SARS-CoV-2 Spike Protein Induces Hemagglutination: Implications for COVID-19 Morbidities and Therapeutics and for Vaccine Adverse Effects

Experimental findings for SARS-CoV-2 related to the glycan biochemistry of coronaviruses indicate that attachments from spike protein to glycoconjugates on the surfaces of red blood cells (RBCs), other blood cells and endothelial cells are key to the infectivity and morbidity of COVID-19. To provide further insight into these glycan attachments and their potential clinical relevance, the classic hemagglutination (HA) assay was applied using spike protein from the Wuhan, Alpha, Delta and Omicron B.1.1.529 lineages of SARS-CoV-2 mixed with human RBCs. The electrostatic potential of the central region of spike protein from these four lineages was studied through molecular modeling simulations. Inhibition of spike protein-induced HA was tested using the macrocyclic lactone ivermectin (IVM), which is indicated to bind strongly to SARS-CoV-2 spike protein glycan sites. The results of these experiments were, first, that spike protein from these four lineages of SARS-CoV-2 induced HA. Omicron induced HA at a significantly lower threshold concentration of spike protein than the three prior lineages and was much more electropositive on its central spike protein region. IVM blocked HA when added to RBCs prior to spike protein and reversed HA when added afterward. These results validate and extend prior findings on the role of glycan bindings of viral spike protein in COVID-19. They furthermore suggest therapeutic options using competitive glycan-binding agents such as IVM and may help elucidate rare serious adverse effects (AEs) associated with COVID-19 mRNA vaccines, which use spike protein as the generated antigen.


Introduction
Key to the infectivity and morbidity of SARS-CoV-2 are glycans that protrude tangentially from 22 N-linked glycosylation sites on each monomer of its spike protein [1][2][3][4][5]. These N-glycans, several of which are capped with terminal sialic acid (SA) moieties, sweep back and forth across spike protein like windshield wipers, partially shielding it from antibody binding [1,2,6,7]. For SARS-CoV-2 and several other coronaviruses, these N-glycans serve as appendages for the virus to make its initial attachments to glycoconjugates on the host cell surface [1,[8][9][10][11][12][13]. These glycoconjugates are abundantly distributed on the surfaces of host cells such as red blood cells (RBCs) [14,15], platelets [16] and endothelial cells [17] and confer an associated negative electrostatic potential, additionally facilitating attachments by positively charged SARS-CoV-2 spike protein prior to viral fusion to ACE2 for replication [8,[18][19][20]. For endothelial cells of blood vessel linings, for example, the disparity between 28,000 SA-tipped CD147 receptors and 175 ACE2 receptors per cell [21] electrostatic potential of the spike protein of these four variants were studied using molecular modeling and were related to their HA-inducing potencies as experimentally observed.

Tests for Hemagglutination (HA) and for Its Inhibition and Reversal by IVM
In the HA experiment, for the Wuhan, Alpha and Delta lineages, we observed HA at a spike protein concentration of 1.06 ng/µ L and above, but not below. For the Omicron spike protein, we observed HA at a minimum concentration of 0.13 ng/µ L and above, but not below (Figures 1 and 2, Table 1).   electrostatic potential of the spike protein of these four variants were studied using molecular modeling and were related to their HA-inducing potencies as experimentally observed.

