Single-cell RNA sequencing combined with single-cell proteomics identifies the metabolic adaptation of islet cell subpopulations to high-fat diet in mice

Abstract

Aims/hypothesis

Islets have complex heterogeneity and subpopulations. Cell surface markers representing alpha, beta and delta cell subpopulations are urgently needed for investigations to explore the compositional changes of each subpopulation in obesity progress and diabetes onset, and the adaptation mechanism of islet metabolism induced by a high-fat diet (HFD).

Methods

Single-cell RNA sequencing (scRNA-seq) was applied to identify alpha, beta and delta cell subpopulation markers in an HFD-induced mouse model of glucose intolerance. Flow cytometry and immunostaining were used to sort and assess the proportion of each subpopulation. Single-cell proteomics was performed on sorted cells, and the functional status of each alpha, beta and delta cell subpopulation in glucose intolerance was deeply elucidated based on protein expression.

Results

A total of 33,999 cells were analysed by scRNA-seq and clustered into eight populations, including alpha, beta and delta cells. For alpha cells, scRNA-seq revealed that the Ace2^low subpopulation had downregulated expression of genes related to alpha cell function and upregulated expression of genes associated with beta cell characteristics in comparison with the Ace2^high subpopulation. The impaired function and increased fragility of ACE2^low alpha cells exposure to HFD was further suggested by single-cell proteomics. As for beta cells, the CD81^high subpopulation may indicate an immature signature of beta cells compared with the CD81^low subpopulation, which had robust function. We also found differential expression of Slc2a2 in delta cells and a potentially stronger cellular function and metabolism in GLUT2^low delta cells than GLUT2^high delta cells. Moreover, an increased proportion of ACE2^low alpha cells and CD81^low beta cells, with a constant proportion of GLUT2^low delta cells, were observed in HFD-induced glucose intolerance.

Conclusions/interpretation

We identified ACE2, CD81 and GLUT2 as surface markers to distinguish, respectively, alpha, beta and delta cell subpopulations with heterogeneous maturation and function. The changes in the proportion and functional status of islet endocrine subpopulations reflect the metabolic adaptation of islets to high-fat stress, which weakened the function of alpha cells and enhanced the function of beta and delta cells to bring about glycaemic homeostasis. Our findings provide a fundamental resource for exploring the mechanisms maintaining each islet endocrine subpopulation’s fate and function in health and disease.

Data availability

The scRNA-seq analysis datasets from the current study are available in the Gene Expression Omnibus (GEO) repository under the accession number GSE203376.

Graphical abstract

Supplementary Information

The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-022-05849-5.

Introduction

Pancreatic islets, containing various endocrine cells such as alpha, beta, delta, pancreatic polypeptide (PP) and epsilon cells, are responsible for nutrient utilisation and energy metabolism [1]. As insulin is the main regulator of blood glucose, most studies of islet dysfunction have been limited to beta cells. However, increasing evidence has emphasised the synergistic interaction of alpha, beta and delta cells in glucose homeostasis [2–4]. In type 2 diabetes, several studies indicate that progressive dysfunction of alpha, beta and delta cells occurs in the early stages. The metabolic load from obesity and hyperlipidaemia causes insulin resistance and higher insulin demand, which stimulates beta cell compensatory proliferation [5]. A combination of an increased proportion of alpha cells migrating towards the islet centre and inappropriately increased pulsatile glucagon secretion is another signature of type 2 diabetes [6, 7]. Furthermore, significantly decreased delta cell numbers preceding the gradual decline in beta cell numbers was observed in people with diabetes [8, 9]. The above studies indicate that the dynamic balance of each islet subpopulation is disrupted across the disease course. The prevalence of glucose intolerance in Chinese adults was 35.2% in a study based on 2018 data [10]. High energy intake characterised by high fat intake leads to insulin resistance, beta cell dysfunction and type 2 diabetes [11–13]. However, the changes to islet cell subsets in high-fat-induced glucose intolerance remain unclear.

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful approach to explore intercellular heterogeneity, cell–cell interactions and the unique function of individual cells in the context of the cellular microenvironment [14, 15]. Recent studies have reported the heterogeneity and diverse function of beta cells using scRNA-seq. For instance, a novel beta cell subpopulation with stem cell properties and immature characteristics could defend itself against attack from immune cells, surviving type 1 diabetes progression for a long time [16]. Human beta cells exist in three different states according to the activation of unfolded protein response (UPR) and insulin gene expression. Proliferating human beta cells were characterised by low insulin gene expression and high UPR activation [17]. Wang et al also found that the proportions of three beta cell subpopulations in type 2 diabetes varied with age and BMI [18]. In summary, functionally distinct beta cell subsets act independently or collaboratively to satisfy dynamic insulin requirements.

