A Small-Molecule Inducer of PDX1 Expression Identified by High

Chemistry & Biology Article

A Small-Molecule Inducer of PDX1 Expression Identiﬁed by High-Throughput Screening

Yuan Yuan,1,2 Kate Hartland,3 Zarko Boskovic,1,2,4 Yikai Wang,1 Deepika Walpita,1 Philippe A. Lysy,5 Cheng Zhong,1 Damian W. Young,1 Young-kwon Kim,1 Nicola J. Tolliday,3 Etienne M. Sokal,5 Stuart L. Schreiber,1,2,4 and Bridget K. Wagner1,* 1Chemical Biology Program, Broad Institute of Harvard and MIT, Cambridge, MA 02142, USA 2Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA 3Chemical Biology Platform, Broad Institute of Harvard and MIT, Cambridge, MA 02142, USA 4Howard Hughes Medical Institute, Broad Institute of Harvard and MIT, Cambridge, MA 02142, USA 5Laboratory of Pediatric Hepatology and Cell Therapy, Catholic University of Leuven, Brussels 1200, Belgium *Correspondence: [email protected] http://dx.doi.org/10.1016/j.chembiol.2013.10.013

SUMMARY cells and through transdifferentiation from mature non-b cells such as liver, acinar, and ductal cells (Akinci et al., 2012; Ferber Pancreatic and duodenal homeobox 1 (PDX1), a et al., 2000; Kaneto et al., 2005; Karnieli et al., 2007; Kojima et al., member of the homeodomain-containing transcrip- 2002; Kubo et al., 2011; Motoyama et al., 2009; Wu et al., 2007; tion factor family, is a key transcription factor impor- Yamada et al., 2001; Yang et al., 2002). A notable recent tant for both pancreas development and mature b success in the transdifferentiation field was the generation of b cell function. The ectopic overexpression of Pdx1, functional insulin-producing cells from acinar cells in mice Neurog3, and MafA in mice reprograms acinar cells after viral transduction of the pancreas with Pdx1, Neurog3, and MafA (Zhou et al., 2008). These results were repeated in to insulin-producing cells. We developed a quantita- cell culture using the rat exocrine cell line AR42J (Akinci et al., tive PCR-based gene expression assay to screen 2012), suggesting that in vitro cell models can enable the study more than 60,000 compounds for expression of of cellular reprogramming. each of these genes in the human PANC-1 ductal From a therapeutic perspective, viral introduction of exoge- carcinoma cell line. We identified BRD7552, which nous genetic material raises concerns regarding genomic inte- upregulated PDX1 expression in both primary human gration and tumorigenicity (Barrilleaux and Knoepfler, 2011). islets and ductal cells, and induced epigenetic An attractive alternative is to use small molecules that achieve changes in the PDX1 promoter consistent with temporal high expression of transcription factors. Recent transcriptional activation. Prolonged compound efforts to this end have focused on the factors required for the treatment induced both insulin mRNA and protein generation of induced pluripotent stem (iPS) cells (Takahashi and also enhanced insulin expression induced by and Yamanaka, 2006). This approach identified kenpaullone as a potential replacement for Klf4 (Lyssiotis et al., 2009), the three-gene combination. These results provide a transforming growth factor b receptor inhibitor as a potential a proof of principle for identifying small molecules replacement for Sox2 and c-Myc (Ichida et al., 2009), and Src that induce expression of transcription factors to inhibitors as other replacements for Sox2 (Staerk et al., 2011). control cellular reprogramming. These results suggest that novel screening methods can identify compounds that aid in the reprogramming process. Herein we report the development of a gene expression-based INTRODUCTION assay to perform high-throughput screening of 60,752 compounds for endogenous expression of PDX1, NEUROG3,or Pancreatic and duodenal homeobox 1 (PDX1) is a key tran- MAFA. We identified BRD7552, a compound that increased scription factor involved in pancreas development and b cell PDX1 mRNA levels in a dose- and time-dependent manner in function (Babu et al., 2007; McKinnon and Docherty, 2001). human PANC-1 cells, primary human islets, and human duct- Homozygous disruption of Pdx1 in mice and PDX1 in human derived cells (HDDCs). BRD7552 modified histone H3 tail modi- patients causes pancreatic agenesis. Heterozygous mutation fications associated with transcriptional activation, suggesting is associated with both type 2 diabetes and maturity onset dia- that the compound may induce PDX1 expression through betes of the young, type 4 (Ahlgren et al., 1997; Oliver-Krasinski either direct or indirect epigenetic control. Further mechanism et al., 2009; Stoffers et al., 1997). In the adult pancreas, PDX1 is of action studies suggest a role for the transcription factor expressed in b cells and d cells and controls the expression FOXA2 in BRD7552-induced PDX1 transcriptional activation. of key b cell markers such as insulin, glucose transporter 2, BRD7552 can partially replace PDX1 in the genetic induction of MafA, glucokinase, and islet amyloid polypeptide (Babu et al., insulin expression in PANC-1 cells. These results lay a founda- 2007; McKinnon and Docherty, 2001). Furthermore, ectopic tion for the development of novel small molecules as useful overexpression of PDX1 is essential for b cell neogenesis, both tools with which to manipulate the endogenous expression of through direct differentiation from pluripotent or progenitor master regulatory transcription factors.

