We enrolled 82 consecutive participants, aged 18 or older, who survived severe ARDS associated with COVID-19, meeting Berlin criteria and requiring invasive mechanical ventilation for at least 7 days and who were discharged from our hospital between February 2020 and January 2021. SARS-CoV-2 infection was confirmed by positive reverse-transcriptase polymerase chain reaction on nasal swab or tracheal aspirate at the time of ARDS. Patients were associated to COVID-19 initial Wuhan wave and were recruited six-months after ICU-discharge between September 2020 and July 2021 ( Supplementary Figure S1 ). Detailed selection criteria can be found in the Supplementary Methods section. Written consent was obtained from all participants, and the study received approval from the institutional Ethics Committee at La Paz University Hospital (PI-4189).

Demographic, clinical, and ICU management data were recorded. At 6-months post-ICU discharge, anthropometric parameters, smoking status, comorbidities, and current treatment were documented. Respiratory symptoms were assessed using the European Community for Coal and Steel Questionnaire and the modified Medical Research Council dyspnea scale. Prior to lung function tests, a venous blood sample was drawn and EBC was sampled during 10 min of relaxed tidal breathing. Spirometry, lung volumes and diffusing capacity of the lungs for carbon monoxide (DLCO) measurements were conducted using MasterScreen (Viasys, CareFusion, Germany) according to current standardization. Lung diffusing capacity for nitric oxide was also measured following ERS recommendations (9). The membrane component of diffusing capacity (DmCO) and the pulmonary capillary blood volume (Vc) were calculated, and reference equations proposed by Zavorsky et al. were used (9). Lung morphology was assessed with a non-contrast chest CT scan with a quantitative analysis of lung attenuation using automated software.

Measurements of lung volumes and DLCO were repeated at 24-months after ICU discharge.

Detailed information on all clinical, functional and image measurements is available in the Online Supplementary Data and Supplementary Figure 2 .

Blood obtained by venipuncture into ethylenediaminetetraacetic acid (EDTA) tubes was layered on top of 10 mL Ficoll-Paque Plus (Amersham Biosciences, Amersham, UK) and centrifuged for 20 minutes at 1500 rpm at room temperature. Plasma was removed from the upper layer; peripheral blood mononuclear cells (PBMCs) were removed from the interphase and washed in phosphate buffered saline for later RNA extraction.

Human primary fibroblasts of healthy donors were obtained and expanded. Fibroblasts were cultured in Dulbecco’s modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37°C and 5% CO2, subsequent passing was performed using trypsin 0,02% and neutralized in equal volume of FBS. Fibroblasts on passage 8 were seeded in m24 plates (30.000 fibroblast/well), where they rested for 16 hours. Next, fibroblasts were treated with TGF-β1 signaling inhibitor, SB431542 (10µM) and after 3 hours they were stimulated with healthy volunteer or ARDS survivors’ plasma or EBC (10%) and cultured for 24 hours under those conditions. Fibroblasts supernatant was recovered, and cells were harvested for mRNA extraction or protein extraction. Protein extract was obtained by incubating cell pellet for 45 minutes in 200µL of specific lysis buffer from MARCKS ELISA kit (Raybiotech; Ref: PELMARCKS-S152-T) complemented with 1% protease inhibitor and 2% phosphatase inhibitor.

Protein concentration was determined in plasma, EBC and protein extracts using ELISA kits. Free active TGF-β1 isoform was assessed by using the LegendMax™ ELISA kit (BioLegend; ref. 437707; CA, USA), following the manufacturer’s instructions. Additionally, the following ELISA kits were employed following the manufacturer’s instructions: MMP2 (Invitrogen; ref. HC3081; MA, USA), α-SMA (FineTest; ref. EH1509; Wuhan, China), MARCKS (RayBiotech; ref. PELMARCKS-S152-T; GA, USA). EBC protein concentrations were quantified by ELISA as previously performed by other publications (22) and values were normalized to total protein concentration quantified by bicinchoninic acid (BCA) protein assay (Thermofisher Scientific; Ref. 23225; MA, USA). Additionally, α-SMA and phospho-MARCKS were quantified in fibroblasts’ protein extract following MARCKS ELISA kit instructions and reagents.

