Prediction of the effects of the top 10 nonsynonymous variants from 30229 SARS-CoV-2 strains on their proteins

Introduction

A new coronavirus disease known as COVID-19, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was first reported in Wuhan, China in December 2019.¹ SARS-CoV-2 is a positive-sense single stranded RNA virus with a helical nucleocapsid. The genome size of SARS-CoV-2 is about 30 kilobases. There are 11 protein-coding genes from the SARS-CoV-2 genome including four structural genes (spike (S), envelope (E), membrane (M), and nucleocapsid (N) genes) and seven nonstructural genes (ORF1ab, ORF3a, ORF6, ORF7a, ORF7b, ORF8 and ORF10).²

The SARS-CoV-2 virus can rapidly mutate to bypass the immune response of the host.³ These mutations can be synonymous, nonsynonymous, deletions, insertions, or others. Nonsynonymous mutations are expected to have a greater impact than synonymous mutations since nonsynonymous mutations affect the amino acid sequences of a protein, subsequently they may change their structures and functions. According to Kim et al. (2020), a total of 767 synonymous and 1352 nonsynonymous mutations have been identified from SARS-CoV-2 genomes.⁴ In another study, a total of 119 SNPs were identified using 11,183 SARS-CoV-2 genomes, in which there were 74 nonsynonymous mutations and 43 synonymous mutations.⁵ From a study on the analysis of nonsynonymous mutations in structural proteins of SARS-CoV-2, it has been shown that S and N proteins have higher mutation rate per gene compared to that of E and M proteins.⁶ The mutations of SARS-CoV-2 proteins may affect viral transmission, host immune evasion, and disease severity. Many SARS-CoV-2 S protein mutations, for example, D614G, N501Y have been heavily studied in detail because S protein facilitates viral entry into human host through the binding of ACE2 receptor.⁷ S protein D614G mutation increases viral transmission in animal model and human cell line study⁸ while S protein N501Y mutation results in an increase in binding to ACE2 receptor⁹ and an increase in viral replication in human upper-airway cells.¹⁰ However, the biological consequences of some of these mutations on the functions and structures of SARS-CoV-2 proteins remain unclear. In our study, computational analysis of the nonsynonymous mutations of SARS-CoV-2 proteins were performed using different bioinformatics tools including co-mutation analysis, protein structure stability and flexibility analysis, and protein function analysis to predict the effects of the mutations on the structures and functions of proteins.

Methods

Results

The statistics of nonsynonymous mutations in SARS-CoV-2 proteins

From the multiple alignment analysis, we identified 231 nonsynonymous mutations from 30,229 SARS-CoV-2 genome sequences. Figure 1 shows the numbers of the nonsynonymous mutations found in 11 coding sequences of SARS-CoV-2 proteins. ORF1a has the highest numbers of nonsynonymous mutations, followed by S protein and N protein. The top 10 highest frequency nonsynonymous mutations affecting 9 amino acids residues including ORF1a nsp5 P108S, ORF1b nsp12 P323L and A423V, S protein N501Y and D614G, ORF3a Q57H, N protein P151L, R203K and G204R and their frequency percentage in the primary lineages associated with the past and present VOCs are shown in Table 3.

Figure 1. The numbers of nonsynonymous mutations in 11 coding sequences of SARS-CoV-2 proteins.

Table 3. Top 10 nonsynonymous mutations of SARS-CoV-2 proteins and their frequency percentage in the primary lineages of VOCs.

Protein	Nucleotide changes	Amino acid changes	Frequency	Lineage (VOC)
				B.1.1.7 (α)	B.1.351 (β)	P.1 (γ)	B.1.617.2 (δ)	BA.1 (ο)
ORF1a nsp5	C10376T	P108S	4024	-	-	-	-	-
ORF1b nsp12	C14408T	P323L	27953	100%	90%	99%	100%	100%
	C14708T	A423V	3988	-	-	-	-	-
S	A23063T	N501Y	4218	99%	91%	97%	0%	94%
	A23403G	D614G	28022	100%	100%	100%	100%	100%
ORF3a	G25563T	Q57H	5274	0%	98%	0%	0%	0%
N	C28725T	P151L	4007	-	-	-	-	-
	G28881A	R203K	18116	99%	0%	97%	0%	100%
	G28882A	R203K	18092	99%	0%	97%	0%	100%
	G28883C	G204R	18090	91%	0%	97%	0%	99%

