Exploration of Key Genes Combining with Immune Inltration Level and Tumor Mutational Burden in Hepatocellular Carcinoma

Background: Hepatocellular carcinoma (HCC) is still a lethal malignancy because of its heterogenicity and aggressive behavior, with unsatisfactory early diagnosis and poor prognosis. Recently, few somatic mutations have been reported to associated with HCC carcinogenesis and function in predicting HCC progression. Interactions of tumor cells with surrounding immune microenvironment in HCC participate in orchestrating the onset and development. Herein, our aim is to investigate the associations of tumor mutational burden (TMB) with immune microenvironment in HCC. Then we will seek for differential expression genes (DEGs) in terms of immune and TMB scores and discuss whether their latent functions affect HCC prognosis and progression. Methods: The expression, clinical and mutational data were downloaded from TCGA database, then calculated the immune inltration of these HCC samples by R package “ssGSEA”, “CIBERSORT” and “ESTIMATE”. Then the samples were claried to 2 groups according to the immune levels. TMB was also calculated and differential expressional genes (DEGs) in the low and high TMB group were intersected and the different immune level groups. Then the cox analyses and prognostic model were performed and tested by R package “glmnet”. Then the selected genes BCL10 and TRAF3 were tested their expression by qrt-PCR and IHC and tested their clinical correlation by chi-square analyses and their biological processes enriched by GSEA, and their immune inltration by “ssGSEA” individualy. Last, pearson algorithm was employed to judge the relevance of BCL10 and TRAF3. Results: Upregulated degrees of immune inltration correlated with TMB, they synergistically predicted poor prognosis in HCC. DEGs enriched in immune-related pathways could serve as an indicator of therapeutic effect of HCC immunotherapy. Among these DEGs, BCL10 and TRAF3 were highly expressed in HCC tissues, especially in TP53 mutation group. BCL10 and TRAF3 corporately exhibited immunological function, thereby affecting HCC progression and prognosis. Conclusions: We identied that BCL10 and TRAF3 exhibited good prognostic value in predicting the clinical outcome of HCC patients, which may inuence TMB and tumor microenvironment (TME) and help us moving towards immune-based therapies with HCC patients, ultimately improving their long-term survival. immune checkpoint


