All variants are innocent unless proven guilty.

     —  Eric Topol, The Scripps Research Institute, California.

1 Introduction

It is close to 2 years since the COVID-19 pandemic started, but reports of new infections from the SARS-CoV-2 are still on the rise. While multiple vaccines of different types are now available, the cause of concern is complete protection from being infected with the new antigenic variants of the original strain first detected in Wuhan, China. Dorothy Hamre first isolated the human coronavirus from the respiratory tract of students admitted to the infirmary at the University of Chicago in 1966 (Hamre and Procknow 1966). This common cold causing strain was later named 229E, which along with OC43, the strain believed to be the reason for the pandemic in cattle and humans in 1890 (King 2020), has been widely prevalent in the world, and most of us probably carry antibodies against them.

Coronavirus belongs to the order of Nidovirus, which are enveloped positive sense single strand RNA viruses that generate 3′ co-terminal nested sub-genomic mRNAs to code for its structural and accessory proteins and mostly infect animals (Di et al. 2018). Being an RNA virus, they have a tremendous ability to mutate as they replicate within their animal reservoir or after they jump to a new species in order to adapt to a new host. A change in only two amino acids was enough for SARS-CoV-1 to adapt in human cells (Li et al. 2005). Similarly, it is believed that a single amino acid change (T372A) may have helped SARS-CoV-2 to spillover from bats to humans (Kang et al. 2021). To make this emergence a success, the virus must evolve fast to gain in its ability to subvert the human immune system and replicate inside the human cells, and this process might happen through several rounds of infection and reinfection. The more the genome replicates, the more the number of mutations that get accumulated as a result, several thousand of variants are produced as it evolves.

This review describes the different types of SARS-CoV-2 variants that are now prevalent, how they are formed, how it affects their infectivity, COVID-19 disease severity, and transmission. More importantly, how do their presence and emergence impact the effectiveness of the current vaccines and future breakthrough infections?

2 What is a variant?

RNA viruses generally have a very high error rate in replication, which on the one hand, is beneficial for them to adapt to selective pressures in the environment but generates a large number of self-eliminating deleterious mutants on the other. This feature also makes developing any chemotherapeutic or vaccine intervention challenging. The low replication fidelity of the RNA-dependent RNA polymerase is partially compensated by the 3′–5′ exonuclease proofreading activity present in the non-structural protein (nsp14) (Robson et al. 2020). This has resulted in SARS-CoV-2 accumulating mutation at a slower rate of around one in every two weeks, corresponding to 1.1 × 10–3 substitutions per site per year. Until October 2020, most genomes were found to collect around twenty mutations compared to the ancestral strain (Wuhan-Hu-1), suggesting that the virus was constantly changing its genome while replicating and sampling the fitness landscape inside a wide range of immunocompetent humans, its new host. Each time the ancestral sequence is changed, a new variant is born. Few survive while most disappear from the population. The mutations occur randomly distributed across the genome, and the ones that were associated with low-fitness cost or no-cost were tolerated and remained unnoticed. However, a small number of them did give the coronavirus a growth advantage by altering its infection cycle by increasing its infectivity and transmissibility. Importantly, some regions in the genome become less conserved and can tolerate diversity, thus representing the hotspots for generating new variants. In an epidemic, these changes are monitored in a community through genomic surveillance to track the ancestral strain and build the phylogenetic tree in which each persisting branch becomes a lineage. At the start of the COVID-19 pandemic, two clusters comprising strain A (20%) and B (80%) emerged in China, while the progenitor strain A slowly disappeared, strain B became the most dominant and mutated to generate more lineages as it spread outside China (Rambaut et al. 2020). The first report of a non-synonymous mutation leading to an amino acid change from aspartic acid to glycine at position 614 in the most important spike glycoprotein (aka D614G) came in from Scotland in March 2020 (Robertson 2020). This strain was later christened as B.1, was responsible for the biggest outbreak in Lombardy, Italy. The strain B.1 appeared to gain a selective advantage over others in increasing the magnitude of infection by producing more viral load. However, this strain had no impact on the disease severity and went on to produce B.1.1–B.1.1.7 in the United Kingdom (Rambaut et al. 2020). Since then, several national genomic epidemiology efforts have sequenced thousands of isolates and curated a database (GISAID) of the genetic changes associated with major epidemiological events and changes in the virus’s behavior (Elbe and Buckland-Merrett 2017). The GISAID dashboard helped track emerging lineages as they traveled between regions and countries and captured the global pattern of the SARS-CoV-2 genetic diversity (figure 1).

