An RNA-Binding Protein, Hu-antigen R, in Pancreatic Cancer Epithelial to Mesenchymal Transition, Metastasis, and Cancer Stem Cells

Pancreatic cancer has the highest fatality rate among all cancers, with a 5-year overall survival rate less than 8% (1). Whereas advancements in targeted therapies and immunotherapies have greatly improved outcomes in patients with many types of cancers, the benefits have not been gained for patients with pancreatic cancer. First-line chemotherapies such as gemcitabine plus nab-paclitaxel (2), or FOLFIRINOX (3), only achieve a median overall survival of 9 to 13 months in patients with advanced disease, and have multiple significant toxicities.

An enrichment of stem-like cancer cells (CSC) in pancreatic cancer has been proposed to root the poor prognosis and treatment outcomes of this disease (4, 5). The presence of CSCs is highly associated with cancer cell epithelial–mesenchymal transition (EMT) that contributes to chemoresistant tumors prone to metastasis and recurrence. EMT is typified by loss of cell–cell junctions and apico-basolateral polarity, resulting in the formation of migratory mesenchymal cells with invasive properties (6). EMT is a major process underlying the heterogeneity of cancer cells, and is an important initial step for cancer cell dissemination and metastasis. Through EMT, cancer cells dedifferentiate, gain motility and invasiveness, evade senescence, apoptosis, and immunosurveillance, and become resistant to conventional and targeted therapies (7–10). Induction of EMT in neoplastic cells also resulted in the enrichment of CSCs (11). Given the highly aggressive, drug resistant, and metastatic nature of pancreatic cancer, targeting EMT would cast hope for the treatment for this devastating disease.

EMT initiation and progress are regulated by complex signaling networks. HuR (the RBP Hu antigen R), an RNA-binding protein, is increasingly recognized as a pivotal factor in cancer-related gene expression, and is proposed to play a role in EMT regulation. HuR is a member of the embryonic lethal abnormal vision (ELAV) family, and is highly expressed in virtually all malignancies tested, including pancreatic cancer (12, 13). HuR contains three RNA recognition motifs (RRM), of which RRM1 and RRM2 are involved in RNA binding, and RRM3 is needed for cooperative assembly of HuR oligomers on RNA (14). HuR target mRNAs bear adenine- and uridine-rich elements (ARE) in their 3′- or 5′-UTRs (15). Cytoplasmic binding of HuR to these mRNAs generally confers the ARE-mediated mRNA decay and leads to mRNA stabilization and increased translation (16). HuR promotes tumorigenesis by interacting with a subset of mRNAs that encode proteins implement in cell proliferation, cell survival, angiogenesis, invasion, metastasis, and treatment resistance (17). It has been reported that HuR stabilizes mRNAs of MMPs, uPA, and probably Snail (18–20). Snail is a central EMT-promoting transcription factor. This suggested a regulatory role of HuR in cancer cell EMT. However, so far, there is only one published study demonstrating stabilization of Snail by HuR in a breast cancer cell line (20). The mechanisms that HuR could be involved in cancer cell EMT have not been well understood.

HuR is highly expressed in pancreatic cancer as well as many other types of cancers, compared with normal tissues (21). In normal cell, HuR is mostly located in the nucleus, while in cancer cells, the cytoplasmic level is elevated. The translocation of HuR from nucleus to cytoplasm is mediated by a nuclear–cytoplasmic shuttling sequence (NCS) between RRM2 and RRM3 (22). The elevated HuR expression levels and its cytoplasmic translocation enhances proinflammatory and oncogenic protein expressions (23, 24), and is correlated with advanced clinical pathology, and patient prognosis and survival rate in various types of cancers (25). Therefore, HuR is considered a putative target for cancer treatment. However, despite many efforts, there has been limited success in small molecules that directly disrupt the HuR interaction with AREs of its target mRNAs (26–30). In this study, we aim to fill the gap of understanding of the role HuR plays in pancreatic cancer cell EMT and metastasis, and explore a novel approach for pharmacologic inhibition of HuR.

