Abstract β-cell dysfunction and dedifferentiation towards an α-cell-like phenotype are hallmarks of type 2 diabetes. However, the cell subtypes involved in β-to-α-cell transition are unknown. Using single-cell and single-nucleus RNA-seq, RNA velocity, PAGA/cell trajectory inference, and gene commonality, we interrogated α-β-cell fate switching in human islets. We found five α-cell subclusters with distinct transcriptomes. PAGA analysis showed bifurcating cell trajectories in non-diabetic while unidirectional cell trajectories from β-to-α-cells in type 2 diabetes islets suggesting dedifferentiation towards α-cells. Ten genes comprised the common signature genes in trajectories towards α-cells. Among these, the α-cell gene SMOC1 was expressed in β-cells in type 2 diabetes. Enhanced SMOC1 expression in β-cells decreased insulin expression and secretion and increased β-cell dedifferentiation markers. Collectively, these studies reveal differences in α-β-cell trajectories in non-diabetes and type 2 diabetes human islets, identify signature genes for β-to-α-cell trajectories, and discover SMOC1 as an inducer of β-cell dysfunction and dedifferentiation. Subject terms: Cell signalling, Diabetes, Differentiation __________________________________________________________________ β-cell dysfunction and dedifferentiation towards an α-cell-like phenotype are hallmarks of type 2 diabetes. Her,e the authors detect five α-cell subpopulations, find differences between healthy and diabetic donors, and identify SMOC1 as an inducer of human β-cell dysfunction and dedifferentiation. Introduction Glucagon-producing pancreatic α-cells are recognized as important physiological regulators of life-threatening hypoglycemia by counteracting the effects of insulin on glucose homeostasis^[44]1,[45]2. In diabetes, patients display postprandial hyperglucagonemia, which exacerbates hyperglycemia^[46]3–[47]6. Despite these important observations, initial studies on human pancreatic islet cell heterogeneity have focused mainly on insulin-producing β-cells. Recently, however, more studies have emerged describing specific human islet α-cell subpopulations that participate in normal and dysregulated glucose homeostasis^[48]7–[49]10. Indeed, recent evidence indicates that there is variation in glucagon content among human α-cells^[50]7 and that α-cell functional heterogeneity is linked to α-cell maturation in type 2 diabetes (T2D)^[51]8. Furthermore, the subpopulation of glucagon-like peptide-1 (GLP-1) secreting α-cells is increased in human T2D islets^[52]9. Finally, elevated serum amino acids induce a subpopulation of α-cells to initiate pancreatic neuroendocrine tumor formation^[53]10. Collectively, these studies indicate that analyzing α-cell heterogeneity and the mechanisms controlling their identity can be of great importance in health and disease. Single-cell RNA sequencing (scRNA-seq) has revolutionized the identification and analysis of different cell types within heterogeneous cell populations. By algorithmically clustering the data, it is possible to annotate distinct cell types, and with varying hyperparameters for granularity such as Louvain resolution, we can investigate cell subpopulations with distinct transcriptomes in an unbiased manner^[54]11. In the human islet, scRNA-seq has uncovered several α- and β-cell subtypes with different transcriptome profiles that can predict the maturity of these islet cell subtypes in basal and diabetic conditions^[55]12–[56]19. However, how these human α- and β-cell subtypes can transcriptionally transition from one cell subtype to another remains understudied. Furthermore, T2D is characterized by a decrease in functional β-cells, in part due to dedifferentiation and conversion to other endocrine cells, including glucagon-producing α-like-cells^[57]20,[58]21, a concept that may in part support the hyperglucagonemia encountered in diabetes^[59]3–[60]6. However, whether specific β-cell subpopulations can transition into α-cells, and whether there is a specific gene signature involved in this process, is unknown. In the current study using scRNA-seq and snRNA-seq of human islets isolated from adult non-diabetic donors, we identified five GCG-expressing α-cell subtypes with different transcriptome profiles. We found that one of the α-cell subtypes, named “AB cells” is a multihormonal α-cell subpopulation that could potentially transition in a bifurcated manner into either mature α- or β-cells. However, trajectory analysis of scRNA-seq data from human islets isolated from T2D donors obtained from the Human Islet Research Network (HIRN)-Human Pancreas Analysis Program (HPAP) database showed unidirectional trajectories from β-cells to α-cells suggesting potential pressure on β-cells to become less differentiated or converted to α-like cells. Analysis of common genes on the trajectories from β- to α-cells in human islets isolated from adult non-diabetic donors identified SMOC1, PLCE1, PAPPA2, ZNF331, ALDH1A1, SLC30A8, BTG2, TM4SF4, NR4A1 and PCSK2 as signature genes. Among these, SMOC1 (SPARC-related modular calcium-binding protein-1) which encodes for an extracellular glycoprotein of the SPARC (secreted protein, acidic and rich in cysteine)-related modular calcium-binding protein family, was recently identified as one of the top islet-derived genes encoding a secreted protein in obese mice^[61]22–[62]29. However, the role of SMOC1 in islet cells is unknown. Here, we show that SMOC1 is expressed mostly in α-cells in adult human non-diabetic islets and its expression negatively correlated with INS expression. Interestingly, SMOC1 mRNA and protein were detected in T2D β-cells. SMOC1 expression in non-diabetic human islets and EndoC-βH1 cells reduced INS expression, diminished glucose-stimulated insulin secretion (GSIS), decreased the expression of β-cell identity