Introduction
Lung cancer continues to be the leading cause of cancer-related mortality worldwide, responsible for nearly one in five cancer deaths each year [1-3]. Despite advances in early detection, targeted therapies, and immune checkpoint inhibitors, the overall prognosis for lung cancer patients remains poor. This is largely due to a significant proportion of cases being diagnosed at advanced or metastatic stages, as well as the frequent occurrence of therapeutic resistance [4,5]. These clinical challenges underscore the urgent need to identify the molecular determinants that contribute to tumor progression, invasiveness, and the establishment of aggressive tumor phenotypes.
In recent years, there has been growing attention to the tumor secretome, which is defined as the collection of proteins and vesicular components actively secreted or passively released by tumor cells and other cells within the tumor microenvironment [6,7]. The cancer secretome is now recognized as a critical mediator of intercellular communication, orchestrating tumor-stromal interactions and shaping the biological behavior of cancer cells. Increasing evidence demonstrates that tumor-derived secreted factors regulate key oncogenic processes, including extracellular matrix remodeling, angiogenesis, immune modulation, metabolic reprogramming, and resistance to anticancer therapies [7-10]. Consequently, secreted proteins have become an important source of candidate biomarkers and functional mediators of cancer progression.
Among the secreted proteins implicated in cancer, serpin family A member 3 (SERPINA3; α1-antichymotrypsin) is a serine protease inhibitor that belongs to the serpin superfamily. It is widely recognized as an acute-phase protein, with its expression induced in response to inflammatory signals [11-13]. SERPINA3 is transcriptionally induced by cytokine-activated Janus kinase (JAK)/signal transducer and activator of transcription 3 (STAT3) signaling and is widely associated with inflammatory conditions [14]. In the context of cancer, aberrant SERPINA3 expression has been reported in multiple tumor types, suggesting a potential role in tumor biology that extends beyond its conventional function in protease inhibition [11,13].
Functionally, SERPINA3 has been linked to tumor aggressiveness across various malignancies. Elevated expression of SERPINA3 is associated with increased tumor cell migration, invasion, epithelial-mesenchymal transition, and poor clinical outcomes in cancers such as melanoma, triple-negative breast cancer, colorectal cancer, and glioma, highlighting its predominantly pro-tumorigenic role [15-19]. In lung cancer, available evidence remains limited and conflicting. Some studies have reported reduced SERPINA3 expression in tumor tissues, suggesting tumor-suppressive effects upon ectopic overexpression. In contrast, other studies have demonstrated elevated SERPINA3 expression and pro-tumorigenic functions, including increased proliferation and invasion [20,21]. These conflicting observations suggest that the biological role of SERPINA3 is highly context-dependent and not well understood, particularly in non-small cell lung cancer (NSCLC).
In addition to its connection with tumor aggressiveness, recent transcriptomic studies across various tumor types, including glioma, have linked SERPINA3 expression to immune-related programs, especially inflammatory signaling and pathways associated with immune cells [19,22]. This indicates that SERPINA3 expression may relate to different immune states within the tumor microenvironment. However, the significance of SERPINA3 in tumor-immune interactions has not been systematically investigated, and its potential link to immune checkpoint-related tumor states, particularly in lung cancer, remains largely unexamined.
Notably, despite the growing interest in the role of SERPINA3 in cancer biology, most studies to date have been conducted within individual cohorts or cancer-type-specific experimental systems, with limited independent validation across multiple datasets. Comprehensive multi-cohort investigations validating SERPINA3 expression patterns and clinical significance in independent lung cancer cohorts remain scarce. Additionally, the relationships between SERPINA3 expression and key malignant phenotypes—including tumor invasiveness, disease progression, and immune-associated programs—have not been systematically evaluated in an integrated multi-cohort framework.
In this study, we conducted a multi-cohort transcriptomic screen of 1,896 genes encoding secreted proteins to identify prognostic biomarker candidates in NSCLC. Candidates were prioritized based on differential expression between tumor and normal tissues, as well as survival associations across public NSCLC cohorts. SERPINA3 was selected for further validation in The Cancer Genome Atlas (TCGA) NSCLC dataset and an institutional bronchoalveolar lavage fluid (BALF) registry. Our objectives were to determine whether SERPINA3 consistently indicates adverse tumor biology and immune contexture across cohorts, and whether BALF SERPINA3 has potential clinical utility as a minimally invasive biomarker linked to patient outcomes.
