The mRNA-binding protein HuR promotes hypoxia-induced chemoresistance through posttranscriptional regulation of the proto-oncogene PIM1 in pancreatic cancer cells

Abstract

Previously, it has been shown that pancreatic ductal adenocarcinoma (PDA) tumors exhibit high levels of hypoxia, characterized by low oxygen pressure (pO₂) and decreased O₂ intracellular perfusion. Chronic hypoxia is strongly associated with resistance to cytotoxic chemotherapy and chemoradiation in an understudied phenomenon known as hypoxia-induced chemoresistance. The hypoxia-inducible, pro-oncogenic, serine-threonine kinase PIM1 (Proviral Integration site for Moloney murine leukemia virus 1) has emerged as a key regulator of hypoxia-induced chemoresistance in PDA and other cancers. Although its role in therapeutic resistance has been described previously, the molecular mechanism behind PIM1 overexpression in PDA is unknown. Here, we demonstrate that cis-acting AU-rich elements (ARE) present within a 38-base pair region of the PIM1 mRNA 3’-untranslated region mediate a regulatory interaction with the mRNA stability factor HuR (Hu antigen R) in the context of tumor hypoxia. Predominantly expressed in the nucleus in PDA cells, HuR translocates to the cytoplasm in response to hypoxic stress and stabilizes the PIM1 mRNA transcript, resulting in PIM1 protein overexpression. A reverse-phase protein array revealed that HuR-mediated regulation of PIM1 protects cells from hypoxic stress through phosphorylation and inactivation of the apoptotic effector BAD and activation of MEK1/2. Importantly, pharmacological inhibition of HuR by MS-444 inhibits HuR homodimerization and its cytoplasmic translocation, abrogates hypoxia-induced PIM1 overexpression and markedly enhances PDA cell sensitivity to oxaliplatin and 5-fluorouracil under physiologic low oxygen conditions. Taken together, these results support the notion that HuR has prosurvival properties in PDA cells by enabling them with growth advantages in stressful tumor microenvironment niches. Accordingly, these studies provide evidence that therapeutic disruption of HuR’s regulation of PIM1 may be a key strategy in breaking an elusive chemotherapeutic resistance mechanism acquired by PDA cells that reside in hypoxic PDA microenvironments.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDA) is a lethal malignancy with a 5-year survival rate of ⩽ 7%,^1,2 in part, because of tumor-associated acquired drug resistance.^3,4 The selective pressure imposed by the PDA tumor microenvironment favors growth of resilient tumor cells, which tend to become chemoresistant.^4–6 It has been shown that pancreatic tumors are embedded in a highly hypoxic stromal microenvironment wherein clonal populations of PDA cells thrive and expand.^6–8 In response to hypoxia, cells undergo an acute reprogramming that activates pathways responsible for regulating intracellular pH, mitochondrial function, DNA repair and cell survival.^9–11 Recent studies demonstrated that intratumoral hypoxia renders cancer cells refractory to chemotherapy.^12–15

An established regulator of hypoxia-induced chemoresistance is the proto-oncogene PIM1 (Proviral Integration site for Moloney murine leukemia virus 1), a serine-threonine kinase. PIM1 is a prominent modulator of therapeutic resistance in head and neck squamous cell carcinoma, prostate carcinoma and, recently, PDA.^16–19 Histologic analysis of PIM1 expression in PDA revealed a strong correlation with hypoxia markers.²⁰ PIM1 drives chemoresistance by phosphorylating and inactivating key apoptotic and tumor-suppressive proteins.²¹ In the context of tumor hypoxia, the mechanism behind PIM1 regulation is unknown. It has been demonstrated that hypoxia-mediated PIM1 overexpression occurs in a hypoxia-inducible factor 1α- (HIF-1α) independent manner,¹⁸ and no known PIM1 genetic alterations can account for PIM1 abundance in PDA cells. Based on the paucity of data on PIM1 regulation, we sought to determine if PIM1 is regulated via a cancer-specific posttranscriptional mechanism.

An RNA-binding protein central to posttranscriptional gene regulation in cancer, HuR (Hu antigen R; ELAVL1), is a member of the embryonic lethal, abnormal vision, Drosophila-like protein family. It is primarily nuclear localized, where it assists cytoplasmic export of mRNA transcripts, followed by rapid nuclear relocalization.^22,23 Upon stress (e.g., glucose deprivation²⁴ and chemotherapeutic stress²⁵), HuR translocates to the cytoplasm and enhances the stability of specific, prosurvival mRNAs.^24,25 Clinically, we and others have demonstrated that HuR cytoplasmic localization is associated with poor clinical prognosis in many cancers.^26–29 Here, we demonstrate for the first time that cis-acting adenylate uridylate-rich elements (AREs) present in the 3’-untranslated region (3’-UTR) of the PIM1 mRNA mediate its interaction with HuR in hypoxia. This interaction results in PIM1-dependent chemoresistance in PDA cells.

RESULTS

Hypoxia promotes PDA chemoresistance

Oxaliplatin and 5-fluorouracil (5-FU) are both part of an established regimen (FOLFIRINOX) for PDA.³ PDA cell lines incubated in hypoxia became more resistant to oxaliplatin in a dose-dependent manner (Figure 1a). Significantly higher half-maximal inhibitory concentration (IC₅₀) values were observed for cells incubated in hypoxia vs normoxia. Similar results were obtained with 5-FU³ (Supplementary Figure S1A).

