Comprehensive analysis of KRT18 as a novel prognostic biomarker and potential target in lung adenocarcinoma

Abstract

Background: The expression characteristics of Keratin 18 (KRT18) in lung adenocarcinoma (LUAD) remain incompletely elucidated. This study aims to investigate the expression pattern of KRT18 in LUAD and its prognostic significance. Methods: We analyzed the expression status of KRT18 in LUAD and its association with prognosis. Utilizing the UALCAN and STRING databases, we systematically evaluated the clinical phenotypic parameters of KRT18 and its protein–protein interaction network. Through enrichment analysis, we clarified its biological functions and associated signaling pathways, and simultaneously deciphered the association patterns between the tumor immune infiltration landscape and immune checkpoint molecules. Results: High expression of KRT18 was associated with poor prognosis in LUAD patients and was closely correlated with tumor stage and pathological stage. Functional enrichment analysis revealed that KRT18 was significantly enriched in epithelial cell differentiation and intermediate filament pathways. Immune infiltration analysis showed that the expression of KRT18 was associated with tumor immune cell infiltration and immune checkpoints. Immunohistochemistry further confirmed the high expression of KRT18 in LUAD tissues. Conclusion: Therefore, KRT18 may serve as a biomarker for the prognosis and diagnosis of LUAD and also represents a potential target for immunotherapy.

Graphic abstract

Flow diagram of the study

Supplementary Information

The online version contains supplementary material available at 10.1007/s10238-025-01855-0.

Introduction

Lung cancer is the leading cause of cancer-related mortality worldwide. Lung cancer emerged as the most prevalent malignancy worldwide in 2022, with 2.48 million new cases accounting for 12.4% of all cancers globally, surpassing breast cancer to claim the top spot. It also remained the leading cause of cancer-related deaths, responsible for 1.8 million fatalities (18.7% of total cancer deaths), a burden nearly double that of colorectal cancer, the second most deadly malignancy [1]. In non-small cell lung cancer (NSCLC), the predominant subtypes are LUAD and lung squamous cell carcinoma (LUSC) [2–4], each characterized by distinct pathological and biological features. LUAD comprises approximately 40% of all lung cancer cases and ranks among the most aggressive and swiftly fatal forms of the disease, often with an overall survival time of less than five years [5–7].The five-year survival rate ranges from 10 to 20% [8–10]. Over the past decade, treatment strategies for LUAD have evolved, with significant advances in targeted therapies and immunotherapy [11], particularly for advanced cases. However, due to the complex and variable nature of lung cancer and the emergence of drug resistance, only a few patients respond effectively to these treatments [12]. The research landscape of non-small cell lung cancer (NSCLC) is continuously expanding and deepening. In recent years, studies have revealed that mutations in the epidermal growth factor receptor (EGFR), the anti-tumor activities of natural compounds, and molecules associated with epithelial-mesenchymal transition (EMT) all play crucial roles in the development, progression, and treatment of NSCLC [2–4]. Based on studies decoding key gene mutations and molecular pathways, the clinical applications of treatment approaches, including repeated mutations in epidermal growth factor receptor (EGFR) kinase [3], anaplastic lymphoma kinase fusion, and mutations in genes, such as Kirsten rat sarcoma viral oncogene, v-raf murine sarcoma viral oncogene homolog B1, and human epidermal growth factor receptor 2, have significantly changed the treatment approaches for patients with lung cancer [2, 4, 13]. Recently, inhibitors targeting immune checkpoints, such as programmed cell death 1 (PD-1) and programmed cell death ligand 1 (PD-L1), have become pivotal in LUAD treatment as they deactivate the tumor immune escape mechanism and activate the patient’s immune system. Clinical trials have shown that PD-1 and PD-L1 inhibitors significantly enhance the overall and progression-free survival in patients with advanced LUAD, particularly among those with high PD-L1 expression [11]. However, effective indicators for predicting therapy response and monitoring disease progression during immune checkpoint inhibitor therapy are lacking. Therefore, extensive research on biomarkers that accurately identify sensitivity and specificity is urgently required [14, 15].

Keratin 18 (KRT18), or cytokeratin 18 (CK18), is a crucial intermediate filament protein in simple epithelial cells [16]. The protein maintains cytoskeleton integrity and stability, cellular morphology, intercellular adhesion, and cell signaling, thereby influencing various cellular processes, such as cell cycle progression, apoptosis, and mitosis [17, 18]. Initially thought to be present in epithelial and endothelial cells of the gastrointestinal and respiratory tracts [19], KRT18, in the last decade, has been shown to be upregulated in various human tumor tissues, including lung, invasive breast, esophageal, gastric, hepatocellular, bladder, ovarian, oral squamous, pancreatic, and prostate cancers. KRT18 is aberrantly expressed in breast, gastric, colorectal, esophageal squamous cell, and pancreatic cancers, and correlates with clinical progression and poor prognosis [20–27]. High KRT18 expression is associated with advanced tumor grade and stage in renal cell carcinoma and oral cancer [28, 29]. Upregulation of KRT18 expression in breast cancer significantly correlates with advanced clinical and adverse outcomes and promotes cancer cell migration and invasiveness [26, 27, 30]. The upregulated KRT18 expression in gastric cancer promotes tumor cell proliferation, invasion, and migration, inhibits apoptosis, and may serve as a potential target for its treatment [24, 31]. In lung cancer, KRT18 expression is closely associated with clinical staging, lymph node metastasis, number of pathologically positive lymph nodes, recurrence, and metastasis, and it affects the development and progression of NSCLC [32, 33]. While the fundamental properties of KRT18 as an epithelial cell marker have been established, no studies have systematically elucidated the association between KRT18 and the histomorphological features as well as aberrant expression patterns in lung adenocarcinoma, nor its clinical prognostic value. The aim of this study was to investigate the relationship between KRT18 and the prognosis and immune infiltration of LUAD, to observe the histomorphological results of LUAD using HE experiments, and to verify the expression of KRT18 in LUAD using immunohistochemistry experiments to provide an important molecular basis for the early non-invasive diagnosis and immunotherapy of LUAD.