Materials and Methods
1. Study population and patient selection
We screened 178 unique patients with NSCLC from the Severance Hospital respiratory research registry between April 2019 and May 2021 (179 BALF specimens; one patient contributed two specimens). One patient contributed two BALF specimens that would have produced discordant group assignments; to avoid double counting and misclassification, this case was excluded a priori. After exclusions, 162 patients were eligible for analysis. The study was approved by the Institutional Review Board of Severance Hospital (approval number: 4-2021-0228).
2. BALF collection and SERPINA3 measurement
Bronchoalveolar lavage (BAL) was performed following a standardized institutional protocol in the segment corresponding to the affected lesion [23]. Briefly, 30 mL of sterile 0.9% saline was instilled, and approximately 10 mL of BALF was retrieved for analysis. SERPINA3 levels in BALF and plasma were quantified using a quantitative sandwich enzyme-linked immunosorbent assay (ELISA) kit (human alpha-1-antichymotrypsin [SERPINA3/a1ACT] ELISA Kit; MyBioSource, San Diego, CA, USA; catalog no. MBS2702812) according to the manufacturer's instructions. The assay detection range was reported as 3.12-200 ng/mL, with a sensitivity of <1.47 ng/mL. The intra-assay and inter-assay coefficients of variation were reported as <10% and <12%, respectively.
3. Survival analysis
Overall survival (OS) and progression-free survival (PFS) were analyzed using Kaplan-Meier curves with the log-rank test. Patients were divided into SERPINA3-high and SERPINA3-low groups based on the median BALF SERPINA3 concentration as the primary cutoff. As part of exploratory sensitivity analyses, we also examined tertile-based grouping and conducted subgroup analyses restricted to stage IV disease. To enhance interpretability, BALF and plasma SERPINA3 levels were modeled in the Cox analyses per 1 standard deviation increase. For effect estimates, we fitted Cox proportional hazards models to obtain hazard ratios (HRs) and 95% confidence intervals (CIs) in both univariate and multivariable analyses. The proportional hazards assumption was assessed using Schoenfeld residuals, with no significant violations detected. Time was presented in months.
4. Transcriptome differential expression
RNA-seq data and matched clinical information for TCGA lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) cohorts were downloaded from the Genomic Data Commons (GDC) Data Portal on February 24, 2025. For TCGA-based analyses, LUAD and LUSC cases were combined and analyzed as a single NSCLC dataset. Differential expression analysis was performed using DESeq2, employing size-factor normalization and the Wald test, with histologic subtype included as a covariate where applicable to account for cohort-specific differences. P-values were adjusted using the Benjamini-Hochberg method. For microarray datasets, preprocessing and differential expression analysis utilized limma, which included background correction and quantile normalization where appropriate, followed by linear modeling with empirical Bayes moderation. Differentially expressed genes (DEGs) were defined by false discovery rate (FDR) <0.05 and |log2FC| ≥1, unless stated otherwise.
5. Secreted-protein gene set curation and meta-analysis
A curated list of genes that encode secreted proteins was obtained from the Human Protein Atlas [24]. Genewise survival effects were computed for each cohort using Cox models to obtain log(HR) and standard error. These results were then combined across cohorts using the metagen function under fixed- and random-effects models. Between-study variance was estimated using the Paule-Mandel method, and the Hartung-Knapp adjustment was applied only when specified.
Heterogeneity was summarized using I2 and τ2. We reported meta-analysis FDRs using Benjamini-Hochberg correction and required k ≥3 contributing cohorts for inclusion in the main list; I2 <25% was typically applied as an additional criterion.
6. Enrichment analysis
Enrichment of Hallmark and Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets was assessed using gene set enrichment analysis (GSEA) based on ranked differential expression. Multiple testing was controlled using the Benjamini-Hochberg FDR.
7. Immune-cell signature analysis
We quantified the enrichment of tumor-associated immune-cell signatures using logistic regression applied to bulk transcriptomes. Briefly, we assembled immune marker sets derived from single-cell data based on a published atlas [25]. For each signature and sample, we fitted a binomial generalized linear model to predict whether each gene belonged to the signature based on its standardized expression value in that sample. The signed coefficient for membership yielded a per-sample log-odds enrichment score, with significance evaluated using Wald tests and multiple testing controlled by the Benjamini-Hochberg FDR. This overall procedure adhered to the logistic regression framework described previously [26].