genes and enhanced the appearance of T2D α-cell-like and T2D β-cell-like gene features in these cells. Collectively, these studies identify α-cell subpopulations and analyze transcriptional transitions between β-cells and α-cells as a function of the pathophysiological context. We also identify SMOC1 as a gene of previously unrecognized importance and relevance to β-cell dedifferentiation in T2D. Results scRNA-seq and snRNA-seq distinguish five α-cell subtypes in human islets We have previously shown that integrated scRNA-seq and snRNA-seq analysis of human islets can distinguish three β-cell subtypes with different transcriptome profiles^[63]17. Combining the datasets from both RNA-seq platforms increases analytical power by providing additional information on cytoplasmic and nuclear transcriptomes. Using the same dataset and integrated reference, we aimed here to define α-cell subtypes in human islets (Fig. [64]1a, Supplementary Data [65]1). After initial sub-setting (Fig. [66]1b), we subclustered the α-cell cluster with a Louvain resolution of 0.8. We grouped resulting minor α-cell clusters containing fewer than ten cells into clusters with closest proximity, yielding five α-cell subclusters: α1, α2, α3, α4, and AB (an α-cell subpopulation with INS expression) (Fig. [67]1c, d). We employed this unbiased strategy instead of marker-based subcluster assignment since it considers both the entire transcriptome and the multiple PCA dimensions assigned to calculate clusters. The total number of cells analyzed per subcluster and transcriptomics platform in the three human islet preparations appear in Supplementary Fig. [68]1a. The α1, α2 and α3 subclusters comprised most α-cells (77–80%) and their proportions were similar between scRNA- and snRNA-seq datasets. In contrast, α4 and AB subclusters were different in proportion between scRNA- and snRNA-seq datasets (Fig. [69]1e), with a larger proportion of α4 cells (20% vs. 8%) and fewer AB cells (3% vs. 12%) in scRNA-seq than in snRNA-seq, respectively. This suggests that snRNA-seq with pre-mRNA analysis can reveal more AB cells (α-cell with INS expression, see below) than scRNA-seq. Fig. 1. Experimental design, unsupervised clustering, and sub-clustering of human α cells by scRNA-seq and snRNA-seq of islets from adult human non-diabetic donors. [70]Fig. 1 [71]Open in a new tab a Human islet processing and data generation scheme. Created in BioRender. Garcia-Ocana, A. (2025) [72]https://BioRender.com/nizfh1l. b Unsupervised clustering, cell type annotated UMAP and separated α-cell cluster (below). c. Pre-annotated α-cell subclusters by assigning Louvain resolution 0.8. d. Annotated α-cell clusters according to the UMAP location relative to neighboring β-cells^[73]17 and gene expression. UMAP is split by the processing type – scRNA-seq (left) and snRNA-seq (right). e Proportion of the α-cell subtypes in scRNA-seq and snRNA-seq datasets. Gene expression and pathway enrichment analysis of human islet α-cell subtypes GCG and ALDH1A1 were highly expressed across the five α-cell subclusters in both scRNA- and snRNA-seq datasets. TTR and CRYBA2 showed substantially higher expression levels in scRNA-seq than in snRNA-seq datasets, where their expression was marginally detectable or absent in some α-cell subtypes (Fig. [74]2a). This provides further support for our previous observation that the GCG, CRYBA2, ALDH1A1, TTR gene set optimally defines α-cells in scRNA-seq analysis of human islets but is suboptimal for snRNA-seq annotation^[75]17. Therefore, we next analyzed the expression of our recent α-cell gene set derived from snRNA-seq^[76]17 and observed that PTPRT, FAP, PDK4 and LOXL4 showed greater relative expression in clusters in the snRNA-seq than in the scRNA-seq dataset (Fig. [77]2b). Interestingly, these four latter genes showed a progressively decreasing pattern from α1 to AB cells. Fig. 2. Gene expression, DEG, and pathway enrichment analysis of human α-cell subclusters by scRNA-seq and snRNA-seq of islets from adult human non-diabetic donors. [78]Fig. 2 [79]Open in a new tab a Gene expression of 4 canonical α-cell markers, split by processing types, in the different α-cell subclusters. b Gene expression of the previously identified^[80]17 single nucleus α-cell markers in the different α-cell subclusters. c Dot-plot visualization of top 4 differentially expressed genes in each α-cell subcluster. d Differentially expressed genes for α-cell sub-populations in scRNA-seq data (top) and snRNA-seq data (bottom). e Pathway enrichment analysis for α-cell sub-populations. We searched pathways using keywords for pancreatic endocrine cells (α/β/δ/PP/ε), glucagon and hormone signaling/processing/secretion, metabolism, and cellular development (differentiation, precursor, dedifferentiation, development, senescence) and arrange the pathway order according to the enrichment patterns from α1 to AB cells. To better define gene sets that identify the different α-cell subclusters and their transcriptome differences, we performed differential gene expression (DEG) analysis for every cluster against all the remaining clusters (Fig. [81]2c, d, Supplementary Fig. [82]1b, Supplementary Data [83]2). The α1 subcluster displayed selectively higher expression of NEAT1, ACTG1, PEAK1 and ACTB. The α2 subcluster favored the expression of FAP, PCSK2, SLC30A8, and GLS. The α3 subcluster most highly expressed HSPA1A, HSPH1, DNAJB1, and PLCG2. The α4 subcluster favored PCSK1N, GAPDH, TTR, and SNHG29. The AB cell subcluster displayed expression of pan-endocrine hormonal genes such as INS, SST, PPY and IAPP (Fig. [84]2c, Supplementary Fig. [85]1b, Supplementary Data [86]2). Interestingly, most of the differentially expressed genes in AB cells were defined by the snRNA-seq which employs intron-inclusive references. This suggests that a large portion of