Clinical Value and Potential Mechanisms Exploration of GOLGA8B in Hepatocellular Carcinoma: A Comprehensive Investigation Based on Real-Time Quantitative Polymerase Chain Reaction, RNA-seq, Microarray and Immunohistochemistry

Introduction Recent studies found that GOLGA8B plays an essential role in different cancers. However, the role GOLGA8B plays in the carcinogenesis and development of hepatocellular carcinoma (HCC) remains unclear. This study explores the clinical signicance and prospective mechanisms of GOLGA8B in HCC. Materials and methods The expression of GOLGA8B was detected in tissues of HCC and non-HCC controls. A real-time quantitative polymerase chain reaction analysis was performed to evaluate the mRNA expression of GOLGA8B. RNA-sequencing data and microarray chip data were obtained for further analysis. The role GOLGA8B plays in patients with HCC was also evaluated. An immunohistochemistry (IHC) analysis was also performed to evaluate the protein expression of GOLGA8B. The different GOLGA8B expression resources, including mRNA and protein expression, were integrated by calculating standard mean difference (SMD) and summary of the receiver operator characteristic (sROC). Genes coexpressed GOLGA8B were predicted. Enrichment analyses including Gene ontology (GO) and biological pathway were performed to investigate the essential molecular mechanisms. Hub genes were screened out by a protein-protein interactions network. MicroRNAs which target GOLGA8B at a posttranscriptional level were also predicted. Results According to different resources, GOLGA8B manifested a higher expression in tissues of HCC than in non-HCC controls and exhibited clinical values for HCC. The RT-qPCR analysis revealed an increasing trend of GOLGA8B expression in HCC. GOLGA8B expresssion was signicantly increased in RNA-sequencing and 7 of 13 microarray chip. IHC analysis also revealed signicantly higher expression of GOLGA8B protein. Moreover, GOLGA8B expression was correlated with pathologic tumors and stages according to RNA-sequencing data and IHC analysis. The integrated SMD and sROC of different resources was 0.893 (P=0.004) and 0.79. A total 1303 co-expressed genes were gathered to perform enrichment analyses. The most signicant biological pathways of co-expressed genes were spliceosome and mitogen-activated protein kinase signalling (MAPK). Four hub genes (SF3B1,

Golgin A8 family member B (GOLGA8B) is located at 15q14. The protein it encodes is a highly coiled-coil protein called golgin-67, with a molecular weight of 67 kDa. GOLGA8B belongs to the golgin family and functions as maintaining structural integrity of the Golgi complex [6][7][8][9]. Previous studies have demonstrated the relationships between golgins and variant cancers [10][11][12]. As for GOLGA8B, Baine et al. (2011) found that its expression level was decreased between pancreatic cancer patients and healthy controls through the whole genome cDNA microarray analysis [13]. Sugarbaker et al. (2008) found that GOLGA8B had mutations in the malignant pleural mesothelioma tumours [14]. However, the role GOLGA8B plays in cancers still remains unknown, and there are no any relevant studies about GOLGA8B and HCC.
We hope this study will clear the potential molecular mechanisms and contribute to the clinical values of GOLGA8B. To achieve this goal, resected tissues from HCC cases were obtained from the Department of Pathology, First A liated Hospital of Guangxi Medical University. Real-time quantitative polymerase chain reaction (RT-qPCR) analysis, together with data excavation of different databases, was used to verify the mRNA expression of GOLGA8B. Sequencing data obtained from the Cancer Genome Atlas (TCGA) and Oncomine and microarray chip data obtained from Gene Expression Omnibus (GEO) and ArrayExpress were integrated for further analyses. Gene Expression Pro ling Interactive Analysis (GEPIA) was used to perform the survival analysis. We also conducted the immunohistochemistry (IHC) analysis to detect the protein expression level and clinical signi cance of GOLGA8B. Furthermore, coexpressed genes of GOLGA8B were chosen to perform enrichment analyses-including gene ontology (GO) enrichment, biological pathway and the protein-protein interactions (PPI) network analysis to investigate potential pathways and hub genes. MicroRNAs (miRNAs) which may target GOLGA8B were also predicted.
In summary, based on the detection of mRNA and protein expression, our study used multiple methods and a large sample size to evaluate the expression and clinical value of GOLGA8B. We hope this study may discover the clinical signi cance of GOLGA8B and clarify its underlying molecular mechanism. The owchart of this study is shown in Figure 1.