Figure 1.

The proto-oncogene, serine–threonine kinase PIM1 promotes hypoxia-induced chemoresistance. (a) PDA cell lines were treated with the indicated doses of oxaliplatin in normoxia or hypoxia for 72 h. Cell survival was calculated by measurement of dsDNA content using PicoGreen, and dose–response curves were fitted based on a sigmoidal dose–response model. Each data point represents the mean of four independent experiments ± s.e.m. The table depicts IC₅₀ values derived from dose–response curves. (b) MiaPaCa2 cells were incubated in hypoxia for the indicated times and PIM1 mRNA expression was determined by qPCR using 18S rRNA as a loading control (*P ⩽ 0.01, error bars s.e.m.). PIM1 protein expression was analyzed by western blot using α-tubulin as a loading control. Densitometric analysis was carried out by ImageJ. (c) Human PDA specimens were stained for PIM1 and carbonic anhydrase 9. Scale bar = 200 μM. (d) PDA cell lines were incubated in normoxia (N) or hypoxia (H) for 24 h and PIM1 expression was assayed by western blot analysis. The fold induction of PIM1 expression (densitometry by ImageJ) was plotted against the hypoxia IC₅₀ values and fitted to a linear regression to determine the Pearson’s r. The two-tailed P-value indicates the statistical significance of the relationship between PIM1 induction and hypoxia IC₅₀ values. (e) MiaPaCa2 cells were transfected in triplicates with siRNA against PIM1 (siPIM1). PIM1 mRNA and protein expression were measured by qPCR and western blot analysis, respectively (*P ⩽ 0.01, error bars s.e.m.). (f) PicoGreen-based drug sensitivity assays to oxaliplatin were carried out in MiaPaCa2 cells transfected with siPIM1 and incubated in normoxia or hypoxia for 72 h. IC₅₀ values derived from dose–response curves are shown in the table.

PIM1’s role in hypoxia-induced chemoresistance

PIM1 has been implicated in hypoxia-induced chemoresistance in PDA models.^18,30 Quantitative PCR (qPCR) and western blot analyses demonstrate that the PIM1 mRNA and protein expression are induced in a time-dependent manner in hypoxic conditions (Figure 1b). Analysis of human PDA specimens by immunohistochemisty of carbonic anhydrase IX and PIM1 revealed that PIM1 expression occurred in hypoxic niches of the tumor, whereas normoxic areas exhibit low-to-undetectable PIM1 expression (Figure 1c). The contribution of hypoxia-mediated induction of PIM1 expression to chemoresistance was determined in a panel of seven PDA cells lines where hypoxia induced PIM1 expression significantly and this correlated with a decrease in oxaliplatin sensitivity (Pearson’s r = 0.77, P = 0.0467) (Figure 1d).

To determine if PIM1 expression contributed to hypoxia-induced chemoresistance, MiaPaCa2 cells were transfected with small interfering RNA (siRNA) against PIM1. Although PIM1 knockdown (Figure 1e) had no effect on oxaliplatin sensitivity in normoxia, it promoted a significant leftward shift in IC₅₀ in hypoxia (Figures 1e and f).

Hypoxia-induced PIM1 expression is independent of HIF-1α and HIF-2α To validate previous findings that PIM1 expression was not regulated by HIF-1α,¹⁸ MiaPaCa2 cells were transfected with HIF-1α shRNA (Supplementary Figure S1B) and PIM1 expression was evaluated. In the context of HIF-1α knockdown, PIM1 expression remained unaltered in normoxia, and HIF-1α knock-down did not abrogate hypoxia-mediated induction of PIM1 expression (Supplementary Figures S1C and D). Similarly, HIF-2α knockdown had no impact on PIM1 expression (Supplementary Figure S1E), indicating that PIM1 is regulated in an HIF-1α- and 2α-independent manner.

Hypoxia induces HuR nucleocytoplasmic shuttling

Recent studies revealed that HuR nucleocytoplasmic shuttling promotes chemoresistance in PDA cell lines.²⁵ Once in the cytoplasm, HuR stabilizes its cargo mRNA transcripts, thus resulting in enhanced mRNA translation and protein expression.^{23–25,29,31–33}

To determine if hypoxia-induced chemoresistance is mediated by HuR-mediated stabilization of the PIM1 transcript, MiaPaCa2 cells were subjected to hypoxia for a course of 24 h and HuR cytoplasmic expression was assayed by immunofluorescence and western blot. HuR begins to translocate after 2 h of hypoxic stress, and increases in a time-dependent manner (Figure 2b).

Figure 2.

