towards shallow water. Second, a high density of individuals has been documented in shallow waters (Theres et al. 2012) and this could lead to intense competition. If this is the case, an ecological limit would prevent lineages from colonizing waters close to the surface. While this suggests that the low transition rate to shallow water is due to the high degree of specialization required in both ecology and physiology, it does not explain why the reverse shift is very unlikely. Perhaps, the shift to shallow water comes at a cost and fish specializing to shallow water also undergo additional irreversible modifications that make them unfit for different water depths. For instance, visual adaptation to a well-lit environment could be associated with the loss of rod cells (i.e., rod cell density, Hunt et al. 2015) which can be difficult to regain in order to transit to a poorer-lit water band.
In summary, our study demonstrates that the evolution of a trait (e.g., depth range) can influence speciation rates, but occurs independently of the speciation events themselves. These findings show that mechanisms responsible for anagenesis and cladogenesis are not necessarily the same or the extent of their contributions differs.
ACKNOWLEDGEMENTS
This work was financially supported by Consejo Nacional de Ciencia y Tecnologia (CVU 385304 L. H.- A.) and the Netherlands Organisation for Scientific Research (NWO-VICI grant awarded to R.S.E.).This manuscript was enriched by constant discussions with members of Theoretical & Evolutionary Community Ecology. We thank the Center for Information Technology of the University of Groningen for their support and for providing access to the Peregrine high performance computing cluster.
REFERENCES
Araujo M., Ferri-Yañez F., Bozinovic F., Marquet P.A., Valladares F., Chown S.L. 2013. Heat freezes niche evolution. Ecol. Lett. 16:1206–1219.
Bailey R., Stewart J. 1977. Cichlid fishes from lake Tanganyika: additions to the zambian fauna including two new species. Occas. Pap. MUSEUM Zool. Univ. MICHIGAN.:1:30. Beaulieu J.M., O’Meara B.C. 2016. Detecting hidden diversification shifts in models of trait-dependent speciation and extinction. Syst. Biol. 65:583–601.
Bouckaert R.R., Drummond A.J. 2017. bModelTest : Bayesian phylogenetic site model averaging and model comparison. BMC Evol. Biol.:DOI 10.1186/s12862-017-0890-6. Bouckaert R.R., Heled J., Kuhnert D., Vaughan T., Wu C., Xie D., Suchard M.A., Rambaut A., Drummond A.J. 2014. BEAST 2 : A Software Platform for Bayesian Evolutionary Analysis. PloS Comput. Biol. 10:1–6.
Boughman J.W. 2002. How sensory drive can promote speciation. Trends Ecol. Evol. 17:571–577.
Capella-Gutiérrez S., Silla-martínez J.M., Gabaldón T. 2009. trimAl : a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics. 25:1972–1973. Carleton K.L., Dalton B.E., Escobar-camacho D., Nandamuri S.P. 2016. Proximate and Ultimate Causes of Variable Visual Sensitivities : Insights From Cichlid Fish Radiations. Genesis. 54:299–325.
Cornell H. V. 2013. Is regional species diversity bounded or unbounded? Biol. Rev. 88:140–165.
Fitzjohn R.G. 2012. Diversitree: Comparative phylogenetic analyses of diversification in R. Methods Ecol. Evol. 3:1084–1092.
3
transformation. Philos. Trans. R. Soc. B Biol. Sci. 353:219–230.Genner M.J., Seehausen O., Cleary D.F.R., Knight M.E., Michel E., Turner G.F. 2014. How Does the Taxonomic Status of Allopatric Populations Influence Species Richness within African Cichlid Fish Assemblages? J. Biogeogr. 31:93–102.
Goldberg E.E., Igić B. 2012. Tempo and mode in plant breeding system evolution. Evolution (N. Y). 66:3701–3709.
Hata H., Ochi H. 2016. Depth and substratum differentiations among coexisting herbivorous cichlids in Lake Tanganyika. R. Soc. Open Sci. 3:1:10.
Hata H., Shibata J., Omori K., Kohda M., Hori M. 2015. Depth segregation and diet disparity revealed by stable isotope analyses in sympatric herbivorous cichlids in Lake Tanganyika. Zool. Lett.:1–11.
