Abstract

   Epithelium‐specific ETS transcription factor 1 (ESE1) has been
   implicated in epithelial homeostasis, inflammation, as well as
   tumorigenesis, and cancer progression. However, numerous studies have
   reported contradictory roles—as an oncogene or a tumor suppressor of
   ESE1 in different cancers, and its function in the development and
   progression of pancreatic ductal adenocarcinoma (PDAC) has remained
   largely unexplored. Herein, we report that ESE1 was found upregulated
   in primary PDAC compared to normal pancreatic tissue, but high
   expression of ESE1 correlated to better relapse‐free survival in
   patients with PDAC. Interestingly, ESE1 was found to exhibit dual roles
   in regulation of malignant properties of PDAC cells in that its
   overexpression promoted cell proliferation, whereas its downregulation
   enhanced epithelial‐mesenchymal transition (EMT) phenotype. In the
   context of TGF‐β‐induced EMT, ESE1 is markedly downregulated at
   post‐transcriptional level, and reconstituted ESE1 expression partially
   reversed TGF‐β‐induced EMT marker expression. Furthermore, we identify
   AGR2 as a novel transcriptional target of ESE1 that participates in
   TGF‐β‐induced EMT in PDAC. Collectively, our findings reveal an
   ESE1/AGR2 axis that interacts with TGF‐β signaling to modulate EMT
   phenotype in PDAC.

   Keywords: AGR2, epithelial‐mesenchymal transition, ESE1, pancreatic
   cancer, transcription factor

1. INTRODUCTION

   Pancreatic ductal adenocarcinoma (PDAC), the fourth leading cause of
   cancer‐associated deaths in the United States, has a 5‐year survival
   rate of below 10%.[48] ^1 , [49]^2 Major factors that contribute to the
   poor prognosis of PDAC include late diagnosis, poor response to
   chemoradiotherapy, as well as epithelial–mesenchymal transition
   (EMT)‐related early invasion, and metastasis.[50] ^3 Therefore, a
   better understanding of the regulatory mechanisms underlying the EMT
   process will facilitate the development of novel therapeutic strategies
   against PDAC.

   Epithelium‐specific ETS transcription factor 1 (ESE1), also known as
   ELF3, Ert, Esx, and Jen,[51] ^4 , [52]^5 , [53]^6 , [54]^7 belongs to
   the ETS‐domain transcription factor superfamily.[55] ^8 As the name
   suggests, ESE1 is primarily expressed in epithelial‐rich tissues and
   regulates epithelial cell proliferation and differentiation, such as
   that of intestinal, lung, bladder, breast, and squamous epithelia.[56]
   ^6 , [57]^9 , [58]^10 , [59]^11 Although ESE1 deregulation is
   frequently associated with cancer development and progression, its role
   in the pathogenesis of epithelial cancer is highly variable with
   regards to its impact on cell proliferation and EMT. For instance, ESE1
   is reported to negatively regulate EMT in bladder and ovarian cancer
   cells,[60] ^12 , [61]^13 whereas in non‐small cell lung cancer
   (NSCLC)[62] ^14 and hepatocellular cancer (HCC),[63] ^15 it promotes
   cell proliferation and EMT. Similar conflicting results have been
   reported in breast cancer cells. For example, ectopic expression of
   ESE1 decreases estrogen (E2)‐dependent MCF7 cell proliferation by
   inhibiting the transcriptional activity of estrogen receptor α
   (Erα),[64] ^16 and plays a tumor suppressor role. In human epidermal
   growth factor receptor 2 (HER2)^+ breast cancer cells, knockdown of
   ESE1 inhibits cell proliferation and tumor sphere formation by
   inhibiting HER2‐dependent signaling, therefore displaying a
   pro‐tumorigenic role.[65] ^17 , [66]^18 , [67]^19 These results imply
   that the effects of ESE1 is dependent on the cellular context or cancer
   type. Moreover, although the function of ESE1 in several types of
   epithelial cancers has been studied, its role and mechanism in the
   development and progression of PDAC have remained largely unexplored.

   Anterior gradient protein 2 (AGR2) is a member of the protein disulfide
   isomerase (PDI) family predominantly localized to the endoplasmic
   reticulum (ER). AGR2 expression is also mainly restricted to epithelial
   cells, and is associated with cancer cell proliferation, survival and
   tumor growth, as well as invasion and metastasis.[68] ^20 , [69]^21 ,
   [70]^22 , [71]^23 AGR2 contributes to maintain epithelial cell
   phenotype, inhibits EMT induction, and enhances the rate of adhesion to
   plastic substratum, thus playing an undeniable role in tumor
   development and progression.[72] ^24 , [73]^25 However, the
   transcriptional regulation of AGR2 in the setting of cancer cell
   proliferation, migration, and invasion has not yet been fully
   elucidated.

   Here, we investigated whether ESE1 contributes to TGF‐β‐induced EMT in
   PDAC cells. In addition, we identified potential targets of ESE1
   mediating such activity. Our results indicated that ESE1 plays an
   important role in inhibiting TGF‐β‐induced EMT in PDAC cells via the
   transcriptional activation of AGR2.

