The Tumor Suppressor Gene, RASSF1A, Is Essential for Protection against Inflammation -Induced Injury

Ras association domain family protein 1A (RASSF1A) is a tumor suppressor gene silenced in cancer. Here we report that RASSF1A is a novel regulator of intestinal inflammation as Rassf1a+/−, Rassf1a−/− and an intestinal epithelial cell specific knockout mouse (Rassf1a IEC-KO) rapidly became sick following dextran sulphate sodium (DSS) administration, a chemical inducer of colitis. Rassf1a knockout mice displayed clinical symptoms of inflammatory bowel disease including: increased intestinal permeability, enhanced cytokine/chemokine production, elevated nuclear factor of kappa light polypeptide gene enhancer in B-cells (NFκB) activity, elevated colonic cell death and epithelial cell injury. Furthermore, epithelial restitution/repair was inhibited in DSS-treated Rassf1a−/− mice with reduction of several makers of proliferation including Yes associated protein (YAP)-driven proliferation. Surprisingly, tyrosine phosphorylation of YAP was detected which coincided with increased nuclear p73 association, Bax-driven epithelial cell death and p53 accumulation resulting in enhanced apoptosis and poor survival of DSS-treated Rassf1a knockout mice. We can inhibit these events and promote the survival of DSS-treated Rassf1a knockout mice with intraperitoneal injection of the c-Abl and c-Abl related protein tyrosine kinase inhibitor, imatinib/gleevec. However, p53 accumulation was not inhibited by imatinib/gleevec in the Rassf1a−/− background which revealed the importance of p53-dependent cell death during intestinal inflammation. These observations suggest that tyrosine phosphorylation of YAP (to drive p73 association and up-regulation of pro-apoptotic genes such as Bax) and accumulation of p53 are consequences of inflammation-induced injury in DSS-treated Rassf1a−/− mice. Mechanistically, we can detect robust associations of RASSF1A with membrane proximal Toll-like receptor (TLR) components to suggest that RASSF1A may function to interfere and restrict TLR-driven activation of NFκB. Failure to restrict NFκB resulted in the inflammation-induced DNA damage driven tyrosine phosphorylation of YAP, subsequent p53 accumulation and loss of intestinal epithelial homeostasis.


Introduction
The Ras association domain family protein 1A (RASSF1A) is frequently lost in human cancers by promoter-specific methylation [1]. RASSF1A (but not RASSF5A/Nore1A) was previously shown to associate with tumor necrosis factor a receptor 1 (TNF-R1) to promote cell death [2]. Although RASSF5A lacks association with TNF-R1, Rassf5a 2/2 mice do have defective TNFa responses suggesting that RASSF5A may influence TNFa signaling and apoptosis through alternate pathways such as interactions with sterile-20 like pro-apoptotic kinases, MST1/2 (mammalian Hippo) [3]. Rassf1a 2/2 mice are viable and fertile, have normal cell cycle parameters and sensitivity to DNA-damaging agents and show no signs of gross genomic instability [4,5]. They do, however, develop tumors in response to chemical carcinogens and develop spontaneous tumors by 18 months of age. Several animals had B-cell lymphomas and others had tumors localized within the gastrointestinal tract suggesting an important role for RASSF1A in gastrointestinal physiology [4,5]. Recently, it was demonstrated that the p53 2/2 /Rassf1a 2/2 double knockout mice are viable, fertile and developmentally normal [6]. However, the loss of p53 (even in the heterozyogous state) led to .60% of the animals developing tumors in the absence of Rassf1a [6]. These tumors include several adenocarcinomas and sarcomas localized within the lungs. However, there was no evidence of colonic hyperplasia [6]. The p53 2/2 /Rassf1a 2/2 double knockout mice also revealed enhanced cytokinesis failure and chromosomal abnormalities leading to aneuploidy which suggests importance in mitotic regulation and tumor suppressor function.
It has been determined that 80% of sporadic colorectal cancer (CRC) tumors have somatic and germline mutations within the tumor suppressor gene, adenomatous polyposis coli (APC) [7]. In the absence of APC, b-catenin freely translocates to the nucleus where it signals an intracellular cascade resulting in the transcription of hundreds of genes [8][9][10]. The loss of RASSF1A in the context of the APC min/+ mice (Rassf1a 2/2 APC min/+ mice) has been reported to result in increased b-catenin nuclear localization, intestinal adenomas and poorer survival than single knockout littermate controls [11]. These data would suggest the importance of the early loss of RASSF1A during premalignant formation of adenocarcinomas that may lead to increased malignant transformation and tumorigenesis. The loss of both APC and RASSF1A function may be key initiating events in the formation of CRC and these observations highlight the importance of RASSF1A in intestinal physiology [10,12].
It has been demonstrated that TNF-R1 is a unique death receptor that can activate a cell death pathway through RASSF1A, triggering Bax activation and mitochondrial-driven apoptosis [1,2,13,14]. In addition, TNF-R1 can also promote a pro-inflammatory response through nuclear factor kappa B (NFkB) pathways [15]. NFkB is an important mediator of inflammation and the pathogenesis of intestinal inflammation and inflammatory bowel disease (IBD) [16]. TLRs can also activate a potent NFkB response and defend against invading pathogens [17]. Of the RASSF family members, RASSF6 and 8 were demonstrated to modulate NFkB activity but a mechanism was not proposed [18,19]. Recently, Song et al. (2012) demonstrated that RASSF2 can associate with the IKK complex to inhibit NFkB [20] and several reports indicate that RASSF1C and 8 may influence the function of the Wnt/b-catenin pathways, pathways important for the appearance of intestinal tumorigenesis [19,21].
It is currently unknown if RASSF1A can influence NFkB activation via TNF-R1 or TLR pathways. However, epigenetic analysis does reveal promoter specific methylation of RASSF1A resulting in loss of expression in ulcerative colitis (UC) patients [22], a form of IBD. Interestingly, IL-6 can drive epigenetic loss of RASSF1A via DNA methyltransferase 1 (DNMT1) up-regulation [23,24] suggesting that inflammation can drive expression loss of RASSF1A. In this study, we demonstrate an in vivo function for RASSF1A in restricting NFkB-dependent inflammation linked to the TLR pathway. Furthermore, we can demonstrate that NFkBdependent inflammation can modulate epithelial proliferation and the appearance of tyrosine phosphorylation of the Hippo pathway target, Yes-associated protein (YAP). We believe that tyrosine phosphorylated YAP can up-regulate pro-apoptotic genes to drive intestinal epithelial apoptosis. Restricting NFkB-dependent inflammation and tyrosine phosphorylation of YAP will maintain epithelial homeostasis, promote epithelial restitution and allow recovery from inflammation induced injury.

Experimental Procedures
All animal experiments/husbandry have been approved and follow the guidelines of the Canadian Animal Care and Use Committee and the animal ethics board at the University of Alberta (permit numbers #461 and 639). All animals (except for the Rassf1a IEC-KO and Rassf1a IEC-WT ) were on the C57BL/6 background. Rassf1a IEC-KO and Rassf1a IEC-WT were on the C57BL/6-129 background. Although the loss of Rassf1a can lead to tumor formation, the loss of Rassf5a (Rassf5a 2/2 mice) has not been documented to display an overt phenotype nor tumor formation as they age. We therefore used the Rassf5a 2/2 mice as a related control [3]. Villin-Cre transgenic mice [B6.SJL-Tg(Vil cre)997Gum/J] was obtained from Jackson Laboratories.
Genotyping of intestinal specific knockout mice PCR was utilized to detect the conditional allele of Rassf1a (RSF-C/RSF-3), for the deletion of exon 1a of RASSF1A (RSF-5/ RSF-3) [5] and the presence of the Villin Cre-transgene (IMR1878/IMR1879/IMR0015 and IMR0016) (Jackson Laboratories) as described previously. RSF-3 (reverse primer), 59 PCR cycle conditional allele or deleted gene: 94uC for 3 min (1 cycle) followed by 94uC for 30 s, 65uC for 1 min, 72uC for 30 s (30 cycles) with a final cycle of 72uC for 10 min and keep at 4uC till the end. PCR for Cre Transgene, 1 cycle at 94uC for 3 min followed by 35 cycles of 94uC for 30 s, 62uC for 1 min, 72uC for 1.5 min, with a final step of 72uC for 1 min and keep at 4uC till the end. RSF-C/RSF-3 produces a 400 bp fragment for wild-type allele of Rassf1a and a 480 bp fragment for conditioned allele. RSF-5/RSF-3 produces a 400 bp fragment only in the mouse tissue that has Exon 1a of Rassf1a deleted. The Cre Transgene genotyping produces a 200 bp fragment for the wild-type allele and an 1100 bp fragment for the Cre Transgene.

