Combination of Sulindac and Bexarotene for Prevention of Intestinal Carcinogenesis in Familial Adenomatous Polyposis

Colorectal cancer remains the third leading cause of cancer-related deaths affecting both men and women (1). The lifetime risk of developing colorectal cancer in the general population nears 5%; however, patients with hereditary cancer syndromes have a 10- to 20-fold higher risk (2). Familial adenomatous polyposis (FAP) is an autosomal dominant hereditary cancer syndrome associated with pathogenic germline mutations in the Adenomatous polyposis coli (APC) gene (3), which is commonly mutated in colorectal cancer on a somatic basis (4). Patients with FAP develop hundreds of adenomatous polyps throughout the gastrointestinal tract, including the duodenum, small intestine, colon, and rectum, which develop into tumors by the third and fourth decade of life. The current standard of care for FAP management includes endoscopic polypectomy. When polyp burden cannot be controlled endoscopically, patients typically seek a prophylactic colectomy with the creation of an intestinal pouch or ileo-rectal anastomosis; however, rectal polyps often present more frequently after prophylactic surgery (5).

Patients with FAP require prophylactic, risk-reducing procedures given the lack of potency and long-term efficacy of most single-agent therapies to control polyposis. For decades, the use of NSAIDs has provided promising benefits as a colorectal cancer chemopreventive agent due to its effects on the inhibition of protumorigenic inflammation. Several studies have demonstrated the clinical benefits of NSAIDs to reduce colorectal polyposis in FAP via the inhibition of the inflammatory cyclooxygenase 2 (COX2) pathway (6, 7). Preclinical trials using NSAIDs and COX2 inhibitors revealed a reduction in polyp burden in Apc^Min/+ mice, which is considered the best model for small intestine polyposis (8, 9). Therefore, combinations of different anti-inflammatory drugs present potential opportunities for cancer interception strategies to target an array of mRNAs and miRNAs playing pivotal roles in inflammation and signal transduction pathways in colorectal carcinogenesis (10). The main focus of the work presented in this manuscript is to assess the activity of the combination of sulindac and bexarotene. Sulindac, a nonselective COX inhibitor, was the first NSAID shown to decrease the number and size of polyps in patients with FAP (11). Bexarotene, a synthetic rexinoid, has been reported to reduce inflammation via COX2 inhibition within breast cancer and FAP animal models (12, 13). Furthermore, bexarotene reduced tumor progression in non–small cell lung cancer and skin squamous cell carcinoma animal models via the inhibition of the EGFR and WNT signaling pathways (14, 15).

Recent studies have shown that miRNAs are key drivers for the initiation and progression of COX, EGFR, and WNT-mediated carcinogenesis, and thus have a prognostic value of disease outcome (16, 17). Among them, miR-20a, miR-21, miR-103a, miR-106, miR-143, and miR-215 are predictive biomarkers for stage II colorectal cancer (18). It has been reported that miR-135a/b miRNAs that are overexpressed in colorectal cancer inhibit APC expression and activate WNT signaling, whereas, miR-21, a highly expressed and the most prominent biomarker for colorectal cancer, promotes Prostaglandin E2 (PGE2) levels by inhibiting the expression of the tumor suppressor, PDCD4 (19). In addition, miRNA modulators such as miR-7 or miR-133b have improved the potency of anti-EGFR therapies in colon cancer cells (20, 21). Yet, the action of these miRNAs as interventions to inhibit oncogenic miRNA in colorectal cancer needs further investigation.

Overall, these studies suggest that the combination of sulindac and bexarotene may improve cancer interception at early stages of carcinogenesis through inhibition of COX-2, WNT, and EGFR signaling pathways, and miRNAs. Therefore, the overarching goal of this study was to examine the use of an NSAID and rexinoid combination as a novel chemopreventive strategy for colonic polyps in the context of FAP and to understand the molecular mechanisms underlying the preventive activity of these drugs in colon cancer mouse models.

Written informed consent was obtained from all study participants under The University of Texas MD Anderson Cancer Center (MDACC) Institutional Review Board (IRB) protocol (IRB #PA12–0327) and the U.S. Common Rule regulations. All animal experiments were conducted in compliance with the NIH guidelines for animal research and approved by MDACC Institutional Animal Care and Use Committee (Protocol #488-RN02).

