Keywords Evolution Host race Insect-plant interactions Speciation. Access to Document Link to publication in Scopus.
Fingerprint Dive into the research topics of 'Sympatric speciation in phytophagous insects: Moving beyond controversy? Together they form a unique fingerprint. View full fingerprint. Annual Review of Entomology , 47 , In: Annual Review of Entomology , Vol. Annual Review of Entomology. Google Scholar. Darwin C Random House: New York. Dobzhansky T Genetics and the Origin of Species.
Columbia University Press: New York. Hawthorne D, Via S Lexer C et al Evolution 57 : — Maynard Smith J Am Nat : — Mayr E Systematics and the Origin of Species. Rieseberg LH et al Am J Bot 82 : — Science : Savolainen V et al Nature advanced online publication. West-Eberhard MJ Developmental Plasticity and Evolution. Oxford University Press: New York.
II: Assortative mating and host-plant preferences for oviposition. Heredity 94 : — Helbig AJ Evolutionary genetics: a ring of species. Heredity 95 : — Mallet J Speciation in the 21st century. Ollerton J Speciation: flowering time and the Wallace effect. Download references. You can also search for this author in PubMed Google Scholar. Correspondence to D Ortiz-Barrientos. Reprints and Permissions.
Ortiz-Barrientos, D. Speciation: Splitting when together. Heredity 97, 2—3 Download citation. Published : 17 May Issue Date : 01 July Speciation is the main driver of biological diversity and how species arise is a central question in evolutionary biology. For speciation to occur in sexually reproducing organisms the exchange of genetic material gene flow between populations has to be reduced. Ultimately this has to be due to genetically determined reproductive incompatibilities between species.
Yet, whether an initial period of geographic isolation is necessary for these incompatibilities to evolve has been subject to one of the most persistent debates in evolutionary biology. Sympatric speciation is the most extreme case of primary divergence-with-gene-flow and lies at the heart of this question.
However, only few empirical examples of sympatric speciation are generally accepted and in most of these cases some ambiguities and doubts remain. This study provides evidence that the Nicaraguan crater lake cichlids can indeed be considered a valid example of sympatric speciation in the sense that the species themselves probably started to diverge in the absence of geographic barriers.
However, the data also suggests that this divergence in sympatry may have been facilitated by genetic variants that evolved during a time of isolation between an initial founding population and a secondary wave of colonizers stemming from the same source population. This highlights the limitations in the definitions of sympatric speciation when the mosaic nature of genomes is taken into account: some of the genetic regions driving divergence may have evolved in allopatry while the populations themselves diverged in sympatry.
PLoS Genet 12 6 : e 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. Data files for all population genomic and phylogenetic analyses as well as site frequency spectra files are provided as Supporting Information S1 Dataset.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Understanding how populations can diverge and become distinct species in the presence of gene flow is a central objective in evolutionary biology [ 1 — 3 ]. That gene flow poses a problem for speciation has for long been known [ 4 — 6 ].
Gene flow and recombination homogenize the genomes of diverging populations and break down associations of loci relevant for ecological adaptations and assortative mating; a condition usually required for speciation [ 7 — 9 ].
Yet, a growing body of research has shown that speciation can progress in the presence of gene flow [ 2 , 10 — 13 ]. This distinction is important as the latter involves a period of geographic isolation in which the abovementioned problem of gene flow and recombination does not arise [ 2 , 15 ]. The evolution of reproductive incompatibilities in geographic isolation allopatry is well understood and not controversial, while primary divergence-with-gene-flow in the absence of strong geographic barriers demands other explanations [ 16 ].
From a population genetic perspective, the most extreme case of primary divergence-with-gene-flow is sympatric speciation [ 17 ]. In a biogeographic sense, sympatric speciation can be broadly defined as speciation in the complete absence of geographic external barriers [ 18 ]. The two definitions are not always in concordance [ 19 — 21 ], but the ultimate question that relates both and motivates the study of sympatric speciation is whether and to what extent speciation requires the mediating effects of a period of geographic isolation.
