Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Multiomics in silico analysis identifies TM4SF4 as a cell surface target in hepatocellular carcinoma

  • Kah Keng Wong ,

    Roles Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    * kahkeng@usm.my

    Affiliation Department of Immunology, School of Medical Sciences, Universiti Sains Malaysia, Kubang Kerian, Kelantan, Malaysia

    ⨯
  • Suzina Sheikh Ab. Hamid

    Roles Funding acquisition, Project administration, Resources, Writing – review & editing

    Affiliations Tissue Bank Unit, School of Medical Sciences, Universiti Sains Malaysia, Kubang Kerian, Kelantan, Malaysia, Department of Otorhinolaryngology-Head & Neck Surgery, School of Medical Sciences, Universiti Sains Malaysia, Kubang Kerian, Kelantan, Malaysia

    ⨯

Abstract

The clinical application of cellular immunotherapy in hepatocellular carcinoma (HCC) is impeded by the lack of a cell surface target frequently expressed in HCC cells and with minimal presence in normal tissues to reduce on-target, off-tumor toxicity. To address this, an in silico multomics analysis was conducted to identify an optimal therapeutic target in HCC. A longlist of genes (n = 12,948) expressed in HCCs according to The Human Protein Atlas database were examined. Eight genes were shortlisted to identify one with the highest expression in HCCs, without being shed into circulation, and with restrictive expression profile in other normal human tissues. A total of eight genes were shortlisted and subsequently ranked according to the combination of their transcript and protein expression levels in HCC cases (n = 791) derived from four independent datasets. TM4SF4 was the top-ranked target with the highest expression in HCCs. TM4SF4 showed more favorable expression profile with significantly lower expression in normal human tissues but more highly expressed in HCC compared with seven other common HCC therapeutic targets. Furthermore, scRNA-seq and immunohistochemistry datasets showed that TM4SF4 was absent in immune cell populations but highly expressed in the bile duct canaliculi of hepatocytes, regions inaccessible to immune cells. In scRNA-seq dataset of HCCs, TM4SF4 expression was positively associated with mitochondrial components and oxidative phosphorylation Gene Ontologies in HCC cells (n = 15,787 cells), suggesting its potential roles in mitochondrial-mediated oncogenic effects in HCC. Taken together, TM4SF4 is proposed as a promising cell surface target in HCC due to its high expression in HCC cells with restricted expression profile in non-cancerous tissues, and association with HCC oncogenic pathways.

Citation: Wong KK, Ab. Hamid SS (2025) Multiomics in silico analysis identifies TM4SF4 as a cell surface target in hepatocellular carcinoma. PLoS ONE 20(2): e0307048. https://doi.org/10.1371/journal.pone.0307048

Editor: Mahmood S. Choudhery, University of Health Sciences Lahore, PAKISTAN

Received: June 28, 2024; Accepted: January 21, 2025; Published: February 25, 2025

Copyright: © 2025 Wong, Ab. Hamid. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The codes used for the scRNA-seq datasets analysis are available at: https://github.com/kahkengwong/TM4SF4_HCC_scRNA-seq

Funding: This work was supported by the Fundamental Research Grant Scheme (FRGS) (FRGS/1/2023/SKK10/USM/02/14 and grant no: 203.PPSP.6171388) awarded to KKW by the Ministry of Higher Education (MoHE) Malaysia. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer where it accounts for approximately 80–90% of the cases [1,2]. Although HCC is the sixth most common cancer globally, HCC is the second leading cause of cancer-related deaths and its incidence is increasing worldwide [1,3,4]. In terms of mass-directed biopsies, HCC is the most frequently occurring liver lesion [5]. Majority of HCC patients are diagnosed at an advanced stage due to absence of signs and symptoms in the early stages, and the 5-year survival of HCC is dismal at 18–21% [6,7].

Approximately half of HCC patients are treated with systemic therapies, i.e., the kinase inhibitors sorafenib or lenvatinib as the first-line treatment, and regorafenib (kinase inhibitor), cabozantinib (tyrosine kinase inhibitor) or ramucirumab (anti-VEGFR2 antibody) as the second-line treatment [8,9]. In the past five years, multiple immune checkpoint inhibitors (ICIs) such as atezolizumab, bevacizumab, pembrolizumab and ipilimumab, either as monotherapy or in combination with other drugs, have been approved as they enhance the overall survival (OS) of HCC patients [3,10]. Nonetheless, ICI regimens mainly confer numerical improvements in median OS compared with systemic therapies (median OS ~ 19 months vs ~ 8–14 months) [3], and only ~ 20% of HCC patients present with objective radiological responses when treated with anti-PD-1/PD-L1 antibodies [11].

Immunotherapies such as chimeric antigen receptor (CAR) T cell therapy are under investigation for HCC. However, their clinical translation is impeded by the heterogeneity of cell surface protein expression and off-tumor toxicities due to target expression in normal tissues [12]. One example is glypican-3 (GPC3; a heparan sulfate proteoglycan), where its expression is low in normal liver tissues but expressed in around 70% of HCC patients [13,14]. However, GPC3 is expressed in multiple normal tissues including the lungs, mesothelium, mammary epithelium, ovaries and the brain [15–17]. Therapeutic T cells targeting antigen expressed in normal critical tissues including the brain and lung tissues often result in toxicities [18–20]. Furthermore, serum GPC3 is commonly increased in HCC due to cell surface GPC3 shedding, and shed GPC3 (sGPC3) is a negative regulator of cell surface GPC3 leading to inhibition of HCC cell growth [21,22], and blocking anti-GPC3 CAR T cells binding, reducing treatment efficacy [23].

In this in silico study, an exhaustive exploration for the optimal cell surface target highly expressed in HCC but absent from majority of normal human tissues was conducted according to the combination of multiple publicly available transcriptomics and proteomics datasets. Potential therapeutic targets were objectively shortlisted, and TM4SF4 was identified as the most promising candidate. By leveraging multiple independent datasets, as well as comparison with GPC3 and other common surface targets studied in HCC, the therapeutic potential of TM4SF4 was established. Furthermore, the potential functional roles of TM4SF4 in HCC were explored, further supporting its future therapeutic development against the deadly malignancy.

Materials and methods

Datasets of the project

This is an in silico study utilizing publicly available multiomics datasets, to identify and characterize potential cell surface therapeutic targets in HCC. Inclusion criteria for datasets used in this study were: 1) Transcriptomics and proteomics datasets with open access to raw and processed data; 2) Datasets that specifically include HCC and matched control NTL group to allow for paired analysis; 3) Datasets with a sample size of at least 30 per group for quantitative comparisons; 4) Proteomics datasets utilizing mass spectrometry or similar high-throughput methods to quantify protein levels; 5) Datasets with sufficient metadata, including clear labeling of tumor and non-tumor cases, and clinical or demographic data where available. Exclusion criteria for datasets used in this study were: 1) Datasets not related to gene or protein expression, such as genomic mutation-only studies; 2) Datasets without clear annotations of tumor or non-tumor samples; 3) Studies focusing on other liver diseases (e.g., hepatitis, cirrhosis) without sufficient HCC cases; 4) Datasets containing metastatic or recurrent HCC samples only without sufficient primary HCC cases; 5) Datasets with incomplete metadata (e.g., missing labels of tumor or non-tumor cases) or lacking information about experimental protocols (e.g., the specific platform used in their high-throughput analysis).

The liver cancer transcriptome comprised of 12,948 genes according to the Human Protein Atlas (HPA) database (version 20.1) [24–28] and each of the gene’s features were retrieved from the database including cellular location (nuclear, cytoplasmic, or plasma membrane), transcript expression levels in normal human tissues, brain tissues and cancers, as well as concentration of the encoded protein in the blood if detected. In terms of cellular localization, the data were based on HPA’s immunofluorescent staining in cell lines and UniProt cellular localization annotation. The transcript expression levels in normal human tissues were according to HPA’s Consensus Dataset whereby it consisted of the combination of three transcriptomics datasets including HPA, Functional Annotation of the Mammalian Genome (FANTOM5) [29,30] and Genotype-Tissue Expression (GTEx) [31,32]. For normal human brain tissues, their transcript expression profiles were based on HPA’s Consensus Human Brain Dataset, i.e., combination of the human brain transcriptomics datasets of GTEx and FANTOM5. The RNA-seq data of 17 human cancer types by HPA were in accordance to The Cancer Genome Atlas (TCGA) [33,34]. Presence of encoded protein in the blood was according to the mass spectrometry dataset from the PeptideAtlas [35,36].

In addition to the aforementioned datasets, separate transcriptomics and proteomics datasets were also obtained. The transcriptomics dataset of HCC cases (n = 366) from TCGA were retrieved separately from the cBioPortal database [37,38]. The Chinese Human Proteome Project (CNHPP) RNA-seq dataset of 35 fresh paired tissues, i.e., 35 adjacent non-tumor liver (NTL) cells and 35 HCC cases were retrieved from the supplementary information of its publication [39]. An independent transcriptomics dataset (microarray) of HCC (n = 225) and paired NTL (n = 220) were obtained from the Gene Expression Omnibus database (ID: GSE14520) [40,41]. For proteomics dataset, the Chinese HCC (CHCC) dataset of 166 paired tissues, i.e., 166 HCC and 166 NTL cases whose proteins were measured by mass spectrometry (MS) were retrieved from the Proteomic Data Commons database (ID: PDC000198) [42]. Summary of the framework of this study is presented in Fig 1. This in silico study did not involve the direct collection of data from human participants, nor did it utilize human specimens or tissues in a manner requiring ethical approval. All data used in this study were obtained from publicly available sources and anonymized datasets. Therefore, additional review and approval by an institutional review board or ethics committee was not required for this study.

thumbnail
Download:
Fig 1. Workflow to identify the optimal cell surface target in HCCs and downstream analysis to examine its potential functions in HCCs.

https://doi.org/10.1371/journal.pone.0307048.g001

The shortlisting procedures were initiated with the list of 12,948 genes expressed in liver cancers as annotated by The Human Protein Atlas database (v20.1) before the next shortlisting steps were conducted to shortlist for genes absent from normal brain and lung tissues, restricted expression in other normal tissues but highly expressed (>50 FPKM by RNA-seq) in HCCs, and genes that potentially encode cell surface proteins without being shed into circulation. A total of eight genes were shortlisted before the ranking procedures based on the combination of four transcriptomics and proteomics HCC datasets. The top-ranked target was subsequently selected to compare its expression profile in normal human tissues and HCC cases vs GPC3 and other common therapeutic targets in HCCs. The potential functions of TM4SF4 were examined via GO enrichment analysis (in scRNA-seq dataset) and heatmap construction of genes contributing to the enriched GOs.

