Widespread Immune and Inflammatory Changes Linked to Long-Term Lung Problems After COVID-19

We established a longitudinal research cohort during the Omicron BA.5 outbreak in China in December 2022 [15], conducting 2 follow-up visits at 4 months (4 M, n = 57) and 7 months (7 M, n = 54) postacute infection (Figure 1A and Table 1). At each follow-up, systematic PASC symptom assessments were administered by the respiratory physician, and respiratory parameter evaluations were conducted including pulmonary function tests and chest computed tomography (CT) scan. Among 12 PASC symptoms evaluated, the highest incidence rates were observed for dyspnea (46% at 4 M and 39% at 7 M), fatigue (40% at 4 M and 46% at 7 M), impaired cognitive function (26% at 4 M and 41% at 7 M), and chronic cough (22% at 4 M and 20% at 7 M) (Table 1), which are consistent with the symptom incidence rates reported in previous PASC studies [6, 16]. Further categorizing the symptom clusters into organ/systems, the data indicated that PASC predominantly affects the respiratory system (53% at 4 M and 41% at 7 M) and the nervous system (53% at 4 M and 46% at 7 M) (Figure 1B and Table 1). However, during the timeframe from 4 to 7 M, most rates of PASC organ involvements or symptoms remained persistent without significant relief (Table 1). These data suggest that the potential pathogenic factors leading to PASC may persist, or their residual effects may last for an extended period.

The respiratory system, as the primary site of SARS-CoV-2 invasion, is also a major system affected by PASC. Therefore, we focused on R-PASC and divided the cohort into R-PASC group (n = 30 at 4 M, n = 22 at 7 M) and non-R group (individuals without respiratory sequelae symptoms; n = 27 at 4 M, n = 32 at 7 M) based on the physician-administered PASC symptomatology (Table 1). Then, a subset of hospitalized COVID-19 survivors (n = 39) underwent pulmonary function tests at 4 months (Figure 1C). The results revealed impairments in multiple pulmonary function parameters among patients with R-PASC, who exhibited significantly lower diffusion capacity (DLCO%) and restrictive ventilation parameters (forced vital capacity [FVC%] and first second of forced expiration [FEV1%]) compared to the non-R group (P < .05, Figure 1D–1F), whereas no difference was observed in the obstructive ventilation parameters FEV1/FVC (P = .9108, Figure 1G). Paired data also suggest that the lower pulmonary function in patients with R-PASC was already present during the acute infection phase and remained stable from the acute phase to 4 M for both groups (Supplementary Figure 1A and 1B).

Persistent lung imaging abnormalities in COVID-19 survivors have been reported in multiple studies [17, 18]. We then employed the Chest CT Score [9, 19] to assess the relationship between pulmonary lesions and R-PASC (Figure 1H). The data indicate that patients with R-PASC exhibit prolonged and more severe pulmonary lesions compared to non-R individuals, persisting at least 7 M postinfection (P = .054 at 4 M and P = .008 at 7 M, Figure 1I). Longitudinal imaging investigations revealed that, although both groups exhibited significant resolution of pulmonary lesions from the acute phase to 4 M, the lesion area remained stable with no significant change for both groups from 4 to 7 M (Figure 1J), whereas lesion types transitioned from ground-glass opacity and consolidation during the acute infection to reticulation (pulmonary fibrosis) or complete normal during convalescence (Supplementary Figure 1C).

