Expression level, cellular compartment and metabolic network position all influence the average selective constraint on mammalian enzymes

The regulatory genome constrains protein sequence evolution: implications for the search for disease-associated genes

The development of explanatory models of protein sequence evolution has broad implications for our understanding of cellular biology, population history, and disease etiology. Here we analyze the GTEx transcriptome resource to quantify the effect of the transcriptome on protein sequence evolution in a multi-tissue framework. We find substantial variation among the central nervous system tissues in the effect of expression variance on evolutionary rate, with highly variable genes in the cortex showing significantly greater purifying selection than highly variable genes in subcortical regions (Mann-Whitney U p = 1.4 × 10^-4). The remaining tissues cluster in observed expression correlation with evolutionary rate, enabling evolutionary analysis of genes in diverse physiological systems, including digestive, reproductive, and immune systems. Importantly, the tissue in which a gene attains its maximum expression variance significantly varies (p = 5.55 × 10^-284) with evolutionary rate, suggesting a tissue-anchored model of protein sequence evolution. Using a large-scale reference resource, we show that the tissue-anchored model provides a transcriptome-based approach to predicting the primary affected tissue of developmental disorders. Using gradient boosted regression trees to model evolutionary rate under a range of model parameters, selected features explain up to 62% of the variation in evolutionary rate and provide additional support for the tissue model. Finally, we investigate several methodological implications, including the importance of evolutionary-rate-aware gene expression imputation models using genetic data for improved search for disease-associated genes in transcriptome-wide association studies. Collectively, this study presents a comprehensive transcriptome-based analysis of a range of factors that may constrain molecular evolution and proposes a novel framework for the study of gene function and disease mechanism.

Gene connectivity and enzyme evolution in the human metabolic network

BACKGROUND

Determining the factors involved in the likelihood of a gene being under adaptive selection is still a challenging goal in Evolutionary Biology. Here, we perform an evolutionary analysis of the human metabolic genes to explore the associations between network structure and the presence and strength of natural selection in the genes whose products are involved in metabolism. Purifying and positive selection are estimated at interspecific (among mammals) and intraspecific (among human populations) levels, and the connections between enzymatic reactions are differentiated between incoming (in-degree) and outgoing (out-degree) links.

RESULTS

We confirm that purifying selection has been stronger in highly connected genes. Long-term positive selection has targeted poorly connected enzymes, whereas short-term positive selection has targeted different enzymes depending on whether the selective sweep has reached fixation in the population: genes under a complete selective sweep are poorly connected, whereas those under an incomplete selective sweep have high out-degree connectivity. The last steps of pathways are more conserved due to stronger purifying selection, with long-term positive selection targeting preferentially enzymes that catalyze the first steps. However, short-term positive selection has targeted enzymes that catalyze the last steps in the metabolic network. Strong signals of positive selection have been found for metabolic processes involved in lipid transport and membrane fluidity and permeability.

CONCLUSIONS

Our analysis highlights the importance of analyzing the same biological system at different evolutionary timescales to understand the evolution of metabolic genes and of distinguishing between incoming and outgoing links in a metabolic network. Short-term positive selection has targeted enzymes with a different connectivity profile depending on the completeness of the selective sweep, while long-term positive selection has targeted genes with fewer connections that code for enzymes that catalyze the first steps in the network.

REVIEWERS

This article was reviewed by Diamantis Sellis and Brandon Invergo.

