IL-22–Independent Protection from Colitis in the Absence of Nkx2.3 Transcription Factor in Mice

The immune system of the intestine must cope with a massive load of alimentary and microbial Ags and distinguish between those to tolerate and those to eliminate. These functions are carried out by two different types of intestinal lymphoid tissues (1). Peyer patches (PPs) in the small intestine and colonic patches in the large intestine form during embryonic development, whereas solitary intestinal lymphoid tissue (SILT) components develop postnatally. The SILT spectrum consists of cryptopatches (CP) containing mainly hematopoietic lineage⁻ cells at the bottom of intestinal crypts (2). Upon environmental stimulation, these leukocyte congregates transform into isolated lymphoid follicles (ILFs) (3, 4). During this process, lymphocytes are recruited to form follicles via MAdCAM-1 expressed on high endothelial venules (HEVs) and lamina propria vessels (5, 6).

In mice, the vascular specification of PPs is critically determined by the Nkx2.3 homeodomain transcription factor in a tissue-specific manner. In the absence of Nkx2.3 in mice, PPs and the spleen lack most of MAdCAM-1 (7, 8) and exhibit a vasculature reminiscent of peripheral lymph nodes with peripheral node addressin (PNAd)⁺ HEVs (9, 10). In addition to its role in vascular specification of intestinal lymphoid tissues, Nkx2.3 also affects the maturation of villi, as its lack causes aberrant villus formation during the embryonic and early postnatal period of gut development. Meanwhile, in adulthood, small intestines of mice lacking Nkx2.3 present with significantly enlarged villi (11). In humans, single nucleotide polymorphisms in the coding region of Nkx2.3 were found to be associated with both Crohn disease (CD) and ulcerative colitis, the two main types of inflammatory bowel disease (IBD) (12, 13). Nkx2.3 has been shown to regulate the expression of PTPN2, a risk gene for CD (14), and VEGF signaling and endothelin 1 production in intestinal microvascular cells (15).

Although these results point to the involvement of Nkx2.3 in IBD, the details of Nkx2.3-mediated mechanisms underlying intestinal inflammatory processes are still unresolved. In this article, we show that Nkx2.3-deficient mice (Nkx2.3^−/−) harbor a SILT spectrum with impaired maturation of ILFs and are protected from dextran sodium sulfate (DSS)–induced colitis. Alleviated colitis is independent of MAdCAM-1, as MAdCAM-1–deficient mice (MAdCAM-1^−/−) develop severe colitis compared with Nkx2.3^−/− mice, whereas it is coupled with enhanced colonic epithelial regeneration. Diminished colitis in Nkx2.3-deficient mice is associated with increased numbers of colonic RORγt⁺ group 3 innate lymphoid cells (ILC3s), Th17 cells, and Foxp3⁺ regulatory T cells (T_regs). Increased IL-22 levels in DSS-treated Nkx2.3^−/− mice did not contribute to protection as IL-22 blockade had no effect on disease pathogenesis. Nkx2.3^−/− hematopoietic cells failed to provide protection, highlighting the importance of Nkx2.3-expressing intestinal stromal cells in the functionality of the colonic epithelium and various lymphoid cells.

BALB/cJ and C57BL/6J mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Nkx2.3^−/− mice (11) were backcrossed through at least 20 generations onto BALB/c background (9). MAdCAM-1^−/− (Madcam1^tm1.2Nwag) mice (16), mCD19CherryLuciferase (CD19CL) transgenic mice (17), and LacZ-Nkx2.3^+/− reporter mice (8) were described previously. eGFP-TgBALB/c mice (18) were maintained in our laboratory. Prior to experiments, mice were accustomed to standard housing conditions for at least 1 wk. Mice were provided ad libitum with a VRF1 (P) diet (Special Diets Services) and fresh water and were kept on aspen bedding (Abedd). Eight- to ten-week-old mice were used throughout experiments. To induce colitis, mice received 2.5% DSS (AppliChem) in drinking water for 7 d. Mice were sacrificed on day 7 for acute or day 14 for subacute colitis. For controls, either littermates were used, or bedding was regularly exchanged between cages. All procedures involving live animals were carried out in accordance with the guidelines set out by the Ethics Committee on Animal Experimentation (University of Pécs, Pécs, Hungary) under license number BA02/2000-16/2015, with approval for the use of genetically modified organisms under license number SF/27-1/2014 issued by the Ministry of Rural Development, Budapest, Hungary. Throughout the experiments, the authors adhered to Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines.

