Lake level fluctuations and divergence of cichlid fish ecomorphs in Lake Tanganyika
- 522 Downloads
Lake Tanganyika has undergone substantial climate-driven lake level fluctuations that have repeatedly changed the distribution and extent of habitat for endemic fishes. Here we consider whether patterns of population genetic structure and phenotypic divergence within the cichlid fish Telmatochromis temporalis have been affected by changing lake levels. The species has a large-bodied rock-living ecomorph and a small-bodied shell-living ecomorph, and both are found in close proximity in littoral habitats. Using mtDNA sequences we found that geographically distant (>50 km) populations within the southern lake region diverged approximately 130,000–230,000 years ago, suggesting that the regional genetic structure persisted through a low stand of over 400 m ~106,000 years ago that ended with a rise to present levels ~100,000 years ago. We also found signatures of large population expansions since this rise across the study region, suggesting that the populations positively responded to new habitat as lake levels rose to present levels. Finally, we found that geographically adjacent (<10 km) ecomorphs exhibit both significant genetic differentiation and signatures of gene flow after the lake level rise. The results suggest that local ecomorph divergence progressed with gene flow after the last major rise in lake level ~100,000, potentially facilitated by new ecological opportunities.
KeywordsEcological speciation Parallel evolution Population genetics Demographic history
There is increasing evidence that divergent natural selection operating on ecological traits plays a central role in adaptive radiation (Funk et al., 2006), and that the extent of adaptive radiation may be closely linked to the availability of niche space. New ecological opportunity has been inferred to be a trigger of rapid radiation in both experimental systems (Rainey & Travisano, 1998), and phylogenetic studies of natural systems, including plants (Hughes & Eastwood, 2006), lizards (Harmon et al., 2008; Mahler et al., 2010), mammals (Tran, 2014) and fishes (Rüber et al., 2003; Salzburger et al., 2005; Siwertsson et al., 2010; Wagner et al., 2012). A role for ecological opportunity in facilitating adaptive radiation is also supported by evidence that the rate of adaptive evolution diminishes when niches are filled (Phillimore & Price, 2008; Price et al., 2014). Thus, the availability of suitable habitat and food resources may be the key factors that allow lineages with appropriate genetic variation to undergo rapid speciation and adaptive evolution (Yoder et al., 2010).
Multiple African lakes contain adaptive radiations of cichlid fishes, and macro-evolutionary analyses suggest that the physical dimensions of these lakes are reliable predictors of the presence and extent of adaptive radiations within them (Wagner et al., 2012). Within these radiations species often differ dramatically in their depth distributions, substrate preferences and diet (Konings, 1998; Konings 2007). Moreover, these differences are often intrinsically correlated with differences in morphology (Rüber & Adams, 2001; Muschick et al., 2012), and breeding systems (Sefc, 2011). In an increasing number of cases the functional genes related to these traits have been identified (Sugawara et al., 2002; Gerrard & Meyer, 2007). Together, these patterns are suggestive of a strong role for ecologically mediated speciation in these lakes, in addition to the role of sexual selection (Wagner et al., 2012).
This study focuses on the Lake Tanganyika cichlid fish Telmatochromis temporalis Boulenger 1898, part of the species-rich tribe Lamprologini, which contains over 90 species. Of these, around 80 are endemic to Lake Tanganyika, while the remainder are restricted to the main Congo River system (Schelly et al., 2006). All species are brood-guarding substrate spawners, but species differ substantially in their habitat preferences, depth distributions and dietary preferences (Konings, 1998; Muschick et al., 2012). Our focal species, T. temporalis, has a lake-wide distribution and occurs in two distinct ecomorphs each associated with a distinct habitat type. The ‘normal’ sized rock ecomorph is abundant on the rocky shorelines, whereas the ‘dwarf’ shell ecomorph occurs on large aggregations of empty shells of the gastropod Neothauma tanganyicense, which is endemic to Lake Tanganyika (Takahashi, 2004). These shell beds are found throughout the lake, but are patchily distributed and comparatively less common than the rocky habitat (Takahashi et al., 2009). Multiple origins of the shell ecomorph have been suggested based on population genetic evidence from nuclear microsatellite loci and mtDNA sequences (Takahashi et al., 2009). This has been supported by additional population genetic evidence from nuclear AFLPs and mtDNA sequences by Winkelmann et al. (2014), alongside evidence that competition for breeding substrate mediates gene flow between ecomorphs.