Figure 1
figure 1

Genomic epidemiology of the emerging lineages of SARS-CoV-2: phylogenetic divergence of 3869 strains sequenced between December 2020 and July 2021. Figure created from https://nextstrain.org/ and colored as per clades.

Full size image

3 How many SARS-CoV-2 variants are there?

As the virus spread across the continents, it replicated unfettered and generated as many mutants as possible to produce a large number of variants, which differs from each other by the presence of a set of distinctive mutations in its genome. The Pango lineage database, which analyzes the genome sequences deposited in GISAID, has classified the genomes into 1277 variants or lineages (O'Toole and Pybus 2021). These mutations are spread across the whole genome of the coronavirus, but not all of the changes have contributed in altering the viral physiology. For example, the B.1.525 (Eta) variant first detected from Nigeria reported the maximum number of mutations in the genome but with no proportional changes in the viral behavior (Haseltine 2021). Some mutations have caused a significant change in the shape or number of the viral surface proteins, which have resulted in either increasing the virus’s infectivity and transmissibility or its ability to cause severe disease. In addition, these mutations may cause interference in diagnostics and more importantly, may increase chances of reinfection and bring down vaccine efficacy if the modified viral proteins are able to escape from being recognized by the neutralizing antibody generated from previous infections and vaccines. Several dominant strains have become prevalent by natural section and have given away to newer variants over the course of the pandemic (figure 2; table 1). However, these beneficial mutations have become worrisome for us as they get reported from large cluster outbreaks and raise concerns of the protection provided by the currently available vaccines.

Figure 2
figure 2

Dominance of the B.1.617.2 (Delta) variant: prevalence of SARS-CoV-2 lineages over time in India. Line plot for seven day average and shaded region corresponding to the 95% CI. Created from https://outbreak.info server.

Full size image
Table 1 Effect of the different mutations on transmission and vaccine efficacy in SARS-CoV-2 lineages.
Full size table

The D614G strain (B.1) that emerged in China became the first ’variant of concern’ as its prevalence increased from 10% in March 2020 to 78% of all the sequences analyzed by May (Robertson 2020). This D to G change was always accompanied by three other mutations, including an amino acid change in the RNA-dependent RNA polymerase. When checked for its infectivity, pseudovirus expressing the D614G modified spike protein produced nine fold more infectious titres, justifying the higher magnitude of infection it caused in the real world (Korber et al. 2020). Until September 2020, the SARS-CoV-2 genome accumulated ten mutations, when the B.1.1.7 (Alpha) strain with 23 mutations (8 in the spike protein) was detected in Kent, United Kingdom, and was reported to be 50% more transmissible (Rambaut et al. 2020). However, it did not show any reduction in vaccine efficacy. Like its ancestor B.1 strain, it also produced a high viral load leading to a low Ct value in the RT-PCR-based diagnosis (Collier et al. 2021). It is hypothesized that B.1.1.7 may have evolved in patients with weak immunity who received convalescent plasma therapy containing antibodies, contributing to selecting this variant (figure 3).

Figure 3
figure 3

Mutations in Spike protein S1 subunit have implications in increasing infectivity and reducing vaccine efficacy in COVID-19. Mutations which increase the affinity of the receptor binding domain to the human ACE2 receptor (red) and help escape binding to neutralizing antibody (blue). D614G substitution (orange) stabilizes prefusion conformation and contributes in tighter binding to ACE2. Mutations responsible for increased viral fusion (cyan) by increasing cleavage at the furin cleavage site (yellow). Receptor binding motif at the CTD of the S1 subunit in magenta while all other mutations are depicted in green.