Although the introduction of scRNA-seq has achieved promising findings, these results are restricted to the transcriptome level. A comprehensive protein expression profile is essential in the field of islet cell subpopulations. However, traditional proteomics of islet cell subsets is challenged by low yields of FACS-sorted cell subpopulations due to limited islet cells in mouse pancreas. Thus, there is a great demand for the single-cell proteomics, which can sense and quantify more than 1000 proteins in a single mammalian cell [19]. The single-cell protein profile of circulating human CD34⁺ cells obtained from an efficient and straightforward mass-adaptive coating-assisted single-cell proteomics (Mad-CASP) revolutionises our knowledge of different CD34⁺ cell subsets [20]. Yet, single-cell proteomics studies of islet cell subpopulations are still lacking. Moreover, as previous studies mainly focused on beta cells, the heterogeneity of alpha and delta cells is not clear at present. Recent studies have shown diverse electrophysiological response of alpha cells to glucose, suggesting the existence of different alpha cell subsets [21].

In the present study, to comprehensively evaluate the adaptive changes and compensatory mechanisms, we performed scRNA-seq and single-cell proteomics to identify alpha, beta and delta cell subpopulations in a high-fat diet (HFD)-induced mouse model of glucose intolerance and provided a theoretical basis for the precise treatment and prevention of type 2 diabetes.

Methods

Mice and treatment

Four-week-old male C57BL/6J mice (GemPharmatech, China) were bred at specific-pathogen-free conditions under a 12 h/12 h light/dark cycle and randomised into groups with free access to the chow diet (CD; 13.5% fat; 5001; LabDiet, USA) or HFD (60% fat; D12492; Research Diets, USA) for 8 weeks. Bodyweight was monitored every week. Blinding was not carried out in this study. All experimental animal procedures were approved by The Medicine Animal Care Committee of Nanjing Medical University (IACUC1804001).

Glucose and insulin tolerance tests

After 8 weeks’ feeding, mice were subjected to IPGTT and ITT as described in the electronic supplementary material (ESM) Methods.

Isolation of primary islets

Human islets and pancreatic sections were provided from Tianjin First Central Hospital, China (see ESM Methods and Human Islets Checklist). All human studies were conducted according to the principles of the Declaration of Helsinki and approved by the Research Ethics Committee of Nanjing Medical University. Mouse islets were isolated as described in the ESM Methods.

Immunofluorescence staining

Immunofluorescence was performed on the paraffin-embedded pancreas sections. Pancreases were fixed, processed, stained and quantified as previously reported [22]. See ESM Methods for further details.

FACS of islet cells

FACS was used to obtain alpha, beta and delta cell subgroups. Dispersed mouse or human islets were sorted and analysed as described in the ESM Methods. Antibodies are shown in ESM Table 1.

Single-cell library preparation, RNA sequencing and dataset analysis

Islet cells from the HFD- and CD-fed mice were dispersed as single cells and scRNA-seq was performed. The reads were processed as described previously [23]. See ESM Methods for further details.

Identification of cell types and alpha, beta and delta cell subpopulations and differentially expressed gene analysis

The data were scaled and principal component analysis (PCA) was based on the top 2000 highly variable, differentially expressed genes (DEGs), as described in the ESM Methods.

Sample preparation for single-cell proteomics of islet cells

Single-cell proteomics samples were prepared based on mass-adaptive coating of synthetic peptides as previously reported [20]. See ESM Methods for further details.

LC-MS/MS parameters

The peptide samples were subjected to LC-MS/MS using a Fusio Lumo Tribrid mass spectrometer (Thermo Scientific, USA) as previously reported [20]. See ESM Methods for further details.

Single-cell proteomics data analysis

The raw data were processed by MaxQuant (version 1.5.3.0) and fold-changes of differentially expressed proteins (DEPs) were assessed using Welch’s t tests. Single-cell proteomics samples with less than 5000 identified peptides were excluded from data analysis. See ESM Methods for further details.

Quantitative PCR

Sorted primary islet cells were lysed in TRIzol Reagent (Invitrogen, catalogue no. 15596026, USA). RNA and protein were isolated according to the manufacturers’ instructions. Quantitative PCR (qPCR) was performed as described in the ESM Methods. The primers are presented in ESM Table 2.

Western blotting

Proteins from sorted primary islet cells were isolated using TRIzol Reagent. Western blots were performed as described in the ESM Methods.

Statistical analysis

FACS data were analysed using FlowJo software (version 10.0.7, BD Biosciences, USA). Comparisons of data between two groups were carried out using two-tailed t tests. The data are presented as mean ± SD. A p value<0.05 was considered statistically significant.

Results

ScRNA-seq of mouse islet cells

An 8 week HFD induced obvious obesity, impaired glucose homeostasis, and induced insulin resistance in C57BL/6J mice (ESM Fig. 1a). Consistent with our previous research [24], HFD resulted in a compensatory hyperplasia in islet mass, especially regarding beta cells (ESM Fig. 1b). High-throughput scRNA-seq profiled an estimated 15093 cells from islets from CD-fed mice and 18906 from islets from HFD-fed mice and clustered into four principal endocrine cells, annotated based on predominant hormone gene expression: 4069 alpha (Gcg), 14,108 beta (Ins1), 1698 delta (Sst) and 666 PP (Ppy) cells. In addition, small numbers of endothelial, immune, mesenchymal and ductal cells were identified based on previously reported marker genes (Fig. 1a–d).