Functional characterization of SMN evolution in mouse models of SMA

Spinal Muscular Atrophy (SMA) is a monogenic neurodegenerative disorder and the leading genetic cause of infantile mortality. While several functions have been ascribed to the SMN (survival motor neuron) protein, their specific contribution to the disease has yet to be fully elucidated. We hypothesized that some, but not all, SMN homologues would rescue the SMA phenotype in mouse models, thereby identifying disease-relevant domains. Using AAV9 to deliver Smn homologs to SMA mice, we identified a conservation threshold that marks the boundary at which homologs can rescue the SMA phenotype. Smn from Danio rerio and Xenopus laevis significantly prevent disease, whereas Smn from Drosophila melanogaster, Caenorhabditis elegans, and Schizosaccharomyces pombe was significantly less efficacious. This phenotypic rescue correlated with correction of RNA processing defects induced by SMN deficiency and neuromuscular junction pathology. Based upon the sequence conservation in the rescuing homologs, a minimal SMN construct was designed consisting of exons 2, 3, and 6, which showed a partial rescue of the SMA phenotype. While a significant extension in survival was observed, the absence of a complete rescue suggests that while the core conserved region is essential, additional sequences contribute to the overall ability of the SMN protein to rescue disease pathology.

Influence of pathway topology and functional class on the molecular evolution of human metabolic genes

Metabolic networks comprise thousands of enzymatic reactions functioning in a controlled manner and have been shaped by natural selection. Thanks to the genome data, the footprints of adaptive (positive) selection are detectable, and the strength of purifying selection can be measured. This has made possible to know where, in the metabolic network, adaptive selection has acted and where purifying selection is more or less strong and efficient. We have carried out a comprehensive molecular evolutionary study of all the genes involved in the human metabolism. We investigated the type and strength of the selective pressures that acted on the enzyme-coding genes belonging to metabolic pathways during the divergence of primates and rodents. Then, we related those selective pressures to the functional and topological characteristics of the pathways. We have used DNA sequences of all enzymes (956) of the metabolic pathways comprised in the HumanCyc database, using genome data for humans and five other mammalian species. We have found that the evolution of metabolic genes is primarily constrained by the layer of the metabolism in which the genes participate: while genes encoding enzymes of the inner core of metabolism are much conserved, those encoding enzymes participating in the outer layer, mediating the interaction with the environment, are evolutionarily less constrained and more plastic, having experienced faster functional evolution. Genes that have been targeted by adaptive selection are endowed by higher out-degree centralities than non-adaptive genes, while genes with high in-degree centralities are under stronger purifying selection. When the position along the pathway is considered, a funnel-like distribution of the strength of the purifying selection is found. Genes at bottom positions are highly preserved by purifying selection, whereas genes at top positions, catalyzing the first steps, are open to evolutionary changes. These results show how functional and topological characteristics of metabolic pathways contribute to shape the patterns of evolutionary pressures driven by natural selection and how pathway network structure matters in the evolutionary process that shapes the evolution of the system.

Metabolic Determinants of Enzyme Evolution in a Genome-Scale Bacterial Metabolic Network

Different genes and proteins evolve at very different rates. To identify the factors that explain these differences is an important aspect of research in molecular evolution. One such factor is the role a protein plays in a large molecular network. Here, we analyze the evolutionary rates of enzyme-coding genes in the genome-scale metabolic network of Escherichia coli to find the evolutionary constraints imposed by the structure and function of this complex metabolic system. Central and highly connected enzymes appear to evolve more slowly than less connected enzymes, but we find that they do so as a by-product of their high abundance, and not because of their position in the metabolic network. In contrast, enzymes catalyzing reactions with high metabolic flux-high substrate to product conversion rates-evolve slowly even after we account for their abundance. Moreover, enzymes catalyzing reactions that are difficult to by-pass through alternative pathways, such that they are essential in many different genetic backgrounds, also evolve more slowly. Our analyses show that an enzyme's role in the function of a metabolic network affects its evolution more than its place in the network's structure. They highlight the value of a system-level perspective for studies of molecular evolution.