Full size image

Similarly, another lineage of the B.1 variant, B.1.351 (Beta), evolved in South Africa and became a dominant strain in the region with 47% of all the sequences analyzed (Tegally et al. 2021). Its genome showed two additional mutations, which allowed its spike protein to bind tightly to the ACE receptor and at the same time evade binding to the neutralizing antibodies present in convalescent plasma and that were produced by the Astrazeneca vaccines by around a factor of ten (Wang et al. 2021). Monoclonal antibodies formulated by Eli Lilly and one present in the Regeneron cocktail also showed markedly reduced activity against the mutated spike protein, ‘Eek mutation (E484K)’ present in this variant (Annavajhala et al. 2021).

Among all variants of concern, the super-infectious B.1.617, which along with the Alpha variant was responsible for the collapse of the Indian health care system, has now become the most dominant lineage in India and 80 other countries (Yadav et al. 2021). This strain has 13 mutations, including two new in the spike protein (E484Q and P681R), attributed to its 50% higher transmissibility compared to the alpha variant. Since then, three subtypes B.1.617.1/2/3 have emerged and over time, the B.1.617.2 (Delta) have replaced the other two less transmissible lineages to become the dominant variant (Bolze et al. 2021). Despite lacking mutations at position N501 and E484 in its spike protein, this variant has also been found to spread faster within the body and in vitro was found to be less sensitive to the BNT162b2 (BioNTech/Pfizer) vaccine (Collier et al. 2021; Wall et al. 2021). These increased features of ACE2 binding and S1/S2 cleavage have been attributed to the two new mutations at positions L452R and E484Q in the receptor binding domain of the delta variant (Cherian et al. 2021).

While in South America, another variant with 17 mutations became predominant. This strain P.1 (Gamma), also evolved from B.1.1.28 lineage and shared the three important mutations present in the receptor binding domain of the Beta variant that are associated with high transmissibility and ability to re-infect (Faria et al. 2021a). The Gamma variant reached the United States around the same time as another variant B.1.429, which evolved and dominated the state of California before it declined to become a ‘variant of interest’ (WHO 2021b). Several other strains (B.1.526, B1.617.1, P.3), which showed higher affinity for receptor binding and potential for immune escape in vitro due to the presence of the same mutations have been classified as variants of interest but have not been reported to be more contagious. Intriguingly, the B.1.526 variant was able to cause reinfection in New York in the face of active immunity developed from the first 2020 wave. Therefore, it becomes essential to investigate how these mutations are being formed. Are they present in the immune suppressed population, or whether they develop from persistent infections? Monitoring of new variants such as the B.1.617.3 generated in a region and assessing their phenotypic properties must continue in order to evaluate vaccine efficiency.

4 SARS-CoV-2 infection cycle

To understand how each of these genetic changes can impact the spread of COVID-19, it is important to know the steps involved in SARS-CoV-2 infection and pathogenesis. Upon entry through the nasal route, the virus encounters and colonizes the olfactory supporting cells present in the upper part of the nose, which turns out to be its unique host cellular target and determines its niche. The colonization is achieved by adhering to the host cells by latching the viral surface spike glycoprotein to the ACE2 receptors present in high numbers on the surface of the sustentacular cells (Flerlage et al. 2021). A virus, being a non-living entity, is an obligate intracellular pathogen that needs to gain access and hijack the host cellular machinery in order to replicate and, as a collateral benefit, allow them to escape immune surveillance.

