Concerted cell and in vivo screen for pancreatic ductal adenocarcinoma (PDA) chemotherapeutics

PDA is a major cause of US cancer-related deaths. Oncogenic Kras presents in 90% of human PDAs. Kras mutations occur early in pre-neoplastic lesions but are insufficient to cause PDA. Other contributing factors early in disease progression include chronic pancreatitis, alterations in epigenetic regulators, and tumor suppressor gene mutation. GPCRs activate heterotrimeric G-proteins that stimulate intracellular calcium and oncogenic Kras signaling, thereby promoting pancreatitis and progression to PDA. By contrast, Rgs proteins inhibit Gi/q-coupled GPCRs to negatively regulate PDA progression. Rgs16::GFP is expressed in response to caerulein-induced acinar cell dedifferentiation, early neoplasia, and throughout PDA progression. In genetically engineered mouse models of PDA, Rgs16::GFP is useful for pre-clinical rapid in vivo validation of novel chemotherapeutics targeting early lesions in patients following successful resection or at high risk for progressing to PDA. Cultured primary PDA cells express Rgs16::GFP in response to cytotoxic drugs. A histone deacetylase inhibitor, TSA, stimulated Rgs16::GFP expression in PDA primary cells, potentiated gemcitabine and JQ1 cytotoxicity in cell culture, and Gem + TSA + JQ1 inhibited tumor initiation and progression in vivo. Here we establish the use of Rgs16::GFP expression for testing drug combinations in cell culture and validation of best candidates in our rapid in vivo screen.


Results
Alterations in HDAC activity occur in numerous cancers and have prompted the search for pharmacological agents capable of inhibiting these enzymes 24,25 . Several studies have reported elevated expression of HDACs and BETs in PDA. HDAC1, 2, 3, 4, and 7 were reported to be upregulated in PDA, whereas HDAC 2 and 3, along with SIRT1, have been reported to be involved in cancer invasion and chemo-resistance [31][32][33][34][35] . Thus, we assessed the differential expression profile of all HDACs and BETs in human PDA tissue samples in the TCGA database and compared these to mouse models of caerulein-treated pancreatitis, PDA (KIC), and primary PDA cells from KIC mice.  www.nature.com/scientificreports/ HDACs and BET proteins are highly expressed in human and mouse PDA. We analyzed the differential expression of HDACs and BETs at various stages of disease progression in mice. First, we compared expression in normal untreated (UT) pancreas of adult mice to pancreata collected from mice injected (i.p.) with caerulein 2, 4, and 7 days post-treatment (Fig. 1A). Acinar-to-ductal metaplasia (ADM) is greatest at d2 postcaerulein, and morphology gradually returns to normal as the exocrine pancreas recovers by day 7 36 . Expression of nearly all HCAC and BET genes reflects this pattern, showing highest expression at day 2, sequentially declining towards normal levels at days 4 and 7. HDAC and BET expression in early stage ADM was compared to primary PDA cells from KIC mice (Fig. 1B) and human tumor samples (Fig. 1C) in the TCGA dataset (n = 72) collected predominately from stage 1 and 2 patients. In mice, the relative expression of HDAC and BET proteins is similar, except for the reciprocal pattern of HDAC1 and HDAC11 in ADM and primary PDA cancers (Fig. 1B). Primary cancer cells isolated from KIC-Rgs16::GFP mice were sorted for high and low expression of Rgs16::GFP (2.6-fold differential expression following 12 h incubation in 50% or 5% FBS, respectively) (Fig. 1B). By contrast, in human PDA tumors, containing cancer and stromal cell types, nearly all HDACs (except SIRT4) and BETs (except BRDT, BRWD3, and CECR2) were highly expressed (Fig. 1C).