Network Architecture and Mutational Sensitivity of the C. elegans Metabolome

A fundamental issue in evolutionary systems biology is understanding the relationship between the topological architecture of a biological network, such as a metabolic network, and the evolution of the network. The rate at which an element in a metabolic network accumulates genetic variation via new mutations depends on both the size of the mutational target it presents and its robustness to mutational perturbation. Quantifying the relationship between topological properties of network elements and the mutability of those elements will facilitate understanding the variation in and evolution of networks at the level of populations and higher taxa. We report an investigation into the relationship between two topological properties of 29 metabolites in the C. elegans metabolic network and the sensitivity of those metabolites to the cumulative effects of spontaneous mutation. The correlations between measures of network centrality and mutability are not statistically significant, but several trends point toward a weak positive association between network centrality and mutational sensitivity. There is a small but significant negative association between the mutational correlation of a pair of metabolites (r_M) and the shortest path length between those metabolites. Positive association between the centrality of a metabolite and its mutational heritability is consistent with centrally-positioned metabolites presenting a larger mutational target than peripheral ones, and is inconsistent with centrality conferring mutational robustness, at least in toto. The weakness of the correlation between r_M and the shortest path length between pairs of metabolites suggests that network locality is an important but not overwhelming factor governing mutational pleiotropy. These findings provide necessary background against which the effects of other evolutionary forces, most importantly natural selection, can be interpreted.

Correlates of evolutionary rates in the murine sperm proteome

Background

Protein-coding genes expressed in sperm evolve at different rates. To gain deeper insight into the factors underlying this heterogeneity we examined the relative importance of a diverse set of previously described rate correlates in determining the evolution of murine sperm proteins.

Results

Using partial rank correlations we detected several major rate indicators: Phyletic gene age, numbers of protein-protein interactions, and survival essentiality emerged as particularly important rate correlates in murine sperm proteins. Tissue specificity, numbers of paralogs, and untranslated region lengths also correlate significantly with sperm genes’ evolutionary rates, albeit to a lesser extent. Multifunctionality, coding sequence or average intron lengths, and mean expression level have insignificant or virtually no independent effects on evolutionary rates in murine sperm genes. Gene ontology enrichment analyses of three equally sized murine sperm protein groups classified based on their evolutionary rates indicate strongest sperm-specific functional specialization in the most quickly evolving gene class.

Conclusions

We propose a model according to which slowly evolving murine sperm proteins tend to be constrained by factors such as survival essentiality, network connectivity, and/or broad expression. In contrast, evolutionary change may arise especially in less constrained sperm proteins, which might, moreover, be prone to specialize to reproduction-related functions. Our results should be taken into account in future studies on rate variations of reproductive genes.

Electronic supplementary material

The online version of this article (10.1186/s12862-018-1157-6) contains supplementary material, which is available to authorized users.

Diet shifts provoke complex and variable changes in the metabolic networks of the ruminal microbiome

BACKGROUND

Grazing mammals rely on their ruminal microbial symbionts to convert plant structural biomass into metabolites they can assimilate. To explore how this complex metabolic system adapts to the host animal's diet, we inferred a microbiome-level metabolic network from shotgun metagenomic data.

RESULTS

Using comparative genomics, we then linked this microbial network to that of the host animal using a set of interface metabolites likely to be transferred to the host. When the host sheep were fed a grain-based diet, the induced microbial metabolic network showed several critical differences from those seen on the evolved forage-based diet. Grain-based (e.g., concentrate) diets tend to be dominated by a smaller set of reactions that employ metabolites that are nearer in network space to the host's metabolism. In addition, these reactions are more central in the network and employ substrates with shorter carbon backbones. Despite this apparent lower complexity, the concentrate-associated metabolic networks are actually more dissimilar from each other than are those of forage-fed animals. Because both groups of animals were initially fed on a forage diet, we propose that the diet switch drove the appearance of a number of different microbial networks, including a degenerate network characterized by an inefficient use of dietary nutrients. We used network simulations to show that such disparate networks are not an unexpected result of a diet shift.

CONCLUSION

We argue that network approaches, particularly those that link the microbial network with that of the host, illuminate aspects of the structure of the microbiome not seen from a strictly taxonomic perspective. In particular, different diets induce predictable and significant differences in the enzymes used by the microbiome. Nonetheless, there are clearly a number of microbiomes of differing structure that show similar functional properties. Changes such as a diet shift uncover more of this type of diversity.