The spike protein is trimeric in nature, with the three receptor binding domain (S1 subunit) forming the head, which is rather extensively glycosylated to escape recognition by antibody (Wrapp et al. 2020; Xia et al. 2020). The stalk (S2 subunit) connecting to the viral envelope is very flexible due to the presence of hinges, allowing optimal ACE2 receptor binding and thus contributes to its enhanced infectivity compared to SARS-CoV-1 (Pierri 2020; Turoňová et al. 2020). Post adherence, proteolysis and dissociation of the spike protein is necessary for the fusion of the viral and the host cell membrane. For this step, the virus is entirely dependent on the transmembrane serine protease (TMPRSS2), present on the host surface, to recognize and cleave the spike protein into two subunits at the poly-basic cleavage site between the S1/S2 subunits (Tang et al. 2020). Together with TMPRSS2, the neuropilin-1 protease found on the surface of the olfactory cells may make an alternate cleavage at the S2’ hydrophobic fusion peptide site at the S2 domain (Cantuti-Castelvetri et al. 2020). If the coronavirus takes the endocytic route for entry, the vesicular cathepsin-L makes these cleavages before the membrane fusion can occur (Qiu et al. 2006). Most of the time, the spike protein is preprocessed before new virions are even assembled. This is enabled by the furin protease, present inside the Golgi in most host cell types. Thus, a key factor in SARS-CoV-2 infection is tight recognition of the ACE2 receptor and the protease cleavage of the spike protein. This host-virus interaction also introduces tropism in coronaviruses by increasing the target space among the host cells at multiple organs, expressing the ACE2 receptor and the cell surface protease in abundance.

As the virus fulfills its sole intention, to make as many copies of it, the host-induced innate response gets triggered to reduce the pathogen burden. Sensing the presence of the viral genome, the intracellular signaling mechanism induces the interferon-stimulated genes and cytokine response, eventually leading to a shutdown of the cellular processes, cleavage of host and the viral genome by RNase-L leading to cell death (Birdwell et al. 2016). In addition, cytokine induced inflammatory response stops viral replication and also helps recruit phagocytes. However, the coronaviruses have evolved ways to persist by using their proteins to subvert host defense. Several of the viral non-structural proteins either bind or cleave their genome-derived triggers (dsRNA) or inhibit the host-signaling cascade (Li et al. 2021). This helps to antagonize the induced innate response and allows colonization in novel and normally sterile sites. The abundance of furin and other proteases in several host tissues thus promotes cellular tropism while an exuberant host inflammatory response induces tissue pathology, loss of organ function and facilitates the spread of infection. The ability to generate a high viral load in the nasopharynx and infect the host beyond the upper respiratory tract renders the coronavirus its high transmissibility.

5 Infectiousness

At the host population level, this high infectivity translates to its high infectiousness. Unlike a slowly developing chronic disease of the lower respiratory system such as tuberculosis, SARS-CoV-2 infection transits very rapidly from a susceptible state to recovery and does not persist long within an individual. Furthermore, since the nasal load achieved in the primary site of infection is enough to disseminate through viral shedding, SARS-CoV-2 also does not require optimizing its virulence with its reproduction so as to leave the body before the host dies. This also results in over-dispersion and precludes the virus from maintaining a large critical community size to persist in the population. So, now its infectiousness is determined by its ‘basic reproduction number (R0)’ or the number of susceptible hosts whom an infected host can pass on the disease (Endo et al. 2020). The ancestral Wuhan stain started with an R0 of 2.4–2.6, which increased to 3 and 4 in the European strains B.1 and B.1.1.7, respectively. It must be mentioned that the R0 number is not an intrinsic property of the virus and depends largely on host susceptibility and behavior. As a result, the recent and most successful strains of the Delta variants (B.1.617.2 and AY.1) have managed to achieve an R0 between 5 and 8 in countries with slow vaccination and low compliance to the COVID guidelines (Pueuo 2021).

6 Infectivity and immune escape of spike variants

The S1 subunit of the immunogenic spike protein has been identified as a target for many vaccines (Salvatori et al. 2020; Sternberg and Naujokat 2020). This subunit comprises of N-terminal domain (NTD), C-terminal domain and receptor binding domain (RBD). While the fusogenic S2 subunit contains a fusion peptide, two heptapeptide repeat sequences (HR1, HR2) and remains connected to the viral membrane with the help of a transmembrane domain (Huang et al. 2020). Accessibility of the spike trimer to the ACE2 receptor depends on the presence of one of the RBD in its up-state configuration, one of the three perfusion configurations that the spike protein exists (Wrapp et al. 2020). Cleavage at the furin site then irreversibly changes the conformation of the spike protein to a postfusion configuration, thus promoting membrane fusion. The spike protein can also undergo this transformation in the absence of binding to the ACE receptor, leading to S1 subunit shedding but no fusion (Koenig and Schmidt 2021). The S1 subunit has since accumulated multiple mutations, with NTD and the receptor binding motif being relatively more tolerant. Mutations at both of the regions significantly impacted its sensitivity to both monoclonal and vaccine stimulated neutralizing antibodies (Collier et al. 2021).

