Background
Celiac disease (CeD) is a human leukocyte antigen-linked autoimmune-mediated enteropathy, which is triggered by the ingestion of gluten in genetically predisposed individuals. With a worldwide prevalence of 1%, this disorder presents with broad clinical manifestations, including intestinal and extra-intestinal symptoms, and is therefore a considerable global public health concern [
1]. Currently, the only available therapy for CeD patients is a life-long exclusion of gluten from the diet, but it is associated with several possible challenges, including inferior quality of life, inadequate response, possible negative effects, and a heavy economic and societal burden [
2]. Therefore, more novel therapeutic approaches are urgently needed.
Previous studies on the pathogenesis of CeD have suggested that CD4
+ T cells play an important role in recognising deamidated gluten peptides by binding to HLA-DQ molecules HLA-DQ2.5, HLA-DQ2.2, and HLA-DQ8 in antigen-presenting cells. Then, gluten-specific CD4
+ T cells support cytotoxic CD8
+ T cells in expressing high levels of activated NK receptors to kill enterocytes and inducing B cells to produce gliadin-associated antibodies [
3,
4]. Moreover, it has been shown that the percentage of γδTCR
+IELs was increased in duodenal biopsies from CeD patients [
5], which might play a proinflammatory role in the process of gut damage [
6]. A chronic site-specific inflammation triggered by gluten permanently reconfigures the tissue-resident γδTCR
+ IEL compartment in patients with CeD, accompanied by the expansion of gluten-sensitive, interferon-γ-producing γδTCR
+ IEL [
7]. In addition, there are many other immune cells in the gut that are involved in the development of CeD, and more detailed mechanisms of these cells need to be explored further. Therefore, this study involved a comprehensive analysis by integrating multiple‐microarray and transcriptome sequence analysis, then constructed an immune atlas and the pathogenesis network of CeD and attempted to provide potential targets for the clinical diagnosis and treatment of this disorder.
Methods
Establishment of celiac disease (CeD) animal model
Female pregnant Wistar rats, purchased from the Laboratory Animal Center of Southern Medical University (Guangzhou, China), were maintained in specific pathogen free conditions and had free access to food and water. The animal handling protocols were approved by The Institutional Animal Care and Use Committee of Southern Medical University (Guangzhou, China).
Shortly after spontaneous birth, rat pups were randomly distributed into two groups. In detail: (1) “Control group” (artificially reared without treatment) (n = 6); (2) “IFN-γ/gliadin group” (sensitized by intraperitoneal injection of 1000U interferon-γ (IFN-γ) and treated with Pepsin-trypsin gliadin 50 μg/day for 10 days) (n = 6). All newborn pups were fed on a rat milk formula every 4 h until 10 days. A dose of 1000U IFN-γ was derived from recombinant rat IFN-γ (PeproTech, China). Pepsin-trypsin gliadin (PTG, gliadin from wheat gluten, Sigma-Aldrich) was obtained following a previously published procedure [
8]. After 10 days, the anesthetized animals were sacrificed for further studies.
Cell cultures and treatments
Human colon adenocarcinoma-derived cells (Caco-2) obtained from the ATCC (Rockville, USA) were cultured in DMEM, supplemented with 10% fetal bovine serum, 100 units penicillin–streptomycin/mL, and 1 mmol/L glutamine (all these products are Gibco Invitrogen, Milan, Italy). Cells were maintained in a humidified atmosphere (95%) of air and 5% CO2 at 37 °C.
To knock GSDMD down, GSDMD target siRNA (GTGTGTCAACCTGTCTATCAA) or control nonspecific siRNA (Ribobio, Guangzhou, China) were transfected into Caco-2 cells with Advanced DNA RNA Transfection Reagent (ZETA LIFE, USA) following the manufacturer’s protocol. After 24 h, the cells were treated with or without IFN-γ (Peprotech, USA) in different concentrations, and the cells and culture supernatants were collected after 24 h.