Colons were embedded in OCT medium (Bio-Optica, Milan, Italy), snap-frozen, and sectioned at 8 μm. H&E stainings of Swiss roll colons were performed at the Department of Pathology, University of Pécs, according to standard protocols. Histopathological scoring of H&E samples was performed by an experienced pathologist in a blinded fashion based on a modified version of standard protocols (19), with evaluation of submucosal edema (0–3), polymorphonuclear cell infiltration (0–4), number of goblet cells (0–3), and epithelial integrity (0–3). For SILT ratio analysis, sections were made from colonic Swiss rolls at four different planes distributed evenly apart. SILT structures were identified as follows: CP, Thy-1⁺/B220⁻; immature ILF, Thy-1^−/+/B220⁺/ complement receptor type 2 (CR2)^−/+; and mature ILF, B220⁺/CR2⁺/peanut agglutinin (PNA)⁺. Immunofluorescence was performed as previously described (10) using the following commercially available Abs: goat anti-mouse CXCL13 (R&D Systems) developed with donkey anti-goat FITC (SouthernBiotech), biotinylated CD11c (clone N418; BioLegend) developed with Streptavidin PE (BD Biosciences), and biotinylated PNA (Sigma-Aldrich) developed with AMCA Streptavidin (Vector Laboratories). The following Ab clones were obtained from the American Type Culture Collection or produced in our laboratory, and then the Abs were purified and labeled in our laboratory: CD45-FITC (clone IBL-3/16), rat anti-mouse MAdCAM-1–AF555 (clone MECA-367), rat anti-mouse PNAd-Dylight594 (clone MECA-79, kindly provided by Dr. E. Butcher), B220-CF647 (clone 6B2), Thy-1–AF594 (clone IBL-1), and CD21/35-FITC (clone 7G6). Sections were viewed using an Olympus FLUOVIEW FV1000 laser scanning confocal imaging system or an Olympus BX61 fluorescent microscope.

Swiss rolls containing colonic samples prepared from LacZ-Nkx2.3 reporter mice (8) and BALB/c mice were cryosectioned at 8-μm thickness and fixed in cold acetone. After drying, the endogenous peroxidase activity was quenched by 1 mg/ml phenyl-hydrazine-hydrochloride in PBS for 30 min, followed by washing. Following saturation with 5% BSA/PBS, sections were incubated with rat mAbs against vascular associated protein-1 (VAP-1) Ag (clone 7-88/1, kindly provided by Dr. S. Jalkanen), EpCAM (clone G8.8), endothelial marker IBL-20 (produced at the Department of Immunology and Biotechnology, University of Pécs), or control rat IgG (each at 5 μg/ml) in PBS for 45 min, followed by washing. Bound Abs were detected using the ImmPRESS-HRP goat anti-rat IgG kit (Vector Laboratories) with DAB/H₂O₂ according to the vendor’s recommendations. After washing, slides were incubated with X-gal (1 mg/ml; Boehringer-Mannheim) dissolved in 5 mM K-hexacyanoferrate, 5 mM K-ferrocyanide, and 2 mM Mg-chloride (all from Sigma-Aldrich) in PBS overnight at 37°C, followed by extensive washing in PBS.

Epithelial cell proliferation was detected with the Click-iT Plus EdU Alexa Fluor 488 Imaging Kit (Life Technologies) according to the manufacturer’s protocol. Following 5-ethynyl-2-deoxyuridine (EdU) detection, epithelial cells were identified using rat anti-mouse EpCAM (clone G8.8) developed with PE-labeled anti-rat (BD Biosciences).

Epithelial cell apoptosis was measured using the Click-iT Plus TUNEL Assay for In Situ Apoptosis Detection, Alexa Fluor 488 (Life Technologies), according to the manufacturer’s protocol. Following the TUNEL assay, epithelial cells were identified using rat anti-mouse EpCAM (clone G8.8) developed with PE-labeled anti-rat (BD Biosciences).

Serum IL-22 was detected with the Mouse/Rat IL-22 Quantikine ELISA Kit (R&D Systems) according to the manufacturer’s protocol in duplicates. A standard curve was generated using a four-parameter logistic curve-fit, and concentrations were calculated for each sample.

