Search
Close this search box.

Heterozygous missense variant in GLI2 impairs human endocrine pancreas development – Nature Communications

Family with a history of diabetes carries a heterozygous missense variant in GLI2

In a consanguineous family with diabetes of unknown etiology, we identified an uncharacterized heterozygous GLI2 missense variant (hg38 chr2:120990575; NM_001371271.1:c.4661 C > T; p.P1554L; rs767802807) (Fig. 1a, b; Supplementary Table 1). The index child developed diabetes at the age of 5 years, the individual currently in adult age requires multiple daily insulin injections and already had complications onset. The younger sibling also carries the variant and had onset of insulin-dependent diabetes at age of 3 years. Although both parents harbor the c.C4661T variant in GLI2, only the father developed diabetes. Moreover, the grandmother from the paternal side also had diabetes. However, we were not able to determine her genotype as she was deceased. Finally, no mutation in known, diabetes-causing genes have been found in a heterozygous or homozygous manner in any of the family members.

Fig. 1: Identification of heterozygous GLI2 mutation in a family with diabetes of unknown etiology.
figure 1

a Schematic representation of human GLI2 protein and the position of the single amino acid change p.P1554L in the Transcription Activation Domain (TAD). b Family tree of patients with puberty-onset diabetes. The heterozygous GLI2 p.P1554L (c.C4661T) variant was found in four individuals (red outline) of a consanguineous family with incomplete penetrance; diabetes is indicated by black shading. Females are represented by circles and males by squares. Consanguinity is represented by a horizontal parallel double bar. The missense variant was identified in patient 4 (index; arrow) by next generation sequencing55. The other family members were tested by Sanger sequencing. c The amino acid residue P1554 is highly conserved among species according to the University of California Santa Cruz (UCSC) Genome Browser. d Luciferase-based reporter assay with GLI-responsive construct in HEK 293 T cells. Wild-type (8 × 3’Gli-BSδ51LucII) or mutated (8xm3’Gli-BSδ51LucII) luciferase reporter plasmid was co-transfected with a Renilla luciferase control plasmid and indicated DNA expression vectors (GLI2CTRL or GLI2P>L). Results were normalized for transfection efficiency using Renilla luciferase and are represented as Firefly/Renilla activity ratio. The experiment was carried out three independent times with similar results. The average relative light units (RLU) of one representative experiment are shown. Two-tailed Student’s t-test, P = 0.05. Values shown are mean ± SD. ND, not detected. Source data are provided as a Source Data file. e Representative immunofluorescent (IF) images with indicated antibody combinations on human pancreas at Carnegie Stage (CS) 20, 12 post-conception week (pcw) and 20 pcw. At CS20 arrows indicate GLI2/PDX1-positive cells; arrowheads indicate GLI2-positive cell cluster next to PDX1-positive epithelium. NKX6.1 was detected at 12 pcw and 20 pcw. Arrows indicate a subset of pancreatic progenitors positive for GLI2, PDX1 and NKX6.1. Nuclei were labeled with Hoechst (Hoe). Scale bar, 20 μm.

So far, mutations in GLI2 have not been associated with diabetes in publicly available databases. To assess the effect of the identified GLI2 c.4661 C > T variant, we applied various independent in-silico tools (Supplementary Table 1). The GLI2 missense variant results into a single amino acid change (p.P1554L) in the transcription activation domain (TAD) of the protein and the affected residue is highly conserved across species with a genomic evolutionary rate profiling score >3 (Fig. 1a, c). Moreover, the p.P1554L missense mutation showed high CADD score and was predicted to be deleterious by all applied bioinformatics tools25 (Supplementary Table 1). Mutations that are pathogenic for early-onset diabetes are expected to be very rare in the population. Accordingly, the allele frequency (AF) of p.P1554L was 0.000044 in gnomAD and MAF < 0.0125.

Next, we examined the spatio-temporal pattern of expression of GLI2 in human pancreas (Fig. 1e and Supplementary Fig. 1l). We found that GLI2 is present in human early fetal pancreas, specifically in a subset of PDX1+ pancreatic progenitor cells at Carnegie Stage (CS) 20 and colocalizes with PDX1 and NKX6.1 at later stages (Fig. 1e). Consistently, GLI2 was expressed in human iPSCs undergoing differentiation into pancreatic cells, especially abundant at pancreatic progenitor (PP) and β-like cell stages (Supplementary Fig. 1l). Together, our results suggest that GLI2 is part of a complex regulatory network regulating human β-cell development.

