A new review paper by Gerlach et al. (2025) systematically inventories ten structural, functional, and genomic peculiarities of the SARS-CoV-2 spike protein – and arrives at a remarkable finding: not a single one of the eight protein-related pathobiological features has been lost over the course of variant evolution. Five of them are even reinforced in late Omicron variants.
The paper deliberately adopts a source-agnostic stance: it does not assess how these features arose, but rather what they are, how well they are documented, and whether they are still present.
Features at the protein level
1. Furin cleavage site (FCS)
The SARS-CoV-2 spike protein possesses a polybasic furin cleavage site (RRAR motif at positions 681–685) that is absent in the closest known related coronaviruses. Furin is a ubiquitously present cellular enzyme that cleaves spike already during biosynthesis. This generates free S1 subunits that can circulate systemically – and only thereby activate several of the pathobiological properties described below.
The RRAR core motif is absolutely conserved among all SARS-CoV-2 lineages sequenced to date. In Omicron variants, mutations in the immediate vicinity even lead to increased cleavability.
2. Amyloidogenic epitopes
Nyström and Hammarström (JACS 2022) identified seven peptide sequences within the spike protein that can form amyloid fibrils in vitro. Particularly relevant: the epitope Spike685 – located immediately after the furin cleavage site – induces fibrinogen amyloidization into fibrinolysis-resistant fibrin structures. Blood clots formed in the presence of Spike685 amyloid resist degradation by plasmin far more strongly than normal fibrin clots.
All seven epitopes are located in conserved regions of the spike protein and are preserved in late Omicron variants. The amyloidogenicity of the receptor-binding domain (RBD) is even experimentally increased in Omicron.
3. Superantigen-like motif and neurotoxin-homologous sequences
The sequence Y674–R685 of the spike protein shows similarity to staphylococcal enterotoxin B (SEB), a known superantigen. This motif has been discussed as a possible contributor to polyclonal T-cell activation in severe COVID-19 and MIS-C.
Independently of this, the receptor-binding region (S375–L390) contains a second region with structural homology to neurotoxins of various snake genera (cobra, krait). Molecular binding analyses suggest a possible interaction with nicotinic acetylcholine receptors, although direct functional evidence is still pending.
Both regions are fully conserved in current variants.
4. Prion-like domains (computationally determined)
The receptor-binding region contains asparagine/glutamine-rich sequence stretches (Q474, N481, Q493, Q498, N501) that are algorithmically classified as prion-like domains. SARS-CoV-2 is the only known coronavirus in which such domains are located directly in the receptor-binding region.
Recent binding assays show that the Omicron BA.5 RBD binds α-synuclein – the protein central to Parkinson's disease – with roughly twice the affinity of the wild-type RBD. The clinical significance of this observation is still unclear and requires further research.
5. Extended S1/S2 loop and Neuropilin-1 binding
The QTQTN motif (positions 675–679) is a five-amino-acid extension that distinguishes SARS-CoV-2 from other sarbecoviruses. After furin cleavage, a so-called CendR motif arises that binds Neuropilin-1 (NRP1) – a protein strongly expressed in particular in olfactory neurons and endothelial cells. This is a discussed factor in COVID-related loss of smell and vascular pathology.
The QTQTN motif is absolutely conserved in all variants to date.
6. N-glycosylation and DC-SIGN interaction
The spike protein carries 22 N-linked glycosylation sites that can function as ligands for DC-SIGN (CD209) and related C-type lectin receptors on antigen-presenting cells. Remarkably, all 22 sites are fully preserved despite more than 30 amino-acid mutations in the Omicron RBD. At the same time, the glycan composition has shifted toward oligomannose-rich structures – the preferred DC-SIGN ligand.
7. Optimized ACE2 binding affinity
Already upon its first detection at the end of 2019, SARS-CoV-2 displayed a 10–20-fold higher binding affinity for the human ACE2 receptor than SARS-CoV-1. This affinity has been preserved across all variants and has been further increased in late Omicron subvariants (e.g., KP.3 with F456L+Q493E).
8. Bacteriophage-like behavior and toxin-like peptides (preliminary)
A series of studies from an Italian research group as well as the Joint Research Centre of the EU Commission describe two unusual observations:
- SARS-CoV-2 material is said to be able to enter certain gut bacteria, where bacterial ribosomes translate viral RNA into spike protein.
- In blood, urine, and stool samples from COVID-19 patients, toxin-like peptides were detected that exhibit structural similarity to animal venoms (including alpha-conotoxin-like peptides, phospholipase A2, short neurotoxin).
The JRC tested spike protein and these peptides on human iPSC-based 3D neuronal models and observed dose-dependent effects on electrical activity and on the expression of developmentally relevant genes.
The authors explicitly emphasize: the bacteriophage-like behavior so far stems from a single laboratory and must be independently replicated. The toxin-like peptides were detected by two groups but likewise remain preliminary.
Features at the genomic level
9. CGG-CGG codon doublet
The two arginine positions 682–683 in the furin cleavage motif are encoded by a CGG-CGG codon pair. CGG is the rarest arginine codon in coronaviruses (approx. 3% usage), but at the same time the most common arginine codon in human genes. This codon pair is invariantly preserved in all SARS-CoV-2 isolates of all variants sequenced to date.
10. BsaI/BsmBI restriction endonuclease pattern
The SARS-CoV-2 genome contains a pattern of BsaI and BsmBI cut sites that, upon complete digestion, generates six nearly evenly distributed fragments. The authors present this finding as a genomic observation without naming an origin hypothesis. In Omicron variants this pattern is altered by mutations; it is therefore wild-type-specific.
Overarching finding: conservation of the pathobiological features
The paper summarizes the findings in a conservation matrix. The central result: no protein-related pathobiological feature has been lost over the course of variant evolution. Of the eight features at the protein level, five are reinforced in late Omicron variants and three are preserved unchanged.
The authors explain this structurally: immune-evasion mutations concentrate on the receptor-binding domain – the region that neutralizing antibodies attack. The pathobiological features, by contrast, lie in structurally highly conserved regions that are indispensable for viral fitness: furin cleavage site, glycosylation sites, extended loop.
A particular consequence of Omicron evolution: the increased furin cleavability together with reduced membrane fusion means that more free S1 fragments reach the systemic circulation – even though the acute lung disease runs a milder course. These free S1 fragments carry the amyloidogenic, superantigen-like, and neuropilin-binding properties within them.
Implications for mRNA-vaccine-encoded spike
The paper points out that mRNA vaccines encode the wild-type Wuhan spike sequence (with two stabilizing mutations, but an unchanged furin cleavage site). Features 1–7 are therefore also present in the vaccine-encoded spike. The authors emphasize that exposure context and dose differ substantially between infection and vaccination, but consider it relevant to take this overlap into account in mechanistic analyses.
Contextualization
The paper explicitly grades the evidence by strength:
- Strong evidence (features 1–3, 5–7): multiple independent laboratories, animal models, structural biology, clinical correlation
- Moderate evidence (feature 4): in-silico prediction with experimental confirmation, no in-vivo evidence
- Preliminary evidence (feature 8): bacteriophage behavior from one laboratory, toxin peptides from two groups
- Genomic observations (features 9–10): documented findings without interpretive consensus