Lamina propria lymphocytes were isolated using a modified version of the protocol described previously (20). Briefly, colons were cut open longitudinally, washed with PBS, then shaken at 37°C in DMEM containing 15 mM EDTA (Sigma-Aldrich) for 25 min. Intestines were then washed thoroughly with cold PBS, cut into 5-mm pieces, and incubated in DMEM containing 0.6 mg/ml collagenase D and 5 U/ml DNase I (both from Sigma-Aldrich) at 37°C for 20 min. Supernatant was collected and filtered through a 70-μm cell strainer (Greiner Bio-One) and new digestion solution was added. Cycles were repeated until complete digestion of tissues. Mononuclear cells were collected with a 40/80% (w/v) Percoll (Sigma-Aldrich) density gradient centrifugation, washed with PBS, and labeled for flow cytometry using the following commercially available Abs: CD3- allophycocyanin-Cy7 (clone 145-2C11; BD Biosciences), RORγt-AF647 (clone Q31-378; BD Biosciences; intracellular staining was performed according to the manufacturer’s protocol), biotinylated CD11c (clone N418; BioLegend) developed with Streptavidin PE-Cy7 (BioLegend), Thy-1.2 (CD90.2)–PerCPCy5.5 (clone 53-2.1; BioLegend), Foxp3-PE (clone 3G3; Miltenyi Biotec; intracellular staining was performed according to the manufacturer’s protocol), and CD25-PerCpCy5.5 (clone PC61; BD Pharmingen). The following hybridoma clones were obtained from the American Type Culture Collection or developed in our laboratory, and then the Abs were purified and labeled in our laboratory: CD3-FITC (clone KT3), CD4-Cy3 (clone YTS191.1), CD4-AF647 (clone YTS191.1), CD19-FITC (clone 1D3), and CD45-CF647 (clone IBL-3/16). Measurements were performed on a BD FACSCanto II or a FACSCalibur cytometer.

CD19CL transgenic mice (17) crossed with Nkx2.3^−/− mice were anesthetized using 50 mg/kg sodium pentobarbital i.p. and were then i.v. injected with 150 mg/kg d-luciferin (Gold Biotechnology) dissolved in sterile PBS (15 mg/ml). Mice were sacrificed 10 min postinjection, and colons were excised and placed on a petri dish and were then immediately transferred into an IVIS Lumina II (PerkinElmer) imaging system with an imaging chamber set to 37°C. Samples were imaged within 15 min postinjection using the following settings: 180 s acquisition, F/stop = 1, and Binning = 4. Data were analyzed using the Living Image software (PerkinElmer). Regions of interest (ROIs) were drawn automatically using identical luminescence signal thresholds. Total radiance, a calibrated unit of the luminescence (total photon flux per second), was calculated in each ROI, and the total number of ROIs (excluding large ROIs with flux >10⁵/s) per sample was also counted. The ROI area was automatically quantified and expressed as square millimeters.

Four-week-old eGFP-TgBALB/c mice (18) were lethally irradiated at the Department of Oncotherapy, University of Pécs, with 2 × 5.5 Gy administered 6 h apart. Mice received 5 × 10⁶ bone marrow cells in 200 μl of DMEM via tail veins from either BALB/c or Nkx2.3^−/− mice 3 h after the second irradiation. Following reconstitution, mice were provided with ciprofloxacin in drinking water for 2 wk. Chimerism was determined by eGFP/CD45 expression of PBMCs.

Nkx2.3^−/− mice received 2.5% DSS in drinking water for 7 d. Mice were treated with 150 μg of anti–IL-22 mAb (clone 8E11; Genentech) or isotype control (clone GP120:9709; Genentech) i.p. on days 2, 3, 4, 5, and 6. Weights were measured daily. Mice were sacrificed on day 7, and colons were isolated for histology, flow cytometry, and mRNA. Six mice were used per group, and treatments were performed twice.

Total RNA from distal colon homogenates was isolated using NucleoSpin RNA (Macherey-Nagel). Purity and concentration of RNA were analyzed by NanoDrop. cDNA was synthetized using the High Capacity cDNA RT Kit (Life Technologies). RT-PCR was run on an ABI-PRISM 7500 machine in duplicates. TaqMan probes (mGAPDH, mIL-17a, mIL-22, mIFN-γ, mRegIIIβ, mRegIIIγ, mMuc2, mTGF-β, mIL-10) were purchased from Life Technologies. Results are shown as percentage of housekeeping gene (mGAPDH).

Data analysis was performed using IBM SPSS Statistics software (Version 22). Normality of data distribution was assessed by Shapiro–Wilks test. A t test or Mann–Whitney U test was employed to compare two groups with normally distributed and non-normally distributed data, respectively. Data are represented as mean ± SEM. Statistical significance was set at p < 0.05.

Previous studies in our laboratory and elsewhere indicated a substantial alteration of the vascular architecture and lymphocyte composition of PPs in mice lacking Nkx2.3 (7–10). As there are no data on SILT characteristics, we first examined Nkx2.3-deficient mice for the composition of SILT by immunofluorescence analysis of colonic sections. We found that, similarly to PPs, HEVs in colonic patches of Nkx2.3^−/− mice lack MAdCAM-1 (Fig. 1a) and display luminal PNAd, in contrast to abluminal PNAd expression in MAdCAM-1⁺ HEVs of BALB/c mice (Fig. 1b).

