In this article the authors describe an evolutionary convergence in mammals, birds, and reptiles, based on genomic data from NCBI. The evolution of different species and lineages is due to mutations that can appear and accumulate in organisms over time. Those mutations need a high functional potential and have to be conserved in time in order to form new species. The conservation of mutations can occur via selection pressure, mutational compensation, and/or by the separation of members from the same species by geological and environmental events.

In this comprehensive study, the authors describe, a genomic landscape of the parallel evolution by analysing functional nodal mutations (fNMs) by using different types of DNA (mitochondrial and nucleic), the thermostability of mtDNA encoding RNA genes, and the structural proximity of proteins, using the available 3D structures from PDB database. Functional nodal mutations (fNMs) can be separated in single nodal (fSNMs), recurrent nodal mutations (fRNMs), occured independently in unrelated lineages and recurrent combinations of nodal mutations (fRCNMs) recurred independently along with other nodal mutations in combinations in more than a single lineage. The recurrent ones can be taken in consideration the most when we are talking about the convergent adaptive responses, that means the parallel evolution of different species. In this study, one of the aim is to find the best candidate for this adaptive mutations that was present in the evolution of the amniotes. The compensated ones are used to identify the adaptive mutations. The main explanation for the convergent evolution is the presence of the recurrent nodal mutations. Many fNMs are in combination with potential compensatory mutations in RNA and protein-coding genes. The compensation of a functional mutation is the co-occurrence with additional mutations that are “affecting” the original function.

Results

In the article it is claimed that the evidence for parallel evolution is mainly due to the presence of a high number of uncompensated reccurent fNMs. The best candidate to show the parallel evolution is the emergence of body thermoregulation in mammals and birds, that seems to be independent.

The mtDNA, the maternal genetic information was used to identify the fNMs in the amniotes. The study is based on mtDNA from 1003 species and nDNA from 91 species. The mtDNA was used for the structure-base alignment for 24 mtDNA-encoded RNA genes (tRNA and rRNa) and 13 protein-coding gene. To this, they added 4 more mtDNA proteins with the 3D structure: CO1-3 and Cytb, as the cytochromes are highly conserved proteins across various species. The mtDNA genes are usually the same, but what seems to be different it is the order of the genes, that are changed by evolutionary rearrangements. Because of this, they first aligned the genes individually and after this, they concatenated the 37 proteins to the human mtDNA gene order.

The sequence alignment revealed a number of 25234 nodal non-synonimous and RNA gene mutations. To see the potential of this mutations, there were calculating a score that include: evolutionary conservation, physical-properties (of non-synonymous changes) and the molecular thermostability (the free estimated energy (?G) for the two RNA sequences was calculated before and after the mutational event). The score, from 1 to 9 is depending to the level of conservation and physico-chemical properties of the tested amino acid.After calculating the potential function score of all the nodal mutations, there were 3262 non-synonimous fNMs, mainly in RNA genes with mutations related to disease-causing.

The next step was to identify the best candidate for adaptive fNMs by studying the compensated and non-compensated mutations, but the approach chosen by the authors cannot reveal the exact order of compensation process. Meanwhile, there are some compensatory mutations that could gain lower functionality scores than the co-occurring fNMs. In the Figure 1, we can see a demonstration of the potential compensation and a possible adaptation in a protein-coding gene (COX2) through different species. The panel b shows the locations of the fNMs (S155T) and different other co-occurring compensatory mutations. The S155T mutation appears as independently re-occurrent as well as compensatory co-occurring mutations. As we can see, this approach is pure theoretical, because cannot show all the compensations, only the best ones, that got fixed in evolution. The Figure 2 shows the prevalence of different types of mutations that could be compensated or not. The predictive results reveal a high probability of fRCNMs to be compensated for RNA and protein-coding genes. Here are introduced also the information from the nDNA, that is compared with mtDNA in term of prevalence of the compensatory and non-compensatory mutations. Because there was a big difference of the number of species involved in this approach, the evolutionary resolution was reduced. So, the authors decided to analyze the same 91 species for mtDNA and nDNA and reducing the bias. Because of the reduction in the resolution, they redid the analysis by using the most ancient mutations, that occurs in deeper nodes in the case of mtDNA, but this revealed almost the same proccent as they were working with the 91 species (37% for the ancient mutations and 34% by including the younger ones) (Figure 2e & Supplementary 5b,c). So, the older mutations appear to be less compensated and this give more uncompensated mutations that are best candidates in the ancient adaptative mutations. In the supplementary Figures, the authors are using the OXPHOS complexes to compare the fNMs in mtDNA and nDNA by using 91 species. For the intra-mtDNA the albeit is less prominent (31%).

For the nDNA data is used the whole genome of the species. So, the information is much more comprehensive by the presence of a higher number of genes. In comparison with the mtDNA, the compensation prevalence is lower, having a difference of 10%, but in both case the proccent of possible compensation is higher than can be explained by the mutation rate or the chance.

In the end, to determine the best adaptive mutations over the evolution, they used the fRNMs from mtDNA, but maybe because of the low number of the samples, the result did not show any proof of the impact of non-compensated fRNMs in being the main reason for the convergent evolution. Instead, the nDNA revealed a significant pattern with highest number of potential non-compensated fRNMs shared between birds and mammals (N=51). The best candidates resulted by being the mutations in the genes related to the thermoregulation in the birds and mammals.

Conclusion

In this comprehensive study, the authors merged several information, including different types of DNA, from many species, with various physico-chemical parameters. The results of this work reveal, that the ancient functional mutation are the best for being studied, because of their possibility to overcome negative selective. The best candidates for the adaptive nodal mutations are in the end the non-compensated fNMs, that are in a higher presence in the case of old fNM. This seems to be the main helper for the evolution of the thermoregulation in birds and mammals. The protein analysis reinforces the main conclusion: for enriching the adaptative mutations, the non-compensated mutations are the best candidates.