Clinical Samples Collection
We collected formalin-xed, para n-embedded clinical samples from the Department of Pathology, First To identify the expression pattern of GOLGA8B in HCC cells, 282 HCC tissues, 60 para-carcinoma tissues and 26 normal liver tissues were utilised to perform IHC analysis. The GOLGA8B antibody provided by Santa Cruz Biotechnology (Heidelberg, Germany) was utilised for IHC detection.

Statistical Analysis
Statistical Product and Service Solutions 22.0 (SPSS 22.0) was applied to analyse the data we obtained. The expression differences between different groups were analysed by independent sample t-test or oneway analysis of variance (ANOVA). Mean ± standard deviation (SD) was adopted to present all the data.
The area under the curve (AUC) was calculated by plotting receiver operating characteristic (ROC) curves. Stata statistical software version 12.0 (StataCorp, Colleges Station, TX, USA) was used to integrate the results of different resources, which were accessed by the standard mean difference (SMD). The heterogeneity of microarray datasets was assessed by the I2 statistic and Q test. We pooled the SMD by xed effects when small heterogeneity existed (P ≥ 0.05, I2 < 50%); otherwise, a random-effect model would have been chosen. Egger's and Begg's tests were adopted to estimate publication bias. To access the diagnostic value of GOLGA8B, we obtained the summary of the receiver operator characteristic (sROC), diagnostic odds ratio (DOR), sensitivity, speci city, positive likelihood ratio (PLR) and negative likelihood ratio (NLR) based on each resource.
Gene functional enrichment and network analysis 1. Enrichment analyses.
Coexpressed genes were obtained from the Multi Experiment Matrix (http://biit.cs.ut.ee/mem/index.cgi), cBioPortal (http://www.cbioportal.org/) and GEPIA. Genes obtained from cBioportal were selected only when their Spearman score was higher than 0.4. The overlapped genes were removed. Then coexpressed genes were pooled together for further analysis. FunRich 3.0 was utilised to perform enrichment analyses [16]. The Search Tool for the Retrieval Interacting Genes (STRING) (http://www.string-db.org/) database and Cytoscape were used for PPI construction and hub genes identi cation [17]. Hub genes were selected by calculating the degrees of the nodes and regarded as the key coexpressed genes of GOLGA8B in HCC.
A combined score over 0.9 was adopted to access the associations among proteins. The RNA-sequencing expression data of hub genes were downloaded from UCSC xena to evaluate their differences between HCC and paracarcinoma tissues. The correlations between GOLGA8B and these hub genes were also conducted. Moreover, IHC of hub genes were obtained from the Human Protein Atlas (HPA) (https://www.proteinatlas.org/).

The prediction of miRNAs.
The miRWalk 2.0 (http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk2/) was used to predict the potential miRNAs which could in uence GOLGA8B expression. To explore the relationship between miRNAs and GOLGA8B, log2(total_RPM +1) transformed miRNA-sequencing expression data of selected miRNAs were downloaded from UCSC xena for further analysis.

Protein Expression Data
The IHC of HCC and normal liver tissues is shown in Figure 7. The staining intensity of GOLGA8B exhibited strongly in HCC tissues ( Fig. 7a and 7b). In contrast, GOLGA8B exhibited no staining in the normal liver tissues ( Fig. 7c and 7d). As shown in Table 3, GOLGA8B staining exhibited higher scores in HCC tissues than in non-HCC controls (11.8688 ± 0.6593 vs 5.8989 ± 1.8589, P = 0.000). Moreover, IHC staining of GOLGA8B in HCC tissues was stronger in the late stages than in the early stages (P = 0.000).
Gene functional enrichment and miRNA prediction