2. MATERIALS AND METHODS

2.1. Cell culture and selection of stable cell clones

   HEK293T, HeLa, and PDAC lines L3.6, PaTu 8988 t, PANC‐1, MIA PaCa‐2,
   CFPAC‐1, and HPNE, an immortalized human pancreatic epithelial cell
   line, were maintained and transfected as previously described.[74] ^26
   For stable knockdown of the ESE1 mRNA, lentiviral particles generated
   from pLKO‐based lentiviral vector carrying shRNAs targeting human ESE1
   or scramble control were used for cell transduction. The cells were
   selected in puromycin‐containing media as previously described.[75] ^26
   The shRNA sequences can be found in Table [76]S1. For reconstituted
   overexpression of the ESE1 gene, shESE1 L3.6 cells were transfected
   with lentiviral particles carrying the pLenti6.3‐Flag‐ESE1 cDNA. The
   infected cells were selected with blasticidin‐containing media
   (Sigma‐Aldrich, USA). Pooled antibiotic‐resistant cell clones were used
   for indicated experiments.

2.2. Quantitative RT‐PCR (qPCR)

   Total RNA was extracted using TRIzol reagent (Invitrogen, USA) and
   reverse transcribed into cDNA using PrimeScript™ RT reagent Kit
   (TaKaRa, Japan). The cDNA products were used as templates for qPCR
   using iTaq™ Universal SYBR® Green Supermix (Bio‐rad, USA) on a Roche
   Light Cycler 96 System Real‐Time System. All the genes were normalized
   by GAPDH. The primer sequences are listed in the Table S1.

2.3. Western blotting

   The cells were lysed with lysis buffer (Beyotime, China) freshly
   supplemented with phosphatase inhibitors and phenylmethanesulfonyl
   fluoride (Beyotime, China). The samples were fractionated by SDS‐PAGE
   gels and transferred to microporous polyvinylidene difluoride (PVDF)
   membranes (Roche, Swiss). The membranes were blocked with 5% skim milk
   for 1 h and incubated overnight at 4°C with primary antibodies
   including rabbit polyclonal anti‐AGR2[77] ^27 ; ESE1 (Santa Cruz
   Biotechnology, USA); E‐cadherin (CST#3195, USA), N‐cadherin (CST#13116,
   USA), and vimentin (CST#5741, USA); GAPDH (10494‐1‐AP, Proteintech,
   USA). The membranes were then washed and probed with secondary
   antibodies: goat anti‐mouse IgG (H + L)‐HRP or goat anti‐rabbit IgG
   (H + L)‐HRP (Bioworld, USA) before detection with the enhanced
   chemiluminescence (Thermo Fisher Scientific, USA) and imaged using the
   ChemiDoc XRS + System (Bio‐Rad, USA).

2.4. Cell proliferation and migration assays

   Cell proliferation assays were performed as directed by CCK‐8 kit
   (Dojindo, Japan). Briefly, single cell suspensions were seeded in
   96‐well plates at a density of 5000 cells/well and cultured for 0, 24,
   48, and 72 h, respectively, before adding 10% of CCK‐8 solution and
   incubated for another 2 h. The absorbance values were measured at
   450 nm by microplate reader (BioTek, USA). Scratch healing assays were
   performed as previously described.[78] ^28 Cells were cultured
   overnight to reach 90% confluency in a 6‐well plate before scratching
   with a 20‐μl plastic pipette tip. The cells that migrated into the
   wounded areas were imaged with the Microscope (Leica, Germany). Cell
   migration rate was measured as follows: (original wound width ‐ new
   wound width) / original wound width× 100%. For TGF‐β (MCE, USA)
   treatment, serum‐free DMEM supplemented with TGF‐β (15 ng/ml) was added
   to replace the serum‐free DMEM without TGF‐β. Tracked cell migration
   were performed using Operetta CLS system (PerkinElmer, USA). After
   overnight culture, cells were administered with TGF‐β (15 ng/ml) for
   3 days and were tracked for 10 h with the Operetta CLS system. Cell
   distribution, migration speed, and tracks were analyzed with Harmony
   software.

2.5. Luciferase reporter assays

   AGR2 promoter DNA sequence (−1506 to −99 relative to the translation
   start site) was cloned into pGL3‐basic vector (pGL3‐AGR2). The cDNA
   sequences encoding ESE3, ESE1, and its truncation mutants were cloned
   into pM vector (Clontech, USA) in frame with N‐terminal GAL4‐DBD
   (GAL4‐ESE1 or GAL4‐ESE3). The pRL‐TK renilla vector and pM‐GAL4
   reporter plasmid have previously been described.[79] ^29 For AGR2
   promoter luciferase reporter assay, pGL3‐AGR2 reporter was
   co‐transfected with ESE1 or SPDEF cDNA expression plasmid with the
   pRL‐TK vector as internal control for transfection efficiency; to
   determine the transcriptional regulatory activity of ESE1 and ESE3,
   GAL4‐ESE1(and its truncation mutants as indicated), GAL4‐ESE3 plasmids
   were co‐transfected with pGL3 firefly reporter plasmid carrying five
   tandem GAL4 DNA binding sites (GAL4‐luc) as previously described.[80]
   ^29 Transfections were carried out following standard procedures, and
   the dual‐luciferase assays were performed with LucPair™ Duo‐Luciferase
   Assay Kit (GeneCopoeia, USA). To determine the effect of TGF‐β on ESE1
   transcriptional activity, cells were treated with 15 ng/mL of TGF‐β
   (MCE, USA) for 3 days before harvesting for the above assays.