Mouse histology
The descending colon was isolated, fixed in z-Fix (Anatech Ltd) and paraffin-embedded. All inflammation scores were obtained utilizing blinded scoring by a gastrointestinal pathologist (Dr. Aducio Thiesen) based on infiltration of enterocytes, neutrophils, lamina propria cellularity, crypt structure and epithelial hyperplasia (scored as 0 -2 where 2 = maximal injury). [25] Crypt depth was measured using a Leica SP5 confocal microscope at 40X magnification. Immunohistochemistry (IHC) and hematoxylin and eosin (H&E) staining were carried out using standard techniques.
10% Bovine growth serum/1% Penicillin-Streptomycin. Bone marrow derived macrophages (BMDM) were isolated from the femur and grown in DMEM/10% Bovine growth serum/1% Penicillin-Streptomycin for 2-3 days on tissue culture dishes [26]. Nonadherent cells (mainly macrophages) were transferred to Petri dishes and grown in DMEM/10% Bovine growth serum/1% Penicillin-Streptomycin containing 20% L cell conditioned media (supernatant from L cells). Macrophages are the only cells that will adhere to the Petri dish. Purity was determined by flow-cytometry analysis using F4/80 and CD4+ markers. Intestinal crypt cells were isolated as outlined previously [27]. Briefly, colons were flushed with cold 16 PBS, cut open longitudinally and then soaked in 1X PBS with gentle shaking for 20 minutes. The colons were cut into small pieces and incubated with 0.04% sodium hypochlorite for 30 min with gently shaking in a 10 cm 2 Petri dish. Following incubation, the colon pieces were removed and allowed to shake continuously at room temperature for 30 minutes in a solution containing 1X PBS/1 mM EGTA/1 mM EDTA. Cells were then dislodged by pipetting the tissue up and down using a 25 mL serological pipette until the solution became cloudy. The supernatant was removed (containing the crypt cells) and centrifuged at 3000 rpm for 10 minute to collect the crypt cells. The cells were washed in 1X PBS, transferred to an eppendorf tube, pulse spun to pellet the cells followed by nuclear extraction as previously described [26]. Generally, one crypt cell preparation for nuclear extraction was obtained from two colons. Colon protein lysates were prepared by homogenizing in TPER lysis buffer (Fisher Scientific) following isolation.

Mouse colonoscopy
Colonoscopy was carried out using a human ureteroscope from Olympus URF [Uretero-Reno Fiberscope] Type P5 within the Center for Excellence in Gastrointestinal Inflammation and Immunity Research (CEGIIR) group at the University of Alberta under isoflurane anesthesia.

Innate immunity analysis
Animals were administered 3% w/v DSS (#160110, molecular weight of 10000, MP Biomedicals) in the drinking water for 7 days followed by recovery for 7 days. They were monitored for: piloerection, bloatednesss, tremors, lack of movement, rectal bleeding and weight loss (all on a scale of 0-5 with 5 being very severe, adapted from Madsen et al.) [25]. Animals were euthanized once rectal bleeding became grossly apparent. For weight loss, a score of 0 for no weight loss, 1 if ,5% loss, 2 for 5-10% loss, 3 for 10 -15% loss, 4 for 15 -20% loss and a score of 5.20% loss in initial body weight. Disease activity indices (DAI) were the sum of all individual scores. All animals were 10 -12 weeks of age or .25 g in body weight at the beginning of the experiment.

Gut permeability experiments
Following innate immunity activation, animals were starved overnight, FITC-Dextran (0.6 mg/g body weight) was orally administered and 4 hours later blood was collected by cardiac puncture. The resulting serum obtained was analyzed for fluorescein isothiocynate (FITC)-Dextran by fluorimetry (excitation 492 nm; emission, 525 nm; PerkinElmer flurometer). In addition, liver, spleen and mesenteric lymph nodes were removed, homogenized in 1X PBS and 10 mL of the lysate were cultured on MacConkey agar plates to detect Gram-negative bacteria translocated from the gut. Similarly, 10 mL of blood was cultured on MacConkey agar plates.
In vitro kinase assays MST1: MST1 was immunoprecipitated overnight from 600 mg of TPER homogenized colon lysate (rabbit anti-MST antibody, Cell Signaling#3682) was used as described previously [28]. Protein A sepharose beads were used to immobilize the MST1 kinase complex for 1 hour, washed twice with wash buffer and twice with kinase buffer (40 mM HEPES, pH 7.5/20 mM MgCl 2 /20 mM b-glycerophosphate/0.1 mM sodium vanadate). The kinase reaction contained 25 mM ATP/5 mCi 32-P-c-ATP and 2 mg of histone H2B in 20 mL of kinase buffer. It was carried out for 20 minutes at 30uC, samples were separated by SDS-PAGE, dried down using a gel dryer and autoradiographed. Coomassie blue panel reveals the amount of Histone H2B used in the assay. c-ABL: ABL was immunoprecipitated overnight from 600 mg of TPER homogenized colon lysate using 1 mg each of anti-c-ABL antibody from BD Bioscience (#554148) and Santa Cruz (sc-121). Protein A sepharose beads were used to immobilize the c-ABL kinase complex for 1 hour, washed twice with wash buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM sodium vanadate, 1 mM EDTA) and twice with kinase buffer (30 mM HEPES, pH 7.1/10 mM MgCl 2 /2 mM MnCl 2 /5 mM DTT/ 10% glycerol/0.2 mM sodium vanadate). The kinase reaction contained 10% glycerol/1 mM ATP/5 mCi 32-P-c-ATP and 6 mg of lysate from HCT116 cells transiently transfected with FLAG-YAP in 30 mL of kinase buffer. The reaction was carried out for 30 minutes at 30uC, samples were separated by SDS-PAGE, dried down using a gel dryer and autoradiographed. Coomassie blue staining confirmed expression levels of FLAG-YAP (Fig. S7D).

NFkB electromobility shift assay (EMSA)
This was carried out as previously described [26]. Briefly, duplex DNA specific for the NFkB binding site was end-labeled with [c-32 P] ATP by using T4 polynucleotide kinase, purified using a G-50 sephadex column (Roche) and allowed to associate with 4 mg of the nuclear extracts containing NFkB for 30 min at room temperature. The oligonucleotide was AAATGTGG-GATTTTCCCATGA for crypt cell NFkB analysis (the NFkB binding site from the IL-6 promoter) and TCAGAGGG-GACTTTCCGAGAGG for BMDM NFkB analysis (the NFkB binding site from the IL-8 promoter). DNA/protein complexes were then separated by non-denaturing gel electrophoresis, dried onto Whatman filter paper and autoradiographed. Nuclear extracts were prepared using NE-PER (ThermoFisher Scientific) as per manufacturer's instructions). Binding buffer was 100 mM Tris-HCL pH 6.5, 500 mM KCl, 1.2 mM EDTA, 12 mM DTT, 20% glycerol and 1 mg Salmon Sperm DNA. The purity of fractions was confirmed with anti-b-tubulin (cytosolic fraction) or anti-lamin B (nuclear fraction) (data not shown).