Shortlisting for surface targets with low levels of expression in normal human tissues but high levels in HCCs

The following workflow for shortlisting and ranking procedures have previously been published [43] and adopted here for HCC with modifications. Within the longlist of 12,948 genes expressed in liver cancers defined by HPA, the following sequential shortlisting steps were conducted:

  1. (i). Shortlisting for genes not expressed in normal human brain tissues: This is required as cellular immunotherapy such as the approved anti-CD19 CAR T cells have been demonstrated to cause on-target, off-tumor neurotoxicity due to the presence of CD19 expression in brain tissues [44–46]. Hence, genes with the annotation “Detected in all”, “Detected in many”, “Detected in some”, “Detected in single” or with unknown status within the “RNA brain regional distribution” annotation were excluded. This yielded a total of 1,246 genes for the next shortlisting procedure.
  2. (ii). Shortlisting for genes absent from the normal human lung tissues: This is also required as CAR T cells therapy can cause on-target, off-tumor pulmonary toxicities [18,47,48]. From the aforementioned 1,246 shortlisted genes, those genes with expression values of above 1.5 within “Tissue RNA - lung [NX]” (“NX” denotes normalized expression of three independent datasets, i.e., the HPA, GTEx and FANTOM5 datasets) annotation were considered as expressed in lung tissues and thus excluded. This resulted in 731 genes for the next shortlisting step.
  3. (iii). Shortlisting for genes with restricted expression profile in other normal human tissues: From the aforementioned 731 shortlisted genes, those with the annotation “Detected in all” or “Detected in many” within the “RNA tissue distribution” annotation were excluded, resulting in a total of 585 genes.
  4. (iv). Shortlisting for genes highly expressed in HCC cases: From the aforementioned 585 shortlisted genes, the genes annotated with > 50 FPKM for HCC within “RNA cancer specific FPKM” annotation were shortlisted, and > 50 FPKM has been adopted as the cut-off for high expression levels [49,50]. This yielded a total of 99 shortlisted genes.
  5. (v). Shortlisting for genes that encode cell surface proteins and not shed into circulation: From the aforementioned 99 shortlisted genes, to select for genes that potentially encode cell surface proteins, the subcellular locations for each encoded protein were examined according to the annotation by COMPARTMENTS localization data available on GeneCards database [51,52]. A specific gene was considered to potentially encode cell surface protein according to the following criteria: (1) The “plasma membrane” annotation must appear as the top subcellular site with the highest confidence value according to the GeneCards database; (2) When a gene’s top subcellular sites were annotated as “plasma membrane” and “extracellular”, this gene was disregarded because it might be momentarily localized on the plasma membrane for secretion into the bloodstream; (3) Genes without any subcellular location annotation were excluded. A total of ten genes were shortlisted. The final shortlisting step involved the assessment whether each of the ten genes’ encoded protein was detected in the blood by mass spectrometry based on HPA database of re-analyzed mass spectrometry dataset from the PeptideAtlas [35,36]. This yielded a final total of eight shortlisted genes and the basic features of these eight genes were summarized in S1 Table.

Ranking of the eight shortlisted cell surface targets at transcript and protein levels

In order to shortlist for the most ideal cell surface target from the eight genes shortlisted above, the next step was to rank them in accordance with their transcript (TCGA, CNHPP, and GSE14520 datasets) or protein (CHCC dataset) expression levels (median value) in HCC cases. By comparing each gene’s or protein’s median expression values in HCC cases in a particular dataset, each gene or protein was given a rank score (RS). For TCGA (n = 366), GSE14520 (n = 225) and CHCC (n = 165) HCC datasets, the RS ranging from 0.1 to 0.8 was assigned for each of the eight shortlisted genes according to median expression value ranking. For example, a specific gene with the highest median transcript value compared with the rest of the eight shortlisted genes in TCGA HCC cases was assigned with the highest RS of 0.8, followed by the second, third and fourth highest median transcript levels assigned with the RS of 0.7, 0.6 and 0.5, respectively. These can be summarized as follows:

where the underlined alphabet represents its corresponding subscripted suffix.

Then, each RS was divided by the constant value x, where x is the greatest RS value for each distinct dataset, in order to normalize the RS value to a maximum value of 1. This resulted in each dataset’s normalized score (NS) that can be summarized as follows:

For transcriptomics datasets (i.e., TCGA, CNHPP, and GSE14520), each NS was then averaged to obtain the transcript mean normalized score, x̄T, i.e., x̄T =  (NSA + NSC + NSG)/3

For proteomics dataset consisting of sole dataset (i.e., CHCC), its protein mean normalized score, x̄P, is equivalent to NSH.

Finally, both x̄T and x̄P were averaged in accordance with the following weighted average formula:

where m = the number of items in the x̄T set (i.e., 3; NSA, NSC and NSG), n = the number of item in the x̄P set (i.e., 1; NSH).

The x̄TP for each of the shortlisted targets represents the final score for ranking based on the combination of both transcript and protein expression levels in HCCs. Downstream characterization of the selected cell surface target (i.e., TM4SF4) was conducted to determine its expression profile compared with GPC3 and other common cell surface therapeutic targets in HCC, i.e., CD24 [53,54], CD133 (PROM1) [55,56], CD147 (BSG) [57,58], EPCAM [59,60], MET [61,62], and MUC1 [63,64]. Control groups were used extensively in this study. Normal human tissues as controls were included in the comparisons during target shortlisting and validation. These included normal tissues datasets from the FANTOM5 (n = 45), GTEx (n = 54), and the HPA (n = 253). Moreover, non-cancerous human tissues from the HPA immunohistochemistry (IHC) datasets (e.g., gallbladder, duodenum, small intestine, stomach, pancreas, colon) were used as controls for comparison with HCC tissues in terms of TM4SF4 protein expression patterns, as detailed in subsequent sections. Additionally, NTL single cells were used as controls for intra-tumoral comparisons to identify genes specifically correlated with TM4SF4 expression in HCC but not NTL cells, in single-cell RNA-seq (scRNA-seq) analysis as detailed in subsequent sections.

scRNA-seq data analysis

The following scRNA-seq datasets were retrieved from the Gene Expression Omnibus database: 1) Human liver scRNA-seq dataset (GSE149614) of ten HCC patients and eight adjacent NTL liver cells [65]; 2) Human gallbladder scRNA-seq dataset (GSE134355) of two non-cancerous adult individuals [66]. The initial stage of the scRNA-seq data analysis using the Seurat package [67] involved loading gene expression matrices for each dataset into R data frames, which were then transformed into sparse matrix formats. An Seurat object was subsequently created (separately for the liver and gallbladder scRNA-seq datasets), excluding features present in less than 200 cells and filtering out cells with fewer than 500 expressed features or a total RNA count of zero. To exclude cells potentially affected by apoptosis or stress, mitochondrial gene content filtering was conducted. The percentage of mitochondrial genes per cell was calculated and cells with mitochondrial gene content above the 99th percentile were excluded from the analysis. Data normalization was then conducted using log-normalization method, the data were then scaled, and cell cycle effects were regressed out using a predefined list of S and G2M phase genes. Next, doublets were detected and removed using the scDblFinder package. To address batch effects such as variations in sequencing depth or sample handling, the datasets were integrated using Seurat package FindIntegrationAnchors function that identified integration anchors (pairs of cells from different datasets that are similar based on the specified features) between datasets. The anchors were then used by the IntegrateData function to integrate the datasets. Subsequently, Elbow plot was constructed using the ElbowPlot function to aid in determining the number of principal components that guides clusters identification using the FindNeighbors and FindClusters functions. The specific cell type for each cluster was identified according to the most enriched ontology in the Coexpression module of the ToppGene database using the top 50 genes (adjusted p-values) within a cluster. Dimensionality reduction was visualized using Uniform Manifold Approximation and Projection (UMAP) where the UMAP coordinates were extracted, and a data frame was created to associate each cell with its corresponding cluster identity. For visualization, the overall UMAP plots were generated with ggplot2 and colored using the viridis package. In addition, UMAP plots illustrating the expression gradient of TM4SF4 across all cell populations were generated. The gradient ranged from light pink to dark red, representing the spectrum of TM4SF4 expression levels using the colorRampPalette function. Cells that did not express TM4SF4 were depicted in light gray. Additionally, a combined plot integrating both boxplot and jitter plot elements was produced to demonstrate the expression levels of TM4SF4 across all identified cell populations using ggplot2 and colored using viridis. The codes used for the scRNA-seq datasets analysis are available at: https://github.com/kahkengwong/TM4SF4_HCC_scRNA-seq

Gene Ontology (GO) enrichment analysis

The potential functions of TM4SF4 in HCC were investigated via GO enrichment analysis of genes highly correlated with TM4SF4 transcript expression in HCC cases vs adjacent NTL cells. The GO enrichment analysis was conducted according to genes demonstrating high Pearson correlation value (r ≥ 6 as the cut-off) with TM4SF4 expression in the scRNA-seq dataset GSE149614 (HCC n = 10; NTL n = 8) [65]. Cells with zero TM4SF4 transcript count were excluded, and Pearson correlation values for each gene with TM4SF4 expression were calculated in HCC or NTL cells separately. The set of genes demonstrating r ≥ 0.6 with TM4SF4 expression in HCC or NTL cells were used for GO enrichment analysis according to the ToppGene database as described previously [68–73]. Corrected p-value, i.e., q-value according to the Benjamini-Hochberg procedure of < 0.01 was considered as significant enrichment.

Statistical analysis

The Shapiro-Wilk test was conducted to examine the normality of data distribution. The unpaired t-test or Mann-Whitney test was performed to compare numerical variables between two different groups for normally and not normally distributed data, respectively. One-way ANOVA and the Holm-Sidak’s multiple comparisons test were used to examine differences between more than two groups of numerical variables with normal data distribution. Kruskal-Wallis test and the Dunn’s multiple comparisons test were used to compare numerical variables in more than two groups with not normally distributed variables. For all comparisons, two-tailed p < 0.05 was considered as statistically significant. Statistical analyses were conducted by using GraphPad Prism version 9.0 (GraphPad Inc., La Jolla, CA, USA). The variables investigated in this study, across multiple transcriptomics and proteomics datasets, included: 1) Gene and protein expression levels in HCC vs NTL and other normal tissues (such as the brain and lung tissues); 2) Expression levels of TM4SF4 vs other common cell surface markers studied in HCC; 3) Correlation analyses (Pearson’s r values) to shortlist for genes highly correlated with TM4SF4 in HCC but not NTL cells; 4) Functional annotations of these TM4SF4-correlated genes in HCC via ontologies enrichment analysis to uncover TM4SF4-specific processes in HCCs.

