Skip to main content

TMPRSS2 is a tumor suppressor and its downregulation promotes antitumor immunity and immunotherapy response in lung adenocarcinoma

Abstract

Background

TMPRSS2, a key molecule for SARS-CoV-2 invading human host cells, has an association with cancer. However, its association with lung cancer remains insufficiently unexplored.

Methods

In five bulk transcriptomics datasets, one single‐cell RNA sequencing (scRNA-seq) dataset and one proteomics dataset for lung adenocarcinoma (LUAD), we explored associations between TMPRSS2 expression and immune signatures, tumor progression phenotypes, genomic features, and clinical prognosis in LUAD by the bioinformatics approach. Furthermore, we performed experimental validation of the bioinformatics findings.

Results

TMPRSS2 expression levels correlated negatively with the enrichment levels of both immune-stimulatory and immune-inhibitory signatures, while they correlated positively with the ratios of immune-stimulatory/immune-inhibitory signatures. It indicated that TMPRSS2 levels had a stronger negative correlation with immune-inhibitory than with immune-stimulatory signatures. TMPRSS2 downregulation correlated with increased proliferation, stemness, genomic instability, tumor progression, and worse survival in LUAD. We further validated that TMPRSS2 was downregulated with tumor progression in the LUAD cohort we collected from Jiangsu Cancer Hospital, China. In vitro and in vivo experiments verified the association of TMPRSS2 deficiency with increased tumor cell proliferation and invasion and antitumor immunity in LUAD. Moreover, in vivo experiments demonstrated that TMPRSS2-knockdown tumors were more sensitive to BMS-1, an inhibitor of PD-1/PD-L1.

Conclusions

TMPRSS2 is a tumor suppressor, while its downregulation is a positive biomarker of immunotherapy in LUAD. Our data provide a potential link between lung cancer and pneumonia caused by SARS-CoV-2 infection.

Background

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has infected more than 204 million people and caused more than 4.3 million deaths worldwide as of August 12, 2021 (https://coronavirus.jhu.edu/map.html). SARS-CoV-2 invades host cells using its spike glycoprotein (S) [1], which is composed of S1 and S2 functional domains. S1 binds the angiotensin-converting enzyme 2 (ACE2) for cell attachment, and S2 binds the transmembrane protease serine 2 (TMPRSS2) for membrane fusion [1]. Since TMPRSS2 plays a crucial role in the regulation of SARS-CoV-2 invasion, and cancer patients are susceptible to SARS-CoV-2 infection, an investigation into the role of TMPRSS2 in cancer is significant in the context of the current SARS-CoV-2 pandemic. Previous studies have demonstrated the association between TMPRSS2 and cancer [2,3,4,5]. Typically, the TMPRSS2-ERG gene fusion frequently occurs in prostate cancer and is associated with tumor progression [6,7,8]. In a recent study [3], Katopodis et al. revealed that TMPRSS2 was overexpressed in various cancers versus their normal tissues. In another study [4], Kong et al. explored TMPRSS2 expression in lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). This study suggested that TMPRSS2 was a tumor suppresser in LUAD for its significant downregulation in LUAD versus normal tissue. A few studies have examined the association between TMPRSS2 and tumor immunity in cancer. For example, Bao et al. [5] investigated TMPRSS2 expression and its associations with immune and microbiome variates across 33 tumor types. Luo et al. [9] explored the association between TMPRSS2 expression and immune infiltration in prostate cancer. Despite these prior studies, the associations of TMPRSS2 with tumor immunity, oncogenic signatures or pathways, tumor progression and clinical outcomes in lung cancer remain insufficiently explored.

In this study, we analyzed the associations between TMPRSS2 expression levels and the enrichment levels of immune signatures in five LUAD cohorts. The immune signatures included CD8 + T cells, immune cytolytic activity, CD4 + regulatory T cells, myeloid-derived suppressor cells (MDSCs), and PD-L1. We also analyzed the associations between TMPRSS2 expression levels and the activities of several oncogenic pathways, including cell cycle, mismatch repair, and p53 signaling. Moreover, we explored the associations between TMPRSS2 expression and tumor phenotypes (such as proliferation and tumor stemness), genomic features (such as genomic instability and intratumor heterogeneity (ITH)), tumor advancement and prognosis in these LUAD cohorts. Furthermore, we explored the association between TMPRSS2 expression and the response to cancer immunotherapy. We validated the computational findings by performing in vitro experiments in the human lung cancer cell line A549, H1975, and H1299 and in vivo experiments with mouse tumor models. We also validated our findings in LUAD patients we collected from Jiangsu Cancer Hospital, China. Our study demonstrates that TMPRSS2 is a tumor suppressor while its downregulation can promote antitumor immune response and cancer immunotherapy response. This study may provide insights into the connection between lung cancer and pneumonia caused by SARS-CoV-2 infection.

Results

TMPRSS2 expression correlated negatively with the enrichment of immune signatures in LUAD

We found that TMPRSS2 had a significant negative expression correlation with the infiltration levels of CD8 + T cells, which represent the adaptive antitumor immune response, in three of the five LUAD cohorts (Spearman correlation, p < 0.05) (Fig. 1a). Moreover, TMPRSS2 expression levels were negatively correlated with immune cytolytic activity, a marker for underlying immunity [10], in all the five LUAD cohorts. Meanwhile, TMPRSS2 had a significant negative expression correlation with PD-L1 in the five LUAD cohorts (Fig. 1a). TMPRSS2 expression levels were negatively correlated with the infiltration levels of CD4 + regulatory T cells and MDSCs in four LUAD cohorts, which represent tumor immunosuppressive signatures (Fig. 1a).

Fig. 1
figure 1

Association between TMPRSS2 expression and immune signatures in LUAD. a Correlations between TMPRSS2 expression levels and the enrichment levels of CD8 + T cells, immune cytolytic activity, PD-L1 expression levels, and the enrichment levels of CD4 + regulatory T cells and myeloid-derived suppressor cells (MDSCs) in five LUAD cohorts. The Spearman or Pearson correlation coefficients (ρ or r) and p values are shown. b Pearson correlations between TMPRSS2 expression levels and the ratios of immune-stimulatory/immune-inhibitory signatures (CD8 + /PD-L1) in LUAD. c Kaplan–Meier survival curves showing a better disease-free survival in LUAD patients with high ratios of CD8 + /PD-L1 (upper third) than those with low ratios of CD8 + /PD-L1 (bottom third). The log-rank test p value is shown. * p < 0.05, ** p < 0.01, *** p < 0.001, ns p ≥ 0.05. They also apply to the following figures

Taken together, these results suggest a significant negative association between TMPRSS2 abundance and immune infiltration levels in LUAD. Interestingly, TMPRSS2 expression levels showed a significant positive correlation with the ratios of immune-stimulatory/immune-inhibitory signatures (CD8 + T cells/PD-L1) consistently in the five LUAD cohorts (Pearson correlation, p < 0.05) (Fig. 1b). It indicated that TMPRSS2 levels had a stronger negative correlation with immune-inhibitory signatures than with immune-stimulatory signatures. Furthermore, we found that the ratios of immune-stimulatory/immune-inhibitory signatures were positively correlated with disease-free survival (DFS) in The Cancer Genome Atlas of lung adenocarcinoma (TCGA-LUAD) cohort (log-rank test, p = 0.01) (Fig. 1c).

TMPRSS2 downregulation correlates with increased oncogenic signatures, tumor proliferation, stemness, and unfavorable clinical outcomes in LUAD

We found that TMPRSS2 expression levels were inversely correlated with the activities of the cell cycle, mismatch repair, and p53 signaling pathways in the five LUAD cohorts (Spearman correlation, p < 0.001) (Fig. 2a). Moreover, TMPRSS2 showed a negative expression correlation with MKI67, a tumor proliferation marker, in the five LUAD cohorts (Pearson correlation, p < 0.001) (Fig. 2b). Tumor stemness indicates a stem cell-like tumor phenotype representing an unfavorable prognosis in cancer [11]. We observed that TMPRSS2 expression levels were inversely correlated with tumor stemness scores in these LUAD cohorts (Spearman correlation, p < 0.001) (Fig. 2c).