Taken together this study provides new insights into how different lineages and species might have developed over time. It also shows a new way how to combine data from different sources. However, the authors fail in giving an adequate explanation for the fNMs, together with the fact that they lack references that describe this term makes the article difficult to understand, especially for people that are not from the field and this is in fact the contrary of how scientific writing should be done.

Levin & Mishmar, 2017, The genomic landscape of evolutionary convergence in mammals, birds and reptiles. Nature Ecology & Evolution 1: 0041

Introduction

Darwin’s finches from Galapagos and Cocos Island are classic example of young adaptive radiation, entirely intact because none of the species having become extinct as a result of human activity. They have diversified in beak sizes and shapes, feeding habits and diets in adapting to different food resources. Although traditional taxonomy of Darwin’s is based on morphology and has been largely supported by observations of breeding birds finches, in this paper, authors showed the results of whole-genome re-sequencing of 120 individuals representing all of the Darwin’s finch species inhabiting Galapagos archipelago (Fig. 1a) and two close relatives, trying to analyse patterns of intra-and interspecific genome diversity and phylogenetic relationships among the species.

Figure 1a. Sample location of Darwin’s finches

Summary and comments of the paper

The authors analyzed location and phylogeny of Darwin’s finches and found widespread evidence of interspecific gene flow that may have enhanced evolutionary diversification throughout phylogeny. They also reported discovery of a locus with the major effect on beak shape. They generated 10x sequence coverage per individual bird and using 2×100 base-pair (bp) paired-end reads and found evidence of introgression from three sources: ABBA-BABA tests, discrepancies between phylogenetic trees based on autosomal and sex linked loc, and mtDNA. Extensive sharing of genetic variation among populations was evident, particularly among ground and tree finches, with almost no fixed differences between species in each group. Their maximum-likelihood phylogenetic tree based on autosomal genome sequences is generally consistent with current taxonomy showing several interesting deviations (Fig. 1b).

Figure 1b. Phylogeny of Darwin’s finches

Revised and dated phylogeny of Darwin’s finches shows that the adaptive radiation took place in the past million years, with a rapid accumulation of species recently. Genomic characterization of the entire radiation revealed a striking connection between past and present evolution. Evidence of introgressive hybridization is found throughout the radiation, showing that hybridization always gives rise to species of mixed ancestry, which is explained in detail (species and location) in this paper. The most obvious morphological difference among Darwin’s finches concerns beak shape. The authors performed a genome wide scan on the basis of populations that are closely related but show different beak morphology. In this study, they indicated a polygenic basis for beak diversity, discovering 15 regions with strong genetic differentiation between groups of finches with blunt or pointed beaks. Their analysis revealed that ALX homeobox 1 is an excellent candidate for variation in beak morphology, because it encodes a paired-type homeodomain protein (transcription factor), that plays a crucial role in development of structures derived from craniofacial mesenchyme, the first branchial arch and the limb bud, and have influence on migration of cranial neural crest cells, highly relevant to beak development. They observed single nucleotide polymorphisms (SNPs) in ALX1 gene of various finch species and concluded that blunt haplotype has a long evolutionary history because it’s origin predates the radiation of vegetarian, tree and ground finches. The haplotype might have evolved by accumulating both coding and regulatory changes affecting ALX1 function. Natural selection and introgression affecting this locus have contributed to the diversification of beak shapes among Darwin’s finches and hence to their expanded utilization of food resources on different Galapagos islands.

Lamichhaney, S., Berglund, J., Almén, M., Maqbool, K., Grabherr, M., Martinez-Barrio, A., Promerová, M., Rubin, C., Wang, C., Zamani, N., Grant, B., Grant, P., Webster, M., & Andersson, L. (2015). Evolution of Darwin’s finches and their beaks revealed by genome sequencing Nature, 518 (7539), 371-375 DOI: 10.1038/nature14181

To identify the ecological and molecular basis of divergent evolution in the cichlid system, David et al. [1] sequenced the genomes and transcriptomes of five lineages of African cichlids, Pundamilia nyererei (endemic of Lake Victoria); Neolamprologus brichardi (endemic of Lake Tanganyika); Metriaclima zebra (endemic of Lake Malawi); Oreochromis niloticus (from rivers across northern Africa); Astatotilapia burtoni (from rivers connected to Lake Tanganyika). These five lineages diverged primarily through geographical isolation, and three of them subsequently underwent adaptive radiations in the three largest lakes of Africa. Authors comprehensively investigate the features from these massive genomic data. Here is some interesting finding:

Accelerated gene evolution was assessed by non-synonymous/synonymous ratio. Compare with stickleback fish, O. niloticus has significant higher ranks. And three gene, a ligand (bmp4), a receptor (bmpr1b) and an antagonist (nog2) in the BMP pathway, all known to influence cichlid jaw morphology, show accelerated rates of protein evolution in haplo-chromine cichlids.

East African cichlids, including O. niloticus, possess an unexpectedly large number of gene duplicates. The author found 280 duplication events in the lineage leading to the common ancestor of the radiations. And that was 4.5- to 6-fold increase in gene duplications relative to other clades, normalizing by the branch length. But again, same as high dN/dS analysis, there is no significant enrichment for particular gene pathway.

For the transposable elements insertion in different lineage, the authors claimed that there were three waves of TE insertions. And the TE inserted near the 5’ UTR increased gene expression significantly. Surprisingly, none of the five cichlid genomes showed any deficit of sense-oriented LINE insertions, which correspond to a time of transposable element insertions in the common ancestor of the haplo-tilapiine cichlids. This suggests that ancestral East African cichlids went through an extended period of relaxed purifying selection.

For people who interested in small RNA, the authors also found surprising excess number of novel microRNA emerge in cichlid and with wet lab experiment confirmation, these novel miRNAs were believed to alter gene expression in multiple organs.