To determine the effect of the Nkx2.3 deficiency on the global distribution of colonic B cell clusters, we crossed Nkx2.3^−/− mice with CD19CL transgenic reporter mice (17) with B cell–restricted bioluminescence. Ex vivo measurement of B cell–restricted luciferase activity in whole colon samples revealed a significantly decreased average number of B cell clusters in colons of Nkx2.3-null mutants (Nkx2.3^−/− × mCD19Luc⁺) compared with heterozygous control littermates (Nkx2.3^+/− × mCD19Luc⁺) (Fig. 1c, 1d). However, the average flux (photons per second) of these regions did not differ between strains, and the average area of regions was significantly higher in Nkx2.3^−/− mice compared with heterozygous controls.

To delineate the presence of various SILT components and their maturation stages, multicolor immunofluorescence labeling was employed for Thy-1/CD90, B220, and CR2/CD21 Ags in combination with PNA binding. Using this approach, we identified SILT components at different maturation stages grouped into 1) Thy-1⁺/B220⁻ CPs, 2) Thy-1^+/−/B220⁺ immature ILFs, 3) B220⁺/CR2⁺ immature ILFs containing follicular dendritic cells (FDCs), and 4) B220⁺/CR2⁺/PNA⁺ mature ILFs harboring germinal centers, respectively (Fig. 1e, Supplemental Fig. 1A), in both wild-type and Nkx2.3-deficient mice. Staining for the dendritic cell (DC) marker CD11c and the homeostatic chemokine CXCL13 also revealed normal SILT morphology (Supplemental Fig. 1B, 1C). However, the frequency of mature versus immature components showed a shift toward an immature composition (Fig. 1e) in Nkx2.3^−/− mice. Thus, the percentage of CPs was 14.4 ± 2.2% (mean ± SEM) in BALB/c controls compared with 48.6 ± 4.2% in Nkx2.3^−/− colons. Immature ILFs without FDCs represented 48.2 ± 4.8% in BALB/c compared with 27.8 ± 1.2% in Nkx2.3^−/− mice. Immature ILFs containing FDCs comprised 28.0 ± 3.6% compared with 18.3 ± 3.3% in Nkx2.3^−/− colons. Mature, PNA⁺ ILFs constituted 9.4 ± 2.2% of SILT components in wild-type but only 5.3 ± 2.2% in the absence of Nkx2.3. These results indicate that CPs readily form in the absence of Nkx2.3, but their subsequent differentiation into more mature SILT structures is impaired.

To investigate whether this altered SILT ratio is due to the absence of endothelial MAdCAM-1 in Nkx2.3-deficient mice, we also examined SILT components in MAdCAM-1–null mutant mice on C57BL/6 background (16). In MAdCAM-1^−/− colonic patches, we found consistently detectable luminal PNAd expression, similar to that observed in Nkx2.3^−/− colons (Fig. 1f, 1g). SILT elements with structural similarity to wild-type mice could be identified in MAdCAM-1^−/− mice (Supplemental Fig. 1B, 1C). Interestingly, MAdCAM-1^−/− mice showed an even greater inhibition of the formation of both immature and mature ILFs (Fig. 1e).

The transition of colonic CPs into mature ILFs is associated with a characteristic exchange of leukocyte subsets, in which adult lymphoid tissue inducer-type (LTi) innate lymphoid cells and CXCL13-producing DCs are replaced by B cells (21). In agreement with our previous findings, flow cytometric analysis of colonic cells confirmed significantly higher numbers of Lin⁻/Thy-1⁺/CD4⁺/RORγt⁺ LTi cells (Fig. 1h, 1i) in Nkx2.3^−/− mice and increased CD11c⁺ DCs (Fig. 1j) and lower absolute numbers of CD19⁺ B cells (Fig. 1k) in both Nkx2.3^−/− and MAdCAM-1^−/− mice compared with the respective wild-type controls. The reduction of B cells was more severe in MAdCAM-1^−/− than in Nkx2.3^−/− mice, indicating again the more profound block in the recruitment of B cells to MAdCAM-1^−/− colons. In contrast, both relative and absolute numbers of Th cells were significantly higher in Nkx2.3-deficient mice, whereas in MAdCAM-1^−/− mice we noticed a significant increase in the relative Th cell frequencies only (Supplemental Fig. 1D–F). These findings demonstrate that, in the absence of MAdCAM-1, either indirectly (as a consequence of Nkx2.3 absence) or directly (in MAdCAM-1^−/− mice), the colonic SILT maturation toward ILF is blocked, and the lymphocyte distribution is altered.

Colitis induces the transition of CPs to mature ILFs (22, 23). Therefore, we next used the DSS-induced colitis model to examine SILT development under inflammatory conditions. Nkx2.3-deficient mice were given 2.5% DSS in drinking water for 7 d. Samples were collected on day 7 (acute colitis) and day 14 (subacute colitis). Surprisingly, we found that Nkx2.3^−/− mice were protected from DSS-induced colitis based on their physical symptoms compared with controls. Mutant mice lost only minimal weight, macroscopic rectal bleeding was less frequent, the shortening of the colon was minimal, and survival was 100% (Fig. 2a–e). Histological scoring of H&E-stained Swiss roll sections of colons also revealed a significantly lower pathological score both at day 7 and day 14 in Nkx2.3-deficient mutants compared with BALB/c mice (Fig. 2f).

