Identification and validation a costimulatory molecule gene signature to predict the prognosis and immunotherapy response for hepatocellular carcinoma

Primary liver cancer is an aggressive malignant tumor with high mortality worldwide [1]. Hepatocellular carcinoma (HCC) is the most common histological subtype and the fourth leading cause of cancer-related mortality, accounts for approximately 90% of all primary liver cancer. At present, the traditional treatment methods for HCC are systemic chemotherapy, local ablation, TACE (Transhepatic Arterial Chem Otherapy and Embolization) and surgical resection [2]. However, the therapeutic effect of these methods is away from satisfactory. The effect of anti-tumor agents was not consistent in different clinical trials. Portal vein thrombosis and cholestasis or biliary fistula still needed to be addressed in ablation process, when the target lesion was close to vessels and bile ducts [3, 4]. The recurrence rate of liver cancer after resection was up to 70% at 5 years [5]. In recent years, some clinical trials related immunotherapy showed different outcomes in improving the prognosis of HCC patients [6‐8]. Therefore, it is urgently required to explore novel prognostic signature for HCC that can predict survival and the response to immunotherapy.

The component of immune microenvironment in HCC is the target for many therapeutic advances, including immunotherapy [9]. Most recently, immunotherapies targeting the adaptive immune system, specifically, T cells, have improved tumor control [10]. Activating T cells involves many signals, among which costimulatory molecules are important [11, 12]. HCC could utilize immune checkpoint and evade anti-tumor immune responses by expressing the corresponding costimulatory ligands [13]. B7-CD28 superfamily is a pivotal signal in co-stimulation of T cell activation, and PD-1/PD-L1 also belong to it, which demonstrated the critical effect of costimulatory molecules in HCC [14, 15]. Besides, accumulating evidence has shown that TNF superfamily, another costimulatory signals, plays a central role in cancer immune regulation [16]. The OX40-OX40L axis, a member of the TNF superfamily, has been shown to improve anti-tumor effects of immune cells and effect for cancer immunotherapy [17‐19]. Previous studies also have shown that costimulatory molecules can regulate the tumor immune microenvironment (TME), mainly affecting the activation and proliferation of T cells [20]. Thus, these molecules possibly could provide novel insights in TME. However, the functions of costimulatory molecules in HCC remain unclear.

In this systematic study, we evaluated the expression levels of costimulatory molecules genes in HCC tissues and normal tissues from The Cancer Genome Atlas (TCGA) database. Then a costimulatory molecules-related prognostic signature was constructed for HCC patients and we explored the associations between the prognostic signature and clinicopathological features. Furthermore, we also analyzed the potential roles of this prognostic signature in the immune microenvironment, tumor mutation analysis and response to immunotherapy in different subgroups.

Preliminary data from trials of ICIs in the treatment of HCC led to encouraging results. Nevertheless, with the rapid augment in the utilization of ICIs, immune-related adverse events of HCC arose [30, 31]. Multiple studies found that the usage of ICIs strategies targeting costimulatory molecules for HCC management was promising [32]. Therefore, it was necessary to improve the effect on immunotherapy by selecting the suitable HCC patients according to costimulatory molecules expression patterns. In this study, we analyzed the mRNA expression patterns of costimulatory molecules-related in HCC and selected six genes with prognostic values. Then, we constructed the first costimulatory molecule-related prognostic signature for HCC patients. We found that prognostic signature was strongly associated with clinical characteristics. Additionally, our signature was significantly correlated with tumor immune microenvironment and the response to immunotherapy. Univariate and multivariate cox regression analysis indicated our signature could be an independent prognostic factors for the survival of HCC patients. These findings suggested that CMGs risk signature may indicate some insights to personalized targeted treatment in clinical practice.

