Abstract
Objectives: To create a prognostic model based on differentially expressed genes (DEGs) in early lung squamous cell carcinoma (LUSC) and characterize the relationship between risk scores and tumor immune infiltration. Methods: We identified DEGs in normal and tumor tissues that overlapped between LUSC-related data sets from the Gene Expression Omnibus and the Cancer Genome Atlas and evaluated their roles in the diagnosis and prognosis of LUSC by Kaplan-Meier survival analysis, receiver operating characteristic (ROC) analysis, meta-analysis and nomogram analysis. We then constructed a risk model based on Cox regression analysis and the Akaike information criterion and identified the relationship between LUSC risk scores and immune infiltration. Results: Sixty-two overlapping DEGs were involved with keratinocyte differentiation, epidermal cell differentiation, neutrophil migration, granulocyte chemotaxis, granulocyte migration, leukocyte aggregation, and positive regulation of nuclear factor-κB (NF-κB) activity. Overexpression of family with sequence similarity 83 member A (FAM83A) and MYC target 1 (MYCT1), kallikrein related peptidase 8 (KLK8), and downregulation of ADP ribosylation factor like GTPase 14 (ARL14), caspase recruitment domain family member 14 (CARD14), cystatin A (CSTA), dickkopf WNT signaling pathway inhibitor 4 (DKK4), desmoglein 3 (DSG3), and keratin 6B (KRT6B) were associated with a poor prognosis in LUSC and had significant value for LUSC diagnosis. The expression of CSTA, FAM83A, and MYCT1 and high-risk scores were independent risk factors for a poor prognosis in LUSC. A risk nomogram revealed that risk scores could predict the prognosis of LUSC. The risk score was associated with neutrophils, naive B cells, helper follicular T cells, and activated dendritic cells. Conclusions: The expression levels of CSTA, FAM83A, and MYCT1 are related to the diagnosis and prognosis of LUSC and may have potential as therapeutic targets in LUSC. A risk model and nomogram based on CSTA, FAM83A, and MYCT1 can predict the prognosis of LUSC.
Keywords: Differentially expressed genes, receiver operating characteristic, overall survival, lung squamous cell carcinoma, risk model
Introduction
Lung squamous cell carcinoma (LUSC) is a common subtype of non-small cell lung carcinoma, a highly prevalent disease that causes substantial morbidity and mortality [1-5]. In recent years, the prognosis of cancer has improved significantly with advances in treatment methods. Patients with early-stage LUSC can achieve long-term survival with surgical treatment; however, most LUSC is diagnosed at advanced stages with the tumor located in the hilar of the lungs and thus cannot be treated surgically. Furthermore, chemotherapy is ineffective in patients with advanced LUSC, resulting in a poor prognosis [5,6]. Targeted therapy has achieved good results in patients with lung adenocarcinoma (LAC). Therefore, it is important to develop novel molecular targets in patients with LUSC for early diagnosis to improve their prognosis and quality of life.
Elevated levels of abnormal gene expression, microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and other factors are involved in the occurrence and development of LUSC [3,7-11]. For example, lncRNA nicotinamide nucleotide transhydrogenase antisense RNA 1 (NNT-AS1) and forkhead box protein M1 (FOXM1) are frequently up-regulated, while miR-22 is frequently down-regulated in LUSC tissues and cells. Furthermore, NNT-AS1 deletion was found to inhibit LUSC cell migration and invasion, causing apoptosis and inhibiting carcinogenesis by controlling the miR-22/FOXM1 signaling axis [9]. In another study, LUSC cells showed reduced expression of lncRNA STAR Related Lipid Transfer Domain Containing 13 antisense RN (STARD13-AS), and STARD13-AS overexpression could delay the growth and invasion of LUSC cells by controlling the miR-1248/complement C3 (C3A) signaling axis [10]. In contrast, lncRNA RP11-116G8.5 was overexpressed in LUSC cells, and its inhibition could inhibit LUSC cell proliferation, migration, and invasion while speeding up apoptosis. RP11-116G8.5 regulates the expression of PHD finger protein 12 (PHF12) and forkhead box P4 (FOXP4) by acting as a sponge for miR-3150b-3p and miR-6870-5p. However, overexpression of PHF12 and FOXP4 in LUSC cells was shown to reverse the inhibitory effect of RP11-116G8.5 knockdown in cancer cells [11].
Risk models and nomograms are used to assess cancer prognosis [2,12-14]. Here, we screened differentially expressed genes (DEGs) in normal lung tissues and LUSC that were present in Mascaux et al. data in the Gene Expression Omnibus (GEO) [15] and the Cancer Genome Atlas (TCGA) databases. We then evaluated the clinical values of DEGs that are critical in the progression of LUSC using the Kaplan-Meier (K-M) survival analysis, the receiver operating characteristic (ROC) analysis, and the risk model construction. These investigations provided novel diagnostic target molecules and prognostic biomarkers for better management of patients with LUSC. We also constructed a risk model to predict the prognosis of LUSC.
Materials and methods
LUSC gene expression and clinical data
We retrieved and downloaded the Series Matrix File (s) in the GSE33479 dataset from the GEO database, which was generated using the GPL6480: Agilent-014850 Whole Human Genome Microarray 4 × 44K G4112F platform and includes gene expression data from 27 normal tissues, 14 tissues with squamous metaplasia, 13 carcinoma in situ tissues and 14 LUSC tissues. In addition, we obtained Fragments per Kilobase Million (FPKM)-type gene expression data and clinical data from the TCGA database. After removing entries with missing values or incomplete clinical information, our study sample included 49 normal tissues, 502 LUSC tissues, and clinical data from 490 patients.
Overlapping DEGs of the GSE33479 data set
We use the limma package to identify genes that were differentially expressed during progression from normal lung tissues to squamous metaplasia tissues, carcinomas in situ, or LUSC, with a fold change of 1 and an adjusted P value < 0.05 as the screening criteria. The DEGs that overlapped the three groups of cancer tissues were visualized using a Venn diagram and a heat map.
The biological functions, signaling pathways, and protein-protein interaction network of the DEGs
Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) are commonly used to analyze biological functions and signaling mechanisms involving multiple genes [2,16]. The GO type includes three types of information: biological processes, cellular components, and molecular functions. We used GO and KEGG analyses to identify the enriched biological processes and signaling mechanisms of overlapping DEGs, with an adjusted P value < 0.05 as the screening criterion. Furthermore, we constructed a protein-protein interaction (PPI) network between the overlapping DEGs and DEGs in the STRING database and visualized it using Cytoscape software (version 3.8.2). The critical DEGs in the PPI network were visualized using the CytoHubba plug-in.
Identification of DEGs in LUSC tissues
We obtained expression data of the identified DEGs in 49 normal lungs and 502 LUSC tissues from TCGA. The expression levels of the DEGs in unpaired LUSC tissues were investigated using the Wilcoxon rank sum test, with P < 0.05 as the screening criterion. In addition to the unpaired samples, 49 pairs of matched normal lung and LUSC tissues were matched. The expression levels of DEGs in unpaired LUSC tissues were compared with those of paired LUSC tissues, with P < 0.05 as the screening criterion.
K-M survival analysis
We paired and merged the DEG expression data with the survival data of patients with LUSC in TCGA. We then performed a K-M survival analysis to investigate the impact of high or low DEG expression on the prognosis of patients with LUSC [2,16], using P < 0.05 as the filter criterion.
Determination of the diagnostic value of LUSC prognostic genes
ROC analysis is typically used to determine the diagnostic value of gene expression levels in cancer, with an area under the curve (AUC) between 0.5 and 1.0 as the evaluation standard. The higher the AUC, the greater the diagnostic value. We used ROC analysis to investigate the importance of expression of ADP ribosylation factor like GTPase 14 (ARL14), caspase recruitment domain family member 14 (CARD14), cystatin A (CSTA), dickkopf WNT signaling pathway inhibitor 4 (DKK4), desmoglein 3 (DSG3), family with sequence similarity 83 member A (FAM83A), kallikrein related peptidase 8 (KLK8), keratin 6B (KRT6B), and MYC target 1 (MYCT1) in LUSC.
Establishment of a risk model
We used univariate Cox regression to investigate the association between the expression levels of ARL14, CARD14, CSTA, DKK4, DSG3, FAM83A, KLK8, KRT6B, and MYCT1 and the prognosis of LUSC, with P < 0.05 as the elimination criterion. Furthermore, we employed multivariate Cox regression and Akaike information criterion (AIC) to identify the relationship between the expression levels of CSTA, FAM83A, and MYCT1 and the prognosis of LUSC. We also incorporated risk scoring with LUSC tissue samples to construct a risk model for patients with LUSC.