Introduction
Liver hepatocellular carcinoma (LIHC) occupies nearly 90 percent of primary hepatic carcinoma possesses high morbidity and mortality, which is the fth most prevalent malignancies and the second primary cause of cancer -related death globally [1]. World Health Organization speculated that in 2030, about one million patients with hepatopathy will die from hepatocellular carcinoma [2]. LIHC is attributable to chronic hepatitis or cirrhosis arise from various etiologies, including infection with hepatitis B and C virus, as well as alcoholic and nonalcoholic fatty hepatic disease [3]. Satisfactory sanative therapies for hepatocellular carcinoma (HCC) comprised of surgical resection, chemoradiotherapy, ablation, immunotherapy and liver transplantation [4]. However, the treatment of HCC is still challenging owing to the lack of early diagnosis, potent prognosis markers and its high recurrence rate and metastasis [5]. Thus, discovering reliable molecular mechanisms for HCC is imminently needed to improve long-term survival of patients.
Gene mutations are abrupt, heritable variations in genomic DNA molecules. Mutants and viruses in the environment stimulate the activation of hepatocyte division response pathway, causing cell point mutation and gene translocation, which accelerate the malignant transformation of cancer cells [6].
Tumor mutational burden (TMB) is de ned as the total amount of somatic mutations per megabase (Mb) within the tumor genomic sequence [7]. These genetic mutations include nonsynonymous mutations mostly comprised of missense mutations. In addition, synonymous mutations, insertions or deletions and copy number gains and losses are also included in some cases [8]. Recently, TMB is considered to serve as a biomarker across diverse solid cancers for prediction of the neoantigen load, tumor immunogenicity and prognosis after clinical treatment [9]. Interestingly, TMB levels in patients varies in a tumor type dependent manner, thus contribute to opposing prognosis and survival in uences across tumor types [10,11]. Considering that the heterogeneity between individual tumors, the frequency of TMB can vary across tumor types [12]. Therefore, it is high time to take tumor/microenvironment interaction into account when analyzing the correlation of TMB with prognosis. Nonetheless, the value of TMB is still unclear and no speci c relationship has been uncovered between TMB and prognosis in patients with HCC, which need further elucidation.
Currently, immune-related TME has been recognized to exert a pivotal in uence during tumor progression [13]. Immune in ltration in TME has been emphasized by previous researches to be responsible for the early recurrence and metastasis in HCC [14]. Tumor-in ltrating immune cells are reported to contribute to impaired anti-tumor immunoreaction in response to disordered metabolic activity of tumor cells, which is speculated to participate in immune suppression and malignant progression of HCC [15]. Notably, immunotherapy such as immune checkpoint inhibitor (ICI) is recognized as a therapy for various malignancies [16]. Additionally, TMB has been discovered to closely correlated to immunotherapy in several cancer types. The impact of TMB on prognosis and its connection with immune in ltration varied across different cancer types [17]. Because few studies had concentrated on TMB and immune in ltration with respect to LIHC, so we conducted this research to inquire into the prognostic value of TMB and its latent correlation with immune in ltration in LIHC.
In our study, we use databases to evaluate the different degrees of immune in ltration on prognosis and mutation in LIHC. In addition, the correlation between TMB and immune in ltration and their jointly predictive value were also studied by using multiple bioinformation analysis methods. Next, we selected different expression genes based on immune and TMB score in HCC and performed functional enrichment and univariate Cox regression analysis. A prognostic signature was also constructed and the superiority of the model was assessed. Then, we found that BCL10 and TRAF3 expression were evidently elevated in TP53 mutation group in HCC tissues and were associated with several clinical prognostic factors. Finally, functional enrichment, immune in ltration and correlation analysis of BCL10 and TRAF3 were performed. Therefore, BCL10 and TRAF3 are involved in TMB and tumor immune microenvironment, thus in uencing the progression and prognosis of HCC.

Patients and sample collection
This study was approved by the ethics committee of the a liated hospital of Nantong University (AHNTU). Total HCC tissues (n = 20) and surrounding normal tissues (n = 20) were obtained from the A liated Hospital of Nantong University for IHC and qRT-PCR. Permission was approved by the Human Research Ethics Committee of AHNTU, and written informed consent was obtained from each patient.

Data downloading and processing from TCGA LIHC cohort
Transcriptional data, including the counts data and FPKM data, and correspondingly clinical data, including overall survival status, overall survival time, T stage, N stage, M stage and Ajcc stage from TCGA LIHC cohort were obtained by the R package "TCGAbiolinks" [18]. Somatic mutation data from TCGA LIHC cohort were obtained with the selected "mutect2" dataset and visualized by the R package "maftools" [19].

Tumor Immune microenvironment purity assessment
Implementation of Single-Sample Gene Set Enrichment Analysis (ssGSEA) [20], CIBERSORT algorithm [21], ESTIMATE algorithm were facilitated to evaluate the immune in ltration subsets of each sample individually. A reference document containing the key gene expression of 22 leukocyte subtypes were supplemented in the CIBERSORT algorithm.

Consensus clustering Analyzing
The 374 HCC specimens from TCGA LIHC cohort were classi ed into 2 clusters by the R package "ConsensusClusterPlus" [22]according to the immune in ltration level from CIBERSORT.

Tumor mutational burden (TMB) calculating
TMB function as the mature biomarkers in immune checkpoint inhibitors, which is considered to directly measure the number of mutations in tumor tissues. Here, we selected the calculating methods recently reported by Zachary [23]. After assessing the TMB score for each sample in TCGA LIHC dataset, the samples were divided to the high TMB score and the low TMB score group, for subsequent detecting of the differential expressed genes (DEGs).

DEGs and enrichment measurements
Under the screening criteria of DEGs: |log2Foldchange| >1, adjust P value < 0.05, R package "Deseq2" was utilized to get the DEGs between the ICR1 and the ICR2 cluster, and between the different levels of TMB group [24]. 369 genes in total were obtained from these two differential analyses. Then gene ontology (GO), including the MF, BP, CC pathways and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were employed to analyze underlying biological characteristics and signi cance of theses DEGs [25].

