GATA4 Mutations Are a Cause of Neonatal and Childhood-Onset Diabetes

The GATA family of transcription factors, including GATA4 and GATA6, plays important roles in the development of several endodermally and mesodermally derived organs. Heterozygous mutations in GATA6 have previously been associated with congenital heart malformations (1), and, more recently, they were found to be the major cause of pancreatic agenesis in humans (2). Further studies have shown that patients with GATA6 mutations have a variable spectrum of diabetes manifestations, ranging from complete pancreatic agenesis to adult-onset diabetes without exocrine insufficiency (3).

GATA4 is a zinc finger transcription factor closely related to GATA6. Mutations in GATA4 have been described in familial and sporadic cases of heart malformation (4,5), although nonpenetrance is common (6). More than 120 cases of chromosome rearrangements resulting in interstitial or terminal deletions of chromosome 8p including GATA4 have been reported. The phenotypes associated with these deletions include congenital heart malformations, dysmorphic features, and mental retardation (reviewed in Wat et al. [6]).

Mouse studies have suggested a role for Gata4 in pancreatic development. Gata4^−/− mice die prior to pancreas specification due to defective extraembryonic tissues formation (7,8), but investigation of Gata4^−/− mice rescued from embryonic lethality shows agenesis of the ventral pancreas (9). Gata4^+/− mice are phenotypically normal (7,8).

In human embryos, GATA4 is detected in the area of the developing pancreas at 29–31 dpc (CS12), at the same time of the early transcription factor PDX1, suggesting a role of GATA4 in human pancreas development (10). There is one report of a GATA4 missense mutation (p.R319W) in a child with pancreatic agenesis and an atrial septal defect (11). The proband’s father and sister were also heterozygous for the mutation, which was shown to impair protein function in vitro. Both had a congenital heart defect but no diabetes, so a causal link between GATA4 and pancreatic agenesis could not be proven and evidence to confirm a key role for GATA4 in development of the human pancreas is lacking.

We investigated the role of GATA4 in human pancreatic development by studying three patients with diabetes and chromosome 8p deletions spanning the gene and searching for deletions or intragenic GATA4 mutations in 186 patients with neonatal diabetes (NDM) of unknown etiology.

Patients were defined from two sources: 1) One hundred eighty six probands diagnosed with NDM before 6 months in whom no genetic diagnosis had been established were selected from a cohort of 867 patients with NDM who were referred from 73 countries (the relevant clinical information was provided by the referring clinician); and 2) a cohort of patients with diabetes and known chromosomal abnormalities (n = 13).

We used a custom targeted next-generation sequencing (NGS) panel test (12) to screen the coding sequence and exon-intron boundaries of GATA4 in the NDM cohort. The assay is able to detect partial and whole-gene deletions in addition to base substitutions and indels. In two cases, there was insufficient DNA for targeted NGS and mutation testing was carried out by Sanger sequencing. We also sequenced the GATA4 pancreatic and endodermal enhancers described by Rojas and colleagues (13,14). Primer sequences are available on request. Sequencing reactions were run on an ABI 3730 capillary machine (Applied Biosystems) and analyzed using Mutation Surveyor v3.98 (SoftGenetics) (reference sequence NM_002052.3). Parent-proband relationships were confirmed using the PowerPlex kit (Promega).

Deletions of GATA4 were either cytogenetically visible rearrangements including chromosome 8p in patients from the known chromosomal abnormalities cohort or detected by reduced copy number on the targeted NGS assay in the NDM cohort. The extent of small GATA4 deletions was determined using the Affymetrix Genome-Wide Human SNP Array 6.0 (Santa Clara, CA), containing nearly 1 million copy number probes. Genotyping and within-batch copy number calling were performed by AROS Applied Biotechnology (Aarhus, Denmark). We visualized regions of reduced copy number state within Affymetrix Chromosome Analysis Suite and identified probes marking deletion breakpoints.

