Endocrine islet β-cell subtypes with differential function are derived from biochemically distinct embryonic endocrine islet progenitors that are regulated by maternal nutrients

Abstract

Endocrine islet b cells comprise heterogenous cell subsets. Yet when/how these subsets are produced and how stable they are remain unknown. Addressing these questions is important for preventing/curing diabetes, because lower numbers of b cells with better secretory function is a high risk of this disease. Using combinatorial cell lineage tracing, scRNA-seq, and DNA methylation analysis, we show here that embryonic islet progenitors with distinct gene expression and DNA methylation produce b-cell subtypes of different function and viability in adult mice. The subtype with better function is enriched for genes involved in vesicular production/trafficking, stress response, and Ca²⁺-secretion coupling, which further correspond to differential DNA methylation in putative enhancers of these genes. Maternal overnutrition, a major diabetes risk factor, reduces the proportion of endocrine progenitors of the b-cell subtype with better-function via deregulating DNA methyl transferase 3a. Intriguingly, the gene signature that defines mouse b-cell subtypes can reliably divide human cells into two sub-populations while the proportion of b cells with better-function is reduced in diabetic donors. The implication of these results is that modulating DNA methylation in islet progenitors using maternal food supplements can be explored to improve b-cell function in the prevention and therapy of diabetes.

Figure 1. Postnatal b cells derived from M⁺N⁺ and M⁻N⁺ progenitors have different function

(A-C) b-cell proliferation assays. (A) Islets were picked, dissociated, and cytospun onto slides for Ki67 staining and quantification. Ki67 signal (white) marks cycling cells; tdT (red) identifies the descendants of M⁺N⁺ progenitors; insulin (green) marks b cells. White arrows, tdT⁻Ki67⁺ cells; red arrows, tdT⁺Ki67⁺ cells; yellow arrow, a non-b-Ki67⁺ cells. (B) The % of tdT⁺ or tdT⁻ b cells expressing Ki67 in each mouse (m: males. f: females). p value is from paired student t-test (two-tail type 2 error). (C) A dividing tdT⁺ b cell with two separating red nuclei (arrow). (D-H) Insulin secretion assays in pseudo-islets from b-cell subsets. tdT⁺ and tdT⁻ islet cells were separated by FACS (D), made into pseudo-islets (E), and assayed for insulin secretion in response to G2.8 (2.8 mM glucose), G20, or (G20 + 30 mM KCl). The % of total insulin secreted during a 45 minutes window was presented. (F, G) Secretion from pseudo-islets of 2-months old MNA male or female mice, respectively. (H) Pseudo-islets secretion from 8-months old MNA mice, with males and females mixed. Presented are (mean + SEM). p values are calculated with unpaired t-test. (I-O) b-cell death assays in pseudo-islets. (I) Examples of newly-formed pseudo-islets (5 days in hanging drops) from eYFP⁺ and eYFP⁻ b cells of MNYI mice. (J) Relative pseudo-islet size (represented by pseudo-islet area measured from microscopic images). (K) Images of typical live pseudo-islet (white arrow points at one example) 7-days after culture at 11 mM glucose, stained with trypan blue to visualize dead cells (red arrows). The numbers (L) and areas (M) of the surviving pseudo-islets (white arrow) were used to compare the viability of two types of pseudo-islets. Three batches of pseudo-islets were used for this assay. (M) The proportions of islets (using area as a parameter) that survived. (N, O) Pseudo-islet cell death induced by 24-hours treatment by 20 mM glucose and 0.6 mM palmitate in eYFP⁺ and eYFP⁻ pseudo-islets. Presented in (M) are (mean + SEM), with p values calculated with unpaired t-test. Scale bars, 20 mm. For all assays, at least two batches of pseudo-islets were used, derived from different islet isolation, dissociation, cell sorting, and reaggregation plates.

Figure 2. ScRNA-seq identifies DEGs in newly born b-cell subtypes.

