Single-cell and pseudobulk transcriptomic profiling reveals immune-enriched disease- associated signatures in human pancreatic islets from type 2 diabetes

Technical Summary: Single-cell and Pseudobulk Transcriptomic Profiling in Human Pancreatic Islets from Type 2 Diabetes

Problem Statement
Type 2 diabetes (T2D) is characterized by progressive dysfunction of pancreatic islet cells, yet the molecular mechanisms underlying this dysfunction remain complex due to the cellular heterogeneity of the islet microenvironment. Traditional bulk transcriptomic approaches often obscure cell type-specific disease signals, particularly when alterations are subtle or confined to less abundant populations such as immune cells or specific endocrine subtypes. While single-cell RNA sequencing (scRNA-seq) offers higher resolution, previous studies have highlighted challenges in interpreting T2D-associated changes due to variability in donor composition, sequencing depth, and analytical design. This study addresses the question of how T2D alters the cellular landscape and cell-type-specific transcriptional programs in human pancreatic islets at single-cell resolution, specifically seeking to determine if disease signals are global or compartment-specific.

Methodology
The study performed a secondary analysis of publicly available scRNA-seq data (GEO accession GSE221156) comprising 37 human pancreatic islet samples (20 T2D donors and 17 non-diabetic donors). The analytical workflow included:

1. Preprocessing and Quality Control (QC): Raw count matrices were processed using the Seurat package in R. Stringent cell-level filtering retained cells with 500–10,000 detected genes, 1,000–50,000 UMIs, and <10% mitochondrial content. This process reduced the dataset from ~260,000 to 115,404 high-quality cells.
2. Integration and Clustering: To harmonize technical variation across samples while preserving biological heterogeneity, the authors utilized Seurat v5's IntegrateLayers function with Canonical Correlation Analysis (CCA). Graph-based clustering identified 35 global clusters and 21 refined clusters within an islet-enriched subset.
3. Annotation: Cell types were annotated using canonical marker genes (e.g., INS for beta cells, PRSS1 for acinar cells, PTPRC for immune cells) and validated through feature plots, violin plots, and heatmaps.
4. Differential Expression (DE) Analysis:
   • Cell-level: Wilcoxon rank-sum tests were performed within annotated cell types to identify DEGs between T2D and non-diabetic cells.
   • Pseudobulk: To account for biological replication and avoid treating cells from the same donor as independent observations, raw counts were aggregated at the sample level. DESeq2 was used to model these pseudobulk counts with a negative binomial generalized linear model, testing for disease status effects within specific cell compartments.
5. Functional Enrichment: Significant marker genes and exploratory DEGs were submitted to DAVID Bioinformatics Resources for functional overrepresentation analysis.

Key Results

• Global Landscape: Integrated analysis revealed substantial cellular heterogeneity but no clear global separation between T2D and non-diabetic samples. Disease status was not the dominant driver of global variance; instead, cell identity and inter-donor variability structured the data.
• Cell-Level DE Analysis: No robust disease-associated genes were identified in major cell populations (beta-like, acinar-like, ductal-like) after multiple-testing correction.
• Pseudobulk Analysis:
  • In beta-like, acinar-like, and ductal-like compartments, no genes met the stringent significance threshold (FDR < 0.01 and |log2FC| > 2).
  • In the immune-enriched compartment, an exploratory signal was detected. The chemokine gene CXCL3 was upregulated in T2D samples with a log2 fold change of 2.521 and an adjusted p-value of 0.03854 (FDR < 0.05), though it did not meet the stricter FDR < 0.01 threshold.
  • Other genes (e.g., CHI3L1, IL33, STK38L) showed nominal significance in immune cells but failed to pass high-confidence FDR corrections.
• Functional Enrichment: DAVID analysis of significant marker genes highlighted signatures related to the extracellular matrix, secreted proteins, protein digestion/absorption, and proteolytic activity. The immune-enriched exploratory gene list also enriched for "Secreted" and "Extracellular region" terms.

Significance and Claims
The paper claims that transcriptional alterations in human T2D islets are subtle, compartment-specific, and best detected through donor-aware, cell-type-resolved analysis rather than global comparisons. The authors argue that the absence of broad, FDR-significant differentially expressed genes in major cell types suggests that T2D does not induce a uniform, global transcriptomic shift. Instead, disease-related changes appear distributed across specific cellular niches.

The study supports a model where the disease signal is masked by inter-donor variability and cellular heterogeneity, necessitating pseudobulk approaches to detect subtle effects. The identification of CXCL3 in the immune-enriched compartment is presented as a candidate signal reflecting localized inflammatory communication within the islet microenvironment, consistent with the known role of inflammation in T2D pathophysiology. However, the authors maintain a modest tone, characterizing these findings as "exploratory" and noting that the dataset lacks the statistical power to definitively call individual genes significant after stringent FDR correction. The study concludes that larger cohorts, independent datasets, and orthogonal validation are required to confirm whether these signals represent reproducible mechanisms of human T2D islet pathology.