In this study, we enrolled 5 patients with AD and 5 healthy volunteers who received treatment at the Guangdong Provincial Hospital of Chinese Medicine between December 2021 and July 2022. Ethical approval for this research was obtained from the Ethics Committee of the Second Affiliated Hospital of Guangzhou University of Chinese Medicine, as indicated by Approval ID BF2021-220-01. Prior to participation, all individuals involved in the study provided written informed consent, ensuring adherence to ethical research standards. Comprehensive clinical information for all patients is detailed in Supplementary Table S1 .

The samples were processed as follows: Firstly, the tissue was washed with precooled phosphate-buffered saline (PBS, #C10010500BT, Gibco, Carlsbad, USA) and immediately sectioned into 1 mm³ fragments on ice. These fragments were subjected to enzymatic digestion at 37°C using a mixture of 0.5 U/mL dispase II (#LS02109, Worthington, Lakewood, USA), 50 U/mL DNase I (#LS002007, Worthington), 285 U/mL collagenase I (#LS004196, Worthington), and 355 U/mL collagenase II (#LS02109, Worthington) with gentle agitation for 45 minutes. Following digestion, the samples were passed through a 70 µm cell strainer and then centrifuged at 300 × g for five minutes. The resulting cell pellet was then suspended in red blood cell lysis solution (#130094183, Miltenyi Biotec, Shanghai, China) to remove any remaining erythrocytes, followed by washing with PBS containing 0.04% BSA and resuspension in the same buffer. To ensure the isolation of single cells, the cell suspension was further filtered through a 35 μm cell strainer. Finally, the isolated single cells were stained with acridine orange/propidium iodide (AO/PI), and their viability was assessed using a Countstar fluorescence cell analyzer, ensuring the preparation of high-quality single-cell suspensions for subsequent analyses.

To generate the scRNA-Seq libraries, we utilized the 10× Genomics Chromium Controller Instrument and Chromium Single Cell 3’V3 Reagent Kits (10× Genomics, Pleasanton, USA). In this process, the cells were adjusted to a concentration of 1000 cells/μL. We then loaded approximately 8,000 cells into each channel of the instrument, aiming to barcode approximately 5,000 single cells per sample within the gel bead-in-emulsion (GEM) system. Following reverse transcription, the GEMs were broken down, and the barcoded cDNA was isolated. This cDNA then underwent a series of processing steps: it was purified, amplified, and subjected to fragmentation and A-tailing. The final library quality was assessed using the Qubit high-sensitivity DNA assay (Thermo Fisher Scientific, Waltham, USA), and the size distribution was analyzed with a Bioanalyzer 2200 (Agilent, Santa Clara, USA) high-sensitivity DNA chip. Sequencing was performed on an Illumina sequencer (Illumina, San Diego, USA) with a 150 bp paired-end run, ensuring a depth of 50,000 reads per cell.

Single-cell RNA-seq data analysis was conducted using the NovelBrain Cloud Analysis Platform provided by NovelBio Co., Ltd. (www.novelbrain.com). The procedure commenced with the utilization of the fastp tool (25), set to its default parameters, for the purpose of filtering out adaptor sequences and eliminating low-quality reads, thereby yielding data of enhanced quality. Subsequently, the data were aligned to the human genome reference GRCh38 (Ensembl104) using CellRanger v6.1.1, thereby generating feature-barcode matrices. To ensure uniformity across all sequenced samples, a downsampling analysis based on the number of mapped barcoded reads per cell was performed, which led to the formation of a final aggregated matrix. Only cells expressing more than 200 genes and displaying mitochondrial UMI rates less than 70% were deemed high quality and retained. After the removal of mitochondrial genes, the expression data were normalized and regressed using the Seurat package (version 4.0.3, https://satijalab.org/seurat/), accounting for UMI counts and mitochondrial rates to produce scaled data. Principal component analysis of these data was carried out, concentrating on the 2000 most variably expressed genes, with the top 10 principal components utilized for tSNE and UMAP plotting. A graph-based clustering method was applied for unsupervised cell clustering, drawing on these principal components. Marker genes were identified through the FindAllMarkers function using the Wilcoxon rank sum test under the following criteria: lnFC > 0.25, p value < 0.05, and min.pct > 0.1. For in-depth identification of diverse cell types, selected cell clusters were subjected to further reanalysis involving retSNE, graph-based clustering and marker gene assessment.

