6.0±5.2 cm with the majority located in the pancreatic body/tail (44/47, 93.6%). Finalhistology demonstrated 38 nonmalignant (28 without dysplasia, 8 low-grade, 2 moderate-grade) and 9 malignant MCNs (5 high-grade dysplasia, 4 adenocarcinoma). There was nodifference in mean cyst size (5.62±4.29 vs. 7.62±7.98 cm; p=0.62), proportion of cysts ≥6 cm (34.2 vs. 44.4%; p=0.57), or proportion ≥ 3 cm (65.8 vs. 77.8%; p=0.49) whencomparing nonmalignant vs. malignant MCNs. Similarly, no difference in patient age wasfound between nonmalignant vs. malignant MCNs (46.9±12.7 vs. 48.3±11.4; p=0.75) ornoninvasive vs. invasive MCNs (47.4±12.6 vs. 44.0±9.8; p=0.55). The presence of muralnodules in the cyst on pathology was associated with malignancy (5.3% nonmalignant vs.33.3% malignant; p=0.014). However, no differences were found with regards to presenceof mass on endoscopic ultrasound (34.8 [8/28] vs. 33.3% [2/6]; p=0.95), presence of cystcytological atypia via fine-needle aspiration (6.7 [1/15] vs. 0% [0/4]; p=0.60), or mediancyst carcinoembryonic antigen (CEA) level (564, IQR 296-1090 [n=13] vs. 12977 ng/mL, IQR 407-47618 [n=4], p=0.31) when comparing nonmalignant vs. malignant MCNs.Conclusions: 1. Cyst size does not predict malignancy in MCNs. 2. Patient age does notpredict malignancy in MCNs. 3. Not all MCNs are dysplastic. Given the unclear naturalhistory of MCNs, guidelines for pre-operative risk stratification are needed to avoid unneces-sary resections.

Su1856

TH17-Related Cytokines and Excessive Activation of STAT3 Sustain ColorectalCancer Cell GrowthVeronica De Simone, Eleonora Franzè, Giulia Ronchetti, Alfredo Colantoni, Massimo C.Fantini, Davide Di Fusco, Giuseppe Sica, Pierpaolo Sileri, MacDonald Thomas, FrancescoPallone, Giovanni Monteleone, Carmine Stolfi

Background and aim: Colorectal cancer (CRC) is massively infiltrated with immune/inflamma-tory cells, which can either foster or suppress the survival and growth of transformed cells,mostly through the production of cytokines. This study was aimed at characterizing theinflammatory cytokine response in sporadic CRC and determining the signaling pathwaysby which cytokines modulate CRC cell growth. Methods: Tumor infiltrating cells (TIC) andlamina propria mononuclear cells (LPMC) were isolated from CRC samples and adjacent,normal, colonic mucosa respectively and analyzed for the expression of IL-17A, IL-17F, IL-21, IL-22, IL-6 and TNF-α by flow-cytometry and/or ELISA. HT-29 and DLD-1 cells werecultured in the presence or absence of culture supernatants of TIC and LPMC with orwithout Th17 cytokine neutralizing antibodies or STAT3 inhibitor (i.e., BP-1-102) and cellgrowth was assessed by BrdU assay and flow-cytometry. Activation of STAT3 was evaluatedin CRC cells by Western blotting. Th17-related cytokines and STAT3 activation were evalu-ated in colonic tissues taken from Apc (Min/+) mice by real-time PCR and immunohistoche-mistry respectively. Apc(Min/+) mice were given BP-1-102 by oral gavage and the numberand size of colonic lesions was evaluated by endoscopy. Results: The neoplastic area of CRCsamples was massively infiltrated with CD4+ T cells producing Th17-related cytokines andTIC-derived supernatants induced STAT3 activation and a robust proliferation of CRC cells.Neutralization of STAT3 but not that of single Th17-related cytokines significantly reducedTIC-derived supernatants-mediated CRC cell growth. Th17-related cytokines increased inthe colonic lesions of Apc(Min/+) mice as compared with the adjacent non-tumoral areasand this was associated with enhanced STAT3 activation in both transformed epithelial andimmune cells. Mice given BP-1-102 showed reduced colon tumorigenesis. Conclusions:Transition from the normal colonic mucosa to the neoplastic area is marked abundance ofTh17-related cytokines and excessive activation of STAT3. Data suggest that inhibition ofSTAT3 rather than blockade of single Th17 cytokines could be therapeutically useful inCRC patients.