The mRNA stability factor HuR regulates PIM1 expression in response to hypoxic stress. (a) Immunofluorescence of HuR localization in MiaPaCa2 cells subjected to hypoxia for 24 h (green signal, HuR). Cellular nuclei were stained with Hoechst33258 (blue signal). Magnification x40, scale bar = 10 μM. (b) Western blot analysis of fractionated MiaPaCa2 lysates prepared from cells subjected to hypoxia for the indicated times. Lamin A/C and α-tubulin were used as controls to determine the integrity of the lysates. Densitometric analysis was accomplished using ImageJ. (c) Messenger ribonucleoprotein immunoprecipitation (mRNP-IP) assay was accomplished by isolation of cytoplasmic fractions of MiaPaCa2 cells subjected to hypoxia or normoxia for 24 h. HuR was immunoprecipitated (IP) using a rabbit polyclonal antibody, and IP was validated by a mouse monoclonal antibody through western blot analysis. α-Tubulin was used as a loading control for the input and a negative control for the IP samples. Lamin A/C was used as a control to detect nuclear contamination in the input. The relative abundance of PIM1 mRNA bound to HuR was assayed in triplicates by qPCR using 18S rRNA as a loading control. Results were normalized to IgG isotype controls and represented as means of three independent experiments ± s.e.m. GAPDH (glyceraldehyde 3-phosphate dehydrogenase) was used as a negative control and dCK as a positive control (*P ⩽ 0.05, **P ⩽ 0.01, NS not significant). (d) MiaPaCa2 cells were transfected with control siRNA (siControl) or siRNA against HuR (siHuR), and incubated in normoxia or hypoxia for 24 h. Actinomycin D (5 μg/ml) was added for the indicated times 20 h after hypoxic treatment, after which PIM1 mRNA stability was assayed by qPCR using 18 S rRNA as a loading control. Each value represents an average of triplicates ± s.e.m. (e) MiaPaCa2 cells were transfected with control siRNA (siControl) or siRNA against HuR (siHuR), and incubated in normoxia or hypoxia for 24 h. HuR and PIM1 protein expression was assayed by western blot using α-tubulin as a loading control. (f) MiaPaCa2 cells were co-transfected with a control siRNA or siHuR and luciferase reporter constructs without the PIM1 3′-UTR (LucΔ3′-UTR) or fused to the full-length PIM1 3′-UTR (Luc+PIM1 3′-UTR) and subjected to hypoxia or normoxia for 24 h. Renilla luciferase activity was normalized to firefly luciferase activity and is the average of three experiments ± s.e.m. (***P ⩽ 0.001, NS not significant).

HuR binds to PIM1 mRNA and regulates PIM1 expression

PAR-CLIP and RIP-Chip analyses have identified PIM1 as a predicted target transcript of HuR.^34,35 Messenger ribonucleoprotein immunoprecipitation assays^24,25,32,36 were carried out to determine whether HuR directly binds the PIM1 mRNA. Although a basal level of interaction was detected in normoxia, hypoxic treatment significantly enhanced the interaction between HuR protein and the PIM1 mRNA (Figure 2c, right). Glyceraldehyde 3-phosphate dehydrogenase was used as a negative control. This interaction was further confirmed in pancreatic-specific HuR transgenic overexpressing mice (Sawicki, unpublished and Supplementary Figure S2A).

The mRNA-stabilizing effect of HuR on the PIM1 transcript was determined by mRNA stability assays. mRNA stability was assayed by treating cells with the transcriptional inhibitor actinomycin D for the indicated times followed by qPCR-based analysis of PIM1 expression. HuR knockdown resulted in a 1.8-fold decrease (P ⩽ 0.001) in the half-life of the PIM1 mRNA transcript in normoxia (Figure 2d). Similarly, silencing of HuR in hypoxia decreased PIM1 mRNA stability (1.3-fold, P ⩽ 0.001) (Figure 2d). Importantly, reduced PIM1 mRNA stability in the context of HuR silencing resulted in attenuated PIM1 protein expression in normoxia, and completely abrogated hypoxia-mediated induction of PIM1 protein overexpression, both at the mRNA and protein level (Figure 2e;mRNA left panel, protein right panel).

HuR binds to a discrete sequence within the PIM1 3’-UTR

HuR-mediated mRNA stabilization occurs through direct binding to AREs typically located in UTRs of target transcripts.^34,37 Computational predictions indicated the presence of five putative AREs within the PIM1 3’-UTR (Supplementary Figure S3A). Luciferase assays performed in MiaPaCa2 cells transfected with a reporter construct bearing the full-length PIM1 3’-UTR (Luc+3’-UTR) confirmed the predictions that HuR regulates PIM1 via its 3’-UTR (Figure 2f). This regulatory interaction was observed in the context of HuR silencing in hypoxia (Figure 2f) and forced HuR overexpression in normoxia (Supplementary Figure S3B).

To identify the HuR binding sequence within the 3’-UTR, a deletion series of the PIM1 3’-UTR constructs were each transiently co-transfected with a HuR-overexpressing construct into MiaPaCa2 cells. Deletion of putative site no. 3 (Supplementary Figure S3C) abrogated HuR-mediated stabilization of the luciferase reporter construct, indicating that a predicted 38-bp AU-rich region in the 3’-UTR mediates a regulatory interaction with HuR. Deletions of this binding site resulted in the loss of luciferase activity (Supplementary Figure S3D), indicating that a 38-bp region is the required HuR binding site.

HuR cytoplasmic expression correlates with PIM1 overexpression in clinical PDA specimens

PDA tumors on a tissue microarray were confirmed by histopathology and stained for HuR and PIM1. Based on the scoring distribution, samples were divided into low- and high-expressing tumors. Cytoplasmic HuR expression was high in 73% of the samples, whereas 84% of specimens exhibited high PIM1 expression. A significant association was observed between cytoplasmic HuR localization and PIM1 overexpression (Figures 3a and b, two-tailed P-value 0.011) in hypoxic niches of the tumor (P = 0.0198, Supplementary Table S1).