Six-week-old male BALB/c mice were procured from Zhuhai BesTest (license number: SCXK 2020-0051). These animals were maintained in a controlled environment characterized by a temperature of 23 ± 3°C and relative humidity (RH) of 55 ± 5% under a 12-hour light/dark cycle. The mice had ad libitum access to standard laboratory food and water. All breeding activities were conducted at the Research Centre of Basic Integrative Medicine, School of Basic Medical Sciences, Guangzhou University of Chinese Medicine (license number: SYXK 2023-0182). The animal experiments conducted in this study were in strict compliance with the relevant regulations pertaining to the ethics of experimental animal use. Ethical approval for these experiments was granted by the Animal Experiment Management and Use Committee of Guangzhou University of Chinese Medicine, as indicated by the experimental approval reference number 20230615027.

The AD mouse model was adapted from previous methods with modifications (26–28). After a week-long acclimation phase, the mice were methodically segregated into two groups, each comprising eight groups: (1) a control group without oxazolone (#862207, Sigma–Aldrich, Saint Louis, USA) and (2) a model group with 1% oxazolone. On the initial day of the experiment, the mice were anesthetized with 2% isoflurane, and a dorsal section measuring 3.5 cm × 3.5 cm was carefully shaved. A sensitizing mixture containing 3% oxazolone in 100 μL of vehicle solution (consisting of acetone and olive oil at a 4:1 ratio) was applied to the hairless dorsal area for 7 days. The control group was administered 100 μL of vehicle. On the eighth day, 10 μL of 1% oxazolone solution was applied to both sides of the left ear, initiating a series of seven challenges. On the final day of application (Day 14), the mice were euthanized. The ear and intestinal tissues were then excised for subsequent experimental analysis.

For surface staining, IELs and spleen lymphocytes (SPLs) were initially treated with an Fc receptor blocking reagent and then incubated with specific antibodies for 30 minutes at 4°C. For intracellular staining, the cells were subjected to a 5-hour pretreatment with phorbol myristate acetate (PMA, 50 ng/mL, #P8139, Sigma–Aldrich), ionomycin (1 µg/mL, #I3909, Sigma–Aldrich) and Brefeldin A (5 µg/mL, #B7651, Sigma-Aldrich), followed by surface marker staining. After fixation and permeabilization, intracellular antibodies were applied. Details regarding the antibodies used are listed in Supplementary Table S2 . All flow cytometry experiments were carried out with a flow cytometer (LSRFortessa, BD Biosciences, Franklin Lakes, USA), and the data were analyzed using FlowJo software (Tree Star, Ashland, USA).

The mice were euthanized, and the intestines were promptly excised and immediately placed in precooled PBS. Subsequent to the careful removal of residual mesenteric fat tissue, Peyer’s patches were delicately dissected. The intestines were then longitudinally opened and rigorously washed in ice-cold PBS. Next, the intestinal segments were sectioned into 1.5 cm pieces. These pieces were then subjected to two incubation periods in 5 ml of 5 mM EDTA dissolved in Hank’s Balanced Salt Solution (HBSS, #PB180323, Pricella, Wuhan, China), each lasting 15–20 minutes at 37°C with gentle agitation at 100 rpm. After each incubation, the solution was subjected to vigorous vortexing and then filtered through a 70 µm cell strainer, followed by the addition of fresh EDTA solution. The supernatants obtained from IELs isolation from a single small intestine were pooled, and the cells were washed once in cold PBS. The cell suspension was then reconstituted in 10 ml of lymphocyte separation medium. This mixture was centrifuged for 20 minutes at 2500 rpm at room temperature to achieve separation. The cells were subsequently washed once more and resuspended in T-cell medium, after which they were immediately utilized for experimental purposes. The spleens were excised, immersed in cold PBS and then filtered through a 100 µm strainer. Erythrocytes were lysed using RBC lysis buffer, and the isolated SPLs were prepared for further experimentation.

