In-silico analysis of TMEM2 as a pancreatic adenocarcinoma and cancer-associated fibroblast biomarker, and functional characterization of NSC777201, for targeted drug development

Abstract

Pancreatic adenocarcinoma (PAAD), known as one of the deadliest cancers, is characterized by a complex tumor microenvironment, primarily comprised of cancer-associated fibroblasts (CAFs) in the extracellular matrix. These CAFs significantly alter the matrix by interacting with hyaluronic acid (HA) and the enzyme hyaluronidase, which degrades HA - an essential process for cancer progression and spread. Despite the critical role of this interaction, the specific functions of CAFs and hyaluronidase in PAAD development are not fully understood. Our study investigates this interaction and assesses NSC777201, a new anti-cancer compound targeting hyaluronidase. This research utilized computational methods to analyze gene expression data from the Gene Expression Omnibus (GEO) database, specifically GSE172096, comparing gene expression profiles of cancer-associated and normal fibroblasts. We conducted in-house sequencing of pancreatic cancer cells treated with NSC777201 to identify differentially expressed genes (DEGs) and performed functional enrichment and pathway analysis. The identified DEGs were further validated using the TCGA-PAAD and Human Protein Atlas (HPA) databases for their diagnostic, prognostic, and survival implications, accompanied by Ingenuity Pathway Analysis (IPA) and molecular docking of NSC777201, in-vitro, and preclinical in-vivo validations. The result revealed 416 DEGs associated with CAFs and 570 DEGs related to NSC777201 treatment, with nine overlapping DEGs. A key finding was the transmembrane protein TMEM2, which strongly correlated with FAP, a CAF marker, and was associated with higher-risk groups in PAAD. NSC777201 treatment showed inhibition of TMEM2, validated by rescue assay, indicating the importance of targeting TMEM2. Further analyses, including IPA, demonstrated that NSC777201 regulates CAF cell senescence, enhancing its therapeutic potential. Both in-vitro and in-vivo studies confirmed the inhibitory effect of NSC777201 on TMEM2 expression, reinforcing its role in targeting PAAD. Therefore, TMEM2 has been identified as a theragnostic biomarker in PAAD, influenced by CAF activity and HA accumulation. NSC777201 exhibits significant potential in targeting and potentially reversing critical processes in PAAD progression, demonstrating its efficacy as a promising therapeutic agent.

Introduction

Pancreatic cancer (PC) is one of the most aggressive and deadliest malignant neoplasms worldwide, resulting from the abnormal and uncontrolled growth of cells in the pancreatic tissue. It is the second leading cause of mortality among malignant cancer-associated diseases [1,2]. Pancreatic adenocarcinoma (PAAD), the commonest form of PC, accounts for approximately 85% of all types of PC and is connected with a poor prognosis [3].

Treatment approaches for PC include surgical resection, radiotherapy, chemotherapy, neoadjuvant therapy, immunology, and targeted molecular therapy, alone or in combination [4]. Overall 5-year survival rates of PC are 34% when cancer remains local and grows in the pancreas, 12% when cancer has spread to nearby lymph tissues and 3% when cancer has metastasized to other organs and lymph nodes [5]. PC has a higher recurrence rate and lower disease-free survival (DFS) even in patients receiving adjuvant chemotherapy after surgical resection [6]. As the overall survival (OS) rate of patients with early tumor recurrence is significantly lower than those of patients without early tumor recurrence [7], it is important to discover novel strategies for predicting recurrence.

Recent studies have shifted focus towards the tumor microenvironment (TME) of pancreatic adenocarcinoma (PAAD), underscoring its complexity with a composition of inflammatory cells, fibroblasts, immune cells, and growth factors within the extracellular matrix (ECM). This dense stromal tissue, accounting for 15% to 85% of the entire tumor component of PAAD [8,9], plays a pivotal role in tumor proliferation, invasion, and the ability to metastasize, largely influenced by the ECM remodelling by fibroblasts [10]. Cancer-associated fibroblasts (CAFs) within the ECM, crucial for promoting tumor progression and chemoresistance, release growth factors and cytokines that stimulate tumor growth and metastasis [11-14], also contributing to the poor prognosis of PAAD by fostering an immunosuppressive environment [15-18]. With a deeper understanding of the TME of PC, TME-based translational therapies may be a breakthrough in future PC treatments. Hyaluronic acid (HA), a major ECM component, crucial in many cancers [19-21], influences cell adhesion, migration, and proliferation, and is associated with a malignant phenotype [22-28]. The degradation of HA by enzymes, including the novel hyaluronidase transmembrane protein 2 (TMEM2), also known as CEMIP2 (cell migration-inducing hyaluronidase 2), impacts a wide variety of cancers [29-35], with its role in PAAD still to be fully understood. Despite advancements in bioinformatics, the association of TMEM2 with the PAAD-TME, especially its interaction with CAFs, remains underexplored. Investigating TMEM2’s role could unveil new therapeutic targets, potentially improving treatment outcomes for PAAD.

Our research previously identified a series of tetracyclic heterocyclic azathioxanthones with significant cytotoxic effects against cancer cells as multi-kinase inhibitors [36], noting the diversity of this representative scaffold of the small molecule [37-40]. In our ongoing study, we focus on evaluating the candidate, NSC777201’s ability to target pancreatic adenocarcinoma (PAAD) by inhibiting the HA-enzyme, TMEM2, which we found to be overexpressed in PAAD through analysis of public GEO database gene expression profiles of cancer-associated fibroblasts (CAFs) and normal fibroblasts (NFs). Treatment with NSC777201 significantly decreased TMEM2 expression in Panc1 cells, a finding supported by data from The Cancer Genome Atlas (TCGA)-PAAD and the Human Protein Atlas (HPA), which revealed overexpression of TMEM2 at both mRNA and protein levels in pancreatic adenocarcinoma (PAAD) compared to normal tissue. Co-expression analysis revealed a strong association between TMEM2 and cancer-associated fibroblast (CAF) markers, particularly the fibroblast activation protein (FAP). Ingenuity Pathway Analysis (IPA) of cells treated with NSC777201 indicated a reduction in fibroblast activity and an induction of senescence, highlighting the compound’s effectiveness. This was further supported by in silico molecular docking, which demonstrated robust binding of NSC777201 to TMEM2. In-vitro assays and rescue experiments confirmed that NSC777201 specifically targets TMEM2. This targeting efficacy was also reinforced by pre-clinical in-vivo studies using a mouse model, confirming NSC777201’s potential to alter the PAAD tumor microenvironment (TME) by inhibiting TMEM2 and modulating CAF activity, ultimately reducing tumor growth. These results underscore the potential of NSC777201 as a promising therapeutic candidate for PAAD, warranting further development.