We further determined the expression of HDACs and BETs during PDA progression in KIC mice. Pancreas from normal wild type mice, 40 day old KIC (early KIC, with early neoplastic lesion initially confirmed by ultrasound and then histology), and 60 day old KIC (late KIC, with advanced tumors) were digested and processed for a single cell RNA sequencing using 10 × Genomics platform 37 . HDAC1 and BRD2 were highly expressed in PDA compared to normal pancreas, both in humans and in epithelial and mesenchymal cancer cells in KIC mice ( Fig. 1D-F). HDAC overexpression has been associated with poor prognosis in PDA patients 31 . Regulators of histone acetylation induce Rgs16::GFP in mouse primary PDA cells cultured from KIC;Rgs16::GFP mice. Rgs16::GFP is expressed in the earliest ductal lesions and throughout PDA progression in KIC;Rgs16::GFP 22 that expresses a constitutively active Kras G12D allele and deletion of the tumor suppressor Ink4a 38 . KIC mice with homozygous deletion of Cdkn2a (Ink4a/Arf) develop PDA tumors by 3 weeks. Half of the mice die by 8 weeks. P48::Cre drives the deletion of Cdkn2a and activation of Kras G12D in pancreas progenitor cells, and all duct, exocrine and endocrine cells in adult mice. Rgs16::GFP expression is Kras G12D -dependent in the earliest ductal neoplasia (P18) in KC and KIC mice 22 .
Rgs16::GFP is not expressed in normal pancreas acinar tissue ( Supplementary Fig. S1A), nor in the islets of mice with normal blood glucose 23 . Pancreatic stress provoked by caerulein injection induces a transient, gradient response of acinar cell dedifferentiation in the pancreas. Rgs16::GFP expression is induced in affected acini during the interval of one to four days after caerulein treatment ( Fig. 2A,D). Rgs16::GFP expression completely disappears by day 14 as the pancreas heals (Fig. 2B,E).
In KC or KIC mice, Rgs16::GFP is expressed throughout PDA progression, marks the earliest lesions, is only expressed in areas of neoplasia, and is proportional to early tumor burden in KIC;Rgs16::GFP mice (Supplementary Fig. S1) 22 . Immunofluorescence analysis of KIC;Rgs16::GFP tumors show GFP is co-expressed with Sox9 positive cancer cells (Fig. 2C,F). Additionally, we have previously shown that orthotopic transplantation of primary cancer cells isolated from KIC;Rgs16::GFP PDA tumors regrow GFP-positive PDA specifically in duct-like structures in recipient NOD-SCID mice 22 . Primary PDA cells in culture do not express Rgs16::GFP when grown in media with 5% FBS (Fig. 2G), but Rgs16::GFP expression is induced within 16 h of treatment with 50% FBS (Fig. 2H). The HAT activator ISX9 39 induces Rgs16::GFP protein in approximately 90% early passage mouse primary PDA cells (50 uM ISX9, 5% FBS; Fig. 2I), and mRNA of the endogenous Rgs16 gene and the Rgs16::GFP transgene within 16 h of treatment (Fig. 2K).
Histone deacetylase inhibitors (HDACi) such as TSA 40 are cytotoxic for pancreatic cancer cell lines and inhibit growth of different cancer types in vivo [41][42][43] . TSA is a potent inducer of Rgs16::GFP in primary PDA cells (Fig. 2J,L), and induced GFP in a dose dependent manner (Fig. 2M). Since BET bromodomain proteins are readers of histone acetylation, we assayed JQ1, a selective inhibitor of BET with efficacy against a number of different cancers, but it did not induce Rgs16::GFP in cultured primary PDA cells (5% FBS). Gem, standard of care chemotherapeutic for PDA, also did not induce Rgs16::GFP (data not shown).

TSA and JQ1 synergistically inhibit growth in mouse primary PDA cells. HDAC inhibitors and
other regulators of epigenetic modifications are cytotoxic for tumor cells and show low toxicity in vivo. TSA stimulated dose-dependent cell death of PDA primary cells in culture (Fig. 3A). Likewise, JQ1 stimulated PDA cell death in a dose-dependent manner (Fig. 3B). Both TSA and JQ1 had greater cytotoxicity in low passage primary PDA cells (P10) than high passage cells (P50 or P100) (Fig. 3A,B; all subsequent studies were done in low passage cells). Additionally, high passage P100 cells did not induce Rgs16::GFP in response to TSA, ISX9, or high serum. Regardless of passage number, TSA was more effective than JQ1 in inhibiting cell growth.
BET inhibitors retard pancreatic ductal adenocarcinoma cell proliferation and enhance Gem cytotoxicity 35 . However, BET therapeutics are often limited by acquired drug resistance 44 . Therefore, we evaluated the synergistic activity of TSA and JQ1 in low passage primary PDA cells. TSA synergizes with JQ1 to potently suppress primary PDA cell growth in a dose-dependent manner (p < 0.0001) as shown in response curve and CI plot ( Fig. 3C-E).