2.6. Immunofluorescent staining and imaging

   Immunofluorescence (IF) imaging was used to determine the subcellular
   localization of endogenous ESE1 as recently described.[81] ^28 To
   determine the localization of overexpressed ESE1, ESE1‐EGFP, or
   ESE1‐Flag expression plasmids were transfected to HeLa cells using
   Lipo2000 regents. Cells were fixed and stained at 24 h after
   transfection. Images were captured using the Leica 2500 Microscope
   System (Leica, Germany).

2.7. Gene expression profiling and survival analysis

   We downloaded the gene expression profiles from [82]GSE16515 and
   GSE16950 databases. The expression of ESE1 was visualized using the
   GraphPad Prism 7 and the GEPIA database. The relapse‐free survival
   (RFS) and overall survival (OS) analyses for ESE1 or AGR2 were
   performed through the KM‐plotter.[83] ^30

2.8. Functional annotation and transcription factor interactome analysis

   Functional annotation analysis was performed using the OmicStudio tools
   available at [84]https://www.omicstudio.cn/tool.

2.9. Gene set enrichment analysis

   The GSEA enrichment of KEGG pathways was performed at
   [85]https://www.omicstudio.cn/tool. The following list of genes were
   used for GSEA: differentially expressed genes (DEGs) in L3.6‐shCtrl
   cells versus L3.6‐shESE1.

2.10. Statistical analysis

   Statistical analysis was performed using the GraphPad Prism software
   (version 7; GraphPad Software, Inc.). Multiple comparisons were
   performed between the groups using one‐way analysis of variance
   (ANOVA), followed by multiple comparison tests using the GraphPad Prism
   software. Data are expressed as mean ± standard deviation. p‐value
   <0.05 indicated a significant difference.

3. RESULTS

3.1. ESE1 is overexpressed in human PDAC

   We found ESE1 highly overexpressed in various human tumors including
   colorectal adenocarcinoma and PDAC compared to maching normal tissues
   based on GEPIA (Figure [86]S1A). We further analyzed two Gene
   Expression Omnibus (GEO) datasets ([87]GSE16515 and GSE16950)[88] ^31 ,
   [89]^32 and found significantly higher ESE1 expression in PDAC or
   precancerous lesions including IPMA, IPMC, and IPMN than in matching
   normal tissues (Figure [90]1A). In addition, consistent results were
   obtained in the TCGA database with GEPIA (Figure [91]1B). The
   immunohistochemistry (IHC) data from the Protein Atlas database
   ([92]https://www.proteinatlas.org/ENSG00000163435) showed a higher
   number of ESE1‐positive cells in PDAC than in normal pancreatic tissues
   (Figure [93]1C). Interestingly, ESE1 expression decreased with tumor
   stage, and low ESE1 expression was correlated with shorter RFS, but
   better OS (Figure [94]1D,E, Figure [95]S1B). These results reflected a
   dichotomy of ESE1 association with pancreatic cancer progression and
   prognosis. The relative expression of ESE1 was further determined in a
   panel of PDAC cell lines with L3.6 showing the highest levels of both
   mRNA and protein expression (Figure [96]1F,G).

FIGURE 1.

   FIGURE 1
   [97]Open in a new tab

   ESE1 is overexpressed in human PDAC. (A) The expression level of ESE1
   in the GEO datasets. (B) The expression level of ESE1 in primary PDAC
   tissues and matched normal tissue based on TCGA dataset. (C) IHC of
   ESE1 based on the data from Protein Atlas database. (D) Statistical
   analysis of ESE1 expression with tumor progression (clinical stages).
   (E) The correlation between ESE1 expression with RFS in stage‐2 PDAC
   patients based on the data obtained and analyzed from Kaplan–Meier
   Plotter database. ESE1 expression in four PDAC cell lines by western
   blotting (F) and qPCR (G). HR, Hazard Ratio; IPMA, intraductal
   papillary mucinous adenoma; IPMC, intraductal papillary‐mucinous
   carcinoma; IPMN, intraductal papillary mucinous neoplasm; PDAC,
   pancreatic ductal adenocarcinoma.

