Genetic and microenvironmental differences in non-smoking lung adenocarcinoma patients compared with smoking patients.

BACKGROUND: Non-smoking-related lung adenocarcinoma (LUAD) has its own characteristics. Genetic and microenvironmental differences in smoking and non-smoking LUAD patients were analyzed to elucidate the oncogenesis of non-smoking-related LUAD, which will improve our understanding of the underlying molecular mechanism and be of clinical use in the future.
METHODS: The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO) databases were used for clinical and genomic information. Various bioinformatics tools were used to analyze differences in somatic mutations, RNA and microRNA (miRNA) expression, immune infiltration, and stemness indices. GO, KEGG, and GSVA analyses were performed with R. A merged protein-protein interaction (PPI) network was constructed and analyzed. A miRNA-differentially expressed gene network was constructed with miRNet. qRT-PCR was used for validation of 4 most significantly differently expressed genes and 2 miRNAs in tumor samples obtained from 20 pairs of non-smoking and smoking patients.
RESULTS: Five hundred and one patients with LUAD were obtained, including 210 in the non-smoking group and 292 in the smoking group. A total of 174 significantly altered somatic mutations were detected, including mutations in tumor protein p53 and epidermal growth factor receptor, which were downregulated in non-smoking-related LUAD. At the RNA level, 231 significantly differentially expressed genes were obtained; 124 were upregulated and 107 downregulated in the non-smoking group. GSVA analysis revealed 42 significant pathways. Other functional and enrichment analyses of somatic mutations and RNA expression levels revealed that these genes were significantly enriched in receptor activity regulation and receptor binding. Differences in microenvironments including immune infiltration (e.g., CD8+ T cells and resting mast cells) and stemness indices were also found between groups. A 79-pair interaction was found between differentially expressed genes and miRNAs, of which miR-335-5p and miR-34a-5p were located in the center. Twenty-one genes, including vitronectin, neurotensin, and neuronatin, were differentially expressed in both non-smoking LUAD patients and DMSO-treated A549 cells. And the different expression of neurotensin, neuronatin, trefoil factor family2, regenerating family member 4, miR-377-5p, miR-34a were verified with the same tendency in our own samples.
CONCLUSIONS: Non-smoking LUAD patients, compared to smokers, have different characteristics in terms of somatic mutation, gene, and miRNA expression and the microenvironment, indicating a diverse mechanism of oncogenesis. 2020 Translational Lung Cancer Research. All rights reserved.

Introduction

Lung cancer is the most common cancer worldwide. Non-small cell lung cancer (NSCLC) (1), the main subtypes of which include adenocarcinoma and squamous cell carcinoma, accounts for approximately 85% of the cases of lung cancer (2). Lung cancer is the most common malignant tumor and has the highest morbidity and mortality (1). According to the World Health Organization Report on the Global Tobacco Epidemic (3), there are 1.4 billion tobacco users aged 15 years and older worldwide. It is well accepted that tobacco smoking is the primary cause of lung cancer, especially lung squamous cell carcinoma (3-5) since Richard Doll and Austin Bradford Hill first proved the link between smoking and lung cancer in the 1950s (6).

However, approximately 25% of lung cancer cases worldwide, mainly adenocarcinoma, cannot be attributed to tobacco smoking; lung cancer in never smokers is the seventh leading cause of cancer deaths worldwide (7). According to clinical experience, different epidemiology and natural history are observed between lung cancers in never smokers and those in smokers (8), suggesting that lung cancer in never smokers is a ‘different’ disease, with specific etiology and molecular differences (9). With the advent of RNA sequencing technology, more studies have focused on specific genetic or molecular mechanisms (10); however, these have remained unclear until now in non-smoker-related lung adenocarcinoma (LUAD).

To study the differences in non-smoking compared with smoking LUAD patients, clinical data, gene expression, somatic mutations, immune infiltration, and stemness indices were analyzed to determine the role of smoking in LUAD. The effects were also confirmed in LUAD cells. This study will improve our understanding of the causes of oncogenesis of non-smoking LUAD and provide a reference for therapeutic decisions in clinical treatments.