RNA sequencing was performed on 36 samples (18 matched pairs of colorectal polyps and adjacent normal mucosa) from 18 patients with FAP followed in MDACC for their standard of care surveillance (Supplementary Tables S1 and S2). Sample preparation, library construction, and sequencing were performed at MDACC Advanced Technology Genomics Core Facility using the Illumina HiSeq2000 platform. All bioinformatic analyses for differential gene expression were performed as described previously (22). The raw sequencing FASTQ files have been used in previous submissions and can be accessed in GSE106500 (9 samples), GSE88945 (1 sample), GSE153385 (22 FAP samples), and GSE156172 (4 FAP samples). To explore and visualize sample distribution in relationship with tissue type, a principal component analysis (PCA) was performed using data normalized by library size and transformed by variance-stabilizing transformation (VST). For gene set enrichment analysis (GSEA), genes were preranked by log2-FoldChange and analyzed with Bioconductor R package fgsea (23), separately on HALLMARK, KEGG, and REACTOME gene sets obtained from the Broad Institute Molecular Signatures Database v6.0, as well as colorectal cancer related pathway gene sets obtained from Guinney and colleagues (24). Significant pathways were selected based on BH-adjusted P-value≤0.05 and displayed in bar plots ordered by normalized enrichment score (NES).

Human colorectal cancer cell lines HCT116 and HT29 (ATCC CCL24 and HTB38) were grown in DMEM with 4.5 g/L glucose (Mediatech) supplemented with 5% FBS (Sigma-Aldrich), and 1% penicillin/streptomycin (HyClone) at 37°C and 5% CO₂ in an incubator. All cell lines were authenticated upon acquisition by the Cytogenetics and Cell Authentication core at MDACC, and periodically tested for mycoplasma contamination.

Crypt isolation and organoid cultures were prepared from human colorectal polyps and normal mucosa FAP samples, collected from the Endoscopy Center at MDACC, as described previously (25) with modifications. In brief, samples were cut into 1 mm³ pieces, washed vigorously with cold PBS containing 0.1% Amphotericin B, and placed in cold PBS with 2.5 mmol/L EDTA for 30 minutes at 4°C with gentle agitation. The released crypts were collected, and the remaining fragments were further dissociated for 1 hour. Crypts collected from both fractions were pelleted at 1,500 rpm for 5 minutes at 4°C, resuspended in Matrigel, and then dispensed onto 24-well plates in 50 μL droplets. After Matrigel polymerization, crypts were overlaid with organoid media, as described: advanced DMEM/F12 was supplemented with penicillin/streptomycin, 10 mmol/L HEPES, 2 mmol/L GlutaMAX, 1 × B27 (Invitrogen), 1 × N2 Supplement (Invitrogen), 100 μg/mL Primocin (Invitrogen), 10 mmol/L PGE2 (Selleckchem), 10 mmol/L Nicotinamide (Sigma-Aldrich), 10 nmol/L gastrin I (Sigma-Aldrich), and 1 mmol/L N-acetylcysteine (Wako), 50 ng/mL mouse recombinant EGF (Life Technologies), 100 ng/mL mouse recombinant Noggin (Peprotech), 0.1% 250 μg/mL Amphotericin B (Gibco), 10% R-spondin-1 conditioned medium (26), 50% Wnt-3A conditioned medium (27), 500 nmol/L A83–01 (Tocris), and 10 μmol/L SB202190 (Sigma-Aldrich). A 10 μmol/L Rock inhibitor, Y-27632 (Sigma-Aldrich), was added to the media upon initial passage and freezing of organoids. Organoids were cultured with fresh media every 2 to 3 days.

Confluent HCT116, HT29 cells, and FAP organoids were plated in flat-bottom, 48-well tissue culture plates in media supplemented with sulindac and/or bexarotene. After 48 hours incubation, the cell viability was assessed using CellTiter-Glo 3D Cell Viability Assay (Promega).