In other words, is geographic isolation necessary to reduce gene flow and initiate population divergence in the first place or can speciation commence in a panmictic population? Thus, sympatric speciation has for long attracted theoreticians and empiricists alike, not because it is believed to occur frequently, but because—being the endpoint of the continuum of primary divergence-with-gene-flow—it may be particularly informative on the ecological conditions and evolutionary mechanisms that can lead to speciation in the presence of gene flow [ 19 , 22 , 23 ].
While theoretical models have shown that sympatric speciation is possible [ 8 , 22 , 24 — 26 ], only few convincing empirical case studies have been published [reviewed in ref.
And even in some of these cases critics remained doubtful [ 27 , 28 ]. This is partly due to the fact that speciation with geographic isolation is generally considered much more plausible, almost like a null hypothesis in speciation.
Sympatric speciation appears thus not only to be rare, but also hard to demonstrate empirically. In their seminal book Coyne and Orr [ 16 ] proposed four criteria that have to be fulfilled to demonstrate that sympatric speciation is the most likely mode of speciation: i sympatric distribution of contemporary species, ii genetically-based reproductive isolation, iii phylogenetic sister relationship, and iv no historic phase of geographic isolation.
Several cases are in concordance with some of these criteria, but almost none unambiguously fit all four [ 18 , 22 ]. Particularly the latter two criteria are inherently difficult to address and demonstrate. This is because a sister relationships between species criterion iii must reflect a true lineage bifurcation event and not simply result from a close genetic relationship due to secondary gene flow of evolutionarily more distantly related taxa.
Especially inferences based on mitochondrial DNA alone are prone to error due to haplotype replacement [ 29 — 31 ], but nuclear markers can lead to false inferences too, if gene flow and incomplete lineage sorting are not accounted for [ 32 , 33 ]. Further, demonstrating that a past allopatric phase of currently sympatrically occurring true sister species is unlikely criterion iv , can be difficult to do in practice.
In the second scenario some of the genetic variation later involved in reproductive isolation could have evolved in the time of separation of the primary founder population and the secondary migrants. Importantly, these genetic variants would not immediately lead to divergence, but be absorbed into the gene pool—potentially leading to a hybrid swarm—and only later be recruited in the speciation process [ 35 , 36 ]. Speciation in this scenario could still be considered sympatric as population divergence happened in sympatry [ 34 ]; yet there is a role of geographic isolation if the admixture event was essential for speciation in sympatry.
The first two scenarios predict equal levels of shared ancestry with outgroups and the source population and no signs of differential admixture i. Distinguishing between these three scenarios is especially difficult if the source population is not known or extinct; an issue that leads to lingering doubts in even the otherwise most convincing cases of sympatric speciation [ 39 ].
The attainability of big genomic data sets as well as theoretical and methodological advances in recent years have, however, markedly increased the power to investigate more complex demographic scenarios of secondary gene flow, admixture, and multiple colonization events [ 40 — 42 ], thereby permitting to now infer if periods of geographical isolation were involved in putative cases of primary divergence-with-gene-flow and sympatric speciation. In this regard, recent evidence for a complex pattern of secondary gene flow and unequal shared outgroup ancestry of sympatric species of Cameroonian crater lake cichlids [ 34 ], has shed some new light on this traditionally considered prime example of sympatric speciation [ 43 ].
Crater lake cichlids in Nicaragua, belonging to the Midas cichlid species complex Amphilophus sp. The two great lakes are both inhabited by two species of Midas cichlids: A.
While most crater lakes harbor only one yet often polymorphic population of Midas cichlids, in two of the crater lakes, Lake Apoyo and L. According to the current taxonomy Crater Lake Apoyo harbors six [ 47 ] and L. The small size of the crater lakes, the fact that they are surrounded by steep crater walls and no water connections exists, and the complete endemism of Midas cichlid species suggested sympatric speciation to be the most parsimonious scenario.
And indeed, genetic data supported the monophyly of Midas cichlids in L. Apoyo [ 50 ]. Yet, this first study was criticized because the different benthic species inhabiting L.
Apoyo were not considered separately and only one of the species, A. Nicaragua was considered in certain analyses [ 27 ]. Furthermore, the different species in L. Apoyo were not equidistant to the source population in genetic space as might be expected after sympatric speciation. Thus, according to the critics, the null hypothesis of multiple colonizations and introgressive hybridization could not be ruled out completely [ 27 ].