Tests for Hemagglutination (HA) and for Its Inhibition and Reversal by IVM
In the HA experiment, for the Wuhan, Alpha and Delta lineages, we observed HA at a spike protein concentration of 1.06 ng/µ L and above, but not below. For the Omicron spike protein, we observed HA at a minimum concentration of 0.13 ng/µ L and above, but not below (Figures 1 and 2, Table 1).    Table 1. In the HA inhibition experiment, IVM added to 2.5% RBC solution to attain a concentration of 1 µM at 30 min prior to spike protein partially inhibited HA, with HA observed at a spike protein concentration of 2.12 ng/µL for the Wuhan, Alpha and Delta, and of 0.27 ng/µL for the Omicron viral lineages. With IVM at 2 µM, complete inhibition of HA is observed for the Wuhan, Alpha and Delta lineages. For Omicron, IVM at 4 µM is needed to totally block HA.
In the HA reversal experiment, concentrations of IVM needed to reverse HA at the highest concentration of spike tested were 1 µM and 2 µM for the Wuhan/Alpha and Delta/Omicron viral lineages respectively.
HA was not observed with cell culture supernatants of any SARS-CoV-2 strain. In the control experiments, RBCs alone did not exhibit HA. RBCs mixed with phosphatebuffered saline (PBS) likewise did not exhibit HA. The addition of IVM to 50 µL of 2.5% RBCs to attain a concentration of 8 µM did not cause hemolysis or induce HA. A solution of 2.5% DMSO and 97.5% water, the solvent for IVM, did not block or reverse HA (Figures 2 and 3). inhibition of HA by IVM (not pictured), with differences described in Section 2.1 and summarized in Table 1. In the HA inhibition experiment, IVM added to 2.5% RBC solution to attain a concentration of 1 µ M at 30 min prior to spike protein partially inhibited HA, with HA observed at a spike protein concentration of 2.12 ng/µ L for the Wuhan, Alpha and Delta, and of 0.27 ng/µ L for the Omicron viral lineages. With IVM at 2 µ M, complete inhibition of HA is observed for the Wuhan, Alpha and Delta lineages. For Omicron, IVM at 4 µ M is needed to totally block HA.
In the HA reversal experiment, concentrations of IVM needed to reverse HA at the highest concentration of spike tested were 1 µ M and 2 µ M for the Wuhan/Alpha and Delta/Omicron viral lineages respectively.
HA was not observed with cell culture supernatants of any SARS-CoV-2 strain.
In the control experiments, RBCs alone did not exhibit HA. RBCs mixed with phosphate-buffered saline (PBS) likewise did not exhibit HA. The addition of IVM to 50 µ L of 2.5% RBCs to attain a concentration of 8 µ M did not cause hemolysis or induce HA. A solution of 2.5% DMSO and 97.5% water, the solvent for IVM, did not block or reverse HA (Figures 2 and 3).

Western Blot and Quantification Analysis
The results for quantification analysis by Western blot of spike protein of the Wuhan SARS-CoV-2 strain in cell culture supernatants are presented in Figure 4. This shows that its spike concentration is below the concentration thresholds for the induction of agglutination. Indeed, for the supernatant of the Wuhan strain in cell culture at 48 h post viral infection, the concentration of N-glycosylated spike protein is approximately 0.7 ng/µ L, which dilutes to half that concentration when added to wells, approximately three

Western Blot and Quantification Analysis
The results for quantification analysis by Western blot of spike protein of the Wuhan SARS-CoV-2 strain in cell culture supernatants are presented in Figure 4. This shows that its spike concentration is below the concentration thresholds for the induction of agglutination. Indeed, for the supernatant of the Wuhan strain in cell culture at 48 h post viral infection, the concentration of N-glycosylated spike protein is approximately 0.7 ng/µL, which dilutes to half that concentration when added to wells, approximately three times lower than the minimum concentration 1.06 ng/µL of recombinant spike found to induce HA in the HA experiment described above.

Molecular Modeling
Since the surface of RBCs is electronegative due to a high expression of anionic SA in membrane proteins and gangliosides, we used molecular modeling simulations to study the electrostatic surface potential of the spike proteins used in the present study. We focused our attention on the central area of the spike trimers, which is formed by the receptor binding domain (RBD) of each monomer. The electrostatic potential of the central area of the spike trimers increased exponentially from the Wuhan initial lineage to the Omicron variant ( Figure 5, lower panel). This was caused by a progressive decrease in electronegative zones (colored in red) and a concomitant increase in electropositivity (blue zones). The increase was modest between Wuhan and Alpha, larger for Delta, and reached very high levels for Omicron. As depicted in the upper panel of Figure 5, these variations in net positive electric charge between the four SARS-CoV-2 strains considered here have implications for the induction of HA, since electropositivity mitigates the electrostatic repulsion between negatively charged RBC surfaces.
times lower than the minimum concentration 1.06 ng/µ L of recombinant spike found to induce HA in the HA experiment described above.