Analyses of coexpressed genes
Three different databases were utilised for GOLGA8B coexpressed genes selection, and a total of 1303 genes were screened out for further enrichment analyses (Fig. 10a). Additionally, the PPI network was constructed (Fig. 10b). As for the biological process (BP), genes were mainly gathered in peptide metabolism mitosis and the regulation of cell cycle. In the cellular component (CC), genes were majorly enriched in the nucleus, nucleolus and cytoplasm. Regarding molecular function, the coexpressed genes were involved in RNA binding, transcription regulator activity and DNA binding. And for biological pathway, genes coexpressed with GOLGA8B were majorly gathered in spliceosome and the mitogenactivated protein kinase (MAPK) signalling pathway (Fig. 11a-d). Hub genes-including SF3B1, HNRNPA2B1, HNRNPA1 and SRRM2-were discovered according to the degrees and interactions. HNRNPA2B1 (13.65 ± 0.01848 vs 13.41 ± 0.03188) and HNRNPA1 (12.17 ± 0.03196 vs 11.50 ± 0.04929) were increased signi cantly in HCC tissues (P = 0.000). SF3B1 (12.03 ± 0.02576 vs 11.91 ± 0.05114) and SRRM2 (12.57 ± 0.03430 vs 12.42 ± 0.04990) showed an upregulated trend in HCC tissues. (P = 0.0932 and 0.1186). However, there was no correlation between GOLGA8B and four hub genes (Fig. 12a-h). According to the HPA database, SF3B1 and SRRM2 showed no obvious staining in both normal liver tissues and HCC tissues (antibodies HPA050275 and HPA041411). HNRNPA2B1 staining was medium in normal liver tissues but strong in HCC tissues (antibody CAB12403). HNRNPA1 indicated strong staining in HCC tissues than in normal tissues, which showed weak staining (HPA001666) (Fig. 13a-h).

The prediction of miRNAs
In our research, the potential target miRNAs of GOLGA8B were predicted by 12 target prediction algorithms of miRWalk. The miRNAs predicted by more than 6 algorithms were selected as the target candidate miRNAs, and 15 miRNAs were screened out. According to the miRNA-sequencing expression data analysis, miR-203a, miR-139-5p, miR-144-3p, miR-369-3p and miR-374b-5p exhibited signi cant lower expressions in HCC. (Fig. 14).

