The adaptation of codon usage of +ssRNA viruses to their hosts.

Viruses depend on their host's cellular structure to survive. Most of them do not have tRNAs, their translation relies on hosts' tRNA pools. Over the course of evolution, viruses needed to optimally exploit cellular processes of their host. Thus, codon usage of a virus should coevolve with its host to efficiently and rapidly replicate. Some viruses can invade a broad spectrum of hosts (BSTVs), while others can invade a narrow spectrum only (NSTVs). Consequently, we test the hypothesis that similarity of codon usage preference and the degree of matching between BSTVs and their hosts will be lower than that of NSTVs, which only need to coevolve with few hosts. We compare the patterns of codon usage in 255 virus genomes to test this hypothesis. Our results show that NSTVs have a higher degree of matching to their hosts' tRNA pools than BSTVs. Further, analysis of the effective number of codons (ENC) infers that codon usage bias of NSTVs is relatively stronger than that of BSTVs. Thus, codon usage of NSTVs tends to better match their host than that of BSTVs. This supports the hypothesis that viruses adapt to the expression system of their host(s).

Introduction

Viruses are pure parasites. They depend on their hosts' cellular structure and metabolism to replicate and assemble, i.e., survive. Most of their genomes do not encode tRNAs, thus their translation of viral proteins relies on the hosts' tRNA pools (Kumar et al. 2016). A successful infection requires that viruses possess the ability to enter the host cell, and efficiently produce new viruses. The degenerate genetic code unequally uses synonymous codons, which code for the same amino acid (Cristina et al. 2016; Kanaya et al. 2001; Shackelton et al. 2006; Tsai et al. 2007). The redundancy of the genetic code provides the opportunity to shape the efficiency and accuracy of protein production, while maintaining the same amino acid sequence (Chaney and Clark 2015; Plotkin and Kudla 2011; Stoletzki and Eyre-Walker 2007). Considering that the translation of viral proteins relies on the host's pool of tRNAs, codon usage of a virus must coevolve with its host to efficiently use host resources. It is expected that higher similarity of codon usage pattern will better facilitate their replication.

The extent of codon usage among viruses and their hosts has been suggested to affect viral survival, fitness, and evasion from host's immune system (Burns et al. 2006; Costafreda et al. 2014; Mueller et al. 2006). Because the virus relies on the host's cellular machinery for its replication, codon usage bias was suggested to play a role in the adaptation of a virus to its host. Codon usage bias is common in viruses (Butt et al. 2014; Castells et al. 2017; Cristina et al. 2016; He et al. 2017; Li et al. 2017; Moratorio et al. 2013; Singh et al. 2016; Su et al. 2017; Xu et al. 2017; Zang et al. 2017; Zhao et al. 2016). Efficient replication seemingly requires that a virus and host have similar codon usage patterns to share a tRNA pool. Co-evolution between a certain RNA virus and its susceptible hosts at codon usages have been observed in many viruses (Franzo et al. 2017; Rahman et al. 2017; Simón et al. 2017).

Some viruses have a broad ranges of hosts (BSTVs), such as arbovirus. These can infect mammals, birds, and insects. Other viruses have narrow host ranges (NSTVs), which can infect a limited number of hosts only. Because BSTVs must fit to multiple hosts and their diverse tRNA pools and their codon usage has a relationship with their host, a tradeoff exists regarding the extent of codon usage. BSTVs must fit to their diverse hosts and, thus, the extent of matching for codon usage would be lower than that of NSTVs, which must fit few, similar hosts. To test this hypothesis, we analyze 255 viruses from 20 genera of positive-sense single-stranded RNA (+ssRNA) viruses (e.g., Flavivirus, Alphavirus, Coronavirus, Torovirus, Arterivirus, Rubivirus, Pestivirus, Hepacivirus, Alphamesonivirus). Viruses that can infect vertebrates and invertebrates, such as most of the Flavivirus and Alphavirus, were classified as BSTVs, while other viruses that can infect either vertebrate or invertebrate were classified as NSTVs.

Materials and methods

Data collection and preliminary analyses

The complete genome sequences of 255 virus strains from the 20 genera of +ssRNA viruses were obtained from the GenBank database (http://www.ncbi.nlm.nih.gov). The information of host range was determined from NCBI (https://www.ncbi.nlm.nih.gov/taxonomy/) and the Ninth Report of the International Committee on Taxonomy of Viruses (ViralZone Database: http://viralzone.expasy.org/). Accession numbers and other detailed information of these viruses, such as strain names, isolate hosts and host ranges were also retrieved (Supplementary Table 1).

