Clinical and prognostic implications of an immune‐related risk model based on TP53 status in lung adenocarcinoma

Abstract TP53 mutation is the most widespread mutation in lung adenocarcinoma (LUAD). Meanwhile, p53 (encoded by TP53) has recently been implicated in immune responses. However, it is still unknown whether TP53 mutation remodels the tumour microenvironment to influence tumour progression and prognosis in LUAD. In this study, we developed a 6‐gene immune‐related risk model (IRM) to predict the survival of patients with LUAD in The Cancer Genome Atlas (TCGA) cohort based on TP53 status, and the predictive ability was confirmed in 2 independent cohorts. TP53 mutation led to a decreased immune response in LUAD. Further analysis revealed that patients in the high‐index group had observably lower relative infiltration of memory B cells and regulatory T cells and significantly higher relative infiltration of neutrophils and resting memory CD4+ T cells. Additionally, the IRM index positively correlated with the expression of critical immune checkpoint genes, including PDCD1 (encoding PD‐1) and CD274 (encoding PD‐L1), which was validated in the Nanjing cohort. Furthermore, as an independent prognostic factor, the IRM index was used to establish a nomogram for clinical application. In conclusion, this IRM may serve as a powerful prognostic tool to further optimize LUAD immunotherapy.


Introduction
Non-small cell lung cancer (NSCLC) accounts for 85% of all lung cancers, and lung adenocarcinoma (LUAD) is the most frequent NSCLC subtype 1 . Immunotherapy has been integrated into the rst-and second-line treatment strategies for NSCLC, reviving enthusiasm in explaining the prognostic and pathophysiological role of the tumor microenvironment (TME) [2][3][4][5] . However, it is not known whether the immune signature of lung cancer could act as a biomarker to reliably estimate disease prognosis and patient survival.
TP53 is a tumor suppressor gene, and it is one of the 5 most conspicuous mutations in LUAD, though there has not yet been approval for any related molecular inhibitors by the Food and Drug Administration. TP53 encodes p53, a master regulatory transcription factor that controls multiple core programs in cells, including cell cycle arrest, apoptosis, senescence, fertility, and metabolism [6][7][8][9][10] . Recently, p53 has been implicated in immune responses 11 . TP53 mutation remarkably affects the expression of immune checkpoints and activated T cell immune response and may thus serve as a potential predictive factor for guiding immunotherapy 11 . However, the relationship between TP53 mutation and the regulation of immune signaling and responses is still unknown. We hypothesized that TP53 mutation may remodel the TME to in uence tumor progression and prognosis in LUAD.
Recently, computational methods based on transcriptome data were proposed to characterize the immune landscape in the sequenced tumor tissue 12 . In this study, we systemically screened the expression pro les of RNA sequencing data from public databases and developed a 6-gene immunerelated risk model (IRM) based on TP53 mutation to predict the survival of patients with LUAD. We further validated the 6-gene IRM index in the meta-Gene Expression Omnibus (GEO) and Nanjing cohorts.
Through multiple veri cation methods, we determined the IRM index to be an independent prognostic biomarker that could accurately predict the 5-and 7-year overall survival (OS) of patients with LUAD. The potential mechanism, immune in ltration, and expression of immune-related checkpoints of the IRM index were also explored. We believe that this robust prognostic IRM index will improve risk strati cation, provide more exact judgment for individualized clinical management, and serve as a potential immunotherapy biomarker for patients with LUAD.

