The BET bromodomain inhibitor JQ1 suppresses growth of pancreatic ductal adenocarcinoma in patient-derived xenograft models

Clinical characteristics of tumors of origin

The models used in this study (UAB-PA3, -PA4, -PA10, -PA30 and -PA36) were derived from tumors of patients who underwent surgery as a standard of care at the University of Alabama at Birmingham Hospital.²² Models were derived from tumors of patients with stage II PDAC, with three of the five patients having lymph node involvement at the time of surgical resection (Fig. 1A). All primary tumors had mutations in codon 12 of the KRAS gene. These models were chosen for their readily detectable expression of c-Myc by immunohistochemistry (IHC) (Fig. 1B).

PDAC primary tumors express higher levels of c-Myc than normal pancreas.

IHC analysis of FFPE tissues demonstrated minimal-to-high (1+ to 3+) nuclear expression and intermediate-to-high (2+ to 3+) cytoplasmic expression of c-Myc protein in all primary tumors of origin (Fig. 1B). Normal pancreatic (NP) tissue had foci of predominantly cytoplasmic c-Myc in both ductal and acinar tissue. Overall, c-Myc expression was relatively low in NP (expression index 80) compared to primary tumors (expression indices 180-270). IHC analyses demonstrated similar levels of c-Myc in tumorgraft models and their tumors of origin, with the exception of UAB-PA4 tumors which expressed less nuclear c-Myc protein than the tumor from which it was derived.(Fig. 1C). UAB-PA4 also had a lower ratio of nuclear:cytoplasmic (N:C) c-Myc protein, but the impact of this change in ratio of cell phenotype is unclear. Immunoblot analyses confirmed a readily detectable level of c-Myc protein in all models (Fig. 1D). Because the cytotoxicity of JQ1 has been reported to depend on down-regulation of c-Myc,^19,23 we next evaluated the effect of JQ1 on tumor cell morphology, tumor growth in vivo and on c-Myc expression in these tumors.

JQ1 suppressed tumor growth in vivo in preclinical models of PDAC.

To evaluate the efficacy of JQ1, we administered a well-tolerated regimen of JQ1 (50 mg/kg daily x 28 or 21 days) to mice bearing PDAC tumorgrafts. JQ1 inhibited tumor growth by 40-62% compared to vehicle control (VC) in all five independently derived models (Fig. 2A). Model UAB-PA36 was the least responsive to JQ1 treatment with a difference in tumor volume being observed only on the final day of treatment. Time to detection of a JQ1-induced decrease in tumor volume ranged from day 14 (UAB-PA4) to day 28 (UAB-PA36).

We corroborated the effect of JQ1 on tumor volume by immunostaining for the proliferation marker Ki67. Proliferation index was defined as percent of nuclei with detectable Ki67 expression, using photomicrographs imaged using Photoshop CS4 (Figs. 2B, 2C). Ki67 staining of UAB-PA3, -PA4, -PA10, and -PA30 tumors was decreased in JQ1- compared to VC-treated mice, indicating that fewer tumor cells were traversing the cell cycle in the JQ1-treated animals. Consistent with the minimal effect of JQ1 on UAB-PA36 tumors, no difference was seen in the proliferation index of tumors from treated versus control mice bearing this tumor. The data demonstrate that JQ1 inhibited the growth of five PDAC tumor models and decreased the proliferation index in four of the five models of independent origin. The fifth model, UAB-PA36, showed no decrease in proliferation index and was the least sensitive to the cytotoxicity of JQ1 in vivo.

JQ1 induced apoptosis in a subset of models.

