Introduction
Diabetes is a metabolic disorder characterised by the inability of pancreatic beta cells to control blood glucose levels. The aetiology of the most common form, type 2 diabetes, is heterogeneous, although population genetic studies have identified numerous type 2 diabetes-associated genetic variants in loci of genes expressed in the beta cell [
1,
2]. Pathogenic variants in these beta cell genes lead to monogenic diabetes, presenting either as neonatal diabetes or MODY [
3]. Regulatory factor X 6 (RFX6) is a winged-helix transcription factor that regulates genes required for the development of the pancreas and other intestinal organs [
4‐
7]. Autosomal recessive variants in
RFX6 cause Mitchell–Riley syndrome (OMIM, 615710;
www.omim.org), characterised by intrauterine growth retardation, annular or hypoplastic pancreas, permanent neonatal diabetes, gall bladder hypoplasia or agenesis, intestinal stenosis with malabsorptive diarrhoea and, in some cases, pancreatic exocrine insufficiency [
6,
8,
9]. Recently, a less severe manifestation of Mitchell–Riley syndrome was documented, involving compound heterozygous variants that are not fully inactivating. This condition presents as childhood-onset diabetes, typically occurring between the ages of 2 and 5 years [
10,
11].
In mice,
Rfx6 is initially expressed in the definitive endoderm, before becoming progressively confined to the gut and dorsal pancreatic bud, and then islet progenitor cells where it is required for the differentiation of all islet cell types, except for pancreatic polypeptide (PP)-producing cells. Mice with loss of
Rfx6 exhibited similarities to the human phenotype, presenting with neonatal diabetes and intestinal obstruction, albeit with variable pancreatic hypoplasia [
6]. While homozygous
Rfx6 mutant mice had severe symptoms and died shortly after birth, heterozygous mutants did not show signs of diabetes [
6]. Conditional
Rfx6 ablation in adult mouse beta cells resulted in impaired glucose-stimulated insulin secretion (GSIS), without compromised beta cell mass or insulin content. The defective insulin secretion was attributed to the downregulation of beta cell maturation genes
Gck,
Abcc8,
Ucn3 and voltage-dependent calcium channel (VDCC) genes, concomitant with the upregulation of beta cell disallowed genes such as
Slc16a1,
Ldha, and
Igfbp4 [
12]. Similarly, knockdown of
RFX6 in the human beta cell insulinoma line EndoC-βH2 resulted in reduced insulin gene transcription and defective GSIS through reducing VDCC gene expression [
13]. A recent study also demonstrated that
RFX6 knockdown in primary human islets reduced GSIS to the level seen in islets from donors with type 2 diabetes, through transcriptionally dysregulated vesicle trafficking, exocytosis and ion transport pathways [
14].
Heterozygous
RFX6 pathogenic variants have been linked to MODY with reduced penetrance in humans [
15‐
21]. Moreover, genome-wide association studies have associated variants of
RFX6 with type 2 diabetes [
22,
23]. In adult primary islets, RFX6 was shown to be a hub transcription factor that was downregulated in islets from donors with early type 2 diabetes, correlated with reduced GSIS [
14]. Additionally, genetic variants that increase type 2 diabetes risk are predicted to disrupt RFX-binding motifs [
24].
The precise mechanism of how heterozygous pathogenic
RFX6 variant carriers are predisposed to develop diabetes remains unknown. Therefore, we sought to use patient-derived and embryonic stem cells combined with CRISPR-based genetic engineering to create isogenic allelic series models of a specific
RFX6 frameshift variant, circumventing the use of unphysiological systems of complete gene knockout or knockdown. Human pluripotent stem cells have been extensively used to model monogenic diabetes genes [
25], including
RFX6 [
9,
26,
27], as they can be differentiated into stem-cell-derived islets (SC-islets) that closely mimic native human islets developmentally and functionally. Employing our optimised protocol to generate highly functional SC-islets [
28,
29], we elucidate the impact of homozygous and heterozygous
RFX6 pathogenic variants on pancreatic endocrine development and beta cell function.
Discussion
A hierarchy of gene regulatory networks control the development and function of beta cells. Single pathogenic variants in more than 20 transcription factors of this network can lead to diabetes. Importantly, species differences in the developmental and functional processes of beta cells impede their precise study with mouse models. Here, we sought to overcome this limitation using genetically engineered human SC-islets, to investigate the developmental and functional defects of beta cells upon RFX6 perturbation.
Our findings demonstrate that
RFX6 p.His293LeufsTer7 is a loss-of-function variant with a transcript subjected to nonsense-mediated decay. We confirmed that the lower
RFX6 gene dose in heterozygous
RFX6 SC-islets led to
RFX6 haploinsufficiency, which did not compromise the differentiation capacity of the cells in terms of beta cell number or insulin content, but impaired beta cell function. There is no association between
Rfx6 haploinsufficiency and diabetes in mice, indicating that this phenotype cannot be faithfully modelled in rodents [
6]. Similar discrepancies between diabetes development in humans and mice have been reported with variants in
HNF1A,
HNF1B and
PAX4. While pathogenic heterozygous variants in all of these genes cause MODY in humans, heterozygous null mice lack any diabetes phenotype [
70‐
72]. However, our findings are in line with the beta cell-specific
Rfx6 knockout in adult mice [
12] and confirm the reported reduction of insulin secretion in islets from donors with type 2 diabetes as a result of reduced
RFX6 levels [
14]. Reduced RFX6 expression in
RFX6+/− SC-islets led to decreased levels of the beta cell marker
UCN3 and VDCC expression, lower [Ca
2+]
i and reduced insulin secretion in basal and high glucose levels. Moreover,
RFX6+/− SC-islets showed upregulation of disallowed genes that are reported to negatively impact insulin secretion, such as
IGF2 [
65],
CACNB3 [
66,
67] and
KCNQ1 [
68,
69].