Validation of risk gene expression in LUSC tissues
We collected tumor tissues and normal tissues from seven patients with LUSC diagnosed with pathology that were surgically treated at our hospital from July to August 2022. Patients gave their signed informed consent, and the ethics committee approved of this study at Wuhan Central Hospital (WHZXKYL2022-192). We extracted and quantified total RNA from tissue samples and performed reverse transcription according to the instructions of the reverse transcription RNA kit. We then performed reverse transcription quantitative polymerase chain reaction (RT-qPCR) and calculated the relative expression levels of CSTA, FAM83A, and MYCT1 in the tissue samples. Table 1 shows the primers for CSTA, FAM83A, and MYCT1.
Table 1.
PCR primers used in the study
| Gene | Forward primer | Reversed primer |
|---|---|---|
| CSTA | 5’-AATGATACCTGGAGGCTTATCT-3’ | 5’-TTTATTATCACCTGCTCGTACC-3’ |
| FAM83A | 5’-CCCATCTCAGTCACTGGCATT-3’ | 5’-CCGCCAACATCTCCTTGTTC-3’ |
| MYCT1 | 5’-GCCAGAAAACTTTTGGGAGGA-3’ | 5’-ATCCAGTTCTGTTGAGGCCG-3’ |
Note: PCR, Polymerase Chain Reaction; CSTA, Cystatin A; FAM83A, Family With Sequence Similarity 83 Member A; MYCT1, MYC Target 1.
Construction of a prognostic factor-related nomogram
We constructed a nomogram of ARL14, CARD14, CSTA, DKK4, DSG3, FAM83A, KLK8, KRT6B, and MYCT1 expression based on the expression levels in cancer tissues, diagnosis, and prognostic values of ARL14, CARD14, CSTA, DKK4, DSG3, FAM83A, KLK8, KRT6B, and MYCT1.
The values of risk factors and construction of a risk score-related nomogram
We identified the relationship between high and low expression of CSTA, FAM83A, and MYCT1, categorized according to the median expression level of each gene and the clinicopathological characteristics of patients with LUSC using the meta-analysis and the K-M survival analysis functions in the online lung cancer Explorer and the UALCAN database. We also analyzed univariate Cox regression to explore the relationship between risk score and LUSC prognosis. Finally, we constructed a prognostic nomogram related to the clinical stage, T stage, age, and risk score of patients with LUSC.
The relationship between the risk score and LUSC immune infiltration
We used the CIBERSORT algorithm to calculate the levels of immune cell infiltration in the LUSC samples in TCGA. We then divided the samples into high-risk and low-risk groups based on the scores from our risk model and used the limma package to analyze the relationship between high- and low-risk scores and LUSC-infiltrating immune cells, with P < 0.05 as the screening criterion.
Statistical analysis
Gene expression levels in LUSC were investigated using the Wilcoxon rank sum test and limma package. We use ROC analysis to assess the diagnostic value of gene expression levels, with AUC between 0.5 and 1.0 as the evaluation standard. The higher the AUC, the more significant the diagnostic value. We used univariate and multivariate Cox regression analyses to investigate prognostic risk factors in patients with LUSC. We analyzed the relationship between risk scores and immune cell infiltration of LUSC by correlation analysis, with P < 0.05 as the threshold for statistical significance.
Results
Overlapping DEGs in normal tissues and squamous metaplasia tissues, carcinomas in situ, and LUSC tissues
There were 150 significant DEGs in tissues with squamous metaplasia compared to normal tissues (Table 2). Among them, 132 were overexpressed in metaplasia tissues, and 18 had reduced expression in metaplasia tissues compared to normal tissues. There were 1427 significant DEGs in carcinoma in situ tissues compared to normal tissues (Table S1), among which 996 were overexpressed in carcinoma tissues, and 431 had reduced expression in carcinoma tissues. There were 3137 significant DEGs in the LUSC tissues compared with the normal tissues (Table S2), of which 1758 were overexpressed in the LUSC tissues, and 1379 had reduced expression in the LUSC tissues compared with the normal tissues (Table S2). The conversion of 70 DEGs from genes that overlapped among the three groups revealed 62 unique overlapping DEGs (Figure S1 and Table 3).
Table 2.
Differentially expressed genes in squamous metaplasia tissues
| id | logFC | id | logFC | Id | logFC |
|---|---|---|---|---|---|
| A_24_P175519 | 1.305668979 | A_32_P31744 | -1.273349042 | A_23_P207213 | 2.375811323 |
| A_23_P73097 | 1.676109788 | A_32_P63113 | 1.874350357 | A_23_P106806 | 1.479353625 |
| A_23_P60248 | 1.453737069 | A_23_P216052 | 1.723853713 | A_24_P412088 | 1.253757953 |
| A_23_P48350 | 1.257334441 | A_23_P4335 | 1.046156459 | A_23_P69537 | 2.34741673 |
| A_24_P220947 | 1.093857676 | A_24_P673063 | 1.362193272 | A_23_P58266 | 1.193511551 |
| A_23_P170233 | 1.959207116 | A_23_P41114 | 2.406915532 | A_32_P315178 | 1.7071763 |
| A_23_P370635 | 1.00577998 | A_32_P149158 | -1.016032997 | A_23_P81190 | 85.19582085 |
| A_23_P135257 | 1.791428294 | A_24_P212086 | 1.444105599 | A_23_P155711 | 1.144977224 |
| A_23_P369343 | 3.707520566 | A_23_P353524 | 1.957193091 | A_32_P198978 | 1.926114993 |
| A_32_P112452 | 3.777406989 | A_23_P166269 | 1.001215093 | A_23_P24129 | 2.312543966 |
| A_24_P355006 | -1.318793118 | A_23_P360329 | 1.044416316 | A_24_P916782 | -1.232122428 |
| A_32_P71032 | 2.691356836 | A_23_P4494 | 1.05280376 | A_32_P200238 | 1.152174573 |
| A_23_P23048 | 1.672129448 | A_23_P111766 | 5.614856196 | A_32_P62963 | 1.252592288 |
| A_24_P589301 | 2.552030506 | A_23_P115478 | 1.780520598 | A_23_P17134 | 2.900480239 |
| A_24_P348118 | 1.254147483 | A_24_P7642 | 1.324696291 | A_32_P52153 | 1.067986958 |
| A_24_P55092 | 1.104058615 | A_23_P310274 | 3.939282425 | A_32_P176790 | -1.013091836 |
| A_23_P78248 | 1.054512256 | A_23_P92562 | 1.276041648 | A_24_P306896 | 1.592324703 |
| A_32_P174121 | 1.832379447 | A_23_P163338 | 1.562685678 | A_32_P158272 | 1.014369349 |
| A_24_P152845 | 1.798337561 | A_23_P3038 | 1.215441259 | A_32_P141948 | 1.08559652 |
| A_23_P108062 | 1.60850786 | A_32_P67266 | 1.828537263 | A_23_P92161 | 1.812216887 |
| A_32_P161855 | 1.259332336 | A_32_P190303 | -1.104816576 | A_32_P119830 | -1.122679204 |
| A_24_P152968 | 1.199453308 | A_23_P153120 | 1.134293665 | A_23_P11644 | 1.07581238 |
| A_23_P113793 | 1.583887983 | A_32_P168973 | 1.468658268 | A_24_P945059 | -1.066996871 |
| A_23_P356494 | 1.813166276 | A_24_P282266 | 1.558407964 | A_23_P74001 | 1.689310735 |
| A_24_P129341 | 1.596110416 | A_32_P186364 | -1.213002845 | A_24_P913146 | 1.853254949 |
| A_23_P208126 | 1.22910464 | A_24_P252155 | 1.287260405 | A_23_P16523 | 1.088036451 |
| A_23_P258190 | 1.349123989 | A_24_P226755 | -1.005791877 | A_32_P37867 | 1.160838789 |
| A_24_P859859 | 2.760650021 | A_23_P136724 | 1.153515082 | A_23_P108216 | -1.008609446 |
| A_23_P500010 | 2.002885752 | A_32_P94444 | 3.755643047 | A_23_P8801 | 2.06619658 |
| A_23_P163336 | 1.070258728 | A_24_P104689 | 1.442011501 | A_32_P34138 | 3.507513336 |
| A_32_P204676 | 1.259019153 | A_23_P213050 | 1.224115167 | A_23_P76743 | 1.027252127 |
| A_23_P59877 | 1.194315142 | A_23_P94275 | 1.571666444 | A_23_P128574 | -1.064032554 |
| A_24_P238250 | 1.243422754 | A_24_P68908 | 1.617632117 | A_23_P204947 | 1.072236955 |
| A_23_P18751 | 2.804584864 | A_24_P43810 | 1.624736366 | A_23_P214935 | -1.10396774 |
| A_23_P209995 | 1.430272841 | A_32_P51855 | 1.258839681 | A_32_P471485 | 1.824972945 |
| A_23_P30126 | 1.982165789 | A_24_P918065 | 1.106835785 | A_23_P45751 | 1.589627155 |
| A_23_P38537 | 1.133831182 | A_23_P151975 | 1.200385538 | A_23_P351148 | 1.869926295 |
| A_23_P140928 | 1.06922676 | A_23_P124095 | 2.292832348 | A_23_P256425 | 1.126512375 |
| A_23_P93641 | 1.593320408 | A_32_P71710 | 2.505112783 | A_23_P74723 | 1.14070976 |
| A_23_P432978 | 1.190852292 | A_24_P153035 | 1.478314805 | A_24_P31627 | -1.289108403 |
| A_23_P52067 | 1.542073304 | A_24_P48495 | 1.050669737 | A_23_P55198 | -1.113960254 |
| A_32_P141338 | 1.289107314 | A_24_P360674 | 1.040875417 | A_23_P320070 | 1.03797418 |
| A_23_P434809 | 6.448052031 | A_23_P430718 | 1.743908998 | A_32_P128174 | -1.559302387 |
| A_23_P500000 | 2.56451442 | A_24_P408736 | 1.007930946 | A_23_P358917 | 2.00037122 |
| A_23_P52410 | 1.236176037 | A_24_P71781 | -1.37130085 | A_32_P157208 | -1.172676541 |
| A_24_P416645 | 1.825724421 | A_23_P66739 | 2.506683538 | A_24_P313895 | 1.150806867 |
| A_23_P324754 | 1.247052108 | A_23_P76249 | 1.00272057 | A_23_P250385 | 1.699721426 |
| A_23_P217498 | 2.365430447 | A_23_P119015 | 1.173560213 | A_23_P155660 | 1.012706161 |
| A_23_P159406 | 1.742366668 | A_24_P245379 | 1.855535138 | A_24_P399490 | 1.055003049 |
| A_23_P201706 | 1.768636607 | A_23_P123234 | 1.132425178 | A_32_P150891 | 1.040569253 |
Note: DEGs, Differentially Expressed Genes; FC, Fold Change.