To examine whether the lack of colitis in adult Nkx2.3-deficient mice was related to the absence of endothelial MAdCAM-1, we investigated MAdCAM-1^−/− mice. Interestingly, in contrast to Nkx2.3^−/− mice, the generalized absence of MAdCAM-1 did not inhibit the development of intestinal inflammation, as MAdCAM-1^−/− mice exhibited severe colitis comparable to wild-type counterparts (Fig. 2a–f).

In wild-type controls, DSS treatment increased the ratio of mature SILT components containing CR2⁺ FDCs with or without PNA⁺ germinal centers (Fig. 2g, 2h). However, in Nkx2.3^−/− mice we observed a delayed increase of mature components by day 7, although the difference between mutant and wild-type mice was less prominent by day 14 (Fig. 2g). Interestingly, MAdCAM-1-deficient mice exhibited a more profound blockade in SILT maturation, as the ratio of CPs decreased and B cell–containing structures increased only by day 14 after the start of treatment (Fig. 2h). These results on different kinetics of colonic B cell expansion were also supported by flow cytometry. The absolute number of B cells showed a low but significant increase by day 7 of colitis in Nkx2.3^−/− mice, whereas DSS-induced colitis in MAdCAM-1^−/− mice led to a significant increase in absolute B cell numbers only by day 14 (Fig. 2i). The absolute B cell numbers were lower in both mutant strains throughout treatments in comparison with the wild-type controls. Furthermore, the absolute number of Th cells was significantly higher in both mutant genotypes than in wild-type mice at day 7. At day 14, Nkx2.3^−/− mice displayed similar absolute numbers of Th cells as BALB/c mice, whereas Th cells were still significantly higher in MAdCAM-1-deficient mice (Supplemental Fig. 2B).

Because both Nkx2.3^−/− and MAdCAM-1^−/− mice exhibit a similar luminal PNAd expression on colonic vessels but react oppositely to DSS treatment, these results indicate that, although the absence of endothelial MAdCAM-1 may be responsible for the immature status of SILT in both mutants, it does not explain the lack of colitis in Nkx2.3^−/− mice.

The imbalance between RORγt⁺ cells (including CD45⁺/CD3⁺/CD4⁺/RORγt⁺ Th17 cells and CD45⁺/CD3⁻/RORγt⁺ ILC3s) and CD3⁺/CD4⁺/Foxp3⁺/CD25⁺ T_regs has been extensively studied in both human IBD and murine models of colitis (24–26). We thus extended our analysis to RORγt⁺ lymphoid cells and T_regs. With flow cytometry, we observed a significantly higher frequency and absolute number of RORγt⁺ Th17 and ILC3 populations (Fig. 3a, 3b, Supplemental Fig. 2C) in the colons of Nkx2.3^−/− mice compared with BALB/c controls at all examined time points. Detailed analysis of colonic ILC3 subsets showed significantly higher relative and absolute numbers of CCR6⁺/RORγt⁺/CD3⁻ cells in Nkx2.3-deficient mutants, whereas we found a reduced frequency of Nkp46⁺/RORγt⁺/CD3⁻ ILC3s compared with wild-type (Fig. 3c). Meanwhile, the frequency and absolute number of CD3⁺/CD4⁺/Foxp3⁺/CD25⁺ T_regs was significantly higher compared with BALB/c mice at day 7 of colitis (Fig. 3d).

We next performed quantitative PCR (qPCR) from colon samples to measure mRNA levels for signature cytokines produced by ILC3s, Th17 cells, and T_regs. We found that both IFN-γ and IL-17a [characteristic of ILC3 and Th17 (27, 28)] were significantly higher in untreated Nkx2.3^−/− mice compared with BALB/c controls (Fig. 3e, 3f). However, treatment of Nkx2.3^−/− mice with DSS did not further increase mRNA levels of these cytokines, in contrast to wild-type mice, and at day 14 IFN-γ and IL-17a were significantly higher in BALB/c controls compared with Nkx2.3^−/− mutant mice. In contrast, in acute colitis (day 7) we observed significantly elevated IL-22 [characteristic of ILC3 and Th17 and shown to have protective effects in colitis (29, 30)] transcript levels in the absence of Nkx2.3 (Fig. 3g). However, during the course of colitis, IL-22 levels were reversed and were higher in wild-type mice than in Nkx2.3-deficient animals at day 14. Meanwhile, mRNA levels for IL-10 (characteristic of T_regs) were significantly higher in Nkx2.3^−/− colons versus BALB/c in untreated controls and at day 7 of DSS treatment, whereas TGF-β2 was lower in untreated mutant mice and was similar to controls throughout DSS treatment (Fig. 3h, 3i).

