This was a prospective cohort study of adult ARDS patients admitted to the medical floor or the medical ICU of Xiangyang Central Hospital from November 1, 2022 through January 31, 2023, with follow-up having been performed through March 31, 2023. Patients were considered eligible for the study if they had been admitted to the hospital for symptoms consistent with ARDS caused by pneumonia, aspiration, trauma, and sepsis, among other conditions. ARDS was diagnosed as per the Berlin Definition.

ARDS disease severity was used to classify patients into severe and nonsevere groups (13). Patients were classified in the severe group if they were admitted to the ICU and exhibited any: (1) respiratory failure, as characterized by a respiratory rate ≥ 30 breaths/min and resting arterial oxygen saturation ≤ 93% mainly for patients without invasive mechanical ventilation; and/or (2) an arterial blood oxygen partial pressure/fraction of inspiration O₂ (PaO₂/FiO₂) ≤ 150 mmHg mainly for those with invasive mechanical ventilation. Patients were classified in the nonsevere group if they were hospitalized and admitted to the internal medicine or pneumology wards without the need for critical care. Any ARDS patients not meeting the criteria set forth for the severe group were included as nonsevere cases.

ARDS patients were excluded from this study if they met any of the following criteria: (1) patients ≤ 18 years of age; (2) patients that died within 24 h following admission; (3) patients with a history of chronic respiratory disease (such as chronic obstructive pulmonary disease and chronic pulmonary fibrosis), immunodeficiency, or malignant tumors; (4) patients for whom severity status was not established; or (5) patients missing details regarding age, sex, or the results of key clinical laboratory tests. In addition, 52 healthy subjects who had undergone physical examinations in the Physical Examination Center and who generally matched the age and sex distributions of the enrolled ARDS patients were included in this study as HCs.

Basic ARDS patient clinical data including demographic, clinical, and laboratory data were extracted from electronic medical records by trained research staff using standardized forms.

Laboratory results collected within 24 h of patient admission were collected for analysis, including results pertaining to the following: white blood cells (WBCs), hemoglobin (HGB), hematocrit (HCT), lymphocytes, lymphocyte % (LYM%), monocytes, monocyte % (MON%), neutrophils, neutrophil % (NEU%), platelet lymphocyte ratio (PLR), monocyte lymphocyte ratio (MLR), neutrophil-lymphocyte ratio (NLR), platelets (PLTs), platelet distribution width (PDW), mean platelet volume (MPV), C-reactive protein (CRP), procalcitonin (PCT), activated partial thromboplastin time (APTT), prothrombin time (PT), thrombin time (TT), D-dimer, international normalized ratio (INR), and erythrocyte sedimentation rate (ESR). Oxygenation status in ARDS patients was evaluated based on the PaO₂/FiO₂ ratio, while the numbers of organ failures on ICU admission were assessed based on Sequential Organ Failure Assessment (SOFA) scores (14).

The biochemical test results for these patients detailed above were collected through analyses conducted with an Atellica Solution Immunoassay and Clinical Chemistry Analyzer (Siemens Healthcare Diagnostics, Erlangen, Germany). WBCs and lymphocytes were estimated with a Sysmex XE-2100 (TOA Medical Electronics, Kobe, Japan), while a Sysmex CS-5100 System (Siemens Healthcare Diagnostics, Erlangen, Germany) was used to measure APTT, PT, INR, fibrinogen, and D-dimer levels. Levels of CRP were quantified with an i-CHROMA instrument (BodiTech Med Inc., Chuncheon, Korea), whereas the Vacuette ESR System (Greiner Bio-One GmbH, Frickenhausen, Germany) was used to quantify ESR values.

The blood samples analyzed for this study were obtained on enrollment from remaining blood used for routine complete blood counts in the clinic after admission to the ICU or clinical ward. Additionally, blood samples from 61 ARDS patients were obtained when patients were both experiencing severe disease and in stable remission. “Remission” was defined as successful completion of the SBT test for 120 min and with no need for re-intubation after extubation for at least 48 h or tracheotomy status after invasive mechanical ventilation without ventilator support lasting longer than 48 h.