Determination of reactive oxygen species (ROS)
To determine the degree of ROS generated, a fluorometric assay, utilizing the unique intracellular oxidation of 29,79dichlorofluorescin diacetate (DCF-DA), was used. Freshly isolated crypt cells were seeded into 96-well plates in triplicate. Cells were immediately treated with 5 mM DCF-DA and fluorescence was monitored over 45 minutes in the dark. Fluorescence was measured using a Synergy H4 Microplate Reader (Biotek Instruments) set to 37uC. Measurements were made using a 485/20 nm excitation and a 528/20 nm emission filter pair and a gain sensitivity setting of 55%. Readings were made from the bottom every 30 seconds for a total of 45 minutes.

Statistics
Statistical analyses were performed using one-way ANOVA and Students t-test (two-tailed), as indicated using the GraphPad Prism 5 software. All experiments were carried out at least three times. Error bars in all graphs represent the standard error. Statistics for individual experiments are indicated with the graph or figure legends.

Results
Decreased survival of Rassf1a 2/2 mice following exposure to DSS Maintaining intestinal homeostasis relies on the ability to control insults from resident microbes in the intestinal lumen by maintaining and preserving intestinal permeability function. DSS is a chemical irritant of the colonic mucosa resulting in epithelial barrier damage, microbial invasion and activation of TLRexpressing epithelial cells triggering NFkB activation. Administration of DSS in the drinking water results in colitis in rodents, mimicking human symptoms of UC. These symptoms include weight loss, diarrhea and rectal bleeding. Upon DSS treatment, Rassf1a 2/2 mice became sick within 6 -9 days post-treatment ( Fig. 1A) with reduced activity, piloerection, a general moribund state/hunched posture, lack of grooming, severe rectal bleeding and significantly reduced survival. Wild-type and Rassf5a 2/2 mice displayed only mild symptoms of DSS-induced intestinal injury and demonstrated full recovery from DSS exposure (Fig. 1A). Rassf1a 2/2 mice revealed .25% loss of body weight (Fig. 1B), increased DAI (Fig. 1C), significantly shortened colon (Fig. 1D) and decreased crypt depth (Fig. S1A) in contrast to wild-type and Rassf5a 2/2 mice. Histological sections stained with H&E revealed severe colonic disruption with mucosal/crypt damage (cryptitis and crypt abscesses) in Rassf1a 2/2 mice with evident cellular infiltrates composed of immune cells ( Fig. 1E and Fig. S1B). DSStreated wild type and Rassf5a 2/2 mice revealed minimal colonic morphology changes and significantly less cellular infiltrates (Fig. 1C). Surprisingly, the Rassf1a +/2 mice also demonstrated similar symptoms to the Rassf1a 2/2 mice suggesting haploinsufficiency at the Rassf1a locus (Fig. 1A). Since DSS-treated Rassf1a +/2 mice phenocopy the Rassf1a 2/2 mice data henceforth is presented for the Rassf1a 2/2 mice for most of this article. Genotyping and western blot confirmation of loss of Rassf1a is shown in Fig. S1C.

DSS-treated Rassf1a 2/2 mice have enhanced inflammation
Histological sections from DSS-treated Rassf1a 2/2 colons also revealed mild epithelial hyperplasia as reflected by increased inflammation scores in Rassf1a 2/2 treated mice ( Fig. 2A). Using mouse colonoscopy, we observed significant DSS-induced injury by day 8 in Rassf1a 2/2 versus wild-type or Rassf5a 2/2 mice. Reduction of proper colonic striations and disappearance of a normal vascular pattern together with the presence of erythema was observed in the Rassf1a 2/2 mice indicating active inflammation (Fig. 2B). Together, these findings suggest that RASSF1A is important for protection against DSS-induced intestinal inflammation.
To determine the mechanism of DSS-induced injury in Rassf1a 2/2 mice, levels of inflammatory biomarkers were assessed. Serum levels of IL-6, IL-12 and MCP-1 (but not IL-1b, TNFa, IFNc or IL-23) were elevated in Rassf1a 2/2 mice compared to wild-type or Rassf5a 2/2 mice (Fig. 2C -E). Following DSS treatment, NFkB specific activity in bone marrow cells from Rassf1a 2/2 animals was significantly higher than in wild-type or Rassf5a 2/2 bone marrow cells (Fig. 2F). The detected activity was specific for NFkB (Fig. 2F, lane 4) and lost in the presence of a cold competitor oligonucleotide or an NFkB mutant DNA binding oligonucleotide (Fig. 2F, lanes 4 and 5, respectively). We further characterized inflammation injury in Rassf1a 2/2 versus wild-type mice by measuring myeloperoxidase (MPO, a glycoprotein rapidly released by activated neutrophils) and hyaluronic acid (HA) release. Following DSS-induced intestinal injury, HA synthase (HAS) upregulates HA production in macrophages and lamina propria lymphocytes [29], targeting TLR2 and TLR4, and subsequently promoting TLR4/2-driven NFkB activity [30]. We observed a significant increase in both serum and colon tissue levels of MPO (Fig. S1D) and serum HA levels ( Fig. S1E) in DSStreated Rassf1a 2/2 mice (a 2 -3 fold increase in both markers was observed). Elevated HA levels may indicate a sustained need to activate TLR4-driven protective responses to promote epithelial repair in the continued presence of inflammation-induced damage [30]. In contrast, elevation of MPO may be detrimental to the recovery from DSS-induced injury and may have significantly contributed to the decreased survival of Rassf1a 2/2 mice.