Fig. 2
figure 2

Associations between TMPRSS2 expression and oncogenic pathways, tumor phenotypes and prognosis in LUAD. The inverse correlations between TMPRSS2 expression levels and the activities of oncogenic pathways (a), MKI67 expression levels (b), and stemness scores (c) in LUAD. The Spearman or Pearson correlation coefficients (ρ or r) and p values are shown. d Comparisons of TMPRSS2 expression levels between late-stage (Stage III-IV) and early-stage (Stage I-II), between large-size (T3-4) and small-size (T1-2), and between N1-3 (lymph nodes) and N0 (without regional lymph nodes) LUADs. The Student’s t test p values and fold change (FC) of mean TMPRSS2 expression levels are shown. e The lung cancer data from Jiangsu Cancer Hospital showing that TMPRSS2 expression levels are significantly lower in late-stage (Stage IV) than in early-stage (Stage I-II) LUADs. f Kaplan–Meier survival curves showing that low-TMPRSS2-expression-level (bottom third) LUAD patients have worse OS and/or DFS than high-TMPRSS2-expression-level (upper third) LUAD patients. The log-rank test p values are shown. OS, overall survival. DFS, disease-free survival. g Multivariate Cox proportional hazards regression analysis show that stage, age and CD4 + regulatory T cells enrichment have a significant inverse correlation with OS, and that TMPRSS2 expression and CD8 + T cells enrichment have a significant positive correlation with OS in TCGA-LUAD cohort. The “AGE”, “CD4 + regulatory T cells enrichment”, and “CD8 + T cells enrichment” are continuous variables, and the “TMPRSS2 expression” (high versus low) and “STAGE” (early-stage (stage I-П) versus late-stage (stage III-IV)) are binary variables. h Comparisons of TMPRSS2 expression levels between EGFR-mutated and EGFR-wildtype LUADs and between three LUAD transcriptional subtypes. TRU, terminal respiratory unit. PI, proximal-inflammatory. PP, proximal-proliferative. i Comparisons of TMPRSS2 expression levels among different classes of LUAD single cells in two LUAD scRNA-seq datasets (GSE131907 [12] and Maynard corhort [13])

We detected that TMPRSS2 expression levels significantly decreased with tumor advancement in LUAD (Fig. 2d). For example, in the TCGA-LUAD cohort, TMPRSS2 expression levels were significantly lower in late-stage (Stage III-IV) than in early-stage (Stage I-II) LUADs (Student’s t test, p < 0.001; fold change (FC) = 1.6), in large-size (T3-4) than in small-size (T1-2) LUADs (p = 0.007; FC = 1.5), in LUADs with lymph nodes (N1-3) than in those without regional lymph nodes (N0) (p = 0.02; FC = 1.3), and in LUADs with metastasis (M1) than in those without metastasis (M0) (p = 0.07; FC = 1.6). In other two LUAD cohorts (GSE30219 and GSE50081) with tumor size and lymph nodes data available, TMPRSS2 expression levels were also significantly lower in large-size than in small-size LUADs (p < 0.001; FC = 6.4) in GSE30219 and were significantly lower in N1-3 than in N0 LUADs in both GSE30219 (p = 0.02; FC = 2.83) and GSE50081 (p = 0.02; FC = 1.6) (Fig. 2d). Furthermore, the lung cancer data from Jiangsu Cancer Hospital supported that TMPRSS2 expression levels were reduced in late-stage (Stage IV) than in early-stage (Stage I-II) LUADs (p < 0.001; FC = 1.6) (Fig. 2e). Survival analyses showed that TMPRSS2 downregulation was correlated with worse overall survival (OS) and/or DFS in these LUAD cohorts (log-rank test, p < 0.05) (Fig. 2f). To explore whether the positive association between TMPRSS2 expression and OS prognosis was impacted by other confounding variables, we performed multivariate (TMPRSS2 expression, age, stage, CD8 + T cells enrichment, and CD4 + regulatory T cells enrichment) survival analyses using the multivariate Cox proportional hazards model. This analysis showed that TMPRSS2 expression remained a positive prognostic factor (P = 0.0031; hazard ratio (HR) = 0.5637 and its 95% confidence interval (CI): [0.3856, 0.824]) in LUAD (Fig. 2g). As expected, CD8 + T cells enrichment was also a positive prognostic factor (P = 0.0055), and CD4 + regulatory T cells enrichment was likely to be an adverse prognostic factor (P = 0.0552). Both age (P = 0.009) and stage (P < 0.001) were shown to be risk factors for OS prognosis in LUAD.

It has been shown that EGFR-mutated LUADs have a better prognosis than EGFR-wildtype LUADs [14]. We found that TMPRSS2 was more lowly expressed in EGFR-wildtype than in EGFR-mutated LUADs (p = 0.006; FC = 1.5) (Fig. 2h). Besides, LUAD harbors three transcriptional subtypes: terminal respiratory unit (TRU), proximal-inflammatory (PI), and proximal-proliferative (PP), of which TRU has the best prognosis [15]. We found that TMPRSS2 expression levels were the highest in TRU (TRU versus PP: p = 8.68 × 10–14, FC = 2.98; TRU versus PI: p = 1.07 × 10–11, FC = 3.16) (Fig. 2h).

We further analyzed two LUAD single‐cell RNA sequencing (scRNA-seq) datasets (GSE131907 [12] and Maynard corhort [13]) to validate the findings in the tumor bulks. We found that TMPRSS2 expression levels were significantly higher in EGFR-mutated than in EGFR-wildtype LUAD single cells in both datasets (p < 0.05) (Fig. 2i). In GSE131907, TMPRSS2 expression levels followed the pattern in the LUAD single cells: poorly differentiated < moderately differentiated < well differentiated (p < 0.001) (Fig. 2i). In Maynard cohort, the single cells in metastatic tumors displayed significantly lower expression levels of TMPRSS2 than those in primary tumors (p < 0.001); in the same cohort, TMPRSS2 expression levels followed the pattern in the LUAD single cells: progressive disease < TKI naive < residual disease (p < 0.001) (Fig. 2i) that conformed to results of the proliferation potential of LUAD single cells following an opposite pattern: progressive disease > TKI naive > residual disease, as shown in the original publication [13]. Overall, the results from the LUAD scRNA-seq datasets confirmed the tumor suppressor role of TMPRSS2 in LUAD.

Taken together, these results suggest that TMPRSS2 downregulation is associated with worse outcomes in LUAD.

TMPRSS2 downregulation correlates with increased genomic instability in LUAD

Genomic instability plays prominent roles in cancer initiation, progression, and immune evasion [16] by increasing tumor mutation burden (TMB) [17] and aneuploidy or somatic copy number alterations [18]. In the TCGA-LUAD cohort, TMPRSS2 expression levels had a negative correlation with TMB (Spearman correlation, ρ = -0.31; p = 2.58 × 10–12) (Fig. 3a). Homologous recombination deficiency (HRD) may promote chromosomal instability and aneuploidy levels in cancer [19]. We found that TMPRSS2 expression levels were inversely correlated with HRD scores [19] in LUAD (ρ = -0.27; p = 5.76 × 10–10) (Fig. 3b). DNA repair (DR) deficiency can lead to genomic instability [20]. Knijnenburg et al. [19] identified deleterious gene mutations for nine DR pathways in TCGA cancers. We divided LUAD into pathway-wildtype and pathway-mutated subtypes for each of the nine DR pathways. The pathway-wildtype indicates no deleterious mutations in any pathway genes, and the pathway-mutated indicates at least a deleterious mutation in pathway genes. Interestingly, we found that TMPRSS2 expression levels were significantly lower in the pathway-mutated subtype than in the pathway-wildtype subtype for seven DR pathways (p < 0.05; FC > 1.5) (Fig. 3c). The seven pathways included base excision repair, Fanconi anemia, homologous recombination, mismatch repair, nucleotide excision repair, translesion DNA synthesis, and damage sensor. These results suggest a correlation between TMPRSS2 downregulation and DR deficiency.