RESULTS molecules that induce the endogenous expression of PDX1. We therefore developed a gene expression-based assay using Transcription Factor Target Validation in PANC-1 Cells real-time qPCR (Arany et al., 2008) to identify potential small- In order to develop an in vitro model for high-throughput chem- molecule inducers. Using transfection of all three genes as a ical screening, we assessed the suitability of human PANC-1 positive control, we observed that assay Z0-factor values, a ductal adenocarcinoma cells. Although not an acinar cell line, measure of statistical effect size and assay quality (Zhang PANC-1 cells are amenable to high-throughput screening and et al., 1999), were greater than 0.5 for each of the three genes provide a good framework for b cell neogenesis. After cotrans- (Figure 2A), indicating that the assays were suitable for high- fection of full-length PDX1, NEUROG3, and MAFA and antibiotic throughput screening. selection over 2 weeks, we observed an increase in C-peptide We then screened 60,752 compounds to identify inducers of immunofluorescence in a majority of cells (Figure 1A). Each transcription factor expression in PANC-1 cells after 3-day transcription factor was highly expressed in these cells (Figures treatment (Figures S2A and S2B). The screening collection 1B–1D). Interestingly, at the end of 2 weeks, the localization of included compounds from libraries prepared by diversity- each transcription factor was both nuclear and cytoplasmic, oriented synthesis (DOS) (Schreiber, 2009), as well as kinase- suggesting that protein export from the expected nuclear biased, chromatin-biased, commercial compound, bioactive, location may occur during this time. Cytokeratin-19 (CK19), a and natural product libraries. Although no hits were identified ductal marker highly expressed in PANC-1 cells, was signifi- in the NEUROG3 or MAFA screens, we identified BRD7552, cantly downregulated in these cells, consistent with reports of a glucose-derived DOS compound, as a hit in the PDX1 its negative regulation by PDX1 (Deramaudt et al., 2006)(Fig- screen (Figures 2B and S2C). BRD7552 increased PDX1 ure 1E). Quantitative PCR (qPCR) revealed that each of the mRNA levels two- to fourfold, with a maximal effect at 5 mM three transcription factors was highly upregulated 3 days after (Figure 2C). The expression of MAFA was only marginally transfection but decreased to lower levels after 2 weeks in increased by BRD7552, but no NEUROG3 induction was de- culture (Figure 1F). Insulin mRNA levels were induced greater tected (Figures S2D and S2E). than 1,000-fold in transfected cells, albeit from an undetectable Five- and nine-day treatment of cells with BRD7552 further basal state. On the other hand, hormones expressed in other increased PDX1 expression (Figure 2C), but effects were pancreatic endocrine cell types, such as glucagon, somato- observed as early as six hours after compound treatment statin, and pancreatic polypeptide, were induced less than and gradually increased over time (Figure 2D). After 3-day

10-fold (Figure 1F). Examination of the threshold cycle (Ct) compound washout, the expression level of PDX1 returned values of control PANC-1 cells, as well as the DCt values relative to original levels (Figure S3A), suggesting that the effects to glyceraldehyde 3-phosphate dehydrogenase (GAPDH), of BRD7552 are reversible. Immunofluorescence revealed reveals that, with the exception of PDX1 (Ct 28.0), these hor- PDX1 to be cytoplasmic (Figure S3B), perhaps reflecting an mones and transcription factors are more or less not expressed immature or stressed state (Guo et al., 2013; Kawamori in PANC-1 cells. However, the 1,000-fold increase in PDX1, et al., 2003; Macfarlane et al., 1999). We then assessed the NEUROG3, and MAFA, as well as the 2-week 1,000-fold upregu- effects of BRD7552 on endocrine cells. We observed an lation of INS, would put these Ct values nearer to those of increase in PDX1 expression in mouse aTC cells (Figure S3C), GAPDH. These results suggest that an initial burst of expression which express levels of PDX1 similar to those of PANC-1 of PDX1, NEUROG3, and MAFA may be sufficient to induce cells. However, mouse bTC cells, which already have high insulin expression in PANC-1 cells. levels of PDX1 expression, showed no further induction (Fig- Despite the fact that these cells demonstrated some signs ure S3D), suggesting that BRD7552 may impinge on a proof reprogramming, the process appeared to be incomplete. cess already active in b cells. In accordance with the qPCR Insulin levels were not comparable to normal b cells, and neither results, BRD7552 induced a dose-dependent increase in basal insulin secretion nor glucose-stimulated insulin secretion PDX1 protein levels after BRD7552 treatment in PANC-1 cells was detected (data not shown). Furthermore, transfected cells (Figures 2E, 2F). Protein levels of CK19, a ductal marker started dying after 3 weeks in culture, which is consistent with negatively regulated by PDX1, were modestly decreased by results described in AR42J cells (Akinci et al., 2012). The pro- compound treatment as well (Figures 2E). We confirmed that liferation marker Ki67 was reduced to a great extent after the relatively high concentrations of BRD7552 used had no transfection and selection (Figure 1G), and total cell number effect on PANC-1 apoptosis (Figures S3E and S3F). Because was reduced by about 45% compared to cells transfected PANC-1 cells express some PDX1 under basal conditions, with a control vector (Figure 1H). Anchorage-independent cell these data suggest that even a moderate induction may be growth and cell invasion, general properties of human cancer sufficient to promote the downstream effects of this tran- cells, were reduced by about 25% by transcription factor scription factor. overexpression (Figure 1I and Figure S1 available online). These data suggest that overexpression of PDX1, NEUROG3, and BRD7552 Increases Insulin Expression MAFA suppresses cell proliferation and may adopt some Next, we sought to determine whether BRD7552 has an endocrine cell-like features. effect on insulin levels. Nine-day treatment of PANC-1 cells with BRD7552 caused a dose-dependent increase in insulin Screening for Induction of PDX1 Expression mRNA expression (Figure 3A). We also tested the effects of Given the induction of insulin expression in PANC-1 cells under BRD7552 in primary human islets using a cell culture system these conditions, we used this model system to identify small developed for high-throughput screening (Walpita et al.,

Figure 1. Three Transcription Factors Induce b Cell-like Characteristics in Human PANC-1 Cells Immunoﬂuorescence analysis of control and triply overexpressed (TOE) PANC-1 cells after 2 weeks in cell culture. (A) C-peptide. (B) NEUROG3. (C) PDX1. (D) MAFA. (E) Cytokeratin-19 (CK19). Nuclei were stained with Hoechst 33342 dye. Scale bars, 100 mm. (F) qPCR analysis of insulin, MAFA, NEUROG3, PDX1, PAX4, GCG, SST, and PPY gene expression in TOE cells after 3-day and 2-week cultures. Gene expression was normalized by GAPDH expression and scored relative to control vector-transfected cells. Ct (Ctrl) indicates basal threshold cycle (Ct) value of each gene in

DMSO-treated PANC-1 cells. DCt refers to the GAPDH-normalized levels of expression of these genes in control PANC-1 cells. Data represent the mean ± SD of three independent experiments. (G) Immunofluorescence analysis of Ki67 (green) performed on TOE cells 2 weeks after transfection and selection. Nuclei were stained with Hoechst 33342 dye. (H) Total nuclei of TOE cells in each well were detected by Hoechst fluorescence and counted by automated microscopy. Data represent the mean and standard error of eight independent replicates. *p < 0.0001 (t test). (I) Quantification of PANC-1 colony formation in soft agar by nucleic acid detection. RFU, relative fluorescence units. *p < 0.01 (t test). See also Figure S1 and Table S1.

