Immunological and Clinical Markers of Post-acute Sequelae of COVID-19: Insights from Mild and Severe Cases 6 Months Post-infection

Abstract

Post-acute sequelae of COVID-19 (PASC) are a complex clinical condition that requires a better understanding of its underlying biological mechanisms. In this study, we assessed hundreds of virological, serological, immunological, and tissue damage biomarkers in two cohorts of patients who had experienced either mild (n = 270) or severe (n = 188) COVID-19, 6 to 9 months post-initial infection, and in which 40% and 57.4% of patients, respectively, developed PASC. Blood analysis showed that the main differences observed in humoral, viral, and biological biomarkers were associated with the initial COVID-19 severity, rather than being specifically linked to PASC. However, patients with PASC displayed altered CD4⁺ and CD8⁺ memory T cell subsets, with higher cytokine-secreting cells and increased terminally differentiated CD45RA⁺ effector memory T cells (TEMRA). Elevated SARS-CoV-2-specific T cells responsive to nucleocapsid/membrane proteins with a TEMRA phenotype were also observed. A random forest model identified these features and initial symptom duration as top variables discriminating PASC, achieving over 80% classification accuracy.

Keywords: SARS‐CoV‐2; T cell phenotyping; biological biomarkers; humoral; postacute sequelae of COVID‐19 (PASC); virological.

FIGURE 1

Persistent symptoms and anti‐RBD IgG levels in patients with or without PASC. Included patients were classified according to their initial form of COVID‐19, whether severe or mild, and the persistence of symptoms (with or without PASC), as indicated. (A) Proportions of PASC in each cohort, and of persistent symptoms self‐reported during the interview. Comparisons were performed using Fisher's exact test. (B) Serum anti‐RBD IgG levels according to the severity of initial COVID‐19 episode. Serum anti‐RBD IgG levels according to PASC status in patients with initial mild (C) and severe (D) COVID‐19 episodes. Results are expressed as BAU/mL and the positivity cut‐off value applied was defined by the manufacturer at 20.33 BAU/mL. Comparisons were performed using the nonparametric Mann–Whitney U‐test. *p < 0.05; **p < 0.01; ***p < 0.001. BAU, binding antibody unit; IgG, immunoglobulin G; RBD, receptor‐binding domain. #Data were missing for two patients in both cohorts.

FIGURE 2

Plasmatic levels of cytokines and tissue damage biomarkers according to PASC status and severity of the initial COVID‐19 episode. Plasmatic levels of IL‐12 (A), IL‐6 (B), IL‐8 (C), IFN‐β₁ (D), IFN‐λ₁ (E), IFN‐γ (F), CXCL9 (G), CXCL10 (H), suPAR (I), VEGFR2 (J), Nf‐L (K), and GFAP (L). Comparisons were performed between patients grouped according to severity of initial COVID‐19 and PASC status (mild COVID‐19: squares, severe COVID‐19: circles, with PASC: black symbols, without PASC: grey symbols). Results were expressed as pg/mL. #Levels were below the limit of detection for IFNβ₁ in 40 of the 85 mild patients and 35 of the 90 severe patients, and for IFNλ₁ in 43 of the 85 mild patients and 28 of the 90 severe patients, respectively. Comparisons were performed using the nonparametric Mann–Whitney U‐test. *p < 0.05; **p < 0.01; ***p < 0.001. GFAP, Glial fibrillary acidic protein; IFN, interferon; IL, interleukin; IP, interferon protein; MIG, monokine induced by gamma interferon; Nf‐L, neurofilament light; suPAR, soluble urokinase plasminogen activator receptor; VEGFR2, vascular endothelial growth factor receptor 2.

FIGURE 3

Flow cytometry analysis of immune subsets from 6 months postinfection. (A) Flowchart of the cytometry data analysis: PBMCs were labeled using high‐dimensional flow cytometry panels, allowing the analysis of main subsets within PBMCs as well as the detailed phenotype of T cells. Manual gating or FlowSOM unsupervised analysis was performed on the flow cytometry analysis platform OmiQ. The differential state of cell clusters was analyzed using the diffcyt algorithm and the R software. (B) Cellular count for the different immune populations defined by manual gating, comparing patients with and without PASC. Statistical analysis was performed using the non‐parametric Mann–Whitney U‐test. *p < 0.05; **p < 0.01; ***p < 0.001. CD, cluster of differentiation; Ig, immunoglobulins; ILC, innate lymphoid cell; NK, natural killer; PASC, postacute sequelae of COVID‐19.

FIGURE 4

Differential T cell profiles in patients with and without PASC. PBMCs from patients with (44 samples) and without (44 samples) PASC were analyzed by spectral flow cytometry using the “T cell” antibody panel. Unsupervised clustering was performed using FlowSOM, and differential expression of phenotypic markers was defined using diffcyt. Results presented in the figure highlight the most significant differences according to PASC status. Among CD4⁺ T cells (A) and CD8⁺ T cells (D) CD45RA/CCR7 expression, abundance of cells per cluster, and (B, C, and E) frequency of cells expressing the indicated markers among identified T cell clusters are shown. Data were represented as mean ± SD. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test. *p < 0.05; **p < 0.01; ***p < 0.001. CD, cluster of differentiation; CXCR, C‐X‐C chemokine receptor; PBMC, peripheral blood mononuclear cells; TEM, Effector memory cells; TEMRA, terminally differentiated effector memory.

FIGURE 5

Distribution and phenotypic characteristics of SARS‐CoV‐2‐specific CD4⁺ and CD8⁺ memory T cell responses. PBMCs from PASC and non‐PASC patients were incubated, or not, with SARS‐CoV‐2 peptide pools from the S or N and M proteins, and T cell cytokine production was analyzed by flow cytometry. (A) Frequency of IFN‐γ⁺/TNF‐α⁺ secreting cells at the basal state or following stimulation with S or N/M peptide pools. (B) Relative proportions of TEMRA, CD57, Tbet, CD27, and CXCR3‐positive cells within N/M‐responsive CD4⁺ T cells according to PASC status. To define the phenotype of SARS‐CoV‐2‐T cells, a cut‐off of 15 events was used to ensure the accuracy of the measurements. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test. *p < 0.05; **p < 0.01; ***p < 0.001. CD, cluster of differentiation; CXCR, C‐X‐C chemokine receptor; IFN, interferon; NM, nucleocapsid and membrane; PBMC, peripheral blood mononuclear cells; S, spike; TEMRA, terminally differentiated effector memory; TNF, tumor necrosis factor.

FIGURE 6

Evaluation of clinical and biological parameters to discriminate patients according to PASC status. (A) A random forest approach was applied using clinical and biological parameters assessed in this study to identify the factors that best discriminate patients with PASC from those without. The top 10 parameters are shown. Predictor variables were ranked based on the Gini Importance index, calculated using the Random Forest algorithm within the R statistical environment. Higher scores indicate a stronger contribution to predicting the dichotomous outcome (PASC vs. non‐PASC). (B) The confusion matrix shows the classification performance of these 10 parameters. The matrix confronts the conditions predicted by the algorithm to the true conditions. Each row of the matrix represents the patient status (with or without PASC), while each column represents the predicted status. Results were expressed as a percentage (%). The mean error rate on the 1000 iterations of the random forest is given for each group of patients. Ag, antigen; CD, cluster of differentiation; CXCR, C‐X‐C chemokine receptor; GFAP, Glial fibrillary acidic protein; Sp, specific TEM, Effector memory cells; TEMRA, terminally differentiated effector memory.