Functional lncRNA-miRNA-mRNA Networks in Response to Baicalein Treatment in Hepatocellular Carcinoma

Introduction Baicalein has been shown to have antitumor activities in several cancer types. However, its acting mechanisms remain to be further investigated. This work is aimed at exploring the functional long noncoding RNA (lncRNA)/microRNA (miRNA)/messenger RNA (mRNA) triplets in response to baicalein in hepatocellular carcinoma (HCC) cell to understand the mechanisms of baicalein in HCC. Methods Differentially expressed lncRNAs (DELs) and miRNAs (DEMs) in HCC cell treated with baicalein were first screened using GSE95504 and GSE85511, respectively. miRNA targets for DELs were predicted and intersected with DEMs, after which the miRNA expression was validated using ENCORI and its prognostic value was assessed using Kaplan-Meier plotter. Potential miRNA targets were predicted by 3 prediction tools, after which expression level was validated at UALCAN and Human Protein Atlas. Kaplan-Meier plotter was used to evaluate the effects of these genes on overall survival and recurrence-free survival of HCC patients. Enrichment analyses for these genes were performed at DAVID. Results Here, we identified 14 overlapping DELs and 26 overlapping DEMs in the baicalein treatment group than those in the DMSO treatment group. Subsequently, by analyzing expression and clinical significance of miRNAs, hsa-miR-4443 was found as a highly potential miRNA target. Then, targets of hsa-miR-4443 were predicted and analyzed, and we found AKT1 was the most potential target for hsa-miR-4443. Hence, the lncRNAs-hsa-miR-4443-AKT1 axis that can respond to baicalein was established. Conclusion Collectively, we elucidated a role of lncRNAs-hsa-miR-4443-AKT1 pathway in response to baicalein treatment in HCC, which could help us understand the roles of baicalein in inhibiting cancer progression and may provide novel insights into the mechanisms behind HCC progression.


Introduction
Hepatocellular carcinoma (HCC) is the main type of liver cancer and with a high recurrence and morbidity rate, which is a heavy health burden nowadays [1]. It is now clear that exposure to aflatoxin, alcohol history, and infection of hepatitis B virus or hepatitis C virus can increase the risk of HCC [2]. For the further malignant development of HCC, genetic and epigenetic changes will occur [3]. HCC at early stages can be treated with surgical resection method; most of them can experience recurrence [4]. It should be noticed that Chinese herbs are valuable resources for the development of novel cancer treatment reagents and have been shown that these could efficiently inhibit tumor growth and metastasis with a relative low toxicity [5,6].
Baicalein is an active flavonoid compound extracted from root of a traditional Chinese herb, Scutellaria baicalensis Georgi [7]. In recent years, baicalein has been proved to have various pharmacological roles in human including antiinflammatory, anticancer effect, and antioxidant [8][9][10]. In nasopharyngeal carcinoma, the treatment of baicalein can reverse the radioresistance of cancer cell via downregulating autophagy [11]. In prostate cancer, baicalein was found that it could suppress cancer growth by arresting cell cycle and       BioMed Research International stimulating cell apoptosis via regulating the CDK6/FOXM1 axis [12]. The findings in recent decades have proved that Chinese herbs including curcumin and paclitaxel exert anticancer roles by disrupting the expression of noncoding RNAs (ncRNAs) [13,14]. However, it is unclear to date whether baicalein also exerts its effects on tumor progression by affecting ncRNA expression. The improvements of high-throughput sequencing technology have revealed that more than 90% of human genome transcripts did not encode proteins or peptides, which are called ncRNAs [15]. ncRNAs can be generally classified into two types based on nucleotides length: long ncRNAs (lncRNAs) and microRNAs (miRNAs). Aberrant expressions of ncRNAs including pseudogenes, lncRNAs, and miR-NAs have been widely studied in HCC with a large number of ncRNAs been identified [16][17][18]. Importantly, the competing endogenous RNA (ceRNA) theory has linked the functions of protein-coding genes and noncoding genes in diseases [19].

Materials and Method
2.4. miRNA Target Prediction for lncRNA. miRNA targets for DELs were predicted at LncBook. miRNAs that overlapped with DEMs were selected for following the analyses.

Validation of miRNA Expression Level and Clinical
Significance in HCC. Expression levels of these identified   [22]. After entering the name of gene, the expression box plot was automatically generated and the P value was automatically calculated on the webpage. The clinical significance of miRNAs on overall survival of HCC patients was analyzed at Kaplan-Meier plotter (http://kmplot.com/analysis/index .php?p=service&default=true) [23].

Functional Enrichment.
To understand the function of these genes, we performed gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment at Database for Annotation Visualization and Integrated Discovery (DAVID) (https://david.ncifcrf.gov/) [27].
2.8. Construction of lncRNA-miRNA-mRNA Network. Furthermore, we constructed a lncRNA-miRNA-mRNA network that responds to baicalein treatment based on the target prediction results. Then, the network was visualized using Cytoscape software.