Heled J., Drummond A.J. 2010. Bayesian Inference of Species Trees from Multilocus Data Research article. Mol. Biol. Evol. 27:570–580.
Herrera-Alsina L., van Els P., Etienne R.S. 2018. Detecting the Dependence of Diversification on Multiple Traits from Phylogenetic Trees and Trait Data. Syst. Biol. 0:1– 12.
Hoekstra L.A., Siddiq M.A., Montooth K.L. 2013. Pleiotropic Effects of a Mitochondrial – Nuclear Incompatibility Depend upon the Accelerating Effect of Temperature in Drosophila. Genetics. 195:1129–1139.
Hunt D.E., Rawlinson N.J.F., Thomas G.A., Cobcroft J.M. 2015. Investigating photoreceptor densities, potential visual acuity, and cone mosaics of shallow water, temperate fish species. Vision Res. 111:13–21.
Janzen T., Alzate A., Muschick M., Maan M.E., Plas F. Van Der, Etienne R.S. 2017. Community assembly in Lake Tanganyika cichlid fish : quantifying the contributions of both niche- based and neutral processes. Ecol. an.:1–11.
Karvonen A., Lundsgaard-hansen B., Jokela J., Seehausen O. 2012. Differentiation in parasitism among ecotypes of whitefi sh segregating along depth gradients. Oikos. 122:122–128.
Karvonen A., Wagner C.E., Selz O.M., Seehausen O.L.E. 2018. Divergent parasite infections in sympatric cichlid species in Lake. J. Evol. Biol. 31:1313–1329.
Katoh K., Standley D.M. 2013. MAFFT Multiple Sequence Alignment Software Version 7 : Improvements in Performance and Usability Article Fast Track. Mol. Biol. Evol. 30:772–780.
Keller I., Seehausen O. 2012. Thermal adaptation and ecological speciation. Mol. Ecol.
21:782–799.
Kullander S.O., Karlsson M., Karlsson M. 2014a. Chalinochromis cyanophleps, a new species of cichlid fish (Teleostei: Cichlidae) from Lake Tanganyika. Zootaxa. 3790:425– 438.
Kullander S.O., Norén M., Karlsson M., Karlsson M. 2014b. Description of Neolamprologus timidus , new species , and review of N . furcifer from Lake Tanganyika ( Teleostei : Cichlidae ). Ichthyol. Explor. Freshwaters. 24:301–328.
Kuwamura T. 1986. Laboratory of Biology , Faculty of Liberal Arts , only for Asprotilapia ( Haplochromis ) burtoni , a Mouthbrooders. J. Ethol.:129:146.
Maan M.E., Sefc K.M. 2013. Colour variation in cichlid fish : Developmental mechanisms , selective pressures and evolutionary consequences. Semin. Cell Dev. Biol. 24:516–528. Maddison W.P., Midford P.E., Otto S.P. 2007. Estimating a binary character’s effect on speciation and extinction. Syst. Biol. 56:701–710.
Mank J.E. 2007. Mating preferences, sexual selection and patterns of cladogenesis in ray-finned fishes. J. Evol. Biol. 20:597–602.
Markert J.A., Arnegard M.E., Danley P.D., Kocher T.D. 1999. Biogeography and population genetics of the Lake Malawi cichlid Melanochromis auratus: Habitat transience, philopatry and speciation. Mol. Ecol. 8:1013–1026.
Marvier M., Kareiva P., Neubert M.G. 2004. Habitat Destruction , Fragmentation , and Disturbance Promote Invasion by Habitat Generalists in a Multispecies Metapopulation SPECIES TEND TO BE HABITAT. Risk Anal. 24:869–878.
McMahan C.D., Chakrabarty P., Sparks J.S., Smith W.M.L., Davis M.P. 2013. Temporal patterns of diversification across global cichlid biodiversity (Acanthomorpha: Cichlidae). PLoS One. 8.
Meyer B.S., Matschiner M., Salzburger W. 2016. Disentangling Incomplete Lineage Sorting and Introgression to Refine Species-Tree Estimates for Lake Tanganyika Cichlid Fishes. Syst. Biol. 66:531–550.
