Country specific mutational profile of SARS-CoV-2 in pre- and post-international travel ban: Effect on vaccine efficacy

2.1 SARS-Cov-2 sequences

A total of ∼2,20,000 SARS-CoV-2 sequences were downloaded from GISAID (https://www.epicov.org/epi3/), available as of November 28^th 2020. For each of the ORFs/proteins, partial sequences were filtered out, which accounted for less than 1% of the total isolates. The protein sequences were individually aligned using MUSCLE. The SARS-CoV-2 strain, originating in Wuhan, China (GenBank accession ID ‘NC_045512’) was considered as the reference strain. Since countries started implementing international travel ban from the third week of March extending up to early April, the isolates were segregated into two groups, based on whether they were sequenced before May (Group B for ‘before’) or after May (Group A for ‘after’) 2020. We were left with a total of ∼2,15,000 sequences, out of which 1,35,000 sequences are in Group A and 80,000 sequences are in Group B.

2.2 Selection of countries for carrying out mutational profiling

We have selected those countries for which there were a minimum of 100 complete sequences, from May to November 28th, provided that these countries also contained a minimum of 30 sequences in the span of January to April. By this measure, we obtained a total of 35 countries in which the mutational characterization was carried out, arranged according to their geographical locations (Figure 1)

• Download figure
• Open in new tab

Figure 1:

Heat map representing the amino acids substitutions in all SAS-CoV-2 proteins, observed with a frequency of 10% or higher in at least one country in (a) both Group A and Group B; (b) Group B, and (c) Group A only.

2.3 Identification of frequently mutated residues for each of the viral proteins

Non-synonymous mutations were identified independently for each of the selected countries for Group A and B. A complete summary with all the unique variants, each characterized by mutations in one or more amino acid positions, along with the frequency of occurrences are calculated. We have defined a substitutions having at least 2% occurrence in all the variants as a frequently mutated residue. An overall summary table depicting the frequencies for all the 35 selected countries, covering each of the identified frequently mutated residues are provided separately for the two groups (Supplementary Tables1a and 1b).

2.4 Comparison of non-synonymous mutation in pre- and post-international travel ban

In order to obtain a bird’s eye view of the impact of the international travel ban in shaping the mutations in SARS-CoV-2, we have calculated a weighted average of the number of frequently mutated residues for each structural and non-structural protein in 35 selected countries (Table 1). The prevailing global trends with the most mutable proteins in both the groups were compared. For the individual countries, both before and after the travel ban were impinged, the number of frequent mutations per 100 amino acids were calculated (Supplementary Table 2). The variations in the mutation patterns for one or more countries from the global framework were observed.

2.5 Classifying the different mutations on account of their temporal variations

The mutated residues were grouped into three classes. The first of these occurred with a frequency of 10% and higher in at least one of the countries in group A and B were considered to have originated before the stipulation of inter-country travel ban and continued to prevail, either in the previous location or in other regions around the globe. On the contrary, mutations which transpired in 10% of the strains in the early months of the pandemic but were observed to dwindle down after May were bracketed in the second class, whereas mutations that were virtually non-existent before May but emerged with higher frequencies during the ensuing months were placed in the third class. A heatmap depicting the frequencies of these mutations in their respective groups was generated using the R package ‘gplots’, where hierarchical clustering of the various substitutions was carried out using ‘Ward.D2’ method (Figure 1).

2.6 Tracking the evolution of the mutated residues over time in a bi-monthly basis

For each of the mutations comprising the three groups, we have calculated their frequency of occurrences in every two months, from January to October 2020. Since the amount of sequencing in each country is largely inconsistent, in order to capture the mutational frequencies with a fair degree of confidence, we chose to include the total sequences for every two months for frequency calculation instead of conducting the study on a monthly basis. The frequencies were calculated only for those countries in which the mutation was observed to occur with a minimum of 10% abundance in at least one of the time-points considered.