Last but not the least, they also did a lot of population genetic analysis in closely related species of the genera Pundamilia, Mbipia and Neochromis, all of which are endemic to Lake Victoria. Because Lake Victoria is where the most recent radiation happened. Several hundred endemic species emerged within the past 15,000–100,000 years. Their results from Fst comparing suggests that (1) variation in coding sequence is most likely to be involved in the divergence of physiological and/or terminally differentiated traits like color; (2) regulatory variation is more important in morphological changes involving genes that have pleiotropic effects in developmental networks.

Conclusion:

Sometimes with massive interesting point, it is hard to get the simple answer for the ultimate question, why some species diversify so dramatically, some species did not. Here is the case for cichlid, which they try to address the question of what is the genomic substrate for adaptive radiation. The author’s conclusion is neutral and adaptive processes both make important contributions to the genetic basis of cichlid radiations.

Reference:

1. Brawand D, Wagner CE, Li YI, Malinsky M, Keller I, Fan S, Simakov O, Ng AY, Lim ZW, Bezault E, Turner-Maier, J. Johnson J, Alcazar R, Noh HJ, Russell P, Aken B, Alföldi J, Amemiya C, Azzouzi N, Baroiller J-F, Barloy-Hubler F, Berlin A, Bloomquist R, Carleton KL, Conte MA, D’Cotta H, Eshel O, Gaffney L, Galibert F, Gante HF, et al.: The genomic substrate for adaptive radiation in African cichlid fish. Nature 2014.

Protein abundances mainly determined by the balance of transcriptional and translational regulation. Because of the limited technology for the translational research, however, gene expression evolution was based almost entirely on studies of transcriptional regulation. With the quickly development of ribosome profiling–isolating and sequencing short fragments of mRNA bound by actively translating ribosomes–now we can study translational regulation conveniently and efficiently.

Simultaneous detection of regulatory divergence at two levels

In this paper, firstly, in order to assess the relative contributions of regulatory elements evolution to the changes in mRNA abundance and translation rate, the authors applied ribosome profiling and RNAseq to two species of Saccharomyces yeast (S. cerevisiae and  S. paradoxus )and their interspecific hybrid (figure 1).

Within hybrids, both alleles share the same trans-acting cellular environment. Therefore, different mRNA abundance or translation efficiency is caused by cis-regulatory divergence. By applied these methods, the authors showed cis-regulatory divergence in both transcription and translation are abundant, almost 35% orthologs have significant divergence in translational efficiency, as compared with 61% with significant divergence in mRNA abundance. Because they identified cis-regulatory elements change at two regulatory levels simultaneously, an interesting question will be asked is whether changes at the two levels could be reinforcing (acting at the same direction) or opposing (acting in opposite directions). Compared with transcriptional divergence, surprisingly, they found the majority of translation rate divergence has an opposed effect (figure 2).In other words, it means that translational divergence acts to buffer changes in mRNA abundance, leading to maintenance of similar protein abundances between species. This phenomenon makes me quite impressive, because it shows that measuring mRNA abundance to study the expression evolution is not appropriate for some genes its protein abundance are also determined by translation rate. At the same time, the authors found that trans-acting regulatory divergence is also widespread at both regulatory levels and has an opposing pattern between the two levels.

Polygenic selection at two levels of gene regulation

Secondly, the authors applied a recently developed approach to detect expression adaptation evolution (independently at the level of transcription and translation, as well as among all orthologs with reinforcing direction between the two regulatory levels) across functionally related groups of genes. The basic theory of this approach is based on the null hypothesis that under neutral divergence of cis-regulation, no consistent bias is expected in the relative parental direction of ASE (allele specific expression) among genes within a functional category. So, consistent directional bias across a functional group indicates that multiple independent cis-regulatory changes have altered gene expression in a coordinated fashion, and is evidence of lineage specific selection (figure 3). Totally, the authors tested 591 gene sets for deviation from neutral expected frequencies by means of a x2 test, and used a permutation framework to control for the number of tests performed. Then, the authors took S. cerevisiae strain S288c is more resistant to heavy metals than S. paradoxus strain because higher levels of both mRNA and translation in S. cerevisiae among genes whose loss leads to heavy metal sensitivity as an example to show how nature selection affected phenotype by shaping genotype. Their finding of natural selection on both levels of regulation, in some cases targeting the same gene sets, highlights the importance of considering both levels simultaneously.

Identification of conserved C-terminal peptide extensions

Liken alternative splicing, infrequent stop-codon readthrough–involves the ribosome inserting an amino acid into the growing peptide at a stop-codon position and continuing in-frame translation–is another way to increase peptide diversity. Taking advantage of multispecies riboprofiling data, next, the authors focused on searching for direct evidence of translation in putative C-terminal extensions at the transcriptome wide level. Results showed that putative C-terminal extension was detected in one or both species in 109 and 81 cases, respectively. For example, translation initiation factor eIF1A (TIF11) has conserved C-terminal extension between both species shown in (Figure 4). Tif11 is an essential protein that is involved in start codon identification whose C terminus interacts with Fun12, a GTPase also involved in initiation of translation. Stop-codon readthrough could potentially play a role in the regulation of this interaction. At the some time, they also observed several species-specific readthrough events, suggesting this may be an unappreciated source of regulatory divergence.

 Personal opinions

Because the transcription variation was buffered by translation evolution, it proves this variation has some kind of  harmful effect. In other words, it means the mutations caused transcription divergence between the two species are deleterious. So, I am quite wondering why these deleterious mutations got fixed among the population. In my opinion, two hypotheses may explain it.