Moser F.N., Van Rijssel J.C., Mwaiko S., Meier J.I., Ngatunga B., Seehausen O. 2018. The onset of ecological diversification 50 years after colonization of a crater lake by haplochromine cichlid fishes. Proc. R. Soc. B Biol. Sci. 285.
Muschick M., Indermaur A., Salzburger W. 2012. Report Convergent Evolution within an Adaptive Radiation of Cichlid Fishes. Curr. Biol. 22:2362–2368.
Nagai H., Terai Y., Sugawara T., Imai H., Nishihara H., Hori M., Okada N. 2011. Reverse evolution in RH1 for adaptation of cichlids to water depth in Lake Tanganyika. Mol. Biol.
3
Evol. 28:1769–1776.Nakai K., Yanagisawa Y., Sato T., Niimurat Y., Gashagaza M.M. 1990. Lunar synchronization of spawning in cichlid fishes of the tribe Lamprologini in Lake Tanganyika. J. Fish Biol.:589–598.
Van Oppen M.J.H., Turner G.F., Rico C., Deutsch J.C., Ibrahim K.M., Robinson R.L., Hewitt G.M. 1997. Unusually fine-scale genetic structuring found in rapidly speciating Malawi cichlid fishes. Proc. R. Soc. B Biol. Sci. 264:1803–1812.
Pearse W.D., Purvis A. 2013. phyloGenerator : an automated phylogeny generation tool for ecologists. Methods Ecol. Evol. 4:692–698.
Poll M. 1956. Exploration hydrobiologique du lac Tanganika (1946-1947)Vol. III, fasc. 5 B. Inst. R. des Sci. Nat. Belgique. III.
Porcelli D., Butlin R.K., Gaston K.J., Joly D., Snook R.R. 2015. The environmental genomics of metazoan thermal adaptation. Heredity (Edinb). 114:502–514.
Rabosky D.L., Goldberg E.E. 2015. Model inadequacy and mistaken inferences of trait-dependent speciation. Syst. Biol. 64:340–355.
Rabosky D.L., Huang H. 2016. A Robust Semi-Parametric Test for Detecting Trait-Dependent Diversification. Syst. Biol. 65:181–193.
Recknagel H., Elmer K.R., Meyer A. 2014. Crater lake habitat predicts morphological diversity in adaptive radiations of cichlid fishes. Evolution (N. Y). 68:2145–2155.
Ribbink A.J., Marsh A.C., Ribbink A.C., Sharp B.J. 1983. A preliminary survey of the cichlid fishes of rocky habitats in Lake Malawi. South African J. Zool. 18:149–310. Seehausen O. 2015. Process and pattern in cichlid radiations – inferences for understanding unusually high rates of evolutionary diversification. New Phytol. 207:307– 312.
Smith A.R., D’Annunzio L., Smith A.E., Sharma A., Hofmann C.M., Marshall N.J., Carleton K.L. 2011. Intraspecific cone opsin expression variation in the cichlids of Lake Malawi. Mol. Ecol. 20:299–310.
Snoeks J., Ruber L., Verheyen E. 1994. The Tanganyika problem: comments on the taxonomy and distribution patterns of its cichlid fauna. arch. hydrobiol. beih. ergebn. limnol.:355–472.
Steenberge M. Van, André J., Raeymaekers M., Hablützel P.I., Pieterjan M., Vanhove M., Koblmüller S., Snoeks J. 2018. Delineating species along shifting shorelines : Tropheus ( Teleostei , Cichlidae ) from the southern subbasin of Lake Tanganyika. Front. Zool. 15:1–18.
Sturmbauer C., Salzburger W., Duftner N., Schelly R., Koblmüller S. 2010. Evolutionary history of the Lake Tanganyika cichlid tribe Lamprologini ( Teleostei : Perciformes ) derived from mitochondrial and nuclear DNA data. Mol. Phylogenet. Evol. 57:266–284. Teske P.R., Sandoval-castillo J., Golla T.R., Emami-khoyi A., Tine M., Heyden S. Von Der, Beheregaray L.B. 2019. Thermal selection as a driver of marine ecological speciation. Proc. R. Soc. B Biol. Sci. 286.