Tumors were harvested 24 or 48 hours after termination of treatment, and their histology compared using FFPE sections stained with H&E (Fig. 3A). JQ1 increased the number of apoptotic (white arrows) and necrotic (yellow arrows) tumor cells in the UAB-PA10 and -PA30 models, each of which was responsive to JQ1. JQ1 also induced a slight reduction in peritumoral stroma of tumorgraft models UAB-PA3, -PA4, and -PA10 (white circles) (Fig. 3A), and an increase in stromal hyalinization in UAB-PA3, -PA10 and -PA30 (green circles). Consistent with minimal decrease in tumor growth, UAB-PA36 tumors from JQ1-treated mice had decreases in nuclear atypia and increases in N:C ratio, but no apparent increase in apoptotic or necrotic tumor cells. Of interest, JQ1 appeared to induce tumor cell differentiation in UAB-PA4 tumors (Fig. 3A, 40x image, right panel), with tumor cells having more well preserved nuclear polarity, less nuclear polymorphism and stratification, and a lower N:C ratio than tumor cells from VC-treated mice. This observation is consistent with a previous report showing that JQ1 induced squamous cell differentiation of 797NUT midline carcinoma (NMC) cells.¹⁹ However, UAB-PA4 was the only model that showed differentiation after JQ1 treatment. As JQ1 has been reported to exert its antitumor effect by suppressing c-Myc expression in hematological and solid tumors ^{15, 17, 19, 24-28} we next examined whether suppression of tumor growth was coincident with decreased expression of c-Myc in PDAC tumorgrafts.

JQ1 inhibited expression of c-Myc RNA, but had no consistent effect on c-Myc protein.

We evaluated c-Myc expression in tumorgrafts exposed to JQ1 compared to vehicle control by expression array (NanoString), qRT-PCR, and immunoblot. Firstly, array results failed to demonstrate a difference in c-Myc RNA levels in tumors from JQ1 compared to VC treated mice (data not shown). Secondly, a qRT-PCR method²⁹ and human-specific c-Myc primer set showed that JQ1-treated tumors expressed ~17-30% less c-Myc RNA than VC-treated tumors (Fig. 3B), values that were regarded as minimal but that were statistically significant in four of the five models. The discrepancy between expression array data and qRT-PCR data was attributed to the marginal decrease, if any, in c-Myc RNA induced by JQ1. Thirdly, to determine whether the putative decrease in c-Myc RNA reflected decreased protein expression, we assessed c-Myc by immunoblot and immunohistochemistry. Immunoblots detected no effect of exposure to JQ1 on c-Myc protein, and increased expression of this oncogene in tumorgrafts compared to normal pancreas (Fig. 3C). IHC detected a decrease (~1.8-fold) in c-Myc protein only in one model (UAB-PA10) (Figs. 3D, 3E). The other four models showed <10% nuclear immunoreactivity and expression indices of 100-200 with moderate staining intensity of tumors from both JQ1- and VC-treated mice. We concluded that JQ1 had little effect on levels of c-Myc protein in these PDAC models. Further, although JQ1 is thought to inhibit cell growth by inhibiting BRD4 function rather than expression, we verified that BRD4 protein levels were also unaffected by exposure to JQ1. IHC data detected no consistent changes in level of expression or subcellular localization of c-Myc or BRD4 protein (Figs. 3D, 3E).

The minimal decrease in RNA encoding c-Myc suggested by qRT-PCR data was not seen in the NanoString assay or by immunoblots, as mentioned above. However, to rule out an effect of JQ1 on c-Myc function, we evaluated the effect of JQ1 on transcriptional targets of c-Myc, and found that only one of nineteen gene products that have been reported to be transcriptional targets of c-Myc was decreased in all five models, following exposure to JQ1 (Fig. 4A). This observation suggests that c-Myc function was unaffected and is unlikely to be critical to the cytotoxic mechanism of JQ1 in PDAC tumors. Of note, the list of putative transcriptional targets of c-Myc in Fig. 4A was compiled from multiple references that demonstrate the likelihood of regulation by c-Myc in various types of cells and tissues. ^30-48 The genes listed do not necessarily represent transcriptional targets of c-Myc in PDAC tumors. The single exception to this observation may be CDC25B, and is discussed below. The regulatory region of CDC25B has been reported to have a single Myc/Max binding site,⁴⁷ and is regarded as a potential c-Myc transcriptional target.

The data suggest that inhibition of growth of PDAC tumors may be independent of down-regulation of c-Myc, as has been suggested by findings using ex vivo systems, higher doses of JQ1 (100-500 nM, 2.5 μM), and lung adenocarcinoma and glioblastoma cells.^{21, 49} To identify alternative gene products whose up- or down-regulation might be affected by BRD4 inhibition and that might contribute to the mechanism of cytotoxicity of JQ1, we ranked the NanoString data to identify the 30 genes whose expression was decreased to the greatest extent, following exposure to JQ1. We also investigated the effect of exposure of JQ1 on CDC25B expression.