Materials and methods

Acquisition of RNA expression dataset

RNA expression data from GSE172096, including five CAF and three NF samples derived from human pancreatic ductal adenocarcinoma, were downloaded from the publicly available GEO database (http://www.ncbi.nlm.nih.gov/geo/). For further validation, the RNA-sequencing raw counts, and clinical data of patients with a pancreatic tumor (PAAD) and normal samples (n=178) were downloaded from TCGA (https://portal.gdc.cancer.gov/) database, stored, and used. Furthermore, online analysis databases, which use TCGA data, such as (a) http://gdac.broadinstitute.org/, (b) http://gepia.cancer-pku.cn/, and (c) http://ualcan.path.uab.edu/cgi-bin/ualcan-res.pl, were used for analysis and visualization of key genes.

Identification of DEGs

Expression raw data and annotation matrix were downloaded from the GEO database, the circular(circ) RNA IDs were correlated and matched with parent genes, circRNA and parental genes were closely associated. The expression data in the expression matrix were analyzed with DESeq2, an R package [41], a Bioconductor package for differentially expressed genes (DEGs) analysis of expression data, to determine the DEGs in between CAFs and NFs dataset, the criteria of |log2fold change| >1.25 and the adjusted p values of <0.05 were used. With thousands of genes tested, multiple comparison adjustments were necessary so, the Bonferroni method was applied for filtering DEGs; this controls the mean number of false positives, that can be used for multiplicity adjustment [42]. Hierarchical clustering with a heatmap and a principal component analysis (PCA) were respectively generated using Heatmap.2 and the scatterplot3d function tool in R package gplots [43].

GO enrichment and kyoto encyclopedia of genes and genomes (KEGG) analysis of DEGs

GO, KEGG pathway, and cnet plot enrichment analyses were performed for CAF- vs. NF-identified DEGs using clusterProfiler, an R package [44].

Cell culture and reagents

The human pancreatic cancer cell line, PANC1 and SUIT2 was obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA). The cell line was cultured in Dulbecco’s modified Eagle medium (#12491023; GIBCO, Life Technologies Corp., Carlsbad, CA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) and antibiotics (100 µg/ml streptomycin and 100 IU/ml penicillin) at 37°C in a 5% CO₂ incubator. To maintain the cell lines, subcultures were performed every 48~72 hrs for maintenance. NSC777201 is one of our in-house drugs synthesized as previously described in a US patent application (H.S. Huang, D.S. Yu, T.C. Chen, Vol. US Patent No. 8,927,717B1, US, Jan. 6, 2015) [36]. For a stock solution, NSC777201 was dissolved in 10 mM dimethyl sulfoxide (DMSO; Sigma Aldrich, St. Louis, MO, USA) and kept at -20°C. The stock solution was further immediately diluted in a sterile medium to the required concentrations.

Sulforhodamine B (SRB) assay

PC cells (PANC1 and SUIT2) were seeded in 96-well plates (at 3000 cells/well). After the cells had attached to the plate (by 24 h of incubation), cells were randomly divided into control and treatment groups. The control group was treated with DMSO, while the treatment groups were treated with different doses of NSC777201. After 48 h of incubation, the medium was removed, and 100 μL 10% trichloroacetic acid (TCA) was added to each well. After incubation for 1 h at 4°C, TCA was removed, and 100 μL of the SRB reagent was added to each well followed by incubation for 1 h at room temperature. All plates were washed with 1% acetic acid. After the plates were dried in an oven for 20 min at 60°C, 200 μL of Tris (20 nM) was added to each well. The absorbance was measured using spectrophotometry (at a wavelength of 565 nm). Absorbance values are reported as percent (%) cell viability (of treatment groups relative to the control group).

Tumor sphere-formation assay

The tumor sphere-formation assay was performed according to a previously described method with modifications. In short, PC cells (PANC1 and SUIT2) were seeded (2500 cells/well) in six-well ultra-low attachment plates (Corning, Corning, NY, USA) in serum-free media consisting of Dulbecco’s modified Eagle medium (DMEM)/Ham’s F12 (1:1), human epidermal growth factor (hEGF, 20 ng/ml), basic fibroblast growth factor (bFGF; 10 ng/ml (PeproTech, Rocky Hill, NJ, USA), 2 μg/ml 0.2% heparin (Sigma, St. Louis, MO, USA), and 1% penicillin/streptomycin (P/S, 100 U/ml, Hyclone, Logan, UT, USA)). Cells were then allowed to aggregate and grow for at least 7 days. Cells (diameter >50 µm), characterized by compact, non-adherent spheroid-like masses, were considered a tumor-sphere and counted with an inverted phase-contrast microscope.

Wound-healing migration assay

PC cells were resuspended in a complete medium, plated in individual culture-inserts (ibdi, Munich, Germany), appropriated for a 2D migration assay, and maintained at 37°C in a 5% CO₂ atmosphere until confluence. These culture inserts were composed of two chambers separated by a biocompatible silicone material, which after removal allowed cells from each edge to migrate towards the center of the gap. After the barrier was removed, confluent cancer cell monolayers were washed with PBS to remove non-adherent cells and treated with NSC777201. Treated and untreated cells were maintained at 37°C in a 5% CO₂ atmosphere for 24 h. Cell migration was evaluated every 2 h with the BioTek Lionheart FX automated cell imaging system to capture and monitor wound closure with a phase-contrast microscope.

Gene expression sequencing and siRNA knockdown analysis

After NSC777201 treatment, Panc1 cells were collected in TRIzol reagent, and total RNA was isolated and purified using a TRIzol-based protocol (Life Technologies, USA) as per the manufacturer’s instructions. The RNA concentration and purity were determined with a NanoDrop 1000 spectrophotometer (Nyxor Biotech, Paris, France). Two micrograms of total RNA were sent to Welgen Biotech Taiwan (New Taipei City, Taiwan; https://www.welgene.com.tw/main) for sequencing. The experimental flow for sequencing is illustrated in Figure S1. Analysis was performed as raw intensity with background correction, then quantile normalized intensity between samples done, for differential expression analysis between treated and control samples, was analyzed with limma, an R package, a Bioconductor packages for differentially expressed genes (DEGs) analysis of expression data, to determine the DEGs in between NSC777201 treated and control dataset, the criteria of |log2fold change| >1.2 was applied. Furthermore, for RT-PCR analysis, one microgram of total RNA was reverse-transcribed using a Qiagen OneStep RT-PCR Kit (Qiagen), and the PCR was performed using a Rotor-Gene SYBR Green PCR Kit (400, Qiagen). Details of qPCR primers used for this study are listed in Table S1. siRNA-mediated mRNA knockdown, a population of 1 × 10*6 PC cells was cultured on a 10 cm plate. On the following day, siRNAs specifically aimed at TMEM2 (sourced from Integrated DNA Technologies (IDT), siRNA#1, and siRNA#2) were introduced into the cells along with a control siRNA (procured from Thermo Scientific; Control siRNA). This was completed using the Lipofectamine reagent, adhering strictly to the guidelines provided by the manufacturer. After allowing 48 hours post-transfection for the process to take effect, the cells were harvested.

