The COVID-19 pandemic, caused by the SARS-CoV-2 coronavirus, is one of the most challenging global health crisis within the last century (Chan et al., 2020). The virus emerged as a result of a zoonotic shift (Lam et al., 2020; Lu et al., 2020). It is a member of the betacoronaviruses family (Cui et al., 2019), related to coronaviruses found in Bats (Zhou et al., 2020), and to SARS CoV that cause severe respiratory syndrome (Drosten et al., 2003) as well as other widely circulating members of the family (HCoV 229E, NL63, OC43, and HKU1) that cause the common cold (Chibo and Birch, 2006; Ren et al., 2015; van der Hoek et al., 2004; Vijgen et al., 2005; Woo et al., 2005).
Coronaviruses (CoVs) have the largest genomes among RNA viruses (Gorbalenya et al., 2006). Nonstructural protein 14 (nsp14), a subunit of the replicase polyprotein encoded by CoVs is thought to provide a form of proofreading activity that could support the expansion of large CoVs genomes to their current size. One result of such proofreading activity is that CoVs genomes are less mutable compared to other RNA viruses (Eckerle et al., 2010; Smith et al., 2013), and thus the sequence diversity of SARS-CoV-2 is quite low (Fauver et al., 2020).
In response to the SARS-CoV-2 pandemic, many approaches for antibody (Ab) therapies, and vaccines are being explored (Krammer, 2020). Almost all vaccination approaches aim to use the glycoproteins or spike protein (S) of the virus in its trimeric form (Yu et al., 2020) or vaccinate with the full (inactivated) virus (Gao et al., 2020). The spike, a class I fusion protein, mediates entry to the host cell by binding to the angiotensin-converting enzyme 2 (ACE2) receptor (Ou et al., 2020) and is the main target of Ab response (Robbiani et al., 2020; Wu et al., 2020). These therapeutic approaches, hopefully, would be able to elicit strong Ab and T cell response against the virus. In particular, Abs against the spike receptor-binding domain (RBD) have been shown to have neutralization and protective capabilities (Robbiani et al., 2020; Wu et al., 2020).
Since SARS-CoV-2 virus introduction into humans is recent, it probably has not yet evolved extensively to acquire escape mutations from the commutative Ab pressure of the human population (Li et al., 2020). One mutation at the spike (D614G) is now widespread and is thought to support a high viral growth rate (Korber et al., 2020). However, other members of the coronavirus family have been circulating in human populations for many years (Hulswit et al., 2016) and evidence of antigenic drift are seen in SARS-CoV-1 (Guan et al., 2003; Song et al., 2005), and among common cold coronaviruses OC43 (Ren et al., 2015; Vijgen et al., 2005), 229E (Chibo and Birch, 2006). Hence, given the prevalence of SARS-CoV-2, to inform vaccine design and understand how the fitness landscape of the virus evolves, it is important to recognize antigenic drift due to Ab pressure if it were to occur. More generally, antigenic drift due to Ab pressure is common in other RNA viruses such as the seasonal influenza virus (Das et al., 2013; Wu and Wilson, 2017).
Here we sought to understand and predict, from first principle, to what extent the mutability of the spikes of influenza and close relatives of SARS-CoV-2 could be attributed to Ab pressure. The magnitude (titers) of Ab response against a given epitope is a direct consequence of the B immunodominance hierarchy patterns of an immunogen, which are the result of various aspects of the humoral response to antigen (Angeletti et al., 2017; Angeletti and Yewdell, 2018; Mahanty et al., 2015). Amongst them is the B cell repertoire – the number of B cell clones targeting different epitopes (Abbott et al., 2018; Amitai et al., 2020; Krammer et al., 2018b; McGuire et al., 2014; Nabel and Fauci, 2010; Sangesland et al., 2019), their germline affinity (Amitai et al., 2020; Dosenovic et al., 2018), and T cell help to B cell (Tan et al., 2019). Here, we concentrate on the geometric presentation of the spike to Abs. We have previously shown using coarse-grained molecular dynamics simulations, that the geometry of the immunogen spike presentation on the virus recapitulates the known immunodominance of hemagglutinin (HA) head compared to its stem (Amitai et al., 2020).
We developed here an in-silico approach to estimate the Ab targeting - a proxy for B cell immunogenicity (Amitai et al., 2020), of residues on the spike surface, and the differential accessibility to antigenic epitopes due to the geometrical presentation of spikes on the surface of the virus. Superimposed on the spike surface, the immunogenicity score gives the Ab affinity maps of influenza and corona spikes, which we applied to predict how the antigenic space is explored unevenly across the surface of these glycoproteins. We then used sequences from public repositories (www.ncbi.nlm.nih.gov, www.gisaid.org) to evaluate the mutability maps of those glycoproteins. Next, we developed a computational approach based on spectral clustering to compare these maps. We found that about 50% of the mutability maps variability of the S protein of the severe acute respiratory syndrome-related betacoronavirus (sarbecovirus), and 67% of the variability in the mutability of the seasonal influenza spike (HA) can be attributed Ab pressure, as estimated from the model. This suggests average, polyclonal Ab pressure was consequential in the diversification of the coronavirus sarbecovirus spike and the seasonal flu spike. Moreover, our data suggest that the geometry of spike presentation on the viral surface is a major factor determining its mutability.
We further studied the time evolution of SARS-CoV-2 spike mutability up to October 1st, 2020. While the overall correlation between our model and the mutability map is still very low, we find that it has been gradually increasing over time. While very preliminary, it could suggest that some variants with Ab escape mutations are establishing in the population. Overall, our approach allows us to recognize from first principle, based on the 3D structure of glycoprotein and cryo-EM images of the viral surface, whether their mutational landscape has features suggesting Ab evasion, and rank surface residues according to their likelihood to acquire Ab-escaping mutations in the future. Importantly, this approach can detect early signs of SARS-CoV-2 and influenza adaptation to evade immune pressure by memory B cells.