PTG were dissolved in 100 μl of DMSO, and then diluted it in 99.9 ml of DMEM to prepare 200 μg/ml of PTG medium solution. To evaluate the effect of IFN-γ and PTG, Caco-2 cells were divided into 4 groups and incubated with 50 ng/ml IFN-γ only, 200 μg/ml PTG only, both 50 ng/ml IFN-γ and 200 μg/ml PTG, and neither of them, respectively. After 24 h, the cells and media were harvested for further studies.
Cell counting kit-8 (CCK-8) assay
Caco-2 cells were mixed with 10 µl CCK-8 solution (Fude, Hangzhou, China) and incubated at 37 °C for 2 h, and their absorbances at 450 nm were analyzed via the microplate reader (Molecular Devices, USA) to determine the cell viability.
Enzyme-linked immunosorbent assay (ELISA)
ELISA kit (JSBOSSEN, BS-E3947H2, China) was used to measure the production of IL-1β in culture supernatants following the manufacturer’s protocol. Briefly, testing samples and standards were implanted in 96-well plates and incubated with the conjugate reagents for 1.5 h. Then the TMB solution was added to each well, and the samples were incubated at 37 °C for another 15 min. Lastly, the stop solution was added to terminate the reaction, and the absorbance at 450 nm was measured using the microplate reader (Molecular Devices, USA).
Immunofluorescence
According to the standard protocols, paraffin-embedded duodenal tissue sections (4 μm) underwent an appropriate heat-induced antigen retrieval. Samples were blocked for 1 h, and incubated overnight at 4 °C with FITC-anti-rat-CD3 and PE-anti-rat-TCRγδ (Biolegend, San Diego, CA, USA). The next day samples were washed and stained with DAPI for 10 min, and imaged via fluorescent microscopy (Olympus, Japan).
Western blotting
Protein extracts were obtained from cells and rat duodenal tissue samples employing total lysis buffer (Beyotime, Shanghai, China) supplemented with protease and phosphatase inhibitors (Fudebio, Hangzhou, China). After homogenization and centrifugation at 14,000 rpm for 15 min at 4 °C, concentration of protein supernatant was measured using a standard Bradford assay (ThermoFisher Scientific, Rockford, USA). The samples were resolved using SDS-PAGE, then transferred onto PVDF membranes (EMD Millipore, Billerica, Massachusetts, USA), and blocked with 5% skim milk for 120 min. The blocked membranes were incubated at 4 °C overnight with corresponding primary antibodies targeting GSDMD full length and cleaved GSDMD (C Teminal) (Abcam, USA), Caspase-1 (Proteintech, Wuhan, China), IL-1β (Abcam, USA), β-Tublin (Abmart, Shanghai, China), Cleaved Caspase-1 (CST, USA)) followed by secondary antibodies (Proteintech, Wuhan, China) at room temperature for 120 min. Lastly, the detection was performed using the enhanced chemiluminescence detection kit (Yeasen, Shanghai, China).
Quantitative reverse transcription PCR (qRT-PCR)
Total RNA of cell were extracted by TRIzol RNA extraction agent (Thermo Fisher Scientific, USA) according to manufacturer manual. RNA concentration were measure by absorbance at 260 nm, then reverse transcribed via reverse transcriptase kit (Vazyme Biotech Co.,Ltd, China) followed by real-time PCR amplification by use of SYBR Green Premix ExTaq (Vazyme Biotech Co., Ltd, China). Data were interpreted using the 2−ΔΔCt method, with GAPDH serving as the reference gene for normalization. The primer of GSDMD were: GTGTGTCAACCTGTCTATCAAGG (forward strand) and CATGGCATCGTAGAAGTGGAAG (reserved strand). The primer of GAPDH were: GGAGCGAGATCCCTCCAAAAT (forward strand) and GGCTGTTGTCATACTTCTCATGG (reserved strand).