2012). Three of ﬁve donor samples (Figure S4A) tested showed of this primary culture system, we could not determine which signiﬁcant dose-dependent induction of PDX1 (Figure 3B) after cell types were responsible for the increase in gene expres- 3 or 5 days and of insulin (Figure 3C) after a 5-day treatment. It sion; however, we observed no changes in cell numbers is important to note that, because of the heterozygous nature upon 3-day treatment with BRD7552 (Figure S4B), suggesting

Figure 2. BRD7552 Induces PDX1 Gene and Protein Expression (A) Assay development of gene expression in 384- well plates. Triply overexpressed (TOE) cells were compared to DMSO-treated PANC-1 cells for expression of indicated genes. Z0 factors for each assay are listed below each gene name. (B) Chemical structure of BRD7552. (C) qPCR analysis of PDX1 expression in PANC-1 cells following 3-, 5-, and 9-day treatment with

the indicated concentrations of BRD7552. DCt (DMSO) is the relative expression level of the target genes compared to the control gene, GAPDH, as an indication of the basal levels of expression in PANC-1 cells. The fold change was scored relative to control cells treated with DMSO. Data represent the mean ± SD of three independent experiments. (D) qPCR analysis of PDX1 expression induced by 5 mM BRD7552 at the indicated time points. *p < 0.001 (t test). (E) Assessment of protein levels of PDX1 and cytokeratin-19 (CK19) in PANC-1 cells treated with the indicated concentrations of BRD7552 for 5 days. Actin was used as an internal loading control. (F) Quantiﬁcation of PDX1 protein levels induced by 3-day BRD7552 treatment as measured by ELISA. Data represent the mean ± SD of three independent experiments. See also Figures S2 and S3 and Table S1.

a lack of a toxic effect on certain cell populations that could H3K9me3 (Figure 4D), results that are consistent with transcrip- have explained these results. In order to more faithfully mimic tional activation of PDX1. H3K27 methylation was unaffected our screening conditions, we tested the effects of BRD7552 in by compound treatment (Figure 4E). We also observed similar HDDCs, which were derived in vitro directly from purified changes in a portion of the insulin promoter (Figure S5). The CA19-9+ ductal cells through epithelial-mesenchymal transition proximal promoter of PDX1 in PANC-1 cells was hypomethy- and had been shown to be capable of differentiation toward lated, and no significant change in DNA methylation status b-like cells both in vitro and in vivo (Lysy et al., 2012). After was observed upon compound treatment (Figure S6). We 3-day treatment, BRD7552 induced PDX1 expression in two measured cellwide changes in histone modification in PANC-1 donor samples in a dose-dependent manner, and in a third cells and observed that BRD7552 increases total H3K9 and donor sample at the lowest concentration (Figure 3D), indi- H3K27 trimethylation levels (Figure S7). Remarkably, these cating that this compound retains activity in primary human changes are in the opposite direction from the effects on the pancreatic cells. PDX1 and insulin promoters, suggesting that the latter effects are not due to generic cellular effects on chromatin status. BRD7552 Changes the Epigenetic Status These results suggest that BRD7552 acts at the chromatin level of the PDX1 Promoter to induce PDX1 expression in PANC-1 cells. In order to investigate whether BRD7552 upregulates PDX1 expression by modifying its promoter region, we performed FOXA2 Is Involved in BRD7552 Activity bisulfite sequencing and chromatin immunoprecipitation In order to further understand the mechanism of action of (ChIP) qPCR to assess DNA methylation levels and histone BRD7552, we performed gene expression profiling in PANC-1 modifications in BRD7552-treated PANC-1 cells (Figure 4A). cells treated with DMSO or BRD7552. Genes involved in Histone H3 acetylation and trimethylation at lysine 4 insulin production were not increased; however, gene set (H3K4me3) are enriched in active promoter regions (Barski enrichment analysis (GSEA) (Subramanian et al., 2005) revealed et al., 2007; Eberharter and Becker, 2002; Koch et al., 2007), that genes downregulated by the forkhead box transcription whereas that at lysine 9 (H3K9me3) and lysine 27 (H3K27me3) factor FOXA2 were positively enriched in control cells (Fig- are associated with transcriptional repression (Barski et al., ure 5A), suggesting that BRD7552 increases either expression 2007; Koch et al., 2007). BRD7552 increased H3 acetyla- or activity of FOXA2. Accordingly, knockdown of FOXA2 tion (Figure 4B) and H3K4me3 (Figure 4C) and decreased (Figure S8) abolished PDX1 induction by BRD7552 (Figure 5B),

BRD7552 Partially Complements PDX1 in PANC-1 Cells To test whether BRD7552 could truly replace PDX1, we treated cells with 5 mM BRD7552 and the combination of MAFA and NEUROG3, or all three transcription factors. Because the induction effect of BRD7552 was reversible (Figure S2A), the compound was administrated every 3 days. After 2-week culture and antibiotic selection, the combination of plasmids and BRD7552 resulted in higher insulin mRNA expression levels compared to plasmids alone (Figure 7A). Immunoﬂuores- cence analysis revealed that all three genes were still required to detect C-peptide expression, but that the addition of BRD7552 increased those levels further (Figure 7B). These data demonstrate that PDX1 can be partially replaced by BRD7552.