The players may change but the game remains: network analyses of ruminal microbiomes suggest taxonomic differences mask functional similarity

By mapping translated metagenomic reads to a microbial metabolic network, we show that ruminal ecosystems that are rather dissimilar in their taxonomy can be considerably more similar at the metabolic network level. Using a new network bi-partition approach for linking the microbial network to a bovine metabolic network, we observe that these ruminal metabolic networks exhibit properties consistent with distinct metabolic communities producing similar outputs from common inputs. For instance, the closer in network space that a microbial reaction is to a reaction found in the host, the lower will be the variability of its enzyme copy number across hosts. Similarly, these microbial enzymes that are nearby to host nodes are also higher in copy number than are more distant enzymes. Collectively, these results demonstrate a widely expected pattern that, to our knowledge, has not been explicitly demonstrated in microbial communities: namely that there can exist different community metabolic networks that have the same metabolic inputs and outputs but differ in their internal structure.

Patterns of variation during adaptation in functionally linked loci

An understanding of the distribution of natural patterns of genetic variation is relevant to such fundamental biological fields as evolution and development. One recent approach to understanding such patterns has been to focus on the constraints that may arise as a function of the network or pathway context in which genes are embedded. Despite theoretical expectations of higher evolutionary constraint for genes encoding upstream versus downstream enzymes in metabolic pathways, empirical results have varied. Here we combine two complementary models from population genetics and enzyme kinetics to explore genetic variation as a function of pathway position when selection acts on whole-pathway flux. We are able to qualitatively reproduce empirically observed patterns of polymorphism and divergence and suggest that expectations should vary depending on the evolutionary trajectory of a population. Upstream genes are initially more polymorphic and diverge faster after an environmental change, while we see the opposite trend as the population approaches its fitness optimum.

Flagellated algae protein evolution suggests the prevalence of lineage-specific rules governing evolutionary rates of eukaryotic proteins

Understanding the general rules governing the rate of protein evolution is fundamental to evolutionary biology. However, attempts to address this issue in yeasts and mammals have revealed considerable differences in the relative importance of determinants for protein evolutionary rates. This phenomenon was previously explained by the fact that yeasts and mammals are different in many cellular and genomic properties. Flagellated algae species have several cellular and genomic characteristics that are intermediate between yeasts and mammals. Using partial correlation analyses on the evolution of 6,921 orthologous proteins from Chlamydomonas reinhardtii and Volvox carteri, we examined factors influencing evolutionary rates of proteins in flagellated algae. Previous studies have shown that mRNA abundance and gene compactness are strong determinants for protein evolutionary rates in yeasts and mammals, respectively. We show that both factors also influence algae protein evolution with mRNA abundance having a larger impact than gene compactness on the rates of algae protein evolution. More importantly, among all the factors examined, coding sequence (CDS) length has the strongest (positive) correlation with protein evolutionary rates. This correlation between CDS length and the rates of protein evolution is not due to alignment-related issues or domain density. These results suggest no simple and universal rules governing protein evolutionary rates across different eukaryotic lineages. Instead, gene properties influence the rate of protein evolution in a lineage-specific manner.

Three independent determinants of protein evolutionary rate

One of the most widely accepted ideas related to the evolutionary rates of proteins is that functionally important residues or regions evolve slower than other regions, a reasonable outcome of which should be a slower evolutionary rate of the proteins with a higher density of functionally important sites. Oddly, the role of functional importance, mainly measured by essentiality, in determining evolutionary rate has been challenged in recent studies. Several variables other than protein essentiality, such as expression level, gene compactness, protein-protein interactions, etc., have been suggested to affect protein evolutionary rate. In the present review, we try to refine the concept of functional importance of a gene, and consider three factors-functional importance, expression level, and gene compactness, as independent determinants of evolutionary rate of a protein, based not only on their known correlation with evolutionary rate but also on a reasonable mechanistic model. We suggest a framework based on these mechanistic models to correctly interpret the correlations between evolutionary rates and the various variables as well as the interrelationships among the variables.