3.2. Differential impact of ESE1 on PDAC cell proliferation and EMT phenotype

   To explore the function of ESE1 expression in PDAC cells, we selected
   L3.6 with high ESE1 expression for subsequent experiments. ESE1 stable
   knockdown cell lines were obtained by transducing cells with pLKO‐based
   lentiviral shRNA targeting the human ESE1 mRNA (with a non‐targeting
   shRNA as the control) (Figure [98]2A,B). Cell proliferation ability of
   L3.6 cells was significantly reduced by shRNA‐mediated suppression of
   endogenous ESE1 compared with scramble control, whereas reconstituted
   expression of ESE1 mostly restored cell proliferation potential
   compared with the shESE1 group in L3.6 cells (Figure [99]2C). Similar
   results were obtained with colony formation assays (Figure [100]2D,E).
   Contrary to its inhibition of cell proliferation and colony formation,
   the shESE1 stable cells exhibited higher migration capabilities
   (Figure [101]2F,G), whereas reconstituted overexpression of ESE1
   largely reversed this enhanced migration ability compared with the ESE1
   knockdown control. Therefore, ESE1 contributed to promote the
   proliferation, but inhibited the cell migration ability of L3.6 cells.
   Next, we investigated the potential influence of ESE1 on EMT process
   and found ESE1 knockdown markedly increased the expression level of the
   mesenchymal marker vimentin and decreased that of E‐cadherin, whereas
   the changes in the expression levels of both markers were largely
   reversed upon reconstituted expression of ESE1 (Figure [102]2H,I).
   Considering that ESE1 showed high expression in low‐grade and
   low‐to‐undetectable expression in high‐grade PDACs,[103] ^33 we
   selected another well differentiation pancreatic cell line CFPAC‐1 to
   explore the function of ESE1 expression.[104] ^34 Similar to the
   findings in L3.6 cells, ESE1 suppression also inhibited CFPAC‐1
   proliferation (Figure [105]S2A‐C), while promoting its EMT phenotype
   (Figure [106]S2D). These results suggest that ESE1 is a critical
   component of the EMT machinery in these PDAC cells, leading to the
   hypothesis that downregulation of ESE1 promoted a EMT phenotype.

FIGURE 2.

   FIGURE 2
   [107]Open in a new tab

   Differential impact of ESE1 on PDAC cell proliferation and EMT
   phenotype. (A, B) Validation of ESE1 knockdown in L3.6 cells by Western
   blotting and qRT‐PCR. (C) The effect of ESE1 on L3.6 cell proliferation
   was determined with CCK‐8 assay. (D) Representative images of colony
   formation assay in L3.6 cells transfected with scrambled control
   (shCtrl), ESE1 specific shRNA (shESE1), and reconstituted ESE1
   expression in ESE1 knockdown cells (shESE1‐OE). (E) Statistical
   analysis of colony formation assay result. (F) Representative images of
   wound‐healing assay in L3.6 cells from shCtrl, shESE1, and shESE1‐OE
   groups, respectively. (G) Statistical analysis of the wound‐healing
   assay results shown in (F). (H) Immunoblot detection of EMT marker
   expression in L3.6‐shESE1 and shESE1‐OE cells. (I) Quantitative
   analysis of the target protein bands using ImageJ software. Data are
   presented as means ± SEM (n = 3 per group). *p < 0.05, ***p < 0.001,
   ****p < 0.0001. Scale bars: 500 μm (F).

3.3. Knockdown of ESE1 expression affects the TGF‐β network in PDAC cells

   To identify the potential transcriptional targets of ESE1, we next
   performed RNA sequencing analysis in L3.6 cells with ESE1 knockdown
   versus non‐targeting control (GSE 206999). A total of 546 downregulated
   genes and 263 upregulated genes were found to be associated with the
   suppression of ESE1 (p < 0.01; log2 fold change >1; Figure [108]3A, GSE
   206999). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway
   enrichment analysis was applied for pathway annotations and showed that
   DEGs were enriched in cancer signaling pathways, including pathways
   related to bladder cancer, prostate cancer, gastric cancer, and small
   cell lung cancer (Figure [109]3B). ESE1‐induced DEGs were further
   analyzed by functional annotations in Gene Ontology (GO) term, and the
   results revealed that genes associated with carcinogenic pathways are
   enriched, including inflammatory response, extracellular matrix,
   positive regulation of cell migration, wound healing (Figure [110]3C).
   Because of the intricate association between ESE1 and cell migration,
   and the enrichment of wound healing and DEGs in the cancer pathways, we
   hypothesized that ESE1 could be a target of EMT‐related pathways in
   PDAC cells, such as TGF‐β. The KEGG pathway correlated with the DEGs
   was then studied using gene‐set enrichment analysis (GSEA) to further
   investigate the biological functions of DEGs
   ([111]https://www.omicstudio.cn/tool). Indeed, the DEGs were found to
   be enriched in the TGF‐β signaling pathway and also related to
   pancreatic cancer (Figure [112]3D,E). The GSEA of the microarray data
   and the fact that TGF‐β signaling is a predominant promoter of EMT and
   PDAC progression[113] ^35 , [114]^36 prompted us to examine the role of
   ESE1 in TGF‐β‐induced EMT process in pancreatic cancer.

FIGURE 3.

   FIGURE 3
   [115]Open in a new tab

   Knockdown of ESE1 expression affects the TGF‐β network in PDAC cells.
   RNA‐seq analysis was performed in L3.6‐shCtrl versus L3.6‐shESE1 cells
   to determine the DEGs. (A) Volcano map of DEGs. The blue dots
   represented down‐regulated genes while red dots represented
   up‐regulated genes with the number of down‐ and up‐regulated genes
   shown blow. (B) Bubble map of DEGs in L3.6‐shCtrl versus L3.6‐shESE1
   cells based on KEGG pathway analyses. (C) Bubble map of GO pathway
   analyses on DEGs in L3.6‐shCtrl versus L3.6‐shESE1 cells. GSEA analysis
   based on DEGs in L3.6‐shCtrl versus L3.6‐shESE1 cells to compare the
   changes related to TGF‐β signaling pathway (D) and pancreatic cancer
   (E) using the KEGG gene sets.