Pancreatic cancer cell lines were from the ATCC. hTERT-HPNE cells (immortalized human pancreatic ductal epithelial cells) were donated by Dr. Anant at the University of Kansas Medical Center (Kansas City, KS). MTT assay was used for cell viability detection, with starting cell number in 96-well plate of 3,000/well (for 72-hour treatment) or 5,000/well (for 48-hour treatment).

Wound healing assay was performed by scratching confluent monolayer with a 100 μL pipette tip. Wound recovery was calculated by 100% - (remaining area ÷ original area) × 100% at each time point.

Matrigel invasion assay was performed using Boyden chambers (BD Biosciences) either precoated or uncoated with 0.1 mg/mL Matrigel, with 0.5% FBS inside and 10% FBS outside. Starting cell density was 1 × 10⁴/well.

For tumor spheres formation, single-cell suspension was plated into 96-well ultra-low attachment plates (Corning Inc.) at 100 cells/well in stem cell media, supplemented with B27 Supplement, 20 ng/mL human basic FGF, 20 ng/mL EGF, 100 U/mL penicillin/streptomycin (Invitrogen), and 4 μg/mL heparin calcium salt (Thermo Fisher Scientific). Tumor spheres were counted after 14 days, and size was measured using Image J software.

Total RNA was extracted using TRIzol reagent (Invitrogen). cDNA synthesis was performed with 1 μg RNA using Omniscript RT Kit (Qiagen), and diluted 1:5 for further use. Real-time PCR was performed using Bio-Rad iQ iCycler detection system with iQ SYBR Green Supermix (Bio-Rad Laboratories Ltd.). Data were normalized to 18S rRNA.

To detect the decay of mRNAs, cells were treated with 5 μg/mL actinomycin D to block transcription (at −0.5 hour). Total RNA was extracted at 0, 0.5, 1, 2, and 3 hours. KH-3 (2 μmol/L) was added 30 minutes after actinomycin D (at 0 hours).

Recombinant pcDNA3.1 HuR-flag Plasmid (pHuR) was provided by Dr. Dixon at the University of Kansas Cancer Center. The vector pcDNA3.1+ (pVec) was purchased from Addgene, and HuR siRNA from Qiagen. Plasmids were transfected by Lipofectamine 3000 reagent for 48 hours, and siRNA by Lipofectamine RNAiMAX reagent for 24 hours (Invitrogen). HuR levels were verified by Western blot analysis.

CRISPR/Cas9 deletion of HuR gene was performed using the lentiCRISPRV2 vector (AddGene). The control single guide RNAs (sgRNAs) and HuR sgRNAs were cloned into the vector following reported procedures (31). The HuR lentiviral sgRNA or control sgRNA were cotransfected into HEK293FT cells with the packaging plasmids pMD2.G and psPAX2 (AddGene). MIA PaCa-2 cells were infected with virus-containing medium and then selected with 1.0 μg/mL puromycin. Single clones were generated by limited dilution.

Total cell lysate was used for immunoprecipitation with anti-HuR or normal rabbit IgG (Cell Signaling Technology), using the Immunoprecipitation Kit (Protein G; Roche), supplemented with RNaseOUT Recombinant Ribonuclease Inhibitor (Invitrogen) in all steps (100 U/mL). In the KH-3 treatment groups, KH-3 (2 μmol/L) was supplemented in all steps. Total RNA was then extracted from the immunoprecipitation products by TRIzol reagent and subjected to qRT-PCR analysis.

The full-length Snail mRNA 3′-UTR was synthesized by Genewiz. The two truncated Snail mRNA 3′- UTRs (ΔAREs, and AREs) were cloned from total RNA of MIA PaCa-2 cells and amplified by PCR, and then constructed into the pmirGLO dual luciferase reporter plasmid (Fig. 2C). MIA PaCA-2 HuR KO cells were cotransfected with pmirGLO dual luciferase reporter with or without the constructions (full-length, ΔAREs, AREs, or empty reporter; Promega) and pCDNA-3.1+-HuR (or empty vector) using Lipofectamine 3000 reagent (Invitrogen). KH-3 was added at 24 hours, and the dual-glo luciferase reporter assay was performed at 48 hours using Dual-Glo Luciferase Assay System (Promega).