Fig. 3
figure 3

Association between TMPRSS2 expression and genomic instability in LUAD. Spearman correlations between TMPRSS2 expression levels and tumor mutation burden (TMB) (a) and homologous recombination deficiency (HRD) scores (b) in TCGA-LUAD. TMB is the total somatic mutation count in the tumor. The HRD scores were obtained from the publication [19]. c Comparisons of TMPRSS2 expression levels between pathway-wildtype and pathway-mutated LUAD subtypes for seven DNA repair (DR) pathways in TCGA-LUAD. The pathway-wildtype indicates no deleterious mutations in any pathway genes, and the pathway-mutated indicates at least a deleterious mutation in pathway genes. BER, base excision repair. FA, Fanconi anemia. HR, homologous recombination. MMR, mismatch repair. NER, nucleotide excision repair. TLS, translesion DNA synthesis. DS, damage sensor. d Comparisons of TMPRSS2 expression levels between TP53-mutated and TP53-wildtype LUADs. Expression correlations between TMPRSS2 and DR-associated genes (e) and proteins (f) in LUAD. g Spearman correlation between TMPRSS2 expression levels and intratumor heterogeneity (ITH) scores. The ITH scores were evaluated by the DEPTH algorithm [21]

TP53 mutations often leads to genomic instability because of the important role of p53 in maintaining genomic stability [22]. We found that TMPRSS2 displayed significantly lower expression levels in TP53-mutated than in TP53-wildtype LUADs (p = 0.006; FC = 1.5) (Fig. 3d). Moreover, we found numerous DR-associated genes having significant negative expression correlations with TMPRSS2 in these LUAD cohorts (Pearson correlation, p < 0.05), including MSH2, MSH6, POLE, PCNA, and RAD51 (Fig. 3e). Furthermore, we observed significant negative expression correlations between TMPRSS2 and DNA mismatch repair proteins MSH6 (Pearson correlation, r = -0.30; p = 6.6 × 10–9) and PCNA (r = -0.25; p = 1.5 × 10–6) in the TCGA-LUAD cohort (Fig. 3f). These results indicated an association between TMPRSS2 downregulation and the upregulation of DR molecules, the signature of increased genomic instability.

Genomic instability can promote tumor heterogeneity, which is associated with tumor progression, immune evasion, and drug resistance [23]. We used the DEPTH algorithm [21] to score ITH for each TCGA-LUAD sample and found a significant negative correlation between TMPRSS2 expression levels and ITH scores in LUAD (ρ = -0.55; p < 0.001) (Fig. 3g). It indicates a significant association between TMPRSS2 downregulation and increased ITH in LUAD.

Taken together, these results suggest that TMPRSS2 downregulation is associated with increased genomic instability in LUAD.

Co-expression networks of TMPRSS2 in LUAD

We found 150 and 135 genes having strong positive and negative expression correlations with TMPRSS2 in the TCGA-LUAD cohort, respectively (Pearson correlation, |r|> 0.5) (Fig. 4a; Supplementary Table S3). Gene set enrichment analysis (GSEA) [24] revealed that the cell cycle, p53 signaling, mismatch repair, and homologous recombination pathways were significantly associated with the 135 genes with strong negative expression correlations with TMPRSS2. This conforms to the previous findings that TMPRSS2 downregulation was correlated with increased activities of these pathways.

Fig. 4
figure 4

Co-expression networks of TMPRSS2 in LUAD. a 150 and 135 genes having strong positive and negative expression correlations with TMPRSS2 in TCGA-LUAD, respectively (|r|> 0.5). b Gene modules and their representative gene ontology terms highly enriched in high- (upper third) and low-TMPRSS2-expression-level (bottom third) LUADs identified by WGCNA [25]

Weighted gene co-expression network analysis (WGCNA) [25] identified six gene modules (indicated in blue, turquoise, brown, magenta, purple, and pink color, respectively) highly enriched in the high-TMPRSS2-expression-level LUADs. The representative gene ontology (GO) terms associated with these modules included cell projection, chromosome segregation, response to endogenous stimulus, cell adhesion, cellular response to lipopolysaccharide, and micro-ribonucleoz complex. In contrast, three gene modules (indicated in green, black, and green-yellow color, respectively) were highly enriched in the low-TMPRSS2-expression-level LUADs (Fig. 4b). The representative GO terms for these modules included extracellular matrix (ECM), small molecule metabolic process, and postsynapse (Fig. 4b). The ECM signature plays a crucial role in driving cancer progression [26]. Its upregulation in the low-TMPRSS2-expression-level LUADs is in accordance with the correlation between TMPRSS2 downregulation and LUAD progression.

Validating the mRNA-based findings at the protein level

We analyzed a proteomics dataset for LUAD from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) [27] to validate the previous findings at the protein level. Consistently, lower expression of TMPRSS2 correlated with worse OS (p = 0.062) and metastasis-free survival (MFS) (p = 0.089) in CPTAC-LUAD (Figure S1a). Likewise, the protein TMPRSS2 showed a negative expression correlation with Ki-67 which is encoded by MKI67 (r = -0.43; p = 3.9 × 10–6) (Figure S1b). TMPRSS2 expression correlated inversely with the enrichment of the cell cycle, mismatch repair, and p53 signaling pathways and the stemness signature in LUAD (Figure S1b). TMPRSS2 had significantly higher expression levels in EGFR-wildtype than in EGFR-mutated LUADs (p = 0.001; FC = 2.38) (Figure S1c). At the protein level, TMPRSS2 downregulation also showed a significant correlation with increased genomic instability in LUAD, as evidenced by: (1) TMPRSS2 was downregulated in TP53-mutated LUADs relative to TP53-wildtype LUADs (p = 0.019; FC = 1.85); and (2) TMPRSS2 had negative expression correlations with DR-associated proteins (MSH2, MSH6, and PCNA) in LUAD (p < 0.05) (Figure S1d). Furthermore, the correlation between TMPRSS2 abundance and the enrichment of immune signatures was negative at the protein level, consistent with the result at the mRNA level. That is, TMPRSS2 expression correlated inversely with the enrichment of CD8 + T cells, cytolytic activity, PD-L1, and MDSCs (Figure S1e). Collectively, these results validated the findings by analyzing the transcriptome data.

Experimental validation of the bioinformatics findings

To validate the findings from the bioinformatics analysis, we performed in vitro experiments with the human LUAD cell line A549, H1975, and H1299, and in vivo experiments with mouse tumor models. We found that TMPRSS2 knockdown markedly promoted proliferation and invasion potential in the three cells (Fig. 5a and Supplementary Figure S2) and increased tumor volume and progression in Lewis tumor mouse models (Fig. 5b). This is consistent with the previous results showing that TMPRSS2 downregulation is associated with tumor progression and unfavorable prognosis in LUAD. Furthermore, in vitro experiments showed that MSH6 expression was upregulated in TMPRSS2-knockdown versus TMPRSS2-wildtype A549 cells (Fig. 5c). This is in line with the previous finding of the significant negative correlation between TMPRSS2 expression levels and MSH6 abundance in LUAD.