RNA was extracted from PBMCs or fibroblasts using High Pure RNA Isolation Kit (Roche Diagnostics, Switzerland). 0.25 μg of RNA was retrotranscribed using High-Capacity cDNA Reverse Transcription kit (Applied Biosystems; MA, USA). Real time quantitative PCR (RTqPCR) was performed using NZY Supreme qPCR Green MasterMix (Nzytech; Lisboa, Portugal) and specific primers for targeted genes synthesized by Eurofins Genomics Srl (Vimidrone, Italy) ( Supplementary Table S1 ). RTqPCR was performed with CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories; CA, USA) for 40 cycles and results normalized to housekeeping gene 18S expression. Primer specific conditions for RTqPCR are specified in Supplementary Table S2 .

Detailed information regarding statistical analysis and replicates can be found in figure legends. Generally, categorical variables are presented as numbers with percentages, and continuous variables as mean ± standard deviation or median (interquartile range) according to their distribution. Comparisons between subgroups were analyzed by Fisher exact test or Mann-Whitney test. For the in vitro model, group differences were analyzed by two-way ANOVA with Sidak’s multiple comparison test. The relationship between variables was assessed with Spearman correlation analysis. Due to multiple hypotheses testing, p-values have been adjusted by using Benjamini-Hochberg correction for false discovery rate.

To assess the discrimination ability of long-term impairment in total lung capacity (TLC) and DLCO, the area under the receiver operating characteristic curve (AUROC) was estimated and the Youden’s method was used to find an optimal cut-off point. Data were analyzed with GraphPad Prism v.8 and SPSS 25.0 software. Differences were considered statistically significant with P-values.

Detailed characteristics of survivors who experienced severe ARDS secondary to COVID-19 are outlined in Table 1 . Additionally, Supplementary Table S2 and Supplementary Figure S2 provide a summary of symptomatic, functional, and morphological alterations observed at 6-months post-discharge, along with details regarding the ventilatory pattern during EBC collection. At this moment, 36 patients (43.9%) had a DmCO/VA ratio below the lower limit of normal. Further comparison between subgroups of patients with normal or reduced DmCO/VA ratio is also presented in both Table 1 , Supplementary Table S2 . Additionally, 21 patients (25.6%) exhibited reduced DLCO at 6-months post-discharge.

At 6-months of the ICU discharge, patients with reduced DmCO/VA exhibited higher plasma levels of active TGF-β1 isoform and matrix metalloproteinase 2 (MMP2) in comparison with those subjects with normal DmCO/VA ( Figures 1A, B ). As a complementary approach, we observed that the mRNA expression of TGF-β1 and MMP2 in PBMCs from patients with reduced DmCO/VA was also higher than in those with normal DmCO/VA ratio ( Figures 1C, D ). Furthermore, when stratifying ARDS survivors according to DmCO/VA ratio, those with a decreased ratio showed higher levels of active TGF-β1 in EBC than patients with normal DmCO/VA ( Figures 1E, F ). Altogether, these data suggest that post-ARDS patients with an early fibrotic stage characterized by a reduced DmCO/VA ratio exhibited higher levels of TGF-β1 in plasma and EBC compared to patients with preserved lung function.

Among survivors of ARDS secondary to COVID-19, individuals experiencing clinically relevant dyspnea at 6-months post-ICU discharge (mMRC ≥2) exhibited higher plasma TGF-β1 levels compared to the remained subjects ( Figure 2A ). However, no significant differences were observed in plasma levels of MMP2 ( Figure 2B ). Furthermore, in post-ARDS patients, the presence of elevated lung parenchymal attenuation, determined by high-attenuation lung volume (HAV), showed a direct proportional relationship with plasma levels of active TGF-β1 ( Figure 2C ) and MMP2 ( Figure 2D , Supplementary Table S3 ). These findings underscore the potential clinical relevance of elevated plasma levels of active TGF-β1 in survivors of ARDS.