Cells were lysed with RIPA buffer (Sigma Al), and total protein was subjected to Western blotting. BCA method was used for protein quantification (Pierce BCA Protein Assay Kit). Blots were established using either Pierce ECL substrate or Pierce ELC+ substrate (Thermo Fisher Scientific).

Immunofluorescence detection of protein expression was performed with cells grown on 6-well chamber slides as routine. Blocking was performed using 5% Goat serum + 0.3% Triton X-100. Nucleus were stained with ProLong Gold Antifade Reagent containing DAPI (Cell Signaling Technology).

IHC was performed with paraffin-embedded tissue sections (5-μm thick), as routine. DAB were used to develop the sections [HRP/DAB (ABC) detection IHC Kit, Abcam]. All the sections were then counterstained with hematoxylin.

The compound KH-3 was synthesized in-house by a collaborator. The synthesis and structure validation were reported in another publication (32).

All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Kansas Medical Center under the protocol #2015–2247. A pilot MTD (MTD) experiment was carried out to determine the dose regimen to be used in treatment. Mice (n = 3) were started at 50 mg/kg body weight of KH-3 daily, intraperitoneal injection for 3 consecutive days. No clinical signs of toxicities were observed. Dose was then escalated to 75 mg/kg for 3 days and when no signs of toxicities were observed; dose was increased to 100 mg/kg for 5 consecutive days. On day 5 of 100 mg/kg, reduction in activities were observed in mice. Dose was reduced to 3 times weekly at 100 mg/kg. No signs of toxicities were observed. The 100 mg/kg 3 times weekly intraperitoneally was determined as the treatment dose. All treatment concerning KH-3 used this dose regimen.

A subcutaneous tumor model was used to determine tumor formation rate. MIA PaCa-2 HuR WT cells or MIA-PaCa-2 HuR KO cells were inoculated into the flank of female Ncr nu/nu mice at the number of 2 × 10⁶ cells in PBS. Tumor formation was monitored daily, and tumor size was measured 3 times/week using a digital caliper.

An orthotopic pancreatic tumor model was used to determine treatment effects of KH-3. Luciferase-expressing PANC-1 cells (PANC-1–Luc, multi-clones) were established by the Preclinical Proof of Concept Core Laboratory (University of Kansas Medical Center). A small subcostal laparotomy was performed in female Ncr nu/nu mice to expose the pancreas, and 2 × 10⁵ PANC-1-Luc cells in 50 μL PBS were injected into the tail of pancreas. After 11 days, the localized tumors inside the pancreas of these donor mice were removed and minced into small pieces of 1 mm³ cube. One tumor cube was implanted into the pancreas of one recipient nude mouse by laparotomy. After 11 days, the recipient mice were scanned for xenograft formation using an IVIS imaging system upon intraperitoneal injection of 150 mg/kg d-luciferin. Mice were grouped based on tumor burden and treatment commenced as described, with weekly follow-up imaging. Body weight was measured 3 times weekly, and mice were monitored for clinical signs of toxicity during treatment, including guarding, abnormal appearance (hunched), restlessness, and reluctance to move. Treatment lasted for 5 weeks, and gross necropsy was performed at the end of treatment.

Statistical analysis was performed using SPSS software for Student t test, log-rank test, one-way ANOVA with Turey method, or Mann–Whitney U test as each condition applies. A difference was considered statistically significant at the P < 0.05 level. Correlation was analyzed by Pearson test.

To study the role of HuR in pancreatic cancer cell EMT, HuR expression was first silenced by transfecting siRNAs targeting HuR mRNA (siHuR), and downregulation of HuR protein was validated by Western blots. In two human pancreatic cancer cell lines PANC-1 and MIA PaCa-2 transfected with siHuR, the cellular morphology changed to a more epithelium-like state compared with each of their parent cells, characterized by less spindle-like cells, shortened cell length, and/or enlarged cell diameter (Supplementary Fig. S1). Consistent with this phenotypical change, the expressions of signature EMT genes in both cells were altered (Fig. 1A): the epithelial marker Claudin1 was significantly upregulated, the mesenchymal marker Vimentin was downregulated, and the EMT enhancing transcription factor Snail was significantly decreased. We then performed permanent deletion of HuR gene in MIA Paca-2 cells by a CRISPR/Cas-9 method. As expected, the deletion caused a depletion in HuR protein in the cells (Fig. 1A). The morphology of HuR-deleted cells (HuR KO) showed a more epithelium-like state compared with the control cells (Supplementary Fig. S1), again with increase in Claudin1, decrease in Vimentin and Snail, consistent with the results in siHuR transfection (Fig. 1A). Claudin1 expression was further confirmed by immunofluorescence staining in the siHuR transfected MIA PaCa2 cells. Results clearly showed increase of Claudin1 expression (Fig. 1B).