Materials and methods

Venndiagram intersection analysis

Analyzing RNAseq data from the TCGA (https://portal.gdc.cancer.gov/) database (GDC Data Release 38.0) [34], DESeq2 was utilized to identify differentially expressed genes in TCGA-LUAD. With strict thresholds of log2 fold change (logFC) > 1, adjusted P value (padj) < 0.05, and raw P.value < 0.001, we discovered 3296 genes with significant expression differences. Additionally, employing univariate Cox regression from the R packages [35] survival and survminer [36], we pinpointed 1972 prognostic genes linked to Lung Adenocarcinoma, using the criteria of HR (Hazard Ratio) > 1 and P.value < 0.05.Both univariate Cox regression analysis and screening of differentially expressed genes were performed with FDR correction using the Benjamini–Hochberg method. We have procured a comprehensive set of cancer-related genes from the Cancer Genetics Web, totaling 2,168 genes. Additionally, we have obtained a collection of genes associated with Epithelial-Mesenchymal Transition (EMT) from the dbEMT 2.0 database, which comprises 1,184 genes. These gene sets exhibit interactions with one another. To further elucidate the potential role of keratin genes in lung adenocarcinoma, we performed a Venn diagram analysis intersecting the 17 KRT family genes with the aforementioned gene sets. Our analysis involves delineating the unique and shared components among each group. Furthermore, we employ the ggplot2 package [37] and the VennDiagram package to visualize the results of this analysis.

Analysis of DEmRNA and gene co-expression analysis

Based on the differential analysis of TCGA-LUAD data, null values were removed and only mRNAs were retained, a total of 18,600 genes were obtained, and the parameters LogFC threshold 1 and p threshold 0.05 were set to screen the genes, and a total of 3,486 upregulated genes and 1,944 down-regulated genes were obtained and the results of the differential analysis were visualized using the R package ggplot2 [37] and a heat map of target gene with its co-expressed mRNA.

Data acquisition

In TCGA-LUAD, the clinical and RNAseq data of 581 LUAD patients were obtained, including 522 LUAD tissues in the study, as well as 59 paracarcinoma tissues, along with T-stage, age, N-stage, gender, Mstage, smoking and lymphatic invasion.

Localization and expression profile analysis of KRT18

KRT18 expression was analyzed using TCGA data-base (https://portal.gdc.cancer.gov), and RNA sequencing (RNAseq) data from the TCGA-LUAD projects were converted to transcripts per million (TPM) values. We specifically extracted paired samples of tumor and adjacent non-tumor (paratumor) tissue to ensure that the corresponding values were paired. We compared the expression profiles of the KRT18 gene in lung adenocarcinoma (LUAD) and tissues with those in tumor-paraneoplastic tissues to determine the differential expression patterns. The KRT18 expression levels were analyzed after logarithmic transformation [log2 (value + 1)]. Differential KRT18 expression in LUAD and tissues as well as normal lung tissues was determined by using a Wilcoxon rank-sum test for analysis. All survival analyses were performed in R 4.2.1. First, proportional hazards (PH) assumption tests were performed for each covariate individually and for the model as a whole. If the PH assumption was violated (P < 0.05), the variable was either stratified by time or incorporated into the model as a time-varying covariate. Survival regression was then fitted with Cox proportional hazards models (survival::coxph), employing the Efron method for handling ties and right-censored data. Hazard ratios (HR) with 95% confidence intervals (CI) and Wald P values were reported. Multiple-testing correction across genes used the Benjamini–Hochberg false-discovery rate (FDR). Kaplan–Meier curves were plotted with survminer::ggsurvplot, all graphics were finalized in ggplot2 [36]. Data were visualized using the R software “ggplot2” and the “stats” package. “pROC” package was used to analyze ROC [38]. Immunofluorescence staining images from the HPA (The Human Protein Atlas, https://www.proteinatlas.org/) dataset demonstrate the expression and localization of KRT18 in two human cancer cell lines (A-431 and U-251MG). For the A-431 and U-251MG cell lines used in immunofluorescence experiments (with fluorescent images sourced from the Human Protein Atlas), the standard passage ratios recommended by ATCC are 1:3 to 1:8 for A-431 and 1:2 to 1:3 for U-251 MG, respectively. The basal medium used for A-431 is Dulbecco’s Modified Eagle Medium (Catalog No. 30-2002), and the complete growth medium is prepared by supplementing the basal medium with 10% fetal bovine serum. For U-251 MG, the routinely used medium consists of 89% MEM medium (Gibco, Catalog No. 11095080) and 10% fetal bovine serum (FBS, Gibco). In addition, details of the antibody used for both A-431 and U-251 MG in this result are as follows: Atlas Antibodies (Cat# HPA001605; RRID: AB_2666381), including its concentration (0.0325 mg/ml) and purification method (affinity purification using PrEST antigen as the affinity ligand).

