To systematically analyze the occurrence of analogous polyfunctional sequence domains across a comprehensive set of relevant virus species, we turned to the curated ‘betacoronavirus spike glycoprotein’ collection of overlapping homologous superfamilies (InterPro entry IPR042578). This collection represented the entire genus Betacoronavirus through 2,465 evolutionarily diverse spike protein sequences across 34 virus species. We then extracted all predicted FCS motifs (see, Methods) within a constant frame of twenty amino acid residues (Gu, 2020) which were subsequently automatically filtered for pat7 NLS motifs. After removing sequence fragments and duplicates, this procedure resulted in a set of twenty representative sequences (Table 1; number 1-19 and 21) to which two related betacoronavirus sequences (Table 1; 20, 22) were manually added as pat7 negatives. This analysis thus produced only two betacoronavirus positive hits outside of the SARS-CoV-2 clade: one human merbecovirus and one human embecovirus (Table 1; 19 and 21).
To test the sensitivity of this outcome on the size of the sequence search space, the output was also independently verified through NCBI BLASTP searches, across all 10,766 betacoronavirus protein sequences outside of the SARS-CoV-2 clade in that database. This number was an order of magnitude larger than the 1,179 betacoronavirus InterPro/UniProt sequences of that kind used above. This test confirmed the non-random and spike S1/S2 specific pat7/FCS motif design (see, Methods and Supplemental Material, Data File 1) as no other spike pat7/FCS sequence motif representatives were found than those already given in Table 1, numbers 19 and 21. Nuclear localization signals were tested with the standard PSORT prediction algorithm (Nakai and Horton, 1999), which unequivocally detected a pat7 NLS in SARS-CoV-2 spike S1/S2 sequences, including the original reference sequence from Wuhan Hu-1, in the cognate sequences of MERSMA30 CoV as well as in the human coronavirus HKU1 (Table 1). In addition, the presence of the adjacent The/Ser O-glycosite residue pair was homology inferred within the SARS-CoV-2 clade, and for other sequences individually checked with the standard prediction software NetOGlyc4.0 (Steentoft et al., 2013), where positive hits were recorded (Table 1) only for MERS betacoronavirus first isolate EMC/2012 and for MERS betacoronavirus isolate MA30 (also denoted as MERSMA30).
While most resulting virus consensus sequences (Table 1; 1-18) corresponded to within-clade SARS-CoV-2 spike variants that tightly preserved the entire polyfunctional TXXPRRXRSX consensus motif, the two additional betacoronavirus sequences detected (Table 1; 19, 21) were from the spike of human embecovirus HKU1 (HCoV-HKU1, with genotype B, GenBank ABD75545); and from a murine adapted MERS merbecovirus isolate MA30 (GenBank MT576585). The human CoV HKU1 spike sequence presented the canonical FCS motif RRKR embedded into a complete pat7 motif PSSRRKR; however, there was no Thr/Ser O-glycosite pair at or near the expected flanking positions, and therefore the sequence represented not a full functional analog of the corresponding SARS-CoV-2 domain. By contrast, the MERSMA30 sequence comprised the entire polyfunctional domain, 744TLTPRRVRSV753 , with robust O-glycosite predictions for Thr744 and for Ser752. Also, unlike SARS-CoV-2 and MERSMA30 CoV, in the HKU1sequence any FCS dependent proteolytic cut would be outside of the pat7 NLS, due to a double amino acid downstream shift of the FCS sequence location, and so the two motifs did not effectively interfere in HKU1 CoV S primary structure. In a further difference, while any loss of the RXXR FCS would also abrogate pat7 NLS in the SARS-CoV-2/MERSMA30 consensus motif, PRRXRSX, in HKU1 CoV S such strict motif dependency does not exist. These data indicated that within a broad and representative set of betcoronaviruses S sequences MERSMA30 spike was the only instance of a complete pat7/FCS/O-glycosite motif analogous to the S1/S2 polyfunctional sequence domain of SARS-CoV-2 spike (Wuhan Hu-1 genotype).