Mammalian expression vector pIRES(+)-GATA4 was used to perform DpnI mediated site−directed mutagenesis to generate GATA4 N273K (c.819C>A) mutation. Two independent clones were created, sequenced, and assayed in transient transfection assays in quadruplicates on three independent experiments, as described in Boj et al. (15). Lipofectamine 2000 was used to transfect HeLa cells with 0.15 µg of pGL3-WNT2 promoter luciferase construct (16) and 2-ng pRL Renilla luciferase reporter vector, in conjunction with different amounts of empty vector (pIRES-hrGFP), pIRES-GATA4, or pIRES-GATA4-N273K, to reach 0.5-µg total DNA. Firefly and Renilla luciferase activity were assayed using the Dual-Luciferase Reporter Assay System (Promega). Binding of nuclear lysates from cells expressing GATA4 and GATA4-N273K protein to ³²P-labeled oligonucleotides, including binding sites for GATA4, was performed as described (15). Sequences of oligonucleotides used in this assay include a predicted GATA binding site in the pancreatic (P2) HNF4A proximal promoter. Specificity of retardation complex was assessed by preincubating nuclear extracts with 100-fold excess wild-type or mutant unlabeled oligonucleotides, or GATA4 antiserum (GATA4 C-20; Santa Cruz Biotechnology, sc-1237). Immunoblot analysis was used to verify that the expression efficiencies for vectors encoding wild-type and N273K GATA4 were similar.

We report three patients with diabetes and a known chromosomal abnormality resulting in GATA4 haploinsufficiency. Cases 1 and 2 have terminal 8p deletions of 12.5 or 17 Mb, and Case 3 has an unbalanced translocation resulting in a large (∼15 Mb) terminal 8p deletion together with a duplication of the tip of chromosome 9 (Fig. 1). The possibility of an intragenic or regulatory GATA4 mutation on the nondeleted 8p allele or a GATA6 mutation was excluded by sequence analysis.

We then tested 186 patients with NDM for intragenic mutations or deletions of GATA4 and found two mutations. A whole-gene deletion in Case 4 was detected by targeted NGS assay and shown to be a 1 Mb interstitial deletion (chr8:10796092–11934660) that includes 8 additional genes (Fig. 1) by copy number analysis of genome-wide SNPs. A novel heterozygous missense mutation (c.819C>A, p.N273K) that affects a highly conserved amino acid located in the second zinc finger of the GATA4 protein was found in Case 5 (Fig. 2). The mutation is predicted as likely to be pathogenic by Alamut (Interactive Biosoftware, Rouen, France). Mutations in GATA6, PDX1, and PTF1A had previously been excluded in both cases.

All five patients were born to unaffected parents and testing in four cases showed that the GATA4 deletion or missense mutation had arisen de novo (Table 1). Four of the five patients were diagnosed with diabetes during the first week of life, giving an overall frequency of GATA4 mutations in our NDM cohort of 4/867 (0.5%).

To test whether the p.N273K mutation truly affects GATA4 function, we examined the ability of wild-type and mutant proteins to bind and activate target sites. Electromobility shift assays showed that wild-type GATA4 exhibits high affinity binding to a GATA binding site in the pancreatic HNF4A promoter and that the p.N273K mutation impairs this interaction (Fig. 3A). Furthermore, transfected plasmids expressing GATA4 with the p.N273K mutation showed a 4.3-fold decrease in the ability to activate expression of a promoter that responds to GATA4 (16) (Fig. 3B). These results indicate that the p.N273K mutation affects GATA4 transactivational activity by disrupting the ability of this protein to bind target recognition sequences.

