Understanding the causes of disparities in species diversity across taxonomic groups and regions is a fundamental aim in evolutionary biology. Addressing these questions is difficult because of the need for densely sampled phylogenies and suitable empirical systems.
Here we investigate the cichlid fish radiation of Lake Tanganyika and show that per lineage diversification rates have been more than six times slower than in the species flocks of Lakes Victoria and Malawi. The result holds even at peak periods of diversification in Lake Tanganyika, ruling out the age of the lake as an explanation for slow average rates, and is robust to uncertainties over the calibration of cichlid radiations in geological time. Moreover, Lake Tanganyika lineages, irrespective of different biological characteristics (e.g. sexually dichromatic versus sexually monochromatic clades), have diversified at similar rates, falling within typical estimates across a range of plant and animal clades. For example, the mostly sexually dichromatic haplochromines, which have speciated explosively in Lakes Victoria and Malawi, have displayed modest rates in Lake Tanganyika (where they are called Tropheini).
Our results show that either the Lake Tanganyika environment is less conducive for cichlid speciation or the remarkable diversifying abilities of the haplochromines were inhibited by the prior occupancy of older radiations. Although the results indicate a dominant role for the environment in shaping cichlid diversification, differences in the timing of diversification among the Tanganyikan tribes indicate that biological differences were still important for the dynamics of species build-up in the lake. While we cannot resolve the timing of the radiation relative to the origin of the lake, because of the lack of robust geological date calibrations for cichlids, our results are consistent with a scenario that the different clades reflect independent adaptive radiations into different broad niches in the lake.
Citation: Day JJ, Cotton JA, Barraclough TG (2008) Tempo and Mode of Diversification of Lake Tanganyika Cichlid Fishes. PLoS ONE 3(3): e1730. https://doi.org/10.1371/journal.pone.0001730
Academic Editor: Oliver G. Pybus, University of Oxford, United Kingdom
Received: August 20, 2007; Accepted: February 5, 2008; Published: March 5, 2008
Copyright: © 2008 Day et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: JJD is supported by a Dorothy Hodgkin Royal Society fellowship.
Competing interests: The authors have declared that no competing interests exist.
Explaining why some regions and taxonomic groups contain more species than others is a key goal of evolutionary biology. Different categories of explanations have been offered. For example, if resources place a limit on the number of species in a region, then diversity might often arise by adaptive radiation, in which lineages diversify rapidly to occupy different niches following the invasion of a new region or a shift in habitat usage , . Alternatively, species richness may depend on rates of speciation and extinction: some organisms may evolve reproductive isolation more readily or have a lower risk of extinction than others, irrespective of any ecological limits on biomass . Factors determining net diversification rates could include biological traits , , the environment –, or the interaction between traits and the environment , .
One way to evaluate these alternatives is to explore the timing of diversification using molecular phylogenetics. For example, adaptive radiation predicts fast early diversification followed by slower rates of species-turnover once available niches have been filled . Similarly, comparisons among taxa and regions can be used to test the effects of biological traits versus the environment –. Do diversification rates tend to be more similar between related taxa in different environments, between unrelated taxa in the same environments, or do they depend on an interaction between traits and environment? Such comparisons remain rare because of the need for phylogenetic trees containing nearly all species in the regions and taxa of interest.
Here, we use a near-complete phylogenetic tree of cichlid fish species from Lake Tanganyika (LT) to test these ideas in a lake fauna. Cichlids in the East African great lakes have provided classic examples of endemic radiations. Each lake harbors a species flock of many hundred species displaying astonishing levels of ecological, phenotypic and behavioural diversity. Phylogenetic and geological evidence indicates exceptionally rapid radiation from single ancestral species over very short evolutionary timescales in Lake Victoria (LV), 447–535 species and Lake Malawi (LM), 451–600 species , based on age estimates assuming near or total desiccation for these lakes at 12.4 Ka  and 1 Mya - 0.57 Ka , respectively. The same tribe, the haplochromines, has radiated in both lakes, where they comprise the majority of cichlid species. A major cause of the explosive radiations appears to be rapid speciation through sexual selection associated with shifts in male breeding coloration and associated female preferences e.g. –.
The cichlid fauna of LT is also thought to have radiated rapidly , . However, LT represents a very different scenario regarding its geological history and cichlid fauna. Despite being the oldest rift lake (9–12 Mya) , and larger than LM, it contains only one third as many cichlid species (200 reported in this study). LT has experienced fluctuations in lake level , but unlike LM and LV, is not thought to have experienced near or complete desiccation. The fauna derives from several distinct lineages (classified in up to 16 tribes ) derived from several invasions rather than from a single ancestor . Many LT tribes are entirely sexually dichromatic, such as Cyprichromini, Bathybatini. Others are partially sexually dichromatic, such as Ectodini and Tropheini (the LT endemic haplochromines). Conversely, the most species rich lineage, Lamprologini (∼90–100 species) are sexually monochromatic. Therefore, LT provides the opportunity to compare the tempo of diversification among biologically distinct lineages occupying the same environment, and among members of the same lineage (the haplochromines) in different lakes.
In order to investigate patterns and rates of diversification within a clade, ideally nearly all the species should be sampled . Previous comparisons have been limited by incomplete sampling and by a lack of statistical evaluation of expected patterns under hypothesized scenarios. To date, a single study , using lineage-through-time (LTT) plots, based on published linearised trees of five tribes (Limnochromini, Perissodini, Cyprichromini, Bathybatini, Ectodini) has attempted to evaluate patterns of LT cichlid diversification. The LTT plots of all five tribes were interpreted to show a speciation burst around the same time followed by a period of stasis, but the two most species rich tribes were not included and the conclusion was made from graphical interpretation rather than statistical analysis.