Results

TM4SF4 is the top-ranked surface target highly expressed in HCCs

A total of eight genes (SLC2A2, SLC10A1, SLC22A7, SLC38A4, SLCO1B1, TFR2, TM4SF4, and TM4SF5) were shortlisted according to the aforementioned shortlisting methodologies. Each gene was absent from normal human brain and lung tissues, exhibited a restricted transcript expression profile in other normal human tissues, highly expressed in HCCs, and potentially localized on the plasma membrane without evidence of being shed into circulation.

The expression levels of each gene in HCC cases were initially compared at the transcript level in order to identify which of the eight shortlisted genes had the highest expression in HCCs. This was performed in the TCGA (n = 366), CNHPP (n = 35) and GSE14520 (n = 225) datasets. Transmembrane 4 L six family member 4 (TM4SF4) was the top-ranked gene with the highest transcript expression levels in two of the datasets (i.e., CNHPP and GSE14520) and it was ranked the second in the TCGA dataset according to the mean normalized score of transcript expression (x̄T) (Table 1). In terms of protein levels by MS in the CHCC dataset (n = 165), TM4SF5 (another member of the transmembrane 4 superfamily) demonstrated the highest expression levels followed by TM4SF4 and other proteins according to the mean normalized score of protein expression (x̄P) (Table 1 and S1 Fig). Finally, the weighted combination of x̄T and protein x̄P, i.e., x̄TP (combination of both transcript and protein levels ranking) was used as the consolidated ranking metric whereby TM4SF4 was the top-ranked target (x̄TP: 0.938) followed by SLC2A2 (x̄TP: 0.750), TFR2 (x̄TP: 0.719), TM4SF5 (x̄TP: 0.594) and the rest of the targets (Table 1). Hence, TM4SF4 was the selected cell surface target for downstream analysis.

thumbnail
Download:
Table 1. The mean normalized score (NS; derived from rank score, RS) of the eight shortlisted genes at the transcript (x̄T), protein (x̄P) or both transcript and protein (x̄TP) levels in four HCC datasets (combined total of 791 HCC cases).

https://doi.org/10.1371/journal.pone.0307048.t001

Comparison of TM4SF4 with other cell surface HCC therapeutic targets in normal human tissues and HCC cases

To validate the robustness of TM4SF4 as an optimal therapeutic target in HCC, the expression of TM4SF4 in normal human tissues and HCC cases were compared with those of seven other cell surface proteins commonly being researched for HCC therapeutic development, i.e., CD24, CD133 (PROM1), CD147 (BSG), EPCAM, GPC3, MET, and MUC1. In normal human tissues, TM4SF4 expression was significantly lower than each of the aforementioned seven genes in the FANTOM5 (n = 45 different tissues; p < 0.05 for each comparison), GTEx (n = 54; p < 0.01 for each comparison), and HPA (n = 253; p < 1 x 10-11 for each comparison) datasets (Fig 2A-2C). In these three datasets, TM4SF4 transcript was highly expressed in the gastrointestinal (GI) tract and pancreas. Hence, TM4SF4 protein expression profile in these tissues was examined according to the immunohistochemistry (IHC) data by HPA in non-cancerous human tissues. The IHC by HPA was conducted using anti-TM4SF4 rabbit polyclonal antibody (HPA046430) that had been validated by orthogonal method whereby the antibody was mainly consistent with the transcript expression data across 45 human tissues by HPA.

thumbnail
Download:
Fig 2. Comparison of TM4SF4 expression levels with seven other therapeutic targets in normal human tissues.

https://doi.org/10.1371/journal.pone.0307048.g002

Transcript expression levels of the genes were compared with TM4SF4 in FANTOM5 (n = 45) (A), GTEx (n = 54) (B), and The Human Protein Atlas (n = 253) (C) normal human tissues datasets. The violin plots within each graph were arranged from left to right according to descending median expression values of the genes. Within each violin plot, the thick horizontal line represents the median, while the lower and upper horizontal lines represent the first and third quartile, respectively. Each p-value represents comparison of TM4SF4 expression vs each of the other gene expression in all tissues within a specific dataset. Sm.: Small; Sig.: Sigmoid; STPM: Scaled transcript per million; TPM: Transcript per million.

In scRNA-seq of normal gallbladder (n = 3,790 cells), a total of 11 clusters were identified consisting of mesenchymal stromal cells, fibroblasts (two clusters, suggesting the presence of two subtypes of fibroblasts), neutrophils, cholangiocytes, macrophages (two clusters), NK cells, antigen presenting cells, NK and NKT cells, and endothelial cells (Fig 3A). For qualitative observations, TM4SF4 was distinctly expressed in cholangiocytes (cluster no. 4) with sporadic low-level expression in other cell clusters (Fig 3B). Quantitative comparison showed that TM4SF4 was significantly more highly expressed in cholangiocytes compared with all other cell clusters (p < 0.001) (Fig 3C). In scRNA-seq of normal liver (n = 27,246 cells), a total of 16 different clusters were identified comprising of NK cells with cytotoxic T lymphocytes (CTLs), NK and NKT cells, CTLs (three clusters), Kupffer cells (two clusters), dendritic cells, liver sinusoidal endothelial cells, hepatocytes, B cells, macrovascular endothelial cells, cholangiocytes, plasma cells, smooth muscle cells, and NKT cells (Fig 3D). Qualitatively, TM4SF4 was almost exclusively expressed in hepatocytes (cluster no. 9) and cholangiocytes (cluster no. 13) (Fig 3E). Quantitatively, TM4SF4 was significantly more highly expressed in either hepatocytes or cholangiocytes compared with the rest of the 14 liver cell types (p < 0.001) (Fig 3F). Notably, TM4SF4 was absent or very lowly expressed in both the innate (neutrophils, NK cells, macrophages, dendritic cells) and adaptive (CTLs, B cells, plasma cells, NKT cells) immune cell populations in the normal gallbladder and liver.

thumbnail
Download:
Fig 3. Expression of TM4SF4 in normal gallbladder and liver cells in scRNA-seq datasets.

https://doi.org/10.1371/journal.pone.0307048.g003

(A) Eleven different cell clusters in normal gallbladder (total number of cells: 3,790). APCs: Antigen presenting cells; ECs: Endothelial cells; MSCs: Mesenchymal stromal cells; (B) Intensity of TM4SF4 expression in normal gallbladder. The number of TM4SF4 + cells in each cluster and the percentage are presented next to the plot; (C) Comparison of TM4SF4 expression in cholangiocytes versus the rest of the cell clusters (p < 0.001 for all comparisons). TM4SF4 expression levels in median and mean are presented next to the plot; (D) Sixteen different cell clusters in normal liver (total number of cells: 27,246). CTLs: Cytotoxic T lymphocytes; LSECs: Liver sinusoidal endothelial cells; MVECs: Macrovascular endothelial cells; SMCs: Smooth muscle cells; (E) Intensity of TM4SF4 expression in normal liver; (F) Comparison of TM4SF4 expression in cholangiocytes or hepatocytes versus the rest of the cell clusters (p < 0.001 for all comparisons).

At the protein levels, in non-cancerous gallbladder, TM4SF4 protein was strongly expressed in glandular cells with membranous staining on all sides of the plasma membrane domains (i.e., apical, lateral and basal) where the basal domain was in contact with adjacent cells (i.e., the lamina propria consisting of lymphocytes, loose connective tissues and blood vessels) (Fig 4). In contrast, TM4SF4 protein was highly expressed on the apical domain of enterocytes plasma membrane (i.e., the microvilli of enterocytes) in the duodenum and small intestines without contact with the adjacent lamina propria. TM4SF4 protein expression was absent in the lamina propria, lymphocytes, Goblet cells, Paneth cells, intestinal stem cells, smooth muscle cells, and endothelial cells. Membranous TM4SF4 was also present in stomach’s glandular cells but on the apical and lateral domains of plasma membrane, pancreatic intercalated ducts (apical), and intralobular ducts (apical and lateral) without contact with the lamina propria (Fig 4).

thumbnail
Download:
Fig 4. TM4SF4 protein expression by IHC in GI tract and pancreas obtained from The Human Protein Atlas database.

https://doi.org/10.1371/journal.pone.0307048.g004

Cell populations positive or negative for TM4SF4 were labeled with boxes, outline or dotted lines. The cell populations positive for TM4SF4 were stated as membranous or cytoplasmic. For membranous staining, the plasma membrane domain demonstrating TM4SF4 positivity is also stated, i.e., apical, lateral and/or basal. Next, TM4SF4 expression in HCC cases were compared with the aforementioned seven HCC therapeutic targets in three independent transcriptomics datasets, i.e., TCGA (n = 366), GSE14520 (n = 225) and CNHPP (n = 35), as well as in the proteomics CHCC (n = 165) dataset. TM4SF4 expression was significantly higher than six of the genes in TCGA dataset (p < 1 x 10-6 for each comparison) (Fig 5A), four of the genes in CNHPP dataset (p < 1 x 10-5 for each comparison; CD24 was not present in the platform) (Fig 5B), all seven genes in GSE14520 dataset (p < 0.01 for each comparison) (Fig 5C), and four of the proteins in CHCC proteomics dataset (p < 0.05 for each comparison; CD24 was not present in the platform) (Fig 5D).

thumbnail
Download:
Fig 5. Comparison of TM4SF4 expression levels with seven other therapeutic targets in HCC cases.

https://doi.org/10.1371/journal.pone.0307048.g005

Transcript expression levels of the genes were compared with TM4SF4 in TCGA (n = 366) (A), CNHPP (n = 35) (B), and GSE14520 (n = 225) (C) datasets, while comparison at protein levels was conducted in the CHCC (n = 165) (D) dataset. The violin plots within each graph were arranged from left to right according to ascending median expression values of the genes or proteins. Within each violin plot, the thick horizontal line represents the median, while the lower and upper horizontal lines represent the first and third quartile, respectively. Each p-value represents comparison of TM4SF4 expression vs each of the other gene expression in all HCC cases within a specific dataset.