Fig. 1.

Identification of islet cell types. (a) UMAP visualisation (in two dimensions [UMAP_1 and UMAP_2]) of 15,093 islet cells from 5 CD-fed mice and 18,906 islet cells from 5 HFD-fed mice (different colours mark distinct cell types annotated based on the expression of signature genes; each dot represents one cell). (b) Distribution of each cell subpopulation in CD and HFD groups. (c, d) UMAP visualisation (c) and violin plots (d) of the expression of marker genes in islet cell types (depth of purple in c indicates expression level). (e–g) Top 20 enriched KEGG pathways of DEGs in (e) alpha, (f) beta and (g) delta cells between HFD and CD groups. TPM, transcripts per million; UMAP, uniform manifold approximation and projection

Based on the DEGs in alpha, beta and delta cells in the HFD group compared with the CD group (ESM Fig. 1c–e, ESM Table 3), protein processing in the endoplasmic reticulum was significantly enriched in alpha, beta and delta cells (Fig. 1e–g). Inflammation-related pathways (such as Th17 cell differentiation and IL-17 [25] and the NOD-like receptor signalling pathways [26]) were particularly enriched in alpha cells (Fig. 1e), suggesting the importance of inflammation responses to HFD-induced metabolic stress. In beta cells, calcium transport and ion channel regulation pathways (including arrhythmogenic right ventricular cardiomyopathy, hypertrophic cardiomyopathy, dilated cardiomyopathy, and the cGMP–PKG signalling pathway) were top-ranked (Fig. 1f), indicating the relevance of secretory function. Additionally, in both beta and delta cells, oxidative respiratory chain and energy production and metabolism pathways were highly significantly enriched. In delta cells, other notable pathways that were markedly influenced by HFD were protein degradation and recycling-related pathways (Fig. 1g).

Ace2^low alpha cells may represent immature alpha cells with beta cell-like characteristics

To analyse the heterogeneity of alpha cells, alpha cells were clustered into eight clusters using scRNA-seq data (Fig. 2a,b). To characterise them, we analysed the marker gene expression in the eight clusters (Fig. 2c). Interestingly, we found that Ace2, a pivotal gene associated with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, varied noticeably among the alpha cell clusters. According to the expression level of Ace2, we divided the eight alpha cell clusters into Ace2^low (clusters 3 and 5) and Ace2^high (clusters 0, 1, 2, 4, 6, and 7) subpopulations (Fig. 2d,e). ScRNA-seq and immunofluorescence confirmed the existence of alpha cell subgroups in both CD and HFD mouse islets (Fig. 2f,g,j). Heterogeneous expression of ACE2 also exists but ACE2 was only expressed in a small number of human alpha cells in both healthy and type 2 diabetic individuals (ESM Fig. 2).

Fig. 2.

Heterogeneous expression of Ace2 in alpha cells. (a, b) UMAP visualisation of alpha cell subpopulations (a) and violin plot of Ace2 expression distribution in these subpopulations (b). (c) Heatmap of cell marker expression in alpha cell subpopulations. (d, e) UMAP visualisation of Ace2^high and Ace2^low alpha cell subpopulations (d) and violin plot of Ace2 expression distribution in these subpopulations (e). (f, g) UMAP visualisation of alpha cells in the CD and HFD groups (f) and violin plot of Ace2 expression distribution in Ace2^high and Ace2^low alpha cell subpopulations in the CD and HFD groups (g). (h) Heatmap of differential gene expression between Ace2^low and Ace2^high groups. (i) Enriched KEGG pathways of DEGs in (h). (j) Glucagon (green) and ACE2 (red) co-staining of pancreatic sections (white arrows indicate ACE2^low alpha cells, yellow arrows indicate ACE2^high alpha cells) (scale bar, 20 μm). (k) Representative western blot of ACE2 and the beta cell function-related proteins UCN3 and GLUT2 from sorted ACE2^high and ACE2^low alpha cells. (l) Violin plots of key genes associated with alpha cell maturation and function: Mafb, Irx1, Fev, Ttr, Chgb and Pyy were upregulated in the Ace2^high group, while Ucn3, Ero1lb and Slc2a2 were upregulated in the Ace2^low group. (m–q) qPCR analysis of Ace2, alpha cell identity-related gene Mafb, and beta cell identity-related genes Mafa, Ucn3 and Ero1lb in sorted ACE2^high and ACE2^low alpha (α) cells isolated from C57BL/6J mice (per RNA sample from n=10 mice, ≥5 replicates). *p<0.05; **p<0.01; ***p<0.001; Student’s t test). TPM, transcripts per million; UMAP, uniform manifold approximation and projection