RNA sequencing data and immune-related genes
The nonsynonymous mutation status of 512 patients with LUAD (work ow type: Mutect2 pipeline), gene expression data (work ow type: HTSeq-Counts), and corresponding clinical metadata from The Cancer Genome Atlas (TCGA) website (https://gdc.cancer.gov/) were downloaded using the "TCGAbiolinks" R package (version 2.14.1; https://bioconductor.org/packages/release/bioc/html/TCGAbiolinks.html). Of these patients, 499 with RNA sequencing data and TP53 mutation status data were included in the following analysis. Using the "maftools" R package (version 2.2.10; https://www.bioconductor.org/packages/release/bioc/html/maftools.html), we identi ed the mutation status of these patients, including 239 TP53 MUT patients and 260 TP53 WT patients. Entrez IDs were converted to gene symbols using the Bioconductor package "org.Hs.eg.db" (version 3.10.0; https://bioconductor.org/packages/release/data/annotation/html/org.Hs.eg.db.html). Genes with low abundance expression were removed from the pro le.

Differential expression analysis
Differential expression analysis was performed using the "DESeq2" R package (version 1.26.0; https://bioconductor.org/packages/release/bioc/html/DESeq2.html) with the standard comparison mode between the 2 experimental conditions. To lter for differentially expressed genes (DEGs) between TP53 mut and TP53 wt patients, log2|fold change| >1 and adj. P < 0.05 were set as the cut-point.

Construction of the IRM index
Immune-related differentially expressed genes (IRDEGs) were collected from the ImmPort gene list (http://www.immport.org), which is a gene set sponsored by the National Institute of Allergy and Infectious Diseases (NIAID) that includes annotated genes related to immune activity processes. After identifying IRDEGs, a least absolute shrinkage and selector operation (LASSO) algorithm was built using the "glmnet" R package (version 3.0; https://cran.r-project.org/web/packages/glmnet/index.html) to select candidate genes. Kaplan-Meier survival curves were created using the "survival" R package (version 3.5; https://cran.r-project.org/web/packages/survival/index.html). Finally, the "timeROC" R package (version 0.4; https://cran.r-project.org/web/packages/timeROC/index.html) was used to conduct a timedependent receiver operating characteristic (ROC) curve analysis.

Microarray data
The gene expression matrixes from GSE29013, GSE30219, and GSE31908 based on the GPL570 platform, including 131 patients with LUAD, were downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/gds). The gene expression data for the 3 matrixes were subjected to log 2 transformation. The scale method of the "limma" R package (version 3.42.2; https://bioconductor.org/packages/release/bioc/html/limma.html) was used to normalize the data.

Patients in the Nanjing cohort and sample collection
To further evaluate the clinical effectiveness of the IRM index, we enrolled a cohort including 92 patients who underwent radical surgery without neoadjuvant chemotherapy and who were diagnosed with LUAD at the Jiangsu Cancer Hospital (Nanjing, China). The patients' characteristics are presented in Table S1.
The patients in the Nanjing cohort were represented by formalin-xed, para n-embedded (FFPE) specimens collected from radical surgery from 2012 to 2018.

RNA extraction and quantitative reverse-transcription polymerase chain reaction
Total RNA was extracted from 4-µm-thick FFPE specimens using the RNeasy FFPE Kit (Qiagen, Hilden, Germany). The complementary DNA (cDNA) synthesis was performed using PrimeScript RT Master Mix (RR036A) (Takara, Dalian, China). The quantitative reverse-transcription polymerase chain reaction (qRT-PCR) assays were performed using the ViiA 7 Dx RT-PCR System (Applied Biosystems, Foster City, USA) with PowerUp SYBR Green Master Mix (Applied Biosystems, Vilnius, Lithuania). The expression of target genes was normalized with the housekeeping gene GAPDH using the 2-ΔCT method. The primer sequences are provided in Table S2.

Function enrichment analysis
Metascape (https://metascape.org/) 13 , which is an online tool, was used to annotate the functional and pathway enrichment analyses to determine the potential molecular mechanisms and biological processes of the candidate genes. Metascape was used to perform Gene Ontology (GO) and The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses for DEGs and IRDEGs. Select the most enriched set of genes in the cluster as one of the representative clusters.