Notably, regression model analysis demonstrated that symptomatology-based R-PASC grouping was significantly associated with pulmonary function impairment and CT-detectable lung abnormalities, indicating that these symptoms are not entirely subjective but reflect multiple objective physiological dysfunctions (). Supplementary Table 1

To investigate the potential association between R-PASC and inflammatory responses, we conducted longitudinal profiling of 7 inflammatory mediators, while controlling for reinfection confounders through paired XBB-specific neutralization assays [15]. The results revealed that the R-PASC group exhibited significantly elevated levels of interleukin (IL)-6, IL-8, interferon gamma-induced protein (IP)-10, macrophage inflammatory protein (MIP)-1α, and interferon γ (IFNγ) compared to the non-R group at 4 M postacute infection (P < .05, Figure 2A). The inflammatory responses persist over the long term, with significantly higher IL-6 and MIP-1α still observed at 7 M in patients with R-PASC (P < .05), although overall difference levels between the 2 groups have diminished to some extent (Figure 2B). Furthermore, abnormal inflammatory responses in the R-PASC group were significantly correlated with respiratory parameters (Figure 2C, left panel), with IL-6, MIP-1α, and monocyte chemoattractant protein (MCP)-1 positively correlating with more severe pulmonary lesions (Chest CT scores), and IL-6, IL-8, IP-10, MIP-1α, and MCP-1 significant negative correlating with various lung function parameters, whereas this effect was scarcely present in the non-R group (Figure 2C, right panel). Moreover, given that the SARS-CoV-2 virus itself may be a trigger for PASC, we retrospectively assessed the acute upper respiratory viral load in cohort patients (<20 days postsymptom onset) and found that individuals who later developed R-PASC exhibited significantly higher SARS-CoV-2 RNA loads (Figure 2D), suggesting that they experienced a higher dose of antigen exposure.

To profile the SARS-CoV-2–specific T-cell response related to PASC, particularly R-PASC, we subsequently measured the number of virus-specific (IFNγ⁺) T cells in longitudinal peripheral blood mononuclear cells (PBMCs) samples (Figure 3A). The results showed that virus-specific CD4⁺ and CD8⁺ T cells in patients with R-PASC were trendily higher (1.2-1.8-fold in CD4; 2.3-2.7-fold in CD8) than those in the non-R group at both 4 and 7 M (Figure 2B and 2C). Furthermore, we observed that the numbers of IFNγ⁺ CD4⁺ and CD8⁺ T cells increased with the number of organ systems affected by PASC (Figure 3D–3G; grouping details are provided in the Methods). Counts in group 3 (most extensive involvement) were significantly higher than in group 1 (unaffected) at both 4 and 7 M (P < .05) and were elevated compared to group 2 (fewer affected systems) with a trend at 4 M and became significant at 7 M (P < .05).

Multifunctional virus-specific T cells have been reported to mediate the early clearance of SARS-CoV-2 and be associated with mild infections [20]. Our data suggest that, compared to patients with R-PASC, the virus-specific T-cell population in non-R individuals has greater multifunctionality (IFNγ⁺TNF⁺), particularly within the CD4⁺ T cell subsets (P = .0091, Figure 3H). More importantly, total and multifunctional virus-specific T-cell responses, especially CD4⁺ subsets, were significantly positively correlated with improved pulmonary function parameters FEV1% and FVC% in the non-R group rather than the R-PASC group (Figure 3I and 3J, and Supplementary Figure 2A–2D), suggesting a potential contribution to the protection for lung ventilatory function. Multifunctional virus-specific T cells in non-R are also significantly negatively correlated with the inflammatory cytokine IL-8 (Figure 3K). No correlation was observed between virus-specific T cells and the pulmonary lesions in either group at 4 and 7 M (Supplementary Figure 2E and 2F). In addition, using 2 memory markers (CD45RA and CCR7, Figure 3L), we also assessed the memory phenotypes in both groups, finding that patients with R-PASC exhibited significantly lower naïve T cells and higher effector memory T cells in both total and virus-specific T-cell subsets compared to the non-R group (Figure 3M–3P).