DISCUSSION

The identification of chemical inducers of master transcription factors has become an attractive strategy to perturb and study Figure 3. BRD7552 Induces Insulin Expression in PANC-1 Cells biological systems. We developed a qPCR-based assay for (A) qPCR analysis of insulin expression on PANC-1 cells following 3-, 5- and 9- high-throughput detection of PDX1 expression and used this day treatment with the indicated concentrations of BRD7552. Gene expres- assay to screen over 60,000 compounds in human PANC-1 sion was normalized by GAPDH and actin expression and was scored relative cells. One compound, BRD7552, reproducibly increased to the control cells treated with DMSO. Data represent the mean ± SD of three PDX1 expression in a dose- and time-dependent manner and independent experiments. (B and C) qPCR analysis of PDX1 (B) and insulin (INS) expression (C) in induced insulin expression after longer treatment times. dissociated human islet cells from donor 3 (Figure S3), which were treated with Changes in chromatin modification status in the PDX1 and in- the indicated concentrations BRD7552 for 3 or 5 days. Gene expression was sulin promoters are consistent with the transcriptional changes normalized by GAPDH and actin expression and was scored relative to the in these genes. To evaluate the mechanism of action of control vector-transfected cells. Data represent the mean ± SD of three in- BRD7552, we performed gene expression profiling and dependent experiments. GSEA, which revealed that genes downregulated by the (D) qPCR analysis of PDX1 expression in HDDCs at passage 3 or 4 after in- transcription factor FOXA2 were also downregulated by cubation with BRD7552 at the indicated concentrations for 3 days. Gene expression was normalized by TBP expression and scored as the fold change BRD7552. Knockdown of FOXA2 abolished BRD7552 effects relative to 0.1% DMSO-treated cells. P, passage. on PDX1 expression. Interestingly, FOXA2 was recently found See also Figure S4 and Table S1. to occupy the PDX1 promoter at several domains and to be especially important during development (Gao et al., 2008). Further studies are required to deconvolute the mechanism indicating that FOXA2 is involved in the activity of BRD7552. and identify the primary target of BRD7552, as well as to deter- These results suggest that BRD7552 acts in a FOXA2-depen- mine the requirements for inducing a fully glucose-responsive dent manner to increase PDX1 expression in human ductal b cell state. cells. We observed an increase in insulin and PDX1 expression in human islets and HDDCs after compound treatment. Most Structural Features of BRD7552 Are Required for BRD7552 serves as a partial chemical complement for PDX1, Activity increasing insulin expression in the presence of NEUROG3 We assessed the structure-activity relationship of BRD7552 and MAFA overexpression. Further studies designed to based on a collection of 102 analogues present within the determine the precise intracellular target of BRD7552 will screening collection, as well as on newly synthesized analogues. enable identification of additional targets for screening to BRD0749, the stereoisomer of BRD7552, had very weak effects induce more potent increases in PDX1 expression. With regard on PDX1 expression after 5-day treatment (Figure 6), suggesting to cancer, it is noteworthy that, by targeting a master regulatory that the R configuration of C2 is crucial for this activity. We next transcription factor for upregulation by a small molecule, we explored the carbamate substituent at C4; its removal resulted have pushed a cancer cell line into a nonproliferative state. in a complete loss of activity (YW2), whereas other ester It remains to be seen whether this transdifferentiation analogues (YW5 and YW6) had effects similar to those of strategy is feasible for developing novel approaches to target- BRD7552 (Figure 6). YW3 and YW4, however, the synthetic ing cancer cells. intermediates of YW5 and YW6, were not active, suggesting Transcription factors have long been considered ‘‘undrug- that the primary hydroxyl group at C6 was essential. Further gable’’ targets for therapeutic development (Dandapani and exploration of the C4 carbamate substituent revealed that Marcaurelle, 2010; Payne et al., 2007). More conven- many small changes resulted in a loss of activity (Figure 6). tional targets for high-throughput screening have included Thus, there appears to be little flexibility in the structural G protein-coupled receptors and enzyme classes such as requirements for BRD7552 activity. kinases and histone deacetylases (HDACs). These results

Figure 4. Chromatin Modiﬁcations Induced by BRD7552 at the PDX1 Promoter Are Consistent with Transcriptional Activation (A) Location of the primer pairs used for ChIP- qPCR for the human PDX1 gene. PANC-1 cells were treated with DMSO (light gray bars) and 5 mM BRD7552 (dark gray bars) for 3 days. (B–E) The following modiﬁcations were analyzed by ChIP-qPCR: acetylated H3 (B), H3K4me3 (C), H3K9me3 (D), and H3K27me3 (E). Data represent the mean ± SDs of four independent qPCR experiments. *p < 0.05.

novel high-throughput screening aimed at less conventional targets. Other small-molecule modulators of epigenetic regulation, such as the DNA methyltransferase inhibitor 50-azadeoxy- cytidine, increase NEUROG3 expression in PANC-1 cells (Lefebvre et al., 2010). Additionally, HDAC inhibitors amplify endocrine progenitors during development in mice (Haumaitre et al., 2008). Nevertheless, cellular toxicity of general HDAC inhibitors has prevented their application in differentiation, transdifferentiation, and other related studies. This study provides a path through provide evidence that small-molecule modulators of transcrip- which to identify small-molecule tools to manipulate the ex- tion factor expression can be identified by high-throughput pression of master regulatory transcription factors. BRD7552, screening. The extremely low hit rate in our assay might indi- a small molecule with a distinct chemical structure, can cate that transcription factor expression is only rarely targeted induce PDX1 expression and demonstrates the potential by current screening collections (Burke and Schreiber, 2004; to enhance a b cell-like phenotype in a human pancreatic Schreiber, 2009). In fact, the only hit we identified was a ductal cell line in vitro. Such a compound may have particular member of a library of compounds prepared by DOS. Such utility in improving the efficiency of pancreatic differentiation compounds have novel structures not well represented in from human iPS cells. Remaining challenges include fully other common screening collections (Clemons et al., 2011). uncovering the compound’s mechanism of action, further Identification of the structural features critical for effects on increasing levels of PDX1 and insulin expression, and improving transcription factors, coupled with screening collections potency and pharmacokinetic properties through medicinal that fill the gaps in chemical diversity space, will be useful for chemistry.