First of all,  a mutation on a trans factor may affect the gene expression of many genes.  If the net fitness effect is positive, the mutation is favored by the positive selection, even though some genes’ expression may become suboptimal. Compensatory mutations provide a strategy to further improve the fitness. Alternatively, this kind of compensatory regulation was evolved to mitigate the trade-off (caused by pleiotropy) between gene expression abundance  and some other traits (such as expression noise, timing, location and so on). Taking the expression abundance  and noise as an example, because the regulation mechanisms of the both traits are coupled, if nature selection shapes the promoter architecture for decreasing expression noise(beneficial), it will also change the transcription abundance(deleterious). So, in order to compensate the deleterious effect caused by transcription level variation, translation rate evolved on an  opposing direction.

Artieri, C., & Fraser, H. (2013). Evolution at two levels of gene expression in yeast Genome Research, 24 (3), 411-421 DOI: 10.1101/gr.165522.113

 

Despite being the world’s most genetically diverse continent, only a handful of studies attempted to understand the genetic risks for diseases of the African populations. This study shines light not only on the genetic diversity to help learn more about the variants that are associated with malaria and hypertension, but also on the population history across sub-Saharan African populations. Beside the comprehensive map of the African variants obtained from genotypes of 1,481 individuals and whole-genome sequences of 320 individuals, authors offered a design of the array suitable to capturing variants of African populations.

Summary and comments of the paper

Population structure in SSA. Comparing ~2.2 million variants of 18 ethno-linguistic groups from sub-Saharan Africa (SSA), authors found modest differentiation among SSA populations (mean pairwise Fst = 0.019) and among Niger-Congo language groups (mean pairwise Fst = 0.009). In the article, authors suggested that the modest differentiation among Niger-Congo language group showed evidence for ‘Bantu expansion’. However, the Fig1.a shows sample distribution mostly next to the Western, East and South African coasts, rather then inside of continent where the Bantu expansion occurred, therefore indicating the sampling bias.

Furthermore, the authors found a high proportion of unshared and novel variants in Ethiopian population raising the importance of sequencing individuals across Africa.

Extending the analysis on population history in Africa, authors performed PCA analysis among African populations. The results suggested Euroasian gene flow and possible hunter-gatherer (HG) ancestry. To support the results from PCA analysis, unsupervised ADMIXTURE analysis (Fig1.c) showed similar results, with Euroasian admixture in Ethiopian population (Oromo, Amhara, Somali) and HG admixture in Biaka and Mbuti rainforest HG. Also, it is noticeable that Western Europeans and Central/Eastern Asian are well separated, indicating two branches of migration. ADMIXTURE analysis also pointed out the heterogeneous American population. The authors found that the most probable number of clusters of worlds populations in ADMIXTURE analysis is k = 18. Unfortunately, it is not clearly seen from the supplementary data that the CV error of clusters was lowest for k = 18.

The authors were interested in the more detailed gene flow effect among the African populations by masking Euroasian admixture. The results showed reduced population differentiation, suggesting that Euroasian admixture has a significant impact on those populations. Nevertheless, the authors did not discuss other possibilities of gene flow effects, such as allele surfing or allele fixation.

Population admixture in SSA. Using three population tests (f3 statistics), authors identified greatest proportion of Euroasian admixture in East Africa and HG admixture among Zulu and Sotho populations in South Africa. In the Fig2., authors showed that ancient Euroasian admixture appears in Yoruba population (~7,500-10,500 years), which gives support to Neanderthal ancestry in this African population.

Beside the observed HG admixture in South African samples, a HG admixture was also detected in Igbo populations and more recent in East Africa. The explanation of HG admixture in West and South Africa is related to Khoe-San populations, while in the East Africa is related to Mbuti rainforest HG populations dating to ~3,000 years ago.

Moreover, in the Fig 2. is observed an overlap of Euroasian and HG admixtures in East African populations (Barundi, Banyarwanda and Baganda) both dating to ~2,400-3,900 years ago. However, it was not commented in article do these populations have a presence of both admixtures or not and how is it possible.

Positive selection in SSA. The authors observed highly differentiated SNPs in two population structure approaches to inspect the positive selection due to local adaptive forces.

One approach was to observe highly differentiated SNPs between Euroasian and African populations. Beside some other locus-specific differentiations, they found evidence of differentiation in CR1 gene (chemokine receptor 1), previously reported as a gene implicated in malaria susceptibility. The authors also identified locus-specific differentiation within genes active in osmoregulation, specifically in hypertension. Given these results, the authors speculate that changes in these gene regions give basic support in differences of salt sensitivity and hypertension in sub-Saharian African populations.

Second approach observed highly differentiated SNPs among the African populations when Euroasian admixture was masked. It has not escaped to notice that the most of Euroasian admixture had main proportion in Ethiopian populations (as seen in Fig2. and Fig1c.). For that reason, masked Euroasian admixture might affect only Ethiopian population, but certainly cannot be generalized for other African populations that actually might have had a process of local adaptation. Consequently, the quote from paper “This suggests that a large proportion of differentiation observed among African populations could be due to Euroasian admixture, rather than adaptation to selective forces.” should be taken cum grano salis. The speculative reason why there is an observed Euroasian admixture in Ethiopian population is that nomadic groups survived the migration from North and cross the Sahara to inhabit current Eastern African territory.

However, the analysis of African populations with masked Euroasian admixture revealed 56 loci, together with highly differentiated variant in CSK gene region, involved in hypertension. The variant in CKS gene region showed complete linkage disequilibrium (LD) with another risk allele that correlates with latitude, giving the evidence of temperature local adaptation as a mechanism of hypertension.

Next, the authors were interested in comparison of populations situated in endemic and non-endemic regions to distinguish loci related to infectious diseases. They identified set of loci signals in gene regions for malaria, Lassa fever, trypanosomiasis and trachoma.

Designing medical genetics studies in Africa. Taking into consideration that there is a high genetic diversity on African continent, the importance to build the reference genome panel across African populations cannot be stressed enough since it enable us to shed light on most of the worlds variation. Current reference genome panels, such as HapMap and 1000Genome, were mostly built on European, American and Asian populations and they miss the African polymorphisms. This makes more difficult to recognize certain polymorphic biomarkers associated to spectrum of diseases in African populations.