Gene set enrichment analysis and principal components analysis
To determine the potential immune-related pathways and genes between TP53 mut and TP53 wt patients with LUAD in the TCGA cohort, gene set enrichment analysis (GSEA) (version 4.0.0; https://www.gseamsigdb.org/gsea/index.jsp) was performed. An annotated gene set le (msigdb.v7.0.entrez.gmt) was selected for use as the reference gene set. Principal components analysis (PCA) was carried out using the "pca3d" R package (version 0.10.1; https://cran.r-project.org/web/packages/pca3d/) to investigate the relative in ltration patterns of high-and low-index patients.

CIBERSORTx analysis
The CIBERSORTx website (https://cibersort.stanford.edu/index.php) provides an algorithm to quantify the relative in ltration of several immune-in ltrating cells in the TME 12 , including T cell subtypes, naive and memory B cells, myeloid cell subsets, natural killer (NK) cells, and plasma cells. The data of immunein ltrating cell in ltration levels in patients with LUAD were extracted from CIBERSORTx to investigate the correlation with the IRM index.

Nomogram construction and validation
Univariate and multivariate Cox analyses were used to evaluate the IRM index as an independent prognostic factor. Then, a concise nomogram to predict the OS of LUAD was established using the "rms" R package (version 2.10; https://cran.r-project.org/web/packages/rsm/index.html), including 4 factors. We conducted 5-and 7-year OS alignments to determine the prognostic ability of the nomogram model.

Statistical analysis
Statistical analyses were performed using R (version 3.6.1). The Student's t-test was used to determine differences for 2-group comparisons.

Immune landscape based on TP53 mutations in LUAD
In LUAD, TP53 mutation is the most widespread somatic nonsynonymous mutation (Fig. 1A). To comprehensively evaluate the correlation between immune status and TP53 mutation in patients with LUAD, a owchart was constructed to reveal the analyzing process (Fig. 1B). In the TCGA cohort, LUAD samples were divided into 2 groups: TP53 WT (260 patients) and TP53 MUT (239 patients). Subsequently, we performed DEG analysis based on TP53 mutations, which revealed that 1829 genes were statistically signi cantly differentially expressed in the TP53 MUT group compared with the TP53 WT group, including 1230 downregulated genes and 599 upregulated genes (Fig. S1A, Table S3).
Metascape was used to annotate potential functional characteristics, and it showed that the DEGs mentioned above were mainly involved in pathways related to immunology (Fig. S1B). Next, to further prove the correlation between DEGs and tumor immunology, GSEA was used to show that TP53 wt patients with LUAD were enriched in 324 biological processes, including 2 immune-related functional pathways: GO_REGULATION_OF_HUMORAL_IMMUNE_RESPONSE and GO_INNATE_IMMUNE_ RESPONSE_ACTIVATING_CELL_SURFACE_RECEPTOR_SIGNALING_PATHWAY (Fig. 1C), which further con rmed the potential correlation between immunology and TP53 status.

