SARS-CoV-2 spike protein-associated sialoglycoconjugates induce nanoscale filipodia to facilitate micro-size platelet clotting

SARS-CoV-2, spike protein, platelet, blood clotting, aggregation, thrombosis, 12 glycan, sialic acids, mannose, galactose, open canalicular system, 3D volume electron 13 microscopy, focused ion beam scanning electron microscopy (FIB-SEM) 14 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 2

15 COVID-19 disease is associated with thrombosis, but the pathogenic mechanism remains 16 unclear. Here, we investigate how SARS-CoV-2 spike protein causes platelet activation and 17 aggregation. Our three-dimensional ultrastructural analyses showed that invaginated platelet 18 structures, open canalicular system (OCS), expanded upon activation, trapping viral particles 19 in the process. Binding with platelet OCS concealed SAR-CoV-2 spike-coated particles from 20 virion detection in platelet-depleted blood plasma. Both SARS-CoV-2 spike coated-particles 21 and recombinant spikes specifically induced platelet aggregation with nanoscale filipodia 22 extensions, with the terminal sialic acids of the SARS-CoV-2 spike protein-associated 23 sialoglycoconjugates being the key determinant in platelet activation. Our work illustrates 24 that virus-associated sialic acids, not proteins, are functionally responsible for SARS-CoV-2 25 induced thrombotic events, providing a mechanistic insight on how glycosylation contributes 26 to disease severity in COVID-19. This study lays the foundation for the development of 27 glycan-modified vaccines with reduced risks of thrombosis. 28 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 3

29 Coagulopathy-associated microvascular thrombosis and vascular dysfunction are hallmarks 30 of hospitalised patients with severe acute respiratory syndrome coronavirus (SARS-CoV-2) 31 disease 1-3, in which high incidence of thromboembolic events is found in autopsy of 32 deceased coronavirus disease 2019 (COVID-19) patients 4. As SARS-CoV-2 is more difficult 33 to detect in blood than in nasopharyngeal swabs 5, platelet-associated haemostatic 34 abnormalities in COVID-19 are thought to be a secondary consequence of inflammation or 35 cytokine storm from SAR-CoV-2 infection 6-9. SARS-CoV-2 effectively infect endothelial 36 cells 1, while platelets amplify the endotheliopathy of SARS-CoV-2 thereafter 10. Despite the 37 benefit of SARS-CoV-2 vaccination outweighing the diseases-associated risks 11-13, vaccine-38 related thrombotic events occur rarely at 0.21 cases per 1 million COVID-19 vaccinated 39 person-days 14. Amongst these thrombotic events, a sub-group of this thrombotic disorder is 40 known as vaccine-induced immune thrombotic thrombocytopenia (VITT) that is associated 41 with pathogenic anti-platelet factor-4 (anti-PF4) antibodies 15-17. These anti-PF4 antibodies 42 are elicited through adenoviral vector associated adenoviral core protein in conjunction with 43 somatic hypermutation with immunoglobulin light-chain allele IGL V3-21*02 or *03 found in 44 selected human population 18. Independent from VITT, a low incidence of venous (VTE) and 45 arterial (ATE) thrombotic events have also been reported from SARS-CoV-2 mRNA 46 vaccination 14, implicating a second mechanism plus a direct role of SARS-CoV-2 spike 47 protein to induce thrombotic events. Subsequent study further shows a direct correlation 48 between elevated plasma SARS-CoV-2 spike protein and mRNA vaccination-associated 49 myocarditis in a cohort of otherwise healthy young males 19. 50 51 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 4 One of the unique features of platelets is its surface-connected open canalicular system (OCS) 52 that supports passive uptake of sub-micron size particles in platelets 20,21, a feature that is 53 distinct from active phagocytic engulfment capability in neutrophils and macrophages. 54 Although previous studies have shown virion particles can been seen with platelets using 55 thin-section transmission electron microscopy (TEM) 22,23, it is unknown whether these virion 56 particles are actively engulfed into platelets or passively trapped within OCS. 57 58 Published evidence has demonstrated that there are linkages between ABO blood types and 59 COVID-19 infection (including risk of coronary artery disease) 24,25. It is perhaps less 60 appreciated that the determinants of blood type are based on surface glycans, which modulate 61 interaction with erythrocytes via sialoglycoconjugates 26. In contrast with the high-mannose 62 enriched glycan shield in HIV surface proteins 27 that moonlights as molecular Velcro to 63 facilitate infection 28, the SARS-CoV-2 spike proteins are enriched with sialoglycoconjugates 64 across many of the 22 N-linked glycosylation sites per protein monomer 29. The high density 65 of sialic acid termini of SARS-CoV-2 spike resembles the abundant sialoglycoconjugate on 66 human V on Willebrand factor (VWF) 30,31 from its 12 N-glycan sites and 10 O-glycan sites 67 per protein monomer. VWF acts as a haemostatic agent during platelet activation 32. 68 Lowering the levels of VWF sialylated glycoconjugates represses VWF function and plasma 69 half-life 30, but it is not clear whether sialic acids associated with theSARS-CoV-2 spike 70 protein contribute to the pathogenetic functional impacts of SARS-CoV-2. A recent genome-71 wide association study (GWAS) has shown a nucleotide polymorphism of β-galactoside α-72 2,6-sialyltransferase (ST6GAL1), encoding an enzyme that adds sialic acid to glycoproteins, 73 is strongly associated with SARS-CoV-2 infection, but the mechanistic contribution of 74 ST6GAL1 to SARS-CoV-2 infection and pathogenesis remains an enigma 33. 75 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 5

76 SARS-CoV-2 spike protein-coated particles are trapped and remodelled invaginated 77 platelet structures 78 As platelet activation has been linked to SARS-CoV-2 pathogenesis 22,34, we examined 79 whether viruses could activate platelets to reorganise OCS structures (Fig. 1A-E). 80 Quantitative ultrastructural analysis showed a significant increase with not only the size of 81 platelets (Fig. 1C) but also the OCS-to-platelet area ratio, following SARS-CoV-2 spike 82 coated particles activation (Fig. 1D). This activation was directly linked to an elongation of 83 platelet OCS shape, with a 41.7 % and 26.8 % increase in both long- and short-axis, 84 respectively (Fig. 1E), reducing platelet circularity from 0.34 to 0.26 (p<0.05, Fig. S1) as a 85 consequence. Our data demonstrated that SARS-CoV-2 spike protein coated particles 86 induced substantial membrane network remodelling that is consistent with cytoskeletal 87 rearrangement and membrane extension. 88 89 Earlier studies showed that SARS-CoV-2 particles 22, as well as other viruses 23,35, can be 90 identified within internal platelet structures using 2-dimensional thin section TEM, but it was 91 unclear whether these particles were actively engulfed into internal vacuoles or passively 92 captured via platelet OCSs. We employed a correlative light and electron microscopy (CLEM) 93 workflow to track mCherry-tagged SARS-CoV-2 spike protein coated particles within 94 platelets. A schematic overview of the experimental pipeline is shown in Fig. 1F-I. Platelets 95 mixed with fluorescent-tagged virion particles were fixed onto an alphanumerically labelled 96 EM gridded support (Fig. 1F). Spatial locations of platelets were first identified with low 97 magnification confocal microscopy imaging (Fig. 1G), which was followed by selecting a 98 few candidate virion-associated platelets (Fig. 1H). The chosen regions were subjected to 99 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 6 focused ion beam scanning electron microscopy (FIB–SEM) milling (Fig. 1I) in accordance 100 with an alphanumeric fiducial coordinates positioning. Serial FIB–SEM 20 nm think slice 101 imaging was performed to collate a volumetric view of platelet ultrastructure (Fig. 1J), 102 thereby generating three-dimensional (3D) datasets spanning >100 sequential sections 103 covering over 2 μ m depth in slicing direction (Fig. 1K-L). Reconstruction and segmentation 104 of these datasets revealed the presence of viral particles trapped within the 3D space of 105 platelet OCS. Orthogonal views (XY , XZ, YZ) extracted from the reconstructed volume (Fig. 106 1L) confirmed that these particles were localised within the 3D architecture of the 107 invaginated OCS structures (Fig. 1M, Fig. S2-4 and Video S1). This volumetric validation 108 provided direct spatial evidence of viral sequestration within an invaginated membrane 109 platelet OCS system. Importantly, our 3D reconstruction of platelet aggregates demonstrated 110 that viral particles were trapped within activated platelet aggregates (Fig. 1L). It has been 111 suggested that SARS-CoV-2 spike protein engages with integrin for platelet activation 36. 