Compiling mtDNA sequence data for 152 (76%) endemic species, plus selected species from elsewhere [Table S1], we reconstruct the timing of diversification of the LT species flock using a Bayesian relaxed molecular clock approach. We also consider the potential effects of missing species on estimates of the timing and rate of diversification. We show that the tempo and mode of cichlid diversification is highly contingent on the environment. LT cichlids have diversified six times more slowly than endemic radiations in the other lakes, even during peak episodes of diversification. This result is robust to uncertainties in age of the cichlid flocks. Diversification rates are remarkably similar among LT tribes considering 95% confidence limits, despite variation in general niche (e.g. littoral versus deep-water) and breeding system (e.g. sexually monochromatic versus dichromatic). We do not find strong evidence for an initial rapid radiation followed by a uniform slow-down in rates. Instead, faster diversification occurred at intermediate periods, possibly coinciding with periods of changing lake levels e.g. ,  or with successive invasions and separate adaptive radiations. The different tempo and mode of diversification in LT cichlids could reflect either its different physical environment or the constraining effects of earlier radiations on the remarkable diversifying abilities of the haplochromines.
Results and Discussion
Phylogeny and Age Estimates of the LT Cichlid Radiation
Maximum likelihood and Bayesian analyses recover trees with good support for relationships that are largely congruent with taxonomy (Figure S1, Text S1). The majority of species derive from a single well-supported clade (P, Figure S1), comprising two main lineages, the C-lineage , which contains several tribes including the Haplochromini, and a clade we call the L-lineage, which includes the Lamprologini and Eretmodini. The endemic Tanganyikan haplochromines, termed Tropheini, are closely related to the haplochromine radiations in LM and LV .
Dating cichlid radiations in geological time is problematic because of the lack of fossils and other robust sources of independent calibration. Because of this uncertainty, we adopt, in turn, two widely different approaches used by previous authors. First, we use a date of 12 Mya for the root of the LT radiation (Figure 1), which is the proposed maximum geological age of LT  and which is similar to the estimated age for this node based on fossil cichlid calibrations . Second, we use a date of 28 Mya for the same node based on calibrations of Gondwanan fragmentation for cichlids . The cichlid timescale based on continental fragmentation supports a much older age (Cretaceous), of origin for cichlids, as opposed to the fossil record of the group, which dates back to the Eocene. This consequently has implications for the origin of LT cichlid faunas, in that Gondwanan dates imply the lake would have been colonised independently by the main lineages –, rather than the accepted view based on the younger calibration, which assumes that early branching events occurred within the lake , . We report results for both calibrations, but our main conclusions depend on relative dates, rather than absolute dates, and so are robust to these uncertainties. Dating the tree using a Bayesian relaxed molecular clock model in BEAST  resulted in the ultrametric consensus tree depicted in Figure 1.
Timescales for the two alternative calibrations are shown. The gray bars on each node indicate 95% credible intervals for the node ages. Black circle indicates the node used to date the tree. P = Principle lineage, C = C-lineage, L = L-lineage. Gray branches indicate non-endemic taxa. LM = Lake Malawi species.
Net Diversification Rates in LT
We used models assuming constant per lineage rates of speciation and extinction to estimate average net diversification rates (speciation rate minus extinction rate) for the LT flock and constituent clades (Table S2) . None of the clades displayed evidence for a non-zero extinction rate (Table S2). The average estimate of the per lineage diversification rate for the entire LT endemic flock, assuming a root age of 12 Mya and a Yule Prior for the distribution of node ages, is 0.31 species/Myr (we report per lineage rates throughout). The confidence interval incorporating both the stochastic nature of lineage branching and the uncertainty of relative date estimates across the Bayesian samples is 0.27 to 0.37 (Figure 2). To evaluate the effect of incomplete species sampling, we used taxonomic information about missing species to include them in our phylogeny. Each missing species was added to the tree in turn randomly with equal probability along the branches belonging to its likely clade (Figure S1 for placements). Addition of missing species increases the estimates to 0.36 (C.I. 0.30–0.40) for the entire flock (Table S2). We report results for the amended trees including the missing species but conclusions are qualitatively unaffected by their presence or absence. Our rate estimates are lower than previous ones for the LT flock, for example 0.9  and 1.42 species/Myr . Instead, they fall within the range of typical diversification rates estimated across a wide range of plant and animal taxa , ,  . Diversification rate estimates enforcing an alternative prior, proportional-to distinguishable (PD), are slower than those assuming the Yule prior e.g. 0.26 (C.I. 0.22–0.29) for the LT flock, but there remained no significant evidence for non-zero background extinction rates (Table S2). Similarly, assuming an older age of divergence of the LT radiation (28 Myr), rate estimates fall to 0.15 species/Myr (C.I. 0.13 to 0.17, Yule prior).
Comparison with Lake Malawi and Lake Victora
Although the uncertainty in calibration dates prevents robust inference of absolute rates, we use the position of LM endemic species in our tree for a conservative comparison of diversification rates between LT and the LM and LV species flocks. We assume that the node separating the LM species from their nearest related LT lineage, i.e. their stem group age, defines an extreme maximum date, t, for the single common ancestor of the LM and LV species, which comprise a monophyletic group together with several riverine taxa . Our approach, based on the younger calibration, uses an age of 3.1 Myr (C.I. 2.6 to 4.2) whereas the LV flock is believed to be much younger [e.g. 35]. We then estimate the net diversification rate as log(N)/t, where N is the number of species and the 95% confidence interval due to the stochasticity of the branching process is −log(1−0.0251/N)/t to −log(1−0.9751/N)/t . Across the Bayesian samples, using the younger calibration yields an average diversification rate of 2.02 species/Myr (C.I. 1.16 to 3.12) with N = 898 (minimum species number for LM and LV flocks) , rising to 2.09 species/Myr (C.I. 1.71 to 3.19) with N = 1135 (maximum species number of LM and LV flocks) . Using the older calibration yields average diversification rates of 0.87 (C.I. 0.70 to 1.33) and 0.90 (C.I. 0.73 to 1.37) respectively. The confidence intervals do not overlap with the C.I. of any of the LT clades (Figure 2). Therefore, the LM and LV species flocks have diversified at more than five times the rate of the LT flock. This conclusion applies even for the Tropheini, which are closely related to the LM and LV flocks and share their biological characteristics (see next section). Unlike previous comparisons, the conclusion is also independent of how we calibrated the radiations in geological time, because our comparisons derive from a single phylogenetic reconstruction, and is likely to be very conservative. Using a date for the LV flock of 0.12 Myr based on the older, conservative calibration  would yield a staggering per lineage rate estimate of over 50 species/Myr.