Table 3.
Differentially expressed genes associated with lung squamous cell carcinoma progression after gene name conversion
| id | Gene | Id | Gene |
|---|---|---|---|
| A_23_P106806 | PRSS27 | A_23_P73097 | RGS20 |
| A_23_P108216 | FXYD1 | A_23_P74001 | S100A12 |
| A_23_P113793 | ZBED2 | A_23_P76249 | KRT6B |
| A_23_P11644 | SPRR2D | A_23_P92161 | ARL14 |
| A_23_P124095 | CALML5 | A_23_P93641 | AKR1B10 |
| A_23_P128574 | ENOX1 | A_23_P94275 | DKK4 |
| A_23_P136724 | LOC344887 | A_24_P129341 | AKR1B10 |
| A_23_P151975 | RHCG | A_24_P152968 | AKR1C1 |
| A_23_P153120 | DSG3 | A_24_P220947 | AKR1C1 |
| A_23_P155660 | PPP2R2C | A_24_P226755 | TOX |
| A_23_P155711 | NEIL3 | A_24_P31627 | KCNB1 |
| A_23_P159406 | SPRR1B | A_24_P355006 | ADAM22 |
| A_23_P170233 | CSTA | A_24_P412088 | MCM10 |
| A_23_P18751 | TMPRSS11E | A_24_P43810 | FAM83A |
| A_23_P201706 | S100A2 | A_24_P48495 | LYPD3 |
| A_23_P209995 | IL1RN | A_24_P673063 | FABP5 |
| A_23_P216052 | FAM83A | A_24_P68908 | LOC344887 |
| A_23_P23048 | S100A9 | A_24_P7642 | FABP5 |
| A_23_P24129 | DKK1 | A_24_P945059 | MYCT1 |
| A_23_P256425 | ADAMDEC1 | A_32_P112452 | LOC100652944 |
| A_23_P258190 | AKR1B1 | A_32_P119830 | PEG3-AS1 |
| A_23_P310274 | PRSS2 | A_32_P149158 | PLCL1 |
| A_23_P320070 | CARD14 | A_32_P150891 | DIAPH3 |
| A_23_P353524 | IVL | A_32_P157208 | LOC572558 |
| A_23_P369343 | KLK8 | A_32_P168973 | KRT16P3 |
| A_23_P38537 | KRT16 | A_32_P190303 | LONRF2 |
| A_23_P434809 | S100A8 | A_32_P200238 | UCA1 |
| A_23_P4494 | DSC2 | A_32_P204676 | FABP5 |
| A_23_P52067 | GRHL3 | A_32_P34138 | FAM25A |
| A_23_P55198 | CNTD1 | A_32_P62963 | KRT16P2 |
| A_23_P59877 | FABP5 | A_32_P94444 | PRSS2 |
Note: PRSS27, Serine Protease 27; FXYD1, FXYD Domain Containing Ion Transport Regulator 1; ZBED2, Zinc Finger BED-Type Containing 2; SPRR2D, Small Proline Rich Protein 2D; CALML5, Calmodulin Like 5; ENOX1, Ecto-NOX Disulfide-Thiol Exchanger 1; RHCG, Rh Family C Glycoprotein; DSG3, Desmoglein 3; PPP2R2C, Protein Phosphatase 2 Regulatory Subunit Bgamma; NEIL3, Nei Like DNA Glycosylase 3; SPRR1B, Small Proline Rich Protein 1B; CSTA, Cystatin A; TMPRSS11E, Transmembrane Serine Protease 11E; S100A2, S100 Calcium Binding Protein A2; IL1RN, Interleukin 1 Receptor Antagonist; FAM83A, Family With Sequence Similarity 83 Member A; S100A9, S100 Calcium Binding Protein A9; DKK1, Dickkopf WNT Signaling Pathway Inhibitor 1; ADAMDEC1, ADAM Like Decysin 1; AKR1B1, Aldo-Keto Reductase Family 1 Member B; PRSS2, Serine Protease 2; CARD14, Caspase Recruitment Domain Family Member 14; IVL, Involucrin; KLK8, Kallikrein Related Peptidase 8; KRT16, Keratin 16; S100A8, S100 Calcium Binding Protein A8; DSC2, Desmocollin 2; GRHL3, Grainyhead Like Transcription Factor 3; CNTD1, Cyclin N-Terminal Domain Containing 1; FABP5, Fatty Acid Binding Protein 5; RGS20; Regulator Of G Protein Signaling 20; S100A12, S100 Calcium Binding Protein A12; KRT6B, Keratin 6B; ARL14, ADP Ribosylation Factor Like Gtpase 14; AKR1B10, Aldo-Keto Reductase Family 1 Member B10; DKK4, Dickkopf WNT Signaling Pathway Inhibitor 4; AKR1C1, Aldo-Keto Reductase Family 1 Member C1; TOX, Thymocyte Selection Associated High Mobility Group Box; KCNB1, Potassium Voltage-Gated Channel Subfamily B Member 1; ADAM22, ADAM Metallopeptidase Domain 22; MCM10, Minichromosome Maintenance 10 Replication Initiation Factor; FAM83A, Family With Sequence Similarity 83 Member A; LYPD3, LY6/PLAUR Domain Containing 3; MYCT1, MYC Target 1; PLCL1, Phospholipase C Like 1; DIAPH3, Diaphanous Related Formin 3; KRT16P3, Keratin 16 Pseudogene 3; LONRF2, LON Peptidase N-Terminal Domain And Ring Finger 2; UCA1, Urothelial Cancer Associated 1; FAM25A, Family With Sequence Similarity 25 Member A; KRT16P2, Keratin 16 Pseudogene 2.
The roles and signaling mechanisms of DEG enrichment and ppi network construction
The functions of the overlapping DEGs from the GEO database included epidermis development, cornification, skin development, keratinization, keratinocyte differentiation, epidermal cell differentiation, glycoside metabolism, neutrophil chemotaxis, and migration, granulocyte chemotaxis and migration, leukocyte aggregation and migration involved in an inflammatory response, positive regulation of nuclear factor-κB (NF-κB) transcription factor activity, secondary metabolic processes, and protein nitrosylation, among others (Figure S2A-S2C; Table S3). The overlapping DEGs were associated with signaling mechanisms for folate biosynthesis, galactose metabolism, fruit and mannose metabolism, pentose and glucuronate interconversions, and glycerolipid metabolism (Figure S2D). Figure S3A depicts the PPI network between the overlapping DEGs from the GEO data and the DEGs in the STRING database. The key DEGs in the PPI network included SPRR1B, KRT16, IVL, CSTA, and S100A8 (Figure S3B).