112 Luciferase reporter virus-bound platelet assay showed SARS-CoV-2 spike coated particles 113 strongly associated with pre-chemically-fixed platelets upon mixing (Fig. S5). The ability of 114 inactivated (chemically-fixed) platelets to capture particles is consistent with the passive 115 covercyte mechanism 20 of OCS (Fig. S5). Trypsin-induced uncoupling of virion-platelet 116 association illustrated viral particles were in a solvent accessible environment outside of the 117 plasma membrane, including both the outer surface and the spatial cavities in OCS (Fig. S5). 118 The ability of virion particles to bind with platelets reduces the sensitivity of virus detection 119 in blood, particularly given blood plasma (platelet-depleted) is the gold standard in sampling 120 blood-associated viral pathogens. 121 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 7 122 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 8 Figure 1 . SARS-CoV-2 spike-protein coated particles induced remodelling of platelet 123 ultrastructure and spike protein-coated particles were sequestered into the open canalicular 124 system (OCS). Thin-section transmission electron microscopy (TEM) images of (A) resting- 125 and (B) activated-platelets were shown, in which OCS were highlighted with blue arrows. 126 Quantitative comparison showed different 2D areas of resting- (light orange) and activated- 127 (dark orange) platelets (C, p<0.01). OCS area to platelet area ratio (D, p<0.0001) and OCS 128 dimension comparisons (E, long- [p<0.01] and short-axis [p<0.0001]) between resting- (light 129 blue) and activated (dark blue) platelet were shown. (F) Schematic diagram shows transfer of 130 gridded coverslip alphanumeric coordinates from confocal imaging to the resin block for 131 targeted EM. Correlation of (G-H) confocal microscopy image and (I) focused ion beam–132 scanning electron microscopy (FIB–SEM) images coordinates using the alphanumeric 133 fiducial, ‘7G’ was highlighted for visualisation purposes. (H) is the zoom-in section with 134 higher magnification for fluorescent image (targeted platelets highlighted in green square 135 corresponding to L), while (I) was the same section post-FIB milling. (J) Serial FIB–SEM 20 136 nm slice image acquisition and (K) volumetric datasets spanning >100 sequential sections of 137 platelet for volumetric ultrastructure analyses (L) 3D reconstruction of platelet aggregates. 138 Green rectangle in (L) highlighted the section of platelet underwent the orthogonal slice view 139 extraction in (M). (M) The orthogonal slice views from three vantage points (XY , XZ, YZ) 140 confirm localisation of spike-coated particles within the three-dimensional architecture of the 141 platelet OCS (highlighted with blue circumference and arrow). 142 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 9 SARS CoV-2 spike protein-coated particles drive platelet association and aggregation 143 Our 2D TEM data (Fig 1A-B) and 3D FIBS-SEM (Fig. 1K-M) showed specifics on the 144 ultrastructural relationship between platelets and SARS-CoV-2 spike protein-coated particles. 145 To evaluate the population-relationship between platelets and virion particles, confocal 146 imaging showed that SARS-CoV-2 spike protein-coated particles substantially increased: (i) 147 platelet-platelet interactions; and (ii) fluorescent-tagged virion–platelet association (Fig. 2A-148 B), when SARS-CoV-2 spike protein-coated particles were present. The significant increase 149 of platelet-associated fluorescence intensity with SARS-CoV-2 spike protein-coated particles 150 (Fig. 2C, p<0.05, N=6) suggested a role for the SARS-CoV-2 spike glycoprotein in the 151 process. Qualitative differences in SEM determined platelet morphology were observed upon 152 incubation with virion particles (Fig. 2D-H). Platelets incubated with SARS-CoV-2 spike 153 protein-coated particles displayed visible aggregation (Fig. 2F-H), while untreated- (Fig. 2D) 154 and naked (SARS-CoV-2 spike protein-free) particles treated- (Fig.2E) platelets remained 155 largely dispersed. Platelets from both untreated- (Fig. 2D) and naked particles treated- 156 (Fig.2E) samples exhibited minimal filopodia extension (Fig. 2D-E, Fig. S6). In contrast, 157 platelets exposed to virion particles coated with either ancestral (Wuhan)- or Omicron- spike 158 protein resulted in pronounced platelet aggregation (Fig. 2F-G), which were characterised by 159 the presence of multiple platelets clustering and extensive nanoscale filopodia protrusions 160 with heterogeneous morphology (Fig. 2H). To quantify the aggregation effects, platelet 161 populations were classified based on the aggregated area sizes using SEM datasets from 162 multiple donors (N = 9), incorporating 1825+ events per experimental condition (n=1825 to 163 2491). Platelet aggregates were binned into 4 size categories (i) 0–50 μ m2; (ii) 50–100 μ m2; 164 (iii) 100–150 μ m2; and (iv) >150 μ m2 (Fig. 2K), equating to the joining of (i) 1-9; (ii) 9-17; 165 (iii) 17-25, and (iv) >25 individual platelet ‘particles’, respectively. The majority (99.9%) of 166 platelet population in both untreated- and naked particles-treated samples were confined to 50 167 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 10 μ m2 (or <9 platelets), with the remaining 0.1% platelet aggregates assumed the size of 50–168 100 μ m2 (Fig. 2K). In contrast, treatments of platelets with either ancestral- or Omicron-spike 169 protein-coated particles led to the presence of large platelet aggregates exceeding 100 μ m2 170 (Fig. 2K), which accounted for 0.1% or more of the platelet populations. The potential of 171 Omicron spike protein-coated particles as a stronger inducer of platelet aggregates over 172 ancestral (Wuhan) spike protein-coated particles requires further analysis (Fig. 2K). The 173 ‘average size of platelet aggregates’ and the ‘fractional area occupancy of platelet aggregates 174 within the platelet population’ were analysed by pooling samples derived from 9 donors, in 175 which ~4650 platelets aggregates were detected in the platelet population across all 176 conditions (Fig. 2L-M). We defined platelet aggregate as a connected area consisting of two 177 or more platelets. Our data showed that the mean platelet aggregate area significantly 178 increased from 9 / 11 μ m2 (for no- / naked-particles- treatment, Fig. 2L) to 13.5 / 14.5 μm2 179 (ancestral- / Omicro-spike protein-coated particles treatment, Fig. 2L, p<0.01). No mean area 180 difference was observed between no- and naked-particles- treated platelet aggregates (Fig. 181 2L). There were over a 60% increase of ‘fractional areas occupied by platelet aggregates 182 within the platelet population’ upon treatments with ancestral- (Fig. 2M, from 40% to 67%, 183 p<0.01) or Omicron- (Fig. 2M, from 40% to 65% p<0.05) spike protein-coated particles. 