Figure 5. FOXA2-Targeted Gene Set Is Linked to BRD7552 Treatment in PANC-1 Cells by Gene Set Enrichment Analysis (A) GSEA enrichment plot showing expression enrichment of a set of FOXA2-downregulated genes. Bars represent individual genes in a ranked data set list. Expression of the majority of these genes was enriched in the DMSO-treated PANC-1 cells compared to BRD7552-treated cells. Heat map of the 27 genes in the leading edge of the gene set. (B) qPCR analysis of PDX1 expression induced by 5 mM BRD7552 and FOXA2 knockdown in PANC-1 cells. Data represent the mean ± SD. *p < 0.001 (t test). See also Figure S8 and Table S1.

Figure 6. Most Structural Features of BRD7552 Are Required for Activity The common chemical structure of all analogues is shown at the top. Alterations of the stereochemistry at the C2 position, as well as the identities of R (red) and R0 (blue) groups, are indicated below. PDX1 induction of each small molecule was tested at 10 mM after 3- and 5-day treatments. The fold change was normalized by GAPDH expression and scored relative to DMSO-treated control cells. Data represent the mean of two independent experiments.

SIGNIFICANCE these transcription factors, followed by prolonged culture, induces phenotypic changes toward a b cell-like state. We The ability to induce cellular reprogramming in an on- then performed chemical screening to identify compounds demand fashion with small molecules has the potential capable of replacing these factors. High-throughput to transform cell biology and discovery of therapeutics. screening using qPCR as a readout has made possible the In particular, the ability to replenish b cell mass by repro- rapid identification of novel compounds to induce target gramming other cells in the pancreas would represent gene expression. We identified BRD7552, a compound that a significant advance toward novel diabetes therapies. In induces PDX1 expression in PANC-1 cells in a dose-depen- this study, we aimed to identify small molecules capable of dent manner. The epigenetic changes induced by BRD7552 inducing the expression of the transcription factors PDX1, are consistent with transcriptional activation, and NEUROG3,orMAFA, all of which are master regulatory BRD7552 acts in a FOXA2-dependent manner. Prolonged genes involved in b cell development. First, we observed treatment with BRD7552 also induces expression of insulin, that transfection of a human ductal cancer cell line with and this compound is active in primary human islet cells

Figure 7. BRD7552 Can Partially Comple- ment PDX1 in PANC-1 Cells (A) qPCR analysis of insulin expression of transfected PANC-1 cells in the presence or absence of 5 mM BRD7552 after 2 weeks in culture. Gene expression was normalized by GAPDH and actin expression and was scored relative to the control vector-transfected cells. Data represent the mean ± SD of three independent experiments. (B) Immunoﬂuorescence analysis of C-peptide expression (red) in transfected PANC-1 cells in the absence or presence of 5 mM BRD7552 after 2 weeks in culture. Nuclei were stained blue with Hoechst 33342 dye. Scale bars, 100 mm. See also Table S1.

and in primary HDDCs. This study lays the foundation for and scramble siRNA were transfected with lipofectamine RNAiMAX reagent the eventual development of a toolkit of small-molecule (Invitrogen) at a final concentration of 100 nM. replacements for reprogramming factors. High-Throughput Screening PANC-1 cells were plated and allowed to attach overnight before the com- EXPERIMENTAL PROCEDURES pound pin-transferred from concentrated DMSO stocks at a final concentration of 20 mM. Cells were treated with the compound for 3 days before the Cell Cultures and Compounds qPCR assay. The Cells-to-CT Kit (Ambion) was used for cell lysate and reverse PANC-1, aTC cells, and bTC cells (American Type Culture Collection) were transcription. A qPCR assay was performed using the Twister II Microplate cultured in DMEM containing 10% fetal bovine serum (FBS) and 100 U ml 1 Handler (Caliper Life Sciences) and the LightCycler 480 System (Roche penicillin/streptomycin (P/S). Human islets were obtained from the Integrated Applied Science). The data were uploaded and analyzed using Genedata Islet Distribution Program and were cultured as described previously (Walpita software. Approximately 300 compounds scoring greater than two standard et al., 2012). Donor information is provided in Figure S4. HDDCs were deviations from the DMSO distribution were selected for retesting. To confirm expanded from cultures of CA19-9+ human ductal cells purified from screening positives, dose-response curves were generated using GAPDH islet-depleted pancreatic tissues using immunomagnetic bead separation as as a control. Compound purity and identity were determined by ultraperform- previously described (Yatoh et al., 2007). Human cadaveric donor cells were ance liquid chromatography mass spectrometry (Waters). A detailed protocol obtained from the University of Chicago. In brief, freshly isolated CA19-9+ for the high-throughput qPCR assay is provided in the Supplemental Experi- ductal cells were initially seeded onto tissue culture-treated dishes in CMRL mental Procedures. medium (Gibco) containing 10% FBS and 1% P/S (Cellgro). After 72–96 hr, the medium was changed for EGM-2 MV BulletKit medium (Lonza) without Western Blot Analysis hydrocortisone to stimulate proliferation of HDDCs. The medium change Cells were lysed on ice in radioimmunoprecipitation assay buffer (20 mM was then performed twice weekly. Proliferating cells were passaged at Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM Na2 EDTA, 1 mM EGTA, 1% Nonidet 80% confluence using 0.25% trypsin (Cellgro) and subcultured at 5,000 P-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM 2 cells/cm . Cells were cultured at 37 C in a humidified atmosphere containing 1 b-glycerophosphate, 1 mM Na3VO4,1mgml leupeptin, protease inhibitors, 3 4 2 5% CO2. For differentiation assays, HDDCs were plated at 2 10 cells/cm and phosphatase inhibitors). Total protein was separated by 4%–12% SDS- on uncoated, tissue culture-treated, 12-well dishes in DMEM/F-12 containing PAGE and transferred to a polyvinylidene fluoride membrane. Blots were 0.1% FBS. The compounds BRD7552, BRD0749, BRD8583, BRD3185, developed using SuperSignal Chemiluminescent Substrate (Thermo Fisher). BRD6666, BRD7781, BRD5843, BRD7929, BRD5733, YW5, YW6, YW3, Images were quantified using ImageJ software (National Institutes of Health). YW4, YW2, and CZ1 were synthesized in our laboratory. The synthetic route Antibodies used for western blotting were goat anti-PDX1 antibody from and chemical characterization of BRD7552 are included as Supplemental Abcam (1:500), mouse anti-CK19 antibody from Sigma-Aldrich (1:500), and Information. mouse anti-b-actin antibody from Sigma-Aldrich (1:5,000).

