Patients were prospectively included in a prospective observational study in a medico-surgical intensive care unit (ICU) from April 12^th, 2021 to May 25^th, 2022. Eligible patients were hospitalized for hypoxemic pneumonia caused by SARS-CoV-2 infection. Variants identification was performed through complete sequencing of the spike gene. Ethical approval was obtained from the Comité de protection des personnes du sud-ouest et outre-mer (CPP2020-04-042a/2020-A00972-37/20.04.08.64705) and all patients gave their consent to participate, in accordance with the Declaration of Helsinki. Patients with malignant blood diseases or hemostatic disorders, pregnant women, minors or individuals under protective measures were excluded. Baseline characteristics including hemostasis parameters (obtained from the Laboratory of haematology, Rangueil Academic Hospital of Toulouse, France), and outcomes were recorded upon admission to the ICU. Blood was collected 3.8 ± 3.2 and 3.2 ± 1.4 days after ICU admission and 15.6 ± 10.1 and 9.1 ± 4.4 days after symptom onset for patients infected with the Delta and the Omicron variants, respectively. SARS-CoV-2 serology was assessed using the SARS-CoV-2 multi-antigen serology module (5 antigens analyzed simultaneously: S1 RDB, Nucleocapsid, S2 subunit, S1 subunit, Spike) allowing detection of IgG antibodies in plasma by Simple Western capillary electrophoresis (22). Seropositivity and seronegativity was determined according to the cut-offs provided by the manufacturer.

Platelet function tests were not realized if acetyl salicylic acid or corticosteroids were taken within the 10 days prior inclusion.

The recruitment of healthy donors was approved by the Toulouse Hospital Bio-Resources biobank, registered with the Ministry of Higher Education and Research (DC2016-2804). These donors were tested negative for PCR at inclusion, and had not taken antiplatelets or steroids within the 10 days before blood sampling. The cohort consisted of 50% females, with median age of 38 years (interquartile range, 26-69).

Collagen-related peptide (CRP) was obtained from Prof. R. Farndale Laboratory (Cambridge, UK), and the collagen reagent Horm^® (equine) suspension was purchased from Takeda (St. Peter Strasse, Austria). Anti-CD62P conjugated FITC, anti-CD63 conjugated FITC, anti-PAC1 conjugated FITC, antibodies were obtained from BD Biosciences and the anti-GPVI antibody was obtained from Biocytex (Marseille, France). The SARS-CoV-2 (2019-nCoV) Spike S1(D614G)-His recombinant protein and SARS-CoV-2 BQ.1.1 (Omicron) Spike S1+S2 trimer proteins were sourced from SinoBiological (Beijing, China). All other reagents were purchased from Sigma Aldrich (Saint-Louis, USA). Supplementary Materials provide further details regarding the sources and references of the antibodies.

Whole blood samples were collected using Vacutainer tubes containing 3.2% citrate (Becton Dickinson). Platelet-rich plasma (PRP), platelet-poor plasma (PPP), and washed platelets were obtained through differential centrifugation as described previously (22). PPP, used for measuring soluble markers, was stored at -80°C. Washed platelets were resuspended in modified HEPES Tyrode’s buffer with a pH of 7.4 and containing 2 mM CaCl₂ at 2x10⁸ platelets/mL for confocal microscopy and western-blot analysis.

Platelet aggregation was monitored with a turbidimetric method (SD Medical, STAGO, France) following previously described protocols (22). PRP was stimulated with various agonists (CRP 9 μg/mL, thrombin receptor agonist peptide (TRAP) 50 μM, collagen 0.75 μg/mL, thromboxane A2 analog U46619 1 μM) for 10 minutes under continuous stirring conditions (1000 revolution min^-1) at 37°C using siliconized glass cuvettes. The thrombosoft 1.6 software (SD Medical) was used to determine the maximal platelet aggregation percentage.

Flow cytometry was performed using FACSVERSE (BD Biosciences), and FASCsuites software (BD Biosciences) was used for data analysis. Platelet granule secretion was evaluated by measuring surface expression of CD62P and CD63, which serve as markers for α and δ granules, respectively. The activation status of GPIIbIIIa was assessed using PAC1 antibody. PRP was incubated in non-shaking conditions with TRAP at a concentration of 50 μM, U46619 at 5 µM, or CRP at 9 µg/mL for a duration of 10 minutes. Subsequently, PRP was incubated with the appropriate antibodies at room temperature for 15 minutes. After the addition of phosphate buffer saline (PBS), the samples were subjected to cytometer analysis. The results were expressed as median fluorescence intensity (MFI). The number of copy of GPIa, GPIb, GPIIIa, and GPVI on the platelet surface was quantified using a specific commercial kit (Platelet GP Screen, Biocytex) following the manufacturer’s instructions (22).