Relationship between KRT18 expression and clinicopathological features in LUAD

The association between the KRT18 RNA transcription levels of LUAD from the TCGA database and clinical pathological parameters (TNM staging, lymph node metastasis, smoking, gender, age) were analyzed using the University of Alabama at Bir-mingham Cancer Data analysis Portal (https://ualcan.path.uab.edu/) web resource.

Immune‑related analysis of KRT18

Spearman analysis was used to assess the correlation between KRT 18 expression and immune cells as well as immune checkpoint-programmed cell death protein 1 (PD-1)-PDCD 1, cytotoxic T-lymphocyte-associated protein 4 (CTLA 4) and programmed cell death ligand 1 (PD-L1). We visualized the correlations using the “ggplot 2” package and the “pheatmap” package [39]. We evaluated the enrichment of immune infiltrating cells in LUAD patients with high and low KRT 18 expression using the Wilcoxon rank-sum test.

Biological function analysis of KRT18 in LUAD

RNAseq data of LUAD from the TCGA database using a limiting threshold of padj < 0.05 and a fold change of logFC > 1 was used to screen differentially expressed genes related to KRT18. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathway enrichment analyses were performed using the “clusterProfiler” package [40] in R software. The Search Tool for the Retrieval of Interacting Genes (STRING) online database (https://string-db.org/) was used to construct an interaction network of the top 10 hub genes according to the relationship of KRT18 in humans.

Immunohistochemistry

This study was approved by the Ethics Committee of the First Affiliated Hospital of Hebei North University (approval No.: K2024132). Paraffin-embedded tissue samples of LUAD were obtained from the Department of Pathology, First Affiliated Hospital of Hebei North University, in accordance with the following inclusion criteria: (1) Tissues were pathologically diagnosed as lung adenocarcinoma; histological examination of each tissue was verified by two pathologists; clinical staging was determined based on the 7^th edition of the American Joint Committee on Cancer (AJCC) TNM staging system. (2) The specimens included 6 cases of paracancerous normal lung tissues (defined as normal lung tissues located 2 cm from the tumor edge) and 6 cases of lung adenocarcinoma tissues. All patients underwent surgery or biopsy at the First Affiliated Hospital of Hebei North University and had not received chemotherapy prior to surgery [Supporting information: see Supplementary File 1]. Briefly, after fixation in 4% paraformaldehyde, the tissues were embedded in wax blocks and cut into 4-μm sections. After dewaxing, rehydration, and antigen retrieval (performed by heating 1 × sodium citrate antigen retrieval solution to boiling, completely immersing the sections, then heating at low temperature using the microwave defrost mode for 20 min; after which the sections were allowed to cool naturally to room temperature for 1 h and washed with PBS), the sections were incubated with the primary antibody (1:150 dilution, catalog No.: ET1603-8) and stained using a DAB chromogenic kit. The sections were then counterstained with hematoxylin. Additionally, hematoxylin–eosin staining was performed for histopathological evaluation. As previously described, scores were determined based on staining intensity and the proportion of labeled tumor cells. (All reagents mentioned above were purchased from Hangzhou Huaan Biotechnology Co., Ltd, Three independent repeated experiments).

Statistical analysis

Student’s t-test was used to compare the differences between the two groups, and data are reported as mean values ± SD. The p-value < 0.05. Statistical analyses were performed using GraphPad Prism version 9.

Results

Screening of differential prognostic proteins for lung adenocarcinoma in TCGA data

We downloaded 2,168 cancer-related genes from in Cancer Genetics Web site, 1,184 EMT-related genes from dbEMT 2.0 site, 3,296 LUAD differential genes screened from TCGA database with 1,972 prognostic genes, and finally of the 54 KRT genes in that literature [41], 37 very low expression (TMP < 0.1) were excluded from subsequent analyses, in the remaining 17 KRT genes will be used as analyses. Keratin gene family, cancer-related genes, EMT-related genes, and LUAD differential genes from the TCGA database in the literature were intersected with prognostic genes by taking a Wayne diagram, which showed that a total of 2 related genes were intersected, and the results showed that KRT18 and KRT8 were overlapping target genes (Fig. 1a). The number of overlapping genes in each dataset is provided in detail in Supplementary File 2. We obtained 3486 upregulation and 1944 down-regulation after division, volcano distribution map (Fig. 1b). By comprehensive comparison, KRT18 was finally selected as the target gene. Heat map of KRT18 and its co-expressed mRNAs (Fig. 1c).

Fig. 1.