Islets from P2 MNA mice were isolated, dissociated, and used in inDrop-seq. The b cells were then grouped as tdT⁺ and tdT⁻ sub-populations for comparison. (A) UMAPs showing the clustering of different cell types within islets (left) and the b cell subtypes with tdT expression highlighted (right). (B, C) Terms that are enriched in the P2 tdT⁺ (B) and tdT⁻ (C) b-cell subtypes, analyzed using DAVID. (D, E) Heat maps showing the relative expression of several genes that are not (D) or are (E) differentially expressed in tdT⁺ and tdT⁻ b cells. Presented are log-transformed expression levels which were normalized against their mean across the four replicates. To highlight the level differences between tdT⁺ and tdT⁻ cells within each replica (S1-S4), they were presented side-by-side. Note that there are large inter-sample variations, likely reflecting the volatile nature of heterogeneity in individual mouse.

Figure 3. DEGs between P60 b-cell subtypes have established roles for b cells.

Islets from P60 MNA mice were hand-picked and used for scRNA-seg. The b cells were then grouped as tdT⁺ and tdT⁻ sub-populations for comparison. (A) A UMAP showing sub-clustering of P60 b cells with tdT expression highlighted. (B) The ratio of tdT⁺/tdT⁻ cells at P2 and P60, determined via scRNA-seq. Mean + SEM were presented. P value was calculated using unpaired t-test. (C, D) Pathways/terms that are enriched in the tdT⁺ (C) and tdT⁻ (D) b-cell subtypes, determined using DAVID. (E-H) The expression levels of several candidate genes in b-cell subtypes (*: p<0.05. ***: p<0.001. ****: p<0.0001, determined using a Kruskal-Wallis test followed by a post hoc Mann-Whitney U). (I) Pathways/terms that are enriched in the DEGs shared by P2 and P60 samples.

Figure 4. Several DEGs between P60 b-cell subtypes contribute to their different secretory function.

(A) Heat maps showing the relative expression of several genes that are differentially expressed in P60 tdT⁺ and tdT⁻ b cells. Data presentation followed that in Fig. 2D, E. (B-E) The functional effects of reducing Myt TFs dosage in mouse islets. (B) Immunofluorescence (IF) of typical islets from 4-weeks old control (Myt1^F/+; Myt1L ^F/+; St18 ^F/+) and Myt1^F/+; Myt1L ^F/+; St18 ^F/+; Pdx1 ^Cre (het) mice. (C, D) IPGTT results of 6-weeks old male (C) and 3-months old female mice (D). Mean + SEM were presented. p values were calculated using 2way ANOVA. (E) GSIS of islets from control and Myt TF het mice. % of insulin secretion within a 45-minutes window was presented (mean + SEM). The conditions used were: G2.8, G20, and (G20 + 30 mM KCl). P values were calculated using unpaired t-test. (F-I) The effects of reduced Syt7 production in islet cells. (F) The strategy to reduce Syt7 expression, using two gRNAs to recruit fusion repressor protein dCas9-KRAB to the promoter regions of Syt7 for gene repression. (G) Morphology of 4-wks old islets from control and mutant mice. (H, I) GSIS of control and Syt7-KD islets. The % of the total insulin secreted within a 45 minutes window was measured. Mean + SEM were presented. p values were calculated using unpaired t-test. (J, K) Differences in glucose-induced Ca²⁺ entry in 2-months old tdT⁺ and tdT⁻ b cells. (J) A few snap shots of real-time GCaMP6 recording, which reports the level of cytoplasmic Ca²⁺. Basal glucose used was 1 mM; stimulatory glucose, added at time “0”, is 11 mM. (K) Quantification of Ca²⁺ influx in tdT⁺ and tdT⁻ b cells. The relative increase of Ca²⁺ signals were presented, normalized to that between −3 to 0 minutes before glucose administration. Cells with oversaturating GCaMP6 (arrows at 15 m) were not included for quantification. Mean + SEM were presented at each time point. p values were calculated using 2way ANOVA, with repeated measurement. Scale bars, 20 mm.

Figure 5. M⁺N⁺ and M⁻N⁺ progenitor-derived b cells have differentially methylated regions that may regulate gene expression.