3.4. TGF‐β‐induced EMT is associated with ESE1 post‐transcriptional
downregulation

   EMT is a critical biological process that contributes to the tumor cell
   migration and invasion.[116] ^37 We investigated whether ESE1 regulated
   the EMT process in L3.6 cells. We first analyzed the expression levels
   of EMT markers, including E‐cadherin, N‐cadherin, and vimentin. The
   results showed that TGF‐β increased the expression levels of the
   mesenchymal markers N‐cadherin and vimentin, while decreasing the
   expression level of the epithelial marker E‐cadherin. Furthermore, the
   protein expression of ESE1 was significantly reduced after the
   treatment with TGF‐β (15 ng/mL) on the first day accompanied by induced
   N‐cadherin expression, whereas the expression of vimentin started
   increasing on the second day and becoming more apparent on the third
   day (Figure [117]4A). Similarly, cells treated with different
   concentrations of TGF‐β showed reduced expression of ESE1 before
   increased expression of vimentin (Figure [118]4B). These findings
   suggested that TGF‐β contributed to downregulate ESE1 protein
   expression in L3.6 cells. However, qRT‐PCR analysis revealed no
   significant difference at the mRNA levels of ESE1 in L3.6 cells
   following TGF‐β treatment suggesting that TGF‐β did not affect the
   transcription of ESE1 (Figure [119]4C). The results also showed
   increased expression of the mesenchymal marker vimentin and decreased
   expression of epithelial marker E‐cadherin in CFPAC‐1 cells with TGF‐β
   treatment. Furthermore, the protein expression of ESE1 was
   significantly reduced after the treatment with TGF‐β, which was
   consistent with above findings on TGF‐β‐treated L3.6 cells
   (Figure [120]S2E).

FIGURE 4.

   FIGURE 4
   [121]Open in a new tab

   TGF‐β‐induced EMT is associated with ESE1 post‐translational
   downregulation. Immunoblot analysis to determine ESE1 and EMT‐related
   marker expression in L3.6 cells treated with TGF‐β (15 ng/mL) for
   indicated time (A) and doses (B). (C) The mRNA expression of ESE1 and
   EMT markers was determined by qRT‐PCR analysis in L3.6 cells treated
   with and without TGF‐β for 3 days. (D) Representative images of
   wound‐healing assay from shCtrl, shESE1 and shESE1‐OE groups with and
   without TGF‐β treatment. (E) Quantitative analysis of the
   scratch‐healing assay result by measuring the cell migration rate. (F)
   Cell track analysis in shCtrl, shESE1, and shESE1‐OE cells without (a,
   c, and e) and with TGF‐β treatment (b, d, and f) by Operetta CLS. (G)
   Quantitative analysis of the cell displacement with Harmony software.
   (H) Immunoblot analysis of ESE1 and EMT marker protein expression in
   L3.6 cells with and without TGF‐β treatment. Data are presented as
   means ± SEM (n = 3 per group). *p < 0.05, ****p < 0.0001. Scale
   bars:500 μm (D).

   To confirm that ESE1 is a key component of TGF‐β‐induced EMT, we
   investigated whether altered ESE1 expression could influence
   TGF‐β‐induced EMT in L3.6 cells. Scratch wound healing assays were
   carried out on cells that had been pre‐treated with TGF‐β for 72 h to
   initiate the EMT process. The results demonstrated that the migration
   ability of shESE1 cells was dramatically increased following TGF‐β
   treatment, whereas rescued overexpression of ESE1 (shESE‐OE) attenuated
   this effect (Figure [122]4D,E). These results were further validated
   using cell‐tracking assays performed using the Operetta CLS high
   content analysis system (Figure [123]4F,G). Compared with the shCtrl
   group, the shESE1 group demonstrated significantly increased
   displacement. Compared with the shESE1 group, the shESE1‐OE group
   showed significantly reduced migration. Similarly, the expression of
   vimentin and N‐cadherin was significantly increased in ESE1 knockdown
   cells treated with TGF‐β, whereas overexpression of ESE1 relieved the
   increased levels of vimentin and N‐cadherin in shESE1‐OE cells
   (Figure [124]4H). Immunofluorescent staining revealed mesenchymal‐like
   morphologies of shESE1 cells treated with TGF‐β in L3.6 cells
   (Figure [125]S3). Similarly, compared with the shCtrl group, the shESE1
   group demonstrated significantly increased displacement in CFPAC‐1
   cells (Figure [126]S2F). These data indicate that TGF‐β induces EMT in
   PDAC cells by downregulating ESE1.