Comparison of LT Cichlid faunas
Separate estimates of net diversification rate for the endemic tribes reveal a relatively uniform average speed of diversification among tribes when 95% confidence intervals are considered (Figure 2). Mean values for tribes ranged from 0.19–0.60 species/Myr, assuming 12 Myr calibration, falling to 0.07–0.24 species/Myr using the older calibration (Table S2). To test whether rates varied significantly among tribes, we multiplied the time period between successive nodes in each tribe by the number of lineages present during that interval . Under a constant speciation rate model, these transformed internode distances are expected to be equal to the inverse of the diversification rate , . An anova with clade as a factor revealed no significant difference in diversification rate either between the C and L lineages (F1,176 = 0.1, p>0.5) or among tribes (F5,158 = 1.6, p>0.1).
The fastest rates were recorded in the Tropheini (0.60 species/Myr) and Lamprologini (0.41 species/Myr), assuming a 12 Myr calibration, slowing to 0.16 and 0.24 species/Myr respectively, when the 28 Myr calibration is enforced. That these tribes share similar rates is of interest as they have contrasting evolutionary histories, with the former having a more recent origin than the latter, as well as displaying marked differences in breeding behavior. Lamprologines are substrate brooders and are monochromatic (∼5% of species exhibit extreme sexual dimorphism) . Conversely, the Tropheini, as with all haplochromines, are mouthbrooders, and while most species are sexually dichromatic, some genera (e.g. Tropheus) are mainly monochromatic. The slowest rate is recorded in a deep-water tribe of strongly sexually dichromatic mouthbrooders, the Bathybatini, (irrespective of inclusion of the basal species, B. minor), with rates estimated from 0.19 – 0.08 species/Myr, assuming a 12 or 28 Myr calibration respectively. Therefore, contrary to previous findings (16), speciation is not uniformly faster in sexually dimorphic groups than in sexually monomorphic groups. The lamprologines also display much greater ecological diversity in trophic morphology and habitat than the Tropheini, consistent with a mode of ecological speciation . Interestingly, a large proportion of lamprologines have exploited gastropod shells to live in and/or breed . Occupation of this niche and the associated dwarfism it entailed may have triggered additional diversification (shell-brooding is significantly associated with increased diversification rate in the lamprologines (Table S1, Text S1). Whatever the mechanism of diversification in each tribe, it is clear that clades with very different biological characteristics have diversified to similar degrees in the lake.
The Timing of Diversification
Adaptive radiation, driven by ecological opportunity, predicts rapid early speciation, followed by a subsequent slowdown , , whereas speciation by non-ecological sexual selection does not predict this trend , and may instead predict a speed up of net diversification rate, as recorded in LV cichlids , [after 42]. However, if several lineages colonized the lake and radiated independently, we might expect bursts of diversification following colonization but not necessarily congruent between different clades if they colonized at different times and filled different broad niches. Other scenarios might generate similar patterns, for example the appearance of new niches in the lake, such as shell beds, or if environmental events such as changing lake levels ,  triggered an increased speciation rate .
While distinguishing all these scenarios is difficult, we test whether diversification rates have declined over time and whether any changes have been synchronous among clades. Lineage-through-time (LTT) plots show that rates have slowed down somewhat: a straight line is expected if there has been a constant speciation rate with no extinction (Figure 3) . Under both priors (Yule and PD), a significant decline is recorded (Pybus and Harvey γ statistic ) for the entire endemic flock (Figures 3a, S6a, γ = −2.25, p = 0.06; S6e -4.25, p = 0.000) and the principle radiation (Figures 3a, S6a, γ = −3.20, p = 0.007; S6e -4.52, p = 0.000). This is also true for the L-lineage under the PD prior (S6e -3.55, p = 0.002), while the decline enforcing the Yule prior is marginally non-significant (Figure 3b, S6a, γ = −2.77, p = 0.02). Among the tribes only the Lamprologini (which comprise the majority of the L-lineage) display a significant decline in rates over time (Figure 3c, Table S2). Declines among the other tribes and C-lineage are non-significant (Table S2) and overall there is not the strong plateau expected of an early adaptive radiation and slow-down model. Instead, the curve for the C-lineage is more sigmoid in shape (Figure 3b).
A single LLT plot is highlighted for all tribes, but is only displayed for the latter three tribes for clarity. Timescales for the two alternative calibrations are shown.
General Additive Models (GAM) were used to explore the net change in diversification rates over time in more detail (Figure 4) . These show that, across the entire flock, diversification rates were highest at intermediate periods, around 3 to 4 Mya (12 Myr calibration) or 7 to 8 Mya (28 Myr calibration, Figure 4a). These findings appear to contradict those of the Pybus and Harvey test , which implied a significant slowdown. While the robustness of GAMs has been questioned in the past, in this case similar conclusions are reached from trees derived from the two different priors of node age distributions and from the repeated samples from each MCMC analysis. Therefore, we believe the GAMs are fitting real features of the distribution of node ages in our sample: the average overall decline in diversification rate is largely due to decline from intermediate to recent periods, which encompass most of the nodes contributing information to the test statistics.
Scale bars (x-axis) represent both 28 Myr and 12 Myr calibrations and the y-axis is the per lineage diversification rate in species per million years. Dashed lines show standard errors combining those across Bayesian samples and those due to uncertainty in the model in a), but only those across Bayesian samples in b). In the latter case, standard errors due to uncertainty in the model are extremely large.
The pattern differs marginally between the C- and L-lineages (ANOVA comparing GAMs with clade as a factor or not, median F across Bayesian samples = 2.1, median p = 0.033). The C-lineage is typified by an early and late peak in diversification rates, separated by a period of low net rates, during which time the L-lineage sustained higher net rates (Figure 4b). Both lineages reveal a final decline in rates towards the present, which could reflect either a true decline, over-conservative placement of the species missing from our sample, or a taxonomic artifact of failure to recognize recently diverged species. If clades have different peak periods of diversification, even though average diversification rates are relatively similar among tribes, this would suggest that biological differences among tribes did matter in the build-up of present-day diversity. Either clades were affected differentially by different environmental events or they filled broadly different niches in the lake. If biological differences were irrelevant, we would have expected any changes in diversification rate to be congruent among the different clades. However, even with a near complete sample of species, and potentially large differences in average diversification rate at given times, the differences between the two clades are only marginally significant in our present analyses.