CD4⁺ T cells were isolated using a mouse CD4⁺ T-cell cell isolation kit (#70901, Beaver, Suzhou, China) according to the manufacturer’s protocol. Once purified, the cells were cultured in RPMI-1640 medium (#PM150110A, Pricella) supplemented with 10% fetal bovine serum (FBS, #164210, Pricella) and maintained at 37°C for 24 hours. During this period, PMA (50 ng/mL, Sigma–Aldrich) and ionomycin (1 µg/mL, Sigma–Aldrich) were added for the final 6 hours of incubation, while the interleukin (IL)-15/IL-15R complex (50 ng/mL, #HY-P70655, MedChemExpress, Monmouth Junction, USA) was added during the entire 24 hours period. Afterward, the cells were centrifuged to separate the cellular components, and the supernatants were collected for subsequent analysis.

Subconfluent monolayers of HT-29 cells obtained from Wuhan Pricella Biotechnology Co., Ltd. (#CL-0118, Pricella) were cultured in McCoy’s 5A medium (#PM150710, Pricella) supplemented with 10% FBS and 1% penicillin/streptomycin (#PB180120, Pricella) at 37°C in a humidified 5% CO₂ atmosphere. DEFA5 was procured from Abmart Shanghai Co., Ltd. (#RG310219, Abmart, Shanghai, China). On the day before initiating DEFA5 treatment, the cells were subjected to starvation. HT-29 cells were then treated with DEFA5 at concentrations of 1 µg/ml and 2 µg/ml for 24 hours, after which the cells were harvested for subsequent analyses.

Intestinal explants were extracted following the method previously outlined (29). In brief, the terminal ileum was sliced into 0.5 cm segments along its length, rinsed with cold PBS, and then placed into 1 mL of culture medium made up of RPMI 1640 (Thermo Fisher Scientific, Waltham, USA), protease inhibitor cocktail (Sigma-Aldrich, St. Louis, USA), and Penicillin/Streptomycin solution (Waltham, MA, USA). Afterward, the explants were co-cultured with CD4⁺ IELs (1*10⁶ cells/well), which were either stimulated or unstimulated, in 24-well plates at 37°C for 24 hours. Following the incubation period, explants from each group were harvested and preserved at -80°C for subsequent analysis.

SCID mice were modeled for AD using the previously described method, and adoptive transfer was performed on day 7. CD4⁺ IELs were isolated from the small intestine of Balb/c mice as described in the Tissue Preparation and Cell Isolation section. After isolation, 1*10⁶ cells were injected i.p. into SCID mice. For the IELs adoptive transfer experiment, SCID mice were harvested on day 7 post-transfer.

Cultured cells and animal tissue samples were subjected to three washes in ice-cold PBS, followed by lysis using RIPA buffer supplemented with protease and phosphatase inhibitors for protein extraction. The proteins extracted were separated using 10–12% SDS–PAGE and then transferred to PVDF membranes (Bio-Rad, Hercules, USA). After transfer, the membranes were blocked with 5% skim milk for 2 hours and incubated with primary antibodies at 4°C overnight. After they were washed, they were exposed to the corresponding secondary antibodies for 2 hours at room temperature. Protein band visualization was conducted with Tanon 5200 Multi (Tanon, Shanghai, China), and quantitative analysis was performed using ImageJ software. Details regarding the antibodies used are listed in Supplementary Table S2 .

To investigate the colocalization of DEFA5 and CD4, CD4⁺ IELs were incubated overnight at 4°C with primary antibodies against CD4 (1:200, Proteintech, Rosemont, USA) and DEFA5 (1:100, Abbexa, Cambridge, UK) in PBS. Following primary antibody incubation, the sections were washed, and the nuclei were stained with DAPI/Hoechst (1:10000, Life Technologies, Carlsbad, USA) for 5 minutes at RT. After a final wash, confocal fluorescence microscopy was performed using a Zeiss LSM800 (Zeiss, Jena, Germany), with imaging conducted at 488 nm and 594 nm wavelengths using a 20× magnification lens.

RNA was isolated from 1 million CD4⁺ IELs, CD4⁺ SPLs using TRIzol™ Reagent (#10296010, Invitrogen, Carlsbad, USA). The primer sequences synthesized by Sangon are detailed in Supplementary Table S3 of the Supplementary Material .

