Introduction
Lung cancer ranks among the most prevalent and fatal malignant tumors globally, with over 2.2 million new cases reported worldwide, including more than 800,000 occurrences in China each year. Due to advanced-stage diagnosis, the five-year survival rate for lung cancer patients stands at only 17.6% [
1]. Lung squamous cell carcinoma (LUSC) represents one of the predominant pathological subtypes of lung cancer, displaying an increasing incidence trend in the Chinese population in recent years [
2,
3]. Notably, early-stage LUSC patients exhibit a higher 5-year survival rate due to slower growth, delayed metastasis, and enhanced opportunities for surgical resection [
4,
5]. Nonetheless, a comprehensive understanding of the mechanisms underlying the onset as well as the progression of LUSC at molecular level is lacking. It is imperative to categorize LUSC patients based on prognosis, thereby improving their outcomes and investigating prognosis-affecting indicators. This endeavor holds paramount significance in enhancing the quality of life for LUSC patients and alleviating disease burdens. The induction of tumor cell apoptosis or other programmed cell death to impede tumor growth serves as the foundation for all cancer treatment protocols [
6]. While apoptotic tumor cells were conventionally considered non-immunogenic and immune-tolerant [
7,
8], studies have demonstrated immunogenic properties, termed immunogenic cell death (ICD), in certain apoptotic tumor cells [
9,
10]. ICD denotes a unique form of regulated cell death (RCD) driven by stress [
11] and functions to stimulate the immune system, eliciting an immune response [
12,
13]. To date, a total of 33 genes, denoted as ICD-related genes, have been identified to be associated with ICD [
14]]. Many scholars have utilized these 33 ICD-related genes to construct prognostic models for various malignancies, such as head and neck squamous cell carcinoma [
15], lower-grade glioma [
16], as well as ovarian cancer [
17]. Hence, the ability to induce immunogenic tumor cell death constitutes a pivotal factor influencing treatment effectiveness and tumor prognosis.
Long non-coding RNA (lncRNA) refers to noncoding RNA exceeding 200 nucleotides, playing pivotal roles in diverse biological processes such as dosage compensation, epigenetic regulation, cell cycle modulation, and cell differentiation regulation [
18‐
20]. Previous research has emphasized the relative abundance and high stability of lncRNA in circulation, rendering it more dependable than other well-characterized genes [
21,
22]. At the same time, lncRNAs had been found to affect chemotherapy sensitivity [
23,
24]. Additionally, lncRNAs have been shown to impact chemotherapy sensitivity and immunotherapy tolerability in lung cancer [
25]. Moreover, significant prognostic differences have been observed in lung cancer patients based on distinct lncRNA expression profiles [
26,
27]. Notably, several lncRNAs, including Necroptosis-Related lncRNA [
27], N6-Methyladenosine (m6A)-Related lncRNA [
22], and ferroptosis-related lncRNAs [
28] ], have been linked to the clinical outcomes of LUSC. However, the influence of ICD-related lncRNAs on the prognosis of LUSC patients is still largely unexplored.
In our current investigation, we aimed to identify ICD-related lncRNAs possessing independent prognostic value for constructing a risk model.
Materials and methods
Data collection and compilation
We retrieved RNA-seq datasets comprising 51 normal or paracancerous tissues and 502 LUSC tumor tissues from The Cancer Genome Atlas (TCGA) database (
https://portal.gdc.cancer.gov/). Additionally, we obtained 504 relevant clinical information datasets. These datasets were organized into a matrix file using the Perl programming language. After excluding patients with incomplete clinical information (comprising age, sex, TNM staging, survival status, and survival time), 187 patients were randomly allocated to training cohorts, while 184 patients were assigned to testing cohorts. Furthermore, we sourced 33 ICD-related genes from a previous study [
14].
Utilizing the “limma” package (
www.bioconductor.org/packages/release/bioc/html/limma.html) [
29] based on the R programming language, we conducted differential expression analysis for lncRNAs and the 33 ICD-related genes. Subsequently, Pearson correlation analysis was employed to investigate the relationship between ICD-related genes and differentially expressed lncRNAs (Filter condition: correlation coefficient = 0.4, P < 0.001). The interaction network of ICD-related genes-lncRNAs was visualized utilizing the “igraph” package (
https://igraph.org/).
