A consecutive sample of AIS patients (>18 years of age) with COVID-19, admitted to the emergency department of our teaching hospital within 12 h of symptom onset, from March 2020 to May 2021, was included in the study. Diagnosis of SARS-CoV-2 infection was made by COVID-19 real-time reverse transcriptase-polymerase chain reaction (rRT-PCR) test on nasopharyngeal swab. For each patient, we collected demographic data, pre-stroke mRS, past medical history, vascular risk factors, and pre-stroke medications. Pre-stroke functional ability and stroke severity were assessed by using the modified Rankin scale (mRS) and National Institutes of Health Stroke Scale (NIHSS), respectively. COVID-19 severity was assessed by the Sequential Organ Failure Assessment (SOFA) score with a score of >1 indicating a severe disease (15).

As controls, we included a cohort of AIS patients without COVID-19, mostly admitted during the pre-pandemic period, whose clinical/radiological data were previously recorded in a dedicated database and whose serum and plasma samples were stored at −80°C for research purposes after obtaining patient informed consent.

Eligible patients were treated with IV thrombolysis or mechanical thrombectomy according to the current national and international guidelines (16–18). In both cases and controls, brain computed tomography (CT)/CT-Angiography (CTA) were performed in the acute phase, while multimodal 1.5 Tesla Magnetic Resonance Imaging (MRI) was obtained, where applicable, within 24 h from symptom onset. Neuroimaging data (including type of occluded vessel, infarct side and volume, presence of hemorrhagic transformation, collateral status by the Menon regional leptomeningeal score on CTA (19), and vessel recanalization assessed by using the Thrombolysis in Cerebral Infarction [TICI] score for patients receiving mechanical thrombectomy) was analyzed by two independent expert neuroradiologists. In particular, infarct volume was calculated on the MRI T2-FLAIR images through the A^*B^*C/2 formula (20). When it was not possible to perform MRI, CT scan was used as an alternative. Regarding collaterals, Menon scores of 0–10, 11–16, or 17–20 indicated poor, moderate, or good collateral circulation, respectively. Chest CT scan was also performed in all COVID-19 patients to evaluate the presence of SARS-CoV-2-related pulmonary infection.

Trial of ORG 10172 in Acute Stroke Treatment (TOAST) criteria were used to categorize stroke etiology (21).

All study participants gave a written informed consent. The study was approved by the Ethical Committee of our Institution and was conducted according to the Declaration of Helsinki.

Venous blood samples for routine examination were collected at admission. Due to organizational issues, it was possible to perform a supplementary battery of blood tests only in a subset of 10 non-consecutive AIS patients per group. In these patients we analyzed coagulation parameters and biomarkers of endothelium dysfunction as follows: coagulation factors VIII and XIII; total levels of von Willebrand factor antigen (vWF:Ag), and its capability to adhere to platelet glycoprotein complex GPIb-IX-V (VWF:RCo). Furthermore, we also measured the following serum antibodies to PF4/polyanion by using a commercial enzyme immunoassay (IgG/IgA/IgM, Immucor, Lifecodes, Waukesha, WI, cat. # X-HAT45); serum SP by ELISA (My BioSource, San Diego, CA, USA, cat.#MBS7608267); Nucleocapside protein (NP) in the COVID-19 patient group alone by CLEIA (Lumipulse SARS-CoV 2-Ag—CND W0105040619—Fujirebio, cat.# 260340).

Antibody titers were measured using the LIAISON^® SARS-CoV-2 Trimeric S-IgG kit (DiaSorin S.p.A., Saluggia, Italy, cat# 311510) an indirect chemiluminescence immunoassay (CLIA) technology for the detection of serum IgG antibodies directed against SARS-CoV-2 trimeric SP. The assay quantification range is between 4.81 and 2080 BAU/mL (Binding Antibody Units/ml), with a cut-off value of 33.8 BAU/ml.

Washed platelets [from 3 healthy donors (HDs)] were obtained by serial centrifugation and combined to a final concentration of 5 × 10⁸ cells/ml. Sera obtained from stroke patients or HDs were inactivated at 56°C for 30 min. After cooling and a short spin, sera were treated or not with 4 μg/ml of an antibody directed against the SARS-CoV-2 SP S1 subunit (Invitrogen, clone P06DHuRb) for 15 min at 37°C and then incubated for 10 min with washed platelets, preincubated or not with 20 μg/ml of the FcγIIa receptor-blocking antibody (Boster Biological Technology, clone IV.3). The platelet adhesive and pro-coagulant capacity were assessed on a BD Accuri C6 flow cytometer, by measuring the binding of PAC1-FITC (BD Bioscience), that recognizes the active form of integrin αIIbβ3, and of Annexin V-PE (Sony Biotechnology) that binds phosphatidylserine (Supplementary Figure S1).