8. Statistical analysis
All analyses were performed in R version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria). Baseline characteristics were compared using the Mann-Whitney U test for continuous variables and Fisher’s exact test for categorical variables. All tests were two-sided with α=0.05.
Results
1. Multi-cohort selection of prognostic secreted-protein candidates
To nominate secreted-protein biomarker candidates with prognostic relevance in NSCLC, we implemented a multi-cohort transcriptomic selection workflow (Figure 1A). We started with 1,896 genes annotated as encoding secreted proteins, curated from the Human Protein Atlas. Of these, 146 genes exhibited differential expression (tumor vs. adjacent normal) across three independent NSCLC microarray cohorts (GSE19804, GSE32863, GSE75037), identified through the threeway intersection of DEGs (Figure 1B and Supplementary Table S1, including full gene lists and statistics).
Next, using four NSCLC cohorts with survival data (GSE31210, GSE37745, GSE41271, GSE42127), we conducted a Cox proportional hazards meta-analysis for each gene (fixed-effect inverse-variance weighting of log-HRs, with a random-effects model also computed). Applying pre-specified filters (k ≥3 cohorts contributing, I² <25%, FDR <0.05 for both fixed and random models), 124 genes were associated with OS (Supplementary Table S2). Intersecting these 124 genes with the 146 tumor-normal DEGs yielded 24 final candidates (Figure 1A).
An UpSet plot summarizes the number and identity of survival cohorts in which each gene showed individual significance for the 24 candidates (Figure 1C and Supplementary Table S2 for cohort-level Cox results). Among these candidates, eight genes previously implicated in cancer biology were identified: CCL23, S100A9, SPARCL1, VEGFD, SERPINA3, VWF, IL1B, and SFTPC. Notably, VEGFD, SERPINA3, and SFTPC were also identified as tumor-normal DEGs in ≥2 of the three tumor/normal datasets. Among the 24 final candidates, SERPINA3 was prioritized for follow-up due to its reproducible tumor-normal differential expression, consistently directional survival associations across cohorts with minimal heterogeneity, translational relevance as a secreted protein measurable in BALF, and relative underexploration in NSCLC.
Pooled across cohorts, SERPINA3 expression was associated with worse survival (Figure 1D and Supplementary Table S2): fixed-effect HR 1.20 (95% CI, 1.09 to 1.33); random-effects HR 1.20 (95% CI, 1.09 to 1.33); I²=0%. While not every cohort reached nominal significance by log-rank, the direction of the effect was concordant, and the meta-analytic FDR passed the threshold. Cohort-specific Kaplan-Meier curves are provided in Supplementary Figure S1. Together, these data indicate that SERPINA3 is differentially expressed in tumor vs normal tissues and is associated with poorer OS, supporting its candidacy as a secreted prognostic marker in NSCLC.
2. TCGA NSCLC transcriptome and pathway analysis stratified by SERPINA3
To validate the association between SERPINA3 and prognosis, we analyzed the TCGA NSCLC dataset (Figure 2). Across all samples, SERPINA3 expression was lower in tumors than in adjacent normal tissues (p<0.001) (Figure 2A). However, when tumors were categorized based on tumor-intrinsic SERPINA3 expression, a clear heterogeneity emerged: the SERPINA3- high group exhibited significantly higher SERPINA3 expression compared to normal tissues, while the SERPINA3-low group showed significantly lower expression than normal (both p<0.001) (Figure 2B), indicating two biologically distinct tumor subpopulations.
Differential expression analysis between the two tumor groups revealed a prominent innate immune and acute-phase signature (volcano plot, Figure 2C and Supplementary Table S3). In the SERPINA3-high group, the genes OLFM4, LTF, LBP, SERPINA5, and SERPINA4 were upregulated, consistent with neutrophil-associated innate immune and acute-phase responses. Conversely, the markers FGFBP2 (associated with cytotoxic lymphocytes and natural killer cells) and DEFA5 were enriched in the SERPINA3-low group.
Kaplan-Meier analyses indicated that patients in the SERPINA3-high group had significantly shorter survival compared to those in the SERPINA3-low group: OS (median 39.2 months [95% CI, 32.9 to 55.2] vs. 55.7 months [95% CI, 48.8 to 70.1], log-rank p=0.016) and PFS (median 30.6 months [95% CI 25.5-37.9] vs. 44.2 months [95% CI 37.8-51.5], log-rank p=0.018) (Figure 2D).