In order to explore the TGF-β1 role from plasma or EBC, we have performed an in vitro assay using primary human fibroblasts to analyze a specific marker of pulmonary fibrosis such as phosphorylated myristoylated alanine-rich C-kinase substrate (phosphor-MARCKS) (23). In particular, cells were cultured in presence of 10% of plasma or EBC from ARDS survivors or healthy volunteers for 24 hours. We observed an increase in phosphorylated MARCKS relative to total MARCKS in fibroblasts stimulated with ARDS survivors’ plasma ( Figure 3A ). However, when fibroblasts were pre-treated for 3 hours with 10µM SB431542 (TGF-β receptor inhibitor) (24, 25), the effect of ARDS survivors’ plasma was limited ( Figure 3A ). In addition, we performed similar experiment using patients’ or volunteers’ EBC instead of plasma as EBC presents more limited concentration of TGF-β1 and additionally may represent more accurately the pulmonary microenvironment. Fibroblasts treated with patients’ EBC also presented increased expression of phosphor-MARCKS compared to the ones treated with healthy volunteers’ EBC ( Figure 3B ). In accordance with previous plasma results, the effect of patients’ EBC was diminished when fibroblasts were pre-treated with SB431542 ( Figure 3B ). Moreover, we analyzed mRNA expression in fibroblasts stimulated with EBC and found that MARCKS and alpha smooth muscle actine (α-SMA) mRNA expression were higher in patients’ EBC-stimulated fibroblasts in a TGF-β receptor dependent manner ( Figures 3C, D ). In particular, α-SMA is a well-known and characterized protein used for assessment of activated fibroblasts in several tissues including the lung (26, 27). Collectively, our in vitro data suggested the capacity of patients’ plasma and EBC to promote profibrotic changes in human fibroblasts via TGF-β receptor.

As fibrotic lung lesions progress over the long term, patients experience gas exchange impairment and restrictive disorder. To explore these long-term alterations, we compared the levels of active TGF-β1 and other fibrotic markers 6-months after ICU discharge among patients who developed such conditions at 24-months after ICU discharge.

At 24-months post-ICU discharge, 23 patients (28.0%) with ARDS exhibited gas exchange impairment characterized by a reduced DLCO. Notably, patients with decreased DLCO at 24-months had higher plasma levels of active TGF-β1 at 6-months post-ICU discharge ( Figure 4A ). Furthermore, these patients also displayed elevated concentrations of active TGF-β1, MMP2, phosphor-MARCKS, and α-SMA in EBC at 6-months after ICU discharge ( Figure 4B ).

Similarly, 34 out of 82 ARDS survivors evaluated (41.5%) developed a restrictive disorder at 24-months post-ICU discharge, characterized by decreased total lung capacity (TLC). Compared to those maintaining normal TLC, these patients had higher plasma levels of active TGF-β1 at 6-months ( Figure 4C ) and elevated EBC concentrations of active TGF-β1, MMP2, and phosphor-MARCKS at 6-months post-ICU discharge ( Figure 4D ).

The plasma concentration of active TGF-β1 at 6-months post-ICU discharge discriminates patients who, at 24-months, experienced a reduction in both DLCO and TLC (AUROC: 0.718 ± 0.081, p=0.0059) ( Figure 4E ). Using the best-identified cut-off point (28.49pg/mL) yields a sensitivity of 82.4% (95% CI, 55.8-95.3%) and a specificity of 64.6% (95% CI, 51.7-75.8%). Moreover, the concentration of active TGF-β1 in EBC at 6-months post-ICU discharge demonstrates superior discriminative ability for identifying patients who developed long-term gas exchange impairment and restrictive disorders (AUROC: 0.818 ± 0.066, p=0.0025) ( Figure 4E ). The optimal cut-off point of 0.7673pg active TGF-β1/µg total protein results in a sensitivity of 100% (95% CI, 65.5-99.1%) and a specificity of 55.9% (95% CI, 38.1-72.4).

Taken together, our findings emphasize the relevance of fibrotic markers, especially TGF-β1, in ARDS patients. The medium-term levels of these markers appear to be associated with the progression of the fibrotic response, leading to both gas exchange impairment and restrictive disorders in the long term.