To explore the molecular features of Ace2-based alpha cell heterogeneity, we identified 379 DEGs between the Ace2^low and Ace2^high subpopulations (Fig. 2h, ESM Table 4). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed that the DEGs between the Ace2^low and Ace2^high subpopulations were preferentially enriched in oxidative phosphorylation, metabolic pathways, alpha cell secretory function-related pathways (including cGMP–PKG, glucagon, calcium and cAMP signalling pathways) and programmed cell death and proliferation pathways (including ferroptosis, oocyte meiosis, mammalian target of rapamycin [mTOR] signalling, Rap1 signalling and phosphoinositide 3-kinase [PI3K]–Akt signalling) (Fig. 2i, ESM Table 4), indicating that Ace2 expression could distinguish between alpha cells with different functional and survival status. We also found that alpha cell-specific transcription factors (Mafb and Irx1) and alpha cell-related genes (Ttr, Chgb, Fev and Pyy) were significantly downregulated in the Ace2^low group. Instead, Ace2^low alpha cells started to express genes with beta cell characteristics (such as Ucn3, Ero1lb [also known as Ero1b], and Slc2a2) (Fig. 2l). Some of the results were confirmed by qPCR (Fig. 2m–q) and western blotting (Fig. 2k) on the sorted alpha cell populations. It suggested that the Ace2^low subpopulation may represents immature alpha cells that appear to have beta cell-like characteristics.

Single-cell proteomics analysis of islet cell subpopulations

As mRNA expression does not fully reflect the biological activity of cells or organs, we performed single-cell proteomics analyses of islet cell subpopulations (ESM Fig. 3a). A total of 6955 proteins and 50245 peptides were identified in 36 islet cell subpopulation samples. The CVs of peptides and proteins within the groups were small and stable, and the correlation coefficients between replicates were all >0.7, indicating the excellent quality of the single-cell proteomics data (ESM Fig. 3b–d). A detailed heatmap and enriched KEGG pathways of DEPs for each cell subpopulation are shown in ESM Fig. 3e, f and ESM Tables 5 and 6.

ACE2^low alpha cells may represent a state of low-capacity and fragility and susceptibility to HFD stress

FACS of islet single-cell suspensions was used to divide alpha (glucagon [GCG]⁺) cells into ACE2^high and ACE2^low groups. Notably, the proportion of ACE2^high alpha cells was significantly lower in HFD mice (65.3%) than CD mice (80.0%) (Fig. 3a,b), indicating that ACE2-based alpha cell heterogeneity may be involved in the pathophysiology of diabetes. FACS-sorted ACE2^low and ACE2^high alpha cell subpopulations from CD and HFD mice (2000 cells per sample) were collected for single-cell proteomics analysis. The number of proteins detected varied from 3042 to 4021 and with good reproducibility (ESM Fig. 4a). PCA could also clearly differentiate ACE2^low and ACE2^high cells (Fig. 3c).

Fig. 3.

Single-cell proteomics indicating alpha cell subpopulation function. (a) Representative dot plots of the flow cytometry gating strategies for alpha cells, with respective (FMO) control. GCG⁺ alpha cells were sorted into ACE2^low and ACE2^high subpopulations. Representative distributions of ACE2 expression in CD and HFD groups are shown on the right. (b) Proportions of ACE2^low and ACE2^high alpha cell subpopulations in CD and HFD groups (n≥8 mice per group). Student’s t test: *p<0.05. (c) PCA plots of ACE2^low and ACE2^high single-cell proteomics samples (n=10 CD and 10 HFD mice per pair of ACE2^low and ACE2^high alpha cell single-cell proteomics samples, 3 replicates). (d, f) Volcano plots of DEPs between ACE2^low and ACE2^high subpopulations in (d) CD and (f) HFD groups. (e, g) Enriched KEGG pathways of DEPs between ACE2^low and ACE2^high subpopulations in (e) CD and (g) HFD groups. The top x-axis refers to the DEP number and the bottom x-axis refers to the p value. FMO, fluorescence minus one; GCG, glucagon; PC, principal component

In CD mice, there were 29 up- and 80 downregulated DEPs in ACE2^low compared with ACE2^high alpha cells (Fig. 3d, ESM Table 7). The KEGG pathways affected included the Ras signalling pathway and the mitogen-activated protein kinase (MAPK) signalling pathway (Fig. 3e, ESM Table 7). In HFD mice, there was a clear increase in DEPs in ACE2^low compared with ACE2^high alpha cells, with 17 up- and 301 downregulated DEPs (Fig. 3f). Pathways associated with nutrient and energy metabolism (including metabolic pathways, carbon metabolism, glycolysis/gluconeogenesis, pentose phosphate pathway, biosynthesis of amino acids, fructose and mannose metabolism, glutathione metabolism, starch and sucrose metabolism, and galactose metabolism) and protein degradation and recycling pathways (including endocytosis, protein processing in endoplasmic reticulum, proteasome, and lysosome pathways) were all markedly downregulated in the ACE2^low group (Fig. 3g). Furthermore, cell proliferation and survival pathways (including Rap1, oocyte meiosis, neurotrophin and Hippo signalling pathway) were also downregulated. These findings strongly hinted that the metabolic and other biological activities of ACE2^low alpha cells were diminished, especially in HFD mice.