IRM index construction and evaluation of its prognostic ability
To determine the relationship between TP53 mutation status and immune-related functional pathways, 75 IRDEGs were selected for further analysis from the 1829 DEGs based on the Immport database (Table   S4). PCA indicated that these 75 IRDEGs could distinguish between the TP53 WT and TP53 MUT groups ( Fig. 2A). Next, we attempted to evaluate the prognostic performance of the 75 IRDEGs. Six genes with non-zero regression coe cients, including CRHR2, BPIFB2, INHA, SSTR5, SCGB3A1, and BPIFB1, were found to have a maximum prognostic value according to LASSO and Cox regression analysis (Fig. 2B,   2C). Ultimately, a 6-gene IRM was built, and the risk index of each patient was calculated using the following formula for further analysis: IRM index = (-0.121586693 × CRHR2 expression) + (-0.048377824 × BPIFB2 expression) + (0.045865315 × INHA expression) + (-0.018472910 × SSTR5 expression) + (-0.011651852 × SCGB3A1 expression) + (-0.002185301 × BPIFB1 expression) (Fig. 2D). The optimal cutpoint was − 0.9595243. We annotated the immune-related pathways enriched by the 6 IRM-related genes, which included transforming growth factor (TGF) family members, cytokines, cytokine receptors, and antimicrobials (Fig. 2E). Cancer cells and several immune-suppressive cells in the TME recruit multiple cytokines, such as TGF-β, which facilitates tumor progression and mediates T cell dysfunction 14 . This suggests that our IRM may be related to the immunosuppressive state of LUAD, which we will study next.
The cut-point in the TCGA cohort was used to assign patients with LUAD to the high-or low-index groups across all 3 cohorts. The Kaplan-Meier analysis illustrated that a high index was related to worse prognosis (Fig. 3A). The IRM index distribution, OS, and 6-gene expression heatmap are shown in Fig. 3B. 3.3 Validation of the predictive ability of the IRM in the meta-GEO and Nanjing cohorts To identify the robustness of the IRM index, its capability was evaluated in an independent validation cohort (the meta-GEO cohort), which included 131 patients with LUAD. Using the same formula and cutpoint as in the TCGA cohort, the patients in the meta-GEO cohort were divided into high-and low-index groups. The results revealed that high-index patients had outstandingly worse OS compared to patients with a low index, which was consistent with the results of the TCGA cohort (Fig. 3D). The IRM index distribution, OS, and 6-gene expression heatmap are shown in Fig. 3E. The time-dependent ROC curve analysis of the IRM index in the meta-GEO cohort revealed the robustness of the prognostic capability of the IRM index for OS (3 years, AUC = 0.643; 5 years, AUC = 0.635; 7 years, AUC = 0.620; Fig. 3F). To better validate the clinical value of the IRM index, we assessed its prognostic capability in the Nanjing cohort, which included 92 patients who were diagnosed with LUAD at the Jiangsu Cancer Hospital and underwent radical surgery without neoadjuvant chemotherapy. The patients in the Nanjing cohort were divided into high-and low-index groups. The Kaplan-Meier analysis revealed consistent results in the Nanjing cohort (Fig. 3G). The IRM index distribution, OS, and 6-gene expression heatmap are shown in Fig. 3H. The time-dependent ROC curve analysis of the Nanjing cohort revealed the robustness of the prognostic capability of the IRM index for OS (3 years, AUC = 0.705; 5 years, AUC = 0.608; 7 years, AUC = 0.609; Fig. 3I).

Validation of the IRM in different clinical subgroups
TP53 status has a remarkable relationship with the clinical outcome of patients with LUAD. Strati cation analyses were used to test the prediction ability of the IRM index in the TP53 mut and TP53 wt subgroups.
As shown in Fig. 4A and 4B, both in the TP53 mut and TP53 wt subgroups, patients with a high index had worse OS than patients with a low index. Apart from TP53 status, factors such as age, sex, and tumornode-metastasis (TNM) stage may also affect LUAD prognosis. We illustrated that the IRM index was a robust biomarker for estimating OS in younger (≤ 65 years) or older (> 65 years), male or female, and early tumor stage (TNM stage I) or advanced tumor stage (TNM stage II, III, and IV) patients (Fig. 4C-4H).