Identification of crucial DEGs in LUSC
ADAM metallopeptidase domain 22 (ADAM22), ADAM like decysin 1 (ADAMDEC1), aldo-keto reductase family 1 member B (AKR1B1), aldo-keto reductase family 1 member B10 (AKR1B10), aldo-keto reductase family 1 member C1 (AKR1C1), ARL14, calmodulin like 5 (CALML5), CARD14, CSTA, diaphanous related formin 3 (DIAPH3), dickkopf WNT signaling pathway inhibitor 1 (DKK1), dickkopf WNT signaling pathway inhibitor 4 (DKK4), desmocollin 2 (DSC2), desmocollin 3 (DSG3), family with sequence similarity 25 member A (FAM25A), FAM83A, FXYD domain containing ion transport regulator 1 (FXYD1), grainyhead like transcription factor 3 (GRHL3), interleukin 1 receptor antagonist (IL1RN), involucrin (IVL), potassium voltage-gated channel subfamily B member 1 (KCNB1), KLK8, keratin 16 (KRT16), keratin 16 pseudogene 2 (KRT16P2), keratin 16 pseudogene 3 (KRT16P3), KRT6B, LON peptidase N-terminal domain and ring finger 2 (LONRF2), LY6/PLAUR domain containing 3 (LYPD3), minichromosome maintenance 10 replication initiation factor (MCM10), MYCT1, nei like DNA glycosylase 3 (NEIL3), phospholipase C like 1 (PLCL1), protein phosphatase 2 regulatory subunit Bgamma (PPP2R2C), serine protease 2 (PRSS2), serine protease 27 (PRSS27), regulator of G protein signaling 20 (RGS20), Rh family C glycoprotein (RHCG), S100 calcium binding protein A12 (S100A12), S100 calcium binding protein A2 (S100A2), S100 calcium binding protein A9 (S100A9), small proline rich protein 1B (SPRR1B), small proline rich protein 2D (SPRR2D), thymocyte selection associated high mobility group box (TOX), urothelial cancer associated 1 (UCA1), and zinc finger BED-type containing 2 (ZBED2) showed significant changes in expression between unpaired LUSC tissues and normal lung tissues (Figures 1 and S4). ADAMDEC1, AKR1B1, AKR1B10, AKR1C1, ARL14, CALML5, CARD14, CSTA, DIAPH3, DKK1, DKK4, DSC2, DSG3, FAM25A, FAM83A, FXYD1, GRHL3, IL1RN, IVL, KCNB1, KLK8, KRT16, KRT16P3, KRT6B, LONRF2, LYPD3, MCM10, MYCT1, NEIL3, PLCL1, PPP2R2C, PRSS2, PRSS27, RGS20, RHCG, S100A12, S100A2, S100A9, SPRR1B, SPRR2D, TMPRSS11E, TOX, UCA1, and ZBED2 had significantly different expression between the 49 LUSC tissues and the 49 paired normal lung tissues (Figures 2 and S5).
Figure 1.
The differentially expressed genes in unpaired lung squamous cell carcinoma tissues of the Cancer Genome Atlas database. A. ADAM22; B. ADAMDEC1; C. AKR1B1; D. AKR1B10; E. AKR1C1; F. ARL14; G. CALML5; H. CARD14; I. CSTA; J. DIAPH3; K. DKK1; L. DKK4; M. DSC2; N. DSG3; O. FAM25A; P. FAM83A; Q. FXYD1; R. GRHL3; S. IL1RN; T. IVL; U. KCNB1; V. KLK8; W. KRT6B; X. KRT16; Y. KRT16P2. Note: LUSC, Lung Squamous Cell Carcinoma; ADAM22, ADAM Metallopeptidase Domain 22; ADAMDEC1, ADAM Like Decysin 1; AKR1B1, Aldo-Keto Reductase Family 1 Member B; AKR1B10, Aldo-Keto Reductase Family 1 Member B10; AKR1C1, Aldo-Keto Reductase Family 1 Member C1; ARL14, ADP Ribosylation Factor Like Gtpase 14; CALML5, Calmodulin Like 5; CARD14, Caspase Recruitment Domain Family Member 14; CSTA, Cystatin A; DIAPH3, Diaphanous Related Formin 3; DKK1, Dickkopf WNT Signaling Pathway Inhibitor 1; DKK4, Dickkopf WNT Signaling Pathway Inhibitor 4; DSC2, Desmocollin 2; DSC3, Desmocollin 3; FAM25A, Family With Sequence Similarity 25 Member A; FAM83A, Family With Sequence Similarity 83 Member A; FXYD1, FXYD Domain Containing Ion Transport Regulator 1; GRHL3, Grainyhead Like Transcription Factor 3; IL1RN, Interleukin 1 Receptor Antagonist; IVL, Involucrin; KCNB1, Potassium Voltage-Gated Channel Subfamily B Member 1; KLK8, Kallikrein Related Peptidase 8; KRT6B, Keratin 6B; KRT16, Keratin 16; KRT16P2, Keratin 16 Pseudogene 2; DSG3, Desmoglein 3.
Figure 2.
Differentially expressed genes in paired lung squamous cell carcinoma tissues from the Cancer Genome Atlas database. A. ADAMDEC1; B. AKR1B10; C. AKR1C1; D. ARL14; E. CALML5; F. CSTA; G. CARD14; H. DIAPH3; I. DKK1; J. DKK4; K. DSC2; L. DSG3; M. FAM25A; N. FAM83A; O. FXYD1; P. GRHL3; Q. IL1RN; R. IVL; S. KLK8; T. KCNB1; U. KRT6B; V. KRT16; W. LYPD3; X. MCM10; Y. KRT16P3. Note: LUSC, Lung Squamous Cell Carcinoma; ADAMDEC1, ADAM Like Decysin 1; AKR1B10, Aldo-Keto Reductase Family 1 Member B10; AKR1C1, Aldo-Keto Reductase Family 1 Member C1; ARL14, ADP Ribosylation Factor Like Gtpase 14; CALML5, Calmodulin Like 5; CARD14, Caspase Recruitment Domain Family Member 14; CSTA, Cystatin A; DIAPH3, Diaphanous Related Formin 3; DKK1, Dickkopf WNT Signaling Pathway Inhibitor 1; DKK4, Dickkopf WNT Signaling Pathway Inhibitor 4; DSC2, Desmocollin 2; DSC3, Desmocollin 3; FAM25A, Family With Sequence Similarity 25 Member A; FAM83A, Family With Sequence Similarity 83 Member A; FXYD1, FXYD Domain Containing Ion Transport Regulator 1; GRHL3, Grainyhead Like Transcription Factor 3; IL1RN, Interleukin 1 Receptor Antagonist; IVL, Involucrin; KCNB1, Potassium Voltage-Gated Channel Subfamily B Member 1; KLK8, Kallikrein Related Peptidase 8; KRT6B, Keratin 6B; KRT16, Keratin 16; LYPD3, LY6/PLAUR Domain Containing 3; MCM10, Minichromosome Maintenance 10 Replication Initiation Factor; KRT16P3, Keratin 16 Pseudogene 3; DSG3, Desmoglein 3.
Construction of a prognostic nomogram for patients with LUSC
FAM83A, KLK8, and MYCT1 overexpression levels were each correlated with the short overall survival (OS) of patients with LUSC. In contrast, lower expression levels of ARL14, CARD14, CSTA, DKK4, DSG3, and KRT6B were each associated with a short OS in LUSC patients according to the K-M survival plots (Figure 3). ROC analysis showed that the expression levels of FAM83A, MYCT1, ARL14, CARD14, CSTA, DKK4, DSG3, KLK8, and KRT6B could be used to diagnose LUSC (Figure 4). The AUCs of ARL14, CARD14, CSTA, DKK4, DSG3, FAM83A, KLK8, KRT6B and MYCT1 expression in LUSC were 0.717, 0.981, 0.889, 0.847, 0.956, 0.94, 0.836, 0.959, and 0.997, respectively. A nomogram was constructed for DEGs related to LUSC prognosis and diagnosis to assess the prognosis of patients with LUSC (Figure S6).
Figure 3.
Kaplan-Meyer survival analysis of patients with lung squamous cell carcinoma stratified by expression levels of candidate prognostic genes. A. ARL14; B. CARD14; C. CSTA; D. DKK4; E. DSG3; F. FAM83A; G. KLK8; H. KRT6B; I. MYCT1. Note: ARL14, ADP Ribosylation Factor Like Gtpase 14; CARD14, Caspase Recruitment Domain Family Member 14; CSTA, Cystatin A; FAM83A, Family With Sequence Similarity 83 Member A; KLK8, Kallikrein Related Peptidase 8; KRT6B, Keratin 6B; DKK4, Dickkopf WNT Signaling Pathway Inhibitor 4; DSG3, Desmoglein 3; MYCT1, MYC Target 1.