The proinflammatory cytokines interleukin-6 (IL-6), interleukin-8 (IL-8), and TNF alpha were quantified in human plasma using micro-ELISA using ELLA technology (ProteinSimple). Soluble P-selectin, and soluble CD40L were measured in human plasma using ELISA kits (Mybiosources, San Diego, USA; Life Technologies SAS, Thermo Fischer, Courtaboeuf, France) following the manufacturer’s instructions. The capillary western blot assay was performed using all reagents provided by ProteinSimple, except for the primary antibodies, which were obtained separately. Platelet homogenates were prepared and analyzed as previously described (30).

For western blotting, washed platelets from patients or controls were subjected to analysis using specific antibodies according to standard procedures.

Platelets in 1 mL of PRP were fixed for 1 hour at 37°C with 2.5% glutaraldehyde in 19 ml of 0.1 M Sorensen phosphate buffer (pH 7.4). TEM was performed and data analyzed as previously described (22).

Platelets (2.10⁸/mL) were prepared and processed as described previously (22). Confocal images were acquired using a 63X objective and an LSM900 confocal laser microscope (Carl Zeiss). Super-resolution imaging was performed using the Airyscan module. For the quantification of colocalization, the FIDJI software with the JACOP plug-in was used to calculate Pearson’s correlation coefficient and Manders’ M1 coefficient (22).

Washed platelets resuspended in modified HEPES Tyrode’s buffer containing 2 mM CaCl2 (2x10⁸ platelets/mL) were incubated in polypropylene tubes with or without (basal) spike proteins from Delta or Omicron variants (5 µg/mL final concentration), for 1, 10, and 30 minutes at 37°C without stirring. Platelets were then analyzed as we recently described (31). FACS analysis was conducted in a single run per primary antibody (up to 16 barcoding possibilities) using a BD LSR Fortessa™ flow cytometer (BD Biosciences). Platelets were gated on their scatter properties and 500,000 events per tube were analyzed for MFI (5-10 minutes per tube).

Statistical analysis was performed using nonparametric tests, including the Mann-Whitney test to compare patients, the Wilcoxon test for experiments with recombinant proteins and the Fisher test for TEM analysis. P<0.05 was statistically significant. GraphPad Prism (La Jolla, CA, USA) was used to realize the figures.

Twenty patients admitted to the ICU from April 12^th, 2021 to May 25^th, 2022 infected with SARS-CoV-2 Delta variant (n=11) or Omicron variant (n=9), as assessed by RT-PCR and spike gene sequencing, were enrolled as well as 10 healthy donors. The biological and clinical data of the patients are provided in Supplementary Table 1 . There was no differences between the groups in terms of patient history and biological data. The SAPS II severity score tended to be higher in patients in the Delta group compared to the Omicron group. Regarding the ARDS severity (PaO2/FiO2 ratio), the Omicron group patients had a lower median ratio than those in the Delta group [89 (11–131) vs. 125 (68–235) p<0.03] at admission ( Table 1 ). Patients in the Delta group were more ventilated and had longer durations of ICU and hospital stay compared to the Omicron group, but their mortality rates were similar ( Supplementary Table 1 ). Consistent with previous findings for the Delta variant (22), the median concentrations of pro-inflammatory cytokines (IL-6, IL-8 and TNF alpha) were significantly higher in patients with Omicron compared to controls ( Table 1 ).

Surface expression of several platelet glycoproteins indicated that, while the level of GPIb remained unchanged, the expression of GPIaIIa, GPIIbIIIa (α_IIbβ3) and GPVI significantly decreased in platelets from severe patients infected with Omicron, suggesting the occurrence of a shedding process ( Table 1 ). Moreover, electron microscopy analysis revealed signs of prior platelet activation. Indeed, platelets from severe COVID-19 patients infected with Omicron had a reduction in the number of microtubules associated with more pseudopodia ( Supplementary Figure 1 ). Soluble GPVI (sGPVI), soluble P-selectin (sCD62P) and soluble CD40 ligand (sCD40L) were significantly increased in the plasma of patients infected with the Omicron variant compared to controls ( Table 1 ). The increase in these specific soluble platelet activation markers was not related to the seropositive or seronegative status of the patients (Supplementary Figure 2 ) and confirmed a shedding process.