Mice were produced by breeding Apc^LoxP/+ mice (The Jackson Laboratory; stock #009045) with Cdx2P-CreER^T2 mice (The Jackson Laboratory; stock# 022390). Six-week-old gender-matched transgenic Apc^LoxP/+-Cdx2 mice were injected with tamoxifen (20 mg/mL, 100 μL), resulting in deletion of all 15 exons, as described previously (28). Mice were treated with drugs through oral gavage for 6 weeks. Apc^Min/+ mice were randomized to vehicle/sesame oil (control), 7.5 mg/kg/day of sulindac (50 ppm in 100 μL), 15 mg/kg/day of bexarotene (100 ppm in 100 μL); Apc^LoxP/+ mice to 7.5 mg/kg/day of each sulindac and bexarotene (50 ppm in 100 μL), and both models to the combination of sulindac plus bexarotene (each 7.5 mg/kg/day). Both sulindac and bexarotene doses were based on 50 mg/day Human Equivalent Dose, where bexarotene is typically administered clinically to cancer patients at 300 mg/m² per day and sulindac clinically at 200 to 400 mg/day (8, 13). Bodyweight was taken every week with daily supervision of animal behavior. At the completion of the trial, mice were humanely sacrificed, and a final polyp assessment was obtained by longitudinally opening the colon and manually counting and measuring polyps under a dissection microscope.

The mucosal lining of Apc^LoxP/+-Cdx2 mice (n = 7 per group) and Apc^Min/+ mice (n = 3 per group) was stripped for subsequent analysis. The tissue was homogenized in radioimmune precipitation assay buffer with complete Mini protease and PhosSTOP phosphatase inhibitor (Roche Diagnostics). The tissue lysates were separated on polyacrylamide gels (Bio-Rad) for Western blot analysis. The following primary antibodies were used: anti-COX2, anti-Bcl-2, anti-β-catenin, anti-EGFR, anti-p-EGFR, and anti-Cyclin D1 (Cell Signaling Technology), anti-p-NF-κB p65, and anti-RXRα (Santa Cruz Biotechnology). Goat anti-mouse (1:6,000) or goat anti-rabbit (1:6,000) secondary antibodies were used (Abcam). The membranes were detected by the Enhanced Chemiluminescence Detection Kit (GE Healthcare)

Total RNA was extracted from colonic tissues using Direct-zol RNA MiniPrep Plus (Zymo Research). Complementary DNA was generated using the High Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) and used for qRT-PCR, using Fast SYBR Green Master Mix (Thermo Fisher Scientific) and the gene-specific primer sets (Supplementary Table S3). miRNA was isolated from mouse colon/small intestine (normal and polyp tissue) and blood using the miRNeasy Extraction Kit (Qiagen). miRNA was used to measure miRNA expression using SYBR green-based qRT-PCR relative to a universal qPCR primer as ROX reference dye. The primer sequence is included in Supplementary Table S4.

The mean ± SEM was reported for the data analysis using GraphPad Prism (GraphPad Software, Inc.). Statistical analysis was performed with the Student t test when only two value sets were compared, and with a 2-way ANOVA followed by Tukey multiple comparisons test when the data involved three or more groups. P-value<0.05 was considered statistically significant.

In a first step, RNA-seq data were utilized to identify potential targets for colorectal prevention in FAP. To enhance the power to detect differentially expressed genes, colorectal polyps and matching normal mucosa samples within each patient in a cohort of patients with FAP were analyzed. Figure 1A shows hierarchical clusters that clearly demarcate segregation among the normal colonic mucosa from the corresponding adenomatous polyps, thus confirming the results from the PCA (Supplementary Fig. S1). Among the most differentially expressed genes (DEG) in FAP polyps, 12 of the top 48 were associated with the regulation of the immune system, inflammation, and canonical WNT/β-catenin pathways (Fig. 1B). Of the top immune-related genes, CLDN2, MMP7, MMP13, and CLDN1 displayed a log2FC of 6.26, 5.37, 3.13, and 3.08, respectively, in FAP polyp samples when compared with the normal colonic mucosa. GSEA detected activation of CRC-related pathways such as WNT, Myc, E2F, mTOR, and cell cycle (Fig. 1C). Notably, the aberrant activation of the classical WNT/β-catenin signaling pathway was expected in FAP specimens due to the presence of germline and somatic mutations in APC. Furthermore, the highly expressed genes in FAP mucosa indicated the activation of several critical pathways related to inflammatory responses mediated by COX2 and RXR gene signatures. Consistently, the expression levels of RXRα and RARRES2 genes were decreased by log2FC 2.2 and log2FC 1.45, respectively, in adenoma when compared with normal colonic mucosa of the patients with FAP. Therefore, CLDN1, MMP7, and MMP13 were selected as candidate biomarkers of the chemopreventive effect of sulindac and bexarotene, which targets COX, RXR, and inflammatory pathways, in a preclinical trial using mouse models of intestinal and colorectal polyposis.

