COVID-19 is an endothelial disease with two key aspects: endotheliopathy causing coagulopathy and angiogenic disorder, particularly in the lungs [1,2,3]. The condition known as Long COVID, also termed post-COVID-19 condition (PCC) by the WHO [4,5,6], is characterized by the persistence or development of new symptoms at least three months after the initial SARS-CoV-2 infection, without any other discernible explanation. We recently reported that in patients with impaired respiratory function, the angiogenic biomarker VEGF-A was linked to impaired lung function parameters [7]. However, no correlation has been found between circulating plasma biomarkers and persistent clinical symptoms such as fatigue and dyspnea. Circulating endothelial cells (CECs) have been associated with pulmonary vascular disorders, including increased lesions or the proliferation of the endothelium, as well as acute COVID-19 severity [8]. Hence, this study aimed to explore the possible link between CECs and persistent clinical symptoms, lung function, as well as additional plasma biomarkers after the acute phase of SARS-CoV-2 infection.

Our research project was a prospective, single-center cohort analysis conducted at the Georges Pompidou European Hospital as previously described [7]. We defined long COVID/PCC as the occurrence of one or more lingering new symptoms occurring at least three months post-COVID-19 infection that could not be attributed to any pre-existing conditions or other diagnoses. Participants were systematically recruited from November 2020 to June 2022, between three to 24 months following their initial COVID-19 episode. This investigation is a segment of the SARCODO cohort study (2020-A01048-31 A, NCT04624997). CECs were measured on whole blood (EDTA tubes) by immunomagnetic separation as reported earlier [8,9,10]. In the univariate analysis, the OR and 95% CIs of the association between concentration of CECs and clinical symptoms were assessed with separate binary logistic regression models. Then, in a multivariate analysis, models were adjusted for sociodemographic factors along with all factors with a P-value < 0.2 in the univariate analysis. A random effect for patient visit was added (generalized estimated equation models) to account for multiple measures in some patients and inter-time period variability. A P-value < 0.05 was considered significant. Analysis was performed with R software (version 4.1.2 (2021-11-01)).

The study included 137 patients with long COVID/PCC, mostly male (68%), with a median age of 55 years, as previously described [7]. A total of 194 CECs measurements were performed between months 3 and 24 after an episode of SARS-CoV-2 infection. Among the 137 patients at the time of inclusion, 89 (64.9%) reported experiencing persistent fatigue, 91 (66.4%) dyspnea. Pooling all patient visits together, among the 194 CECs measurements, 100 (51.4%) were above the cutoff of 10 described previously in the consensus as a cut-off for normal values (Fig. 1A). No significant differences were noted based on the time of sampling (between months 3 and 24, Fig. 1B). Although no significant differences in CEC levels were found in patients with dyspnea, impaired diffusing capacity of the lung for carbon monoxide (DLCO), or abnormal pulmonary function tests (PFTs), a notable increase in CEC levels was observed in patients with persistent fatigue (median:14.0 CECs/mL IQR [5.50–24.50]) compared to those without fatigue (median: 6.0 CECs/mL IQR [3.0–15.0], p = 0.008, Fig. 1C). As demonstrated in Fig. 1D, persistent fatigue was associated with higher CEC levels in patients who have been hospitalized during acute phase of COVID-19, regardless of whether the initial severity was critical (ARDS) or non-critical. This association was not significant when considering outpatients only (p = 0.06). No difference was observed between patients with fatigue according to non-ARDS or ARDS status (p = 0.08).

Circulating Endothelial Cells (CECs) in long COVID Patients. A CECs Counts Across Study Participants: Scatter plot showing CECs/mL in study participants. Each dot represents an individual participant’s CECs count. The green dashed line represents the normal CECs value as described in the consensus paper. B Longitudinal Analysis of CECs Over Time: Scatter plot showing CECs/mL measured at different time points (M3, M6, M9, M12, M24) post-illness. No significant differences were observed over time (p = 0.548). C CECs Counts in Relation to Clinical Symptoms and Pulmonary Function: Scatter plots showing CECs/mL in relation to dyspnea, fatigue, DLCO (diffusing capacity of the lungs for carbon monoxide) percentages, and PFT (pulmonary function test) results. Fatigue + group shows a significant difference (p = 0.008), while other comparisons are not significant. D CECs in Relation to Disease Severity and Fatigue: Scatter plot showing CECs/mL in outpatients, non-ARDS (acute respiratory distress syndrome), and ARDS groups, with further stratification by presence of fatigue. Significant differences were found in non-ARDS (p = 0.009) and ARDS (p = 0.0002) groups with fatigue+. E Correlation Between DLCO and VEGF-A Levels: Scatter plot showing a negative correlation between DLCO (%) and VEGF-A (pg/mL) levels (r = -0.30, p < 0.001). F Correlation Between DLCO and CECs Counts: Scatter plot showing no significant correlation between DLCO (%) and CECs/mL (r = -0.05, p = 0.421). G Correlation Between VEGF-A and CECs Counts: Scatter plot showing no significant correlation between VEGF-A (pg/mL) and CECs/mL (r = -0.05, p = 0.512). H: Multivariate analysis of persistent fatigue showing an association with CECs independently of age, sex, severity and DLCO in patients initially hospitalized for a COVID-19