Further, we analysed the DEPs and enriched KEGG pathways between the HFD and CD groups in both ACE2^low and ACE2^high alpha cells. There were more downregulated DEPs in ACE2^low alpha cells than ACE2^high alpha cells, and these downregulated DEPs were enriched in more metabolism-related pathways (ESM Fig. 4b–e, ESM Table 7). This suggested that ACE2^low alpha cells are more susceptible to HFD stress and could be a group of low-capacity and fragile alpha cells, while ACE2^high alpha cells may represent the main functional cell population in mice, and they could better withstand HFD stress.

Cd81^high beta cells may represent immature beta cells with weak function

To analyse the heterogeneity of beta cells, we clustered the beta cells into nine clusters and identified the top ten marker genes (Fig. 4a–c). We found that Cd81 was differentially expressed in beta cell clusters, in line with a recent study [27]. We divided the nine beta cell clusters into Cd81^low (clusters 0–4 and 7) and Cd81^high (clusters 5, 6 and 8) subpopulations (Fig. 4d–g). Immunostaining also confirmed the obvious heterogeneity of CD81 distribution in beta cells from both CD and HFD mice (Fig. 4j), which was verified in human islets (ESM Fig. 5). What’s more, the proportion of CD81^high cells was elevated in the type 2 diabetes group (ESM Fig. 5).

Fig. 4.

Heterogeneous expression of CD81 on beta cells. (a, b) UMAP visualisation of beta cell subpopulations (a) and violin plot of Cd81 expression distribution in these subpopulations (b). (c) Heatmap of cell marker expression in beta cell subpopulations. (d, e) UMAP visualisation of Cd81^high and Cd81^low beta cell subpopulations (d) and violin plot of Cd81 expression distribution in these subpopulations (e). (f, g) UMAP visualisation of beta cells in the CD and HFD groups (f) and violin plot of Cd81 expression distribution in Cd81^high and Cd81^low beta cell subpopulations in the CD and HFD groups (g). (h) Heatmap of differential gene expression between Cd81^low and Cd81^high groups. (i) Enriched KEGG pathways of DEGs between Cd81^low and Cd81^high groups. (j) Insulin (green) and CD81 (red) co-staining of pancreatic sections (white arrows indicate CD81^low beta cells, yellow arrows indicate CD81^high beta cells) (scale bar, 40 μm). (k) Representative western blot of CD81 and the beta cell-related protein GLUT2 from sorted CD81^low and CD81^high beta cells isolated from CD and HFD C57BL/6J mice. (l) Violin plots of key genes associated with beta cell maturation and function; Mafa, Ucn3, Slc2a2, G6pc2, Sytl4, and Ero1lb were upregulated in the Cd81^low group, and Mafb, Rbp4, Chga and Chgb were upregulated in the Cd81^high group. (m–p) qPCR analysis of Cd81, the alpha cell identity-related gene Mafb, and beta cell identity-related genes Mafa and Ucn3 in sorted CD81^low and CD81^high beta (β) cells isolated from CD and HFD C57BL/6J mice (per pair of RNA samples from n=5 mice, ≥5 replicates, *p<0.05; ***p<0.001; Student’s t test). TPM, transcripts per million; UMAP, uniform manifold approximation and projection

According to the DEGs and enriched KEGG pathways in the Cd81^low subpopulation relative to the Cd81^high subpopulation (Fig. 4h,i, ESM Table 4), cellular metabolism and related functional pathways along with a series of stress-related and inflammatory pathways were significantly enriched. The key beta cell transcription factor Mafa and beta cell maturation markers (Ucn3, Slc2a2, G6pc2, Sytl4, and Ero1lb) had high mRNA expression in the Cd81^low subpopulation, whereas immature beta cell markers (Chga, Chgb, Mafb, and Rbp4) were highly expressed in the Cd81^high subpopulation (Fig. 4l), suggesting that the Cd81^high subpopulation may represent a group of weakly functioning and immature beta cells. Some of the results were confirmed by qPCR (Fig. 4m–p) and western blotting (Fig. 4k) on the sorted beta cell subpopulations.

To further explore the effects of HFD on beta cell subpopulations, FACS showed that the proportion of CD81^high beta cells decreased in HFD mice compared with CD mice (Fig. 5a,b). Then, FACS-sorted CD81^low and CD81^high beta cells were analysed using single-cell proteomics (10,000 cells per sample). The number of proteins detected varied from 3435 to 4946 and with good reproducibility (ESM Fig. 6a). PCA showed that CD81^low and CD81^high beta cells clustered into two separate groups (Fig. 5c).