Identification of prognosis-associated ICD-related lncRNAs
Clustering of prognosis-associated ICD-related lncRNAs
Construction and validation of risk models for ICD-associated lncRNAs
Using a dataset of 371 LUSC patients with complete information, we conducted Cox regression analysis to construct risk models related to ICD-associated lncRNAs. The risk score calculation formula comprised coefficients derived from the expression levels of multiple lncRNAs. We used univariate and multivariate Cox regression analysis to assess the effect of risk score on the prognosis of LUSC patients. We then divided LUSC patients into the high-risk / low-risk cluster based on the median risk score. The “survivor” package was employed to compare survival differences between these clusters, with the “timeROC” package assessing the prognosticvalue of risk models for LUSC patients.
Patients and samples
Between January 2020 and June 2022, 43 patients diagnosed with LUSC underwent surgical treatment, with their tumor tissues, paracancerous tissues, and normal lung tissues frozen in liquid nitrogen. The patient cohort was composed by 31 males as well as 12 females (age range: 48–72 years). Tissue differentiation levels were distributed as follows: 16 cases of low, 20 cases of moderate, and 7 cases of high tissue differentiation. Tumor-Node-Metastasis (TNM) staging revealed 17 cases in stage I, 18 cases in stage II, and 8 cases in stage III.
Informed consent was obtained from all subjects, and the present study was approved by the ethics committee of Cangzhou Central Hospital.
Real time fluorescence quantitative polymerase chain reaction PCR (RT-qPCR)
Total RNA extraction from tissues was carried out utilizing an RNA extraction Kit (RC101-01, Vazyme, China). Subsequently, cDNA synthesis was performed according to the instructions of the PrimeScript RT reagent (RR047A, Taraka, Japan). A 20 ul RT-qPCR system was prepared following the instructions of a qPCR master mix kit (A6001, Promega, USA). The PCR primer sequences for lncRNA are provided in Supplement Table
1.
Table 1
Patient demographic of training cohort and test cohort
Age (mean ± SD, years) | 67.65 ± 8.42 | 67.03 ± 8.91 | 1.431 | 0.319 |
Gender (n (%)) |
Male | 140 (74.87) | 131 (71.20) | 0.635 | 0.426 |
Female | 47 (25.13) | 53 (28.80) |
TNM stage (n (%)) |
I | 96 (51.34) | 81 (44.02) | 2.030 | 0.566 |
II | 60 (32.08) | 67 (36.41) |
III | 29 (15.51) | 34 (18.48) |
IV | 2 (1.07) | 2 (1.09) |
Alive (n (%)) |
yes | 64 (34.22) | 81 (44.02) | 3.739 | 0.053 |
no | 123 (65.78) | 103 (55.98) |
Survival time (mean ± SD, years) | 1.64 ± 1.34 | 1.53 ± 1.63 | 0.742 | 0.458 |
Positron emission tomography/computed tomography (PET-CT) examination methods and image analysis
All patients underwent a PET-CT scan (Biograph 16 h, Siemens) before surgery, with intravenous 2’-deoxy-2’-[18F] fluoro-D-glucose (18F-FDG) (3.70–5.55 MBq/kg) administered. The region of interest was delineated along the edge of the primary lesion of LUSC, and the maximum standard uptake value (SUVmax) and metabolic volume (MTV) of the primary lesion were automatically obtained using the fixed threshold method (SUVmax = 2.5 was the threshold value).
Statistical analysis
Statistical analysis was conducted using SPSS 20.0 (IBM, USA). Group differences were compared using unpaired t-tests, student’s t-tests, or chi-square tests. Pearson methods were utilized to analyze the correlation of risk scores with SUVmax or MTV in 43 LUSC patients. P < 0.05 was considered statistically significant.
Discussion
Globally, more than 400,000 people have been reported to die from LUSC and its related complications annually [
31,
32]. Despite the slow progression of LUSC and its high potential for surgical resection, treatment options are constrained due to its lower responsiveness to radiotherapy and chemotherapy compared to lung adenocarcinoma. Notably, patients with LUSC exhibit distinct clinical features, including smoking history, complications, age, and molecular characteristics, which differ significantly from patients with lung adenocarcinoma, resulting in a poorer prognosis for advanced LUSC patients [
5,
33]. Thus, the discovery of novel biomarkers has the potential to guide treatment strategies and aid in the prognostic prediction of LUSC patients.