Co-mutation analysis of SARS-CoV-2 proteins

Some nonsynonymous mutations may be random and have no or little biological impact on viral transmission and pathogenesis. If a single nonsynonymous mutation co-mutates with other mutations, they may evolve together and interact functionally. To study co-mutation between different nonsynonymous mutations, the concurrence ratio of co-mutations in the top 10 nonsynonymous mutations was retrieved from GESS database website as shown in Table 2. The visualization of co-mutations in the top 10 nonsynonymous mutations generated with Circos table view is shown in Figure 2. In this chord diagram, connection ribbons represent co-mutations and each ribbon between row and column segments represents the value of concurrence ratio in each top 10 nonsynonymous mutations. Single colours encoded in circular arranged segments represent its own specific mutation whereas rainbow colours represent co-mutation in each mutation. The size of circular arrangement segments is proportional to the total value of concurrence ratio in a row or column. The circular size segment of ORF3a Q57H (G25563T) with the smallest segment size means the total value of concurrence ratio in row or column of ORF3a Q57H (G25563T) having the lowest concurrence ratio. A high concurrence ratio shows high co-mutation between each mutation with thicker ribbon size. S protein D614G (A23403G) with all other nine nonsynonymous mutations had concurrence ratios greater than 99%. On the other hand, low concurrence ratio shows low co-mutation with thinner ribbon size, for example, mutation ORF3a Q57H (G25563T) had the lowest concurrence ratio, only having a high concurrence ratio with S protein D614G (A23403G) and ORF1b nsp12 (P323L) C14408T, the top 2 nonsynonymous mutations which were present in more than 90% of the reported sequences.

Figure 2. Visualization of co-mutation in top 10 nonsynonymous mutations in SARS-CoV-2 proteins.

Discussion

The top 10 highest frequency nonsynonymous mutations of SARS-CoV-2 identified from 30,229 SARS-CoV-2 genome sequences were further analyzed with co-mutation analysis, prediction of the protein structure stability and flexibility analysis, and prediction of the protein function analysis. To determine if two nonsynonymous mutations of SARS-CoV-2 proteins co-mutate, concurrence ratio was calculated. Many nonsynonymous mutations showed a high concurrence ratio, suggesting these mutations may evolve together and interact functionally. The top 2 nonsynonymous mutations, S protein D614G and ORF1b nsp12 P323L (as known as RNA-dependent RNA polymerase) showed very high concurrence ratio with other mutations since they emerged in the early phase of the pandemic.²³ Previously it has been shown that S protein D614G co-evolved with ORF1b nsp12 P323L.²³ The combination of both mutations may enhance viral fitness based on epidemiological data, although the molecular mechanisms of this evolutionary advantage remain elusive.²³ Interestingly other mutations with high or medium concurrence ratio are found in more than 90% of the lineage B.1.1.7 in the alpha variant, in which its frequency peaked between March-May 2021. However, the ORF3a Q57H mutation with the lowest concurrence ratio is absent in the lineage B.1.1.7. Unfortunately, the information on the frequency percentage of ORF1a nsp5 P108S, ORF1b nsp12 A423V and N protein P151L mutations in the primary lineages of different VOCs are not available. In another study, it has been predicted that multiple SARS-CoV-2 genes may have epistatic interactions linked to viral fitness.²⁴ The effects of a mutation can be neutral, harmful, or beneficial to the virus. It is expected that most single mutations have a small effect on viral fitness. It remains an arduous task to associate a specific phenotype with a single viral mutation since it is possible that a specific phenotype is contributed to by the effects of multiple mutations.

There are huge numbers of single nucleotide polymorphisms (SNPs) present in the SARS-CoV-2 genome, hence evaluating the biological functions of all SNPs using experimental approaches is not feasible. Therefore, prediction of the effects of SNPs allows us to prioritize variants which may have some significant biological functions. Our study used the meta-prediction approach to perform functional predictions of nonsynonymous mutations to minimize the false positive rate. When two or three tools are combined, the prediction accuracy increases and reaches greater performance, however, the sensitivity is subsequently decreased as more tools are combined.²⁵

Of all these nine protein mutations, only two mutations namely ORF1a nsp5 P108S and S protein D614G were predicted to reduce their stability whereas only S protein D614G may have more a flexible conformation compared to the wild type. S protein binds to human ACE2 receptors to gain access to the host cell.⁷ D614G mutation is found at S1 domain which is involved in receptor binding.²⁶ Two independent studies of S protein D614G mutant structures derived from cryo-electron microscopy analysis has demonstrated that the G614 mutant adopts a more open conformation compared to D614 wild type.^27,28 Interestingly, an in vitro study has shown that S protein D614G mutation may enhance virus infectivity by promoting the packing of S protein into the virion, not by enhancing the binding of S protein to the ACE2 receptor.²⁹ On the other hand, ORF1a nsp5, also known as 3C-like protease is responsible for cleaving viral polypeptides during replication.² A study by Abe et al., (2021) has showed that ORF1a nsp5 protein P108S mutation associated with the clade, 20B-T (lineage B.1.1.284), diminished its activity, possibly leading to a reduction in disease severity.³⁰