Figure 3.

PIM1 overexpression correlates with cytoplasmic HuR status in PDA tumors. (a) Immunohistochemical detection of HuR and PIM1 in formalin-fixed, paraffin-embedded (FFPE) PDA tumors. (b) Blinded quantification and distribution of immunoreactivity scores. (c) Hierarchical cluster analysis of changes in total, cleavage or phosphorylation status of cancer-associated proteins by RPPA. (d) Changes in total protein expression, cleavage or phosphorylation status between normoxia and hypoxia were classified by function and represented in a pie chart. (e) Venn diagram demonstrating overlaps between proteins whose expression, cleavage or phosphorylation status were significantly increased in hypoxia, and negated by HuR or PIM1 silencing. (f) Identification of the four overlapping hits identified in (e).

HuR’s regulation of PIM1 promotes PDA cell survival signaling pathways

Given the strong clinical correlation between cytoplasmic HuR localization and PIM1 expression, and that HuR regulates the PIM1 mRNA stability in vitro, we sought to determine how the HuR-PIM1 axis protects cells from hypoxic stress. Immunocytochemistry-based reverse-phase protein array (RPPA)^38,39 determined changes in expression and phosphorylation status of cancer-related proteins. When MiaPaCa2 cells were subjected to hypoxic stress for 24 h, changes occurred in expression, cleavage or phosphorylation in 76 out of 111 proteins assayed (68%). Of these, 52 out of 76 (67%) events were upregulated in hypoxia. In this particular subset, 50% of events were associated with cell survival, 22% were antiapoptotic, 15% were proapoptotic and 13% were directly associated with DNA repair and drug resistance (Figures 3c and d). A complete list of these proteins is available in Supplementary Table S2. Next, we sought to determine whether HuR or PIM1 silencing negated changes in phosphorylation status of these 52 proteins. Supplementary Table S3 depicts targets significantly changed in hypoxia and subsequently negated by HuR silencing, and Supplementary Table S4 depicts targets significantly changed in hypoxia and subsequently negated by PIM1 silencing. A total of four events were identified to be upregulated in hypoxia, and negated by HuR or PIM1 silencing (Figures 3e and f). The most significant event was an increase in phosphorylation of the proapoptotic protein BAD at serine 112, which is then negated by HuR or PIM1 silencing. Previous studies demonstrated that BAD phosphorylation (S112) occurs in a PIM1-dependent manner and serves to deactivate the apoptotic effector.⁴⁰ Although further studies are needed to determine the significance of the other identified hits, these results suggest that hypoxia-mediated induction of PIM1 protects cells from hypoxia-induced apoptosis through phosphorylation and inactivation of key apoptotic effectors (e.g., BAD). Additionally, based on RPPA, the HuR-PIM1 axis phosphorylates MEK1/2⁴¹ (S217, S221), a key downstream signal transducer of oncogenic Ras,⁵ thereby functionally supporting cell survival.

The HuR-PIM1 axis effects on cellular phenotype

To further characterize the significance of the HuR-PIM1 axis with respect to survival, cellular proliferation and survival were analyzed. PIM1 silencing in MiaPaCa2 (Figure 4a) and Capan-1 (Supplementary Figure S4A) cells impacted cell growth in normoxia and, more prominently, in hypoxia. Proliferative index assessment was performed to phenotypically compare PIM1 and HuR knockdown. As with PIM1 knockdown, HuR silencing reduced proliferation of MiaPaCa2 cells. Importantly, HuR silencing was detrimental for cells in hypoxia where cellular proliferation was markedly reduced (Figure 4b).

Figure 4.

HuR-mediated regulation of PIM1 promotes PDA cell survival under hypoxia. (a) MiaPaCa2 cells were transfected with control siRNA or siPIM1 and grown in normoxia or hypoxia for 5 days. PicoGreen analysis of cell survival was performed at the indicated time points. (b) MiaPaCa2 cells were transfected with control siRNA, siHuR or siPIM1 and cultured in normoxia or hypoxia for 24 h. Upon harvest, cells were immunostained with anti-Ki67 antibody to visualize proliferative cells (green signal). Hoechst was used to visualize nuclei. Bar = 10 μM. Cells positive for Ki67 were quantified by determining the average number of Ki67-positive cells per x40 field of view (FOV) for 10 individual fields. Results are represented as an average of the percentage of Ki67-positive cells per x20 FOV ± s.e.m. (c) MiaPaCa2 cells transfected as in (a) were subjected to western blot analysis of HuR, PIM1 and cleaved caspase-3. α-Tubulin was used as a loading control. (d) MiaPaCa2 cells were transfected with control siRNA or siHuR for 48 h and subjected to hypoxia or normoxia for 24 h. HuR, PIM1, total BAD, p-BAD (Ser-112), total MEK and p-MEK (S217/221) were analyzed by western blot using α-tubulin as a loading control.