Single-gene for Gene Set Enrichment Analysis (GSEA)
With the purpose of studying possible pathways associated with BCL10 and TRAF3 from the TCGA_LIHC dataset, respectively, GSEA generated an ordered list of all genes from TCGA LIHC FPKM transcriptional data based on their correlation with BCL10 and TRAF3 expression, respectively. The "h.all.v6.2.entrez.gmt" were chosen as reference gene sets. Then, the distinction in gene enrichment between the high-level BCL10 and TRAF3 group and the low-level BCL10 and TRAF3 group were analyzed using GSEA Version 2.10.1 software[26].

Prognostic model construction and inspection
First, univariate Cox regression model was constructed to analyze the 369 DEGs which can predict the prognosis of HCC patients and the visualized forest plot manifested the results. Then the LASSO Cox regression model was constructed and tted with by R package "glmnet" [27]. Under this progression, 19 key genes were concluded in the formula to calculate the risk scores (RS). Then we generated ROC curve by "ROCR" to assess the e ciency of RS to predict prognosis [28]. In order to better predict the prognosis of HCC patients by combining our predicted RS with clinical indexes, Nomogram was integrated with RS, T stage, N stage, M stage and age, gender and generated [29]. Lastly, C-Index, Net reclassi cation index (NRI) and Decision Curve Analysis (DCA) were calculated to evaluate the nomogram.

RNA extraction and quantitative real time-polymerase chain reaction (qRT-PCR)
We carried out RNA extraction and qRT-PCR procedure as previously depicted [30]. The BCL10 and TRAF3 primer sequences are listed below.

The landscape of immune in ltration in HCC patients from TCGA
In order to study the immune in ltration of liver cancer patients, we facilitated the data from TCGA LIHC dataset and performed CIBERSORT to quantify the immune cells levels of HCC. A landscape of immune cell interaction in HCC was illustrated by the correlation coe cient heatmap (Fig. 1A). In order to understand whether the samples could contribute to the degree of immune in ltration, we sought the consensus clustering analysis to stratify 374 samples, which were divided into 2 groups according to optimal cluster number (k = 2) (Fig. 1B). Then ssGSEA was used and we found that the immune cluster 1(ICR1), which was de ned as the high immune in ltration, was linked to activated B cell, activated CD4 T cell, activated CD8 T cell, activated dendritic T cell, century memory CD8 T cell, effector memory CD4 T cell, effector memory CD8 T cell, Gamma delta T cell, Immature B cell, Immature dendritic cell and macrophage compared with the ICR2 group ( Fig. 1C, P < 0.001). Estimate scores, immune scores and stromal scores were procured via applying ESTIMATE algorithm, and they were higher in ICR1 compared with ICR2 ( Fig. 1D, P < 0.001). Also, patients' overall survival (OS) in ICR1 group was signi cantly reduced compared to that of patients in ICR2 group, as visualized by Kaplan-Meier survival curves (Fig. 1E, P = 0.0079). Together, these data revealed that higher immune in ltration level was positively correlated with poor prognosis.

Comparisons of mutation pro les in HCC
The somatic mutation pro les of 74 HCC samples from ICR1 and 285 HCC samples from ICR2 were downloaded from TCGA. Then the results based on the mutation data were visualized in VCF format utilizing the "maftools" package. Mutation information for each gene in each sample was shown within waterfall plot, where colored annotations at the bottom signi ed different types of mutation ( Fig. 2A, B). Interestingly, 94.59% (70 of 74) of patients in the ICR1 group had mutations, while 82.81% (236 of 285) of patients in the ICR2 group had mutations. And the mutation rate of TP53 in the ICR1 group was 50%, while in the ICR2 group it was 25%. Thus these ndings indicate that different immune in ltration levels are associated with mutations in HCC.