3.5. Downregulation of ESE1 and AGR2 highly correlates to EMT phenotype

   Overexpression of AGR2 was associated with advanced clinical stage,
   advanced tumor, but a better prognosis based on the TCGA database and
   KM‐plotter database in PDAC (Figure [127]S4A‐C), suggesting a complex
   effect of AGR2 on cancer growth and survival resembling that of ESE1.
   Similar to ESE1, the mRNA and protein expression of AGR2 were found to
   be higher in L3.6 than the others tested (Figure [128]1F,
   Figure [129]5A,B). Consistent with the previous reports that TGF‐β
   downregulates the expression of AGR2 in a cell context‐dependent
   manner,[130] ^38 we found that AGR2 was significantly downregulated
   during TGF‐β‐induced EMT in L3.6 and CFPAC‐1 cells (Figure [131]5C,D,
   Figure [132]S2E). Therefore, we hypothesized that AGR2 may participate
   in ESE1‐mediated inhibition of EMT. To test this hypothesis, we
   performed protein–protein interaction (PPI) network functional
   enrichment analysis for ESE1 protein using the Search Tool for the
   Retrieval of Interacting Genes/Proteins (STRING) database
   ([133]https://version11.string‐db.org/cgi/network.pl?taskId=VZEEHh0QUtm
   m). The analysis showed a positive relationship between ESE1 and AGR2,
   implying that ESE1 interacts with AGR2 in PDAC (Figure [134]5 E,F).
   Bioinformatics analysis of transcriptome sequencing data showed that
   compared with shCtrl, the expression of AGR2 was significantly
   downregulated in shESE1 cells (GSE 206999). Similar to the effect of
   ESE1 knockdown, shRNA‐mediated suppression of AGR2, as demonstrated in
   Figure [135]S4D,E, also reduced the colony formation, but promoted the
   migration of L3.6 cells, further revealing its resemblance of ESE1 in
   differential regulation of PDAC cell proliferation and migration
   (Figure [136]S4F,G). Moreover, similar to the effect of ESE1 knockdown
   on EMT markers, we found AGR2 knockdown markedly increased the
   expression level of the mesenchymal marker vimentin and decreased that
   of E‐cadherin in shAGR2 L3.6 cells (Figure [137]S4H). The correlation
   between AGR2 and ESE1 was further supported by Spearman's correlation
   coefficient analysis
   ([138]http://gepia.cancer‐pku.cn/detail.php?clicktag=correlation)
   (Figure [139]5G). We next determined the relationship between
   endogenous ESE1 and AGR2 in L3.6 cells. The results showed that the
   levels of AGR2 mRNA and protein were significantly reduced upon
   suppression of endogenous ESE1 (Figure [140]5H,I, Figure [141]S5A).
   Similarly, the levels of AGR2 mRNA (data not shown) and protein were
   significantly reduced upon suppression of endogenous ESE1 in CFPAC‐1
   cells (Figure [142]S2D). These results suggest AGR2 a potential
   transcriptional target of ESE1.

FIGURE 5.

   FIGURE 5
   [143]Open in a new tab

   Downregulation of ESE1 and AGR2 highly correlates to EMT phenotype. The
   relative protein and mRNA expression of AGR2 in PDAC cell lines and
   immortalized normal pancreatic epithelial cell line (HPNE) by Western
   blotting (A) and qRT‐PCR (B). Immunoblot analysis to determine AGR2,
   ESE1 and EMT‐related marker expression in L3.6 cells treated with TGF‐β
   for indicated time (C) and doses (D). (E) Network analysis of ESE1
   correlation by STRING database. (F) Analysis of the ESE1 interaction
   gene score. (G) The correlation analysis between AGR2 and ESE1 based on
   Spearman correlation analysis of GEPIA database. The protein (H) and
   mRNA (I) expression of AGR2 in L3.6‐shCtrl, L3.6‐shESE1, and
   L3.6‐shESE1‐OE groups. Data are presented as means ± SEM (n = 3 per
   group). ***p < 0.001, ****p < 0.0001.

3.6. TGF‐β inhibits ESE1‐mediated transactivation of AGR2 promoter

   To elucidate whether ESE1 functions as a cytoplasmic regulator like
   previously reported in breast cancer or as a nuclear transcriptional
   factor like most other ETS proteins, we examined the expression of
   endogenous ESE1 localization by IF and immunoblot analysis of the
   nuclear and cytoplasmic extracts in L3.6, PANC‐1, and CFPAC‐1. We found
   ESE1 predominantly localized in the nuclear (Figure [144]6A,D,
   Figure [145]S5B). Similar results were observed in HeLa cells with
   exogenously overexpressed ESE1 either with a C‐terminally fused GFP or
   an N‐terminal flag tag (Figure [146]6B,C). GAL4‐based heterologous
   reporter assay detected much stronger activation of luciferase reporter
   by ESE1, compared to its close homolog ESE3 (Figure [147]6E). Further
   analysis of ESE1 truncation mutants indicated that N‐terminal half of
   ESE1 containing the pointed and the transcription activation domains
   (PNT + TAD) were able to robustly transactivate the GAL4‐Luciferase
   reporter in all tested PDAC lines including BxPC3, L3.6, PANC‐1, and
   Patu8988T cells (Figure [148]S6A,B). Importantly, TGF‐β treatment
   significantly reduced ESE1‐mediated transactivation of the reporter
   (Figure [149]6F). Given the crucial role of ESE1 in maintaining the
   expression of the endogenous AGR2 (Figure [150]5H, Figure [151]S5A), we
   next examined whether ESE1 could directly activate AGR2 promoter.
   ESE1‐flag expression vector was co‐transfected with pGL3‐AGR2
   luciferase reporter constructs in HeLa cells with Flag empty vectors as
   negative control, and the SAM pointed domain‐containing ETS
   transcription factor (SPDEF), a gene reported to transcriptionally
   activate AGR2,[152] ^39 , [153]^40 as the positive control. ESE1 was
   found to more robustly increased the activity of AGR2 promoter than
   SPDEF (Figure [154]6G). These results indicated that ESE1 is a strong
   activator of AGR2 promoter transcription, and its activity is regulated
   by TGF‐β signaling.