Interpreting the causes of the diversification pulses is even more difficult. With the younger calibration based on lake age and fossils, then the first peak in the C-lineage corresponds roughly with the onset of full lacustrine conditions and the second peak with a shift in global climate and the aridification of Africa , when lake levels are assumed to have dropped. With the older calibration based on Gondwanan fragmentation, then the initial peak would correspond to an early radiation generating the ancestors of the different tribes, perhaps in a now extinct lake , and the second peak with the onset of full lacustrine conditions. Similarly, the L-lineage may have sustained high rates from basin formation to the onset of full lacustrine conditions (older calibration) or after these events (younger calibrations). The older calibration is appealing because peak diversification across the entire flock (ca. 8 to 9 Mya) would then be congruent with basin formation and development, but this requires that diversification leading to the different tribes occurred at an earlier time. However, the older calibration leads to rather slow estimates of net diversification rates among LT cichlids (although not LM and LV). Clearly, further evidence is needed to resolve these alternatives. Given the difficulty in finding robust calibrations for cichlids, and the limits of sample size being reached for distinguishing alternative diversity models, one possibility would be to compare the timing of diversification across a broad suite of LT endemic taxa, especially those with fossil records or alternative geological calibrations e.g. , .
We conclude that LT cichlid fishes have diversified much more slowly than those of LV and LM, even the endemic LT haplochromines. In contrast to LM and LV, the LT species flock derives from a prolonged accumulation of species, rather than rapid, recent radiation. Clearly, the environment plays a major role in determining cichlid diversification rates. One possibility is that the physical environment of LT is less conducive for cichlid speciation, perhaps by inhibiting whatever special mechanism causes fast speciation in LM and LV. Turbid waters have been shown to impede sexual selection in LV , raising the possibly that a similar phenomenon may have occurred during the history of LT, although currently these waters are very transparent . Alternatively, the remarkable diversifying abilities of haplochromines could have been inhibited because the lake was already occupied by older radiations, such as the lamprologines and ectodines, which occupy similar habitats. The similarity of net diversification rates among tribes with very different biological characteristics might seem to imply that traits had minimal influence on the progress of diversification in the lake. However, there is some evidence that clades did vary in the timing of peak rates of diversification: lineage identity was important for determining diversification potential over large periods of the lake's history. Whether these differences reflect differential responses to episodes of climate change or successive adaptive radiations into different broad niches is difficult to resolve until more robust calibrations are available. However, a plausible explanation is that the lake was colonized several times leading to a series of independent adaptive radiations into different broad niches followed by a slow-down in rate in each descendent clade.
Materials and Methods
Previous phylogenetic studies of LT cichlids consider subsets of the data (< one third of sampled species) , . However, recent proliferation of sequence data represents an important source for comparative studies on LT cichlids. Densely sampled trees are important in diversification studies as well as providing information on relative timing of cladogensis . Increased taxon sampling has been shown to improve both the precision and accuracy of phylogenetic reconstruction –. We reconstruct the evolutionary relationships of 161 cichlid species (152 LT endemics, 9 non-endemics), representing 76% of all LT species (assuming a LT cichlid flock of 200 endemic species, Table S1), based on two mtDNA genes: the protein coding, ND2 (∼1047 bp) and the non-coding, control region (CR), (partial sequence data of 347 bp from the 5′-end) using published sequence data (Table S1). Alignment of ND2 sequences was unambiguous, in comparison to the CR, which varied slightly in length among sequences (337-335, with the longest gap consisting of a three bp indel). The CR data was initially aligned using CLUSTAL X  and subsequently both datasets were optimised by visual inspection in a manual alignment editor . Modeltest v.3.7  selected the evolutionary model GTR+I+Γ, using Akaike Information Criteria and was implemented in subsequent analyses. Bayesian Inference, using Metropolis-coupled Markov Chain Monte Carlo (MCMCMC), implemented in MrBayes v.3.1.2 , was run for 2×106 generations, sampling every 100 generations (four chains, temperature 0.2), to ensure convergence, with the first 250,000 generations discarded as burn-in. This analysis was partitioned to account for the different behaviour of the genes as well as the third codon of ND2. Maximum likelihood inference was implemented in GARLI v.0.942 (Genetic Algorithm for Rapid Likelihood Inference)  also sampling 2×106 generations for multiple runs to ensure similar trees and lnL scores. Branch support was evaluated using non-parametric bootstrapping (BS) consisting of 1000 pseudoreplicates (using GARLI), and Bayesian posterior probabilities (BPP). The tree topologies recovering using Bayesian Inference and Maximum Likelihood were evaluated using the approximately unbiased test , implemented in the program CONSEL  using the site likelihood scores from PAUP*  see Text S1.
A likelihood ratio test  performed on the concatenated data set [molecular clock enforced (-ln L 41653.91), not enforced (-ln L 41352.23), χ2 = 603.36, df = 157, p = 0.00,)] rejected overall constancy of rates of evolution. Divergence time analyses using a log-normal distributed relaxed molecular clock  were performed using BEAST v1.4.6 , which uses Bayesian inference and an MCMC procedure to estimate the posterior distribution of rates and times. We used a constant-rate Yule (speciation process) prior, and all other priors and operators were the default settings, except that the root was constrained with a point prior, and the starting tree was the maximum posterior probability topology form the Bayesian phylogenetic analysis, and this topology remained fixed throughout the MCMC. Three independent chains were run for 2×106 generations each, and convergence checked visually by comparing these runs after a burn-in of 2×105 generations was discarded. As there is no Bayesian MCMC implementation of speciation models that accommodate extinction, we also performed the same analyses using a Proportional-to-distinguishable (PD) prior. The PD arrangements prior puts an equal probability on every different labelled tree . To investigate the sensitivity of divergence time estimates, we performed the same analyses using exponential relaxed clock and strict clock models in BEAST. Results from these analyses were largely congruent with those from the analysis reported here.