In this study we investigate whether the existing population genetic structure and timing of divergence of T. temporalis ecomorphs has followed lake level rises. The lake has been subjected to large lake level changes since formation (Cohen et al., 1997; Scholz et al., 2003), including a major low stand of at least 435 m below present levels ~106,000 years ago, before a rise to current levels ~100,000 years ago (McGlue et al. 2008), and another less substantial low stand (~260 m) during the Last Glacial Maximum 32,000–14,000 years ago (McGlue et al., 2008). Population-level genetic studies have shown that these dramatic lake level changes have strongly influenced the population connectivity and demography of multiple Lake Tanganyika cichlids associated with hard substrates (Verheyen et al., 1996; Rüber et al., 1999; Sturmbauer et al., 2001; Duftner et al., 2006; Sefc et al., 2007; Koblmüller et al., 2011; Nevado et al., 2013).
Major lake level changes will have changed the extent and distribution of littoral habitat available for populations of both ecomorphs of T. temporalis. As water levels rose, new rock habitats will have been exposed and colonised, while new Neothauma shell habitats will have formed. Additionally, due to strong depth limits of this littoral species (maximum depth 28 m, LR and KW pers. obs.), populations will have been lost from former habitats with rising water levels. Thus, we suggest that the most recent lake level rises may have generated a new metapopulation structure over local geographic scales, and it is also plausible that ecomorph divergence took place following the rise in water levels to those of the present day.
Here we investigated the spatial and temporal context of T. temporalis ecomorph divergence by first quantifying genetic (mtDNA) differences within and between populations. We then investigated if habitat differences or geographic distances were the more reliable predictors of population genetic structuring. Next, we quantified migration between regions, and between ecomorphs within regions. Finally, we estimated the timing of divergence events between regions, and the timing of individual population expansions. Together these data are interpreted in relation to geologically inferred lake level rises.
Materials and methods
Sampling and laboratory methods
Population sample sizes (N) and mtDNA control region diversity
Ndole Bay (offshore)
Mitochondrial DNA sequencing
Genomic DNA was extracted using DNeasy Blood & Tissue Kit and a ~900 bp section of the mtDNA control region (D-loop) was amplified using the forward primer 5′-ARA GCR YCG GTC TTG TAA TCC G-′3 and reverse primer 5′-TGG CTA AAT TYA CAC ATG C-′3. Polymerase chain reaction (PCR) was performed in 25.4 µl reactions containing 0.2 µl Taq DNA polymerase (Bioline), 0.5 µl of each primer (10 µM each), 1 µl dNTPs (1 mM each dNTP), 3 µl MgCL (25 mM stock), 5 µl of 5X PCR reaction buffer, 14 µl double-distilled water and 1.2 µl of the extracted DNA. PCR used the following conditions: 3 min at 94°C, then 35 cycles of 1 min at 94°C, 1 min at 54°C and 1.5 min at 72°C, followed by 72°C for 5 min. Sequencing was performed using a Big Dye terminator v.1.1 on a 3730xl DNA Analyzer (Applied Biosystems).
Genetic differences among populations
Sequences were aligned using MAFFT 6.814b in Geneious Pro 5.5.6 (Biomatters Ltd., Auckland, New Zealand) using default settings, and the resulting alignment was checked by eye. The number of haplotypes (H), haplotype diversity (He) and nucleotide diversity (π) for each population were calculated in DNASP (Rozas, 2003). Genetic divergences between populations (ΦST.) were measured in Arlequin 3.5 (Excoffier & Lischer, 2010), and statistical significance was tested using 10,000 permutations. Analyses of Molecular Variance (AMOVA) in Arlequin 3.5 were used to quantify within and between-population genetic variance. For AMOVA analyses regions were defined as “northern” (populations A–E) and “southern” (populations I–P).