Therefore, authors investigated imputation accuracy of two African populations using two different reference genome panels – 1000Genome project and ‘merged’ 1000Genome project with 320 whole genome-sequenced African individuals, respectively. They observed the slight improvement in imputation accuracy of the Sotho and Igbo populations using ‘merged’ reference genome panel (Fig3.).

Moreover, the authors compared the usefulness of current array chips to define the most favorable array design capturing African variants. Their results showed efficiency of HumanOmni2.5M array capturing >80% of common variation. Surprisingly, authors did not mention future possibilities of whole-genome sequencing in Africa that play a crucial role in modern research nor the drawbacks of microarray noisy data. The dropping costs of sequencing technology and its development would certainly bring more precise results.

Conclusion

In spite of the nicely presented results with plenty of supplementary data, the article raises lots of speculations and thoughtful discussions on migration of African populations. Furthermore, the PCA analysis in extended and supplementary data are hard to read due to many different symbols and colors. Easier representation of PCA analysis would help to distinct the patterns of African populations. However, the study provides invaluable resource of variant association information for several diseases that will increasingly improve medical diagnostics in African populations.

Gurdasani, D., Carstensen, T., Tekola-Ayele, F., Pagani, L., Tachmazidou, I., Hatzikotoulas, K., Karthikeyan, S., Iles, L., Pollard, M., Choudhury, A., Ritchie, G., Xue, Y., Asimit, J., Nsubuga, R., Young, E., Pomilla, C., Kivinen, K., Rockett, K., Kamali, A., Doumatey, A., Asiki, G., Seeley, J., Sisay-Joof, F., Jallow, M., Tollman, S., Mekonnen, E., Ekong, R., Oljira, T., Bradman, N., Bojang, K., Ramsay, M., Adeyemo, A., Bekele, E., Motala, A., Norris, S., Pirie, F., Kaleebu, P., Kwiatkowski, D., Tyler-Smith, C., Rotimi, C., Zeggini, E., & Sandhu, M. (2014). The African Genome Variation Project shapes medical genetics in Africa Nature, 517 (7534), 327-332 DOI: 10.1038/nature13997

Herbivorous stick insect (Timema cristinae) is an endemic species to California and adapted to different host plants, Adenostoma fasciculatum and Ceanothus spinosus. Researchers investigated the whole genome divergence by parallel speciation in a nice experimental set up with this species. T.cristinae individuals were sampled from four replicate population pairs where 3 of them were adjacent and one was 6.4 km far away. They annotated the reference genome for the species and resequenced 160 individuals sampled from field transplant experiment.

In the first part of the study, the data obtained was used to analyse the effects of adaptation on genomic divergence in different scales and aspects. The results showed that the divergence between ecotype pairs varied geographically. Principal components analysis and phylogenomic trees clustered the individuals by geography, not by host (Figure 1). Genome-wide fixation index was higher for the geographically separated population and lower for the adjacent populations. The genomic differentiation found in this study was lower than other studies that investigated the consequences of divergence. Afterwards, the researchers tested whether divergence between replicate population pairs frequently involved the same genomic regions by using the highly divergent single nucleotide polymorphisms (SNPs) between each population pair. The results showed that most of the SNPs (83%) were divergent only in one single population pair and divergence was non-parallel. The researchers discussed different evolutionary forces could gave rise to observed results. The remaining 17% of SNPs were represented in two or more population pairs and the pattern was in congruence with HMM.

In the second part of the study, the researchers performed a field transplant experiment. The design of the experiment let the researchers to maintain an “ancestor” population, a derived population hosted by Ceanothus and another by Adenostoma in 5 replicate blocks (Figure 3). The transplanted and “ancestor” insects were sampled after 1 year which corresponds to one generation of T.cristinae in nature. The genetic comparison between “ancestor” and transplanted individuals identified that most of the SNPs exhibit weak/moderate divergence. 213 SNPs exhibited larger allele frequency changes between hosts and were present in each block. ~15% of these SNPs were also supported by HMM and distributed across the genome.

Researchers examined the function of these genomic regions that might exhibit parallel divergence. The SNPs that pronounced as “parallel divergence SNPs” were present in four natural population pairs and located in the genes involved in metabolism and signal transduction (metal and calcium ion binding) pathways. They exhibited a 1.5-fold enrichment for being in coding regions of genes compared with all SNPs. Then, researchers performed the analysis for SNPs that were divergent between only a single population pair and also found the genes related to metal binding – non-parallel divergence? Researchers concluded these results as “(i)the result of adaptive divergence between host ecotypes, (ii) a case of parallelism at the functional level”.

Although some regions of the genome exhibited parallel divergence, the data showed that parallel speciation in Timema cristinae involve non-parallel genetic divergence.

Personal comments

The paper provides valuable information for speciation and annotation of the Timema cristinae genome will open new horizons. The sampling sites for the aim of the study are well planned and field transplant experiment is well designed. However, the paper is not so easy to read and follow because of massive background information. Also figures are not so helpful — I would not expect to see pie-charts between main figures. For example Figure2 and Figure 3, in which “quantification of parallel and nonparallel divergence across population pairs” and “allele frequency changes across the genome in a flied transplant experiment” was shown, are a bit redundant and can be in supplementary documents. Conversely, some figures in the supplementary material are worth to be in the paper since they help the reader to follow the story easily and visualise more valuable information than Figure 2 and Figure 3.