Later studies taking several or all six described species into account and using different genetic markers concluded sometimes in favor of monophyly of the L. Apoyo flock and thus sympatric speciation [ 49 , 51 , 52 ] and sometimes not [ 53 ]. In addition the assignment of individuals to the proposed six-species taxonomy did not match in many cases [ 49 , 53 ]. Generally, L. Nonetheless, also L.
But, a comprehensive investigation of the plausibility of sympatric speciation in L. In addition to the questions of monophyly and sympatric speciation there have been discrepancies in the inferred order of speciation events based on different markers and types of analyses [ 49 , 52 ]. Most importantly, none of the abovementioned studies did explicitly take admixture between lakes, intralacustrine gene flow, and population size changes into account.
Nonetheless, Midas cichlids still feature as one of the most prominent examples of sympatric speciation [ 18 ]. In this study we use genome-level analyses and demographic modeling in a coalescent framework to reconstruct the evolutionary history of the two parallel radiations of Midas cichlids in L.
Apoyo and L. More specifically we address all major points of previous criticism and more recent doubts concerning sympatric speciation in Midas cichlids [ 27 , 34 ]. To this end, we take all described species of Midas cichlids in the source and crater lakes into account and objectively assign individuals to genetic clusters to then i test for signs of unequal shared outgroup ancestry and differential admixture of sympatric species, ii establish the evolutionary relationships among species, and iii infer the demographic history of the two radiations to evaluate the evidence for primary divergence-with-gene-flow with or without secondary colonizations or secondary contact as outlined in the three scenarios of putative sympatric speciation above.
Previous studies of Midas cichlids had been partially hampered by difficulties concerning the taxonomic classifications. Thus as a first objective we investigated the population structure in our comprehensive data set. We were interested in both signs of genetic exchange and relationships among lake populations as well as population structure and individual ancestry within crater lakes.
In concordance with the geographic proximity and the assumed colonization history, the genetic cluster of L. Apoyo was closer to L. Nicaragua and L. Managua, while the two great lake populations were in close proximity in the two-dimensional genetic space. Interestingly, two distinct genetic clusters could be identified for L. Managua than the other one.
Individuals in this cluster corresponded exclusively to the two species A. In the case of twelve clusters, four of the clusters corresponded to the two species A. Managua and L. Nicaragua while individuals from L. Apoyo were assigned to four different clusters each S1A Fig. Notably, there were no signs of admixture between the lake populations anymore. Indicated on the geographic map are the locations of the two great lakes and the two focal crater lakes of this study.
Superimposed are the first two main axes of genetic variation principal components based on 17, SNPs. PC1 and PC2 explain 3. Dots mark the position of individuals in two-dimensional genetic space and are color-coded by lake of origin. In the intralacustrine Admixture analysis of L.
Apoyo the occurrence of four and five clusters had the highest support S2 Fig. Yet, 19 individuals, which are of strongly admixed ancestry in the case of four clusters S1 Fig , formed a distinct cluster in the case of five clusters Fig 2C.
Thus, our set of samples from L. Apoyo seemed to be best described by five genetic clusters. The main axis of variation PC1 clearly differentiated the limnetic A. However, the delineation of the benthic individuals into the four different genetic clusters did in many cases not fit their species assignment based on morphology.
Only in the case of A. Note that from here on we will essentially adopt a genetic cluster species concept [ 57 ] and use the terms species and cluster interchangeably. Furthermore, a few individuals from all genetic clusters exhibited signs of admixed ancestry.
Genetic clustering and individual ancestry of individuals within the two crater lakes both in form of A B the first three axes of genetic variation and C D the most supported number of clusters in an Admixture analysis bottom panel.
Analyses are based on 7, and 11, SNPs for L. Groups are labelled by species, if applicable, or genetic clusters as used in this study. Sample sizes are given in parentheses. Fish images next to species illustrate representative individuals. For L. Apoyo the five described benthic species are shaded in gray as they do not entirely match the genetic clusters.
Only in seven out of cases three A. However, the admixture plot also revealed a substantial amount of hybridization; eighteen individuals exhibited varying degrees of admixed ancestry between A. Putative hybrids are expected to occupy positions in genetic space along fictive lines connecting the species clusters [ 55 ]. Our rationale for this was that the inclusion of such a number of obviously admixed individuals also based on morphology, see below might have had a strong impact on the phylogenetic and demographic analyses and in many cases it would have been difficult to decide to which species they should be assigned to.