We present the following article in accordance with the MDAR reporting checklist (available at http://dx.doi.org/10.21037/tlcr-20-276).

Methods

The Cancer Genome Atlas (TCGA) data collection

Detailed information for patients with LUAD was downloaded from TCGA. Clinical data, RNA expression data, microRNA (miRNA) expression and somatic mutation (VarScan MAF files) data were downloaded from NCI’s Genomic Data Commons (https://gdc.cancer.gov/) (TCGA-LUAD). After matching mRNA expression, miRNA expression, and somatic mutation data with clinical data, 502 cases of RNAs, 376 cases of miRNAs, and 501 cases of somatic mutations were obtained to explore the differences in the genomes of smoking and non-smoking LUAD patients at different levels (see for details). According to the smoking data described in TCGA, patients who had never smoked or who had quit smoking for more than 15 years were classified as non-smokers while those who were current smokers, who quit smoking less than 15 years ago, or quit smoking but the years were unknown were classified as smokers ().

Figure 1

Flow diagram of whole design.

GEO data collection

The mRNA expression data of LUAD from current, former, and never smokers were download from Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). GSE31210, GSE50081, and GSE68465 datasets (11-13), published on February 20, 2008, were collected to determine the coding RNA expression signature of cigarette smoking.

The GSE69770 dataset (14), published on February 7, 2018, performed RNA sequencing of dimethyl sulfoxide (DMSO)- or cigarette smoke condensate (CSC)-treated A549 human LUAD cells. Cells were treated for 48 h or 2 weeks.

Immune infiltration data

Immune infiltration data was collected from a previous study (15) and matched with the clinical data downloaded from the TCGA database. A total of 503 samples were finally enrolled; 210 were classified as non-smokers and 293 were classified as smokers.

Stemness indices

Stemness indices were collected from a previous study (16) and matched with the clinical data downloaded from the TCGA database. A total of 496 samples were finally enrolled; 207 were classified as non-smokers and 289 were classified as smokers.

Tissue samples

All samples were obtained from patients at the Department of Thoracic Surgery, Zhongshan Hospital, Fudan University. Tumor samples were derived from surgical resection of LUAD between January 2019 and April 2019. All samples were quickly frozen in liquid nitrogen after removal and then stored at −80 °C before used. Finally, 20 pairs of non-smoking LUAD tissue and smoking adenocarcinoma (smoking index ≥20 pack-years) tissues from 40 patients were obtained for qRT-PCR validation.

RNA, miRNA preparation and qRT-PCR analysis

To detect the expression of NTS, NNAT, TFF2, REG4, miR-512, miR-372 in 20 pairs of samples, RT-qPCR was carried out on an ABI Prism 7500 real-time PCR system (Applied Biosystems) with proper PCR parameters.

Total RNAs were extracted by TRIzol (Beyotime, China). The first-strand cDNA was synthesized using PrimeScriptTM RT reagent Kit with gDNA Eraser (Real Time Perfect) (TaKaRa, Tokyo, Japan) according to the manufacturer’s instructions. Then SYBR Premix Ex TaqTM II (Tli RNaseH Plus) (TaKaRa) was used with the following PCR parameters, 1 cycle of 30 s at 95 °C, 40 cycles of 5 s at 95 °C and 34 s at 60 °C. β-actin was used as the reference. Primers used in this study are listed in .

Table S1

The sequences and melting temperature (Tm) of the primers used in our research, whether they span exon junctions, PCR efficiency and correlation with dilution series (R2)

Gene	Primers sequence 5'-3' (forward, reverse)	Tm (°C)	Exon junction span	PCR efficiency (%)	Correlation with dilution series (R2)
NTS	TGCTTTAGATGGCTTTAGCTTGG	59.9	Yes	101.8	0.998
	TTCCTGGATTAACTCCCAGTGT	59.8
NNAT	ACTGGGTAGGATTCGCTTTTCG	59.9	Yes	99.8	0.999
	ACACCTCACTTCTCGCAATGG	59.2
TFF2	GCTGTTTCGACTCCAGTGTCA	64.0	Yes	101.5	0.999
	CCACAGTTTCTTCGGTCTGAG	60.2
REG4	CTGCTCCTATTGCTGAGCTG	58.9	Yes	101.2	0.996
	GGACTTGTGGTAAAACCATCCAG	58.6

