An intestinal stem cell niche in Apc mutated neoplasia targetable by CtBP inhibition

C-terminal binding protein 2 (CtBP2) drives intestinal polyposis in the Apc min mouse model of human Familial Adenomatous Polyposis. As CtBP2 is targetable by an inhibitor of its dehydrogenase domain, understanding CtBP2’s role in adenoma formation is necessary to optimize CtBP-targeted therapies in Apc mutated human neoplasia. Tumor initiating cell (TIC) populations were substantially decreased in Apc min Ctbp2+/- intestinal epithelia. Moreover, normally nuclear Ctbp2 was mislocalized to the cytoplasm of intestinal crypt stem cells in Ctbp2+/- mice, both Apc min and wildtype, correlating with low/absent CD133 expression in those cells, and possibly explaining the lower burden of polyps in Apc min Ctbp2+/- mice. The CtBP inhibitor 4-chloro-hydroxyimino phenylpyruvate (4-Cl-HIPP) also robustly downregulated TIC populations and significantly decreased intestinal polyposis in Apc min mice. We have therefore demonstrated a critical link between polyposis, intestinal TIC’s and Ctbp2 gene dosage or activity, supporting continued efforts targeting CtBP in the treatment or prevention of Apc mutated neoplasia.


INTRODUCTION
Sporadic colon cancer is frequently characterized by mutation of the APC tumor suppressor, while autosomal dominant inheritance of a mutant APC allele, as in Familial Adenomatous Polyposis (FAP), results in early onset massive colonic polyposis that uniformly progresses to colorectal cancer unless prophylactic total colectomy is performed [1]. Recently, we have shown that the emerging oncogene and drug target, C-terminal binding protein 2 (CtBP2), is a key driver of neoplasia in the Apc min/+ (Apc min ) mouse model of human FAP [2]. C-terminal binding proteins 1 and 2 (CtBP) are paralogous transcriptional co-regulators frequently overexpressed and associated with worse outcome and aggressive tumor features in colon, [3] breast [4,5], gastric [6], ovarian [7], and prostate [8] cancers. Indeed, the majority of colon and breast tumors demonstrate overexpression of CtBP1 and/or CtBP2 proteins (64% [3] and 92% [4], respectively). CtBP upregulation leads to inappropriate cell survival as well as enhanced migratory and invasive properties due to the ability of CtBP to repress transcription of Bik [9], Brca1 [4], PTEN [10], the epithelial adhesion protein E-cadherin [11], and many other tumor suppressive genes [11,12], as well as co-www.oncotarget.com Oncotarget, 2018, Vol. 9, (No. 65), pp: 32408-32418 Research Paper www.oncotarget.com activate the migration-associated gene Tiam1 [13] and the drug efflux pump MDR1 [14]. In vitro overexpression of CtBP is oncogenic in a manner similar to mutant H-Ras [2], transforming primary mouse embryo fibroblasts to anchorage-independent growth, which is a strong predictor of tumor growth in mouse xenograft models [15]. Importantly, CtBP transcriptional co-regulation is activated by an increase in NADH concentration, as is often the case in hypoxic and/or glycolytically active tumors [16], due to NADH-dependent oligomerization of CtBP's conserved dehydrogenase domain [17].
CtBP2's role in driving a tumor initiating cell (TIC) niche in solid tumors is emerging [19]. TICs contribute to intra-tumoral heterogeneity, metastasis and chemoresistance in a variety of solid cancers, including colon, pancreatic and ovarian, among other cancers [20][21][22][23][24], and an ideal therapy would target this population to overcome metastatic or local relapse from treatmentresistant TICs. Both normal stem cell and TIC populations from intestinal epithelia exhibit the same cell surface markers, though underlying molecular events, such as APC allelic loss, transform normal intestinal stem cells into TIC's [21,25]. A number of TIC-related cell surface markers have been identified in the intestine, including CD44, CD24, CD133, and CXCR4 (23,(28)(29)(30)(31)(32). CD44+/ CD24+ populations obtained from colorectal tumors can initiate growth of colonospheres in vitro and colorectal tumors in vivo, and similar cells exhibit TIC activity across a variety of solid tumors [26]. The