Insulin content of the heterozygous cells was not affected, suggesting that the difference in insulin secretion more likely reflects impaired stimulus–secretion coupling or granule trafficking. However, exocytosis measured either as changes in capacitance or by TIRF imaging of granules at the plasma membrane was not reduced in response to depolarising stimuli. This is in line with intact K
+-induced insulin secretion in static and dynamic perifusion experiments. Despite slightly reduced gene expression levels of some voltage-gated Ca
2+ channel components, there was no difference in Ca
2+ currents, unlike the reduced Ca
2+-channel activity that was reported in
RFX6 knockdown in the EndoC-βH2 cell line [
13]. The reduced basal [Ca
2+]
i and the lower [Ca
2+]
i after membrane depolarisation may instead reflect differences in Ca
2+ transport or buffering. The lower basal [Ca
2+]
i is consistent with reduced insulin secretion at low glucose. Since the substimulatory [Ca
2+]
i influences granule priming [
73], reduced resting [Ca
2+]
i may indirectly impair secretion also at elevated glucose. However, the intact K
+-induced insulin secretion, and the reduced GSIS despite insignificantly altered [Ca
2+]
i, more likely point towards a role for
RFX6 in controlling the metabolic amplifying pathway. The [Ca
2+]
i elevation triggered by K
+ depolarisation was significantly reduced in
RFX6+/− cells, yet Ca
2+ influx near insulin granules might still reach a level sufficient to induce a maximum insulin secretion response, as evidenced in both static and dynamic insulin secretion assays. Based on these observations, we suggest insulin secretagogues and/or GLP-1 receptor agonists as treatment to enhance the secretory capacity of beta cells in heterozygous
RFX6 variant carriers.
The heterozygous
RFX6 SC-islets sustained lower insulin secretion in vivo, resulting in delayed lowering of blood glucose post implantation. All of our findings from the heterozygous cell model are consistent with the increased risk of variant carriers to develop diabetes, as shown for both gestational and type 2 diabetes, and previously for MODY with reduced penetrance [
15]. Further exploration of the clinical phenotype in variant carriers is warranted. However, two small studies have indicated elevated fasting glucose [
15] or 2 h glucose [
47] during an OGTT in carriers compared with non-carriers. Additionally, lower fasting and stimulated GIP levels were observed [
15]. Consistent with an impact on insulin secretion, carriers displayed a younger age at diagnosis of diabetes and a lower BMI.
The complete loss of RFX6 led to reduced expression of
PDX1,
NKX6.1,
FEV,
ISL1, NEUROD1 and
PAX6 with gradual loss of pancreatic progenitor identity. These genes are in line with some of the reported targets of ChIP-seq analysis of whole pancreatic tissue from adult mice (e.g.
Pdx1,
Neurod1 and
Nkx6.1) [
74]. Although, the pancreatic markers
NEUROG3,
NKX2.2 and
PAX4 were reduced at an earlier stage of the differentiation, they recovered at the endocrine precursor stage. This phenotype is similar to that reported in
Rfx6 knockout mice [
6,
12], with a difference of reduced
NKX6.1 expression in our model. While gene expression levels of
PPY were increased in
RFX6−/− endocrine precursors, PP cells formed in extremely limited numbers in vitro and were not found after 3 months following implantation in vivo, recapitulating the loss of pancreatic polypeptide in the plasma of the Mitchell–Riley syndrome patient, in contrast to the increased numbers of PP cells in the
Rfx6−/− mouse model [
6]. We demonstrate persistent expression of NEUROG3 and SOX9 at the pancreatic endocrine stage in
RFX6−/− cells, suggesting a negative regulatory role of RFX6 on SOX9 and NEUROG3 expression. Most of the emerging endocrine cells exhibited enterochromaffin cell identity with a pancreatic endocrine specification barricade, followed by increased cell death upon further differentiation in vitro. Reduced generation of the pancreatic progenitor pool and increased apoptosis phenotype could explain the pancreatic hypoplasia seen in homozygous individuals. Implantation of the homozygous cells confirmed persistent SOX9 expression in the majority of the grafted cells, while CHGA-expressing cells were extremely rare. This suggests that RFX6 functions to limit SOX9 expression to allow full endocrine differentiation. Human C-peptide was not detected in mice implanted with
RFX6−/− cells. The presence of small numbers of insulin-positive and glucagon-positive cells may explain the very low C-peptide levels observed in the patient but not the normal glucagon levels. These normal glucagon levels might be explained by potential extrapancreatic secretion of the hormone [
75].
In summary, we highlight the critical role of RFX6 in augmenting and maintaining the pancreatic progenitor pool, with an endocrine roadblock upon its loss, persistent NEUROG3 and SOX9 expression and increased cell death. We demonstrate that RFX6 haploinsufficiency does not affect beta cell number or insulin content but does impair function, predisposing carriers to diabetes. Our allelic series isogenic SC-islet models represent a powerful tool to elucidate specific aetiologies of diabetes in humans, enabling the sensitive detection of aberrations in both beta cell development and function.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.