Nevertheless, this paper provides novel insights to the field and adds a genetic time-stamp on the complex process of speciation. I recommend reading this paper but allocating more time than usual to be able digest it.
Soria-Carrasco, V., Gompert, Z., Comeault, A., Farkas, T., Parchman, T., Johnston, J., Buerkle, C., Feder, J., Bast, J., Schwander, T., Egan, S., Crespi, B., & Nosil, P. (2014). Stick Insect Genomes Reveal Natural Selection’s Role in Parallel Speciation Science, 344 (6185), 738-742 DOI: 10.1126/science.1252136

Electric organs – organs that are capable of creating and discharging electricity – have evolved independently in at least six different lineages of fish (Torpediniformes, Rajiformes, Mormyroidea, Euteleostei, Siluriformes, Gymnotiformes) and play an important role in communication, navigation, defense and predation.

To investigate whether the convergent evolution of these organs has a common genetic basis, Jason Gallant and his coworkers studied the transcriptome of five species of electrogenic fish in three different lineages: Electrophorus electricus, Sternopygus macrurus, Eigenmannia virescens (Gymnotiforme), Malapterurus electricus (Siluriforme) and Brienomyrus brachyistius (Mormyroidea).

Electric organs are comprised of arrays of electrocytes – asymmetric cells that are enriched in cation-specific ion channels on one and sodium pumps on the opposing side. The resulting ion flux slowly charges the electrocyte membrane and upon activation by a neuronal stimulus, the voltage is discharged, generating an electrical pulse from the fish.

Although the morphology of electric organs and electrocytes varies substantially amongst these species, they are all muscle-derived tissue and originate developmentally from muscle progenitor cells.

Since this evolution of muscular to electrogenic tissue has occurred several times independently, the authors investigated, whether the underlying genetic mechanisms are shared.

To address this question, Gallant et al. first sequenced and assembled the genome of the electric eel, E. electricus. The authors further performed transcriptome analysis on multiple tissues of E. electricus as well as on pairs of skeletal muscle and electric organ tissue of two species within the same lineage (S. macrurus and E. virescens) and two species of distinct lineages (B. brachyistius, M. electricus).

Main findings

Across the species they observed common patterns of differential gene expression between electric organs versus skeletal muscles, which they attributed to the following five key mechanisms for the evolution of electrogenic tissue:

• Alteration of the expression of myogenic transcription factors
• Increased excitability by upregulation of transporters and ion channels
• Enhanced isolation and direction of electrical currents by the upregulation of proteins in the connective tissue
• Decrease in contractility by down-regulation of sarcomere associated genes
• Increase of cell size by up-regulation of factors in the Insulin-like growth factor signaling pathway

Gallant et al. propose a convincing set of changes in gene expression to explain the functional differences between electric organ and muscle tissue. The fact that these mechanisms seem to be conserved in five species of electrogenic fish is an intriguing, yet not entirely surprising observation: presumably there are strong constraints on keeping muscle function intact while opening the potential for specialization to electric tissue – it would be interesting to inquire if – and to what extent – these shared expression differences are reflected on the genetic level.

Given the ambitious goal of uncovering the basis of electric organ evolution, I think the sampling of only one individual per species is problematic, despite the authors´ main interest in inter-species similarities. For an evolutionary approach the importance of intra-species variations should not be neglected and certainly requires a larger number of individuals. Including specimens from the other electrogenic lineages (e.g. Torpediniformes, Rajiformes) or (genetic) comparisons between electrogenic and non-electrogenic descendants within a lineage would have further strengthened the evolutionary aspect.

Lastly one could suspect that phylogenetically “older” electric organs have undergone a more advanced tissue specialization, resulting in a reduced “muscle profile” but the authors do neither raise this question, nor provide any information on this aspect.

Nevertheless I can highly recommend reading and discussing the paper – the ideas and methodology are presented in a clear language, the figures are appealing and – apart from the histological pictures – informative and well explained.

Although the results on what is in charge of the evolution of electric organs holds no shocking surprise yet, the research is still electrifying.

Gallant, J., Traeger, L., Volkening, J., Moffett, H., Chen, P., Novina, C., Phillips, G., Anand, R., Wells, G., Pinch, M., Guth, R., Unguez, G., Albert, J., Zakon, H., Samanta, M., & Sussman, M. (2014). Genomic basis for the convergent evolution of electric organs Science, 344 (6191), 1522-1525 DOI: 10.1126/science.1254432

Summary

By applying a population genomic framework the authors analyzed 89 complete nuclear genomes of polar bears and brown bears. They showed that two species diverged 479-343 thousand years ago (kya) and found 16 genes under strong positive selection on the polar bear in comparison with the brown bear. They analyzed more precisely nine of these genes that are known to be associated with high risk of cardiomyopathy and vascular diseases in humans. However, in polar bears these genes are responsible for an important reorganization of the cardiovascular system which allowed them to survive in extreme life’s conditions within Arctic Circle (e.g. very low temperatures, high physical activity in cold water, high demand of the energy, hyperlipidemic diet, etc).

Principal results and discussion

In brief, authors analyzed 79 polar bears and 10 brown bears from different areas. First of all, they sequenced and de novo assembled a polar bear reference genome. The data analysis was performed by population genomic framework.

They determined joint demographic history of polar bear and brown bear, inferred effective population sizes and estimated the divergence time of two species around 479-343 kya by using two complementary methods – identity by state (IBS) tracts of DNA and diffusion approximation for demographic interference (????) (Fig. 2, taken from original publication). Both of these approaches based on past population size changes. This principal point allows avoiding mistakes from a simple isolation-with-migration model which does not consider ancient population size changes and overestimates the divergence time. The same methods together with D statistics (the ABBA-BABA test) also allowed authors to investigate the patterns and direction of gene flow between two populations from their split.

Liu and colleagues discussed evidences suggesting population bottlenecks and reduction in effective population size of polar bear that was accompanied by period of migration. Moreover, on the Figure 2a they indicate an important size reduction of joint past population predating the divergence of polar bear and brown bear. In turn, this divergence was also followed by decrease of population size in polar bear that was greater and longer than in brown bear. However, later we observe an increase of both population sizes which started earlier in brown bears.