The relationship of TM4SF4 with clinico-demographical characteristics and survival of HCC patients

Two datasets contained comparable clinico-demographical and survival data of the HCC patients, GSE14520 (HCC n = 221) and CHCC (HCC n = 159), hence they were utilized to examine the relationship of TM4SF4 expression with those characteristics at the transcript and protein levels, respectively. TM4SF4 transcript and protein levels were not associated with the clinico-demographical characteristics examined including age, gender, cirrhosis, AFP, and ALT, except lower TM4SF4 transcript expression in stage III-IV compared with stage I-II patients (p = 0.014) but this was not significant at the protein levels (p = 0.477) where median TM4SF4 protein levels were higher in stage III-IV than stage I-II HCC patients (S2 Fig). In terms of survival, TM4SF4 expression was not associated with OS or relapse-free survival (RFS) at both transcript and protein levels (S3 Fig).

TM4SF4 expression is higher in HCC compared with NTL cases

Three datasets contained matched NTL cases, i.e., CNHPP (n = 35 for each HCC and NTL group), GSE14520 (NTL n = 220; HCC n = 225), and CHCC (n = 166 for each HCC and NTL group), and hence they were utilized for comparison of TM4SF4 expression between both groups. TM4SF4 expression was significantly higher in HCC cases compared with NTL cases in both CNHPP (p = 2.58 x 10-9) (Fig 6A) and GSE14520 (p = 3.82 x 10-9) (Fig 6B) datasets. Consistent with these observations, TM4SF4 protein was also more highly expressed in HCCs compared with NTLs in the CHCC proteomics dataset (p = 1.16 x 10-11) (Fig 6C).

thumbnail
Download:
Fig 6. TM4SF4 expression in NTL compared with HCC cases at transcript and protein levels.

https://doi.org/10.1371/journal.pone.0307048.g006

Comparison of TM4SF4 expression in NTL vs HCC cases in transcriptomics (A-C) datasets with paired control NTL cases. Comparison of TM4SF4 protein expression by IHC in NTL (D) and HCC (E) cases obtained from The Human Protein Atlas database. The frequency (in percentage) and intensity (moderate or strong) of TM4SF4 staining in HCC cells were stated with each HCC case ID. Yellow arrow: Cholangiocyte; Green arrow: Endothelial cell; Red arrow: Bile duct canaliculi; Pink arrow: Kupffer cell; Orange arrow: Hepatic stellate cell.

In terms of TM4SF4 protein levels by IHC, the NTL (n = 3) and HCC (n = 7) cases immunostained for TM4SF4 were retrieved from the HPA database. In the three available NTL cases, TM4SF4 protein was strongly expressed on the plasma membrane domain of hepatocytes that form bile duct canaliculi, while demonstrating weak cytoplasmic staining in hepatocytes (Fig 6D). TM4SF4 was also strongly expressed in cholangiocytes with membranous (all sides of the plasma membrane domain) and cytoplasmic staining. TM4SF4 expression was absent in hepatic stellate cells, Kupffer cells, and endothelial cells (Fig 6D). In HCC cases, TM4SF4 was expressed in all seven cases whereby membranous TM4SF4 expression in HCC cells on all sides of the plasma membrane domains were present in 85.7% (n = 6/7) of the HCC cases, in accordance with HPA’s interpretation of “Distinct membranous immunoreactivity was observed in liver and pancreatic cancers”. Three representative HCC cases with membranous TM4SF4 expression are shown in Fig 6E. A schematic diagram illustrating bile duct canaliculi positive for TM4SF4 staining and their surrounding structures in NTL is presented in Fig 7.

thumbnail
Download:
Fig 7. Schematic diagram depicting the bile duct canaliculi and their surrounding structures in normal liver tissues according to Tam et al.[74].

https://doi.org/10.1371/journal.pone.0307048.g007

Immune cells such as lymphocytes and macrophages that arrive from the hepatic portal vein and hepatic artery patrol the surrounding liver cells such as hepatocytes, cholangiocytes and the canal of Hering that are partially lined by hepatocytes and cholangiocytes. TM4SF4 is expressed on the surface of hepatocytes that form the bile duct canaliculi that are not exposed to immune cells. Meanwhile, TM4SF4 is expressed on the surface of cholangiocytes, potentially exposing them to immunotherapy. The inset TM4SF4 IHC staining image is an NTL case obtained from HPA (patient ID: 2429).

TM4SF4 expression is associated with mitochondrial components and oxidative phosphorylation GOs in HCC cells

To gain insights on the potential functions of TM4SF4 in HCCs, GO enrichment analysis was conducted in the scRNA-seq dataset (GSE149614) of HCC cells (n = 15,787 cells derived from ten HCC patients) and matched, adjacent NTL cells (n = 1,932 cells from eight control individuals). The scRNA-seq dataset was utilized to avoid bulk transcriptome datasets that may contain mixture of different cells apart from HCC cells, e.g., cholangiocytes, immune cells and/or endothelial cells.

A total of 49 and 32 distinct genes were highly correlated (i.e., Pearson r > 0.6) with TM4SF4 expression in NTLs and HCCs, respectively (S2 Table). Subsequently, GO enrichment analysis was conducted for these two sets of genes separately. In NTLs, five groups of GOs were significantly enriched (q < 0.01) mainly pertaining to plasma membrane components, cellular adhesion, myosin complexes, and secretory vesicles (Fig 8). In contrast, seven groups of GOs were significantly enriched (q < 0.01) for HCCs whereby four of the seven GO groups comprised of mitochondrial components and energy production GOs, and the other three GO groups consisted of cellular detoxification, lipid processes, and nucleotide synthesis or metabolism GOs (Fig 8). The complete list of enriched GOs and the list of genes contributing to each GO enrichment in NTLs or HCCs are presented in S3 Table. The top eight genes with the highest r values with TM4SF4 expression are shown in the scatter plot (Fig 8). Five of these top genes (CLU, SCP2, AGT, ALB, and SDHC) were described in terms of their normal functions and potential association with TM4SF4, and such descriptions for all 32 genes are presented in S4 Table.

thumbnail
Download:
Fig 8. GO enrichment analysis according to the genes with Pearson correlation value (r) of > 0.6 with TM4SF4 expression in NTL cells (n = 49 genes) or HCC cells (n = 32 genes).

https://doi.org/10.1371/journal.pone.0307048.g008

The enrichment analysis was conducted in the scRNA-seq dataset (GSE149614) of HCC cells (n = 15,787 cells derived from ten HCC patients) and adjacent NTL cells (n = 1,932 cells from eight HCC patients). The top eight genes with the highest r values with TM4SF4 expression are shown in the scatter plot. The bar graphs correspond to the groups of enriched GOs, and three representative GOs (the GO ID is displayed above each bar graph) that formed each GO group labeled below the graph are demonstrated. The bar graph was drawn on the -log q-value scale, and the dotted horizontal line represents the q-value cut-off of 0.01 to be considered as significant enrichment. Notes: One gene with r > 0.6 (i.e., ANPEP) in HCCs was not included in the HCCs GO enrichment analysis due to it exhibited the lowest r value that satisfied the 0.6 cut-off in HCCs, and its r value was higher in NTLs than HCCs (0.630 vs 0.602). It was thus included in the NTLs GO enrichment analysis only to avoid any overlapping gene for the analysis enriched specifically in NTLs or HCCs.

In terms of TCGA cases, each gene’s expression values (transcript per million, TPM) in HCC cases (n = 369) vs non-HCC liver cases (n = 160) were compared according to the Gene Expression Profiling Interactive Analysis 2 (GEPIA2) database [75]. Six of the 32 genes (MPC2, NDUFC2, COX6A1, TMCO1, COPZ1, and COX8A), as well as TM4SF4, showed significantly higher expression values in HCC compared with non-HCC liver cases (all p < 0.001; S4 Table). The rest of the genes (n = 26) did not show significant difference between the two groups, however 22 of these genes showed higher expression values in the HCC group (S4 Table).

Finally, heatmap of the 32 genes highly correlated with TM4SF4 expression in HCCs as well as the GO groups containing these genes are presented in Fig 9. In particular, 17 of the 32 genes (53.1%) contributed to the enrichment of mitochondrial components and energy production GOs (i.e., CLU, SCP2, TMBIM6, SDHC, ACAA2, AMBP, MPC2, NDUFC2, NDUFB2, GSTK1, MGST3, COX6A1, PRDX6, COX8A, CFH, PHYH, and ATP5A1).

thumbnail
Download:
Fig 9. Heatmap and hierarchical clustering of HCC cells (n = 15,787 cells from ten HCC patients; GSE149614).

https://doi.org/10.1371/journal.pone.0307048.g009

This was according to the expression of the 32 genes positively associated with TM4SF4 expression (i.e., r > 0.6) as shown in the bar graph on the left of the heatmap. The number(s) in brackets for each gene corresponds to the seven GO groups (defined below the heatmap) where a specific gene contributed to their enrichment in HCCs. Genes without such number denote absence from contributing to the enrichment of any GO.

Discussion

In this in silico study, through a sequence of impartial shortlisting and ranking methodologies, TM4SF4 was shortlisted and finally selected as an optimal cell surface therapeutic target for HCCs. Similar framework had previously been devised and implemented in other types of cancers, including colorectal cancer (CRC) and diffuse large B cell lymphoma, to identify key cell surface targets in these malignancies [43], and transcriptomics as well as proteomics datasets have regularly been utilized to build algorithms for uncovering crucial biomarkers in cancers [76,77]. TM4SF4 is not expressed in the critical tissues for survival including the brain, lung, heart and kidney tissues, while overexpressed in HCCs compared with matched and adjacent NTL cells. TM4SF4 also compares more favorably than other common HCC therapeutic targets in terms of its expression profile, i.e., lower in normal human tissues but high in HCC cases.