Figure 4.
Receiver operating characteristic analysis of the diagnostic values of differentially expressed genes in LUSC. A. ARL14; B. CARD14; C. CSTA; D. DKK4; E. DSG3; F. FAM83A; G. KLK8; H. KRT6B; I. MYCT1. Note: LUSC, Lung Squamous Cell Carcinoma; ARL14, ADP Ribosylation Factor Like Gtpase 14; CARD14, Caspase Recruitment Domain Family Member 14; CSTA, Cystatin A; FAM83A, Family With Sequence Similarity 83 Member A; KLK8, Kallikrein Related Peptidase 8; KRT6B, Keratin 6B; DKK4, Dickkopf WNT Signaling Pathway Inhibitor 4; DSG3, Desmoglein 3; MYCT1, MYC Target 1.
Constructing a prognostic risk model of LUSC
We used univariate Cox regression to examine the relationship between the expression levels of ARL14, CARD14, CSTA, DKK4, DSG3, FAM83A, KLK8, KRT6B, and MYCT1 and the prognosis of patients with LUSC. We found that CSTA, FAM83A, and MYCT1 each had a significant prognostic value (Figure 5A). Furthermore, multivariate Cox regression and AIC screening revealed that the expression levels of CSTA, FAM83A, and MYCT1 independently influenced the prognosis of LUSC (Figure 5B and Table 4). Using the expression levels of these three genes, we constructed a risk score as follows: risk score = (CSTA × -0.099643117) + (FAM83A × 0.117245801) + (MYCT1 × 0.246912866). Figure 5C-E shows the relationship between risk scores and OS in patients with LUSC, confirming that patients with high-risk scores had poor prognoses.
Figure 5.
A prognostic-related risk model for LUSC. A. Prognostic risk factors identified by univariate Cox regression analysis. B. Multivariate Cox regression and AIC screening confirmed the relationship between risk factors and patient prognosis. C, D. The relationship between overall survival of patients with lung squamous cell carcinoma and risk factors. E. Patients with high-risk scores in the risk model have a poor prognosis. F, G. Cox regression analysis identified prognostic factors in lung squamous cell carcinoma. Note: LUSC, Lung Squamous Cell Carcinoma; CSTA, Cystatin A; FAM83A, Family With Sequence Similarity 83 Member A; MYCT1, MYC Target 1.
Table 4.
Univariate Cox regression analysis revealed prognosis-related genes
| Gene | HR | P |
|---|---|---|
| CSTA | 0.915229739 | 0.015172199 |
| FAM83A | 1.099601051 | 0.010161734 |
| MYCT1 | 1.335569478 | 0.015580685 |
Note: HR, Hazard Ratio; CSTA, Cystatin A; FAM83A, Family With Sequence Similarity 83 Member A; MYCT1, MYC Target 1.
Prognostic risk model nomogram
PCR confirmed that the expression of CSTA was reduced in 71.43% (5/7) of patients with LUSC from our hospital (Figure S7A), while the expression of FAM83A and MYCT1 increased in 100% and 71.43% (5/7) of the patients, respectively (Figure S7B and S7C). Meta-analysis and K-M survival analysis using Lung Cancer Explorer and UALCAN databases showed that high expression of MYCT1 and FAM83A and low expression of CSTA were associated with poor prognosis in patients with LUSC (Table 5). CSTA expression in TCGA LUSC tissues was related to progression-free interval endpoint events in patients with LUSC after the samples were grouped according to the median expression level of CSTA (Table 6). On the contrary, expression levels of FAM83A and MYCT1, similarly grouped by their median values, were associated with primary therapy outcomes of patients with LUSC (Tables 7 and 8). Furthermore, there were significant differences in the expression levels of CSTA, FAM83A, and MYCT1 in the risk model score grouping (Figure S8), indicating that the risk model based on the expression of CSTA, FAM83A, and MYCT1 could predict disease progression and prognosis in patients with LUSC.
Table 5.
Expression of MYCT1, FAM83A, and CSTA was associated with a poor prognosis in patients with lung squamous cell carcinoma
| Gene | HR (95% CI) | P | Method | Ref |
|---|---|---|---|---|
| MYCT1 | 1.01 (0.89-1.15) | 0.84 | Meta-analysis | LCE database |
| FAM83A | 1.14 (1.04-1.25) | ** | Meta-analysis | LCE database |
| CSTA | 0.90 (0.82-0.99) | * | Meta-analysis | LCE database |
| MYCT1 | NA | * | K-M survival | UALCAN database |
| FAM83A | NA | *** | K-M survival | UALCAN database |
| CSTA | NA | 0.64 | K-M survival | UALCAN database |
Note: HR, Hazard Ratio; CI, Confidence Interval; LCE, Lung Cancer Explorer; K-M, Kaplan-Meier; CSTA, Cystatin A; FAM83A, Family With Sequence Similarity 83 Member A; MYCT1, MYC Target 1; *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Table 6.
Clinicopathological characteristics of patients with high and low CSTA expression in LUSC
| Characteristic | Low expression of CSTA | High expression of CSTA | P |
|---|---|---|---|
| T stage | 0.120 | ||
| T1 | 60 (12%) | 54 (10.8%) | |
| T2 | 135 (26.9%) | 159 (31.7%) | |
| T3 | 43 (8.6%) | 28 (5.6%) | |
| T4 | 13 (2.6%) | 10 (2%) | |
| N stage | 0.679 | ||
| N0 | 161 (32.5%) | 159 (32.1%) | |
| N1 | 65 (13.1%) | 66 (13.3%) | |
| N2 | 20 (4%) | 20 (4%) | |
| N3 | 1 (0.2%) | 4 (0.8%) | |
| M stage | 0.723 | ||
| M0 | 208 (49.6%) | 204 (48.7%) | |
| M1 | 3 (0.7%) | 4 (1%) | |
| Pathologic stage | 0.625 | ||
| Stage I | 119 (23.9%) | 126 (25.3%) | |
| Stage II | 78 (15.7%) | 84 (16.9%) | |
| Stage III | 47 (9.4%) | 37 (7.4%) | |
| Stage IV | 3 (0.6%) | 4 (0.8%) | |
| Primary therapy outcome | 0.494 | ||
| PD | 19 (5.3%) | 12 (3.3%) | |
| SD | 7 (1.9%) | 10 (2.8%) | |
| PR | 2 (0.6%) | 3 (0.8%) | |
| CR | 151 (41.8%) | 157 (43.5%) | |
| Gender | 0.155 | ||
| Female | 73 (14.5%) | 58 (11.6%) | |
| Male | 178 (35.5%) | 193 (38.4%) | |
| Race | 0.890 | ||
| Asian | 4 (1%) | 5 (1.3%) | |
| Black or African American | 14 (3.6%) | 16 (4.1%) | |
| White | 176 (45.2%) | 174 (44.7%) | |
| Age | 0.482 | ||
| ≤ 65 | 91 (18.5%) | 100 (20.3%) | |
| > 65 | 155 (31.4%) | 147 (29.8%) | |
| Smoker | 0.824 | ||
| No | 10 (2%) | 8 (1.6%) | |
| Yes | 236 (48.2%) | 236 (48.2%) | |
| OS event | 0.241 | ||
| Alive | 136 (27.1%) | 150 (29.9%) | |
| Dead | 115 (22.9%) | 101 (20.1%) | |
| DSS event | 0.276 | ||
| Alive | 177 (39.3%) | 184 (40.9%) | |
| Dead | 50 (11.1%) | 39 (8.7%) | |
| PFI event | 0.040 | ||
| Alive | 166 (33.1%) | 188 (37.5%) | |
| Dead | 85 (16.9%) | 63 (12.5%) |
Note: CSTA, Cystatin A; CR, Complete Response; PR, Partial Response; SD, Stable Disease; PD, Progressive Disease; OS, Overall Survival; DSS, Disease-Specific Survival; PFI, Progression-Free Interval; LUSC, Lung Squamous Cell Carcinoma.
Table 7.