Also, authors indicate that the gene flow corresponds to four migration waves between polar bear and brown bear populations (dated 319 – 148 kya) which probably continue till the present. However, they didn’t exclude that this migration could have occurred earlier because IBS method has some limitations to detect migration close to split time. On the Figure 2b authors constructed a best-fit model reflected that a great part of the gene flow took place from polar bear to brown bear. Apparently, these data do not correspond to results obtained with isolation-with-migration model proposed earlier (Fig.2c).

Next big part of paper is devoted to investigation of divergence and of polymorphism between polar bear and brown bear by analyzing of polar bear’s genes under positive selection that managed the reorganization of their cardiovascular system and facilitated adaptation to extreme living conditions of the High Arctic after divergence. Also, authors tried to find different evolutionary changes in protein coding sequences as well as they performed analysis of the coding regions of 19.822 genes and they used the giant panda as an outgroup. Interestingly, most of genes under positive selection were associated with vital processes such as cardiovascular function, heart development, blood coagulation, adipose tissue development and metabolism of fatty acids. However, in humans these genes are responsible for cardiomyopathy and vascular diseases.

Conclusions and personal comments

On my opinion, this paper touches a very interesting issue. Especially, I appreciated that by using genomic approaches Liu with colleagues suppose genetic causes underlying development of human cardiovascular diseases. It is an interesting point even if it needs to be proved by farther investigations, for example, by analyzing other species which are more closely related to humans.

Personally, I think that the genomic approach chosen by authors allows us to elucidate the divergence time, speciation and evolutionary history of two species of bears that were not so clear by the present time. However, from my point of view, a deeper analysis could be performed to investigate the current state of both brown bear and polar bear populations because in the paper authors indicate the possibility of recent gene flow between polar bear and brown bear.

Reference

Liu, S., Lorenzen, E., Fumagalli, M., Li, B., Harris, K., Xiong, Z., Zhou, L., Korneliussen, T., Somel, M., Babbitt, C., Wray, G., Li, J., He, W., Wang, Z., Fu, W., Xiang, X., Morgan, C., Doherty, A., O’Connell, M., McInerney, J., Born, E., Dalén, L., Dietz, R., Orlando, L., Sonne, C., Zhang, G., Nielsen, R., Willerslev, E., & Wang, J. (2014). Population Genomics Reveal Recent Speciation and Rapid Evolutionary Adaptation in Polar Bears Cell, 157 (4), 785-794 DOI: 10.1016/j.cell.2014.03.054

Linnen et al. are interested in coloration of the deer mice (Peromyscus maniculatus) present on the light-colored soils of the Nebraska Sand Hills. Adaptation for crypsis is the strongest hypothesis to explain prevalence of the light morph compared to the black morph,  and they wanted to dissect the genetic basis of this adaptation. This study is composed of two main parts : first to understand and to evaluate the complexity of coloration phenotype and then to find the mutations responsible of those variation in traits and on which morph selection is acting on. First of all, they implemented an experimentation with plasticine models to count the number of attacks on each coloration morph. As they expected, statistical test reveals that the dark models are significantly more attacked than light models. Closer inspection reveals multiple pigmentation traits and pattern that differ between the light and the dark morph to compose complex coloration phenotype (particularly for dorsal hue and brightness, ventral color, dorsal ventral boundary and tail stripe). In previous study, they found that recent change in dorsal fur to light color is mainly caused by a change at the Agouti locus.

Before looking at point mutation in this locus, they wanted to see if color and color pattern are, or not, dependant. Principal component analysis (PCA) reveals that the phenotypes in this wild population were largely independent suggesting multiples independent genetic control. To test this hypothesis, they used NGS to generate polymorphism data for ~2100 unlinked regions and a smaller region containing Agouti and all known regulatory elements. Single-SNP linear regressions allowed first to find which mutation is associated with the different color traits. Then, using the residues of those regressions, multiple-SNP analysis are done with the other SNPs to look for dependent effect of mutations. We must keep in mind that choice of SNP for the first and the following regressions matters (see explanation in figure 1). Their results are really interesting as they find that most of the color traits are associated with a unique set of SNPs (except for one deletion associated with both ventral color and tail stripe), and that no single set of polymorphisms could account for variation across the five traits. Most interestingly, many of those SNPs fell in or near regions containing regulatory elements suggesting that multiple molecular mechanisms are involved in color adaptation in these Sand Hills mice.

One remark can be made about figure 2 panel C of the article, as it is difficult to see differences between the gray and white circles. Or it is something important as the gray circles represent significant SNPs after correction of false discovery…. It is important to note that no gray circles are found for ventral color trait, and that only one red circle is found (significant SNP after false discovery and bonferroni correction). Moreover, PVE (percent of variation of traits explain by SNPs) is 16%, which the smaller value of all traits. This could mean that this trait variation could be under control of other genes that were not sequenced here.

Which lead them to answer the two questions:

• Does single mutation have pleiotropic effect?

The response is mainly no.

• Do mutations have small and independent effects?

The response is yes, with SNPs falling in coding and non-coding (regulatory elements) regions.