progeny of CD133+ cells are capable of giving rise to all major intestinal cell types, as well as neoplasms [21,27], and CD133+ marks long-lived multipotent intestinal stem cells [28]. CXCR4, a chemokine receptor, is expressed both in normal tissue stem cells of the breast, lung, and prostate gland, as well as tumors formed in those organs [29][30][31][32] Consistent with a role of colon TIC's, colon cancer micrometastases require CXCR4 to initiate proliferation, and CXCR4/CD133 dual positive cells demonstrate enhanced tumorigenic capabilities over unsorted cells [33,34]. CXCR4 is also found to be overexpressed in the majority of colon cancer cases [35,36].
In this work we have interrogated the effect of Ctbp2 haploinsufficiency on the intestinal stem cell niche of both wild type and Apc min mice, to better understand Ctbp2's biologic role in the expanded stem cell population in Apc min mice that serve as precursors to polyps, which are enriched for cells with stem cell-like markers that are characterized as tumor initiating cells (TIC's). We show that Ctbp2 loss or inhibition with 4-Cl-HIPP both reduce normal stem cell and TIC populations in Apc min intestine, with a surprising finding of Ctbp2 protein misolocalization to the cytoplasm of stem cells and TIC's when haploinsufficient. This mislocalization could explain the profound effect of Ctbp2 haploinsufficiency on polyp number and survival in Apc min mice and supports further therapeutic development of CtBP as a target in APC mutated neoplasia.
Expression of CD133 in normal crypts of aforementioned genotypes revealed that CD133 was equivalently expressed throughout the lower half of the intestinal crypts of Apc wildtype and Apc min small intestines (See empty arrows, CD133 panel, Figure 2A-2D, 2E), especially in transit amplifying cells and early progenitors, as previously reported [27].
Consistent with Ctbp2's nuclear role in coregulation of Wnt target genes [37], Ctbp2 expression in Apc min (See arrows, DAPI and Ctbp2 panels, Figure 2A) and Apc wildtype (See arrows, DAPI and Ctbp2 panels, Figure 2B) normal crypt intestinal stem cells, or the "stem cell zone" www.oncotarget.com (See circled area indicating stem cell zone in Merge panel, Figure 2A), was robust and predominantly nuclear (See arrows indicating nuclear Ctbp2 in Merge panel, Figure  2A). Staining for Ctbp2 in normal intestinal crypts from Ctbp2 haploinsufficient genotypes (Apc min or Apc WT) demonstrated markedly less nuclear Ctbp2 expression and mostly cytoplasmic expression in the stem cell zone (See arrows, DAPI and Ctbp2 panels Figure 2C and 2D). Moreover, the stem cell zone of Ctbp2 haploinsufficient crypts where cytoplasmic Ctbp2 was observed, also showed partial loss of CD133 expression (See empty arrows in CD133 panel and circles in Merge panel, Figure  2C and 2D), with the average number of CD133+ cells in Ctbp2 +/crypts significantly lower than in Ctbp2 +/+ crypts ( Figure 2E). This indicates a tight correlation between stemness (i.e. CD133 expression) and Ctbp2 nuclear localization, suggesting that the decrease in polyps in Ctbp2 +/-Apc min/+ mice [2] and the lower abundance of stem cells in Ctbp2 haploinsufficient polyps ( Figure 1D-1E), may, in part, be due to cytoplasmic mislocalization of Ctbp2 that correlates with loss of crypt stemness ( Figure  2C-2D).
The expression of CD133 was otherwise intact in the stem cell zone of intestinal crypts from Ctbp2 WT mice (both Apc min and Apc wildtype), suggesting that both alleles of Ctbp2 must be present to sustain expression of CD133 in the stem cell zone. By inference, there is a need for a full complement of Ctbp2 to maintain a phenotypically normal crypt stem cell compartment. Also consistent with this finding in normal crypts, Ctbp2 was predominantly cytoplasmic in adenomatous cells from Ctbp2 +/-Apc min/+ small intestinal polyps as compared with its usual nuclear expression in Ctbp2 +/+ Apc min/+ polyps ( Figure 3A-3B), suggesting the mechanism that led to Ctbp2 cytoplasmic localization in normal crypt stem cells, carried forward when stem cells transformed to form TIC's and adenomatous polyps.

