GLI2-dependent c-MYC upregulation mediates resistance of pancreatic cancer cells to the BET bromodomain inhibitor JQ1.

abstract CC BY

JQ1 and I-BET151 are selective inhibitors of BET bromodomain proteins that have efficacy against a number of different cancers. Since the effectiveness of targeted therapies is often limited by development of resistance, we examined whether it was possible for cancer cells to develop resistance to the BET inhibitor JQ1. Here we show that pancreatic cancer cells developing resistance to JQ1 demonstrate cross-resistance to I-BET151 and insensitivity to BRD4 downregulation. The resistant cells maintain expression of c-MYC, increase expression of JQ1-target genes FOSL1 and HMGA2, and demonstrate evidence of epithelial-mesenchymal transition (EMT). However, reverting EMT fails to sensitize the resistant cells to JQ1 treatment. Importantly, the JQ1-resistant cells remain dependent on c-MYC that now becomes co-regulated by high levels of GLI2. Furthermore, downregulating GLI2 re-sensitizes the resistant cells to JQ1. Overall, these results identify a mechanism by which cancer cells develop resistance to BET inhibitors.

jq1-resistant pancreatic cancer cells are resistant to brd4 knockdown and demons CC BY

Recently, we demonstrated that BET inhibitors are effective against pancreatic cancer cells growing in three-dimensional collagen (Fig. 1a)13. Since cancer cells can eventually develop resistance to therapeutic agents18, we treated CD18 pancreatic cancer cells with increasing concentrations of JQ1 over a prolonged period of time to generate CD18 cells resistant to JQ1 (CD18-JQ1R). These cells, in contrast to parental CD18 cells (CD18-P), continued to grow in 3D collagen in the presence of increasing concentrations of JQ1 (Fig. 1a). Significantly, CD18-JQ1R cells were also resistant to the structurally-related BET inhibitor I-BET151 (Supplementary Fig. S1). Since the effects of JQ1 in CD18 cells are primarily mediated by inhibition of BRD413, we examined whether there was increased expression of BRD4 protein in CD18-JQ1R. The levels of BRD4 protein were in fact lower in CD18-JQ1R cells (Fig. 1b). Moreover, while BRD4 siRNA repressed growth of CD18-P cells, CD18-JQ1R cells were resistant to the effects of BRD4 knockdown (Fig. 1b). We next examined the levels of JQ1-target genes in CD18-P and CD18-JQ1R cells. As shown previously13, expression of c-MYC, FOSL1 and HMGA2 was repressed following acute treatment of CD18-P cells with JQ1 (Fig. 1c). In contrast, CD18-JQ1R cells treated continuously with JQ1 demonstrate minimally decreased levels of c-MYC and in fact have increased FOSL1 and HMGA2 levels (Fig. 1d).

jq1-resistant pancreatic cancer cells demonstrate decreased cell-cell and cell-m CC BY

While CD18-P cells grow in cohesive clumps, CD18-JQ1R cells grow primarily as single cells and failed to express E-cadherin (Fig. 2a), suggesting that JQ1 resistance was accompanied by an epithelial-mesenchymal transition (EMT). Although there was no significant change in the levels of Snail and Slug, both regulators of EMT2122, in CD18-JQ1R cells compared to CD18-P cells (data not shown), CD18-JQ1R cells express increased ZEB1 mRNA and protein levels (Fig. 2a). The CD18-JQ1R cells also express significantly lower levels of miR-200a and miR-200c (data not shown), microRNAs that target ZEB123. As activation of integrin signaling can contribute to resistance to anti-tumor therapies24, we examined whether there was increased integrin activation and cell adhesion to collagen in CD18-JQ1R cells compared to CD18-P cells. However, CD18-JQ1R cells have in fact decreased adhesion to collagen matrix (Supplementary Fig. S2a). CD18-JQ1R cells also showed reduced cell-surface expression of the collagen-binding α2-integrin without significantly affecting cell-surface levels of β1-integrin (Supplementary Fig. S2b). Moreover, CD18-JQ1R cells have significant reduction in FAK phosphorylation compared to CD18-P cells (Supplementary Fig. S2c). These results indicate that CD18-JQ1R cells in fact have decreased signaling downstream of collagen-binding integrins. Since it was shown that reverting EMT could restore sensitivity of lung cancer cells to targeted therapy20, we examined the effect of ZEB1 siRNA on the response of CD18-JQ1R cells to JQ1. Transfection with ZEB1 siRNA restored E-cadherin levels and enhanced cell-cell adhesion (Fig. 2b). ZEB1 siRNA also restored adhesion of CD18-JQ1R cells to collagen matrix and increased expression of α2-integrin (Supplementary Fig. S2d, e). However, despite reversing these aspects of the EMT, ZEB1 siRNA failed to re-sensitize CD18-JQ1R cells to JQ1 treatment (Fig. 2c). Moreover, ZEB1 siRNA had minimal effects on c-MYC, FOSL1 and HMGA2 mRNA and protein levels in CD18-JQ1R cells (Fig. 2d).