The clinical characteristics are shown in Table 1. The endocrine phenotype was variable. Two patients had NDM diagnosed in the first week of life, which remitted temporarily but relapsed after 6 months and 7 years. Two patients had permanent diabetes diagnosed at 1 week or 13 years. Case 1 was diagnosed at 13 years and presented with an HbA_1c of 114.2 mmol/mol (12.6%) following a 6-month history of polyuria and polydipsia, but was not ketotic. No pancreatic autoantibodies were detected and C-peptide production was preserved 3 years’ postdiagnosis (stimulated C-peptide 3.8 ng/mL). Case 5 was born prematurely (34 weeks’ gestation), diagnosed with diabetes on the first day of life, and died of multiple organ failure at four days of age. The four patients with NDM had low birth weight (<3rd centile) consistent with insulin deficiency in utero. All five were treated with insulin.

The pancreatic exocrine phenotype was also variable. There was complete absence of the pancreas on postmortem examination of Case 5. Case 4 had exocrine pancreatic insufficiency (fecal elastase = 17 μg/g) and received exocrine supplementation. To assess subclinical exocrine insufficiency in the three patients not receiving exocrine supplementation, we measured fecal elastase and performed pancreatic imaging. There was reduced pancreatic size in two patients and low fecal elastase in one (132 μg/g; normal range >200 μg/g).

Congenital heart malformations were present in the four cases with GATA4 deletions and ranged from septal defects associated with pulmonary stenosis to atrioventricular canal defect, but they were not seen in Case 5 at postmortem. Developmental delay and neurocognitive defects were reported in all five patients.

Our study reports five patients with GATA4 mutations and a variable phenotype of transient or permanent diabetes diagnosed in neonates or during childhood. The exocrine pancreatic phenotype ranged from complete agenesis to hypoplasia with subclinical exocrine insufficiency or normal exocrine function. Additional features included neurocognitive defects and congenital heart malformations. The variable phenotype was not correlated with the size of the deletion, consistent with the variable penetrance of GATA4 mutations/deletions reported in the literature. Both patients with pancreatic agenesis (this study and D’Amato et al. [11]) had missense mutations rather than deletions, but the absence of diabetes in two heterozygous relatives suggests that a dominant-negative effect is unlikely (11). Our results indicate that GATA4 mutations/deletions are a rare cause of NDM, accounting for 0.5% of our international series (4/867 cases). The observation that postzygotic GATA4 mutations in embryonic heart tissue can cause congenital heart defects (16) implies a similar mechanism for NDM, but is impossible to prove without access to pancreatic tissue.

The novel GATA4 missense p.N273K mutation is highly likely to be pathogenic because it affects a highly conserved amino acid within one of the zinc fingers and has arisen de novo, and cell-based studies of the mutant protein demonstrated impaired DNA binding and reduced transactivational activity. A different mutation affecting the same residue (p.N273S) was previously reported in the heart tissue of a deceased patient with severe cardiac abnormalities (17), consistent with a pathogenic role of this mutation albeit with interindividual differences in the phenotypic expressivity.

The GATA4 deletions in our cohort ranged from 1 to 17 Mb with an overlapping interval that includes 8 additional genes. No mutations in these genes have previously been reported to cause NDM, although previous reports describe an association between diabetes and this chromosome 8p region, with a possible critical region at 8p23 (18). Several candidate genes have been proposed, including PPP1R3B (19) and BLK (20), but not GATA4. BLK sequencing was included within the targeted NGS assay but no mutations were found. Follow-up studies of individuals with 8p deletions and intragenic GATA4 mutations are required to ascertain the incidence of diabetes outside the neonatal period.

Conditional simultaneous knockout of Gata4 and Gata6 in mouse pancreas causes pancreatic agenesis, whereas only mild early developmental defects are observed in mice in which only Gata6 or Gata4 are ablated, suggesting redundant roles for Gata4 and Gata6 in mouse pancreatic development (21,22). Recent evidence also suggests that mouse pancreas development is sensitive to Gata4 and Gata6 dosage, as mice retaining only one Gata4 or Gata6 allele exhibit pancreatic developmental defects that are not observed with double homozygous mutant mice. These observations in mice should be contrasted with our results showing that in humans haploinsufficiency of either GATA6 (2,3) or GATA4 can cause impaired pancreatic development resulting in a wide spectrum of diabetes manifestations. Similar discrepancies in human/mouse phenotypes have been observed in other endodermal transcription factors involved in pancreas development such as Hnf1a, Hnf1b, and Hnf4a. Existing knowledge suggests that although the transcription factors that regulate pancreas development in humans and mice are very often the same, haploinsufficiency appears to cause much more severe manifestations during human as opposed to mouse pancreas development.

