Abstract
The inflammatory response is a direct factor leading to changes in the microenvironment of renal tissues. The immune cells and stromal cells infiltration were the essential characteristics of diabetic nephropathy (DN). Based on the differentiation of the microenvironment, describing the heterogeneity of DN may provide a new approach to explore the mechanisms of disease progression. This study was aimed to classify DN samples based on the infiltration levels of immune cells and stromal cells, describe the microenvironment heterogeneity of DN samples, explore the potential mechanisms of phenotypic differentiation, screen key pathogenic genes, construct a quantitative scoring model to describe the microenvironment, and investigate the role of key pathogenic genes of DN. We downloaded RNA sequencing datasets of DN tissue and normal kidney tissue (GSE142025 and GSE96804) from the Gene Expression Omnibus (GEO) database. The RNA sequencing data was transformed into immune cell and stromal cell infiltration data using the xCell algorithm. Ward's method was used for consensus clustering to identify different phenotypes of DN. We screened out the key pathogenic genes associated with phenotypes, and established a scoring model through principal component analysis which was tested the reliability and accuracy in the training cohort and the validation cohort. We explored the influence of key pathogenic genes on the biological behavior of human mesangial cells. Based on the heterogeneity of the diabetic nephropathy (DN) microenvironment, this study identified 15 key pathogenic genes: FN1, EGR1, TPM1, CCND2, COL1A2, TGFB2, COL6A3, ITGA11, ABCC9, THBS2, TNC, COL3A1, C7, C1QC, and ITGB6. The PCA score constructed from these genes (i.e., their first principal component, PC1) demonstrated good efficacy in distinguishing normal samples from DN samples (AUC = 0.90, 95 % CI [0.82-0.97]), differentiating DN subtypes with distinct microenvironments (AUC = 0.99, 95 % CI [0.97-1.00]), and stratifying DN samples at different disease stages. This score showed significant positive correlations with immune score (r = 0.75, P = 1.6e-7) and stromal score (r = 0.96, P < 2.2e-16). Under high-glucose stimulation, the protein and mRNA expression of ABCC9 in mesangial cells increased over time. Knockdown of ABCC9 partially counteracted the high glucose-induced increase in apoptosis and enhancement of migration in mesangial cells. In an external validation cohort, both the PCA score (AUC = 0.91 for distinguishing normal from DN; AUC = 0.95 for differentiating DN subtypes) and ABCC9 expression (AUC = 0.93 for distinguishing normal from DN; AUC = 0.90 for distinguishing early from advanced DN) further confirmed their association with the DN microenvironment and disease progression. FN1, EGR1, TPM1, CCND2, COL1A2, TGFB2, COL6A3, ITGA11, ABCC9, THBS2, TNC, COL3A1, C7, C1QC, and ITGB6 are potential key pathogenic genes in DN. The PCA score constructed based on these genes can distinguish between normal tissue phenotype and DN phenotype, quantitatively describing the immune microenvironment and stromal microenvironment of DN, reflecting the progression of DN. Knocking out ABCC9 can counteract the increased apoptosis and migration of glomeruli mesangial cells induced by high glucose.