For TMEM2 stable and longer expression a construction of TMEM2 Overexpression and Silencing Vectors, and Transfection Procedures, a TMEM2 overexpression (TMEM2-OE) vector was created by inserting the TMEM2 coding sequence into the pcDNA3.1+ vector. For shRNA-mediated TMEM2 silencing, the specific shRNA sequence (shTMEM2: GTGAGAAACTATGAAAATCATAG) was incorporated into the pLKO.1 vector (Genepharm). In overexpression studies, Panc1 cells were cultured until they reached 70 to 90% confluency and then transfected with the TMEM2 overexpression vector using Lipofectamine 2000 reagent (Thermo Fisher Scientific) over 48 hours. For knockdown assays, Panc1 were grown to 80% confluency and transfected either with scramble shRNA (control) or shTMEM2 utilizing the same Lipofectamine 2000 reagent for 48 hours. Post-transfection, Overexpression and silencing of the TMEM2 gene in Panc1 cells were confirmed by quantitative RT-PCR and Western blotting, then the cells were collected and subsequently, these cells were utilized in various assays.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blot analysis

Total protein lysates from PC cells (parental, tumor-spheres, and transfected) were extracted after treatment in different experiments and were separated by SDS-PAGE using the Mini-Protean III system (Bio-Rad, Taiwan) and transferred onto polyvinylidene difluoride membranes using the Trans-Blot Turbo Transfer System (Bio-Rad). Membranes were incubated with the primary antibody to react overnight at 4°C. Details of the primary antibody and dilutions used for these studies are listed in Table S1. Then membranes were incubated with the horseradish peroxidase-labelled secondary antibody. Proteins of interest were detected and visualized using enhanced chemiluminescence (ECL) detection kits (ECL Kits; Amersham Life Science, NJ, USA). Images were captured and analyzed using the UVP BioDoc-It system (Upland, CA, USA).

Survival, risk score and diagnostic/prognostic significance of DEGs

Survival analysis and prognostic value of the messenger (m)RNA in the PAAD signature for both survival and risk between groups were analyzed using the online database SurvExpress (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp) [45]. High- and low-risk groups were divided by the risk score algorithm embedded in the platform. The pROC R package [46] was used to plot receiver operating characteristic (ROC) curves for key genes in PAAD, and values of the area under the curve (AUC) were calculated to assess their diagnostic values in PAAD.

The HPA: immunohistochemical (IHC) quantification, visualization, and subcellular localization

Expressions of key proteins and their correlations at the protein level were checked on the HPA (HPA; http://www.proteinatlas.org/) using the R package programs, hpar and HPAanalyze [47].

Ingenuity pathway analysis (IPA)

Ingenuity Pathway Analysis (IPA) was run to identify the canonical pathway network [48], accessed on 6^th Nov 2021, it uses the popular activation z-score analytical method, as proposed by Kraèmer et al. in 2014 [48], which measures activation states (either increased or decreased) of pathways affected by the DEGs. We used a statistical method to define a quantitative z-score, which determines whether the biological function has been significantly more “increased” predictions than “decreased” predictions (z-score >0) or vice versa (z-score <0). In general practice, an absolute z-score of >2 or <-2 may be significant.

Molecular docking analysis

Studies were performed using the automated CB-Dock server (http://clab.labshare.cn/cb-dock/php/index.php; accessed on 10 November 2021) [49] with default parameters to investigate interactions between NSC777201 and TMEM2. The crystal structure of TMEM2 was not available; therefore, homologous modelling (comparative protein structure modelling) was used to find the three-dimensional (3D) structure of the TMEM2 protein obtained from the uniport database (https://www.uniprot.org/uniprot/Q9UHN6; accessed on 8 November 2021), and the NSC777201 3D structure was drawn in Sybyl mol2 using the Avogadro molecular builder and visualization tool vers. 1.1.0 [50]. Once the molecular docking experiments were completed and five configurations for each protein-ligand complex were generated for NSC777201 and TMEM2, the lowest binding affinity (kcal/mol) complex was considered to be the most stable docking pose. The interaction between the ligand and proteins was then prepared, visualized, and analyzed using the Discovery studio visualizer vers. 21.1.0.20298 (BIOVIA, San Diego, CA, USA) [51].

Animal studies

All the animal experiments and maintenance complied with the Animal Use Protocol at Taipei Medical University (protocol LAC-2017-0161). Five-week-old female NOD/SCID mice were purchased from BioLASCO (Taipei, Taiwan). The mice were maintained under pathogen-free conditions and were provided with sterilized food and water. Cells (1 × 10⁶) were suspended in 0.2 mL serum-free DMEM and were injected subcutaneously into the right flank near the hind leg of each mouse. Tumor size was calculated using the formula V = width² × length/2. When the tumors became palpable (the tumor volume was ~100 mm³), the mice were then randomly divided into three groups, i.e. control, sh-TMEM2 and NSC777201 (NSC777201, 10 mg/kg, five times/week) treated (only) group. The changes in the tumor volume, body weight (BW), and survival were monitored and recorded every week. The animals were humanely euthanized after the experiments were terminated, and the tumor samples were harvested for further analysis.

Statistical analysis

GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA) was used to draw the figures. Pearson’s Chi-squared test was used to compare categorical variables. The student’s t-test was used to analyze the normal distribution of continuous variables. Kaplan-Meier (KM) plots with the log-rank test were used to estimate survival differences. The diagnostic significance of selected genes for PAAD was evaluated by the ROC curve. Statistical significance was indicated p or adjusted P<0.05.

Results

Identification of DEGs between PAAD-derived CAFs and NFs

To determine the gene expression patterns between PAAD-derived CAFs and NFs from the publicly available GEO database (GSE172096, n=8 total samples), details of sample and sequencing platform shown in Table S2, hierarchical clustering analysis of top genes was performed; data demonstrated that as shown in the heatmap, Figure 1A. Principal component analysis (PCA) was utilized to visualize the spatial distribution of the samples, which distinguishes CAF samples from NF samples (Figure 1B), In total, 416 DEGs were obtained, including 339 upregulated and 77 downregulated DEGs in CAFs vs. NFs, based on the cut-off criteria |log2FC| >1.25 and p-adj. <0.05, shown in the volcano and heatmap of Figure 1C, 1D. The complete list of DEGs is available in Table S3, all DEGs were included in the further analysis. The identified DEGs in CAFs and NFs were further analyzed to identify the associated GO and KEGG pathways, using the “clusterProfiler” package [53]. The GO enrichment analysis classified the DEGs into three functional groups, including biological processes (BPs), cellular components (CCs), and molecular functions (MFs) (The Gene Ontology Consortium, 2018) [54]. As shown in Figure S2A-C. In the BP category, the top three most enriched terms were “DNA integrity checkpoint”, “DNA replication”, and “chromatin remodelling”. In the CC category, the top three most enriched terms were “cell leading edge”, “lamellipodium”, and “microtubule end”. In the MF category, the top three most enriched terms were “ATPase activity”, “Ras GTPase binding”, and “guanyl-nucleotide exchange factor activity”. Moreover, as shown in Figure S2D, the top three most enriched terms in the KEGG analysis were “Pathways in cancer”, “N-glycan biosynthesis”, and “Protein processing in endoplasmic reticulum”. The GO-all, as shown in Figure S2E “cnetplot”, depicts linkages of genes and biological concepts as a network, with “positive regulation of transcription”, “DNA-templated”, “protein targeting to peroxisome”, “peroxisomal matrix”, and “ubiquitin-protein transferase” being key pathways demonstrating roles in PAAD.