The tetraspanin superfamily comprises of around 33 proteins localized in the plasma membrane where each protein consists of four transmembrane domains, short intracellular amino and carboxy tails, a small intracellular loop, and two extracellular loops [78,79]. Tetraspanins exert diverse biological functions such as cell adhesion, motility, invasion and signal transduction through their unique abilities to associate with other proteins (e.g., integrins and receptor tyrosine kinases) to form tetraspanin-enriched microdomains [79,80]. The TM4SFs is a branch of the tetraspanin superfamily consisting of TM4SF1, TM4SF4, TM4SF5, TM4SF18, TM4SF19 and TM4SF20 where they significantly differ with other tetraspanins structurally [81,82]. TM4SF4, originally identified in human intestine epithelium and liver tissues [83], is specifically expressed in the gastrointestinal tract and pancreas. In the intestines, TM4SF4 plays physiological roles in the absorption and uptake of thiamine by interacting with the human thiamine transporter-2 [84], and in the liver, TM4SF4 is required for regenerating liver cells through regulation of cellular proliferation [85]. In the pancreas, TM4SF4 is highly expressed in pancreatic α cells, and the protein is used as a cell surface marker to differentiate pancreatic α from β cells [86,87].

In the past decade, TM4SF4 has been implicated in promoting the oncogenic processes of HCCs. TM4SF4 protein expression was significantly upregulated in HCC tissues compared with paired non-cancerous liver tissues [88], comparable with the observations in this study that TM4SF4 protein was upregulated in the CHCC proteomics dataset of HCCs vs paired NTLs. Functionally, TM4SF4 overexpression in HCC cell lines (QGY-7701 and BEL-7404) induced cell growth and colony formation of the cells, and TM4SF4 downregulation led to inhibited HCC cells proliferation [88]. In a separate report, TM4SF4 protein was also overexpressed in HCCs compared with paired non-tumor tissues by utilizing an in-house anti-TM4SF4 polyclonal antibody. TM4SF4 protein was specifically localized on all sides of the plasma membrane domains of HCCs but only present in the bile duct canaliculi of normal liver tissues, consistent with the IHC data by HPA as reported in this study. Knockdown of TM4SF4 in HCC cells impaired growth potential in vitro and repressed tumor growth in a mouse xenograft model of human HCC [89], similar with another study showing attenuated proliferation and migration of TM4SF4-silenced HCC cells [90]. Collectively, these studies indicate that TM4SF4 functions as an oncoprotein in HCCs.

Accumulating evidence has also established the roles of TM4SF4 as an oncoprotein in other solid tumors. In CRC, higher TM4SF4 expression is associated with liver metastasis, worse survival, tumor development and epithelial-mesenchymal transition (EMT) program regulated by TGFβ/Snail, TNFα/NFκB, and thymidylate synthase pathways [91–93]. In lung cancer, TM4SF4 expression is upregulated where it promotes tumor growth, migration, and invasion through activation of PI3K/AKT and JAK2/STAT3 pathways, as well as preserving cancer stem cell features [94–97]. TM4SF4 is also overexpressed in pancreatic cancers as a hub gene and upregulated in chemoresistant ovarian cancer cells [98,99]. Taken together, TM4SF4 exerts oncogenic effects in multiple types of solid tumors, supporting its potential as a therapeutic target beyond HCC.

Mitochondria are the key organelles that generate energy for cellular processes via the production of adenosine triphosphate (ATP). ATP is mainly produced by oxidative phosphorylation which takes place on the inner mitochondrial membrane of the cell [100–102]. Liver cells are remarkably rich in mitochondria to support their numerous metabolic functions whereby each hepatocyte contains approximately 1,000–2,000 mitochondria [103,104], and deregulated mitochondria signaling could lead to liver pathologies including HCCs. Mitochondrial oxidative phosphorylation defects cause misfolded proteins to accumulate in the mitochondrial matrix, causing increased expression of oncoproteins that fuel HCCs development and metastasis [105].

In this study, TM4SF4 expression was positively associated with mitochondrial components, oxidative phosphorylation, and ATP synthesis GOs. Mitochondria, often found in close proximity to the PM for Ca2 + signaling or to provide ATP, play key roles in liver metabolism and lipid hormone responses [106–108]. In HCC, the expression of mitochondrial citrate transporter and PM citrate transporter is elevated, and inhibitors targeting these transporters induce apoptosis of HCC cells [109]. More importantly, TM4SF4 overexpression accelerates carbon tetrachloride (CCl4)-mediated liver injury, and CCl4-mediated liver injury is caused by mitochondrial permeability alterations [110]. Collectively, as TM4SF4 is found on the PM of HCCs, the potential interaction of TM4SF4 with mitochondria in HCC cells may lead to oncogenic effects such as mitochondrial abnormalities, but this proposal remains to be validated experimentally.

Four of the genes that contributed to the enrichment of mitochondrial components (MPC2, NDUFC2, COX6A1, and COX8A) were significantly overexpressed in TCGA HCC cases as described previously in S4 Table. As TM4SF4 was also overexpressed in TCGA HCC cases, this suggests that TM4SF4 signaling may lead to upregulation of any of these four genes’ expression. Although there have not been studies showing the direct association between TM4SF4 with these four genes, co-expression patterns in independent studies may support this observation. In particular, MPC2 may appear as a stronger candidate among the other genes where the expression of both MPC2 and TM4SF4 were suppressed in human malignant epithelial cells infected by SARS-CoV-2 versus controls [111,112]. However, copy-number variation analyses showed that MPC2 was frequently amplified in a specific molecular subtype of HCC termed as the “mixed group” subtype characterized by higher variability in metabolic activity, compared to the quiescent subtype defined by low metabolic activity and cholesterol subtype characterized by reliance on cholesterol synthesis pathways [113]. Hence, direct mechanistic studies of TM4SF4 regulation of genes expression are required to demonstrate the downstream signaling controlled by TM4SF4 in HCC.

On the other hand, TM4SF4 expression was utilized as a marker for normal cholangiocytes to separate them from normal hepatocytes, fibroblasts, lymphocytes, and endothelial cells in liver tissues based on recent scRNA-seq data [114] that tally with the scRNA-seq observations in this study. TM4SF4 protein is highly expressed in normal cholangiocytes that form the hepatic bile ducts, while absent from multiple other cell types in the liver. Although TM4SF4 protein is expressed in hepatocytes, it is localized on the surface of these cells that form the bile duct canaliculi that are not in contact with infiltrating lymphocytes and other immune cells. As such, immunotherapy such as CAR T cells targeting TM4SF4 is proposed to have a lower likelihood in causing significant toxicities on normal hepatocytes. Likewise, TM4SF4 expression in the intestinal enterocytes’ apical surface (specifically, the microvilli of enterocytes lining the intestinal lumen) are unlikely to be affected by anti-TM4SF4 immunotherapy due to absence of contact with the lamina propria where immune cells including lymphocytes patrol. TM4SF4 highly expressed on the surface of HCC cells are not necessarily tumor-associated antigens due to lack of actual immunotherapeutic agent demonstrating potency against TM4SF4 in HCCs.

The limitations of this study are acknowledged as follows: 1) TM4SF4 is present in normal bile ducts (i.e., cholangiocytes) and the gallbladder epithelial cells where they are in contact with the adjacent stroma containing immune cells and lymphocytes. Hence, anti-TM4SF4 immunotherapy may cause toxicity to these sites. However, immune-related injury of bile ducts after immune-checkpoint inhibitor treatment can be treated with immunosuppressive agents such as prednisolone or mycophenolate mofetil [115,116]; 2) Targeting TM4SF4 systematically can cause toxicity to pancreatic α cells that express TM4SF4 [117,118]. However, our findings have demonstrated that TM4SF4 expression is more restricted compared with other more commonly investigated HCC surface therapeutic targets (e.g., GPC3 and CD147) which are highly expressed in tissues critical for survival such as the lungs and brain. Essentially, the localization of TM4SF4 in the bile duct canaliculi of the hepatocytes where immune cells are unable to access, suggests a lower likelihood of off-tumor effects when anti-TM4SF4 therapies are locally administered near HCC tumor sites; 3) The potential roles of TM4SF4 in mitochondrial-mediated oncogenic effects in HCCs require experimental verification.

Conclusions

This study provides evidence to support the development of anti-TM4SF4 immunotherapies such as CAR T cells against HCCs. The more restricted expression profile of TM4SF4 in normal human tissues versus other common therapeutic targets in HCC further supports the use of anti-TM4SF4 targeted agents against the disease. Moreover, the association of TM4SF4 expression with mitochondrial processes and energy production GOs in scRNA-seq dataset of HCC cells are reported for the first time. Finally, the shortlisting and ranking methodologies proposed in this study can be implemented independently to uncover the optimal cell surface targets for therapeutic development in other malignancies.

Supporting Information

S1 Fig. Transcript expression levels of the eight shortlisted targets in HCC cases.

Violin plots of the transcript expression levels in TCGA (n = 366) (A), CNHPP (n = 35) and GSE14520 (n = 225) HCC cases, and the protein expression levels in CHCC (n = 165) HCC cases (D). The violin plots within each graph were arranged from left to right according to ascending median transcript or protein expression values for each gene. For each violin plot, the thick black bar in the center denotes the median while the upper and lower lines denote the third and first quartile, respectively.

https://doi.org/10.1371/journal.pone.0307048.s001

(PDF)

S2 Fig. The relationship of TM4SF4 transcript and protein expression with clinico-demographic characteristics of HCC patients.

(A) Association of TM4SF4 transcript levels with clinico-demographic characteristics of HCC patients in the GSE14520 dataset (n = 221); (B) Association of TM4SF4 protein levels with clinico-demographic characteristics of HCC patients in the CHCC dataset (n = 159).

https://doi.org/10.1371/journal.pone.0307048.s002

(PDF)

S3 Fig. The relationship of TM4SF4 transcript and protein expression with the survival of HCC patients.

(A) Association of TM4SF4 transcript levels with the survival of HCC patients in the GSE14520 dataset (n = 221); (B) Association of TM4SF4 protein levels with the survival of HCC patients in the CHCC dataset (n = 159).

https://doi.org/10.1371/journal.pone.0307048.s003

(PDF)

S1 Table. Characteristics of the eight shortlisted genes highly expressed in HCC cases but with restricted expression profiles in normal human tissues according to The Human Protein Atlas dataset.

https://doi.org/10.1371/journal.pone.0307048.s004

(XLSX)

S2 Table. The complete list of genes and their Pearson r values with TM4SF4 expression in NTL or HCC cells in scRNA-seq dataset (GSE149614).

https://doi.org/10.1371/journal.pone.0307048.s005

(XLSX)

S3 Table. The list of enriched GOs (q < 0.01) according to the genes with Pearson r > 0.6 with TM4SF4 expression in HCC cells (n = 15,787 cells derived from ten HCC patients) or paired NTL (n = 1,932 cells from eight HCC patients) in scRNA-seq dataset (GSE149614).

https://doi.org/10.1371/journal.pone.0307048.s006

(XLSX)

S4 Table. Normal functions and potential association with TM4SF4 of the genes highly correlated with TM4SF4 expression in HCC cells (n = 32).

https://doi.org/10.1371/journal.pone.0307048.s007

(DOCX)

Acknowledgments

The authors wish to acknowledge the School of Medical Sciences and the Research Creativity and Management Office (RCMO), Universiti Sains Malaysia, for their invaluable support.