As EMT promotes cancer cells migration and invasion, we expected HuR downregulation would inhibit pancreatic cancer cell migration and invasion, and HuR overexpression would enhance them. Indeed, siHuR significantly decreased the migration of PANC-1 and MIA PaCa-2 cells in a wound healing assay (Fig. 1C). Consistently, HuR KO MIA PaCa-2 cells also had decreased ability to migrate (Fig. 1C). Migration/invasion were further assessed using Matrigel uncoated and coated Boyden chambers. siHuR inhibited migration and/or invasion in both MIA PaCa2 cells and PANC-1 cells, and HuR gene deletion in MIA PaCa2 cells greatly impaired cell migration and invasion (Fig. 1D). Because HuR also regulates cell proliferation, there is a possibility that the inhibition in migration/invasion detected here was due to inhibition in proliferation. To address this question, we examined the proliferation of siHuR and HuR KO cells. SiHuR MIA PaCa-2 cells had the same growth rate as the SiCtrl cells and untreated cells (Ctrl) up to 72 hours (Supplementary Fig. S2A), suggesting the inhibition in migration/invasion was independent of cell proliferation. A difference in growth was observed between the HuR KO MIA PaCa-2 cells and the wild-type cells at 48 and 60 hours. At 60 hours, HuR KO cells had approximately 40% growth inhibition compared with wild-type cells (Supplementary Fig. S2B) and could contribute to the inhibition of gap-closing detected in the HuR KO cells. Considering HuR KO cells were permanently and completely deleted with HuR, while siHuR was temporary and not a 100% inhibition, it is likely that inhibition of HuR may first influence EMT and migration/invasion in the tested cells, and when the depletion of HuR is more severe, growth/proliferation was affected.

We then examined the cancer stem-like cell population (CSC) using tumor spheroid formation assay. Data showed that the number and size of tumor spheres were both significantly reduced in PANC-1 and MIA PaCa2 cells with siHuR transfection (Fig. 1E), indicating inhibition in CSCs. The HuR gene deletion also decreased the number of spheres, but did not influence the sizes of the spheres formed (Fig. 1E).

We then reexpressed HuR in the HuR KO MIA PaCa2 cells and examined the EMT markers, migration, and CSCs. HuR reexpression decreased the epithelial markers Claudin1 and ZO-1 and increased the mesenchymal marker Vimentin, and Snail (Fig. 1F). The restore of HuR expression also enhanced migration (Fig. 1G) and increased number of tumor spheres (Fig. 1H) compared with HuR KO cells while the size of the formed spheres slightly decreased.

As the number of CSCs is responsible for tumorigenicity in vivo, we compared the tumor formation rate of MIA PaCa2 HuR KO cells to that of the CRSPR/Cas9-control cells (HuR WT cells) in nude mice. At the inoculation number of 2 × 10⁶ cells subcutaneously, the HuR WT cells yield 100% (16/16) tumor formation in 8 days after injection (day 8). The HuR KO cells had a tumor formation rate of 25% (4/16) at day 8, and only reached a final tumor formation rate of 37.5% (6/16) at day 21 (Fig. 1I). At day 20, the average tumor volume of HuR KO tumors was significantly smaller than that of the WT tumors (Fig. 1J).