Patients were followed from the time of hospital admission to discharge or death, with no interference with medical decision-making for the enrolled patients. The follow-up for the patients included in this study continued until March 31, 2023. The endpoint for follow-up for patients with severe ARDS was death or recovery from illness. The outcome-related information was collected for further analysis, including survival status, discharge date, days of hospitalization, and date of death date.

Blood samples remaining in the clinic after routine complete blood count (CBC) examinations were collected for the separation of PBMCs. The PBMCs were isolated from 2 mL whole blood by density centrifugation at 3000 rpm for 20 min with 3 mL of Human Lymphocyte Separation Medium (Solarbio, P8610), with collection of the PBMCs from the second layer (layers from the top to the bottom: plasma, PBMCs, separation medium, and erythrocyte layers). The PBMCs were resuspended in 1 mL of TRIzol reagent (Invitrogen, 15596026) prior to storage at – 80 °C.

The PBMC samples in TRIzol were removed from the –80°C freezer and were thawed on ice before the addition of 200 μL of chloroform, followed by thorough mixing and incubation on ice for 15 min. The samples were then centrifuged at 12 000 rpm for 15 min at 4°C. The upper aqueous phase was transferred to another EP tube, taking care not to aspirate the interphase, and an equal volume of isopropanol was added, followed by thorough mixing and incubation on ice for 15 min. This was followed by further centrifugation (12 000 rpm, 15 min, 4°C), the supernatant was discarded, and the RNA was allowed to pellet at the bottom of the tube. After the addition of 1 mL of 75% ethanol, the pellet was resuspended with gentle shaking, and centrifuged (7500 rpm, 5 min, 4°C). The supernatant was discarded as much as possible and the pellet was air-dried at room temperature for 5–10 min. The RNA was dissolved in 40 μL of 0.1% DEPC water (adding DEPC water according to the amount of pellet) and stored at -20°C until use.

The RNA was reverse-transcribed to cDNA using an iScript cDNA Synthesis Kit (Bio-Rad, 1708891) according to the provided directions. The reverse transcription was performed in a reaction volume of 20 µL, including 15 μL of RNA solution together with 1000 ng RNA, 4 μL of 5x reverse-transcription reaction mix, and 1μL of iScript reverse transcriptase at 25°C for 5 min, followed by 46°C for 5 min and 95°C for 1 min. After the reaction, the cDNA was collected and stored at 4°C or -20°C.

Relative gene expression levels were calculated using the 2^-ΔΔCt method, using GAPDH as a reference as in prior reports (15). Primer sequences used for this study are presented in Supplementary Table S1 . qPCR was performed on the ABI7500 system by using the iTaq Universal SYBR Green Supermix (Bio-Rad, 1725122). The reaction volume was 20 µL consisting of 3.0 μL of diluted cDNA, 1.0 μL of the forward and reverse primers (10 μM), 10 μL of iTaq Universal SYBR Green Supermix (Bio-Rad, 1725122), and 5 µL of water with the following conditions: 95°C for 5 min; 40 cycles at 95°C for 10 s; 60°C for 20 s.

Western blotting was performed as previously described (16). The primary antibodies used for the reactions included anti- GAPDH (GoodHere, AB-P-R 001, China), and anti-TM9SF1 (PA5–84406, Invitrogen, USA).

Continuous data are presented as means ± standard deviation, while qualitative data are presented as N (percentage), and these data types were respectively compared with Student’s t-tests or chi-square tests, with one-way ANOVAs being used to compare 3 or more groups. Linear correlation analyses were used to assess correlations between TM9SF1 expression and cytokines. The association between TM9SF1 levels and ARDS severity was examined through univariate and multivariate logistic regression analyses in which odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. The composite outcome of disease progression and death was analyzed with Cox regression analyses, calculating hazard ratios (HRs) and corresponding 95% CIs. Analyses were adjusted for demographic characteristics including age, sex, smoking status, drinking status, and history of disease.