Theres M., Marius D., Adrian R., Marco I., Colombo M., Gschwind M., Keller I., Kovac R., Barluenga M., Muschick M., Salzburger W. 2012. Depth-dependent abundance of Midas Cichlid fish ( Amphilophus spp .) in two Nicaraguan crater lakes Depth-dependent abundance of Midas Cichlid fish ( Amphilophus spp . ) in two Nicaraguan crater lakes. Hydrobiologia. 686:277:285.
Verburg P., Bills R. 2007. Two new sympatric species of Neolamprologus (Teleostei: Cichlidae) from Lake Tanganyika, East Africa. Zootaxa.:1:21.
Verne J. 1870. 20000 leagues under the sea. .
Wagner C.E., Harmon L.J., Seehausen O. 2014. Cichlid species-area relationships are shaped by adaptive radiations that scale with area. Ecol. Lett. 17:583–592.
Yokoyama S., Yokoyama R. 1996. Adaptive Evolution of Photoreceptors and Visual Pigments in Vertebrates. Annu. Rev. Ecol. Syst. 27:543–567.
3
SUPPLEMENTARY MATERIALAppendix 1. GenBank accession numbers and references for sequence information. Our full dataset consisted of two mitochondrial genes: the NADH dehydrogenase subunit 2 (ND2 gene, sequences from (Clabaut et al., 2005; Day et al., 2007; Duftner et al., 2005; Klett and Meyer, 2002; Koblmüller et al., 2016, 2007; Kocher et al., 1995; Kullander et al., 2014; Mabuchi et al., 2007; Morita et al., 2014; O’Quin et al., 2010; Schelly et al., 2006; Schwarzer et al., 2009; Shirai et al., 2014; Sturmbauer et al., 2010; Wagner et al., 2012, 2009; Weiss et al., 2015) and the cytochrome b (cytb) gene (sequences from (Genner et al., 2007; Kullander et al., 2014; Mabuchi et al., 2007; Matschiner et al., 2016, 2011; Morita et al., 2014; Nevado et al., 2009; O’Quin et al., 2010; Salzburger et al., 2002; Shirai et al., 2014; Takahashi et al., 2007, 2009; Wagner et al., 2009) and three nuclear genes: the recombinase activating protein 1 (rag1, sequences from (Clabaut et al., 2005; Friedman et al., 2013; Koblmüller et al., 2016; Kullander et al., 2014; Meyer et al., 2016; Nevado et al., 2009; Shirai et al., 2014), the ribosomal protein S7 (rps7, sequences from Schelly et al. 2006; Meyer et al. 2016)) gene and the rod opsin gene (RH1, sequences from (Meyer et al., 2015; Nagai et al., 2011; Spady et al., 2005; Sugawara et al., 2002). GenBank access numbers for the used sequences can be found in the table below. References
Clabaut, C., Salzburger, W., Meyer, A., 2005. Comparative phylogenetic analyses of the adaptive radiation of Lake Tanganyika cichlid fish: nuclear sequences are less homoplasious but also less informative than mitochondrial DNA. J. Mol. Evol. 61, 666– 81. doi:10.1007/s00239-004-0217-2
Day, J.J., Santini, S., Garcia-Moreno, J., 2007. Phylogenetic relationships of the Lake Tanganyika cichlid tribe Lamprologini: the story from mitochondrial DNA. Mol. Phylogenet. Evol. 45, 629–42. doi:10.1016/j.ympev.2007.02.025
Duftner, N., Koblmüller, S., Sturmbauer, C., 2005. Evolutionary relationships of the Limnochromini, a tribe of benthic deepwater cichlid fish endemic to Lake Tanganyika, East Africa. J. Mol. Evol. 60, 277–289. doi:10.1007/s00239-004-0017-8