Estimates of codon usage, the relative synonymous codon usage (RSCU) (Sharp and Li 1986), and the effective number of codons (ENC) (Wright 1990), were calculated using CodonW (available at http://sourceforge.net/projects/codonw). In this study, Aedes, Culex and Ixodes represented the main arthropod hosts, and Gallus, Homo and Mus the three major groups of vertebrates. Coding sequences of the hosts were obtained from the Ensembl database (available at: http://www.ensembl.org) (Yates et al. 2016). Copy number of tRNAs for transmission vectors and hosts were obtained from GtRNAdb (http://gtrnadb.ucsc.edu/).

Estimating the adaptability of viruses to their hosts' tRNA pools

Optimizing codon usage of viruses according to that of highly expressed host genes has been proved to increase the production of viral proteins (Chithambaram et al. 2014; Ngumbela et al. 2008) or transgenic genes (Koresawa et al. 2000). The degree of similarity for overall codon usage between viruses and their hosts' tRNA pools was estimated with a parameter based on optimized codon usage and the extent of matching between viral ORF's codon usage bias and their hosts' tRNA pools. ORFs were optimized on the basis of tRNA copy number characteristics of their hosts' expression system. Online optimization software (http://genomes.urv.es/OPTIMIZER/) (Puigbo et al. 2007) was utilized. The matching degree (MD) was calculated as follows:where M was defined as the number of the different bases before and after optimized sequence, and N was the total length of open reading frame. This value could have ranged from zero to 1.

To better quantify the effect of the overall codon usage of the host on the formation of the overall codon usage of the virus, the similarity index D(A,B) reported by the previous study was introduced into our work (Zhou et al. 2013). The D(A,B) represented the potential effect of the overall codon usage of the host on that of virus. This value potentially ranged from zero to 1.0.

Results

The matching degree of +ssRNA viruses to their hosts' tRNA pools

MD values were calculated for viral ORFs. Unfortunately, some viruses lacked data of tRNA copy numbers, and coding sequence of their hosts. Therefore, only 255 viruses (+ssRNA) were used in this analysis. The 101 strains of arboviruses that belonged to BSTVs were optimized according to their hosts' tRNA pool expression systems (host: arthropods, mammals, Gallus gallus). MD values (mean ± SD) were 0.7388 ± 0.0121, 0.7383 ± 0.0083, and 0.7427 ± 0.0045 in the three hosts, respectively (Supplementary Table 2). The MD values of 154 NSTVs strains had a mean of 0.7617 ± 0.0168 (Supplementary Table 3). Wilcoxon & Mann-Whitney U test obtained statistically significant higher MD values in NSTVs than BSTVs (Z = −9.99, p < 0.001; Z = −10.25, p < 0.001; Z = −5.59, p < 0.001, respectively. Fig. 1a).

Fig. 1

Matching degree (MD) of viruses to their hosts' tRNA pools.

(a) Group 1 represents MD of BSTVs to arthropod's tRNA pool. Group 2 shows the MD of BSTVs to mammal's tRNA pool. Group 3 is the MD of BSTVs to Gallus gallus's tRNA pool. Group 4 shows the MD of NSTVs to their hosts' tRNA pools. (b) The MD of 20 virus genera (+ssRNA) to their hosts' tRNA pools.

Among the different genera of arbovirus (Fig. 1b), the extent of matching for Flavivirus to their hosts' tRNA pools was higher than that of Alphavirus. In addition, among the 18 genera of NSTVs, the MD values of Togaviridae (rubivirus) were the highest, followed by Picronaviridae (Aphthovirus, Cardiovirus, Casavirus, Enterovirus, Hepatovirus, Kobuvirus), Mesoniviridae (Alphamesonivirus), Caliciviridae (Lagovirus, Norovirus, Sapovirus, Vesivirus), Arterviridae (Artervirus), Astroviridae (Astrovirus), Coronaviridae (Coronavirus, Torovirus), and the last was Flaviridae (Pestivirus, Hepacivirus).

Similarity of overall codon usage between +ssRNA viruses and their hosts

Similarity index values D(A,B) were obtained for each strain in relation to its host(s) (Supplementary Table 2, Supplementary Table 3). As shown in Fig. 2a, the indices of the fourth group (NSTVs vs hosts) were higher than those of the groups 1, 2, and 3 (BSTVs vs arthropods, BSTVs vs mammals, BSTVs vs Gallus gallus).

Fig. 2

Similarity of overall codon usage pattern of viruses to their hosts.

(a) Group 1 is the similarity degree between BSTVs and arthropods. Group 2 is the similarity degree between BSTVs and mammals. Group 3 is the similarity degree between BSTVs and Gallus gallus. Group 4 is the similarity degree between NSTVs and a particular host. (b) The similarity degree of the overall codon usage bias between 20 virus genera (+ssRNA) and the hosts.