Fig. 5.

Single-cell proteomics identification of beta cell subpopulation functions. (a) Representative dot plots of the flow cytometry gating strategies for beta cells, with respective FMO control. INS⁺ beta cells were sorted into CD81^low and CD81^high subpopulations. Representative CD81 expression distributions in CD and HFD groups are shown on the right. (b) Proportions of CD81^low and CD81^high beta cell subpopulations in CD and HFD groups (n≥8 mice per group). Student’s t test: *p<0.05. (c) PCA plots of CD81^low and CD81^high single-cell proteomics samples (per pair of CD81^low and CD81^high beta cell single-cell proteomics sample from n=5 CD or 5 HFD mice, 3 replicates). (d, e) Volcano plots of DEPs between CD81^low and CD81^high subpopulations in the (d) CD and (e) HFD groups. (f, g) Enriched KEGG pathways of DEPs between CD81^low and CD81^high subpopulations in the (f) CD and (g) HFD groups. The top x-axis refers to DEP number and the bottom x-axis refers to p value. (h) CD81 (green), PDX1 (red) and insulin (white) co-staining of pancreatic sections (white arrows indicate CD81^low PDX1^high beta cells, yellow arrows indicate CD81^high PDX1^low beta cells) (scale bar, 20 μm). (i) Representative western blot of PDX1 from sorted CD81^high and CD81^low beta cells isolated from CD and HFD C57BL/6J mice was shown on the right. FMO, fluorescence minus one; INS, insulin; PC, principal component

Comparing CD81^low with CD81^high beta cells, 869 up- and 373 downregulated DEPs were detected in CD mice, and 707 up- and 312 downregulated DEPs were detected in HFD mice (Fig. 5d,e, ESM Table 8). In both CD and HFD mice, pathways associated with energy metabolism (including oxidative phosphorylation and Alzheimer’s disease) were downregulated in the CD81^low subpopulation (Fig. 5f,g). The downregulated DEPs involved in these pathways were also decreased in scRNA-seq between HFD and CD beta cells (ESM Table 3). The results reflect the consistency of the transcriptome and proteome, confirming the reliability of our single-cell RNA-seq and proteomics. Furthermore, in both CD and HFD mice, pancreatic and duodenal homeobox 1 (PDX1), a key transcription factor associated with beta cell maturation and functional maintenance, was significantly upregulated in CD81^low cells relative to CD81^high cells (ESM Table 8), the expression was further validated by immunofluorescence (Fig. 5h) and western blot analyses (Fig. 5i). Additionally, DEPs enriched in endocytosis, phagosomes, proteasomes and synaptic vesicle cycling pathways were mostly upregulated in CD81^low beta cells (Fig. 5f,g), suggesting a more mature status of CD81^low beta cells.

Notably, there were 324 DEPs between the HFD and CD mice in CD81^low cells and 202 DEPs between the HFD and CD mice in CD81^high cells (ESM Fig. 6b, d, ESM Table 8), which is much fewer than the DEPs between the CD81^low and CD81^high cell subpopulations (1242 DEPs in CD mice, and 1019 in HFD mice). However, the enriched pathways all seem to have no obvious relationships with beta cell maturation or functional status in either CD81^low or CD81^high cells (ESM Fig. 6c, e). This suggests that the effect of HFD on CD81-divided beta cell subpopulations is to change the subpopulation proportion rather than a direct effect on the subpopulation function.

HFD enhanced the function of mature GLUT2^low delta cells

Although Slc2a2 (encoding GLUT2) is a beta cell signature gene, we still found a small group of delta cells with relatively high expression of Slc2a2 in scRNA-seq. Therefore, we divided the delta cells into Slc2a2^low and Slc2a2^high subpopulations (ESM Fig. 7a, b). According to the DEGs and enriched pathways in the Slc2a2^low vs Slc2a2^high subpopulations (ESM Fig. 7c, d), upregulated pathways were related to calcium ion and hormone secretion, suggesting that the Slc2a2^low subgroup may be more functionally mature. However, the 8 week HFD did not alter the proportions of the two delta cell subpopulations (Fig. 6a,c). Immunofluorescence confirmed the existence of delta cell subgroups in both mouse and human islets (Fig. 6b and ESM Fig. 8).

Fig. 6.