The GLI2 p.P1554L patient variant exhibits decreased HH signaling activation

All three GLI proteins (GLI1, GLI2 and GLI3) share a conserved C2H2-type zinc finger DNA-binding domain15. GLI1 functions exclusively as transcriptional activator, while GLI2 and GLI3 can act as both activators or repressors, depending on the levels of HH16. To start investigating the newly identified GLI2 p.P1554L (hereinafter referred to as GLI2P>L) patient variant, we first tested whether it interferes with the basic transactivation property of the GLI2 transcription factor using a GLI-responsive luciferase reporter assay26. The GLI2P>L variant displayed significantly reduced transcriptional activity compared to wild-type (WT) GLI2 activity (Fig. 1d).

Next, to study the GLI2P>L patient variant in pancreatic β-cells, we established human isogenic iPSC lines with either the heterozygous or homozygous variant and differentiated them along the pancreatic cell lineage (Supplementary Fig. 1a). We used an iPSC line, which carries a doxycycline (DOX)-inducible Cas9 expression cassette inserted into the AAVS1 locus, allowing the de novo introduction of mutations into otherwise healthy iPSCs27. This strategy not only bypasses the limitations of iPSC generation from patients that requires biopsy and reprogramming, but also overcomes the issues related to the genetic background difference of patients derived-iPSCs versus healthy controls, which can substantially influence the phenotype12. Both heterozygous (GLI2P>L HET) and homozygous (GLI2P>L HOMO) gene-edited iPSCs displayed normal morphology and expressed pluripotency markers (Supplementary Fig. 1b–e). Moreover, GLI2P>L HET iPSCs displayed reduced endogenous transcriptional activation of the GLI-responsive luciferase reporter in comparison to control iPSCs (referred to as GLI2CTRL) (Supplementary Fig. 2a). These results are in line with the luciferase assay in HEK293T cells (Fig. 1d), indicating that GLI2P>L variant impairs HH signaling.

Defective pancreatic progenitor and endocrine fate specification in iPSCs carrying the GLI2P>L patient variant

The patient-like heterozygous GLI2P>L HET iPSCs were then differentiated into pancreatic β-like cells using a suspension-based differentiation protocol, as previously reported (Fig. 2a)28,29. Of note, this differentiation protocol29 does not require the addition of any HH signaling pathway inhibitors, unlike other protocols7,30,31,32.

L HET-derived iPSCs undergoing pancreatic endocrine differentiation.”>

Fig. 2: Characterization of GLI2P>L HET-derived iPSCs undergoing pancreatic endocrine differentiation.
figure 2

a Schematic representation of the differentiation protocol of iPSCs into pancreatic β-like cells29. Cells were differentiated in suspension as 3D clusters. b Representative flow cytometry plots of PDX1+ cells (shown as %) in WT GLI2CTRL and GLI2P>L HET-derived cells at day (D) 5 of differentiation. The differentiation experiment was carried out at least three independent times. c RT-qPCR analysis of selected gene transcripts in GLI2CTRL and GLI2P>L HET differentiated cells at D5 and D9. Values are normalized to GAPDH. Data are shown as fold change relative to undifferentiated cells (D0). Values shown are mean ± SEM. n = 3 differentiation experiments on two GLI2P>L HET independent clones. Two-tailed Student’s t-test; exact P-values are reported in the figure. d Representative flow cytometry plots of NKX6.1+ and PDX1+ cells (shown as %) in GLI2CTRL– and GLI2P>L HET-derived endocrine progenitor (EP) cell clusters at D14. e RT-qPCR analysis of selected gene transcripts in GLI2CTRL and GLI2P>L HET differentiated cells at D14. Data are represented as fold change relative to undifferentiated cells (d0). Values shown are mean ± SEM. n = 3 differentiation experiments on two GLI2P>L HET independent clones. Two-tailed Student’s t-test; exact P-values are reported in the figure. f Representative IF images for indicated markers on GLI2CTRL– and GLI2P>L HET-derived endocrine progenitor cells. Nuclei were labeled with Hoechst (Hoe). Scale bar, 20 μm. g RT-qPCR analysis of selected gene transcripts in differentiated cells at D21. Data are represented as fold change relative to undifferentiated cells (D0). Values shown are mean ± SEM. n = 3 differentiation experiments on two GLI2P>L HET independent clones. Two-tailed Student’s t-test; exact P-values are reported in the figure. h Representative static glucose stimulation insulin secretion (GSIS) assay of D21 GLI2CTRL and GLI2P>L HET derived β-like cells. On Y-axis ratio of insulin secreted at high versus low glucose [Glc] conditions. Values shown are mean ± SD. n = 3 differentiation experiments on two GLI2P>L HET independent clones. Two-tailed Student’s t-test; exact P-values are reported in the figure. Source data are provided as a Source Data file.