The predictive performance of TM9SF1, other key indicators, and developed predictive models as tools for assessing ARDS severity and mortality was evaluated using receiver operating characteristic (ROC) curves. Predictive models for ARDS severity and mortality were developed through a multivariate analysis of those variables significant in univariate logistic regression analyses (P < 0.05), incorporating all variables that remained significant (P < 0.05) after adjustment for a range of covariates. Multivariate analysis results were also used to construct a nomogram related to ARDS severity. Nomogram performance was evaluated through discrimination analyses based on the concordance index and through the use of calibration curves and the Hosmer–Lemeshow calibration test, with bootstrapping (1,000x resampling).

The threshold for significance was a two-sided P-value < 0.05. Figures were constructed with GraphPad Prism 5 and R v3.6.3, while all statistical analyses were performed with SAS 9.4 software (SAS Institute, NC, USA).

Between November 1, 2022 and January 31, 2023, 239 total adult ARDS patients who were admitted to Xiangyang Central Hospital were enrolled in this study. Of these patients, 123 and 116 were respectively diagnosed with severe and nonsevere ARDS. ARDS patient demographic and clinical characteristics are presented in Table 1 . Sex did not differ significantly between groups (female: 42.2% vs. 47.2%, P = 0.445), whereas the average age of severe ARDS patients was higher than that of individuals with nonsevere disease (73.67 ± 13.26 vs. 65.98 ± 17.03 , P < 0.001). Of these ARDS patients, 68.2% exhibited at least one comorbidity, among which hypertension (43.9%), diabetes mellitus (26.4%), and cardiovascular disease (18.0%) were the most common. Patients with severe disease were more likely than nonsevere patients to have hypertension ( Table 1 ). The median duration of hospitalization for the overall ARDS patient cohort was 15 [8–23] days, with a significantly longer duration of hospitalization among individuals with severe disease as compared to those with nonsevere disease (19 [6–45] vs. 12 [8–18] days, P = 0.017).

Laboratory findings from patients grouped according to disease severity are shown in Table 1 . Routine blood analyses revealed that patients with severe ARDS presented with higher WBCs (P = 0.008) and neutrophil counts (P = 0.001) together with reduced lymphocyte counts as compared to nonsevere patients (P = 0.001). HCT, lymphocyte %, monocyte %, neutrophil %, NLR, PLR, and MLR values also differed significantly among groups (all P values < 0.001). With respect to platelet and coagulation indices, individuals with severe disease presented with higher APTT, PT, and D-dimer levels (all P values < 0.05). CRP and ESR levels were also significantly elevated in individuals suffering from severe disease as compared to those with nonsevere disease (all P-values < 0.001). The PaO₂/FiO₂ ratio was also significantly reduced among severe patients (P < 0.001), while SOFA scores exhibited the opposite trend to those for the PaO₂/FiO₂ ratio.

All patients with ARDS completed the follow-up period through March 31, 2023, and were included in these analyses. Over the course of this follow-up interval, 64 (26.8%) patients died and 175 (73.2%) patients recovered. Similar differences in clinical and demographic characteristics were observed between patients in the severe ARDS group that died and recovered ( Table 2 ). Relative to patients in the severe group that progressed, through who died exhibited several worse laboratory parameters including higher WBCs, neutrophil %, NLR, PLR, MLR, D-dimer, CRP, and SOFA score values together with decreases in lymphocyte %, monocyte %, and PaO₂/FiO₂ ratio (all P-values < 0.05).