Total miRNAs were extracted with miRcute miRNA Isolation Kit (TianGen Biotech, Beijing, China) and then reversed by miRcute Plus miRNA First-Strand cDNA Kit (TianGen). miRcute Plus miRNA qPCR Kit (SYBR Green) (TianGen) was used in the PCR system with the following PCR parameters, 1 cycle of 2 min at 94 °C, 40 cycles of 20 s at 94 °C and 34 s at 60 °C. U6 was used as the reference. All miRNA primers were obtained from TianGen.

All the samples were repeated three times.

Statistical analysis

The distribution of patients’ characteristics (e.g., sex, race, age group, primary site, differentiation grade, and chemotherapy) was summarized using counts and percentages. Statistical analysis was performed with SPSS 23.0 software (IBM). Differential analysis of RNA, miRNA, and immune infiltration data was performed with R (version 3.6.1). All data were normalized and standardized by constructing relevant expression matrices using edgfR. Differential genes [P<0.05, false discovery rate (FDR) <0.05 for somatic mutations, miRNAs, and others] were sorted according to logFoldChange values (|logFC| >1) to identify significantly different expression between smoking and non-smoking patients. A merged protein-protein interaction (PPI) network was constructed and analyzed with Cytoscape software (version 3.7.2). Differences in stemness indices and verification were determined by Student’s t-test with Graphpad Prism 8. The test level was α=0.05, and differences were considered statistically significant at P<0.05. Survival analysis was performed with the Kaplan-Meier method with R (version 3.6.1).

All procedures performed in this study were in accordance with the Declaration of Helsinki (as revised in 2013) and was approved by the Ethics Committee of Zhongshan Hospital, Fudan University, China (B2019 – 377R). Because of the retrospective nature of the re-search, the requirement for informed consent was waived.

Results

shows the basic clinical characteristics of the patients (e.g., sex, age group, location, and stage) summarized using counts and percentages. The smokers were younger than the non-smokers. Most of the cases of LUAD in non-smokers were acinar, papillary predominant, and invasive mucinous pathology subtype. Smokers had lower 3-year survival rates [smokers (S), 72.11% vs. non-smokers (NS), 77.56%], although this difference was not statistically significant ().

Table 1

Clinical features of the TCGA samples

Characteristics	Nonsmokers (n=210)			Smokers (n=292)		P value
	n	%		n	%
Age						<0.001
≤60	36	17.14		118	40.41
61–70	69	32.86		102	34.93
≥71	92	43.81		66	22.60
NA	13	6.19		6	2.05
Sex						0.469
Female	117	55.71		153	52.40
Male	93	44.29		139	47.60
Anatomic location						0.587
Left	84	40.00		113	38.70
Right	122	58.10		169	57.88
NA	4	1.90		10	3.42
Stage						0.366
Stage I	119	56.67		153	52.40
Stage II	46	21.90		71	24.32
Stage III	32	15.24		48	16.44
Stage IV	10	4.76		16	5.48
NA	3	1.43		4	1.37
Pathology subtype1						0.143
Acinar predominant ADC	37	17.62		32	11.00
Colloid ADC	0	0.00		1	0.34
Lepidic predominant ADC	6	2.86		5	1.72
Micropapillary predom ADC	10	4.76		14	4.81
Papillary predominant ADC	14	6.67		9	3.09
Solid predominant ADC	20	9.52		37	12.71
Invasive mucinous	7	3.33		3	1.03

1, only those with specific identification of pathological subtypes were included in this column. TCGA, The Cancer Genome Atlas; ADC, adenocarcinoma.

Figure 2

Survival analysis: survival time analysis of 509 patients with smoking status.