Molecular Modeling
Since the surface of RBCs is electronegative due to a high expression of anionic SA in membrane proteins and gangliosides, we used molecular modeling simulations to study the electrostatic surface potential of the spike proteins used in the present study. We focused our attention on the central area of the spike trimers, which is formed by the receptor binding domain (RBD) of each monomer. The electrostatic potential of the central area of the spike trimers increased exponentially from the Wuhan initial lineage to the Omicron variant ( Figure 5, lower panel). This was caused by a progressive decrease in electronegative zones (colored in red) and a concomitant increase in electropositivity (blue zones). The increase was modest between Wuhan and Alpha, larger for Delta, and reached very high levels for Omicron. As depicted in the upper panel of Figure 5, these variations in net positive electric charge between the four SARS-CoV-2 strains considered here have implications for the induction of HA, since electropositivity mitigates the electrostatic repulsion between negatively charged RBC surfaces.

Discussion
For all four SARS-CoV-2 lineages tested, Wuhan, Alpha, Delta and Omicron BA.1, spike protein mixed with human RBCs induced HA. This result provides an in vitro counterpart to the clumps of stacked RBCs (rouleaux) found, for example, in the blood of most [26,27] or a third [28] of COVID-19 patients in three clinical studies, and reinforces indications that such blood cell aggregation is key to the morbidities of this disease. HAassociated vascular obstruction in COVID-19 was further demonstrated in zebrafish embryos [47], which have capillary diameters [48] and blood cell glycosylation patterns

