Single-cell transcriptional profiling dissects the cellular ecosystem of chronic atrophic gastritis
To determine the cellular heterogeneity profile of chronic atrophic lesions, three pairs of CAG biopsies located in the gastric body, and neighboring normal mucosa were collected (Additional file 5: Table S1). For each biopsy, we prepared single cell solutions without the selection of certain cell types and applied the SeekOne platform to generate the scRNA-seq data. After removing low-quality cells, a total of 62,542 cells were retained for subsequent analysis, which yielded a median of 2,431 genes detected per cell (Fig. 1A). The number of cells from each biopsy is provided in Additional file 5: Table S1. The histological examination of CAG lesions through hematoxylin and eosin (HE) staining revealed the presence of pathological changes (Fig. 1B).
After the normalization of read counts and sample integration, we performed dimensionality reduction and unsupervised cell clustering and finally obtained 12 cell clusters in this dataset. Based on the expression of canonical marker genes and the top differently expressed genes of these clusters, we further annotated these clusters as T cells, B cells, myeloid cells, mast cells, fibroblasts, pericytes, endothelial cells, epithelial cells (Fig. 1C and 1E). We further separated cells by disease conditions and found that the cell types identified in the two groups were similar, and no cell type specific to a certain disease stage was found in our study (Fig. 1D). In our dataset, we found an increased abundance of T cells and epithelial cells and a decreased abundance of fibroblasts, endothelial cells, mast cells, and pericytes in CAG samples (Fig. 1F).
Identification of altered atrophic gastric epithelial cell phenotypes
We observed high cellular heterogeneity in the gastric epithelial cells. Projected into the two-dimensional principal-component analysis (PCA) space, the gastric epithelial cells were further clustered into 6 gastric lineages, namely, surface mucous cells (MUC5AC and TFF1), mucous neck cells (MUC6), chief cells (PGA4 and PGA5), parietal cells (ATP4A and ATP4B), enteroendocrine cells (CHGA), and glial cells (PLP1 and S100B)[10, 19] (Fig. 2A and 2C). We observed that the cell types found under the two conditions were the same. However, the proportions of cells in these two conditions are different. In both groups, the major cell population was surface mucous cells, followed by mucous neck cells, chief cells, parietal cells, enteroendocrine cells and glial cells. As expected, the loss of chief cells was observed in CAG group. The proportion of mucous neck cells also decreased in the CAG group, and surface mucous cells emerged in the CAG group (Fig. 2B). This finding suggested that the expansion of surface mucous cells may occur in CAG lesions. We further investigated the potential function of surface mucous cells and found that the surface mucous cells were enriched in digestion pathway and maintenance of the gastrointestinal epithelium, and more importantly, their expressed genes were involved in different metabolic processes, including xenobiotic metabolic processes and olefinic compound metabolic processes (Additional file 6: Table S2). This may suggest that in the state of CAG, surface mucous cells expand to undergo various metabolic processes to maintain normal gastric functions.
To illustrate the change in the expression program in each epithelial cell subtype during progression to atrophic gastritis, we then focused on the difference in the gene expression of each epithelial cell subtype between the two conditions. Interestingly, we observed the downregulation of the antimicrobial molecules DUOX2 and LCN2 in the surface mucous cells of the CAG group, suggesting an attenuated antimicrobial defense ability of surface mucous cells in the CAG group. We also detected some upregulated genes that were involved in the cell cycle G1/S phase transition in mucous neck cells, including SMARCC2, BCL7C, CDKN1A, indicating that mucous neck cells acquired an aberrant cell proliferation phenotype in the atrophic gastritis stage (Fig. 2E). We examined the transcriptional changes in chief cells and parietal cells and found that the metal detoxification genes MT1H, MT1M and MT1F were downregulated in chief cells and that the antioxidant genes GPX4, PRDX1 and PRDX3 were downregulated in parietal cells, indicating an impaired epithelial detoxification system in these two cell subpopulations (Fig. 2F).
We generated atrophic gastritis susceptibility gene scores based on the gene set provided by atrophic gastritis in the Harmonizome 3.0 database (https://maayanlab.cloud/Harmonizome/). We found that the score was higher in the epithelial compartment of the gastric mucosa (Additional file 1: Figure S1A). We further examined atrophic gastritis susceptibility gene score in different epithelial subtypes and found that the susceptibility gene score was higher in surface mucous cells, parietal cells, mucous neck cells and chief cells (Additional file 1: Figure S1B). More importantly, the susceptibility gene score was significantly higher in the chief cells and mucous neck cells of CAG tissue than in those of normal gastric tissue (Additional file 1: Figure S1C), which indicates that the atrophic gastritis susceptibility genes may be related to the aberrant alteration of the expression program in chief cells and mucous neck cells when progressing to CAG.
Chronic atrophic gastritis is associated with an increased exhausted T cell phenotype and a decreased cytotoxic T cell phenotype.