Single-cell proteomics identification of delta cell subpopulation functions. (a) Representative dot plots of the flow cytometry gating strategies for delta cells, with respective FMO control. SST⁺ delta cells were sorted into GLUT2^low and GLUT2^high subpopulations. Representative GLUT2 expression distributions in CD and HFD groups are shown on the right. (b) Somatostatin (green) and GLUT2 (red) co-staining of pancreatic sections (white arrows indicate GLUT2^low delta cells, yellow arrows indicate GLUT2^high delta cells) (scale bar, 20 μm). (c) Proportions of GLUT2^low and GLUT2^high delta cell subpopulations in CD and HFD groups (n≥6 mice per group). (d) PCA plots of GLUT2^low and GLUT2^high single-cell proteomics samples. (e, f) Volcano plots of DEPs in GLUT2^low relative to GLUT2^high cells in (e) CD and (f) HFD groups (per pair of GLUT2^low and GLUT2^high delta cell single-cell proteomics samples from n=10 CD and 10 HFD mice, 3 replicates). (g, h) Enriched KEGG pathways in (g) CD and (h) HFD groups. The top x-axis refers to DEP number and the bottom x-axis refers to p value. FMO, fluorescence minus one; PC, principal component; SST, somatostatin

FACS-sorted GLUT2^low and GLUT2^high delta cell subpopulations from CD and HFD mice (3000 cells per sample) were used for single-cell proteomics analysis, with high reproducibility among the three replicates in each group (ESM Fig. 7e). PCA data showed that GLUT2^low and GLUT2^high samples clustered into two distinct populations (Fig. 6d).

In the CD mice, there were 548 up- and 171 downregulated DEPs between the GLUT2^low and GLUT2^high subpopulations; meanwhile, there were more upregulated (832) and fewer downregulated (102) DEPs in the HFD mice (Fig. 6e,f, ESM Table 9). KEGG analysis revealed that similar to beta cells, oxidative phosphorylation and Alzheimer’s disease pathways were significantly downregulated in the GLUT2^low cells (relative to the GLUT2^high cells) in the CD mice (Fig. 6g). Multiple nutrient metabolic pathways were also significantly upregulated in the GLUT2^low cells in both CD and HFD mice (Fig. 6g,h). The same was true for endocytosis and protein processing in the endoplasmic reticulum-related pathways, indicating a more active metabolism in the GLUT2^low subpopulation.

The number of DEPs between HFD and CD mice in GLUT2^low (117) and GLUT2^high (177) subpopulations was much lower than between GLUT2^low and GLUT2^high subpopulations in HFD (934) and CD (719) mice (ESM Fig. 7f, g, ESM Table 9). In GLUT2^low delta cells, comparing HFD and CD mice, upregulated pathways such as AMPK, cGMP–PKG and mTOR signalling pathways dominated. In addition, the calcium signalling pathway, pancreatic secretion, and insulin and glucagon signalling pathways were significantly upregulated (ESM Fig. 7h), indicating that HFD promoted energy metabolism in GLUT2^low delta cells. Meanwhile, some metabolic synthesis pathways were downregulated in GLUT2^high delta cells (HFD vs CD) (ESM Fig. 7i). In conclusion, GLUT2^low delta cells may be more functional and metabolically active than GLUT2^high delta cells, and HFD further stimulates GLUT2^low delta cell function and metabolism.

Discussion

The development of scRNA-seq technology and the discovery of new cell markers deepened the understanding of beta cell heterogeneity in health and disease [28, 29]. However, no surface markers have been reported to distinguish mature alpha cells. Therefore, the identification of the surface marker ACE2 is meaningful because this protein not only marks immature alpha cells but can also be used to isolate living alpha cell subtypes. Furthermore, ACE2 is widely expressed in several organs, but its expression in pancreatic islets is controversial [30–33]. We found differential distribution of ACE2 in human and mouse islets. In mouse islets, ACE2 is highly expressed in alpha cells but less in beta cells and exocrine cells, whereas in human islets, it is predominantly expressed in the core of the islet, with only a few positive alpha cells.

A recent study found that alpha cell dysfunction in type 2 diabetes is associated with a ‘beta cell-like’ electrophysiological signature, which was linked to cell maturation state and impaired alpha cell identity [34]. We found some hints of this in our own data: immature alpha cells may be functionally impaired and have beta cell characteristics. Based on this, ACE2 likely marks mature alpha cells as its reduced expression levels are associated with decreased alpha cell function maintenance-related genes and increased beta cell maturation-related genes [35–37]. Consistent with Xuan et al, who showed that beta cell dedifferentiation occurred in response to metabolic stress and was accompanied by ACE2 reduction [38], we also found an increased proportion of the ACE2^low subpopulation after HFD.

We found that ACE2^low alpha cells have significantly reduced metabolism and other biological activity and are less able to resist HFD stress, that is, they may be low-capacity and fragile alpha cells. In contrast, ACE2^high alpha cells may be the major functional cell population in mice and are more able to withstand HFD stress. The increase in the proportion of the ACE2^low subpopulation after an 8 week HFD suggests that HFD decreases alpha cell function, which corresponds to the scRNA-seq result that HFD alpha cells are enriched in inflammation-related pathways. We speculated that the phenomenon might be due to the elevated insulin demand caused by insulin resistance in obesity, as alpha cells achieve glucose homeostasis by reducing glucagon secretion.