Clinicopathological characteristics of patients with high and low FAM83A expression in LUSC
| Characteristic | Low expression of FAM83A | High expression of FAM83A | P |
|---|---|---|---|
| T stage | 0.262 | ||
| T1 | 60 (12%) | 54 (10.8%) | |
| T2 | 150 (29.9%) | 144 (28.7%) | |
| T3 | 28 (5.6%) | 43 (8.6%) | |
| T4 | 13 (2.6%) | 10 (2%) | |
| N stage | 0.762 | ||
| N0 | 161 (32.5%) | 159 (32.1%) | |
| N1 | 70 (14.1%) | 61 (12.3%) | |
| N2 | 18 (3.6%) | 22 (4.4%) | |
| N3 | 2 (0.4%) | 3 (0.6%) | |
| M stage | 0.720 | ||
| M0 | 210 (50.1%) | 202 (48.2%) | |
| M1 | 3 (0.7%) | 4 (1%) | |
| Pathologic stage | 0.736 | ||
| Stage I | 120 (24.1%) | 125 (25.1%) | |
| Stage II | 87 (17.5%) | 75 (15.1%) | |
| Stage III | 40 (8%) | 44 (8.8%) | |
| Stage IV | 3 (0.6%) | 4 (0.8%) | |
| Primary therapy outcome | 0.034 | ||
| PD | 12 (3.3%) | 19 (5.3%) | |
| SD | 5 (1.4%) | 12 (3.3%) | |
| PR | 1 (0.3%) | 4 (1.1%) | |
| CR | 168 (46.5%) | 140 (38.8%) | |
| Gender | 1.000 | ||
| Female | 65 (12.9%) | 66 (13.1%) | |
| Male | 186 (37.1%) | 185 (36.9%) | |
| Race | 0.757 | ||
| Asian | 4 (1%) | 5 (1.3%) | |
| Black or African American | 13 (3.3%) | 17 (4.4%) | |
| White | 177 (45.5%) | 173 (44.5%) | |
| Age | 0.059 | ||
| ≤ 65 | 106 (21.5%) | 85 (17.2%) | |
| > 65 | 140 (28.4%) | 162 (32.9%) | |
| Smoker | 0.810 | ||
| No | 8 (1.6%) | 10 (2%) | |
| Yes | 237 (48.4%) | 235 (48%) | |
| OS event | 0.528 | ||
| Alive | 147 (29.3%) | 139 (27.7%) | |
| Dead | 104 (20.7%) | 112 (22.3%) | |
| DSS event | 1.000 | ||
| Alive | 180 (40%) | 181 (40.2%) | |
| Dead | 45 (10%) | 44 (9.8%) | |
| PFI event | 0.769 | ||
| Alive | 179 (35.7%) | 175 (34.9%) | |
| Dead | 72 (14.3%) | 76 (15.1%) |
Note: FAM83A, Family With Sequence Similarity 83 Member A; CR, Complete Response; PR, Partial Response; SD, Stable Disease; PD, Progressive Disease; OS, Overall Survival; DSS, Disease-Specific Survival; PFI, Progression-Free Interval; LUSC, Lung Squamous Cell Carcinoma.
Table 8.
Clinicopathological characteristics of patients with high and low MYCT1 expression in LUSC
| Characteristic | Low expression of MYCT1 | High expression of MYCT1 | P |
|---|---|---|---|
| T stage | 0.958 | ||
| T1 | 55 (11%) | 59 (11.8%) | |
| T2 | 147 (29.3%) | 147 (29.3%) | |
| T3 | 37 (7.4%) | 34 (6.8%) | |
| T4 | 12 (2.4%) | 11 (2.2%) | |
| N stage | 0.468 | ||
| N0 | 157 (31.7%) | 163 (32.9%) | |
| N1 | 69 (13.9%) | 62 (12.5%) | |
| N2 | 22 (4.4%) | 18 (3.6%) | |
| N3 | 1 (0.2%) | 4 (0.8%) | |
| M stage | 1 | ||
| M0 | 205 (48.9%) | 207 (49.4%) | |
| M1 | 3 (0.7%) | 4 (1%) | |
| Pathologic stage | 0.481 | ||
| Stage I | 116 (23.3%) | 129 (25.9%) | |
| Stage II | 89 (17.9%) | 73 (14.7%) | |
| Stage III | 42 (8.4%) | 42 (8.4%) | |
| Stage IV | 3 (0.6%) | 4 (0.8%) | |
| Primary therapy outcome | 0.036 | ||
| PD | 17 (4.7%) | 14 (3.9%) | |
| SD | 5 (1.4%) | 12 (3.3%) | |
| PR | 5 (1.4%) | 0 (0%) | |
| CR | 166 (46%) | 142 (39.3%) | |
| Gender | 0.067 | ||
| Female | 56 (11.2%) | 75 (14.9%) | |
| Male | 195 (38.8%) | 176 (35.1%) | |
| Race | 0.395 | ||
| Asian | 3 (0.8%) | 6 (1.5%) | |
| Black or African American | 13 (3.3%) | 17 (4.4%) | |
| White | 180 (46.3%) | 170 (43.7%) | |
| Age | 0.508 | ||
| ≤ 65 | 99 (20.1%) | 92 (18.7%) | |
| > 65 | 146 (29.6%) | 156 (31.6%) | |
| Smoker | 0.223 | ||
| No | 6 (1.2%) | 12 (2.4%) | |
| Yes | 240 (49%) | 232 (47.3%) | |
| OS event | 0.321 | ||
| Alive | 149 (29.7%) | 137 (27.3%) | |
| Dead | 102 (20.3%) | 114 (22.7%) | |
| DSS event | 0.605 | ||
| Alive | 188 (41.8%) | 173 (38.4%) | |
| Dead | 43 (9.6%) | 46 (10.2%) | |
| PFI event | 0.922 | ||
| Alive | 178 (35.5%) | 176 (35.1%) | |
| Dead | 73 (14.5%) | 75 (14.9%) |
Note: MYCT1, MYC Target 1; CR, Complete Response; PR, Partial Response; SD, Stable Disease; PD, Progressive Disease; OS, Overall Survival; DSS, Disease-Specific Survival; PFI, Progression-Free Interval; LUSC, Lung Squamous Cell Carcinoma.
The univariate Cox regression analysis revealed that age, clinical stage, T stage, and risk score all impacted the prognosis of LUSC (Figure 5F). Multivariate Cox regression further revealed that age and risk score were independent prognostic factors in LUSC (Figure 5G). We used age, clinical stage, T stage, and risk score to create a prognostic model nomogram (Figure 6), which showed that the clinical stage is the most important prognostic factor, followed by the risk score and T stage.
Figure 6.
A prognosis-related risk model nomogram in LUSC. Note: LUSC, Lung Squamous Cell Carcinoma.
The risk score is related to the infiltrating of immune cells in LUSC
Immune cell infiltration is an essential factor in cancer progression. Therefore, we calculated the levels of immune cell infiltration in TCGA LUSC tissues using the CIBERSORT algorithm. We divided the LUSC tissues into high- and low-risk groups using the risk scores and the median value of the risk score among all samples. We observed that the risk score was significantly correlated with the levels of LUSC immune infiltration consisting of naïve B cells, helper follicular T cells, neutrophils, and activated dendritic cells (Figure 7).
Figure 7.
Immune cell infiltration in lung squamous cell carcinoma of patients with high- and low-risk scores. A. Neutrophils; B. Naïve B cells; C. T cells helper follicular; D. Activated dendritic cells.
Discussion
Patients with LUSC experience high morbidity and mortality [3] and have a worse prognosis than patients with LAC. Therefore, it is imperative to find new treatment targets and methods to improve the prognosis of patients with LUSC. Many studies have reported the role of changes in gene expression during LUSC development, and modulation of gene expression is expected to improve the prognosis of LUSC patients [9,17,18]. For example, fat mass and obesity-associated protein (FTO) influences the prognosis of patients with LUSC and are the main factors causing abnormal mA modification in LUSC. FTO knockdown can effectively promote apoptosis and inhibit the proliferation of L78 and NCI-H520 cells, while overexpression of FTO encourages the malignant phenotype of CHLH-1 cells. Furthermore, FTO can enhance myeloid zinc finger 1 (MZF1) expression by reducing the levels of mA and the stability of MZF1 mRNA transcripts, thus exerting oncogenic functionalities [17]. As another example, lncRNA NNT-AS1 deletion inhibits LUSC cell migration and invasion and induces apoptosis. Overexpression of miR-22 impedes LUSC progression by targeting FOXM expression. NNT-AS1 directly regulates FOXM1 expression by binding to miR-22 in LUSC cells, thus affecting the growth and migration of LUSC cells [9].
The TCGA and GEO databases contain expression data for many cancer genes, miRNAs, lncRNAs, and other RNAs, and are used in many cancer studies [2,19,20]. For example, Takeda et al. used transcriptome data from the TCGA and GEO databases to show that high expression of insulin like growth factor 2 receptor (IGF2R), a tumor suppressor gene, is related to a poor prognosis in cervical cancer. Further research using various cellular models showed that interference with IGF2R expression in cervical cancer cells could induce apoptosis, reduce viability, and increase susceptibility to the anticancer drug cisplatin. IGF2R can also exert carcinogenic effects by transporting M6P-labeled cargo [20]. The transition from normal tissue to carcinoma in situ, or LUSC, takes time. Therefore, the GEO database can discover critical DEGs in LUSC progression, providing novel targets for LUSC diagnosis and treatment. In our study, DEGs that were overlapped with squamous metaplasia, carcinoma in situ, and LUSC were involved the development of the epidermis, cornification, epidermal cell differentiation, glycoside metabolic process, neutrophil chemotaxis, and migration, granulocyte chemotaxis and migration, leukocyte aggregation, migration involved in the inflammatory response, positive regulation of Nf-κB transcription factor activity, secondary metabolic processes, protein nitrosylation, and other roles. The appearance and development of LUSC may be induced by chronic stimulation and injury to columnar epithelial cells of the bronchial mucosa, loss of cilia, and squamous metaplasia basal cells. The overlapping DEGs are enriched with biological functions related to these possible causes of LUSC.