The Correlation between the immune in ltration and Somatic Variants in HCC
Statistical data analyses were all implemented and generated through R software 4.0.3, all the plots were generated by R package "ggplot2" [31]. The value of all experiments were quanti ed by mean ± SD. The correlation coe cient between two groups (BCL10/TRAF3) of HCC and pan-cancer samples in TCGA was generated by Pearson's correlation analysis. P values less than 0.05 was considered statistically signi cant.
Tumor mutation burden (TMB) is de ned as the amount of somatic gene coding errors, base substitutions, gene insertion or deletion errors detected per megabase in total. To further understand the relevance of immune in ltration degrees and mutation, we then assessed the TMB of the ICR1 and ICR2 group and we found The TMB is higher in ICR1 group than ICR2 (Fig. 3A, P < 0.001). The Estimate score was subsequently veri ed to be positively related to the TMB in ICR1 group by using correlation analyses (Spearman coe cient: R 2 = 0.06, p = 0.0037),while have no obvious correlation in ICR2 group (Spearman coe cient: R 2 = 0.02, p = 0.262, Fig. 3B). Next, patients' overall survival (OS) in high TMB group was signi cantly lower than that of patients in low TMB group revealed by Kaplan-Meier survival curves (Fig. 3C, P < 0.001). At the same time, considering that the survival probability was lower in the ICR1 group compared to that of the ICR2 group in result 1, so we used the Kaplan-Meier survival curves and displayed great differences of survival in subgroups ( Fig. 3D). Thus these ndings indicate that the high immune in ltration level might correlated with TMB and could synergize with high TMB level to predict poor prognosis.

Differential expressed genes with immune and TMB score in HCC
In search of the differential expressed genes (DEGs) contribute to different immune in ltration level and TMB level. The "Limma" package was performed, and 1424 DEGs were selected in the high or low TMB score group, while 2181 DEGs were selected in the ICR1 or ICR2 group. There are also 369 DEGs in these two groups at the same time (Fig. 4A). In addition, GO functional enrichment (BP, MF and CC) (Fig. 4B-D) and KEGG pathway analyses (Fig. 4E) was conducted in 369 co-expressed DEGs. Interestingly, these genes were enriched in Positive regulation of macrophage colony − stimulating factor stimulus pathway, B cell apoptotic process pathway, T cell receptor complex pathway, B cell receptor signaling pathway, T cell receptor signaling pathway. These ndings suggested that these genes may function as underlying predictive indicators which can response to immunotherapy in HCC. Meanwhile, univariate Cox regression analysis of 89 signi cant DEGs (P < 0.05) from the 369 DEGs was performed in a forest plot (Fig. 4F-H).

Construction of the combined prognostic risk signature in HCC.
Then, a generalized linear model named LASSO algorithm was carried out to build the prognostic signature. The log 2 transformation of the lambda (λ) value generated a coe cient pro le plot, which was determined when the likelihood deviance was minimum (Fig. 5A). Nineteen DEGs were included and the formula was generated: = -0.043 GHR expression − 0.005 RORC expression − 0.184 CHP1 expression + 0.066 HBEGF expression + 0.004 CXCL8 expression + 0.081 PGF expression + 0.027 SPP1 expression + 0.083 TXLNA expression + 0.010 CSF3R expression − 0.001 ACKR1 expression − 0.176 ENG expression + 0.138 IL15RA expression − 0.088 CD8A expression − 0.143 ZAP70 expression + 0.056 BCL10 expression − 0.043 CD79A expression + 0.006 CD209 expression + 0.199 TRAF3 expression + 0.317 SLC29A3 expression. Also, we calculating the corresponding coe cients using the minimum 10-fold cross-validation mean square error in HCC (Fig. 5B). Each patient's risk score = ∑ gene expression * coe cient (glmnet R package). The results showed that compared with the surviving patients, dead patients' risk score was markedly higher (Fig. 5C, P < 0.001). In order to judge the accuracy of the prediction results, we then generated the ROC, which was 0.722 (Fig. 5D, p < 0.001). Then age, gender, T stage, N stage, M stage and risk score were all included to establish a nomogram for predicting 1-, 2-, 3-year OS. Experimental results demonstrated that T stage contributed the most to prognosis, secondly was the risk score ( Fig. 5E). To evaluate the constructed prognostic model, we worked out that the C-index of our nomogram was 0.736, and Fig. 5F presented the DCA plots of our nomogram. Besides, to evaluate the calibration of the nomogram, we utilized 1000 resamples boot-strapping procedure and obtained revised estimates of predicted and actual values. The calibration graphs showed remarkable accordance between predicted and actual survival rates (Fig. 5G). All these results substantiated the excellent clinical value of the nomogram compared to the risk score in predicting OS.

