A programmed cell death-related gene signature to predict prognosis and therapeutic responses in liver hepatocellular carcinoma

Liver cancer is the sixth most common cancer worldwide and the third leading cause of cancer-related death worldwide [1]. As the most commonly diagnosed histologic types (75–85% of cases), hepatocellular carcinoma (HCC) has a poor prognosis and high recurrence, with 5 year survival rates of 19.6 and 2.5% for patients withHCC and advanced and metastatic disease respectively [2]. In the last decades, great advances including surgery intervention, liver transplantation, local ablation, transarterial chemoembolization (TACE) radiation therapy have been made for treating HCC [3, 4]. However, these therapeutic strategies are recommended based on widely used tumor node metastasis (TNM) stage and clinical grade [5], and accurate staging remains as a great challenge for precision treatment. Although systemic treatments has emerged as promising approaches for HCC based on regorafenib, lenvatinib, ramucirumab, carbozantinib, and immune checkpoint inhibitors (ICIs) application [6]. Nevertheless, only certain HCC patients could benefit from them, which necessitates the identification of biomarkers for prognosis prediction and precision treatments.

Clinical features, such as age and stage, cannot achieve precision treatment because it is based on sequencing. Cell death is of vital importance to human health and most cells die from activated programmed cell death (PCD) pathways [7]. Multiple well-defined types of PCD pathways comprising pyroptosis, apoptosis, necroptosis, PANoptosis, ferroptosis, cuproptosis and autophagy-dependent cell death are crucial to homeostasis and disease [8‐11]. Accumulating studies addressed the potential role of PCD mechanism in cancer [12]. Recent evidence also reveals that PCD pathways are implicated in the modulation of immunosuppressive tumor microenvironment (TME) and associated with outcomes following anti-cancer therapeutics [13]. PCD-related signature has been produced to predict the clinical outcomes, TME alterations and therapeutic responses in lung adenocarcinoma [14] and osteosarcoma [15]. According to integrated analysis, cuproptosis- or pyroptosis-related long non-coding RNA signature could predict prognosis and immunotherapy for HCC patients [16, 17]. Therefore, comprehensive analysis of PCD-related genes is necessary for prognosis prediction and therapeutic options.

In this study, we constructed a classification system based on PCD-related genes and found 69 co-DEGs through Venn diagram in The Cancer Genome Atlas (TCGA) database. After removing highly correlated risk genes using LASSO analysis, a PCD-related prognostic signature was generated based on five PCD-related prognostic genes (MCM2, SPP1, S100A9, MSC and EPO), and its robustness was validated in ICGC-LIRI-JP and GSE14520 datasets. This study uncovered the underlying mechanisms of PCD-related genes in HCC, and provided a promising classification as well as PCD-related biomarkers for prognosis prediction and treatment selection for HCC patients.

With an increasing investigation of PCD in various types of cancers, the identification of PCD-related genes has been widely involved in the study of HCC. Therefore, in the present work, based on PCD-related genes, we identified four molecular subtypes showing different prognosis, genomic landscape, pathway characteristics and immune characteristics. Next, five PCD-related genes were discovered from the DEGs among subtypes and used for developing a PCD-related prognostic signature. This prognostic signature had strong robustness in prediction of prognosis and therapeutic response, showing a significant positive correlation with T cells, CD8 and CD4 T cells, B lineage, fibroblast infiltration, and innate and adaptive immune response pathways, and also had stable predictive values in independent datasets.

Various immune cell infiltration including MDSC, dentritic cell and CAF has been proven to be implicated in TME. It has been reported that increased MDSC promotes the development of HCC and an unfavorable prognosis through impairing DC function [32, 33]. IL-1β-induced excessive cystine transport protein-SLC7A11 enhances the expression of PD-L1 and colony-stimulating factor 1, and promotes tumor-associated macrophage and MDSC infiltration, therefore leading to HCC metastasis [34]. Moreover, CAFs are involved in the immunosuppressive microenvironment of tumors. HCC-related CAFs may induce dysfunction of NK cells via affecting tryptophan metabolism by releasing prostaglandin E2 and indoleamine 2,3-dioxygenase [35]. Thus, the crosstalk between these immune cells contributes to the progression of HCC. Additionally, we found that some inflammatory pathways and proliferation-related pathways were both enriched in subtype C1. Among these pathways, NF-κB is a central coordinator of innate and adaptive immune responses, and plays a crucial role in controlling communication between cancer cells and inflammatory cells [36]. The oncogenic role of PI3K-Akt-mTOR signaling in ovarian cancer is to promote uncontrolled cell proliferation, anti-apoptosis, and tumorigenesis [37]. Enhanced activation of the unfolded protein response, which controls multiple tumor-promoting properties in cancer cells, has also been found in lymphoma, neuroblastoma, prostate cancer, and breast cancer [38]. Notch signaling is aberrantly activated in different solid tumors, and cancer cells can use this aberrant signaling to “educate” surrounding microenvironment cells for pro-tumor behavior [39]. The evidence indicated that the enhanced activity of these pathways may be an important reason for the poor prognosis of C1 subtype. In patients with subtype C1, higher expression of PDCD1 (PD-1), CTLA4 and CD274 (PD-L1) has been observed, which is associated with a dismal prognosis [40]. Our identified molecular subtyping indicated a low response of subtype C1 to immunotherapy, which might improve the current classic classification system and offer treatment guidance.