Importance of spatial and environmental variables for genetic variation of populations
To test for dependence of genetic distance (ΦST between populations) on geographic distance and environmental differences (habitat type, sampling depth), we used distance-based redundancy analysis (dbRDA; Geffen et al., 2004, Legendre & Fortin, 2010) using the capscale function in the R package “vegan” (Oksanen et al., 2013). Geographic information was coded as decimal latitude and longitude coordinates. Substrate was coded as 1 rock, 0 for shell and 0.5 for mixed substrate. Euclidean distances were used within the dbRDA for quantification of environmental and geographic distances. Significance was tested using 100,000 permutations.
Populations of the same ecomorph in the same region with no significant genetic differences were ‘pooled’, generating “northern” and “southern” region datasets for each ecomorph. Migrate-n 3.6.11 (Beerli and Felsenstein 2001) was used to simultaneously estimate the magnitude and direction of historical migration between pairs of the four groups (north rock, north shell, south rock, south shell). For each run we used Bayesian search strategy, and one long chain. In total we recorded 1 million steps separated by 100 step increments, following a burn-in of 10,000 trees. The Theta uniform prior range was 0–0.5, and the M uniform prior range was 0–1,000. Other search parameters were default. We conducted three runs allowing bi-directional migration between all combination of sites and ecomorphs. We also conducted one run allowing only migration between sites but not ecomorphs, and one run allowing only migration between ecomorphs but not sites. Relative likelihoods of models were compared using Bayes factors.
The timing of migration events was estimated using two mtDNA control region substitution rates previously estimated for African cichlids, namely 0.0324 changes per site per million years (Genner et al., 2010) and 0.057 changes per site per million years (e.g. Koblmüller et al., 2011), and a generation time of either 2 or 3 years, based on estimates from the Tanganyika lamprologine Neolamprologus modestus (Hellmann et al., 2015).
Timescale of splitting of the northern and southern population groups
Our analyses were consistent with no gene flow between the northern and southern populations. This enabled us to estimate the timing of the split between the populations using the *BEAST approach (Heled & Drummond, 2010) in BEAST 1.8.2. (Drummond et al., 2005). Each analysis was run for 25 million steps using HKY+Γ substitution model, with parameters logged every 1,000 generations. A strict molecular clock was employed, again employing the substitution rates 0.0324 and 0.057 changes per site per million years.
To detect historical changes in effective population sizes Bayesian skyline plots were calculated in BEAST (Drummond et al., 2005). Again, each analysis was run for 25 million steps using HKY+Γ substitution model, with parameters logged every 1000 generations. Again a strict molecular clock was employed, alongside the substitution rates of 0.0324 and 0.057 changes per site per million years. A coalescent Bayesian skyline tree prior was used between 4 and 10 grouped coalescent intervals, and a UPGMA starting tree. All other settings were as the default. Chain convergence and Bayesian skyline plots were both visualised in Tracer 1.5 (http://tree.bio.ed.ac.uk/software/tracer/). We used the Bayesian skyline plots to identify when population expansion towards the present day started, and the time of maximum population growth, following methods in Genner & Turner (2014).
Genetic diversity and population differentiation
Pairwise population differentiation (ΦST) between sixteen populations based on mtDNA sequences
Genetic variation, geographic distance and environmental distance
Analyses of molecular variance (AMOVA) on rock and shell ecomorph populations in northern and southern regions
Source of variation
Sum of squares
Variance of components
Percentage of variation
Between northern and southern regions
Among populations within regions
Between northern and southern regions
Among populations within regions
Between rock and shell ecomorph
Among populations within ecomorphs
Between rock and shell ecomorph
Among populations within ecomorphs
Tests of the association of genetic distance (ΦST) with geographic variables, substrate type and water depth using distance-based redundancy analysis
Marginal (all variables)
Conditional (latitude and longitude)
Conditional (depth and substrate)
Demographic history, population splitting times and gene flow
Summary of Bayesian skyline plot reconstructions of historic population demography
Neτ (upper 95 %)
Neτ (lower 95 %)
Start population expansion (Ma)*
Maximum population growth (Ma)*
Start population expansion (Ma)**
Maximum population growth (Ma)**
Migrate-n models allowing a full migration matrix (harmonic mean log-likelihood = −3863.0, average of 3 runs) were considerably more likely than the model that allowed migration between ecomorphs but not regions (harmonic mean log-likelihood = −5,365.4, log Bayes Factor = −3,004.75), and the model that allowed migration between regions but not ecomorphs (harmonic mean log-likelihood = −4,229.80, log Bayes Factor = −733.6).