The deleterious impact of HuR or PIM1 silencing in the setting of hypoxia was confirmed by flow cytometry, where an increase in the sub-G1 population was observed (Supplementary Figure S4B). To validate that this increase was indicative of cellular apoptosis, western blot analysis of cleaved caspase-3 was performed. Caspase-3 cleavage was observed in response to HuR or PIM1 silencing in normoxia, and significantly increased in hypoxia (Figure 4c). As indicated previously⁴⁰ and further validated by our RPPA screen, PIM1 promotes cell survival under severe stress by phosphorylating and inactivating the proapoptotic protein BAD on serine 112. HuR silencing resulted in attenuated PIM1 expression and consequently impacted BAD phosphorylation at serine 112 in the setting of hypoxia, with no impact on total BAD protein expression (Figure 4d). Similarly, the impact of HuR-mediated regulation of PIM1 on pMEK1/2, a newly identified target of PIM1, was confirmed by western blot analysis. These results indicate that the HuR-PIM1 axis is engaged under hypoxia to promote cell survival.

HuR’s regulation of PIM1 promotes hypoxia-induced resistance to chemotherapy

To determine the contribution of HuR and PIM1 to hypoxia-induced chemoresistance, MiaPaCa2 cells were transfected with siRNA against HuR and drug sensitivity assays were performed in normoxia and hypoxia. As shown in Figure 5a, HuR knockdown resulted in increased sensitivity to oxaliplatin in normoxia, and negated hypoxia-mediated oxaliplatin resistance. To determine if HuR drives chemoresistance through a PIM1-dependent mechanism, PIM1 overexpression and rescue experiments were carried out in PDA cells. PIM1 overexpression (Figure 5b) promoted oxaliplatin resistance in cells with intact HuR expression. Whereas HuR silencing alleviated hypoxia-induced chemoresistance, forced overexpression of PIM1 in the context of HuR silencing rescued oxaliplatin resistance (Figure 5c).

Figure 5.

HuR regulates PIM1 to promote hypoxia-induced chemoresistance. (a) PicoGreen-based drug sensitivity assays to oxaliplatin were carried out in MiaPaCa2 cells transfected with siHuR and incubated in normoxia or hypoxia for 72 h. Dose–response curves were fitted based on a sigmoidal dose–response model. Each data point represents the mean of four independent experiments ± s.e.m. IC₅₀ values were derived from dose–response curves. (b) PIM1 expression rescue experiments were carried out in MiaPaCa2 by transiently transfecting siRNAs to HuR (or scrambled control, siControl) and overexpressing His-tagged PIM1 (or control empty vector, pcDNA3.1) for 72 h. (c) PicoGreen-based drug sensitivity assays to oxaliplatin were carried out in MiaPaCa2 cells transfected as in (b). Dose–response curves were fitted based on a sigmoidal dose–response model. Each data point represents the mean of four independent experiments ± s.e.m. IC₅₀ values were derived from dose–response curves.

PIM1 has a role in the DNA damage response through phosphorylation on CDC25 proteins.^25,42,43 Immunoprecipitation studies demonstrated that PIM1’s interaction with the cell cycle regulator CDC25a increases in response to hypoxia (Supplementary Figure S4C). Similarly, oxaliplatin treatment enhances PIM1’s interaction with CDC25a, whereas HuR silencing attenuated this interaction (Supplementary Figure S4D). As a consequence of PIM1-mediated phosphorylation of CDC25a, cells undergo a dynamic reprogramming that allows them to orchestrate a chemoresistant response through enhancement of DNA repair.^42,43 As such, the impact of HuR’s regulation of PIM1 on the DNA damage response was assessed by confocal microscopy-based visualization and quantification of DNA double-stranded break foci. DNA damage foci were visualized and counted using the marker γH2AX, as reported previously.²⁵ First, the baseline number of foci was assessed in cells grown in normoxia or exposed to hypoxia for 24 h, in the presence or absence of oxaliplatin. There was a marked decrease in DNA damage foci in cells exposed to hypoxia in non-treated cells. Although oxaliplatin promoted a significant increase in DNA double-stranded breaks in normoxia, it did not impact cells in hypoxia (Figure 6a).

Figure 6.

HuR-mediated regulation of PIM1 prevents DNA damage induced by oxaliplatin. (a) MiaPaCa2 cells were subjected to normoxia or hypoxia for 24 h, during which cells were treated with 1 μM oxaliplatin for 16 h. Double-stranded DNA breaks were visualized by immunofluorescence analysis of γH2AX (green signal), and blindly quantified using single plane images, and plotted ± s.d. (n = 40, NS, nonsignificant, ***P < 0.001). (b) MiaPaCa2 cells were transiently transfected with His-PIM1 for 48 h. Subsequently, cells were subjected to hypoxia or normoxia for 24 h, during which cells were treated with 1 μM oxaliplatin for 16 h. Double-stranded DNA breaks were visualized by immunofluorescence analysis of γH2AX (green signal), and blindly quantified using single plane images, and plotted ± s.d. (n = 40, NS, nonsignificant, ***P < 0.001). (c) MiaPaCa2 cells were transfected with a control non-targeting siRNA, or siRNA against HuR (siHuR) or PIM1 (siPIM1). Cells were subjected to hypoxia or normoxia for 24 h, during which cells were treated with 1μM oxaliplatin for 16 h. Double-stranded DNA breaks were visualized by immunofluorescence analysis of γH2AX (green signal), and blindly quantified using single plane images, and plotted ± s.d. (n = 40, ***P < 0.001).

