INTS8 Predicts a Therapeutic Target for Intrahepatic Cholangiocarcinoma via Integration of Bioinformation Analysis and Experimental Validation

Background & Aims: Intrahepatic cholangiocarcinoma (ICC) remains a rare malignancy, ranking the leading lethal primary liver cancer worldwide. However, biological functions of integrator complex subunit 8 (INTS8) remain unknown in ICC. Thus, this research aimed to explore the potential role of INTS8, as well as develop a novel diagnosed or therapeutic target in ICC. Methods: Differently expressed genes in two GEO datasets were obtained by RRA package in R software. The “maftools” package was implemented to visualize the cholangiocarcinoma (CHOL) mutation data in TCGA database. The expression of INTS8 was detected by preforming quantitative reverse transcription-PCR (qRT-PCR) and immunohistochemistry in cell lines and human samples, respectively. The association between subtypes of tumor-infiltrating immune cells (TICs) and INTS8 expression in CHOL was performed by using CIBERSORT tools. We evaluated a correlation between INTS8 expression and mismatch repair genes (MMRs) and DNA methyltransferases in pan-cancer analyses. Finally, the pan-cancer prognostic signatures of INTS8 were identified by univariate analyses. Results: We obtained the mutation landscape of a RRA gene set in CHOL. The expression of INTS8 was actually upregulated in ICC cell lines and human ICC samples. Furthermore, INTS8 expression is tightly associated with distinct landscape of TICs, the MMRs, and DNA methyltransferase s in CHOL. In addition, the high INTS8 expression group presented a significantly poor outcomes, including overall survival (OS), disease-specific survival (DSS) and disease-free interval (DFI) ( P < 0.05) in multi-cancers. Conclusions: INTS8 contributes to the tumorigenesis and progression of ICC. Our study highlights the significant roles of INTS8 in ICC and pan-cancers, providing a valuable molecular target for cancer research.


Introduction
As one of the hepatobiliary malignancy subtypes, intrahepatic cholangiocarcinoma (ICC) has attracted more attentions due to its increasing global incidence and trends [1] , with the largest age-standardized incidence rates increased in China (average annual percent change: 11.1%) for ICC from 1993 to 2012 [2] . Moreover, the mortality rate from ICC showed a rising trend globally, which was around 1-2/100,000 in most countries [3] . Attributing to the aggressive and asymptomatic features of ICC, many patients are diagnosed at an advanced stage. Surgical resection is regarded as the best treatment strategy for achieving a good prognosis and long survival. However, a 5-year overall survival rate remains limited to 22%-30% after curative hepatectomy due to high recurrence rates [4][5] . Moreover, 40-80% recurrence rate was reported for ICC patients after surgical resection [6] . The combination of gemcitabine and cisplatin was regarded as the standard-chemotherapy regimen, despite showing limited effectiveness for ICC. Therefore, it is urgent to improve the sensitivity of diagnose and effectiveness of targeted therapy for ICC.
Integrator is reported as a transcriptional regulatory complex, which is comprised by at least 14 subunits [7] . Integrator complex subunit 8 (INTS8) serves as one of the major components of RNA polymerase II, has been demonstrated to involve in the cleavage of small nuclear RNAs and transcriptional processes [8][9] . A recent study found that INTS8 was essential to transcription repression, which was induced by recruiting protein phosphatase 2A to prevent transcription elongation and promote transcription termination [10] . Previous study revealed that the INTS8 was robustly increased in neurodevelopmental diseases [11] and numerous tumors [12][13] . Overexpressed INTS8 could facilitate the epithelial-to-mesenchymal transition, which was mediated by TGFβ signaling pathway in hepatocellular carcinoma [14] . However, studies concentrated on the effect of INTS8 in ICC are generally lacking.
In the present study, we performed robust rank aggregation (RRA) method to select differently expressed genes (DEGs) based on Gene Expression Omni (GEO) database. Then, we explored genes at an intersection between DEGs and gene mutation profiles in cholangiocarcinoma cohort (CHOL) of the cancer genome atlas (TCGA), and identified INTS8 as a candidate gene. We verified the overexpression of INTS8 in ICC cell lines and human ICC samples by quantitative reverse transcription-PCR (qRT-PCR) and immunohistochemistry (IHC), respectively. Our study showed that the high INTS8 expression has a tight correlation with poor prognosis in multi-cancers.
Moreover, the underlying mechanism may attribute to tumor-infiltrating immune cells (TICs), the mismatch repair genes (MMRs), and DNA methyltransferases. Therefore, INTS8 was identified to be a therapeutic target for in ICC and pan-caner, which was related to the TICs, and mediated by MMRs and DNA methyltransferases.
GSE26566 included 104 cholangiocarcinoma samples and 6 matched surrounding samples. While GSE32225 contained 149 cholangiocarcinoma samples and 6 matched surrounding samples. All expression profiles were downloaded and processed by "GEOquery" package (www.r-project.org). Considering the differences and batch effects of different platforms, we utilized the "sva" package [17] to avoid these effects and remove other unwanted variation. In addition, transcriptome profiling, mutation and methylation data, as well as clinical information of tumor samples and corresponding samples were obtained from the TCGA database (https://cancergenome.nih.gov/), which were analyzed by utilizing the "TCGA biolinks" package [18] . Transcripts per million (TPM) values were applied for subsequent analyses [19] .