Estimating Diversification Rates
The resulting ultrametric trees obtained using both Yule and PD priors using a log-normal distributed relaxed clock calculated by BEAST were imported into the APE 1.2–2 package  of the R statistical programming language  to generate semi logarithmic LLT plots calculated for 1000 sampled trees (sampling every 2500 generations of the LT radiation and across different taxonomic grouping of LT cichlids We also consider diversification rate (speciation minus extinction rates) using the results from BEAST, implementing different constant speciation and extinction rate models using the birthdeath function in APE, for the same taxonomic groups, similarly calculated from 1000 sampled trees (sampling every 2500 generations). All non-endemic taxa were excluded from these analyses. For each clade we also tested for significant departures from the constant speciation model using the γ statistic . Positive values for this statistic signify that there has been an apparent increase in diversification rate towards the present, whereas negative values indicate a deceleration in diversification rate towards the present. Generalized additive models were constructed using the mgcv library in R . Models of transformed internode intervals as a function of time were fitted assuming a basis dimension (k) of 50 knots. Unbiased risk estimation (UBRE) was then used to estimate smoothing parameters and the effective degrees of freedom for the smooth term . Hypothesis-testing comparing a single and different smoothing functions for the C and L lineages was based on unpenalized GAMs ; for these, k was set to 10 because statistical power falls as k is increased. Nested models (with a single smoothing function or a separate smoothing functions for the two lineages) were compared using ANOVA and F tests. We further applied these models to the data from the PD prior, but found the result was non-significant.
Phylogeny of Lake Tanganyika cichlids based on mtDNA. Bayesian tree recovered from combined analysis of ND2 and control region datasets. Branch support indicated by BPP and BS derived from BI and ML methods (>50%) from left to right respectively. * indicates the same score for both support methods. P = principal lineage, C = C-lineage, L = L-lineage, LM = Lake Malawi. Gray branches indicate non-endemic taxa. Gray circles containing numerical numbers, indicate number of additional species added at that branch, closed gray circle indicates no additional species added beyond that node. Black circle indicates the node used to calibrate the tree. Full details for all species used in this study, including GenBank accession numbers are provided in Table S2.
(0.61 MB EPS)
Cichlid taxa included in phylogenetic analyses and associated Genbank numbers.
(0.13 MB PDF)
Speciation rate statistics for the LT cichlid radiation, major lineages and constituent tribes.
(0.08 MB PDF)
We thank David Gower, Andy Purvis, Oliver Pybus, Lukas Rüber, Mark Wilkinson and anonymous referees for suggestions to improve this manuscript and Andrew Rambaut for help with BEAST.
Conceived and designed the experiments: TB JD JC. Analyzed the data: TB JD JC. Wrote the paper: TB JD JC.
- 1. Simpson GG (1953) The Major Features of Evolution. New York: Columbia University Press. GG Simpson1953The Major Features of Evolution.New YorkColumbia University Press
- 2. Schluter D (2000) The Ecology of Adaptive Radiation. Oxford: Oxford University Press. D. Schluter2000The Ecology of Adaptive Radiation.OxfordOxford University Press
- 3. Purvis A, Hector A (2001) Getting the measure of biodiversity. Nature 405: 212–219.A. PurvisA. Hector2001Getting the measure of biodiversity.Nature405212219
- 4. Barraclough TG, Vogler AP, Harvey PH (1998) Revealing the factors that promote speciation. Phil Trans R Soc London B 353: 241–249.TG BarracloughAP VoglerPH Harvey1998Revealing the factors that promote speciation.Phil Trans R Soc London B353241249
- 5. Coyne J, Orr HA (2004) Speciation. Sunderland, MA: Sinauer. J. CoyneHA Orr2004Speciation.Sunderland, MASinauer
- 6. Cardillo M, Orme CDL, Owens IPF (2005) Testing for latitudinal bias in diversification rates: An example using New World birds. Ecology 86: 2278–2287.M. CardilloCDL OrmeIPF Owens2005Testing for latitudinal bias in diversification rates: An example using New World birds.Ecology8622782287
- 7. Davies TJ, Barraclough TG, Chase MW, Soltis PS, Soltis DE, et al. (2004) Darwin's abominable mystery: Insights from a supertree of the angiosperms. Proc Natl Acad Sci USA 101: 1904–1909.TJ DaviesTG BarracloughMW ChasePS SoltisDE Soltis2004Darwin's abominable mystery: Insights from a supertree of the angiosperms.Proc Natl Acad Sci USA10119041909
- 8. Ricklefs RE (2003) Global diversification rates of passerine birds. Proc R Soc London B 270: 2285–2291.RE Ricklefs2003Global diversification rates of passerine birds.Proc R Soc London B27022852291
- 9. Davies TJ, Savolainen V, Chase MW, Goldblatt P, Barraclough TG (2005) Environment, area, and diversification in the species-rich flowering plant family Iridaceae. Am Nat 166: 418–425.TJ DaviesV. SavolainenMW ChaseP. GoldblattTG Barraclough2005Environment, area, and diversification in the species-rich flowering plant family Iridaceae.Am Nat166418425
- 10. Genner MJ, Seehausen O, Cleary DFR, Knight ME, Michel E, et al. (2004) How does the taxonomic status of allopatric populations influence species richness within African cichlid fish assemblages? J Biogeogr 31: 93–102.MJ GennerO. SeehausenDFR ClearyME KnightE. Michel2004How does the taxonomic status of allopatric populations influence species richness within African cichlid fish assemblages?J Biogeogr3193102