Screening of relevant differential prognostic proteins. (a) Five genes, TCGA-LUAD-PGs (1972), Cancer Genetics Web (2168), TCGA-LUAD-DEGs (3296), dbEMT 2.0, and (1184) Keratin gene (17) were sets were taken for intersection. (b) Volcano maps of differential genes based on TCGA-LUAD data analysis were visualized. (c) Clinically relevant co-expression maps of the associated genes

KRT18 is highly expressed in lung adenocarcinoma

The Timer online database was used to analyze the extensive expression of KRT18 in human pan-cancer tissues, which showed that KRT18 was widely expressed in Bladder uroepithelial carcinoma (BLCA), Breast invasive carcinoma (BRCA), Cervical squamous cell carcinoma and adenocarcinoma (CESC), Cholangiocarcinoma (CHOL), Colorectal carcinoma (COAD), Esophageal carcinoma (ESCA), Head and neck squamous cell carcinoma (HNSC), Kidney clear cell carcinoma (KIRC), Kidney papillary cell carcinoma (KIRP), Hepatocellular carcinoma (LIHC), Prostate cancer (PRAD), pancreatic adenocarcinoma (PAAD), gastric adenocarcinoma (STAD), thyroid adenocarcinoma (THCA), endometrial adenocarcinoma (UCEC) and its expression levels were all higher than normal tissues, and the expression level of KRT18 in LUAD tissues was significantly higher than that in normal tissues (Fig. 2a). We analyzed the RNAseq datasets in the TCGA database and in the TCGA-LUAD project for a total of 598 LUAD patients, including 522 LUAD tissue samples, and 59 paracancerous tissue samples of LUAD. Compared with normal lung tissues, LUAD had higher levels of KRT18 expression.KRT18 can be used as a potential diagnostic biomarker (Fig. 2b, c), and Kaplan–Meier survival curve analysis showed that LUAD patients with high KRT18 expression had a low overall survival rate, which was of important prognostic significance (Fig. 2e). ROC curve analysis revealed an AUC value of 0.822 with a 95% confidence interval (CI) of 0.766–0.866, indicating that KRT18 possesses significant diagnostic value for lung adenocarcinoma (Fig. 2d). Based on data from the human genome (GRCh38.p14), the KRT18 gene was found to be localized on chromosome 12 (Fig. 3a). The Human Protein Atlas (HPA) database revealed that KRT18 is primarily distributed in the cytoplasm of cells (Fig. 3b). Additionally, KRT18 expression was detected in both human epidermoid carcinoma cells (A-431) and human malignant glioblastoma cells (U-251MG) (Fig. 3c, d), suggesting that it may also exert functional roles in other types of tumor cells. HE staining results revealed that the paracancerous normal lung tissue (Normal) exhibited intact alveolar structures, with thin and uniformly—sized alveolar walls and regular cell morphology (Fig. 3e). Immunohistochemical results showed that KRT18 was weakly expressed in paracancerous normal lung tissues (Normal), with faint and sparsely distributed staining signals, and the alveolar structures were clear (Fig. 3f). In contrast, the lung adenocarcinoma tumor tissue (tumor) showed remarkable pathological features. The tumor cells were arranged in a disordered pattern, with obvious cellular atypia, a disturbed nuclear—cytoplasmic ratio, an increased number of mitotic figures, along with structural disorganization and invasive growth (Fig. 3g). In lung adenocarcinoma tumor tissues (tumor), KRT18 was significantly highly expressed, with strong and widely distributed staining signals. The tumor cells were arranged in a disordered manner, and a large number of brown—positive staining areas were observed (Fig. 3h). The locally magnified regions (inset) clearly revealed the cytological differences in KRT18 expression between the two types of tissues, which could intuitively verify the high expression of KRT18 in lung adenocarcinoma tissues.

Fig. 2.

A bad prognosis was linked to upregulation of KRT18 in LUAD. (a) The expression of KRT18 in pan-cancer from Timer database. (b–c) KRT18 expression in LUAD from TCGA database. (d) The ROC curve of KRT18. (e) Overall survival curve of KRT18 from TCGA database. **, p < 0.01, ***, p < 0.001

Fig. 3.

(a) Circular diagram showing the chromosomal localization of human KRT18; (b) Cellular localization of KRT18 from the HPA database; (c, d) Subcellular localization of KRT18 in A-431 and U-251MG cells based on the HPA database; (e, g) Hematoxylin–eosin (HE) staining images of peritumoral tissues and LUAD tissues; (f, h) Immunohistochemical images of peritumoral lung tissues and LUAD

Relationship between KRT18 transcript expression and clinicopathological features in LUAD

Analysis of the TCGA-LUAD dataset showed that KRT18 was transcribed at higher levels in LUAD tissues than in normal tissues. In LUAD patients, KRT18 transcript levels were slightly higher in male patients than in females (Fig. S1a), transcript levels of the KRT18 gene were higher in the smokers and ex-smokers groups than in the non-smokers group (Fig. S1b), and the levels of KRT18 transcript were higher than those of normal lungs in LUAD patients of the age groups 21–40 years/41–60 years/61–80 years/81–100 years tissues, with the highest KRT18 transcript levels in the 21–40/41–60 age groups (Fig. S1c). KRT18 transcript levels were higher in LUAD clinicopathological stages and lymph node metastases than in normal lung tissues, with the highest levels in patients with pathology stage III and lymph node metastases stage N2 (Fig. S1d, e). In addition, mutations in the TP53 gene were also associated with higher levels of KRT18 transcript expression (Fig. S1f).