Immune status between the low-and high-index patients with LUAD
Characterization of the immune in ltration landscape is important to further explain the potential correlation between the IRM index and tumor humoral immunology by the status of the immune microenvironment. First, using Metascape, we annotated the functions of the 75 IRDEGs, which overlapped with the DEGs and Immport database. These IRDEGs were mainly enriched in terms of antimicrobial humoral response, hormone level regulation, hormone secretion, and neuroactive ligand binding (Fig. S2A-S2C). This data thus provided solid evidence of the potential relationship between the IRM index and tumor humoral immunology. Meanwhile, the results prompted us to further explore the correlation between the IRM and immune-related biological processes. Using CIBERSORTx, we identi ed the relationships between the IRM index and 22 in ltrating immune cells, which were acquired in the TCGA cohort (Fig. 5A). High-and low-index patients were segmented into 2 distinct clusters using PCA based on the relative in ltration of the above-mentioned cell subpopulations (Fig. 5B). Additionally, the relative in ltration of the 22 immune-in ltrating cells showed mild to moderate association (Fig. S3). The high-risk patients with LUAD had remarkably lower relative in ltration proportions of memory B cells and regulatory T cells (Tregs; Fig. 5C, 5D) and signi cantly higher relative in ltration proportions of neutrophils and resting memory CD4 + T cells (Fig. 5E, 5F) than low-risk patients with LUAD. Thus, the above results suggested that the heterogeneity of tumor immune-related cell in ltration in LUAD may be a potential prognostic biomarker and target for response after immunotherapy and may have remarkable clinical signi cance.
Furthermore, immune checkpoint inhibitors play an antitumor role, reversing the immunosuppressive effect of the tumor. We investigated the correlation between the IRM index and the expression of crucial immune checkpoints, including PDCD1 (encoding PD-1), LAG3, TIGIT, TIM3, CD274 (encoding PD-L1), and CTLA4. The correlation coe cient is shown in Table S5. We observed that the IRM index had a remarkably positive relationship to the expression of TIM3, TIGIT, PDCD1, and CD274 (P < 0.05; Fig. 6A). In addition, the expression of PDCD1 and CD274 in the high-index group was remarkably higher than in the low-index group (Fig. 6B, 6C). To validate the expression differences in PDCD1 and CD274 between the high-and low-index patients, we detected the PDCD1 and CD274 expression in the Nanjing cohort, and the high-index patients expressed higher levels of PDCD1 and CD274 (Fig. 6D, 6E). These results are consistent with a previous study showing that PD-L1 expression is related to advanced pathological features and worse OS in patients with NSCLC 15 . We also found the expression differences of PD-1 and PD-L1, which are encoded by PDCD1 and CD274, respectively, using IHC samples from 12 patients in the same cohort. These results revealed a signi cantly positive correlation between the IRM index and IHC scores of these 2 immune checkpoint proteins in patients with LUAD ( Fig. 6F-6H).

Development and veri cation of a nomogram based on the IRM index
To assess the prognostic value of the IRM index, a Cox regression analysis was conducted in the TCGA cohort. We rst selected the IRM index and several clinical factors for univariate Cox analysis. As shown in Fig. 7A, factors including lymphatic invasion, distant metastasis, IRM index, age, and TNM stage were signi cantly related to OS in the TCGA cohort (P < 0.05). After multivariate Cox regression analysis of these factors, lymphatic invasion, IRM index, age, and TNM stage were still signi cantly associated with survival (P < 0.05). Based on these results, we further integrated the IRM and multiple independent clinical factors (lymphatic invasion, age, and TNM stage) to develop a nomogram, which gives clinicians a quantitative way to predict the clinical outcome of patients with LUAD (Fig. 7B). The C-index for the nomogram in the TCGA cohort was 0.746 (95% CI: 0.5886-0.8730). Calibration plots comparing the predicted and actual outcomes of 5-and 7-year OS indicated good agreement (Fig. 7C, 7D). Additionally, we used a time-dependent ROC curve analysis to compare the robustness of the prognostic capability between the nomogram, IRM index, age, and TNM stage in 5-and 7-year OS (Fig. 7E, 7F). The nomogram revealed higher prognostic capability at 5 and 7 years with a larger AUC. Furthermore, we veri ed the prognostic accuracy in the Nanjing cohort. As we expected, the nomogram obtained consistent results (Fig. 7G, 7H).

