Value of Biomarkers in Liver Cancer EMT model under different interventions(cid:0)A meta-analysis

There are various interventions to establish the Liver cancer epithelial-mesenchymal transition (EMT) models. However, the ideal biomarkers for unique model are not well established. Further studies are necessary to evaluation of effective EMT biomarkers under different interventions in vitro studies. A meta-analysis was performed to evaluate the performance of different biomarkers in HepG2 cells during EMT under multiple interventions. of intervention, knockout/knockdown, hypoxia, and other tumor microenvironments, as well as non-coding RNA (ncRNA) overexpression and silencing. N-cadherin can effectively evaluate the intervention effect of medication, genetic intervention, hypoxia and other tumor microenvironments, as well as ncRNA overexpression. Vimentin reects the effects of medication, pro-EMT genetic intervention and gene knockout/knockdown, anti-EMT ncRNA overexpression and anti-EMT ncRNA silencing and hypoxia. Snail only responds to the intervention of anti-EMT genetic intervention and gene knockout/knockdown, tumor microenvironments other than hypoxia, anti-EMT ncRNA overexpression and ncRNA silencing. (equal). Li: Conceptualization (equal); Validation (equal); Writing- review & editing(lead).


Background
Liver cancer is one of the deadliest malignancies. According to the global cancer statistics released by the World Health Organization in 2018, about 800,000 people die from liver cancer every year and the global incidence rate is increasing rapidly [1]. China has the most cases of liver cancer resulting in about 100,000 deaths every year [2]. Moreover, the majority of cases are diagnosed at the locally or distantly advanced stage, which contributes to a high rate of postoperative relapse and metastasis [3].
Multiple studies [4][5] have shown that epithelial-mesenchymal transition (EMT) can promote the invasion and metastasis process by enhancing the separation of the original tumor mass and subsequent metastasis, thereby entering the bloodstream or lymphatic ow circulation. The most critical characteristics of cancer cells that undergo EMT is the downregulation of epithelial markers and upregulation of mesenchymal markers [6]. Several studies have targeted these biomarkers to promote or inhibit the EMT process. For example, miR-125a-5p downregulates the epithelial marker E-cadherin and upregulates the stromal markers N-cadherin and Vimentin to promote EMT [7]. CD47 regulates E-cadherin, N-cadherin, and Snail to induce EMT [8].
There are several biomarkers of EMT, among which E-cadherin is the most critical epithelial marker and the main component of cell adhesion [9]. N-cadherin induces the transformation of epithelial cells into broblasts during cancer development, making cells more motile and invasive [10], and Vimentin (an intermediate lamentous protein) plays an important role in the formation of mesenchymal cells [11].
Different studies have used different combinations of biomarkers to evaluate the effect of EMT after intervention. Meng J et al. [12] used E-cadherin, N-cadherin, and Vimentin to evaluate ubiquitin-speci c protease 5 (USP5) intervention on EMT. Wang Y et al. [13] used E-cadherin and Vimentin to determine the effect of lncRNA-CASC2 intervention on EMT. However, the underlying reason for choosing these markers was not explained. Further investigation [14] suggested that different types of EMT exhibit different biomarker pro les. The process of EMT may be regulated by several different signaling pathways such as Wnt/β-catenin, Notch, Hedgehog, and receptor tyrosine kinase-mediated signaling pathways [15]. Therefore, a comprehensive evaluation of effective EMT biomarkers under different interventions is imprecise for guiding future in vitro studies.
Meta-analysis is an effective way to comprehensively summarize the available evidence [16]. A large number of in vitro EMT studies have been performed in the recent years [17][18]. Among them, HepG2 cells are the most commonly used cell line for in vitro studies because of their constant and stable phenotype, easy acquisition, and easy treatment [19]. Several studies have subject this cell line to different kinds of treatments with drugs, genes, ncRNAs, and microenvironmental changes [20][21][22]. In this study, we evaluated the effect of these interventions on the four most common EMT biomarkers (Ecadherin, N-cadherin, Vimentin, and Snail) to evaluate their effectiveness and selection for future studies.

Methods
This meta-analysis followed the PRISMA guidelines strictly [23] and the details of the interventions are provided in Supplemental 1.

Study selection
PubMed, Web of Science, Embase, CNKI, CBM, Wan Fang Data, and VIP databases were systematically searched since inception till June 14, 2020. The retrieval method of combining subject words and free words was carried out. The following terms were used for the search: "Hepatocellular Carcinoma" or "Liver Cancer" or "Liver Neoplasms", "Epithelial-Mesenchymal Transition" or "EMT," and "Hep G2 Cell" or "Hepatoblastoma G2 Cell Line." The search strategies for all the databases are described in Supplemental 2. The original literature were used for study.