Pharmacological inhibition of Ctbp limits polyposis and intestinal stem cell/TIC abundance in Apc min mice
As we have identified Ctbp2 as a key dependency for both Apc min intestinal polyposis and TIC populations, we investigated whether CtBP chemical inhibition, which can also suppress polyposis, likewise reduced TIC populations. The 1 st generation CtBP inhibitor, HIPP, targets the dehydrogenase domain of CtBP, diminishes CtBP function [2], and suppresses polyp formation by 50% in Apc min mice [2]. Given the high dose of HIPP required in vivo, we tested the most potent available HIPP derivative, 4-Cl-HIPP [18], for suppression of polyposis and effect on TIC populations in Apc min mice (Supplementary Figure 2). Previous studies performed with 4-Cl-HIPP suggest that it is a CtBP2 specific inhibitor due to its on-target ability to abrogate CtBP2-mediated transcriptional repression [18]. Using dosing with minimal toxicity to major organs (Supplementary Figure 2A), we saw that 4-Cl-HIPP was as effective as HIPP at suppressing polyposis at 8 weeks, with a 60% reduction noted, and at a dose 60% lower than HIPP. (

Pharmacological inhibition of Ctbp blocks intestinal Wnt signalling
Colorectal cancers often exhibit aberrant activation of the Wnt/ βcatenin pathway leading to activation of downstream oncogenic targets, such as cyclin D1, c-Myc and the intestinal stem cell regulatory gene, Lgr5 [21,25]. Since Ctbp2 plays a key role in Wnt/βcatenin mediated signaling by coactivating TCF4 [37], we sought to assess if pharmacological inhibition of Ctbp2 using 4-Cl-HIPP inhibits oncogenic Wnt signaling in Apc min intestinal epithelia, as we had previously found in Ctbp2 haploinsufficient mice [2]. Indeed, the mRNA expression of c-Myc and Lgr5 was suppressed in 4-Cl-HIPP vs. vehicle-treated Apc min intestinal cells ( Figure  4D). 4-Cl-HIPP was also able to suppress Ctbp2 mRNA

CtBP2 ablation or pharmacologic inhibition attenuates TIC function and Wnt signalling in human colon cancer cells
We have illustrated efficient 4-Cl-HIPP inhibition of intestinal TIC populations in the Apc min polyposis (precancer) model. To determine if 4-Cl-HIPP could similarly inhibit human TIC populations in cancer cells, and mirror effects of CtBP2 deficiency, we utilized human colorectal cancer cell lines that form tumor spheroids (tumorspheres) that are enriched for TIC's [19] to assay the effect of CtBP inhibition or CtBP2 genetic depletion. We generated HCT116 colon cancer cells genetically deficient for CtBP2 using CRISPR knockout techniques ( Figure 5A, 5D), and compared the effect of 4-Cl-HIPP treatment vs. CtBP2 knockout on tumorsphere growth. Notably, 4-Cl-HIPP efficiently disrupted primary sphere formation of HCT116 cells ( Figure 5B). Additionally, HCT116 cells with CRISPR-mediated deletion of both CtBP2 alleles were also unable to form primary tumorspheres ( Figure 5C), suggesting tumorsphere inhibition, and by inference, TIC inhibition, is an ontarget activity of 4-Cl-HIPP.
Translating our TIC marker and Wnt pathway results from Apc min mouse intestinal epithelia to human colon cancer cells, and consistent with CtBP's regulation of TIC-enriched tumorspheres, LGR5 mRNA and/ or protein expression was also effectively disrupted www.oncotarget.com by CtBP2 knockout in HCT116 tumorspheres ( Figure  5E), or by 4-Cl-HIPP treatment in HCT116 or HT29 tumorspheres ( Figure 5F-5G). Along with LGR5 regulation, we have demonstrated that Ctbp2 deficiency is associated with downregulation of Wnt pathway oncogenic effectors [2] and 4-Cl-HIPP likewise inhibited downstream Wnt pathway targets c-Myc and cyclin D1 in Apc min polyps ( Figure 4E-4H). We therefore examined the effect of 4-Cl-HIPP on c-Myc expression in HCT116 tumorspheres, as c-Myc is a downstream oncogenic target of the Wnt/ TCF4 pathway [38], and observed a robust decrease in c-Myc protein levels after 4-Cl-HIPP treatment ( Figure 5G). Thus, 4-Cl-HIPP can phenocopy CtBP2 knockout and inhibit the Wnt target and TIC marker LGR5, and also attenuate expression of the key oncogenic Wnt pathway target c-Myc in human colon cancer cells.