Differences in somatic mutations

The distribution patterns of somatic mutations (the overall pattern is shown in ) were investigated between the non-smoker and smoker groups. After performing enrichment analysis, we identified 174 enriched mutations with FDR <0.05 between the non-smoker and smoker groups (). The somatic mutation rates of tumor protein p53 (TP53; S, 56.31% vs. NS, 35.10%); titin (TTN; S, 57.34% vs. NS, 29.33%); KRAS proto-oncogene (KRAS; S, 31.40% vs. NS, 22.60%), and ryanodine receptor 2 (RYR2; S, 45.39% vs. NS, 22.12%), were higher in smokers, while epidermal growth factor receptor (EGFR; S, 6.83% vs. NS, 19.71%) were higher in smokers. The mutation rate of human epidermal growth factor receptor-2 (HER2; S, 17.41% vs. NS, 8.17%) and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA; S, 8.60% vs. NS, 0) showed no difference between smokers and nonsmokers.

Figure S1

The summary of the LUAD patients’ somatic mutation data, (A) displayed number of variants in each sample as a stacked bar plot and variant types as a boxplot summarized; (B) classified SNPs into transitions and transversions, (a) showed the overall distribution of the six different transformations, (b) classified the SNPs as transitions (Ti) and transversions (Tv), showing their proportion, (c) stacked bar graph of the percent conversion in each sample.

Figure 3

Somatic mutation waterfall map grouped by smoking status, the red band below corresponded to the smoking group, and the blue was the non-smoking group.

Using the somatic interactions function, which performed a pair-wise Fisher’s exact test to detect significant pairs of genes, we detected the relationships among the top 50 genes with different somatic mutations (). Almost all of the mutated genes were tightly linked to each other, except EGFR, which was strongly excluded. Thus, these mutations might be linked to the development of LUAD.

Figure S2

Exclusive/co-occurrence event analysis on top 20 differently mutated genes.

Differentially expressed coding genes and functional annotation in TCGA-LUAD

To determine whether smoking can change biological characteristics, expression profiles were examined for differentially expressed genes (DEGs). Of the 60,483 ensemble numbers, 13,747 were successfully converted to gene names by org.Hs.eg.db, and the edgeR algorithm identified 231 DEGs [210 mRNAs and 21 long non-coding RNAS (lncRNAs)] (P<0.05) with an obvious fold change (|logFC| >1), of which 124 were upregulated and 107 were downregulated (). A PPI network () was also constructed to reveal the association between proteins encoded by DEGs.

Figure 4

Differentially expressed genes in non-Smoking lung adenocarcinoma in TCGA-LUAD dataset. Volcanic map reviewed genes differentially expressed between the smoker and non-smoker groups. Blue dots represented significantly down-regulated genes, red dots represented significantly up-regulated genes, and black dots represented genes that are not differentially expressed. TCGA, The Cancer Genome Atlas; LUAD, lung adenocarcinoma.

Figure S3

Protein-protein interaction (PPI) network of differently expressed genes between smokers and non-smokers in TCGA-LUAD. The size and gradient color of the Node were adjusted by degree, and the thickness and gradient color of edge are adjusted by combined score.

We also applied the same differential analysis method for the data from the GEO databases (GSE31210, GSE50081, and GSE68465) to identify DEGs between smokers and non-smokers (). A total of 81 RNAs were differentially expressed in both the TCGA-LUAD and GEO datasets.

Figure S4

Differentially expressed genes differentially expressed genes in non-smoking lung adenocarcinoma associated with smoking in GEO datasets. GEO, Gene Expression Omnibus.

The expression of genes in LUAD cells has been linked to smoking-related transversion mutations in lung tumors. To identify tobacco smoke-inducible enhancers at the cellular level, we used the same method to analyze the expression profile of A549 human LUAD cells treated with DMSO or CSC for 2 weeks (GSE69770; ). A total of 21 genes (upregulated: MALAT1, RSPO3, CPLX2, BAAT, CLDN2, TM4SF4, NTS, and VTN; downregulated: IGF2, HIST1H3F, HIST1H4C, HIST1H4D, HIST1H4E, HIST1H3B, HIST1H2AI, HIST1H1E, HIST1H2AM, HIST1H2BE, HIST1H3C, HIST1H1D, and HIST1H1B) were both differentially expressed between smoking and non-smoking LUAD patients and between CSC- and DMSO-treated A549 cells, which more reliably revealed that changes in these genes were caused by smoking and their upregulation or downregulation might play a more important role.