Figure 1.

Identification of the candidate differentially expressed genes (DEGs) between cancer-associated fibroblasts (CAFs) and normal fibroblasts (NFs) samples. A. The expression heatmap and hierarchical clustering (n=8) of DEGs in GSE172096 datasets of human pancreatic ductal adenocarcinoma (PAAD). B. Principal component analysis (PCA). C, D. Volcano plot and heatmap in microarray representing top DEGs. DEGs are represented by satisfying the criteria of absolute log2fold changes (|log2FC|) value >1.25 and P<0.05.

Overview of DEGs modulated by NSC777201

Before NSC777201 treatment on Panc1 cells, we identified the novel potential of NSC777201, as a drug candidate, the online SwissADME algorithm developed by the Swiss Institute of Bioinformatics (http://www.swissadme.ch/index.php; assessed on 23^rd January 2022) and ADMETlab 2.0 (https://admetmesh.scbdd.com/service/screening/molecule) was used to predict the PKs, drug-likeness, and adsorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of NSC777201, as shown in Figure S3, the NSC77201 pass the criteria to belong as a drug candidate.

To further investigate changes in gene expressions associated with NSC777201 treatment of Panc1 cells (Figure 2A), the percentage growth inhibition effect and 50% inhibitory concentration (IC50) values are shown in Figure S4, and microarray analysis was conducted for a global gene expression pattern analysis (Figure 2B). In total, 570 DEGs were obtained, including 320 upregulated and 250 downregulated DEGs by applying the |log2FC| >1.2 criteria, as shown by a heatmap in Figure 2B, and a complete list of total NSC777201-related DEGs is shown in Table S4. Overlapping common genes between NSC777201-related DEGs and CAFs-DEGs are shown in a Venn diagram (Figure 2C), and a divergent stacked bar demonstrates nine DEGs that overlapped between NSC777201-related DEGs and CAFs-DEGs according to the fold change values (Figure 2D). These nine overlapping DEGs were used for further analyses.

Figure 2.

Microarray expression analysis of NSC777201 treated Panc1 cells. A. Molecular structure of NSC777201 and treatment of Panc1 cells. B. Heatmap displaying the DEGs between NSC777201-treated and control untreated cells. DEGs are represented by satisfying the criteria of |log2FC| >1.25. C. Venn diagram indicating overlapping DEGs between public dataset cancer-associated fibroblasts (CAFs) vs. normal fibroblasts (NFs) (GSE172096) and NSC777201-treated vs. untreated control cells. D. Horizontal bar graph showing the log2fold changes in expressions of nine common overlapping DEGs.

Prognostic-related gene signatures of nine common overlapping DEGs

Expression levels of nine overlapping DEGs of both NSC777201-related DEGs and CAFs-DEGs are demonstrated in a heatmap in Figure 3A. Co-occurrence expressions of genes are often observed associated with functional inter-relatedness; therefore, we examined if and to what degree these nine overlapping DEGs were culpable in risk of death or recurrence in PAAD patients. We found that expression levels of these nine DEGs were equivocal as to the death risk, and higher expressions of these genes were strongly associated with a higher risk of death (Figure 3B). mRNA expression levels of these nine DEGs in the TCGA-PAAD (n=176) datasets by the risk group are shown in Figure 3B. In corroboration, a KM plot was generated for survival analysis of the co-expression of these nine DEG signatures in the TCGA-PAAD cohort (Figure 3C). Results showed that compared to the low-expression group, patients with higher expression levels of these nine DEGs exhibited worse mid-term to long-term (>5 years) OS ((hazard ratio (HR) =4.9), (95% confidence interval =2.14~11.21); P=0.0001685). Moreover, to evaluate the prognostic model, the ROC curve analysis using the R survival timeROC, and pROC were used to generate the plot. Showed that this risk score model could contribute to determining the relationships of these nine DEGs with OS. AUCs for the time-dependent ROC curve are shown in Figure 3D for the nine DEGs. Furthermore, distributions of expression levels of these nine overlapping DEGs in the normal, tumor, and metastatic TCGA-PAAD samples are presented in Figure 3E. Interestingly, CEMIP2 (TMEM2) was one of the DEGs significantly overexpressed in CAF vs. NF samples and TCGA-PAAD tumor and metastasis samples compared to normal samples, and it remains associated with the higher risk group in PAAD (Figure 3A, 3B, 3D, 3E), and markedly significant suppression of TMEM2 was observed after NSC777201 treatment (Figure 3A). Therefore, we selected TMEM2 for further analysis and study.

Figure 3.

Prognostic value of nine common overlapping differentially expressed genes (DEGs). A. Heat map of nine overlapping DEGs. B. The SurvExpress database (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp; accessed 6 August 2021) was applied to analyze associations of the nine-gene signature with the predictive risk, survival time, and prognosis, with gene expression levels in high- and low-risk groups. C. Kaplan-Meier survival curve showing that patients in the high-risk group had a poor prognosis, i.e., those with higher expression levels of the combined nine DEGs demonstrated lower survival times (hazard ratio =4.9 (95% confidence interval =2.14~11.21); P=0.0001685). An online protocol was used (http://www.progtools.net/gene/results.php; accessed 6 August 2021) to analyze the association of the nine DEG signatures with the predicted risk. D. An ROC analysis was performed to compare the sensitivity and specificity of the survival prediction, and P<0.05 was considered statistically significant. R-package, timeROC, and pROC were used to generate the plot. E. Expressions of the nine DEGs in pancreatic adenocarcinoma (PAAD), metastasis, tumor, and normal samples. TNM plot (https://tnmplot.com/; accessed 6 August 2021) was used to download the expression data. Data were plotted using GraphPad Prism 9. Prog. Idx., Prognosis Index. *P<0.05.

TMEM2 mrna expression in PAAD

TCGA-PAAD data were used to verify the findings of this study. mRNA expression levels of TMEM2, pathological stage, and protein expressions in PC and normal tissues were analyzed. TMEM2 expression levels in 37 different cancer types are demonstrated in Figure 4A. Higher levels of TMEM were detected in PC tissues than in normal tissues (Figure 4B). Among PAAD patients, the relative expression level of TMEM2-mRNA was significantly higher in stage 4 patients than in stage 1 and normal patients (P<0.05; Figure 4C), and TMEM2 was significantly expressed in metastatic tumor than in normal samples (P=0.00000377; Figure 4D). Furthermore, the prognostic value of TMEM2 was examined using SurvExpress and kmplot (https://kmplot.com/analysis/). As shown in Figure 4E, we observed that patients with higher TMEM2 expression were associated with poorer OS compared to those with low TMEM2 expression levels (hazard ratio (HR) =31.64, 95% CI =1.06-2.52; log-rank P<0.023), and patients with higher expression levels of TMEM2 were strongly associated with a higher risk of death Figure 4F (left panel). mRNA expression levels of TMEM2 in TCGA-PAAD datasets are shown by a risk group optimization in Figure 4F (right panel), which suggests that TMEM2’s ability to predict the progression and prognosis of PAAD patients is still largely elusive.