Next, the relative contribution of PIM1 overexpression with regard to oxaliplatin-induced DNA damage was determined. To this extent, forced overexpression of PIM1 in normoxia rendered cells resistant to oxaliplatin-induced DNA damage (Figures 6b and c). As both HuR and PIM1 have been shown to have a critical role in DNA repair,^25,42,43 DNA damage in response to oxaliplatin treatment in PDA cells transfected with siRNAs against HuR or PIM1 in normoxia and hypoxia was assessed. The basal level of double-stranded breaks was significantly lower in hypoxia (P ⩽ 0.001) and oxaliplatin treatment did not result in additional damage (i.e., hypoxia-induced chemoresistance). However, HuR or PIM1 silencing restored sensitivity to oxaliplatin as indicated by the overall increase in the number of foci per cell (Figure 6c).

HuR inhibition breaks hypoxia-induced chemoresistance

The small-molecule HuR inhibitor MS-444 (Novartis, Basel, Switzerland)⁴⁴ has been previously identified and characterized with regard to its mechanism of action and HuR binding affinity.⁴⁴ A dose-response study indicated that concentrations as low as 5 μm successfully inhibited HuR cytoplasmic localization in response to hypoxia (Figure 7a, left), thus resulting in marked attenuation of total PIM1 expression (Figure 7a, right). Dose-response studies determined that 5 μm successfully inhibited HuR cytoplasmic localization with no short-term effect on cell survival or viability (Supplementary Figure S4). Therefore, this dose was selected to determine whether MS-444 synergized with oxaliplatin or 5-FU to overcome hypoxia-induced chemoresistance. To this extent, treatment of MiaPaCa2 cells with low-dose (5 μm) MS-444 enhances drug sensitivity in normoxia, and successfully breaks hypoxia-induced chemoresistance to oxaliplatin and 5-FU (Figures 7b and c). Similar results were observed in Panc-1 cells (Supplementary Figure S5B). The chemosensitizing effect of MS-444 was negated by forced PIM1 overexpression (Supplementary Figure S5C).

Figure 7.

Pharmacological HuR inhibition by the small-molecule MS-444 disrupts HuR-mediated stabilization of PIM1 and sensitizes cells to DNA-damaging chemotherapeutics. (a) MiaPaCa2 cells were treated with the indicated doses of MS-444 for 24 h and subjected to hypoxia. Non-treated cells in normoxia were used as controls. Whole-cell and cytoplasmic protein lysates were collected. HuR expression was assayed by western blot analysis in cytoplasmic lysates using Lamin A/C as a negative control. PIM1 expression was assayed by western blot analysis of whole-cell lysates. (b) PicoGreen-based drug sensitivity assays were carried out in MiaPaCa2 cells cotreated with a sublethal dose of MS-444 (5 μM) and increasing doses of oxaliplatin or 5-FU, respectively. Dose–response curves were fitted based on a sigmoidal dose–response model. Each data point represents the mean of four independent experiments ± s.e.m. (c) IC₅₀ values derived from dose–response curves in (b). (d) A schematic model of hypoxia-mediated HuR cytoplasmic localization and subsequent stabilization of the PIM1 mRNA. The HuR-PIM1 axis promotes chemoresistance through potentiation of cell survival and DNA repair mediated by the oncogenic activity of the PIM1 kinase.

DISCUSSION

An immediate and practical therapeutic approach to treating PDA is to improve FDA-approved regimens that have proven activity in recent clinical trials (e.g., FOLFIRINOX^3,4). We recently put forth a concept that key aspects of the tumor microenvironment (i.e., glucose deprivation²⁴ and DNA-damaging agents²⁵) drive an HuR-dependent PDA cell survival mechanism. Herein, we expand on this work and demonstrate a novel hypoxia-induced chemoresistance.

Previous studies demonstrated that human PDAs are hypovascular and exhibit low O₂ partial pressure and intracellular perfusion (i.e., hypoxia).⁸ Subsequent studies demonstrated that tumor-associated hypoxia drives a multifaceted cellular reprogramming within tumor cells, resulting in the activation of prosurvival pathways.^9–11 Moreover, studies demonstrated that intratumoral hypoxia renders cancer cells refractory to ionizing radiation and chemotherapy.^{12,13,15,30,45,46}

The contribution of PIM1 to hypoxia-induced chemoresistance has been identified previously.^17,18 To our knowledge’ only two previous studies provide mechanisms of PIM1 overexpression in PDA: oncogenic Ras activation⁴⁷ and interleukin-6 stimulation.⁴⁸ Although these mechanisms do not account for hypoxia-mediated induction of PIM1, one study identified heat-shock protein 90 as a posttranslational stabilizer of the PIM1 protein.⁴⁹ Still, this study does not account for changes in PIM1 mRNA expression and translation in response to hypoxia. We propose a model (Figure 7d) wherein HuR-mediated stabilization of the PIM1 mRNA is an adaptable, critical step in hypoxia-induced PIM1 expression. Once stabilized, PIM1 phosphorylates downstream targets to inactivate apoptosis, promote cell survival and enhance DNA repair (Figure 7d). Although our RPPA results indicate that these mechanisms may occur primarily through phosphorylation of BAD and MEK1/2, other valid downstream targets have been implicated.^42,43,50,51

In the clinical setting, numerous studies have linked PIM1 overexpression to negative outcomes across multiple malignancies.^16–19 While PIM kinases were primarily thought to have a role in the initiation and progression of hematological malignancies,^52–57 recent reports have extended the oncogenic role of PIM1 to PDA.^17,18,47,58 Although our studies demonstrate a significant correlation between cytoplasmic HuR and PIM1 over-expression in PDA, and previous work by our group showed that cytoplasmic HuR is associated with poor pathologic features,⁵⁹ we were unable to determine the precise clinical significance of this finding in this current data set. Our study was not adequately powered to delineate clinical correlates to determine if the HuR-PIM1 axis is a novel predictor of treatment outcome. However, our findings warrant further investigations to ascertain the validity of cytoplasmic HuR and/or PIM1 overexpression as biomarkers. Moreover, future studies will determine the direct correlation between cytoplasmic HuR expression and PIM1 levels within hypoxic niches within an individual tumor’s microenvironment.