To further investigate the occurrence of hybridization within crater lakes we performed morphological analyses. Indeed, individuals in the hybrid group exhibited an intermediate morphology S3 and S4 Figs. Thus, in both crater lake radiations we find evidence for distinct genetic clusters, yet also signs for ongoing gene flow. Pairwise levels of overall genetic differentiation among all species in the four lakes are provided in S2 Table. Patterns of genome-wide differentiation across the 24 linkage groups and among all sympatric species within the two crater lake radiations are visualized in S5 and S6 Figs.
Analyses to detect loci putatively under divergent selection are described in S1 Text and detected outlier loci are given in S3 Table. The occurrence of two clusters in L. Managua, would be consistent with two waves of colonization followed by introgressive hybridization. However, clustering methods do not explicitly take the demographic history into account and can thus sometimes falsely indicate admixture [ 58 ].
Thus we performed formal tests of admixture using f3-statistics [ 59 ]. If the two species A. Managua in the PCA—or more accurately their ancestral population—resulted from secondary contact and subsequent introgressive hybridization with the already established crater lake population, tests including one of these two species as a test population and one of the other two species from L. Yet, none of the tests with this constellation returned a significant negative value Table 1.
In fact, we performed the test among all 1, possible three-population combinations considering all populations and lakes in our data set and only three tests returned a negative score, and none of those turned out to be significant. Thus, the f3-statistics do not provide evidence for secondary contact followed by introgressive hybridization. We note, however, that a history of admixture will not always result in negative f3-statistics , especially if the test population has experienced a lot of population-specific drift [ 60 , 61 ].
We further note that tests based on the f3-statistics would not be able to detect an admixture event secondary colonization that occurred before the sympatric species diverged as the test and reference populations of the crater lakes would share equal proportions of admixed genotypes.
Another way to investigate possible admixture events is by placing migration edges on a phylogenetic tree and evaluating whether they improve the fit of the model tree by reducing deviations in the residual covariance matrix: positive residuals indicate populations that exhibit observed covariances that are higher than accounted for by the model [ 61 ].
We used Treemix to build a tree and placed up to four migration edges m on it. Importantly, no stark positive residual covariances between any of the source populations and any of the crater lake species was apparent S7 Fig. The first three putative migration edges were placed between sympatric species within the two crater lakes S7C, S7E and S7G Fig and the fourth one between the ancestor of all L.
The latter migration edge is difficult to interpret as we would expect secondary gene flow from the source population or a related species into a crater lake species to be reflected by a migration edge coming from the lineage leading to the two species in the respective source lake, but not from its own ancestral lineage.
We note that we are not aware of any closely related species that could have hybridized with a Midas cichlid species in the last few thousand years. Furthermore, the small increase in fit provided by the fourth migration edge does not come from a decrease of positive residual covariances between A. Instead, it seems to improve the fit of the relationships among species within L. Thus, rather than indicating secondary gene flow from the source or a related population into A.
That some of the divergence events do not adhere to a strict bifurcating manner was also supported by other phylogenetic analyses that we performed see below. In any case, the f3-statistics did not provide evidence for differential admixture of A. We stress that we used Treemix in an explorative approach, but refer readers to the f3-statistics for formal tests of differential admixture.
Given the high fit of the model with four migration edges and the fact that the highest scaled residual covariance between any two populations was very low with less than 1. Also the phylogenetic sister relationship of L.
Interestingly, apart from the node grouping A. The low bootstrap support of nodes within the crater lake radiations in our Treemix tree led us to further investigate the evolutionary relationships among the sympatric species. To this end, we first built phylogenetic trees using SNAPP , which is explicitly designed to handle biallelic markers such as SNPs and employs the multispecies coalescent [ 62 ].
Due to the computational burden and since we were only interested in the topology as well as relative branching times within the two radiations we built two separate trees, as their respective monophyly was strongly supported. In both trees it was evident that the two species from the source lakes are sister species and are equally distantly related to the crater lakes radiations Fig 3. Within L. Apoyo the cloudogram indicated an almost starlike topology with extremely short internal nodes Fig 3A.