Since the protein function depends directly on the three-dimensional structure of the protein, we wanted to see if these mutations may affect the function of the protein using SIFT 4G and PROVEAN prediction tools. The PROVEAN tool is applicable for all organisms. SIFT4G, instead of SIFT was used since it allows us to build a SARS-CoV-2 genome database with variant annotation. Interestingly ORF1a nsp5 protein P108S mutation was the only mutation found to be deleterious from both SIFT4G and PROVEAN functional analysis. Together with the DynaMut stability result, it has been demonstrated that this mutation may be harmful to the virus itself, and can be less damaging to the human host as reported by Abe et al. (2022).³⁰ On the other hand, ORF3a Q57H and N protein P151L mutations are predicted to be deleterious by the PROVEAN tool only. ORF 3a is an ion channel (viroporin) which is involved in viral egress steps through lysosomal trafficking.^31,32 ORF3a Q57H mutation not only causes a change in amino acid in ORF3a, but also produces a truncated ORF3b due to the overlapping protein-coding sequences shared by ORF3a and ORF3b.³³ However, there are conflicting results about the effect of the ORF3a Q57H mutation on the human host immune response.^33,34 N protein is involved in the liquid-liquid phase separation for the viral genome packaging.³⁵ N protein P151L mutation is located at the RNA binding domain. It has been proposed that this mutation may disrupt the protein-drug interaction.³⁶ Although another two N protein mutations, R203K and G204R were not predicted to be deleterious in our study, they have been identified in the alpha variant, B.1.1.7, gamma variant, P.1, lambda variant, C.37 and omicron variant, BA.1.³⁷ While N protein, T205I mutation has been reported in the beta variant, B.1.351 and Mu variant, B.1.621.³⁷ More recently, another N protein mutation, R203M has been reported in the delta variant, B.1.617.2.³⁷ Interestingly mutants with N protein S202R or R203M mutations can pack more RNA material compared to the wild type based on in vitro studies.³⁸ These observations and experimental results suggest that N protein residues, S202, R203, G204 and T205 may play some role on viral RNA replication.

The SARS-CoV-2 virus genome data from GISAID database ranging from 1st January 20 to 22 March 21 were analyzed in this study. The frequency of alpha variant peaked around March-May 21 in most countries.³⁷ Hence, it is not surprised that five nonsynonymous mutations, including ORF1b nsp12 P323L, S protein N501Y, S protein D614G, N protein R203K and N protein G204R identified in this study were also part of the defining mutations in the alpha variant.³⁷ Note that ORF1b nsp12 P323L is the same mutation as ORF1b nsp12 P314L, nsp12 is located between ORF1a and ORF1b, and the last 9 overlapping amino acid residues from ORF1a, SADAQSFLN were included in ORF1b nsp12 P323L. The mutational profile of SARS-CoV-2 genome is changing very rapidly. However, it is out of the scope of this paper to monitor the mutational changes of SARS-CoV-2 genome since it is impossible to keep up with the exponential growth of these data. Interestingly two preprints on genomic surveillance analysis using specimens collected in late 2021 have showed that a small number of wild deer in North America carry the alpha or alpha-like variant.^39,40 Although the alpha variant circulating in human population has been replaced by other variants, it remains to be seen if the alpha variant would jump back from animal to human. Furthermore, these five nonsynonymous mutations found in the alpha variant were also found in the current dominant variant, the omicron variant as shown in Table 3. Therefore, it is still relevant to study the consequences of these mutations.

Conclusion

In this study, ORF1a nsp5 P108S, S protein D614G, ORF3a Q57H and N protein P151L mutations have been predicted to alter their structures and/or functions. Since all the reported variants of concern contain multiple mutations present in multiple SARS-CoV-2 proteins, it is necessary to evaluate the impact of these mutations in combination on viral transmission and pathogenicity. The biological consequences of these nonsynonymous mutations of SARS-CoV-2 proteins should be further validated with in vivo and in vitro experimental studies in the future.

Ethics and dissemination

No ethical approval is required for data analysis in this study (EA0802021).

Author contribution

CHN contributes to the concept, design, supervision of the project. SBZ, WXB, YYY and SK contribute to the design, methodology, and data collection. SBZ, WXB, YYY and SK contributed to the analysis, and interpretation of data.

All authors were involved in drafting and revising the manuscript and approved the final version.

Data availability

Software availability

The python script used for the identification of SARS-CoV-2 genome mutations can be obtained through GitHub (https://github.com/wxboon98/Mutations-Identification).