Fig 2: Vitamin A enhances epithelial expression of REG3 antimicrobial proteins in the small intestine.(A) Vitamin A deprivation model. Pregnant dams were fed vitamin A–deficient (Vit A–) or vitamin A–replete (Vit A+) diets during gestation. Offspring were maintained on the maternal diet for at least 8 weeks.(B) Single cell RNA sequencing (scRNA-seq) was performed on a single cell suspension from the mouse small intestine. Live cells were bar-coded and sequenced, and cells annotated as epithelial cells were analyzed and are displayed as a two-dimensional reduction plot (UMAP). Both Vit A+ and Vit A– cells are plotted (n=3 mice per group).(C) Distribution of epithelial cell subsets from Vit A+ and Vit A– mice (n=3) projected on the same UMAP as in (B).(D) Heatmap of average normalized expression of known vitamin A-responsive genes from the scRNA-seq analysis of IECs.(E) Volcano plot of vitamin A-dependent changes in gene expression in IECs analyzed by scRNA-seq. Reg3b and Reg3g are highlighted in red.(F) Density plots of Reg3b and Reg3g expression in IECs analyzed by scRNA-seq.(G) The gut microbiota induces expression of the antibacterial proteins REG3β and REG3γ. Microbial molecular patterns activate a dendritic cell–ILC3 signaling relay that drives IL-22 production, which induces Reg3b and Reg3g expression in IECs.(H) qPCR analysis of Reg3g and Reg3b transcript abundance in small intestines from conventional mice fed Vit A+ (n=12) and Vit A– (n=14) diets, and germ-free mice fed a Vit A+ diet (n=9). Mice were from three independent litters. Each data point represents one mouse.(I) Immunofluorescence microscopy of REG3G in the distal small intestine of Vit A+ or Vit A– mice. Sections were stained for REG3G (red) and counterstained with DAPI (blue). Scale bar, 100 μm. Images are representative of three mice per group.(J) Mean fluorescence intensities of at least 65 villi were determined across three mice in each dietary group.Vit A+, vitamin A+; Vit A–, vitamin A–; wk, weeks; scRNA-seq, single cell RNA sequencing; IEC, intestinal epithelial cell; UMAP, Uniform Manifold Approximation and Projection; qPCR, quantitative real-time PCR; REG3B, regenerating islet-derived protein 3β; REG3G, regenerating islet-derived protein 3γ; Conv, conventional; GF, germ-free. Means ± SEM are plotted; ***p < 0.001 by Mann-Whitney test. See also Figure S1.

Fig 3: The vitamin A metabolite retinoic acid promotes epithelial REG3G expression.(A) HT-29 cells were treated with 1 μM retinol and/or 100 ng/mL IL-22. REG3G transcripts were quantified by qPCR 18 hours later. Each data point represents one experimental replicate (n=6 per group).(B) Retinol is converted to RA through a two-step enzymatic reaction catalyzed by retinol/alcohol dehydrogenases and retinaldehyde/aldehyde dehydrogenases. Disulfiram inhibits aldehyde dehydrogenase enzymatic activity (including RALDH).(C) HT-29 cells were treated with 1 μM RA and/or 100 ng/mL IL-22 in the presence of vehicle or 100 μM disulfiram. REG3G transcripts were quantified by qPCR after 18 hours. Each data point represents an independent experimental replicate (n=6 per group).(D) qPCR analysis of REG3G transcripts in HT-29 cells treated with retinol and IL-22 in the presence or absence of disulfiram. Each data point represents one experimental replicate (n=4 per group).(E) qPCR analysis of REG3G transcripts in HT-29 cells treated with retinol and IL-22 in the presence of disulfiram, with rescue by RA. Each data point represents an independent experimental replicate (n=4 per group).(F) The Rdh7−/− mouse carries a global deletion of the gene encoding RDH7, which catalyzes the first step in the retinol-to-RA conversion. Rdh7ΔIEC mice harbor an IEC–specific deletion of Rdh7, generated by crossing Rdh7fl/fl mice with Villin-Cre transgenic mice.(G) qPCR analysis of Reg3g transcripts in the small intestines of conventional wild-type (n=5) and Rdh7−/− (n=5) mice, and germ-free wild-type (n=21) mice. Each data point represents one mouse.(H) Immunoblot of REG3G in small intestines from conventional wild-type (n=3) and Rdh7−/− (n=3) mice. ACTIN was the loading control. Each lane is from one mouse.(I) qPCR analysis of Reg3g transcripts in small intestines from six litters of conventional Rdh7fl/fl (n=11) and Rdh7ΔIEC (n=13) mice and germ-free wild-type (n=21) mice. Each data point represents one mouse.