TSA synergistically potentiates JQ1 and Gem lethality in vitro and in vivo.
Gem is a standardof-care chemotherapy but is only marginally effective for PDA treatment in humans 7 or mice 22 . Because combination therapy is an effective approach, we tested the cytotoxic effect of Gem in combination with TSA and/ or JQ1. Gem is potentiated by both TSA and JQ1. However, Gem in combination with TSA was more effective compared to JQ1 (Fig. 4A). In our in vitro screening, the combination of Gem + TSA + JQ1 is one of the most  www.nature.com/scientificreports/ potent cytotoxic cocktails we have identified ( Fig. 4B-D). Therefore, we tested Gem + TSA + JQ1 in our rapid, two-week in vivo assay of PDA initiation and growth ( Fig. 5A-C; see Ref. 22,45 for comparisons to Gem + Abraxane and related myosin inhibitors). Gem + TSA + JQ1 significantly decreased the initiation and progression of PDA in all KIC;Rgs16::GFP mice tested at P29, compared to untreated or Gem alone (Fig. 5). Three of 8 mice that were treated with TSA + JQ1 + Gem had the lowest tumor burden we have ever observed in KIC mice at P29.

Discussion
Effective therapeutics and early detection strategies are needed to enhance the long-term survival of patients with PDA. A particular challenge is to identify critical signaling pathways that drive PDA formation and progression. Gem is a standard-of-care drug but is only marginally effective for PDA treatment. Combination therapy is one approach to improve treatment. We describe a concerted approach using Rgs16::GFP expression in cell culture and a rapid in vivo assay (RIVA) in a mouse model of PDA to identify improved therapeutics. Signaling through GPCRs, which represents the largest sector of pharmaceutical development, is an attractive area that is understudied in pancreatic cancer. The long-term objective of this work is to identify therapeutics that act synergistically to inhibit the growth of mutant Kras-dependent PDA. Mutant alleles of Kras (such as Kras G12D ) are found in more than 90% of human PDA samples 46 . Kras GEFs can be activated by protein kinase receptors and GPCR 47 . GPCR signaling can be modulated by Rgs proteins, which accelerate the GTPase activity of Gq-and Gi-class alpha subunits 17,18 . GPCR signaling can induce Rgs gene expression as a feedback regulatory mechanism 18 . In KIC;Rgs16::GFP mice, GFP is induced in the earliest pancreatic lesions and in a subset of PDA cells throughout tumor progression 22 . In these mice, at two weeks of age, every cell in the pancreas expresses Kras G12D but only a few foci have undergone acinar to ductal metaplasia (ADM) and express Rgs16::GFP 22 . Simply www.nature.com/scientificreports/ expressing Kras G12D is not sufficient to induce acinar to ductal metaplasia (ADM) and Rgs16::GFP. However, Kras G12D increases the probability that acinar cells will dedifferentiate and release inflammatory cytokines that promote further activation of oncogenic Kras 48,49 . Kras and the master regulatory transcription factor Ptf1a oppose each other's activity. Ptf1a maintains terminally differentiated and functional acinar cells whereas oncogenic Kras promotes ADM and suppresses Ptf1a expression. At later times in tumor progression, Rgs16::GFP expression is restricted to neoplasia in PDA ( Fig. 2; Supplementary Fig. S1). Thus, Rgs16 expression can be effectively used to follow tumor progression in the mouse models of PDA therapy and identify potential new therapeutics. Primary PDA culture cells, obtained from KIC;Rgs16::GFP mice, express Rgs16::GFP and retain the capacity to form GFP + tumors when injected into recipient mice 22 . They display the morphological and gene expression profile of well differentiated epithelial cancer cells 22 . Ductal expression in adult animals appears to recapitulate early embryonic expression of Rgs16 in progenitor cells of the pre-pancreatic bud and epithelium of the developing pancreas and the earliest response to diabetic hyperglycemia 22,23 . :GFP mice at P29 in untreated (Unt, black dots) mice or treated with Gem_1 (G, green dots) alone or in combination (T + J + G, purple dots) with TSA (T) and JQ1 (J). Control (Ctrl, grey dots) mice are non-tumorigenic Rgs16::GFP mice. Tumor burden of each mouse is represented by five non-overlapping micrographs, depicted as dots, lined vertically from the highest to the lowest GFP fluorescence intensity quantified by NIH ImageJ (https ://image j.nih.gov/ij/; publicly available). Third highest (median) intensity micrograph of each mouse is marked with a red horizontal line. Pancreata are aligned from left to right in descending medians, each treatment group separated by vertical dashed lines. Upper and lower horizontal dashed lines represent the 95th and 1st %tile of all untreated micrograph values. Number of mice with micrographs above the 95th and below the 1st %ile is shown for each group above the total sample size. Control group has only one micrograph value per pancreas. Graphpad Prism v7 software (https ://www.graph pad.com/ (current version); licensed) and NIH Image J (https ://image j.nih.gov/ ij/; publicly available) were used for data analysis and visualization 60 www.nature.com/scientificreports/ In primary PDA cells, Rgs16::GFP expression was induced by serum or two regulators of histone acetylation (ISX9, TSA) (Fig. 2G-M). Concentration-dependent FBS induction of Rgs16::GFP is not likely to be cell proliferation dependent as primary PDA cells have a consistent doubling time of about 12 h at FBS concentrations of 10% and above. FBS likely contains GPCR ligands that induce Rgs16::GFP gene expression in a concentrationdependent manner.