The overall consensus root canal suggested that A. Yet, every possible topology within the radiation was represented by some trees. In total different consensus trees were found differing in topology and divergence time. Also for the species that are endemic to L.
Interestingly the hybrid group did not take an intermediate position between A. Cloudograms of the radiation of A L. Apoyo and B L. Thin lines represent individual trees and thick blue lines indicate the overall supported topology root canal.
Due to the computational demand of this method the phylogenetic trees were limited to only four individuals per species and a subset of loci [ 63 , 64 ].
To evaluate whether the phylogenetic results might be influenced by using only few individuals and excluding missing data [ 65 ], we built individual-based phylogenetic split networks including all samples and more markers allowing for missing data see Methods for details. For both radiations they revealed essentially an identical pattern S8 Fig. Apoyo all species seemed to diverge simultaneously, whereas in L.
The hybrid group occupied an intermediate position between the latter two species, which is expected considering that the networks were based on genetic distance. In both analyses the two species in the source lakes were almost not distinguishable and were equally distantly related to the crater lake radiations. The fact that the great lake species were almost not distinguishable is probably due to the fact that the networks were based on genetic distance only.
Overall genetic differentiation between the great lake species was very low S2 Table —presumably due to their relatively large effective population sizes—leading to a low resolution in the networks. In the SNAPP analyses differences in effective population sizes were taken into account and the two species appeared probably therefore clearly diverged in the SNAPP trees in contrast to the networks.
A limitation of the described phylogenetic methods is that they do not take gene flow and changing population sizes into account. Moreover, the f3-statistics may not detect admixture events that happened before the split of the sympatric species. To overcome these limitations and furthermore infer the demographic history of the radiations we used fastsimcoal2 to perform coalescent simulations in pre-defined models and evaluated their fit against our empirical data summarized in the multi-dimensional site frequency spectrum SFS [ 66 , 67 ].
To better account for the complexity of multi-population models, we started with one-population models for both species in both great lakes the source populations. For each of the four populations six different models were tested S9A Fig. Since a signal of recent population expansion could be driven by rare alleles resulting from sequencing and genotyping error we repeated the analyses for A.
Next, we tested each crater lakes species together with A. We used A. Apoyo and are extremely rare in L. Moreover, our phylogenetic analyses suggest that both species in the source lakes are equally distantly related to the crater lake radiations Fig 3. We tested between nine and eleven models for each species and the same class of model differing only in migration was supported for all species S5 Table.
In the case of L. Apoyo gene flow between the lakes was not supported, whereas in L. However, the relative statistical support for the different models with or without migration is not very different S5 Table. Thus, our data strongly support the population size changes and the admixture event, but we have rather low power to distinguish the different migration scenarios.
For all species a model in which the colonization event happened after in forward time the bottleneck in the source populations was superior to a model in which we forced the colonization to happen before the bottleneck S5 Table.
This could indicate a limitation of the inference method, as a bottleneck in the source could lead to a loss of information and bias lineages to coalesce before backwards in time the bottleneck [ 68 ]. In order to test this, we simulated data using the maximum likelihood parameter estimates and data structure of cluster 2 in L.
Apoyo i. Importantly, we were able to infer the correct i. This suggests that theoretically we have enough power to correctly infer divergence times that happened before the bottleneck in the source populations. Finally we analyzed the demographic history in five-population models. Even though the f3-statistics did not provide evidence for secondary contact we wanted to make use of the likelihood framework to explicitly evaluate the evidence for the two main competing hypotheses: sympatric speciation after admixture from the source population and secondary contact followed by introgressive hybridization Fig 4.
In addition, we aimed to evaluate different topologies within the radiations to further investigate the support for simultaneous divergence events.
Furthermore, in L. Apoyo we did not include gene flow between the lakes, whereas in models of L. Managua into the crater lake species. Migration was assumed to be identical, that is only one migration parameter was used. In both radiations we added gene flow between the sympatric species, assuming it again to be identical and symmetrical.
While this assumption may be overly simplistic, including different migration parameters for all twelve possible migration routes would have likely over-parameterized our models. Schematic illustrations of the most supported demographic models of A B sympatric speciation and C D the alternative hypotheses of secondary contact for both radiations.