RNA Isolation and Real-Time PCR ELISA Total RNA was extracted from cells using the RNeasy Plus Mini Kit (QIAGEN). The Human PDX1 ELISA Kit (Cusabio) was used to quantitatively determine Reverse transcription was performed using the High Capacity cDNA Reverse PDX1 concentrations in cell lysates. Brieﬂy, 100 ml of standards or samples Transcription Kit (Applied Biosystems) with RNase inhibitor (QIAGEN) were added into a 96-well and incubated for 2 hr at 37C. Liquid from each according to the manufacturer’s protocol. Real-time qPCR reactions were well was removed, and wells were incubated for 1 hr with biotinylated anti- performed with SYBR Green PCR Master Mix in 384-well format using a body, washed 3 times, and incubated with horseradish peroxidase-avidin. 7900HT plate reader (Applied Biosystems). cDNA primers for FOXA2 3,30,5,50-Tetramethylbenzidine substrate was added to each well and (QT00212786) and FOXO1 (QT00044247) were purchased from QIAGEN. incubated for 15–30 min at 37C, followed by stop solution. The optical Other cDNA primers were obtained from Euroﬁns MWG Operon, the se- density of each well was determined within 5 min using a microplate reader quences of which are listed in Table S1. set to 450 nm.

Plasmid Transfection and siRNA Knockdown Immunoﬂuorescence and Microscopy Human PDX1 (RC222354), NEUROG3 (RC224767), MAFA (RC214782) and Cells were seeded into black optical 96-well plates overnight for attachment, control vector (PS100001) were obtained from Origene. Both 1 mg/ml of followed by compound treatment. Cells were ﬁxed for 20 min in 4% parafor- each plasmid and 3 mg/ml of control plasmid were transfected into PANC-1 maldehyde, permeabilized for 20 min with 0.1% Triton X-100, blocked with cells using Lipofectamine 2000 (Invitrogen), and cells were selected with PBS containing 2% BSA at 4C overnight, and incubated at 4C overnight neomycin (Sigma Aldrich). siRNAs against FOXA2 (GS3170) and FOXO1 with the following primary antibodies: mouse (Ms) C-peptide (Chemicon (GS2308) were purchased from QIAGEN. siRNA against FOXA2, FOXO1, International), goat Pdx1 (Abcam), rat brain (Rb) neurogenin 3 (EMD Millipore),

Rb Mafa (Abcam), Ms CK19 (Sigma-Aldrich), and Ms Ki-67 (Abcam) at 1:100 REFERENCES dilution. Cyanine 3 (Cy3)-labeled or Cy2-labeled secondary antibodies (Jackson ImmunoResearch) were added, along with Hoechst 33342 dye Ahlgren, U., Pfaff, S.L., Jessell, T.M., Edlund, T., and Edlund, H. (1997). (Invitrogen), and observed with an Axiovert 200M fluorescence microscope Independent requirement for ISL1 in formation of pancreatic mesenchyme (Zeiss) at 3400 magnification. Cell number analysis was performed with an and islet cells. Nature 385, 257–260. 3 ImageXpress Micro automated microscope (Molecular Devices) using a 4 Akinci, E., Banga, A., Greder, L.V., Dutton, J.R., and Slack, J.M. (2012). lens objective with laser-based focusing. Nuclear staining was handled in Reprogramming of pancreatic exocrine cells towards a beta (b) cell character Photoshop (Adobe Systems) to improve visibility. Each image in Figure 1 using Pdx1, Ngn3 and MafA. Biochem. J. 442, 539–550. was handled in the same way (brightness level 75, with blue hues reduced Arany, Z., Wagner, B.K., Ma, Y., Chinsomboon, J., Laznik, D., and Spiegelman, to 24). B.M. (2008). Gene expression-based screening identifies microtubule inhibitors as inducers of PGC-1a and oxidative phosphorylation. Proc. Natl. Acad. Anchorage-Independent Colony Formation Assay Sci. USA 105, 4721–4726. PANC-1 cells were trypsinized and tested for soft agar colony formation Babu, D.A., Deering, T.G., and Mirmira, R.G. (2007). A feat of metabolic using a CytoSelect 96-Well Cell Transformation Assay (Cell Biolabs) with proportions: Pdx1 orchestrates islet development and function in the CyQUANT GR Dye (Molecular Probes) to measure total cellular nucleic maintenance of glucose homeostasis. Mol. Genet. Metab. 92, 43–55. acid levels. Fluorescence was detected with an Analyst HT plate reader (LJL BioSystems) using a 485/520 nm filter set. Barrilleaux, B., and Knoepfler, P.S. (2011). Inducing iPSCs to escape the dish. Cell Stem Cell 9, 103–111.