Expression and clinical correlation of BCL10 and TRAF3 in HCC
To nd whether the 19 DEGs were correlated with TP53 mutation, we analyzed the data from TCGA LIHC dataset and discovered the higher BCL10 and TRAF3 expression levels in TP53 mutation group than that in the TP53 wild group (Fig. 6A-B, p < 0.001). Then the scatter and box plot revealed that BCL10 and TRAF3 expression were signi cantly elevated in HCC tissues in the TCGA LIHC dataset (Fig. 6C-D, p = 0.011; p < 0.001) and AHNTU cohorts (Fig. 6E-F, p < 0.001; p < 0.01). To further investigate the BCL10 and TRAF3 protein expression, IHC was performed to reveal the staining distribution in HCC and adjacent tissues (Fig. 6G-H). Mining the clinical data of 234 HCC patients from TCGA, the high BCL10 level was corelated with the status (p = 0.008), Ajcc stage (p = 0.0075) and T stage (p = 0.0076, Fig. 6I), while the TRAF3 expression was associated with stage (p = 0.044, Fig. 6J).

Functional enrichment analysis and correlation of BCL10 and TRAF3
For the sake of understanding the potential characteristics of BCL10 and TRAF3. GSEA analysis of each hub gene with the TCGA LIHC dataset was performed. The samples (n = 374) were partitioned into 2 groups on the basis of the median expression values of BCL10 and TRAF3. BCL10 and TRAF3 were all enriched in functions: apoptosis, IL-2 STAT5 signaling, P53 signaling, PI3K-AKT-MTOR signaling, TGF-β signaling, TNFα signaling via NF-κB and in ammatory response, while negatively enriched in Bile Acid and Fatty Acid metabolism (Fig. 7A-B). Then, to validate the relationship of BCL10 and TRAF3 levels with the immune microenvironment in-depth, the ratio of immune in ltration subsets was analyzed using ssgsea and we found BCL10 and TRAF3 were all correlated with activated CD4 T cell, central memory CD4 T cell, type 2 T helper cell, natural killer T cell, effector memory CD4 T cell, natural killer cell, central memory CD8 T cell, plasmacytoid dendritic cell, immature B cell, T follicular helper cell, MDSC, neutrophil, activated CD8 T cell, and eosinophil cells (p < 0.05; Fig. 7C-D). BCL10 and TRAF3 enrich in various pathways in GSEA analysis hinted that there is a potential connection and correlation between BCL10 and TRAF3. Then the pearson test was used and we found BCL10 is signi cantly correlated with TRAF3 in HCC (r = 0.65, p < 0.0001, Fig. 7E). Further, BCL10 was signi cantly correlated with TRAF3 in pan cancers (r = 0.53, p < 0.0001, Fig. 7F), while Fig. 7G showed TRAF3 is correlated with BCL10 in 31 types of cancers in TCGA. Together, all these results showed TRAF3 and BCL10 can corporately participate in various biological functions and pathways, thus in uencing the tumor immune microenvironment.