Discussion
For all four SARS-CoV-2 lineages tested, Wuhan, Alpha, Delta and Omicron BA.1, spike protein mixed with human RBCs induced HA. This result provides an in vitro counterpart to the clumps of stacked RBCs (rouleaux) found, for example, in the blood of most [26,27] or a third [28] of COVID-19 patients in three clinical studies, and reinforces indications that such blood cell aggregation is key to the morbidities of this disease. HA-associated vascular obstruction in COVID-19 was further demonstrated in zebrafish embryos [47], which have capillary diameters [48] and blood cell glycosylation patterns [49] similar to those of humans. SARS-CoV-2 spike protein injected into a zebrafish embryo vein caused the formation of small RBC clumps and an associated reduction in blood flow velocity within 3-5 min after injection [47]. Also, in various in vivo or in vitro studies, SARS-CoV-2 spike protein S1 was found to cause endothelial, pulmonary and neuronal damage, as well as platelet-thrombi formation and microclots [40,[50][51][52].
In addition to the glycan bindings reviewed above, electrostatic attraction could promote attachments between RBCs and SARS-CoV-2 spike protein and contribute to the observed induction of HA. As indicated here by molecular modeling simulations, the SARS-CoV-2 spike protein central region has a net positive charge, which yields an attractive force to the negatively charged RBC surface and also mitigates the repulsive electrostatic force between RBCs. This result is consistent with prior determinations of a net positive electrostatic potential of the SARS-CoV-2 spike protein [19,53]. Regardless of whether glycan bindings or electrostatic attraction is the predominant underlying force, the HA observed here suggests, more generally, that spike protein attachments to other cells such as platelets and endothelial cells may be implicated in COVID-19 morbidities, these two cells, like RBCs, having dense surface distributions of SA-tipped glycans [1] and corresponding negative surface charges [16,17]. These attachments can manifest clinically in damage to endothelial cells induced by virions or free spike protein, marked by traces of spike protein [23,35,51], and in the formation of fibrin-hardened clots that incorporate platelets, neutrophils and other blood cells [54], potentially seeded by RBC clumps.
The much stronger HA-inducing effect of the spike protein of the SARS-CoV-2 Omicron variant vs. the prior lineages tested here, with thresholds of concentration for HA induction of 0.13 ng/µL for Omicron vs. 1.06 ng/µL for the Wuhan, Alpha and Delta lineages, was an intriguing result. Under the hypothesis that RBC clumping underlies the morbidities of COVID-19, especially those related to diminished efficiency of RBC oxygenation, this finding appears paradoxical, since percentages of patients having respiratory distress or requirements for oxygen or mechanical ventilation were all sharply less for Omicron as compared with prior viral lineages [55]. Yet diminished pulmonary morbidities of Omicron can be explained by its less efficient replication compared with prior SARS-CoV-2 lineages in the lung parenchyma, encompassing the alveolar epithelium, in contrast to Omicron's faster replication in the bronchi [56,57]. The disruption of the alveolar-capillary barrier caused by a COVID-19 infection is a prime route by which the virus enters the bloodstream [58], so limited replication of Omicron in alveolar tissue would limit systemic serum viral loads and potential for associated RBC clumping as well.
Among the attributes of the Omicron variant that are of interest for further study is that although its overall electrostatic potential is more positive than for prior lineages, its Nterminal domain (NTD) is less positive [53,59,60]. Consistent with the greater HA-inducing activity of Omicron, however, its SA-binding affinity as predicted by molecular modeling is greater than for prior lineages [61]. It has also been proposed, again based upon in silico analysis, that decreased binding affinity of mutated residues on the Omicron spike protein to a cellular receptor implicated in thromboembolic and neurological complications of COVID-19, α7nAChR, could also account for decreased morbidity of this viral variant [62].
IVM, an electrically neutral molecule [63], has been found in silico to attach with a strong affinity to 10 glycan binding sites on the spike protein of the Alpha through Delta variants [43], which suggests that competitive inhibition of spike protein attachments to host cell glycoconjugates could be its means of HA inhibition. IVM at concentrations of 1-2 µM for the four different viral lineages inhibited HA when added concurrently with spike protein and also reversed HA that had been induced by the prior addition of spike protein. This HA-reversal effect could account for sharp increases observed within 24 h after administration of IVM of pre-treatment depressed SpO2 levels in severe COVID-19 patients, as summarized in Figure 6 below, which reproduces a figure from the most recent of three clinical studies reporting this effect [64]. In contrast to the sharp, rapid increases in SpO2 levels observed in these studies, moderate and severe COVID-19 patients under standard care typically manifest decreasing SpO2 values in tandem with increasing pulmonary CT abnormalities from the day of onset of disease symptoms through the second week following, as established in several studies that tracked SpO2 values, pulmonary abnormalities, or both [65][66][67][68][69][70][71].
The inhibition of HA by IVM as reported here parallels the prevention of RBC clumping in zebrafish embryos by heparan sulfate, a glycosaminoglycan likewise indicated to strongly bind to SARS-CoV-2 spike protein [49,72], when co-injected with spike protein [49]. It is noteworthy that the scattered RBC clumps observed in this zebrafish embryo study, as with those observed in the blood of COVID-19 patients [26][27][28], are much smaller than the macroscopic-scale HA-clusters of extensively interlaced RBCs-observed in this study. It is therefore likely that smaller concentrations of spike protein and IVM would be required to, respectively, induce and reverse RBC clumps of clinical relevance. Thus, the peak plasma level of IVM of approximately 412 nM, as attained about four hours after a standard oral dose of 200-350 µg/kg [1], appears to be in a range that could achieve clinical effects analogous to HA reversal observed in this study at IVM concentrations of 1-2 µM.
The HA-inducing activity of SARS-CoV-2 spike protein, which is especially potent for Omicron, raises questions as to potential risks for COVID-19 mRNA vaccines, which use spike protein as the generated antigen, even though serious adverse effects (AEs) linked to spike protein, such as myocarditis [73][74][75], are rare. Detectable levels of SARS-CoV-2 spike protein and S1 in serum or plasma have been found to persist for as long as 50 days following such vaccinations [76][77][78]. The possibility that spike protein migrating into the bloodstream could in rare cases prompt such HA-associated AEs is suggested, for example, by a study of 1006 subjects experiencing AEs after receiving a Pfizer/BioNTech or Moderna mRNA vaccination, which found a significant degree of RBC aggregation in the blood of 948 of those subjects [79]. These risks may be increased for younger age groups, with 301 adolescents of 13-18 years of age who received two doses of the BNT162b2 mRNA COVID-19 vaccine in one study having a 29.2% rate of cardiac AEs, ranging from tachycardia or palpitation to myopericarditis [80]. The investigators considered chest pain, which occurred at a 4% incidence, "an alarming side effect," however, myopericarditis cases were mostly mild and temporary. The possibility that the increased HA-inducing activity of the Omicron spike protein might result in increased HA-associated AEs from Omicron booster vaccines can be ruled out by close vaccine safety monitoring to follow up on the animal testing on these booster vaccines performed to date [81].
Additional experiments of interest for follow-up to this study include microscopic detection to check for the initial formation of RBC clumping at spike protein concentrations lower than those which induce HA. Also, the zebrafish study described above [47] could be replicated using IVM instead of heparan sulfate as the blocking agent. Finally, HA could be tested as in this experiment but using RBCs supplemented with human serum albumin (HSA) at a physiological concentration. If IVM were to bind to spike protein glycans at the same molecular region as that which binds to HSA, that could significantly limit its HA-inhibitory effect, since 93% of IVM binds to serum proteins, mainly HSA, in blood [82], and 93% of IVM would then be rendered inactive for this effect. Conversely, if, IVM were to bind to spike protein glycans and HSA each at different regions, then HSA, a large molecule (molecular mass of 66.5 kDa, vs. 875.1 Da for IVM), could considerably boost the HA-inhibitory effect of IVM through steric interference.