Figure 4.

Expression distribution of transmembrane protein 2 (TMEM2 or CEMIP2) in pancreatic adenocarcinoma (PAAD) patients. A. Expression distribution of TMEM2 in different tumors in TCGA database (https://gdac.broadinstitute.org/; accessed 7 August 2021). B. Median expression levels of TMEM2 in tumor and normal human pancreatic tissues in a body map (http://gepia.cancer-pku.cn/; accessed 7 August 2021). C. TMEM2 expression levels in PAAD patients based on individual cancer stages (http://ualcan.path.uab.edu/cgi-bin/ualcan-res.pl; accessed 9 August 2021). D. TMEM2 expression levels in metastasis, tumor, and normal samples (https://tnmplot.com/; accessed 9 August 2021). E, F. KM plot database (https://kmplot.com, accessed 19 April 2024) and the SurvExpress database were used for Kaplan-Meier survival analysis and to predict TMEM2 with the predictive risk, survival time, and prognosis using gene expression levels in high- and low-risk groups (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp; accessed 9 August 2021).

IHC analysis of TMEM2 expression in human

Furthermore, to investigate the expression and distribution of the TMEM2 protein in human PC samples, we retrieved and analyzed the IHC, immunofluorescent, and stained cell data from the HPA (http://www.proteinatlas.org/; accessed 15 November 2021) using the R packages hpar and HPAanalyze. The importance of these data is that they provide a detailed description of protein expressions in human cells and tissues, providing a basis for tissue-based diagnostic and translational research. From this analysis, we observed that TMEM2 at the protein level was highly or moderately expressed in liver cancer (35% high and 65% moderate), PC (25% high and 75% moderate), and thyroid cancer (25% high and 75% moderate) patients compared to other different cancers (Figure 5A). We also observed medium expression levels of the TMEM2 protein in normal pancreatic tissues (Figure 5B); from further analysis, we observed that TMEM2 was mostly expressed in vesicles and plasma membranes, suggesting its enzymatic role as a predominant mediator of contact-dependent hyaluronan (HA) and focal adhesion (FA). HA degradation and remodelling of the microenvironment favour tumor growth and invasion [52]. IHC samples from the HPA database corresponded to our previous observations, i.e., the protein level of TMEM2 was upregulated in PC tissue samples compared to normal tissues (Figure 5C), bar plot (also shown below) to demonstrate the number of patients showing negative, weak, moderate or strong expression, we observed cancer samples are showing strong (n=3) and moderate (n=8) expression of TMEM2 as compared to normal (moderate, n=3). Confirming the cancer showing higher expression of TMEM2 at the protein level.

Figure 5.

Detection of TMEM2 and cancer-associated fibroblast (CAF) marker expressions by IHC in human cancer tissues. A. Histogram of TMEM2 and CAF marker expressions in 20 different human cancers from the Protein Atlas (www.proteinatlas.org/, accessed 10 September 2021). IHC staining was evaluated as high/medium/low staining or not detected, and the histogram and tissue/cell subcellular expression intensities were visualized using the HPAanalyze R package. TMEM2 was highly expressed in pancreatic cancer compared to other cancer types. B. Expression levels of TMEM2 in tissues and cells, subcellular location, and expression proportions in pancreatic cancer. C. Representative IHC staining of TMEM2 in pancreatic adenocarcinoma (PAAD) and normal samples present in the HPA database, Bar charts depict the number of cases with different IHC staining intensities.

TMEM2 correlated with CAF expression in PAAD

In the second phase of this study, a thorough analysis was conducted to investigate the association of TMEM2 expression and the impact of CAF-associated genes in PAAD. A shortlist of important known CAF-associated and -related markers, viz., ACTA2, FAP, FN1, PDGFA, PDGFB, S100A4, SMAD2, TGFB1, and VIM, were analyzed, and their expression patterns at both the mRNA and protein levels were compared and correlated with TMEM2 expression in PAAD samples. As shown in Figure 6A, a heatmap (left panel), and combined correlation plot (right panel) denoted the strong correlations of TMEM2 with CAF markers, especially FAP (r=0.482; P=9.83E-12), FN1 (r=420; P=1.20E-08), and ACTA2 (r=0.420; P=2.60E-07). Furthermore, in Figure 6B, CAF-associated markers were observed to be strongly co-expressed in the high-risk PAAD patient group compared to the low-risk group, especially expressions of FAP, FN1, S100A4, and SMAD2. CAF genes were significantly overexpressed in the high-risk group of patients compared to the low-risk group. In Figure 6C, interestingly, the IHC analytical data of HPA demonstrated that co-expression correlations at the protein level also followed the trend, that is, at the protein level expression, FAP was strongly associated with TMEM2 expression in human PC compared to other CAF markers. The co-expression correlations of these markers were also evaluated in normal pancreatic samples, and results demonstrated that all the markers were not detected or detected at very low levels in normal tissues, except for SMAD2, PDGFA, PDGFB, and VIM. These results denoted that in a cancerous condition, the markers of S100A4, FN1, FAP, ACTA2, and TGFβ1, which were not detected in normal tissues, may play oncogenic roles, especially FAP expression might modulate the expression of TMEM2 to pave a favourable path for tumor growth and metastasis (Figure 6C; lower panel).

Figure 6.

Correlation and risk associated with expressions of TMEM2 and CAF markers and associated genes in TCGA-pancreatic adenocarcinoma (PAAD) datasets. A. Correlation plot of TMEM2 and CAF marker expressions in PAAD patients, and a heatmap showing correlation coefficients. B. A box plot confirming higher expressions of TMEM2 and CAF-related genes in the high-risk group than in the low-risk group using a t-test. C. Histogram of TMEM2 and CAF marker expressions in PAAD from the HPA; the histogram and tissue/cell subcellular expression intensities were visualized using the HPA analyze R package.