JQ1 altered the expression of nine gene products in all five models.

The 30 gene products most-affected by JQ1, as evaluated by NanoString, are listed in Supplementary data (Table. S1). The values in the Table represent the average change among the five models, but do not necessarily reflect similar changes in each of the five models. A Partek Genomic Suite program identified six RNA species commonly up-regulated and three RNA species commonly down-regulated in the five models (Figs. 4B, 4C). Fig. 4B shows a heatmap of these nine genes (blue = down-regulation; red = up-regulation, details in Methods). Exposure to JQ1 increased expression of PTEN, CASP2, MSH6, BRAF, NOTCH11, ERCC4 by 1.1- to 1.8-fold (P = 0.026, 0.028, 0.033, 0.035, 0.036, and 0.048, respectively), and decreased expression of LMO2, TIMP3, and CDC25B by 1.3- to 1.4-fold (P = 0.026, 0.032, and 0.048, respectively) (Fig. 4C). The change in expression of each of these RNAs was less than 2-fold, and none appears in the list of most-affected genes. The potential involvement of CDC25B was investigated further based on the following considerations. Heatmap results indicated a uniform down-regulation of CDC25B in all five PDAC tumorgraft models. CDC25B has been reported to be an oncogene ⁵⁰ and to be overexpressed in PDAC tumors,⁵¹ and targeted CDC25B inhibitors inhibit the growth of PC cell lines.⁵¹ CDC25B (cell division cycle 25B) belongs to a family of cell cycle phosphatases (CDC25A, CDC25B, and CDC25C) that regulate cell cycle progression by dephosphorylation of cyclin dependent kinases (CDK), thereby activating CDK/cylin complexes and stimulating cell proliferation. CDC25B accumulates in the nucleus during the G2 phase of the cell cycle to initiate entry into mitosis.^52,53

Expression of CDC25B, a member of the CDC25 phosphatase family.

qRT-PCR analyses confirmed a JQ1-dependent reduction in expression of CDC25B RNA in three models (Fig. 5A; p<0.05); but as was seen with c-Myc expression the degree of decrease was less than 2-fold. Immunoblot data demonstrated that JQ1 decreased CDC25B protein expression by 33% (UAB-PA3) to 99% (UAB-PA4) in the four tumorgraft models most sensitive to JQ1 (Fig. 5B). Expression indices demonstrated that exposure to JQ1 decreased expression of nuclear CDC25B from 33% (UAB-PA10) to 83% (UAB-PA4) (Fig. 5C). No decrease in CDC25B protein was observed in the model least sensitive to JQ1 (UAB-PA36).

Further, we demonstrated a dose response relationship between concentration of JQ1, CDC25B protein expression, and inhibition of cell growth using an in vitro model. We exposed BxPC3 pancreatic cancer cells to JQ1 to determine the IC₅₀ and to evaluate the level of expression of CDC25B protein over a range of JQ1 concentrations (10⁻⁷ to 10⁻⁴ M) (Fig. 6). Growth inhibition data (Fig. 6A) demonstrate an IC₅₀ of 3.5 × 10⁻⁶ M. Immunoblot data (Fig. 6B) demonstrate a ~34-94% decrease in CDC25B protein at concentrations of JQ1 (M) of 1.0 × 10⁻⁶ and 1.0 × 10⁻⁵. The data demonstrate a simultaneous inhibition of cell growth and decrease in CDC25B expression.

In summary, data from in vivo models demonstrate that exposure to JQ1 inhibited CDC25B expression in tumor models most sensitive to this agent. In vitro data demonstrated dose-dependent decreases in BxPC3 cell viability and CDC25B protein by JQ1. The data also suggest that decreased expression of c-Myc is unlikely to be a primary mechanism by which JQ1 inhibits PDAC cell growth, as we observed no consistent change in c-Myc RNA or protein expression, or function as assessed by the level of expression of transcriptional targets of this transcription factor. In vivo data demonstrating JQ1-mediated inhibition of tumor progression indicate that bromodomain inhibitors merit further investigation for treatment of PDAC tumors as single agents and in combination.