We also found distinct patterns of single-nucleotide variants among the four subtypes, with the greatest difference in prognosis between C1 and C4 and the greatest difference in tumor mutational burden (TMB) between these two subtypes. The gene with the highest mutation rate in C1 was TP53, a tumor suppressor gene whose mutation not only impaired its antitumor activity but also conferred oncogenic properties on the mutated p53 protein [41]. Mutations in CTNNB1 and TP53 are mutually exclusive and represent two major groups of HCC with distinct phenotypic features: alcohol-induced HCC and HBV-induced HCC [42‐44]. The mutation rate of TP53 in C1 was significantly higher than that of CTNNB1 (65 vs 12%), and that of CTNNB1 in C4 was significantly higher than that of TP53 (60 vs 21%). It is therefore likely that C1 is HBV-induced HCC and C4 is alcohol-induced HCC, respectively. We discovered five PCD-related prognostic genes (MCM2, SPP1, S100A9, MSC and EPO). Minichromosome maintenance complex (MCM) genes play important roles in the process of DNA replication and cell cycle. MCM2 belongs to MCM family and participate in the regulation of cell proliferation and the development of cancers [45]. Elevated MCM2 has been found in pan-cancer and is associated with TMB, stage, immunotherapy response and dismal prognosis, indicating that MCM2 acts as a potential target for cancer immunotherapy [45]. MCM2 also stimulates the stemness and resistance to sorafenib of HCC cells through regulating hippo pathway [46]. Previous study demonstrated that upregulated SPP1 contributes to cell proliferation, migration and invasion of lung cancer and is relates to cisplatin resistance, whereas inhibition of SPP1 may improve the survival [47]. Highly expressed SPP1 enhances tumor cell proliferation through inhibiting autophagy and apoptosis [48]. Recently, SPP1 has been identified as one of the autophagy-related genes and the autophagy-related gene signature can predict clinical outcomes and therapy response in prostate cancer [49]. Calprotectin S100A9 is implicated in inflammatory reactions and neoplastic process [50]. Notably, SPP1, S100A9 and EPO have been discovered as immune-associated genes associated with prognosis and therapeutic response in HCC patients [51]. MSC, also known as ABF-1, is a helix-loop-helix transcription factor gene that can inhibit plasma cell differentiation and promotes initiation of memory B cell [52]. Deficiency of MSC after silencing cytidine deaminase may trigger anti-proliferative effects on ceritinib-resistant non-small-cell lung cancer [53]. In this study, these five PCD-related genes were integrated into a model to estimate the mortality risk of patients by defining a risk score for HCC. Patients with a high risk of death had a higher degree of adaptive immune cell infiltration, immune checkpoints overexpression, and adverse immunotherapy response, immune and inflammatory regulatory pathways included NF-kappa B signaling pathway, cGAS-STING signaling pathway, Toll-like receptor signaling pathway, and the activation intensity of MAPK signaling pathway was significantly increased compared with those with a low risk. Therefore, we hypothesized that PCD may partly regulate the immune system of HCC through these five genes, thus affecting the prognosis and immunotherapy response of HCC. This study first identified these PCD-related might contribute to the development of HCC and provided novel a prognostic signature to predict prognosis and therapeutic response for HCC patients.

Cisplatin is a first-line chemotherapeutic drug in the treatment of multiple cancers. Hepatic arterial infusion chemotherapy containing Cisplatin and 5-fluorouracil is a frequently used therapeutic strategy for treating advanced HCC patients [54]. Previous study has revealed that the elevated expression of MCM2 and MCM3 is remarkably associated with acquired resistance of Bel-7402 cells to 5.Fluorouracil [55]. Moreover, Bel-7402/5-Fu cells are not only resistant to 5-Fu, but also resistant to Cisplatin even they have never been exposed to. These findings suggest that aberrant MCM2 is associated with the resistance of Cisplatin and 5.Fluorouracil in HCC. Shen et al. constructed an immune-related gene signature (DCK,CDK4, BIRC5, IL1RN, SPP1, HSPA4, PSMD2, STC2, PGF, and HSP90AA1), and they found that high-risk HCC patients were more sensitive to 5-Fluorouracil, VX-11e and sapitinib [56]. S100A9 has been reported as a potential biomarker for chemoresistance in many tumors [57], nevertheless, its role in Cisplatin and 5-fluorouracil resistance in HCC remains unclarified. In this study, the PCD-related gene signature developed by integrating MCM2, SPP1, S100A9, MSC and EPO provided a risk stratification for HCC patients, and HCC patients with high risk were found to be more sensitive to Cisplatin and 5.Fluorouracil.

Some limitations in the present study should be noted. Firstly, although we have validated the prognostic value of the PCD-relate gene signature in external datasets, the reliability and accuracy should be further validated before clinical application. Secondly, more clinical cohorts with large sample size could be used to verify the identified molecular subtypes and prognostic signature. Finally, further in vivo and in vitro experimental studies using qRT-PCR, western blotting, flow cytometry, immunofluorescence assay, and immunohistochemical analysis are required to investigate the underlying mechanism of PCD-related genes in HCC development and drug resistance of Cisplatin and 5.Fluorouracil.

In conclusion, we identified four PCD-related subtypes of LIHC, which could improve the clinical classification of HCC patients. Based on the identified molecular subtypes, the prognostic signature of five PCD-related genes was developed and validated to be able to predict survival, clinicopathologic features, tumor immune phenotypes and therapeutic responses. This study provided new insights into the subtyping of HCC and directions of personalized treatment for HCC patients.