Introduction
Insulin-producing beta cells are central to the modulation of glucose homeostasis, and their impaired function, loss of identity or lowered numbers result in type 2 diabetes [
1]. Previous studies have provided an understanding of the transcriptional machinery that orchestrates beta cell development from early pancreatic and endocrine precursors [
2]. To bolster these transcriptional programmes in vivo, chronic regulation via the epigenome appears to step in to maintain beta cell identity in the long term [
3‐
7].
Early reports [
8,
9], and more recent studies based on single-cell transcriptomic profiling [
10‐
12], electrophysiology [
13] and functional imaging [
14‐
17], have demonstrated functional heterogeneity amongst individual beta cells (reviewed in [
18]). Identified subpopulations have been associated with known markers or maturation states (
Flattop/
Cfap126 [
16], polysialyated-neural cell adhesion molecule [PSA-NCAM] [
19], CD81 [
20], CD24 [
7,
21], tyrosine hydroxylase [TH] [
6], neuropeptide Y (NPY) [
22], CD63 [
23]), are ‘virgin’ beta cells [
24] or are defined by their roles in coordinating islet-wide Ca
2+ dynamics (e.g. ‘hubs’ [
14], ‘leaders’ [
15,
25] and ‘first responders’ [
26]). Furthermore, loss of beta cell heterogeneity or intercellular connectivity may contribute to the development of type 2 diabetes [
14,
15,
27]. Importantly, functional subpopulations have also been demonstrated in human beta cells [
11,
12,
28], and the distribution of antigenically defined sub-groups (based on CD9 and ST8 α-
N-acetylneuraminide α-2,8-sialyltransferase [ST8SIA1] positivity) is altered in type 2 diabetes [
10]. The features underlying beta cell heterogeneity include pathways governing glucose sensing and metabolism [
14‐
16], insulin content and secretory competence [
5,
13,
24,
29,
30], and cilia activity and localisation within the islet [
25]. Recently, two discrete populations of ‘CD63
hi’ and ‘CD63
lo’ cells have been described [
23], with ‘CD63
hi’ cells enriched for CD63 and for insulin content and glucose-stimulated insulin secretion (GSIS). Epigenomically defined (by histone methylation, H3K27me3) CD24-positive ‘β
HI’ beta cells with enhanced insulin content and GSIS [
7] may partly overlap the ‘CD63
hi’ population [
23].
Imprinted genes are expressed from a single allele in a parent-of-origin-specific manner and their expression is also controlled via epigenetic modifications, notably DNA methylation. Imprinted genes often play key physiological roles, particularly in early (fetal and postnatal) growth and development, controlling a wide range of cellular processes. Thus, human imprinting disorders involving altered expression from specific imprinted loci are associated with severe childhood developmental and metabolic complications (reviewed in [
31]).
Imprinted genes play key functional roles in pancreatic beta cells by modulating insulin secretory machinery or beta cell mass [
32]. Correspondingly, imprinted gene expression is dysregulated both in beta cells with diminished GSIS and in pancreatic islets from individuals with type 2 diabetes, and single nucleotide polymorphisms (SNPs) at imprinted loci are associated with type 2 diabetes risk [
18]. Monoallelic expression of imprinted genes is maintained trans-generationally by differential methylation between parental alleles at imprinted loci in the germline [
33], with additional ‘somatic’ or ‘secondary’ differentially methylated regions (DMRs) also established post fertilisation [
34].
We have previously shown [
35] that the paternally expressed, imprinted gene
Nnat (encoding neuronatin [NNAT]) is nutrient-regulated in pancreatic beta cells, and controls insulin content and GSIS by modulating early insulin precursor processing at the signal peptidase complex (SPC) [
35]. At extra-pancreatic sites, changes in
Nnat expression also modulate appetite and metabolism [
36‐
38]. Here, we explore the possibility that differential methylation of the
Nnat gene contributes to the functional heterogeneity of embryonic and adult pancreatic beta cells.
Methods
For details, please refer to electronic supplementary material (ESM)
Methods.
Discussion
We describe here a novel subgroup of beta cells characterised by transient methylation of a second DMR (but not the gametic DMR) in the
Nnat locus. This is consistent with previous findings that methylation at gametic DMRs is thought to be relatively stable [
50], whereas ‘secondary’ imprinting regions have been shown to be more sensitive to nutrient- or physiology-based changes [
6,
51]. We have previously described the importance of
Nnat for beta cell insulin content and secretion [
35] and therefore hypothesise that NNAT
+ and NNAT
− cells display differences in secretory function. Correspondingly, transcriptomic analysis of purified adult NNAT
+ and NNAT
− beta cells demonstrated that the NNAT
+ fraction is enriched for functional pathways including translation initiation, ETC/oxidative phosphorylation pathways and co-translational protein ER membrane targeting.