Differential TM9SF1 expression was assessed in ARDS patients and HCs ( Figure 1 ). Blood samples from 61 ARDS patients were obtained during both periods of severe and stable remission, and samples were collected from 52 HCs at the same time. Significant increases in TM9SF1 mRNA levels were detected among patients with severe ARDS relative to individuals with nonsevere disease (0.21 ± 0.03 vs. 0.08 ± 0.02, P < 0.001). When TM9SF1 expression levels were analyzed in individuals during both severe disease and stable remission (61 pairs), TM9SF1 expression was found to be significantly reduced with disease resolution (0.28 ± 0.04 vs. 0.10 ± 0.02, P < 0.001) ( Figure 1B ). There were also significantly higher TM9SF1 expression levels in ARDS patients relative to HCs (0.15 ± 0.02 vs. 0.06 ± 0.01, P = 0.003) ( Figure 1C ). Independent-sample t-tests revealed that TM9SF1 expression levels were also significantly higher among deceased ARDS patients as compared to surviving patients (0.29 ± 0.06 vs. 0.10 ± 0.01, P < 0.001) ( Figure 1D ).

In addition, we performed Western blotting to verify the TM9SF1 protein levels in PBMCs from 4 patients with ARDS and 4 healthy controls. It was found that the protein levels of TM9SF1 were upregulated 5.19-fold in patients with ARDS compared to healthy controls ( Figures 1E, F ). Additionally, TM9SF1 protein levels in PBMCs from patients with severe ARDS were also upregulated 2.02-fold compared to those from patients with non-severe ARDS ( Figures 1G, H ).

The release of excessively high cytokine levels is a hallmark of ARDS that can result in severe lung damage (17). To probe the relationship between TM9SF1 mRNA levels and ARDS patient cytokine levels, correlation analyses were next conducted ( Figure 2 ). This approach revealed that TM9SF1 mRNA levels were significantly positively correlated with interferon (IFN)-γ (r = 0.577, P < 0.001) and tumor necrosis factor (TNF)-α (r = 0.857, P < 0.001) levels, with the same also being true for interleukin (IL)-6 (r = 0.664, P < 0.001), IL-17A (r = 0.844, P < 0.001), and the transcription factor forkhead box P3 (FOXP3) (r = 0.731, P < 0.001). When these same correlation analyses were only conducted in patients with severe ARDS, these correlations remained intact, with particularly strong correlations between the expression of TM9SF1 and IFN-γ (r = 0.821, P < 0.001), TNF-α (r = 0.640, P < 0.001), and IL-17A (r = 0.613, P < 0.001). These data support a relationship between TM9SF1 expression and plasma cytokines in ARDS patients, identifying this gene as a potential regulator of ARDS pathogenesis. The cytokine levels during remission and severe disease in patients with ARDS are presented in Supplementary Figure S1 .

The link between the expression of TM9SF1 and ARDS severity was next probed through univariate and multivariate analyses ( Supplementary Table S2 ). When these analyses were adjusted for patient demographic characteristics including age, sex, smoking status, drinking status, and history of disease, ARDS severity rose with increases in TM9SF1 expression (OR = 2.43, 95% CI = 2.15–3.72, P = 0.005). When median TM9SF1 expression levels were used to stratify patients into low- and high-risk groups, high-risk patients presented with a 3.03-fold increase in ARDS severity as compared to low-risk patients (OR = 4.03, 95% CI = 3.75–6.36, P < 0.001).

Cox regression analyses were further used to probe the link between the expression of TM9SF1 and patient mortality ( Supplementary Table S3 ). When these analyses were adjusted for demographic characteristics including age, sex, smoking status, drinking status, and history of disease, ARDS patient mortality risk rose as TM9SF1 levels increased (HR = 2.27, 95% CI = 2.20–4.39, P = 0.001). Relative to patients with lower levels of TM9SF1 expression, the survival time of patients expressing higher levels of this gene was reduced (HR = 3.10, 95% CI = 2.03–4.32, P = 0.008).