DEGs and RRA Analysis
The DEGs were determined between cholangiocarcinoma samples and matched surrounding samples by using the "limma" package [20] . Robust rank aggregation was performed for gene list integration [21] , and the score < 0.05 was regarded to determine RRA gene set. These data were visualized by heatmap and volcano plot.

Identification of INTS8 by mutation analysis and ROC curves
We administrated the "maftools" package [22] to visualize the CHOL mutation data.
To assess mutated genes presented in RRA gene set, we obtained the intersection of RRA genes and CHOL mutation genes. To evaluate the diagnostic performance of muted RRA genes, ROC curves was performed by using pROC. Next, we selected optimal efficacy indicator for the further research according to the areas under the curve (AUCs). The patients were stratified into two groups according to the median expression of INTS8. DEGs between high/low INTS8 groups were confirmed by using "Limma" package. The cutoffs for the DEGs were as follows: |Log2FC| > 1, and false discovery rate (FDR) < 0.05. A heat map by using ggplot2 packages was used to visualize the DEGs.

Functional enrichment of mutated RRA genes and INTS8 related genes
To identify possible pathways and biological functions of 5 mutated RRA genes and differential INTS8-associated genes, we applied the "clusterProfiler" package [23]

CIBERSORT estimation
In order to characterize the tumor microenvironment, we applied "CIBERSORT" (R package) (http://cibersort.stanford.edu/) [24] to explore the relative proportions and absolute fraction score of 22 subtypes of TIICs in CHOL tissues. Moreover, the association between the INTS8 expression levels and the infiltration of TIICs was also assessed and visualized by heatmaps and violin plots.

The association of INTS8 gene expression with clinical outcome in different tumors
To explore the influence of expression level of INTS8, we carried out analyses by publicly available database. we retrieved data from The Cancer Cell Line Encyclopedia (CCLE, https://portals.broadinstitute.org/ccle) [25] and The Genotype-Tissue Expression (GTEx) [26]

The identification of the robust DEGs in GEO
Based on the results of DEGs, a total of 710 up-regulated and 903 down-regulated significant DEGs were confirmed in GSE26566, and 432 up-regulated and 566 downregulated significant DEGs were found in GSE32225. The DEGs were shown by heatmaps and volcano plots ( Figure 1A-D). Furthermore, these DEGs were integrated by RRA method with the score < 0.05. And then RRA gene set was visualized by a heatmap ( Figure 1E). In a result, a RRA gene set was obtained for further investigation.  Figure 2B). Our result showed that the RRA gene set was associated with some metabolic pathways.

Mutation landscape of RRA gene set in CHOL
To identify mutational landscape in CHOL patients, the "maftools" package in R software was implemented. Missense mutations were the predominate type of mutation in patients with CHOL ( Figure 2A). Single nucleotide polymorphism had a more frequent occurrence compared with insertion or deletion ( Figure 2B). Particularly, C >T remained the most common mutation type of single nucleotide variants in CHOL ( Figure 2C). Moreover, the mutation types were displayed for CHOL ( Figure 2D-E).