Healthy GLI2CTRL iPSCs robustly differentiated along the pancreatic cell lineage, yielding >20% of insulin-secreting β-like cells at day (D) 21 of differentiation (Supplementary Fig. 1f–k). Patient-like GLI2P>LHET iPSCs differentiated successfully to definitive endoderm and gut tube stage, as shown by cell morphology and expression of stage-specific markers (Fig. 2b, c), with >90% of the cells PDX1-positive, as quantified by flow cytometry analysis (Fig. 2b). Patient-like GLI2P>L HET iPSCs also progressed to the pancreatic and endocrine progenitor stages as the GLI2CTRL cells (Fig. 2c, d). However, the level of expression of PDX1 and NKX6.1 was significantly lower in GLI2P>LHET-derived progenitors (Fig. 2c) and this was accompanied by ~70% reduction in the fraction of PDX1/NKX6.1-double positive cells as compared to control cells (Fig. 2d). Additionally, the expression levels of other endocrine transcription factors, such as PAX4, NGN3 and its downstream targets NEUROD1 and NKX2.2, were reduced in GLI2P>L HET-derived endocrine progenitors (Fig. 2e). Consistently, the number of NKX6.1/INSULIN-double positive cells was significantly lower in GLI2P>L HET-derived cell clusters compared to controls (Fig. 2f and Supplementary Fig. 2b). No significant changes in proliferation were measured in the GLI2P>L HET mutant cells compared to controls (Fig. 2f and Supplementary Fig. 2d, f), while the number of apoptotic Cleaved Caspase-3 (cCAS3)+ cells was significantly increased at D9 (Supplementary Fig. 2e, g). Altogether, our results demonstrate that the GLI2P>L patient variant impairs the activation of genes essential for pancreatic progenitors and subsequent endocrine development, without inducing other endodermal lineages (Supplementary Fig. 2j).

Dose-dependent effect of GLI2P>L variant in β-cell differentiation

Co-expression of PDX1 and NKX6.1 in endocrine progenitors is a key step necessary for the transient expression of NGN3 and subsequent generation of glucose-responsive β-like cells33. Since GLI2P>L patient variant affected the differentiation of iPSCs into pancreatic and endocrine progenitors, we asked whether this had any functional consequences on β-cell development. Heterozygous GLI2P>L-derived endocrine progenitor cells were cultured in suspension for an additional week to promote the formation of β-like cells and then analyzed for their differentiation state and functional properties. Consistent with the findings at earlier stages, GLI2P>L HET β-like cells showed reduced expression of essential β-cell markers, including key transcription factors (PDX1, NKX6.1, NEUROD1, NKX2.2), genes important for insulin secretion (KIR6.2) and endocrine hormones (INSULIN, GLUCAGON) (Fig. 2g). Moreover, GLI2P>L HET β-like cells showed a striking decrease of insulin release in response to glucose stimulation (Fig. 2h). Thus, these findings support impaired β-cell differentiation in GLI2P>L HET cells, which might be responsible for causing diabetes in individuals carrying the c.C4661T variant in GLI2.