Intragenic GATA4 mutations or deletions causing GATA4 haploinsufficiency have been described in more than 140 patients with congenital heart malformations. In contrast, diabetes only has been reported in six patients with GATA4 mutations or deletions (our study and D’Amato et al. [11]), suggesting a lower penetrance of the pancreatic phenotype compared with the cardiac phenotype. The genetic and/or environmental factors underlying this phenotypic variability are not understood. Animal studies have shown that knockout of one copy of Gata4 in mouse strains with different genetic background results in a marked difference in the cardiac phenotype observed, ranging from absence of pathogenic phenotype to complex congenital heart defects (23). This evidence suggests that additional genetic factors contribute to the penetrance of the cardiac phenotype. A similar mechanism in humans would explain the difference in the pancreatic and extrapancreatic phenotypes observed in patients with GATA4 mutations or deletions. In our study, we were not able to identify any variant in the coding or regulatory regions of GATA4 or GATA6 that could account for the presence of the pancreatic phenotype in our patients. Further studies will be needed to assess the possible contribution of other genes to the variable phenotype associated to mutations in this transcription factor.

In summary, our results show that GATA4 mutations are a rare cause of NDM and pancreatic agenesis. This is an important diagnosis to make as it confers an increased risk for future offspring being affected with NDM and/or congenital heart defects. Testing for intragenic mutations or deletions of GATA4 should therefore be considered in patients with NDM or childhood-onset diabetes and congenital heart defects in whom no GATA6 mutation has been identified. This study confirms a role for GATA4 in the development of the human pancreas.

Acknowledgments. The authors thank the families who participated in this study.

Funding. The research leading to these results received funding from the European Community’s 7th Framework Programme (FP7/2007-2013) under grant agreement number FP7-PEOPLE-ITN-2008 (Marie Curie Initial Training Network, Biology of Liver and Pancreatic Development and Disease) and grant agreement number 223211 (Collaborative European Effort to Develop Diabetes Diagnostics), Ministerio de Economía y Competitividad (SAF2011-27086), Diabetes UK (ref. 11/0004193), and the Wellcome Trust. S.E. and A.T.H. are employed as core members of staff within the National Institute for Health Research–funded Exeter Clinical Research Facility. S.E., J.F., and A.T.H. are Wellcome Trust Senior Investigators and A.T.H. is a National Institute for Health Research Senior Investigator. M.B. and W.M. are supported by National Science Center, Poland (NCN) grant 2011/01/M/NZ5/02815 and by Innovative Economy Operational Programme–Activity 1.2 (the TEAM Programme coordinated by the Foundation for Polish Science).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.S.-S. researched the clinical data, contributed to discussion, and wrote the manuscript. E.D.F. researched the molecular genetics data, contributed to discussion, and wrote the manuscript. H.L.A., S.E.F., M.Bo., G.P.d.N., and I.W. researched the molecular genetics data and reviewed the manuscript. M.Ba. and J.F. performed the functional studies, contributed to discussion, and reviewed the manuscript. C.E.T., J.v.A.-v.d.V., J.C.-R., Z.M., and G.D. researched the clinical data and reviewed the manuscript. W.M. researched the clinical and molecular data, contributed to discussion, and reviewed the manuscript. A.T.H. directed the research, contributed to discussion, and reviewed the manuscript. S.E. directed the research, contributed to discussion, and wrote the manuscript. S.E. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.