DISCUSSION
Years of study in cell culture and in human tumors have implicated CtBP1 and 2 as oncogenic, by virtue of observed activities in TIC regulation, EMT induction, chemoresistance and common overexpression in solid tumors linked to poor outcome [39]. Our recent observations that CtBP2 is a key dependency in Apcmutated neoplasia, was the first in vivo evidence of CtBP2's neoplastic potential [2]. In this work, we have determined the biologic basis of CtBP2's neoplastic activities in Apc-mutated neoplasia as linked to its regulation of TIC populations and transcriptional regulation of downstream Wnt targets, such as LGR5, c-Myc and cyclin D1. Moreover, relative deficiency of Ctbp2, or functional inhibition, in the early progenitor niche in crypts, may suppress Apc-deficiency associated neoplasia, as Ctbp2 is a key component of the pathway between Apc loss and neoplastic transformation, via its regulation of c-Myc and cyclin D1.
The surprising findings relative to Ctbp2's unique cytoplasmic localization when genetically haploinsufficient, could explain the drastically lower TIC and polyp count in Ctbp2 +/-Apc min/+ mice, as Ctbp2 localized in the cytoplasm would be partially or fully inactive to perform transcriptional coregulation of Wnt target genes [40]. As a result, crucial Wnt targets that play a driver role in cell proliferation and TIC function, such as c-Myc and cyclin D, would be expressed at lower levels, as already observed in Ctbp2 haploinsufficient Apc min polyps [2], and thus render stem cell precursors to TIC's less likely to transform. Cytoplasmic accumulation of Ctbp2 in the stem cell zone of normal (non-adenomatous) crypts, along with partial loss of CD133 in the setting of Ctbp2 haploinsufficiency, is evidently compatible with normal intestinal development, but could have implications in slower recovery of intestinal crypt regeneration and higher sensitivity of the gut to stress or injury, as the stem cell reserve may be limited. Still unclear, is exactly why Ctbp2 is found in the cytoplasm when it is haploinsufficient, which requires further study, but may be related to stoichiometry of its binding to APC protein (which like Ctbp2, also exhibits a dot like pattern in the cytoplasm [41]). Thus, Ctbp2 is potentially a key driver of transformation of normal crypt CD133+ stem cells to adenomatous polyps, and thus, a driver of the intestinal TIC phenotype in Apc mutated intestinal neoplasia.
Our evaluation of the 2 nd generation CtBP inhibitor 4-Cl-HIPP demonstrates that targeting CtBP is safe and efficacious in the Apc min model, and moreover, its effect on TIC populations in vivo and in vitro phenocopies Ctbp2 knockout, consistent with on-target activity. Further evaluation of 4-Cl-HIPP and related inhibitors, alone or in combination with standard therapies, is warranted in colorectal cancer. Our data also suggests that anti-CtBP therapy, in general, may serve as a novel anti-TIC therapy in settings where TIC's represent a mechanism for cancer relapse and/or chemoresistance [42].

MATERIALS AND METHODS
Cell culture and sphere assay HCT116 cells and HT29 cells (ATCC) were maintained in RPMI 1640 medium in tissue culture treated plates and passaged using trypsin-EDTA upon 70% confluency. For tumorsphere assays, stem cell media (SCM) was prepared as follows; DMEM/ F12 media supplemented with 1% penicillin/ streptomycin, 20ng/ml epidermal growth factor, 10ng/ml fibroblast growth factor and B27 were used. Tumorsphere culture was maintained using 200 cells/ well of an ultra-low attachment plate in SCM and measured for protein and mRNA at the end of 5 days. For primary tumorspheres, cells were seeded in SCM on day one and passaged on day 5 for secondary tumorspheres. On day 10, tumorspheres were passaged for tertiary spheres and harvested on day 15.

Mouse small intestinal mucosal cell isolation
Mouse small intestines were isolated and flushed with HBSS with 2% glucose solution 2 times in closed conformation. Intestinal slices were cut longitudinally along the length and cleaned for debris and fecal matter. The intestines were cut and chopped into fine pieces of about 2mm in size and stored in HBSS+2% glucose on ice. Tissue was resuspended in HBSS-glucose-dispasecollagenase solution and placed on a shaker for 25 min at 25˚C. The digested tissue was further disaggregated by hand pipetting for 3 min. This was followed by 3 slow centrifugations at 300rpm with resuspension of pellets in HBSS+2% glucose to further purify the intestinal mucosal cell population.