Figure S5

Differentially expressed genes between A549 cells treated with DMSO and cigarette smoke condensate (CSC) in GSE69770 datasets.

Gene functional analysis

Gene set variation analysis (GSVA) of RNA sequencing data converts biological function annotations (GO and KEGG pathway analyses) into new expression matrices and treats them as general matrix files for differential expression analysis. This analysis revealed 42 significant pathways between smokers and non-smokers (); e.g., a Reactome for ethanol oxidation and the Amundson DNA damage response TP53 were downregulated and Ly aging middle DN, and Montero thyroid cancer poor survival up were upregulated in nonsmokers.

Figure 5

Heatmap of gene set variation analysis for and RNA-Seq data (GSVA).

We also performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of the 231 genes that had differential somatic mutations or were differentially expressed based on the TCGA data. Gene functional enrichment analysis showed that six GO functional groups exhibited significant differences between smokers and non-smokers: extracellular matrix structural constituent, receptor ligand activity, structural constituent of cytoskeleton, actin filament binding, hormone activity, and receptor regulator activity (Figure S6A). For KEGG pathway analysis, four pathways differed between smokers and non-smokers: systemic lupus erythematosus, alcoholism, thyroid hormone synthesis, and protein digestion and absorption (Figure S6B).

Differentially expressed miRNAs and functional annotation in TCGA-LUAD

Of the 1,881 miRNAs expressed in the TCGA-LUAD with clinical information, 38 miRNAs were significantly upregulated (P<0.05, |logFC| >1) and 7 were downregulated between smokers and non-smokers using the edgeR algorithm ().

Figure S7

Differentially expressed miRNA in non-smoking lung adenocarcinoma in TCGA-LUAD dataset.

To determine the regulatory relationship between DEGs and miRNAs, we looked for links between the genes and miRNAs using the miRNet platform and obtained 7,945 pairs of miRNA-DEGs,while in consideration of the appearance of the figure, only the closest 85 pairs were shown in (differential genes were defined as P<0.001 and |logFC| >1 and differential miRNAs were defined as P<0.05). miR-335-5p and miR-34a-5p are located at the core of the network, suggesting these miRNAs and their binding genes may play an important role in the differences between smoking and non-smoking lung cancer mechanisms.

Figure 6

Seventy-nine microRNAs-differentially expressed gene (miRNA-DEG) pairs consisting of 19 miRNAs and 36 mRNAs by miRNet.

Validation of mRNA and miRNA expression by qRT-PCR

After comprehensively considering the significance of differences and the FC value, we selected 4 highly significantly differential genes and 2 miRNAs. NTS, NNAT and miR-377-5p, miR-34a-3p are the most up-regulated factors in non-smoking adenocarcinoma and the TFF2 and REG4 are the most down-regulated in TCGA data as shown in and . Therefore, we used qRT-PCR to measure the expression levels of them in 20 pairs of non-smoking LUAD and smoking adenocarcinoma samples. As shown in , the alteration of the factors’ expression in the results of qRT-PCR was generally consistent with that shown in the bioinformatics analyses though the FCs were not as large as those in the bioinformatics results.

Figure 7

Validation of comparatively more significantly different mRNAs and miRNAs expression by qRT-PCR. *, P<0.05; **, P<0.01; ***, P<0.005.

Microenvironmental differences

Tumor immune cell infiltration refers to the migration of immune cells from the blood to the tumor tissue, where they exert their function. Infiltrating immune cells can be isolated from the tumor tissue. Resting mast cells, M2 macrophages, memory resting CD4+ T cells, and resting dendritic cells were upregulated in non-smokers, while CD8+ T cells, plasma cells, activated CD4+ T cells, M1 macrophages, and follicular helper T cells were downregulated (). Furthermore, we looked for differences in immune-related gene expression. Nine genes related to antimicrobials (), six genes related to cytokines (), granzyme B (GZMB) for natural killer (NK) cell cytotoxicity (), and nuclear receptor subfamily 0 group B member 2 (NR0B2) for cytokine receptors () were found to be significantly differentially expressed, indicating that cytotoxic and antimicrobial function may be one reason for the differences in the immune microenvironment between smokers and non-smokers. For programmed death-ligand 1 (PD-L1) (CD274), a significant difference in expression was also observed between smokers and non-smokers (), but with a low fold change value (logFC =0.09).