6.1 D614G

Since its identification, this mutation has become dominant in the circulating strains and is now present in every ‘variant of concern’. This substitution does not directly affect the receptor binding but allows the usually disordered ‘630 loop’ in the S1 subunit to insert into a wider gap between the NTD and CTD created by the presence of a smaller amino acid glycine. (Zhang et al. 2021) Thereby stabilizing the RBD up-state configuration in the prefusion complex and preventing premature and futile S1 dissociation (Yurkovetskiy et al. 2020). The increased availability of the functional spike protein increases the viral infectivity but has not reported any increase in disease severity thus far.

6.2 N501Y

This mutation is present in the receptor binding motif of RBD. It is observed in all the highly transmissible variants like B.1.1.7, B.1.351, and P.1 (Cerutti et al. 2021; Faria et al. 2021a; Tang et al. 2021). N501 residue present in wild type version interacts with Y41 of ACE2 through a hydrogen bond. The substitution disrupts this interaction but allows the spike protein to have two new hydrogen bonds with ACE2 (Liu et al. 2021). Tighter binding of the RBD to the ACE2 receptor is the main reason for the high infectivity. In addition, the mutation also showed a significant loss of neutralization by monoclonal antibodies with reduced sensitivity to B.1.1.7 variant (Collier et al. 2021).

6.3 ∆H69/∆V70

First identified in the Alpha variant, this deletion has been observed to co-occur with the other receptor binding motif mutations N501Y, N439K and Y453F (Meng et al. 2021). The mutation alone has shown to impart two-fold higher infectivity in pseudotyped lentiviral assays by increasing affinity to ACE2. The deletion has even helped rescue a binding defect observed in mutants formed in vitro and in vivo (D796H) (Kemp et al. 2021; Meng et al. 2021). Since H69-V70 is located in a disordered loop of the NTD near the recognition site of antibodies, the deletion might alter its accessibility affecting the binding of neutralizing antibodies targeting the NTD (Collier et al. 2021).

6.4 E484K

Identified first in the RBD of the B.1.351 variant. This mutation reduces the affinity of spike protein to ACE2 when present alone. E484 of the spike glycoprotein interacts with K31 of ACE2, and disruption of this interaction reduces affinity (Lan et al. 2020). However, when present with N501Y it increases affinity to ACE2 (Cheng et al. 2021). This substitution in the immunodominant epitope causes a substantial loss in binding to neutralizing antibodies induced by the two mRNA vaccines currently in use (Wang et al. 2021; Wibmer et al. 2021) and is the main reason for immune evasion to most of RDB targeting antibodies (Gobeil et al. 2021).

6.5 K417N/T

K417 lies near RBD and like E484 mutation, it reduces affinity to ACE2 when present alone but is compensated in the background of N501Y (Cheng et al. 2021). However, affinity to antibodies is markedly reduced due to the loss of several non-covalent interactions between the aliphatic chain of K417 with monoclonal antibodies isolated from convalescent serum.

6.6 P681H/R

Mutation at this position, near the S1/S2 cleavage site, was first observed in the UK variant. The mutation was predicted to affect the furin mediated cleavage but with no impact on viral entry in the Alpha variant (Lubinski et al. 2021). However, in the new Delta variant of concern, substitution at the same position has been observed but with an R residue (P681R). This additional arginine was found to enhance furin mediated cleavage, followed by enhanced membrane fusion.Surprisingly, the spike protein found inside and on the surface of cells infected with the B.1.617.2 (Delta) variant was predominantly in its cleaved form (Mlcochova et al. 2021). Cleavage of the spike prior to the release of the virion is the primary reason for it increased uptake, faster replication resulting in a thousand fold higher infectivity compared to the original Wuhan strains A/B (Li et al. 2021). Furthermore, like P681H, in vitro cell culture infection P681R variant showed larger syncytia formation, thus hinting at a switch in replication from cell-free infection to cell-cell fusion thus, explaining the cause of its rapid spread in the body (Saito et al. 2021). Thus, mutation at this position (P681H/R) is possibly associated with an increase in the fusogenic potential of the spike protein.