To quantify the degree of similarity of the overall codon usage pattern between 20 different virus genera and their hosts, the similarity index D(A,B) was calculated to all strains (Fig. 2b). The MD and D(A,B) values showed a similar trend in that values of NSTVs were the highest.

Finally ENC's values were calculated for the 255 strains to gain insights into codon usage bias. All viruses had ENC values >40, except Human coronavirus HKU1 (ENC = 35.66), Human coronavirus NL63 (ENC = 36.77), Rubella virus (ENC = 38.45), Human hepatitis A virus (ENC = 39.23), Canine kobuvirus 1 (ENC = 39.38), Aichi virus (ENC = 39.52), Simian hepatitis A virus (ENC = 39.91), and Kobuvirus dog/AN211D/USA/2009 (ENC = 39.92). The codon preferences of NSTVs were stronger than for the BSTVs (Fig. 3 ).

Fig. 3

Effective number of codons (ENCs) of 20 virus genera.

(a) BSTVs and NSTVS. (b) 20 virus genera (+ssRNA).

Discussion

The degeneracy of the genetic code implies that several triplets can code for the same amino acid. The use of synonymous codons in gene coding regions dos not occur randomly, and codon usage bias is very common in viruses (Butt et al. 2014; Cristina et al. 2016; He et al. 2017; Moratorio et al. 2013; Singh et al. 2016). Codon usage is among the determinant factors that influence gene expression levels (Chaney and Clark 2015; Zhou et al. 2016). Because viruses do not have tRNAs, and rely on host cell machinery for replication, co-evolution between a certain RNA virus and its susceptible hosts at codon usages have been observed (Franzo et al. 2017; Rahman et al. 2017; Simón et al. 2017). However, ambiguity remains in the co-evolution patterns of different viruses. Some viruses infect a broad range of species (BSTVs), whereas others infect only a single host (NSTVs). Viruses have very diverse hosts, and different hosts have very diverse tRNA pools. The MD and D(A,B) values of NSTVs are significantly higher than those of the BSTVs vs. Anopheles gambiae, Homo sapiens, Gallus gallus, and Macaca mulatta (Fig. 1, Fig. 2). Thus, and as our hypothesis predicts, NSTVs appear to be more precisely adapted to their hosts' codon usage pattern and tRNA pools than BSTVs. Each NSTV can infect one host only. Therefore, these viruses need to fit only one host's tRNA pool. They appear to have evolved more consistent codon usage patterns with their hosts' expression systems. In contrast, BSTVs can infect and replicate in mammals, birds, and insects. Thus, adaptation in BSTVs may involve a tradeoff between precise and functional matching to fit the diverse tRNA pools of multiple hosts. This may explain the relatively lower matching of BSTVs to their hosts.

Enterovirus, Hepacivirus and Arterivirus are NSTVs and their D(A,B) values are relatively low (Fig. 2b). This indicates a relatively low extent of similarity in overall codon usage between these viruses and their host. These viruses may not need to replicate rapidly. Other factors, such as mutational pressure, may also play a role in determining codon usage bias (Gu et al. 2004; Rahman et al. 2017; Wang et al. 2011; Wong et al. 2010).

To quantify the extent of variation in codon usage, the ENc values were calculated (Wright 1990). Most viruses have ENC values >40, which represents weak codon bias. This may be beneficial for efficient replication of viruses in host cells with potentially distinct codon preferences. The codon preference of NSTVs is relatively stronger than for BSTVs (Fig. 3). This is could be due to weak codon preference being advantageous in the adaptation of BSTVs to multiple host expression systems.

The ability to enter the host-cell and efficiently replicate itself is essential for viral infection. Viruses have coevolved many pathways to transcribe their own genetic material in their hosts (Harwig et al. 2017). Codon usage in BSTVs may involve a tradeoff between precise and functional matching to fit the diverse tRNA pools of multiple hosts. As expected, our analysis show that generally NSTVs are more adapted to their hosts' codon usage pattern and tRNA pools than BSTVs. This may help the virus to use the host transcript machinery more efficiently and, therefore, replicate faster.

The following are the supplementary data related to this article.

Supplementary Table 1

Virus strains used in this study.

Supplementary Table 2

The matching degree and similarity index of BSTVs to their hosts.

Supplementary Table 3

The matching degree and similarity index of NSTVs to their hosts.

Author contributions

YS: designed the study; LT and XS: analyzed the data; LT, RWM and YS: drafted the manuscript; All authors read and approved the final manuscript.