jq1-resistant pancreatic cancer cells demonstrate increased gli2 expression CC BY

In order to characterize the resistance mechanism, we evaluated the extent to which CD18-JQ1R cells remained dependent on c-MYC, FOSL1 or HMGA2 for growth in 3D collagen using siRNA to knockdown each protein. CD18-JQ1R cells require c-MYC, but not FOSL1 or HMGA2, for growth in 3D collagen (Fig. 3a). Thus, we screened additional JQ1-independent pathways that may regulate c-MYC levels in CD18-JQ1R cells. The hedgehog (Hh) signaling pathway, which is both an early and a late mediator of pancreatic cancer tumorigenesis2526, can regulate c-MYC expression27. Initially, we evaluated whether there was alteration in Hh signaling by examining relative expression of Patched1, Patched2, GLI1 and GLI2 in CD18-P and CD18-JQ1R cells. As CD18-P and CD18-JQ1R cells do not express GLI1 or Patched2 mRNA (data not shown), we compared the relative levels of Patched1 and GLI2 in CD18-P and CD18-JQ1R cells. Patched1 levels were modestly higher in CD18-JQ1R than in CD18-P cells; however, GLI2 mRNA was ~7-fold higher in CD18-JQ1R than in CD18-P cells (Fig. 3b). There was also increased expression of GLI2 protein in CD18-JQ1R cells (Fig. 3b). Since the GLI2 promoter has SMAD and lymphoid enhancer factor/T cell factor binding sites28, we examined the role of Smad and β-catenin signaling in mediating GLI2 expression in CD18-JQ1R cells. Down-regulating β-catenin, but not Smad4, attenuated GLI2 expression in CD18-JQ1R cells (Fig. 3c). We next evaluated the extent to which GLI2 regulated c-MYC levels in CD18-JQ1R cells. Initially, we conducted chromatin immunoprecipitation (ChIP) to evaluate whether GLI2 bound to c-MYC promoter in CD18-JQ1R cells. ChIP with anti-GLI2 antibody demonstrated binding of GLI2 to the c-MYC promoter in CD18-JQ1R cells, but not in CD18-P cells (Fig. 3d). Additionally, GLI2 knockdown resulted in ~50% reduction in c-MYC levels in CD18-JQ1R cells (Fig. 3e); however, GLI2 siRNA failed to block growth of CD18-JQ1R cells in 3D collagen. These results indicate that partial repression of c-MYC by GLI2 siRNA may not be sufficient to affect growth of CD18-JQ1R cells.

targeting gli2 restores sensitivity of jq1-resistant cells to bet inhibitors CC BY

We next examined whether combination of JQ1 and GLI2 siRNA would result in further repression of c-MYC and thereby limit growth of CD18-JQ1R cells in 3D collagen. In agreement with a recent paper demonstrating that GLI2 is a target of JQ1 in medulloblastoma cells29, treatment of CD18-JQ1R cells with JQ1 resulted in partial repression of GLI2. Although GLI2 siRNA alone repressed c-MYC mRNA by ~40%, combination of JQ1 and GLI2 siRNA resulted in ~80% inhibition of c-MYC mRNA levels starting at the 0.125 μM JQ1 (Fig. 4a). JQ1 also enhanced the repressive effect of siGLI2 on c-MYC protein levels in CD18-JQ1R cells (Fig. 4b). Additionally, co-transfection of BRD4 siRNA with GLI2 siRNA led to ~75% inhibition of c-MYC mRNA levels in CD18-JQ1R (Fig. 4c). These results indicate that the c-MYC levels in CD18-JQ1R cells are now co-regulated by both BRD4 and GLI2. Significantly, GLI2 siRNA restored sensitivity of CD18-JQ1R cells to growth inhibition by JQ1 (Fig. 4d).