We previously demonstrated that increased VEGF-A levels in plasma were associated with impaired lung function parameters including DLCO in patients with long COVID (Fig. 1E), whereas no correlation was observed between plasma biomarkers and clinical symptoms, particularly dyspnea or persistent fatigue [7]. Then we further explored the association between CEC with VEGF-A and DLCO. The Spearman correlation between CECs and DLCO or VEGF-A in the entire cohort was not significant (p = 0.512 for VEGF-A and p = 0.421 for DLCO; Fig. 1F and G). Moreover, in a univariate logistic regression model (Table 1), persistent fatigue was significantly associated with increases in CECs (above 10 cells/mL, as previously described in the consensus of the technique [10]) but also with higher CRP levels, higher age and male sex. Finally, for patients previously hospitalized during acute phase of COVID-19 in whom an association between persistent fatigue and CEC levels was found, a multivariate analysis was performed to adjust for potential confounding variables, such as respiratory or other post-ARDS sequelae. The factors analyzed include age, sex, CEC levels, DLCO and initial severity (ARDS vs. non-ARDS). This multivariate analysis (Fig. 1H) showed that impaired DLCO (OR: 0.57, 95% CI: 0.25-1.3O, P = 0.11) was not significantly associated with persistent fatigue, whereas CECs ≥ 10 (OR: 5.16, 95% CI: 2.27–11.75, P < 0.001) were strongly associated with it. This indicates that higher levels of CECs are significantly associated with increased odds of persistent fatigue, even when accounting for age, sex, initial severity and DLCO in patients who initially experienced a severe COVID-19.

Several research groups have indicated that persistent endothelial dysfunction plays a significant role in long COVID. Plasma von Willebrand factor (VWF): Ag and VWF propeptide levels have been inversely correlated with 6-min walk tests in a cohort of 50 patients [11]. Jacobs et al. discovered that 24 months after COVID-19, among 145 individuals (51% of whom were hospitalized), long COVID symptoms, particularly fatigue and dyspnea, were associated with endothelial dysfunction, with a notable link to elevated endothelin-1 levels [12]. Recently, two high-throughput proteomics approaches have been applied in a prospective cohort of 113 patients with long COVID (70% having required hospitalization for severe acute COVID-19) [13]. They identified persistent complement-mediated immunopathology with increased VWF levels in long COVID patients, unaffected by fatigue [13]. The TUN-EndCOV study revealed that dysfunction of the endothelium emerged as an independent predictor for long COVID with non-respiratory symptoms [14]. We previously described that increased VEGF-A levels were associated with impaired DLCO and pulmonary dysfunction in long COVID, but angiogenic biomarkers were not significantly associated with clinical symptoms like dyspnea or fatigue [7]. The results of the present study describe, in patients previously hospitalized in acute phase of COVID-19, the association between CECs and persistent fatigue independent of dyspnea, impaired DLCO or relationships with VEGF-A. The TUN-EndCOV study and our findings align, suggesting two forms of long COVID linked to endothelial dysfunction: respiratory symptoms likely tied to abnormal angiogenesis, and non-respiratory symptoms like fatigue associated with endothelial dysfunction. Our study had limitations, including insufficient data on non-respiratory symptoms besides fatigue, lack of detailed cognitive symptom questionnaires, and no record of patients’ physical activity levels, which could confound the association between chronic fatigue and endothelial dysfunction [15,16,17].

All in all, our study reveals a link between persistent fatigue post-COVID-19 and endothelial dysfunction, especially in patients hospitalized for severe cases. Elevated CECs and fatigue suggest a distinct long COVID profile driven by vascular issues. These findings were absent in outpatients with normal lung function. We recommend a nuanced approach to managing long COVID, focusing on vascular health and targeted interventions for those hospitalized with pulmonary COVID-19.