References

  1. 1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–49. pmid:33538338
  2. 2. Donne R, Lujambio A. The liver cancer immune microenvironment: therapeutic implications for hepatocellular carcinoma. Hepatology. 2023;77(5):1773–96. pmid:35989535
  3. 3. Llovet JM, Castet F, Heikenwalder M, Maini MK, Mazzaferro V, Pinato DJ, et al. Immunotherapies for hepatocellular carcinoma. Nat Rev Clin Oncol. 2022;19(3):151–72. pmid:34764464
  4. 4. Samant H, Amiri HS, Zibari GB. Addressing the worldwide hepatocellular carcinoma: epidemiology, prevention and management. J Gastrointest Oncol. 2021;12(Suppl 2):S361–73. pmid:34422400
  5. 5. Algashaamy K, Montgomery EA, Garcia-Buitrago M. Liver mesenchymal neoplasms: something old, something new. Pathology (Phila). 2022;54(2):225–35. pmid:34965900
  6. 6. Lee YT, Wang JJ, Luu M, Noureddin M, Kosari K, Agopian VG, et al. The mortality and overall survival trends of primary liver cancer in the United States. J Natl Cancer Inst. 2021;113(11):1531–41. pmid:34010422
  7. 7. Burton A, Tataru D, Driver RJ, Bird TG, Huws D, Wallace D, et al; HCC-UK/BASL/NCRAS Partnership Steering Group. Primary liver cancer in the UK: Incidence, incidence-based mortality, and survival by subtype, sex, and nation. JHEP Rep. 2021;3(2):100232. pmid:33748727
  8. 8. Facciorusso A, Tartaglia N, Villani R, Serviddio G, Ramai D, Mohan BP, et al. Lenvatinib versus sorafenib as first-line therapy of advanced hepatocellular carcinoma: a systematic review and meta-analysis. Am J Transl Res. 2021;13(4):2379–87 pmid:34017396
  9. 9. Shang R, Song X, Wang P, Zhou Y, Lu X, Wang J, et al. Cabozantinib-based combination therapy for the treatment of hepatocellular carcinoma. Gut. 2021;70(9):1746–57. pmid:33144318
  10. 10. Cheng AL, Qin S, Ikeda M, Galle PR, Ducreux M, Kim TY, et al. Updated efficacy and safety data from IMbrave150: atezolizumab plus bevacizumab vs. sorafenib for unresectable hepatocellular carcinoma. J Hepatol. 2022;76(4):862–73. pmid:34902530
  11. 11. Rimola J, Da Fonseca LG, Sapena V, Perello C, Guerrero A, Simo MT, et al. Radiological response to nivolumab in patients with hepatocellular carcinoma: a multicenter analysis of real-life practice. Eur J Radiol. 2021;135:109484. pmid:33383399
  12. 12. Guo J, Tang Q. Recent updates on chimeric antigen receptor T cell therapy for hepatocellular carcinoma. Cancer Gene Ther. 2021;28(10-11):1075–87. pmid:33500535
  13. 13. Gao W, Ho M. The role of glypican-3 in regulating Wnt in hepatocellular carcinomas. Cancer Rep. 2011;1(1):14–9 pmid:22563565
  14. 14. Huang SL, Wang YM, Wang QY, Feng GG, Wu FQ, Yang LM, et al. Mechanisms and clinical trials of hepatocellular carcinoma immunotherapy. Front Genet. 2021;12:691391. pmid:34306031
  15. 15. Luxardi G, Galli A, Forlani S, Lawson K, Maina F, Dono R. Glypicans are differentially expressed during patterning and neurogenesis of early mouse brain. Biochem Biophys Res Commun. 2007;352(1):55–60. pmid:17107664
  16. 16. Zhou F, Shang W, Yu X, Tian J. Glypican-3: a promising biomarker for hepatocellular carcinoma diagnosis and treatment. Med Res Rev. 2018;38(2):741–67. pmid:28621802
  17. 17. Herb BR, Glover HJ, Bhaduri A, Casella AM, Bale TL, Kriegstein AR, et al. Preprint: Single-cell genomics reveals region-specific developmental trajectories underlying neuronal diversity in the prenatal human hypothalamus. bioRxiv. 2021;1–23.
  18. 18. Thistlethwaite FC, Gilham DE, Guest RD, Rothwell DG, Pillai M, Burt DJ, et al. The clinical efficacy of first-generation carcinoembryonic antigen (CEACAM5)-specific CAR T cells is limited by poor persistence and transient pre-conditioning-dependent respiratory toxicity. Cancer Immunol Immunother. 2017;66(11):1425–36. pmid:28660319
  19. 19. Morgan RA, Chinnasamy N, Abate-Daga D, Gros A, Robbins PF, Zheng Z, et al. Cancer regression and neurological toxicity following anti-MAGE-A3 TCR gene therapy. J Immunother. 2013;36(2):133–51. pmid:23377668
  20. 20. Brudno JN, Kochenderfer JN. Toxicities of chimeric antigen receptor T cells: recognition and management. Blood. 2016;127(26):3321–30. pmid:27207799
  21. 21. Zittermann SI, Capurro MI, Shi W, Filmus J. Soluble glypican 3 inhibits the growth of hepatocellular carcinoma in vitro and in vivo. Int J Cancer. 2010;126(6):1291–301. pmid:19816934
  22. 22. Feng M, Kim H, Phung Y, Ho M. Recombinant soluble glypican 3 protein inhibits the growth of hepatocellular carcinoma in vitro. Int J Cancer. 2011;128(9):2246–7. pmid:20617511
  23. 23. Sun L, Gao F, Gao Z, Ao L, Li N, Ma S, et al. Shed antigen-induced blocking effect on CAR-T cells targeting Glypican-3 in hepatocellular carcinoma. J ImmunoTher Cancer. 2021;9(4):e001875.
  24. 24. Uhlen M, Fagerberg L, Hallstrom BM, Lindskog C, Oksvold P, Mardinoglu A, et al. Proteomics. Tissue-based map of the human proteome. Science. 2015;347(6220):1260419. pmid:25613900
  25. 25. Uhlen M, Zhang C, Lee S, Sjostedt E, Fagerberg L, Bidkhori G, et al. A pathology atlas of the human cancer transcriptome. Science. 2017;357(6352):eaan2507. pmid:28818916
  26. 26. Uhlen M, Karlsson MJ, Hober A, Svensson AS, Scheffel J, Kotol D, et al. The human secretome. Sci Signal. 2019;12(609):eaaz0274. pmid:31772123
  27. 27. Thul PJ, Akesson L, Wiking M, Mahdessian D, Geladaki A, Ait Blal H, et al. A subcellular map of the human proteome. Science. 2017;356(6340):eaal3321. pmid:28495876
  28. 28. Sjostedt E, Zhong W, Fagerberg L, Karlsson M, Mitsios N, Adori C, et al. An atlas of the protein-coding genes in the human, pig, and mouse brain. Science. 2020;367(6482):eaay5947. pmid:32139519
  29. 29. Abugessaisa I, Ramilowski JA, Lizio M, Severin J, Hasegawa A, Harshbarger J, et al. FANTOM enters 20th year: expansion of transcriptomic atlases and functional annotation of non-coding RNAs. Nucleic Acids Res. 2021;49(D1):D892–8. pmid:33211864
  30. 30. Forrest AR, Kawaji H, Rehli M, et al; FANTOM Consortium and the RIKEN PMI and CLST (DGT). A promoter-level mammalian expression atlas. Nature. 2014;507(7493):462–70. pmid:24670764
  31. 31. GTEx Consortium. Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science. 2015;348(6235):648–60. pmid:25954001
  32. 32. GTEx Consortium. The Genotype-Tissue Expression (GTEx) project. Nat Genet. 2013;45(6):580–5. pmid:23715323
  33. 33. Weinstein JN, Collisson EA, Mills GB, Shaw KR, Ozenberger BA, et al; Cancer Genome Atlas Research Network. The cancer genome atlas pan-cancer analysis project. Nat Genet. 2013;45(10):1113–20. pmid:24071849
  34. 34. Hutter C, Zenklusen JC. The Cancer Genome Atlas: Creating Lasting Value beyond Its Data. Cell. 2018;173(2):283–5. pmid:29625045
  35. 35. Deutsch EW, Sun Z, Campbell D, Kusebauch U, Chu CS, Mendoza L, et al. State of the Human Proteome in 2014/2015 As Viewed through PeptideAtlas: Enhancing Accuracy and Coverage through the AtlasProphet. J Proteome Res. 2015;14(9):3461–73. pmid:26139527
  36. 36. Schwenk JM, Omenn GS, Sun Z, Campbell DS, Baker MS, Overall CM, et al. The Human Plasma Proteome Draft of 2017: Building on the Human Plasma PeptideAtlas from Mass Spectrometry and Complementary Assays. J Proteome Res. 2017;16(12):4299–310. pmid:28938075
  37. 37. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012;2(5):401–4. pmid:22588877
  38. 38. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. 2013;6(269):pl1. pmid:23550210
  39. 39. Jiang Y, Sun A, Zhao Y, Ying W, Sun H, Yang X, et al; Chinese Human Proteome Project (CNHPP) Consortium. Proteomics identifies new therapeutic targets of early-stage hepatocellular carcinoma. Nature. 2019;567(7747):257–61. pmid:30814741
  40. 40. Roessler S, Jia HL, Budhu A, Forgues M, Ye QH, Lee JS, et al. A unique metastasis gene signature enables prediction of tumor relapse in early-stage hepatocellular carcinoma patients. Cancer Res. 2010;70(24):10202–12. pmid:21159642
  41. 41. Zhao N, Dang H, Ma L, Martin SP, Forgues M, Ylaya K, et al. Intratumoral gammadelta T-Cell Infiltrates, Chemokine (C-C Motif) Ligand 4/Chemokine (C-C Motif) Ligand 5 Protein Expression and Survival in Patients With Hepatocellular Carcinoma. Hepatology. 2021;73(3):1045–60. pmid:32502310
  42. 42. Gao Q, Zhu H, Dong L, Shi W, Chen R, Song Z, et al. Integrated Proteogenomic Characterization of HBV-Related Hepatocellular Carcinoma. Cell. 2019;179(2):561–77.e22. pmid:31585088