Figure 8

Differences of microenvironment in non-smoking lung adenocarcinoma. (A) Comparation of each leukocyte fraction between smokers and nonsmokers; (B) differential genes related to antimicrobials; (C) Differential genes related to cytokines; (D) differential genes related to natural killer (NK) cell cytotoxicity; (E) differential gene related to cytokine receptors; (F) differential gene related to PD-L1. PD-L1, programmed death-ligand 1. *, P<0.05; **, P<0.01; ***, P<0.005; ****, P<0.001.

The stemness indices (mRNAsi) were calculated for the TCGA data. mRNAsi is an index calculated based on expression data, which ranges from 0 to 1; a value close to 1 indicates a lower degree of cell differentiation and stronger stem cell characteristics (16). We found a significant difference in stemness indices between the groups (496 samples; 207 non-smokers and 289 smokers, P<0.05, ), which confirmed the original results that non-smokers or long-term reformed smokers exhibit lower stemness than current and recently reformed smokers, as previously reported (16).

Figure 9

Differences of stemness between smokers and non-smokers in LUAD. LUAD, lung adenocarcinoma.

Discussion

The study of cancer predisposition and cancer-specific genes is an important research direction. In this study, we observed quantitative differences between the LUAD genome of never smokers compared with smokers. Many studies before concluded and verified that smoke would change human bronchial and pulmonary alveolar epithelial cells via affecting mitochondrial function and regulating the expression of several genes, especially PP2A, which might be connected with oncogenesis (17-20). However, these studies still focused more on how smoking affects lung acinar epithelial cells and causes chronic lung disease rather than tumors.

We proved that these differences existed at different levels of the genome, including somatic mutations, RNAs, and miRNAs, and some, mainly RNAs, showed significant increases or decreases at the cellular level.

Boeckx et al. (21) found that the frequency of EGFR mutations is higher, and the frequency of TP53 mutations, is reduced in never smokers compared with smokers. However, in our study, these genes only showed differences at the somatic level, and not the RNA level, which is in contrast to previous studies of smoking-related genomic and/or transcriptional alterations. These differences may be due to the selection of different patient populations, tumor characteristics, and cohort sizes (22). In our previous study (23), we found that over half of our patients carried EGFR mutations, followed by KRAS and HER2 mutations, the first two of which also show different rate of mutation in this study, while HER2 rem ALK, RET, and ROS1 translocations, PIK3CA, BRAF, and NRAS mutations were rare and all were identified in less than 2% of patients. PD-1 positive tumors were identified in 14 (1.9%) patients, and PD-L1 positive tumors in 95 (13.0%) patients.

In the present study, NTS and NNAT were upregulated in non-smoking LUAD patients, DMSO-treated A549 LUAD cells and our own samples. NTS, along with its receptor NTSR, has been shown to be important in lung cancer outcomes (24,25) and therapies against NTS can decrease tumor growth and metastasis (26,27). These data suggested that the overexpression of NTS in non-smokers may be a trigger for the development of LUAD. NNAT mRNA is mainly expressed in endocrine and adipose tissues in a hormone- and nutrient-sensitive manner (28-30) and has the potential to be used as a marker differentiating between Large-Cell Neuroendocrine Carcinoma (LCNEC) and small cell lung cancer (SCLC) as it more highly expressed in LCNEC, making it different from smoker NSCLC LUAD (31,32).

TFF2 and REG4 were both downregulated in non-smoking LUAD in TCGA and our own samples. Lung macrophages rely upon TFF2 to promote epithelial proliferation, while in the absence of TFF2, lung epithelia were unable to proliferate and expressed reduced lung mRNA transcript levels (33,34). In non-smoking LUAD patients, down-regulation of TFF2 might indicate weaker lung repair ability, which promotes tumor development. The overexpression of REG4, which was shown in the smoking group previously, was proven to be closely correlated to the carcinogenesis in some types of cancer (35). Also, REG4 plays an important role in KRAS-driven lung cancer pathogenesis and is a novel biomarker of LUAD subtype (36). However, whether its expression can be used to distinguish non-smoking from smoking LUAD needs further experiments to verify, as it showed unstable CT values when verifying.

