Formalin-fixed paraffin-embedded (FFPE) colon samples were obtained from IBD patients who underwent small intestine/colon resection due to the severity of disease. There were eight patients with CD (four female and four male patients, 27–84 years of age [average age, 45 years]) and five patients with UC (four female and one male patients, 18–51 years of age [average age, 38 years]). Control FFPE colon samples were obtained from patients undergoing colon resection due to carcinoma (eight non-IBD control subjects: one female and seven male control subjects, 50–90 years of age [average age, 72 years]). Disease activity in IBD patients was evaluated using a published histologic score method (Naini and Cortina, 2012), excluding neuroendocrine hyperplasia (Supplementary Table S1A). Human tissue specimens used for lamina propria mononuclear cell (LPMC) isolation for Western blot and flow cytometry analysis included surgical colon resections and mucosal biopsies from 5 UC patients (four female and one male patients, 39–77 years of age [average age, 53 years]) and two CD patients (male, aged 35 and 67 years).

Two DSS experiments were included into the study. Ten C57Bl6N male mice (Charles River, Italy) were exposed to 3% DSS (MP Biomedicals, 38–50 kDa) in drinking water every working day for 3 weeks (5 days a week DSS and 2 days without DSS). Nine control animals were drinking only water. At the beginning of the studies, mice were 7–8 weeks old and weighed 19–22 g (median, 21 g). Disease activity in the DSS model was evaluated using a histologic score method published by Nishitani et al. (2009) (Supplementary Table S1B).

Three adoptive transfer experiments have been performed. An adoptive transfer model of chronic colitis was induced in C.B-17 female SCID mice (Charles River, France). At the beginning of the experiments, mice were 7–10 weeks old and weighed 15–21 g (median, 18 g). Purified CD4⁺CD25⁻CD62L⁺ T cells were isolated from spleen cells of female BALB/c mice (Charles River) using “CD4+CD25⁻CD62L⁺ T Cell Isolation Kit II, mouse” (Miltenyi Biotec) according to manufacturer instructions. Purified CD4⁺CD25⁻CD62L⁺ cells (>95% pure and >90% viable) were resuspended in sterile phosphate-buffered saline (PBS) (5 × 106 cells/mL) and transferred into SCID mice intraperitoneally (i.p.). Mice with more than 5% of CD4⁺ T cells within total blood leukocytes at day 14 (D14) after transfer were included in the study. Twenty control SCID naive animals and 35 diseased animals were sacrificed at 4, 6, and 8 weeks after transfer. Disease activity in the adoptive transfer model was evaluated using a histologic score method published by Laroux et al. (2004) (Supplementary Table S1C).

Immunohistochemistry (IHC) was performed on FFPE tissue using the following primary antibodies: claudin-1 (Abcam, ab56417, 1:250), claudin-2 (Abcam, ab125293, 1:75), claudin-3 (GeneTex, GTX15102, 1:50), claudin-4 (Abcam, ab53156, 1:1,000), and claudin-8 (Abcam, ab183738, 1:1,000) followed by DAKO EnVision Detection System Kit or R&D System Kit (Supplementary Figure S1). Whole slide for each human sample/animal sample was evaluated. Human samples were at least 1 cm long. Animal samples represented the whole colon. Colon was sampled as proximal and distal parts; each part was cut into three to four pieces; whole material was embedded. On one slide, there have been three to four transversal cuts from proximal colon and three to four cuts from the distal colon.

Total RNA was isolated from the FFPE blocks with the use of RNeasy FFPE Kit, as recommended by manufacturer (Qiagen). First-strand cDNA was obtained by reverse transcription using a commercial SuperScript III First-Strand Synthesis System according to the manufacturer’s protocol (Invitrogen) and random hexamers as primers.

Quantitative real-time polymerase chain reaction (qRT-PCR) for detection of claudin-1, claudin-2, claudin-3, claudin-4, and claudin-8 mRNA expression was performed on an Applied Biosystems 7300 Real-Time PCR system using the TaqMan method and glycerin-aldehyde-3-phosphate (GAPDH) as a housekeeping gene. Only expression of human claudin-3 expression was measured using SyberGreen technology. Primers and probes (Table 1) were designed in Primer Express software version 3.0 and used at validated concentrations. All primers and TaqMan probes were used at a final concentration of 100 nM qRT-PCR was performed in a 30 µL final volume containing 1× Applied Biosystems TaqMan PCR Master Mix or 1× Applied Biosystems SYBR Green PCR Master Mix, 100 nM primers and probes, and 1 µL of cDNA. Amplification was performed using the following cycling conditions: 2 min at 50°C, 10 min at 95°C, and 45 two-step cycles of 15 s at 95°C and 60 s at 60°C. Triplicate qRT-PCR analyses were executed for each sample, and the obtained threshold cycle (CT/Ct) values were averaged. The relative level of tested genes mRNA was then normalized to the GAPDH CT value to give the ΔCT value for each sample.