The corresponding TXXPRRXRSX consensus motif between the spikes of SARS-CoV-2 and MERSMA30, which overall share a level of 32.7% in protein sequence identity, was not a product of convergent evolution. This was evident from two distinct observations:
First, the closest natural parental strain to MERSMA30 was the MERS first isolate EMC/2012 (Genbank NC_019843), which lacks the pat7 spike motif (744TLTPRSVRSV753; see Table 1). Only during adaptation and rational selection after serial passage in artificially humanized mice the pat7 precursor sequence PRSVRSV changed into a full pat7 NLS through a non-synonymous mutation (Ser749Arg) in MERS S (Li et al., 2017). MERSMA30 S pat7/FCS is therefore a product of directed evolution/adaptation and rational selection in bioengineered host cells (Figure 1A).
Second, the closest genomic relative to SARS-CoV-2, the bat coronavirus RaTG13, is devoid of the entire polyfunctional domain (678TNS----RSV683; see Table 1) but already includes the fixed and highly conserved residue positions 678/682 and 681, which are functionally inactive O-glycosite and FCS precursors, respectively. If, on this conditional (Thr678/Ser682 and Arg681) background, natural selection first proceeds to a complete FCS and then progresses into a full pat7, then this FCS would produce a selective disadvantage for pat7 because FCS proteolytic cleavage would abrogate the pat7 phenotype. Conversely, a full pat7 motif without FCS would produce a selective disadvantage for FCS, because S protein cellular localization shifts to the nucleus, i.e. away from the normal sites of furin expression and activity (see, UniProt entry P09958).
This evolutionary incompatibility argument is supported by three independent lines of evidence: (a) in MERS CoV, continued serial passage of MERSMA30 in the humanized (hDPP4 knock-in) murine model lead to an abrogation of pat7 NLS through a P747H mutation (Li et al., 2017), which demonstrated that the pat7/FCS MA30 genotype was not fixed during adaptation and ultimately was selected against and, more broadly, through the absolute lack of evidence of any naturally occurring MERS CoV with a pat7/FCS spike (Table 1); (b) in the betacoronavirus HKU1 spike sequence, which does not present the corresponding glycosites (Table 1), any cleavage induced by the canonical RRKR↓ motif is not incompatible with pat7 as such proteolytic cut (↓) would occur, in contrast to MERSMA30 CoV and SARS-CoV-2, after the last residue of the pat7 motif; (c) for SARS-CoV-2 and in further analogy with MERSMA30 CoV, all globally dominant SARS-CoV-2 variants (in chronological order: Alpha, Delta, Omicron, and their later recombinants) that had emerged after the original 2019 Wuhan Hu-1 isolate included spike P681(H/R) mutations which ablate pat7 NLS (Figure 1B). Specifically, the presence and absence of the P681(H/R) substitution was, among all characteristic spike mutations within the pandemic variants of concern Alpha, Beta, Gamma, Delta, Omicron, the only one which was strictly correlated with global dominance of the Alpha, Delta and Omicron, and with the lack thereof for the marginal Beta and Gamma variants (Figure 1B), respectively. Moreover, based on an extensive analysis of 9,310,717 SARS-CoV-2 genomic sequences collected worldwide between January 2020 and November 2023 (Figure 1C), the earliest time point of dominance (prevalence >50%) of a P681(H/R) spike variant coincided with the earliest time point of emergence of the first globally dominant variant of concern Alpha, followed by Delta, and Omicron. In a negative control, all other amino acid mutations within the pat7/FCS motif remained marginal with prevalence below 0.5% (Figure 1C). Thus the P681(H/R) genotype was an accurate indicator of both pat7 loss and of a selective sweep from globally emerging SARS-CoV-2 variants. Of note, such specific substitution of the characteristic pat7 Pro residue is known to be associated with a functional loss of nuclear translocation in analogous pat7 NLS (Korb et al., 2013). This overall analysis suggests that on this genetic background a combined pat7/FCS spike S1/S2 motif is directly selected against and that its recent emergence as a SARS-CoV-2 Wuhan Hu-1 genotype during the course of natural betacoronavirus evolution alone was unlikely.