The data were analyzed using GraphPad Prism 9.0 software (GraphPad Software, Inc., San Diego, USA). The quantitative data represent at least three independent experiments. Unpaired Student’s t tests were utilized for comparisons between two groups. Multiple comparisons were analyzed by one-way analysis of variance (ANOVA) with Bonferroni post hoc correction. A p-value less than 0.05 was considered to indicate statistical significance.

To investigate the immune composition in the intestine of patients with AD, we collected terminal ileum tissue samples from five adult AD patients (ADs) and five normal control individuals (NCs) for single-cell RNA-seq analysis. Detailed clinical information about the cohort is provided see ( Supplementary Table S1 ). Following stringent quality control measures (see Supplementary Figures S1A, B ), we obtained a total of 75,912 cells, including 43,085 cells from ADs and 32,827 cells from normal controls for further analysis. Utilizing cluster-specific marker genes and unsupervised t-distributed stochastic neighbor embedding (t-SNE), we identified a total of 22 cell clusters ( Figure 1A ). To explore the transcriptomic heterogeneity between NCs and ADs, we delved deeper into the analysis of these 22 cell clusters. The tSNE plots revealed the cellular distribution within the NC and AD groups ( Figure 1B ). In total, five clusters, including clusters 4, 6, 7, 19, and 21 were classified as T_NK cells based on differential gene expression of CD3D, CD3E and PTPRC ( Supplementary Figures S1C–E ).

We further delineated the cell populations within the T_NK cell cluster through tSNE analyses. The T_NK cells identified were categorized into CD4⁺ T cells, CD8⁺ T cells, natural killer T (NKT) cells, NK cells, and γδT cells ( Figure 1C ). The NKT cells (Cluster 4), NK cells (Clusters 8 and 9), and γδT cells (Clusters 0, 1, 2, 3, 5, 6, and 7), which constituted a minor proportion, were classified into eight distinct clusters ( Figure 1C ). We further conducted differential gene expression analysis on CD4⁺ T cells ( Figure 1D ) and CD8⁺ T cells ( Figure 1E ). It was found that DEFA5 was among the top upregulated genes in ADs compared to NCs. In light of evidence showing the ability of CD8⁺ T cells to secrete a variety of α-defensins (16), we hypothesize that CD4⁺ T cells in the gut may have similar function. Independent analyses of DEFA5 expression in CD4⁺ T cells from the donors confirmed the increased expression of this gene in CD4⁺ T cells from ADs compared to these from NCs ( Figure 1F ).

Next, we analyzed the expression of DEFA5 among several key cell types within the T_NK cells, including CD4⁺ T cells, CD8⁺ T cells and γδT cells, and found significantly increased expression of DEFA5 in all types of cells in ADs compared to NCs ( Figure 2A ). Subsequent Gene Ontology enrichment analysis of CD4⁺ T cells revealed significant enrichment of defensin-associated gene sets, including “defense response to virus”, “host regulation of defense response to virus” and “membrane disruption in another organism” ( Figure 2B ). Furthermore, pathway analysis in the DEFA5-positive subset of CD4⁺ T cells revealed that the top differentially enriched pathways include the PPAR signaling pathway, MAPK signaling pathway, and NOD signaling pathway ( Figure 2C ). Particularly, the PPAR signaling pathway was the top upregulated pathway in the CD4⁺ T cells of ADs in comparison to NCs ( Figure 2D ).

