Pervasive Positive Selection on Virus Receptors Driven by Host-Virus Conflicts in Mammals

ABSTRACT

Viruses hijack cellular proteins, known as viral receptors, to initiate their infections. Viral receptors are subject to two conflicting directional forces, namely, negative selection to maintain their cellular function and positive selection resulted from the ever-changing host-virus arms race. Much remains unclear about how viral receptors evolved in mammals and whether viral receptors from different mammal groups experienced different strengths of natural selection. Here, we perform evolutionary analyses of 92 viral receptors in five major orders of mammals, including Carnivora, Cetartiodactyla, Chiroptera, Primates, and Rodentia. In all five mammal orders, signals of positive selection are detected for a high proportion of viral receptors (from 41% in Carnivora to 65% in Rodentia). Many positively selected residues overlap the host-virus interaction interface. Compared with control genes, we find viral receptors underwent elevated rates of adaptive evolution in all five mammal orders, suggesting that host-virus conflicts are the main driver of the adaptive evolution of viral receptors in mammals. Interestingly, the overall strength of natural selection acting on viral receptors driven by the host-virus arms race is largely homogenous and correlated among different mammal orders, with bats and rodents, zoonosis reservoirs of importance, being unexceptional. Taken together, our findings indicate host-virus conflicts have driven the elevated rate of adaptive evolution in viral receptors across mammals and might have important implications in zoonosis surveillance and prediction.

IMPORTANCE Viral receptors are cellular proteins hijacked by viruses to help their infections. A complete picture on the evolution of viral receptors in mammals is still lacking. Here, we perform a comprehensive evolutionary analysis of the evolution of 92 viral receptors in five mammal orders, including Carnivora, Cetartiodactyla, Chiroptera, Primates, and Rodentia. We find that positive selection pervasively occurred during the evolution of viral receptors, and viral receptors exhibit an elevated rate of adaptive evolution compared to that of control genes in all five mammal orders, suggesting host-virus conflicts are a major driver of the adaptive evolution of viral receptors. Interestingly, the strength of positive selection acting on viral receptors is similar among the five mammal orders. Our study might have important implications in understanding the evolution of host-virus interaction.

INTRODUCTION

Viral receptors are host cellular proteins hijacked by viruses to initialize their entry into host cells (1, 2). Virus and viral receptor are not simply a one-to-one relationship (2). Several different viruses might leverage a common cellular protein as their receptor; for example, several New World arenaviruses, mouse mammary tumor virus (MMTV) (a retrovirus), canine parvovirus, and feline panleukopenia virus, use transferrin receptor (TFRC), the cellular surface receptor for iron-loaded transferrin, for cellular entry (3–6). The cellular entry of a virus might be mediated by one receptor or another (5); for example, at least two viral receptors, low-density lipoprotein receptor (LDLR) and CD81, can mediate the cellular entry of hepatitis C virus (7–9). Therefore, the relationship between viral receptors and viruses is extremely complex.

Essentially host cellular proteins, viral receptors are theoretically expected to evolve under two conflicting directional forces, namely, negative selection to maintain their own cellular functionality and positive selection resulted from the ever-changing host-virus arms race (1, 2, 5, 10). Our previous study shows that viral receptors have experienced an elevated rate of adaptive evolution in primates, which are likely to have been driven by a host-virus arms race (10). Moreover, some virus receptors have been found to be subject to positive selection in mammals outside of primates, for example, Niemann-Pick C1 protein (NPC1, the receptor for Ebola virus), TFRC (5, 11), and angiotensin-converting enzyme 2 (ACE2, the receptor for SARS-CoV and SARS-CoV-2) (12–14). However, much remains unclear how diverse viral receptors evolved across mammals. Our understanding of viral diversity in mammals remains fragmented and incomplete, but virus richness (the total number of viral species) seems to vary greatly among the taxonomic orders of mammals (15, 16). It remains unclear whether viral receptors from different mammal orders evolved under different strengths of natural selection.