HuR typically stabilizes its targeting mRNAs and promotes translation by binding to adenine- and uridine-rich elements (ARE) located in the 3′untranslated region (UTR) of the target mRNA. We examined whether HuR binds to the mRNAs of important regulators of EMT and CSCs, using ribonucleoprotein immunoprecipitation (RNP-IP) assay (33). Pull-down products from MIA PaCa2 total cell lysate using anti-HuR antibody were quantified for RNA components by qRT-PCR. mRNAs of a panel of EMT/CSC regulators showed strong association with HuR protein, that is, Snail, Slug, Zeb1, and β-catenin, as well as the mRNAs of the known HuR targets Msi1 and HuR itself (Fig. 2A; refs. 34, 35). In HuR KO cells, this panel of mRNAs were not pulled down (Fig. 2A).

We postulated that the binding to HuR stabilized these mRNAs. HuR WT and HuR KO MIA PaCa2 cells were treated with actinomycin D to block transcription, and then the stability of these mRNAs was detected during time. Data showed significantly enhanced degradation of Snail mRNA (Fig. 2B), but the decay of the mRNAs of Slug, Zeb1, and β-catenin did not change by the knockdown of HuR (Supplementary Fig. S3A) despite binding of their mRNAs to HuR. Consistent with these results, the protein expression of Snail was decreased with HuR knockdown (Fig. 1A), whereas the protein levels of Slug, Zeb1, and β-catenin were minimally influenced (Supplementary Fig. S3B).

The direct interaction of HuR with Snail mRNA 3′-UTR was examined with a luciferase reporter assay. The full-length 3′-UTR, and two truncated Snail mRNA 3′- UTRs (ΔAREs and AREs) were each constructed into the pmirGLO vector, which contains a firefly luciferase gene under the PGK promoter (Fig. 2C). The sequence of ΔAREs did not contain the AU-rich HuR-binding elements, and the sequence of AREs contained the major part of the AU-rich elements in the 3′-UTR (Fig. 2C). MIA PaCa2 HuR KO cells were then cotransfected with HuR and the pmirGLO plasmid containing each of the constructed Snail UTRs. Data clearly showed that only with the full-length 3′-UTR and the AREs, HuR transfection could enhance luminescence signal, and when there was a lack of HuR binding elements (ΔAREs), the luminescence signal did not change with HuR transfection (Fig. 2D).

To further determine the functional importance of Snail in the HuR-regulated EMT and migration. We reexpressed Snail in HuR KO MIA PaCa2 cells and detected the migration ability of the cells. The restoration of Snail significantly increased migration of the HuR KO cells (Fig. 2E).

Using a reported fluorescent polarization assay (30), a novel compound KH-3 (Fig. 3A) was identified that bound to HuR and interrupt HuR–mRNA interaction. The direct interaction of KH-3 to HuR was identified and validated using multiple in vitro binding assays: surface plasmon resonance (SPR) analysis, fluorescence polarization assay (FP), and amplified luminescent proximity homogeneous assay (Alpha assay). The Alpha assay and FP assay showed that KH-3 interfered HuR binding to one of its known target Msi1 mRNA (Ki ∼300–700 nmol/L). The SPR assay identified that KH-3 directly bound to the RNA Recognition Motifs of HuR (RRM1/2). These data are reported in another article (32).

Pancreatic cancer cell lines with different endogenous HuR expression levels were then treated with serial concentrations of KH-3 for 48 hours. KH-3 induced cytotoxicity in pancreatic cancer cells, with the sensitivity correlated to endogenous HuR protein levels (Fig. 3B and C). MIA PaCa2 cells have the highest HuR protein abundance among the tested cell lines and were the most sensitive to HuR treatment (IC₅₀ = 5 μmol/L). PANC-1 cells have the lowest HuR expression level and were the most resistant among the tested cancer cells (IC₅₀ = 25 μmol/L). BxPC-3 cells, another human pancreatic cancer cell line, have HuR expression level in the middle, and the IC₅₀ of KH-3 was in the middle (10 μmol/L). A noncancerous human pancreatic ductal epithelial cell line (hTERT-HPNE) was tested under the same conditions. hTERT-HPEN cells have the lowest abundance of HuR protein compared with the cancer cells, and the cytotoxicity of KH-3 to these cells was minimal (IC₅₀ >> 40 μmol/L). There is an inverse correlation in the tested cell lines between the HuR expression levels and the sensitivity to KH-3 treatment (Fig. 3C; R = −0.71 by Pearson tests).