To investigate functional differences, we performed GSEA comparing SERPINA3-high versus -low tumors. In the Hallmark collection (Figure 2E and Supplementary Table S4), SERPINA3-high tumors showed enrichment for tumor necrosis factor-α (TNF-α) signaling via nuclear factor-κB (NF-κB; q=4.32×10-6), inflammatory response (q=0.004), and interleukin 6 (IL-6)-JAK-STAT3 signaling (q=0.047). In the KEGG pathway analysis (Supplementary Table S5), IL-17 signaling was also significantly enriched (q=0.00097). Collectively, these findings indicate that high SERPINA3 expression is associated with an immune-inflammatory transcriptional program, delineating a biologically distinct subset of NSCLC linked to poorer clinical outcomes.
3. Immune microenvironment of SERPINA3- high NSCLC: T-cell infiltration with checkpoint upregulation and features of T-cell dysfunction
To identify the immune compartments associated with pathway enrichment in SERPINA3-high tumors, we conducted logistic regression analyses using tumor-associated immune-cell signatures derived from single-cell references, as detailed in the Materials and Methods (Figure 3A and Supplementary Table S6). SERPINA3-high status was positively correlated with plasma cells, memory T cells, total lymphocytes, and monocytes/macrophages, as well as exhausted CD8+ T-cell signatures (G6/G9). In contrast, the cytotoxic lymphocyte signature (G8) was negatively correlated (Figure 3A).
We then compared the expression of T-cell trafficking chemokines, including C-X-C motif chemokine ligand 9 (CXCL9), CXCL10, CXCL11, CXCL13, C-C motif chemokine ligand 2 (CCL2), CCL4, and CCL5; inhibitory immune checkpoints, such as programmed cell death protein 1 (PD-1), cytotoxic T-lymphocyte associated protein 4 (CTLA-4), T cell immunoglobulin and mucin domain-3 (TIM-3), lymphocyte activating 3 (LAG-3), V-set immunoregulatory receptor (VSIR), T cell immunoreceptor with Ig and ITIM domains (TIGIT), and B- and T-lymphocyte attenuator (BTLA); and immunosuppressive mediators related to T-cell dysfunction, including transforming growth factor-β (TGF-β), IL-10, indoleamine 2,3-dioxygenase (IDO), and the adenosine-A2A receptor pathway. Chemokines were broadly upregulated in SERPINA3-high tumors; however, multiple checkpoints and suppressive mediators were also significantly increased (Figure 3B-D). These findings suggest that while T-cell recruitment remains intact, heightened inhibitory signaling is associated with features of T-cell dysfunction in SERPINA3-high tumors.
4. Baseline characteristics of the NSCLC BALF registry cohort
To translate the TCGA findings into a clinical setting, we analyzed an institutional respiratory research registry of patients with NSCLC (Figure 4). Of 178 unique patients (179 BALF specimens), one duplicate BALF case and 15 additional patients were excluded per protocol, yielding 162 eligible patients. Patients were dichotomized into SERPINA3-high and SERPINA3-low groups based on the median BALF SERPINA3 concentration (208.36 ng/mL) as the cutoff (n=81 per group). Baseline characteristics are summarized in Table 1. The SERPINA3-high group had a lower proportion of men than the SERPINA3-low group (53.1% vs. 71.6%, p=0.023); all other demographic and clinical variables—including age, stage, histology, programmed death-ligand 1 (PDL1) status, and epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) status—did not differ significantly between groups. PFS and OS were used as outcomes and analyzed in Tables 2, 3.
5. Cox regression for PFS and OS
In multivariable Cox models adjusted for age, sex, stage, histology, EGFR status, and plasma SERPINA3 , higher BALF SERPINA3 , modeled per 1 standard deviation increase, was independently associated with shorter PFS (Table 2). For OS, higher BALF SERPINA3 was also linked to worse outcomes, along with male sex and stage IV disease (Table 3). In an exploratory subgroup analysis restricted to stage IV disease, Kaplan- Meier analyses did not show statistically significant separation by BALF SERPINA3 group; however, in adjusted Cox analysis, higher BALF SERPINA3 remained associated with shorter PFS (Supplementary Figure S2 and Supplementary Tables S7, S8). Due to the limited size of this subgroup, these findings should be considered exploratory. In an exploratory sensitivity analysis using tertile-based grouping of BALF SERPINA3 , the overall patterns of OS and PFS were broadly similar to those observed in the primary median-based analysis, although the differences did not reach statistical significance (Supplementary Figure S3).