Mutation-scaled effective population sizes (ϴ), as estimated in three replicate Bayesian runs of Migrate-n with a full migration matrix
θ (average ± SD of modal values from 3 runs)
K, N, L
Bayesian estimations of mutation-scaled migration rates (M), as estimated in three replicate runs of Migrate-n with a full migration matrix
Migration rate (M) population 2 to 1 (average ± SD of modal values from 3 runs)
Migration rate (M) population 2 to 1 (average ± SD of modal values from 3 runs)
Estimates of the average timing of all migration events as estimated in as the average timing of events across three replicate runs of Migrate-n, using the full migration matrix
Generation time (years)
Mean time (± standard deviation) of migration from 3 runs (years)
North Shell > North Rock
North Rock > North Shell
South Shell > South Rock
South Rock > South Shell
South Rock > North Rock
North Rock > South Rock
South Shell > North Shell
North Shell > South Shell
South Shell > North Rock
South Rock > North Shell
North Shell > South Rock
North Rock > South Shell
Using *BEAST and the substitution rate of 0.0324 change/site/Ma, we estimated that the divergence of the northern and southern populations took place 230,000 years ago (95% Highest Posterior Density intervals 163,000–303,000 years). Using the substitution rate of 0.057 change/site/Ma, we estimated that the divergence of the northern and southern populations took place 130,700 years ago (95% Highest Posterior Density intervals 92,600–172,200 years).
Major drivers of population genetic structuring
The results clearly demonstrate a strong signal of geographic structuring, consistent with expectations of limited dispersal among fragmented habitats within both ecomorphs of the species. This general spatial pattern is compatible with previous work on Lake Tanganyika rock cichlids, including representatives of the Tropheini (Wagner and McCune 2009), Eretmodini (Rüber et al., 2001; Taylor et al., 2001; Sefc et al., 2007), Perissodini (Koblmüller et al., 2009) and Lamprologini (Duftner et al., 2006). Selection presumably favours philopatry in these cichlids due to the benefits of persisting in local known environment relative to the costs of movement across unfamiliar and less structured habitat, such as sand or deep water.
In general, close associations between ecomorphology and breeding habitat can reduce gene flow and facilitate speciation (Edelaar et al., 2012; Webster et al., 2012, Malinsky et al., 2015). Although in T. temporalis the dominant factor affecting gene flow was geographic proximity, there was also evidence of restricted gene flow between ecomorphs in neighbouring habitats, similar to the findings of Takahashi et al. (2009) and Winkelmann et al. (2014). In this species, substrate use of individuals appears strongly linked to the availability of shelter and predation threat. Adults of both ecomorphs are believed to be highly vulnerable to multiple predators that characterise hard substrate environments of Lake Tanganyika (Takahashi et al., 2012), including piscivorous fish (catfishes, mormyrids, cichlids, mastacembelid eels), birds (kingfishers and cormorants), mammals (spotted-neck otters) and reptiles (water cobra) (Konings 1998). It has been found that body size matches available shelter size in T. temporalis, and that in transplant experiments rock ecomorphs are unable to make use of empty shells as shelter against predators (Takahashi et al., 2012). Thus, rock females may be unable to use shell habitat, while shell females could in principle use rock habitat, and this may explain the apparent greater migration from shell to rock habitat observed with maternally inherited mtDNA. Laboratory work suggests that competition is important in determining habitat use of this species, with large rock ecomorph cichlids forcing smaller individuals to use less favoured shell habitat (Winkelmann et al., 2014). Taken together, the evidence is consistent with natural selection operating against migrants with non-adapted phenotypes, and at least partially restricting gene flow.