The effect of sulindac and bexarotene activity on cell viability was assessed to determine the drug combination's efficiency and its synergistic effect in colorectal cancer cell lines, HT29 and HCT116. Cytotoxic analysis of sulindac treatment showed IC₅₀ values of 1,951 ± 86 μmol/L and 1,718 ± 250 μmol/L for HT29 and HCT116, respectively (Supplementary Fig. S2). Conversely, bexarotene treatment showed IC₅₀ values of 121 ± 22 μmol/L and 96 ± 8 μmol/L for HT29 and HCT116 cells, respectively. Combinatorial analysis for 48 hours demonstrated that HCT116 cells sustained 49.5% survival at the combination of 40 μmol/L bexarotene and 900 μmol/L sulindac with a CI of 0.62, and HT29 cells sustained 62.1% survival with CI of 0.577 (Supplementary Fig. S3), thus indicating that the HCT116 (microsatellite instable, MSI) cell line was more sensitive than HT29 (microsatellite stable, MSS) to both drug treatments. However, drug sensitivity could be also driven by β-catenin (CTNNB1) mutations in HCT116 cells when compared with HT29 cells, which are mutant in APC. Further determination of factors contributing to drug sensitivity may be necessary for future studies. Nevertheless, these data led us to hypothesize that the synergistic effect of a low-dose combination of sulindac and bexarotene will lead to a more clinically favorable scenario with long-term tolerance compared with high-dose therapies that are more toxic, while maintaining potency for both MSI and MSS colorectal models.

The drug combination was tested ex vivo using an organoid platform, which mimics the in vivo structure and physiology of intact parent tissue, generated from colorectal adenomas (polyp-organoid) and normal mucosa (normal-organoid) from patients with FAP (Fig. 2). Cell viability analysis showed that sulindac treatment was slightly more effective on polyp-organoids than normal-organoids. Cell viability of polyp-organoids was 60% ± 1.7 with IC₅₀ values of sulindac at 348.6 μmol/L, whereas normal-organoids was 69.5% ± 8.7 with IC₅₀ values of sulindac at 383.1 μmol/L (Fig. 2A). The cell viability analysis of bexarotene treatment showed 51.3% ± 5.6 in the polyp-organoids with IC₅₀ values of bexarotene at 29 μmol/L, whereas normal-organoids showed 82% ± 8.2 with IC₅₀ values of bexarotene at 58.5 μmol/L (Fig. 2B). The bexarotene cell viability data of normal and polyp samples suggest that bexarotene is more effective on polyp-organoids than normal-organoids. When a combination of 300 μmol/L of sulindac and 40 μmol/L of bexarotene was tested together, cell viability of polyp-organoids was significantly decreased (57.7% ± 9) when compared with normal-organoids (81.6% ± 5.7, Fig. 2C). Figure 2D depicts colonic organoids from patients with FAP with reduced growth of polyp-organoids compared with normal-organoids treated with 300 μmol/L sulindac, 40 μmol/L bexarotene, and the drug combination of 300 μmol/L sulindac with 40 μmol/L bexarotene. Overall, these data provide a rationale for a clinically-relevant therapeutic opportunity of combinatorial drug treatment as a chemopreventive strategy to reduce the growth of colonic polyps in patients with FAP.