Ethics statement

This study included human subjects and vertebrate animals. All protocols and procedures were approved by the University of Alabama at Birmingham Institutional Review Board (IRB) or the University of Alabama at Birmingham Institutional Animal Care and Use Committee (IACUC).

Patient-derived xenograft (tumorgraft) models

Female SCID CB 17^−/− mice, 4-6 weeks old (Taconic Farms, Newtown, MA, USA) were housed under barrier conditions with a 12-hour light/dark cycle and ad libitum access to food and water. Methods for development, transplantation, and characterization of tumorgrafts generated directly from surgical specimens were reported previously.²²

Treatment of tumor-bearing mice

We prepared JQ1 according to published procedures.¹⁹ When transplanted tumors reached ~100-200 mm3 in size, we randomized mice (8 or 10 tumors/group), and administered 50 mg/kg JQ1 or vehicle control (VC) intraperitoneally (i.p.) daily for 21 (UAB-PA4) or 28 (UAB-PA3, -PA10, -PA30, and -PA36) days. We measured tumor size with Vernier calipers (Fowler/Slyvac, Newton, MA, USA) twice weekly, and calculated tumor volume assuming a perfect sphere and the equation v = (π/6)*d3. Twenty-four (UAB-PA3, -PA4, -PA30, -PA36) or forty-eight (UAB-PA10) hours after completion of treatment, mice were euthanized; and tumors were fixed in formalin and embedded in paraffin (FFPE), or frozen in liquid nitrogen and stored at -80°C.

Immunoblots

Frozen specimens were homogenized using a liquid nitrogen-cooled biopulverizer (MidSci, St. Louis, MO, USA) and placed in cell lysis buffer (Cell Signaling Technology, Danvers, MA, USA) containing protease inhibitors (Complete Mini, Roche Diagnostics, Indianapolis, IN, USA). SDS-PAGE analysis of lysates was done using standard methods. Primary antibodies were: anti-human c-Myc (AHO0062; Life Technologies, Grand Island, NY, USA), CDC25B (ab70927; abcam, Cambridge, MA, USA or Cell Signaling; 9525), and β-actin (A5316; Sigma-Aldrich, St. Louis, MO, USA). Blots were quantitated using Adobe Photoshop CS5 (San Jose, CA, USA) or Image Studio Lite (Li-Cor Biotechnology; Lincoln, NE, USA).

NanoString nCounter® RNA expression analysis

Total RNA was isolated from frozen specimens by Trizol-choloroform. RNA samples (400 ng) were analyzed in the UAB Nanostring Laboratory http://www.uab.edu/medicine/radonc/en/nanostring) using an nCounter® Analysis System (NanoString Technologies, Seattle, WA, USA.⁶⁵ Briefly, two human sequence-specific capture probes comprised of a human sequence-specific oligonucleotide plus a short common sequence linked to biotin were constructed. RNA + capture probe + fluorescence-labeled reporter probes complementary to capture probes were hybridized under conditions facilitating formation of triplexes (mRNA + capture probe + reporter probe). Excess probe and partial complexes were removed by affinity purification. Probe/mRNA complexes were loaded into streptavidin coated cartridges, to immobilize triplexes on the cartridge. Current applied to the cartridge aligned and extended the complexes to facilitate imaging. An nCounter Digital analyzer acquired digital images of the fluorescence-labeled target mRNA affixed to the cartridge using a CCD camera and microscope objective. Digital images of 230 cancer related genes (Cancer Reference Panel) were analyzed to detect unique “barcodes” corresponding to the gene of interest. Expression level was quantitated by comparison with six reference genes.

Data analysis was conducted with nSolver v1.1 software (NanoString Technologies, Seattle, WA, USA), after normalizing for variations in binding efficiency, hybridization, and purification, using spiked-in positive controls and normalizing background values using negative control probes. Data are reported as fold-difference in expression between JQ1 and VC treated tumors.

