Nonself Mutations in the Spike Protein Suggest an Increase in the Antigenicity and a Decrease in the Virulence of the Omicron Variant of SARS-CoV-2

Despite extensive worldwide vaccination, the current COVID-19 pandemic caused by SARS-CoV-2 continues. The Omicron variant is a recently emerged variant of concern and is now taking over the Delta variant. To characterize the potential antigenicity of the Omicron variant, we examined the distributions of SARS-CoV-2 nonself mutations (in reference to the human proteome) as 5 amino acid stretches of short constituent sequences (SCSs) in the Omicron and Delta proteomes. The number of nonself SCSs did not differ much throughout the Omicron, Delta, and Reference Sequence (RefSeq) proteomes but markedly increased in the receptor binding domain (RBD) of the Omicron spike protein compared to those of the Delta and RefSeq proteins. In contrast, the number of nonself SCSs decreased in non-RBD regions in the Omicron spike protein, compensating for the increase in the RBD. Several nonself SCSs were tandemly present in the RBD of the Omicron spike protein, likely as a result of selection for higher binding affinity to the ACE2 receptor (and hence higher infectivity and transmissibility) at the expense of increased antigenicity. Taken together, the present results suggest that the Omicron variant has evolved to have higher antigenicity and less virulence in humans despite increased infectivity and transmissibility.


28
Despite worldwide efforts for vaccination, the COVID-19 pandemic still prevails as 29 of December 2021, two years after its pathogenic emergence caused by a novel corona- 30 virus, SARS-CoV-2 [1][2][3][4]. Recently, a new variant emerged from South Africa, which was 31 announced on 25 November 2021 [5] and was designated the Omicron variant (B.1.1.529), 32 one of the variants of concern announced by the World Health Organization (WHO) on 33 26 November 2021 [6]. A risk assessment of the Omicron variant was urgently released on 34 2 December 2021 [7]. Currently, the Omicron variant is spreading worldwide, including 35 in Denmark [8] and the United States of America [9], displacing a previous variant of con- 36 cern, the Delta variant (B.1.617.2) [6]. The Omicron variant contains a large number of 37 unique mutations and appears to have higher infectivity and transmissibility than previ-38 ous variants, raising a public health concern [10][11][12][13]. Prompt characterization of the Omi- 39 cron variant is of high importance for public health. 40 Possible functional changes associated with mutations in the Omicron variant have 41 already been evaluated by several computational studies. Several mutations have been 42 localized in the receptor binding domain (RBD) of the spike (S) protein, possibly contrib- 43 uting to higher affinity to the ACE2 receptor and lower affinity to pre-existing infection-44 induced or vaccine-induced antibodies [13][14][15]. An increase in hydrophobic amino acid 45 residues [15] or in electrostatic interactions [16,17] introduced by mutations has been 46 linked to potentially higher infectivity and transmissibility. Computational analyses, 47 including a structural analysis [13], an analysis based on artificial intelligence (AI) trained 48 with numerous experimental data [18], and an analysis with amino acid interaction (AAI) 49 networks [19], have revealed potential resistance of the Omicron variant against pre-ex-50 isting antibodies. A large-scale SARS-CoV-2 genome analysis has suggested that Omicron 51 mutations have been selected for vaccine resistance [20]. Consistent with these in silico 52 studies, in vitro functional analyses of the Omicron variant have suggested significant 53 escape from pre-existing infection-induced or vaccine-induced antibodies [21,22]. How-54 ever, pre-existing T-cell immunity has still been effective [22,23]. These studies have al-55 ready addressed some of the concerns associated with the Omicron variant, but alterna-56 tive and complementary methods to further characterize new variants will also be helpful. 57 Here, we employed a novel method simply based on amino acid sequences of SARS-CoV-58 2 and human proteomes to evaluate the antigenicity of the Omicron variant. 