184 These data showed that virion-associated SARS-CoV-2 spike proteins activated platelets by 185 increasing both the size and the frequency of platelet aggregation (Fig. 2K-M), in which the 186 SARS-CoV-2 spike protein was a major contributor to platelet activation.187 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 11 188 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 12 Figure 2 . SARS-CoV-2 spike protein-coated particles promote platelet–platelet interaction 189 and aggregation. Confocal imaging shows association of mCherry-labelled virion particles 190 with platelets following incubation with (A) spike protein-deficient particles and (B) 191 Omicron spike protein-coated particles. (C) Quantification of fluorescence intensity 192 associated with platelets in the presence of spike protein-deficient and spike protein-coated 193 particles (N=6). (D) SEM image of platelets (no particle), (E) SEM image of platelets 194 incubated with spike protein-deficient particles (F) SEM image of platelets incubated with 195 ancestral spike protein-coated particles (F) SEM image of platelets incubated with omicron 196 spike coated particles (H) Higher resolution SEM images showing morphology of aggregated 197 platelets and filipodia (K) Size-distribution analysis of platelet aggregates across donors (N = 198 9; >1800 events per condition) (L) Statistical comparison of mean aggregate area and (M) 199 statistical comparison of fractional area occupancy of platelet aggregates within the platelet 200 population. Capital ‘N’ is for Donor numbers and little ‘n’ is for number of events. Scale bars 201 sizes are listed. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. 202 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 13 SARS CoV-2 spike protein alone induces dose-dependent platelet aggregation 203 To determine whether the spike protein alone was sufficient to drive platelet activation, 204 platelets were exposed to increasing concentrations (from 0.25 μg/mL to 25 μg/mL) of 205 purified recombinant SARS-CoV-2 spike protein for morphological analyses with SEM. Data 206 collected from platelets (across from 8 donors) consistently showed dose dependent 207 recombinant spike protein-induced platelet aggregations (Fig.3A), illustrating the specificity 208 of SARS-CoV-2 spike protein to promote platelet activations. The observed recombinant 209 spike protein-induced morphological changes in platelets (Fig.3A, S6) were indistinguishable 210 from those seen with platelet treatment using intact SARS-CoV-2 spike protein-coated 211 particles (Fig. 2D-G). These observations suggest that the SARS-CoV-2 associated platelet 212 aggregation capacity was primarily driven by the viral spike protein. Population analyses 213 showed 0.25 μg/mL of recombinant SARS-CoV-2 spike was sufficient to increase the amount 214 of 50-100 μm2 size platelet aggregates 5-fold (from 0.04% to 0.8%, Fig. 3B), while up to 25-215 fold increase of platelet aggregates (or 2% of the platelet population) exceeding 50 μm2 were 216 detected with 25 μ g/mL recombinant spike protein stimulation (Fig. 3B). Roughly 0.1% of 217 platelets reached 150 μ m2 or larger in size upon 25 μg/mL recombinant spike protein 218 treatment (Fig. 3B), which was associated with an increase to 15 μm2 (p<0.01) mean platelet 219 aggregate area from 8 μm2 in the untreated control (Fig. 3C). The fractional area occupancy 220 of platelet aggregate within the platelet population had an increase from 50% in untreated 221 control to roughly 65%, 70%, and 75% (Fig. 3D) upon treatments with 0.25 μg/mL, 2.5 222 μg/mL and 25 μ g/mL, respectively, of the recombinant spike proteins (Fig. 3D). Our results 223 demonstrated that platelet activation and aggregation were driven by direct molecular 224 interactions with the SARS-CoV-2 spike protein that was independent from intact viral 225 particles. Our findings are consistent with previous studies showing SARS-CoV-2 can 226 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 14 directly interact with and activate platelets, thereby contributing to thrombo-inflammatory 227 responses observed in COVID-19 6,22,34. Furthermore, our observation provided a scientific 228 rationale to account for the molecular mechanism on how otherwise healthy male subjects 229 experienced myocarditis from mRNA-based COVID vaccination in the absence of SARS-230 CoV-2 infection 19.231 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 15 232 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 16 Figure 3. SARS-CoV-2 spike protein alone induces dose-dependent platelet aggregation. (A) 233 Representative SEM images of platelets under control conditions and following incubation 234 with increasing concentrations of purified ancestral SARS-CoV-2 spike protein (0.25, 2.5, 25 235 μ g/mL). (B) Size distribution of platelet aggregates classified into bins (0–50, 50–100, 100–236 150, >150 μ m²). Statistical analysis of (C) mean aggregate area and (D) fractional area 237 occupancy of platelet aggregate within the platelet population. Capital ‘N’ is for Donor 238 numbers and little ‘n’ is for number of events. Scale bars sizes are listed. *p<0.05, **p<0.01, 239 ***p<0.001, ****p<0.0001. 240 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 17 Spike protein terminal sialylation drives platelet aggregation 241 N-linked glycosylation of proteins can be manipulated by inactivation of host cell 242 glycosylation enzymes during protein synthesis or exogenous glycosidases in purified 243 recombinant proteins (See schematic in Fig. S7). To identify molecular determinants within 244 the SARS-CoV-2 spike proteins that were responsible for platelet activation, we leveraged the 245 N-acetylglucosaminyl transferase I defective (GnTI /i2 ) cells line to produce glycosylation-246 modified SARS-CoV-2 spike protein coated-particles 28. The SARS-CoV-2 spike protein 247 produced in GnTI /i2 cells contained high-mannose N-linked glycans across all 66 N-glycan 248 sites per each of the SARS-CoV-2 trimer (Fig. 4A, S7), in complete contrast to the sialic acid-249 rich complex glycans found on authentic SARS-CoV-2 spike protein produced in the control 250 cell line (Fig. 4B). Platelets incubated with Omicron spike protein-coated particles containing 251 high mannose glycans remained largely as individual ‘units’ with minimal aggregation (Fig. 252 4A). In contrast, platelets exposed to sialic acid-enriched spike protein-coated particles 253 formed many platelet-platelet interactions that were associated with elevated levels of 254 fluorescent signal (Fig. 4B). Quantitation of particle-associated fluorescence across platelets 255 from 6 donors between two types of SARS-CoV-2 glycan spike protein-coated particles were 256 not significance (Fig. 4C). Similarly, employing an in vitro luciferase-reporter virus-platelet 257 binding assay with platelets derived from 3 donors did not show a significant difference in 258 the capacity of these two types of SARS-CoV-2 spike protein glycans to interact with 259 platelets. The qualitative differences platelet clustering between these two types of glycan-260 decorated SARS-CoV-2 spike protein-coated particle suggests a role of glycans in platelet 261 activation (Fig. 4A-B). SEM analyses showed platelets incubated with high-mannose SARS-262 CoV-2 spike protein-coated particles exhibited background to low-level of detectable 263 activation (Fig. 4E, S8) in comparison to a no particle incubation control (Fig. 4D, S8). In 264 contrast, sialic acid-containing glycosylated SARS-CoV-2 spike protein-coated particles 265 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 18 promoted platelet activation resulting in more frequent nano-size filopodia protrusions (Fig. 266 4F). Population analyses on the sizes of platelet aggregates confirmed high-mannose SARS-267 CoV-2 spike protein-coated particles were not effective in activating platelets in vitro (Fig. 268 4G, N=9, n=1898), exhibiting either no changes in mean platelet aggregate area from control 269 at ~10 μm2 (Fig. 4H) or non-significant differences with 40-50% fractional area occupancy of 270 platelet aggregate (Fig. 4I). The distinct impacts between (i) high mannose- and (ii) complex 271 sialylated glycans-, SARS-CoV-2 spike protein-coated particles on platelet aggregation (Fig. 272 4D-I) suggested the precise composition of terminal glycans in the SARS-CoV-2 spike 273 glycosylation could be a determinant in platelet clotting induction and SARS-CoV-2 274 associated thrombosis. 