Extending these studies to humans, NNAT deficiency in human EndoC-βH1 beta cells severely blunted GSIS, likely via inhibition of signal peptidase-mediated processing [
35]. We show here that E17.5
Nnat+ beta cells are enriched for expression of the SPC and translocon apparatus (
Spcs1,
Spcs2,
Sec11a,
Sec11c,
Sec61) and that the SPC/NNAT interaction is via SPCS1, and likely the means through which NNAT influences cellular insulin content. Importantly, we demonstrate that heterogeneous expression of NNAT is also a feature of the adult human pancreatic islet. Finally, NNAT
+ cells were de-enriched for highly connected ‘hubs’ and, thus, are likelier to belong to the ‘follower’ population [
14]. Interestingly, Ca
2+ responses to 11 mmol/l glucose were significantly higher in
Nnat-deficient vs wild-type mouse islets, in contrast to previous findings where differences were not observed in response to 16.7 mmol/l glucose [
35]. A similar strong tendency was also seen when comparing NNAT
+ vs NNAT
− cells in the same islet in
Nnat-eGFP reporter mice, and suggests that NNAT
+ cells are less responsive to metabolic stimulation by glucose, in line with their enrichment in a ‘follower’ subset of cells, and consistent with a role in insulin production rather than glucose detection [
14,
15]. The molecular underpinnings of the weaker Ca
2+ responses are unclear, but do not appear to involve differences in the levels of transcripts encoding
Slc2a2 (
Glut2) or
Gck. Moreover, the transcriptome of NNAT
+ cells does not show enrichment for genes enriched in ‘hub’ [
15] or ‘leader’ cells [
25].
Interestingly, bimodal expression of
Nnat was already clearly evident at embryonic stages, with
Nnat+ beta cells enriched for markers of late-stage beta cell differentiation, as well as
Ins1 and
Ins2 mRNAs. These differences were, however, less marked in the adult islet, though CD24a [
7] was more weakly expressed in the NNAT
+ population. De-enrichment of markers of other islet cell types (and the immaturity marker
Cd81) was also a common feature of both embryonic and adult NNAT
+ beta cells. Interestingly, both embryonic and adult NNAT
+ beta cells were enriched for
Npy, suggesting that these cells were not fully matured. NNAT
+ beta cells, however, had significantly higher insulin content, suggesting a possible, at least partial, overlap with the recently described ‘CD63
hi’ population [
23], and the ‘β
HI’ population described in reference [
7].
Our work also provides evidence that the epigenome controls the fate of specific beta cell subpopulations. CpG methylation plays a crucial role in early beta cell developmental maturation, including the silencing of ‘disallowed’ genes such as
Hk1,
Mct1 (
Slc16a1) and
Ldha via DNMT3A [
4,
48]. Here we show that islets transition from a state wherein the majority of beta cells are NNAT
+ in late embryogenesis to comprising a restricted subpopulation of NNAT
+ beta cells by P7. Whether this represents changes at the level of individual beta cells, or the turnover of the NNAT
+ population and replacement with a largely NNAT
− population, was not determined, and would require investigation using other approaches, including fate mapping (lineage tracing).
Adult NNAT+ beta cells are virtually unmethylated at the Nnat promoter whereas robust promoter methylation was apparent in NNAT− cells. These findings, and the fact that the Nnat promoter is differentially methylated between sperm and oocytes (see the Results), indicate that promoter methylation during this transition is likely to be acquired on the paternal allele (from which Nnat is selectively expressed via genomic imprinting). Our studies also show that specific deletion of the de novo methyltransferase DNMT3A at the pancreatic progenitor stage resulted in partial loss of NNAT-based beta cell heterogeneity. Thus, DNA methylation modulates restricted NNAT expression in specific beta cells during early maturation.
Might these findings be relevant for the pathogenesis of type 2 diabetes? Whether demethylation of the second DMR at the
Nnat locus may occur in this disease is an interesting possibility which remains to be explored. In rodents, even mild hyperglycaemia deregulates
Nnat expression alongside that of several other critical beta cell identity genes [
52], and we describe here a significant reduction of NNAT
+ beta cells in
db/db mice. Moreover, altered CpG methylation is a common observation in islets from individuals with diabetes at imprinted and non-imprinted loci (reviewed in [
18,
32]).
In conclusion, the present work demonstrates how an effector gene in pancreatic beta cells,
NNAT/Nnat, is controlled at the level of the epigenome (DNA methylome), contributing to a functional hierarchy between cells. Chemical modification of the epigenome may in future provide an attractive therapeutic angle not only for beta cell replacement or regeneration [
53] but also to modulate beta cell function and cell–cell connectivity in type 2 diabetes.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.