A conserved mammalian protein interaction network

Physical interactions between proteins mediate a variety of biological functions, including signal transduction, physical structuring of the cell and regulation. While extensive catalogs of such interactions are known from model organisms, their evolutionary histories are difficult to study given the lack of interaction data from phylogenetic outgroups. Using phylogenomic approaches, we infer a upper bound on the time of origin for a large set of human protein-protein interactions, showing that most such interactions appear relatively ancient, dating no later than the radiation of placental mammals. By analyzing paired alignments of orthologous and putatively interacting protein-coding genes from eight mammals, we find evidence for weak but significant co-evolution, as measured by relative selective constraint, between pairs of genes with interacting proteins. However, we find no strong evidence for shared instances of directional selection within an interacting pair. Finally, we use a network approach to show that the distribution of selective constraint across the protein interaction network is non-random, with a clear tendency for interacting proteins to share similar selective constraints. Collectively, the results suggest that, on the whole, protein interactions in mammals are under selective constraint, presumably due to their functional roles.

The relationship between the hierarchical position of proteins in the human signal transduction network and their rate of evolution

Background

Proteins evolve at disparate rates, as a result of the action of different types and strengths of evolutionary forces. An open question in evolutionary biology is what factors are responsible for this variability. In general, proteins whose function has a great impact on organisms’ fitness are expected to evolve under stronger selective pressures. In biosynthetic pathways, upstream genes usually evolve under higher levels of selective constraint than those acting at the downstream part, as a result of their higher hierarchical position. Similar observations have been made in transcriptional regulatory networks, whose upstream elements appear to be more essential and subject to selection. Less well understood is, however, how selective pressures distribute along signal transduction pathways.

Results

Here, I combine comparative genomics and directed protein interaction data to study the distribution of evolutionary forces across the human signal transduction network. Surprisingly, no evidence was found for higher levels of selective constraint at the upstream network genes (those occupying more hierarchical positions). On the contrary, purifying selection was found to act more strongly on genes acting at the downstream part of the network, which seems to be due to downstream genes being more highly and broadly expressed, performing certain functions and, in particular, encoding proteins that are more highly connected in the protein–protein interaction network. When the effect of these confounding factors is discounted, upstream and downstream genes evolve at similar rates. The trends found in the overall signaling network are exemplified by analysis of the distribution of purifying selection along the mammalian Ras signaling pathway, showing that upstream and downstream genes evolve at similar rates.

Conclusions

These results indicate that the upstream/downstream position of proteins in the signal transduction network has, in general, no direct effect on their rates of evolution, suggesting that upstream and downstream genes are similarly important for the function of the network. This implies that natural selection differently distributes across signal transduction networks and across biosynthetic and transcriptional regulatory networks, which might reflect fundamental differences in their function and organization.

Selection for higher gene copy number after different types of plant gene duplications

The evolutionary origins of the multitude of duplicate genes in the plant genomes are still incompletely understood. To gain an appreciation of the potential selective forces acting on these duplicates, we phylogenetically inferred the set of metabolic gene families from 10 flowering plant (angiosperm) genomes. We then compared the metabolic fluxes for these families, predicted using the Arabidopsis thaliana and Sorghum bicolor metabolic networks, with the families' duplication propensities. For duplications produced by both small scale (small-scale duplications) and genome duplication (whole-genome duplications), there is a significant association between the flux and the tendency to duplicate. Following this global analysis, we made a more fine-scale study of the selective constraints observed on plant sodium and phosphate transporters. We find that the different duplication mechanisms give rise to differing selective constraints. However, the exact nature of this pattern varies between the gene families, and we argue that the duplication mechanism alone does not define a duplicated gene's subsequent evolutionary trajectory. Collectively, our results argue for the interplay of history, function, and selection in shaping the duplicate gene evolution in plants.