Initially, we attempted to classify DEFA5-expressing CD4⁺ T cells into Th1 (IFNG ⁺, CXCR3 ⁺, CCR5 ⁺, IL2 ⁺), Th2 (IL10 ⁺, IL4 ⁺, IL5 ⁺, IL13 ⁺), and Th17 (IL17A⁺ , IL17F⁺ , IL22⁺ ) subsets. However, DEFA5 was predominantly enriched in a previously uncharacterized cell population ( Supplementary Figures S2A, B ). To further characterize DEFA5-expression CD4⁺ T cells, we analyzed and identified six distinct CD4⁺ T-cell subclusters ( Figure 2E ), including CD4⁺ naïve T cells (CCR7 ⁺, SELL ⁺, LEF1 ⁺, and TCF7 ⁺) (30), Follicular helper T (Tfh) cells (CXCR5 ⁺, CD200 ⁺, TOX ⁺, and TOX2 ⁺) (31, 32), CD4⁺ central memory T (Tcm) cells (CCR7 ⁺, TCF7 ⁺, and CD69 ⁺) (33, 34), CD4⁺ effector memory T (Tem)/TH1-like cells (IFNG ⁺, CCL5 ⁺, and GZMK ⁺) (35, 36), CD4⁺ resident memory T (Trm) cells (CXCR6 ⁺ and KLRB1 ⁺) (37, 38), and Treg cells (FOXP3 ⁺, IKZF2 ⁺, and CTLA4 ⁺) (39, 40). As depicted in Figures 2F, G , there was a notable increase in DEFA5 expression in the intestinal CD4⁺ T cells of the ADs relative to that in the NCs. In addition, DEFA5 expression was primarily enriched in Tcm and Trm cells. Based on the notable transcriptomic characteristics exhibited by CD4⁺ T cells, we conducted a pseudotime trajectory analysis. The results suggested that the intestinal T cells differentiated from naïve CD4⁺ T cells may diverge into two distinct evolutionary pathways: one leading toward Treg/Tfh cell states and the other toward Trm cells ( Figures 2H, I ). To further understand the biological characteristics of CD4⁺ T-cell clusters, we applied quantitative set analysis for gene expression (QuSAGE). The results revealed that the PPAR signaling pathway was predominantly enriched in Trm cells (Cluster 4) ( Supplementary Figure S3 ).

To further understand the underlying mechanisms of DEFA5 production by intestinal CD4⁺ T cells in AD, we established an oxazolone-mediated AD mouse model ( Figure 3A ). Compared with the control group, oxazolone-treated mice exhibited pronounced redness and swelling in both the ears and the back, along with a considerable increase in ear thickness ( Figures 3B, C ). Thymic stromal lymphopoietin (TSLP) is considered an a key marker that plays a significant role in advancing AD and is linked to gut microbiota imbalance related to the disease (41). Western blot analysis of TSLP showed increased levels of this protein in the ears of the treated mice compared to that in control group ( Figure 3D ), confirming the establishment of the AD in the mice. IELs constitute the most abundant lymphatic system within the intestine, and CD4⁺ T cells within IELs are predominantly located in the distal part of the small intestine (42). IELs are also considered to be one of the most abundant and intestine-specific subsets of Trm cells (43). Thus, we isolated single CD4+ IELs from the mouse intestine and detected them using confocal microscopy. The findings indicated that stimulated CD4+ IELs generated more DEFA5 than unstimulated CD4+ IELs, with CD4 and DEFA5 clearly colocalizing ( Supplementary Figure S4A ). In addition, flow cytometry results demonstrated a significant increase in CD25 and CD69 ( Figures 3E–G ) within the CD4⁺ IELs, indicating aberrant activation of CD4⁺ IELs in mice with AD.

To test if activation could lead to DEFA5 expression in CD4⁺ T cells from other tissues, we evaluated DEFA5 expression in CD4⁺ T cells from the spleen and IELs through PCR upon activation. Moreover, DEFA5 expression was markedly elevated in stimulated CD4⁺ IELs as opposed to unstimulated CD4⁺ IELs ( Supplementary Figure S4A ). The results showed that CD4⁺ IELs, but not CD4⁺ T cells from the spleen, were capable of expressing DEFA5 ( Figure 4A ). Moreover, DEFA5 expression was markedly elevated in stimulated CD4⁺ IELs as opposed to unstimulated CD4⁺ IELs ( Supplementary Figure S4B ). In addition, epithelial cells, particularly Paneth cells, are recognized as key producers of DEFA5. To address potential contamination by Paneth and other epithelial cells, we examined the expression of Paneth cell-specific markers and epithelial cell-specific markers, including Sox9, Epcam, ctnnb1, Lyz2, Lyz1, and Cd24a in the isolated CD4⁺ IELs. The results showed that these genes were undetectable in the isolated IELs ( Figures 4B–G ), suggesting that the CD4⁺ IELs obtained were unlikely contaminated by Paneth cells or epithelial cells. Furthermore, flow cytometric evaluation of the purified CD4⁺ IELs indicated that 99.8% of the CD45⁺ cells were CD4⁺, showing minimal presence of non-immune or non-CD4⁺ T cells ( Supplementary Figure S5 ).