To further evaluate the transcriptional changes in the T cell population during the development of CAG, we reclassified the T cell population into three subpopulations according to the specific marker expressed by each cell subtype, namely, CD4+ exhausted T cells, CD8+ effector T cells, and CD8 + stress response T cells (Fig. 3A and 3B, Additional file 2: Figure S2 A-C). We further compared the distribution of T cell subclusters. The percentage of CD4+ exhausted T cells was greater in the CAG group, the percentage of CD8+ stress response T cells remained similar under both conditions, and the percentage of CD8+ effector cells was correspondingly decreased, suggesting infiltration of the exhausted phenotype and a reduction in effector functions involved in the immune compartment of CAG (Fig. 3C). Notably, CD8+ effector T cells have high expression levels of chemokine molecules, including CCL4, CCL5, and CCL3, which suggests that CD8+ effector T cells are a source of immune recruiting cells that release chemoattractants and enhance the immune response [20]. Additionally, CD8+ effector T cells express high levels of cytotoxic molecules, including GZMA, NKG7, GNLY, GZMK [21], the antimicrobial molecule IFNG [22] and the antiviral molecule IFITM2 [23]. As expected, CD4+ exhausted T cells express high levels of the inhibitory molecules CTLA4, TOX2, ICOS, CD28, PDCD1, and TOX [21]. We found that there is a subpopulation of T cells that are in a state of stress response and that highly express the molecules BTG1, BTG2, and ZFP36 to restrain T cell activation and proliferation [24, 25] (Fig. 3D). In the search for upstream regulatory mechanisms of different T cell subtypes, we inferred TF activity using the Dorothea algorithm [26, 27]. We observed differences in TF activity among the three T cell subtypes. TFs with higher activity scores, including IRF4, GATA6 and STAT3, in the CD4+ T exhausted cell subtype were observed to have low activity in the CD8+ T effector cell subtype. However, the CD8+ T effector cell subtypes had high TF activity scores for IRF1, IRF3 and STAT4 (Additional file 2: Figure S2D). This indicates differences in transcription between exhausted CD4+ T cells and effector CD8+ T cells. We further compared the TF activity of each T cell subtype between the CAG and control conditions. We found that the active TFs of each cell subtype under the two conditions were similar, but in the CAG state, exhausted T cells had higher TF activity scores, and effector T cells had lower TF activity scores than did the control cells (Additional file 2: Figure S2E). This result suggested that the transcription program related to the exhausted phenotype is more active in the CAG state and that the transcription program related to the effector phenotype is more active in the control state.
Next, we conducted cell differential expression analysis between the two conditions, and higher expression levels of exhausted-related molecules, including PDCD1 and CTLA4, and lower expression levels of the cytotoxic molecules GZMA and NKG7 were detected in the T cells than in the control cells (Fig. 3E). We also observed an infiltration of PDCD1-positive and CTLA4-positive T cells in the tissue microenvironment of CAG lesions. The cell ratio of PDCD1-positive T cell in the CAG group was 0.175; however, in the control group, the ratio decreased to 0.097. The ratio of CTLA4-positive T cells in the CAG group was 0.202; in the control group, the ratio was 0.093. The number of cytotoxic T cells in the tissue microenvironment decreased. The cell ratio of NKG7-positive T cells was 0.393 in the CAG group and 0.538 in the control group, and the ratio of GZMA-positive T cells was 0.33 in the CAG group and 0.470 in the control group (Fig. 3F). We compared the signature scores between the CAG and non-CAG groups, and found that the exhaustion score of CD4+ exhausted cells was significantly elevated in the CAG group, while the cytotoxic score of CD4+ effector cells was significantly decreased in the CAG group (Fig. 3G). These results suggested that the development of atrophic gastritis is associated with an exhausted phenotype and reduced cytotoxic activity of T cells.
Chronic atrophic gastritis is characterized by impaired phagocytic ability of C1Q + macrophages
We further investigated the potential roles of the innate immune system involved in the CAG. Myeloid cells were further divided into C1Q+ macrophages, CLEC17A + DCs and IRF4 + DCs based on the unsupervised clustering and canonical cell markers (Fig. 4A and 4B). To systematically study the functional implications of myeloid cells in CAG, we identified DEGs and further subjected them to GO analysis, which revealed a striking difference in pathway enrichment between the CAG group and the control group. Granulocyte chemotaxis, granulocyte migration, neutrophil migration, phagocytosis, receptor-mediated endocytosis, cellular response to lipopolysaccharide, and cellular response to molecules of bacterial origin were observed to have high pathway activity in the control group, indicating that these pathway activities diminished in the CAG state. In the CAG group, we observed enrichment of genes related to T cell differentiation, regulation of T cell activation, positive regulation of antigen receptor-mediated signaling pathway, the Fc receptor signaling pathway, and the Fc-gamma receptor signaling pathway (Fig. 4C, Additional file 9: Table S5). Specifically, the lower expression of MRC1 and FPR1 in C1Q+ macrophages in the CAG group indicated the impaired phagocytic activity of C1Q+ macrophages in the CAG group compared with the control group [28]. We also observed an infiltration of dendritic populations expressing FCMR and FCER2 (Fig. 4D and 4E). Next, we investigated the pattern recognition system and chemokine system of C1Q+ macrophages in CAG. We generated gene scores for Toll-like receptors and chemokines and found decreased expression levels of TLR family molecules, including TLR4 and TLR2, and chemokine molecules, including CCL3, CCL4 and CXCL2, in C1Q+ macrophages cells of chronic atrophic tissue (Additional file 3: Figure S3A).