Ex vivo analysis of Rassf1a 2/2 stimulated bone marrow cells reveal abnormal NFkB regulation
Our in vivo results implicate NFkB in intestinal inflammation in DSS-treated Rassf1a 2/2 mice. Activation of intestinal epithelial cells (IECs) generally results in signals transmitted to the bone marrow and lymph nodes, resulting in activation of macrophages and lymphocytes (that is, the adaptive immune response). The reported changes in NFkB DNA-binding activity in bone marrow cells (Fig. 2F) suggested that DSS-driven activation of IECs can be transmitted to the bone marrow to promote an adaptive immune response and amplify NFkB activity. Alternatively, the intrinsic loss of RASSF1A in hematopoietic cells could also cause hyperactivation of NFkB in the immune cells and reflect an important role for RASSF1A in these cells. This issue was addressed by culturing bone marrow derived macrophages (BMDM), splenocytes or crypt cells ex vivo (from untreated animals) and stimulating them with lipopolysaccharide (LPS) or DSS (both activators of TLR4) [29,30]. Normally, NFkB is inhibited by IkBawhich is phosphorylated and targeted for degradation following LPS or DSS stimulation, allowing nuclear translocation and activation of NFkB. Analysis in Rassf1a 2/2 BMDM revealed a sustained increase in IkBa phosphorylation following LPS stimulation (even up to 2 hours post LPS-addition [data not shown]); while both wild-type and Rassf5a 2/2 BMDM showed transient expression of IkBa phosphorylation (Fig. 3A). This indicated that there was an inherent failure of BMDM from Rassf1a 2/2 mice to downmodulate NFkB activity when compared to LPS or DSS-treated BMDM from wild type or Rassf5a 2/2 mice.
Not surprisingly, ex vivo LPS-stimulated Rassf1a 2/2 BMDM revealed elevated production of NFkB-regulated pro-inflammatory elements (IL-6 and MCP-1) ( Fig. 3B -C) as well as IL-12 and TNFa ( Fig. S2A -B). Both IL-6 and IL-12 were also secreted from Rassf1a 2/2 BMDM upon treatment with 3% DSS (Fig. 3D -E). IL-10 (an anti-inflammatory cytokine) and MCP-1 were unchanged upon DSS stimulation ( Fig. S2C and data not shown, respectively) nor was IL-10 changed with LPS stimulation (data not shown). The elevated production of IL-6 and IL-12 (upon LPS or DSS stimulation) and MCP-1 (upon LPS stimulation) can be inhibited by utilizing the NFkB inhibitor, (E)-2-Fluoro-49-methoxystilbene (methoxyresveratrol, IC 50 = 150 nM) ( Fig. 3B -E and Fig. S2A). Methoxyresveratrol has been demonstrated to effectively inhibit LPS-induced COX-2 mRNA production as well as TNFa-driven NFkB activation in cultured cells [31]. These data  is 0.0008 and wild type (+DSS) vs 5a 2/2 is 0.290 (n = 5 -10). (B) Colonoscopy images of 3% DSS-treated animals. The Rassf1a 2/2 mice (+DSS) were euthanized on day 9 due to disease, whereas all others recovered from 3% DSS treatment. Untreated Rassf1a 2/2 or Rassf5a 2/2 mice had images similar to wild type untreated mice. (C -E) Blood serum levels of the indicated cytokines. The p value of DSS treated wild type vs Rassf1a 2/2 = 0.0001 for IL-6 and MCP-1; 0.005 for IL-12; n = 10-14 per group. DSS-treated Rassf1a +/ 2 revealed as similar increases in IL-6 ( Figure 8C) and IL-12 (data not shown). (F) NFkB DNA binding in nuclear extracts derived from bone marrow cells upon DSS treatment. Cold wild type or mutant NFkB oligonucleotide was added to determine the specificity of the observed band (lanes 4 and 5, respectively). Lanes 3 and 7 are the same treatment and genotype but BMDM was obtained from different animals. For (A) and (F), all untreated results for Rassf1a 2/2 and Rassf5a 2/2 were similar to wild type (untreated). doi:10.1371/journal.pone.0075483.g002 strongly suggested that NFkB activity is driving cytokine production ex vivo in BMDM cells from Rassf1a 2/2 mice upon LPS or DSS stimulation. In addition, it also suggests that RASSF1A may function to restrict NFkB activation within the TLR4 pathway. We also observe the abnormal activation of NFkB in ex vivo stimulated colon crypt cells from Rassf1a 2/2 mice upon DSS stimulation that can also be inhibited by methoxyresveratrol (Fig. 3F). No changes were observed between wild-type/Rassf5a 2/ 2 mice versus Rassf1a 2/2 mice for the levels of secreted IL-1b, IFNc? nor IL-23 from ex vivo LPS or DSS-stimulated BMDM (data not shown). BMDM stimulation with the TLR9 activator, CpG DNA (but not the TLR2 ligand PAM 3 CSK 4 ) resulted in the elevated production of TNFa ( Fig. S2D and data not shown respectively) and we can also observe increased cytokine/ chemokine production in isolated Rassf1a 2/2 splenocytes stimulated with CpG DNA (Fig. S2D) or LPS (Fig. S2E -F). Taken together, our ex vivo data suggests that RASSF1A may have a universal role in modulating cytokine/chemokine production in immune cells as well as in epithelial cells and that RASSF1A may influence both TLR4 and TLR9 pathways linked to NFkB activation.

RASSF1A regulates intestinal epithelial homeostasis following DSS-induced inflammation
In order to specifically explore the role for RASSF1A in intestinal physiology, an intestinal epithelial cell specific knockout to RASSF1A, Rassf1a IEC-KO , was generated by mating Rassf1a-LoxP conditional mice [5] with the Villin-Cre transgenic mouse and the knockout allele detected by PCR (Fig. 4A). Villin is an intestinalrelated cytoskeletal protein that is associated with brush border microvilli. RASSF1A knockout was confirmed by PCR utilizing genomic DNA from colonic sections as previously published [5]. Similar to the Rassf1a 2/2 mice, DSS-treated Rassf1a IEC-KO animals displayed ,25% survival (  5F) and colonic tissue HA levels (Fig. S3B). Taken together, these data strengthen the protective role of RASSF1A in restricting unwanted NFkB and following DSS-induced inflammation injury.
Not surprisingly, ex vivo analysis of LPS or DSS stimulated BMDM or splenocytes from Rassf1a IEC-KO mice revealed a wildtype like response to LPS or DSS (compare results in Fig. S3C -D versus Fig. 3B -E). This is consistent with RASSF1A only being deleted in IECs and not in immune cells in the Rassf1a IEC-KO mice. It is thus the loss of RASSF1A within the intestinal epithelial cells that is promoting intestinal inflammation and amplifying the response by activating BMDM. The elevated responses and high cytokine production possibly explains the poor survival of Rassf1a knockout mice. RASSF1A may function to restrict NFkB activation in both colon epithelial and immune cells in order to modulate the inflammatory response. It should be noted that in all Rassf1a IEC-KO animals, we detected the knockout allele in the kidney as well (Fig. 4, right panel) since villin is also expressed in the epithelia of renal proximal tubules [32]. However, RASSF1A immunoblotting revealed the complete loss of RASSF1A from IEC cells but only partial loss of RASSF1A in kidney lysates from Rassf1a IEC-KO mice (Fig. S3E). Importantly, serum creatinine, systolic blood pressure and renal morphology were similar in Rassf1a IEC-KO and wild-type mice before or after exposure to DSS ( Fig. S4A -C). This excludes renal failure as the cause of the poor survival of Rassf1a IEC-KO animals upon DSS treatment.

DSS-treated Rassf1a knockout mice have increased gut permeability
Epithelial barrier breakdown and increased permeability of the gut occurs during severe inflammation and IBD and are associated with increased production of cytokines [33,34]. Fortunately, IEC layer integrity is rapidly re-established by epithelial restitution, wound healing and/or increased epithelial proliferation after inflammation-induced injury in order to maintain a healthy gut. These events prevent direct exposure of lamina propria immune cells to the intestinal microflora and a harmful inflammatory response [35]. Thus, the epithelial integrity of DSS-treated Rassf1a 2/2 mice was examined, assessing cell repair and intestinal permeability. Following a 7 day treatment with DSS and a one day recovery period with regular water, animals were administered FITC-dextran by oral gavage. Serum levels of FITC-dextran were significantly higher in DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice versus wild-type/Rassf5a 2/2 mice, suggesting a role for RASSF1A in maintaining permeability during the recovery phase of DSS-induced inflammation injury (Fig. 6A). Interestingly, spleen, liver and mesenteric lymph node extracts from DSStreated Rassf1a knockout animals revealed .50% bacterial translocation also indicating increased gut permeability (Fig. 6B). DSS-treated Rassf1a knockout mice thus have an impaired ability to contain commensal bacterial in the colon upon inflammationinduced insult.