Inclusion and exclusion criteria
Studies that ful lled the following criteria were included: (1) in vitro studies that were published in both English and Chinese; (2) the HepG2 cell line was used, under any intervention; and (3) expression level of the biomarker were measured by western blotting (WB). Studies were excluded from meta-analysis if (1) the quality of experiments was poor, with no speci c experimental methods and steps, lack of information pertaining to cell sources, and culture details; (2) the experiments were repeated less than three times and no mean and standard deviation values could be obtained after contacting the corresponding author; and (3) the full text of the published articles was not available.

Data extraction
An MS Excel data extraction table was created by two independent reviewers (Jing Yan and Bei Xie). The following information was obtained from the studies: (1) authors, published year; (2) the name, dose, and duration of the intervention; (3) source of the cell line and culture medium; (4) the expression of EMT biomarkers including E-cadherin, N-cadherin, Vimentin, and Snail.

Quality assessment
Two independent reviewers (Yan Jing and Shuli Zou) performed quality evaluation of the collected data. To assess the risk of bias in vitro cell-based studies, we comply with the standard established by Golbach LA et al. [24]. Based on our study, the following 11 factors were evaluated: cell source, intervention time, intervention dose, cell culture, cell viability assessment, experimentation process, measurement process of results, randomization and blinding of study, control, standarlized reagent and instrument. Based on the above information, the study was rated as "high quality" "low quality" or "unclear quality."

Data analysis
Stata15 was used for performing meta-analysis. The mean difference (MD) was calculated for continuous variables. The standardized mean difference (SMD) was calculated to analyze continuous data. 95% con dence interval (CI) was calculated for every variables. Heterogeneity among studies was tested by I 2 . An I 2 value of ≤ 50% indicated no signi cant heterogeneity and the data were combined by a xed-effect model; otherwise, a random-effect model was used. P-value of <0.05 was considered as statistically signi cant.
To explore potential sources of heterogeneity and obtain further information, we performed a subgroup analysis based on different interventions and effect.
(2) Gene: Genetic intervention group A (pro-EMT genetic intervention) and group B (anti-EMT genetic intervention), gene knockout or knockdown.

Literature search
As shown in Figure 1, a total of 849 studies were identi ed with research method metioned above. After removing duplicates, 660 studies were considered for title and abstract screening. Of these studies, 266 were selected for full-text screening and 58 were nally included in the meta-analysis.

Risk of Bias Characteristics
The results of the risk of bias assessment are presented in Supplemental 4. Cell source was not reported in 9.52% (6/63) studies, 14.29% (9/63) studies did not implement the cell vitality assay, 38.1% (24/63) studies did not report intervention time, 65.08% (41/63) studies did not report intervention concentration, and 1.59% (1/63) studies did not report the cell culture process.   Of these, as shown in Figure  The expression of N-cadherin in the pro-EMT ncRNA silence (SMD = 8.14, 95% CI [-9.77, 26.05], P >0.05) was higher than that in the control group, whereas the expression of N-cadherin in the anti-EMT ncRNA silence (SMD = -6.05, 95% CI [-12.25, 0.14], P >0.05) was lower than that in the control group, but without signi cant difference.

Publication bias
Publication bias for the meta-analysis with regard to the expression of E-cadherin in the genetic intervention group, indicated that there was no obvious publication bias in the included literature (Egger test P = 0.96).

