Aging-related CD8⁺ T cell alterations and calcium/calmodulin dependent protein kinase 1D activation in the pathogenesis of diabetic kidney disease

Single-cell RNA sequencing data analysis

Single-cell RNA sequencing (scRNA-seq) analysis was conducted using two publicly available CKD datasets from the GEO database: GSE183276 and GSE211785. GSE183276 comprises 14 DKD patients and 20 controls, while GSE211785 consists of 13 DKD patients and 26 negative controls. Six DKD patients, six young individuals, and six elderly individuals were selected from these datasets for integrated and analysis (details provided in Table S1). Data quality control was performed using Seurat (version 5.1.0), applying filtering criteria that included genes expressed in each cell ranging between 500 and 4,000, and mitochondrial content less than 10%. Standardization and normalization were then performed, followed by dimensionality reduction and clustering based on the top 2,000 highly variable genes. Doublets were removed, and cell annotation was carried out using the sc-Type method (Ianevski, Giri & Aittokallio, 2022). Batch-effect correction was applied using Harmony (version 1.2.1).

Sc-type cell type annotation

To annotate cell types for different cell clusters (p), given an scRNA-seq dataset (X) with m genes and n cells, Sc-Type first normalizes the expression profile of each gene as a z-score across all cells. Subsequently, it compares only the positive and negative marker genes corresponding to distinct cell types within the specified tissue (immune cell subtype markers were obtained from the ScType database, details provided in Table S2) (Lake et al., 2023). Next, the expression level of each gene is then multiplied by its cell-type-specific score ($S_i^t$):${\mathrm{X}}^{{}^{\prime }}=\left({\left(\mathrm{Z}\left({\mathrm{X}}^{\mathrm{T}}\right)\right)}^{\mathrm{T}}\subseteq {\mathrm{M}}_t\right)\cdot S_i^t$ where X′ represents the scRNA-seq expression matrix for n cells and M_t denotes the vector of marker genes for all cell types in tissue t. Z denotes the z-score transformation. The transformed expression values for each specific cell type (c) in the tissue are aggregated into a cell-type-specific marker enrichment score: ${\chi }_c\prime =\frac{{\sum }_{i=1}^j{\chi }_i^{{}^{\prime }}}{\sqrt{j}}$-$\frac{{\sum }_{k=1}^l{\chi }_k\prime }{\sqrt{l}}$. Here, χ′ represents the unique column of X′ corresponding to a single cell; i and j are indices for positive marker genes of a given cell type, whereas k and l are indices for negative marker genes not expected to be expressed in the cell type. This transformation results in a normalized expression matrix of dimensions c×n, where each row corresponds to a cell type and each column corresponds to a cell. Finally, for each specific cluster p, values across the corresponding row (cell type) are summed to calculate the group-level enrichment score (ScType score): ScType score${}_{\mathrm{c}}={\sum }_{z\in p}{\chi }_c^z$. The cell type with the highest ScType score is assigned to the respective p cluster.

Bulk RNA sequencing data acquisition

Bulk RNA sequencing (bulk RNA-seq) data for DKD were retrieved from the GEO database. By using the keyword “Diabetic nephropathy” and specifying Homo sapiens as the organism, the dataset GSE166239 was identified, which includes six DKD samples and six healthy controls.