In this study, we performed comprehensive evolutionary analyses of 92 viral receptors in five mammalian orders, namely, Carnivora (carnivores), Cetartiodactyla (whales and even-toed ungulates), Chiroptera (bats), Primates (primates), and Rodentia (rodents). We found that a high proportion of viral receptors are subject to episodic positive selection in all five mammal orders. Compared with control genes, viral receptors experienced an elevated rate of adaptive evolution likely driven by an ever-changing host-virus arms race. The strength of natural selection appears to be similar among different mammal orders. Our study provides novel insights into the evolution of viral receptors in mammals.

RESULTS

Pervasive positive selection acts on viral receptors in mammals.

To explore natural selection acting on viral receptors in mammals, we analyzed a total of 92 viral receptor genes (see Table S1 in the supplemental material) in mammal orders with more than 10 species whose genomes have been sequenced. Our study includes a total of 84 representative mammal species within five major mammal orders, namely, Carnivora (16 species), Cetartiodactyla (14 species), Chiroptera (16 species), Primates (20 species), and Rodentia (18 species) (Fig. S1). Initially, we employed the branch-site unrestricted statistical test for episodic diversification (BUSTED) method to detect whether a viral receptor gene experienced episodic positive selection on at least one site on at least one branch (17). We detected signals of episodic positive selection in 38 (41.3%), 39 (43.3%), 59 (64.1%), 52 (56.5%), and 60 (65.2%) viral receptor genes for carnivores, cetartiodactyls, bats, primates, and rodents, respectively (Fig. 1). In cetartiodactyls, we only analyzed 90 viral receptors, because orthologs for glycophorin A (GYPA) or tumor necrosis factor receptor superfamily member 14 (TNFRSF14) were not identified, likely due to either gene loss or gaps in the assembly. Our study suggests that pervasive positive selection has acted on viral receptors in mammals (Fig. 1A and C). The viral receptor genes under positive selection substantially overlap among the five mammal orders; for instance, 20 (20/92; 21.7%) viral receptor genes are subject to episodic positive selection in all the five mammal orders (Fig. 1B). For 8 viral receptors, the signal of positive selection was not detected in any mammal order. For each mammal order, the numbers of virus receptor genes under positive selection appears to be different on different branches (Fig. S1 and dx.doi.org/10.6084/m9.figshare.6025748).

FIG 1.

Viral receptors under positive selection in five mammal orders. (A) The total number of viral receptors subject to positive selection in five mammal orders. Signals of positive selection were detected by the BUSTED method. (B) Relationship of viral receptor genes under positive selection among five mammal orders. (C) Positive selected genes in five mammal orders. The viral receptor genes under positive selection are labeled in filled circles. The signals of positive selection were detected using the BUSTED method. NA indicates that no selection analysis was performed due to a limited number of genes.

Many positively selected sites reside in host-virus interaction interface.

To investigate the relationship between sites subject to positive selection and host-virus interaction, we used a site model (M7 versus M8 implemented in PAML) to detect positively selected sites in viral receptors and mapped these positively selected sites onto the structures of several receptor and virus protein complexes. For at least five viral receptors, we found positively selected sites overlap known host-virus interaction interface. (i) Human immunodeficiency virus type 1 (HIV-1) interacts with T-cell surface glycoprotein CD4 to invade host cells. We found that 7, 4, 21, 5, and 7 positively selected sites are located in the virus and receptor interaction interface in Carnivora, Cetartiodactyla, Chiroptera, Primates, and Rodentia, respectively (Fig. 2A) (18). (ii) Membrane cofactor protein CD46 is the receptor of adenovirus, and positively selected sites are detected in the virus-receptor interaction regions in all five orders (Fig. 2B) (19). (iii) Many positively selected sites overlap the interaction interface between complement decay-accelerating factor CD55 and enterovirus in mammals (Fig. 2D) (20). (iv) Carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) acts as the receptor of mouse hepatitis virus (MHV), and some positively selected sites are located in the virus-receptor interaction regions (Fig. 2E) (21). (v) Intercellular adhesion molecule 1 (ICAM1) acts as the receptor of coxsackievirus A21 and rhinovirus. Our results show that 2, 4, 15, 4, and 4 sites are in the interface of Carnivora, Cetartiodactyla, Chiroptera, Primates, and Rodentia, respectively (Fig. 2G) (22, 23). For at least three viral receptors, we found that positive selection sites located on interface of virus-receptor interaction are only specific to certain mammalian orders. (i) Aminopeptidase N (ANPEP) is the receptor of coronavirus 229E. Positively selected sites in host-virus interaction regions were detected in Carnivora, Cetartiodactyla, Chiroptera, and Rodentia (Fig. 2C). (ii) Dipeptidyl peptidase 4 (DPP4) is the receptor of Middle East respiratory syndrome (MERS) coronavirus. Positively selected sites overlapping host-virus interaction interface were identified in Carnivora, Cetartiodactyla, Chiroptera, and Rodentia (Fig. 2H). (iii) For TFRC, positively selected sites located in its interaction interface with MMTV and canine parvovirus were found in Carnivora, Chiroptera, and Rodentia (Fig. 2F) (5, 11, 24, 25). Together, these results indicate that the confrontation between viruses and receptor proteins might have driven the adaptive evolution of many viral receptors, although only a few structures of virus protein and viral receptor complexes are available.