(J) Rdh7ΔIEC mice received two intraperitoneal injections of vehicle or retinoic acid (1 μM), administered 12 hours apart. Mice were sacrificed 12 hours after the last injection.(K) qPCR analysis of Reg3g transcripts in the intestines of two litters of Rdh7fl/fl (n=3) and Rdh7ΔIEC (n=6) littermates injected intraperitoneally with RA or vehicle. Each data point represents one mouse.(L) Immunofluorescence microscopy of REG3G in the small intestines of Rdh7fl/fl and Rdh7ΔIEC littermates injected via the intraperitoneal route with retinoic acid (RA) or vehicle. Sections were stained for REG3G (red) and counterstained with DAPI (blue). Scale bar, 100 μm. Images are representative of at least three fields per sample from two independent experiments (three littermates per group).(M) Mean fluorescence intensities of at least 150 villi from the images represented in (L) were quantified across at least two mice in each experimental group.IEC, intestinal epithelial cell; REG3G, regenerating islet-derived protein 3γ; qPCR, quantitative real-time PCR; RDH, retinol dehydrogenase; RALDH, retinaldehyde dehydrogenase; Conv, conventional; GF, germ-free; i.p., intraperitoneal; RA, retinoic acid. Means ± SEM are plotted; *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant by Mann-Whitney test. See also Figure S2.

Fig 4: Retinoic acid receptors bind promoter RAREs to activate mouse Reg3g and human REG3G transcription.(A) RARs bind to retinoic acid response elements (RAREs) in target gene promoters to induce RA-dependent transcription.(B) The mouse Reg3g promoter was scanned for putative RAREs using the JASPAR motif database and quantified with the Transcription Factor Binding Site R Package.56(C) IECs were isolated from wild-type mice, and RAR-bound chromatin was immunoprecipitated. Enrichment of Reg3g promoter regions and an off-target control promoter (Vstm2a) was quantified by qPCR using primers centered on the RARE at −4195 bp upstream of the transcription start site. Control ChIP used a non-specific IgG isotype control antibody. The data were from two mice, with two to three technical replicates per mouse.(D) ChIP-sequencing read tracks from HepG2 cells (human hepatocarcinoma-derived cells) expressing FLAG-tagged RARα, obtained from a publicly available ENCODE Consortium dataset36. Base pair positions are shown relative to the REG3G transcription start site. ENCODE candidate cis-regulatory element promoter signature is denoted by a blue box.(E) ChIP was performed on HCT-116 cells treated overnight with retinol and IL-22. RARs were immunoprecipitated using a pan-RAR antibody, and associated DNA was quantified by qPCR using primers centered on the RARα binding peak encompassing −212 bp to 8 bp relative to the REG3G start site. Binding to an off-target promoter (LINC01643) was also assessed. Control ChIP was done using a non-specific IgG isotype control antibody. Each data point represents an independent experimental replicate (n=6 per group).(F) Transcription reporter assays. HEK293 cells were transfected with a plasmid containing the native REG3G promoter region (2 kb) cloned upstream of a firefly luciferase reporter. The human REG3G promoter contains three predicted RAREs (RARE −8, +33, and +75). The cells were treated with 1 μM retinol and/or co-transfected with a plasmid that expresses RAR403, which suppresses RAR activity (see Fig. 3D).(G) HEK293 cells were transfected with the REG3G promoter–luciferase reporter plasmid to establish that luminescence is dependent on the presence of the plasmid and retinol. Luciferase luminescence was measured in duplicate and each data point represents the average of two readings from one tissue culture well. (n=8 experimental replicates per group).(H) HEK293 cells were transfected with the REG3G promoter–luciferase reporter plasmid in the presence of retinol and the presence or absence of RAR403. Luminescence was measured in duplicate and each data point represents average readings from one tissue culture well. (n=6 experimental replicates per group).