275 276 To dissect the molecular contribution of SARS-CoV-2 spike protein glycans to platelet 277 aggregation and thrombosis, specific glycosidases were used to trim off selected groups of 278 glycans to assess their capacity to induce platelet aggregation. Peptide:N-glycosidase F 279 (PNGase F) was used to remove SARS-CoV-2 spike protein N-linked glycans at the base of 280 the N-linked glycan ‘tree’ resulting in the complete removal of N-glycans (Fig. S7). A 281 sialidase ( α2-3,6,8,9 Neuraminidase A) was employed to snip off the terminal sialic acid 282 residues on the N-linked glycans (Fig. S7), thereby maintaining the majority of N-glycan tree, 283 while exposing galactose from previously sialic acid-capped structures to be the new terminal 284 N-glycans structures. Polyacrylamide gel electrophoresis (PAGE) protein separation coupled 285 with protein staining, immunoblot, and lectin probing were then carried out to confirm that: (i) 286 the majority removal of N-glycans by PNGase F; and (ii) the terminal sialic acids were 287 cleaved off from the N-glycans. These glycan modifications resulted in: (a) a faster migration 288 of recombinant spike protein upon PNGase F treatment; (b) a lower detectable level of sialic 289 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 19 acids from both PNGase F and sialidase treatment; and (c) an enhanced detectability of 290 galactose terminal residues post-sialidase treatment (Fig. 4J). 291 292 Using (i) no-spike protein stimulated platelets and (ii) wild-type sialic acid-containing 293 glycosylated SARS-CoV-2 spike protein stimulated platelets as negative and positive controls, 294 respectively, in SEM analyses (Fig. 4K-L, S9), it was noted that the ultrastructural 295 morphologies of platelets upon stimulation with either PNGase F-treated- (Fig. 4M, S9) or 296 sialidase-treated- (Fig. 4N, S9) SARS-CoV-2 spike protein were closer aligned with the no-297 spike protein stimulated platelets negative control. PNGase F-treated-spike protein stimulated 298 platelets (Fig. 4M, S9) had limited signs of activation in comparison with the sialic acid-299 containing complex glycan SARS-CoV-2 spike protein treated positive control (Fig. 4L, S9). 300 Population analyses on the sizes of platelet aggregates showed N-glycan-depleted 301 recombinant SARS-CoV-2 spike have only 0.1% platelet aggregates that are greater than 50 302 μm2 in size (Fig. 4O), which was at the background levels observed in no-spike protein 303 stimulated platelets (Fig. 4O). Both ‘mean platelet aggregate area’ (Fig. 4P) and ‘fractional 304 area occupancy of platelet aggregate’ (Fig. 4Q) from N-glycan-depleted recombinant SARS-305 CoV-2 spike protein treated platelets were compatible with background controls but 306 significantly different from wild-type complex glycan SARS-CoV-2 spike protein stimulated 307 platelets (p<0.001 Fig. 4P , p<0.01 Fig. 4Q). These data showed that the N-glycans of the 308 SARS-CoV-2 spike protein were directly involved in platelet activation and aggregation. 309 310 More strikingly, precision removal of terminal sialic acid residues from the SARS-CoV-2 311 spike protein N-glycans resulted in an absence of platelet aggregates above 50 μm2 (Fig. 4O) 312 and less visible nanoscale filopodia extrusion (Fig. S9). Both ‘mean platelet aggregate area’ 313 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 20 (Fig. 4P) and ‘fractional area occupancy of platelet aggregate’ (Fig. 4Q) from sialic acid free 314 (or galactose terminal-capped) spike protein-stimulated platelets were significantly different 315 from the wild-type spike protein-induced platelet control (p<0.001 Fig. 4P , p<0.01 Fig. 4Q). 316 These data provided direct evidence that the terminal sialic acid residues on the SARS-CoV-2 317 spike protein were responsible for platelets aggregations in vitro, suggesting a role for SARS-318 CoV-2 spike protein’s sialic acid residues in COVID-19 associated thrombo-inflammatory 319 pathology. 320 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 21 321 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 22 Figure 4. SARS-CoV-2 spike protein terminal sialic acid residues induce platelet aggregation. 322 (A-B) Confocal images of platelets incubated with mCherry-labeled Omicron spike protein-323 coated particles possessing (A) high-mannose (GnTI-/- HEK293T) or (B) complex-type 324 (HEK293T) glycans. (C) Quantification of VLP-platelet association with particles with high 325 mannose or particles with complex glycan. (D-F) Scanning electron microscopy (SEM) 326 images of human platelets that have been incubated with: (D) no particles, (E) Omicron spike 327 high-mannose glycoprotein-coated particles, and (F) Omicron spike complex glycoproteins-328 coated particles. (G) Quantification of mean platelet aggregate areas (G) and platelet 329 aggregation index (H) from SARS-CoV-2 spike protein-coated particles induced platelet 330 aggregation SEM analyses. (I) Size distributions of platelet aggregates from D-F. (J-M) SEM 331 images of human platelets that have been incubated with different recombinant spike proteins: 332 (J) Recombinant protein blots visualised by Coomassie blue, anti-his antibody, or lectins 333 (sambucus nigra lectin [SNA], wheat germ agglutinin [WGA], and peanut agglutin [PNA]) 334 for the detection of recombinant spike protein and associated glycans. (K) no spike protein; 335 (L) complex glycan spike protein; (M) glycan diminished PNGaseF treated spike protein; (N) 336 sialic acid diminished sialidase-treated spike protein. (O) Size distributions of platelet 337 aggregates from K-N. Quantification of mean platelet aggregate areas (P) and fractional area 338 occupancy of platelet aggregate in platelet population (Q) from recombinant SARS-CoV-2 339 spike protein induced platelet aggregation SEM analyses (K-N). Capital ‘N’ is for Donor 340 numbers and little ‘n’ is for number of events. Scale bars sizes are listed. *p<0.05, **p<0.01, 341 ***p<0.001, ****p<0.0001. 342 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 23

343 Platelet hyperactivation and thrombotic events are hallmarks of severe SARS-CoV-2 344 infection, yet the molecular determinants underlying these processes remain incompletely 345 defined. Our data showed that SARS-CoV-2 spike protein-coated particles and the spike 346 protein directly induced platelet aggregation. We further identified spike protein 347 glycosylation as a critical regulator of this effect. Using imaging and quantitative analyses, 348 we showed that spike protein-coated particles promote platelet aggregation, and spike protein 349 alone was sufficient to cause platelet aggregation in a concentration-dependent manner. 350 Importantly, using glycan modified SARS-CoV-2 spike protein-coated particles and 351 glycosidase trimmed recombinant SARS-CoV-2 spike protein, we directly demonstrated that 352 the terminal sialic acid residues of SARS-CoV-2 spike protein are the key determinant which 353