Protein-Protein Interaction (PPI) Network Construction.
STRING (https://string-db.org/) [28] was employed to analyze PPI interactions of the targets for miRNA. Cytoscape software was used to construct PPI network and analyze the hub genes using the CytoHubba plug-in. The gene in central nodes might be core genes that contribute to HCC progression.  (Figure 1(a)), while 29 DELs were identified in the DMSO vs. 80 μM baicalein group (Figure 1(b)). Venn diagram indicated that there are 14 intersected DELs (Figure 1(c)), and the detailed information of these lncRNAs is listed in Table 1. In addition, heatmap for the expression of these 14 lncRNAs in the DMSO and baicalein treatment groups is shown in Figure 1(d).   Figure 2(b)). Furthermore, we showed there are 26 overlapping miRNAs (Figure 2(c)). The detailed information of these 265 miRNAs are presented in Table 2, and their expressions displayed in a heatmap manner are shown in Figure 2(d).

Identification of
3.3. Prediction miRNA Targets of DELs. miRNA targets for the identified 14 DELs were predicted at LncBook (Table S1) and then extracted the overlapped miRNAs with the 26 DEMs for following the analysis. The detailed results are presented in Table 3, and we found that 24 out of 26 DEMs may interact with DELs.

Validation of the Expression of miRNAs Using ENCORI.
Furthermore, we explored the expression levels of these 24 miRNAs in HCC tissues and normal tissues using ENCORI. The detailed expression results are presented in Table 4; we showed that 7 out of these 24 miRNAs were indeed abnormally expressed in HCC tissues compared with normal tissues. By comparing miRNA expression level in HCC tissues and in HCC cell after baicalein treatment, hsa-miR-4443 and hsa-miR-675-5p were selected for following the analyses since these two miRNAs have opposite expression pattern in these two conditions.

Validation of the Clinical Significance of hsa-miR-4443
and hsa-miR-675-5p in HCC. Then, we explored the effects of hsa-miR-4443 and hsa-miR-675-5p expression on the overall survival of HCC patients using the Kaplan-Meier plotter. We found that low hsa-miR-4443 was a predictor     10 BioMed Research International for overall survival of HCC patients (Figure 3(a)), while hsa-miR-675-5p did not have strong correlation with HCC patients' overall survival (Figure 3(b)). To further understand the roles of hsa-miR-4443 expression in HCC, we explored correlation of has-miR-4443 expression and clinical characteristics of HCC patients using Kaplan-Meier plotter database. Low hsa-miR-4443 was associated with worse overall survival in tumor stage at 1 to 3 (Table 5). In addition, low hsa-miR-4443 expression was correlated with overall survival in grades 2 and 3 of HCC patients but was not associated with overall survival in grade 1 (Table 5). In addition, we explored the expression of hsa-miR-4443 in different tumor stages and tumor grades. Our results confirmed that HCC patients at late tumor stage or high tumor grade tend to have low hsa-miR-4443 expression (Figures 4(a) and 4(b)).

Function Enrichment of the Targets.
Subsequently, we performed GO and KEGG analyses to understand the roles of hsa-miR-4443 targets. As shown in Figure 6(a), we showed 10 enriched biological processes (BP) for the predicted 796 genes and showed AKT1 linked to 4 of 10 BP terms. In addition, KEGG enrichment analysis indicated these genes were also can be enriched in pathways related to cancer including pathways in cancer, non-small-cell lung cancer, small-cell lung cancer, and so on (Figure 6(b)).

Identification of Hub
Gene in PPI Network. PPI network was constructed after analyzing using STRING and visualized at Cytoscape ( Figure S1). CytoHubba was used to identify the hub genes in this network (Figure 7(a)), and

11
BioMed Research International the detailed information of these top 10 hub genes are shown in Table 6. Subsequently, we selected an enriched pathway (hsa05200: pathways in cancer) and intersected with the top 10 hub genes, and we showed that there are 4 overlapping genes (AKT1, MAPK8, AR, and MDM2) (Figure 7(b)).
3.9. Validation of AKT1, MAPK8, AR, and MDM2 Gene Expression and Clinical Significance in HCC. UALCAN was used to explore AKT1, MAPK8, AR, and MDM2 expression, and we found that they were all abnormally expressed in HCC tissues compared with normal tissues (Figure 8(a)). However, high AKT1 and MAPK8 were associated with both   13 BioMed Research International poorer overall survival and recurrence-free survival of HCC patients (Figures 8(b) and 8(c)). IHC assay confirmed that AKT1 and MAPK8 protein levels were higher in HCC tissues compared than in normal tissues (Figure 8(d)). Interestingly, we showed AKT1 expression level was decreased by baicalin treatment compared with DMSO treatment (Figure 9(a)), whereas MAPK8 expression level was slightly increased by baicalin but the difference was not significant (Figure 9(b)). Hence, AKT1 was selected for following the analyses. As indicated in Table 7, high AKT1 expression was associated with overall survival and recurrence-free survival. Univariate and multivariate analyses showed that AKT1 expression along with tumor stage could be used as independent factors to predict the overall survival of HCC patients (Table 8).