We found that the expression of FAM83A, MYCT1, ARL14, CARD14, CSTA, DKK4, DSG3, KLK8, and KRT6B were associated with LUSC prognosis and had diagnostic potential. Various studies have confirmed associations between ARL14, CARD14, DKK4, DSG3, KLK8, FAM83A, MYCT1, and CSTA expression levels and cancer progression [21-34]. For example, ARL14 expression levels are associated with a prognosis in patients with LAC. ARL14 silencing inhibits LAC cell proliferation, cell cycle progression, migration, and invasion ability. It also reduces radiation damage to cancer cells but does not affect normal lung cell proliferation. Interference with ARL14 expression can effectively block the extracellular signal-regulated kinase (ERK)/p38 signaling pathway [21]. The expression of CARD14 was higher in breast cancer samples than in normal breast tissues, and its inhibition can delay cell proliferation and migration, leading to cell cycle arrest in the G/S phase and promoting apoptosis [23]. DKK4 is related to cancer progression and negatively regulates the Wnt/β-catenin signaling pathway. The expression of DKK4 in A549/DTX cells increased compared to that in A549 cells. DKK4 overexpression increases the resistance of A549 cells to docetaxel, whereas interference with DKK4 expression can inhibit growth and reduce colony formation and invasion properties of A549/DTX cells. Furthermore, because it is associated with caspase-3 activation and BCL-2 down-regulation, DKK4 suppression enhances docetaxel’s apoptosis-promoting ability [26]. These findings suggest that the prognostic and diagnostic genes identified in our study play an important role in LUSC and may be able to predict the prognosis of patients with LUSC.
The expression of FAM83A in cervical cancer tissues is significantly increased compared to that in normal cervical tissues. This expression of FAM83A is related to the differentiation, stage of TNM, lymph node metastasis, and prognosis of cervical cancer. Interference with FAM83A expression can inhibit the proliferation, colony formation, and invasion of cervical cancer cells. In lung cancer, overexpression of FAM83A promotes the epithelial-mesenchymal transition (EMT) and Wnt signaling pathways [30] and is associated with poor patient survival. Interference with FAM83A expression can inhibit the proliferation, migration, and invasion of H1355 and A549 lung cancer cells and promotes the inactivation of the epidermal growth factor receptor/mitogen-activated protein kinase (MAPK)/choline kinase alpha signaling pathway [29]. MYCT1 inhibits the adhesion and migration of laryngeal cancer cells by regulating the expression of the COL6 target [31]. Down-regulation of CSTA is associated with high tumor grade, lymph node metastasis, and short OS in patients with oral squamous cell carcinoma (OSCC). CTSA overexpression can also inhibit OSCC cell migration and invasion in vitro [33]. CSTA is down-regulated in lung cancer cells compared to normal lung epithelial cells, and its high expression is correlated with low tumor grade. Stable CSTA transfection reduces cathepsin B activity; inhibits colony formation, migration, and invasion; and enhances gemcitabine-induced apoptosis. CSTA overexpression also reduces ERK, p38, and AKT activities and inhibits the ERK/MAPK pathway to block EMT [34]. Two of the genes that contribute to our risk score, FAM83A, and CSTA, have been reported in previous studies of lung cancer and LAC. However, MYCT1 has not been reported in the literature in lung cancer. Our results using TCGA and GEO data, as well as patient samples collected in our hospital, together with previous findings in the literature, indicate that FAM83A, CSTA, and MYCT1 are of significant biological importance in LUSC.
The immune microenvironment is closely related to cancer progression, and immunotherapy is an effective approach that promises to improve cancer outcomes. The immune microenvironment can influence the efficacy of immunotherapy [35]. We analyzed the relationship between our risk score and the LUSC immune microenvironment and found that the risk score was significantly correlated with the levels of LUSC immune infiltration consisting of naïve B cells, helper follicular T cells, neutrophils, and activated dendritic cells. However, the relationship between our risk model and the immune microenvironment must be further confirmed using basic research.
Our study based on PCR results and data from the TCGA and GEO databases had a large sample size and it is therefore expected to be highly reliable and provides novel molecular targets for diagnosing LUSC and determining LUSC prognosis. A nomogram is also required to determine the prognoses of patients, and more research is needed to confirm our findings. The roles of immune cell infiltration in LUSC progression and metastasis are worth exploring. Therefore, we explore the functions and signaling mechanisms of CSTA, FAM83A, and MYCT1 in the infiltration of certain immune cells based on the relationship between risk scores and immune cells identified in LUSC tissues. More LUSC tissue samples and clinical data will need to be collected to verify the clinical values of ARL14, CARD14, DKK4, DSG3, KLK8, FAM83A, MYCT1, and CSTA in LUSC. Additionally, a LUSC cell model must be constructed, and the impact of CSTA, FAM83A, and MYCT1 on the growth and migration of LUSC cells must be validated through future proliferation, apoptosis, and migration experiments. However, our results show that CSTA, FAM83A, and MYCT1 are abnormally expressed in LUSC tissues and significantly related to LUSC diagnosis and prognosis, suggesting that they may have potential as molecular targets for the treatment of LUSC. Furthermore, our risk model and the nomogram based on CSTA, FAM83A, and MYCT1 expression can potentially evaluate the prognosis of patients with LUSC.
Acknowledgements
This work was supported by the Wuhan Municipal Health Commission Foundation (No. wx21Q38).
Disclosure of conflict of interest
None.
Table S1
Table S2
Table S3
Figures S1-S6
References
- 1.Duruisseaux M, Esteller M. Lung cancer epigenetics: from knowledge to applications. Semin Cancer Biol. 2018;51:116–128. doi: 10.1016/j.semcancer.2017.09.005. [DOI] [PubMed] [Google Scholar]
- 2.Zhang YQ, Yuan Y, Zhang J, Lin CY, Guo JL, Liu HS, Guo Q. Evaluation of the roles and regulatory mechanisms of PD-1 target molecules in NSCLC progression. Ann Transl Med. 2021;9:1168. doi: 10.21037/atm-21-2963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ke D, Guo Q, Fan TY, Xiao X. Analysis of the role and regulation mechanism of hsa-mir-147b in lung squamous cell carcinoma based on the cancer genome atlas database. Cancer Biother Radiopharm. 2021;36:280–291. doi: 10.1089/cbr.2020.4187. [DOI] [PubMed] [Google Scholar]
- 4.Woodman C, Vundu G, George A, Wilson CM. Applications and strategies in nanodiagnosis and nanotherapy in lung cancer. Semin Cancer Biol. 2021;69:349–364. doi: 10.1016/j.semcancer.2020.02.009. [DOI] [PubMed] [Google Scholar]
- 5.Liu Y, Jia W, Li J, Zhu H, Yu J. Identification of survival-associated alternative splicing signatures in lung squamous cell carcinoma. Front Oncol. 2020;10:587343. doi: 10.3389/fonc.2020.587343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wu J, Xu C, Guan X, Ni D, Yang X, Yang Z, Wang M. Comprehensive analysis of tumor microenvironment and identification of an immune signature to predict the prognosis and immunotherapeutic response in lung squamous cell carcinoma. Ann Transl Med. 2021;9:569. doi: 10.21037/atm-21-463. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Liu HS, Guo Q, Yang H, Zeng M, Xu LQ, Zhang QX, Liu H, Guo JL, Zhang J. SPDL1 overexpression is associated with the 18F-FDG PET/CT metabolic parameters, prognosis, and progression of esophageal cancer. Front Genet. 2022;13:798020. doi: 10.3389/fgene.2022.798020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Haque S, Raina R, Afroze N, Hussain A, Alsulimani A, Singh V, Mishra BN, Kaul S, Kharwar RN. Microbial dysbiosis and epigenetics modulation in cancer development-a chemopreventive approach. Semin Cancer Biol. 2021;86:666–681. doi: 10.1016/j.semcancer.2021.06.024. [DOI] [PubMed] [Google Scholar]