in platelet aggregation induction. Together, these findings establish a mechanistic link 354 between SARS-CoV-2 spike protein-associated sialic acid residues and platelet aggregation 355 thrombosis biology. 356 357 A recent study has shown that platelet OCS can trap cell-free DNA 21. Diagnostic detection of 358 rare foetal DNA or cancer-derived DNA is more sensitive with lysed whole blood in 359 comparison with ‘conventional gold standard’ platelet-depleted plasma 21. The ability of 360 platelets to trap SARS-CoV-2 spike protein-coated particles in OCS is likely to: (i) impair 361 blood detection of SARS-CoV-2 using standard platelet-depleted plasma; and (ii) enable 362 platelets to serve as a Trojan horse reservoir to delay virus clearance in convalescence 363 individuals, thereby prolong the duration of SARS-CoV-2 infection in the host. 364 365 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 24 It is estimated that 14.4 million life-years have been saved with the SARS-CoV-2 vaccine 37, 366 yet the unappreciated biology that contributes to SARS-CoV-2 related thrombosis 1,2,4 and 367 vaccine-associated thrombotic events 14,16,38,39 have dampened vaccine confidence. While the 368 mechanism of pathogenetic anti-PF4 antibodies induction in VITT is now resolved 18, the 369 source of SARS-CoV-2 mRNA vaccine-related thrombotic events 14 and myocarditis 19 370 remain a mystery. Building from our prior work on glycan-glycan interaction in viral 371 pathology 28, we noted that the high density of terminal sialic acid residues is found on the 372 surfaces of both the SARS-CoV-2 spike proteins 29 and the host haemostasis initiator V on 373 Willebrand factor (VWF) 30,31. As N-glycan-associated terminal sialic acid residues contribute 374 to VWF functions and prevent premature plasma clearance 40-42, we postulated this shared 375 feature enables the SARS-CoV-2 spike protein to mimic host cell VWF to activate platelets 376 that leads to the formation of microscopic scale clots resulting in thrombotic 377 thrombocytopenia. 378 379 Our proposed mechanism on viral protein-associated sialic acids as the causation of SARS-380 CoV-2 induced thrombotic thrombocytopenia is in line with recent work showing that a 381 sialyltransferase gene ST6GAL1 is associated with SARS-CoV-2 infection 33. The 382 contribution of ST6GAL1 intron single nucleotide polymorphism (SNP) to SARS-CoV-2 383 infection could be through: (i) altering the stability of ST6GAL1 mRNA for protein 384 expression; and/or (ii) interfering with the functions of other glycosylation machinery 33. 385 With up to 2% of the human genome (over 400 glyco-genes plus 950 pathway-process related 386 genes) directed towards glyco-modifications, personalised glycosylation does not only 387 determine human blood types 43 but also dictates the individualised glycan profiles of the 388 SARS-CoV-2 spike protein upon infection. If the patterns of SARS-CoV-2 spike protein 389 terminal sialic acid residues and their relationships with host biology are the keys to initiate 390 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 25 COVID-19 thrombotic pathogenesis, it is perhaps not surprising to observe the vast 391 differences on the level of disease pathogenesis across SARS-CoV-2 infected individuals. 392 393 Despite the high-level safety standard being achieved with the adenoviral vector-based 394 SARS-CoV-2 vaccine 11, the concerns with rare vaccine-induced thrombotic 395 thrombocytopenia have forced several effective COVID vaccines out of the market, 396 prematurely. Identified the linkage between adenoviral core proteins and immunoglobulin 397 light-chain allele, IGL V3-21*02 or *03 will help mitigate risks of CVDs through adenoviral 398 vector vaccine-related VITT 18 but does not offset rare SARS-CoV-2 mRNA vaccine-399 associated thrombotic events 14 or myocarditis 19. As we have shown that glycan-modified 400 SARS-CoV-2 spike protein is ineffective in platelet aggregation induction, glycan-modified 401 vaccines have the potential to reduce risks of vaccine-related thrombotic events. Recent 402 studies have shown that introduction of glycosylation machinery inhibitors during viral 403 protein production can disable SARS-CoV-2 function by interfering with the glycan profiles 404 of progeny viruses 44. Local co-injection of glycosylation machinery inhibitors with SARS-405 CoV-2 mRNA vaccine could alter the glycan profiles of soluble SARS-CoV-2 spike protein 406 to reduce the risk of SARS-CoV-2 mRNA vaccine-associated thrombotic events or 407 myocarditis, while still enabling the production of the protein antigen to elicit protective 408 immunity. Such a ‘glycan-modified mRNA vaccine’ approach is likely to have wider 409 applications to other glycan-dependent pathogens, including the potential to crack open the 410 HIV glycan shield and to expose broadly neutralising epitopes, thereby facilitating the 411 elicitation and the maturation of protective immunity against HIV . 412 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 26

413 Human Ethics statement 414 Human blood samples were obtained from healthy adult donors following informed consent 415 in accordance with the Declaration of Helsinki. Ethical approval for this study was granted by 416 the Griffith University Human Research Ethics Committee (HREC approval number: 417 HREC/2026/0129). All participants provided written informed consent prior to blood 418 collection, and all experiments involving human blood were conducted in accordance with 419 approved institutional guidelines. 420 Blood collection and platelet isolation 421 Human blood was collected in Acid Citrate Dextrose (ACD) solution containing BD 422 Vacutainer® glass whole blood 'ACD-A' tube from the de-identified donors. Platelet-rich 423 plasma (PRP) was prepared from whole blood collected from healthy adult donors. All 424 donors were medication-free for at least 10 days prior to blood collection and had not taken 425 antiplatelet or anti-inflammatory drugs during this period. Blood was collected using standard 426 venipuncture techniques with minimal stasis to avoid artifactual platelet activation. To 427 minimise spontaneous activation, PRP was prepared under controlled conditions using low-428 speed centrifugation without brake, and all handling steps were performed gently using wide-429 bore pipette tips. Samples were processed at room temperature and used within a standardised 430 time frame following collection. 431 Immediately after blood was drawn from the donors, it was centrifuged for 15 /i2 min at 432 100/i2 ×/i2 g at room temperature. The top 2/3 of the platelet-rich plasma (PRP) was transferred 433 to a fresh falcon tube. After gently inverting 3-5 times to homogenise the concentration and 434 distribute to 1.5 ml tubes each with 95 ul. Immediately, spike protein was added to the PRP 435 with no pipetting or swirling the pipette tip. After incubating the PRP and spike protein mix 436 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 27 in RT/37 °C for 15-30 minutes, samples for SEM imaging were fixed with 2.5% 437 glutaraldehyde (end concentration) 438 Production of SARS-CoV-2 spike-coated- and naked- virus-like-particles (VLPs) 439 SARS-CoV-2 spike-coated- and naked (spike-free)- virus-like-particles (VLPs) were 440 produced using the HEK 293T cell line or GlcNAc transferase I defective 293S GnTI-/- cells 441 via transfection. All cells were cultured under standard conditions as previously described 442 28,45. The plasmids used to generate the nanoluciferase VLPs for the cell surface binding assay 443 were: (i) NL4-3 based proviral DNA construct expressing Gag and Env, but with deletions in 444 RT, IN, Vif and Vpr (pNL4-3 Δ RTΔ INΔ VifΔ Vpr); and (ii) nanoluciferase-Vpr expression 445 construct under control of the CMV promoter. Each transfection utilized a total of 12 μ g 446 plasmid DNA, combined in an 8:1 ratio of each plasmid listed above. The plasmids used to 447 generate the VLPs for fluorescent imaging were: (i) pNL43Δ PolΔ Env; (ii) pNL43Δ PolΔ Env-448 Gag-mEOS2; and (iii) SARS-CoV-2 spike protein expression vector. Each transfection 449 utilized a total of 12 μ g plasmid DNA, combined in a 3:1:1 ratio of each plasmid listed above. 450 For VLP production, HEK 293T cells were transiently transfected using polyethylenimine 451 (PEI MAX). PEI/DNA complexes were formed by adding PEI to plasmid DNA diluted in a 452 small volume of fresh DMEM. The mixtures were vortexed vigorously and incubated at room 453 temperature for 30 min prior to transfection. Culture supernatants of transfected HEK 293T 454 cells were harvested 48 h post transfection and clarified by centrifugation at 1500 x g for 10 455 min. Cell debris was removed by passing the supernatants through a 0.45 μ m filter. VLPs 456 were then concentrated by ultra-centrifugation over a 20% sucrose cushion. Concentrated 457 VLPs were resuspended in DPBS and stored at −80 °C. VLP production, where possible, was 458 quantified using a p24 ELISA (Xpress Bio). 