FIG 2.

Positively selected sites overlap the virus-receptor interacting interfaces. Positively selected sites on the virus-receptor interfaces detected by PAML with M7 and M8 are labeled in colored, filled circles. (A) CD4 and human immunodeficiency virus interacting interface; (B) CD46 and adenovirus interacting interface; (C) ANPEP and coronavirus 229E interacting interface; (D) CD55 and enterovirus interacting interface; (E) CEACAM1 and mouse hepatitis coronavirus interacting interface; (F) TFRC and mouse mammary tumor virus or canine parvovirus interacting interfaces; (G) ICAM1 and coxsackievirus or rhinovirus interacting interfaces; and (H) DPP4 and MERS coronavirus interacting interface.

Rate of adaptive evolution is elevated in viral receptors.

To further explore whether adaptive evolution in viral receptor genes has been driven by host-virus conflicts, we compared the strength of positive selection between viral receptors and control genes, which share similar gene ontology (GO) biological process categories with viral receptor genes (10). Within 238 GO categories in which viral receptor genes are enriched, a control gene was chosen based on two criteria: it shares at least 50% of GO items of the corresponding viral receptor, and the total number of GO items of a control gene does not exceed 150% of the corresponding viral receptor. We found the ratio of nonsynonymous to synonymous substitution rates (estimated dN/dS rates) of virus receptors are significantly higher than that of control genes for all five mammal orders, including carnivores (0.25 versus 0.09; P < 0.001, Mann-Whitney U test), cetartiodactyls (0.27 versus 0.11; P < 0.001, Mann-Whitney U test), bats (0.32 versus 0.10; P < 0.001, Mann-Whitney U test), primates (0.31 versus 0.11; P < 0.001, Mann-Whitney U test), and rodents (0.27 versus 0.10; P < 0.001, Mann-Whitney U test) (Fig. 3A to E). Significantly more branches were found to be subject to positive selection in viral receptor genes than in control genes in all five mammal orders, including carnivores (2.55 versus 1.54; P = 0.012, Mann-Whitney U test), cetartiodactyls (2.56 versus 1.58; P = 0.025, Mann-Whitney U test), bats (2.79 versus 1.26; P = 0.002, Mann-Whitney U test), primates (5.08 versus 2.28; P < 0.001, Mann-Whitney U test), and rodents (2.38 versus 1.20; P = 0.003, Mann-Whitney U test) (Fig. 3F to J). Moreover, a higher proportion of sites experienced episodic positive selection in viral receptor genes than in control genes in all five mammal orders, including carnivores (0.0077 versus 0.0019; P < 0.001, Mann-Whitney U test), cetartiodactyls (0.0076 versus 0.0022; P < 0.001, Mann-Whitney U test), bats (0.0214 versus 0.0050; P < 0.001, Mann-Whitney U test), primates (0.0070 versus 0.0022; P < 0.001, Mann-Whitney U test), and rodents (0.0236 versus 0.0062; P < 0.001, Mann-Whitney U test) (Fig. 3K to O). All these results suggest that viral receptor genes experienced elevated rates of adaptive evolution in all five mammal orders, which is likely to have been driven by recurrent evolutionary conflicts with viruses.

FIG 3.