The classical Apc^Min/+ model spontaneously develops polyps primarily within the small intestine and provides a durable model for studying FAP-associated duodenal polyposis (8, 29). Conversely, the Apc^LoxP/+-Cdx2 mouse model better recapitulates FAP-associated colonic polyposis of the large intestine in humans.

Apc^Min/+ mice were randomized into four different groups and treated for a total of 6 weeks. Both polyp multiplicity and polyp size were assessed at the completion of treatment to determine the effect of drug intervention. Sulindac alone caused the most significant reduction (75%) in small intestine polyp count (9 ± 1.2) compared with untreated (control) mice (36.7 ± 4, Fig. 3A). Bexarotene alone also reduced polyp count (23 ± 3) by 40% compared with control mice; however, to a lesser extent, when compared with sulindac. The combination of sulindac and bexarotene provided a moderate effect (14.2 ± 1) with a 60% reduction in polyp burden. The effect of bexarotene appeared comparable with sulindac in reducing polyps of the large intestine derived from Apc^Min/+ mice (Fig. 3B); however, this trend was not observed in the Apc^LoxP/+-Cdx2 mice when compared with single-agent sulindac. Further, sulindac significantly reduced the prevalence of polyps greater than 2 mm in size, although the effect of bexarotene was moderate in reducing polyps in the small intestine (Fig. 3C). Nonetheless, these results support that, although single-agent sulindac demonstrated the most beneficial chemopreventive agent for reducing the small intestine polyp burden, the combination of sulindac and bexarotene might be a better chemopreventive strategy for simultaneous inhibition of polyposis in both the small and large intestine.

Following tamoxifen injection to conditionally inactivate the Apc gene in the large intestine, Apc^LoxP/+-Cdx2 mice were randomized for drug treatment similar to Apc^Min/+ mice. Assessment of drug efficacy in Apc^LoxP/+-Cdx2 mice revealed that the combination of sulindac and bexarotene afforded the most significant reduction in polyps (0.2 ± 0.2) of the large intestine when compared with the control group (3.6 ± 0.2) However, both sulindac and bexarotene administered as single-agents provided significant polyp reduction (Fig. 3B and D). Furthermore, single-agent sulindac treatment appeared to be more effective in preventing polyposis within the small intestine of the Apc^LoxP/+-Cdx2 mice (Fig. 3A and C). These results from both mouse models, the Apc^Min/+ and Apc^LoxP/+-Cdx2 mice, indicate that single-agent sulindac administered at 7.5 mg/kg/day protects against small intestinal polyposis, and single-agent bexarotene largely prevents large intestinal polyposis development. Furthermore, the combination of sulindac and bexarotene appeared to reduce polyposis across the gastrointestinal tract, both small and large intestine, and was well tolerated by both mouse models, as indicated by no adverse changes in animal body weight (Supplementary Figs. S4A–S4D).

The molecular effect of the combination of sulindac and bexarotene was evaluated to determine the modulation of key regulators involved in colonic inflammation and tumorigenesis. Protein levels of COX-2 decreased more in the combination of sulindac and bexarotene than in either sulindac or bexarotene or the control animal groups (Fig. 4A, left). Because retinoid X receptor α (RXRα) is expressed at very low levels in colonic tumors (13), the data demonstrated that levels of RXRα were significantly increased by single-agent bexarotene and the combination of sulindac and bexarotene compared with single-agent sulindac and control groups. Levels of NF-κB were remarkably reduced in the sulindac and bexarotene combination groups. The levels of cyclin D1 protein, required for progression through the G₁ phase of the cell cycle, were significantly decreased within the sulindac and bexarotene combination group as compared with the other groups. The protein expression of Bcl-2, an anti-apoptosis regulator, shown to be increased by COX2 expression (30), was significantly reduced when the Apc^LoxP/+-Cdx2 mice were treated with the dual-agent combination. These results suggest that the combination of sulindac and bexarotene reduced polyp development and tumorigenesis by suppressing inflammation and cell-cycle arrest via decreased protein expression of COX-2, NF-κB, and cyclin D1, as well as by increasing apoptosis via Bcl-2 inhibition within Apc^LoxP/+-Cdx2 mice. In addition to inflammatory and cell-cycle regulators, a significant decrease in the levels of EGFR and phosphorylated EGFR (pEGFR) with bexarotene alone and with the sulindac and bexarotene combination treatment groups were observed (Fig. 4A, middle). Moreover, the levels of activated β-catenin also decreased with bexarotene alone and combined sulindac and bexarotene treatment. Thus, the combination of sulindac and bexarotene decreased the expression of COX-2, p-NF-κB, Bcl-2, EGFR, and activated β-catenin levels. Although single agents had significant effects, the dual-agent regimen appeared to provide broader effects across all biomarkers assessed compared with either single agent alone.