Source and Preparation of Red Blood Cells
Red blood cells (RBCs) were provided by the "Establishment Français du Sang" (EFS) from blood bag donors qualified as "non-therapeutic blood bag" (Convention N°7828). Twenty milliliters of whole blood (with anticoagulant EDTA) were added to a fiftymilliliter conical tube and filled with thirty milliliters of PBS at pH = 7.2 before centrifugation at 800× g for 10 min. Then, the supernatant was discarded and replaced with fresh PBS. This procedure was again repeated twice. After the last centrifugation, RBCs were diluted in PBS at a final concentration of 2.5%. RBCs were then stocked for one week at 4 °C.

Spike Proteins Preparation
Recombinant spike proteins (BioServUK, Sheffield, UK) of these four following SARS-CoV-2 strains were used for this experiment: Wuhan, Alpha, Delta and Omicron BA.1 (BSV-COV-PR-33, BSV-COV-PR-65, BSV-COV-PR-97 and BSV-COV-OM-0.1, respectively). Four concentrations of spike protein were prepared for each of these viruses by diluting stock suspension in PBS. Spike protein of each viral strain was dissolved in

Source and Preparation of Red Blood Cells
Red blood cells (RBCs) were provided by the "Establishment Français du Sang" (EFS) from blood bag donors qualified as "non-therapeutic blood bag" (Convention N • 7828). Twenty milliliters of whole blood (with anticoagulant EDTA) were added to a fifty-milliliter conical tube and filled with thirty milliliters of PBS at pH = 7.2 before centrifugation at 800× g for 10 min. Then, the supernatant was discarded and replaced with fresh PBS. This procedure was again repeated twice. After the last centrifugation, RBCs were diluted in PBS at a final concentration of 2.5%. RBCs were then stocked for one week at 4 • C.