2.7 Country specific co-occurring amino acid substitutions

We carried out analysis of co-occurring amino acid substitutions, considering those that have occurred with 10% or higher frequencies in each country, encompassing all viral proteins and have analysed for the residues which have co-occurred with higher abundance. This allowed us to identify the viral variants which were showing greater dominance in a certain region along with the co-occurring mutations in the respective variants. We have also discerned the substitutions which had occurred simultaneously in the countries prone to those substitutions, and have further corroborated the fact from the bi-monthly frequencies of such variants.

2.8 Overlapping the frequent mutations in the S-protein with predicted epitopes from external studies

A number of studies have been carried out for speedy engineering of vaccines against the SARS-CoV-2 virus. We selected eight studies which have experimentally determined potential immune-responsive epitopes owing to several viral proteins, of which we are interested especially with the epitopes delegated towards the Spike protein (Farrera-Soler, Daguer et al. 2020, Li, Lai et al. 2020, Mishra, Huang et al. 2020, Poh, Carissimo et al. 2020, Shomuradova, Vagida et al. 2020, Wang, Wu et al. 2020, Yi, Ling et al. 2020, Musico, Frigerio et al. 2021) (Supplementary Table 5a). Several studies have used in-silico approaches to predict T-cell and B-cell epitopes that could potentially be employed to trigger immune responses in the host. We have selected three such independent studies which used different methodologies to predict for putative epitopes that could be implemented in designing safe and effective vaccines (Cun, Li et al. 2020, He, Huang et al. 2020, Yarmarkovich, Warrington et al. 2020) (Supplementary Table 5b). These potential epitopes were intersected with all frequently mutated S-protein substitutions obtained in our study. The mutations found to be harboured within one or more of the epitopic sites were observed along with the respective countries in which they were dominant. To further ascertain the progression of these mutants over time, we studied the bi-monthly frequencies of these variants.

3.1 Frequently mutated residues in SARS-CoV-2 during pre- and post-travel ban

To assess the effect of international travel restrictions in shaping the mutational profile in 35 countries, all protein sequences of SARS-CoV-2 comprising ∼2,15,000 isolates were analysed. We identified 334 amino acid substitutions that were observed in at least 2% of the total number of isolates, for at least one country in the period from January to April 2020, while 656 such frequently mutated residues were obtained in the interval from May to November 2020 (Supplementary Tables 1a and 1b). Among the structural proteins, the Envelope and Membrane proteins were the most resilient to non-synonymous changes. On the other hand, the Spike and Nucleocapsid proteins displayed greater plasticity, respectively accounting for 41 and 37substitutions during the initial four months (group B, for ‘before’), and 80 and 69 substitutions during post travel ban (group A, for ‘after’). ORF3a, ORF8 and ORF7a were the most mutable in their respective phases. Expectedly, ORF1ab, which constitutes around 70% of the viral genome, represented by 16 NSPs had encountered the most number of substitutions, cumulatively equalling to 214 and 408substitutions in Group B and A respectively. Probing into the substitutions sustained by the NSPs, NSPs 2, 3 and the replication transcription complex (RTC) proteins NSP-12, 13 and 14 were the highest in the group. The summary tables containing the frequencies of each of the frequently mutated residues, as observed in 35 countries for both groups are presented in Supplementary Tables 1a and 1b.Overall, we found that the number of mutations in Group A had at least doubled compared to their status as observed in Group B.

3.2 Non-synonymous mutations per 100 amino acids in the pre- and post-lockdown

A weighted average of the frequently mutated residues per 100 amino acids (≥ 2% of all variants in a country) across all countries was calculated separately for both Group B and A (Table 1). For the months preceding the travel ban restrictions, the global trend suggested that the mutation rate per 100 aa was the highest for ORF3a(0.84%)and ORF8 (0.94%)for the non-structural proteins while Nucleocapsid (0.93%)exhibited the highest rate among the structural proteins. When calculated for the following months, Nucleocapsid(1.56%), Spike proteins (0.4%), ORF3a (1.22%)and ORF7b(0.61%)were found to have the highest rates of mutations. Nucleocapsid, ORF3a and ORF7b showed the largest surge in mutation rates over the two phases, while ORF8 depicted opposite behaviour with an overall decrease in rate by 0.75 residues per 100 aa(Table 1).The structural proteins E and M continued to display remarkable robustness and remained largely conserved over the entire period.