At rest, the surface expression of α and dense granules secretion markers P-selectin (CD62P) and CD63, respectively, as well as the activation of αIIbβ3 integrin, did not show significant differences in controls and in patients infected with the Omicron variant ( Table 1 ). However, platelet stimulation by TRAP, U46619 or CRP efficiently increased the membrane expression of CD62P and CD63 as well as αIIbβ3 integrin activation, albeit with a lesser intensity in platelets from patients compared to controls. These results indicate that platelets from severe COVID-19 patients infected with the Omicron variant exhibit signs of preactivation and are partially hyporesponsive, similar to what has been shown previously in severe COVID-19 patients infected with the Delta variant (22, 32).

Infection has been shown to modify the transcriptome of megakaryocytes and platelets, leading to the induction of an antiviral response phenotype (26). For example, interferon (IFN)-induced transmembrane protein 3 (IFITM3) known to inhibit viral replication, is upregulated in megakaryocytes and platelets from patients infected with the Delta variant (17). Interestingly, platelets from patients infected with Omicron also exhibited IFITM3 expression, although to a lesser extent compared to those infected with Delta ( Figure 1A ). Additionally, as markers of Toll-like receptors (TLR) activation, interleukin-1 receptor-associated kinase 4 (IRAK4) was upregulated and phosphorylated in platelets from Delta variant-infected patients, while it showed weak upregulation and poor phosphorylation in platelets from Omicron-infected patients ( Figure 1B ).

Platelets from six severe Omicron-infected patients were analyzed by TEM. Characteristic crown-like structures exhibiting a size consistent with SARS-CoV-2 particles were observed ( Figure 1C ). These viral-like particles were detected in all tested patients but not controls ( Figure 1D ). The percentage of platelets containing viral-like particles was 27.7 ± 3.4%, slightly higher than the percentage we previously reported for the Delta variant (22).

A distinguishing feature of platelets from severe Delta-infected subjects was the existence of enlarged vesicles that were distinct from the open canalicular system (OCS) (22). Electron microcopy analysis of complete platelet cross-sections showed that platelets from Omicron-infected patients exhibited enlarged vesicles that were smaller than those seen in Delta infection ( Figure 1E ). Another difference was the distribution of these vesicles, which were more peripheral and closer to the plasma membrane in platelets from Omicron-infected patients ( Figures 2A, B ). Interestingly, super-resolution confocal microscopy also indicated that spike-positive structures were localized near the plasma membrane in platelets from Omicron-infected patients ( Figures 2C, D ). These findings suggest a distinct mechanism of viral entry or processing in platelets.

We have previously demonstrated that platelets can activate an intrinsic antiviral defense mechanism involving the degradation of viral material through autophagosomes and autophagolysosomes (22). TEM analysis revealed that SARS-CoV-2 Omicron infection also induces selective autophagy in platelets from infected patients, characterized by an increased number of autophagosomes and autophagolysosomes ( Figures 3A–C ). However, compared to Delta, Omicron showed limited efficiency in inducing the formation of elongation membranes (phagophores) ( Figure 3C ) suggesting a different mechanism in the initiation of autophagy. The expression level of LC3B type I and LC3B type II, markers of autophagosomes and autolysosomes, were significantly elevated in platelets from Omicron-infected subjects compared to controls ( Figures 3D, E ).

Super-resolution microscopy showed a weaker colocalization between the spike protein and LC3B in platelets from Omicron-infected patients, compared to Delta. The colocalization between spike protein and EEA1 (early endosome marker) was relatively low and similar, while the colocalization between spike protein and Rab7 (a late endosome marker) was significantly increased in platelets from Omicron-infected patients ( Figures 4A–E ). Compared to Delta variant, involving LC3B-associated phagocytosis and selective autophagy (22), the strong colocalization between spike protein and Rab7 suggests the existence of endosomal pathway-derived elements in platelets from Omicron-infected subjects. Furthermore, spike proteins from Omicron or Delta variants were separately added to platelets from healthy donors. Subcellular localization analysis indicated that, compared to the spike protein from the Delta variant, the spike protein carrying mutations present in Omicron exhibited a significantly higher colocalization with Rab7 and a lower colocalization with LC3B ( Supplementary Figures 3A–D ).

Subsequently, we investigated whether the differences in the intraplatelet processing of the two variants would correspond to distinct signaling potential of their respective spike proteins. Washed control platelets were incubated for 1, 10 and 30 minutes with the two spike proteins in non-shaking conditions, and the phosphorylation of signaling intermediates was assessed by phosphoflow cytometry (31). Both spike proteins induced a significant increase in AKT, P38MAP-kinase, LIM-kinase and SLP76 phosphorylation ( Figure 5 ). However, the signal intensity was lower compared to that elicited by CRP or TRAP. Interestingly, the Delta spike protein exhibited a slow activation of signaling pathways (30 minutes), whereas the Omicron spike protein triggered a much faster response (within 1 minute), suggesting a distinct mode of action associated with the acquisition of Omicron spike protein mutations.