- 9.Ma J, Qi G, Li L. LncRNA NNT-AS1 promotes lung squamous cell carcinoma progression by regulating the miR-22/FOXM1 axis. Cell Mol Biol Lett. 2020;25:34. doi: 10.1186/s11658-020-00227-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Li G, Guo X. LncRNA STARD13-AS blocks lung squamous carcinoma cells growth and movement by targeting miR-1248/C3A. Pulm Pharmacol Ther. 2020;64:101949. doi: 10.1016/j.pupt.2020.101949. [DOI] [PubMed] [Google Scholar]
- 11.Li H, Zhao Q, Tang Z. LncRNA RP11-116G8.5 promotes the progression of lung squamous cell carcinoma through sponging miR-3150b-3p/miR-6870-5p to upregulate PHF12/FOXP4. Pathol Res Pract. 2021;226:153566. doi: 10.1016/j.prp.2021.153566. [DOI] [PubMed] [Google Scholar]
- 12.Shi H, Zhong F, Yi X, Shi Z, Ou F, Xu Z, Zuo Y. Application of an autophagy-related gene prognostic risk model based on TCGA database in cervical cancer. Front Genet. 2020;11:616998. doi: 10.3389/fgene.2020.616998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Zhang Y, Tang M, Guo Q, Xu H, Yang Z, Li D. The value of erlotinib related target molecules in kidney renal cell carcinoma via bioinformatics analysis. Gene. 2022;816:146173. doi: 10.1016/j.gene.2021.146173. [DOI] [PubMed] [Google Scholar]
- 14.Sun L, Zhang Z, Yao Y, Li WY, Gu J. Analysis of expression differences of immune genes in non-small cell lung cancer based on TCGA and ImmPort data sets and the application of a prognostic model. Ann Transl Med. 2020;8:550. doi: 10.21037/atm.2020.04.38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Mascaux C, Angelova M, Vasaturo A, Beane J, Hijazi K, Anthoine G, Buttard B, Rothe F, Willard-Gallo K, Haller A, Ninane V, Burny A, Sculier JP, Spira A, Galon J. Immune evasion before tumour invasion in early lung squamous carcinogenesis. Nature. 2019;571:570–575. doi: 10.1038/s41586-019-1330-0. [DOI] [PubMed] [Google Scholar]
- 16.Wu Y, Wang J, Ge L, Hu Q. Significance of a PTEN mutational status-associated gene signature in the progression and prognosis of endometrial carcinoma. Oxid Med Cell Longev. 2022;2022:5130648. doi: 10.1155/2022/5130648. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Liu J, Ren D, Du Z, Wang H, Zhang H, Jin Y. MA demethylase FTO facilitates tumor progression in lung squamous cell carcinoma by regulating MZF1 expression. Biochem Biophys Res Commun. 2018;502:456–464. doi: 10.1016/j.bbrc.2018.05.175. [DOI] [PubMed] [Google Scholar]
- 18.Chen TJ, Zheng Q, Gao F, Yang T, Ren H, Li Y, Chen MW. MicroRNA-665 facilitates cell proliferation and represses apoptosis through modulating Wnt5a/β-Catenin and Caspase-3 signaling pathways by targeting TRIM8 in LUSC. Cancer Cell Int. 2021;21:215. doi: 10.1186/s12935-021-01913-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wu Z, Liu Z, Jiang X, Mi Z, Meng M, Wang H, Zhao J, Zheng B, Yuan Z. Depleting PTOV1 sensitizes non-small cell lung cancer cells to chemotherapy through attenuating cancer stem cell traits. J Exp Clin Cancer Res. 2019;38:341. doi: 10.1186/s13046-019-1349-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Takeda T, Komatsu M, Chiwaki F, Komatsuzaki R, Nakamura K, Tsuji K, Kobayashi Y, Tominaga E, Ono M, Banno K, Aoki D, Sasaki H. Upregulation of IGF2R evades lysosomal dysfunction-induced apoptosis of cervical cancer cells via transport of cathepsins. Cell Death Dis. 2019;10:876. doi: 10.1038/s41419-019-2117-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Guo F, Yuan D, Zhang J, Zhang H, Wang C, Zhu L, Zhang J, Pan Y, Shao C. Silencing of ARL14 gene induces lung adenocarcinoma cells to a dormant state. Front Cell Dev Biol. 2019;7:238. doi: 10.3389/fcell.2019.00238. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Lim JY, Kim SW, Kim B, Park SJ. Knockdown of CARD14 inhibits cell proliferation and migration in breast cancer cells. Anticancer Res. 2020;40:1953–1962. doi: 10.21873/anticanres.14150. [DOI] [PubMed] [Google Scholar]
- 23.He S, Shen J, Hu N, Xu X, Li J. DKK4 enhances resistance to chemotherapeutics 5-Fu and YN968D1 in colorectal cancer cells. Oncol Lett. 2017;13:587–592. doi: 10.3892/ol.2016.5461. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Yang X, Liu Y, Li W, Li A, Sun Q. DKK4-knockdown enhances chemosensitivity of A549/DTX cells to docetaxel. Acta Biochim Biophys Sin (Shanghai) 2017;49:899–906. doi: 10.1093/abbs/gmx086. [DOI] [PubMed] [Google Scholar]
- 25.Cai X, Yao Z, Li L, Huang J. Role of DKK4 in tumorigenesis and tumor progression. Int J Biol Sci. 2018;14:616–621. doi: 10.7150/ijbs.24329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Chao TB, Li CF, Lin CY, Tian YF, Chang IW, Sheu MJ, Lee YE, Chan TC, He HL. Prognostic significance of DSG3 in rectal adenocarcinoma treated with preoperative chemoradiotherapy. Future Oncol. 2016;12:1457–67. doi: 10.2217/fon-2016-0071. [DOI] [PubMed] [Google Scholar]
- 27.Chen YJ, Lee LY, Chao YK, Chang JT, Lu YC, Li HF, Chiu CC, Li YC, Li YL, Chiou JF, Cheng AJ. DSG3 facilitates cancer cell growth and invasion through the DSG3-plakoglobin-TCF/LEF-Myc/cyclin D1/MMP signaling pathway. PLoS One. 2013;8:e64088. doi: 10.1371/journal.pone.0064088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Zhang M, Luo C, Lin D, Cui K, Chen Z, Liu J. Human tissue kallikrein 1 is downregulated in elderly human prostates and possesses potential in vitro antioxidative and antifibrotic effects in rodent prostates. Oxid Med Cell Longev. 2021;2021:8877540. doi: 10.1155/2021/8877540. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Liu PJ, Chen YH, Tsai KW, Yeah HY, Yeh CY, Tu YT, Yang CY. Involvement of MicroRNA-1-FAM83A axis dysfunction in the growth and motility of lung cancer cells. Int J Mol Sci. 2020;21:8833. doi: 10.3390/ijms21228833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Lan C, Liu CC, Nie XC, Lei L, Xiao ZX, Li MX, Tang XN, Jia MY, Xu HT. FAM83A promotes the proliferative and invasive abilities of cervical cancer cells via epithelial-mesenchymal transition and the wnt signaling pathway. J Cancer. 2021;12:6320–6329. doi: 10.7150/jca.62563. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Wang PP, Ding SY, Sun YY, Li YH, Fu WN. MYCT1 inhibits the adhesion and migration of laryngeal cancer cells potentially through repressing collagen VI. Front Oncol. 2020;10:564733. doi: 10.3389/fonc.2020.564733. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Kabir AU, Subramanian M, Lee DH, Wang X, Krchma K, Wu J, Naismith T, Halabi CM, Kim JY, Pulous FE, Petrich BG, Kim S, Park HC, Hanson PI, Pan H, Wickline SA, Fremont DH, Park C, Choi K. Dual role of endothelial Myct in tumor angiogenesis and tumor immunity. Sci Transl Med. 2021;13:eabb6731. doi: 10.1126/scitranslmed.abb6731. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wang Y, Wang L, Li X, Qu X, Han N, Ruan M, Zhang C. Decreased CSTA expression promotes lymphatic metastasis and predicts poor survival in oral squamous cell carcinoma. Arch Oral Biol. 2021;126:105116. doi: 10.1016/j.archoralbio.2021.105116. [DOI] [PubMed] [Google Scholar]
- 34.Ma Y, Chen Y, Li Y, Grün K, Berndt A, Zhou Z, Petersen I. Cystatin A suppresses tumor cell growth through inhibiting epithelial to mesenchymal transition in human lung cancer. Oncotarget. 2018;9:14084–14098. doi: 10.18632/oncotarget.23505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Yu L, Ding Y, Wan T, Deng T, Huang H, Liu J. Significance of CD47 and its association with tumor immune microenvironment heterogeneity in ovarian cancer. Front Immunol. 2021;12:768115. doi: 10.3389/fimmu.2021.768115. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.