Fig. 5
figure 5

In vivo and in vitro experimental validation of the bioinformatics findings. TMPRSS2-knockdown tumors display increased tumor-infiltrating lymphocytes, expression of immune checkpoint molecules, and sensitization to immune checkpoint inhibitors. a TMPRSS2 knockdown markedly promoted proliferative and invasive abilities of A549 cells. b TMPRSS2 knockdown increased tumor volume and progression in Lewis tumor mouse models. Lewis tumor cells transfected with ShCon or ShTMPRSS2 lentivirus were subcutaneously injected into mice. The tumor volumes were measured every three days from the fifth day to the fifteenth. Data represent mean ± SEM. SEM, standard error of mean. ShTMPRSS2 versus ShCon group, n = 6 for each group, two-tailed Student’s t test, * p < 0.05, ** p < 0.01, *** p < 0.001. c TMPRSS2 knockdown increased MSH6 expression in A549 cells, as evidenced by Western blotting. d TMPRSS2 knockdown enhanced the expression of MHC class I genes (HLA-A, HLA-B, and HLA-C) in A549 cells, as evidenced by real-time qPCR. e NK cells co-cultured with TMPRSS2-knockdown A549 cells showing higher proliferation capacity than NK cells co-cultured with TMPRSS2-wildtype A549 cells, as evidenced by the EDU proliferation assay. f CD8, CD49b, and PD-L1 immunofluorescence staining in Lewis orthotopic tumors and H-score analysis. ShTMPRSS2 versus shCon group, n = 6 for each group, two-tailed Student’s t test, *** p < 0.001. g-j Comparisons of TNF-α, IFN-γ, PD-1, and LAG3 expression on CD8 + T cells from tumor-infiltrating lymphocytes (TILs) in tumor-bearing mice between TMPRSS2-knockdown and TMPRSS2-wildtype group (ShTMPRSS2 versus ShCon group, n = 6 for each group, two-tailed Student’s t test, * p < 0.05, ** p < 0.01, *** p < 0.001). TILs were stained with CD3, CD8, TNF-α, and IFN-γ and were then analyzed by flow cytometry. Lymphocytes were gated according to forward scatter and side scatter. CD3 and CD8 staining was used to identify CD8 + T cells. k-m TMPRSS2-knockdown tumors formed by subcutaneous injection of Lewis cells, as mentioned in (b). shCon and shTMPRSS2 tumor-bearing mice were divided into vehicle and BMS-1 groups. The vehicle and BMS-1 groups of mice were treated with solvent and BMS-1, respectively. k Representative images of tumor-bearing mice shown on the left. The right graph showing the change of tumor size in the tumor-bearing mice over time. Data represent mean ± SEM (n = 6 for each group, two-tailed Student’s t test, * p < 0.05, ** p < 0.01, *** p < 0.001); Comparison of the volume ratios of mice tumors after and before treatment with BMS-1 between TMPRSS2-knockdown and TMPRSS2-wildtype groups (two-tailed Student’s t test, *** p < 0.001). Comparisons of TNF-α (l) and IFN-γ (m) expression on CD8 + T cells from TILs in tumor-bearing mice (n = 6 for each group, two-tailed Student’s t test, * p < 0.05, ** p < 0.01, *** p < 0.001)

Our bioinformatics analysis revealed a significant inverse correlation between TMPRSS2 abundance and immune infiltration levels in LUAD. Consistently, the MHC class I genes (HLA-A, HLA-B, and HLA-C) showed significantly higher expression levels in TMPRSS2-knockdown than in TMPRSS2-wildtype A549 cells, demonstrated by real-time qPCR (Fig. 5d). NK cells co-cultured with TMPRSS2-knockdown A549 cells displayed significantly stronger proliferation ability than NK cells co-cultured with TMPRSS2-wildtype A549 cells, evident by the EdU proliferation assay (Fig. 5e). Furthermore, in vivo experiments showed that infiltration of CD8 + T cells and NK cells significantly increased in TMPRSS2-knockdown tumors (Fig. 5f). Moreover, on CD8 + T cells from tumor-infiltrating lymphocytes (TILs) in TMPRSS2-knockdown tumors, the expression of TNF-α and IFN-γ were significantly upregulated (Fig. 5g, h), indicating that TMPRSS2 knockdown can enhance the activity of CD8 + TILs. Meanwhile, the expression of PD-1 and LAG3 also significantly increased on CD8 + TILs in TMPRSS2-knockdown tumors (Fig. 5i, j), indicating that TMPRSS2 deficiency can also promote the exhaustion of CD8 + TILs.

Our bioinformatics analysis revealed a significant negative correlation between TMPRSS2 and PD-L1 expression levels. This result was confirmed by both in vitro and in vivo experiments; knockdown of TMPRSS2 increased PD-L1 expression in A549 cells, as evidenced by Western blotting (Fig. 5c); TMPRSS2-knockdown tumors had significantly enhanced PD-L1 expression (Fig. 5f). Furthermore, bioinformatics analysis revealed a significant positive correlation between TMPRSS2 expression levels and the ratios of CD8 + T cells/PD-L1. This was confirmed by that TMPRSS2-knockdown tumors displayed a lower level of increases in CD8 + T cell infiltration than in PD-L1 abundance (Fig. 5f). Because PD-L1 expression is a predictive biomarker of response to immune checkpoint inhibitors (ICIs) in cancer [28], we anticipated that knockdown of TMPRSS2 would promote the response to ICIs in LUAD. As expected, the volume of the TMPRSS2-knockdown tumors had a significantly higher level of decreases than that of TMPRSS2-wildtype tumors after treatment with BMS-1, an inhibitor of PD-1/PD-L1 (Fig. 5k); this result supports that knockdown of TMPRSS2 can enhance the sensitivity of LUAD to the PD-1/PD-L1 inhibitor. Furthermore, the activities of CD8 + TILs and NK TILs markedly increased in TMPRSS2-knockdown tumors after treatment with BMS-1; they were significantly higher in TMPRSS2-knockdown than in TMPRSS2-wildtype tumors after treatment with BMS-1 (Fig. 5l, m). These results support that the PD-1/PD-L1 inhibitor promotes immune elimination of tumor cells by inhibiting the exhaustion of CD8 + TILs and NK TILs in TMPRSS2-depleted LUAD.

To summarize, bioinformatics analysis revealed a negative correlation between TMPRSS2 abundance and immune infiltration levels in LUAD. Experimental results demonstrated that this relationship was a causal relationship. That is, reduced TMPRSS2 abundance can boost immune infiltration for LUAD.

Discussion

As a pivotal molecule in the regulation of SARS-CoV-2 invading human host cells, TMPRSS2 is attracting massive attention in the current SARS-CoV-2 pandemic [29,30,31]. Because SARS-CoV-2 has and is infecting large numbers of people, including many cancer patients, an investigation into the role of TMPRSS2 in cancer may provide valuable advice for treating cancer patients infected with SARS-CoV-2. Previous studies of TMPRSS2 in cancer mainly focused on its oncogenic role in prostate cancer [6,7,8]. In this study, we focused on LUAD, considering that it is the most common histological type in lung cancer and that the lungs are the primary organ SARS-CoV-2 attacks. TMPRSS2 plays a tumor suppressive role in LUAD, as we have provided abundant evidence. First, TMPRSS2 downregulation correlates with elevated activities of many oncogenic pathways in LUAD, including cell cycle, mismatch repair, p53, and ECM signaling. Second, TMPRSS2 downregulation correlates with increased tumor cell proliferation, stemness, genomic instability, and ITH in LUAD. Finally, TMPRSS2 downregulation is associated with tumor advancement and worse survival in LUAD. Furthermore, both in vitro and in vivo experiments demonstrated that TMPRSS2 downregulation markedly promoted the proliferation and invasion capacity of LUAD cells, supporting the tumor suppressor role of TMPRSS2 in LUAD.

Our analysis indicates a significant association between TMPRSS2 expression and DR pathways’ activity. That is, TMPRSS2 expression is downregulated in the DR pathway-mutated LUAD patients relative to those DR pathway-wildtype patients. Several factors could be responsible for this association. First, TMPRSS2 downregulation can markedly promote cell cycle and proliferation abilities of LUAD cells to alter the activity of DR pathways. Second, TMPRSS2 could directly interact with key proteins regulating the DR pathways. For instance, we have unveiled a significant negative correlation between TMPRSS2 expression and the expression of RAD51, a key factor for homologous recombination repair. Lastly, TMPRSS2 downregulation may promote the expression of DR-associated molecules, such as MSH2, MSH6, POLE, and PCNA, thereby altering the activity of DR pathways.