459 Recombinant spike protein expression and purification 460 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 28 Plasmid encoding for Hexaprolin spike FL trimer was transiently transfected into HEK293-T 461 cells at a final concentration of 3 µg/mL in Freestyle™ 293 Expression Medium using 462 polyethylenimine (PEI, Polysciences® ) as a transfection reagent. After 24 h, the cells were 463 diluted 1:1 with ExCell® 293 Serum-Free Medium for HEK 293 cells supplemented with 2.2 464 mM valproic acid to promote additional endocytic uptake of the plasmid-PEI complex. After 465 4 days, the supernatant was clarified by centrifugation for 30 min at 3000 × g at 4ºC and 466 treated with benzonase® (Merck) for 4 h at 4ºC before downstream purification. 467 A pre-packed Ni2+-sepharose column (HisTrap Excel, GE Life Sciences) was used to purify 468 His-tagged proteins from the cell culture supernatant by immobilised metal ion affinity 469 chromatography (IMAC). The Ni2+-column was first equilibrated with 50 mM Na2HPO4, 300 470 mM NaCl, and 10 mM Imidazole pH 8 for capturing of the His-tagged Hexaprolin spike FL 471 protein. Following, the protein was eluted with 250 mM imidazole in the same buffer and 472 detected by UV absorbance at 280 nm. TEV protease cleavage was achieved, before the 473 protein was concentrated and buffer exchanged into 20 mM Tris and 200 mM NaCl (pH 8.0) 474 over a 30 KDa cut-off Amicon Centrifuge Filter. The concentrate was loaded on a HiLoad 475 26/600 Superdex pg 200 column (GE Life Sciences) for size exclusion chromatography and 476 collected in fractions of 1 mL. The fractions were checked for protein presence and purity by 477 SDS-PAGE with colloidal Coomassie blue staining to visualise protein bands After 478 expression and purification, Hexprolin spike FL protein was concentrated to 15 mg/mL and 479 stored at – 20 °C until further use. 480 SEM imaging and image analysis 481 For scanning electron microscopy (SEM) imaging, platelet-rich plasma (PRP) samples (95 482 µL) were incubated with either spike-coated particles (5 µL; 40 ng/µL final particle stock 483 concentration) or soluble spike protein for 20–30 minutes at room temperature. For dose-484 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 29 dependent experiments, spike protein was added at final concentrations of 0.05 mg/mL, 0.1 485 mg/mL, and 0.5 mg/mL. In all other experiments, spike protein was used at a final 486 concentration of 0.5 mg/mL. Control samples received 5 µL of gel filtration buffer, matching 487 the buffer used for spike protein preparation. Immediately after incubation, samples were 488 fixed by adding 2.5% glutaraldehyde to reach a final volume of approximately 400–500 µL 489 and fixed for 30 minutes. This dilution step with glutaraldehyde was critical, as fixation 490 without dilution caused PRP samples to become gel-like; maintaining a liquid suspension 491 ensured efficient platelet settling onto substrates. Following 30 minutes fixation, platelet 492 suspensions were deposited onto poly-L-lysine–coated Thermanox coverslips and allowed to 493 settle and adhere to the substrate and continue fixing overnight. Samples were washed twice 494 with PBS anddehydrated with serial increasing concentrations of ethanol (30%-50%-70%-495 90%-100%) for 10 minutes at each ethanol concentration, and twice with 100% 496 ethanbol.Samples were then dehydrated using a critical point dryer (Leica EM 497 CPD300). The dried samples on Thermanox substrates were mounted onto 498 aluminium specimen stubs using conductive carbon tape, and the edges of the 499 substrates were additionally covered with carbon tape to improve electrical 500 conductivity. Samples were subsequently sputter-coated with gold prior to imaging. 501 SEM imaging was performed under identical, standardised acquisition conditions using a 502 Thermo Fisher Phenom XL SEM. At least six random, non-overlapping fields were captured 503 per donor for each experimental condition. Following image acquisition, all SEM images 504 were analysed using ImageJ. A platelet aggregate was defined as the association of any two 505 or more platelets (total area > 6 µm²). The surface areas of individual platelets and platelet 506 aggregates were quantified using ImageJ, and at least 6 images (technical replicates) platelets 507 area were averaged per donor prior to statistical analysis. 508 Thin-section TEM imaging of platelets 509 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 30 Platelet-rich plasma (PRP) was prepared from whole blood by low-speed centrifugation (100 510 × g). Platelet activation was induced by adding 5 µL of Omicron spike–coated particles (80 511 ng/µL) to 95 µL of PRP, followed by incubation at 37 °C for 30 minutes. Resting platelet 512 controls were prepared by incubating PRP under identical conditions without the addition of 513 any agonist for 30 minutes at 37 °C. Following incubation, samples were fixed overnight at 514 4 °C in 2.5% glutaraldehyde prepared in PBS. During fixation, PRP samples were diluted 515 approximately 3–5-fold with fixative to ensure proper preservation of platelets as in liquid 516 suspension. Samples washed twice with PBS and followed by 1% osmium tetroxide post-517 fixation using a Pelco Biowave processor following our methods reported elsewhere 46. 518 Samples were dehydrated through a graded ethanol series (30–100%) and infiltrated with 519 LX112 resin, followed by polymerisation at 60 °C for 48 h. Resin blocks were trimmed and 520 sectioned (~70 nm thickness) using a Leica Ultracut UC7 ultramicrotome with a DiATOME 521 diamond knife. Sections were collected onto Formvar-coated TEM grids, stained with 5% 522 uranyl acetate and Reynolds’ lead citrate, air-dried, and imaged using either a JEOL 523 JEM-1400 Flash or Hitachi HT7700 operating at 80 kV . Thin-section TEM images of 524 platelets and their OCSs parameters were analysed using the ImageJ software. 525 Focus Ion Beam – Scanning Electron Microscopy (FIB-SEM) analysis 526 mCherry-tagged Omicron spike–coated particles were incubated with platelet-rich plasma 527 (PRP) for 30 min at 37 °C and fixed in 4% paraformaldehyde (PFA). PFA fixed samples were 528 transferred onto gridded glass-bottom dishes (MatTek gridded glass-bottom dish) to enable 529 coordinate-based correlation between fluorescence and electron microscopy images. 530 Confocal fluorescence imaging was performed to identify regions of interest (ROIs). 531 Following confocal imaging, samples were post-fixed in 2.5% glutaraldehyde and processed 532 using a workflow like that used for thin-section transmission electron microscopy sample 533 preparation in this work. To enhance membrane contrast for focused ion beam scanning 534 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 31 electron microscopy (FIB–SEM), samples were further post-fixed in 1% osmium tetroxide 535 overnight at 4 °C, washed twice with PBS, and subjected to an additional osmium tetroxide 536 staining step using a Pelco BioWave microwave processor. 537 Samples were then embedded in LX112 resin and polymerized at 60 °C for 24–48 hours. 538 After resin curing, the plastic edges surrounding the glass coverslip were trimmed, and the 539 resin block was released from the coverslip by alternating cycles of liquid nitrogen and hot 540 water (2–5 cycles), yielding a grid-transferred resin block approximately 3–5 mm thick. Prior 541 to imaging, the resin block surface was sputter-coated with platinum and transferred to a FIB-542 SEM. Low-magnification SEM overview images were acquired to identify alphanumeric grid 543 coordinates corresponding to previously imaged ROIs from confocal microscopy. After 544 reidentifying of the ROI in FIB-SEM, protective carbon and platinum layers were deposited 545 over the block surface to minimize ion-beam damage during milling. Initial coarse trench 546 milling was performed manually. Once platelet structures were exposed, automated serial 547 slice-and-view acquisition was initiated. Three-dimensional datasets were collected at a voxel 548 size of 5.4 × 5.4 × 20 nm with a slicing thickness of 20 nm. 549 Luciferase reporter virus-like-particles (VLPs) – Platelets binding assay. 