Upon examining the rate of mutations for each country(Supplementary Table 2), some differences were observed. In the months leading to the lockdowns where ORF3a had the highest mutation rates in 13 countries, followed by ORF3a and Nucleocapsid in 8 countries each. However, in the period from May to November, Nucleocapsid displayed the highest mutation rates in 18 of the 35 countries, while in 7 countries ORF3a had encountered the highest number of mutations (Supplementary Table 2). Country-wise mutation rates were most stark in case of ORF8, which had remained immune to changes in 21 countries while effecting fairly high rates of substitutions in the remaining countries. Apart from these, ORF7a in South Korea, ORF7b in UK and Latvia, and ORF8 in Kenya and South Africa showed the highest rate of mutations.

3.3 Mutations that have occurred with relatively high frequencies (≥10%) in one or both groups under study

3.3 Mutations that have newly emerged or increased sharply post-travel ban

We next surveyed the mutations that were virtually non-existent before May, but gained prominence in specific locations post travel ban (Figure 1c). Interestingly, we noticed that the European countries formed a separate cluster comprised of4 mutations in the S protein, 3 each in Nucleocapsid and ORF3a, and others encompassing NSP4, NSP6, NSP7, NSP9, NSP12, NSP13 and NSP16. These were clearly not pervasive before May and had come into effect in the wake of the travel bans. Apart from these, other noteworthy mutations that originated in specific countries post-lockdown were found in Nucleocapsid and NSPs3, 5 and 12 in Japan, three mutations comprising the S protein, NSP3 and NSP15 in China. Additionally, substitutions exclusive to South Korea among NSP7 and NSP16, mutations in S and NSP3 specific to Czech Republic, two mutations in ORF7a and NSP3 prevalent only in Kenya, and two variations within NSP3 and Sin Canada were among the ones that were absent in the pre-lockdown era.

3.4 Temporal survey of the mutated variants on a bi-monthly basis

Upon identifying mutations specific to certain countries, we wanted to see whether these variations had come into being or declined sharply at a distinct time point. Estimation of bi-monthly frequencies for the corresponding countries (Supplementary Table3) allowed us to perceive three distinct trends with which the mutations developed in the course of time. The substitution S_D614G was seen to have attained close to 100% concordance in most countries by October, with North American and European countries gaining the mutation at a faster rate compared to Middle East and Asian countries (Figure 2a-c). In contrast, the co-occurring substitutions R203K and G204R in Nucleocapsid showed more diverse variations (Figure 2d-f). In most of the countries, these substitutions showed greater dominance over the wild type, with countries like Australia, Mexico, Russia, Japan, Bangladesh and Latvia, where the corresponding frequencies rose to ∼90% and higher. However, in several European countries namely Belgium, Denmark, France, Germany, Ireland, Netherlands, Norway, Spain, Switzerland and UK, and also in USA and Brazil, this variant has not gained a firm footing over its wild-type counterpart, remaining more or less constant over the specified time period or plummeting after brief ascension. ORF3a substitution Q57H continued to reign disproportionately in different countries, where it can be seen to have attained a semblance of stability in USA, Canada, Mexico and Sweden (Figure 2g), while low abundance or plunging rates were observed in Czech Republic, Germany and Australia among others (Figure 2h). Contrarily, Belgium, France, Switzerland and China were among the nations where this variant was seen to be on the ascension after a brief hiatus around July-August (Figure 2i).

• Download figure
• Open in new tab

Figure 2:

Bi-monthly frequency plots of substitutions (a, b and c) D614G in Spike protein, (d, e and f) R203K in Nucleocapsid and (g, h and i) Q57H in ORF3a both before and after international lockdowns. S_D614G (a, b and c) and N_R203K (d, e and f) plots have been segregated according to geographical locations, while the three distinct trends of mutation of ORF3a_Q57H (g, h and i) have been shown as separate plots.