Finally, to assess the molecular effect of drug intervention in the Apc^Min/+ model, the expression of two key targets of sulindac and bexarotene was examined: COX-2 and RXR-α, respectively. The data indicated a significant reduction in COX-2 expression of mice treated with sulindac, which was further reduced by the drug combination (Fig. 4A, right). Similar to the Apc^LoxP/+-Cdx2 model, increased expression of RXR-α in Apc^Min/+ mice treated with the drug combination was observed, as expected with drug synergism. Taken together, the combination of sulindac and bexarotene reduced polyp development and tumorigenesis by inhibiting the levels of COX-2, EGFR, and β-Catenin within the Apc^LoxP/+-Cdx2 mice. These results were also confirmed in the Apc^Min/+ model, as shown by decreased COX-2 expression and increased RXR-α expression.

The effect of sulindac and bexarotene treatment was assessed on the expression levels of genes selected based on expression results observed in human FAP specimens, and that could be used as predictive biomarkers of drug activity. The results indicate that transcript levels of Cldn1, Mmp7, and Mmp13 were strongly reduced in the polyps of Apc^LoxP/+-Cdx2 mice that received sulindac and bexarotene alone or in combination when compared with transcript levels in the polyps from untreated control mice (Fig. 4B). The expression of Cldn1, Mmp7, and Mmp13 was also measured in the Apc^Min/+ model (Fig. 4C); however, a significant decrease in expression was only observed for Mmp7 in Apc^Min/+ mice treated with bexarotene.

To further identify a potential preventive biomarker of colorectal cancer, this study focused on the oncomiRNAs that are known to play key roles in colon tumorigenesis. Among oncomiRNAs, miR-21, miR-26a, miR-29, miR-30, and miR-103 are associated with colorectal cancer progression, and specifically, miR-21 has been reported to be highly upregulated and associated with tumor size and poor overall survival (16, 31). Thus, the expression levels of these miRNAs in a cohort of FAP patients, Apc^Min/+ mice, and Apc^LoxP/+-Cdx2 mice, as well as the effect of sulindac and bexarotene on the expression levels of these miRNAs were examined. The data presented reveals miR-21 was highly expressed in the serum (Fig. 5A) and polyp samples (Fig. 5B) of patients with FAP with more than 50 polyps. Therefore, these data suggest that miR-21 expression may be correlated with polyp occurrence in patients with FAP, thus highlighting its prognostic and clinical significance of FAP-mediated colon tumorigenesis (Fig. 5A). Consistent with this notion, miR-21 levels were also strongly elevated in polyps of the Apc^Min/+ mice when compared with levels in WT (C57BL/J6) mice (Fig. 5C). To ascertain whether miR-21 is a preventive biomarker of FAP polyposis upon drug intervention, miR-21 levels were evaluated in treated Apc^Min/+ and Apc^LoxP/+-Cdx2 mice. As shown in Fig. 5C, both sulindac and the combination of sulindac and bexarotene significantly reduced the expression levels of miR-21 in the blood serum of mice receiving the intervention. However, there were no significant changes in the expression levels of miR-26a, miR-29, miR-30, and miR-103 between normal mucosa and polyps from a cohort of patients with FAP (Supplementary Fig. S5). In the Apc^LoxP/+-Cdx2 model, miRNA extracted from mucosa was used for analysis. As shown in Fig. 5D, miR-21 was significantly reduced in all three treatment groups compared with untreated control mice. These results from both mouse models are indicative of miR-21 as a potential biomarker of polyp burden in patients already diagnosed with FAP-associated polyposis.