550 NanoLuc-tagged virus-like particles (VLPs) pseudotyped with Omicron spike or spike-551 deficient control were generated in HEK293T cells or GnTi- cells and quantified by p24 552 ELISA prior to use. Platelet-rich plasma (PRP) was prepared from citrated whole blood by 553 centrifugation at 200 × g for 15 min at room temperature without brake, and platelets were 554 maintained under resting conditions throughout preparation. Washed platelets (1 × 10 /i2 per 555 condition) were incubated with NanoLuc-tagged VLPs at defined particle-to-platelet ratios 556 (approximately 100:1) in a final volume of 200 µL at 37 °C for 30 min with gentle mixing. 557 Following 30 minutes incubation at 37 °C, platelets were pelleted at 1200 × g for 10 min and 558 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 32 washed five times with Tyrode’s buffer to remove unbound particles. Platelet pellets were 559 then divided into three fractions to quantify total associated, surface-bound, and internalised 560 VLPs. Total platelet-associated NanoLuc signal was measured directly after resuspension in 561 buffer and addition of substrate. Surface-bound and partially internalised VLPs were removed 562 by brief treatment with 0.05% trypsin at 37 °C for 3–5 min followed by washing 3 times. 563 PNGase F and Sialidase treatment of spike protein under non-denaturing conditions 564 Recombinant His-tagged SARS-CoV-2 spike protein was diluted to 0.556 mg mL /i2 ¹ prior to 565 enzymatic treatment. Deglycosylation reactions were performed using PNGase F (for the 566 removal of N-linked glycans) and sialidase (for the removal of terminal sialic acid residues), 567 and the mock (without enzyme treatment) respectively. All reactions were incubated at 37 °C 568 for 1 h. Following the manufacturer's recommended procedure, For PNGase F treatment, 100 569 µg spike protein (180 µL) was mixed with 20 µL 10× GlycoBuffer and 30 µL PNGase F 570 enzyme. For sialidase treatment, 30 µg spike protein (54 µL) was mixed with 6 µL 10× 571 GlycoBuffer and 30 µL sialidase enzyme. Mock-treated spike controls were prepared by 572 incubating spike protein with 10× GlycoBuffer only under identical conditions. Following 573 incubation, reaction mixtures were subjected to buffer exchange and enzyme removal using 574 100 kDa molecular weight cut-off Amicon centrifugal filters (0.5 mL). Samples were 575 centrifuged at 5000 ×g for 5 min at 4 °C, followed by replenishment with gel-filtration buffer 576 to the retained volume. This wash step was repeated five times to ensure removal of free 577 enzyme and reaction buffer components. Protein concentrations following deglycosylation 578 and buffer exchange were determined by absorbance at 280 nm using a Nanodrop 579 spectrophotometer and calculated using: 580 /g1829/g3404/g1827/g3400/g1839/g1849 /g2013 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 33 where A is absorbance at 280 nm, MW is the molecular weight of spike protein (540 kDa 581 trimer equivalent), and ε is the extinction coefficient (428,225 M/i2 ¹ cm/i2 ¹). 582 Following concentration measurement, all enzyme-treated and mock-treated (as control) 583 spike protein preparations were adjusted to a final concentration of 0.5 mg mL /i2 ¹ prior to 584 downstream platelet activation experiments. Successful deglycosylation following PNGase F 585 and sialidase treatment was verified by SDS-PAGE mobility shift and lectin blotting. 586 Western and lectin blot analysis 587 Enzyme-treated and mock-treated Spike protein run through SDS-PAGE 10% Bis-Tris 588 NuPAGE (Invitrogen, NW00105BOX), one set left for Coomassie blue staining, the rest set 589 of gels transferred to nitrocellulose membrane (GE Healthcare). Membranes were blocked 590 with 5% (w/v) skim milk in Tris Buffered Saline, 0.05% Tween20 (TBST) for western blot, 591 washed and probed with anti-his monoclonal antibody. HRP conjugated anti-mouse (Dako, 592 P0161) was used as secondary antibodies, and blots was imaged by chemiluminescence 593 (SuperSignal TM, Thermo scientific, 34580). Western blots were imaged using BioRad 594 ChemiDoc XRS+. Membranes were blocked with 3% BSA in TBST for lectin blot, and 595 incubated with Sambucus nigra Lectin (SNA, FITC conjugated) for sialic acid detection, 596 Wheat germ agglutinin (WGA, FITC conjugated) for the detection of sialic acid and N-597 acetylglucosamine, peanut agglutin (PNA, FITC conjugated) for the detection of galactose. 598 Lectin blots washed and imaged with using BioRad ChemiDoc XRS+ with 488 nm laser 599 excitation. 600 Statistical Analysis 601 Statistical analyses were performed using GraphPad Prism. Two tailed student's t test and 602 One-way ANOV A with Bonferroni's post entry test were used in the data analysis. Statistical 603 significance is expressed as p <0.05 (∗ ), p <0.01 (∗∗ ), p <0.001 (∗∗∗ ) for all tests. 604 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 34 Acknowledgments 605 General: We express our gratitude to Dr Oren Cooper at IBG, Griffith University, for 606 generously providing us with lectins used in lectin blots in this study. EM analyses were 607 performed at the Centre for Microscopy and Microanalysis at the University of Queensland 608 (CMM UQ), Australia. We thank Drs Nicole Schieber and Hui Diao at CMM UQ for their 609 assistance in FIB-SEM set up. 610 Funding: This work was supported by the Australian Centre for HIV and Hepatitis Virology 611 Research (to JM), the US National Institute of Health (NIAID R21AI172534 to JM), the 612 Australian National Health and Medical Research Council (GNT2018895 to JM, 613 GNT1196520 and GNT 2042634 to MvI ). The content is solely the responsibility of the 614 authors and does not necessarily represent the official views of the Funders. 615 Author contributions 616 AB, BLS, and JM designed the study. AB, AS, EB, VM and BLS performed the experiments. 617 LD and MvI contributed to reagents. AB and JM wrote the manuscript. All authors 618 contributed to the discussions on the experiments, the collected data, and the finalisation of 619 the manuscript. 620 Declaration of interests 621 The authors declare no competing interests. 622 Data and materials availability 623 All data needed to evaluate the conclusions in the paper are present in the paper and / or in 624 the Supplementary Materials. Additional data related to this may be requested from the 625 corresponding author. 626 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 35

627 1 Varga, Z. et al. Endothelial cell infection and endotheliitis in COVID-19. Lancet 628 395, 1417-1418 (2020). https://doi.org/10.1016/S0140-6736(20)30937-5 629 2 Ackermann, M. et al. Pulmonary vascular endothelialitis, thrombosis, and 630 angiogenesis in Covid-19. New England Journal of Medicine 383, 120-128 631 (2020). 632 3 Knight, R. et al. Association of COVID-19 With Major Arterial and Venous 633 Thrombotic Diseases: A Population-Wide Cohort Study of 48 Million Adults in 634 England and Wales. Circulation 146, 892-906 (2022). 635 https://doi.org/10.1161/CIRCULATIONAHA.122.060785 636 4 Wichmann, D. et al. Autopsy Findings and Venous Thromboembolism in 637 Patients With COVID-19: A Prospective Cohort Study. Annals Of Internal 638 Medicine. 173, 268-277 (2020). https://doi.org/10.7326/M20-2003 639 5 Wang, W. et al. Detection of SARS-CoV-2 in Different Types of Clinical 640 Specimens. Jama. 323, 1843-1844 (2020). 641 https://doi.org/10.1001/jama.2020.3786 642 6 Manne, B. K. et al. Platelet gene expression and function in patients with 643 COVID-19. Blood, The Journal of the American Society of Hematology 136, 644 1317-1329 (2020). 645 7 Mehta, P . et al. COVID-19: consider cytokine storm syndromes and 646 immunosuppression. Lancet 395, 1033-1034 (2020). 647 https://doi.org/10.1016/S0140-6736(20)30628-0 648 8 Jose, R. J. & Manuel, A. COVID-19 cytokine storm: the interplay between 649 inflammation and coagulation. Lancet Respir Med 8, e46-e47 (2020). 