Calorie restriction increases insulin sensitivity to promote beta cell homeostasis and longevity in mice

Abstract

Caloric restriction (CR) can extend the organism life- and health-span by improving glucose homeostasis. How CR affects the structure-function of pancreatic beta cells remains unknown. We used single nucleus transcriptomics to show that CR increases the expression of genes for beta cell identity, protein processing, and organelle homeostasis. Gene regulatory network analysis reveal that CR activates transcription factors important for beta cell identity and homeostasis, while imaging metabolomics demonstrates that beta cells upon CR are more energetically competent. In fact, high-resolution microscopy show that CR reduces beta cell mitophagy to increase mitochondria mass and the potential for ATP generation. However, CR beta cells have impaired adaptive proliferation in response to high fat diet feeding. Finally, we show that long-term CR delays the onset of beta cell aging hallmarks and promotes cell longevity by reducing beta cell turnover. Therefore, CR could be a feasible approach to preserve compromised beta cell structure-function during aging and diabetes.

Fig. 1. Short-term caloric restriction (CR) improves glucose homeostasis in male mice.

A Schematic diagram of mice subjected to ad libitum (AL), 20% CR, or HFD for 2 months starting at 8 weeks of age. B Mouse body mass over 2 months on AL, HFD, or CR diet. C Body mass change after 2 months on diet. D Ratio between fat mass and lean mass after 2 months on diet AL, HFD, or CR diet. E Blood glucose levels during the meal tolerance test (MTT) after 2 months on diet and respective area under-curve (AUC) measurements. F Insulin levels during the MTT and G the respective fold change from baseline values. H Ratio between the insulin and glucose values obtained during the MTT. I HOMA-IR calculated from fasting glucose and insulin values after 2 months on diet. J Blood glucose values obtained during an intraperitoneal insulin tolerance test (ipITT) and the respective decay of the glucose rate per minute (kITT). K Mouse body mass after 2 months on AL, CR, or 20% diluted diet (DL). L Ratio between fat mass and lean mass of AL, CR, or DL mice. M, N Blood glucose levels and circulating insulin levels of AL, CR, and DL mice during a meal tolerance test (MTT) after 2 months on diet. Each dot represents the mean measurement for individual mice. B, C n = 60 AL, n = 75 CR, and n = 20 HF mice per group; D n = 7 mice per diet group; E–I n = 9 AL, n = 11 CR, and n = 5 HF mice per group; J n = 5 mice per diet group; K–M n = 10 AL, n = 10 DL, and n = 15 CR mice per group; N n = 8 AL, n = 9 DL, and n = 13 CR mice per group. P values for all significantly different comparison are shown. F, H the asterisks indicate *p < 0.05, **p < 0.01. Statistical analysis was conducted using one-way ANOVA with Tukey’s post-hoc test C, E, I, or followed by Dunnett’s post-hoc test (K and N) or Benjamini, Krieger, and Yekutieli’s post-hoc test (L). Data on (F–H, M) were analyzed with two-way ANOVA with Sidak’s post-hoc test or Tukey’s post-hoc test. For two groups comparison, unpaired two-tailed Student’s t test was performed (D, J). All data presented as mean ± 95% confidence intervals (C.l.). A was created with BioRender.com. and released under a Creative Commons Attribution-Non-Commercial NoDerivs 4.0 International license.

Fig. 2. CR disrupts beta cell adaptive response during HFD.

A Mouse body mass and body mass change over the course of 4 months, before and after switching diets. Mice were placed on AL or CR diets for 2 months and then shifted to AL (CR-AL) or HFD (AL-HF or CR-HF) for another 2 months. Two groups of AL (AL-AL) or CR (CR-CR) mice were maintained on their original diets as controls. B Ratio between fat mass and lean mass at the end of the diet switch experiment. C, D Blood glucose levels during the meal tolerance test (MTT) and area under-curve (AUC) measurements at the end of the diet switch experiment. E Insulin levels during the MTT shown in (C). All p values for select comparisons are listed at the bottom. F HOMA-IR was calculated from fasting glucose and insulin values at the end of the diet switch experiment. G Bulk-RNA-seq pathway enrichment analysis graph displaying interconnected nodes representing pathways downregulated in CR-HF mice versus AL-HF. H Pancreas beta cell mass in AL-HF versus CR-HF mice at the end of the diet switch experiment. P values for all significantly different comparison are shown, and each dot represents the mean measurement for individual mice. A n = 29 AL-AL, n = 19 AL-HF, n = 24 CR-CR, n = 40 CR-AL, and n = 27 CR-HF mice per group; B n = 17 AL-AL, n = 12 AL-HF, n = 14 CR-CR, n = 15 CR-AL, and n = 17 CR + HF mice per group; C–F n = 5 AL-AL, n = 5 AL-HF, n = 11 CR-CR, n = 12 CR-AL, and n = 14 CR-HF mice per group; H n = 9 mice per group. Statistical analysis was conducted using one-way ANOVA with Tukey’s post-hoc test (B, D), or followed by Benjamini, Krieger, and Yekutieli’s post-hoc test (F). E two-way ANOVA with Benjamini, Krieger, and Yekutieli’s post-hoc test was performed. For two groups comparison, unpaired two-tailed Student’s t test was performed (H). All data presented as mean ± 95% confidence intervals (C.l.).