59 The human immune system recognizes foreign proteins based on short amino acid 60 sequences presented as peptides by MHC molecules [24][25][26]. In the present study, a stretch 61 of an amino acid sequence, when it exists as a part of a protein, is called a short constituent 62 sequence (SCS; pronounced as [es/si:/es] or [ʃɔks]). Operationally speaking, the human 63 immune system stores memories of all possible SCSs from the human proteome, which 64 are here called self SCSs. Every peptide presented by MHC molecules is collated with a 65 dataset of self SCSs. When a sequence of a given peptide presented by MHC molecules is 66 found in the dataset, it is recognized as "self", a part of a human body. In contrast, when 67 a sequence of a peptide is not found in the dataset, it is recognized as "nonself", a foreign 68 object to be eliminated by the immune system. These SCSs in a protein are called nonself 69 SCSs, and they are antigenic by definition, although their antigenicity in vitro and in vivo 70 should be determined experimentally. When a given peptide sequence increases rapidly 71 to emergency levels when a microbe or virus infects a human body, the immune system 72 likely produces antibodies against it even if it is found in a dataset. However, self SCSs 73 may still be less antigenic than nonself SCSs. 74 We reasoned that the relative abundance of self and nonself SCSs in a virus can be 75 used as an indicator to evaluate the antigenicity and thus the virulence of that virus be-76 cause a nonself SCS is more antigenic than a self SCS for the host immune system to avoid 77 autoimmunity. When the number of nonself SCSs increases in a viral proteome during 78 evolution, this means that the virus is more discoverable by the host immune system be-79 cause nonself SCSs can be easily recognized as foreign objects. As a result, viral virulence 80 may decrease. Although protein analysis methods based on SCSs have been developed 81 since 2005 [27][28][29][30][31][32][33], we introduced the self/nonself concept in SCS-based analysis in a pre-82 vious study for the first time [34]. In that study, we showed that when using 5 amino acid 83 stretches as SCS units, most SCSs in the SARS-CoV-2 proteome are self SCSs in reference 84 to the human proteome [34]. In other words, nonself SCSs are scattered in a sea of self 85 SCSs. We also discovered nonself SCS clusters in the spike protein that may serve as an 86 excellent candidate epitope for vaccines that efficiently induce immunity without induc-87 ing autoimmunity [34]. In the present study, we attempted to develop a simple method 88 for predicting antigenicity based on self/nonself SCSs to understand the relationship be-89 tween the Omicron variant and the human immune system.

91
The human reference proteome and the SARS-CoV-2 proteome sequences were ob-92 tained from NCBI (the National Center for Biotechnology Information, Bethesda, MD, 93 USA) as described elsewhere [34]. In addition to the SARS-CoV-2 proteome reference se-94 quence (ASM985889v3), the Delta and Omicron variant proteomes were obtained simi-95 larly (accessed on 30  SCSs (containing 5 amino acids) were extracted from the SARS-CoV-2 proteomes by 105 sliding one amino acid residue at a time from the N-terminus to the C-terminus. All SARS-106 CoV-2 SCSs were categorized into either self (existent in the human reference proteome) 107 or nonself (nonexistent in the human reference proteome) SCSs as described elsewhere 108 [34]. We assigned each SCS in the SARS-CoV-2 proteome a 0 (self; invisible from the host 109 immune system) or 1 (nonself; visible from the host immune system) at the first position 110 of its amino acid in a protein sequence [34]. The numbers of nonself SCSs were counted 111 and assigned in sequence maps manually based on the self (0) or nonself (1) assignments 112 calculated by our program and exported into Microsoft Excel [34]. For analyses, ORL7b 113 was excluded because some proteome files did not show this protein as being translated. 114 ORF1a was also excluded because its sequence was completely redundant with that of 115 ORF1ab. 116 In this study, a nonself mutation indicates a mutation that causes a self-to-nonself 117 SCS status change. As a result of such a mutation, a nonself SCS is produced. A nonself 118 SCS is considered antigenic by definition. Similarly, a self mutation indicates a mutation 119 that causes a nonself-to-self SCS status change, which produces a self SCS. A self SCS is 120 considered nonantigenic. An increase in nonself SCSs in a protein or in a proteome means 121 a decrease in self SCSs and vice versa.