Identification of DEGs between high and low INTS8 expression groups
The ROC analysis was applied to determine the diagnostic efficacy of the 5 mutated genes from RRA gene set. Moreover, the INTS8 had the highest AUC value To explore the underlying mechanism of INTS8 in ICC, patients were divided into two groups according to median expression value of INTS8. DEGs were displayed between high and low INTS8 expression groups ( Figure 3B). Furthermore, we found the mRNA expression of INTS8 was actually upregulated in 3 ICC cell lines compared with HIBEC in vitro ( Figure 3C). Meanwhile, the protein levels of INTS8 by using IHC were also verified to be obviously increased in ICC patients tissue compared with the normal samples ( Figure 3D). The experimental results were consistent with that of the bioinformatic analysis.

Functional enrichment of INTS8 in CHOL
To identify the biological function and key candidate pathways of the INTS8 related genes, we performed GO and KEGG analysis. As shown in Figure

Association between tumor-infiltrating immune cells and INTS8 expression in CHOL
TIICs significantly impact the development and progression of many types of cancers, including CHOL. By applying CIBERSORT tools, we observed a high level of M0 macrophages, M2 macrophages, Monocytes, and resting CD4+ memory T cells, while a lower level of activated Dendritic cells, Eosinophils, Neutrophils and activated CD4+ memory T cells in CHOL ( Figure 5A-B). Moreover, we assessed the relationship between TIIC and INTS8 expression in CHOL. We found that high INTS8 expression group presented a unique TIICs landscape, including statistically significantly high level of M0 macrophages but low level of M2 macrophages, elevated resting CD4+ memory T cells but low level of CD4 native T cells, increased resting mast cell but low activated mast cell. In addition, low expression of gamma delta T cells and monocytes were also found in high INTS8 expression group ( Figure 5C-D).

The INTS8 expression in multiple dimensions
Considering the extensive mutational heterogeneity of cancers, we systematically performed the large-scale profiling of INTS8 expression in 21 cell lines and 31 related tissues based on CCLE and GTEx. As shown in Figure 6A-B, the expression levels of INTS8 in diverse cancer tissues and cell lines were elevated to differing degrees, including biliary tract, liver, bone marrow. In addition, we also found INTS8 harbored most prevalent mutations such as missense, truncating and fusion in different tumors ( Figure 6C).

The association between INTS8 and clinicopathologic characteristics and survival information
As show in  Figure 3-5). These findings suggested that INTS8 expression showed the potent potential prognostic biomarker for pan-cancers.

The MMRs and DNA methylation involved in CHOL
Increasing evidence has demonstrated that MMRs play an important role in maintaining genomic stability, and DNA methylation regulates gene expression. The loss of key gene function of MMR could induce DNA replication errors, resulting in high level of somatic mutations. It has been reported that MMR pathway was potently activated during G1/S-phase [27] .  Figure 7A).
DNA methylation serves as an epigenetic modification, which could regulate gene expression. Attributed to the function of DNA methyltransferase, DNA methylation was processed when the covalent bonding of a methyl group to the 5'carbon position of the cytosine in genome CpG dinucleotide [28] . As shown in Figure 7B, an epigenetic signature was discovered, and showed a high correlation between INTS8 and DNA