As for beta cell subpopulation markers, there are similarities as well as differences between our results and those of Salinno et al. The analogous conclusion is that CD81 marks immature beta cells in both Salinno et al [27] and our study. This also shows the reliability of the method and the stability of the CD81 marker. The difference in our study is that we used a high-fat mouse model, which is closer to the natural history of glucose intolerance, and found a decreased proportion of CD81^high cells after 8 weeks of HFD feeding. Salinno et al [27] found an elevated proportion of CD81^high cells in the islets of streptozotocin (STZ)-treated mice and db/db mice. We hypothesise that the HFD-induced glucose-intolerant islets were in a compensatory proliferative state and required more mature CD81^low beta cells to meet the body’s insulin requirements. Recent research has also found that CD81 deficiency leads to obesity, insulin resistance and adipose tissue inflammation [39]. Nevertheless, the STZ and db/db mouse models simulate severe hyperglycaemia in diabetes in which islet function fails to compensate, so the proportion of the immature CD81^high cells increases, which could also explain the increased proportion of CD81^high cells that we observed in islets from donors with type 2 diabetes.

Our study is unique in that we found that most of the metabolic and exocytosis pathways related to beta cell maturation [40] were upregulated in CD81^low beta cells, with some difference in DEPs (such as PDX1 [41]), through single-cell proteomics, findings which are not replicated in scRNA-seq. Since the protein is the executor of biological function, these could suggest that CD81^low beta cells are functionally mature. We also found that the effect of HFD on CD81-divided beta cell subpopulations is to change the subpopulation proportion rather than directly affecting subpopulation function, explaining the influence of metabolic stress on beta cell subsets.

GLUT2 mediates glucose and fructose transport and is one of the beta cell glucose-sensing mechanisms [42, 43]. We found that GLUT2^low delta cells are the main delta cells found in islets that have improved function and are more metabolically active than GLUT2^high delta cells. HFD further stimulates GLUT2^low delta cell function and metabolism. This is consistent with the basal hypersecretion of somatostatin in HFD-fed mouse islets in our previous experiments [24]. In GLUT2^high cells, according to single-cell proteomics, insulin/INS2 expression was upregulated, accompanied by downregulated nutrient metabolic pathways, which may be indicative of immature delta cells with partial beta cell characteristics. Since HFD did not alter the delta cell subpopulation ratio, the effect of HFD on delta cells mainly involved enhancing the function of the major GLUT2^low subpopulation.

Several limitations of our study should be recognised. Although the subpopulation markers we identified are all membrane proteins, permeabilisation is still required to stain islet hormones, and functional validation of live islet subpopulations is still needed. Moreover, although we demonstrated the existence of subpopulations in human islets, it will be interesting to explore whether they are involved in the progression of type 2 diabetes, as well as the functional consistency between human and murine islets.

In conclusion, we found that immature islet cells partly lost their identity maintenance genes and began to show the characteristics of other endocrine cells. In HFD-induced glucose intolerance, the proportion of low-capacity and immature ACE2^low alpha cells increased with the increasing ratio of functionally active and mature CD81^low beta cells. The mature GLUT2^low delta cells were functionally enhanced but in constant proportions. These dynamic changes reveal the metabolic plasticity of pancreatic islets in response to HFD stress. This plastic adaptation can counteract metabolic stress and restore glucose homeostasis by inhibiting the hyperglycaemic effect of alpha cells, enhancing the hypoglycaemic effect of beta cells, and increasing the regulation of alpha and beta cells by delta cells (Fig. 7). Therefore, targeting the above markers to change the distribution of islet subpopulations in order to strengthen islet function may be a useful approach to treating diabetes.

Fig. 7.

Islet cell (alpha, beta and delta) subpopulations in mice with HFD-induced islet dysfunction. In HFD-induced glucose intolerance, the proportion of low-capacity and immature ACE2^low alpha cells increased at the same time as the ratio of functionally active and mature CD81^low beta cells increased. The mature GLUT2^low delta cells were functionally enhanced but in constant proportions. These dynamic changes reveal the metabolic plasticity of pancreatic islets in response to HFD stress. This plastic adaptation can counteract metabolic stress and restore glucose homeostasis by inhibiting the hyperglycaemic effect of alpha cells, enhancing the hypoglycaemic effect of beta cells, and increasing the regulation of alpha and beta cells by delta cells. The figure was created with BioRender.com

Abbreviations

CD

Chow diet

DEG

Differentially expressed gene

DEP

Differentially expressed protein

HFD

High-fat diet

KEGG

Kyoto Encyclopedia of Genes and Genomes

mTOR

Mammalian target of rapamycin

PCA

Principal component analysis

PDX1

Pancreatic and duodenal homeobox 1

PP

Pancreatic polypeptide

qPCR

Quantitative PCR

scRNA-seq

Single-cell RNA sequencing

UPR

Unfolded protein response

Funding

This work was supported by the National Natural Science Foundation of China (81830023, 82070803, 82100838 and 82100837).

Data availability

The scRNA-seq analysis datasets during the current study are available in the Gene Expression Omnibus (GEO) repository under the accession number GSE203376 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE203376). Other datasets are available from the corresponding authors upon reasonable request.