Relationship between KRT18 expression and clinicopathological parameters

KRT18 expression was associated with clinicopathological factors, and the chi-square test showed significant correlations between KRT18 and T-stage (P = 0.012), N-stage (P  = 0.018), and pathological stage (P = 0.027) of LUAD patients. And there were no significant correlations with gender, age, whether or not they smoked, and the number of pack-years of smoking, as shown in Table 1.

Table 1.

KRT18 expression correlates with clinicopathological features (chi-square test)

Characteristics	Low expression of KRT18	High expression of KRT18	P value
n	269	270
Pathologic T-stage, n (%)			0.012
T1	102 (19%)	74 (13.8%)
T2	140 (26.1%)	152 (28.4%)
T3	16 (3%)	33 (6.2%)
T4	9 (1.7%)	10 (1.9%)
Pathologic N-stage, n (%)			0.018
N0	181 (34.6%)	169 (32.3%)
N1	49 (9.4%)	48 (9.2%)
N2	25 (4.8%)	49 (9.4%)
N3	2 (0.4%)	0 (0%)
Pathologic M stage, n (%)			0.626
M0	179 (45.9%)	186 (47.7%)
M1	11 (2.8%)	14 (3.6%)
Pathologic stage, n (%)			0.027
Stage I	162 (30.5%)	134 (25.2%)
Stage II	58 (10.9%)	67 (12.6%)
Stage III	31 (5.8%)	53 (10%)
Stage IV	12 (2.3%)	14 (2.6%)
Gender, n (%)			0.180
Female	152 (28.2%)	137 (25.4%)
Male	117 (21.7%)	133 (24.7%)
Age, n (%)			0.725
<  = 65	126 (24.2%)	131 (25.2%)
> 65	133 (25.6%)	130 (25%)
Smoker, n (%)			0.113
No	45 (8.6%)	32 (6.1%)
Yes	218 (41.5%)	230 (43.8%)
Number pack years smoked, n (%)			0.652
< 40	87 (23.6%)	101 (27.4%)
>  = 40	88 (23.8%)	93 (25.2%)

All content highlighted in bold in the tables indicates that the corresponding p-values meet the criterion of statistical significance (i.e., p<0.05).

As shown in Table 2, to investigate the effects of KRT18 expression and clinicopathological parameters on survival, we used univariate and multivariate Cox regression analyses. Among the variables with p < 0.05 in the univariate Cox regression model, KRT18 was statistically significantly associated with T-stage, N-stage, M-stage, and pathological stage in LUAD patients. In the present study, the After univariate and multivariate analyses of clinicopathological parameters in patients with lung adenocarcinoma (LUAD), we found that patients with pathological T-stage (size or extent of the primary tumor) T2 (p = 0.023), T3&T4 (p < 0.001), pathological N-stage (status of regional lymph node involvement) N1 (p < 0.001) and N2 (p < 0.001), pathological M-stage (distant metastases) M1 (p = 0.005), and overall cancer staging of stages III & IV (p < 0.001) showed a significantly higher risk of death in univariate analysis. However, in multivariate analyses, only T3 & T4 in T-stage, and pathological N-stage N1 and N2 were significantly associated with a higher risk of death. Gender, age and pack-years of smoking did not show a significant effect in univariate analysis.

Table 2.

Univariate and multivariate analyses of clinicopathological parameters in patients with LUAD

Characteristics	Total(N)	Univariate analysis		Multivariate analysis
		Hazard ratio (95% CI)	P value	Hazard ratio (95% CI)		P value
Pathologic T-stage	527
T1	176	Reference		Reference
T2	285	1.507 (1.059– 2.146)	0.023	1.534 (0.974– 2.415)		0.065
T3&T4	66	3.095 (1.967– 4.868)	< 0.001	2.688 (1.466– 4.928)		0.001
Pathologic N-stage	514
N0	345	Reference		Reference
N1	96	2.293 (1.632– 3.221)	< 0.001	1.987 (1.339– 2.949)		< 0.001
N2	71	3.133 (2.154– 4.559)	< 0.001	2.140 (1.012– 4.526)		0.047
N3	2	0.000 (0.000– Inf)	0.994	0.000 (0.000– Inf)		0.993
Pathologic M stage	381
M0	356	Reference		Reference
M1	25	2.176 (1.272–3.722)	0.005	1.196 (0.557–2.565)		0.646
Pathologic stage	522
Stage I&Stage II	415	Reference		Reference
Stage III&Stage IV	107	2.710 (1.994–3.685)	< 0.001	1.460 (0.679–3.139)		0.333
Gender	530
Female	283	Reference
Male	247	1.087 (0.816–1.448)	0.569
Age	520
<  = 65	257	Reference
> 65	263	1.216 (0.910–1.625)	0.186
Number pack years smoked	363
< 40	183	Reference
>  = 40	180	1.073 (0.753–1.528)	0.697

All content highlighted in bold in the tables indicates that the corresponding p-values meet the criterion of statistical significance (i.e., p<0.05).