Data acquisition
In the Gene Expression Omnibus database (GEO
https://www.ncbi.nlm.nih.gov/geo/), we selected several datasets related to CeD for analysis. Among them, three microarray datasets, GSE72625 [
9], GSE112102 [
10], GSE164883 [
11] were based on the GPL10558 platform and were merged into the derivation cohort, including 48 CeD patients and 51 healthy control (HC)’s duodenal tissues. Validation cohort was developed using three transcriptome high-throughput sequencing datasets (GSE131705 [
12], GSE134900 [
13] and GSE146190 [
14]), consisting of duodenal tissue of 90 CeD patients and 70 HCs. Dataset GSE123649 [
7] contained samples of TCRγδ
+ intraepithelial lymphocytes in the duodenum of 18 active CeD patients and 8 CeD patients receiving gluten-free diet (GFD) and 8 HCs. Dataset GSE106260 included duodenal epithelial cell samples of 4 CeD patients and 4 HCs [
15]. External datasets associated with pyroptosis, including GSE125625 [
16] and GSE191015 [
17], were used to further verify reliability of pyroptosis enrichment score. Single cells RNA sequence dataset GSE195780 [
18] (including epithelial cells and lamina propria CD45
+ immune cells in CeD duodenal tissues), were included for further analysis. The dataset GSE145358 [
19] contained 15 duodenal tissues from CeD patients received GFD and 15 duodenal tissues from GFD patient accepted gluten challenge (10 weeks, 4 g of gluten daily).
Data processing and identification of differentially expressed genes
Datasets were merged and removed the batch effect with the R package “sva” [
20], and then visualized in a boxplot and examined by principal component analysis (PCA). Differential expression genes (DEGs) analysis of microarray data were performed using the R package “limma”[
21]. In high-throughput sequencing datasets, the “DEseq2” R package [
22] was used for detecting DEGs. An adjust
P-value < 0.05 and log Fold Change (logFC) > 2 were considered statistically significant.
GO, KEGG and reactome enrichment analysis
Construction of protein–protein interaction (PPI) network and recognition of hub genes
The Search Tool for the Retrieval of Interacting Genes (STRING) online database (
http://string-db.org; version11.0) was used to construct PPI networks for the DEGs, and an interaction score > 0.4 was regarded as statistically significant. Subsequently, the molecular interaction network was visualized using Cytoscape software (
https://cytoscape.org/). Furthermore, we used the MCODE plugin app (
https://sourceforge.net/projects/mcode/) within Cytoscape to identify hub genes and calculate the degrees of interaction between DEGs to visualize the ranking.
Single sample gene set enrichment analysis (ssGSEA) and definition of pyroptosis enrichment score (PES)
To quantitatively describe the pyroptosis degree of each sample, the parameter PES was constructed using the ssGSEA algorithm based on the R package “GSVA” [
26]). Briefly, after identifying all differentially expressed pyroptosis genes between CeD and HC samples, the “pro-pyroptotic” genes and the “anti-pyroptotic” genes were defined as upregulated and downregulated genes in CeD samples, respectively. The positive pyroptosis pathway components were calculated by the ssGSEA scores of upregulated PRGs, and the negative pathway components were calculated by the ssGSEA scores of downregulated PRGs. The PES was calculated as normalized positive pyroptosis pathway components minus normalized negative pyroptosis pathway components. In addition, the enrichment score of “Response to IFN-γ” pathway was also calculated empolying ssGSEA algorithm based on the GO database (GO:0034341).
Immune cell infiltration estimation and co-expression network analysis
By applying
xCell [
27] to the microarray data, the estimated proportion of immune and stromal cell types can be obtained for each duodenal sample in the derivation and validation cohorts. The cut-off value for the cell analysis was
P < 0.05. Cell types were categorized into lymphoid, myeloid, stromal, epithelial and stem cells. Furthermore, to elucidate the interactions between immune cells and pathways and explore critical immune cells in CeD pathogenesis, we identified co-expression patterns based on Spearman correlation analysis.
Weight gene correlation network analysis (WGCNA)
WGCNA were performed in the gene expression profiles from the derivation cohort, while the characteristics matadata were consisted of PES, infiltration of γδT cells level and IFN-γ response score. At first, the
hclust function was used for hierarchical clustering analysis. Then, the soft thresholding power value was screened during module construction using the
pickSoftThreshold function. Candidate power (1 to 30) was used to test the average connectivity degrees of different modules and their independence. A suitable power value was selected if the independence degree was > 0.8. The “WGCNA” R package [
28] was used to construct co-expression networks (modules). The minimum module size was set to 12, and each module was labelled with a different color. Finally, the gene module that most relevant to γδT cells infiltration, IFN-γ response and pyroptosis was extracted to identifying hub genes.