Defective epithelial restitution in Rassf1a knockout mice following DSS-induced injury
The increased epithelial permeability and decreased survival of DSS-treated Rassf1a 2/2 mice could be explained by the damage caused by disruption of the epithelial barrier (and damaged paracellular space), elevated cell death (apoptosis) or defects in epithelial cell repair and lack of crypt cell proliferation [36]. Due to increased colonic permeability and mucosal/crypt damage in Rassf1a 2/2 and Rassf1a IEC-KO mice following DSS insult (Fig. 1D -E and Fig. 4D -E) we speculate that RASSF1A may influence pathways involved in epithelial restitution and/or cell repair to regain normal epithelial architecture. Proliferating cell nuclear antigen (PCNA) staining of tissue colon sections from Rassf1a 2/2 and Rassf1a IEC-KO mice treated with DSS confirmed reduced crypt cell proliferation versus wild-type or Rassf5a 2/2 (Fig. S5A). We also observed significant PARP cleavage (a marker of late apoptosis) in colon lysates (Fig. 6C) and increased TUNELpositive staining in colonic sections (a marker of condensed nuclei and indicative of apoptosis) (Fig. 6D) from DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice. The substantial increase in apoptosis in DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice is likely a significant contributing factor to the decreased survival observed for these mice in Fig. 1A and 4B.
DSS-treated Rassf1a 2/2 mice triggers tyrosine phosphorylation of yes-associated protein (YAP) following DSS-induced injury RASSF1A can normally promote cell death via TNF-R1 and the loss of RASSF1A should result in decreased cell death. We, in fact, can detect enhanced cell death by PARP and TUNEL analysis (Fig. 6C -D). We thus explored alternative explanations for how the loss of RASSF1A may result in increased intestinal apoptosis. RASSF1A is a component of the Rassf/Salvador/ Hippo pathway that can modulate cell death, organ size and   [37][38][39]. It is thought that in mammalian cells, RASSF1A can function to activate MST1/2, resulting in the activation of LATS and downstream serine phosphorylation events [40]. Activated LATS can lead to the inactivation (by cytoplasmic re-localization and retention) of the Yes-associated protein (YAP), a key diver of proliferation (known as Yorkie [Yki] in Drosophila) linked to the TEAD family of transcription factors [41] Removal of either Yki from intestinal stem cells in Drosophila [42] or YAP in Yap 2/2 mice [43] revealed poor survival and decreased epithelial cell proliferation in response to DSS treatment (characteristics similar to what we have observed). However, we are arguing for the important role of RASSF1A in restricting NFkB and downstream effects that hyperactivation of NFkB can cause. Although inflammation was not investigated in either of these studies, it does suggest an important role for both RASSF1A and YAP in modulating crypt cell proliferation and survival following DSS-induced inflammation injury.
In the absence of RASSF1A, there was a significant reduction in S127 phosphorylation of YAP (Fig. 6E, top panel) and enhanced nuclear presence of p73 associated YAP (Fig. 6E, bottom panel). Reduced S127-phosphorylated YAP and increased nuclear presence of YAP would be expected to result in increased YAPdriven transcription to drive proliferation of intestinal crypt cells in the canonical Hippo pathway. Enhanced proliferation can be observed in wild-type mice upon DSS treatment but not in DSStreated Rassf1a 2/2 and Rassf1a IEC-KO mice (Fig. S5A). DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice clearly revealed enhanced nuclear levels of YAP but reduced crypt proliferation (as determined by PCNA staining in Fig. S5A). However, increased crypt apoptosis (as determined by elevated PARP cleavage and TUNEL staining) can be observed in line with reduced crypt proliferation (Fig. 6C -D). Furthermore, MST kinase activity was dramatically reduced in DSS-treated of Rassf1a 2/2 and Rassf1a IEC-KO mice (Fig. S5B) in agreement with enhanced nuclear presence of YAP.
Further detailed analysis revealed increased tyrosine phosphorylation (pY) of YAP in DSS-treated Rassf1a 2/2 mice in colon lysates/nuclear fractions (Fig. 6F) and in colonic sections (Fig. 7A). A clear difference in pY-YAP between DSS-treated wild type and Rassf1a 2/2 mice was observed as early as day 5 post DSS addition that continued to increase towards day 9 (Fig. 7A, right panel). Previous studies have suggested a functional role for pY-YAP/p73 complex to ultimately drive pro-apoptotic gene expression, especially Bax [44,45]. In support of these arguments, we detected enhanced nuclear presence of a pY-YAP/p73 complex (Fig. 6E, bottom panel), increased Bax expression by immunohistochemistry and immunoblotting (Fig. 7B) in intestinal crypt cells from DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice but not in samples from wild-type or Rassf5a 2/2 treated mice. This suggests that a pY-YAP/p73 driven up-regulation of Bax may promote intestinal cell death leading to increased gut permeability, lack of effective epithelial repair and poor survival of DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice.
Previous studies have suggested a functional role for c-Abldriven tyrosine 357 phosphorylation of YAP in response to DNA damage [46,47]. This resulted in the formation of a pY-YAP/p73 complex to drive pro-apoptotic gene expression, especially Bax [44,45]. We detected enhanced Y357 phosphorylation in Rassf1a knockout mice using a Y357 YAP antibody (Fig. 6F and Fig. 7A).
In addition, we also observed increased DNA damage as early as day 3 (Fig. 7C and Fig. S5C), oxidative damage by day 5 -9 ( Fig. 7D) as well as significant ROS production in colon samples from in vivo DSS-treated Rassf1a knockout mice (Fig. S5D). DNA and oxidative damage of colonic cells have been characterized as significant factors in the pathogenesis of IBD [48,49]. Similar to IBD patients, inflammation-induced injury in our Rassf1a knockout mice has associated DNA and oxidative damage. Since we can observe DNA damage as early at day 3, we speculate that this may be driving the tyrosine phosphorylation of YAP observed by day 3 -5 (Fig. 7A). YAP tyrosine phosphorylation following DSSinduced inflammation injury will in turn drive colonic cell death (via p73 association and transcriptional activation of cell death genes) and poor survival of DSS-treated Rassf1a knockout animals.
Imatinib can protect Rassf1a +/2 mice but not Rassf1a 2/2 mice from DSS-induced inflammation injury Next, we wanted to inhibit the protein tyrosine kinase (PTK) for YAP in order to prevent a pY-YAP/p73 driven transcription of pro-apoptotic genes (such as Bax) and possibly aid in promoting the enhanced survival of DSS-treated Rassf1a knockout mice. As mentioned earlier, YAP has been shown to be tyrosine phosphorylated by c-Abl in response to DNA damage. Imatinib/gleevec was developed to specifically inhibit c-Abl (IC 50 of 25 nM) [50] and has been successful utilized to treat leukemia patients. However, it can also inhibit c-Kit and PDGF-R at 400 nM [50]. Imatinib treatment of our Rassf1a +/2 mice (IP injections of 60 mg/kg body weight of imatinib on day 3 and 6 of an acute 3% DSS treatment) resulted in .80% survival (Fig. 8A), low disease activity indices (Fig. S6A), low histopathological scoring of ,2 (Rassf1a +/2 mice normally have .6, Fig. 8B), reduced IL-6 ( Fig. 8C), reduced cell death ( Fig. 8D and Fig. S6B), reduced DNA and oxidative damage (Fig. S6C -D, purity of nuclear extracts is shown in S6E), regain normal crypt architecture (Fig, 8E and Fig.  S7A) and increased PCNA staining (increased proliferation, Fig.  S7A). Imatinib significantly reduced pY-YAP as determined by immunohistochemical staining (Fig. 8E) and immunoblotting (Fig.  S7B). Furthermore, there was increased c-Abl kinase activity towards FLAG-YAP in colon lysates from the Rassf1a +/2 mice that was inhibited by imatinib (Fig. 8F). The increased kinase activity of c-Abl was not inhibited by a direct modulation by RASSF1A as exogenous addition of GST-RASSF1A did not interfere with the kinase activity of c-Abl as determined by our in vitro method (Fig. 8F, top panel, last two lanes on the right). Thus c-Abl activity may be indirectly modulated by RASSF1A and increased c-Abl activity is most likely a result of increased DNA damage during intestinal inflammation. Taken together, our data suggested that inhibition of the c-Abl class of PTKs was effective in reversing the detrimental effects of a 3% DSS treatment in DSS-treated Rassf1a +/2 mice.
Further analysis of this discrepancy, revealed that the cleavage of c-Abl was more apparent in DSS-treated Rassf1a 2/2 mice ( Fig. 9A) and highly elevated in imatinib treated Rassf1a 2/2 mice (Fig. 9B), especially the presence of the predominant 25 kDa cleaved (but active) fragment of c-Abl ( Fig. 9A and B, marked with ''*'') [51,52]. Interestingly, DSS-treated wild type animals revealed almost undetectable levels of p53 whereas DSS-treated Rassf1a +/2 or Rassf1a 2/2 mice (Fig. 9C) revealed highly elevated or accumulated levels of p53 (as early as day 3 for the DSS-treated Rassf1a 2/2 mice). The accumulation of p53 was more apparent by day 9 and increased in imatinib-treated animals (Fig. 9C) whereby p53 levels not only accumulated but p53 appears to be modified to a slower migrating form around 75 kDa (Fig. 9 C, lanes 9 -12). Under DSS or imatinib treatment, p53 mRNA levels do not appear to be significantly different, eliminating transcriptional control (Fig. 9D). Evidence for p53 stabilization can be seen in Fig.  S7E and does reveal that imatinib can interfere with p53   dependent ubiquitination and increased in the DSS and imatinibtreated Rassf1a 2/2 mice (Fig. S7D -E). Since c-Abl can be caspase cleaved and its resultant active fragment can phosphorylate the p53 inhibitor/E3 ligase, Mdm2 [53], we speculate that in the absence of RASSF1A, DSS-induced inflammation injury can result in accumulated p53 and significantly higher levels of cell death and poor recovery following DSS-induced inflammation injury.