We performed a descriptive analysis; continuous variables were expressed as mean (±SD) or median (interquartile range [IQR]) as appropriate, while dichotomous and categorical variables were reported as counts, and proportion were calculated by dividing the numbers of events by the total number of patients, excluding missing or unknown cases. Student's t-test or Mann-Whitney U test and χ² or Fisher exact test were used as appropriate for the comparison of demographics, clinical, radiological, and laboratory characteristics between the two groups of AIS patients with and without COVID-19. Correlation coefficients (Spearman ρ) were derived to quantify the association between biomarkers and infarct volume. Multivariate linear regression models (including a clinical model and clinical/biomarkers model) were used to adjust for the effects of potential confounders and to evaluate whether the presence of COVID-19 was an independent predictor of stroke lesion volume. Variables with p-values < 0.05 at univariate analysis for infarct volume were included in the multivariate models as well as other variables potentially influencing infarct volume regardless their univariate p-value. Given the small sample of patients included in the study, multivariate analysis should be considered as exploratory with a potential risk of model overfitting and the results should be interpreted with caution and deemed as not conclusive.

Statistical significance level was set at p < 0.05 for all analyses. Statistical analysis was performed using SPSS statistical software (IBM Corp, SPSS Statistics for Windows, Version 25, Armonk, NY).

A total of 39 AIS patients were enrolled in this study: 22 (female 5; mean ± SD age 68.4 ± 16.1 years) with and 17 (female 5; mean ± SD age 71.4 ± 14 years) without COVID-19. Clinical and demographic characteristics are summarized in Table 1. Vascular risk factors were similar between the two groups except for a significant higher prevalence of smoking and atrial fibrillation among patients without COVID-19 (52.9 vs. 4.5%; p = 0.001 and 52.9 vs. 18.2%; p = 0.022, respectively). Overall, the most frequently reported stroke risk factor was arterial hypertension with similar proportions in both COVID-19 and non-COVID-19 patients (p = 0.704).

COVID-19 patients had stroke on awakening or of unknown time of onset significantly more frequently than non-COVID-19 patients (63.6 [14/22] vs. 17.6% [3/17], p = 0.004). The baseline median (IQR) NIHSS score was 15 [5–21.0]) with similar degree in both groups (11.50 [3.50–21.25] AIS with COVID-19 vs. 16 [10.50–20.50] AIS non-COVID-19, p = 0.263).

According to the TOAST classification, cardioembolic stroke tended to be more prevalent in non-COVID-19 compared to COVID-19 patients (52.9 [9/17] vs. 22.7% [5/22], p = 0.054); conversely, the latter had more frequently a cryptogenic stroke (50.0 [11/22] vs. 5.9% [1/17], p = 0.004) (Table 1).

Patients with COVID-19 received less frequently IV tissue-type plasminogen activator (tPA) (4.5 [1/22] vs. 64.7% [11/17], p < 0.001) or mechanical thrombectomy (22.7 [5/22] vs. 64.7% [11/17], p = 0.008) with longer median time intervals from stroke onset to endovascular treatment (581 vs. 265 min., p = 0.036) (Table 1).

Among the COVID-19 AIS patients, the mean time from COVID-19 diagnosis and stroke onset was 4.6 days (±7.1 SD) (Supplementary Table S1). Typical ground glass pneumonia was detected in 17/22 (77.3%) COVID-19 patients, with 7/17 (41.2%) exhibiting more than 50% of lung parenchyma involvement. None of the COVID-19 patients met the sepsis criteria according to the Sepsis-3 International Consensus (15), while 8/19 (42.1%) of them had a SOFA score ≥ 1.

Only one patient had received the first dose vaccination against SARS-CoV-2 (Pfizer BioNTech) 10 days before stroke onset. The patient had also a pneumonia and SOFA score of 2.

Demographics, clinical characteristics, and laboratory data of a subset of 10 non-consecutive AIS patients with COVID-19 for whom additional specific biomarkers were measured, were overall similar to those of the remaining 12 COVID-19 patients except for a higher COVID-19 severity (SOFA score > 1: 75 [6/8] vs. 18.2% [2/11], p = 0.024; mean [SD] SOFA score: 2.2 [1.4] vs. 0.82 [1.0], respectively) (Supplementary Table S2).

Concerning the routine laboratory tests at admission, in the COVID-19 cohort we observed significantly lower lymphocyte count compared with the non-COVID-19 counterpart (mean [SD] 1.35 [0.92] × 10³/μL vs. 2.30 [1.25] × 10³/μL; p = 0.035) and significantly higher levels of LDH (461.56 [404.06] vs. 219.57 [47.18] mg/dL; p = 0.034) (Supplementary Table S3).