Construction of the least absolute shrinkage and selection operator (LASSO) regression model
The LASSO regression model was established to further verify the diagnostic or predictive effectiveness of PES, pyroptosis-associated hub genes or γδT cells infiltration in CeD. In this model, variables from the derivation cohort were used to construct the LASSO model by the R package “glmnet” (
https://CRAN.R-project.org/package=glmnet). The reliability of the LASSO model was evaluated by Receiver Operating Characteristic (ROC) analysis using the R package “ROCR”. The repeatability of the variables included in the LASSO model was verified in the validation cohort using the Ridge regression model.
Gene set enrichment analysis (GSEA)
GSEA is a calculation method used to determine whether predefined genomes between two groups exhibit significant differences. Therefore, we used the GSEA software (
https://www.gsea-msigdb.org/gsea/index.jsp) to screen GO terms and KEGG pathways that may be related to γδT cells in duodenal tissues in the derivation cohort. The GSEA analysis was performed according to default parameters.
P value < 0.05 was considered significant.
Single-cell sequencing analysis (scRNAseq)
In this study, one sample “GSM5850285RCDII-2” from dataset GSE195780 were selected for scRNAseq analysis. Seurat package v4.0.3 (
https://satijalab.org/seurat/articles/pbmc3k_tutorial.html) was used in R version 4.1.3. The Seurat object was constructed under the criteria that retain only genes expressed in at least 3 cells and cells with at least 200 detected genes. After calculating mitochondrial gene proportion, a second quality control was carried out to remove cells with unique gene cell counts fewer than 200 or more than 5000, while mitochondrial gene percentage was greater than 15%. Next, the data was transformed through log-normalization and linear scaling. Two thousand most highly variable features were selected for PCA. The top 18 PCA dimensions were used for graph-based clustering. With a resolution of 0.9, the clusters were dimensionally reduced and visualized employing a Uniform Manifold Approximation and Projection (UMAP) algorithm.
Statistical analysis
The SPSS 26.0 software (SPSS, Chicago, IL, USA) was used for statistical analysis. Results were described as mean ± standard deviation (SD) or median contains upper and lower quartiles. The Kolmogorov–Smirnov test was used to assess whether the data accorded with normal distribution. Statistical comparison was performed using the Student t-test or one-way analysis of variance (ANOVA) for normal distribution variables, and Kruskal–Wallis test and the Mann–Whitney U test for Non-normal distribution variables. P < 0.05 indicated a statistically significant difference.
Discussion
Current studies illustrated that gluten intolerance was the key factor in the pathogenesis of CeD. However, the mechanism of CeD pathogenesis has not been clarified [
33]. This study systematically analysed the differential genes and functional enrichment changes in CeD tissue using scientifically combined multiple microarray and transcriptome sequence data; it further confirmed that γδT cells, a subgroup of T cells, were involved in the pathogenesis of CeD and mediated epithelial pyroptosis via IFN-γ.
Histologically, the duodenum has long intestinal villi, and many microvilli are distributed on the epithelial luminal side, which create a tremendous surface area for efficient absorption of a variety of nutrients [
34]. The most significant pathological change in CeD condition is the duodenal villi atrophy, leading to malabsorption, diarrhoea, weight loss, and malnutrition. Although the root cause of duodenal villi atrophy is not clear, repeated mucosal inflammation and programmed epithelial cell death are considered as vital causes [
35]. DEGs analysis revealed that downregulated genes in CeD tissue were primarily concentrated on nutrients transport and absorption and changes in metabolic process, confirming that malnutrition in patients with celiac disease was associated with duodenal villi atrophy.