Discussion
An underlying hallmark of many inflammatory diseases is the enhanced activation of transcription factors, including NFkB and elevated production of NFkB -regulated cytokines. These responses are needed to promote innate immunity activation and stimulate the repair process. No single animal model can explain the pathogenesis of IBD suggesting that IBD has multifactorial etiologies [54,55]. The present study argues that RASSF1A is an important regulatory element to restrict NFkB activity. Unrestricted NFkB activity can significantly contribute to poor recovery following inflammation-induced injury and inhibit the ability to promote efficient epithelial repair and survival (Fig. 10).
We have demonstrated that DSS-treated Rassf1a 2/2 and Rassf1a IEC-KO mice have elevated NFkB activity. Mechanistically RASSF1A appears to restrict complex formation of MyD88/ TRAF6/IRAK and thus interfere with NFkB activation (Fig. 10). RASSF1A forms robust associations upon LPS or DSS stimulation with the membrane proximal complexes MyD88/TLR4 and TRAF6/IRAK2 (Fig. S8A -B) but not with TAK1, IKK or TRADD/RIP1/TNF-R1 (data not shown) in order to restrict NFkB activity. Since both LPS and DSS can stimulate TLR4, we believe that RASSF1A functions through TLR4 to inhibit IKK and the cytoplasmic/nuclear shuttling of NFkB. Indeed we have observed a significant reduction in the production of the NFkBdriven cytokine, IL-8, when RASSF1A was overexpressed in colon cancer epithelial cells (Fig. S8C). Furthermore, enhanced ubiquitination of TRAF6 can be observed in colon lysates from DSStreated Rassf1a 2/2 but not wild-type mice (Fig. S8D) indicative of elevated membrane proximal activation of the TLR pathway to drive TRAF6 ubiquitination and subsequent NFkB activation (Fig. 10).
We also have in vivo evidence to demonstrate that YAP becomes tyrosine phosphorylated in response to DSS-induced inflammation injury most likely by the c-Abl class of PTKs. YAP was originally identified by its association with the YES Src tyrosine kinase [56] and has been demonstrated to be a transcription factor whose cytoplasmic/nuclear shuttling is controlled by post-translational phosphorylation events [57] [58]. During DSS-induced inflammation injury in both Rassf1a +/2 and Rassf1a 2/2 mice, we observed elevated level of nuclear tyrosine phosphorylated YAP, enhanced YAP/p73 associations (Fig. 6E) and elevated levels of Bax expression to promote apoptosis ( Fig. 6 and 7). Recently, van der Weyden et al. [59] demonstrated that the combined loss of Rassf1a and Runx2 promoted the formation of oncogenic YAP1/ TEAD transcriptional complexes to promote tumorigenesis. We observed reduced S127 phosphorylated YAP (possibly indicating reduced YAP/TEAD transcriptional complexes) in DSS-treated Rassf1a 2/2 mice. We propose that our results demonstrate that the early loss of Rassf1a can lead to enhanced inflammation, a key precondition to cancer and inflammation driven activation of YAP/ p73 complexes to promote cell death. YAP1/TEAD complexes may arise after prolonged chronic inflammation to promote proliferation in support of the findings of van der Weyden et al. [59]. It has been demonstrated that IL-6 can promote the activation of DNA methyltransferase 1 (DNMT1) [24,60] and DNMT1 can epigenetically silence RASSF1A [23,61] to provide a mechanism whereby inflammation may drive epigenetic silencing of RASSF1A and loss of the function of this tumor suppressor protein.
The c-Abl tyrosine kinase can phosphorylate YAP upon DNA damage, selectively promoting a YAP/p73 complex to drive the expression of pro-apoptotic genes [47]. Since we cannot observe inhibition of c-Abl kinase by exogenous addition of RASSF1A, we speculate that NFkB may transcriptionally up-regulate a PTK or that NFkB dependent-inflammation will promote DNA damage to drive the tyrosine phosphorylation of YAP via c-Abl or c-Abl-like kinases (Fig. 10). Either mode of regulation would result in an abnormal tyrosine phosphorylation of YAP to yield increased epithelial apoptosis and p53 accumulation to promote further apoptosis. These events would result in poor survival following intestinal inflammation induced injury. Inhibiting this abnormal phosphorylation of YAP using the c-Abl class of PTK inhibitors (imatinib) reversed the damaging effects of DSS induced intestinal inflammation but only in a Rassf1a +/2 background. In a Rassf1a 2/ 2 background, imatinib partially reversed the effects of DSSinduced inflammation but did not allow the animal to recover due a robust accumulation of p53 towards day 9 that may have resulted in more apoptosis and poor recovery. Interestingly, in day 9 imatinib-treated Rassf1a 2/2 mice, p53 was accumulated and modified (Fig. 9C, lane 9). We have not identified what this modification is (most likely one of many p53 phosphorylated residues) but we are currently exploring this and, if unique, may be an interesting biomarker for disease prognosis.
It is known that caspase cleaved c-Abl can phosphorylate the p53 E3 ligase, Mdm2, to inhibit Mdm2-dependent degradation of p53 and allow it to accumulate [53]. Accumulated p53 can then associate with the anti-apoptotic inhibitor, Bcl-2 to promote cell death [62]. This may explain why imatinib-treated Rassf1a 2/2 mice are not protected from DSS-induced inflammation and have similar or enhanced cell death. Curiously, we cannot absolutely eliminate tyrosine phosphorylation of YAP using imatinib (Fig.  S7B) nor most of the biomarkers of inflammation or cell death. This would suggest that other PTKs may also influence tyrosine phosphorylation of YAP upon inflammation induced injury in the absence of RASSF1A. YES, the other known PTK for YAP, has two NFkB binding sites within its promoter region (as documented by http://www.genecards.org/cgi-bin/carddisp.pl?gene = YES1) and could be up-regulated by NFkB activation due to the failure of RASSF1A to restrict NFkB activation (Fig. 10). Preliminary results do reveal increased c-YES expression upon DSS-induced stimulation in Rassf1a +/2 and Rassf1a 2/2 colon lysates (data not shown). We are currently exploring how NFkB may regulate YES expression, if YES kinase activity is elevated in DSS-treated Rassf1a knockout mice and if the YES inhibitor, dasatinib, can promote the survival of DSS-treated Rassf1a knockout mice.