Elevated rate of adaptive evolution of viral receptors in mammals. (A to E) The comparison of estimated dN/dS rates between viral receptor genes and control genes in five mammal orders. (F to J) The comparison of the branches subject to positive selection between viral receptor genes and control genes. (K to O) The comparison of the proportion of sites with adaption detected by the MEME method between viral receptor genes and control genes. Viral receptor genes and control genes are labeled in red and blue, respectively. Panels A to J and panels K to O are calculated by PAML-codeml and MEME, respectively.

Strength of positive selection in viral receptors is similar among mammal orders.

Next, we explored whether the strength of positive selection in virus receptor genes is different among mammal orders and whether viral receptors in certain mammal orders underwent stronger natural selection. The estimated dN/dS rates, the number of positively selected sites identified, and the number of positively selected branches are not directly comparable among different mammal orders, because the number of taxa used and divergence time might affect selection pressure analyses (26–28). To control the effect of sampling, we developed a statistic, E, to represent the excess ratio of virus receptor genes to control genes in terms of the values quantifying the strength of positive selection (namely, estimated dN/dS rates, the proportion of positively selected sites, and the number of positively selected branches). The excess ratio was defined as $E=(R-C)/\overline{C}$, where R and C represent the values quantifying positive selection in viral receptor genes and control genes, and $\overline{C}$ represents the mean values quantifying positive selection in all the control genes. The E statistic reflects the fraction of adaptive evolution driven by host-virus conflicts. We found that the E based on estimated dN/dS rates is not significantly different among the five major mammal orders (Fig. 4A). Moreover, no significant difference of E was observed among the five major mammal orders in terms of the number of branches subject to positive selection and the proportion of positively selected sites (Fig. 4B to D). All these results indicate that the strength of positive selection that acts on viral receptor genes resulting from the host-virus arms race is overall similar among the five mammal orders. It should be noted that we are not denying that a certain viral receptor might have evolved under different selective pressures among different mammal orders.

FIG 4.

Degree of adaptive evolution of virus receptors is homogenous among five mammal orders. (A) The comparison of excess ratio based on estimated dN/dS rates of viral receptor genes among five mammal orders. (B) The comparison of excess ratio based on branches subject to positive selection in viral receptors among five mammal orders. (C) The comparison of excess ratio based on positively selected sites in viral receptors among five mammal orders. The positively selected sites were detected using the MEME method. (D) The comparison of excess ratio based on the sites subject to positive selection in viral receptors among five mammal orders. The positively selected sites were detected using the FUBAR method.

Selection pressure on viral receptors is correlated between mammal orders.

To further explore the relationship of positive selection-acting viral receptor genes among different mammal orders, Pearson correlation coefficient was used to quantify the correlation between the estimated dN/dS rates of virus receptors in mammal orders. We found the selection strength measured by estimated dN/dS rates in viral receptor genes are highly correlated between mammal orders (Carnivora versus Cetartiodactyla, r = 0.95, P < 0.001; Carnivora versus Chiroptera, r = 0.95, P < 0.001; Carnivora versus Primates, r = 0.89, P < 0.001; Carnivora versus Rodentia, r = 0.94, P < 0.001; Cetartiodactyla versus Chiroptera, r = 0.94, P < 0.001; Cetartiodactyla versus Primates, r = 0.89, P < 0.001; Cetartiodactyla versus Rodentia, r = 0.91, P < 0.001; Chiroptera versus Primates, r = 0.89, P < 0.001; Chiroptera versus Rodentia, r = 0.96, P < 0.001; Primates versus Rodentia, r = 0.90, P < 0.001) (Fig. 5). The strong correlation of estimated dN/dS rates between mammal orders implies that viral receptor genes evolved under largely similar selection pressure across different mammal orders. This pattern suggests that viruses, either closely or distinctly related, that use the same receptor might have infected different mammalian orders.

FIG 5.

Positive correlation of selection pressure of receptor genes between mammalian orders. The diagram represents pairwise correlation based on the estimated dN/dS rates of virus receptors between mammal orders. (A) Carnivora versus Cetartiodactyla, (B) Carnivora versus Chiroptera, (C) Carnivora versus Primates, (D) Carnivora versus Rodentia, (E) Cetartiodactyla versus Chiroptera, (F) Cetartiodactyla versus Primates, (G) Cetartiodactyla versus Rodentia, (H) Chiroptera versus Primates, (I) Chiroptera versus Rodentia, and (J) Primates versus Rodentia.