Our bioinformatics analysis revealed significant negative associations between TMPRSS2 expression and immune signatures, including both immune-stimulatory and immune-inhibitory signatures, in LUAD (Fig. 1a). Nevertheless, TMPRSS2 expression tended to have a stronger negative correlation with immune-inhibitory signatures than with immune-stimulatory signatures in LUAD (Fig. 1b). The significantly different levels of correlations of immune-stimulatory and immune-inhibitory signatures with TMPRSS2 expression could be a factor responsible for the worse prognosis in LUAD patients with TMPRSS2 deficiency. Furthermore, the associations between TMPRSS2 and tumor immunity in LUAD were completely verified by both in vitro and in vivo experiments. That is, knockdown of TMPRSS2 significantly increased tumor immunogenicity and immune cell infiltration in LUAD. On the other hand, both computational and experimental data showed that TMPRSS2 downregulation significantly enhanced PD-L1 expression in LUAD. Because both inflamed tumor microenvironment and PD-L1 expression are determinants of cancer responses to immunotherapy [32], TMPRSS2-depleted LUAD would respond better to immunotherapy than TMPRSS2-wildtype LUAD. This was supported by our in vivo experiments showing that TMPRSS2-knockdown tumors were more sensitive to the PD-1/PD-L1 inhibitor. Thus, TMPRSS2 downregulation is a positive biomarker of immunotherapy for LUAD. In addition, because TMPRSS2 downregulation often occurs in advanced LUAD, it indicates that advanced LUAD could benefit more from immunotherapy than early-stage LUAD. To summarize, enhanced PD-L1 expression, TMB and tumor immune infiltration collectively promote immunotherapy response in the TMPRSS2-depleted LUAD subtype.

It is crucial to prevent COVID-19 patients with lung cancer from acute progress in the beginning stage of SARS-CoV-2 infection, since the pneumonia caused by SARS-CoV-2 infection’s acute progress will damage the function of the lungs that pose a major threat to lung cancer patients’ life. TMPRSS2 inhibition has been indicated as a strategy for treating SARS-CoV-2 infection for the essential role of TMPRSS2 in the SARS-CoV-2 invasion [30, 33]. However, our data suggest that this strategy may not be a good option for lung cancer patients in terms of the tumor suppressor role of TMPRSS2 in LUAD. Interestingly, we found that TMPRSS2 displayed significantly higher expression levels in non-smoker than in smoker LUAD patients (Student’s t test, p < 0.05, FC > 1.5) (Fig. 6a). This result indicates that non-smoker LUAD patients likely have a better prognosis than smoker LUAD patients. Meanwhile, it indicates that non-smoker LUAD patients could be more susceptible to SARS-CoV-2 infection than smoker LUAD patients. It is in line with some reports that smoking is associated with a lower risk of SARS-CoV-2 infection [34, 35]. Therefore, the use of TMPRSS2 inhibition strategies in COVID-19 patients with lung cancer should be cautious. As expected, non-smoker LUAD patients had significantly lower TMB and antitumor immunity than smoker LUAD patients (Fig. 6b), consistent with findings from previous studies [36, 37].

Fig. 6
figure 6

Comparisons of TMPRSS2 expression levels, TMB, and immune signatures between non-smoker and smoker LUADs. Non-smoker LUAD patients showing significantly higher TMPRSS2 expression levels (a) and lower TMB and immune signature scores (b) than smoker LUAD patients. The two-tailed Student’s t test and one-tailed Mann–Whitney U test p values are shown in (a) and (b), respectively

Conclusions

TMPRSS2 is a tumor suppressor in LUAD, as evidenced by its downregulation correlated with increased tumor proliferation, stemness, genomic instability and ITH, tumor progression, and unfavorable clinical outcomes in LUAD. However, TMPRSS2 downregulation is a positive biomarker of immunotherapy for LUAD. Our data provide implications in the connection between lung cancer and pneumonia caused by SARS-CoV-2 infection, as well as significant clinical implications for LUAD therapy.

Methods

Datasets

We downloaded RNA-Seq gene expression profiling (level 3 and RSEM normalized), protein expression profiling, and clinical data for the TCGA-LUAD cohort from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/). We downloaded microarray gene expression profiling (normalized) and clinical data for other four LUAD cohorts (GSE12667 [38], GSE30219 [39], GSE31210 [40], and GSE50081 [41]) from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/). Moreover, we downloaded two scRNA-seq data for LUAD, including GSE131907 [12] and Maynard corhort [13]. The proteomic dataset CPTAC-LUAD was downloaded from CPTAC (https://gdc.cancer.gov/about-gdc/contributed-genomic-data-cancer-research/clinical-proteomic-tumor-analysis-consortium-cptac). In addition, we collected 100 blood samples from LUAD patients and 20 blood samples from healthy persons from Jiangsu Cancer Hospital, China. The studies were “approved by Jiangsu Cancer Hospital.” According to the diagnosis and treatment guidelines for non-small cell lung cancer (CSCO 2020), LUAD patients in this study were divided into two groups: 50 patients in early stage (stage I) and 50 patients in late stage (stage III-IV). We log2-transformed the RNA-Seq gene expression values before further analyses. A description of these datasets is shown in Supplementary Table S1.

Patient and public involvement

The study was done in accordance with both the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines and was approved by the institutional review board.

scRNA-seq data pre-processing

We analyzed two LUAD scRNA-seq datasets GSE131907 [12] (10x) and Maynard cohort [13] (smart-seq2). In GSE131907, the gene expression values were the unique molecular identifier (UMI) data which we normalized using the “NormalizeData()” function in the R package “Seurat” (v4.0.6) with the default parameters. That is, the UMI value of each cell was normalized by size-factor 10,000 and then ln(x + 1) transformed. For the Maynard cohort dataset, we used the normalized count values of gene expression.

Gene-set enrichment analysis

We quantified the enrichment levels of immune signatures, pathways, and tumor phenotypes in tumors by the single-sample gene-set enrichment analysis (ssGSEA) [24] of their marker gene sets. The ssGSEA was performed with the R package “GSVA” [24]. The marker gene sets are presented in Supplementary Table S2. We used GSEA [42] to identify KEGG [43] pathways significantly associated with a gene set with a threshold of adjusted p value < 0.05. We used WGCNA [25], an R package, to identify gene modules and their associated GO terms enriched in the high- (upper third) and low-TMPRSS2-expression-level (bottom third) LUADs.

Survival analysis

We compared OS and DFS between the high- (upper third) and low-TMPRSS2-expression-level (bottom third) LUAD patients. Kaplan–Meier curves were utilized to display survival time differences, whose significances were evaluated by the log-rank test. We performed the survival analyses using the R package “survival”. Moreover, we performed multivariate survival analysis using the Cox proportional hazards model to explore the correlation between TMPRSS2 expression and OS prognosis after correcting confounding variables, including TMPRSS2 expression, age, tumor stage, and enrichment levels of immune cells (CD8 + T cells and CD4 + regulatory T cells). The “age”, “CD8 + T cells enrichment”, and “CD4 + regulatory T cells enrichment” were continuous variables, and both “TMPRSS2 expression” (high versus low) and “tumor stage” (early versus late) were binary variables. We implemented the multivariate survival analysis using the function “coxph” in the R package “survival”.

Statistical analysis

We used the Spearman correlation to evaluate associations between TMPRSS2 expression levels and ssGSEA scores of gene sets; the Spearman correlation coefficients (ρ) and p values were reported. In addition, we used the Pearson correlation to evaluate associations between TMPRSS2 expression levels and gene or protein expression levels and the ratios of immune signatures; the Pearson correlation coefficients (r) were reported. The ratios between immune signatures were the log2-transformed values of the ratios between the geometric mean expression levels of all marker genes in immune signatures. In comparisons of TMPRSS2 expression levels between different groups of samples, we used the two-tailed Student’s t test for two groups and the one-way ANOVA test for more than two groups. We performed the statistical analyses using the R programming software (https://cran.r-project.org/).