- 11. Johnson TC, Scholz CA, Talbot MR, Kelts K, Ricketts RD, et al. (1996) Late Pleistocene desiccation of Lake Victoria and rapid evolution of cichlid fishes. Science 273: 1091–1093.TC JohnsonCA ScholzMR TalbotK. KeltsRD Ricketts1996Late Pleistocene desiccation of Lake Victoria and rapid evolution of cichlid fishes.Science27310911093
- 12. Delvaux D (1996) Age of Lake Malawi (Nyasa) and water level fluctuations. Mus R Afr Centr Tervuren (Belg) Dept Geol Min Rapp Ann 1995–1996: 99–108.D. Delvaux1996Age of Lake Malawi (Nyasa) and water level fluctuations.Mus R Afr Centr Tervuren (Belg) Dept Geol Min Rapp Ann1995–199699108
- 13. Dominey WJ (1984) Effects of sexual selection and life history on speciation: Species flocks in African cichlids and Hawaiian Drosophila. In: Echelle AA, Kornfield I, editors. Evolution of Fish Species Flocks. Orono, Maine: University Maine at Orono Press. pp. 231–250.WJ Dominey1984Effects of sexual selection and life history on speciation: Species flocks in African cichlids and Hawaiian Drosophila.AA EchelleI. KornfieldEvolution of Fish Species FlocksOrono, MaineUniversity Maine at Orono Press231250
- 14. Danley PD, Kocher TD (2001) Speciation in rapidly diverging systems: lessons from Lake Malawi. Mol Ecol 10: 1075–1086.PD DanleyTD Kocher2001Speciation in rapidly diverging systems: lessons from Lake Malawi.Mol Ecol1010751086
- 15. Knight ME Turner GF (2004) Laboratory mating trials indicate incipient speciation by sexual selection among populations of the cichlid fish Pseudotropheus zebra from Lake Malawi. Proc. R. Soc. London B. 271: 675–680.GF Knight ME Turner2004Laboratory mating trials indicate incipient speciation by sexual selection among populations of the cichlid fish Pseudotropheus zebra from Lake Malawi.Proc. R. Soc. London B.271675680
- 16. McCune AR, Lovejoy NR (1998) The relative rate of sympatric and allopatric speciation in fishes: tests using DNA sequence divergence between sister species and among clades. In: Howard D, Berlocher S, editors. Endless Forms: Species and Speciation. Oxford: Oxford University Press. pp. 172–185.AR McCuneNR Lovejoy1998The relative rate of sympatric and allopatric speciation in fishes: tests using DNA sequence divergence between sister species and among clades.D. HowardS. BerlocherEndless Forms: Species and SpeciationOxfordOxford University Press172185
- 17. Turner GF (1999) Explosive speciation of African cichlids. In: Magurran AE, May RM, editors. Evolution of Biological Diversity. Oxford: Oxford University Press. pp. 113–129.GF Turner1999Explosive speciation of African cichlids.AE MagurranRM MayEvolution of Biological DiversityOxfordOxford University Press113129
- 18. Cohen AS, Soreghan MJ, Scholz CA (1993) Estimating the age of formation of lakes - an example from Lake Tanganyika, East-African rift system. Geology 21: 511–514.AS CohenMJ SoreghanCA Scholz1993Estimating the age of formation of lakes - an example from Lake Tanganyika, East-African rift system.Geology21511514
- 19. Cohen AS, Lezzar KE, Tiercelin JJ, Soreghan M (1997) New palaeogeographic and lake-level reconstructions of Lake Tanganyika: Implications for tectonic, climatic and biological evolution in a rift lake. Basin Res 9: 107–132.AS CohenKE LezzarJJ TiercelinM. Soreghan1997New palaeogeographic and lake-level reconstructions of Lake Tanganyika: Implications for tectonic, climatic and biological evolution in a rift lake.Basin Res9107132
- 20. Takahashi T (2003) Systematics of Tanganyikan cichlids fishes (Teleostei: Perciformes). Ichthy Res 50: 367–382.T. Takahashi2003Systematics of Tanganyikan cichlids fishes (Teleostei: Perciformes).Ichthy Res50367382
- 21. Salzburger W, Mack T, Verheyen E, Meyer A (2005) Out of Tanganyika: Genesis, explosive speciation, key-innovations and phylogeography of the haplochromine cichlid fishes. BMC Evol Biol 5: 17.W. SalzburgerT. MackE. VerheyenA. Meyer2005Out of Tanganyika: Genesis, explosive speciation, key-innovations and phylogeography of the haplochromine cichlid fishes.BMC Evol Biol517
- 22. Barraclough TG, Nee S (2001) Phylogenetics and speciation. Trends Ecol Evol 16: 391–399.TG BarracloughS. Nee2001Phylogenetics and speciation.Trends Ecol Evol16391399
- 23. Seehausen O (2006) African cichlid fish: a model system in adaptive radiation research. Proc R Soc London B 273: 1987–1998.O. Seehausen2006African cichlid fish: a model system in adaptive radiation research.Proc R Soc London B27319871998
- 24. Cane MA, Molnar P (2001) Closing of the Indonesian seaway as a precursor to East African aridification around 3–4 million years ago. Nature 411: 157–162.MA CaneP. Molnar2001Closing of the Indonesian seaway as a precursor to East African aridification around 3–4 million years ago.Nature411157162
- 25. Tiercelin JJ, Mondeguer A (1991) The geology of the Tanganyika trough. In: Martens K, Goddeeris B, Coulter G, editors. Lake Tanganyika and its life. Oxford: Oxford University Press, Oxford. pp. 7–48.JJ TiercelinA. Mondeguer1991The geology of the Tanganyika trough.K. MartensB. GoddeerisG. CoulterLake Tanganyika and its lifeOxfordOxford University Press, Oxford748
- 26. 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: 666–681.C. ClabautW. SalzburgerA. Meyer2005Comparative 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 Evol61666681