To further characterize the phenotype and assess a possible dose-dependent effect of the GLI2P>L variant, we differentiated homozygous GLI2P>L iPSC lines towards pancreatic progenitors and β-like cells. Mutant and control lines differentiated efficiently into definitive endoderm and pancreatic foregut stages (Fig. 3a, b). Quantification by flow cytometry showed that over 90% of GLI2CTRL and GLI2 P>L HOMO-derived cells were positive for PDX1 (Fig. 3a). However, after gut tube stage, the number of GLI2P>L HOMO cell clusters gradually decreased and differentiation arrested at endocrine progenitor stage, failing to reach β-like cell stage. Differences between GLI2CTRL and GLI2P>L HOMO became evident at pancreatic progenitor stage and exacerbated at endocrine progenitor stage, with only ~10% of GLI2P>L HOMO-derived clusters being positive for both PDX1 and NKX6.1 (Fig. 3c, e). Taking together, the homozygous GLI2P>L iPSCs show a more severe phenotype than GLI2P>L HET iPSCs, failing to differentiate into β-like cells (Fig. 3). These findings suggest a dose-dependent effect of the GLI2P>L variant on endocrine lineage differentiation. Notably, no patients homozygous for the GLI2P>L variant were reported so far.

L HOMO cells during pancreatic cell differentiation.”>

Fig. 3: Characterization of homozygous GLI2P>L HOMO cells during pancreatic cell differentiation.
figure 3

a Representative flow cytometry plots of PDX1+ cells (shown as %) in GLI2CTRL and GLI2P>L HOMO-derived cells at day (D) 5 of differentiation. n = 3 differentiation experiments on two GLI2P>L HOMO independent clones. b RT-qPCR analysis of selected gene transcripts in GLI2CTRL– and GLI2P>L HOMO differentiated cells at D5. Values were normalized to GAPDH and shown as fold change relative to undifferentiated cells (D0). Values shown are mean ± SEM. n = 3 differentiation experiments on two GLI2P>L HOMO independent clones. c Representative FACS plot of NKX6.1+ and PDX1+ cells (shown as %) in GLI2CTRL– and GLI2P>L HOMO-derived cells at D14. n = 3. d RT-qPCR analysis of selected gene transcripts in GLI2CTRL and GLI2P>L HOMO differentiated cells at D14. Values were normalized to GAPDH and shown as fold change relative to undifferentiated cells (D0). Values shown are mean ± SEM. n = 3. Two-tailed Student’s t-test; exact P-values are reported in the figure. Source data are provided as a Source Data file. e Whole-mount immunostaining for PDX1 and NKX6.1 in GLI2CTRL and GLI2P>L HOMO-derived endocrine progenitor clusters. Immunostaining experiments were performed 2 independent times, and each experiment showed similar results. Nuclei were labeled with Hoechst (Hoe). Scale bars, 50 μm.

Finally, GLI2P>L HOMO-derived pancreatic progenitor cells failed to progress along endocrine differentiation also when exposed to a distinct differentiation protocol which involves inhibition of HH signaling pathway (Supplementary Fig. 3)32. Glucagon expression was not affected in this condition (Supplementary Fig. 3c), suggesting a more specific impairment of the β-cell lineage. Thus, independently of the differentiation protocols applied, GLI2P>L HOMO cultures showed decreased expression of endocrine transcription factors and β-cell genes and overall reduction in bonafide NKX6.1/INSULIN-positive β-cells (Figs. 2, 3 and Supplementary Fig. 3).

Inhibition of non-canonical WNT pathway partially rescues GLI2P>L HET-derived endocrine progenitors