(I) Luciferase reporter assay comparing transcriptional activity of the native human promoter and promoter variants with individual disruptions of the RAREs shown in (F). Assays were performed in the presence of retinol. Each data point represents an independent experimental replicate (n=4 per group).(J) Luciferase reporter assay comparing transcriptional activity of the native human promoter and promoter variants with individual RARE disruptions, assessed in the presence or absence of retinol. Each data point represents an independent experimental replicate (n=4 per group).(K) ChIP-sequencing read tracks from ENCODE datasets showing binding of tagged RARα, RARβ, and STAT3 upstream of and within the REG3G locus. Data were generated in human epithelial cells lines engineered to express FLAG-tagged RARα (HepG2), RARβ (A549), or endogenous STAT3 (MCF-10A-derived).(L) Chromatin was isolated from cultured HCT-116 cells and STAT3-bound chromatin was immunoprecipitated. Enrichment of REG3G promoter sequences and an off-target control promoter (SEPTIN2) were quantified by qPCR. Control ChIP used a non-specific IgG isotype control antibody.(M) Chromatin was isolated from IECs recovered from wild-type mice and STAT3-bound chromatin was immunoprecipitated. Enrichment of Reg3g promoter sequences (centered on RARE −4195) and an off-target control promoter (Vstm2a) were quantified by qPCR. Control ChIP used a non-specific IgG isotype control antibody. The data were generated from two mice, with two to three technical replicates per mouse.RA, retinoic acid; RAR, retinoic acid receptor; RARE, retinoic acid response element; REG3G, regenerating islet-derived protein 3γ; IEC, intestinal epithelial cell; ChIP, chromatin immunoprecipitation; qPCR, quantitative real-time PCR; Chr2, chromosome 2; Luc, luciferase; STAT3, Signal Transducer and Activator of Transcription 3. Means ± SEM are plotted; *p < 0.05; **p < 0.01; ***p <0.001; ns, not significant by Mann-Whitney test. See also Figure S4.

Fig 5: Retinoic acid receptors drive transcription of mouse Reg3g and human REG3G genes.(A) Vitamin A-derived retinol is metabolized to RA, which activates retinoic acid receptors (RARs). RARs bind target gene promoters to drive RA-dependent transcription.(B) HT-29 cells (human intestinal epithelial cells) were treated with the RAR antagonist BMS493 or the RAR agonist Ch55 and simultaneously stimulated overnight with RA and IL-22. REG3G transcripts were quantified by qPCR. Each data point represents an independent experimental replicate (n=6 per group).(C) HCT-116, a transfection-competent human intestinal epithelial cell line expressing RARΑ and RARG (Fig. S3B), was treated for 24 hours with an siRNA targeting either gene and then stimulated overnight with retinol and IL-22. REG3G transcripts were quantified by qPCR. Each data point represents an independent experimental replicate (n=4 per group).(D) IEC-specific disruption of RAR signaling using a dominant-negative RAR (dnRAR) knock-in allele. dnRAR mice harbor a loxP-flanked STOP cassette upstream of a dominant-negative RAR open reading frame. The dnRAR is derived from a mutant human RARα (RAR403) lacking the ligand-dependent transactivation domain and functions as a pan-RAR inhibitor. Crossing dnRAR mice with Villin-Cre transgenic mice excises the STOP cassette in IECs, resulting in IEC-selective expression of dnRAR and inhibition of RAR signaling.(E) qPCR analysis of Reg3g expression in small intestines of conventional dnRARfl/fl (n=12) and dnRARIEC (n=17) mice from five litters, and germ-free wild-type mice (n=21).(F) Immunofluorescence microscopy of REG3G in small intestines of dnRARfl/fl and dnRARIEC mice. Sections were stained for REG3G and counterstained with DAPI. Scale bar, 100 μm. Images are representative of at least three fields per sample and two independent experiments (three littermates per group).(G) Mean fluorescence intensities of at least 150 villi from the images represented in (F) were quantified across at least two mice of each genotype.RAR, retinoic acid receptor; RA, retinoic acid; IEC, intestinal epithelial cell; REG3G, regenerating islet-derived protein 3γ; siRNA, small interfering RNA; dnRAR, dominant negative retinoic acid receptor; Conv, conventional; GF, germ-free. Means ± SEM are plotted; *p < 0.05; **p < 0.01; ***p<0.001; ns, not significant by Mann-Whitney test. See also Figure S3.