Our results indicate that HuR potentiates a chemoresistant phenotype through stabilization of the PIM1 mRNA. However, the pleiotropic binding of HuR to transcripts with AREs suggests that other targets likely have a role. Our studies indicate that HuR binds to other previously published targets (WEE1, DR5)^25,36 (Supplementary Figure S2B). Although it is difficult to measure exact relative binding affinities from these assays, WEE1 and DR5 appear to not bind to the extent of PIM1 mRNA to HuR under hypoxia (Supplementary Figure S2B).^25,36 These data suggest that HuR may prioritize different targets under different stressors. Determining the significance of HuR’s regulation of each target when induced by different stressors poses many technical challenges, and is part of our ongoing work.^24,25,36,60 Furthermore, further quantitative studies are needed to determine whether HuR’s regulation of PIM1 is important to sustain the angiogenic switch, a phenomenon wherein HuR’s regulation of other classic hypoxia-associated mRNAs (e.g., VEGF, HIF-1α⁶¹) have been shown to be critically important. To this point, the overall prioritization of various targets, and the relative abundance of HuR available to each target, is still unclear.

We demonstrate the translational potential of targeting hypoxia with the specific HuR inhibitor, MS-444,⁴⁴ yet its clinical utility needs to be further assessed through more extensive studies. Careful consideration should be given when developing HuR inhibition strategies⁶² as this protein exerts protective functions.^63–65 In an attempt to overcome the potential toxicity of global pharmacological HuR inhibition, our findings present a novel therapeutic window where HuR could be targeted only in tumor cells where cytoplasmic HuR is abundant.^{27,28,31,32,59,66,67} Ongoing studies will determine the generality of this mechanism in other tumor types^26,28,31,68 where hypoxic-induced chemoresitance likely has a role.^{12,13,15,18,30,45}

MATERIALS AND METHODS

Cell culture

Human PDA cell lines were obtained from ATCC (Manassas, VA, USA) and cultured as recommended. To establish hypoxia, cells were placed in a hypobaric chamber flushed with a gas mixture of 5% CO₂, 94% N₂ and 1% O₂ (Invivo2 300; Baker Ruskin, Sanford, ME, USA).

Drug sensitivity

Oxaliplatin and 5-FU were purchased from Sigma-Aldrich (St Louis, MO, USA). MS-444⁴⁴ was obtained from Novartis Institutes for Biomedical Research and solubilized in dimethyl sulfoxide using an extinction coefficient of 2200/M/cm at 328 nm. Cells were seeded at subconfluent levels and treated with indicated doses for 72–96 h in hypoxia or normoxia. Cell survival was assayed using Quant-it PicoGreen dsDNA Assay Kit (Invitrogen, Grand Island, NY, USA) as described previously.³² A dose-response curve was fitted to the data using the KaleidaGraph 4.5.0 software (Synergy Software, Reading, PA, USA), from which the IC50 was derived.

DNA and siRNA transfections

HuR overexpression was accomplished as described previously.²⁵ The coding region of the 33 kDa isoform of the human PIM1 gene was PCR amplified from MiaPaCa2 cDNA using primers (available upon request) containing an N-terminal polyhistidine (6xHis) tag. His-PIM1 was subcloned into the NheI and XbaI sites of pcDNA3.1 (Invitrogen) and transfected. Transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. HuR knockdown studies were performed using an siRNA oligo as reported previously.²⁵ A commercially available siRNA (catalog no. s10527; Life Technologies, Grand Island, NY, USA) against PIM1 and a control siRNA (catalog no. AM4635; Life Technologies) were used.

Luciferase reporter assays

The full-length PIM1 3’-UTR was subcloned into the XhoI and NotI sites of the psiCheck2 vector (Promega, Madison, WI, USA). A deletion series of the PIM1 3’-UTR, based on HuR binding predictions, was generated and assays were performed using Dual-Luciferase Reporter Assay System (Promega).

mRNA expression and half-life assays

Total RNA was extracted using TRIzol (Invitrogen). cDNA synthesis and qPCR analysis were performed as described previously.⁶⁹ Primer sequences are available upon request. Fold changes in mRNA expression were normalized according to the ΔΔCt method. The PIM1 mRNA half-life (t_½) was determined in MiaPaCa2 cells treated with actinomycin D (5 μg/ml; Fisher Scientific, Fair Lawn, NJ, USA) as described previously.⁶⁹