  43. 43. Wong KK. Integrated transcriptomics and proteomics data analysis identifies CDH17 as a key cell surface target in colorectal cancer. Comput Biol Chem. 2023;105:107897. pmid:37247573
  44. 44. Parker KR, Migliorini D, Perkey E, Yost KE, Bhaduri A, Bagga P, et al. Single-Cell Analyses Identify Brain Mural Cells Expressing CD19 as Potential Off-Tumor Targets for CAR-T Immunotherapies. Cell. 2020;183(1):126–42.e17. pmid:32961131
  45. 45. Shalabi H, Nellan A, Shah NN, Gust J. Immunotherapy Associated Neurotoxicity in Pediatric Oncology. Front Oncol. 2022;12:836452. pmid:35265526
  46. 46. Karschnia P, Strubing F, Teske N, Blumenberg V, Bucklein VL, Schmidt C, et al. Clinicopathologic Findings in Fatal Neurotoxicity After Adoptive Immunotherapy With CD19-Directed CAR T-Cells. Hemasphere. 2021;5(3):e533. pmid:33615147
  47. 47. Morgan RA, Yang JC, Kitano M, Dudley ME, Laurencot CM, Rosenberg SA. Case report of a serious adverse event following the administration of T cells transduced with a chimeric antigen receptor recognizing ERBB2. Mol Ther. 2010;18(4):843–51. pmid:20179677
  48. 48. Portillo AL, Hogg R, Poznanski SM, Rojas EA, Cashell NJ, Hammill JA, et al. Expanded human NK cells armed with CAR uncouple potent anti-tumor activity from off-tumor toxicity against solid tumors. iScience. 2021;24(6):102619. pmid:34159300
  49. 49. Bidkhori G, Benfeitas R, Elmas E, Kararoudi MN, Arif M, Uhlen M, et al. Metabolic Network-Based Identification and Prioritization of Anticancer Targets Based on Expression Data in Hepatocellular Carcinoma. Front Physiol. 2018;9:916. pmid:30065658
  50. 50. Li B, Qing T, Zhu J, Wen Z, Yu Y, Fukumura R, et al. A Comprehensive Mouse Transcriptomic BodyMap across 17 Tissues by RNA-seq. Sci Rep. 2017;7(1):4200. pmid:28646208
  51. 51. Binder JX, Pletscher-Frankild S, Tsafou K, Stolte C, O’Donoghue SI, Schneider R, et al. COMPARTMENTS: unification and visualization of protein subcellular localization evidence. Database (Oxford). 2014;2014:bau012. pmid:24573882
  52. 52. Ben-Ari Fuchs S, Lieder I, Stelzer G, Mazor Y, Buzhor E, Kaplan S, et al. Geneanalytics: an integrative gene set analysis tool for next generation sequencing, RNAseq and microarray data. OMICS. 2016;20(3):139–51. pmid:26983021
  53. 53. Altevogt P, Sammar M, Huser L, Kristiansen G. Novel insights into the function of CD24: A driving force in cancer. Int J Cancer. 2021;148(3):546–59. pmid:32790899
  54. 54. Zhang P, Ha M, Li L, Huang X, Liu C. MicroRNA-3064-5p sponged by MALAT1 suppresses angiogenesis in human hepatocellular carcinoma by targeting the FOXA1/CD24/Src pathway. FASEB J. 2020;34(1):66–81. pmid:31914639
  55. 55. Wang Y, Chen M, Wu Z, Tong C, Dai H, Guo Y, et al. CD133-directed CAR T cells for advanced metastasis malignancies: A phase I trial. Oncoimmunology. 2018;7(7):e1440169. pmid:29900044
  56. 56. Liu F, Qian Y. The role of CD133 in hepatocellular carcinoma. Cancer Biol Ther. 2021;22(4):291–300. pmid:33899676
  57. 57. Zhang RY, Wei D, Liu ZK, Yong YL, Wei W, Zhang ZY, et al. Doxycycline Inducible Chimeric Antigen Receptor T Cells Targeting CD147 for Hepatocellular Carcinoma Therapy. Front Cell Dev Biol. 2019;7:233. pmid:31681766
  58. 58. Tian H, Huang Y, He J, Zhang M, Ni P. CD147 monoclonal antibody targeted reduction-responsive camptothecin polyphosphoester nanomedicine for drug delivery in hepatocellular carcinoma cells. ACS Appl Bio Mater. 2021;4(5):4422–31. pmid:35006854
  59. 59. Hwang HS, Yoo JE, Han DH, Choi JS, Lee JG, Joo DJ, et al. Circulating cancer stem cells expressing EpCAM/CD90 in hepatocellular carcinoma: a pilot study for predicting tumor recurrence after living donor liver transplantation. Gut Liver. 2022;16(3):443–55. pmid:34737243
  60. 60. Zhang P, Shi B, Gao H, Jiang H, Kong J, Yan J, et al. An EpCAM/CD3 bispecific antibody efficiently eliminates hepatocellular carcinoma cells with limited galectin-1 expression. Cancer Immunol Immunother. 2014;63(2):121–32. pmid:24177984
  61. 61. Jiang W, Li T, Guo J, Wang J, Jia L, Shi X, et al. Bispecific c-Met/PD-L1 CAR-T Cells Have Enhanced Therapeutic Effects on Hepatocellular Carcinoma. Front Oncol. 2021;11:546586. pmid:33777728
  62. 62. Yu J, Chen GG, Lai PBS. Targeting hepatocyte growth factor/c-mesenchymal-epithelial transition factor axis in hepatocellular carcinoma: Rationale and therapeutic strategies. Med Res Rev. 2021;41(1):507–24. pmid:33026703
  63. 63. Tang Q, Li X, Chen Y, Long S, Yu Y, Sheng H, et al. Solamargine inhibits the growth of hepatocellular carcinoma and enhances the anticancer effect of sorafenib by regulating HOTTIP-TUG1/miR-4726-5p/MUC1 pathway. Mol Carcinog. 2022;61(4):417–32. pmid:35040191
  64. 64. Kumar AR, Devan AR, Nair B, Nair RR, Nath LR. Biology, Significance and Immune Signaling of Mucin 1 in Hepatocellular Carcinoma. Curr Cancer Drug Targets. 2022;22(9):725–40. pmid:35301949
  65. 65. Lu Y, Yang A, Quan C, Pan Y, Zhang H, Li Y, et al. A single-cell atlas of the multicellular ecosystem of primary and metastatic hepatocellular carcinoma. Nat Commun. 2022;13(1):4594. pmid:35933472
  66. 66. Han X, Zhou Z, Fei L, Sun H, Wang R, Chen Y, et al. Construction of a human cell landscape at single-cell level. Nature. 2020;581(7808):303–9. pmid:32214235
  67. 67. Satija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol. 2015;33(5):495–502. pmid:25867923
  68. 68. Brown PJ, Wong KK, Felce SL, Lyne L, Spearman H, Soilleux EJ, et al. FOXP1 suppresses immune response signatures and MHC class II expression in activated B-cell-like diffuse large B-cell lymphomas. Leukemia. 2016;30(3):605–16. pmid:26500140
  69. 69. Wong KK, Gascoyne DM, Soilleux EJ, Lyne L, Spearman H, Roncador G, et al. FOXP2-positive diffuse large B-cell lymphomas exhibit a poor response to R-CHOP therapy and distinct biological signatures. Oncotarget. 2016;7(33):52940–56. pmid:27224915
  70. 70. Wong KK, Mervin LH, Mazzolari A, Bender A, Yaacob NS. Towards the mode of action of Strobilanthes crispus through integrated computational and experimental analyses. J Plant Biochem Biotechnol. 2017;26(4):451–66.
  71. 71. Mohd Shukri ND, Farah Izati A, Wan Ghazali WS, Che Hussin CM, Wong KK. CD3(+)CD4(+)gp130(+) T Cells Are Associated With Worse Disease Activity in Systemic Lupus Erythematosus Patients. Front Immunol. 2021;12:675250. pmid:34149710
  72. 72. Nur Husna SM, Md Shukri N, Tuan Sharif SE, Tan HTT, Mohd Ashari NS, Wong KK. IL-4/IL-13 Axis in Allergic Rhinitis: Elevated Serum Cytokines Levels and Inverse Association With Tight Junction Molecules Expression. Front Mol Biosci. 2022;9:819772. pmid:35372516
  73. 73. Balasubramaniam SD, Wong KK, Oon CE, Balakrishnan V, Kaur G. Comparative transcriptomic profiling in HPV-associated cervical carcinogenesis: Implication of MHC class II and immunoglobulin heavy chain genes. Life Sci. 2020;256:118026. pmid:32615187
  74. 74. Tam PKH, Yiu RS, Lendahl U, Andersson ER. Cholangiopathies - Towards a molecular understanding. EBioMedicine. 2018;35:381–93. pmid:30236451
  75. 75. Tang Z, Kang B, Li C, Chen T, Zhang Z. GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis. Nucleic Acids Res. 2019;47(W1):W556–60. pmid:31114875
  76. 76. Azevedo ALK, Gomig THB, Giner IS, Batista M, Marchini FK, Lima RS, et al. Comprehensive analysis of the large and small ribosomal proteins in breast cancer: Insights on proteomic and transcriptomic expression patterns, regulation, mutational landscape, and prognostic significance. Comput Biol Chem. 2022;100:107746. pmid:35961236
  77. 77. Zhang Z, Sun C, Liu ZP. Discovering biomarkers of hepatocellular carcinoma from single-cell RNA sequencing data by cooperative games on gene regulatory network. J Comput Sci. 2022;65.
  78. 78. Fu F, Yang X, Zheng M, Zhao Q, Zhang K, Li Z, et al. Role of transmembrane 4 L Six family 1 in the development and progression of cancer. Front Mol Biosci. 2020;7:202. pmid:33015133
  79. 79. Deng Y, Cai S, Shen J, Peng H. Tetraspanins: Novel Molecular Regulators of Gastric Cancer. Front Oncol. 2021;11:702510. pmid:34222025
  80. 80. Robert JH, Amoussou NG, Mai HL, Loge C, Brouard S. Tetraspanins: useful multifunction proteins for the possible design and development of small-molecule therapeutic tools. Drug Discov Today. 2021;26(1):56–68. pmid:33137483