Relationship between AKT1 Expression and TIICs in HCC.
We employed TIMER to explore the association of AKT1 expression and TIICs in HCC, and the results in Figure 10 showed 3.11. Drugs against AKT1 Predicted at DrugBank. According to DrugBank, the approved or experimental drugs that could act on ACTB and VEGFA as shown in previous studies are summarized in Table 9.
3.12. Proposed Highly Potential lncRNA-miRNA-mRNA Network Participates in the Roles of Baicalin on HCC Progression. Based on the above analysis results, we proposed a HSALNT0171251/HSALNT0103092/HSALNT0167051hsa-miR-4443-AKT1 ceRNA network that can respond to baicalin treatment and hinder HCC tumorigenesis ( Figure 11).

Discussions
Novel ncRNAs are continually to be identified in recent years due to the improvements of high-throughput transcriptome analysis methods [15]. The abnormally expressed lncRNAs were demonstrated that could either stimulate or inhibit cancer progression. Functional pattern of lncRNA-miRNA-mRNA regulatory network has been revealed to participate in the progression of a variety of cancer types [14,33]. Several studies have been performed to investigate the acting mechanisms of baicalin in cancers [10,12]. However, the investigation of lncRNA-miRNA-mRNA network that can respond to baicalin treatment remains largely unknown. In this work, our purpose was to construct potential lncRNA-miRNA-mRNA ceRNA triplets in baicalin treatment HCC cells. Using comprehensive bioinformatic analyses methods, several lncRNAs, miRNA, mRNA, and then the lncRNA-miRNA-mRNA triplets that altered after baicalin treatment were identified in HCC.
In this study, we identified 14 overlapping DELs and 26 DEMs via limma package by comparing the microarray data with baicalin or DMSO treatment. By analyzing the miRNA targets of lncRNA and extracting the overlapping miRNAs with DEMs, we screened out 24 miRNAs for following the analyses. After analyzing the expression level and clinical significance of miRNA in HCC, only one miRNA, hsa-miR-4443, was selected for further exploration. hsa-miR-4443 has been revealed with decreased expression in ovarian cancer and stimulated metastasis and tumorigenesis, indicating a tumor-suppressive role [34]. In HCC, hsa-miR-4443 was revealed to be a target of lncRNA FEZF1-AS1 to suppress the aggressive behaviors of cancer cells [35]. In addition, hsa-miR-4443 was also found to be a tumor-suppressive miRNA in other cancers including glioblastoma and osteosarcoma and regulated by lncRNAs [36,37]. Here, we also showed hsa-miR-4443 decreased expression in HCC and associated with poorer overall survival of cancer patients, suggesting hsa-miR-4443 also function as a tumorsuppressive miRNA in HCC.
Based on the ceRNA hypothesis, lncRNA can serve as decoy for miRNA via miRNA binding site to affect target gene expression. This led us to explore the targets of lncRNA using three prediction tools. A total of 796 overlapped targets were identified and reported to be associated with cancer progression via KEGG enrichment analysis. By identifying     15 BioMed Research International positive correlation of AKT1 expression level and infiltration level of CD4+ T cells. Activation of AKT1 was previously reported to be a crucial step for the production of chemoresistant phenotypes, while the inhibition on AKT1 can improve the chemotherapy sensitivity to induce apoptosis [38]. In addition, AKT1 was revealed that it could be upregulated by SET domain containing 5 in breast cancer to stimulate tumor growth and metastasis [39]. Baicalein was found which can regulate cell growth, metastasis, apoptosis, and autophagy in cancer [40,41]. Hence, the identification of AKT1 as a potential target for baicalein may be used to explain the biological functions of baicalein.
There are several limitations in this work. For example, the analysis of DELs and DEMs was derived from a single dataset obtained from one HCC cell. Therefore, further validation in other HCC cells is necessary. Moreover, the effects of the lncRNA-miRNA-mRNA network we identified in this work should also be validated using in vivo animal experiments.

Conclusion
In summary, our study performed comprehensive analysis of aberrantly expressed lncRNAs and miRNAs in HCC cell after baicalein treatment. We constructed ceRNA network which contains 3 lncRNAs, 1 miRNA, and 1 target gene that respond to baicalein treatment. Hence, the current study not only helped us to understand the acting mechanisms of

Data Availability
The data used in this work is available at GEO database or other online tools as described in Materials and Methods.

Conflicts of Interest
There are no conflicts of interest from any of the authors.