Su1857

Use of Surveillance Colonoscopy in Non-Selected PopulationMaria Rodriguez-Soler, Miriam Juarez-Quesada, Cecilia M. Egoavil, Carla Guarinos, EvaHernandez-Illan, Lucia Castaño-Soler, Lucía Pérez-Carbonell, Juan Martínez, FranciscoRuiz, Luis Compañy, Jose Ramon Aparicio, Juan Antonio Casellas, Cristina Alenda,Rodrigo Jover

INTRODUCTION Surveillance after polyp removal is recommended at the minimum fre-quency necessary for colorectal cancer (CRC) prevention. The timing of surveillance colonos-copy in individuals with adenomas is derived of a risk estimation based on number, sizeand histology of these adenomas. The aim of this retrospective study is to analyze the useand the yield of surveillance colonoscopies in our community. METHODS We retrospectivelyincluded all patients with diagnosis of adenomas in colonoscopy performed in 2007 at atertiary hospital. Of these patients we recollected data until 2012. Patients with diagnosisof CRC at the moment of this colonoscopy and patients without surveillance were excluded.Colonoscopy results and the time of surveillance colonoscopies were confirmed by reviewingcolonoscopy reports. Histology of polyps was reviewed by an expert pathologist of ourHospital. We considered High Risk Adenoma (HRA) when 3 or more adenomas were foundor at least one ≥1cm or villous or high grade dysplasia were present. RESULTS During2007, 217 patients were diagnosed with adenomatous polyps. Of them, 114 (52.5%) patientshad HRA and 103 (47.5%) had Low Risk Adenoma (LRA) at this basal colonoscopy. In 117patients at least one surveillance colonoscopy was done, independent of the grade of riskof basal adenoma (HRA 51.8% vs LRA 56.3%; p=0.5). When we analyzed the group ofpatients with surveillance colonoscopy, 74 patients were male (63.2%) and the mean agewas 62±11 years. At the baseline colonoscopy the median of adenomas were 1 (1-8) and48.7% of patients had adenomas larger than 1cm. 49.6% of patients had LRA and 50.4%had HRA at baseline colonoscopy. Adenomas were found in successive colonoscopies in39.3% of patients. In the first surveillance colonoscopy no adenoma was found in 70.9%of patients whereas LRA in 22.2% and HRA in 6.8%. The presence of new adenomas insuccessive surveillance colonoscopies during 5 years was independent of the grade of riskof the initial adenoma. Patients with HRA at basal colonoscopy had more HRA in the firstsurveillance colonoscopy (HRA 11.9% vs LRA 1.7%; p=0.03). No CRC was detected inthese surveillance colonoscopies. The time of surveillance colonoscopies was similar in bothgroups (HRA: 1 year 51.7%, 2 years 19%, 3 years 24.1% and 4 years 5.2% vs LRA: 1 year

S-487 AGA Abstracts

54.2%, 2 years 25.4%, 3 years 18.6% and 4 years 1.7%; p=0.5). Only few cases with earlyrepeated colonoscopy can be explained by incomplete or deficient colon cleansing (18.8%of all baseline colonoscopies). CONCLUSION There is a high overuse of surveillance colonos-copy after adenoma removal, especially in patients with LRA at baseline. These surveillancecolonoscopies showed a low yield for advanced lesions.

Su1858

Adrenergic Signaling Promotes Pancreatic Cancer Initiation and ProgressionBernhard W. Renz, Christoph B. Westphalen, Xiaowei Chen, Yoku Hayakawa, Daniel L.Worthley, Marina Macchini, Samuel Asfaha, Axel Kleespies, Helen Remotti, Kenneth P.Olive, Timothy C. Wang