Numbers represent fold induction relative to phospho-c-H2AX levels in untreated wild type animals. (D) Real time PCR analysis for an oxidative damage marker hemexogenase-1 (HO-1) utilizing colonic mRNA (n = 4 for all and D designates the day that the tissues were harvested). P value of wild type (+DSS) versus Rassf5a 2/2 is 0.027 (Day 3), 0.006 (day 5) and 0.0008 (day 9) with n = 4 -6 per day per sample. The Rassf1a IEC-KO result was comparable to the result in the DSS treated Rassf1a 2/2 mice (data not shown). doi:10.1371/journal.pone.0075483.g007 Figure 8. The PTK inhibitor, imatinib, inhibits the appearance of pY-YAP and promoted increased survival of Rassf1a +/2 but not Rassf1a 2/2 mice animals following inflammation-induced injury. (A) Rassf1a +/2 or Rassf1a 2/2 mice were intraperitoneally injected with the PTK inhibitor, imatinib at 50 mg/g body weight on days 3 and 6 following 3% DSS addition. P-value between survival of DSS-treated wild type and Rassf1a +/2 was ,0.0001 (n = 17) and between DSS-treated Rassf1a +/2 (+ imatinib) versus DSS-treated Rassf1a +/2 was 0.0086 (n = 17). No significance difference was observed between DSS-treated Rassf1a +/2 and DSS-treated Rassf1a 2/2 (+ imatinib) mice (please see Fig. 1A for the survival curve of DSS-treated Rassf1a 2/2 mice). Following DSS/gleevec treatment, (B) histological analysis of colonic sections, (C) serum IL-6, (D) cell death via PARP (late marker of apoptosis); (E) phospho-YAP by IHC, and (F) In vitro kinase activity was carried out for c-Abl using colon lystes from DSS-treated wild Our results demonstrate that the loss of Rassf1a resulted in the loss of MST1 kinase activity (Fig. S5B) and the increased nuclear presence of tyrosine phosphorylation of YAP ( Fig. 6F and 7A). The studies by van der Weyden et al. [59], Ren et al. [42], Cai et al. [43], Zhou et al. [63] and Matallans et al. [64] explored MST/ RASSF1A dependent-regulation of YAP serine phosphorylation. However, during intestinal inflammation in our system, YAP transcriptional activities drive apoptosis in the absence of MST1/2 activity (see Fig. S5B). Recently, it was shown that YAP1 does not require MST1/2 to drive the proliferation of keratinocytes [65] in support of MST1/2 independent YAP functions. Furthermore, although Matallans et al. [64] demonstrated that RASSF1A can promote a YAP/p73 association they did not determine YAP phosphorylation linked to S127 or Y357. We argue that intestinal inflammation is driving a RASSF1A indirect affect on YAP through elevated NFkB activity to transcriptionally up-regulate a PTK to tyrosine phosphorylate YAP or, alternatively, in promoting enhanced DNA damage to activate a YAP PTK. Either way, it is an MST1/2 independent effect.
As mentioned earlier, DSS-treated YAP 2/2 mice have decreased survival and proliferation and increased cell death upon DSS treatment that is very similar to our inflammation phenotype [43]. However, Cai et al. [43] did not explore tyrosine phosphorylation of YAP but did show reduced YAP phospho-S127 with increased time post-DSS addition [43]. We have evidence that increased c-Abl cleavage (most likely by caspases) and accumulated p53 may occur in parallel to p73/YAP driven cell death following DSS-induced inflammation injury (based on the results in Fig. 9C). This may explain why DSS-treated YAP 2/2 mice have elevated apoptosis and poor survival in the absence of YAP. It has been shown in chronic UC patients that p53 is upregulated in crypt cells and neutralizing antibodies to TNFa reduce the levels of p53 and IBD-related symptoms of UC patients [66] in support of the important role for p53 in intestinal epithelial-dependent cell death. In addition to p53, PUMA (p53 up-regulated modulator of apoptosis) has been shown to be upregulated in intestinal cells following DSS treatment in mice and in colonic biopsies from UC patients suggesting an important role for other pro-apoptotic proteins in the pathogenesis of colitis [67]. Preliminary microarray analyses on colonic RNA isolated from DSS-treated wild-type and Rassf1a 2/2 mice revealed up-regulation of several pro-apoptotic genes. These genes were increased 2 -4 fold and include the death associated protein 3 (DAP3), programmed cell death 10 (PDCD10), PRKC apoptosis WT1 regulator (PAWR) and death effector domain-containing (DEDD) (data not shown). It will be interesting to explore if any of these are transcriptionally regulated by YAP/p73 in response to intestinal inflammation.
Efficient epithelial proliferation/repair is an integral part of recovery following inflammation-induced injury [36]. Preliminary data suggest that we can observe altered b-catenin/E-cadherin adherens junctions on colon sections from DSS-treated Rassf1a knockout mice during the recovery phase following DSS addition (day 9, data not shown). Altered b-catenin/E-cadherin associations together with elevation of tyrosine phosphorylated YAP, will result in abnormal formation of epithelial junctions and the appearance of a ''leaky gut''. Interestingly, YAP is overexpressed in CRC patients and RASSF1A is epigenetically silenced in CRC patients [12]. The loss of RASSF1A may, therefore, affect other aspects of epithelial restitution with YAP activity an integral part of this process. Failure to efficiently reseal and re-establish epithelial cell homoestasis will lead to an unnecessary inflammatory response, poor recovery and an early phase of colitis-associated carcinogenesis [68,69].
An important element for recovery from inflammation-induced damage is efficient repair. During the pathogenesis of IBD, it has been documented that DNA damage [70], oxidative damage [71,72] and abnormal ROS levels [72] are elevated in a similar manner to what we have observed for DSS-treated Rassf1a knockout mice. Reactive oxygen species (ROS) are generated by inflammatory cells, neutrophils and macrophages and surround the inflamed mucosa [72]. High levels of ROS can lead to changes in colonic membrane permeability, decreased mucosal barrier of intestinal epithelial cells and to increased DNA and oxidative damage. Both DNA and oxidative damage can lead to c-Abl activation to drive the formation of tyrosine phosphorylated YAP and a YAP/p73-dependent transcriptional program. Inhibitors of ROS generation or DNA damage may be interesting angles to pursue in the future.
In addition to c-Abl and YES as potential PTKs for YAP, preliminary microarray analyses on colonic RNA isolated from DSS-treated wild-type and Rassf1a 2/2 mice revealed a 7.5 fold increase in TNK2 (a tyrosine kinase, non receptor type 2 also known as ACK1) and a 2.3 fold increase in c-Mer proto-oncogene tyrosine kinase (MERTK) (data not shown). MERTK has several potential NFkB binding sites (as documented by http://www.ge necards.org/cgi-bin/carddisp.pl?gene = MERTK) and can modulate the autophosphorylation of TNK2. MERTK can also regulate cytokine production and clearance of apoptotic cells. [73] Very little is known about TNK2 other than its association with cdc42 in order to maintain it in a GTP bound form [74] and TNK2dependent tyrosine phosphorylation of the androgen and epidermal growth factor receptor [75]. These observations provide interesting candidates to pursue that may function to drive the tyrosine phosphorylation of YAP and a YAP/p73 transcriptional program. Inhibiting the devastating effect of tyrosine phosphorylated YAP/p73-driven activation of pro-apoptotic genes and epithelial cell damage in the absence of RASSF1A may aid in designing future treatment schemes for patients with chronic inflammation. In fact, there has been one report to document the beneficial effects of protein tyrosine kinase inhibitors in treating IBD [26,76].