ChIP-qPCR Assay for Histone Tail Modifications Barski, A., Cuddapah, S., Cui, K., Roh, T.Y., Schones, D.E., Wang, Z., Wei, G., PANC-1 cells were treated with DMSO or 5 mM BRD7552 for 3 days in 15-cm Chepelev, I., and Zhao, K. (2007). High-resolution profiling of histone tissue culture dishes. ChIP was performed using the Magna ChIP G Kit methylations in the human genome. Cell 129, 823–837. (EMD Millipore) according to the manufacturer’s protocol. The sonication pro- Burke, M.D., and Schreiber, S.L. (2004). A planning strategy for diversity- tocol (16 3 10 s on with a 2-min refractory period, duty cycle 10, and output 5) oriented synthesis. Angew. Chem. Int. Ed. Engl. 43, 46–58. was optimized for the Branson Sonifier 250. ChIP was performed for acety- Clemons, P.A., Wilson, J.A., Dancı´k, V., Muller, S., Carrinski, H.A., Wagner, lated histone H3 (Rb anti-acetyl histone H3; EMD Millipore), H3K4me3 B.K., Koehler, A.N., and Schreiber, S.L. (2011). Quantifying structure and (Ms anti-histone H3 trimethyl K4; Abcam), H3K9me3 (Rb anti-histone H3 performance diversity for sets of small molecules comprising small-molecule trimethyl K9; Abcam), and H3K27me3 (Ms anti-histone H3 trimethyl K27; screening collections. Proc. Natl. Acad. Sci. USA 108, 6817–6822. Abcam). Histone H3 (Rb anti-histone H3; Abcam) was used as the positive control, and normal rabbit immunoglobulin G (EMD Millipore) was used as Dandapani, S., and Marcaurelle, L.A. (2010). Grand challenge commentary: the negative control. qPCR was performed as described before. The primer accessing new chemical space for ‘undruggable’ targets. Nat. Chem. Biol. sequences used are listed in Table S2. 6, 861–863. Deramaudt, T.B., Sachdeva, M.M., Wescott, M.P., Chen, Y., Stoffers, D.A., and Rustgi, A.K. (2006). The PDX1 homeodomain transcription factor DNA Methylation Assay negatively regulates the pancreatic ductal cell-specific keratin 19 promoter. Four enriched CpG islands were selected for analysis (Ma et al., 2010). J. Biol. Chem. 281, 38385–38395. PANC-1 cells were treated with DMSO and 5 mM BRD7552 for 9 days. The cells in frozen medium were sent to SeqWright for bisulfite sequencing. Eberharter, A., and Becker, P.B. (2002). Histone acetylation: a switch between repressive and permissive chromatin: second in review series on chromatin dynamics. EMBO Rep. 3, 224–229. Microarray and GSEA Analysis PANC-1 cells were treated with DMSO and 5 mM BRD7552 for 72 hr. mRNA Ferber, S., Halkin, A., Cohen, H., Ber, I., Einav, Y., Goldberg, I., Barshack, I., was extracted from cells using the RNeasy Plus Mini Kit (QIAGEN), and the Seijffers, R., Kopolovic, J., Kaiser, N., and Karasik, A. (2000). Pancreatic and GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix) was used for duodenal homeobox gene 1 induces expression of insulin genes in liver and gene expression profiling. GSEA was performed using the GSEA 2.0 software ameliorates streptozotocin-induced hyperglycemia. Nat. Med. 6, 568–572. package (http://www.broad.mit.edu/tools/software.html). Gao, N., LeLay, J., Vatamaniuk, M.Z., Rieck, S., Friedman, J.R., and Kaestner, K.H. (2008). Dynamic regulation of Pdx1 enhancers by Foxa1 and Foxa2 is SUPPLEMENTAL INFORMATION essential for pancreas development. Genes Dev. 22, 3435–3448. Guo, S., Dai, C., Guo, M., Taylor, B., Harmon, J.S., Sander, M., Robertson, Supplemental Information includes Seupplemental Experimental Procedures, R.P., Powers, A.C., and Stein, R. (2013). Inactivation of specific b cell eight figures, and two tables and can be found with this article online at transcription factors in type 2 diabetes. J. Clin. Invest. 123, 3305–3316. http://dx.doi.org/10.1016/j.chembiol.2013.10.013. Haumaitre, C., Lenoir, O., and Scharfmann, R. (2008). Histone deacetylase inhibitors modify pancreatic cell fate determination and amplify endocrine ACKNOWLEDGMENTS progenitors. Mol. Cell. Biol. 28, 6373–6383. Ichida, J.K., Blanchard, J., Lam, K., Son, E.Y., Chung, J.E., Egli, D., Loh, K.M., We thank Joshua Bittker and Michelle Palmer for helpful suggestions on Carter, A.C., Di Giorgio, F.P., Koszka, K., et al. (2009). A small-molecule assay development and high-throughput screening; Stefan Kubicek, Jacob inhibitor of Tgf-b signaling replaces Sox2 in reprogramming by inducing Hecksher-Sørensen, Dina Fomina-Yadlin, Paul Clemons, Jeremy Duvall, Nanog. Cell Stem Cell 5, 491–503. Andrew Stern, and Alykhan Shamji for helpful discussions and advice; and Kaneto, H., Nakatani, Y., Miyatsuka, T., Matsuoka, T.A., Matsuhisa, M., Hori, Piotr Witkowski and Ryosuke Misawa for providing human pancreatic tissues M., and Yamasaki, Y. (2005). PDX-1/VP16 fusion protein, together with for ductal cell isolation. Funding for this project was provided by the Juvenile NeuroD or Ngn3, markedly induces insulin gene transcription and ameliorates Diabetes Research Foundation and by National Institute for General Medical glucose tolerance. Diabetes 54, 1009–1022. Science grant GM38627 (to B.K.W. and S.L.S., respectively). S.L.S. is an investigator at the Howard Hughes Medical Institute. Karnieli, O., Izhar-Prato, Y., Bulvik, S., and Efrat, S. (2007). Generation of insulin-producing cells from human bone marrow mesenchymal stem cells Received: June 22, 2013 by genetic manipulation. Stem Cells 25, 2837–2844. Revised: September 30, 2013 Kawamori, D., Kajimoto, Y., Kaneto, H., Umayahara, Y., Fujitani, Y., Miyatsuka, Accepted: October 9, 2013 T., Watada, H., Leibiger, I.B., Yamasaki, Y., and Hori, M. (2003). Oxidative Published: November 27, 2013 stress induces nucleo-cytoplasmic translocation of pancreatic transcription