We subsequently investigated the role of antiviral antibodies in PASC and found that the titer levels of the BA.5 neutralizing antibody (Nab) and cross-reactive XBB.1.9 Nab were significantly higher in the R-PASC group compared to the non-R group at 4 M (Figure 4A and 4B). However, for wild-type Nab (Figure 4C), which is commonly regarded as a baseline type associated with previous vaccination, as well as the binding antibodies S + N IgG (Figure 4D), no significant differences were observed between the 2 groups at 4 M. The findings indicate enhanced variant-specific humoral responses in early-stage R-PASC, likely related to their ongoing immune reactions, but not across all antibody types. Notably, for the long-term time point of 7 M, no differences were found in any antibody type between the R-PASC and non-R groups (Figure 4A–4D), nor between the groups classified by the number of PASC-affected systems (Supplementary Figure 3A–3C). Furthermore, we analyzed the correlation between antiviral antibody responses and respiratory parameters or inflammatory responses, finding that none of the antibody components showed significant correlations with lung function parameters or the pulmonary lesions in both groups (Figure 4E and 5F blue frame, Supplementary Figure 3D–3H), which further suggests that SARS-CoV-2 antibodies may play a minimal or nondirect role in the R-PASC pathophysiology.

The complement system is another important immunological factor in the PASC pathophysiology. We assessed multiple complement proteins in cohort serum (Figure 5A) and found that patients with R-PASC exhibited a wide and excessive complement activation compared to non-R individuals at 4 M (Figure 5B), with significantly elevated (P < .05) markers in terminal pathway (sC5b-9, C5a, C3a), the classical pathway (C1q), and the alternative pathway (Factor B), whereas the lectin pathway (mannose-binding lectin [MBL]) showed no significant differences between the 2 groups (P = .462). Correlation analyses between the complement profiles and respiratory parameters revealed excessively activated sC5b-9 and C5a in the R-PASC group negatively correlated with pulmonary function (DLCO% and FVC%) (P < .05), whereas no correlation of complement responses and lung function impairment was found in the non-R group (Figure 5C and 5D). For lung tissue lesions, both groups exhibited a positive correlation between sC5b-9, C5a, and chest CT scores (P < .05, Figure 5C and 5E), which is consistent with previous studies that observed terminal complement components causing lung injury in hospitalized patients with COVID-19 [21]. Concurrently, the complement components in both groups were positively associated with the inflammatory responses to some extent (Figure 5F), suggesting a potential mechanism underlying their injury effects.

Furthermore, as predicted, we observed that as the number of PASC-affected organ systems increased, the activation degree of the complement system correspondingly escalated (Figure 5G–5I and Supplementary Figure 3G and 3H), with levels of broad markers in group 3 and group 2 significantly higher than those in group 1 (P < .05). Interestingly, the key component of the lectin pathway, MBL, exhibits a distinct role: its level is highest in group 1, which has the least number of affected organ systems (Figure 5J), and shows positive correlation with improvements in pulmonary function, lung tissue lesions (Figure 5K and 5L), and demonstrates negative correlation with inflammatory response in R-PASC (Figure 5M). Overall, our data suggest that prolonged activation of the complement response, particularly via the terminal, classical, and alternative pathways, may contribute to the progression of PASC, especially R-PASC, while the lectin pathway may play a protective role.

Previous evidence indicates that severe SARS-CoV-2 infection during the acute phase is associated with the elevated risk of subsequent PASC among hospitalized patients [2, 6, 22]. Therefore, we further analyzed the clinical data from the acute infection phase of individuals with the subsequent respiratory subtype of PASC, and found that compared to the non-R group, patients with R-PASC had higher rates of hospitalization, greater disease severity, and increased requirements for oxygen supplementation (Supplementary Table 2). In addition, a retrospective analysis of clinical laboratory data from the acute infection phase was also performed, regarding parameters including induced sputum cytology, fractional exhaled nitric oxide, complete blood count, blood biochemistry, coagulation profile, and autoantibody profile. Most measured parameters were comparable between the 2 groups (Supplementary Table 2), with the exception of significantly higher lactate dehydrogenase levels in the R-PASC cohort—a marker associated with persistent airway injury following SARS-CoV-2 infection [23]. Collectively, these findings suggest that the acute-phase disease status and related laboratory parameters may serve both biological and predictive roles in the pathophysiological process of R-PASC.