Discussion
To the best of our knowledge, this is the rst meta-analysis study that explored the expression of biomarkers under diverse pro-EMT and anti-EMT interventions in HepG2 cells. It based on the preintervention and post-intervention self-control of HepG2 cells. Our study incorporated 58 studies with all different ways of interventions and contained more detailed information about experimental and detection process, and excluded the poor-quality articles that did not provide details about cell origin or experimental process. Primarily, we classi ed these 58 studies according to the interventions, and then subgrouped some intervention groups according to the inconsisitent outcomes (enhancement or inhibition of EMT), thus systematically evaluated the effectiveness of biomarkers under interventions with different outcomes.
Our study showed that E-cadherin responds well to the intervention of medication, genetic intervention, gene knockout/knockdown, hypoxia, and other tumor microenvironments, as well as ncRNA overexpression and silencing. N-cadherin can effectively evaluate the intervention effect of medication, genetic intervention, hypoxia and other tumor microenvironments, as well as ncRNA overexpression. Vimentin re ects the effects of medication, pro-EMT genetic intervention and gene knockout/knockdown, anti-EMT ncRNA overexpression and anti-EMT ncRNA silencing and hypoxia. Snail only responds to the intervention of anti-EMT genetic intervention and gene knockout/knockdown, tumor microenvironments other than hypoxia, anti-EMT ncRNA overexpression and ncRNA silencing.
E-cadherin can better evaluate the effect of medicine, genes, microenvironment, and ncRNAs on EMT intervention. The reason may be that medicine, genes, microenvironment, and ncRNAs can directly or indirectly act on the signaling pathways to activate the upstream protein E-cadherin and therefore its expression. E-cadherin, the rst member of the cadherin family, is an active inhibitor responsible for invasion and growth in many epithelial cancers [10]. conclusion, E-cadherin can effectively re ect the EMT process in HepG2 cells under various interventions. Therefore, we recommend it to be the rst choice in future in vitro studies as a EMT biomarker for medicine, gene, microenvironment, and ncRNA interventions in HepG2 cells.
Our study also showed that N-cadherin expression changes signi cantly to EMT process induced by medicine, genetic intervention, tumor microenvironment and ncRNA overexpression intervention. Ncadherin is a calcium-dependent single-chain transmembrane glycoprotein that mediates homotype and heterotype adhesion between cells [88]. Overexpression of N-cadherin changes cell polarity and inter-cell adhesion, making cancer cells more prone to metastasis [89]. A previous study showed that during EMT, cadherin changes from E-cadherin to N-cadherin, so the upregulated expression of N-cadherin is considered an important EMT biomarker [90]. it may be that the types of genes and ncRNAs involved in the included experiments were not consistent, as well as the complex signaling pathways such as TGF-β1, Wnt/β-catenin, EGFR, and NF-κB ultimately led to signi cant differences in the e cacy of the intervention and resulted in greater heterogeneity among the included studies. Therefore, we suggest the following: (1) when researchers select medicine, genetic intervention, and tumor microenvironment, and overexpression of ncRNA to in uence the EMT effect on HepG2 cells, N-cadherin can be used as a secondary marker; (2) More studies involving the expression of N-cadherin regulated by gene knockout/knockdown and ncRNA silence are needed to establish the research system of gene/ncRNA-N-cadherin-signaling pathway to reduce the heterogeneity and improve the conversion and utilization.
We also found that medicine, pro-EMT genetic invervention, gene knockout/knockout, anti-EMT ncRNA overexpression/silencing, and hypoxia signi cantly affects Vimentin. Expression of Vimentin increases as epithelial cells transforms to mesenchymal cells during the EMT process [91]. Liu et al. [94] showed that Vimentin participates in the EMT of cancer cells by mediating cytoskeletal tissue and local adhesion maturation. Vimentin is upregulated in HCC and then induced in the EMT process [92]. Dan et al. [93] found that Vimentin acetylation is involved in SIRT5-mediated migration in liver cancer. Most antitumor drugs can inhibit the growth and invasion of tumor cells by directly or indirectly inhibiting Vimentin. Satelli A et al [94] found that a Vimentin-binding mini-peptide can bind to Vimentin, target, and interact with Vimentin to interfere with various signaling pathways such as Erk, AKT1, Axl and PI3K and then cell functions. Liu et al. [27] also reported that Fucoidan inhibited the expression of vimentin by inhibiting the activity of the PI3K/AKT signaling pathway. However, interventions for anti-EMT genetic intervention, tumor microenvironments other than hypoxia, pro-EMT ncRNA overexpression/silencing does not signi cantly change Vimentin expression. In conclusion, Because of the limitation in the number of included studies (1-2), there is not enough data to support the evaluation effect of Vimentin on the above three interventions. In summary, we believe that the following: (1) Vimentin could be a valuable marker for interventions of medicine, pro-EMT gene overexpression/knockdown/knockout, hypoxia, anti-EMT ncRNA overexpression/silencing act on HepG2 cells to affect the EMT; (2) Future studies should pay more attention to the effects of intervention for tumor microenvironments other than hypoxia, pro-EMT ncRNA overexpression/silencing needs further investigation.
In terms of Snail, interventions of tumor microenvironments other than hypoxia, anti-EMT gene overexpression, gene knockdown/knockdown, anti-EMT ncRNA overexpression, and ncRNA silencing induced signi cant changes of Snail expression. Snail family of proteins is one of the most important EMT downstream signaling pathway transcription factors. It's a key inducer of EMT, which can regulate invasion and migration [95]. The possible reasons are as follows: During EMT, its downstream signaling pathway transcription factors (EMT-TFS) are activated, mainly including zinc nger transcription factors family (Snail1 and Snail2), Twist family (Twist1 and Twist2), and ZEB family related to zinc nger E-box (ZEB1 and ZEB2), etc. Among them, the Snail family of proteins is a key inducer of EMT, which can regulate invasion and migration [95]. Its main role is to directly binds to the E-Box sequence located in the promoter region of the E-cadherin gene and inhibit its transcription [96], hence promoting EMT. There are many studies about the relationship between the Snail pathway and EMT [97][98], but few studies evaluated its potential as EMT biomarker. (Only 21 articles were included in this paper) Therefore, we believe that Snail could be used as a reference marker in future studies to further investigation are needed to explore its value under various interventions.
Our strength is that we included all different kinds of interventions and performed subgroup analysis to reduce the effects of heterogeneity between studies. The limitation is the heterogeneity of in vitro study, intervention dose and processing time, cell source, culture medium, etc. 65% of the included studies did not provide the concentration of the intervention. That's why SMD was used in the pooled analysis of the data. In addition, the different sources, culture details and intervention time of HepG2 cells in the included experiments all had a certain effect on the subsequent analysis. Therefore, it is suggested that the origin,