factor PDX-1 through activation of c-Jun NH2-terminal kinase. Diabetes 52, Payne, D.J., Gwynn, M.N., Holmes, D.J., and Pompliano, D.L. (2007). Drugs 2896–2904. for bad bugs: confronting the challenges of antibacterial discovery. Nat. Koch, C.M., Andrews, R.M., Flicek, P., Dillon, S.C., Karao¨ z, U., Clelland, G.K., Rev. Drug Discov. 6, 29–40. Wilcox, S., Beare, D.M., Fowler, J.C., Couttet, P., et al. (2007). The landscape Schreiber, S.L. (2009). Organic chemistry: molecular diversity by design. of histone modifications across 1% of the human genome in five human cell Nature 457, 153–154. lines. Genome Res. 17, 691–707. Staerk, J., Lyssiotis, C.A., Medeiro, L.A., Bollong, M., Foreman, R.K., Zhu, S., Kojima, H., Nakamura, T., Fujita, Y., Kishi, A., Fujimiya, M., Yamada, S., Kudo, Garcia, M., Gao, Q., Bouchez, L.C., Lairson, L.L., et al. (2011). Pan-Src family M., Nishio, Y., Maegawa, H., Haneda, M., et al. (2002). Combined expression kinase inhibitors replace Sox2 during the direct reprogramming of somatic of pancreatic duodenal homeobox 1 and islet factor 1 induces immature cells. Angew. Chem. Int. Ed. Engl. 50, 5734–5736. enterocytes to produce insulin. Diabetes 51, 1398–1408. Stoffers, D.A., Zinkin, N.T., Stanojevic, V., Clarke, W.L., and Habener, J.F. Kubo, A., Stull, R., Takeuchi, M., Bonham, K., Gouon-Evans, V., Sho, M., (1997). Pancreatic agenesis attributable to a single nucleotide deletion in the Iwano, M., Saito, Y., Keller, G., and Snodgrass, R. (2011). Pdx1 and Ngn3 human IPF1 gene coding sequence. Nat. Genet. 15, 106–110. overexpression enhances pancreatic differentiation of mouse ES cell-derived Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., endoderm population. PLoS ONE 6, e24058. Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and Lefebvre, B., Belaich, S., Longue, J., Vandewalle, B., Oberholzer, J., Gmyr, V., Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based 0 Pattou, F., and Kerr-Conte, J. (2010). 5 -AZA induces Ngn3 expression and approach for interpreting genome-wide expression profiles. Proc. Natl. endocrine differentiation in the PANC-1 human ductal cell line. Biochem. Acad. Sci. USA 102, 15545–15550. Biophys. Res. Commun. 391, 305–309. Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells Lyssiotis, C.A., Foreman, R.K., Staerk, J., Garcia, M., Mathur, D., Markoulaki, from mouse embryonic and adult fibroblast cultures by defined factors. Cell S., Hanna, J., Lairson, L.L., Charette, B.D., Bouchez, L.C., et al. (2009). 126, 663–676. Reprogramming of murine fibroblasts to induced pluripotent stem cells with Walpita, D., Hasaka, T., Spoonamore, J., Vetere, A., Takane, K.K., Fomina- chemical complementation of Klf4. Proc. Natl. Acad. Sci. USA 106, 8912– Yadlin, D., Fiaschi-Taesch, N., Shamji, A., Clemons, P.A., Stewart, A.F., 8917. et al. (2012). A human islet cell culture system for high-throughput screening. Lysy, P.A., Weir, G.C., and Bonner-Weir, S. (2012). Concise review: pancreas J. Biomol. Screen. 17, 509–518. regeneration: recent advances and perspectives. Stem Cells Transl. Med. 1, 150–159. Wu, Y., Minger, S.L., Sawyer, G.J., Fabre, J.W., Persaud, S.J., and Jones, P.M. (2007). Pancreatic duodenal homeobox 1 expression is insufficient to Ma, J., Wang, J.D., Zhang, W.J., Zou, B., Chen, W.J., Lam, C.S., Chen, M.H., transdifferentiate liver cells into insulin-secreting cells. Pancreas 34, 272–275. Pang, R., Tan, V.P., Hung, I.F., et al. (2010). Promoter hypermethylation and histone hypoacetylation contribute to pancreatic-duodenal homeobox 1 Yamada, S., Kojima, H., Fujimiya, M., Nakamura, T., Kashiwagi, A., and silencing in gastric cancer. Carcinogenesis 31, 1552–1560. Kikkawa, R. (2001). Differentiation of immature enterocytes into enteroendo- crine cells by Pdx1 overexpression. Am. J. Physiol. Gastrointest. Liver Macfarlane, W.M., McKinnon, C.M., Felton-Edkins, Z.A., Cragg, H., James, Physiol. 281, G229–G236. R.F., and Docherty, K. (1999). Glucose stimulates translocation of the homeodomain transcription factor PDX1 from the cytoplasm to the nucleus Yang, L., Li, S., Hatch, H., Ahrens, K., Cornelius, J.G., Petersen, B.E., and in pancreatic b-cells. J. Biol. Chem. 274, 1011–1016. Peck, A.B. (2002). In vitro trans-differentiation of adult hepatic stem cells into pancreatic endocrine hormone-producing cells. Proc. Natl. Acad. Sci. USA McKinnon, C.M., and Docherty, K. (2001). Pancreatic duodenal homeobox-1, 99, 8078–8083. PDX-1, a major regulator of b cell identity and function. Diabetologia 44, 1203– 1214. Yatoh, S., Dodge, R., Akashi, T., Omer, A., Sharma, A., Weir, G.C., and Motoyama, H., Ogawa, S., Kubo, A., Miwa, S., Nakayama, J., Tagawa, Y., and Bonner-Weir, S. (2007). Differentiation of affinity-purified human pancreatic b Miyagawa, S. (2009). In vitro reprogramming of adult hepatocytes into insulin- duct cells to -cells. Diabetes 56, 1802–1809. producing cells without viral vectors. Biochem. Biophys. Res. Commun. 385, Zhang, J.H., Chung, T.D., and Oldenburg, K.R. (1999). A simple statistical 123–128. parameter for use in evaluation and validation of high throughput screening Oliver-Krasinski, J.M., Kasner, M.T., Yang, J., Crutchlow, M.F., Rustgi, A.K., assays. J. Biomol. Screen. 4, 67–73. Kaestner, K.H., and Stoffers, D.A. (2009). The diabetes gene Pdx1 regulates Zhou, Q., Brown, J., Kanarek, A., Rajagopal, J., and Melton, D.A. (2008). In vivo the transcriptional network of pancreatic endocrine progenitor cells in mice. reprogramming of adult pancreatic exocrine cells to b-cells. Nature 455, J. Clin. Invest. 119, 1888–1898. 627–632.