AG

AA

bst

ract

sfeatures irregular thickening of epithelium and variable width of capillaries, the % of agree-ment was 41.5% and 44.5%, respectively, while estimates of kappa were 0.12 and 0.11,respectively. Discussion: In this study we attempted to create classification criteria foradvanced adenomas of the colon using selected features during a preliminary review ofrandomized pCLE videos. Overall sensitivity was low, albeit improving when diagnoses werechanged to advanced when confidence level was low. Despite this, NPV is high due to thelow prevalence of advanced neoplasia. Further refinement in the classification scheme isrequired before a diagnose and discard strategy can be implemented in routine clinicalpractice. The moderate inter observer agreement for certain feature does have promisingimplications. However, further studies are needed to confirm these findings.

586

Increased Intestinal Epithelial Cell Shedding in a Rodent Model ofInflammatory Bowel Disease - Quantitative Analysis Using Confocal LaserEndomicroscopy, Confocal Microscopy and Light MicroscopyStephanie J. Mah, Jan K. Rudzinski, Hai Y. Bao, Aducio Thiesen, Eytan Wine, KarenMadsen, Richard N. Fedorak, Julia J. Liu

Purpose: Epithelial gaps are formed in the gastrointestinal tract when epithelial cells areshed, and can be observed in patients and rodents using confocal laser endomicroscopy(CLE). Previous studies have shown that epithelial gaps occupied 3% of spaces in the smallintestine of healthy rodents. We hypothesize that the rate of epithelial cell shedding isincreased in IBD model, which can be observed as increased density of epithelial gaps. Thisincrease in gap density in turn, may contribute to increased antigen presentation. Thepurpose of our study was to quantitate intestinal epithelial gap density in a murine IBDmodel (interleukin 10 knockout or IL 10 -/-) using CLE, confocal microscopy (CM) andlight microscopy (LM), and correlate the gap density with macromolecular permeability.Methods: The terminal ileum of 129 SV/Ev (control), IL 10 -/- were imaged with CLE , CMand LM. CLE were performed using Optiscan confocal endomicroscope on exteriorizedintestine stained with acriflavine. For CM, intestinal tissues were stained with DAPI (nuclear)and Phalloidin (actin), and imaged using the Quorum WaveFX spinning disk confocalmicroscope. Cross-sectional CLE/CM images were reconstructed in 3D for gap and cellcounts. Gap density was defined as the total number of epithelial gaps per 1000 cells countedfrom a minimum of 5 villi per animal. For LM, frozen sections of the intestine were stainedwith alcian blue and nuclear fast red, 10 villi from each rodent were counted for gaps andcells for gap density. For macromolecular permeability, FITC - labelled dextran was gavagedafter an overnight fast and blood samples were collected after 4 hr for serum FITC-dextrandetermination. Results: Epithelial gaps appeared as dark, irregular areas surrounded byadjacent cells on CLE and CM. For control(n=10) and IL 10 -/-(n=10) mice, the CLE-determined gap density (mean±SE) was 9.5±1.3 and 20.6±2.1 gaps/1000 cells (P <0.001),respectively; the CM-determined gap density was 7.3±1.3 and 22.8±6.2 gaps/1000 cells (P=0.03), respectively. Spearman's correlation coefficient for CLE and CM gap density was 0.88(95% C.I: 0.72 to 0.95). The LM-determined gap density was 29.2±5.9 and 51.5±6.4 gaps/1000 cells (P <0.0001), respectively. The gap density of 29.2 gaps/1000 cells correspondedto 2.92% of cell space, similar to previous reports of 3% cell spaces. Serum dextran concentra-tion in control and IL 10 -/- mice were 0.59±0.05 ug/mL, and 1.5±0.3 ug/mL (P =0.001),respectively, implying increased permeability and antigen presentation in the IBD mouse.Conclusion: Increased intestinal epithelial gap density was observed in the normal appearingterminal ileum of IBD mice compared to controls. This increased epithelial gap density wassimilarly observed with CLE, CM, and LM, which was further associated with increase inpermeability to macromolecules.