In vitro experiments

Antibodies, reagents and cell lines

All antibodies were used at a dilution of 1:1000 unless otherwise specified. Anti-PD- L1 (ab213480), anti-CD8 (ab22378), anti-CD49b (ab181548), anti-MSH6 (ab92471), anti-TMPRSS2 (ab109131) and anti-GAPDH (ab181603) were purchased from Abcam (Burlingame, CA). Anti-PD-L1 (66248-1-Ig) and anti-MSH6 (66172-1-Ig) in supplementary materials were purchased from Proteintech Group, Inc.PE anti-mouse TNF-α antibody (12-7321-81), APC anti-mouse IFN-γ antibody (17-7311-81), APC anti-mouse CD279 (PD-1) antibody (12-9985-81), and APC anti-mouse CD223 (LAG-3) antibody (12-2231-81) were purchased from eBioscience (San Diego, CA). The human lung cancer cell lines A549, H1975, and H1299 were from the American Type Culture Collection. They were cultured in 90% F12K (GIBCO, USA) supplemented with 10% fetal bovine serum in a humidified incubator at 37 °C and 5% CO2. NK92 cells (KeyGEN BioTECH, Nanjing, China) were cultured in Alpha MEM (GIBCO, USA) with 2 mM L-glutamine, 1.5 g/L sodium bicarbonate, 0.2 mM inositol, 0.1 mM 2-mercaptoethanol, 0.02 mM folic acid, 100–200 U/mL recombinant human IL-2 (PeproTech, Rocky Hill, New Jersey, USA), and a final concentration of 12.5% horse serum and 12.5% fetal bovine serum.

TMPRSS2 knockdown with small interfering RNA (siRNA)

A549 cells were transfected with TMPRSS2 siRNA or control siRNA by using Effectene Transfection Reagent (Qiagen, Hilden, Germany, B00118) according to the manufacturer’s instructions. The medium was replaced after 24 h incubation with fresh medium, and the cells were maintained for a further 24 h. Quantitative PCR or Western blotting were used to detect the transfection efficiency. TMPRSS2 siRNA and control siRNA were synthesized by KeyGEN Biotech (Nanjing, China). Their sequences were as follows: TMPRSS2 siRNA: 1, 5'- GGAC AUGG GCUA UAAG AAU -3' (sense) and 5'- AUUC UUAU AGCC CAUG UCC-3' (antisense); 2, 5'- ACUC CAAG ACCA AGAA CAA -3' (sense) and 5'- UUGU UCUU GGUC UUGG AGU-3' (antisense); 3,5'-GGAC UGGA UUUA UCGA CAA-3'(sense) and 5'-UUGU CGAU AAAU CCAG UCC-3' (antisense); control siRNA: 5'-UUCU CCGA ACGU GUCA CGU dTdT-3' (sense) and 5'-ACGUGACACGUUCGGAGAAdTdT-3' (antisense).

Lentivirus generation and infection

Lentivirus was prepared according to the manufacturer’s instructions. The heteroduplexes, supplied as 58-nucleotide oligomers, were annealed; the downstream of the U6 promoter was inserted into the pLKO.1 plasmid to generate pLKO.1/ShTMPRSS2. Recombinant and control lentiviruses were produced by transiently transfecting pLKO.1/vector and pLKO.1/ShTMPRSS2, respectively. The lentiviruses were transfected into 293 T cells. After 48 h, lentiviral particles were collected and concentrated from the supernatant by ultracentrifugation. Effective lentiviral shRNA was screened by infecting these viruses with Lewis cells, and their inhibitory effect on TMPRSS2 expression was analyzed by quantitative PCR and Western blotting. The lentivirus containing the ShTMPRSS2 RNA target sequences and a control virus were used for the animal study. The coding strand sequence of the shRNA-encoding oligonucleotides was 5’-ACGGGAACGTGACGGTATTTA-3’ for TMPRSS2.

Western blotting

A549, H1975 and H1299 cell extracts were lysed by using lysis buffer supplemented with protease inhibitor cocktail immediately before use. Total proteins present in the cell lysates were quantified by using the BCA assay. Proteins were denatured by addition of 6 volumes of SDS sample buffer and boiled at 95 °C for 5 min and were then separated by SDS-PAGE. The resolved proteins were transferred onto a nitrocellulose membrane after electrophoresis. The membranes were incubated with 5% skimmed milk in TBS containing 0.1% Tween 20 (TBS-T) for 1 h to block the non-specific binding and then incubated overnight at 4 °C with specific antibodies. After 2 h incubation with the HRP-labeled secondary antibody, proteins were visualized by enhanced chemiluminescence using a G: BOX chemiXR5 digital imaging system (SYNGENE, UK). The band densities were normalized to the background, and the relative optical density ratios were calculated relative to the housekeeping gene GAPDH.

Quantitative PCR

The total RNA was isolated by Trizol (Invitrogen, USA) and was reversely transcribed into cDNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher, USA). Quantitative PCR was performed with the ABI Step one plus Real-Time PCR (RT-PCR) system (ABI, USA) using One Step TB Green™ PrimeScript™ RT-PCR Kit II (SYBR Green) (RR086B, TaKaRa, JAPAN). Relative copy number was determined by calculating the fold-change difference in the gene of interest relative to GAPTH. The program for amplification was one cycle of 95 °C for 5 min, followed by 40 cycles of 95 °C for 15 s, 60 °C for 20 s, and 72 °C for 40 s. The relative amount of each gene was normalized to the amount of GAPDH. The primer sequences were as follows: hTMPRSS2: 5'-AACT TCAT CCTT CAGG TGTA-3' (forward) and 5'-TCTC GTTC CAGT CGTCTT-3' (reverse); hGAPDH: 5'- AGAT CATC AGCA ATGC CTCCT-3' (forward) and 5'-ACAC CATG TATT CCGG GTCAAT-3' (reverse).

Cell proliferation assay

A549, H1975 and H1299 cells were plated in 96-well plates at 3 × 104 cells per well and maintained in a medium containing 10% FBS. After 24 h, cell proliferation was determined using the Cell Counting Kit-8 (CCK-8; KeyGEN Biotech, China) following the manufacturer’s instructions. To perform the CCK-8 assay, 10 µl CCK-8 reagent was added to each well and the 96 plates were incubated at 37 °C for 2 h. The optical density was read at 450 nm using a microplate reader. All these experiments were performed in triplicates.

Transwell migration and invasion assays

Cell migratory and invasive abilities were assessed using 24 well transwell chambers (Corning, USA) with membrane pore size of 8.0 µm. A549, H1975 and H1299 cells were seeded into the upper chamber without matrigel at 1 × 105 cells in serum-free medium, while 500 µl medium containing 20% FBS was added to the lower chamber. The chambers were incubated at 37 °C and 5% CO2 for 24 h. The cells on the upper chamber were scraped off with cotton-tipped swabs, and cells that had migrated through the membrane were stained with 0.1% crystal violet at 37 °C for 30 min. The migrated cells were counted at 200x magnification under the microscope using three randomly selected visual fields. All these experiments were performed in triplicates.

Co-culture of tumor cells with NK92 cells

A transwell chamber (Corning, USA) was inserted into a six well plate to construct a co-culture system. A549 cells were seeded on the six well plate at a density of 5 × 104 cells/well, and NK92 cells were seeded on the membrane (polyethylene terephthalate, pore size of 0.4 µm) of the transwell chamber at a density of 5 × 104 cells/chamber. Tumor cells and NK92 cells were co-cultured in a humidified incubator at 37 °C and 5% CO2 atmosphere for 48 h.

EdU proliferation assay

After co-culture of A549 cells with NK92 cells for 48 h, we measured the proliferation capacity of NK92 cells by an EdU (5- ethynyl-2'-deoxyuridine; Invi-trogen, California, USA) proliferation assay. NK92 cells were plated in 96-well plates with a density of 2 × 103 cells/well with 10 µM EdU at 37 °C for 24 h. The cell nuclei were stained with 4',6- diamidino-2-phenylindole (DAPI) at a concentration of 1 µg/mL for 20 min. The proportion of NK92 cells incorporating EdU was detected with fluorescence microscopy. All the experiments were performed in triplicates.

In vivo experiments

In vivo mouse models

Lewis tumor cells were transduced with ShCon (scramble) or ShTMPRSS2 lentivirus and selected by puromycin for 7 days. The stably transfected Lewis tumor cells (1 × 107/ml) were subcutaneously injected into the right armpit of recipient mice after shaving the injection site. After 5 days, when the tumor volume was approximately 4–5 mm3, the mice were randomly divided into six groups, with half of the ShCon and ShTMPRSS2 mice treated with 150 U/L PD1/PDL1 inhibitor BMS-1 (concentration 500 mg/mL; i.p.) (MCE Cat. No. HY-19991) every 3 days. The tumors were isolated from mice after 15 days. Tumor volumes did not exceed the maximum allowable size according to the LJI IACUC animal experimental protocol. The tumor volume was measured every 3 days after the tumor appeared on the fifth day and was calculated as follows: V = 1/2 × width2 × length. The studies were “approved by Nanjing Medical University.”