Histological analysis: hematoxylin and eosin (H&E) staining

Fresh tumor tissue was placed for 24 hours in 10% neutral buffered formalin (Fisher Scientific, Suwanee, GA, USA) immediately after harvest, and then embedded in paraffin. Thin sections (5 μm) were prepared by the Comparative Pathology Laboratory at UAB (Birmingham, AL, USA). H&E staining was performed as previously described.²²

Immunohistochemistry (IHC)

Slides containing thin sections of tumor specimens were baked at 60°C, deparaffinized in xylene, and rehydrated using a decreasing ethanol gradient. After antigen retrieval, endogenous peroxidase activity was blocked with 3% hydrogen peroxide in methanol. Non-specific protein binding was minimized by incubation in 10% horse serum (Vector Laboratories, Burlington, CA, USA), and primary antibody binding was accomplished by overnight incubation. Antibody binding was detected using IMMpress secondary reagent (Vector Laboratories, Burlington, CA, USA) and DAB High Contrast chromogen (Scytek Laboratories, Logan, UT, USA). Tissues were counterstained with Harris Hematoxylin (Fisher Scientific, Suwanee, GA, USA). Photomicrographs were taken using an Olympus BH-2 microscope with DP70 camera and DPS-BSW v3.1 software (Center Valley, PA, USA). Proliferation Indices⁶⁶: 10 random high power fields (HPF = 400x) were taken, and positive nuclei counted and divided by the total number of nuclei in two independent experiments. Expression index = % of positive cells x staining intensity (1+ 2+ or 3+).⁶⁷

Real-time RT-PCR (qRT-PCR)

Total RNA (2 μg) was used as the template for cDNA synthesis, using the Superscript® III synthesis kit (Invitrogen, Carlsbad, CA, USA). 2x SsoFast™ EvaGreen® cocktail (Bio-Rad Laboratories, Hercules, CA, USA) was diluted with nuclease free water and combined with cDNA and appropriate primers. c-MYC Forward: 5’-CGACTCTGAGGAGGAACAAG-3’, c-MYC Reverse: 5’GTGATCCAGACTCTGACCTT-3’, GAPDH Forward: 5’AACATCATCCCTGCTTCCAC-3’, GAPDH Reverse: 5’-GACCACCTGGTCCTCAGTGT-3’; CDC25B Forward: 5’- GTGATCCCAGGTGACGACAT-3’; CDC25B Reverse: 5’-TTTGATTCCACACCCTCGCAC-3’. Reaction conditions were: denaturation at 94°C for 1 min; 40 cycles of 15 sec at 94°C, 10 sec at 50°C; extension at 72°C for 10 sec. Reactions were done using a CFX96 System and analyzed using Bio-Rad CFX manager software v1.5 (Bio-Rad Laboratories, Hercules, CA, USA).

In vitro cell viability assays

Cell viability assays were done using standard methods.⁶⁸ Briefly, cells were plated in 96-well plates at 1.5 × 10³ cells per well. After 24 hours, cells were treated with either DMSO (1%) or 10 fold serial dilutions of JQ1 for 72 hours. Cell viability was assessed by standard alamarBlue (Fisher Scientific, Suwanee, GA, USA) protocols. Fluorescence was measured at 590 nm using a microplate reader (Molecular Devices, LLC. Sunnyvale, CA, USA) to calculate IC₅₀s (GraphPad Prism software).

Bioinformatics

Raw values from NanoString were normalized using Partek Genomic Suite (PGS, St. Louis, MI, USA) to house-keeping genes prior to paired group comparison (JQ1 versus vehicle control). The log2-transformed fold-difference between JQ1-treated and VC-treated mice was averaged across all 5 models for each gene and expression ranked from lowest to highest. Data were imported into GENE-E (version 3.0.214; Broad Institute, Cambridge, MA, USA) to generate a heatmap. The color intensity on the heat map reflects global expression with a minimum (25%) in blue and a maximum (75%) in red. Cluster analysis, using one-minus Pearson correlation of rows only and distance for the clustering, was calculated using an average linkage algorithm.⁶⁹

Statistics

Tumor volumes were compared by Two-Way-ANOVA and all values are presented as mean ± SEM. RNA expression (a minimum of three independent experiments for qRT-PCR) was compared by 2-tailed t test using GraphPad Prism 5.0 software (San Diego, CA, USA). P < 0.05 was considered significant.