Biological function analysis of KRT18 in LUAD

For the TCGA-LUAD RNAseq data set, GO enrichment analysis indicated that KRT18 was involved in epidermal cell differentiation and development, played a crucial role in the intermediate filament cytoskeleton, and was associated with serine-type endopeptidase activity (Fig. 4a). KEGG enrichment analysis showed that in LUAD cells, KRT18 was associated with pathways including Neuroactive ligand-receptor interaction, Gastric cancer, and Estrogen signaling pathway (Fig. 4b). Using STRING online databases, protein interaction networks involving the human KRT18 gene included proteins, such as KRT8, KRT19, and EGFR, suggesting a close association between KRT8, KRT19, EGFR, and KRT18 (Fig. 4c).

Fig. 4.

(a) Gene ontology (GO) functional enrichment of keratin 18 (KRT18) gene in lung adenocarcinoma (LUAD); (b) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway of KRT18 in LUAD; (c) Top 10 hub genes of human KRT18 gene interaction network from Search Tool for the Retrieval of Interacting Genes (STRING) line database

Correlation between KRT18 gene transcription and tumor immune microenvironment in LUAD

Using the TCGA database, specifically the TCGA-LUAD dataset, we evaluated the correlation between the transcriptional levels of the KRT18 gene and the relative abundance of 24 immune cell types in LUAD. Our results demonstrated that the transcriptional levels of the KRT18 gene were significantly correlated with the enrichment abundance of 15 immune cell types (Fig. 5a). Among the eight immune cell types that exhibited the strongest correlation with KRT18 transcriptional levels, Th2 cells (R = 0.179, p < 0.001) and NK CD56bright cells (R = 0.179, p < 0.001) showed significant positive correlations with KRT18 gene expression (Fig. 5b, c). In contrast, T helper cells (R = − 0.268, p < 0.001), central memory T cells (Tcm, R = − 0.194, p < 0.001), B cells (R = − 0.249, p < 0.001), T cells (R = − 0.225, p < 0.001), macrophages (R = − 0.205, p < 0.001), and Th1 cells (R = − 0.202, p < 0.001) exhibited significant negative correlations with KRT18 gene expression (Fig. 5d–i).

Fig. 5.

(a) keratin 18 (KRT18) expression correlates with abundance of 24 immune cells; (b) KRT18 expression correlates with T helper (Th)2 cell enrichment; (c) KRT18 expression correlates with NK CD56bright cell enrichment; (d) KRT18 expression correlates with Th cell enrichment; (e) KRT18 expression correlates with central memory T cells (Tcm) enrichment; (f) KRT18 expression in relation to B cell enrichment; (g) KRT18 expression in relation to T cell enrichment; (h) KRT18 expression in relation to macrophage enrichment; (i) KRT18 expression in relation to Th1 enrichment. **p < 0.01, ***p < 0.001

Based on the results of immune infiltration analysis, we identified a correlation between KRT18 and the eight most relevant immune cell types. To validate this hypothesis, we performed a differential analysis between high and low expression groups for these eight immune cell types using the Wilcoxon rank-sum test. Wilcoxon rank-sum test was used to detect the enrichment of immune cells in KRT18 high and low expression groups. The results showed that KRT18 high-expressing LUAD showed a high enrichment of Th2 and NK CD56bright cells compared to KRT18 low-expressing LUAD (Fig. 6a, b). T helper cells, Tcm, B cells, T cells, macrophages and Th1 were less enriched (Fig. 6c–h). The results are consistent with the aforementioned immune infiltration analysis, confirming that KRT18 is closely associated with these eight immune cell types.

Fig. 6.

Differences in immune cell enrichment in lung adenocarcinoma (LUAD) with high versus low keratin 18 (KRT18) expression. (a) T helper (Th)2 cells, (b) NK CD56bright cells had significantly higher enrichment scores in the high KRT18 expression group than in the low KRT18 expression group. (c) Th cell; (d) central memory T cells (Tcm); (e) B cell; (f) T cell; (g) macrophage; (h) Th1 enrichment scores were lower in the low expression group than in the high expression group. **p < 0.01, ***p < 0.001

The expression of KRT18 in LUAD was negatively correlated with that of CD274 (R = − 0.144, p < 0.001) (Fig. 7a), was not significantly correlated with the expression of PDCD1 (R = − 0.016, p = 0.719) (Fig. 7b), negatively correlated with the expression of Cytotoxic T Lymphocyte-Associated Antigen-4(CTLA4) (R = − 0.223, p < 0.001) (Fig. 7c).

Fig. 7.

Tumor immune checkpoints and KRT18 expression. Keratin 18 (KRT18) expression significantly correlates with the presence of programmed cell death ligand 1 (PD-L1) (CD274), CTLA4 (a) (c); not significant with programmed cell death 1 (PD-1) (PDCD1) (b)

Discussion

Recently, molecular-targeted therapy and immunotherapy have significantly improved treatment outcomes for patients with LUAD [42–45]. However, the overall response rate is 20%, with only a small percentage of patients deriving benefits [46]. The effective application of cancer immunotherapy and enhancement of therapeutic efficacy still encounter challenges, and the research and exploration of relevant biomarkers for guiding immunotherapy remains a key focus [47].