Mendelian randomization analysis of key genes

Differentially expressed genes (DEGs) specific to naive CD8⁺ T cells were identified by comparison with other cell subpopulations. Gene ID conversion was performed using the ‘clusterProfiler’ and ‘org.Hs.eg.db’ R packages. Two-sample MR analysis was conducted to investigate the potential causal relationship between these genes and DKD. Summary-level MR data were retrieved from the IEU Open GWAS database (https://gwas.mrcieu.ac.uk/). The training dataset was ebi-a-GCST90018832, and the validation dataset was finn-b-DM_NEPHROPATHY. These GWAS datasets were selected based on their inclusion of large European-ancestry cohorts with well-characterized DKD phenotypes, providing adequate statistical power and population comparability to the transcriptomic datasets derived primarily from Caucasian donors in the GEO database. MR analysis was conducted using ‘TwoSampleMR’ R package (version 0.6.8) and ‘ieugwasr’ R package (version 1.0.1) ensuring that selected single nucleotide polymorphisms (SNPs) met the following three assumptions: (i) relevance assumption: genetic instruments must be associated with exposure; (ii) independence assumption: genetic instruments must be independent of confounding factors; and (iii) exclusion restriction assumption: genetic instruments must influence the outcome only through exposure. SNPs strongly associated with exposure (p < 5  × 10⁻⁸) were selected as instrumental variables (IVs), with R² < 0.1 and kb = 10,000 applied to remove linkage disequilibrium effects. An F-statistic > 10 was was used to exclude weak instrument bias (Larsson, Butterworth & Burgess, 2023; Palmer et al., 2012). Five distinct MR methods were applied: inverse variance weighted (IVW) as the primary method, with MR Egger, weighted median (WM) simple mode, and weighted mode as Supplementary Methods. Sensitivity analyses were also performed, including Cochran’s Q test for heterogeneity, the MR-Egger intercept for pleiotropy, and a leave-one-out analysis to detect and correct potential biases or influential outliers.

Reverse mendelian randomization analysis

To assess whether DKD has a causal effect on the expression of key genes, a reverse MR analysis was conducted using the TwoSampleMR (version 0.6.8). In this analysis, DKD-associated SNPs were used as IVs, while the identified key genes were treated as outcome variables.

Clinical tissue samples

Human kidney tissues were obtained from patients with biopsy-confirmed DKD at different pathological stages (II, III, and IV), classified according to the Renal Pathology Society (Tervaert et al., 2010) criteria. All specimens were collected at the Affiliated Hospital of Nanjing University of Chinese Medicine (Jiangsu Province Hospital of Chinese Medicine). Each sample represented one disease stage and was used to access the expression and localization of CAMK1D in DKD renal tissue. Because kidney biopsy is rarely performed in healthy individuals, normal control samples were not available. Clinical and laboratory data at the time of biopsy, including eGFR, urinary albumin-to-creatinine ratio (UACR), and glycated hemoglobin (HbA1c), are summarized in Table S3. This study was approved by the Ethics Committee of Affiliated Hospital of Nanjing University of Chinese Medicine, Jiangsu Province Hospital of Chinese Medicine. All patients provided written informed consent (Ethical Application Ref: 2024NL-120-02).

Double immunofluorescence staining

Paraffin-embedded tissue sections were deparaffinized and underwent antigen retrieval using EDTA antigen buffer (Servicebio, Beijing, China). After blocking with 3% BSA for 30 min, a mixture of CAMK1D and CD8 primary antibodies was applied, followed by overnight incubation at 4 °C. Secondary antibodies were added and incubated at room temperature for 50 min, after which DAPI staining was performed for 10 min. An autofluorescence quencher (Servicebio, Beijing, China) was applied before mounting with an anti-fluorescence quenching mounting medium. Fluorescent images were captured using a fluorescence microscope (Nikon, Tokyo, Japan). Details of the primary antibodies are provided in Table 1.

scRNA-seq analysis of DKD, young, and elderly kidney tissues

A flow chart of the study design is presented in Fig. 1. scRNA-seq data were retrieved from the GEO datasets GSE211785 and GSE183276, from which kidney samples were selected from six DKD patients, six young healthy individuals, and six elderly healthy individuals (Table S1). In total, 48,980 single kidney cells were analyzed, comprising 20,905 from DKD patients, 9,160 from young controls, and 18,915 from elderly controls (Table 2). As these single-cell data sets originated from distinct datasets, batch effects were corrected using the “Harmony” algorithm. After initial quality control, 24 major cell clusters were identified at a resolution of 0.8. After doublet removal, a total of 26,041 cells were included for downstream scRNA-seq analysis, comprising 10,667 cells from DKD patients, 4,870 from young individuals, and 10,504 from elderly individuals (Table 3, Figs. S1A–SH). A total of 19 major cell clusters were identified and annotated using the “ScType” algorithm (Fig. S2A). These clusters were classified into nine distinct kidney cell populations: connecting tubule cells, endothelial cells, intercalated cells, pericytes, podocytes, principal cells, proximal tubule cells, thick ascending limb cells, and immune cells (Figs. 2A, 2B). Notably, immune cell infiltration was significantly increased in both DKD and elderly individuals, suggesting that immune activation is a common feature of both DKD and age-associated renal alterations.