Friedman, M., Keck, B.P., Dornburg, A., Eytan, R.I., Martin, C.H., Hulsey, C.D., Wainwright, P.C., Near, T.J., 2013. Molecular and fossil evidence place the origin of cichlid fishes long after Gondwanan rifting. Proc. R. Soc. B Biol. Sci. doi:10.1098/rspb.2013.1733
Genner, M.J., Seehausen, O., Lunt, D.H., Joyce, D.A., Shaw, P.W., Carvalho, G.R., Turner, G.F., 2007. Age of cichlids: new dates for ancient lake fish radiations. Mol. Biol. Evol. 24, 1269–82. doi:10.1093/molbev/msm050
Klett, V., Meyer, A., 2002. What, if anything, is a Tilapia?-mitochondrial ND2 phylogeny
of tilapiines and the evolution of parental care systems in the African cichlid fishes. Mol. Biol. Evol. 19, 865–883. doi:10.1093/oxfordjournals.molbev.a004144
Koblmüller, S., Duftner, N., Sefc, K.M., Aibara, M., Stipacek, M., Blanc, M., Egger, B., Sturmbauer, C., 2007. Reticulate phylogeny of gastropod-shell-breeding cichlids from Lake Tanganyika--the result of repeated introgressive hybridization. BMC Evol. Biol. 7, 7. doi:10.1186/1471-2148-7-7
Koblmüller, S., Nevado, B., Makasa, L., van Steenberge, M., Vanhove, M.P.M., Verheyen, E., Sturmbauer, C., Sefc, K.M., 2016. Phylogeny and phylogeography of Altolamprologus: ancient introgression and recent divergence in a rock-dwelling Lake Tanganyika cichlid genus. Hydrobiologia 1–16. doi:10.1007/s10750-016-2896-2
Kocher, T.D., Conroy, J. a, McKaye, K.R., Stauffer, J.R., Lockwood, S.F., 1995. Evolution of NADH dehydrogenase subunit 2 in east African cichlid fish. Mol. Phylogenet. Evol. doi:10.1006/mpev.1995.1039
Kullander, S.O., Norén, M., Karlsson, M., Karlsson, M., 2014. Description of Neolamprologus timidus, new species, and review of N . furcifer from Lake Tanganyika (Teleostei: Cichlidae). Ichthyol.Explor.Freshwaters 24, 301–328.
Mabuchi, K., Miya, M., Azuma, Y., Nishida, M., 2007. Independent evolution of the specialized pharyngeal jaw apparatus in cichlid and labrid fishes. BMC Evol. Biol. 7, 10. doi:10.1186/1471-2148-7-10
Matschiner, M., Hanel, R., Salzburger, W., 2011. On the origin and trigger of the notothenioid adaptive radiation. PLoS One 6. doi:10.1371/journal.pone.0018911
Matschiner, M., Musilová, Z., Barth, J.M.I., Starostova, Z., Salzburger, W., Steel, M., Bouckaert, R., 2016. Bayesian Phylogenetic Estimation of Clade Ages Supports Trans-Atlantic Dispersal of Cichlid Fishes. Syst. Biol. doi:10.1093/sysbio/syw076
Meyer, B.S., Matschiner, M., Salzburger, W., 2015. A tribal level phylogeny of Lake Tanganyika cichlid fishes based on a genomic multi-marker approach. Mol. Phylogenet. Evol. 83, 56–71. doi:10.1016/j.ympev.2014.10.009
Meyer, B.S., Matschiner, M., Salzburger, W., 2016. Disentangling incomplete lineage sorting and introgression to refine species-tree estimates for Lake Tanganyika cichlid fishes. Syst. Biol. syw069, 1–20. doi:10.1093/sysbio/syw069
Morita, M., Awata, S., Yorifuji, M., Ota, K., Kohda, M., Ochi, H., 2014. Bower-building behaviour is associated with increased sperm longevity in Tanganyikan cichlids. J. Evol. Biol. 27, 2629–2643. doi:10.1111/jeb.12522
Nagai, H., Terai, Y., Sugawara, T., Imai, H., Nishihara, H., Hori, M., Okada, N., 2011. Reverse evolution in RH1 for adaptation of cichlids to water depth in Lake Tanganyika.