discussion CC BY

Since epigenetic changes have been implicated in every aspect of cancer development and progression, including response to therapy and recurrence3031, there is an increasing interest in targeting these changes in solid tumors and blood cancers32333435363738. We have previously reported that human pancreatic tumors demonstrate increased histone acetylation in areas of fibrosis and that targeting histone acetyltransferases can sensitize pancreatic cancer cells to chemotherapy16. We recently reported that targeting readers of acetylation with BET inhibitors can also limit the growth of cancer cells in the collagen microenvironment13. In this report, we show that pancreatic cancer cells that develop resistance to BET inhibitors upregulate GLI2 and that targeting GLI2 re-sensitizes the pancreatic cancer cells to BET inhibitors. Significantly, GLI levels are modulated in many forms of resistant cancer. For example, GLI levels are increased in cancer cells isolated from chemo-resistant ovarian cancer cells compared to matched primary tumors39. Accordingly, inhibition of GLI2 re-sensitizes ovarian tumors obtained from recurrent platinum-resistant patients to cisplatin3940, and also reverses taxane resistance in ovarian cancer41. Mechanistically, GLI proteins can upregulate not only BCL24243, a key molecule involved in the suppression of the intrinsic apoptotic pathway, but also mediate resistance to extrinsic apoptotic pathway by increasing cFLIP expression44. In JQ1-resistant cells, we found that GLI2 along with BRD4 maintains expression of c-MYC, which is required for continued growth of these cells in the collagen microenvironment. In support of our findings, ectopic GLI2 expression was recently demonstrated to rescue growth inhibition by JQ1 in Hh-driven medulloblastoma cells29, suggesting the presence of GLI2-responsive promoters that do not require BRD4 for transactivation. However, our results suggest that c-MYC promoter is dependent on both GLI2 and BRD4 for transactivation in the JQ1-resistant cells. We also show that the JQ1-resistant cells demonstrate evidence of EMT, which is also associated with resistance to therapy in pancreatic cancer cells454647. Expression profiling of chemo-resistant cell lines have shown a strong association between EMT and chemotherapy resistance47. The EMT transcription factor ZEB1 is upregulated in chemotherapy resistant cell lines and silencing ZEB1 with siRNA reverses EMT and restores chemo-sensitivity47. EMT also plays a role in modulating resistance to targeted biologic therapies as well. Cells that express high levels of Snail or ZEB1 demonstrate significantly decreased growth inhibition in response to treatment with the EGFR inhibitor erlotinib compared to cells with an epithelial phenotype48. Despite the importance of ZEB1 and EMT in mediating resistance to chemotherapy and targeted therapies, we show that reversal of EMT through downregulation of ZEB1 failed to sensitize the JQ1-resistant cells to BET inhibitors. Integrin signaling can also mediate resistance to chemotherapy and targeted therapies. Increased adhesion to ECM and activation of β1-integrin inhibits apoptosis in response to chemotherapy in breast and lung cancer cells2449. Increased integrin activation also mediates resistance to anti-HER2 targeted therapy in breast cancer cells and to EGFR inhibitors in lung cancer cells5051. We have previously shown that 3D collagen I contributes to resistance to gemcitabine in pancreatic cancer cells17. Others showed that increased adhesion to laminin and collagen IV also decreases sensitivity of pancreatic cancer cells to gemcitabine through FAK activation52. Despite the importance of integrin signaling in mediating resistance to chemotherapy and targeted therapies, we found that JQ1-resistant cells have decreased integrin expression and signaling, indicating that this mechanism was not the cause of drug resistance. Overall, we demonstrate that cancer cells developing resistance to BET bromodomain inhibitors demonstrate EMT, decreased integrin signaling, and activation of the hedgehog pathway (Fig. 4e, Model). Although EMT does not mediate resistance to BET inhibitors, blocking the expression of GLI2 restores JQ1 sensitivity. Our findings demonstrate that a better understanding of the mechanism of resistance to BET inhibitors may allow for identification of additional therapeutic targets that extend the efficacy of BET inhibitors.