123
We first characterized the numbers of nonself SCSs in the proteomes of the Delta and 124 Omicron variants in comparison with that in the RefSeq (reference sequence) proteome. 125 The numbers of nonself SCSs in the proteomes of the Delta and Omicron variants did not 126 differ much from that in the RefSeq proteome ( Table 1). The number of nonself SCSs in 127 the Delta variant proteome increased by just one, and the number of nonself SCSs in the 128 Omicron variant proteome decreased by just one, although these data did not indicate 129 that there were no self/nonself status changes in these variants. The increase or decrease 130 in the number of nonself SCSs in each protein, ΔN, in reference to RefSeq was not remark-131 able; mostly either 0 or ±1.   (Figure 2). Focusing on the RBD, the Delta variant had 2 nonself SCSs created by 153 mutations (self-to-nonself status changes), and no nonself SCSs found in RefSeq disap-154 peared (nonself-to-self status changes) (Figures 1 and 3). Thus, the net increase in nonself 155 SCSs was +2. In contrast, the Omicron variant had 7 nonself SCSs (due to the following 7 156 mutations, G339D, S375F, S477N, T478K, Q498R, N501Y, and Y505H, among 15 muta-157 tions), and 3 nonself SCSs found in RefSeq disappeared (Figures 2 and 3). Thus, the net 158 increase in nonself SCSs was +4. 159 Interestingly, this tendency of an increase in the number of nonself SCSs in the RBD 160 was not observed in the non-RBD regions. Instead, the net changes in the number of non-161 self SCSs in the non-RBD region of the Omicron and Delta variants were -3 and -1, respec-162 tively ( Figure 3). The differences in the net changes between the RBD and non-RBD re-163 gions of the Omicron and Delta variants, ΔN, were +7 and +3, respectively (Figure 3). In 164 both variants, an increase in the number of nonself SCSs in the RBD was compensated for 165 by a decrease in the number of nonself SCSs (an increase in the number of self SCSs) in 166 the non-RBD regions (Figure 3), resulting in a net spike change of just +1 (Table 1, Figure 167 3). 168 Notably, in the receptor binding motif (RBM) of the Omicron variant, 3 novel nonself 169 SCSs (YQAGN, NKPCN, and KPCNG) were localized immediately at the N-terminal side 170 of the potential vaccine epitope identified in a previous study [34], extending the epitope 171 region toward the N-terminal side (Figure 2). Furthermore, 2 novel nonself SCSs (FRPTY 172 and GVGHQ) were localized immediately at the C-terminal side of the potential epitope, 173 extending the epitope region toward the C-terminal side up to the end of the RBM (Figure 174 2). On the other hand, one nonself SCS (QSYGF) found in RefSeq and the Delta variant 175 disappeared (Figures 1 and 2). Together, in a 34 amino acid stretch of the C-terminal end 176 of the RBM, 8 nonself SCSs (YQAGN, NKPCN, KPCNG, PCNGV, GFNCY, FNCYF, 177 FRPTY, and GVGHQ) were concentrated tandemly.