650 https://doi.org/10.1016/S2213-2600(20)30216-2 651 9 Barrett, T . J. et al. Platelets contribute to disease severity in COVID-19. J 652 Thromb Haemost 19, 3139-3153 (2021). https://doi.org/10.1111/jth.15534 653 10 Barrett, T. J. et al. Platelets amplify endotheliopathy in COVID-19. Sci Adv 7, 654 eabh2434 (2021). https://doi.org/10.1126/sciadv.abh2434 655 11 Voysey, M. et al. Safety and efficacy of the ChAdOx1 nCoV-19 vaccine 656 (AZD1222) against SARS-CoV-2: an interim analysis of four randomised 657 controlled trials in Brazil, South Africa, and the UK. Lancet 397, 99-111 (2021). 658 https://doi.org/10.1016/S0140-6736(20)32661-1 659 12 Polack, F. P . et al. Safety and Efficacy of the BNT162b2 mRNA Covid-19 660 Vaccine. New England Journal Of Medicine. 383, 2603-2615 (2020). 661 https://doi.org/10.1056/NEJMoa2034577 662 13 Baden, L. R. et al. Efficacy and Safety of the mRNA-1273 SARS-CoV-2 663 Vaccine. New England Journal Of Medicine. 384, 403-416 (2021). 664 https://doi.org/10.1056/NEJMoa2035389 665 14 Smadja, D. M., Yue, Q. Y ., Chocron, R., Sanchez, O. & Lillo-Le Louet, A. 666 Vaccination against COVID-19: insight from arterial and venous thrombosis 667 occurrence using data from VigiBase. Eur Respir J 58 (2021). 668 https://doi.org/10.1183/13993003.00956-2021 669 15 von Hundelshausen, P., Lorenz, R., Siess, W. & Weber, C. Vaccine-Induced 670 Immune Thrombotic Thrombocytopenia (VITT): Targeting Pathomechanisms 671 with Bruton Tyrosine Kinase Inhibitors. Thromb Haemost 121, 1395-1399 672 (2021). https://doi.org/10.1055/a-1481-3039 673 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 36 16 Greinacher, A. et al. Vaccine-induced immune thrombotic thrombocytopenia 674 (VITT): Update on diagnosis and management considering different resources. 675 J Thromb Haemost 20, 149-156 (2022). https://doi.org/10.1111/jth.15572 676 17 Greinacher, A. et al. Anti-platelet factor 4 antibodies causing VITT do not 677 cross-react with SARS-CoV-2 spike protein. Blood. 138, 1269-1277 (2021). 678 https://doi.org/10.1182/blood.2021012938 679 18 Wang, J. J. et al. Adenoviral Inciting Antigen and Somatic Hypermutation in 680 VITT. New England Journal Of Medicine. 394, 669-683 (2026). 681 https://doi.org/10.1056/NEJMoa2514824 682 19 Yonker, L. M. et al. Circulating Spike Protein Detected in Post-COVID-19 683 mRNA Vaccine Myocarditis. Circulation 147, 867-876 (2023). 684 https://doi.org/10.1161/CIRCULATIONAHA.122.061025 685 20 White, J. G. Platelets are covercytes, not phagocytes: uptake of bacteria 686 involves channels of the open canalicular system. Platelets 16, 121-131 687 (2005). https://doi.org/10.1080/09537100400007390 688 21 Murphy, L. et al. Platelets sequester extracellular DNA, capturing tumor-689 derived and free fetal DNA. Science. 389, eadp3971 (2025). 690 https://doi.org/10.1126/science.adp3971 691 22 Zhang, S. et al. SARS-CoV-2 binds platelet ACE2 to enhance thrombosis in 692 COVID-19. J Hematol Oncol 13, 120 (2020). https://doi.org/10.1186/s13045-693 020-00954-7 694 23 Real, F. et al. Platelets from HIV-infected individuals on antiretroviral drug 695 therapy with poor CD4(+) T cell recovery can harbor replication-competent 696 HIV despite viral suppression. Science translational medicine 12 (2020). 697 https://doi.org/10.1126/scitranslmed.aat6263 698 24 Hilser, J. R. et al. COVID-19 Is a Coronary Artery Disease Risk Equivalent 699 and Exhibits a Genetic Interaction With ABO Blood Type. Arterioscler Thromb 700 Vasc Biol 44, 2321-2333 (2024). 701 https://doi.org/10.1161/ATVBAHA.124.321001 702 25 Zietz, M., Zucker, J. & Tatonetti, N. P. Associations between blood type and 703 COVID-19 infection, intubation, and death. Nature communications 11, 5761 704 (2020). https://doi.org/10.1038/s41467-020-19623-x 705 26 Cohen, M., Hurtado-Ziola, N. & Varki, A. ABO blood group glycans modulate 706 sialic acid recognition on erythrocytes. Blood. 114, 3668-3676 (2009). 707 https://doi.org/10.1182/blood-2009-06-227041 708 27 Reitter, J. N., Means, R. E. & Desrosiers, R. C. A role for carbohydrates in 709 immune evasion in AIDS. Nat Med 4, 679-684 (1998). 710 28 Spillings, B. L. et al. Host glycocalyx captures HIV proximal to the cell surface 711 via oligomannose-GlcNAc glycan-glycan interactions to support viral entry. 712 Cell reports 38, 110296 (2022). https://doi.org/10.1016/j.celrep.2022.110296 713 29 Watanabe, Y ., Allen, J. D., Wrapp, D., McLellan, J. S. & Crispin, M. Site-714 specific glycan analysis of the SARS-CoV-2 spike. Science. 369, 330-333 715 (2020). https://doi.org/10.1126/science.abb9983 716 30 Ward, S., O'Sullivan, J. M. & O'Donnell, J. S. von Willebrand factor sialylation-717 A critical regulator of biological function. J Thromb Haemost 17, 1018-1029 718 (2019). https://doi.org/10.1111/jth.14471 719 31 Canis, K. et al. Mapping the N-glycome of human von Willebrand factor. 720 Biochem J 447, 217-228 (2012). https://doi.org/10.1042/BJ20120810 721 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint 37 32 Ruggeri, Z. M. Von Willebrand factor, platelets and endothelial cell interactions. 722 J Thromb Haemost 1, 1335-1342 (2003). https://doi.org/10.1046/j.1538-723 7836.2003.00260.x 724 33 Geller, F. et al. Central role of glycosylation processes in human genetic 725 susceptibility to SARS-CoV-2 infections with Omicron variants. Nat Genet 58, 726 299-306 (2026). https://doi.org/10.1038/s41588-025-02484-9 727 34 Zaid, Y . et al. Platelets Can Associate with SARS-Cov-2 RNA and Are 728 Hyperactivated in COVID-19. Circ Res (2020). 729 https://doi.org/10.1161/CIRCRESAHA.120.317703 730 35 Koupenova, M. et al. The role of platelets in mediating a response to human 731 influenza infection. Nature communications 10, 1780 (2019). 732 https://doi.org/10.1038/s41467-019-09607-x 733 36 Kuhn, C. C. et al. Direct Cryo-ET observation of platelet deformation induced 734 by SARS-CoV-2 spike protein. Nature communications 14, 620 (2023). 735 https://doi.org/10.1038/s41467-023-36279-5 736 37 Watson, O. J. et al. Global impact of the first year of COVID-19 vaccination: a 737 mathematical modelling study. Lancet Infect Dis 22, 1293-1302 (2022). 738 https://doi.org/10.1016/S1473-3099(22)00320-6 739 38 Pavord, S. et al. Clinical Features of Vaccine-Induced Immune 740 Thrombocytopenia and Thrombosis. New England Journal Of Medicine. 385, 741 1680-1689 (2021). https://doi.org/10.1056/NEJMoa2109908 742 39 Klok, F. A., Pai, M., Huisman, M. V. & Makris, M. Vaccine-induced immune 743 thrombotic thrombocytopenia. Lancet Haematol 9, e73-e80 (2022). 744 https://doi.org/10.1016/S2352-3026(21)00306-9 745 40 Grewal, P. K. et al. The Ashwell receptor mitigates the lethal coagulopathy of 746 sepsis. Nat Med 14, 648-655 (2008). https://doi.org/10.1038/nm1760 747 41 Ward, S. E. et al. A novel role for the macrophage galactose-type lectin 748 receptor in mediating von Willebrand factor clearance. Blood. 131, 911-916 749 (2018). https://doi.org/10.1182/blood-2017-06-787853 750 42 Byrne, C. et al. Enhanced alpha2-3-linked sialylation determines the extended 751 half-life of CHO-rVWF. Blood. 145, 2768-2773 (2025). 752 https://doi.org/10.1182/blood.2024027038 753 43 Stanley, P ., Wuhrer, M., Lauc, G., Stowell, S. R. & Cummings, R. D. in 754 Essentials of Glycobiology (eds A. Varki et al.) 165-184 (2022). 755 44 Thaler, M. et al. Epi-Cyclophellitol Cyclosulfate, a Mechanism-Based 756 Endoplasmic Reticulum alpha-Glucosidase II Inhibitor, Blocks Replication of 757 SARS-CoV-2 and Other Coronaviruses. ACS Cent Sci 10, 1594-1608 (2024). 758 https://doi.org/10.1021/acscentsci.4c00506 759 45 Kishor, C. et al. Calcium Contributes to Polarized Targeting of HIV Assembly 760 Machinery by Regulating Complex Stability. JACS Au 2, 522-530 (2022). 761 https://doi.org/10.1021/jacsau.1c00563 762 46 Bremaud, E. et al. Calcium-phosphate bridge is a novel phosphorylation 763 switch that stabilises protein-complexes during HIV assembly. bioRxiv , 764 2026.2002. 2025.708077 (2026). 765 766 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 20, 2026. ; https://doi.org/10.64898/2026.04.19.719510doi: bioRxiv preprint