Discussion
We conducted a multi-cohort transcriptomic screen of genes encoding secreted proteins, integrating differential expression and Cox meta-analysis, which identified SERPINA3 as a candidate prognostic biomarker in NSCLC. In the TCGA NSCLC dataset, SERPINA3 distinguished tumors into two transcriptomic phenotypes: while normal lung tissue exhibited higher average levels of SERPINA3 than bulk tumors, the SERPINA3-high subset surpassed normal levels, whereas the SERPINA3-low subset fell below normal levels, indicating biologically distinct subgroups. Patients with SERPINA3-high tumors experienced shorter OS and PFS according to Kaplan-Meier analysis.
To reconcile the seemingly discrepant prior findings, we considered the inter-patient variability in tumor SERPINA3 expression. A prior NSCLC analysis likewise found lower SERPINA3 in TCGA tumors than in normal lung [20], mirroring our tumor-versus-normal contrast. However, such between-tissue comparisons do not explain why, within the tumor cohort, the SERPINA3-high subset fares worse. Across public cohorts, we observed a consistent two-level pattern: datasets such as GSE19804, GSE32863, and GSE75037 demonstrate an overall downregulation of SERPINA3 in tumor tissue compared to normal tissue. In contrast, other datasets (e.g., GSE31210, GSE37745, GSE41271, GSE42127) associate higher tumor SERPINA3 levels with poorer prognosis. Our TCGA validation reconciles these observations by revealing significant inter-patient variability in tumor SERPINA3 expression. Although bulk tumors average lower SERPINA3 levels than normal lung tissue, stratifying tumors into SERPINA3-high and SERPINA3-low subsets uncovers distinct immune programs and survival trajectories. This provides a biologically coherent rationale for the adverse outcomes observed in the SERPINA3-high subset.
An additional point worth considering is the origin of the supranormal SERPINA3 expression observed in the SERPINA3-high subset. Since SERPINA3 is an inflammation-related acute-phase protein that can be induced by cytokine signaling, including IL-1/IL-6-STAT3, one possibility is that this expression pattern reflects an intense inflammatory tumor microenvironment rather than a purely tumor-cell-intrinsic program [11]. At the same time, prior studies in other malignancies support a tumor-cell-intrinsic role for SERPINA3 [16,27]. Because our analyses were based on bulk transcriptomic data, we cannot determine whether the SERPINA3-high state is primarily due to malignant-cell overexpression, inflammatory microenvironmental factors, or a combination of both.
This interpretation aligns with findings in other malignancies, including breast cancer, glioma, melanoma, and endometrial cancer, where elevated SERPINA3 is associated with adverse outcomes [15,16,19,28]. By contrast, a prior lung cancer study described tumor-suppressive effects of SERPINA3 largely from in vitro and murine models, with limited patient-level survival data [20]. In the TCGA dataset, GSEA indicated enrichment of TNF-α/NF-κB, inflammatory response, IL-6-JAK-STAT3, and IL-17 signaling in the SERPINA3-high group. This pattern aligns with previous reports, which suggest that in melanoma, SERPINA3 is involved in tumor migration and invasion through STAT3-dependent induction [14], and in glioma, a SERPINA3-high subset shows enrichment of TNF-α/NF-κB signaling [19]. Together, these data indicate a conserved inflammatory circuitry that may contribute to NSCLC biology. In this context, SERPINA3-high NSCLC exhibited a T-cell-infiltrated phenotype characterized by elevated checkpoint markers, consistent with T-cell dysfunction and the upregulation of immunosuppressive mediators (TGF-β, IL-10, IDO). This suggests an association with T-cell dysfunction and potential immune escape.
Prior work reported elevated extracellular vesicles (EVs) containing SERPINA3 in the plasma of patients with NSCLC compared to healthy controls, suggesting circulating detectability [29]. However, that study evaluated EV-associated SERPINA3 in immunocaptured plasma tumor-derived EVs, while we measured SERPINA3 using ELISA in unfractionated plasma and BALF. Despite these methodological differences, our findings build on previous observations by demonstrating that SERPINA3 levels in BALF, but not in plasma, are linked to clinical outcomes. In multivariable Cox models, higher BALF SERPINA3, modeled per 1 standard deviation increase, remained independently associated with both shorter PFS and worse OS, whereas plasma SERPINA3 did not (Tables 2, 3). This supports the relevance of SERPINA3 at the tumor-airway interface. Additionally, when considered alongside transcriptomic data from publicly available datasets, our results suggest a localized secreted-protein signal at the tumor-airway interface that is captured by BALF, highlighting the potential of SERPINA3 as a candidate prognostic biomarker for risk stratification.