- 27. Genner MJ, Seehausen O, Lunt DH, Joyce DA, Shaw PW, et al. (2007) Age of cichlids: New dates for ancient lake fish radiations. Mol Biol Evol 24: 1269–1282.MJ GennerO. SeehausenDH LuntDA JoycePW Shaw2007Age of cichlids: New dates for ancient lake fish radiations.Mol Biol Evol2412691282
- 28. Nishida M (1997) Phylogenetic relationships and evolution of Tanganyikan cichlids: a molecular perspective. In: Kawanabe H, Hori M, Nagoshi M, editors. Fish Communities in Lake. Tanganyika: Kyoto University Press. pp. 3–23.M. Nishida1997Phylogenetic relationships and evolution of Tanganyikan cichlids: a molecular perspective.H. KawanabeM. HoriM. NagoshiFish Communities in LakeTanganyikaKyoto University Press323
- 29. Terai Y, Takahashi K, Nishida M, Sato T, Okada N (2003) Using SINEs to probe ancient explosive speciation: “Hidden” radiation of African cichlids? Mol Biol Evol 20: 924–930.Y. TeraiK. TakahashiM. NishidaT. SatoN. Okada2003Using SINEs to probe ancient explosive speciation: “Hidden” radiation of African cichlids?Mol Biol Evol20924930
- 30. Salzburger W, Meyer A, Baric S, Verheyen E, Strumbauer C (2002) Lake level fluctuations synchronize genetic divergences of cichlid fishes in African lakes. Syst Biol 1: 1–23.W. SalzburgerA. MeyerS. BaricE. VerheyenC. Strumbauer2002Lake level fluctuations synchronize genetic divergences of cichlid fishes in African lakes.Syst Biol1123
- 31. Drummond AJ, Ho SYW, Phillips MJ, Rambaut A (2006) Relaxed phylogenetics and dating with confidence. PLOS Biol 4: 699–710.AJ DrummondSYW HoMJ PhillipsA. Rambaut2006Relaxed phylogenetics and dating with confidence.PLOS Biol4699710
- 32. Nee S, May RM, Harvey PH (1994) The reconstructed evolutionary process. Phil Trans R Soc Lond B 344: 305–311.S. NeeRM MayPH Harvey1994The reconstructed evolutionary process.Phil Trans R Soc Lond B344305311
- 33. Baldwin BG, Sanderson MJ (1998) Age and rate of diversification of the Hawaiian silversword alliance (Compositae) Proc Natl Acad Sci USA 95: 9402–9406.BG BaldwinMJ Sanderson1998Age and rate of diversification of the Hawaiian silversword alliance (Compositae)Proc Natl Acad Sci USA9594029406
- 34. McCune AR (1997) How fast is speciation: molecular, geological and phylogenetic evidence from adaptive radiations of fishes. In: Givnish TJ, Sytsma K, editors. Molecular Evolution and Adaptive Radiation. Cambridge: Cambridge University Press. pp. 585–610.AR McCune1997How fast is speciation: molecular, geological and phylogenetic evidence from adaptive radiations of fishes.TJ GivnishK. SytsmaMolecular Evolution and Adaptive RadiationCambridgeCambridge University Press585610
- 35. Nagl S, Tichy H, Mayer WE, Takezaki N, Takahata N, et al. (2000) The origin and age of haplochromine fishes in Lake Victoria, East Africa. Proc R Soc London B 267: 1049–1061.S. NaglH. TichyWE MayerN. TakezakiN. Takahata2000The origin and age of haplochromine fishes in Lake Victoria, East Africa.Proc R Soc London B26710491061
- 36. Purvis A, Nee S, Harvey PH (1995) Macroevolutionary inferences from primate phylogeny. Proc R Soc London B 260: 329–333.A. PurvisS. NeePH Harvey1995Macroevolutionary inferences from primate phylogeny.Proc R Soc London B260329333
- 37. Nee S (2001) Inferring speciation rates from phylogenies. Evolution 55: 661–668.S. Nee2001Inferring speciation rates from phylogenies.Evolution55661668
- 38. Barraclough TG, Vogler AP (2002) Recent diversification rates in North American tiger beetles estimated from a dated mtDNA phylogenetic tree. Mol Biol Evol 19: 1706–1716.TG BarracloughAP Vogler2002Recent diversification rates in North American tiger beetles estimated from a dated mtDNA phylogenetic tree.Mol Biol Evol1917061716
- 39. Konings A (1998) Tanganyika Cichlids in their Natural Habitat. El Paso, TX: Cichlid Press. A. Konings1998Tanganyika Cichlids in their Natural Habitat.El Paso, TXCichlid Press
- 40. Yanagisawa Y, Ochi H, Gashagaza MM (1997) Habitat use in cichlid fishes for breeding. In: Kawanabe H, Hori M, Nagoshi M, editors. Fish Communities in Lake. Tanganyika: Kyoto University Press. pp. 151–173.Y. YanagisawaH. OchiMM Gashagaza1997Habitat use in cichlid fishes for breeding.H. KawanabeM. HoriM. NagoshiFish Communities in LakeTanganyikaKyoto University Press151173
- 41. Gavrilets S, Vose A (2005) Dynamic patterns of adaptive radiation. Proc Natl Acad Sci USA 102: 18040–18045.S. GavriletsA. Vose2005Dynamic patterns of adaptive radiation.Proc Natl Acad Sci USA1021804018045
- 42. Meyer A, Kocher TD, Basasibwaki P, Wilson (1990) Monophyletic origin of Lake Victoria cichlid fishes suggested by mitochondrial-DNA sequences. Nature 347: 550–553.A. MeyerTD KocherP. BasasibwakiWilson1990Monophyletic origin of Lake Victoria cichlid fishes suggested by mitochondrial-DNA sequences.Nature347550553
- 43. Sturmbauer C, Baric S, Salzburger W, Rüber L, Verheyen E (2001) Lake level fluctuations synchronize genetic divergences of cichlid fishes in African lakes. Mol Biol Evol 18: 144–154.C. SturmbauerS. BaricW. SalzburgerL. RüberE. Verheyen2001Lake level fluctuations synchronize genetic divergences of cichlid fishes in African lakes.Mol Biol Evol18144154
- 44. Pybus OG, Harvey PH (2000) Testing macro-evolutionary models using incomplete molecular phylogenies. Proc. R. Soc. Lond. B 267: 2267–2272.OG PybusPH Harvey2000Testing macro-evolutionary models using incomplete molecular phylogenies.Proc. R. Soc. Lond. B26722672272
- 45. Binda-Emonds ORP, Cardillo M, Jones KE, MacPhee RDE, Beck RMD, et al. (2007) The delayed rise of present-day mammals. Nature 446: 507–512.ORP Binda-EmondsM. CardilloKE JonesRDE MacPheeRMD Beck2007The delayed rise of present-day mammals.Nature446507512