There were instances of mutated variants that had gained a brief upsurge pre-travel-ban but were seen to diminish in stature in the ongoing months. D936Y and I119V substitutions in S protein specific to Sweden and Peru showed dwindling characteristics (Figure 3a). A similar trend was observed for S197L substitutions in Nucleocapsid in Spain and Australia, which were coincident with F308Y substitution in NSP4 in the same countries (Figure 3b).Similarly,N_P13L and NSP3_T1198K variants that were appreciably high in India and Singapore before May, sharply reduced concurrently in the later periods(Figure 3c). Moreover, S183Y, Q240K and P344S in Nucleocapsid, characteristic of New Zealand, Ireland and UAE had taken a much sharper nosedive (Figure 3d). NSP2_T85I with the exception of USA had peaked in the interval of March-April before following a downward trajectory (Figure 3e). Among other noteworthy mutations that were on the descent after a brief period of prevalence are NSP12_L323P and ORF3a_G251V as seen in myriad nations (Figure 3f, g).

• Download figure
• Open in new tab

Figure 3:

Bi-monthly plots of selected mutations seen to prevail pre-lockdown but have sharply reduced in abundance in the following months. Various mutations have been shown in representative countries corresponding to the (a-d) S and N proteins, (e) NSP2, (f) NSP12 and (g) ORF3a.

Several new mutations were found to have emanated after May and were seen to be on the rise in specific countries. Our interest was specifically arrested by two substitutions, A222V (Figure 4a) and S477N (Figure 4b) in S protein that were seen to emerge after July and progress rapidly thereafter, the only exception being Norway (Figure 4a). These mutations were exclusively obtained in high percentages largely among the European nations, with the exception of the 477N variant also being widespread and on the rise in Australia (Figure 4b). Among other variants, Spike protein mutations on the ascension post-travel-ban stage included S98F exclusive to Belgium and Netherlands, L18F in the UK and V1122L in Sweden (Figure 4c). N-protein variantsR40C specific to Latvia and P151L in Japan were seen to have reached 50% abundance by October. Among the NSPs, significant mutations that were on the ascendency include NSP2 substitution I120F exclusive to Australia and Bangladesh, coincidentNSP3 substitutions Q203R and S284G in Latvia andI441V in Belgium and S543P in Japan, NSP5_L89F and NSP5_P108S in USA and Japan respectively, P59T variant of NSP10 occurring in Latvia, A423V and V720I mutations of NSP12 in Japan and Czech Republic respectively (Figure 4d).

• Download figure
• Open in new tab

Figure 4:

Bi-monthly plots depicting the mutations that had emerged in specific countries after imposition of travel ban. Spike protein mutations (a) A222V and (b) S477N were found to have increased specific to European countries. Remaining mutations in the (c) Spike and Nucleocapsid proteins, and (d) non-structural proteins corresponding to distinct countries.

3.5 Mutations that have co-occurred with identical or nearly identical frequencies

Now that we had identified prominent mutations in diverse SARS-CoV-2 proteins, we wanted to find the substitutions which have co-occurred in different geographical locations. We have considered variants with a prevalence of at least 10% for a particular country to be the dominant variant. These prevailing variants arising from the merger of various proteins along with the corresponding frequencies are presented in Supplementary Table 4. We have noticed that with the exception of China, all the substitutions have occurred in the background of the Spike protein D614G variant, which in all likelihoods preceded the inception of the other mutations. Apart from the known co-occurring mutations in the Nucleocapsid protein at 203^rd and 204^th residues, we noted several other mutations that have appeared concurrently as also verified by having similar frequencies in one or more countries. Spike protein variant A222V went hand-in-hand with the A220V variant in the Nucleocapsid, seen exclusively in the European nations. F308Y substitution in NSP4 co-occurred with S197L variation in the Nucleocapsid, as observed in Spain and Australia. However, these resulting variants did not prosper and were seen to be obliterated over time. Substitutions G15S and T428I in NSP5 and NSP3 respectively were observed in Peru, Canada, South Africa and New Zealand with appreciable frequencies. These mutations were gained after May but were seen to be decreasing as of October. Substitutions M234I and A376T in the Nucleocapsid, M324I in NSP4, A185S and V776L in NSP12 alongside K218R and E261D in NSP13 were seen to have arisen with identical measures after June and have reached appreciable levels in Belgium, Denmark, France and Switzerland.NSP13 also encountered simultaneous substitutions in 504^th and 541^st residues as seen in USA. ORF3a had similarly manifested corresponding mutations in 172^nd and 202^nd residues observed in Belgium and Netherlands from the month of July (Supplementary Table 3).