This prospective cohort study was conducted at The First People's Hospital of Yunnan Province during China's Omicron BA.5 outbreak. Our cohort primarily consisted of convalescent individuals who were hospitalized following their first SARS-CoV-2 infection and had previously received 2 to 3 doses of inactivated vaccines (). Follow-up evaluations were performed at 4 and 7 M after initial infection. Each follow-up included: (1) physician-administered PASC symptom assessments via the PASC Symptom Assessment Form, (2) peripheral blood sample collection, and (3) clinical respiratory evaluations (pulmonary function tests and chest CT). Participation in follow-ups and respiratory assessments was voluntary. Supplementary Table 3

PASC were defined according to the World Health Organization criteria (October 2021) [14]: (1) confirmed SARS-CoV-2 infection by quantitative reverse transcription polymerase chain reaction (PCR) or antigen testing; (2) symptoms emerging or worsening at least 3 months after COVID-19 onset, persisting for ≥2 months; and (3) exclusion of alternative diagnoses through clinical evaluation and medical history. Participants were categorized into organ/system-specific PASC groups based on the sustained symptoms aligned with corresponding physiological systems [38]: (1) respiratory system PASC (chronic cough, dyspnea); (2) cardiovascular system PASC (palpitation, chest pain); (3) neurological PASC (headache, cognitive impairment, fatigue, insomnia); (4) digestive system PASC (nausea, anorexia); (5) sensory system PASC (hyposmia, hypogeusia); and (6) musculoskeletal system PASC (myalgia, arthralgia). Notably, assignment to a specific organ-system PASC group (eg, respiratory PASC) was based on the presence of defining symptoms for each system, without excluding individuals with concurrent symptoms from other systems. This approach was chosen to reflect the clinical reality of frequent symptom overlap (Supplementary Table 4) in PASC while allowing us to focus on the pathophysiology of the targeted PASC system.

Serial Chest CT was performed for hospitalized COVID-19 convalescents in acute-phase, 4 and 7 M with commercial multidetector CT scanners (SOMATOM, Siemens, Germany). Two certified radiologists independently analyzed CT images using a semiquantitative scoring system (Chest CT Score) validated in prior studies [17, 18, 39, 40]. Each lung was divided into 5 lobes (right upper, middle, lower; left upper, lower), with individual lobe scores assigned as follows: 0 (no lesions), 1 (<5% involvement), 2 (5–25%), 3 (26–50%), 4 (51–75%), or 5 (>75%). Total scores ranged from 0 to 25, reflecting overall disease burden. Lesion morphology was characterized per Fleischner Society guidelines [41], including ground-glass opacity, consolidation, and fibrosis.

Lung function assessments were conducted during acute SARS-CoV-2 infection and at 4 M after infection, with voluntary participation in follow-up testing. Measurements were performed in a standardized laboratory using a Masterscreen Jaeger spirometer (Germany), adhering to American Thoracic Society/European Respiratory Society guidelines [42]. Participants underwent 3 repeated tests at 5-minute intervals to minimize fatigue effects. Analyzed parameters included: FVC%, FEV1%, FVC/FEV1 ratio (ventilation), and DLCO%.

The nasopharyngeal swab samples collected during the acute phase of SARS-CoV-2 infection were tested for viral RNA using a commercial kit (Shengxiong Biotechnology Co., Ltd., Changsha, China) via quantitative reverse transcription PCR. This kit targets the ORF1ab and N genes and has been approved by the National Medical Products Administration of China. Samples were included for analysis within 20 days postsymptom onset.