Discussion
ICC is an extremely aggressive biliary neoplasm with increasing incidence and poor prognosis worldwide [29] . The current diagnostic and prognostic marker for ICC remained unsatisfactory. In this study, TCGA and GEO database were applied to systematically analyze the mutational status of RRA genes in ICC, and 5 mutant genes were found by intersection analysis. Based on the diagnosis efficacy of the 5 mutant genes, we selected the INTS8 with the largest AUC value for follow-up research, which showed a significant role in ICC even across pan-cancer.
Various evidence suggested that the integrator complex played an essential role in the aspect of RNA processing and transcription regulation. Previous studies have showed that INTS8 mutation could induce a severe neurodevelopmental syndrome [11] , and multi-cancers [30] . In this study, we described that INTS8 was significantly overexpressed in ICC compared to normal samples, which was consistent with the results of IHC and PCR. Our results showed that overexpressed INTS8 was positively related to the poor prognosis in many tumor types.
The GO enrichment analyses showed that high INTS8 expression was mainly associated with organic anion transport, organic acid transport, carboxylic acid transport and acute inflammation response. In addition, retinol metabolism, chemical carcinogenesis, drug metabolism-cytochrome P450, metabolism of xenobiotics, drug metabolism-other enzymes, fatty acids degradation were most significantly enriched in CHOL patients with high INTS8 expression than low INTS8 expression group. Retinol belongs to fat-soluble nutrients, which is essential to maintain physiological functions in many tissues [31] . Retinol metabolism abnormality caused by genetic or environmental factors could induce developmental pathologies, including mammalian placental and embryonic development [32] , ovary disease [31] and fatty liver disease [33] .
Previous study showed that the administration of retinol facilitated hepatocarcinogenesis development during its early stages [34] . Cytochrome P450 had a vital role in mediating hepatic metabolism, especially its clinical significance in the metabolism of some opioid and hypnotic drugs [35] . Drug metabolism-cytochrome P450 was related to DNA methylation-driven genes in prostate adenocarcinoma [36] . In addition, previous data showed that hepatic cytochrome P450 (CYP) family enzymes, especially increased CYP2A6 and diminished CYP2E1 might participate the progression of ICC [37] . Lipid metabolism is newly recognized as a hallmark of cancer, and inhibiting fatty acid availability could control the development of malignancy [38][39] . Li et al. found that ICC tumorigenesis was insensitive to FASN deprivation, which contributed to high expression of fatty acid uptake and resulted in rapid tumor growth.
Therefore, promoting fatty acid degradation may be a novel therapeutic approach for the ICC treatment options [40] .
DNA damage and repair provide protection for mutation avoidance, which play central roles in maintaining genome stability [41][42] . Up to date, it has been reported that 4 major DNA repair pathways involved in maintaining the gene expression, including nucleotide excision repair, base excision repair, MMR, and double strand break repair [43] . To estimate an association of INTS8 and MMRs, we assayed an association of INTS8 and MSH2, MSH6, PMS2, and EPCAM. The expression of INTS8 was positively correlated to MSH2, MSH6, PMS2, but not associated with MLH1 and EPCAM. By using the immunohistochemical analysis [44] , the results showed that there is no loss of the expression of DNA repair enzymes/MMR proteins (MLH1, MSH2, PMS2, and MSH6) in both Occupational cholangiocarcinoma (CCA) [45] and cohorts with cholangiocarcinoma [46] .
MMR gene mutations or tumor MLH1 promoter methylation was concerned as a main cause of microsatellite instability (MSI) in patients with Colorectal cancer (CRC) [47] . After analyzing 96 cases of intra-and extrahepatic cholangiocarcinomas for MMR status by using IHC, Jennifer Y Ju et al. found that 6% of samples presented with MMR deficiency [48] . Although microsatellite instability relatively frequently occurred in in intraductal papillary neoplasms of the biliary tract, the methylation of hMLH1 was not detected [49] . EpCAM, a stemness-related marker, is positively with the aggressive prognosis in ICC and HCC [50][51] . However, we didn't observe an association between

INTS8 and EpCAM in ICC.
Recently, Epigenetic alterations is characterized by any heritable modification of chromatin DNA or histone proteins but without change in the DNA sequence [52][53] , which can be observed in many human cancers, and cooperate with genetic alterations to dominate the formation of cancers [54] . DNA methylation is one of the main presence of epigenetic changes, which is specifically mediated by the DNA methyltransferase (DNMT) family (including DNMT1, DNMT1, DNMT3A and DNMT3B) [55] . DNMTs could establish and maintain DNA methylation patterns, which induced gene silencing, transcriptional activation and post-transcriptional regulation mediated by DNMT2- discovered the γδ T cell-based immunotherapy showed no adverse effects, and could positively regulated peripheral immune functions for the patient with ICC [57] . we identified low gamma delta T cells level was present in high INTS8 expression group in ICC. Based on this promising finding, INTS8 could be developed into a promising therapy for ICC. Previous studies proved that CD4 regulatory T cells is prominent immunosuppressive characteristics in ICC [58] . Although Tregs was relatively high in ICC, but had no association with INTS8.