HDAC and BET family proteins are highly expressed in human PDA and mouse models of PDA, and inhibitors of these epigenetic regulators have some therapeutic value 26,30 . Treatment with the Brd4 inhibitor JQ1 alone is also not very effective 50,51 but has shown improvement in combination with Gem. We tested inhibitors of HDAC and BET proteins alone and in combination for their cytotoxic effects on primary PDA cells isolated from KIC;Rgs16::GFP mice. TSA was the most potent inducer of Rgs16::GFP in primary mouse PDA cells in culture, but was not a particularly effective cytotoxic agent by itself. The time course of TSA induction of Rgs16::GFP expression and subsequent cell death suggested that stressed primary PDA cells expressed Rgs16::GFP 12-24 h prior to death. Therefore, we tested whether TSA would potentiate cytotoxic drugs used to treat PDA. Indeed, TSA acted synergistically with Gem and JQ1 to kill primary PDA cells during three days of treatment in culture (Fig. 4). Low passage primary PDA cells were more responsive than high passage cells for Rgs16::GFP expression and sensitivity to cytotoxic drugs (Fig. 3) with the exception that the highest concentrations of TSA killed all cells in cultures of high passage PDA cells that had undergone crisis (P50 and P100, Fig. 3A Figs. S2, S3). There was no correlation between Rgs16 expression and Kras allele status in 45 human PDA cell lines, but there were only 4 WT cell lines available for comparison. Therefore, we extended the analysis to other cancers with more WT Kras alleles, and again there was no correlation. These results are all from stable cell lines, whereas Rgs16 was induced in low passage primary PDA cells, but not after these cells were passaged 100 times (Fig. 3). The capacity for Rgs16 induction was probably lost in human cancer cell lines that were passaged innumerable times. By contrast, the seven TCGA tumor samples with highest Rgs16 expression all harbored oncogenic Kras alleles ( Supplementary Fig. S2, S3) and were among the clades most related to mouse PDA cells by whole transcriptome analysis 19 . Thus, primary cell culture is optimal for using Rgs16 gene expression to screen for novel therapeutics.
In summary, we have developed an innovative methodology of Rgs16::GFP reporter gene expression to identify new PDA chemotherapy in mouse primary PDA cells and validate their efficacy in vivo 22 . The validation step in Rgs16::GFP mice is quantitative, cost effective, and rapid-10 min per mouse to quantitate tumor burden following a 2 week in vivo assay prior to weaning in a spontaneous model of PDA in transgenic mice. Furthermore, the initial screen in cultured primary cells can be used to identify and characterize new GPCRs and ligands or adapted for high throughput screens to identify new chemo-preventative therapy for early neoplastic lesions in high risk patients and/or adjuvant chemotherapy after successful resection. The combined approach of HTS and in vivo validation using our Rgs16::GFP reporter is a powerful innovation in pancreatic cancer research. In future studies, we will characterize the biological mechanism of GPCR antagonists and other drugs that inhibit PDA initiation and progression in KIC mice. This work suggests a general approach for using Rgs genes to identify novel ligands and GPCRs and develop biomarkers and drugs for inflammatory diseases and cancers. Because Rgs regulation of GPCR signaling is ubiquitous during embryonic and childhood development, and adult life, the approach of using Rgs reporter genes to understand PDA is broadly applicable to identifying therapeutic interventions for the treatment of many cancers and disease conditions.