ROC curve analyses were used to gauge the relative performance of TM9SF1 and a range of other clinical indicators (WBCs, lymphocytes, lymphocyte %, monocytes, monocyte %, neutrophils, neutrophil %, NLR, MLR, PLR, CRP, D-dimer, PaO₂/FiO₂ and SOFA) when predicting ARDS severity ( Table 3 ). Of the analyzed variables, TM9SF1 exhibited the highest area under the ROC curve (AUC) when predicting disease severity (0.871, 95% CI = 0.785–0.924). The corresponding AUC values for MLR and D-dimer when predicting ARDS severity were 0.798 (95% CI = 0.645–0.847) and 0.811 (95% CI = 0.764–0.883). These results thus suggested that TM9SF1 was the most efficacious marker for use when distinguishing between ARDS patients with severe and nonsevere disease. When using a TM9SF1 expression cut-off value of 0.07, the respective sensitivity and specificity values were 87.3% and 84.7%.

Time-dependent ROC analyses were further used to probe the performance of TM9SF1 and these other variables when predicting patient mortality ( Table 4 ). In these analyses, TM9SF1 levels again exhibited superior predictive utility with a C-index of 0.846 (95% CI = 0.752–0.917) for patient mortality, with corresponding sensitivity and specificity values of 87.5% and 82.6% at a cut-off value of 0.15. The respective C-index values for CRP, PaO₂/FiO₂, and SOFA scores when predicting ARDS patient mortality were 0.783 (95% CI = 0.646–0.905), 0.816 (95% CI = 0.719–0.853), and 0.797 (95% CI = 0.711–0.852). These results demonstrate that TM9SF1 exhibits predictive performance better than that of a wide range of other clinical indicators when assessing ARDS patient disease severity and mortality risk. As such, TM9SF1 may offer value as a novel biomarker to predict ARDS severity and patient survival.

Multivariate logistic regression analyses incorporating all results that were significantly associated with ARDS severity in univariate analyses (P < 0.05) were next conducted to develop a predictive model. Those variables that were significant in this model following adjustment for covariates (P < 0.05) were considered independently associated with ARDS severity, and included TM9SF1 levels, age, D-dimer, and CRP levels, with respective ORs of 2.31 (95% CI = 1.12–4.67, P = 0.017), 1.07 (95% CI = 1.02–1.11, P = 0.024), 1.48 (95% CI = 1.31–2.05, P = 0.018), and 1.02 (95% CI = 1.01–1.05, P = 0.035).

Multivariable Cox proportional hazards regression analyses similarly identified elevated TM9SF1 expression (HR = 2.47, 95% CI = 2.18–5.49, P = 0.011), older age (HR = 1.36, 95% CI 1.22–4.19, P = 0.038), elevated NLR levels (HR = 1.81, 95% CI 1.01–3.25, P = 0.048), and elevated D-dimer levels (HR = 1.51, 95% CI = 1.08–2.12, P = 0.017) as being independently associated with the risk of mortality among patients with severe ARDS.

A nomogram for use in predicting severe ARDS development was established incorporating TM9SF1 levels, age, CRP levels, and D-dimer levels based on the above multivariate analysis results ( Figure 3 ). The odds of developing severe ARDS were computed for individual patients by summing the scores for each of these individual risk factors. Similarly, a predictive nomogram for severe ARDS patient mortality risk was developed incorporating TM9SF1 expression, age, NLR, and D-dimer levels ( Figure 4 ).

ROC curves were used to validate the nomogram designed to predict ARDS severity, yielding an AUC of 0.887 (95% CI = 0.715–0.943), a sensitivity of 87.9%, and a specificity of 86.8%. The validation of the nomogram used when predicting severe ARDS patient mortality yielded a C-index of 0.890 (95% CI = 0.627–0.957). Patients were separated into low- and high-risk groups based on total points to evaluate prognostic outcomes, revealing an optimal risk stratification value of 112.4 points.