Isolation of TILs

After the tumor tissues were separated aseptically and rinsed with cold PBS for 3 times, they were excised and chopped with tweezers and scissors and were then digested with 2 mg/mL collagenase (type IV, sigma V900893) for 45 min, until no tissue mass was visible. Following digestion, lymphocytes were separated with lymphocyte separation medium, washed with PBS, and counted. The specific protocol was as follows: tumors were filtered through 70 µM cell strainers, and the cell suspension was washed twice in culture medium by centrifugation at 1500 rpm and 4 °C for 10 min. After the washing, the cells were resuspended with PBS and were layered over 3 mL of 30%-100% gradient percoll (Beijing Solarbio Science & Technology, Beijing, China); this was followed by centrifugation at 2600 rpm for 25 min at 25 °C. The enriched TILs were obtained at the interface as a thin buffy layer, were washed with PBS three times, and finally were resuspended in FACS staining buffer for further staining procedures.

Flow cytometry

TILs were stained with CD8 (eBioscience, 11-0081-81), CD49b (eBioscience, 11-5971-81), PD-1 (eBioscience, 12-9985-81), and LAG3 (eBioscience, 12-2231-81) and were analyzed by flow cytometry. TILs were restimulated with cell stimulation cocktail (eBioscience, San Diego, California, USA), and the expression of IFN-γ and TNF-α (Biolegend) was analyzed by flow cytometry. Staining for cell surface markers was performed by incubating cells with antibody (1:100 dilution) in FACS buffer (0.1% BSA in PBS) for 30 min at 4 °C. Surface markers of intracellular cytokines (IFN-γ (eBioscience, 17-7311-81) and TNF-α (eBioscience, 12-7321-81)) were stained before fixation/permeabi-lization (Intracellular Fixation & Permeabilization Buffer Set, ThermoFisher).

Immunofluorescence of CD8, CD49b and PD-L1

Paraffin-embedded mice tumor tissue section (3 µm thick) were subjected to immunofluorescence with CD8 (Abcam, ab22378), CD49b (Abcam, ab181548), or PD-L1 (Abcam, ab2134808) primary antibodies. Before immunostaining, tumor tissue sections were deparaffinized with xylene, rehydrated and unmasked in sodium citrate buffer (10 mM, pH 6.0), and treated with a glycine solution (2 mg/mL) to quench autofluorescence. After antigen retrieval, 3% H2O2-methanol solution blocking inactivated enzymes, and goat serum blocking, tissue slides were incubated in wet box for 2 h at 37 °C with anti-CD8, CD49b, or anti-PD-L1 rabbit primary antibodies (1:100 dilution) in blocking solution, and were then dropped with FITC (1:100 dilution) secondary antibody 50-100ul and incubated at 37° for 1 h in the dark. The immunolabeled slides were examined with a fluorescence microscope after nuclear counterstaining with DAPI. Green, red and blue channel fluorescence images were acquired with a Leica DFC310 FX 1.4-megapixel digital color camera equipped with LAS V.3.8 software (Leica Microsystems, Wetzlar, Germany). Overlay images were reconstructed by using the free-share ImageJ software.

Availability of data and materials

The five LUAD genomic datasets were obtained from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/) and the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/).

Abbreviations

ACE2:

Angiotensin-converting enzyme 2

CCK-8:

The Cell Counting Kit-8

CPTAC:

Clinical Proteomic Tumor Analysis Consortium

DAPI:

4',6- Diamidino-2-phenylindole

DR:

DNA repair

DFS:

Disease-free survival

ECM:

Extracellular matrix

FC:

Fold change

FDR:

False discovery rate

GO:

Gene ontology

GSEA:

Gene set enrichment analysis

HRD:

Homologous recombination deficiency

ICIs:

Immune checkpoint inhibitors

ITH:

Intratumor heterogeneity

LUAD:

Lung adenocarcinoma

LUSC:

Lung squamous cell carcinoma

MDSCs:

Myeloid-derived suppressor cells

OS:

Overall survival

PI:

Proximal-inflammatory

PP:

Proximal-proliferative

RT-PCR:

Real-Time PCR

S:

Spike glycoprotein

SARS-CoV-2:

Severe acute respiratory syndrome coronavirus 2

scRNA-seq:

Single‐cell RNA sequencing

siRNA:

Small interfering RNA

ssGSEA:

Single-sample gene-set enrichment analysis

TCGA:

The Cancer Genome Atlas

TILs:

Tumor-infiltrating lymphocytes

TMB:

Tumor mutation burden

TMPRSS2:

Transmembrane protease serine 2

TRU:

Terminal respiratory unit

UMI:

Unique molecular identifier

WGCNA:

Weighted gene co-expression network analysis

References

  1. Ou X, et al. Characterization of spike glycoprotein of SARS-CoV-2 on virus entry and its immune cross-reactivity with SARS-CoV. Nat Commun. 2020;11(1):1620.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  2. Tomlins SA, et al. Role of the TMPRSS2-ERG gene fusion in prostate cancer. Neoplasia. 2008;10(2):177–88.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  3. Katopodis P, et al. Pan-cancer analysis of transmembrane protease serine 2 and cathepsin L that mediate cellular SARS-CoV-2 infection leading to COVID-19. Int J Oncol. 2020;57(2):533–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Kong Q, et al. Analysis of the susceptibility of lung cancer patients to SARS-CoV-2 infection. Mol Cancer. 2020;19(1):80.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Bao R, et al. ACE2 and TMPRSS2 expression by clinical, HLA, immune, and microbial correlates across 34 human cancers and matched normal tissues: implications for SARS-CoV-2 COVID-19. J Immunother Cancer. 2020;8(2):e001020.

    Article  PubMed  Google Scholar 

  6. Mollica V, Rizzo A, Massari F. The pivotal role of TMPRSS2 in coronavirus disease 2019 and prostate cancer. Future Oncol. 2020;16(27):2029–33.

    Article  CAS  PubMed  Google Scholar 

  7. Trifunovski A, et al. Detection of TMPRSS2-ERG fusion transcript in biopsy specimen of prostate cancer patients: a single centre experience. Pril (Makedon Akad Nauk Umet Odd Med Nauki). 2020;41(1):5–14.

    PubMed  Google Scholar 

  8. Dudka I, et al. Comprehensive metabolomics analysis of prostate cancer tissue in relation to tumor aggressiveness and TMPRSS2-ERG fusion status. BMC Cancer. 2020;20(1):437.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Luo L, et al. TMPRSS2 correlated with immune infiltration serves as a prognostic biomarker in prostatic adenocarcinoma: implication for the COVID-2019. Front Genet. 2020;11:575770.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Takahashi H, et al. Immune cytolytic activity for comprehensive understanding of immune landscape in hepatocellular carcinoma. Cancers (Basel). 2020;12(5):1221.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Miranda A, et al. Cancer stemness, intratumoral heterogeneity, and immune response across cancers. Proc Natl Acad Sci U S A. 2019;116(18):9020–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Kim N, et al. Single-cell RNA sequencing demonstrates the molecular and cellular reprogramming of metastatic lung adenocarcinoma. Nat Commun. 2020;11(1):2285.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Maynard A, et al. Therapy-induced evolution of human lung cancer revealed by single-cell RNA sequencing. Cell. 2020;182(5):1232-1251 e22.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Li WY, et al. The role of EGFR mutation as a prognostic factor in survival after diagnosis of brain metastasis in non-small cell lung cancer: a systematic review and meta-analysis. BMC Cancer. 2019;19(1):145.

    Article  PubMed  PubMed Central  Google Scholar 

  15. Cancer Genome Atlas Research, N., Comprehensive molecular profiling of lung adenocarcinoma. Nature. 2014;511(7511):543–50.