Spike Proteins Preparation
Recombinant spike proteins (BioServUK, Sheffield, UK) of these four following SARS-CoV-2 strains were used for this experiment: Wuhan, Alpha, Delta and Omicron BA.1 (BSV-COV-PR-33, BSV-COV-PR-65, BSV-COV-PR-97 and BSV-COV-OM-0.1, respectively). Four concentrations of spike protein were prepared for each of these viruses by diluting stock suspension in PBS. Spike protein of each viral strain was dissolved in PBS and added to wells at final increasing concentrations of 0.27 ng/µL, 0.53 ng/µL, 1.06 ng/µL and 2.12 ng/µL for the Wuhan, Alpha and Delta lineages, and of 0.07 ng/µL, 0.13 ng/µL, 0.27 ng/µL and 0.53 ng/µL for the Omicron variant.

Cells and SARS-CoV-2 Strains Preparation
Vero E6 cells (ATCC-CRL-1586) were cultured as in previously described conditions [87][88][89] in medium (MEM, Gibco, Gaithersburg, MD, USA) with 2 mM L-glutamine and 10% fetal bovine serum (FBS) at 37 • C in a 5% CO 2 incubator. We infected Vero E6 with four viral strains genotyped by whole genome next-generation sequencing (NGS) as belonging to: Pangolin lineage B.1.1 [90] (the first major lineage following the Wuhan genotype that circulated during the first epidemic period in France, designated as "Wuhan") and

Western Blot and Quantification Analysis
Supernatants from Wuhan SARS CoV-2 infected cells prepared as described above or uninfected cells and commercial spike were lysed with 2× Laemmli Sample Buffer (#1610737, Bio-Rad, Hercules, CA, USA) with DTT (#EU0006-B, Euromedex, Souffelweyersheim, France) added as reducing agent and then heated at 95 • C for 5 min. Proteins were separated by 10% SDS-polyacrylamide gel electrophoresis (Laemmli, 1970) and Westernblotted on a nitrocellulose membrane. After 1 h of saturation in 5% nonfat dry milk with 0.3% Tween-20 in PBS, the membrane was incubated overnight with SARS/SARS-CoV-2 Coronavirus Spike Protein (subunit 1) rabbit polyclonal antibody (Thermo Fisher Scientific, Illkirch-Graffenstaden, France) at a dilution of 1:1000 in the same buffer as for saturation. After this first incubation, the membrane was washed for 10 min three times in PBS 1X-Tween buffer and then incubated for 1 h at room temperature with peroxidase-labeled anti-rabbit donkey antibody (#NIF 824 ECL Rabbit IgG, HRP-linked whole Ab, Sigma-Aldrich Life Science, Burlington, MA, USA) diluted in saturation buffer at 1:1000. After this second incubation, the membrane was rinsed for 10 min three times in PBS 1X-Tween buffer before ECL (Western Blotting Substrate, # W1001 Promega, Madison, WI, USA) revelation by image acquisition with the Fusion Fx chemiluminescence imaging system (Vilber Lourmat, Marne-la Vallée, France). Quantification in each well was calculated by measuring band intensities using ImageQuant TL Version 7.0 analysis software (GE Healthcare, Buc, Yvelines, France). Protein markers (New England Biolabs, #P7719S) were used for molecular mass determination.

IVM Preparation
IVM was supplied by Sigma Aldrich (St Quentin Fallavier, France). Stock solution was diluted in 2.5% of DMSO and 97.5% of water. IVM, 20 µL in volume, was added in designated wells to reach final concentrations of 1, 2, 4 and 8 µM, as specified.