Fig. 3. Single-cell multiome sequencing of islets reveals diet-specific changes to beta-cell heterogeneity.

A Schematic representation of the workflow used for single nuclei multiome sequencing (ATAC + RNA) of pancreatic islets isolated from AL, CR and HFD mice after 2 months on diet. B Uniform Manifold Approximation and Projection (UMAP) of the integrated transcriptome and chromatin dataset. Data from a total of n = 18,741 islet cells from n = 2 mice per diet group. Different cell types are indicated by various colors and labels. Inset, donut plot with the total number of cells in each cell cluster. C UMAPs show the expression of cell marker genes. D Dot plot with relative expression of marker genes in alpha, beta, delta, gamma, acinar, endothelial, stellate, and macrophage cell clusters. E Genome tracks showing ATAC peaks for hormone markers: Gcg (alpha cell), Ins1 (beta cell), Sst (delta cell), and Ppy (gamma cell). F Heatmap showing the correlation between gene expression (RNA-seq) and chromatin accessibility (ATAC-seq) across the identified cell types. G Heatmap with the top differentially expressed genes (RNA-seq) and H top ATAC-seq peaks in transcriptional factors (TF) motifs in beta cells from AL, CR, and HFD mice. I Annotated node map with pathway enrichment analysis of differentially expressed genes and TF motifs in beta cells from CR mice. Analysis was performed using Metascape with an FDR < 0.05. J UMAP projection of beta cells and identified subclusters from AL, CR, and HFD mice. Beta cell populations were defined according to distinct transcriptional states. K Top marker genes from each beta-cell state. A was created with BioRender.com. and released under a Creative Commons Attribution-Non-Commercial NoDerivs 4.0 International license.

Fig. 4. CR reprograms beta-cell gene regulatory networks (GRNs).

A SCENIC heatmap with hierarchical clustering analysis of TF activity in beta cells from AL, CR and HFD mice after 2 months on diet. TFs identified as “ON” are shown in black, while TFs identified as “OFF” are in white. B Pearson correlation matrix of n = 272 TF identified in all mouse beta cells (all diet groups together). Boxes highlight clusters of TFs with a high degree of correlation. C Gene regulatory networks (GRNs) formed by TFs identified using SCENIC in beta cells from AL, CR and HFD mice. TFs are shown as pink nodes, while target protein-coding genes are shown in blue. Node size represents the “betweenness centrality” measurements that report the influence of a given TF within a network. D Dot plot and hierarchical clustering showing the gene expression levels of target genes associated with Atf6 and Mafa GRNs from AL, CR, and HFD beta cells. The dot plot scale shows the relative expression level and percentage of cells expressing a target gene.

Fig. 5. CR beta cells are metabolically fit.

A Schematic diagram of imaging mass spectrometry (MALDI-MS) approach to measure metabolite abundance in AL and CR pancreases after 2 months on diet. B Average mass to power (m/z) spectra of all samples combined. C Receiver Operating Characteristic (ROC) analysis of ions enriched in AL versus CR islets. D Representative images of an islet-enriched ion (cholesterol sulfate, m/z: 465.304 Da). E Representative images of AL or CR islet-enriched ions. C Data pooled from n = 3 sections (1 section per mouse) per diet group. D, E scale bar = 1 mm. A was created with BioRender.com. and released under a Creative Commons Attribution-Non-Commercial NoDerivs 4.0 International license.

Fig. 6. Long-term CR delays beta cell aging signatures.