As regards the coagulation profile, fibrinogen levels were significantly higher in COVID-19 patients than in the non-COVID-19 patients (mean [SD], 486.25 [92.69] vs. 405.21 [108.83] mg/dL; p = 0.026) (Supplementary Table S3 and Figure 1A). Based on the available data, D-Dimer levels were elevated at admission in the majority of both COVID-19 and non-COVID-19 patients, however, they were not significantly different between the two groups (p = 0.150) (Supplementary Table S3). INR and aPTT were normal.

In a subset of 10 AIS patients per group, coagulation Factor VIII was increased in cases compared to controls (196.18 [132.84] vs. 87.89 [52.79]%; p = 0.031) (Supplementary Table S3 and Figure 1B) whereas Factor XIII was within the normal range in both patient groups. Overall, we observed increased circulating levels of VWF:Ag and VWF:RCo, which were however higher in the COVID-19 group than in the non-COVID-19 group with a statistically significant difference reached for the VWF:Ag levels (mean [SD], 275.39 [184.97]% p = 0.041) (Supplementary Table S3 and Figure 1C).

Admission levels of Factor VIII and vWF:Ag positively and significantly correlated with infarct volume in the overall study population (Spearman's rho = 0.539, p = 0.026 and 0.697, p = 0.003, respectively) (Supplementary Figure S2), as well as in the two COVID-19 and non-COVID-19 patients although without reaching statistical significance.

Only 2 out of the 10 AIS/COVID patients with available additional biomarkers showed a measurable level of SP in the serum. Overall, no substantial differences in demographics, clinical and laboratory data were observed between SP-positive and SP-negative patients (Supplementary Table S4).

We could not detect NP in the serum of both cases and controls.

Among the 4 AIS with COVID-19 screened for the IgG against SARS-CoV-2 SP level, only 1 patient, who received 10 days earlier, the first dose of Pfizer BioNTech vaccine against COVID-19, showed high antibody titer (Supplementary Table S1).

Patients' serum of both groups did not show detectable levels of pan antibodies (IgG, IgM and IgA) to PF4-polyanion complexes.

Plasma from AIS patients without COVID-19 induced a significant increase in the basal activation of integrin αIIbβ3 in the platelets of HCs, while plasma from AIS patients with COVID-19 had a modest non-significant effect (Figure 2A). Conversely, plasma from stroke patients either with or without COVID-19 induced a significant increase of phosphatidylserine exposure, which is indicative of a pro-coagulant activity (Figure 2B). Pre-incubation with a blocking antibody for the S1 protein or inhibition of the FcγRIIA, the platelet receptor for immune complexes, did not interfere with the upregulation of the platelet function (Figures 2C, D).

Neuroimaging data are included in Table 1. As regards the infarct volume, we found a borderline statistically significant between-group difference with a median (IQR) infarct volume of 25.95 cm³ (6.80–136.88) in the COVID-19 patients and of 9.0 cm³ (3.08, 25.80) in the non-COVID-19 counterpart (p = 0.123) (Supplementary Figure S3).

Lepto-meningeal collaterals significantly differ between the two groups (p = 0.046) with COVID-19 patients having moderate collateral circulation (Menon score 11–16: 3/3, 100% vs. 3/13, 23.1%; p = 0.016) and non-COVID-19 patients having a larger variety of Menon scores from good (Menon score 17–20: 6/13, 46.2%) to poor (Menon score 0–10: 4/13, 30.8%) collateral status. There were no statistically significant between-group differences in terms of recanalization rate. Hemorrhagic transformation of the index infarct was found in 3/22 (13.6%) COVID-19 patients and in 6/17 (35.3) AIS patients without COVID-19 (p = 0.142).

At univariate analyses infarct volume resulted associated with the diagnosis of COVID-19 with a trend toward the statistical significance (p = 0.123) and significantly associated with the presence of internal carotid artery stenosis >50% (p = 0.009) and a clinical history of valvulopathy (p = 0.035). Furthermore, infarct volume correlated with baseline NIHSS (Spearman's rho = 0.579, p < 0.001), admission levels of endothelium dysfunction biomarkers (FVIII: Spearman's rho = 0.539, p = 0.026; vWF:Ag: Spearman's rho = 0.697, p = 0.003; vWF:RCo: Spearman's rho = 0.547, p = 0.023) and platelet count (Spearman's rho = 0.511, p = 0.002).

In multivariate linear regression analyses, COVID-19 resulted as an independent predictor of infarct volume both in the clinical (B 20.318, Beta 0.576, 95% CI 6.077–34.559; p = 0.011) and in the clinical/biomarker (B 35.440, Beta 0.837, 95% CI 3.320–67.560; p = 0.038) models (Table 2).