  81. 81. Wright MD, Ni J, Rudy GB. The L6 membrane proteins--a new four-transmembrane superfamily. Protein Sci. 2000;9(8):1594–600. pmid:10975581
  82. 82. Kummer D, Steinbacher T, Schwietzer MF, Tholmann S, Ebnet K. Tetraspanins: integrating cell surface receptors to functional microdomains in homeostasis and disease. Med Microbiol Immunol. 2020;209(4):397–405. pmid:32274581
  83. 83. Wice BM, Gordon JI. A tetraspan membrane glycoprotein produced in the human intestinal epithelium and liver that can regulate cell density-dependent proliferation. J Biol Chem. 1995;270(37):21907–18. pmid:7665614
  84. 84. Subramanian VS, Nabokina SM, Said HM. Association of TM4SF4 with the human thiamine transporter-2 in intestinal epithelial cells. Dig Dis Sci. 2014;59(3):583–90. pmid:24282057
  85. 85. Liu Z, Zhao M, Yokoyama KK, Li T. Molecular cloning of a cDNA for rat TM4SF4, a homolog of human il-TMP (TM4SF4), and enhanced expression of the corresponding gene in regenerating rat liver(1). Biochim Biophys Acta. 2001;1518(1-2):183–9. pmid:11267677
  86. 86. Muraro MJ, Dharmadhikari G, Grun D, Groen N, Dielen T, Jansen E, et al. A Single-Cell Transcriptome Atlas of the Human Pancreas. Cell Syst. 2016;3(4):385–94.e3. pmid:27693023
  87. 87. Ghazvini Zadeh EH, Huang Z, Xia J, Li D, Davidson HW, Li WH. ZIGIR, a Granule-Specific Zn(2+) Indicator, Reveals Human Islet alpha Cell Heterogeneity. Cell Rep. 2020;32(2):107904. pmid:32668245
  88. 88. Li Y, Wang L, Qiu J, Da L, Tiollais P, Li Z, et al. Human tetraspanin transmembrane 4 superfamily member 4 or intestinal and liver tetraspan membrane protein is overexpressed in hepatocellular carcinoma and accelerates tumor cell growth. Acta Biochim Biophys Sin (Shanghai). 2012;44(3):224–32. pmid:22236579
  89. 89. Wang L, Feng J, Da L, Li Y, Li Z, Zhao M. Adenovirus-mediated delivery of siRNA targeting TM4SF4 attenuated liver cancer cell growth in vitro and in vivo. Acta Biochim Biophys Sin (Shanghai). 2013;45(3):213–9. pmid:23296076
  90. 90. Mou Y, Ding X. LncRNA ST8SIA6-AS1 facilitates hepatocellular carcinoma progression by governing miR-651-5p/TM4SF4 axis. Anticancer Drugs. 2022;33(8):741–51. pmid:35946523
  91. 91. Li H, Zhong A, Li S, Meng X, Wang X, Xu F, et al. The integrated pathway of TGFbeta/Snail with TNFalpha/NFkappaB may facilitate the tumor-stroma interaction in the EMT process and colorectal cancer prognosis. Sci Rep. 2017;7(1):4915. pmid:28687755
  92. 92. Zhang F, Ye J, Guo W, Zhang F, Wang L, Han A. TYMS-TM4SF4 axis promotes the progression of colorectal cancer by EMT and upregulating stem cell marker. Am J Cancer Res. 2022;12(3):1009–26 pmid:35411242
  93. 93. Huang MY, Wang HM, Chang HJ, Hsiao CP, Wang JY, Lin SR. Overexpression of S100B, TM4SF4, and OLFM4 genes is correlated with liver metastasis in Taiwanese colorectal cancer patients. DNA Cell Biol. 2012;31(1):43–9. pmid:22011044
  94. 94. Choi SI, Kim SY, Lee J, Cho EW, Kim IG. TM4SF4 overexpression in radiation-resistant lung carcinoma cells activates IGF1R via elevation of IGF1. Oncotarget. 2014;5(20):9823–37. pmid:25344917
  95. 95. Choi SI, Kim SY, Lee JH, Kim JY, Cho EW, Kim IG. Osteopontin production by TM4SF4 signaling drives a positive feedback autocrine loop with the STAT3 pathway to maintain cancer stem cell-like properties in lung cancer cells. Oncotarget. 2017;8(60):101284–97. pmid:29254164
  96. 96. Gamerith G, Rainer J, Huber JM, Hackl H, Trajanoski Z, Koeck S, et al. 3D-cultivation of NSCLC cell lines induce gene expression alterations of key cancer-associated pathways and mimic in-vivo conditions. Oncotarget. 2017;8(68):112647–61. pmid:29348853
  97. 97. Jung K, Choi JS, Koo BM, Kim YJ, Song JY, Sung M, et al. TM4SF4 and LRRK2 Are Potential Therapeutic Targets in Lung and Breast Cancers through Outlier Analysis. Cancer Res Treat. 2021;53(1):9–24. pmid:32972043
  98. 98. Liang X, Peng J, Chen D, Tang L, Liu A, Fu Z, et al. Identification of novel hub genes and lncRNAs related to the prognosis and progression of pancreatic cancer by microarray and integrated bioinformatics analysis. Gland Surg. 2021;10(3):1104–17. pmid:33842254
  99. 99. Liu J, Zheng M, Qi Y, Wang H, Liu M, Liu Q, et al. Lewis(y) antigen-mediated positive feedback loop induces and promotes chemotherapeutic resistance in ovarian cancer. Int J Oncol. 2018;53(4):1774–86. pmid:30066907
  100. 100. Kadenbach B. Complex IV - The regulatory center of mitochondrial oxidative phosphorylation. Mitochondrion. 2021;58:296–302. pmid:33069909
  101. 101. Zhang H, Yan Q, Wang X, Chen X, Chen Y, Du J, et al. The role of mitochondria in liver ischemia-reperfusion injury: from aspects of mitochondrial oxidative stress, mitochondrial fission, mitochondrial membrane permeable transport pore formation, mitophagy, and mitochondria-related protective measures. Oxid Med Cell Longev. 2021;2021:6670579. pmid:34285766
  102. 102. Robson B. Computers and preventative diagnosis. A survey with bioinformatics examples of mitochondrial small open reading frame peptides as portents of a new generation of powerful biomarkers. Comput Biol Med. 2021;140:105116. pmid:34896883
  103. 103. An P, Wei LL, Zhao S, Sverdlov DY, Vaid KA, Miyamoto M, et al. Hepatocyte mitochondria-derived danger signals directly activate hepatic stellate cells and drive progression of liver fibrosis. Nat Commun. 2020;11(1):2362. pmid:32398673
  104. 104. Neuman MG, Seitz HK, Tuma PL, Osna NA, Casey CA, Kharbanda KK, et al. Alcohol: basic and translational research; 15th annual Charles Lieber &1st Samuel French satellite symposium. Exp Mol Pathol. 2022;126:104750. pmid:35192844
  105. 105. Lee HY, Nga HT, Tian J, Yi HS. Mitochondrial Metabolic Signatures in Hepatocellular Carcinoma. Cells. 2021;10(8):1901. pmid:34440674
  106. 106. Westermann B. The mitochondria-plasma membrane contact site. Curr Opin Cell Biol. 2015;35:1–6. pmid:25801776
  107. 107. Frieden M, Arnaudeau S, Castelbou C, Demaurex N. Subplasmalemmal mitochondria modulate the activity of plasma membrane Ca2+-ATPases. J Biol Chem. 2005;280(52):43198–208. pmid:16216868
  108. 108. Fowler SL, Akins M, Zhou H, Figeys D, Bennett SA. The liver connexin32 interactome is a novel plasma membrane-mitochondrial signaling nexus. J Proteome Res. 2013;12(6):2597–610. pmid:23590695
  109. 109. Poolsri WA, Phokrai P, Suwankulanan S, Phakdeeto N, Phunsomboon P, Pekthong D, et al. Combination of Mitochondrial and Plasma Membrane Citrate Transporter Inhibitors Inhibits De Novo Lipogenesis Pathway and Triggers Apoptosis in Hepatocellular Carcinoma Cells. Biomed Res Int. 2018;2018:3683026. pmid:29546056
  110. 110. Qiu J, Liu Z, Da L, Li Y, Xuan H, Lin Q, et al. Overexpression of the gene for transmembrane 4 superfamily member 4 accelerates liver damage in rats treated with CCl4. J Hepatol. 2007;46(2):266–75. pmid:17069928
  111. 111. Wyler E, Mosbauer K, Franke V, Diag A, Gottula LT, Arsie R, et al. Transcriptomic profiling of SARS-CoV-2 infected human cell lines identifies HSP90 as target for COVID-19 therapy. iScience. 2021;24(3):102151. pmid:33585804
  112. 112. Vastrad BM, Vastrad CM. Bioinformatics analysis of expression profiling by high throughput sequencing for identification of potential key genes among SARS-CoV-2/COVID 19. 2021.
  113. 113. Deng W, Zhu P, Xu H, Hou X, Chen W. Classification and prognostic characteristics of hepatocellular carcinoma based on glycolysis cholesterol synthesis axis. J Oncol. 2022;2022:2014625. pmid:36213830
  114. 114. Zhang M, Yang H, Wan L, Wang Z, Wang H, Ge C, et al. Single-cell transcriptomic architecture and intercellular crosstalk of human intrahepatic cholangiocarcinoma. J Hepatol. 2020;73(5):1118–30. pmid:32505533
  115. 115. Haanen J, Carbonnel F, Robert C, Kerr KM, Peters S, Larkin J, et al. Management of toxicities from immunotherapy: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28(suppl_4): iv119–iv142.
  116. 116. Murayama A, Tajiri K, Nakaya A, Ito H, Hayashi Y, Entani T, et al. Intrahepatic bile duct injury as a hepatic immune-related adverse event after immune-checkpoint inhibitor treatment. Case Rep Gastroenterol. 2021;15(2):645–51. pmid:34616270
  117. 117. Wendt A, Eliasson L. Pancreatic alpha-cells - The unsung heroes in islet function. Semin Cell Dev Biol. 2020;103:41–50. pmid:31983511
  118. 118. Roder PV, Wu B, Liu Y, Han W. Pancreatic regulation of glucose homeostasis. Exp Mol Med. 2016;48(3):e219. pmid:26964835