Using whole transcriptome analysis, differentially expressed genes in premalignant colorectal lesions (adenomas) from a cohort of patients with FAP were identified as potential targets for cancer interception approaches. First, patient-derived colonic organoids were generated from colorectal polyps and normal tissue of patients with FAP to test the dose-synergism effect of sulindac and bexarotene on cell viability (32). Then, two colorectal cancer mouse models, Apc^Min/+ and Apc^LoxP/+-Cdx2, were used to validate the effects of sulindac and bexarotene. Both mouse models afforded the successful evaluation of sulindac and bexarotene efficacy in reducing polyposis and the levels of inflammatory markers and colorectal cancer drivers. In addition, the levels of miRNA21, highly elevated in a cohort of patients with FAP, were remarkably reduced in both mouse models upon drug intervention. These preclinical findings establish the framework for developing a novel cancer interception strategy of dual administration of sulindac and bexarotene for colonic tumorigenesis in FAP.

The development of intestinal neoplasms is due to disruption of APC and activation of EGFR signaling that promotes the expression of COX2. Subsequently, proinflammatory and protumorigenic responses elevate the levels of PGE2, which results in stimulation of growth and angiogenesis and inhibition of cellular apoptosis via activation of NF-κB and Wnt signaling (33–35). It has been shown that the protumorigenic action of COX-2 is mediated through its role in the production of PGE2 via β-catenin/TCF/LEF pathway activation, which is significantly increased during FAP colorectal tumorigenesis (35, 36). Human FAP trials and mouse model studies showed that a combination of COX and EGFR inhibitors are successful in reducing intestinal adenoma development (10, 29, 37). Furthermore, bexarotene treatment decreased inflammation in breast cancer and FAP animal models (12, 13), most likely through inhibition of EGFR and WNT signaling. Thus, the COX-2 signaling pathway is an ideal druggable target for colorectal cancer prevention. This study presents the combination of sulindac and bexarotene as an effective strategy to reduce COX-2 expression in a FAP mouse model that mimics colonic tumorigenesis. Sulindac more effectively reduced polyposis in Apc^Min/+ mice, which is consistent with previous reports (13, 38). However, another study reported that Targetrin (bexarotene) failed to inhibit small intestinal neoplasms in Apc^Min/+ mice based on histopathologic analyses (39). This combination, in turn, played a key role in reducing the cell viability of FAP polyp-derived colonic organoids as well as successfully reducing polyp growth within the small and large intestine of the FAP animal models.

Clinical studies of patients with FAP treated with NSAIDs to reduce tissue PGE2 levels have demonstrated effective inhibition of polyp development (40, 41). Sulindac is a conventional NSAID that has been shown to reduce polyposis (both in number and size of polyps) via nonselective COX inhibition (8, 9, 42, 43). The results presented in this manuscript indicate that sulindac played a significant role in reducing inflammation via COX-2 inhibition as well as decreasing NF-κB protein levels within the large intestine of Apc^LoxP/+-Cdx2 mice and the small intestine of Apc^Min/+ mice, which might play a primary role in the reduction of polyp burden within this preclinical study.

Retinoic acid plays a significant role in suppressing COX-2 expression levels through activation of the C/EBP-B transcription factor pathway (34). RXRα is a retinoid acid receptor and a potential colon cancer prevention target based on its decreased expression in colonic tumors (38, 44). Anti-inflammatory agents, such as sulindac and bexarotene, are ligands for RXR and inhibit chronic inflammation (45). The synthetic retinoid, bexarotene, is a highly selective RXR agonist responsible for restricting intestinal neoplasia in Apc^Min/+ mice (13), which is confirmed by our data. Further, inhibition of EGFR and WNT/β-catenin in the large intestine of Apc^LoxP/+-Cdx2 mice by bexarotene was observed, confirming previous studies using different mouse models (14, 46). Thus, targeting different pathways simultaneously with a combination of chemopreventive drugs such as sulindac against the COX2 pathway and bexarotene against RXR signaling appeared to reduce intestinal polyposis at different segments of the gastrointestinal tract in two animal models of FAP, and further clinical investigation in humans is warranted.