From a clinical perspective, the value of BALF in this context may stem less from its reduced invasiveness and more from its proximity to the tumor-bearing lung compartment. Compared to plasma, BALF is more likely to capture local immune-inflammatory and tumor microenvironment signals that may be diluted or obscured in the systemic circulation [30,31]. This interpretation aligns with our finding that BALF SERPINA3 levels, but not those in plasma, are associated with clinical outcomes. Currently, these findings support the role of SERPINA3 primarily as a prognostic biomarker candidate for risk stratification, especially when considered alongside established clinicopathologic variables, rather than as a standalone marker for treatment selection. We also observed a sex imbalance between the BALF SERPINA3-high and -low groups. While direct evidence addressing sex differences in SERPINA3 specific to NSCLC is limited, broader studies have described sex-associated transcriptomic differences in NSCLC [32]. Given that SERPINA3 is an inflammation-related acute-phase protein, this finding may reflect sex-related differences in immune or inflammatory states; however, this interpretation remains hypothesis-generating.
Several features strengthen our study. By combining a multi-cohort transcriptomic screen of genes encoding secreted proteins (differential expression plus Cox meta-analysis) with orthogonal validation in the TCGA NSCLC dataset, we minimized bias from single cohorts or methods and nominated SERPINA3 as a candidate prognostic biomarker in NSCLC. Concordant signals across differential expression, survival separation, and pathway enrichment support a mechanism-informed interpretation—specifically, that SERPINA3-high tumors are T-cell-infiltrated yet exhibit high checkpoint levels, consistent with T-cell dysfunction. This lends biological plausibility to the prognostic association. On the translational axis, the SERPINA3-associated state was captured in BALF but not in plasma, indicating compartment specificity at the tumor-airway interface. This supports the feasibility of BALF-based risk stratification and serial monitoring without the need for invasive tissue biopsies, such as transbronchial lung biopsy or computed tomography-guided needle biopsy.
Several limitations warrant consideration. First, our study relied in part on publicly available bulk transcriptomic datasets, which limited cell type-resolved attribution and precluded inferences about cell-cell interactions. The correlations between SERPINA3 and markers and pathways related to T-cell dysfunction are observational and do not establish causality. Second, because BAL involves variable dilution and recovery, the measured BALF SERPINA3 levels may have been influenced by procedural variability. In our study, BAL was performed using a standardized institutional protocol, but no additional normalization or dilution correction for instilled or retrieved volume was applied. Third, although higher BALF SERPINA3 levels were independently associated with shorter PFS, the HR was numerically close to 1, suggesting that its incremental clinical value may be modest and should be interpreted cautiously. Accordingly, BALF SERPINA3 may be better viewed as a candidate biomarker requiring further validation rather than a standalone clinical marker at this stage. Finally, the BALF registry lacked temporally matched tumor tissue from the same patients, limiting direct within-patient correlations between tissue and BALF and motivating the need for paired-sampling studies.
Taken together, the data indicate that inter-patient variability in tumor SERPINA3 expression is associated with distinct immune-functional states and prognostic trajectories. The SERPINA3-high state corresponds with sustained inflammatory cytokine signaling and upregulation of T-cell checkpoints, features that likely reflect—and may reinforce—a myeloid-skewed, protease-rich microenvironment conducive to T-cell dysfunction. In contrast, the SERPINA3-low state is relatively enriched in cytotoxic and natural killer cell-related features, which align with more effective antitumor immunity. Overall, our data support SERPINA3 as a promising prognostic biomarker candidate linked to T-cell dysfunction in NSCLC. Measurements based on BALF may provide clinically relevant tumor-proximal information for patient risk stratification, especially when local immune-inflammatory signals are not easily captured in plasma. The potential role of SERPINA3 as a predictor of response to PD-1/PD-L1 blockade is promising and warrants prospective, pre-specified evaluation in biomarker-defined cohorts. Finally, mechanistic studies should clarify whether SERPINA3 drives T-cell dysfunction or serves as a readout of it, as well as assess its therapeutic potential.
PDF Links
PubReader
ePub Link
XML Download
Data Sharing Statement
Full text via DOI
Supplement1
Print
Download Citation