(A, B) Myt1 and Neurog3 co-expression in pancreatic buds cultured in vitro with or without DNMT inhibition by AzaC. Only co-staining was shown. P value is from t-test, two-tail type 2 errors. Scale bar, 20 mm. (C) Methylation scores of DMRs between P2 eYFP⁺ and eYFP⁻ b-cell subtypes. (D) The locations of DMRs in respect to their predicted target genes. (E) DMRs near the Arx locus. Shown are four tracks (from top to bottom) of CpG methylation of eYFP⁻ b cells, DMRs with higher, DMRs with lower levels of methylation in eYFP⁺ cells, and CpG methylation of eYFP⁺ b cells. Among the two DMRs (red arrows), one has higher and the other lower levels of methylation in eYFP⁺ cells. The number following “X:” indicates the number of CpG dinucleotides within the DMR. (F) DNA motifs that are enriched in the DMRs between b-cell subtypes. The motifs in DMRs with lower or higher methylation were presented separately. (G, H) Pathways/terms that are enriched in the DMR-associated genes, with either lower- (E) or higher levels of (F) methylation. (I) P2 DEGs (down- or up-regulated) that are associated with P2 DMRs (with either lower or higher levels of methylation in eYFP⁺ cells). P values was calculated using Fisher exact test, comparing with random association with all detectable genes in b cells. (J) A DMR in the first intron of Syt7. The annotation is identical to that in panel (B). (K-M) The effects of site-specific methylation increase near the Syt7 locus. An U6 promoter was used to drive the expression of a guide RNA, which was expected to bring fusion protein dCas9-DNMT3a (reported by eGFP expression) to methylate DNA close by (I, J) (Liu et al., 2019). The co-expression of the two constructs in MIN6 cells (rep, or repression) resulted in significant reduction in Syt7 mRNA (K) than controls (con, without guide RNA). Scale bar, 20 mm. Presented in K are real-time RT-PCR assays of three samples (Mean + SEM), with p values calculated with unpaired t-test.

Figure 6. DNA methylome in postnatal b cells is dynamic but adult b-cell subtypes have DMRs that associate with a common set of genes as those at P2.

Genome-wide methylome analysis followed that in Fig. 5. (A) Clustering of HMRs in P2 and P60 b cells based on their levels of DNA methylation. The number of HMRs in each of the six clusters (C1 to C6) are shown. “*” indicate the clusters with >1.8 folds difference in their mean. (B) Methylation scores of DMRs between the P60 b-cell subtypes. (C) Enriched motifs in the DMRs with lower or higher levels of methylation (meth) in eYFP⁺ b cell subtypes. (D) DMR-associated DEGs between the P60 b-cell subtypes. P values is from Fisher exact test. (E, F) Terms that are enriched for DMR-associated DEGs, with lower (E) or higher (F) levels methylation in eYFP⁺ cells. (G-K) DMR/DEG overlaps between P2 and P60 cell subtypes. (G) Clustering of all DMRs at P2 and P60. (H) The number of DMRs that is retained in P2 and P60 cell subtypes. (I) The overlap between DMR-associated genes at P2 and P60. P values were calculated using hypergeometric analysis. (J) An example (DNMT3a) to show that a gene can be regulated by DMRs detected at P2 and P60, although the locations of the DMRs may shift (red arrows). (K) Pathways that are enriched in the P60 DEGs that also associated with P2 and P60 DMRs.

Figure 7. A reduction of b-cell subtype of higher function corresponds to diabetes development in mice and human.

(A) A few examples of genes with changed expression in Neurog3⁺ cells with HFD or control diet (CD) treatment. Shown are the levels of expression relative to that in CD-treated cells. (B, C) Examples and quantification of labeled b cells in P4 MNYI mice after maternal HFD/CD treatment. Bar = 20 mm. p value is from t-test (two-tail type 2). (D) The 20 mouse DEGs selected for human b-cell subpopulation studies. See Table S8 for references that helped with their selection. (E, F) UMAPs of human b-cell populations based on the aggregated expression of the 20 genes listed in (A), using available data in Kaestner et al. (Kaestner et al., 2019) and Xin et al. (Xin et al., 2018). (G) Gini indices and respective p-values to measure if the distribution of 20 genes was equal amongst all b cells. The null hypothesis is that the distribution was equal. (H, I) Processes that are enriched in Pop. 1 cells of Kaestner et al. (E) or Xin et al. (F). The top 9 processes are presented. (J, K) Sub-clustering and quantification of b-cell subpopulations from T2D donors. The T2D b cells from Kaestner et al. (Kaestner et al., 2019) were used. Presented are (Mean ±SEM). P value is calculated using t-test. (L) Pathways that enriched in Pop. 1 T2D donor islets.