Matrix metalloproteinase genes, including MMP7 and MMP13, are known for cell adhesion and migration in many cancers and upregulated during intestinal inflammatory bowel disease and the growth and migration of colorectal carcinoma cells. Furthermore, Mmp7 is a known downstream target of β-catenin-mediated TCF signaling and reported to be associated with distant metastasis and poor outcomes in early colorectal cancer (47). Claudins are a family of tight junction proteins that regulate epithelial barrier functions in the intestine (48). MMP7, MMP13, FOXQ1, CLDN1, and CLDN2 genes have been previously reported to be coexpressed at high levels in colonic tumors with high levels of COX2 (49). To gain mechanistic insights regarding inhibition of FAP colonic polyposis, the results presented here showed that sulindac and bexarotene strongly reduced the expression levels of the potential colon cancer marker genes, Cldn1, Cldn2, Mmp7, and Mmp13, within polyps of the large intestine of Apc^LoxP/+-Cdx2 mice. However, these results were not consistently observed in the Apc^Min/+ model, which may be attributed to significant variation of expression observed within the control group. The expression of these genes was increased in a cohort of FAP patients, based on RNA-seq analysis, which is consistent with other findings (47, 50). However, none of the potential preventive biomarkers (Cldn1, Cldn2, Mmp7, and Mmp13) were shown to have a significant difference in gene expression when compared with the combination of sulindac and bexarotene to each study group.

Previous reports demonstrate that NSAIDs, like sulindac, reduce FAP polyp burden at the beginning of drug treatment (10). However, this study did not examine the efficiency of the drug combination over an extended period. Thus, a further study is needed to validate the drug combination effectiveness in reducing FAP colonic polyposis over an extended period of time.

Another important observation in this study is the downregulation of miR-21, a potential prognostic biomarker for colon cancer, in polyps of the small and large intestine of Apc^Min/+ and Apc^LoxP/+-Cdx2 mice by sulindac and a combination of sulindac and bexarotene. Taken together, it is conceivable that these NSAIDs inhibit the expression of not only oncogenes, but also miRNAs, including miR-21 as well as other signaling pathways, which consequently results in inhibition of polyposis and protection from initiation of colon cancer.

In summary, the RNA-seq analysis of FAP polyps indicated an increase of immune-related genes as compared with normal mucosa samples. The combination of sulindac and bexarotene provided synergism and an effective method of reducing cell viability to colon cancer cells and FAP patient-derived colonic organoids. Finally, the observed reduction in colonic polyp burden and size, together with the inhibition of COX2, EGFR, and WNT signaling, as well as the increase of RXRα activation, provides insights into the molecular mechanism of colonic polyp regression in patients with FAP.

M.W. Taggart reports grants from NIH during the conduct of the study. E. Vilar reports grants from NIH/NCI, SWOG/Hope Foundation, and other support from Feinberg Family Foundation during the conduct of the study; grants and personal fees from Janssen Research and Development and personal fees from Recursion Pharma outside the submitted work. No disclosures were reported by the other authors.

C.M. Bowen: Data curation, formal analysis, visualization, writing–original draft, writing–review and editing. L. Walter: Conceptualization, data curation, investigation, methodology, writing–original draft. E. Borras: Conceptualization, data curation, formal analysis, investigation, methodology. W. Wu: Software, formal analysis, methodology. Z. Ozcan: Software, formal analysis. K. Chang: Data curation, software, formal analysis. P.V. Bommi: Writing–review and editing. M.W. Taggart: Data curation. S. Thirumurthi: Data curation. P.M. Lynch: Data curation. L. Reyes-Uribe: Resources, data curation, formal analysis, visualization, writing-original draft, writing–review and editing. P.A. Scheet: Supervision, writing–review and editing. K.M. Sinha: Formal analysis, supervision, writing–original draft, writing–review and editing. E. Vilar: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was supported by the SWOG/Hope Foundation Impact Award, and a gift from the Feinberg Family to E. Vilar; and P30 CA016672 (US NIH/NCI) to The University of Texas MD Anderson Cancer Center Core Support Grant. We thank the patients and their families for their participation. We thank the staff of the Advanced Technology Genomics Core at MDACC for assistance with mRNA sequencing.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.