Tissue-resident lymphocytes, including αβT and γδT cells, function as immune monitors and regulators [
36]. During the pathogenesis of CeD, the elevated infiltration of γδT cells was observed both in our study and a previous report [
5]. However, unlike most T cells, the function and source of γδT cells have not been fully elucidated. Gluten-induced inflammation might trigger not only the change in the number of γδT cells but also their function through the rearrangement of TCR [
7]. Our study also revealed an increased infiltration of γδT cells in CeD and the functional changes in IFN-γ expression, which was a pivotal factor in CeD pathogenesis.
Pyroptosis, also known as inflammatory necrosis, is a kind of programmed cell death which is characterised by the continuous expansion of cells and rupturing of the cell membrane, resulting in the release of cell contents such IL-1β, leading to a strong inflammatory response [
29]. Since no explicit study has reported the occurrence of pyroptosis in the small intestinal epithelial cells, our study constructively established a relationship between pyroptosis and γδT cells in CeD. Induced by a high level of IFN-γ derived from γδT cells, epithelial cells entered a “pre-pyroptotic” condition, with high expression of GSDMD, which were prone to pyroptosis after stimulation by pepsin-trypsin gliadin or other exogenous toxins. Nonetheless, this hypothesis needs to be further explored employing in vitro and in vivo experiments.
In addition, our study identified 21 hub genes with significant discrepancy expression between CeD and HC patients, including 20 genes that were also distinctly associated with pyroptosis based on WGCNA. Among them, single-cell RNA sequence analysis revealed that IRF1, one of the hub gene higher expressed in CeD epithelial cells, was a transcriptional regulator which regulated the transcription of IFN and IFN-inducible genes and participated in the regulation of many genes expression [
37]. In vitro experiments showed that IFN-γ could upregulate the protein level expression of GSDMD in Caco-2 epithelial cell line, while knockdown of GSDMD would lead to effective inhibition of IL-1β release and reduce the occurrence of pyroptosis. Additionally, IFN-γ was highly expressed in γδT cells, which induced a profound transformation in mucosal immune response. Interestingly, IFN-γ was not only regulated by the downstream factor IRF1 but also an upstream element regulator to activate expression of IRF1 [
38]..The above evidence indicated an IFN-γ/IRF1/GSDMD signal transduction axis which played a nonnegligible role in pyroptosis.
Gluten is a trigger of pyroptosis, which not only regulates the expression of IFN-γ but also influences the infiltration of γδT cells. High PES was demonstrated in duodenal tissue of CeD patients after gluten challenge, suggesting that pyroptosis might be regulated by gluten. After intake of gluten, the infiltration of γδT cells was increased, while the gluten-free diet markedly rebuilt the expression of IFN-γ in γδT cells. In vitro research on Caco-2 cells further claimed that combined administration of gluten and IFN-γ could activate caspase-1, which cleaved GSDMD-FL into GSDMD-N and caused cell perforations, promoting the release of IL-1β and mediating pyroptosis. Above all, we conceived that gluten promoted the upregulation of IFN-γ expression in γδT cells and changed the quantity of γδT cells, which together resulted in epithelial cells pyroptosis.
Due to no specific symptoms, serological marker, and significant endoscopic and pathological changes in the pathogenesis of CeD, its clinical diagnosis (especially mild or early type) is often missed or mistaken. Based on the LASSO regression model, our findings propose new diagnostic strategies including γδTCR and PES, which have shown considerable distinction in predicting the incidence of CeD in both derivation and validation cohorts.
Conclusions
In conclusion, by uniting multiple transcriptome profiles, our study constructed an immune cell landscape of CeD, and established a novel association between immunity, inflammation, and programmed cell death. While individuals with the genetic background of CeD were stimulated by a gluten diet, a significant infiltration of γδT cells would be initiated in their duodenal mucosal. By producing a large amount of IFN-γ, γδT cells promoted the expression of GSDMD in epithelium by upregulating IRF1 and triggered epithelial cells to be in a “pre-pyroptotic” state. After gluten intake, epithelial cells underwent pyroptosis, resulting in an irreversible programmed cell death which eventually led to duodenal villi atrophy and malabsorption of nutrients. These conclusions might provide new insight into the cellular and molecular mechanisms, facilitate the clinical diagnosis, and propose novel ideas for the treatment of CeD.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.