Discussion
In this study, we systematically analyzed the association between TP53 status and immune-related phenotypes in patients with LUAD. A TP53-related IRM index was constructed. It was derived from the TCGA cohort and validated in the meta-GEO and Nanjing cohorts. It was found to be remarkably associated with prognosis. The prognostic value of the 6-gene IRM index was independent of multiple known strong prognostic factors. Furthermore, the IRM index allowed us to divide patients with LUAD into 2 subgroups with different immune-related phenotypes. Consequently, we integrated the IRM with multiple clinical factors into a nomogram with robust OS prediction.
Through multi-functional enrichment analysis of DEGs, we observed obvious enrichment in the humoral immune response. The consistent pathway enrichment analysis results were further veri ed in selected IRDEGs. Furthermore, 6 IRM-related genes were found to participate in multiple immunosuppressionrelated pathways. Afterward, we divided the patients with LUAD into 2 subgroups with different prognosis and immune-related phenotypes. In all 3 cohorts, high-index patients showed worse survival than lowindex patients. Meanwhile, the IRM index was positively correlated with the expression of several important immune checkpoints. In previous research, NSCLC progression was positively associated with the increased expression of T cell exhaustion markers, such as PD-1, TIM3, and CTLA4 16 , which is consistent with the results of our research.
CRHR2 belongs to the G-protein coupled receptor superfamily, regulating corticotropin-releasing hormone to perform biological functions 17 , which widely expressed in the gastrointestinal tract, lung and skeletal muscle 18 . CRHR2 stimulates intracellular cAMP pathway, including activation of nuclear factor-kB and expression of TNF-β in T cells 19,20 . BPIFB1 and BPIFB2 belong to the bactericidal/permeabilityincreasing-fold-containing family 21 . BPIFB1 and BPIFB2 protein are most highly expressed in the trachea and lung 22 and bind to the Gram-negative bacteria and exert antibacterial function 23,24 . Moreover, it has been reported that BPIFB1 is abnormally expressed in tumors, which suggests that it plays a role in tumor development 25 . INHA encodes a member of the TGF-β superfamily of proteins (RefSeq, Aug 2016), which perform the functions of activated cytokine and hormone 26,27 . SSTR5 is a predominant component of somatostatin receptor subtypes, which regulates inhibitory effects on endocrine and exocrine secretions 28 . SSTR5 has a signi cant regulating-in ammatory effect via regulating somatostatin, which demonstrated in different animal models 29,30 . SCGB3A1 belong to secretoglobin gene superfamily, which are cytokine-like small molecular weight secreted proteins and predominantly expressed in lung airway epithelial cells 31 . Secretoglobins are thought to be involved in immunomodulatory 32 . Moreover, SCGB3A1 has been reported as a tumor suppressor in various human tumors including breast, prostate, lung, and pancreatic carcinomas 33,34 .
In chronic diseases, T cells malfunction due to T cell exhaustion, which increases the expression of inhibitory receptors incorporating PD-1, LAG3, TIM3, CTLA4, and TIGIT [35][36][37][38][39] , resulting in fewer cytokines and loss of antitumor capabilities. The limited e cacy of immunotherapy may be due to the production of dysfunctional T cells in the TME 40 . Therefore, regulators that reverse the state of T cell dysfunction are the focus of current research. For example, tertiary lymphoid structure immune activity dysfunction is reversed and antitumor capabilities are enhanced after treatment with anti-PD-1/PD-L1 and anti-CTLA-4 immune checkpoint inhibitors in humans and mice [41][42][43][44] . The PD-L1/PD-1 axis is an important regulatory pathway of T cell exhaustion in tumors. DD1α expression induced by p53, which is encoded by TP53, has been shown to upregulate PD-1 and PD-L1, as cancer cells respond to genotoxic stress and DNA damage, which then promotes the gradual priming of immune surveillance 45 . This nding clari ed the relationship between p53 and immune checkpoint inhibitors, which may indicate the intrinsic molecular mechanism of the relationship between the IRM index and the expression of immune checkpoints. PD-L1 has abundant expression in cancer cells and the tumor extracellular