Immune cell subpopulation analysis in DKD, young, and elderly kidney tissues

Numerous studies have demonstrated that immune cell infiltration and associated signaling pathways play essential roles in the progression of DKD (Fu et al., 2022; Guo et al., 2022; Peng et al., 2024; Tang & Yiu, 2020). In our data, immune cells accounted for a significantly greater proportion in DKD kidney tissues compared to healthy kidneys from young and elderly individuals (Fig. 2B). Immune cell clusters were extracted and subjected to batch correction and dimensionality reduction, resulting in 18 distinct clusters (Figs. S2B–SD). These clusters were re-annotated using the ScType method, and five primiary immune cell subtypes were identified: natural killer T cells, neutrophils, non-classical monocytes, plasma cells, and T cells (Fig. 2C). Among these, T cells were the most abundant immune cell type in DKD samples, followed by those from elderly individuals, and were least abundant in the young cohort (Fig. 2C). However, analysis of the relative proportion of T cells within the total immune cell compartment revealed a decreasing trend from young individuals to elderly individuals and DKD patients (Fig. 2D). These results indicate that although the absolute number of T cells increases, their relative abundance within the immune microenvironment declines with aging and disease, suggesting that T cells may play a progressively active yet tightly regulated role in DKD.

To further investigate T cell, their clusters were reanalyzed after batch effect correction and dimensionality reduction. A total of nine major T cell clusters were identified (Figs. S2E—SG), and three distinct T cell subtypes were annotated using ScType: CD8⁺ T cells, CD4⁺ T cells, and γδ T cells (Fig. 2E). The total number of CD8⁺ T cells followed the same increasing trend as total T cells across the three groups. However, the proportion of CD8⁺ T cells within the T cell population was highest in DKD, followed by the young group, and lowest in the elderly group, which differed from the proportion of T cells observed at the immune cell level (Fig. 2F). In DKD, both the absolute number and proportion of CD8⁺ T cells were the highest among all groups, indicating that CD8⁺ T cells represent the most prominently altered T cell subpopulation within the DKD immune microenvironment and may be involved in the pathogenesis of the disease.

Cell communication between CD8⁺ T cells and other cell types in DKD and aging microenvironments

To investigate potential connections between DKD and aging, we analyzed intercellular communication between CD8⁺ T cells and other cell types in both DKD and elderly kidney tissues using the CellChat package. In elderly kidney samples, CD8+T cells primarily interacted with podocytes, non-classical monocytes, and γδ T cells, via signaling pathways including MIF–(CD74+CD44), GZMA–F2R, and ANXA1–FPR1. In DKD samples, CD8⁺ T cells communicated with collecting duct cells and non-classical monocytes mainly via the ANXA1–FPR1 signaling axis (Figs. 3A–3D). These findings indicate that CD8⁺ T cells engage in ANXA1–FPR1-mediated intercellular crosstalk with non-classical monocytes in both aged and DKD kidney tissues.