EMT signature gene expression was altered by KH-3 treatment in both MIA PaCa2 and PANC-1 cells showing Vimentin and Snail decreases, and Claudin1 increase (Fig. 4A and B), mimicking the consequences of HuR knockdown shown above (Fig. 1A). The alternation indicated EMT inhibition. HuR expression was not changed (Fig. 4A), confirming that KH-3 works through interrupting HuR-mRNA binding but does not alter HuR expression.

KH-3 inhibited MIA PaCa2 and PANC-1 cells migration and invasion in the wound healing assay as well as in the Boyden chamber trans-well assay (Matrigel assay; Fig. 4C and D). To examine the target specificity of KH-3, HuR knockdown cells were used. In both the siHuR cells (MIA PaCa2 and PANC-1) and the CRISPER/Cas9 HuR KO cells (MIA PaCa2), the knockdown of HuR itself resulted in dampened migration compared with the wild-type cells, as expected. Importantly, in the HuR knocked down cells, KH-3 lost its target, and did not show additional effects to the effect of the knockdown (Fig. 4C–F). We then reexpressed HuR in the HuR KO cells by transfecting the cells with an HuR-expressing plasmid. When HuR was restored, KH-3 showed inhibitory effect again to the migration of the cells (Fig. 4G).

Tumor sphere formation was inhibited by KH-3 treatment. In PANC-1 cells, 10 μmol/L of KH-3 eliminated tumor sphere formation, while in MIA PaCa2 cells, 4 μmol/L of KH-3 had the similar effects. As in BxPC-3 cells, 8 μmol/L of KH-3 significantly inhibited both the number and the size of tumor spheres (Fig. 4H).

RNP-IP assay was recruited to examine the interruption of binding between HuR and its target mRNAs with KH-3 treatment. It is expected the HuR downstream EMT-related mRNAs will less likely to be coprecipitated with HuR protein upon KH-3 treatment. Indeed, KH-3 treatment at 2 μmol/L for 24 hours significantly decreased the pull-down amounts of mRNAs of Snail, Slug, Zeb1, β-catenin, HuR, and Msi1 in MIA PaCa2 cells (Fig. 5A), consistent with but slightly less efficient than the HuR KO. Parallel with HuR KO, the KH-3 treatment (2 μmol/L) enhanced Snail mRNA decay (Fig. 5B), and decreased the protein level of Snail (Fig. 4A).

Interruption of KH-3 to the binding of HuR with Snail 3′-UTR was further examined in the luciferase reporter assay. With cotransfection of HuR and full-length Snail 3′-UTR or AREs, KH-3 treatment inhibited the luminescence signal (Fig. 5C), clearly demonstrating interruption of HuR interaction with the 3′-UTR. When there was a lack of HuR-binding elements (with ΔAREs), KH-3 had no influence on the luminescence signal (Fig. 5C).

The in vivo tumor inhibitory effects of KH-3 were tested in a highly metastatic orthotopic model of pancreatic cancer. To avoid peritoneal lesions resulted from leak of injection, 5 mice (donor mice) were injected with luciferase expressing PANC-1 cells (PANC-1-Luc, 2 × 10⁵) and tumors developed in the pancreas of these donor mice were harvested and cut into approximately 1 mm³ and implanted into the pancreatic parenchyma of recipient mice. After 2 weeks, the recipient mice were imaged for tumor development and grouped to have equip average tumor burden (n = 9 for control group, and n = 10 for KH-3–treated group). Treatment then commenced with KH-3 at 100 mg/kg, i.p., 3 times weekly, determined based on MTD as described in Materials and Methods. The treatment continued for 5 weeks, and mice were euthanized, and gross necropsy was performed. The data showed that KH-3 treatment significantly inhibited longitudinal tumor growth and reduced tumor burden compared with the vehicle-treated group (control; Fig. 6A and B). The final tumor weight was significantly reduced (Fig. 6C). In the control group, 5 of 9 mice developed uncountable lesions of metastasis in the liver (56%), whereas in the KH-3–treated group, only 1 of 10 mouse developed metastasis (10%; Fig. 6D). At the end of the study, tumor tissues were examined for EMT alternations by Western blots. The epithelial markers Claudin1 and ZO1 trended toward increase, and Snail trended toward downregulation with the KH-3 treatment (Fig. 6E). This is consistent with the expected EMT inhibition. IHC confirmed the high expression level of HuR in the tumor tissues compared with the adjacent normal pancreatic tissues. KH-3 treatment did not change the expression level of HuR in the tumor tissues (Fig. 6F).