Western blot analysis

Western blot analyses were performed as described previously.⁶⁷ Subcellular fractionation was accomplished by using the NE-PER Kit (Thermo Fisher Scientific, Rockford, IL, USA). Membranes were probed with antibodies against HuR (5261 clone 3A2; Santa Cruz Biotechnologies, Santa Cruz, CA, USA) at 1:20 000 for 1 h at room temperature, PIM1 (ab75776; Abcam, Cambrige, MA, USA) at 1:1000 for 16 h at 4°C and α-tubulin (32–2500; Invitrogen) at 1:10 000 for 1 h at room temperature. For the analysis of caspase-3 activation by cleavage, rabbit polyclonal caspase-3 (Asp175) antibody (9661; Cell Signaling Technologies, Danvers, MA, USA) was used at 1:1000 at 4°C overnight. All antibodies were incubated in Odyssey blocking buffer (Li-Cor, Lincoln, NE, USA) and detected using IR-conjugated secondary antibodies (IRDye goat anti-mouse 800 or IRDye goat anti-rabbit 800CW; Li-Cor). Membranes were scanned using a Li-Cor IR scanner (Odyssey Infrared Imaging System ODY-1181, Lincoln, NE, USA) and quantified using ImageJ (National Institutes of Health, Bethesda, MD, USA).

Immunohistochemistry

A PDA tumor tissue microarray was constructed using formalin-fixed, paraffin-embedded PDA samples of Institutional Review Board-consented patients at the TJU Hospital (Philadelphia, PA, USA).⁷⁰ Surgical pathologists confirmed the diagnosis for each case and representation of the tumor area in the tissue microarray. Detection of HuR, PIM1, and carbonic anhydrase IX were performed in serial sections (4 μm) using the following antibodies: mouse monoclonal 3A2 HuR antibody (Santa Cruz Biotechnologies) at 1:500, rabbit monoclonal anti-PIM1 antibody (Abcam) at 1:200 and rabbit polyclonal anti-CAIX (Novus Biologicals, Littleton, CO, USA) at 1:1000. Standard staining protocols were used.^32,59

Immunoreactivity scoring

Stained tissues were scored for intensity of staining as described previously⁶⁷ by blinded investigators (FFB, MJ, PM, WJ, JRB). For both PIM1 and HuR, staining intensity was scored on a scale of 0 to 3 (0, negative staining; 1, weak staining; 2, moderate staining; or 3, strong staining). PIM1 was only scored by intensity as it is both nuclear and cytoplasmic. To assess HuR subcellular localization, the percentage of cytoplasmic HuR-positive cells was scored on a scale of 0–4 for the percentage of tissue stained: 0 (0% positive cells), 1 (< 10%), 2 (11% to 50%), 3 (51% to 80%) or 4 (>80%). The two scores were multiplied resulting in an immunoreactivity score value ranging from 0 to 12 as described.⁶⁷

Reverse-phase protein array

RPPAs were constructed and analyzed as described previously.^38,39 Briefly, MiaPaCa2 cells transfected in triplicates with control siRNA, or siRNA against HuR or PIM1, respectively, were subjected to normoxia or hypoxia for 24 h. Replicate sets of cell lysates were printed in triplicate spots on nitrocellulose-coated glass slides using an Aushon 2470 arrayer (Aushon Biosystems, Billerica, MA, USA). Reference standard lysates were used as procedural controls and positive controls for antibody staining as described previously.³⁸ Total protein was estimated per sample by Sypro Ruby Protein Blot Stain (Invitrogen), according to the manufacturer’s protocol. Printed slides were treated with 1x BeBlot Mild Solution (Chemicon, Billerica, MA, USA) for 15 min, washed with phosphate-buffered saline and blocked in 2% I-Block (Applied Biosystems, Grand Island, NY, USA). The arrays were probed with a library of ~200 antibodies against total, cleaved and phosphoprotein end points using an automated slide stainer (Dako, Carpinteria, CA, USA). A complete list of antibodies used is available.³⁸ Primary antibody binding was detected using a biotinylated goat anti-rabbit IgG (1:7500; Vector Laboratories, Burlingame, CA, USA) or rabbit anti-mouse IgG (1:10; Dako), followed by streptavidin-conjugated IRDye680 fluorophore (Li-Cor). Acquired images were analyzed with MicroVigene v.4.0.0.0 (VigeneTech, Carlisle, MA, USA). Unsupervised hierarchical clustering was conducted with JMP v.5.1 (SAS Institute, Cary, NC, USA). Statistical analysis was carried out by multiple means, three-way comparison (analysis of variance; SAS Institute) using a significance cutoff of P ⩽ 0.1 (appropriate for a multiple mean comparison with N =3 per group).

Messenger ribonucleoprotein immunoprecipitation

MiaPaCa2 cells were cultured in normoxia or hypoxia for 24 h and qPCR detection of HuR-bound mRNAs was carried out as reported previously.^24,25,71

Immunofluorescence

Cells were grown on coverslips in 24-well tissue culture plates and immunostained with anti-HuR antibody as described previously.^24,25 Confocal acquisition and foci quantification were carried out as reported.⁶⁹

Cell cycle analysis

Cells were transfected as indicated for 72 h, followed by incubation in hypoxia for 24 h. For the final hour of incubation, cells were pulsed with 10 μm bromodeoxyuridine (GE Healthcare, Pittsburgh, PA, USA) for 60min and harvested in cold 70% ethanol in phosphate-buffered saline. Standard staining procedures were performed.²⁵