Population divergence after a major lake level rise
On average populations tended to show pulses of maximum growth approximately 43,000 and 75,000 years ago, depending on the rates of molecular evolution employed. These results are consistent with T. temporalis ecomorphs undergoing expansions with gene flow after the major lake level rise that would have provided new expanses of littoral habitat for the geographically separate “northern” and “southern” population groups. The results are suggestive of the lake level rise providing the opportunity for the development of a new metapopulation structure and phenotypic divergence between ecomorphs driven by local selective pressure. Notably, the populations do not show clear influence of changes in effective population size during the low stand of approximately 260 m during the Last Glacial Maximum 32,000–14,000 years ago (McGlue et al., 2008), suggesting that genetic diversity was maintained in each region despite environmental changes.
Large lake level changes will have fragmented and reunified rock habitats, and altered the locations and extent of shell habitat. The distribution of this habitat will be dependent on availability of suitable substrate for living gastropods, whether hydrodynamic conditions are favourable for shell aggregation, the extent of bioturbation that maintains shell exposure, and the water chemistry that will influence rates of shell erosion. Individual Neothauma shells have been dated up to a maximum of 1,600 years of age (McGlue et al., 2010). However, we know very little of longevity of beds themselves. Nevertheless, it seems plausible that apparent lake level stability for the last 14,000 years, at least, has promoted the generation of a population genetic structure in T. temporalis influenced by both geographic proximity of populations and the nature of the substrate present.
In the East African Great Lakes, water-level fluctuations have been considered to act as species ‘pumps’ (Rossiter, 1995; Salzburger, 2009; Danley et al., 2012), with the changes repeatedly splitting populations and promoting phenotypic divergence in allopatry. An opposing view is that such lake level changes may alternatively act as species ‘dumps’, bringing together formerly separated populations in novel habitat, and leading to ‘reverse speciation’ (Seehausen, 2006; Taylor et al., 2006; Teotonio et al., 2009). The results of this study suggest an alternative perspective on the concept of the species pump. In addition to rising water levels leading to the evolution of new allopatric variants, they may also provide new opportunities for divergence in allopatric, parapatric or sympatric circumstances. Thus, we propose that changes to habitat availability, together with ecological stability over millennial timescales determines whether ecological speciation proceeds.
Associations between genetic and ecological divergence are dependent on spatial scale
Geography was a major predictor of genetic structuring over the spatial scale of the whole study area, while results suggest that habitat plays an additional role for population genetic structuring over more local scales. Therefore, it appears that the ability to detect associations between environmental contrasts and gene flow is strongly influenced by spatial scale in this species, and likely others where parallel evolution of ecomorphs has occurred. A meta-analysis of published studies has demonstrated the ubiquity of isolation-by-ecology in natural systems (Shafer & Wolf, 2013). However, while Shafer & Wolf considered correlations between geographic distance and ecological distances, the changing associations between genetic distances, geographic distances and ecological distances over increasing spatial scales were not explicitly studied. The most important factor governing such patterns is likely to be the dispersal abilities of the studied organism (Sexton et al., 2014). Organisms with large potential dispersal distances, for example birds, may have a strong signal of isolation-by-ecology over the range of hundreds of kilometres (Edelaar et al., 2012). By contrast lamprologine cichlids, which have very limited dispersal abilities, and exhibit clear potential for parallel evolutionary divergence, represent the alternative extreme where isolation-by-ecology must be studied locally.
In conclusion, our study suggests that metapopulation structure and phenotypic diversification followed changes in lake depth. Thus, in this case, lake level changes may have acted as a facilitator of adaptive diversification and contribute to local reproductive isolation of incipient species. Notably, lake depth is a key predictor of species richness in lacustrine cichlid radiations (Wagner et al., 2012), potentially because deeper lakes contain more ecological niches for species to diversify among. Our results hint at the possibility that lake level changes that characterise deep lakes have repeatedly provided new ecological opportunity in allopatric populations that permits diversifying selection. Evidence from neighbouring Lake Malawi would support this concept, as many phenotypically unique populations of littoral fishes and gastropods have been founded following major level rises over the last 90,000 years (Schultheiss et al., 2009; Genner & Turner, 2014).
We thank H. Phiri, D. Sinyinza, J.K. Zima, L. Makasa, R. Shapola and staff at the Lake Tanganyika Research Unit, Mpulungu of the Ministry of Livestock and Fisheries, Republic of Zambia for their field support. We thank H.H. Büscher for help in the field and W. Salzburger for arranging research permits. For comments on the manuscript, we thank S. Hiscock, S. Koblmüller and W. Salzburger. Work was funded by a Natural History Museum Ph.D studentship and the BBSRC Collaborative Scheme for Systematics Research (CoSyst) initiative.