Discussion
In spite of the advances in diagnosis and treatment, HCC patients are still diagnosed at a late stage and possess a poor overall survival rate [20][21]. Therefore, more strategies are needed for HCC early diagnosis and treatment. A growing body of articles has explored numerous molecules year on year, hoping to clear the underlying molecular mechanisms of HCC and contribute to its diagnosis and treatment. Some of them have manifested vital roles in HCC carcinogenesis, such as metadherin [22][23][24]. However, so far, there has been no valid target protein-especially a protein located in the Golgi complex which could be the target for HCC diagnosis and treatment, making it necessary to be explored. Recent studies have concentrated on proteins of the Golgi apparatus in HCC. Yang et al. (2014) focused on the Golgi apparatus proteomes in HCC tissues and adjacent liver tissues; proteins related to the different BPs were found [25]. It is reported that GP73 acts as a key oncogene in HCC by regulating metastasis. For HCC diagnosis, the accuracy of Golgi protein 73 was higher than alpha-fetoprotein, which is used for clinical diagnosis [26][27]. Liu et al. (2018) suggested that GOLGPH3 acts an oncogene in HCC development and progression by activating the mTOR pathway, which is also a potential target for HCC diagnosis and therapy [28]. Lee et al. (2018) indicated that the Golgi transmembrane protein TEME165 contributes to the progression of HCC [29]. GOLGA8B-a gene that encodes protein golgin-67, which is located in the Golgi complex to maintain its structure-may also serve as a key character in HCC carcinogenesis.
We were the rst group to focus on GOLGA8B in HCC. Some previous studies discovered that GOLGA8B manifested different expressions in variant cancers through different methods [11][12][13][14][15]. However, there has been no research so far that has directly studied the relationship between GOLGA8B and cancers. To investigate the accurate function and underlying molecular mechanisms of GOLGA8B, we identi ed the expression and clinical signi cance of GOLGA8B, at mRNA level, in HCC by different data combination. IHC was also performed to access the expression of GOLGA8B at the protein level. Consequently, potential functions and pathways were explored.
First, we investigated the role GOLGA8B plays in HCC by an RT-qPCR analysis. Big data which contained RNA-sequencing data and microarray chip data were also utilised for GOLGA8B expression elucidation. There was an higher expression of GOLGA8B in HCC, according to different resources. Based on RNAsequencing data, GOLGA8B expression was higher in the late pathologic tumour stages (T3-T4) than the early pathologic tumour stages (T1-T2), indicating GOLGA8B might serve as an oncogene in HCC carcinogenesis. However, no correlations were found between GOLGA8B expression and the survival of HCC patients. The expression of GOLGA8B in IHC revealed a similar trend with mRNA expression. GOLGA8B protein exhibited signi cantly higher expressions in HCC tissues. GOLGA8B protein exhibited higher expressions in late stages (III-IV) than in early stages (I-II), which further elucidated its tumour promotor role in HCC development and progression.
We integrated the expression data obtained from different resources at mRNA and protein level. The results showed SMD = 0.893 (95% CI: 0.279-1.506, P = 0.004). A signi cantly higher expression in HCC at mRNA and protein level further con rmed the function of GOLGA8B in tumour promotion. Moreover, GOLGA8B might be of clinical value due to the result of sROC with an AUC of 0.79. However, a study with larger samples is still necessary for GOLGA8B expression and clinical value elucidation.
Enrichment analyses also reveal the vital functions and pathways of GOLGA8B in HCC. Within all the biological pathways, the spliceosome pathway was the most signi cantly enriched term. The spliceosome includes more than 150 different proteins; it yields mature mRNA by removing introns and joining extrons of pre-mRNA, which is responsible for gene expression [30][31]. The spliceosome pathway could participate in the carcinogenesis, cancer development and chemoresistance [32]. By the regulation of mRNA and RNA splicing, phosphorylation of spliceosome proteins may also take effect in the metastasis of HCC [33]. Besides, the spliceosome pathway also takes part in the tumour progression from cirrhosis to HCC [34]. Many genes associated with spliceosome, such as the MYC genes, manifested different expressions in HCC samples compared to the corresponding adjacent healthy liver samples. A number of studies have found that spliceosome could be a target for many anticancer drugs [35].
In addition, classic tumour suppressor phosphatase and tensin homologue could affect proteins on the Golgi apparatus to play a role in tumour suppression by interacting with spliceosome [36]. However, no relevant study related to GOLGA8B and spliceosome was found. In this study, we hypothesised GOLGA8B could be a vital element in the spliceosome pathway, but the potential molecular mechanism of GOLGA8B in HCC needs further con rmation. In addition, the MAPK pathway, which consists of extracellular signal-regulated kinase, functions as regulating fundamental cellular process, which includes survival, proliferation, progression and migration [37]. As reported, in 50-100% of HCC cases, the MAPK pathway is activated and is related to poor prognosis [38]. Carbohydrate-responsive elementbinding protein, which acts an important role in lipid and glucose metabolism in liver, takes effect in the development of HCC by regulating the MAPK pathway [39].
Factors such as anticancer drugs act on HCC regulation through triggering the MAPK pathway. This indicates the MAPK pathway might be a potential therapeutic target [40][41]. The MAPK pathway also participates in liver-disease progression-initiated from in ammation, followed by brosis, cirrhosis and HCC [42]. A number of studies have elucidated that MAPK pathway plays a signi cant role in HCC. In this study, we assumed that GOLGA8B is a vital factor in the MAPK pathway. However, no studies related to GOLGA8B and the MAPK pathway could be found; more experiments are necessary to verify this hypothesis.
Hub genes SF3B1, HNRNPA2B1, HNRNPA1 and SRRM2 were selected for further analyses according to degrees and interactions. SF3B1 was overexpressed in HCC tissues and was identi ed to be a target antigen of HCC-associated antibodies. Also, SF3B1 was signi cantly mutated through DNA sequencing and mutation analyses [43][44]. Both HNRNPA2B2 and HNRNPA1 belong to the hnRNP A/B family, which is a subset of hnRNP proteins [45]. HNRNPA2B1 was overexpressed in HCC tissues. The expression of HNRNPA2B1 was signi cantly correlated with tumour differentiation, microvascular invasion and survival rate. As a splicing factor, HNRNPA2B1 participates in cancer development by different pathwaysincluding the MAPK pathway and NF-κB pathway, which provides an interplay between HNRNPA2B1 and GOLGA8B. Furthermore, HNRNPA2B1 interacts with factors like long noncoding RNA and human telomerase reverse transcriptase to serve as an important marker in HCC [45][46]. Similarly, HNRNPA1 expression was upregulated in HCC cell lines and tissues. Through the regulation of CD44v6, overexpression of HNRNPA1 promotes HCC invasion; this indicates shorter overall survival and a higher tumour recurrence rate [47].
SRRM2 is a factor of spliceosome. The mutation in SRRM2 predisposes people to thyroid carcinoma by affecting the alternative splicing of downstream target genes [48]. However, studies related to SRRM2 and HCC have not been found. These four hub genes were all overexpressed in HCC based on TCGA data and showed an inverse trend with GOLGA8B expression. The expression of GOLGA8B and hub genes in HCC informed us that GOLGA8B may target these hub genes to hinder their expression through different mechanisms or signalling pathways. Further explorations are necessary to elucidate the relationship between GOLGA8B and hub genes in HCC.
miRNAs are endogenous, non-coding small RNAs which participate in posttranscriptional regulation by targeting the 3'-untranslated region (UTR) of mRNAs, causing degradation or translation suppressing of mRNAs, and act as suppressors or oncogenes in tumorigenesis [49][50]. In this study, miRNAs which may in uence GOLGA8B expression were selected only when they were predicted by six algorithms. The expression of those selected miRNAs were obtained for further analysis. miR-203a, miR-139-5p, miR-144-3p, miR-369-3p and miR-374b-5p were signi cantly decreased in HCC tissues. It is reported that miR-203a is downregulated in HCC and is associated with the prognosis of HCC patients [51].  found that miR-203a could suppress the metastasis, migration, invasion and angiogenesis of HCC cells [52]. miR-139-5p exhibited lower expressions in HCC and HCC cell lines. miR-139-5p also plays a role as a tumour suppressor via inhibiting migration, invasion and growth of HCC cells [53][54]. Wu et al. (2017) found that miR-144-3p could suppress the migration and growth of HCC cells [55]. Moreover, the downregulated miR-144-3p was correlated to poor disease-free survival. As for miR-369-3p and miR-374b-5p, they were dysregulated in various types of cancers, but no study has been conducted to explore their functions and mechanisms in HCC[56-60]. These ve miRNAs may target GOLGA8B to function as tumour suppressors. However, in vivo and in vitro experiments are necessary for further elucidation.