Costimulatory molecules played an important role in immunotherapy [20]. Recent findings demonstrated that CD28 co-stimulation was necessary for responses to PD-1 blockade in tumor rejection [33]. Thus, understanding the states of costimulatory molecules in HCC patients will help us determine which patients might benefit in immunotherapy. To explore the expression levels of costimulatory molecules in HCC, we acquired 13 members of the B7-CD28 family and 46 members of the TNF family for HCC patients. Six costimulatory molecular genes (TNFSF4, TNFRSF4, TMIGD2, TNFRSF11A, TNFRSF11B, CD40LG) with prognostic values were selected. The TNFRSF4-TNFSF4 pathway provided crucial co-stimulatory signals for CD4⁺T cell responses [34]. Previous study showed that TNFSF4 was closely related to the unfavorable prognosis of HCC patients [35]. In addition, TNFRSF4 was overexpressed in HCC, associated with a more aggressive phenotype and the activation of multiple immunosuppressive pathways [36]. A phase I clinical research also supported that Ivuxolimab (a TNFRSF4 agonist) showed well tolerance and effective anti-tumor capacity in locally advanced or metastatic cancers, including HCC [37]. Consequently, treatment targeting TNFRSF4-TNFSF4 should be considered in the future. TMIGD2 was mainly expressed in tissue-resident lymphocyte T cells, related to improved tumor prognosis [38]. The different interaction between TMIGD2 and B7-H5 have been identified in certain cancers, such as lung cancer, osteosarcoma, oral squamous cell carcinoma (OSCC), colorectal cancer (CRC) and glioma [39]. The TMIGD2-B7-H5 interaction was involving in Akt-dependent signal pathway, which was a recognized regulation in tumor [40]. Previous study reckoned that HCC patients were likely to be benefited from Everolimus (involved in the Akt-MTOR pathway) only after molecular screening [41]. As such, we could make a reasonable speculation that molecular selection was very necessary for the individualized treatment of HCC and our prognostic signature might provide a new method in distinguishing suitable patients. Of note, our results firstly revealed that TMIGD2 was highly expressed in HCC with favorable prognosis. TNFRSF11A, also known as RANK, was significantly up-regulated in HCC, and can lead directly to migration and invasion by its ligand [42]. Interestingly, genetic deletion of TNFRSF11A in thymic epithelial cells resulted in impaired thymic involution and blunted expansion of natural regulatory T (Treg) cells [43]. Additionally, study showed that HCC patients with high serum TNFRSF11B, also known as osteoprotegerin (OPG), level had poorer survival rates compared with HCC patients with low OPG level [44]. CD40 ligand-expressing dendritic cells could induce regression of HCC [45]. In addition, some clinical trials targeting the agonist or antagonist of CD40/CD40LG has showed promising results in different malignancies. Study showed that adenoviral vectors expressing CD40 ligand (AdCD40L) were safe in vivo and could reduce the number of tumor cell infiltration in bladder cancer [46]. Besides, AdCD40L intratumoral injection increased T-effector/T-regulatory cell ratio by improving systemic immune condition, which was related favorable survival in malignant melanoma patients [47]. Thus, combined other scholar studies and our own bioanalysis, it was possible to improve the prognosis of HCC patients with a similar approach. Moreover, the expression levels of six prognostic genes were verified using qRT-PCR and immunohistochemistry. However, the protein expression of prognostic genes signatures was not completely similar with our previous results, may partly owing to different race and clinical characteristics. At the same time, it also explained that why targeted therapy and immunotherapy did not work for all HCC patients and tumor heterogeneity should be considered in treatment practice. With these six costimulatory molecular genes elucidated in immunity, we hope that the signature constructed by these could predict the response to immunotherapy for HCC patients.

Tumors were complex ecosystems, defined by spatiotemporal interactions among heterogeneous cell types [48]. Subsequently, we compared the associations between our signature and tumor immune microenvironment, the immune cell infiltration and tumor mutation profiles in high-risk and low-risk patients. Our results showed that naïve B cells, plasma cells and regulatory T cells (Tregs) were significantly enriched in low-risk groups. Much of researches regarded Tregs as an immunosuppressive cell, posing anti-tumor immunity in various cancers [49]. Nevertheless, some scholars stated that inhibiting the expression of PD-1 promoted other immune checkpoints, resulting in impaired immune killing ability [50, 51]. Accordingly, fewer Tregs cells in HCC patients with poor prognosis, indicated that those were more likely to be inhibited by PD-1 and activated more immune checkpoints. Correspondingly, high-risk subgroup manifested lower levels of infiltration of immune cells, implicating less process in immune activation.

Tumor mutation burden (TMB) is emerging as a potential biomarker, and participated in immunotherapy-related pathway [52, 53]. We found that the tumor mutation burden (TMB) in the high-risk group was higher than that in the low-risk group with no significant, partly due to the small sample size. TP53 mutation frequency was evidently higher in high-risk group (frequency rate 42%) than low-risk group (frequency rate 14%) according to our mutation results, suggesting more increases genomic instability and complicated major pathway signaling changes in HCC. Additionally, it was important to highlight that different microenvironment-based immune subtypes, based on gene profiling or signatures, and other molecular features, may help identify subgroups of patients more likely to benefit from specific therapies [54]. Some scholars have found that immune-excluded tumors in HCC were proposed to be primarily resistant to ICIs [55]. IPS could predict the response to immunotherapy in cervical cancer and HCC. The prediction of IPS has been demonstrated in different studies [56, 57]. In the present study, we found low-risk group tended to have higher IPS-CTLA4, IPS-PD1/PD-L1/ PD-L2, and IPS-PD1/PD-L1/PD-L2⁺CTLA4, implying that HCC patients with low-risk score could benefit more from immunotherapy than high-risk patients. Therefore, our signature was of great help to clinical immunotherapy decision.

However, there were some limitations in this study. Firstly, we did not explore the exact function of six costimulatory molecule genes in HCC. Thus, it was still necessary to clarify the mechanism of them in the future. Secondly, it was inevitable that there were limited clinical information for HCC patients in public datasets, so the values of the prognostic signature needed to be determined by experimental and prospective studies. Moreover, the risk signature for evaluating the response to immunotherapy was restricted to costimulatory molecule genes and tumor immune microenvironment was highly heterogeneous. Therefore, the prognostic information for HCC patients with immunotherapy were needed to validate the prediction power of our signature clinically.

In our study, we first elucidated the expression of costimulatory molecules for HCC patients, and constructed a six CMGs prognostic signature. The costimulatory molecular-related signature could stratify patients into different subsets with adverse clinical outcomes. In addition, immunotherapy response prediction by our signature explained disparate effect on HCC patients. Consequently, we believed our research manifested the capacity of costimulatory molecules and provided clinicians with applicable treatment.