A group of histones including HIST1H3F, HIST1H4C, HIST1H4D, HIST1H4E, HIST1H3B, HIST1H2AI, HIST1H1E, HIST1H2AM, HIST1H2BE, HIST1H3C, HIST1H1D, and HIST1H1B differentially expressed between smokers and nonsmokers in our study. Histone variants act as transcriptional activators or repressors of cancer-related genes. Several researches about HIST1H2, HIST1H3, HIST1H4 found that they were related to the progression and metastasis of tumor cell (37-39). However, there reports on how smoking regulates the expression of these histones so far.

Overexpression of miR-377-5p, which was more often in the non-smokers in our study, was reported remarkably downregulated in NSCLC tissues, and inhibited cell proliferation and development as a tumor‐suppressor gene (40,41). miR-34a overexpressed among the nonsmokers. Previous study found that p53 regulates PDL1 via miR-34 and SART3 overexpression increased miR-34a levels, which may affect cell cycle progression in NSCLC cells, while miR-34a inhibits NSCLC tumor growth and metastasis through targeting EGFR (42-44). Furthermore, it was also considered to be a target in the therapy of lung cancer (45,46).

We also focused on gene function. The leukocyte fraction varied substantially across the immune subtypes (15). We found that CD8+ T cells, plasma cells, activated CD4+ T cells, M1 macrophages, and follicular helper T cells were comparatively downregulated in non-smokers. Previously Kinoshita et al. found a high FOXP3/CD4 ratio in smokers with adenocarcinoma, a low number of CD20+ B cells in non-smokers was identified as an independent unfavorable prognostic factor in resected NSCLC, and infiltrating CD8+ T cells may not be activated sufficiently in the immunosuppressive microenvironment in non-smokers with adenocarcinoma (47,48). Li et al. (49) concluded that aberrant activation of mast cells and CD4+ memory T cells plays a crucial role in cigarette smoking-induced immune dysfunction in the lung, leading to tumor development and progression. Generally, tobacco smoking patients with NSCLC showed a higher PD-L1 tumor proportion score (50-52) and experience a higher response rate with immunotherapy, mainly pembrolizumab and nivolumab, than non-smokers (50,53-57). In the present study, we also found a significant difference in expression of the PD-L1 gene (CD274) between the two groups but only with a small fold-change value. All of this evidence suggested that smoking created a different immune environment with abnormal immune cells. Smokers appeared to develop LUAD more easily because their immune cells are not as limited as those of non-smokers.

A strong association between mRNAsi and tobacco smoking status in LUAD was shown in this study. This confirmed previous research, which suggested that the stemness of LUAD tumors might be activated in response to environmental stimuli such as smoking, and might influence tumor aggressiveness (16).

There are some limitations to this study. Smoking status data in the TCGA database did not include information regarding whether the patients were frequently exposed to second-hand smoke or kitchen fumes, which are also important causes of lung cancer. Also, our research results were mainly based on the data of TCGA and verified with GEO datasets. We were unable to verify our results with other databases because of a lack of high-quality data on smoking. Lung cancer caused by smoking is a long-term process, so it is difficult to determine the role of these factors. Additional studies will be necessary to further explore and verify these differences.

Conclusions

In summary, our study characterized thorough patterns of genome alterations and the microenvironment in non-smoking LUAD. Compared with smokers, non-smoking LUAD patients had a lower mutation rate in somatic cells, RNAs, and miRNAs with a different microenvironment, including immune cell infiltration and stemness, which revealed the complex association between molecular mechanisms and clinical outcomes. We believe this study will improve our understanding of the mechanism underlying non-smoking LUAD, and effectively prevent, diagnose, and treat LUAD.

Flip graph of the functional enrichment analysis of differential mutation and expression genes between smokers and non-smokers group which was more focused on overlapping of genes between different gene sets (A); bubble chart for all significantly different KEGG pathways (B).

The article’s supplementary files as