Discussion
Recent prior bioinformatics analyses have not evaluated the SARS-CoV-2 viral spike in isolation from the rest of the viral genome. The viral spike protein is relatively invariant, when compared to influenza or HIV, across worldwide strains, with essentially only one missense variant so far in the viral spike at amino acid position 614 (D/G), despite global spread. We here report that the G614 variant correlates strongly with case fatality rate in a global survey of countries, suggesting that a higher CFR possibly emerging in the USA East Coast or other regions affected later in the pandemic may be at least partly a consequence of this virus-autonomous factor.
Molecular modeling suggests that G614 destabilizes the infectious form of the viral spike protein, thereby favoring the ground-state, unliganded, “down” SARS-CoV-2 viral forms that would be expected to infect less readily, due to masking of the host receptor binding site. This suggests that the mechanism by which G614 viruses causes greater fatality is immunologic rather than virologic: namely, that the form that binds the receptor less well is also better shielded from host immune system attack and/or elicits harmful anti-viral-spike antibodies or other harmful immune responses. Indeed, there are several precedents in other viral diseases: the viral spikes of both HIV and RSV exhibit unliganded states that do not elicit protective antibodies via vaccination, and indeed, may elicit antibodies that enhance disease6,7. Furthermore, disease-enhancing, immunodominant, antibody-targeted epitopes in the SARS viral spike have been definitively identified8, and a recent report emphasized that key cryptic antibody-targeted neutralization epitopes in the SARS family of viral spikes are accessible only in the conformation in which the ACE-2 binding site is exposed9.
There are several potential limitations and/or confounding issues in the data we used to calculate country-by-country CFR. First, the data relies heavily on diagnostic testing and fatality reporting, which have both been highly variable in penetrance and accuracy across the countries studies, at present. For example, China recently revised their fatality count upward and the United Kingdom is an outlier with an unusually high CFR, and when included, the correlation weakens (Figure S1 ). This outlier may be caused by the unusually low incidence of active cases being reported in the UK during the 2 weeks under study. Selection bias is unlikely to be a confounder in this analysis, however, since the same reporting bias would have to occur in the same direction and for the same reason across all of the global regions, which is unlikely. Social distancing measures have been implemented in a highly variable fashion across the countries studied in this report, but the case fatality rate should be unaffected by social distancing, as it does not affect the clinical course of viral disease, only acquisition of virus. Measuring this correlation at this early time point in the pandemic may be optimal to avoid the confounder of successful medical treatment to artificially reduce the CFR.
Based on our results, short-term epidemiologic models of the pandemic that attempt to predict the incidence and prevalence of severe disease and fatalities over time that do not take the D614 percentage of viruses in a region into account may be inaccurate. The presence of G614 in a virus detected in a patient may also have prognostic significance.