Discussion
Despite tremendous progresses have been made in the screen, diagnostic and therapeutic techniques and prevention for HCC, the treatment of HCC remains a bottleneck. HCC still has a highly fatality rate, as evidenced by the low 5-year survival rate [32]. Emerging studies have highlighted that the TME contributes fundamentally to initiate and maintain the pathogenesis of HCC [33]. The intricate interaction of hepatoma cells with the TME plays a critical role in regulating HCC behavior [33]. Apart from provide structural support for tumor cells, TME exert pro-tumorigenic effect via interacting dynamically with malignant cells, thus in uences HCC development, metastasis and prognosis by offering inhibitory or stimulatory growth signals [34]. In this regard, treatments targeting TME and its crosstalk with hepatoma cells are underway. In addition, immune checkpoints, for instance, cytotoxic T lymphocyte protein 4 (CTLA-4), program death protein 1 (PD-1) and its ligand PD-L1 expressed on the surface of immune cells, are increased in HCC. They are associated with poorer survival by working as immunosuppressive factors in HCC microenvironment to prevent uncontrolled immune responses. Considering the immune-rich TME in HCC, immune-based therapies are currently studied extensively [35]. Two forms of immunotherapy against HCC have exhibited remarkable clinical outcomes in HCC landscape: immune checkpoint inhibitors (ICIs) which stimulating pre-existent anti-tumor immunity and monoclonal antibodies that directly induce T-cell dysfunction called exhaustion [36]. Our ndings veri ed that different levels of TME have close correlations with the prognosis in HCC patients.
Thanks to advances in large-scale mutational deep-sequencing technologies, many scientists have been able to picture the mutational landscape and identify driver mutations in several solid tumors such as hepatocellular carcinoma [37]. Accumulations of genetic mutations could result in the activation of cancer-associated genes (CAGE) and contribute to hepatocarcinogenesis as evidenced by invasive tumor behaviors, such as metastatic abilities and resistance to anti-tumor drugs [38]. It has been well depicted that TP53 and CTNNB1 are the most commonly mutated CAGE in HCC and mutations of these genes are mutually independent [39]. TP53 ranked among the most frequently mutations across human sporadic tumors con rmed by genomic analysis, such as colorectal, ovarian and triple-negative breast cancer [40].
TP53 serve as "a guardian of the genome" and tumor suppressor because its role in preserving genetic stability via regulating the transcription of target genes and resulting in cell-cycle inhibition, cell aging and apoptosis in answer to DNA damage. Mutations of TP53 leading to loss-of-function of wild-type TP53 and predisposing to environmental carcinogens [41]. TP53 mutations promote oncogenesis and are related to unfavorable therapeutic responses and, hence, arise adverse prognosis in several cancer types [42]. Interestingly, the mutation rate of TP53 is very low in healthy people, while it is up to 30% in HCC as corroborated by the TCGA database. In HCC, the existence of TP53 mutations are signi cantly associated with tumor differentiation, tumor grade, Child-Pugh classi cation and serum AFP levels. HCC patients with TP53 mutations have poor clinical prognostic factors such as disease-free survival (DFS) and overall survival (OS), as well as high malignant behaviors and mortality of HCC [43]. Our study discovered that the overall mutation levels between two groups divided by different immune microenvironmental scores were alternated, especially TP53 mutations. According to previous analysis of the TMB impact on the survival of primary solid tumors from the TCGA database, compared with low TMB, patients with high TMB showed poorer prognosis in several cancers, and HCC is also included. Because TMB is reportedly correlated with survival rate in response to immunotherapies which aimed to kill malignant tumor cells in several cancers, and the immune microenvironment (IME) is critical for prediction of immune responses [44]. Therefore, the obscure correlation between TMB and IME in HCC requires to be explained. K-M and correlation analysis in our study uncovered for the rst time that TMB is positively correlated with IME, higher levels of TMB and IME was related to worse prognosis and they predict clinical poor outcomes in HCC collaboratively.
We identi ed differentially expressed genes (DEGs) of HCC patients from the TCGA LIHC database on the basis of different immune in ltration and TMB levels. Additionally, functional enrichment and pathway analysis unveiled the immune-related functions of these DEGs. We further screened 89 genes associated with prognosis by Cox regression analysis. Prognostic models of 19 DGEs were also established using Lasso regression algorithm. We next generated a risk score models based on 19 hub DEGs of surviving and died patients in HCC cohorts. Finally, combining TNM staging and risk scores, a prognostic nomogram was constructed for prediction of overall survival and found that the nomogram has higher application value than the risk score model. Furthermore, AUC, C-index, DCA and calibration plot was used to demonstrate that the e ciency of constructed prognostic model was relatively good. However, before in-depth clinical practice, the prognostic data from a larger cohort is indispensable to verify these models. Besides, it is noteworthy that speci c TME estimation models are essential due to the mutational environments vary among different tumor types.
B-Cell Lymphoma/Leukemia 10 (BCL10) gene is a member of BCL family, it was rst detected in a translocation of mucosa-associated lymphoid tissue (MALT) lymphoma. Protein encoded by BCL10 (wildtype BCL10) could induce apoptosis and synergically forming a complex (CBM signalosome) with Caspase Recruitment Domain Family Member 11 (CRAD11) and its substrate, the MALT protein, thus initiating its downstream signaling called nuclear factor-kappa B (NF-κB). It also serves as a critical adaptor in adaptive immune responses for maintaining immune homeostasis. What's more, the CBM complex exert both cell-intrinsic and cell-extrinsic effects via in uencing cells in the TME [45]. Previous studies demonstrated that mutated BCL10 still activated NF-κB but lost proapoptotic function and exhibited oncogenic activity. The dysfunction of BCL10 may lead to immunode ciency diseases and malignancies, such as lymphoma, which mainly affects the gastric mucosa and promotes it towards a higher grade by t (1;18) (p21; q21) translocation [45,46]. Such mutations of BCL10 were discovered in other solid tumors, which suggest that BCL10 mutation may be involved in malignant transformation of human tumors. TNF Receptor Associated Factor 3 (TRAF3) gene belongs to the TRAF3 family. TRAF3 proteins interact with CD40 (TNFRSF), which is a member of TNF receptor (TNFR) family, exerted effects on activating T-cell dependent immune responses mainly by positively modulating regulatory T cells (Treg cells), consequently regulating NF-κB signal transduction and cell death [47]. Treg cells is responsible for cancer progression. In addition, TRAF3 is mainly expressed in B cells and restraining its survival.
Inactivation mutations of TRAF3 results in NF-κB signaling activation and subsequent B cells survival, which enhances the tendency of malignant neoplasia [48]. TRAF3 mutations are closely related to human diseases including multiple myeloma, macroglobulinemia and cutaneous herpes infection [49]. Our study discovered for the rst time that BCL10 and TRAF3 were upregulated in HCC and were higher in TP53 mutation group. What's more, TCGA database further indicated that they were associated with clinical stages and grades. However, we need to verify with in a larger sample in subsequent trails. Moreover, we explored the functions of BCL10 and TRAF3 respectively through GSEA enrichment analysis and found several co-existing functions. Considering that BCL10 and TRAF3 both associated with immune responses, ssgsea was performed and revealed that they were associated with numerous immune cells. We further uncovered the positive correlation between BCL10 and TRAF3. Therefore, we concluded that the upregulation of BCL10 and TRAF3 may affect the progression and prognosis of HCC through modulating TMB and tumor immune microenvironment.