matrix, and blocking the PD-L1/PD-1 axis can enhance the antitumor capabilities of T cells 46 . In our study, the IRM index was positively correlated with the expression of immune checkpoints, which shows that as the IRM index grows, T cells are exhausted, and their antitumor abilities decrease. This explains the poor prognosis of high-index patients with LUAD. It can be speculated that immunotherapies that block the pathways that suppress tumor immune responses for patients with LUAD in the high-index group may increase the presentation of cancer-associated antigens, resulting in the recovery of the immune response of CD8 + T cells 47 , which may then result in better immunotherapy effects.
In the immune in ltration analysis between the high-and low-index patients, the high-index patients with LUAD had remarkably higher proportions of neutrophils and resting memory CD4 + T cells and lower proportions of memory B cells and Tregs. Memory resting CD4 + T cells can be further differentiated into multiple cell subpopulations and confer different functions, including blocking CD8 + T cell activation and NK cell killing and suppressing the immune response to autoantigens and exogenous antigens 48 . In human NSCLC, neutrophils play a key role in tumor immunity 49 . In mice, neutrophils in ltrating tumors can either promote carcinogenesis by supporting tumor-related in ammation, angiogenesis, and metastasis and inhibiting T cell activation or restrict tumor growth through the expression of antitumor and cytotoxic mediators 50 . Nevertheless, because the tumor is in an irreversible continuous chronic in ammatory state, suppressive neutrophils are constantly mobilized and become the dominant subpopulation of neutrophils 51 , which may suppress the immune response and promote malignant progression. In our study, the high-index patients with LUAD had signi cantly higher proportions of neutrophils and resting memory CD4 + T cells, restricting the function of CD8 + T cells and NK cells in the tumor development process, resulting in malignant tumor progression and worse OS. It is reasonable to speculate that the high-index patients with LUAD may have improved CD8 + T cell function and receive better e cacy by using an immune checkpoint inhibitor that blocks the PD-1/PD-L1 axis.
Tregs are an inhibitory subpopulation of CD4 + T cells. Cancer-associated broblasts in the extracellular matrix express COX-2, which promotes PGE2 secretion to induce immunosuppressive FOXP3 + Tregs 52 , which then accumulate in primary tumor tissues and the peripheral blood to promote immune evasion 46 . Progressing tumors can inhibit CD8 + T cells through several approaches, including Tregs, which can directly suppress the antitumor functions of CD8 + T cells 53  Our research supplies a novel angle regarding LUAD immune microenvironment groupings and immunotherapy responses. However, since our study is a retrospective study, it has limitations and needs to be further validated by prospective studies. Furthermore, mechanistic studies of the IRM index-related genes and immune-in ltrating cells need to be implemented to explain their clinical application. In subsequent research, we will focus on single-cell transcriptome studies of immune-in ltrating cells in patients with LUAD with different IRM indexes.
In conclusion, the IRM index is a robust clinical biomarker that can assign patients with LUAD to subgroups with signi cantly different prognoses and immune-related phenotypes, which may explain the molecular mechanism of different prognoses from the perspective of immunology. The data sets supporting the conclusions of this article are available from the "TCGAbiolinks" R package (version 2.14.1; https://bioconductor.org/packages/release/bioc/html/TCGAbiolinks.html) and GEO database (https://www.ncbi.nlm.nih.gov/gds).

Ethics approval and consent to participate
This study was approved by the Regional Ethics Committee at Nanjing Medical University. The experiments were undertaken with the understanding and written consent of each patient. The methodologies conformed to the standards set by the Declaration of Helsinki.

Consent for publication:
The consent was obtained from patients.

Con ict of interest
The authors declare that the research was conducted in the absence of any commercial or nancial relationships that could role as a potential con ict of interest.