3
Mol. Biol. Evol. 28, 1769–1776. doi:10.1093/molbev/msq344Nevado, B., KoblmÜller, S., Sturmbauer, C., Snoeks, J., Usano-Alemany, J., Verheyen, E., 2009. Complete mitochondrial DNA replacement in a Lake Tanganyika cichlid fish. Mol. Ecol. 18, 4240–4255. doi:10.1111/j.1365-294X.2009.04348.x
O’Quin, K.E., Hofmann, C.M., Hofmann, H.A., Carleton, K.L., 2010. Parallel Evolution of opsin gene expression in African cichlid fishes. Mol. Biol. Evol. 27, 2839–2854. doi:10.1093/molbev/msq171
Salzburger, W., Baric, S., Sturmbauer, C., 2002. Speciation via introgressive hybridization in East African cichlids? Mol. Ecol. 11, 619–25.
Schelly, R., Salzburger, W., Koblmüller, S., Duftner, N., Sturmbauer, C., 2006. Phylogenetic relationships of the lamprologine cichlid genus Lepidiolamprologus (Teleostei: Perciformes) based on mitochondrial and nuclear sequences, suggesting introgressive hybridization. Mol. Phylogenet. Evol. 38, 426–438. doi:10.1016/j.ympev.2005.04.023
Schwarzer, J., Misof, B., Tautz, D., Schliewen, U.K., 2009. The root of the East African cichlid radiations. BMC Evol. Biol. 9, 186. doi:10.1186/1471-2148-9-186
Shirai, K., Inomata, N., Mizoiri, S., Aibara, M., Terai, Y., Okada, N., Tachida, H., 2014. High prevalence of non-synonymous substitutions in mtDNA of cichlid fishes from Lake Victoria. Gene 552, 239–245. doi:10.1016/j.gene.2014.09.039
Spady, T.C., Seehausen, O., Loew, E.R., Jordan, R.C., Kocher, T.D., Carleton, K.L., 2005. Adaptive molecular evolution in the opsin genes of rapidly speciating cichlid species. Mol. Biol. Evol. 22, 1412–1422. doi:10.1093/molbev/msi137
Sturmbauer, C., Salzburger, W., Duftner, N., Schelly, R., Koblmüller, S., 2010. Evolutionary history of the Lake Tanganyika cichlid tribe Lamprologini (Teleostei: Perciformes) derived from mitochondrial and nuclear DNA data. Mol. Phylogenet. Evol. 57, 266–84. doi:10.1016/j.ympev.2010.06.018
Sugawara, T., Terai, Y., Okada, N., 2002. Natural Selection of the Rhodopsin Gene During the Adaptive Radiation of East African Great Lakes Cichlid Fishes. Mol. Biol. Evol. 19, 1807–1811. doi:10.2967/jnumed.112.106294
Takahashi, R., Watanabe, K., Nishida, M., Hori, M., 2007. Evolution of feeding specialization in Tanganyikan scale-eating cichlids: A molecular phylogenetic approach. BMC Evol. Biol. 7, 1–11. doi:10.1186/1471-2148-7-195
Takahashi, T., Watanabe, K., Munehara, H., RÜber, L., Hori, M., 2009. Evidence for divergent natural selection of a Lake Tanganyika cichlid inferred from repeated radiations in body size. Mol. Ecol. 18, 3110–3119. doi:10.1111/j.1365-294X.2009.04248.x
Wagner, C.E., Harmon, L.J., Seehausen, O., 2012. Ecological opportunity and sexual selection together predict adaptive radiation. Nature 487, 366–9. doi:10.1038/nature11144
Wagner, C.E., McIntyre, P.B., Buels, K.S., Gilbert, D.M., Michel, E., 2009. Diet predicts intestine length in Lake Tanganyika’s cichlid fishes. Funct. Ecol. 23, 1122–1131. doi:10.1111/j.1365-2435.2009.01589.x
Weiss, J.D., Cotterill, F.P.D., Schliewen, U.K., 2015. Lake Tanganyika — A ’ Melting Pot ’ of Ancient and Young Cichlid Lineages ( Teleostei : Cichlidae )? 1–29. doi:10.5061/dryad.8k0g5.Funding