Conclusions
Our study described that higher immune in ltration levels have associations with TMB, and they corporately predicting poor prognosis. Our study may contribute to the unveiling of predictive roles of BCL10 and TRAF3 expression in evaluating hepatic carcinogenesis and clinical prognosis for patients with HCC, which may function by regulating TMB and tumor immune microenvironment.      Construction of the combined prognostic risk pro le for HCC. a A coe cient pro le plot determined by the minimum likelihood deviance was generated. b Nineteen DEGs and corresponding coe cients were exhibited by cross-validation. c The risk score of dead and alive patients. d Receiver operating characteristic (ROC) curve was generated for evaluating the Prediction accuracy. The AUC index of ROC curve was shown in the bracket (0.722). e On the basis of age, gender, TNM stage and risk score, a nomogram was constructed to estimate 1-, 2-, 3-year survival rate for HCC. f The DCA plots re ected the net bene t of the nomogram. g Calibration plots further veri ed the predictive ability of our model.

Figure 6
Page 23 /24 Expression and clinical correlation of BCL10 and TRAF3 in HCC. a, b Differential BCL10 and TRAF3 expression in the TP53 mutation and wild group using TCGA LIHC dataset. c, d, e, f Differential BCL10 and TRAF3 expression in HCC and adjacent normal tissues based on TCGA LIHC and AHNTU cohorts. g, h IHC staining of BCL10 and TRAF3 expression in HCC and adjacent normal tissues. i, j Associations of high BCL10 and TRAF3 expression with the status, Ajcc stage and TMN stage. Figure 7