No clinical signs of toxicity were observed during the treatment. There is a difference in body weight between control and treatment group after 21 days of treatment. However, because of tumor growth, it is yet difficult to attribute the changes in body weight to toxicity. Histologic examination of the liver found no changes in the treatment group (Supplementary Fig. S3).

Furthermore, an in vivo treatment was performed using MIA PaCa2 HuR KO tumors, to examine whether the inhibitory effects of KH-3 were dependent on HuR. Because the HuR KO cells did not form tumors orthotopically, cells were subcutaneously inoculated, and tumor formation and growth were monitored with caliper measurement. The KH-3 treatment started on the same day the cells were inoculated. KH-3 (100 mg/kg, i.p., 3 times weekly) did not influence either tumor formation (Fig. 6G) or tumor growth (Fig. 6H) of the HuR KO tumors. Theses data, together with the data with the orthotopic tumor model, strongly indicated that the KH-3 effects were dependent on HuR.

In pancreatic cancer, higher HuR level is associated with higher tumor T stage in patients (21) and resistance to gemcitabine treatment (36, 37). HuR is not likely to be a tumor initiator in pancreas by itself, but rather facilitates tumor development (38). Pancreas-specific transgenic HuR mouse did not form spontaneous tumor; instead, the elevated intra-pancreas HuR level promoted a pancreatitis-like inflammatory microenvironment that could facilitate tumor development (38). Specific silencing or knockout of HuR inhibited pancreatic cancer cell proliferation, migration, invasion, and disabled in vivo xenograft formation (39). We hypothesized that HuR plays an important role in regulating pancreatic cancer cell EMT and stemness, and this regulation underlines the aggressiveness of the tumor in terms of invasion, metastasis, drug resistance, and new tumor generation. Data here reveal that HuR enhances pancreatic cancer cell EMT, mainly by stabilizing Snail mRNA and enhancing its protein expression. This enhancement in EMT promotes pancreatic cancer cell migration and invasion. The enhancement of EMT process by HuR also has implications in pancreatic CSC formation and maintenance. Inhibition of HuR dampened the ability of pancreatic cancer cell to migrate and invade, and inhibited CSCs. These data add to our knowledge of the role that HuR is playing in tumor metastasis and cancer stem cells, which was not well understood before.

HuR is more and more recognized as a responder in the cells to various stresses, and is involved in many physiological and pathologic processes. Apparently, its function and the downstream genes regulated are different in different tissues and under different conditions. For example, under hypoxia HuR enhances VEGF and HIF-1α expression (40), whereas under oxidative stress, HuR enhances expression of cyclins and sirtuins (41). HuR is also shown to causally linked to the onset of inflammation in kidney disease (42) and contributes to liver fibrosis (43), as well as many other pathologic processes related to inflammation, whereas in adipose tissue HuR protects against diet-induced obesity and insulin resistance (40). The oncologic microenvironment and the cancer cell itself might provide a “stress” condition that upregulates and activates HuR, with mechanisms not yet understood. Data here showed that HuR was differentially expressed in cancerous and normal pancreatic cells and mice tissues, consistent with clinical reports of the association between HuR levels and pancreatic cancer. The results here showed that in pancreatic cancer cells, HuR bound to a panel of mRNAs of regulator genes in the process of cancer cell EMT, but apparently, the binding did not influence protein expressions of all the bound mRNAs. Here, only Snail mRNA was stabilized and protein expression increased. The mechanism of this selectivity remains to be understood. Another unexplored area in this study is the potential interaction and cross-talk between the tumor microenvironment and HuR expression and function in the cancer cells and in the surrounding cells, such as pancreatic stellate cells, immune cells, and fibroblasts.