179
In the present study, we reasoned that the number of nonself SCSs in a proteome or 180 in a protein is a determinant of its antigenicity. We further reasoned that antigenicity is 181 probably reflected directly in the virulence of the virus. Not only the number but also the 182 locations of nonself SCSs in a protein (e.g., in a binding domain, at an active site, or in 183 proximity to one another) would matter to determine the antigenicity of nonself SCSs. 184 Based on the above logic, we endeavored to urgently characterize the Omicron variant of 185 SARS-CoV-2. Although immunological validity should be evaluated in further studies, 186 this study proposes a novel method of predicting viral virulence based on the amino acid 187 sequences of viral proteins. 188 Theoretically, a virus evolves under selection pressure for higher infectivity and 189 transmissibility, and in this evolutionary process, the amino acid residues of the binding 190 site for its receptor are mutated for higher affinity. This type of evolution is here called 191 "offensive" evolution. Offensive evolution may attenuate when any functional disad-192 vantage unavoidably associated with excessive mutations for higher affinity overwhelms 193 the mutational functional advantage. Offensive evolution may also be attenuated when 194 nonself SCSs accumulate for higher antigenicity. 195 Alternatively, a virus theoretically evolves for higher sequence mimicry to the host 196 SCS repertoire defined by the host proteome [34], and in this evolutionary process, nonself 197 SCSs decrease in number for lower antigenicity. This type of evolution is here called "de-198 fensive" evolution. Defensive mimicry of host SCSs by increasing self SCSs is advanta-199 geous for viruses because they cannot be readily detected by the host immune system. 200 Excessive mimicry evolution may also be attenuated due to an unavoidable functional 201 compromise. 202 Given a sufficient period of time, an equilibrium state between offensive and defen-203 sive evolution may be established through a functional compromise of a virus. However, 204 in reality, a virus may evolve within a limited period (i.e., replications). Either offensive 205 or defensive evolution may settle at local maxima of survival, depending on various en-206 vironmental and host conditions. 207 We discovered that the Omicron variant has accumulated self-to-nonself status 208 change mutations at the RBD of the spike protein. The Omicron mutant had 7 new nonself 209 SCSs in the RBD, in contrast to just 2 additions in the Delta variant. The net increase in 210 nonself SCSs within the RBD was +4 in the Omicron variant, in contrast to +2 in the Delta 211 variant. This increase is likely immunologically significant, considering that the RBD is 212 readily accessible to other proteins, such as ACE2, antibodies and T-cell receptors. Inter-213 estingly, there were just 2 newly added nonself SCSs in the non-RBD regions in both the 214 Omicron and Delta variants, but in the Omicron variant, more new self SCSs were intro-215 duced in the non-RBD regions compared with the non-RBD regions in the Delta variant 216 and the RBD of the Omicron variant. This increase in self SCSs in the non-RBD regions in 217 the Omicron variant may compensate for the increase in nonself SCSs in the RBD to make 218 the net nonself SCS increase small for lower antigenicity. Indeed, the differences in nonself 219 SCS changes between the RBD and non-RBD regions, ΔN, were +7 and +3 in the Omicron 220 and Delta variants, respectively. 221 It seems that for the Omicron variant, the selection pressure for higher binding affin-222 ity to ACE2 (i.e., higher infectivity and transmissibility) is so large that an increase in the 223 number of nonself SCSs in the RBD was unavoidably allowed despite its immunological 224 disadvantage for the virus. In other words, an increase in nonself SCSs in the RBD likely 225 means a compromise of the virus, which probably indicates low virulence. In support of 226 this view, nonself mutations in the RBD appear to contribute directly to higher affinity to 227 ACE2 [18]. Therefore, the Omicron variant is probably a product of offensive evolution 228 but with a compromise of high antigenicity. The Delta variant may also be a product of 229 offensive evolution but to a lesser degree. Due to relatively high antigenicity, the current 230 offensive evolution of the Omicron variant may attenuate at some point. 