A Schematic diagram of mice subjected to AL or 20% CR for 12 months starting at the age of 8 weeks. B Body mass after 12 months on diet. C Ratio between fat mass and lean mass after 12 months on diet. D Blood glucose levels during the meal tolerance test (MTT) after 12 months on diet. E Blood insulin levels during the MTT and F the respective fold change from baseline values. G Ratio between the insulin and glucose values obtained during the MTT. H HOMA-IR calculated from fasting glucose and insulin values after 12 months on diet. I Representative images of pancreatic sections from AL and CR male mice after 2 months or 12 months on diet stained with 53BP1 and insulin. Right, quantification of DNA damage by 53BP1+ beta cells. Each dot represents the average from each mouse. A total of 127 AL and 200 CR islets were analyzed. Data are normalized by percentage of 53BP1-positive beta cells per total beta-cell number. Scale bar, 50 microns. J Quantification of beta cell nuclear levels of Lamin B1 in pancreatic islets from AL and CR mice after 12 months on diet. A total of 74 AL and 105 CR islets were analyzed. K Representative images of pancreatic sections from AL and CR male mice after 12 months on diet stained with Cdkn2a and Cdkn1a mRNA probes, and insulin. Right, the quantification of incidence of these markers per beta-cell. A total of n = 5570 AL and n = 4781 CR beta cells were analyzed. Scale bar, 5 microns. L Representative images of pancreatic sections from AL and CR mice after 12 months on diet stained with Lc3I-II or Lamp1. Right, quantification of beta cell area occupied by Lc3I-II or Lamp1. Each dot represents the average measurement from individual islets. A total of n = 108 AL 2mo, n = 89 CR 2 mo, n = 51 AL 12mo, and n = 73 CR 12mo islets were analyzed for Lc3I-II staining. A total of n = 23 AL 2mo, n = 23 CR 2 mo, n = 38 AL 12mo, and n = 57 CR 12mo islets were analyzed for Lamp1 staining. Scale bar, 10 microns. B, C n = 18 mice per diet group; D–H n = 15 AL and n = 18 CR mice per group; I n = 5 AL 2mo, n = 5 CR 2mo, n = 10 AL 12mo, and n = 13 CR 12mo mice per group; J–L Data pooled from n = 5 male mice per diet group. P values for all significantly different comparison are shown. Statistical analysis was conducted using two-way ANOVA with Sidak’s post-hoc test E–G, or one-way ANOVA with Benjamini, Krieger and Yekutieli’s post-hoc test (I, L). For two groups comparison, unpaired two-tailed Student’s t test was performed (B, C, H, J, and K). For all panels, data presented as mean ± 95% confidence intervals (C.l.). A was created with BioRender.com. and released under a Creative Commons Attribution-Non-Commercial NoDerivs 4.0 International license.

Fig. 7. CR increases beta cell mitochondria density and modifies cristae structure.

A Representative images of pancreatic sections from AL and CR male mice after 2 months on diet. Slides were stained with Insulin, Lamp1, and Sdha. Blue arrowheads point to regions of overlap between Sdha and Lamp1 signal. B Colocalization between Sdha and Lamp1 was measured. Each dot represents the average colocalization index in each animal calculated from ~100 beta cells per animal. C Representative images of pancreatic beta cells from AL and CR male mice using electronic microscopy. Blue arrows point to mitochondria density, while pink arrows point to lipofuscin granules in aged beta cells. Mitochondrial density is measured by the total mitochondrial number per beta cell area. Each dot represents a single beta cell analyzed from three different islets per mouse (n = 3 mice per diet group). D Representative images of pancreatic beta cells from AL and CR male mice using high-resolution electron tomography (eTomo). Yellow arrowheads indicate MDVs, whereas the asterisk in each image marks the location of lysosomes. E Representative 3D reconstructions of beta cell crista segmentation generated using deep learning image analysis tools. F, G Cristae surface area and cristae density in AL or CR beta cells. H Dot plot showing the relative expression levels of genes involved in cristae formation and morphology. I Calculation of ATP generation for an average beta cell mitochondrion. Each dot represents data from one eTomo image stack. J Relative frequency of mitochondrial-derived vesicles observed in eTomo images. F–H and I, J each dot represents an individual field of view with an average of five mitochondria per field. eTomo data acquired from n = 3 mice per diet group, five fields of view per animal. Total of 69–70 mitochondria analyzed per diet group. The asterisks indicate *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 using unpaired, two-tailed Student’s t test. C, D Scale bar = 500 nm. A Full view panels have a scale bar = 10 microns. Zoom overlay scale bar = 2 microns. E scale bar = 1 micron. All data presented as mean ± 95% confidence intervals (C.l.).

Fig. 8. CR promotes beta-cell longevity by slowing down beta-cell turnover rates.

A Study design. After 2 months of age, ¹⁵N-labeled mice were kept on AL, 20% CR or HFD for 12 months. B Multi-isotope imaging mass spectroscopy (MIMS) representation technique. C Validation of ¹⁵N incorporation by MIMS imaging and quantification of ¹⁵N/¹⁴N ratio in cortical neurons from a ¹⁵N-labeled mouse after chase with ¹⁴N diet for 12 months. Each dot represents a single neuronal nucleus (n = 21 nucleus from 2 cortical layers of 1 mouse) D MIMS of pancreatic islets from AL, CR, or HFD mice (n = 3 islets per animal, n = 1 mouse per diet group). E Single cell analysis of ¹⁵N/¹⁴N ratios in beta cells at day 0 (no chase) and after 12 months (mo) on AL, HFD, or CR diet. Each dot represents a single beta cell nucleus, pooled from 3 different islets of 1 mouse per diet group. Total of n = 54 AL 0mo, n = 217 AL 12mo, n = 114 HFD 12m0, and n = 194 CR 12mo beta cells were analyzed. The dotted horizontal pink and black lines represent the mean and the lowest ¹⁵N/¹⁴N levels found in cortical neurons shown in (C). F Estimation of beta cells that are estimated to be LLC or young after 12 months on AL, CR, or HFD. B, D scale bar = 10 microns. The p values for all significantly different comparison are shown. Statistical analysis was conducted using one-way ANOVA with Benjamini, Krieger, and Yekutieli’s post-hoc test (B). All data presented as mean ± 95% confidence intervals (C.l.). A was created with BioRender.com. and released under a Creative Commons Attribution-Non-Commercial NoDerivs 4.0 International license.