Our study provides a mechanistic understanding of DSSinduced inflammation injury in the absence of RASSF1A. Biomarker analyses for the Rassf1a +/2 and Rassf1a 2/2 mice revealed similar outcomes suggesting similar mechanisms of disease pathogenesis (except for responsiveness to imatinib). As can be observed in Fig. 1A, we did notice a slight delay in the appearance of clinical symptoms in the Rassf1a +/2 mice, especially in the appearance of rectal bleeding. However, both Rassf1a +/ 2 and Rassf1a 2/2 mice had similar survival rates at the end of the experiment and the delay at the beginning of the experiment was type and Rassf1a +/2 (top panel) and Rassf1a 2/2 (bottom panel) mice with overexpressed FLAG-YAP as substrate. Expression levels of c-Abl were similar in all the lanes (data not shown) and bacterially expressed GST or GST-1A (1A) was used to explore how RASSF1A may directly interfere with c-Abl kinase activity. Expression of FLAG-YAP, GST and GST-1A are shown in Fig. S7D. For (B) p-value between wild type versus Rassf1a +/2 mice (+DSS) was 0.004, wild type versus Rassf1a +/2 mice + DSS + gleevec) was 0.168 and wild type versus Rassf1a 2/2 mice (+DSS + gleevec) was 0.452 (n = 4 -8). For (C), p-value between wild type versus Rassf1a +/2 mice (+DSS) was 0.004 and wild type versus Rassf1a +/2 mice (+ DSS + gleevec) was 0.347 and wild type versus Rassf1a 2/2 mice (+DSS + gleevec) was 0.262 (n = 4 -8). For (E) P values of Rassf1a +/2 mice or Rassf1a 2/2 mice (+DSS 2/+ gleevec) was ,0.001 (n = 10). doi:10.1371/journal.pone.0075483.g008 not statistically significant. We are currently exploring biomarker analyses between day 0 and day 8 to explore molecular differences between Rassf1a +/2 and Rassf1a 2/2 mice. Interestingly, it has recently been demonstrated that p53 functions to recruit DAXX and DNMT1 to CpG methylation sites on RASSF1A and epigenetically silence RASSF1A [77]. p53 is accumulated in DSS-treated animals, in IBD patients and can also silence RASSF1A by direct promote regulation of epigenetic silencing by DNA methylation of its promoter. Furthermore, RASSF1A epigenetic silencing can be detected in ulcerative colitis patients [22] and upon IL-6 production [23]. We, therefore, speculate that chronic inflammation will result in the loss of epigenetic loss of RASSF1A and RASSF1A may be in early change in the pathogenesis of ulcerative colitis. The Rassf1a +/2 , Rassf1a 2/2 and Rassf1a IEC-KO mice may be useful models for understanding IBD and other inflammatory disorders.  . Increased PTK activity (in the absence of Rassf1a) would drive tyrosine phosphorylation of YAP and increased pY-YAP/p73 transcriptional up-regulation of pro-apoptotic genes such as Bax. Increased levels of Bax (and other p73/pY YAP targets) results in apoptosis, intestinal inflammation, oxidative (and DNA damage) and colonic injury. Sustained levels of apoptosis will result in the pro-apoptotic cleavage of c-Abl to stabilize p53. Accumulated p53 can further promote cell death and further colonic injury and poor recovery following inflammation insults. doi:10.1371/journal.pone.0075483.g010 is 0.0043; Rassf1a 2/2 (+DSS) vs Rassf5a 2/2 mice (+DSS) 0.0099. n = 10-15 for each biomarker. (TIF) Figure S2 Ex vivo analysis of BMDM and splenocytes from the indicated genotypes. BMDM (A -D) and splenoctyes (B, D -F) were cultured, stimulated with 0.2 mg/ml LPS for 5 hours or 3% DSS for 16 -20 hours as indicated. Supernatants were harvested and secreted cytokines quantitated by ELISA; n = 6. P-values between wild type and Rassf1a 2/2 LPSstimulated cells are indicated and was found to be the same for both BMDM and splenocytes in (B and D). P values between wildtype and Rassf5a 2/2 in S2E were 0.454. For (C), p-value = 0.7 by One-way Anova analysis. n = 4-6 from 2 -3 independent experiments for all panels. P values for the differences between  Figure S4 Normal kidney function in Rassf1a IEC-KO animals following DSS treatment. Serum creatinine (A) and systolic blood pressure (B) was measured as indicated (day 9 post-DSS treatment). P value was between 0.2 and 0.5 for the differences between Rassf1a IEC-WT vs Rassf1 IEC-KO animals for these measurements (n = 8-10 for the serum creatinine levels and 6 for the blood pressure measurements). (C) Kidney sections from the indicated genotypes were analyzed after H&E staining and were found to be unaltered with and without DSS treatment. (TIF) Figure S5 The loss of RASSF1A results in decreased crypt cell proliferation and increased cell death following DSS-induced inflammation injury. (A) Measurement of PCNA positive proliferation in colonic sections. Biotinylated PCNA staining was detected using diaminobenzidine (DAB) streptavidin. HRP appears as a brown precipitated over the H&E stained sections. ''*'' p value = 0.0001 and ''**'' = 0.0002, n = 12-15 for all genotypes and treatments. P value wild type (+DSS) vs Rassf5a 2/2 (+DSS) = 0.719. Percent PCNA staining was calculated by counting 4 groups of ,100 cells in three independent histological sections from each genotype and treatment. (B) MST1 in vitro kinase assay was carried out on colon lysates from the indicated genotypes using Histone H2B. MST1 expression is shown for the 2/+ DSS-treated Rassf1a 2/2 mice. Similar results were obtained for the other genotypes and experiment was carried out twice with similar results. (C) Time course analysis of DNA damage utilizing expression of phospho-c-H2AX as observed in Fig. 7C. (D) Fluorometric analysis of production of reactive oxygen species (ROS) using intracellular oxidation of 29,79-dichlorofluorescin diacetate (DCF-DA) by freshly isolated colon crypt cells from the indicated genotypes. P values wild-type (+/+, +DSS) vs Rassf1a 2/2 or Rassf1a +/2 mice (+DSS) ,0.0001; wild type (+DSS) vs Rassf1a IEC-KO mice = 0.0006; wild type (+DSS) vs Rassf5a 2/2 mice (+DSS) 0.083 (n = 8 for all genotypes). For (A and C) all untreated results for Rassf1a 2/2 , Rassf5a 2/2 , Rassf1a IEC-KO were similar to wild type (untreated). (TIF) Figure S6 The PTK inhibitor, imatinib reverses the damaging effects of DSS treatment in the Rassf1a +/2 but not Rassf1a 2/2 knockout mice. Imatinib was administered intraperitoneally at 50 mg/g body weight on day 3 and 6 and (A) Disease activity index, (B) cell death using Bax immunoblotting (as an early marker of apoptosis) (in colon lysates), (C) the DNA damage marker phospho-c-H2AX (in colon lysates) and (D) the oxidative damage marker, HO-1was carried out as indicated (source of sample was colonic mRNA). (E) Purity of our nuclear and cytoplasmic fractions was tested as indicated. . Fourty-eight hours after transfection, cells were grown in serum free media for two hours followed by (A) 1 mg/ml LPS addition or (B) 3% DSS in the media, lysed and immunoprecipitated with the indicated antibodies. Associated proteins were separated on SDS-PAGE and immunoblotted. (C) HT-29 (colon epithelial cancer) cells were transiently transfected with vector (-) or HA-tagged RASSF1A (HA-1A). Fourty-eight hours after transfection, cells were grown in serum free media for two hours followed by the addition of 1 mg/ ml LPS or 3% DSS for 5 hours. One mL of the supernatant was harvested and IL-8 levels determined. Inset, expression of HA-1A in HT-29 cells. Similar results were obtained for HCT-116 cells (data not shown). (D) Colon extracts were prepared from the indicated genotypes 2/+DSS treated, immunoprecipitated with anti-TRAF6 and immunoblotted as indicated. Lysates from two animals/treatement/genotype were analyzed as indicated.