KRT18 is a type I intermediate filament protein that extends from the nuclear surface to the cell membrane and is involved in multiple cellular processes, including cell proliferation, cell cycle regulation, apoptosis, motility, and cell signaling [48]. Over the past decade, studies have confirmed that KRT18 is upregulated in various tumors [20]. In this study, bioinformatics analysis was performed on its expression, prognostic value, related pathways, and immune characteristics, providing new insights into the mechanism and treatment of LUAD. The results showed that KRT18 is highly expressed in most cancer tissues, and its upregulation in LUAD was verified by immunohistochemistry. Moreover, high KRT18 expression is associated with poor prognosis, as well as tumor grade and stage in LUAD, suggesting that it could serve as a potential diagnostic and prognostic biomarker. This study clarifies the relationship between KRT18 and the prognosis and immune infiltration of LUAD, thereby providing a molecular basis for early non-invasive diagnosis and immunotherapy of LUAD.

Enrichment analysis revealed that KRT18 is closely associated with processes such as epidermal cell differentiation and development, intermediate filament cytoskeleton, and neuroactive ligand-receptor interaction. Protein interaction analysis indicated that KRT18 is closely related to proteins including KRT8, KRT19, and EGFR. KRT8 (type II), KRT18/19 (type I) are the major keratins in NSCLC [41]. In A549 cells, the KRT8/KRT18 complex is predominant, and low expression of KRT19 can balance the ratio of type I to type II keratins [41, 49]. In LUAD, overexpression of KRT8 is associated with shortened overall survival of patients and resistance to targeted agents such as EGFR-tyrosine kinase inhibitors [50]. It is thus hypothesized that combined detection of KRT8/KRT18 may be more valuable for the treatment and prognostic evaluation of LUAD, though its clinical significance requires further verification.

Over the past decade, research on the tumor microenvironment has advanced significantly. Beyond tumor cells, it contains abundant immune cells, cancer-associated fibroblasts, endothelial cells, other non-malignant cells, and extracellular matrices. Tumor cell staging, initial characteristics, patient clinical features, and the composition and function of various cells in the microenvironment all promote or inhibit tumor progression [51]. Molecular components secreted by these cells are critical for tumor development and therapy [52]. In the tumor immune microenvironment of LUAD, the correlation between KRT18 expression and immune cell enrichment exhibits a clear hierarchical validation logic (Fig. 8). Initial correlation analysis (Figs. 5, 8) based on TCGA-LUAD dataset identified 15 immune cell types significantly associated with KRT18 transcriptional levels, with 8 showing the strongest associations: Th2 cells and NK CD56bright cells displayed significant positive correlations, while T helper cells, Tcm, B cells, T cells, macrophages, and Th1 cells presented significant negative correlations. To validate these correlative findings, we further stratified LUAD patients into high and low KRT18 expression groups and performed Wilcoxon rank-sum test to assess differences in immune cell enrichment (Figs. 6, 8). Consistent with the correlation results, the high KRT18 expression group showed significantly higher enrichment of Th2 cells and NK CD56bright cells, whereas T helper cells, Tcm, B cells, T cells, macrophages, and Th1 cells were significantly less enriched in this group. This progressive analytical strategy—from initial correlation screening to subsequent group-based differential validation—confirms that the association between KRT18 and specific immune cell infiltration patterns is not merely a statistical correlation but represents a robust biological feature in LUAD, providing a more solid basis for exploring the role of KRT18 in regulating the tumor immune microenvironment.Th2 cells, which positively correlate with high KRT18 expression in LUAD, are important in the development of severe cancers and concomitant chronic inflammation, accompanied by an increase in suppressive myeloid cells, pro-inflammatory, growth, and pro-angiogenic factors, resulting in epithelial cell anisotropy, decreased endothelial cell function, and immunosuppression [53]. Coupled with Th1 cells negatively associated with high KRT18 expression in LUAD, this suggests that drift in the Th1/Th2 balance occurs in LUAD and correlates with the expression of KRT18 in LUAD. NK CD56bright cells, a subset of NK cells primarily found in secondary lymphoid tissues, exhibit low CD16 expression and weak cytotoxicity. They are involved in NK cell-mediated tumor killing, immune regulation, and processes related to interferon-γ [54]. Immune cells negatively correlated with high KRT18 expression include adaptive immune cells (Th cells, Tcm, B cells, T cells, Th1 cells) and innate immune macrophages. The low enrichment of these cells in the LUAD microenvironment may impair immune function; for instance, the spatiotemporal functional heterogeneity of macrophages (as observed by Casanova-Acebes et al. [55]) could exacerbate immune deficiencies. KRT18 may influence the landscape of the LUAD immune microenvironment by regulating immune cell recruitment. Whether the positive correlation between NK CD56bright cells and KRT18 synergizes with the low enrichment of other immune cells warrants further investigation, which provides insights into deciphering the role of KRT18 in LUAD immune regulation and its clinical translation. Aberrantly high expression of PD-L1 (CD274) in tumor cells and tumor-associated myeloid cells can impair immune surveillance by binding to the PD-1 (PDCD1) receptor on the surface of adaptive immune cells [56]. As a core approach in tumor immunotherapy, PD-1/PD-L1 inhibitors exhibit heterogeneity in efficacy across cancer types. As a classic biomarker for immunotherapy, PD-L1 TPS primarily reflects the direct interaction of the PD-1/PD-L1 pathway, whereas KRT18 in this study is more focused on regulating the cellular composition of the immune microenvironment (such as enrichment of Th2 cells and low enrichment of immune cells such as T cells), with distinct mechanisms between the two. KRT18 exhibits a negative correlation with CD274 (the gene encoding PD-L1), suggesting that high KRT18 expression may be accompanied by low PD-L1 levels—this association provides a novel perspective for explaining the heterogeneous immune responses in PD-L1 TPS-negative patients. In clinical practice, PD-L1 TPS has limitations (e.g., some negative patients still derive benefits), while KRT18 possesses dual potential for both prognostic evaluation and immune response prediction. It is hypothesized that low KRT18 expression in PD-L1 TPS-negative patients may indicate potential immune benefits, and high KRT18 expression in PD-L1-positive patients may serve as a warning sign of reduced therapeutic efficacy. Combined detection of the two is expected to enhance the precision of immunotherapy stratification in LUAD. Future studies aim to validate the "KRT18 + PD-L1 TPS" combined model in large-sample cohorts, comparing the predictive efficacy of individual versus combined indicators for objective response rate (ORR) and progression-free survival (PFS) to clarify its clinical translational value. Angelo Borsarelli Carvalho Brito’s team confirmed that in NSCLC, PD-1 inhibitors, either as monotherapy or in combination with chemotherapy, are more effective than PD-L1 inhibitors [57]. Our study found that in LUAD, KRT18 expression is negatively correlated with CD274 and shows no significant association with PDCD1. This suggests that patients with high KRT18 expression may have reduced sensitivity to PD-L1 inhibitors but might benefit from PD-1 inhibitors (either alone or in combination with PD-L1 inhibitors). Notably, the positive correlation between KRT18 and CTLA4 indicates that LUAD with high KRT18 expression may concurrently accumulate CTLA4-mediated immunosuppression. CTLA4 inhibits initial T cell activation by competitively binding to CD80/CD86, and together with the PD-1/PD-L1 pathway, promotes tumor immune evasion; their co-expression may enhance the suppressive effect. Therefore, for this subtype, on the basis of considering PD-1 inhibitor regimens, simultaneous evaluation of CTLA4 expression and exploration of the synergistic application of KRT18 targeting and CTLA4 inhibitors may more comprehensively reverse the immunosuppressive microenvironment, providing a new direction for improving immunotherapy responses. However, its mechanisms and value require further verification in preclinical and clinical studies.

