Introduction
Cancer stemness is the capacity for self-renewal and differentiation, which leads to tumor metastasis and relapse [
1]. In addition, cancer stemness is associated with genomic and proteomic signatures that can modulate malignant biological behaviors and support the initiation, differentiation, and proliferation of tumor cells [
2]. Mounting evidence indicates that tumor cells bearing stemness features can differentiate into cancer stem cells (CSCs), which are empowered with increased metastatic capacity and resistance to therapy. Such stem-like cells also exist in various cancers, including lung squamous cell carcinoma (LUSC), and play a critical role in the genetic profile of the tumor microenvironment [
3]. Due to the complexity and heterogeneity of the tumor microenvironment, it remains unclear how stemness features regulate stem cell-related biological programs and shape the tumor microenvironment.
LUSC is the top global cause of death with high mortality rates but lacks effective therapeutic strategies [
4]. Although the death rate of lung cancer has declined over the past few decades, the average five-year survival rate for LUSC is only between 20% and 30% [
5]. Compared with lung adenocarcinoma (LUAD), there are still few effective targeted treatment options for LUSC in the clinic [
6]. Unfortunately, LUSC does not respond well to chemotherapy and radiotherapy as well as other types of cancers [
7]. In recent years, advances in cancer immunotherapy have extended overall survival (OS) in select non-small cell lung cancer (NSCLC) patients with positive PD-L1 expression. Nevertheless, only a small percentage of LUSC patients show a survival benefit from immunotherapy [
8]. Therefore, a subtype classifier based on specific characteristics for survival prediction and therapy response estimation is the first step toward personalized cancer treatment for LUSC patients.
A recent study provided strategies for integrated analysis of cancer stemness features according to the stemness index (mRNAsi), which could classify tumors based on their stemness features and provide predictive biomarkers for treatment response and survival outcome [
9]. The stemness index mRNAsi have proven to be associated with the dedifferentiated oncogenic state and infiltrating immune cells of the tumor microenvironment [
10]. Furthermore, multiple stemness-related genes have been confirmed to be involved in the prognosis and response to different therapies [
11]. However, most existing studies refer to the identification of stemness-related prognostic genes [
12,
13]. The relationship among stemness features, tumor heterogeneity, and treatment responsivity in LUSC patients is still unknown. Thus, further integrated analysis of the genetic features of stemness and stemness-related heterogeneity is important for accurate classification and guiding treatment selection for LUSC patients.
In this study, the stemness index (mRNAsi) of LUSC patients was calculated according to mRNA expression data from The Cancer Genome Atlas (TCGA) and GEO databases. Subsequently, LUSC patients were divided into high-mRNAsi and low-mRNAsi groups based on mRNAsi scores, which exhibited distinct survival outcomes and functional annotations. Next, we applied consensus clustering analysis based on stemness-related differentially expressed genes (DEGs) to classify patients into two subtypes with distinct prognoses. Furthermore, bioinformatic analysis were performed to investigate the differences in functional enrichment, immune profiles, and the response to different treatment strategies between these two stemness subtypes. Finally, we constructed a stemness subtype classifier to distinguish these two subtypes and validate the subtype classifier into three independent GEO datasets. Our study provides a clinical practice tool for survival prediction and screening which patients will respond well to immunotherapy, chemotherapy, and targeted therapy.
Discussion
Due to tumor intrinsic heterogeneity and complex genomics, new cancer treatments for LUSCs have been challenging in recent years [
3]. It is urgent to determine the interplay of oncogenic pathways and develop new therapies available for LUSC patients. Cancer stemness is associated with particular oncogenic pathways that can modulate transcriptional networks and support cancer cell growth, proliferation and metastasis [
30]. Furthermore, cancer stemness can effectively quantify the level of oncogenic differentiation in tumor tissue via the mRNA expression-based stemness index (mRNAsi) [
9]. Recent studies revealed that cancer stemness could affect the treatment response and clinical outcome in different types of cancer, including LUSC [
31,
32]. Our study performed an in-depth analysis of the correlation between cancer stemness and the efficacy of immunotherapy, chemotherapy, and targeted therapy in patients with LUSC. We presented an approach to discriminate tumor subtypes with distinct treatment responses and prognoses according to the stemness index and validated this approach through its application to multiple independent datasets.