Isolation of LPMCs was performed using a lamina propria dissociation kit (Miltenyi Biotec, 130-097-410) following the manufacturer’s instructions. Briefly, the method consisted of epithelial cell removal by incubation of tissue in a solution containing 1 mM DTT and 5 mM EDTA, followed by enzymatic digestion (enzymes included in the kit) and mechanical dissociation using a gentle MACS dissociator (Miltenyi Biotec, 130-093-235). In case of resection specimens, from which higher cell yield was obtained, mononuclear cells were further purified by Lymphoprep density gradient centrifugation. Isolated cells were cryopreserved in medium containing 70% fetal bovine serum (FBS), 20% RPMI 1640, and 10% dimethyl sulfoxide, following the protocol described previously for PBMCs (Kreher et al., 2003).

For Western blot analysis, LPMCs were lysed in PBS with 1% Triton-X (Sigma, T9284), phosphatase inhibitor (Roche, 04906837001), and protease inhibitor (Roche, 04693124001). Protein concentration in lysates was determined using Pierce BCA Protein Assay Kit (23,225; Thermo Fisher Scientific, Waltham, MA, USA) and adjusted to 0.5 mg/mL. Claudin-1, claudin-2, and claudin-4 and GAPDH protein expression was determined by use of Wes System, Wes separation 12–230 kDa capillary cartridges (SM-W004), and standard pack (PS-ST01EZ-8) (all from Protein Simple) according to manufacturer’s instructions. The following primary antibodies were used: GAPDH (Abcam, ab9485, 1:1,000), claudin-1 (Abcam, ab15098, 1:50), claudin-2 (Abcam, ab53032, 1:100) and claudin-4 (Abcam, ab53156, 1:100).

For flow cytometry, LPMCs were thawed and plated in 96-well U-bottom plates for staining. Cells were washed twice with PBS and stained for live/dead discrimination using Invitrogen LIVE/DEAD Fixable Aqua dead cell stain kit (Thermo Fisher Scientific). Blocking of Fc receptors was performed using Human BD Fc Block (Becton–Dickinson, Franklin Lakes, NJ, USA) and was followed by surface staining. Cells were stained with fluorescently labeled monoclonal antibodies directed to claudin-1 BV421 (clone 421203, BD), CD56 BV605 (clone HCD56; BioLegend, San Diego, CA, USA), CD19 BV711 (clone 2H7; BioLegend), CD4 PerCP-Cy5.5 (clone RPA-T4; Thermo Fisher Scientific), claudin-2 PE (polyclonal; Lifespan Biosciences, Seattle, WA, USA), claudin-3 PE (clone REA751; Miltenyi Biotech, Bergisch Gladbach, Germany), claudin-4 PE (clone REA898; Miltenyi Biotech), HLA-DR PE-Cy7 (clone LN3; Thermo Fisher Scientific), CD3 APC (clone SK7; Thermo Fisher Scientific), CD14 AF700 (clone HCD14; BioLegend), and CD45 APC-Cy7 (clone HI30; BioLegend) at 4°C for 30 min. After staining, cells were washed in PBS with 2% FBS and 2 mM EDTA and acquired immediately. Fluorescence-minus-one (FMO) controls were prepared for each claudin. Gates were set using FMO controls with <1% background accepted. Gating strategy for flow cytometry analysis is given in (Supplementary Figure S2). Data were collected using an Attune NxT flow cytometer (Thermo Fisher Scientific) and then analyzed using FlowJo software version 10.7 (FlowJo LLC, Ashland, OR, USA). Graphs were generated using GraphPad Prism software version 9.1 (GraphPad Software, San Diego, CA, USA).

The Mann–Whitney nonparametric test was used for statistical evaluation of histology score data. Differences in claudin gene expression between healthy and UC or CD patients, as well as between naive animals and animals with induced colitis, were deduced using t test. The level of significance was set to p < 0.05 in all cases. GraphPad Prism Software (San Diego, CA, USA) was employed.