Identification of hub genes associated with DKD and aging via scRNA-seq and MR

Based on the scRNA-seq analysis, DEGs in CD8⁺ T cells were identified through comparison with renal and other immune cell populations (Table S4). These DEGs were subsequently mapped to SNP-based exposure data using the “TwoSampleMR” and “ieugwasr” R packages. DKD-associated Genome-wide association studies (GWAS) summary statistics were retrieved from the IEU Open GWAS repository, with the dataset ebi-a-GCST90018832 used as the outcome variable. A total of 680 cis-acting SNPs were initially selected as IVs (Table S5). After linkage disequilibrium (LD) pruning (F > 10, R² < 0.1), 15 independent DEGs were retained for analysis (Table S6), and IVW analysis identified 10 of them as having significant causal associations with DKD (P < 0.05), including eight positive and two negative associations (Figs. 4–5A). Among these, RRP12, TRAF3, TTC21A, CAMK1D, and TRAV4 remained significant according to WM method. However, TRAF3, TTC21A, and TRAV4 did not show consistent results across simple mode, MR-Egger, and Wald ratio tests (P > 0.05), and RRP12 exhibited an opposing effect direction in the simple mode method. Taken together, only CAMK1D exhibited a consistently significant causal association with DKD and was therefore selected for further analysis.

Sensitivity analyses of CAMK1D revealed no significant heterogeneity by Cochran’s Q test (P > 0.05), supporting the stability of the MR results (Table S7, Table 4). To further validate the reliability of the results, funnel plots and forest plots were constructed (Figs. 5B, 5C). The MR-Egger regression intercept suggested no evidence of pleiotropy for CAMK1D (P > 0.05) (Table S8, Table 4, Fig. 5D). Leave-one-out sensitivity analysis confirmed that the causal association between CAMK1D and DKD was stable and not attributable to any single SNP (Fig. 5E).

Furthermore, validation using an independent GWAS dataset (finn-b-DM_NEPHROPATHY) further supported the association between CAMK1D and DKD (Figs. S3A–S3D, S4). Reverse two-sample MR analysis revealed no evidence of a causal effect of DKD on CAMK1D expression (Fig. 5G). The SNP rs883220, located within the CAMK1D locus, emerged as a lead signal significantly associated with DKD, with other SNPs in strong LD potentially contributing collectively to the observed genetic association (Fig. 5F).

Cell–cell communication profiles of CAMK1D+/- CD8⁺ T cells in the renal microenvironment

To elucidate the functional role of CAMK1D within the immune microenvironment, the expression profiles of 15 candidate genes were mapped onto the single-cell dataset. CAMK1D expression was enriched in podocytes, connecting tubule cells, endothelial cells, and immune cells (Figs. 6A, 6B). Within the immune compartment, CAMK1D expression was particularly enriched in neutrophils, T cells, and non-classical monocytes (Figs. 6C, 6D). Stratification of T cells revealed that CAMK1D expression was predominantly localized in CD8⁺ T cells (Figs. 6E, 6F), consistent with the prior findings.

Cell–cell communication analysis was performed on CD8⁺ T cells stratified by CAMK1D expression status (positive vs. negative). As shown in Fig. 7A, both CAMK1D + and CAMK1D- CD8⁺ T cells engaged in ANXA1–FPR1-mediated communication with non-classical monocytes. Additionally, CAMK1D + CD8⁺ T cells exhibited interactions with thick ascending limb (TAL) cells through the TGFB1–(ACVR1B+TGFBR2) signaling axis and with intercalated cells via the NAMPT–INSR pathway (Fig. S3E).

Metabolic pathways enrichment analysis of CAMK1D + CD8⁺ T cells in DKD patients revealed selective upregulation of terpenoid backbone biosynthesis, glycosphingolipid biosynthesis (lacto and neolacto series), and glycosaminoglycan degradation (Fig. 7B).

CAMK1D expression in RNA-seq data and DKD patient kidney tissues

Using the GSE142025 dataset, we analyzed renal gene expression profiles from healthy individuals and DKD patients. The mRNA expression level of CAMK1D was significantly elevated in DKD samples compared to healthy controls (P < 0.05) (Fig. 7C). This finding is consistent with the results of both MR and scRNA-seq analyses, further reinforcing the role of CAMK1D in DKD.

To further validate the involvement of CAMK1D in DKD, kidney biopsy samples were collected from three patients with a confirmed diagnosis of DKD at Jiangsu Provincial Hospital of Chinese Medicine in 2024. Double immunofluorescence staining revealed that both CAMK1D and CD8⁺T cell were expressed in the glomeruli and renal tubules of DKD kidneys, with partial co-localization observed (Fig. 7D).