Given the importance of HuR in EMT and CSCs, along with its roles in other signaling pathways involved in cancer progress, it is intriguing to develop pharmacologic inhibitors of HuR for the treatment of cancer. However, there has been little success in small molecules that inhibit HuR functions, partially due to the unclear structure of the RNA-interacting pocket in the HuR protein (26). The only small molecule advanced to early-phase clinical trial is MS-444, a myosin light chain kinase inhibitor. MS-444 interferes nuclear–cytoplasmic translocation of HuR (26); however, it has multiple issues on specificity and toxicities. Our data showed that a novel compound KH-3 works in a different mechanism from MS-444. KH-3 binds to RNA recognition motifs (RRM1/2) of HuR protein and interrupts HuR–mRNA interaction. Consequently, KH-3 inhibited pancreatic cancer cell EMT, migration, invasion, and CSCs, mimicking the in vitro effects of HuR knockdown. KH-3 also inhibited cell viability dependent on endogenous HuR levels. As an HuR inhibitor, KH-3 may comprehensively impact pancreatic cancer cell proliferation, migration, and EMT.

Importantly, KH-3 showed good target specificity to HuR. Our in vitro data showed that when HuR was knocked down in pancreatic cancer cells, KH-3 lost its inhibitory activity against cell migration. In vivo, KH-3 inhibited a HuR positive tumor growth and metastasis, but did not inhibit the formation of HuR knocked-down tumors, nor inhibited the growth of these tumors. Good target specificity is likely to predict less off-target effects that are undesirable in drug development.

As a single-drug treatment, KH-3 inhibited pancreatic cancer growth and reduced metastasis rate in vivo, with good tolerability in mice at the dose of 100 mg/kg 3 times i.p. weekly. Combination treatments with currently available chemo-regimens are worth testing for synergy. Being the first of its class, KH-3 showed the promise of using target-specific small molecules to inhibit HuR function by direct interruption of HuR–RNA binding. Pharmacologic inhibition of HuR holds the promise to comprehensively inhibit pancreatic cancer progress, metastasis, drug resistance, and tumor recurrence.

X. Wu reports grants from NIH (R01 CA191785) during the conduct of the study, as well as a patent for Therapeutic Uses of Inhibitors of The RNA-Binding Protein HuR, provisional United States Patent Application pending (the manuscript involved a compound covered by this pending patent). L. Xu reports grants from University of Kansas and UNC (grant support cited in the manuscript: R01 CA191785 from NIH) during the conduct of the study, as well as a patent for Therapeutic Uses of Inhibitors of The RNA-Binding Protein HuR, provisional United States Patent Application filed on March 30, 2020, pending (the manuscript involved a compound covered by a pending patent). J. Aubé reports grants from NCI (R01 CA191785) and other from UNC (unrestricted research support) during the conduct of the study, as well as a patent for Therapeutic Uses of Inhibitors of The RNA-Binding Protein HuR pending (Provisional United States Patent Application filed on March 30, 2020). Q. Chen reports grants from University of Kansas Medical Center (R21 grant from NCI) during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

R. Dong: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing, R. Dong conducted most of the experiments and collect original data for this report. P. Chen: Data curation, methodology, project administration. K. Polireddy: Data curation, investigation, writing-review and editing. X. Wu: Data curation, methodology, writing-review and editing. T. Wang: Data curation, methodology, writing-review and editing. R. Ramesh: Validation. D.A. Dixon: Resources, methodology, writing-review and editing. L. Xu: Conceptualization, resources, methodology, writing-review and editing. J. Aubé: Resources, methodology, writing-review and editing. Q. Chen: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing.

This work was supported in part by NIH grant R21 CA198265 (to Q. Chen) and R01 CA191785 (to L. Xu and J. Aubé). We thank Dr. Shrikant Anant at the University of Kansas Cancer Center for providing cells used in our experiments. We thank Dr. Lisa Zhang at the University of Kansas Medical Center for providing HuR antibody, and help with troubleshooting in immune-blots experiments. We thank our previous postdoctoral fellow Dr. Ying Zhang for exploratory work on this project.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.