  16. Ferguson LR, et al. Genomic instability in human cancer: molecular insights and opportunities for therapeutic attack and prevention through diet and nutrition. Semin Cancer Biol. 2015;35 Suppl(Suppl):S5-s24.

    Article  PubMed  Google Scholar 

  17. Palmieri G, et al. Genetic instability and increased mutational load: which diagnostic tool best direct patients with cancer to immunotherapy? J Transl Med. 2017;15(1):17.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Smith JC, Sheltzer JM. Systematic identification of mutations and copy number alterations associated with cancer patient prognosis. Elife. 2018;7:e39217.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Knijnenburg TA, et al. Genomic and molecular landscape of DNA damage repair deficiency across the Cancer Genome Atlas. Cell Rep. 2018;23(1):239-254.e6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. McKinney JA, et al. Distinct DNA repair pathways cause genomic instability at alternative DNA structures. Nat Commun. 2020;11(1):236.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Li M, et al. An algorithm to quantify intratumor heterogeneity based on alterations of gene expression profiles. Commun Biol. 2020;3(1):505.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Eischen CM. Genome stability requires p53. Cold Spring Harb Perspect Med. 2016;6(6):a026096.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Burrell RA, et al. The causes and consequences of genetic heterogeneity in cancer evolution. Nature. 2013;501(7467):338–45.

    Article  CAS  PubMed  Google Scholar 

  24. Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Walker C, Mojares E, Del Rio Hernandez A. Role of extracellular matrix in development and cancer progression. Int J Mol Sci. 2018;19(10):3028.

    Article  PubMed  PubMed Central  Google Scholar 

  27. Gillette MA, et al. Proteogenomic characterization reveals therapeutic vulnerabilities in lung adenocarcinoma. Cell. 2020;182(1):200-225. e35.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Patel SP, Kurzrock R. PD-L1 expression as a predictive biomarker in cancer immunotherapy. Mol Ther. 2015;14(4):847–56.

    Article  CAS  Google Scholar 

  29. Zang R, et al. TMPRSS2 and TMPRSS4 promote SARS-CoV-2 infection of human small intestinal enterocytes. Sci Immunol. 2020;5(47):eabc3582.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Hoffmann M, et al. SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor. Cell. 2020;181(2):271-280 e8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Bestle D, et al. TMPRSS2 and furin are both essential for proteolytic activation of SARS-CoV-2 in human airway cells. Life Sci Alliance. 2020;3(9):e202000786.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Cristescu R, et al. Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy. Science. 2018;362(6411):eaar3593.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Stopsack KH, et al. TMPRSS2 and COVID-19: serendipity or opportunity for intervention? Cancer Discov. 2020;10(6):779–82.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Paleiron N, et al. Impact of tobacco smoking on the risk of COVID-19: a large scale retrospective cohort study. Nicotine Tob Res. 2021;23(8):1398–404.

    Article  CAS  PubMed  Google Scholar 

  35. Li M, et al. Identifying novel factors associated with COVID-19 transmission and fatality using the machine learning approach. Sci Total Environ. 2021;764:142810.

  36. Chae YK, et al. Association of tumor mutational burden with DNA repair mutations and response to Anti-PD-1/PD-L1 therapy in non-small-cell lung cancer. Clin Lung Cancer. 2019;20(2):88-96 e6.

    Article  CAS  PubMed  Google Scholar 

  37. Kinoshita T, et al. Determination of poor prognostic immune features of tumour microenvironment in non-smoking patients with lung adenocarcinoma. Eur J Cancer. 2017;86:15–27.

    Article  PubMed  Google Scholar 

  38. Ding L, et al. Somatic mutations affect key pathways in lung adenocarcinoma. Nature. 2008;455(7216):1069–75.

  39. Rousseaux S, et al. Ectopic activation of germline and placental genes identifies aggressive metastasis-prone lung cancers. Sci Transl Med. 2013;5(186):186ra66.

  40. Okayama H, et al. Identification of genes upregulated in ALK-positive and EGFR/KRAS/ALK-negative lung adenocarcinomas. Cancer Res. 2012;72(1):100–11.

  41. Der SD, et al. Validation of a histology-independent prognostic gene signature for early-stage, non-small-cell lung cancer including stage IA patients. J Thorac Oncol. 2014;9(1):59–64.

  42. Subramanian A, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545–50.

  43. Kanehisa M, et al. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res. 2017;45(D1):D353–D361.

Download references

Acknowledgements

Not applicable.

Funding

This work was supported by the China Pharmaceutical University (grant number 3150120001 to XW), Natural Science Foundation of Jiangsu Province (grant number BK20201090 to ZL), and China Postdoctoral Science Foundation (grant number 2021M691338 to ZL), China International Medical Foundation (Z-2021-46-2101-2023 to ZL).

Author information

Authors and Affiliations

Authors

Contributions

Zhixian Liu: Validation, Formal analysis, Resources, Investigation, Data curation, Visualization, Writing—original draft, Funding acquisition. Qiqi Lu: Software, Formal analysis, Visualization, Writing—review & editing. Zhilan Zhang: Software, Formal analysis, Investigation, Data curation, Visualization. Qiushi Feng: Software, Formal analysis, Visualization. Xiaosheng Wang: Conceptualization, Methodology, Resources, Investigation, Writing—original draft, Writing—review & editing, Supervision, Project administration, Funding acquisition.

Corresponding author

Correspondence to Xiaosheng Wang.

Ethics declarations

Ethics approval and consent to participate

The study was done in accordance with both the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines and was approved by the Ethics Committee of Nanjing Medical University and Experimental Animal Welfare Ethics Committee of Nanjing Medical University.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

12931_2024_2870_MOESM1_ESM.xlsx

Additional file 1: Table S1. A summary of the datasets analyzed. Table S2. The gene sets representing immune signatures, pathways, and tumor phenotypes. Table S3. The genes with strong positive and negative expression correlations with TMPRSS2 in the TCGA-LUAD cohort.

12931_2024_2870_MOESM2_ESM.pdf

Additional file 2: Figure S1. Validation of the mRNA-based findings at the protein level in CPTAC-LUAD. (a) Kaplan-Meier survival curves showing that LUAD patients with lower TMPRSS2 expression levels (bottom third) have worse OS and MFS than those with higher TMPRSS2 expression levels (upper third). The log-rank test p values are shown. OS, overall survival. MFS, metastasis-free survival. (b) The expression of TMPRSS2 correlates inversely with Ki-67 expression, the enrichment of the cell cycle, mismatch repair, and p53 signaling pathways and the stemness signature in LUAD. (c) TMPRSS2 is more highly expressed in EGFR-wildtype than in EGFR-mutated LUADs. (d) TMPRSS2 is more lowly expressed in TP53-wildtype than in TP53-mutated LUADs and shows negative expression correlations with DR-associated proteins (MSH2, MSH6, and PCNA) in LUAD. (e) TMPRSS2 expression correlates inversely with the enrichment of CD8+ T cells, cytolytic activity, PD-L1, and MDSCs. The Pearson or Spearman correlation coefficients and p values are shown in (b, d, e). * p < 0.05, ** p < 0.01, *** p < 0.001, ns p ≥ 0.05.

12931_2024_2870_MOESM3_ESM.pdf

Additional file 3: Figure S2. TMPRSS2 knockdown markedly promoted proliferative and invasive abilities in another two lung adenocarcinoma cells. (A&B). TMPRSS2 knockdown markedly promoted proliferative and invasive abilities of H1975 cells. C. TMPRSS2 knockdown increased MSH6 and PD-L1 expression in H1975 cells. (D&E). TMPRSS2 knockdown markedly promoted proliferative and invasive abilities of H1299 cells. F. TMPRSS2 knockdown increased MSH6 and PD-L1 expression in H1299 cells.

Additional file 4: Figure S3. Full uncropped Gels and Blots images.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Liu, Z., Lu, Q., Zhang, Z. et al. TMPRSS2 is a tumor suppressor and its downregulation promotes antitumor immunity and immunotherapy response in lung adenocarcinoma. Respir Res 25, 238 (2024). https://doi.org/10.1186/s12931-024-02870-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12931-024-02870-7

Keywords