Here, we calculated the stemness index (mRNAsi) of LUSC patients from the TCGA and GEO databases via the OCLR algorithm. The mRNAsi was low in normal samples but high in tumor samples, which was consistent with the point that tumor progression involved the acquisition of oncogenic dedifferentiation and stemness features. After performing an integrated analysis of the connection of mRNAsi with the survival outcome, we observed that the stemness index was positively associated with OS, while no significant difference in PFS was found between the low mRNAsi group and the high mRNAsi group, indicating that high mRNAsi was an indicator of favorable OS for LUSC patients. Interestingly, a negative association between the stemness score and survival was reported in some cancers, such as pancreatic cancer and liver cancer [
33,
34]. The disparate results indicated that the association between the mRNAsi score and survival outcome across different tumor types is complex and likely involves multiple factors, cancer stemness may be linked to the origins of malignant cells and the heterogeneity of tumors in certain types of cancer. In the case of LUSC, several factors may contribute to this finding. Firstly, we observed that patients in the high mRNAsi group tended to be younger than those in the low mRNAsi group. Secondly, our gene set variation analysis (GSVA) revealed a significant enrichment of pathways related to DNA damage repair in the high mRNAsi group, including homologous recombination, mismatch repair, and nucleotide excision repair pathways. It is widely acknowledged that the DNA repair capacity of tumors is an important prognostic factor in cancer patients. Therefore, further research is needed to fully elucidate the underlying mechanisms behind the observed association between mRNAsi scores and prognosis in LUSC.
Mounting evidence suggests that stemness is associated with immune microenvironment variables and the antitumor immune response [
10]. Our results showed that the mRNAsi score was negatively correlated with most tumor-infiltrating immune cells, and the results were truly unexpected. However, several studies have reported that there is a negative association between cancer stemness and immune infiltration [
29,
35]. The tumor immune microenvironment is diverse and complex in terms of immune status, and complex interactions among tumors, immune cells and their microenvironment exist throughout the initiation and development of tumors. Tumor-infiltrating immune cells may perform either protumorigenic or antitumor roles, which could shape their microenvironment and affect tumor development and the prognosis of patients [
36]. Our results demonstrated that most immune cells were increased in the low mRNAsi group, including immunosuppressive cells such as regulatory T cells (Tregs) macrophages and tumor-associated macrophages, which were associated with poor prognosis. Another possible explanation for the high CD8 T-cell abundance in the low-mRNAsi group with poor prognosis may be that several immunosuppressive molecules, including PD-1, TIM3, and LAG3, were also higher in the low-mRNAsi group. High PD-1 and TIM3 expression on CD8 T cells was associated with exhaustion status, which may contribute to the poor prognosis of patients with lung cancer [
37,
38]. In addition, tumor-infiltrating immune cells may vary in their activation status under different stimulators.
Immunotherapy, such as ICIs, has revolutionized the treatment options for LUSC owing to its durable response but manageable side effects and is currently approved as the first-line treatment for patients with advanced LUSC [
4,
8]. However, a large proportion of LUSC patients do not respond to cancer immunotherapy. Based on the above issue of clinical efficacy in immunotherapy, our study constructed a novel LUSC classification according to tumor stemness. LUSC patients were divided into stemness subtype A and stemness subtype B based on the expression of stemness-related DEGs. We observed that patients in stemness subtype A with lower mRNAsi scores responded better to immunotherapy than those in stemness subtype B. Various factors may influence the response to immunotherapy in lung cancer. Taking ICIs as an example, PD-L1 expression, tumor-infiltrating lymphocytes, tumor mutation burden (TMB) and mismatch repair deficiency status were all able to affect the efficacy of ICIs [
39,
40]. Our results showed that stemness subtype A tended to manifest as increased expression of immune coinhibitory/costimulatory genes, including PD-L1, enrichment of immune-related pathways and high immune status, which could explain why patients with stemness subtype A have a better response to immunotherapy.