In this study, we initially categorized LUSC patients into 2 groups using cluster analysis based on 16 ICD-associated lncRNAs linked to LUSC prognosis. We observed notable differences in the cumulative overall survival of these patient types, indicating the potential use of ICD-related lncRNAs as prognostic factors for LUSC patients. Subsequently, we utilized the LOSS regression model to develop a prognostic risk model for LUSC patients, incorporating 12 ICD-related lncRNAs associated with the prognosis of LUSC. Both univariate and multivariate regression analyses, along with total survival analysis and ROC curve evaluations, substantiated the accurate predictive ability of this risk model across different LUSC patient subtypes.
When tumor cells are disrupted by external factors, they can transition from being non-immunogenic to triggering an anti-tumor immune response a phenomenon known as ICD [
34,
35]. During ICD, tumor cells release damage-associated molecular patterns (DAMPs) such as calreticulin on the cell surface, secreted HMGB1, ATP molecules, HSP70, and HSP90 [
36,
37]. These DAMPs bind to pattern recognition receptors on the surface of dendritic cells, initiating cellular responses that activate both innate and adaptive immunity [
38,
39].ICD can be induced by various stressors including intracellular pathogens, traditional chemotherapy drugs, targeted anticancer drugs, and physical therapy [
13,
40]. Building upon these findings, our study developed a risk model based on ICD-related lncRNAs associated with the prognosis in LUSC patients. This model stratified LUSC patients into high-risk and low-risk groups, revealing that those within the high-risk category exhibited lower overall survival rates in contrast with the low-risk group.
Currently, PET/CT has been pivotal for the evaluation of the clinical efficacy and prognosis of cancer patients in China post-treatment. The fundamental principle involves illustrating the biological characteristics of tumor cells using various contrast agents to assess cancer cell activity in patients [
41,
42]. Comparing PET/CT imaging data of cancer patients before and after treatment in China accurately reflects clinical treatment efficacy, facilitating the early identification of ineffective treatments and predicting post-treatment patient prognosis [
43,
44]. However, the high cost and limited availability of equipment/imaging specialists are primary constraints affecting timely PET/CT testing for all cancer patients. In numerous Chinese hospitals, patients often face prolonged waiting periods of weeks for PET/CT tests, with local PET/CT testing costing approximately 5,000–7,000 renminbi (RMB) and whole-body PET/CT testing reaching tens of thousands of RMB, further burdening cancer patients [
45,
46].
In the present study, we observed a positive association of the risk score calculated from the ICD-lncRNA model with the SUVmax and MTV obtained from PET-CT. SUVmax, a widely used metabolic parameter in PET-CT imaging, provides a semi-quantitative measure of
18F-FDG uptake in tumors, reflecting metabolic activity within highly proliferative tissues [
47,
48]. Extensive research has highlighted SUVmax as a vital imaging marker for predicting the prognosis of non-small cell lung cancer patients [
46,
49,
50] Notably, it’s essential to recognize that SUVmax only captures the uptake in a segment of the tumor mass and does not account for the entire tumor. Meanwhile, MTV represents the volume of all pixels within a specific range of SUV on PET-CT images, serving as a metabolic parameter based on tumor volume size [
51]. Recently, there has been increased recognition of the value of metabolic tumor volume as an indicator of metabolic tumor burden, showing promise as a quantitative PET index [
52,
53]. Remarkably, our constructed risk model, requiring only a few hundred RMB and minutes, offers an efficient means to accurately assess patient prognosis compared to PET/CT, which is considerably more costly and time-consuming.
In summary, we identified ICD-related lncRNAs associated with the prognosis of LUSC patients from the TCGA public database and utilized this information to develop a risk model capable of accurately predicting patient prognosis. This risk model not only forecasted patient outcomes in LUSC but also provided guidance for treatment selection. Nevertheless, the limitations of this study should be noted. Specifically, our validation included only 43 LUSC samples, indicating an unavoidable limitation due to the small sample size. Additionally, the dataset of LUSC patients within the TCGA database originated from varied sources, potentially introducing concerns about data heterogeneity given potential differences in data acquisition methodologies across institutions. Lastly, there were potential confounding variables that should be considered.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.