Parameter estimates are provided in Table 3. Species names are abbreviated by their first three letters. Note that migration between sympatric species and from L. Managua into L. Furthermore, growth was modelled to be exponential and not linear as depicted here. Models are not drawn to scale but merely indicate relative differences. Apoyo we tested six different models. Five models of sympatric speciation and one model of secondary contact were evaluated. Incorporating an initial split of A.
However, including another parameter to model an additional split of cluster 5 from the other two benthic species, as weakly indicated in our phylogenetic analysis, did not increase the likelihood. Removing gene flow among the sympatric species or the admixture event into the crater lake population before sympatric speciation strongly decreased the likelihood of the model. With four species a multitude of two-colonization scenarios is conceivable, yet it is computationally unfeasible and biologically not sensible to test all possible models [ 69 ].
Hence, based on the firmly established finding that A. Apoyo we formalized the main competing hypothesis of secondary contact as: an initial colonization by A.
This model was 2. Similar to L. Apoyo, for L. Modeling two intralacustrine divergence events, one between A. However, a sister relationships of A. Gene flow between the sympatric species and an admixture event before the onset of the radiation was again strongly supported Table 2.
Given the seemingly closer genetic affiliation of A. For this type of model we tried two different topologies, one in which A. The latter model was more strongly supported, yet it was about 70 times less likely than the model of sympatric speciation.
According to the maximum likelihood point estimates of the most supported model admixture prior to sympatric speciation of L. Nicaragua decreased from ca. Immediately afterwards A. The current population sizes of the four sympatric species range from about 6, to 43, individuals, with A. Migration i. Note that migration rates cannot be readily converted to number of individuals in growing populations, as the number of migrants is the product of the respective migration rate and the size of the receiving population, which is changing exponentially.
Although the secondary contact model had a lower likelihood we report the parameter estimates for comparability Table 3. For many parameters such as the timing of the bottleneck and population sizes we obtained similar estimates as for the model above, although some of the current sizes deviate. Most importantly, according to this model the first colonization would have happened ca.
The three benthic species would then have diverged approximately generations ago. In the most supported model admixture prior to sympatric speciation of L.
Managua happened ca. The colonization of L. Managua took place ca. Interestingly, the admixture event happened at around the same time as in L. Apoyo, about generations ago, but the admixture proportion was much higher with ca. Again similarly to L.
Apoyo, the first speciation event happened only a few generations after the admixture event. In this case it led to the species A. The current population sizes of A. Migration rates among the sympatric species are 8. Estimates of population sizes and migration rates were similar to those of the sympatric speciation model Table 3.
Whether geographical isolation is required for the initiation of speciation continues to be one of the most controversially discussed topics in evolutionary biology. Sympatric speciation is the most extreme case of primary divergence-with-gene-flow, in which geographic barriers play no role in reducing gene flow [ 17 , 18 ]. While theoretically possible, only few putative empirical examples exist and, together with Palm trees and some other plant lineages on Lord Howe island [ 39 , 70 ], crater lake cichlids in Cameroon and Nicaragua have been among the most widely-accepted examples [ 18 , 22 ].
Moreover, a recent study provided evidence that the divergence of two eco-morphs of cichlids in a crater lake in Tanzania happened in sympatry [ 71 ]. Yet, evidence for complex phases of secondary contact and gene flow among crater lake radiations and riverine populations of Cameroonian cichlids was provided recently [ 34 ] and some criticism of sympatric speciation in Nicaraguan Midas cichlids was expressed initially: the criticism was mainly concerned with the fact that not all species of Midas cichlids in the crater lake and source lake were taken into account and that no explicit explanation for the intermediate position of a species in multivariate genetic space was given [ 27 ].
In this study we took previous and more recent concerns [ 34 , 53 ] into account and evaluated the evidence for putative periods of allopatry in two radiations of Midas cichlids using a comprehensive genomic data set. Since Midas cichlids provide the rare advantage that the actual source populations of the crater lakes are known, we were able to reconstruct the demographic history of two crater lake radiations.
This allowed us not only to test for differential admixture of crater lake species—for which we found no evidence—but to detect a secondary colonization admixture from the same respective source population prior to the onset of the radiations. This admixture event would have been otherwise difficult to detect as it results in equal proportions of shared ancestry among all species within a radiation.
0コメント