At present, platinum-doublet chemotherapy is still the standard of treatment for patients with unresectable LUSC [
41,
42]. We found that patients with stemness subtype A showed higher sensitivity to first-line chemotherapeutic drugs, including cisplatin, gemcitabine and vinorelbine. Our results were consistent with previous reports that a higher stemness index was correlated with chemoresistance due to its self-renewal ability and drug-efflux ability [
43]. Furthermore, DNA damage repair pathways are important determinants of sensitivity to chemotherapeutic agents [
44]. GSVA showed that DNA damage repair pathways, including homologous recombination and nucleotide excision repair pathways, were enriched in stemness subtype B, which may contribute to chemoresistance in the stemness subtype B group. Targeted therapies with EGFR-TKIs have shown very limited clinical benefits in treating LUSC patients [
45]. Our study showed that patients with stemness subtype B were more sensitive to EGFR-TKIs and resistant to VEGFR, PARP1 and PI3K inhibitors. The underlying cancer stemness ability is dependent on multiple molecular targets, including signaling pathways, the tumor microenvironment and stem cell differentiation. These molecular targets may be involved in the efficiency of tumor chemotherapy and targeted therapy, indicating that the combination of chemotherapy, targeted therapy or immunotherapy may provide more efficient management to eliminate cancer stemness in LUSC patients.
To apply our results in clinical practice, we developed a clinically applicable classifier that could easily discriminate the stemness subtype of LUSC patients based on 190 stemness-related DEGs. We identified nine hub genes (AXL, EFEMP2, VIM, EHD2, COL3A1, FSTL3, ALOX5, TNFRSF12A, HSPB8) and defined them as stemness subtype classifiers by LASSO and RF machine learning methods. The AXL protein belongs to the TAM (TYRO3, AXL, and MER) family of receptor tyrosine kinases, which is an essential factor for stemness. Upregulation of AXL expression is correlated with resistance to TKIs and chemotherapeutic agents in various types of cancer, including LUSC [
46]. EFEMP2 and EHD2 have been reported to inhibit the invasion and metastasis of lung cancer cells by regulating the epithelial-mesenchymal transition (EMT) process and MMP activity [
47,
48]. HSPB8 is a stress-related protein that plays an important role in tumor proliferation, invasion and apoptosis in lung cancer [
49]. FSTL3, as an oncogene of the FSTL family, is involved in the occurrence and progression of lung cancer [
50]. Previous reports have demonstrated that FSTL3 is linked to remodeling of the tumor immune microenvironment and may serve as a predictor of sensitivity to immunotherapy and chemotherapy [
51]. COL3A1 is an integral ECM protein that is closely involved in malignant progression and drug resistance by regulating tumor immunity and EMT in a variety of cancers, particularly lung cancer [
52]. ALOX5 encodes a nonheme iron-containing dioxygenase of the lipoxygenase gene family that has been identified as a critical regulator of cancer stem cells from hematological malignancies [
53]. TNFRSF12A is a member of the TNF superfamily of receptors that has been reported to be elevated in different cancers [
54]. At present, the role and clinical value of these two genes in LUSC remain unclear. However, inhibition of these stemness-related hub genes may be a promising approach to gain a better therapeutic effect in LUSC patients.
In recent years, studies have confirmed that cancer stemness and stemness-related genes could serve as diagnostic, prognostic and therapeutic biomarkers in various cancers [
55,
56]. Our classifier could identify two distinct stemness subtypes in LUSC patients and provide a possible method for screening LUSC patients who display an effective response to different treatment strategies. However, there are still several limitations in this study. The major limitation of this study was that all of these results were based on bioinformatics analysis of public databases. Although three GEO datasets were enrolled as an external validation set to verify the predictive efficiency and support the conclusions of our study, a clinical cohort from our own center to confirm the classifier is necessary. Furthermore, additional in vivo or in vitro experiments, such as flow cytometry or preclinical models, are warranted to comprehensively analyze the molecular mechanisms and verify our results.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.