7 Which are the future variants of concern?

Antigenic drift leads to a gradual accumulation of mutations. We witness this each year during the flu season in the case of the influenza virus. Immune resistance mutations at the RBD were observed in the case of the seasonal coronavirus 229E over the years (Wong et al. 2017) and is natural to happen for the new coronavirus as well. So, what is the future of new SARS-CoV-2 variants coming from the past? Is it possible to predict the future mutations that will escape being recognized by the current vaccine induced antibodies or will transmit rapidly in society? Bloom and colleagues have attempted to identify these future changes in the genome by using a yeast display system to produce different versions of the RBD and monitor binding to monoclonal antibodies and convalescent plasma (Greaney et al. 2021b). The process also helps to select mutants by growing different spike variants in the presence of neutralizing antibodies. The escape maps identified few major epitopes or hot spots in the RBD, which often acquire mutations to evade neutralization by antibodies targeting a wide surface, with the E484 residue showing the maximum impact on binding (Greaney et al. 2021a). The immune escape mutations can be recapitulated in the real world by studying the viral evolution as a function of convalescent plasma therapy. In an immunosuppressed individual receiving plasma therapy upon infection by the B.1 strain, a dominant viral genotype with the deletion at H69/V70 was found to emerge together with mutation (D796H) in S2 subunit (Kemp et al. 2021). The same mutations resulted in losing sensitivity to neutralizing monoclonal antibodies and in high infectivity when tested with pseudoviral particles. Encouragingly, treatment with non-competing monoclonal antibody combinations prevented the development of escape variants and delayed the emergence of resistance to antibodies in clinical trials (Copin et al. 2021). Since serum neutralization is an accepted correlate for protection, the impact caused by the individual mutations must be considered for releasing vaccine updates and when choosing a combination of monoclonal antibodies. However, we must also not forget the unchanged protection provided by the non-neutralizing (binding) antibodies and the CD8+ T cells mediated clearance of the host cells infected with SARS-CoV-2 strains (Moore et al. 2021; Tarke et al. 2021).

In conclusion, all the successful management strategies of the pandemic that we have implemented so far have been based on the knowledge of the initial SARS-CoV-2 genome sequence. This information, when collected at scale without geographical inequities, becomes pivotal in understanding how the virus is evolving and how we should change our guard in terms of its impact on accuracy in diagnosis, introducing stringent public health measures, and designing vaccination schedules. Without adequate genomic surveillance, all in vitro estimates of immune escape and new efforts in predicting clade emergence using machine-learning algorithms will remain theory. Until now, we have been able to sequence little more than 1% of the viral genomes from all the cases worldwide, far below the 5% recommended by the World Health Organization. It has since called for accelerated integration of genome sequencing into a regular global health practice to track the new variants and prepare for future threats from pandemics (WHO 2021a). Impactful genomic epidemiology has helped track the evolution of the latest variant of interest C.37 (Lambda), which infected 80% of the unvaccinated population in Peru with the highest mortality rate in the world. A similar effort from the South African surveillance (NGS-SA) identified a new variant C.1.2 closely related to the Lambda variant, but with distinct mutations. In addition to the well known mutations observed in the variants of concern and interest, this variant was reported to accumulate unique mutations at the N terminal domain, within the receptor binding motif and adjacent to the furin cleavage site in the spike protein. We should be concerned and monitor its immune evasion, transmissibility and the disease severity as it spreads beyond the African continents (Scheepers et al. 2021). Fortunately, most of the current vaccines have shown efficacy against all the SARS-CoV-2 variants. There is every reason to believe that they will protect from severe disease and save millions of lives provided we can bring them to the low-income countries. However, with more than 80% of the vaccines made available to 10% of the world’s population, it will not be easy to stop generating new epicenters of variants in the developing world. Vietnam’s failure in controlling the recent outbreak of the delta variant after creating the biggest success story at the beginning of the pandemic is the best example to learn from.