Preliminary single-cell RNA sequencing (scRNA-seq) analyses have shown a significant upregulation of the PPAR signaling pathway in ADs. We therefore evaluated the expression of PPARγ in the mouse intestine. Compared to the control group, a notable increase in PPARγ mRNA and protein levels was observed in the intestines of the mice with AD ( Figures 4H, I ). Previous findings suggest that PPARγ activation plays a key role in preserving defensin expression within the intestine (44). To investigate whether PPARγ is capable of regulating DEFA5, we used AlphaFold3 to visualize the potential for a direct binding interface between DEFA5 and PPARγ ( Supplementary Figure S6A ). Furthermore, we utilized the GEPIA web tool (http://gepia.cancer-pku.cn/) to conduct a Spearman correlation analysis, aiming to investigate the link between DEFA5 and PPARG gene expression ( Supplementary Figure S6C ). The results demonstrated a significant positive correlation between PPARG and DEFA5 expression (R = 0.54, p < 0.001), which indicates that increased expression of PPARγ is frequently linked to the upregulation of DEFA5. Next, PPARγ inhibitor GW9662 was used to test the role of this molecule in DEFA5 expression by CD4⁺ IELs. The results showed that following the addition of GW9662, a significant decrease in DEFA5 expression was observed compared to cells without PPARγ inhibition ( Figure 4J ), indicating that PPARγ plays an important role in regulation of DEFA5 expression by CD4⁺ IELs.

Upon induction of AD, we noted a significant reduction in body weight in the AD group of mice compared to the control group ( Figure 5A ). Additionally, we assessed the mRNA expression of Epcam, Ephb3, and ZO-1, as well as the protein levels of IL-1β in the intestine. It was discovered that mice with AD exhibited reduced expression of Epcam, Ephb3, and ZO-1, but increased expression of IL-1β compared to control ( Figures 5B–E ), suggesting the impairment of intestinal barrier function and the development of intestinal inflammation. Patients with ADs frequently experience gastrointestinal symptoms such as abdominal pain, bloating, and diarrhea (45, 46). This phenomenon has also been validated in various AD animal models (47). Under these circumstances, intestinal inflammation may lead to the aberrant activation of intestinal CD4⁺ T cells, resulting in the production of DEFA5. Paneth cells constitute the principal cellular source of DEFA5 in humans. Single-cell transcriptomic analysis revealed an increased abundance of Paneth cells in the ileum of individuals with AD ( Supplementary Figure S7A ). Additionally, within the intestinal epithelium, the expression of DEFA5 in the ADs was significantly greater than that in the NCs ( Supplementary Figure S7B ). It has demonstrated that elevated local concentrations of DEFA5 are also among the factors that induce apoptosis (48). To test this, we treated the intestinal epithelial cell line HT-29 cells with 1 µg/ml and 2 µg/ml of DEFA5, followed by examination of ZO-1 and IL-1β by western blot. The results showed a dose-dependent reduction in ZO-1 expression and increase in IL-1β ( Figure 5F ), suggesting that DEFA5 may play important roles in the compromise of intestinal barrier function and the development of intestinal inflammation in AD.

To investigate the contribution of CD4⁺ IELs to the development of intestinal inflammation in mice with AD, we adoptively transferred CD4⁺ IELs into SCID mice with or without AD induction. Compared with NC group and AD group without T cell transfer, SCID mice that received CD4⁺ IELs exhibited significantly increased expression of PPARγ, accompanied by elevated levels of IL-1β and IFNγ ( Figure 5G ). Additionally, we co-cultured intestinal explants with either stimulated or unstimulated CD4⁺ IELs. ZO-1 expression in explants was significantly lower when co-cultured with stimulated CD4⁺ IELs than when co-cultured with unstimulated CD4⁺ IELs or without any CD4⁺ IEL co-culture ( Supplementary Figure S4C ). These results suggested that the abnormal increase of CD4⁺ IELs could be a key factor contributing to the development of intestinal inflammation in AD.