To further characterize putative downstream mechanisms underlying the impaired endocrine differentiation of GLI2P>L HET iPSCs, we performed bulk RNA-seq at three stages of differentiation (iPSC, gut tube, and endocrine progenitor stages) (Fig. 4a). At early stages (iPSC and gut tube stage), the transcriptomes of GLI2CTRL and GLI2P>L HET-derived cells were highly comparable, with only 48 genes significantly dysregulated at gut tube stage (Supplementary Fig. 4a; Supplementary Data 3, 4). By contrast, differences at the transcriptome level became evident at endocrine progenitor stage [928 genes upregulated and 1070 downregulated in mutant versus control cells] (Fig. 4b; Supplementary Data 5). Gene ontology (GO) term analysis identified genes classified under ‘endocrine pancreas development’, ‘regulation of insulin secretion’ and ‘glucose homeostasis’ categories to be among the most significantly downregulated ones in GLI2P>L HET endocrine progenitor cells (Supplementary Fig. 4c). Consistently, gene set enrichment analysis (GSEA) showed that gene sets significantly downregulated in mutant GLI2P>L HET cells belong to the ‘Pancreas development’ GO term, including PDX1, NKX6.1, NGN3 and SOX9 (Fig. 4c). Gene network analysis highlighted that significantly downregulated categories were enriched with factors involved in pancreatic secretion, insulin secretion and MODY (Fig. 4e). Furthermore, RNA-seq analysis revealed upregulation of genes encoding non-canonical WNT ligands, such as WNT5A, WNT5B, WNT7A, in GLI2P>L HET endocrine progenitors (Fig. 4b, d and Supplementary Fig. 4b, d). RT-qPCR further validated the dysregulation of WNT ligands (e.g., WNT2, WNT5A, WNT7A) and WNT receptors (e.g., FZD3, FZD7, FZD8) as well as GLI transcripts and components of the HH pathway in GLI2P>L HET endocrine progenitor cells (Fig. 4f). Interestingly, the transcript levels of GLI2 and GLI3 were upregulated in GLI2P>L HET starting from endocrine progenitor stage, while GLI1 expression was downregulated (Supplementary Figs. 2i, 4b; Supplementary Data 4, 5). This could be the result of an attempt of compensating HH downstream inactivation.

L HET iPSCs undergoing differentiation into endocrine progenitor cells.”>

Fig. 4: Whole-transcriptome analysis of GLI2P>L HET iPSCs undergoing differentiation into endocrine progenitor cells.
figure 4

a Schematic of the experimental design. RNA-seq data were obtained from GLI2CTRL and GLI2P>L HET iPSCs at day (D) 0 (undifferentiated stage) and undergoing differentiation at D5 (gut tube stage) and D14 [endocrine progenitor (EP)] stage from two independent biological replicates. b Volcano plots visualizing the global transcriptional change across the groups compared (D14 GLI2CTRL vs GLI2P>L HET). Each data point in the plot represents a gene. Genes with an adjusted P-value <0.05 and a log2 fold change >1 are indicated by red dots. These represent upregulated genes. Genes with an adjusted P-value <0.05 and a log2 fold change <−1 are indicated by blue dots. These represent downregulated genes. Wald test was used as statistical test as implemented in DESeq2 package (see Methods). c, d Gene set enrichment analysis (GSEA) plots of representative gene set downregulated (c) or upregulated (d) in GLI2P>L HET vs GLI2CTRL EPs at D14. GSEA P-values are derived from permutation test and corrected for multiple testing using the FDR method. Nominal P-value = 0.00, FDR = 0.00. e GO term enrichment and pathway term network analysis of DEGs between GLI2CTRL– and GLI2P>L HET-derived EPs. Gene Network showing downregulated genes (FDR  0.05). Term node size increases with term significance and color corresponds to a particular functional group that is based on the similarity of their associated genes. Groups that shared 50% or more of the same genes were merged. The proportion of each node that is filled with color reflects the kappa score. Functionally related groups partially overlap. f RT-qPCR validation of a subset of DEGs in GLI2CTRL– and GLI2P>L HET-derived EPs. All tested genes showed concordant differential expression with RNASeq results. Validation was performed on an independent differentiation experiment. Data were normalized to that of GAPDH and shown as Log2-expression ratio between GLI2P>L HET and GLI2CTRL EP cells. Source data are provided as a Source Data file.

Recent observations pointed out to multiple distinct roles of the non-canonical WNT pathway in endoderm and endocrine cell differentiation, morphogenesis, and maturation, entailing its tight regulation during pancreatic development34,35,36,37.