Material and methods
Ethical considerations. All experimental procedures were reviewed and approved by the UT Southwestern Medical Center Institutional Animal Care and Use Committee (IACUC protocol: 2016-101480) and conducted in accordance with the UT Southwestern Medical Center IACUC guidelines. All efforts were made to minimize animal suffering.
Comparative gene expression. Whole exome sequencing genomic data was obtained from the publicly available TCGA (Cancer Genome Atlas) database for all samples of pancreatic ductal adenocarcinoma (PDA). Only samples that had mutational profiling and RNA sequencing data available were included. The mutation annotation format file used was the latest mc3 + caller version created as part of the recent TCGA pan cancer atlas study. RNA sequencing data was batch normalized using RNA seq by Expectation-Maximization (RSEM). Tissue digestion for single cell RNA sequencing, single-cell cDNA library preparation, sequencing, and bioinformatics analyses for single cell RNA sequencing was performed as described previously 37   Fluorescent microscopy and GFP quantification. Pancreatic expression of Rgs16::GFP in KIC;Rgs16::GFP mice was captured and quantified as described 22 . Briefly, a Zeiss Lumar tissue dissection microscope was used to capture the images. Images were captured via a single-channel camera (Hamamatsu Photonics 60-C, 1″, 1×) in 1344 × 1024 resolution with 1 s exposure and 1 × 1 binning, analog gain = 10, and analog offset = 2 settings. Pancreatic fields representing the tumor burden (five or more) of the pancreas were imaged, covering up to 50% of the organ surface area. Images were quantified using NIH ImageJ (https ://image j.nih.gov/ ij/; publicly available) software with background subtraction with a radius of 50 pixels. A variable and tight lower threshold was set to eliminate residual background. Intensities of all particles with size ≥ 5 pixels were summed to obtain the total light intensity per image. www.nature.com/scientificreports/ GE IN Cell Analyzer 6000 microscope equipped with a Nikon 20 X/0.45 objective was used for GFP expression determination in primary PDA cells. After 72 h of incubation with drug(s), cells were imaged using 405 and 488 lasers lines for Hoechst 33342 and GFP, respectively. Sixteen fields of view per well were captured using a 4-megapixel CMOS camera binned at 2 × 2. Images were analyzed using the GE Analyzer Workstation v3.7.3 software (URL not available for the version used in this paper. URL for currently available version: https ://downl oad.cytiv alife scien ces.com/cella nalys is/downl oad_data/incel l/6500/incel l_6500_downl oad_page.htm; licensed). Nuclei and cellular compartments were segmented using their respective wavelengths. The mean GFP signal intensity for each cellular object was measured.
RNA preparation and quantitative RT-PCR. Cells were collected with Trizol (Invitrogen, CA, USA) and total RNA was extracted using the Direct-zol RNA MiniPrep kit (Zymo Research, CA, USA). Following DNase treatment, cDNA was synthesized using SuperScript II Reverse Transcriptase (Invitrogen). Real-time PCR was performed on 10-50 ng of cDNA in triplicate using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, CA, USA) and Bio-Rad CFX96 Real-Time PCR Detection System. Primers for qPCR were designed using the Primer-BLAST (NCBI Tools) and validated. Data were analyzed using the comparative C T method (2 −ΔΔC T ) 58 using cyclophilin B as a reference gene for normalization. Relative changes in expression of target gene in response to drug treatment were expressed as fold induction compared with the level of expression (given as 1) in non-treated cells.
Statistical analysis. Data are reported as mean ± SEM. Graphs and their statistical comparisons were performed using GraphPad Prism v7 software [Graphpad Software Inc., CA, USA, https ://www.graph pad.com/ (current version); licensed]. Between-subject ANOVAs were used to detect significant changes among multiple groups or conditions followed by tukey's post-hoc test. CompuSyn synergy software (https ://www.combo syn. com/; publicly available) based on the drug combination principles of Chou-Martin 59 was used for combination index (CI) values quantification. An alpha of p < 0.05 was considered statistically significant. www.nature.com/scientificreports/