FIGURE 6.

   FIGURE 6
   [155]Open in a new tab

   TGF‐β inhibits ESE1‐mediated transactivation of AGR2 promoter. (A)
   Detection of endogenous ESE1 expression and localization in L3.6 and
   PANC‐1 cells by IF. ESE1(green), nuclei (DAPI). HeLa cells were
   transfected with ESE1‐GFP (B) or ESE1‐flag expression plasmids (C) to
   reveal their predominant nuclear localization. (D) ESE1 protein
   expression in L3.6 cells was assessed by immunoblot analysis of
   cytoplasmic and nuclear fractions. (E) The ability of GAL4 DBD‐ESE1 and
   GAL4 DBD‐ESE3 fusion proteins to activate the GAL4‐Luc reporter in L3.6
   cells. (F) GAL4‐Luc reporter assay to determine the activity of GAL4
   DBD‐ESE1 fusion proteins in L3.6 cells with or without TGF‐β treatment.
   (G) Luciferase reporter assay to determine the transactivation of AGR2
   promoter by ESE1 relative to the positive control SPDEF. Data are
   presented as means ± SEM (n = 3 per group). ****p < 0.0001 compared
   with control, ^#### p < 0.0001 intergroup comparison. Scale bars:
   20 μm. PNT, TAD, and ETS indicate Sterile alpha motif/Point domain,
   transactivation domain, and ETS DNA binding domain, respectively.
   AGR2‐Luc, AGR2 luciferase reporter.

4. DISCUSSION

   Many studies reported contradictory roles of ESE1 as an oncogene or a
   tumor suppressor depending on the cancer types or different malignant
   properties, highlighting a complexed nature and mechanism of its
   action.[156] ^13 , [157]^41 , [158]^42 , [159]^43 The current work
   illustrated that ESE1 exhibits dual roles in regulation of malignant
   phenotypes in that its overexpression promoted cell proliferation,
   whereas its downregulation enhanced EMT phenotype in the L3.6 and
   CFPAC‐1 PDAC cells. Our data establish ESE1 as a negative regulator of
   TGF‐β‐induced EMT in PDAC. We further identify AGR2 as a novel
   transcriptional target of ESE1, that participates in TGF‐β‐induced EMT.
   These findings provide mechanistic insight into the differential roles
   of ESE1 in regulating different cell malignant properties and offer a
   potential explanation for its seemingly contradicting impact on RFS and
   OS in PDAC patients, as well as some early conflicting results reported
   in different types of cancer. Given the growing interest in
   ETS‐directed therapies, such as with small molecule inhibitors or
   siRNA,[160] ^44 , [161]^45 our results raise necessary precautions for
   applying this type of approach, and a better mechanistic understanding
   will facilitate design of drugs with desired inhibition of tumor
   growth, while eliminating the potential adverse effects on promoting
   EMT and metastasis.

   TGF‐β signaling is a predominant driver of EMT, a key initial step of
   cancer progression in various epithelial cancer including PDAC.[162]
   ^46 Previous studies have reported ESE1 as a transcriptional activator
   of TGF receptor type II (TGFβR‐II). ESE1 directly binds TGFβR‐II
   promoter and robustly increases its expression in breast cancer and the
   mouse embryonal carcinoma cells.[163] ^47 Genetic ablation of mouse
   ESE1 gene was associated with highly reduced TGFβR‐II expression.[164]
   ^48 Significantly, the developmental defect can be partially rescued by
   transgenic expression of Tgf‐βRII in ESE1(−/−) background. Together
   with the crystal structure of a mouse Elf3 complexed with mouse
   TGFβR‐II promoter DNA,[165] ^49 these studies established ESE1 as a
   bona fide and crucial upstream regulator of TGF‐β signaling both in the
   embryonic development and the cancer cells. The current identification
   of ESE1 as a downstream target of TGF signaling reveals a ESE1‐mediated
   feed‐back regulatory mechanism during TGF‐β‐induced EMT in PDAC, and
   perhaps also in other cancer. In contrast to the direct transcriptional
   activation of Tgf‐βRII by ESE1, we found that the expression of ESE1
   protein, but not its mRNA, is markedly down regulated by TGF‐β,
   suggesting that decreased ESE1 expression following TGF‐β‐induced EMT
   is mainly a post‐transcriptional event, but the underlying mechanism
   remains to be further investigated.