IPA and downstream biological function analysis of DEGs associated with the response of Panc1 cells to NSC777201

Next, we performed an IPA of DEGs in Panc1 cells in response to NSC777201. The IPA core analysis allowed us to interpret the dataset in the context of biological entities such as canonical pathways, upstream regulators, causal network master regulators, diseases, and biological functions (all of which passed the selection criteria of P<0.05 and z-score >2), whereas an orange node represents predicted activation (z-score >2) and a blue node represents predicated inhibition (z-score <-2) [48]. As described in the overall summary of the IPA, we observed activation of significant pathways, such as replicative senescence of fibroblast cell lines, replicative senescence of cells, and cytostasis (inhibition of cell growth and multiplication) as denoted by a higher-intensity orange color. Significant inhibition of cell movement of tumor cell lines, cell proliferation of carcinoma cell lines, and colony formation of tumor cells are denoted in blue with a higher intensity (Figure 7A) after NSC777201 treatment of Panc1 cells. Furthermore, based on our analysis, we identified several disease-associated or biological function-associated pathways (Figure 7B). Importantly, the cell-to-cell signaling interaction (P<0.05), cellular movement (P<0.05), cell growth and proliferation (P<0.05), and hematological system development (P<0.05) were the most suppressed (blue; z-score ≤2) pathways. Interestingly, a few pathways were significantly activated (orange; z-score ≥2), such as replicative senescence of fibroblast cell lines (P<0.05), replicative senescence of cells (P<0.05), inhibition of cell movement, and cellular growth and division (P<0.05) (lower panel, Figure 7B). These results suggest that NSC777201 treatment can inhibit cancer cell proliferation and induce the senescence of fibroblast cells in a predictive manner.

Figure 7.

Ingenuity Pathway Analysis (IPA) summary and downstream effector analysis of NSC777201-associated DEGs. A. Overall summary representing the regulatory effect of NSC777201 treatment on Panc1 pancreatic cancer cells. B. Visualization of a hierarchical heat map (TreeMap) depicting affected functional categories based on the DEGs where the major rectangular boxes represent the category of disease and functions. Each individual-coloured box is associated with a particular biological function or disease, and the colour indicates its predicted activation state of induced (orange) or reduced (blue). The size of the rectangles is correlated with the overlap significance. Negative Z-scores indicate the downregulation of a biological function, while positive Z-scores indicate the upregulation of a function. Absolute Z-score values of >2.0 were used to make biological predictions. Significant inhibition of cancer cell proliferation and activated replicative senescence was observed after NSC777201 treatment.

Molecular docking elucidates the binding mode of NSC777201 with TMEM2

As the TMEM2 protein is overexpressed in PAAD, we wanted to elucidate the mechanism of TMEM2 inhibition. Using computer-based drug target prediction software SwissTargetPrediction (http://www.swisstargetprediction.ch/) for NSC777201 as a query molecule, we identified the top predicted targetable proteins, among which were enzymes (26.7%), proteases (26.7%), kinases (13.3%), family A G protein-coupled receptors (13.3%), electrochemical transporters (13.3%), and hydrolases (6.7%) as shown in Figure S5. We, therefore, performed a computational docking analysis of the NSC777201 compound docked in the active site of TMEM2. For docking the TMEM2 target protein (Q9UHN6; left side, Figure 8A), its structure was downloaded from the uniport website, and the chemical 3D structure of the ligand (NSC777201; right side, Figure 8A) was prepared. Protein-ligand docking was performed with the entire protein structure using the automated CB-Dock server [49] and visualized by the Discovery studio visualizer vers. 21.1.0.20298, (BIOVIA, San Diego, CA, USA) [51]. Blind docking was performed to detect suitable binding sites for a ligand onto the protein by adjusting the cavity centre and docking box size. The result of docking demonstrated that NSC777201 stably docked in the binding cavity of TMEM2 (Figure 8B) and docking energies for each docking mode and docking box size were calculated (below, Figure 8B). The legend-receptor complex was stabilized through various hydrogen bonds and alkyl, van der Waal, and carbon-hydrogen bond interactions. Studies showed that NSC777201 interacted with TMEM2 with an utmost binding affinity score of -6.8 kcal/mol, demonstrating a high drug interaction, whereas, as shown in Figure 8C, NSC777201 occupied the active site of the target protein. Furthermore, the docked conformation of NSC777201 with its interacting residues is shown in Figure 8D. NSC777201 interacted with THR(A):246, ASN(A):246, and THR(A):1339 at respective distances of 2.95, 4.54, and 3.77 Å, via conventional hydrogen bonds. Although alkyl and pi-alkyl interactions occurred with LYS(A):278 and LEU(A):280, at respective distances of 4.80 and 5.48 Å with Cl atoms of NSC777201, LYS(A):278 formed another interaction via pi-alkyl bonds at a distance of 6.38 Å. Furthermore, pi-pi T-shaped interactions emerged at respective distances of 6.82 and 7.37 Å with PHE(A):254 and PHE(A):1338. Meanwhile, PHE(A):254 established pi-sulfur bonds at a distance of 6.84 Å of NSC777201, whereas ASP(A):273, LYS(A):1318, THR(A)276, VAL9A):386, LEU(A):1337, ASP(A):387, ARG(A):245, and GLU(A):474 interacted with NSC777201 through van der Waals forces, and carbon-hydrogen bond interactions at a distance of 3.64 Å with the PRO(A):253 residues. Furthermore, there was one unfavourable LYS(A):1336 interaction with NSC777201 at a distance of 7.24 Å. Collectively, the results of our computational study further reinforced that NSC777201 is a novel drug that is a key inhibitor of TMEM2 and can improve PC therapy.

Figure 8.

Visualization of molecular docking analysis of binding of NSC777201 with TMEM2. A. The structure of TMEM2 (CEMIP2) [Uniprot ID: Q9UHN6], and the chemical structure of NSC777201. B. The coupling of NSC777201 with TMEM2 active canter residues, and the binding affinity, cavity size and coordinates. C. The enlarged 3D image shows the binding of NSC777201 with different bonds with acceptor amino acid residues of TMEM2 to stabilize the binding complex in the cavity. D. The 2D plot shows the interaction of the binding pocket residues with the NSC777201.

In-vitro validation of the impact of siRNA and pharmacological (NSC777201) inhibition effect on TMEM2

In-vitro study illustrated in Figure 9, the effects of TMEM2 inhibition on various pivotal cellular processes associated with cancer progression were examined. In Figure 9A and 9B, expression both at protein and mRNA levels analyzed by western blot and real-time RT-PCR, we observed a noticeable dysregulation in the expression of TMEM2 at both protein and mRNA levels post-inhibition, lending confidence to the efficacy and specificity of inhibitory approach. In the cell viability assay depicted in panel Figure 9C, TMEM2 inhibition demonstrated a significant reduction in cellular viability, emphasizing TMEM2’s key role in orchestrating cellular survival dynamics. Further, as explained in Figure 9D, 9E, the inhibition of TMEM2 indicated an evident decline in the tumor sphere formation and cellular migratory activities, highlighting TMEM2’s essential role in driving tumor aggressiveness (stem cell behaviour) and invasiveness. The Figure 9F elucidates that the strategic inhibition of TMEM2 led to a consequential decline in the expression of pivotal CAF markers across the Panc1 and Suit2 cells. Lastly, to validate NSC777201 targeting TMEM2, the rescue experiment was performed as shown in Figure S6A, S6B, sh-TMEM2 transfection treated induced the suppression of sphere-forming abilities of Panc1 tumor sphere, while those can be significantly reversed when treated together with NSC777201-TMEM2-OE (over expression) when compared to control. Interestingly, at the protein level (Figure S6C), TMEM2 expression was rescued by TMEM2-OE, as compared to control and sh-TMEM2. The results showed the inhibitory effect of the sh-TMEM2 treated Panc1 tumor sphere ability, and the rescue effect of TMEM2-OE, abrogating the NSC777201 effect. This observation underscores TMEM2’s influential modulation of the CAF phenotype, thereby revealing new insights into our understanding of the intricate interplay between TMEM2 and the cancer microenvironment.