Background: There is increasing evidence that chronic sympathetic nervous activation canlead to increased noradrenaline levels in the tumor microenvironment. This process has beenassociated with proliferative signals largely mediated by β-adrenergic signaling. However, themechanisms by which adrenergic neurotransmitters are delivered to the tumor microenviron-ment are not well understood. In this study we aimed to investigate the role of locally andsystemically delivered catecholamines to the tumor microenvironment in a well establishedgenetically engineered pancreatic cancer mouse model (PDx1-Cre/KRasG12D (KC) andPDx1-Cre/KRasG12D/Trp53R172H (KPC)). Methods: Adrenergic signaling was induced orinhibited in KC or KPC mice or 3D pancreatic spheres in culture using specific non-selectiveand selective agonists or antagonists. Adrenergic receptor expression was assessed by RT-PCR, Western blot and IHC. Adrenergic signaling was also modeled in vivo applying restraintstress. To elucidate the crosstalk between nerve and cancer cells, pancreatic spheres andcancer cells were co-cultured with dorsal root ganglia and neuronal plasticity was quantifiedby evaluating neurite outgrowth, number of branches and perikaryonal size. Denervationof the pancreas was performed using surgical or chemical neural ablation. Results: Adrenergicsignaling receptors, in particular ADRB2 and ADRB3, are upregulated during pancreaticcancer development. Furthermore, in vitro stimulation of cells harboring an oncogenic KRasmutation displayed a significantly increased sphere forming efficiency. This effect was blockedby the non-specific β-blocker propranolol and the selective β2-blocker ICI 188,551. Stimula-tion of pancreatic sphere formation from KC mice induced by adrenergic signaling fromdorsal root ganglia in vitro was also prevented by selective blockage of the β2-adrenergicsignaling pathway. While surgical or chemical (Botulinum toxin A, 6-hydroxydopamine)denervation of the pancreas appeared not to retard the development precancerous lesionsin KC mice, denervation significantly increased the survival when performed in mice aftera tumor (3-5mm) was detected by high-resolution ultrasound imaging. In contrast, stressaccelerated pancreatic cancer development in KC mice. The effects of stress were preventedby treatment with the selective ADRB2 antagonist ICI 188,551. Conclusions: Taken together,our data suggests that an increased level of stress and an increased level of systemic catechola-mines may promote pancreatic carcinogenesis in its early stages. During this phase, patientsmight benefit from pharmacological inhibition of stress-induced signaling. The progressionof pancreatic cancer seems to depend on the local delivery of catecholamines to the microenvi-ronment, and thus patients might also benefit by additional targeting of the ADRB2 signal-ing pathway.

Su1859

Transforming Growth Factor β Facilitates Early Tumorigenesis Despite PotentTumor Suppressive Effects in Epithelial TissuesDaniel R. Principe, Brian DeCant, Elizabeth A. Wayne, Lauren E. Droske, EmmanMascariñas, Jessica Bauer, Hidayatullah G. Munshi, Paul Grippo, Barbara H. Jung

Background & Aims: In early neoplastic events, TGFβ is often considered a tumor suppressordue to its growth inhibitory effects in epithelial tissues. This is in stark contrast to clinicalobservations suggesting that in advanced disease, TGFβ promotes tumor progression. Whileepithelial TGFβ signaling has been demonstrated to induce anti-proliferative factors suchas p21, we have previously established that animals globally deficient in the type 1 TGFβ-receptor (TGFBR1) are protected against KRAS-induced pancreatic neoplasia. This workidentifies the underlying mechanisms of these phenotypic variations and their relevance tostrategies targeting TGFβ in the clinic. Methods: Murine models with pancreas specificexpression of the human mutant KRAS allele (EL-KRAS) were crossed to mice globallydeficient in TGFBR1 (Tgfbr1+/-) or mice expressing a dominant negative version of TGFBR2in epithelial tissues (MT-TGFBR2DN). Blood, lymphatic and epithelial tissues of these animalswere evaluated for expression of T cell lineage/activation markers. Co-cultures of pancreasor colon epithelial cells, with appropriate organ stroma, were probed for downstream targetsof TGFβ via western blotting. TGFβ pathway activation and downstream p21 inductionwere examined in a cohort of human pancreatic cancer specimens via immunohistochemicalanalysis. Results: For the first time, we present evidence that despite the tumor suppressiverole in regulating the proliferation of neoplastic cells, the effects of TGFβ in the tumormicroenvironment (TME) dominate its epithelial functions. Stromal overexpression of TGFβappears to promote the infiltration of F4/80+ and GR1+ inflammatory cells, yet suppresseslesion specific GranzymeB and cleaved Caspase-3 accumulation, thereby facilitating tumori-genesis by repressing T cell mediated cytotoxicity. Clinically, pancreatic cancers harboringSMAD4 deficiency had highly significant reductions in p21 indicating a loss of tumorsuppressive TGFβ signaling, though this was not related to GranzymeB involvement. Similarstudies in human colon cancer patients are ongoing. Conclusions: Our results indicatethat while EL-KRAS/MT-TGFBR2DN and EL-KRAS/Tgfbr1+/- mice have similar signalingdeficiencies with respect to the pancreatic epithelium, systemic Tgfbr1 haploinsufficiencyrestores functional anti-tumor immunity by disrupting stromal-mediated TGFβ overexpres-sion. Furthermore, we have used this information to identify a subset of patients who havelost the beneficial aspects of epithelial TGFβ signaling, yet harbor pathological TGFβ signalingin the TME. We propose that this group of patients will respond optimally to therapeuticstrategies targeting TGFβ.

AG

AA

bst

ract

s