3.6 Mapping the frequently mutated residues with Spike protein epitopes

We have mapped the frequently mutated residues observed in the post-lockdown period with S protein epitopes obtained from 11 studies (Figure5, Supplementary Table 5a,b). Among the studies concerned with experimentally determined epitopes, one study (Farrera-Soler, Daguer et al. 2020) escaped convergence with any frequently mutated residues. The remaining reports contained at least one site which overlapped with the observed mutations in the Spike protein. The list of experimentally predicted epitopes with the corresponding mutations nestled within them along with the countries are provided in Supplementary Table 6. Bi-monthly frequencies were calculated for each of these mutations (Supplementary Table 7) to trace the course of their advancement through time, and determine those mutants, which might seem innocuous in the early stages but could pose a threat to vaccine potency if left unchecked.

• Download figure
• Open in new tab

Figure 5:

Illustration of the predicted Spike protein epitopes which have overlapped with one or more frequently mutated Spike protein residues (Cun, Li et al. 2020, He, Huang et al. 2020, Li, Lai et al. 2020, Mishra, Huang et al. 2020, Poh, Carissimo et al. 2020, Shomuradova, Vagida et al. 2020, Wang, Wu et al. 2020, Yarmarkovich, Warrington et al. 2020, Yi, Ling et al. 2020, Musico, Frigerio et al. 2021). The countries in which they were found to be prevalent post lockdown have been mentioned alongside. The residues with frequencies greater than 10% have been highlighted along with the corresponding countries.

Most of the studies predicted epitopes within 500 to 700 amino acids positions of S-protein (Figure 5). This region contained perceived mutational changes, chief among them are substitutions at positions 614 and 626 of S protein. The A626S variant exhibited contrasting trends in Czech Republic as opposed to Norway, Italy and Denmark. In both cases this mutant came into existence July onwards, but diminished sharply after peaking in Czech Republic, while was on the rise in the remaining countries. Other noteworthy mutations which may compromise vaccine effectiveness include T632I (New Zealand), G639S (Italy), E654A (Ireland), Q675H (Germany), Q677H (Sweden, UAE) and P681H (New Zealand), all of which were seen to increase in the months after travel ban albeit at varying rates. Moreover, theV1176Fvariant in Brazil warrants immediate attention as these were seen to accumulate substantially in the span of ten months. Mutations D1163Y and G1167V, both identified in Spain should not be discounted as these were also on the ascendancy as of October, while being present within a putative epitope as outlined by at least 2 studies. Another variant P1162S, shared within epitopes predicted by these studies was observed with a frequency greater than 10% in Portugal around the time of travel ban but lack of sufficient sequencing data in the latter months deprives us to get information of its evolution. Another substitution S459Y soaring over 10% frequency in Czech Republic was also overlapped with predicted epitopes by two studies. However, this variant were seen to subside in abundance by October in the respective countries. A solitary epitope contained P272L substitution which was discerned to rapidly gain grounds in many European nations. Similarly, S477N variant present in one epitopic site was already seen to be abundant in Australia and numerous European nations. Other epitope containing mutations that had an upwards curvature included V483A in UAE, seen only after September, F486L in Netherlands and V90F in Norway and Switzerland seen from July onwards.

The in-silico epitope predictions likewise mapped with several substitutions, the details of which are presented in Supplementary Table 8. These included substitutions L18F, seen to be increasing in UK, A222V rampaging among the European population and M1229I specific to Czech Republic. Among other substitutions, S98F variant as seen previously, exhibited an upwards trajectory, more noticeable in Belgium and Netherlands, M153T and S459Y variants were uncovered in Japan and Czech Republic respectively, while S477N was a major mutant that had disseminated across the European countries.