DISCUSSION

As the first battlefield of the confrontation between hosts and viruses, viral receptors have been thought to be under two directional forces due to both functional constraints and host-virus arms race (1, 2, 5, 10). Our previous study shows that virus receptors display elevated rates of adaptive evolution in primates (10). In this study, we investigated the evolution of viral receptor genes in five mammal orders, namely, cetartiodactyls, carnivores, bats, primates, and rodents. We found that adaptive evolution is prevalent in viral receptors across these mammal orders. Moreover, many positively selected sites are located at the host-virus interaction interface. Compared with control genes, viral receptors experienced elevated rates of adaptive evolution in all five mammal orders. All these lines of evidence suggest that host-virus conflicts have driven the elevated rate of adaptive evolution in the five mammal orders, which is not specific to primates. It is reasonable to hypothesize that host-virus conflict also drives the adaptive evolution of viral receptors in other mammal orders.

Our results show that around 20 virus receptors are subject to positive selection in all five mammal orders, most of which appear to be receptors of diverse picornaviruses, such as CMRF35-like molecule 1 (CD300LF), CD55, cadherin-related family member 3 (CDHR3), ICAM1, LDLR, poliovirus receptor (PVR), and P-selectin glycoprotein ligand-1 (SELPLG) (20, 23, 29–33). For CD55 and ICAM1, positively selected sites are located on the virus-receptor interaction interface. Indeed, picornaviruses are widespread in vertebrates and might have a long coevolutionary history with vertebrates (34). Therefore, picornaviruses are likely to have posed strong selective pressure on the evolution of viral receptors in mammals (and vertebrates). Moreover, for four coronavirus receptors, ACE2, ANPEP, CEACAM1, and C-type lectin domain family 4 member G (CLEC4G) (21, 24, 35, 36), signals of positive selection were also detected in all five mammal groups. For CEACAM1 and ANPEP, many positively selected sites were detected in the regions crucial for virus-host interaction. These results indicate that the arms race between coronaviruses and mammals lasted for millions of years. Taken together, our results suggest that viruses provide a nonnegligible selective pressure for the adaptive evolution of virus receptors. Moreover, the findings here might provide novel insights into the host distribution of some viruses.

Virus richness has been thought to vary greatly among the taxonomic orders of mammals (15, 16). If a virus that uses a certain viral receptor is specific to and has long been circulating in a certain mammal order, the viral receptor gene from different mammal orders is expected to evolve differentially among different mammal orders (Fig. 6). In contrast, if a virus (family) infects a wide range of mammals and viruses that use the same viral receptor infect various mammals, the viral receptor from different mammal orders is expected to be subject to similar levels of positive selection (Fig. 6). In this study, we found that the strength of positive selection on viral receptors due to host-virus conflicts is largely similar among mammal orders. One might argue that different viruses infect different mammal orders, which can result in similar excess ratios among the five mammal orders. However, nearly perfect correlation of selection pressure strength was observed between the five mammal orders, which is inconsistent with order-specific infection of viruses. It should be noted that we are not denying that a certain viral receptor might have evolved under different selective pressures among different mammal orders. Nevertheless, our results suggest that a viral receptor frequently can be hijacked by viruses in different mammal orders.

FIG 6.

Scenarios for the evolution viral receptor in different mammal orders. Pattern 1 indicates that a virus that uses a certain viral receptor is specific to and has long been circulating in a certain mammal order may lead to significant differences in the evolution of virus receptors in different mammals. Pattern 2 indicates that if a virus infects a wide range of mammals and viruses that use the same viral receptor infect various mammals, viral receptors from different mammal orders may share similar evolutionary patterns.

Our study comes with several caveats. (i) It is possible that some viral receptors are not authentic viral receptors, and there are potential virus receptors that have not been identified in the control gene. However, our previous study showed that the receptor genes are enriched with GO terms related to viral infection, while the control genes are not (10). The finding that viral receptor genes experienced elevated rates of adaptive evolution in all five mammal orders also indicates the viral receptors used in this study are largely viral receptors in all five mammal orders. (ii) Virus-receptor interaction sites are usually resolved in a specific manner. It is possible that even similar viruses interact with a viral receptor on a different interface and use different viral receptors, and the identification of known viral receptors is mainly based on studies of key organisms. More structures of virus-receptor complexes are needed to decipher the evolution of virus-host interaction.