To gain further insights into the metabolic pathway activity of C1Q+ macrophages in non-atrophic and atrophic samples, we used the scMetabolism package to systematically quantify metabolic activity in our scRNA-seq data. Interestingly, we found that the activity of multiple metabolic pathways was downregulated in atrophic samples (Additional file 3: Figure S3B). We found that the nitrogen metabolism activity and glycerophospholipid metabolism were upregulated in MRC1+ macrophages, FPR1+ macrophages and CD14+ macrophages, and downregulated in atrophic samples (Additional file 3: Figure S3C). Phagocytosis is the critical process by which macrophages ingest and eliminate pathogens, and requires dynamic changes in plasma membrane fusion and fission. Lipid synthesis has been linked to enhanced phagocytosis [29]. Taken together, we concluded that the C1Q+ macrophages in atrophic gastric mucosa were more inactive in cellular metabolism, which may be associated with a decreased phagocytosis phenotype. Nitrogen metabolism activity and glycerophospholipid metabolism are downregulated in atrophic gastric mucosa to limit the phagocytosis of C1Q+ macrophages.
Mast cells constitute a major sensory arm of the innate immune system. In both the normal control group and the CAG group, we detected infiltration of mast cells (Fig. 4F). We further compared the differentially expressed genes. GO analysis revealed that mast cells in the GAG groups mainly expressed genes involved in cellular response to extracellular stimulus, regulation of TOR signaling, positive regulation of pattern recognition receptor signaling pathway, myeloid differentiation and immune response regulation signaling pathway (Fig. 4G, Additional file 10: Table S6). In addition, several mast activation related markers such as IL1RL1, MAPK1, POUF1 and LYN [30] were upregulated in CAG lesions (Fig. 4H). These results suggested that mast cells enhance active sensory ability in the state of CAG.
Fibroblast-immune cell crosstalk promotes the inflammatory microenvironment
Fibroblasts play an important role in regulating inflammation. In our research, we discovered four fibroblast clusters, including CCL11+APOE+ fibroblasts, CXCL14+SOX6+ fibroblasts, HIPP+ myofibroblasts, and RGS5+ pericytes (Fig. 5A and 5B, Additional file 4: Figure S4). CCL11+APOE+ fibroblasts and CXCL14+SOX6+ fibroblasts were the major fibroblast populations found in our samples. We further investigated the potential functions of these fibroblast subtypes. Based on the markers and enrichment analysis of these two fibroblast subtypes, we discovered that CCL11+APOE+ fibroblasts express genes involved in the regulation of immune cell recruitment and the expression of cytokine (IL33, IL34), hence exhibited increased cytokine activity and cellular chemotaxis ability. Moreover, CXCL14+SOX6+ fibroblasts exhibited elevated BMP signaling activity (Fig. 5C).
We further explored the cellular interactions of CCL11+APOE+ fibroblast subpopulations with other immune subtypes in CAG lesions. We used NicheNet to infer ligand-target regulatory potential with a prior model. We observed that fibroblasts expressing CCL2 displayed high regulatory potential with chemokine molecules expressed by C1Q+ macrophages, including CCL3, CCL4 and CXCL2, indicating that these fibroblast subpopulations might increase the immune recruitment ability of C1Q+ macrophages [31]. In addition, these fibroblast subpopulations also have high regulatory potential for ICAM1, which suggests the CCL2 expressing fibroblasts can promote leukocyte trafficking and immune cell effector functions [32]. Specifically, the CCL2 ligand exhibits a high level of regulatory potential with the well-characterized antimicrobial molecule SOD2, indicating CCL2 expressing fibroblasts promote the killing of infectious cells [23] (Fig. 5D). In addition, IL33 expressing fibroblasts exert regulatory effects on the innate immune system by regulating Toll-like receptor signaling pathways (TLR2, TLR4 and LY96) and scavenger receptors (CD14 and CD163) (Fig. 5D). We also observed strong regulation effects of IL15 on FASLG, GZMB and KLRK1, which are well-characterized cytotoxicity-related genes, between fibroblasts and CD8+ effector T cells (Fig. 5E), suggesting that IL15-expressing fibroblasts play active roles in T cells cytotoxicity. We found that both CCL2+ fibroblasts and IL15+ fibroblasts infiltrated in microenvironment of the CAG and control samples, and IL33+ fibroblasts emerged in the CAG samples (Fig. 5F). Therefore, we conclude that based on the gene expression of fibroblasts and predicted target genes, CCL2+, IL33+, and IL15+ fibroblasts might have persistent effects on the recruitment and phagocytosis of C1Q+ macrophages and the cytotoxicity of CD8+ effector cells in the development of CAG.