Figure 9.

In-vitro validation of the impact of TMEM2 inhibition. A, B. Western-blot and qRT-PCR analysis showing the siRNA-mediated inhibition of TMEM2 in Panc1 and Suit2 cells. C. Impact TMEM2 inhibition on the viability of both the cells measured using SRB cell viability assay. D, E. Self-renewal and migratory ability of both the cells were evaluated by tumor sphere and migration assay after TMEM2 knockdown. F. qRT-PCR analysis demonstrated inhibition of TMEM2 in Panc1 and Suit2 cells, respectively inhibiting key CAFs marker expression, and importantly FAP expression. ***P<0.001; **P<0.01; *P<0.05.

In-vivo preclinical validation

After establishing NSC777201’s anti-PAAD effect functions in-vitro, the in-vivo preclinical effects of NSC777201 targeting xenograft mouse Panc1 tumor model. The tumor size over time clearly showed that NSC777201 treatment alone or sh-TMEM2 resulted in significantly delayed tumorigenesis, while the vehicle control groups showed induced tumor growth. Notably, NSC777201 and sh-TMEM2 groups showed the most significant delays in tumorigenesis (Figure 10A, 10B). Using a Kaplan-Meier survival curve, we verified that NSC777201, sh-TMEM2 conferred a significant survival advantage in mice, compared to the vehicle-control groups (Figure 10C). The qPCR analysis of plasma levels of TMEM2 showed the reduced level in NSC777201 and sh-TMEM2 treated pooled blood samples, in comparison to vehicle control (Figure 10D). Comparative western blots and qRT-PCR analysis of tumor samples collected in all groups demonstrated reduced TMEM2 and FAP expression (Figure 10E, 10F). Taken together, these data suggest that the downregulation of TMEM2 expression in PAAD carcinoma cells effectively inhibits the formation and growth of PAAD cancer in-vivo.

Figure 10.

Efficacy evaluation of NSC777201 using a Panc1 cells xenograft mouse model. (A) The insert shows representative photographs of tumor samples from each group, (B) tumor size over a time curve. The tumor growth delay was most significant in the NSC7777201 treatment group, followed by the sh-TMEM2-only group, while vehicle control groups showed a significant difference in tumor size. (C) Kaplan-Meier survival curve, mice receiving NSC777201, or sh-TMEM2 only showed the highest survival ratios, while control mice showed the lowest survival ratios. (D) qPCR analyses of plasma levels of TMEM2. Pooled blood samples from all four groups of mice were analyzed for TMEM2 plasma levels. The NSC777201, or sh-TMEM2 only showed the lowest level followed by control. (E) Tumor sample western blot analysis. Expressions of TMEM2 and FAP were lower in samples from NSC777201, or sh-TMEM2 treated tumors. (F) qPCR analyses of tumor samples from all three groups of mice were analyzed The NSC777201, or sh-TMEM2 only showed the lowest level of TMEM2 and FAP expression. ***P<0.001; **P<0.01; *P<0.05.

Discussion

Recently, the application of novel chemotherapeutic drugs and surgical interventions has partially improved PAAD treatment. However, most patients with PAAD are diagnosed at an advanced stage with poor survival and therapeutic resistance resulting in higher PAAD patient mortality [53,54]. As PAAD is one of the most serious solid tumors, the increased stromal content is one of the PAAD hallmarks features [55]. In the TME, the most dominant components in the tumor stroma are CAFs, which are spindle-shaped cells that build up and remodel the ECM [56]. However, the “CAF population” remains poorly understood in terms of its origin, subtypes, and biology due to higher heterogeneity and lack of specific markers in PAAD [57,58]. Another important member of the ECM is hyaluronan (HA); its abnormal metabolism and accumulation particularly the small HA oligosaccharide (LMM-HA), as an independent prognostic factor for poor survival in PC [59], are key hallmarks of cancer [60,61]. HA is known to be metabolized by HA-enzyme-hyaluronidase enzymes such as HYAL1, -2, and -3, and KIAA1199 is also known as CEMIP [62-64]. Activation of these enzymes reportedly showed aberrant expressions in many cancers [29]. In particular, HYAL1 and KIAA1199 are significantly overexpressed in PC, and inhibition of these enzymes results in the reduced migratory ability of PC cells [30,31]. Nevertheless, how CAFs modulate their expressions and are associated with HA-enzyme activation is still not studied well.

Therefore, in this present exploration, we conducted a comprehensive in silico approach intended to identify a key gene, as well as the potential of the novel NSC777201’s effect on inhibiting PAAD. We screened DEGs from the GEO database (GSE172096), i.e., from CAF compared to NF samples (Figure 1) and performed functional enrichment and pathway analyses of these identified DEGs (Figure S2). Furthermore, the function and pathway enrichment analyses found that the most significant pathway was “pathways in cancer”, whereas those of functional categories were “DNA replication” and “DNA integrity checkpoint”. In addition, to investigate the effect of the novel NSC777201’s effect on PC Panc1 cells, microarray gene expression profiling was performed. NSC777201 treatment modulated expressions of many key DEGs, and after overlapping these NSC777201-associated DEGs with CAF-associated DEGs, we observed nine genes in common (Figure 2). These nine overlapping NSC777201-associated DEGs were further screened by a Cox multiple regression analysis of PAAD RNA-sequencing data from TCGA. The KM survival analysis indicated that patients at high risk corresponded with shorter OS times than patients with low-risk scores (P<0.0001). The AUC of this model was an average of 0.6 at 12 months OS, indicating that the predictive value of the nine-gene signature could be utilized for survival predictions. Compared to other specific medical parameters (including age, sex, tumor stage, and histological type), risk scores were better predictors of patient survival, indicating that the nine-gene signature may be of value in further research (Figure 3).