Fig. 8.

This flowchart presents an analytical logic from "association discovery" to "differential validation": correlation analysis of immune infiltration provides clues to potential associations, while differential analysis of immune infiltration validates the robust relationship between KRT18 expression levels and the infiltration patterns of specific immune cells

This study has limitations: the small sample size may lead to result bias, and it only relies on bioinformatics analysis and partial in vitro validation, lacking multi-dimensional verification with in vivo animal models and clinical samples. The specific molecular regulatory mechanism of KRT18 in LUAD, especially the upstream and downstream pathways involved in its interaction with immune checkpoint molecules and keratin family members, remains unclear. Future studies should expand the sample size and include multi-center data to enhance the reliability and translational potential of the conclusions.

Conclusion

High KRT18 expression in LUAD is highly correlated with the clinical and pathological staging of the disease and the survival prognosis of affected patients. Additionally, KRT18 expression is of significance in guiding the use of PD-L1 inhibitors for patients with LUAD. Therefore, KRT18 as an effective biomarker has the initial potential of guiding the use of immunological drugs and clinical prognosis of LUAD. Subsequent clinical trials encompassing serology-based KRT18 detection, clinical drug use, and prognosis assessment should be conducted to lay a solid foundation for the clinical application of KRT18. This will aid in the treatment of lung cancer and improve overall patient health levels.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Jiajia Xiao collected samples, performed experimental manipulation, analyzed bioinformatics data, and wrote the manuscript. Zhenpeng Zhu performed data visualization. Fan zhang and Zhicong Yang analyzed the bioinformatics data. Zhenpeng Zhu reviewed the manuscript. Dandan Xu performed experimental manipulation. Fan Zhang observed the pathological specimens. Xinsheng Wang designed the study, reviewed and revised the manuscript, and provided resources. All authors have read and approved the final version of the manuscript.

Funding

This work received financial support from Natural Science Foundation of Hebei Province (C2022405023), the 2024 Government Funded Clinical Medicine Excellent Talent Training Project of Hebei Province (ZF2024224), the Key Research and Development Program of Zhangjiakou City (2121098D), and the Research Fund Project of the Health Commission of Hebei Province (20210702).

Data availability

The datasets presented in this study can be found in the online repositories, including TCGA (https:// portal. gdc. cancer. gov/), UALCAN (https://ualcan.path.uab.edu/), the Human Protein Atlas (https://www.proteinatlas.org/) and STRING (https://cn.string-db.org/).

Declarations

Ethics approval

The study was approved by the Ethics Committee of the First Affiliated Hospital of Hebei North University (Approval No. K2024132). Informed consent was obtained from all individual participants included in the study. All human research participants have signed the relevant informed consent and agreed to the publication of the research results. The authors confirm that the human research participants have provided informed consent for the publication of the images in Fig. 3e–h.

Conflict of interest

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. The authors have no relevant financial or non-financial interests to disclose.

Consent for publication

Not applicable.