The SARS-CoV-2 furin cleavage site: natural selection or smoking gun?

SARS-CoV-2 has continued to evolve as it moves through the population, and we have learned more about its pathogenicity and transmission determinants. As previously commented in The Lancet Microbe, one of these determinants is the unusual furin cleavage site (FCS) on its spike protein. ¹ While it has been proposed that the FCS might have been engineered, it is becoming clearer that natural selection is, in fact, the driving factor in its acquisition and functionality, through recombination and epistasis.

Central to this concept is the highly dynamic molecular nature of the SARS-CoV-2 spike, which is synthesised as a trimer with its three receptor (ACE2)-binding domains initially in a closed conformation. When an FCS is present, the spike encounters the intracellular protease furin on its way out of the cell, which processes and separates the S1 and S2 subdomains, imparting a relaxation in the protein and allowing one or more receptor-binding domains to flip up—the open conformation—markedly increasing receptor engagement and driving the transmissibility of the virus.² The spike thus reversibly samples its open-trimer conformations and the FCS promotes this, but not exclusively.³ However, this sampling comes at some cost; if there are too many “up” domains the spike becomes unstable, which poses a problem as the virus can now shed S1 and lose transmissibility. The virus therefore exists in a transmission window, but its FCS has functionally stayed in place—as for Leigh Van Valen's so-called Red Queen hypothesis—through sequence changes elsewhere in the spike. For SARS-CoV-2, the best known of these changes was D614G, which arose and became embedded within lineage B just a few months after the initial outbreak. 614G markedly increases infectivity and is a “gateway mutation” upon which all the specific FCS changes in the variants of concern were built.⁴ Lineage A viruses seemed to have solved the problem slightly differently (Q613H)—eg, with the A.23.1 variant⁵—and other gateway mutations have followed suit (H655Y).⁴ Eight of the latest omicron variants have all three (Q613H, D614G, and H655Y).⁶ These gateway mutations cooperate epistatically to modulate the problematic FCS.

The FCS problem is well illustrated by the rapidity with which it is often lost upon virus adaptation to cell culture, both through point mutations and deletions.⁷ In reality, the FCS is not just an on-off switch but is highly regulated, and has in fact been incrementally optimising itself through mutations at the FCS in the major variants of concern—alpha, delta, and omicron—allowing the virus to stay within its transmission window (although in reality easing forward—ie, not quite the “same place” as with the Red Queen).

SARS-CoV-2 is not unique in this regard and many other betacoronaviruses have an FCS that is highly adaptable. Laboratory strains of HCoV-OC43 have point mutations in the FCS, presumably to also gain stability, and as the coronaviruses causing seasonal endemic infections are further explored, there are clear examples of genetic changes that structurally position the flexible FCS loop to better engage its furin activator. This is demonstrated in a genotype I virus with a four amino acid downstream insertion;⁸ whether this is truly a pathogenesis determinant or a transmission determinant remains to be seen. Recent work on HCoV-HKU1 has also shown the highly dynamic nature of the open-closed conformation of its spike, but in this case with its natural FCS removed for protein expression and the conformational changes regulated by sialoglycan binding.⁹

The spike is adaptable and the FCS clearly makes a difference, but—in the end—it is no smoking gun.

We declare no competing interests.