Immunohistochemistry and analysis data for staining of tight junction proteins

For immunohistochemistry, slides were processed using an automated platform (Leica Bond RXm, Leica Biosystems, Deer Park, IL, USA). Targets were Claudin 1 (Abcam, Cat. #ab15098), Claudin 2 (Invitrogen, Cat. #51-6100), Occludin (Invitrogen, Cat. #33-1500, Clone OC-3F10), E-Cadherin (BD Biosciences, Cat. #610181), and MAC-387 (Invitrogen, Cat. #MA1-80446). Complete methodology for each marker can be found in Supplementary Table 3. Slides were scanned using a NanoZoomer S60 (Hamamatsu Photonics, Shizuoka, Japan) and visualized using NDP.view2 version 2.9.29 (Hamamatsu Photonics). Stained images were scored both objectively and subjectively. For objective scoring, images were uploaded into QuPath version 0.5.1 (Bankhead et al., 2017). Areas of the mucosa were selected by a trained operator (JDY) with supervision by a board-certified pathologist (JPC) to exclude luminal contents and processing artifacts. Tissues were only considered if selected areas totaled 2.37 mm2. Pixelwise H-scores were calculated using modified scripts based on publicly available scripts implemented by Peter Bankhead and used scripts can be found in Supplementary File 2. Subjective scores were generated by a panel of three board-certified pathologists (JPC, JE, KL) following recommendations by Meyerholz and Beck (2018). Pathologists were instructed to score the epithelium and the lamina propria (LP) of each tissue separately. In brief, scores were created based off the approximate percentage of cells with immunoreactivity (0-100%) and the estimated intensity of the immunoreactivity (mild (1), moderate (2), marked (3)). By multiplying the % positivity of cells with the percentage of staining that was mild, moderate, and marked, which were represented by the numerical values 1, 2, and 3, respectively, the H-score for the epithelium and lamina propria were determined by all three pathologists independently.