Conclusions
This study indicates overexpressed GOLGA8B may play an important role in HCC carcinogenesis and is related to pathologic tumor stages. The potential clinical value and underlying molecular mechanisms, however, require further elucidation.

Declarations
Authors' Contribution GZ, FM, HW, QS, ZH, GC and YD: study design, data analysis, writing of the article, and nal approval of the article. GZ, FM: data collection, data analysis, writing of the article and nal approval of the article. All authors read and approved the nal manuscript.  Tables   Table 1. The RT-qPCR data of GOLGA8B.

Figure 1
The ow diagram exhibited the main design of this study.

Figure 2
The expression and clinical features of GOLGA8B based on TCGA data. a. The expression of GOLGA8B based on TCGA. b. The ROC curve of GOLGA8B expression based on TCGA. c. The expression of GOLGA8B between TNM stages I-IV based on 373 HCC tissues.

Figure 12
Hub genes expression based on TCGA. a-d. Hub genes expression in HCC and non-HCC control. e-h.
Correlations between hub genes and GOLGA8B.

Figure 13
The IHC analysis of four hub genes obtained from HPA. a-d. The expression of SF3B1, HNRNPA2B1, HNRNPA1 and SRRM2 in HCC tissues (antibody HPA050275, CAB012403, HPA001666 and HPA041411).