- Danley, P. D., M. Husemann, B. Ding, L. M. Dipietro, E. J. Beverly & D. J. Peppe, 2012. The impact of the geologic history and paleoclimate on the diversification of East African cichlids. International Journal of Evolutionary Biology 2012: Article ID 574851.Google Scholar
- Koblmüller, S., W. Salzburger, B. Obermüller, E. Eigner, C. Sturmbauer & K. M. Sefc, 2011. Separated by sand, fused by dropping water: habitat barriers and fluctuating water levels steer the evolution of rock-dwelling cichlid populations in Lake Tanganyika. Molecular Ecology 20: 2272–2290.CrossRefPubMedGoogle Scholar
- Konings, A., 1998. Tanganyika cichlids in their natural habitat. Cichlid Press, Germany, St. Leon-Rot, Germany.Google Scholar
- Konings, A., 2007. Malawi cichlids in their natural habitat, 4th ed. Cichlid Press, El Paso, TX.Google Scholar
- McGlue, M. M., M. J. Soreghan, E. Michel, J. A. Todd, A. S. Cohen, J. Mischler, C. S. O’Connell, O. S. Castaneda, R. J. Hartwell, K. E. Lezzar & H. H. Nkotagu, 2010. Environmental controls of shell-rich facies in tropical lacustrine rifts: a view from Lake Tanganyika’s littoral. Palaios 25: 426–438.CrossRefGoogle Scholar
- Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens & H. Wagner, 2013. vegan: Community Ecology Package. R package version 2.0-10. http://CRAN.R-project.org/package=vegan
- Price, T. D., D. M. Hooper, C. D. Buchanan, U. S. Johansson, D. T. Tietze, P. Alstrom, U. Olsson, M. Ghosh-Harihar, F. Ishtiaq, S. K. Gupta, J. Martens, B. Harr, P. Singh & D. Mohan, 2014. Niche filling slows the diversification of Himalayan songbirds. Nature 509: 222–225.CrossRefPubMedGoogle Scholar
- Schelly, R., W. Salzburger, S. Koblmüller, N. Duftner & C. Sturmbauer, 2006. Phylogenetic relationships of the lamprologine cichlid genus Lepidiolamprologus (Teleostei: perciformes) based on mitochondrial and nuclear sequences, suggesting introgressive hybridization. Molecular Phylogenetics and Evolution 38: 426–438.CrossRefPubMedGoogle Scholar
- Siwertsson, A., R. Knudsen, K. Kahilainen, K. Praebel, R. Primicerio & P. A. Amundsen, 2010. Sympatric diversification as influenced by ecological opportunity and historical contingency in a young species lineage of whitefish. Evolutionary Ecology Research 12: 929–947.Google Scholar
- Tran, L. A. P. (2014.) The role of ecological opportunity in shaping disparate diversification trajectories in a bicontinental primate radiation. Proceedings of the Royal Society B: Biological Sciences 281: Article 20131979.Google Scholar
- Verheyen, E., L. Rüber, J. Snoeks & A. Meyer, 1996. Mitochondrial phylogeography of rock-dwelling cichlid fishes reveals evolutionary influence of historical lake level fluctuations of Lake Tanganyika, Africa. Philosophical Transactions of the Royal Society B: Biological Sciences 351: 797–805.CrossRefGoogle Scholar
- Winkelmann, K., M. J. Genner, T. Takahashi & L. Rüber, 2014. Competition-driven speciation in cichlid fish. Nature Communications 5: Article 3412.Google Scholar
- Yoder, J. B., E. Clancey, S. Des Roches, J. M. Eastman, L. Gentry, W. Godsoe, T. J. Hagey, D. Jochimsen, B. P. Oswald, J. Robertson, B. A. Sarver, J. J. Schenk & S. F. Spear, 2010. Ecological opportunity and the origin of adaptive radiations. Journal of Evolutionary Biology 23: 1581–1596.CrossRefPubMedGoogle Scholar