To examine whether this altered lymphoid cell composition is due to the endothelial absence of MAdCAM-1, we analyzed MAdCAM-1^−/− mice. DSS treatment increased the relative frequency of RORγt⁺ Th17 cells and ILC3s at day 7 without affecting the absolute cell number, whereas at day 14 we observed a significantly higher absolute number of these cells compared with wild-type (Supplemental Fig. 2D, 2E). In untreated MAdCAM-1^−/− colons, T_reg frequency and absolute number was significantly lower compared with controls, whereas at day 7 we observed a significantly increased frequency but not absolute number of T_regs (Supplemental Fig. 2F). In contrast to Nkx2.3^−/− mice, IL-22 was not higher in MAdCAM-1-deficient animals compared with controls. IFN-γ was not affected by DSS treatment, and IL-17a increased, although to a smaller extent than in control mice (Supplemental Fig. 2G–I). Untreated MAdCAM-1-deficient colons had significantly higher levels of IL-10, but this difference was absent upon DSS treatment. We observed no difference between MAdCAM-1^−/− mice and C57BL/6 controls in the expression levels of TGF-β2 (Supplemental Fig. 2J, 2K).

Taken together, these data indicate a shift in the composition of lamina propria lymphocytes, coupled with an altered cytokine production profile in mice lacking Nkx2.3, independent of the absence of endothelial MAdCAM-1.

IL-22 has been shown to have protective effects in colitis by inducing the production of various Reg (regenerating islet-derived protein) family members and mucins important in mucosal healing (29, 30). Accordingly, we found that mRNA levels of RegIIIβ and RegIIIγ were higher in Nkx2.3^−/− mice than in wild-type mice. This difference was most prominent in acute colitis (Fig. 4a, 4b). The expression of Muc2, a crucial mucin in the colon (31), was markedly higher in untreated Nkx2.3^−/− mice than in wild-type mice but did not differ during acute colitis (Fig. 4c).

IL-22R is also expressed on intestinal stem cells and induces mucosal repair and epithelial cell proliferation upon ligand binding (32, 33). Using EdU staining, we found a significantly higher proliferation rate of colonic epithelial cells in untreated Nkx2.3^−/− mice. Treatment with DSS reduced cell proliferation at day 7 in both Nkx2.3 mutant and wild-type genotypes; however, the rate of cell proliferation was still significantly higher in Nkx2.3^−/− colons (Fig. 4d). Furthermore, whereas in BALB/c mice most EdU⁺ cells localized near the bottom of crypts, in Nkx2.3-deficient large intestines, proliferating cells were visible along the whole length of crypts (Fig. 4e). As intestinal epithelial cell turnover is also influenced by cell death, we next analyzed cell apoptosis. In accordance with previously published data (11), we observed no difference in the number of TUNEL⁺ colonic epithelial cells per crypt between Nkx2.3^−/− and BALB/c mice (Supplemental Fig. 3A). However, DSS treatment led to a significantly lower number of apoptotic cells in Nkx2.3^−/− colons compared with BALB/c controls.

To investigate whether the increased IL-22 mRNA is responsible for protection from colitis in Nkx2.3^−/− mice, we first measured IL-22 serum protein levels. Although we found no differences in untreated mice, serum IL-22 was significantly lower in DSS-treated mice compared with wild-type controls (Supplemental Fig. 3B). To further rule out the possibility that IL-22 elevated locally has a role in preventing colonic inflammation in the absence of Nkx2.3, Nkx2.3^−/− mice received antagonistic anti–IL-22 Ab or isotype control during the course of a 7-d DSS treatment. Surprisingly, we observed no differences in colitis symptoms (Fig. 4f, Supplemental Fig. 3C–E) during treatment. Histological scoring of inflammation severity and SILT spectrum evaluation in Swiss roll sections of colons also revealed no difference (Fig. 4g, 4h). Both absolute and relative Th17 and ILC3 did not differ statistically between groups (Supplemental Fig. 3F, 3G). Furthermore, mRNA for IL-17a (Fig. 4i) was significantly higher in anti–IL-22 treated mice, whereas IFN-γ (Supplemental Fig. 3H) and IL-22 (Fig. 4j) were not. mRNA for RegIIIβ and RegIIIγ was significantly lower in anti–IL-22 treated mice (Fig. 4k, 4l), indicating an efficient blockade by the anti–IL-22 Ab. We also observed a significantly lower proliferation rate of colonic epithelial cells in Nkx2.3^−/− mice receiving anti–IL-22 mAb (Fig. 4m).

These results indicate that blocking IL-22 protein does not induce colitis in Nkx2.3^−/− mice, although it results in reduced expression of Reg family members involved in mucosal healing.