231 Based on the idea that SCS frequencies in proteomes are related to the phylogenetic 232 and parasitic status of organisms [39,40], we speculated previously that the SARS-CoV-2 233 proteome will eventually evolve to accumulate self SCSs to defray molecular and cellular 234 attack from the human immune system [34]. However, this defensive evolution based on 235 the sequence mimicry hypothesis was not detected clearly in the Omicron and Delta var-236 iants. An increase in self SCS was detected only in the non-RBD regions in the Omicron 237 and Delta variants. Gears may be changed at some point toward a defensive mode of 238 evolution, but offensive evolution may settle at local maxima of survival for long periods 239 of time. 240 In addition to the above discussion on the number of nonself SCSs, further discussion 241 can be made in terms of the location of nonself SCSs within the RBM in the RBD. We 242 discovered a candidate stretch of amino acids that contained nonself SCSs within the RBM 243 of the RefSeq spike protein, a potential epitope for vaccine development to avoid vaccine-244 induced autoimmunity [34]. Interestingly, in the Omicron variant but not in the Delta var-245 iant, 3 additional nonself SCSs (YQAGN, NKPCN, and KPCNG) are present at the N-246 terminal side of the potential epitope region, and furthermore, 2 additional nonself SCSs 247 (FRPTY and GVGHQ) are present at the C-terminal side, although one nonself SCS in 248 RefSeq is not present. Together, the candidate epitope region in RefSeq has been extended 249 to both the N-terminal and C-terminal sides. Because this region in the RBM is important 250 for ACE2 binding, the expansion of this nonself SCS epitope region here was probably 251 unavoidable to increase the binding affinity for ACE2 at the expense of an increase in 252 antigenicity. This result supports the previous finding that this region is likely a good 253 (probably the best) epitope candidate for vaccine development [34]. 254 At first glance, the present results may not seem to be consistent with recent studies 255 on the Omicron variant, which suggested an increase in virulence [10][11][12][13][14][15][16][17][18][19][20][21][22]. These studies 256 evaluated whether the pre-existing immunological memory induced by previous vaccines 257 or infection continues to be effective against the Omicron variant, but in the present study, 258 the antigenicity of the Omicron variant itself was evaluated. The present study suggests 259 that the Omicron variant has reduced virulence because of its relatively high antigenicity. 260 Therefore, the Omicron variant may not cause severe symptoms. However, due to its high 261 infectivity and transmissibility, as suggested by other studies [10][11][12][13][14][15][16][17][18][19][20][21][22], the Omicron variant 262 should still be regarded with alarm. 263 Antigenicity defined by nonself SCSs is just a single factor to explain virulence. A 264 virus with low virulence might preferentially affect tissues/organs such as the digestive 265 tract and testes, leading to non-life-threatening but long-term effects, rather than tis-266 sues/organs such as the lungs, the effects of which could be life-threatening. For example, 267 infection may cause infertility if testicular cells expressing ACE2 are preferentially in-268 fected. Moreover, due to high infectivity and transmissibility, even if virulence is low, the 269 absolute number of hospitalized people may not decrease in the Omicron pandemic in 270 comparison with the Delta pandemic. Therefore, the Omicron variant should be consid-271 ered highly threatening in terms of public health. 272 The frequency of random mutations is directly proportional to the number of repli-273 cations (i.e., the number of infected people), but selection pressure shapes the direction of 274 viral evolution. The higher binding affinity to ACE2 and higher antigenicity together 275 caused by the accumulated mutations in the Omicron variant suggest that a hampered 276 transmission state despite a large number of replications was a driving force for the evo-277 lution of the Omicron variant. Ironically, such an unnatural state was created by world-278 wide vaccination, which might have helped the emergence of the Omicron variant, as 279 suggested by a recent study [20].

281
Through an application of the SCS concept to the human-SARS-CoV-2 system to-282 gether with immunological self/nonself considerations, in the present study, a novel 283 method was used to characterize the Omicron and Delta variants. It appears that the Omi-284 cron variant increased its infectivity and transmissibility at the expense of higher antigen-285 icity and lower virulence. Thus, the symptoms of individuals infected with the Omicron 286 variant may be less severe than those of individuals infected with the Delta variant. We 287 also confirmed that a specific stretch in the RBM is a good candidate epitope for vaccines. 288 Since SARS-CoV-2 evolution might have been driven by selection pressure imposed by 289 worldwide vaccination, vaccination-focused strategies against the Omicron variant may 290 further enhance a current mode of evolution. Alternatives to vaccination-focused 291 strategies [41][42][43][44][45][46][47] may also be useful. Continuous caution regarding the Omicron variant 292 is necessary. Data Availability Statement: All data that support the conclusions of the study are included in this 307 paper and the related Supplementary Information. The source codes for human SCS analysis and 308 for SARS-CoV-2 SCS analysis are freely available at https://adslab-uryukyu.github.io/scs-sars-cov-309 2/. 310