- 46. Day JJ, Wilkinson M (2006) On the origin of the Synodontis catfish species flock from Lake Tanganyika. Biol. Letts 2: 548–552.JJ DayM. Wilkinson2006On the origin of the Synodontis catfish species flock from Lake Tanganyika.Biol. Letts2548552
- 47. Marijnissen SAE, Michel E, Daniels SR, Erpenbeck D, Menken SBJ, et al. (2006) Molecular evidence for recent divergence of Lake Tanganyika endemic crabs (Decapoda : Platythelphusidae). Mol. Phylogenet. Evol. 40: 628–634.SAE MarijnissenE. MichelSR DanielsD. ErpenbeckSBJ Menken2006Molecular evidence for recent divergence of Lake Tanganyika endemic crabs (Decapoda : Platythelphusidae).Mol. Phylogenet. Evol.40628634
- 48. Seehausen O, van Alphen JJM, Witte F (1997) Cichlid fish diversity threatened by eutrophication that curbs sexual selection. Science 277: 1808–1811.O. SeehausenJJM van AlphenF. Witte1997Cichlid fish diversity threatened by eutrophication that curbs sexual selection.Science27718081811
- 49. Langenberg VT, Mwape LM, Tschibangu K, Tumba J-M, Koelmans AA, et al. (2002) Comparison of thermal stratification, light attenuation, and chlorophyll-a dynamics between the ends of Lake Tanganyika. J Aquat Ecosyst Health 5: 255–265.VT LangenbergLM MwapeK. TschibanguJ-M TumbaAA Koelmans2002Comparison of thermal stratification, light attenuation, and chlorophyll-a dynamics between the ends of Lake Tanganyika.J Aquat Ecosyst Health5255265
- 50. Flynn JJ, Finarelli JA, Zehr S, Hsu J, Nedbal MA (2005) Molecular phylogeny of the Carnivora (Mammalia): Assessing the impact of increased sampling on resolving enigmatic relationships. Syst Biol 54: 317–337.JJ FlynnJA FinarelliS. ZehrJ. HsuMA Nedbal2005Molecular phylogeny of the Carnivora (Mammalia): Assessing the impact of increased sampling on resolving enigmatic relationships.Syst Biol54317337
- 51. Zwickl DJ, Hills DM (2002) Increased taxon sampling greatly reduces phylogenetic error. Syst Biol 51: 588–598.DJ ZwicklDM Hills2002Increased taxon sampling greatly reduces phylogenetic error.Syst Biol51588598
- 52. Thompson JD, Gibson TJ, Plewniak F, Jeanmougin J, Higgins DG (1997) The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res 25: 4876–4882.JD ThompsonTJ GibsonF. PlewniakJ. JeanmouginDG Higgins1997The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools.Nucleic Acids Res2548764882
- 53. Rambaut A (1996) Se-Al: Sequence Alignment Editor. A. Rambaut1996Se-Al: Sequence Alignment Editor.http://evolve.zoo.ox.ac.uk/. http://evolve.zoo.ox.ac.uk/.
- 54. Posada D, Crandall K (1998) MODELTEST: testing the model of DNA substitution. Bioinform 14: 817–818.D. PosadaK. Crandall1998MODELTEST: testing the model of DNA substitution.Bioinform14817818
- 55. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinform 19: 1572–1574.F. RonquistJP Huelsenbeck2003MrBayes 3: Bayesian phylogenetic inference under mixed models.Bioinform1915721574
- 56. Zwickl DJ (2006) Genetic Algorithm Approaches for the Phylogenetic Analysis of Large Biological Sequence Datasets under the Maximum Likelihood Criterion. Ph.D. Dissertation, University of Texas at Austin. DJ Zwickl2006Genetic Algorithm Approaches for the Phylogenetic Analysis of Large Biological Sequence Datasets under the Maximum Likelihood Criterion. Ph.D. Dissertation, University of Texas at Austin.Available from http://www.zo.utexas.edu/faculty/antisense/garli/Garli.html. Available from http://www.zo.utexas.edu/faculty/antisense/garli/Garli.html.
- 57. Shimodaira H (2002) An approximately unbiased test of phylogenetic tree selection. Syst Biol 51: 492–508.H. Shimodaira2002An approximately unbiased test of phylogenetic tree selection.Syst Biol51492508
- 58. Shimodaira H, Hasegawa M (2001) CONSEL: for assessing the confidence of phylogenetic tree selection. Bioinform 17: 1246–1247.H. ShimodairaM. Hasegawa2001CONSEL: for assessing the confidence of phylogenetic tree selection.Bioinform1712461247
- 59. Swofford DL (2002) PAUP*: Phylogentic analysis using parsimony (* and other methods), version 4.0b10. Sunderland, MA: Sinauer. DL Swofford2002PAUP*: Phylogentic analysis using parsimony (* and other methods), version 4.0b10Sunderland, MASinauer
- 60. Huelsenbeck JP, Crandall KA (1997) Phylogeny estimation and hypothesis testing using maximum likelihood. Ann Rev Ecol Syst 28: 437–466.JP HuelsenbeckKA Crandall1997Phylogeny estimation and hypothesis testing using maximum likelihood.Ann Rev Ecol Syst28437466
- 61. Drummond AJ, Rambaut A (2007) BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol 7: 214.AJ DrummondA. Rambaut2007BEAST: Bayesian evolutionary analysis by sampling trees.BMC Evol Biol7214
- 62. Rosen DE (1978) Vicariant patterns and historical explanation in biogeography. Syst Zool 27: 159–188.DE Rosen1978Vicariant patterns and historical explanation in biogeography.Syst Zool27159188
- 63. Paradis E, Claude J, Strimmer K (2004) APE: Analyses of Phylogenetics and Evolution in R language. Bioinform 20: 289–290.E. ParadisJ. ClaudeK. Strimmer2004APE: Analyses of Phylogenetics and Evolution in R language.Bioinform20289290
- 64. R Development Core Team (2006) R: A Language and Environment for Statistical Computing: Vienna, Austria. R Development Core Team2006R: A Language and Environment for Statistical Computing: Vienna, Austria
- 65. Wood SN (2006) Generalized Additive Models: An Introduction with R:. Boca Raton, Florida: Chapman & Hall/CRC. SN Wood2006Generalized Additive Models: An Introduction with R:Boca Raton, FloridaChapman & Hall/CRC