To better determine whether the altered cytokine profile in Nkx2.3^−/− mice was a lymphoid cell–intrinsic effect or was caused by Nkx2.3-deficient stromal cells interacting with lymphocytes, we created bone marrow chimeras. Four-week-old eGFP-Tg-BALB/c mice were lethally irradiated and reconstituted with either Nkx2.3^−/− or BALB/c bone marrow. Five weeks after transplantation, chimeric mice (with over 90% chimerism based on the absence of eGFP in CD45⁺ peripheral blood leukocytes) received 2.5% DSS in drinking water for 7 d and were then sacrificed. Weight loss was not different between the two groups (Fig. 5a), and also other parameters of colitis were similar (Fig. 5b–d, Supplemental Fig. 4A–L). These results indicate that the lack of Nkx2.3 in hematopoietic cells fails to influence the course of colitis, suggesting that the protection in Nkx2.3^−/− mice is due to the absence of this transcription factor in colonic stromal cells.

To identify Nkx2.3-expressing nonhematopoietic cells within the colon, we next used LacZ-Nkx2.3^+/− reporter mice (8). Using immunohistochemistry combined with X-gal staining, we found that Nkx2.3 is not expressed in epithelial cells or endothelial cells (Fig. 5e, 5f). However, nuclear LacZ signal was observed in cells with a myofibroblast-like morphology and VAP-1 expression mainly in the tunica muscularis mucosae of gut wall (Fig. 5g), in line with recently published human data (34).

IBDs are often associated with lymphoid neogenesis resulting in the formation of intestinal tertiary lymphoid tissues, a process that significantly recapitulates secondary lymphoid organ development (1). Hematopoietic cells important in lymphoid tissue development and the pathogenesis of colitis include the heterogeneous group of RORγt⁺ ILC3s playing diverse roles in response to epithelial damage and the subsequent modulation of SILT organization (35). Intestinal ILC3 migration is mediated by α4β7 integrin, the counter receptor for endothelial MAdCAM-1 (36).

The homeodomain transcription factor Nkx2.3 has been implicated in both IBD pathogenesis and the formation of intestinal lymphoid tissues. It was first described as a tissue-specific morphogenic factor promoting murine spleen and PP development (7). Nkx2.3 was found to be expressed throughout the intestine starting from embryonic day 9 (8). Expression was most prominent in the ileum and localized mainly to smooth muscle and endothelial cells of the intestine and the pancreas, but it was also present in the colonic lamina propria in adult mice. Nkx2.3^−/− mice showed delayed intestinal development coupled with malabsorption (11). Importantly, the development of PPs was also impaired, including the absence of MAdCAM-1 expression on HEVs (7, 8), replaced by PNAd in adult mice (10). In humans, Nkx2.3 was linked to both CD and ulcerative colitis. Because of single nucleotide polymorphisms in its coding region, it may act as a shared susceptibility factor for both types of IBD (12, 13). Recent findings in human samples indicate that Nkx2.3 possibly serves to maintain myofibroblast identity in pericryptal stem cell niches of colorectal mucosa in a complex interaction also involving VAP-1 and TGF-β (34). Interestingly, no data have been available on the effect of Nkx2.3 on the development of intestinal tertiary lymphoid structures and the inducibility of colitis, prompting our investigation of a chemically induced colitis model in mice lacking Nkx2.3.

Our observations in Nkx2.3^−/− mice revealed the impaired capacity of colonic LP to promote the maturation of ILFs, resulting in the accumulation of immature stages of SILT. These results are in line with previous work emphasizing the role of endothelial MAdCAM-1–α4β7 interactions in the transition of CPs to ILFs (5) and were comparable to the phenotype of MAdCAM-1-deficient mice. The lack of detectable Nkx2.3 in mature normal B cells also points to stromal factors responsible for impaired follicle maturation in these mutants (37).

Upon treatment with DSS, SILT components in Nkx2.3^−/− mice show a gradual shift toward the mature state observed in wild-type. However, these mice did not develop colitis. Importantly, these observations are in striking contrast with our findings in MAdCAM-1^−/− mice presenting with severe colitis but only marginal maturation of CPs, despite the similar SILT appearance without treatment. The profound difference in the two mutant strains, both deficient for endothelial MAdCAM-1, suggests that the lack of colitis in Nkx2.3-deficient mice is not mediated by the absence of MAdCAM-1. Although the SILT maturation in C57BL/6 mice is delayed compared with BALB/c during the postnatal period (38), the age of mice used had passed this period. As, by this stage in Nkx2.3-knockout mice, endothelial MAdCAM-1 expression is already extinguished, yet upon DSS treatment Nkx2.3-deficient mice react in a drastically different way than Madcam1^−/− mice, we conclude that the differences in SILT development and response to DSS treatment are not linked to the absence of endothelial MAdCAM-1, per se.