reagents CC BY

Antibodies against c-MYC, FOSL1 and GLI2 were purchased from Cell Signaling, HMGA2 antibody was from Biocheck Inc, while vimentin and BRD4 antibodies were from Abcam. ZEB1, α-tubulin and total FAK antibodies were obtained from Santa Cruz, while E-cadherin antibody was from Invitrogen. pFAK(Y397) antibody was from BD Transduction laboratories, while GAPDH and α2- and β1-integrin antibodies were from Millipore. Secondary antibodies were purchased from Sigma. The EZ-Chip and EZ-Zyme Chromatin Prep kits were from Millipore. The anti-GLI2 rabbit antibody for ChIP assay was purchased from Abcam, while the control IgG rabbit antibody was from Cell Signaling. BET inhibitor JQ1 was obtained from BPS Bioscience, while I-BET151 was acquired from Tocris Bioscience. BRD4, c-MYC, FOSL1, ZEB1 and GLI2 siRNAs were purchased from Life Technologies.

cell culture CC BY

CD18/HPAF-II cells were obtained from American Type Culture Collection (ATCC; Manassas, VA) in 2008. Cells were maintained in DMEM containing 10% FBS and antibiotics (100 U/ml Penicillin and 100 µg/ml Streptomycin). JQ1-resistant CD18 (CD18-JQ1R) cells were generated by treating parental CD18 (CD18-P) cells with increasing concentrations of JQ1 over a period of 3 months. The surviving cells were maintained in 2.5 μM JQ1. The CD18-P and CD18-JQ1R cells were authenticated by STR profiling at the Johns Hopkins Genetic Resources Core Facility in October, 2013.

transfection CC BY

Cells were transfected with siRNA against BRD4, c-MYC, FOSL1, GLI2, HMGA2, ZEB1 or control siRNA using Lipofectamine RNAimax (Invitrogen) according to manufacturer's instructions before plating into collagen.

embedding and examination of cells in three-dimensional type i collagen gels CC BY

Collagen mixture (2 mg/mL) was made by adding the appropriate volumes of sterile water, 10X DMEM and NaOH and kept on ice until needed53. Cells were then suspended in the collagen solution and allowed to gel at 37°C. For RNA extraction, the gel containing cells was processed using RNeasy extraction kit (Qiagen) and then processed for qRT-PCR analysis. For morphological examination of cells, cell colonies in three-dimensional collagen were examined using a Zeiss Axiovert 40 CFL microscope and pictures taken with a Nikon Coolpix 4500 camera53. The relative size of individual colonies was measured using Photoshop.

quantitative real time-pcr analysis CC BY

Quantitative mRNA and microRNA expression was performed with gene specific Taqman probes, TaqMan Universal PCR Master Mix and the 7500 Fast Real-time PCR System from Applied Biosystems. The data were then quantified with the comparative CT method for relative gene expression.

immunoblotting CC BY

Immunoblotting for ZEB1, E-cadherin, vimentin, BRD4, c-MYC, FOSL1, HMGA2, GLI2, GAPDH and α-tubulin was done as previously described53.

adhesion assay CC BY

Equal numbers of CD18-P and CD18-JQ1R cells were seeded onto tissue culture plates pre-coated with type I collagen. Cells were allowed to adhere at 37 °C for 10 minutes, washed once with PBS, and then fixed and stained54. Cells were imaged using a Zeiss microscope and photographed using a Nikon camera.

flow cytometric analysis CC BY

CD18-P and CD18-JQ1R cells in suspension were incubated with anti α2- or β1-integrin antibody and then stained with secondary antibody conjugated to Alexa Fluor 488 prior to analysis with Summit Software 4.3 on a Beckman Coulter fluorescence-activated cell sorter55.

immunofluorescence CC BY

CD18-P and CD18-JQ1R cells plated overnight onto collagen-coated glass coverslips were fixed and incubated with E-cadherin antibody followed by secondary antibody conjugated to Alexa Fluor 488. The cells were stained with DAPI after which the cells were washed, mounted, and observed using a Zeiss Axiovert 200 microscope.

chromatin immunoprecipitation CC BY

CD18-P and CD18-JQ1R cells were treated with formaldehyde to create DNA-protein cross-links. Chromatin fragments were prepared using EZ-Zyme Chromatin Prep kit and then ChIP performed using the EZ ChIP kit and anti-GLI2 antibody or control IgG antibody. Purified DNA was then analyzed by PCR using primers specific for c-MYC locus56, and the PCR products visualized on a 2%-agarose gel.

statistical analysis CC BY

All statistical analyses were done using GraphPad Instat using a two-tailed t-test analysis. Error bars represent standard deviation.