Because of the significant upregulation of WNT5a in GLI2P>L HET iPSCs from gut tube stage onward (Fig. 4b, d, f; Supplementary Fig. 4b, d, e), we hypothesized that its aberrant and sustained activation may contribute to impaired endocrine cell development in the mutant lines. To test this hypothesis, we either stimulated or inhibited the non-canonical WNT signaling pathway in GLI2CTRL and GLI2P>L HET iPSCs, respectively (Fig. 5a). GLI2CTRL cells were treated with WNT5A for 4 days after the induction of PDX1 and acquisition of pancreatic fate, and, subsequently, analyzed by RT-qPCR. After exposure to WNT5A, we observed a decrease in the expression of pancreatic (PDX1, NKX6.1) and endocrine (NGN3, NKX2.2, NEUROD1) gene markers in GLI2CTRL-derive endocrine progenitors (Fig. 5b). RT-qPCR analysis also showed upregulation of WNT5A itself and its downstream targets, LAMC2 and TEAD4, in stimulated GLI2CTRL cells (Fig. 5b). GLI2CTRL cells that were further differentiated to β-like cell stage after exposure to WNT5A showed a decrease in the expression levels of NKX2.2 and NEUROD1, accompanied by a reduction of gene transcripts essential for human β-cell functionality, such as PCSK1, as well as endocrine hormones, including INSULIN, GLUCAGON, SOMATOSTATIN (Fig. 5c). Exposure to WNT5A also affected the functional maturation of the cells, which showed reduced glucose-stimulated insulin secreting (GSIS) capacity as compared to GLI2CTRL-derived β-like cells, but similar to GLI2P>L HET mutant cells (Fig. 5d). Together, these results demonstrate that inappropriate activation of WNT5A signaling impairs differentiation of iPSCs into β-like cells and recapitulates the phenotype of GLI2P>L HET mutant cells.

L HET iPSCs.”>

Fig. 5: Modulation of WNT5A signaling rescues β-cell differentiation in GLI2P>L HET iPSCs.
figure 5

a Schematic representation of the differentiation protocol of iPSCs into β-like cells. Cell clusters were treated in suspension with WNT5A recombinant protein or BOX5 for 4 days at pancreatic progenitor (PP) stage of differentiation. b RT-qPCR analysis of selected gene transcripts in GLI2CTRL– and GLI2P>L HET-derived endocrine progenitors (EP) at day (D) 14 after the indicated treatments. Data are represented as fold change relative to undifferentiated cells (D0). Values shown are mean ± SEM. n = 3 differentiation experiments. Two-tailed Student’s t-test; exact P-values are reported in the figure. c RT-qPCR analysis of selected gene transcripts in GLI2CTRL– and GLI2P>L HET differentiated cells at D21 after the indicated treatments. Data are represented as fold change relative to undifferentiated cells (D0). Values shown are mean ± SEM. n = 3 differentiation experiments. Two-tailed Student’s t-test; exact P-values are reported in the figure. d Representative static GSIS assay of D21 GLI2CTRL– and GLI2P>L HET-derived β-like cells after the indicated treatments. On Y-axis ratio of insulin secreted at high versus low glucose [Glc] conditions. Values shown are mean ± SD. n = 3 differentiation experiments. Two-tailed Student’s t-test; exact P-values are reported in the figure. Source data are provided as a Source Data file.

We next investigated if inhibition of WNT5A could instead rescue β-cell development in GLI2P>L HET iPSCs. Cells were treated with BOX5, a WNT5A antagonist, which was previously shown to attenuate WNT5A-mediated Ca2+ and protein kinase C signaling38 (Fig. 5a). GLI2P>L HET clusters treated with BOX5 exhibited a significant increase in the expression of PDX1, NKX6.1 and other endocrine markers (Fig. 5b). Moreover, the expression of WNT5A and its targets was reduced after addition of BOX5 (Fig. 5b), confirming the activity of the compound. Later at D21, BOX5-treated GLI2P>L HET-derived β-like cells restored the expression of β-cell genes (NKX6.1, NKX2.2, NEUROD1, PCSK1, INSULIN) to levels comparable with control cells. Consistently, we found that exposure to BOX5 ameliorates the functional defects of GLI2P>L HET-derived β-like cells, with the treated cells showing an increase of insulin release in response to glucose stimulation (Fig. 5d). Overall, these findings indicate that WNT5A signaling regulation is critical for pancreatic progenitor development and its inhibition partially rescues the GLI2P>L HET defects that may ultimately contribute to diabetes onset.