Importantly, we found no difference upon DSS treatment of irradiated wild-type mice reconstituted with either wild-type or Nkx2.3^−/− bone marrow, emphasizing Nkx2.3’s stroma-associated function. Despite our repeated attempts using a large cohort of mice, we did not succeed in creating chimeras using Nkx2.3-deficient mice, as all Nkx2.3^−/− recipients exposed to the same dose of radiation needed for myeloablation in BALB/c recipients died shortly after transplantation of bone marrow from syngeneic BALB/c donors. We attribute this failure to the atrophic red pulp (8, 9) with insufficient support capacity during the early period of the hematopoietic reconstitution. Therefore, Nkx2.3-knockout mice are clearly radiosensitive, and the replacement of wild-type hematopoietic cells with Nkx2.3-deficient leukocytes in wild-type recipients did not reduce the inducibility of colitis. This does not rule out the survival of radioresistant host-derived hematopoietic cells, including innate lymphoid cells; however, testing for GFP expression in lamina propria ILC3 cells in reconstituted GFP⁺ BALB/c recipients, we found high-level (at least 80–90%) GFP⁻ donor representation (shown on Supplemental Fig. 4H–J). Furthermore, previous studies showed no detectable expression of Nkx2.3 mRNA in hematopoietic cells (8, 11), thus, even if there are radioresistant hematopoietic cells, they are not directly influenced by Nkx2.3 but rather by their microenvironment.

Protection in Nkx2.3-mutant mice was coupled with significantly higher absolute numbers of RORγt⁺ cells and increased mRNA expression levels of IL-22, leading to previously reported protective consequences (29–31). These findings are similar to those observed in DSS-treated Ret^MEN2B mice with more ILC3s producing IL-22 but not IL-17a (39). However, the lack of inflammation upon administration of the antagonistic anti–IL-22 mAb in Nkx2.3^−/− mice indicates that the protection from colitis is independent of IL-22. Considering the significance of neural regulation of ILC3s (39), possible alterations of the enteric nervous system in Nkx2.3 deficiency may also contribute to the protection from colitis, perhaps involving other Nkx family members such as Nkx2.1 (also known as Thyroid Transcription Factor-1, TITF1), linked to RET mutations and Hirschsprung disease (40).

These observations lead us to hypothesize that under homeostatic conditions the absence of endothelial MAdCAM-1 explains the immature SILT phenotype observed in both Nkx2.3^−/− and MAdCAM-1^−/− mice. However, the significant differences between the two mutant strains during DSS treatment indicate that the absence of MAdCAM-1 on its own is an unlikely cause for these alterations. Instead, we suggest that upon epithelial damage, stromal cells lacking Nkx2.3 create an IL-22–independent protective microenvironment, leading to enhanced mucosal healing. The exact identity, phenotype, and function of these Nkx2.3-dependent cells and their interactions with lymphoid cells remain to be elucidated; however, their precise identification and regulatory pathways influenced by Nkx2.3 may offer potential therapeutic targets. As, in both humans and wild-type mice, Nkx2.3 expression is detectable primarily within the muscle layer underneath the lamina propria (8, 34) (Fig. 5g), the augmented epithelial regeneration in Nkx2.3 mutants may be linked to effects exerted in myofibroblasts influencing intestinal stem cell homeostasis via Wnt signaling (41, 42).

Our work also highlights the ambiguous role of MAdCAM-1–α4β7 interactions in the induction of colitis. Although β7 integrin contributes to colitis, as demonstrated in a T cell–dependent CD45RB T cell transfer model (43) and in the T cell–independent DSS model (44), our findings reveal that the absence of MAdCAM-1 does not necessarily protect from DSS-induced colitis. Considering the differences in the onset of DSS-induced colitis, a detailed investigation of intestinal stroma and the exploration of Nkx2.3 functions within stromal cells clearly warrant further studies. Finding the mechanisms of how either lymphoid and/or epithelial cells are influenced by Nkx2.3⁺ stromal cells may have therapeutic implications in IBD and other chronic inflammatory conditions.

The authors wish to express their gratitude to Dr. Eugene Butcher (Stanford University) and Dr. Sirpa Jalkanen (University of Turku, Turku, Finland) for providing the anti-PNAd-Dylight594 mAb and the VAP-1 mAb, respectively, and the Department of Oncotherapy, University of Pécs, for performing mouse irradiation. The authors gratefully acknowledge the generous donation of the anti–IL-22 (clone 8E11) and isotype control (clone GP120:9709) by Genentech, Inc. LacZ-Nkx2.3^+/− colonic samples were kindly provided by Dr. Cheng-Chun Wang (Institute of Molecular and Cell Biology, Singapore). The authors acknowledge the technical assistance of Gabriella Tuboly-Vincze.

The authors have no financial conflicts of interest.