MATERIALS AND METHODS

Orthologs of viral receptors and control genes in mammals.

Information on viral receptors and the corresponding control genes were retrieved from our previous study (see Table S1 in the supplemental material) (10). Briefly, viral receptor GO enrichment analyses retrieved a total of 238 GO biological process categories related to viral receptors, with a P value of <10⁻⁵ as the threshold. Only these 238 GO terms were used for the following control gene mapping. Genes are regarded as the control of the virus receptor, which must first satisfy that at least 50% of the items are identical to the virus receptor (10). To ensure the specificity of the control gene, the total number of GO items of the control gene should not exceed 150% of the virus receptor. Based on the above-described two criteria, we constructed a control gene library for each receptor and then randomly selected a control gene for each receptor gene from the library (10). We analyzed the evolution of viral receptors in mammal orders with more than 10 species whose genomes have been sequenced. Our study includes a total of five mammal taxonomic orders, namely, Carnivora (16 species), Cetartiodactyla (14 species), Chiroptera (16 species), Primates (20 species), and Rodentia (18 species). The phylogenetic relationships among mammals are based on the literature (Fig. S1) (37). The reference transcripts of 84 species were retrieved from NCBI. The tBLASTn algorithm was utilized to identify the orthologs of viral receptors and control genes among the species with human protein sequence as the query and an E value cutoff of 10⁻¹⁰ (38). Orthologs were identified and verified by phylogenetic analyses. The orthologous genes were aligned with MUSCLE based on the codon evolution model (39). The sequences and trees used in the study were deposited to figshare (dx.doi.org/10.6084/m9.figshare.14784240).

Evolutionary analyses in mammals.

We used the branch-site unrestricted statistical test for episodic diversification (BUSTED) method implemented in HyPhy to test whether the receptor gene has experienced episodic positive selection in mammals (17). The Codeml program in the PAML 4.9 package was employed to quantify selective pressures for receptor and control genes (40). The phylogeny of gene sequences was based on the phylogenetic relationship of each mammal taxonomic order. Free ratio model that calculates one ratio for each branch (with model = 1, NSsites = 0) was used to evaluate the branch-specific dN/dS values. For each gene, two NSsites, models M7 (neutral model, without dN/dS> 1) and M8 (positive selection model, dN/dS > 1 is allowed), were used to detected positively selected sites. The estimated dN/dS rate of M8 was used as a parameter to quantify the global selection of genes in this study. Furthermore, two methods of selection analyses implemented in HyPhy were also performed: the mixed effects model of evolution (MEME) method was used to detect positive selection sites under a proportion of branches (41), and the fast unconstrained Bayesian approximation (FUBAR) method was used to infer positively selected sites among the entire phylogeny (42). Results were deposited to figshare (dx.doi.org/10.6084/m9.figshare.14784255).

Excess ratio of adaptive evolution.

Selection pressure analyses might be affected by the number of taxa used and divergence time; thus, the quantitative data of selection pressure are not directly comparable among different mammal orders. To control the effect of sampling, we developed a statistic, E, to represent the excess ratio of virus receptor genes to control genes in terms of the values quantifying the strength of positive selection (namely, estimated dN/dS rates, the number of positively selected sites, and the number of positively selected branches). The excess ratio was defined as $E=(R-C)/\overline{C}$, where R and C represent the values quantifying positive selection in viral receptor genes and corresponding control genes and $\overline{C}$ represents the mean values quantifying positive selection in all the control genes. The E statistic reflects the fraction of adaptive evolution of receptors among mammal orders.

Statistical analyses.

Pearson correlation coefficient, used to measure the correlation between two variables, was calculated by the ggstatsplot package in R. The comparisons between viral receptors and control genes were analyzed with Mann-Whitney U test using STATISTICA 8.0. Multigroup comparison of E values among mammal orders were performed with Fisher least significant difference test using STATISTICA 8.0. The threshold for significance for all tests was set to 0.05. Significance is indicated by asterisks: *, P < 0.05; **, P < 0.01; ***, P < 0.001 (n.s., not significant; P > 0.05).