It is useful now to test for positive selection on Agouti and SNPs. To do so, they compare a neutral model to a model with selection (created after calculating a coefficient selection) using simulations and a likelihood ratio test. The neutral model is a demographic model that they previously built using dadi. Also in figure 3 panel A, the simulations are done using all haplotypes, but because sweep (recent mutation) is supposed to be associated with light phenotype only, it is important to restricted the simulation after removing the dark haplotypes. On figure 3, y-axis corresponds to values of likelihood ratio (LR). Bigger the values of LR are (big peaks), less the models fit to each other. Also, when LR value is around 0 there is no departure from neutral model to the model under selection. Finally, as you can see on panel B and C, peaks are much more numerous and bigger when restricted to light alleles. Also, panel C is interesting as it zooms on the most strongly associated polymorphism for each trait, and helps to compare results from dark and light morphs (black line = dark haplotypes, colored line = light haplotypes). Peaks in panel B are found to be significant and clustered around the location of SNPs, which is consistent with recent selection acting on, or near, color-associated SNPs in light haplotypes. Finally, results from a comparison between dark and light-associated alleles are concur with multiple targets of selection among the light, but not dark, alleles of Agouti. Last but not least, strength of selection analysis reveal that selection coefficients (estimated using a maximum likelihood approach) are greater in traits linked to light allele compared to traits of dark allele. For example, for d-v boundary traits, selection coefficient s is 0.42 for light allele, and 0.067 for dark allele. Values of s in dark-associated alleles are really small compared to the light-alleles. Moreover, there is a positive correlation between PVE values (percent of variation of traits explain by SNP) and selection coefficient s across all light associated SNP.

To conclude, their results are finally consistent with Fisher geometric model of adaptation. Small and independent effects of mutation can lead to a more important pleiotropy of a gene (as here with agouti locus leading to complex coloration in the deer mice of Nebraska). Finally, Linnen et al.  want us to remember that it is individual mutations, not genes that bring population closer to its phenotypic optimum.

From my point of view this Science letter is well written, clear and concise to answer a question that have important impacts in evolutionary biology. A deeper look allows the reader to appreciate the complexity of the issue and the good work done by those researchers.

Linnen, C., Poh, Y., Peterson, B., Barrett, R., Larson, J., Jensen, J., & Hoekstra, H. (2013). Adaptive Evolution of Multiple Traits Through Multiple Mutations at a Single Gene Science, 339 (6125), 1312-1316 DOI: 10.1126/science.1233213

The aim of this work is to investigate the hypothesis according to which convergent phenotypes are not just a lucky coincidence produced by different point-mutations occurred in different species, but rather that a convergent phenotype is associated with the same mutation in all the species involved, and that these mutations are not happen by chance but are pushed by adaptation to the environment.

In order to do it, this group analysed sequence identities in the genomes of species that developed independently echolocation, certainly a very complex feature that it’s hard to believe it has developed in different species just by chance.

The first step was building the gene set to work on. Therefore, they sequenced the genome of four different bat species (both echolocating and non-echolocating), and acquired online the coding gene sequences (CDSs) of other bat species, a dolphin species and other non-echolocating animals (dog, cow, horse, mouse and human). In order to remove potential error sources, all ambiguous sites/codons were removed, and they selected only those genes that had no missing data and whose homologous were present in at least six of the investigated species. The final gene pool was about 2000 CDSs.

In order to evaluate the genetic convergence of these genes, they considered three different phylogenetic tree hypothesis.

H₀: the real phylogenetic tree

H₁: a fake phylogenetic tree in which all echolocating bats (in brown) are in the same branch

H₂: a fake phylogenetic tree in which all echolocating animals (including dolphins) are in the same branch.

Then they proceeded with the alignment of all the CDSs in these species, and measured their SSLS

SSLS: site-wise log-likelihood support. A score that, based on the alignment of each amino acid in each gene of the investigated species, tells how much that alignment fits to a tree hypothesis.

If echolocating animals really have a sequence convergence in one or more genes, the alignment of these genes in echolocating animals will be better than expected, so it will fit to the fake phylogenetic trees (H₁-H₂) better than to the real tree.

The fitness (SSLS) to the three trees is evaluated, and ?SSLS of each gene is calculated.

?SSLS: difference between SSLS to H₀ and the SSLS to one of the fake trees. If the alignment of a gene fits more to H₁ than to H₀ (sequence convergence), then ?SSLS(H₁) = SSLS(H0) – SSLS(H₁) of that specific gene will be a negative number. This calculation is applied to all the genes of the pool. Then they checked the scores of those genes which were already found to be involved in echolocation.

All the hearing genes known to be involved in echolocation show negative ?SSLS, as well as other genes involved in hearing (green) and vision (blue), supporting the hypothesis of genetic convergence. With H₂, the ?SSLS scores are still negative, but not so much, because the species involved are very distant and the hypothesis is more stringent.

In order to evaluate whether or not these converging mutation were pushed by adaptation, they used the ? ratio as a score of the influence of selection for each site.

If ? > 1, it means that there is adaptation.

For each site, the ? score was correlated with ?SSLS. For instance, if in a gene ?SSLS and ? are inversely correlated (positive ? and negative ?SSLS) it means that adaptation has pushed those particular sites towards the same mutations in the different echolocating species.

As expectable, after screening all the gene pool, all possible kinds of correlations were found. What is interesting is that some genes involved in hearing and vision had a strong correlation between ? and ?SSLS that supports the hypothesis of convergence by adaptation (bottom left in the picture).

However, it’s not really specified if all the genes involved in echolocation showed this correlation.

The paper shows a very interesting approach for correlating genetic paths in evolution in different species, and some of the results strongly suggest the validity of the hypothesis of genetic convergence by adaptation. Nevertheless, it seems that the authors are trying to prove their point skipping some elements that would need to be further investigated.

First of all, it seems that negative ?SSLS is present also in many genes that haven’t shown, so far, any association with echolocation, a fact that might bring some doubts about how much the ?SSLS information is actually important in this context.

Moreover, in the second part of the paper, they show that one of the proteins with the best ? score is Cdk1, saying that it supports their hypothesis because this protein is important for the development of hair cells in the inner ear. But Cdk1 is a ubiquitary protein, necessary in the cell cycle of every type of cells, so it’s not a protein specifically involved in hearing.

Parker, J., Tsagkogeorga, G., Cotton, J., Liu, Y., Provero, P., Stupka, E., & Rossiter, S. (2013). Genome-wide signatures of convergent evolution in echolocating mammals Nature, 502 (7470), 228-231 DOI: 10.1038/nature12511