587

Efficacy of Autofluorescence Imaging and a Transparent Hood for Detection ofColorectal Neoplasms: A 2x2 Prospective Randomized TrialYoji Takeuchi, Noboru Hanaoka, Masao Hanafusa, Hiroyasu Iishi, Ryu Ishihara, NoriyaUedo

Background & Aims: Detecting and removing colorectal neoplasms improve the prognosisof patients with colorectal cancer. Colonoscopy is one of the most reliable methods fordetection of neoplasms, but it can overlook some lesions. In this study, we sought to evaluatethe efficacy of autofluorescence imaging (AFI) with a transparent hood for detection ofcolorectal neoplasms. Methods:We conducted a 2×2 factorial designed, prospective, random-ized controlled trial in a tertiary cancer center to investigate the efficacy of AFI with atransparent hood. Five hundred and sixty-one patients undergoing screening colonoscopy[for investigation of a positive fecal occult blood testing (FOBT)] or who were referred forsurveillance colonoscopy (for follow-up post-endoscopic resection of colorectal neoplasms),were enrolled and allocated to four groups: (1) white light imaging (WLI) alone: colonoscopyusingWLI without a transparent hood; (2) WLI+TH: colonoscopy using WLI with a transpar-ent hood; (3) AFI alone: colonoscopy using AFI without a transparent hood; and (4) AFI+TH:colonoscopy using AFI with a transparent hood. Eight colonoscopists investigated patientsusing each allocated method. The primary endpoint was the difference in neoplasm detectionrate (number of detected neoplasms per patient) between the WLI group and AFI+TH group.Results: Neoplasm detection rate (95% confidence interval) in the AFI+TH group wassignificantly higher than in the WLI alone group [1.96 (1.50-2.42) vs 1.19 (0.93-1.44), P =0.023, Turkey-Kramer multiple comparison test]. Relative detection ratios [RDR (95%CI)]based on Poisson regression model were significantly increased by mounting a TH, foroverall detected lesion [1.49(1.25-1.78)], overall neoplasm [1.45 (1.18-1.77)] and non-neoplastic lesion [1.66 (1.13-2.43)]. RDR for polypoid neoplasms was significantly increasedby mounting a TH [1.69 (1.34-2.12)], and RDR for flat neoplasms was significantly increasedby AFI observation [1.83 (1.24-2.71)]. RDR for neoplasm in proximal was significantlyincreased by AFI observation [1.37 (1.05-1.79)] and RDR for neoplasm in distal colon wassignificantly increased by mounting a TH [1.60 (1.23-2.08)]. Conclusion: AFI colonoscopywith a transparent hood detected significantly more colorectal neoplasms than did conven-tional WLI colonoscopy without a transparent hood. AFI and TH are both efficacious fordetection of colorectal neoplasms, using different complementary mechanisms.(UMIN Clin-ical Trials Registry number, UMIN000001473)

S-108AGA Abstracts

588

External Validation of Novel Probe-Based Confocal Laser Endomicroscopy(pCLE) Criteria for the Diagnosis of Dysplasia in Barrett's Esophagus (BE)Srinivas Gaddam, Julian A. Abrams, Emmanuel Coron, Mathieu de Preville, MontherBajbouj, Neil Gupta, Sachin B. Wani, Amit Rastogi, Ajay Bansal, Jean Paul Galmiche,Charles J. Lightdale, Alexander Meining, Prateek Sharma

Background:Novel pCLE criteria for diagnosis of dysplasia in BE have been recently developedand internally validated. Aim:To externally validate novel pCLE criteria by assessing theirperformance among 6 international BE/pCLE experts by evaluating overall accuracy andinter-observer agreement (IOA). Methods:pCLE (Mauna Kea Technologies, Paris, France)video sequences along with corresponding histology, were acquired from 1 tertiary referralsite evaluating role of pCLE for BE neoplasia. These videos were first reviewed by a pCLEexperienced GI and a GI pathologist, who devised novel pCLE criteria for dysplasia in BE.Criteria were refined by blinded review of 30 pCLE videos by 2 additional experts. 6 criteriadevised for diagnosing dysplasia in BEwere: 1)epithelial surface: saw-toothed 2)cells: enlarged3)cells: pleomorphic 4)glands: not equidistant 5)glands: unequally in size and shape 6)gobletcells: not easily identifiable. These criteria were internally validated among experts and non-experts (previously reported) with good accuracy (78.9%) and IOA (k=0.62). In this study,the same videos were independently evaluated by 6 international BE/ pCLE experts (notpart of the development or internal validation). Training for above criteria was providedusing 10 high quality images [5 dysplastic and 5 nondysplastic). Histologic diagnosis ofbiopsies corresponding to areas of video sequences was the gold standard; diagnoses of highgrade dysplasia (HGD) and cancer were grouped as dysplasia. Each video was evaluated forits quality [1 (poor) to 5 (excellent)], diagnosis (dysplastic vs. nondysplastic), assessor'sconfidence in making diagnosis (high or low), and confidence to “not take a biopsy”. Accuracyof predicting histology was calculated and IOA was assessed using kappa statistics. Results:75videos [non dysplastic BE 45, HGD 28, cancer 2] were assessed by 6 experts. Mean videoquality was 3.6 [range 1-5]. Assessors had high level of confidence in 58.8% videos andanswered “confident not to take a biopsy” in 47.5% of videos. Overall accuracy of pCLEfor dysplasia was 78.4%(95%CI 74.4-81.9); similar to the previously reported accuracy(78.9%). Accuracy rates were higher when endoscopist was “confident not to take a biopsy”(92.9% vs. 65.2%, p<0.01). Overall agreement was moderate with k=0.51 (95%CI 0.29-0.61); not statistically different fromprevious results [k=0.62]. IOA if assessors were confidentof their diagnosis was [(k=0.91(95% CI 0.81-0.97) vs. k=0.43(0.33-0.55)p<0.01] Conclu-sion:This study demonstrates that overall accuracy using these newly developed and validatedpCLE criteria for dysplasia in BE pts was >90% if the endoscopist is confident of makingthe diagnosis along with a moderate degree of agreement. These criteria have been externallyvalidated and should be used in future pCLE studies for diagnosis of dysplasia in BE pts.Comparison of Accuracy and Kappa between the internal and external validation studies