Tests for Hemagglutination (HA) and for Its Inhibition and Reversal by IVM
Three sets of experiments were performed to test for HA induced by SARS-CoV-2 spike protein and then for HA inhibition and reversal by IVM. To test for HA, using a 96 micro-well plate, 50 µL of 2.5% RBCs in PBS was added to wells together with 62 µL of diluted spike proteins at specified concentrations. An additional 20 µL of PBS was added to attain the same total fluid volume as used in the IVM inhibition and reversal experiments. This mixture was let sit for 30 min at room temperature under gentle agitation. Then, the plate was tilted for at least 30 s, after which, if HA had not occurred a teardrop could be observed at the bottom of the well consisting of settled RBCs (Figure 1). This teardrop was not observed if HA had occurred, i.e., if a network of linked, agglutinated RBCs had formed, as described previously [91]. To test for inhibition of HA by IVM, 50 µL of RBCs were mixed with 20 µL of IVM at specified concentrations ranging from 1-8 µM and let sit for 30 min at room temperature under gentle agitation. Then 62 µL of spike protein was added at specified concentrations, wells were let sit for an additional 30 min at room temperature under gentle agitation, and HA was determined as above. Finally, to test for reversal of HA by IVM, 50 µL of RBCs were mixed with 62 µL of spike proteins for 30 min to determine HA as above. Then 20 µL of IVM were added at specified concentrations, wells were let sit for an additional 30 min at room temperature under gentle agitation and then the plate was tilted at least 30 s and wells were rechecked for HA as described above.
The following control experiments were performed. Fifty microliters of RBCs were deposited alone in the wells to check their absence of agglutination. Twenty microliters of PBS were added to fifty microliters of RBCs to verify the absence of HA. The potential for induction of HA by IVM was tested by adding it at the highest concentration used (8 µM) to 50 µL of 2.5% RBCs. In order to test whether DMSO blocked or reversed HA, we also performed the HA inhibition and reversal experiments described above using the solvent for IVM, 2.5% DMSO and 97.5% water, but without IVM.
The HA experiment was then done with viral suspensions for each viral strain by adding to the wells the same volume of viral supernatant as for commercial spike suspensions.
Each experiment was done in triplicate.

Molecular Modeling Simulations
A complete structure of the reference spike protein was generated from the original 20B strain (pdb: 7bnm) as previously described [92]. All gaps in the pdb file were fixed by inserting the missing amino acids with the protein structure prediction service Robetta [https://robetta.bakerlab.org/, accessed on 3 November 2022] [53,93]. This source file model was used to introduce the specific mutational profiles of the indicated Alpha, Delta and Omicron variants with the MUTATE tool of Swiss-PdbViewer [53,94]. Trimeric structures in the closed pre-fusion conformation were constructed with Swiss-PdbViewer by homology with a reference model (pdb: 6VSB). All structures were then submitted to several rounds of energy minimization with the Polak-Robière algorithm [88]. The electrostatic surface potential of the spike trimers was analyzed with Molegro and quantified with ImageJ software as described previously [92].

Conclusions
Spike protein from four lineages of SARS-CoV-2 induced HA in human RBCs, which supports other indications that spike protein-induced RBC clumping, as well as viral attachments to other blood cells and endothelial cells, may be key to the morbidities of COVID-19. IVM, a macrocyclic lactone indicated to bind strongly to multiple glycan sites on SARS-CoV-2 spike protein, blocked HA when added to RBCs prior to spike protein and reversed HA when added afterward, which suggests therapeutic options for COVID-19 treatment using this drug or other competitive glycan-binding agents. The Omicron B.1.1.529 variant had significantly greater HA-inducing activity than the three prior lineages tested, which may relate to the findings from molecular modeling that the electrostatic charge of the central region of its spike protein was considerably more positive than for those of the prior lineages and, from a prior study, that its SA-binding affinity is greater than for prior lineages [61]. Whether increased risks of rare HA-associated serious AEs for Omicron booster vaccines might be correspondingly increased as compared with those for legacy mRNA COVID vaccines can be ruled out by close vaccine safety monitoring to follow up on the animal testing on these booster vaccines performed to date.