   Of note, ESE1 contains a serine‐ and aspartic rich (SAR) motif and an
   HMG‐like AT‐hook domain, which have been reported to confer ESE1 unique
   cytoplasmic localization and additional functional properties in breast
   cancer.[166] ^50 , [167]^51 The SAR domain is found both necessary and
   sufficient to maintain a morphological EMT and metastatic phenotype,
   whereas nuclear localized ESE1 protein contributes to maintain the
   transformed phenotype of breast cancer cells.[168] ^52 A monopartite
   nuclear localization signal (NLS1) in the AT‐hook and a three lysine
   residues (NLS2) within the ETS domain have been mapped.[169] ^53 ,
   [170]^54 Additionally, two CRM1‐dependent nuclear export signals(NES)
   located within the pointed domain and the ETS domain, respectively,
   have also been found.[171] ^53 Therefore, these NLS/NES may work
   together to regulate cytoplasmic‐nuclear shuttling of ESE1. However,
   unlike those reported in the breast cancer cells, we observed
   predominantly nuclear localized ESE1 of both endogenous protein and
   transfected full‐length protein expression plasmid in PDAC cells. We
   suggest that cytoplasmic‐nuclear shuttling represents a key knot
   governing ESE1's dichotomous activities toward promoting cell
   proliferation and suppression of EMT that may be cancer cell type
   dependent and subject to regulation by additional signaling pathways,
   such as TGF‐β as we demonstrated. The precise control mechanism of ESE1
   subcellular localization and its functional properties should be a key
   topic for future research.

   Despite of its role as a crucial guardian of the epithelial state[172]
   ^41 and a negative regulator of EMT in cancer,[173] ^12 , [174]^13 the
   signaling pathways governing the activity and the transcriptional
   targets of ESE1 in mediating its cellular effects remain largely
   elusive. The current work identifies AGR2 as a novel and functional
   ESE1 transcriptionally activated target gene in PDAC. Interestingly,
   AGR2 has previously been identified as a TGF‐β down‐regulated gene with
   an essential role in governing MUC1 expression, and in facilitating
   mPanIN initiation and progression to PDAC.[175] ^55 Reduced AGR2
   expression in human PDAC patients correlates to EMT phenotype,
   aggressive histological grade and adverse outcome highly resembling
   that of ESE1.[176] ^56 Similarly, in human lung adenocarcinoma,
   TGF‐β‐induced EMT leads to dramatic down regulation of AGR2, whereas
   forced overexpression of AGR2 largely reversed the TGF‐β‐induced
   EMT‐phenotype.[177] ^38 An independent study further show that TGF‐β
   inhibits ESE‐1 expression in the same cell lines, suggesting that AGR2
   downregulation is mediated, at least in part, by ESE1.[178] ^57 Our
   data together with the published observations suggest ESE1/AGR2 axis
   serves to maintain epithelial cell identity, which is antagonized by
   TGF‐β signaling in cancer cells to undergo EMT.

   Interestingly, both ESE1 and AGR2 are linked to ZEB1/2 regulation, a
   key EMT driver downstream of TGF‐β. AGR2 is identified as a direct
   target of ZEB1 transcriptional repression, and that AGR2 promotes ZEB1
   mRNA degradation therefore forming a double negative feedback loop in
   the EMT and metastasis process. The balance between AGR2 and ZEB1 is
   suggested to govern the aggressiveness and invasive phenotype of tumor
   cells.[179] ^24 On the contrary, ZEB1/2 and Snail have also been
   identified as ESE1 transcriptional targets of repression in biliary
   tract and colorectal cancer.[180] ^58 , [181]^59 In breast cancer
   cells, ETS1 and ESE1 reciprocally regulate the expression of ZEB1/2,
   and the net outcome is mainly dictated by ERK1/2 activity, which
   positively impacts on ETS1 transactivation of ZEB1/2, while it inhibits
   their transcriptional repression by ESE1.[182] ^60 The current results
   together with the published work reveal an intricate regulatory network
   between TGF‐β signaling and ESE1/AGR2 axis that function together with
   other EMT drivers such ZEBs to govern the normal epithelial homeostasis
   or EMT phenotype in cancer cells.

AUTHOR CONTRIBUTIONS

   Hui‐Jing Xu: Conceptualization (lead); data curation (lead);
   investigation (lead); methodology (lead); writing – original draft
   (lead). Jing Bai: Conceptualization (lead); data curation (lead);
   investigation (lead); methodology (lead). Ye Tian: Data curation
   (equal); investigation (equal); methodology (equal). Xiao Feng: Data
   curation (equal); investigation (equal); methodology (equal). Ai‐Ping
   Chen: Data curation (equal); investigation (equal); methodology
   (equal). Jie Wang: Data curation (equal); investigation (equal);
   methodology (equal). Jin Wu: Data curation (equal); investigation
   (equal); methodology (equal). Xu‐Ru Jin: Investigation (equal);
   methodology (equal). Feng Zhang: Data curation (equal); investigation
   (equal); methodology (equal). Mei‐Yu Quan: Data curation (equal);
   investigation (equal); methodology (equal). Chengshui Chen:
   Conceptualization (equal); funding acquisition (equal); resources
   (equal); supervision (equal). Kwang Yeol Lee: Conceptualization
   (equal); funding acquisition (equal); resources (equal). Jin‐San Zhang:
   Conceptualization (lead); data curation (lead); funding acquisition
   (lead); investigation (lead); methodology (lead); resources (lead);
   supervision (lead); writing – review and editing (lead).

CONFLICT OF INTEREST

   The authors declare that they have no competing associated with the
   manuscript.

ETHICS STATEMENT

   This study did not research on human or animal subject.

Supporting information

   Figure S1

   Figure S2

   Figure S3

   Figure S4

   Figure S5

   Figure S6
   [183]Click here for additional data file.^ (14.7MB, docx)

   Table S1
   [184]Click here for additional data file.^ (18.4KB, docx)

ACKNOWLEDGMENTS