589

IRGM1 Deficiency Selectively Affects Intestinal Paneth Cell Morphology andFunctionBo Liu, Ajay S. Gulati, Lisa C. Holt, Guoling Luo, Gregory A. Taylor, Ryan B. Sartor

Background: Immunity-related GTPaseM (IRGM) gene polymorphisms are linked to Crohn'sdisease. IRGM and its mouse homologue, Irgm1, mediate autophagic defenses againstintracellular pathogens. Perturbed autophagy contributes to intestinal inflammation andalters the function of Paneth cells, which regulate gut microbiota via secretion of antimicrobialpeptides such as lysozyme and defensins. The roles of IRGM and Irgm1 in colitis and Panethcell function have not been explored experimentally. Hypothesis: Irgm1 modulates intestinalPaneth cell morphology and function. Methods: Intestines were harvested from Irgm1deficient (KO) and wild-type (WT) mice following DSS challenge. Paneth cell distribution,granule morphology and autophagasome structure were analyzed by light and transmissionelectron microscopy (TEM). Antimicrobial peptide mRNA expression and autophagy LC3protein were assayed by qRT-PCR and immunohistochemistry. Results: While crypt heightand villus length of the ileum and colon in KO mice were normal, there were obviousPaneth cell morphological abnormalities even before exposure to DSS. Paneth cells in KOtissues were abnormally located in the middle 1/3 of crypts, and intermediate cells (withPaneth and goblet cell features) and granular goblet cells were found diffusely throughoutthe small intestinal crypts and villi. The number of Paneth cells containing small, dysmorphicgranules in each crypt was increased (4.6 ± 1.9 vs. WT 1.2± 0.8, p<0.0001), while cellswith normal-appearing granules were decreased. DSS treatment potentiated aberrant distribu-tion and granule abnormalities of KO Paneth cells compared with untreated KO mice.Lysozyme and LC3 immunohistochemical positive granules localized to crypt-base Panethcells, intermediate cells and goblet cells throughout the crypts and villi in DSS-treated KOmice. The number of LC3 positive granules per crypt was higher in KO mice (19.9± 9 vs.WT 6.3 ± 4.2, p<0.0001), with a reduction of lysozyme positive granules per Paneth cell(4.9 ± 1.7 vs.WT 7.6 ±3.8, P<0.046). DSS-treated KOmice also showed a ~3-fold reduction oflysozyme mRNA expression (p<0.005) and a ~2-fold reduction of Defcr20 mRNA expression(p<0.001). TEM confirmed decreased secretory granule size within KO Paneth cells, expan-sion of the peripheral halo around dense granules and smaller electron-dense cores. Frequentdouble membrane structures (2-6 per granule) were present within the enlarged halo of KOgranules; some were multilamellar and others very dense. These structures appeared to beLC3+ by immunostaining. Conclusion: Marked morphological abnormalities, induction ofautophagy-related structures and decreased antimicrobial peptide production occur in Panethcells of Irgm1 KOmice, suggesting that Irgm1 contributes to Paneth cell mucosal homeostasisvia the autophagy pathway.