In control human colon, TJ proteins claudin-1, claudin-2, claudin-3, claudin-4, and claudin-8 were variably expressed along the crypt axis and differently distributed along the epithelial cell membrane, reflecting the differences of claudin expression by absorptive and secretory epithelial cells (Figure 1A-Control). Claudin-3 and claudin-4 were expressed along the whole crypt axis, whereas claudin-1 and claudin-8 expression increased toward crypt surface (Figure 1A, Control). Claudin-2 was variably present in colon tissue from different donors and ranged from faint to strong. Absorptive cells expressed claudin-1 (Figure 2A), claudin-3 (Figure 2B), claudin-4 (Figure 2C; Supplementary Figure S3D), and claudin-8 (Figure 2D; Supplementary Figure S3E) at TJs and along basolateral membranes. Claudin-3, claudin-4, claudin-8 (Supplementary Figure S3F), and focally claudin-1 (Supplementary Figures S3B,C) formed TJs of goblet cells. In addition, claudin-3 and claudin-4 (Figures 2B,C), as well focally claudin-1 (Supplementary Figure S3C) and claudin-8, were expressed linearly along the basolateral membrane of goblet cells. Interindividual differences were noted in focal/segmental claudin-1 and claudin-8 expression by goblet cells. Tuft cells were positive for claudin-1 (Supplementary Figure S3A). Apart from epithelial cells, claudins were expressed by cells of mesoderm origin. Claudin-2 was expressed by mononuclear cells in lamina propria, a subset of lymphocytes within lymphoid tissue and endothelial cells. Claudin-3 was expressed by a subset of cells within the germinal center of lymphoid follicles (Supplementary Figures S3G,H) and high endothelial venules’ (HEVs’) endothelial cells (Supplementary Figure S3I).

In severe IBD cases analyzed in this study (Figure 3C; Supplementary Table S1A), the epithelial proliferation rate and cell differentiation were profoundly disturbed. This was manifested through a plethora of crypt phenotypes, each characterized by a different claudin signature (Figure 1A, UC, CD). Nevertheless, some characteristic patterns were still observed. Lineage claudin expression signature was not altered by the shift of differentiation toward absorptive cells (Figures 2E–H). Claudin-2 was preponderantly, but not exclusively, expressed by cells within an enlarged crypt proliferative zone (Figure 4A) and at sites of “budding” (Figures 5A,C). Claudin-2 expression pattern was heterogeneous and varied among crypts, even between morphologically identical cells within the same crypt (Figures 5B,D). On the other hand, decreased claudin expression by epithelial cells was observed at sites of neutrophil infiltration. Furthermore, intestinal surface epithelium in a vicinity of mucosal erosions and ulcers, in addition to decreased membranous claudin-3 expression, showed internalization of claudin-3 (Figures 6A,B) and claudin-8 into the cell cytoplasm.

Expression of claudin-1 (Figures 7A,D), claudin-2 (Figures 7B,E), and claudin-4 (Figures 7C,F) by inflammatory cells infiltrating into IBD mucosa was noticed in tissue examined by IHC. Protein expression of claudins in isolated LPMCs was confirmed by Western blot analysis and flow cytometry of biopsies obtained from seven IBD donors (Figure 8). Within ectopic tertiary lymphoid tissue, claudin-2 was expressed by lymphocyte subpopulation(s) (Figures 7B,E), and claudin-3– and claudin-4–positive cells were found in germinal centers (Figures 7C,F). In CD, epithelioid macrophages and multinuclear giant cells within granuloma expressed claudin-2 (Figure 9B), while being claudin-1 (Figure 9A) and claudin-3 (Figure 9C) negative.

In surgically removed intestine from patients with CD, claudin-1 and claudin-2 mRNA levels (Figure 1B) were significantly higher than in non-IBD control intestine. Levels of claudin-4 and claudin-8 mRNA remained unchanged (Figure 1B). Claudin-3 mRNA was below detection level in all CD samples. In resected colon samples obtained from patients with UC, claudin-1, claudin-2, claudin-4, and claudin-8 mRNA levels were comparable to non-IBD control intestine (Figure 1B). On the other hand, a tendency toward the lower level of claudin-3 mRNA was observed in UC patients (Figure 1B).

In C57Bl/6N naive colon (Figure 10A), absorptive epithelial cells expressed claudin-3 along lateral membranes, claudin-4 was present at a very low level within the cytoplasm, and claudin-8 was expressed unevenly within the cytoplasm and/or along lateral membranes. Colon absorptive epithelial cells in C57Bl/6N mice did not express claudin-1. Claudin-1 was present within the cytoplasm of crypt secretory cells. All cell types within crypt expressed claudin-3 along the whole length of the basolateral membranes. Claudin-2 was positive in the stem cell niche at the crypt base, situated at TJs and along the basolateral membrane. The claudin-4 expression pattern in crypt was more pronounced in the proliferative region. Claudin-8 was mostly absent from crypt cells. Furthermore, claudin-4 was unevenly expressed by epithelial cells overlying mucosa-associated lymphoid tissue (MALT). The cell subpopulation within MALT expressed claudin-2 and/or claudin-4. The macrophages in lamina propria were strongly claudin-4 positive. Endothelial cells were claudin-2 positive, but claudin-1, claudin-3, claudin-4, and claudin-8 negative.