Interestingly, among the nine overlapping DEGs, TMEM2 or CEMIP2 expression was negatively correlated with OS times of PAAD patients. TMEM2 is a cell-surface hyaluronidase and a potent modulator of matrix-associated HA, and TMEM2 activity is necessary for cells to achieve robust cell adhesion and migration on HA-containing substrates [52]. TMEM2 has been implicated in the aggressive behavior of many solid cancers, with the studies linking it to poor outcomes in many cancers [65,66]. Importantly, Lee et al. utilized the Gene Expression-based outcome for breast cancer online tool to find that TMEM2 expression correlated with worse prognosis in grade 3 breast tumors, especially within the luminal B and HER2-positive categories via SOX4 regulation [65]. Study reported by L. Gan et al. demonstrated the inhibition of TMEM2 results in the reduced invasion and migration of breast cancer through the modulation of JAK/STAT3 signaling [67]. Similarly, in gliomas, increased TMEM2 expression not only escalates with tumor grade but also serves as a distinct prognostic marker for refining molecular subtypes and predicting outcomes more accurately [66]. Despite these associations, the precise mechanisms through which TMEM2 influences cancer progression warrant further investigation.

Our study contributes to this body of knowledge by demonstrating a significant upregulation of TMEM2 in PAAD tumors compared to normal tissue, a finding consistent across a diverse array of 37 cancer types. Notably, TMEM2 levels were markedly higher in advanced stages of PAAD, correlating with decreased patient survival (Figures 4, 5). These observations position TMEM2 as a critical biomarker for diagnosis, and prognosis, and as a potential target for therapeutic intervention in PAAD, signifying its importance in future research endeavours. Furthermore, our examination extends into the intricate environment of PC, which is characterized by a mix of transformed cancer cells at varying stages of the epithelial-mesenchymal transition (EMT) and an assortment of non-transformed stromal cells. This stromal compartment, comprising cancer-associated fibroblasts (CAFs), macrophages, and a variety of immune, endothelial, and epithelial cells, plays a significant role in the tumor microenvironment (TME) [68,69].

CAFs are one of the prominent and active components of the pancreatic TME, and classical CAF markers are characterized by induced expression of alpha-smooth muscle actin (α-SMA), FAP, fibroblast-specific protein 1 (FSP1) or S100A4, platelet-derived growth factor receptor (PDGFR)-α/β, and vimentin (VIM), all recognized as contributors to carcinogenesis [57,70]. The activation of canonical SMAD2 signaling, culminating in the elevation of TGFβ1 and FN1, further denotes the abundance of CAFs, highlighting their significance in the PC landscape [71,72]. This detailed understanding of the PC microenvironment, along with the pivotal role of TMEM2, underlines the complexity of cancer progression and the potential avenues for therapeutic intervention. Our results demonstrated that TMEM2 expression at both the mRNA and protein levels was strongly correlated with expressions of FAP (r=0.480), FN1 (r=0.420), and S100A4 (r=0.380) in PAAD, and higher expressions of FAP (P=5.38 × 10^-7), FN1 (P=2.46 × 10^-6), and S100A4 (P=8.85 × 10^-9) were significantly correlated with the high-risk group in PAAD (Figure 6). CAF roles are in forming ‘cancerized’ fibrotic stroma favourable to tumor initiation, progression, stemness, dissemination, metastasis, and drug resistance via remodelling of the ECM through activation of hyaluronidase [73]. The activation of TMEM2 was suggested to be indirectly or directly associated with the expression of CAFs mainly with classical CAFs associated with the FAP gene, which plays a pivotal role in tumor initiation through the metabolism of HA. Hence, the identification of novel small molecules with the potential to inhibit the HA enzyme and modulation of CAF expression is crucial. Our team previously reported the importance of NSC777201 in inhibiting lung cancer tumorigenesis [38]. However, the effect of NSC777201 on ECM modulation still needs exploration, especially in PAAD tumorigenesis. Herein, we performed an IPA analysis to investigate the effect of NSC777201 on gene expressions and associated pathway modulation in PAAD, and results demonstrated significant activation of replicative senescence of fibroblast cell lines, cell death, apoptosis, (z-score >2), and inhibition of cellular development, cell proliferation, and the EMT (z-score <-2). The in-silico study results suggested that NSC777201 is a novel molecule that exhibits potential to target PAAD through modulating the CAF-HA-enzyme axis (Figure 7).

Interestingly, to further demonstrate the predictive target of NSC777201, we applied the SwissTargetPrediction online computational tool (http://www.swisstargetprediction.ch/), and the top predicted and significant targetable proteins identified were enzymes (26.7%) and proteases (26.7%). Therefore, an in silico molecular docking analysis of NSC777201 was also performed on TMEM2, and results demonstrated in Figure 8, that NSC777201 showed a higher binding affinity with TMEM2 with the lowest energy conformation (-6.8 kcal/mol), the ligand-and-receptor complex was stabilized through various hydrogen bonds and alkyl, van der Waal, and carbon-hydrogen bond interactions, indicating that NSC777201 can be used. Our comprehensive studies demonstrate the profound impact of TMEM2 inhibition on cancer progression through in-vitro and in-vivo analyses (Figures 9 and 10). TMEM2 dysregulation at both protein and mRNA levels confirmed the efficacy of our inhibitory strategy, with subsequent assays indicating a crucial role for TMEM2 in maintaining cellular survival and promoting tumor aggressiveness (Figure 9A-E). In-vivo, treatments with NSC777201 and sh-TMEM2 in a xenograft mouse model substantially delayed tumorigenesis and enhanced survival, with reduced TMEM2 and FAP expression corroborating the treatment’s effectiveness. Rescue experiments further validated these effects, showing a reversal of TMEM2 suppression through combined NSC777201 and TMEM2 overexpression treatments, which moderated NSC777201’s effects (Figure S6). These results highlight TMEM2’s significant influence on the cancer microenvironment and its potential as a target for therapeutic interventions, offering new insights into cancer biology and treatment strategies.

Conclusion

In conclusion, our findings, as illustrated in Figure 11 (overall study design), highlight the pivotal role of TMEM2 as a significant prognostic and tumorigenic biomarker in the progression of pancreatic adenocarcinoma (PAAD) through its interaction with cancer-associated fibroblasts (CAFs) and modulation of fibroblast activation protein (FAP). These interactions underscore TMEM2’s potential as a novel therapeutic target, which could lead to more effective treatments for PAAD. Additionally, NSC777201 emerges as a promising novel small molecule targeting the HA-enzyme, exhibiting substantial anti-PAAD effects. Given its demonstrated efficacy, NSC777201 can serve as an important small molecule drug worthy of therapeutic implications and warrants further investigation to be therapeutically used against PAAD patients.

Figure 11.

Overall study flow (left to right). Common DEGs identification between public and in-house sequencing data. Key gene (TMEM2) identification and its prognostic values, together with its expression analysis in PAAD from different (both RNA and protein) databases, correlation analysis with CAFs markers, IPA analysis of DEGs identified after the NSC77201 drug treatment in Panc1 cells, to predict key pathways inhibited or activated after treatment. Molecular docking to show predictive binding of drug with TMEM2, in-vitro and in-vivo validation. The current study was designed to explore the importance of TMEM2 a HA enzyme, in the PAAD TME, together with the role of CAFs, interestingly the newly discovered molecule NSC777201 can also modulate the expression of TMEM2, which results in the alteration of PAAD-TME and its progression.

Disclosure of conflict of interest

None.