To assess viral-neutralizing capacity, plasma specimens underwent analysis via a cytopathic response assay conducted in an certified biosafety level 3 facility, as described previously [43]. In brief, serially diluted plasma samples were incubated with viral suspensions (SARS-CoV-2 Omicron BA.5 or XBB 1.9 subvariants) before inoculation onto VERO E6 cells. After a 4-day culture, neutralizing antibody titers were quantified through automated cytopathic effect analysis (Celigo Imaging System, Nexcelom Bioscience), with geometric mean values calculated from duplicate measurements relative to virus and cell controls.

Quantification of SARS-CoV-2–specific IgG antibodies targeting both spike (S) and nucleocapsid (N) proteins (anti-S + N IgG) was conducted using a chemiluminescent immunoassay platform (iFlash 3000, Yhlo Biotech, Shenzhen, China) in accordance with the manufacturer's instructions, as described previously [44].

Plasma concentrations of inflammatory cytokines and chemokines (IL-6, IL-8, IP-10, MIP-1α, MCP-1, IFNγ, TNF) were quantified using the Cytometric Bead Array Human Soluble Protein Flex Set System (BD Biosciences, USA) following manufacturer protocols. Briefly, following serum centrifugation to remove debris, supernatants were analyzed alongside serially diluted cytokine standards. Antibody-conjugated magnetic beads were incubated with samples and detection reagents, followed by flow cytometric analysis (BD Fortessa). Absolute concentrations were determined via FCAP Array software using manufacturer-defined standard curves.

PBMCs stimulation protocols employing SARS-CoV-2 peptide pool mixtures were implemented following established methodologies [15, 43]. In brief, PBMCs were activated ex vivo with SARS-CoV-2 peptide pools (250 nM/peptide), recombinant IL-2 (20 IU/mL; PeproTech, USA), and protein transport inhibitor GolgiPlug (1 μM; BD Biosciences, USA) for 16 hours.

Following overnight culture, cells were stained with Live/Dead-FVS440 (BD, 1:1000, #566332) for viability assessment, followed by surface marker labeling using CD3-BUV395 (BD, 1:200, #564001, SK7), CD4-APC-H7 (BD, 1:150, #560158, RPA-T4), CD8-PerCP-Cy5.5 (BioLegend, 1:200, #344710, SK1), CCR7-APC (BioLegend, 1:150, #353214, G043H7), and CD45RA-FITC (BioLegend, 1:200, #304106, HI100). After fixation/permeabilization (BD Cytofix/Cytoperm), intracellular cytokines were detected with IFN-γ-BV421 (BioLegend, 1:200, #506538, B27) and TNF-α-PE (BD, 1:200, #559321, MAb11). Washed cells were analyzed on a BD Fortessa X-20 flow cytometer, with data processed using FlowJo v10.8.

The levels of complement system components (Sc5b-9, C3a, C5a, C1q, MBL, and Factor B) were quantitatively analyzed in cell-free plasma samples using commercially available enzyme-linked immunosorbent assay kits (Jiangsu Meimian Industrial Co., Ltd., China). Following standardized enzyme-linked immunosorbent assay protocols, diluted plasma samples and reference standards were loaded into designated wells, followed by sequential incubation with horseradish peroxidase–conjugated detection antibodies. After chromogenic reactions initiated by substrate addition, optical density at 450 nm was measured using a Multiskan GO microplate spectrophotometer (Thermo Fisher Scientific, USA).

All data analyses were conducted using GraphPad Prism (v9.5.1) and SPSS (v26), with flow cytometry data processed in FlowJo. Nonnormally distributed data were analyzed using 2-tailed Mann-Whitney U tests, with results presented as median and interquartile range. Normally distributed data were compared via Student t-test and expressed as mean ± standard error of the mean. Antibody titers were reported as geometric mean titers with 95% confidence intervals. Paired nonparametric comparisons used the Wilcoxon signed-rank test. CT pulmonary feature changes between timepoints were assessed using the chi-squared test. Correlations were assessed using Spearman's rank correlation coefficient and linear regression analysis. Statistical significance thresholds were defined as P < .05 (*), P < .01 (**), and P < .001 (***).