DSS-induced murine chronic colitis (Figure 3A) was characterized by weight loss (Figure 3B), colon shortening and chronic active inflammation, mucosa erosions, and formation of ulcers that was reflected through a total histology activity score (Figure 3C, Supplementary Table S1B). Epithelial cells situated at the edge of epithelial defects were claudin-1 and claudin-3 positive along membranes and within the cytoplasm (Figure 6C). In noninflamed mucosa, claudin-3 occasionally formed small intracytoplasmic vacuoles within absorptive epithelial cells. In majority of the epithelial cells, claudin-4 was present at a relatively low level, whereas claudin-8 expression level within epithelial cell compartment increased (Figure 10B). Within the crypt proliferative zone, claudin-2 was overexpressed (Figure 4B; Figure 10B), claudin-3 was localized along lateral membranes, whereas claudin-8 was moderately expressed within the subset of the crypts (Figure 10B). Lymph follicles were strongly claudin-4 positive (Figure 10B; Figure 11C), whereas various infiltrating inflammatory cells in lamina propria were either claudin-1 (Figure 10B; Figure 11A), claudin-2 (Figure 11B), claudin-4 (Figure 11C), or claudin-8 positive (Figure 10B); among those, macrophages expressed claudin-1 (Figure 10B; Figure 11A), claudin-4 (Figure 10B; Figure 11C), and claudin-8 (Figure 10B). qRT-PCR analysis of the mouse colon revealed significantly greater expression of claudin-4 and significantly lower expression of claudin-2, claudin-3, and claudin-8 in DSS-treated mice versus healthy controls when expressed over housekeeping gene (Figure 10C).

In SCID murine nondiseased colon, absorptive epithelial cells expressed claudin-1 and claudin-3 along the lateral membranes (Figure 12A). Claudin-8 was unevenly expressed among surface and crypt cells, mainly within the cytoplasm. All crypt cell-types showed membranous claudin-3 expression (Figure 12A). Cells in the proliferative zone, unlike other crypt cells, were strongly claudin-2 positive. Claudin-2 was situated at TJs, along the whole lateral membrane or within the cytoplasm of cells in crypt proliferative zone (Figure 12A). Microfold (M) cells within the epithelium covering the lymphoid tissue expressed claudin-4 along the lateral membranes and occasionally within the cytoplasm, whereas MALT cells were claudin-4 negative (Figure 12A). On the other hand, MALT-associated epithelial cells and lymphoid cells within the MALT were claudin-8 positive (Figure 12A). Claudin-8 was mainly found within the cell cytoplasm. Endothelial cells were claudin-2 positive.

CD4⁺CD25⁻CD62L⁺ T cell transfer into SCID mice induced colitis (Figure 3A) that was manifested by weight loss (Figure 3B), colon elongation, and by mucosal inflammation accompanied by crypt loss due to lymphoid infiltration into the crypt epithelium, diffuse crypt elongation with increased crypt proliferative zone, and reduced goblet cell number (Figure 12B) and was reflected through a total histology activity score (Figure 3C; Supplementary Table S1C). Absorptive cells lining mucosa surface expressed claudin-1 and claudin-3 along the basolateral membranes (Figure 12B). From 4 weeks after transfer onward, cells lining elongated crypts that were lacking goblet cells acquired claudin-1 and claudin-3 expression pattern characteristic for surface absorptive cells (Figure 12B). Most of the surface, crypt body, and crypt base epithelial cells retained cytoplasmic claudin-8 positivity at variable levels (Figure 12B). Claudin-8 translocated toward the membrane in a subset of those cells. In diseased animals, M cells remained claudin-4 positive. Claudin-4–positive cells “outside” MALT-covering epithelium were found. Among epithelial cells covering mucosa surface and in crypts, small clusters of epithelial cells expressed claudin-4 along basolateral membrane (Figure 12B). The crypt proliferative zone was steadily increasing over the posttransfer observation period and was strongly claudin-2 positive (Figure 4C; Figure 11E; Figure 12B). At sites of lymphocytic invasion, epithelial cells lost claudin expression. In adoptive transfer model of IBD subsets of infiltrating mononuclear cells, among other macrophages were claudin-1 (Figure 11D; Figure 12B), claudin-2 (Figure 11E; Figure 12B), claudin-4 (Figure 11F), and/or claudin-8 positive. qRT-PCR analysis of the mouse colon revealed lower expression of claudin-2 and claudin-3 mRNA in mice with adoptive transfer versus healthy controls when expressed over housekeeping gene, whereas claudin-1, claudin-4, and claudin-8 mRNA expression did not significantly change (Figure 12C).