Immunohistochemistry and immunofluorescence
Mouse intestinal swiss rolls were formalin fixed and paraffin embedded on slides and were subjected to deparaffinization and hydration steps followed by quenching and peroxidase reaction steps [43]. Antigen retrieval was performed using a pressure cooker, followed by blocking for an hour 5% goat serum and incubation with primary antibodies at a dilution of 1:100 in blocking buffer with CtBP2 (Cat no. 612044, BD Biosciences); CD133 (Cat no. 18470-1-AP, Proteintech, Rosemont, IL, USA); c-Myc (N-262, Cat no. sc-764, Santa Cruz Biotechnology, SantaCruz, CA, USA); cyclin D1 (DCS-6, Cat no. 20044, Santa Cruz Biotechnology) overnight. After 3 washes in PBS, secondary antibodies (HRP conjugated anti mouse or anti rabbit IgG, Cat no. K4000, Dako Envision Systems, Santa Clara, CA, USA) were incubated for 1 hour followed by 3 washes in PBS and slide development using DAB chromogen substrate from Dako Envision Systems (Cat no. K3467). Nuclear staining was performed by incubation in hematoxylin for 2 mins and was followed by dehydration and cover-slipping steps. For IF, steps were similar to IHC, except the secondary antibodies used for detection of CD133 and CtBP2 were anti rabbit (Alexa fluor 594; Cat# A11012; Invitrogen) and anti mouse (Cruz fluor 488; Cat# sc-362257; Santa Cruz Biotechnology), respectively, followed by similar wash steps as IHC. After dehydration steps, slides were mounted with mounting media containing DAPI (Cat no. S36938, Invitrogen, Eugene, OR, USA) and air dried before analysis.

CRISPR/Cas9-mediated deletion of CtBP2 in HCT116 cells
Annealed DNA oligonucleotides DP225-5'-TGC AGA CGG GAT GTT GCA CAG TTT T -3'; DP226-5'-TGT GCA ACA TCC CGT CTG CAC GGT G -3') that coded for the target specific crRNa were ligated with linear GeneArt CRISPR Nuclease Vector with Orange Fluorescent Protein (OFP) (Cat No: A21174; Thermo Fisher Scientific, Waltham, MA, USA). Electrocompetent E.coli (DH10B) were transformed and grown overnight on LB-Agar plates supplemented with 100 μg/ml Ampicillin. Individual clones were purified by streaking and their plasmid DNA was verified by sequencing. HCT116 cells at 70% confluency, grown in RPMI 1640 supplemented with 10% FBS, in a 100mm dish were transfected with 6μg DNA using Lipofectamine 2000. The cells were incubated at 34 o C (5% CO 2 ) for 72 hours and OFP positive cells were single cell fluorescence activated cell sorted in a 96 well plate using Aria-BD, (San Jose, CA, USA) FACSAria™ II High-Speed Cell Sorter at λ exc =488nm. The cells were allowed to grow to confluency before splitting them in triplicate plates to screen for mutant clones.
Clones were washed in cold PBS and lysed in 96 well plates using 50 μL RIPA buffer supplemented with Protease inhibitor cocktail (Roche). Samples were clarified by centrifugation, prepared for SDS PAGE, and resolved on 15 well Bis-Tris Gradient polyacrylamide gels (4-12%) (NuPAGE, Thermo Fisher Scientific) using MOPS as a running buffer followed by a wet transfer to a 0.45μm PVDF membrane (Immobilon-FL, CAT. NO: IPFL00010, Millipore). The membranes were incubated overnight with Anti-CtBP2 Mouse mAb (Catalog No.612044 BD Biosciences) and anti-Vinculin (E1E9V) Rabbit mAb (Cell Signalling, Beverly, MA, USA) followed by 1h incubation with Alexafluor (680) labelled secondary antibodies (Thermo Fisher Scientific). The blots were scanned at 685nm (Odyssey CLx scanner, Odyssey, Lincoln, NE, USA).
After screening, mutant clones were grown to confluency in 6 well plates and genomic DNA was extracted using ISOLATE II Genomic DNA Kit

Author contributions
ATC, SRG, and BP designed experiments; ATC, ADC and BS performed experiments; ATC, SRG, MI, and BP analyzed data; ATC drafted manuscript; SRG edited manuscript; KCE provided reagents.