Recessive Mutations in PCBD1 Cause a New Type of Early-Onset Diabetes

Diabetes is classified into type 1 diabetes (T1D) caused by autoimmune β-cell destruction; type 2 diabetes (T2D), caused by relative insulin deficiency in face of insulin resistance; gestational diabetes mellitus; and other specific types, including monogenic diabetes (1). T2D is a polygenic disease with over 60 susceptibility loci and numerous risk variants colocated with genes, causing monogenic diabetes (2). Monogenic diabetes is defined by neonatal, childhood, or a postpubertal age of onset, variable clinical presentation, a lack of autoimmunity, and acanthosis nigricans as well as uncommon obesity and ketoacidosis outside the neonatal period. It affects ∼1% of diabetic patients (1,3). Interestingly, monogenic cases are often accompanied by endogenous insulin production and lack of insulin resistance. Heterozygous GCK, HNF1A, and HNF4A mutations are the most common cause, although more than 20 other genes have been described (3,4). However, the disease-causing variants in numerous families remain obscure. Identifying novel genes would provide insights into pathogenesis and suggest new treatment strategies for rare monogenic diabetes as well as common polygenic T1D and T2D.

We sought to identify additional genes by combining linkage analysis with the whole-genome sequencing of a consanguineous family 1 (Fig. 1A and Table 1) with nonautoimmune diabetes and no pathogenic mutations in HNF1A, HNF1B, HNF4A, INS, ABCC8, or KCNJ11 genes. Our institutional review boards approved the studies and written consent was obtained from the participants or their guardians.

To pinpoint suggestive linkage regions, we performed a haplotype mapping and parametric linkage analysis of seven members of family 1 (Fig. 1A, red borders) using HumanCytoSNP-12 v2.1 BeadChip (Illumina) and Merlin software, assuming recessive inheritance, complete penetrance, and a disease allele frequency of 0.001. We obtained 24 genomic regions with positive logarithm of odds (LOD) scores, including 13 regions on 10 chromosomes with maximal LOD score of 1.3 (Supplementary Fig. 1). We next performed a whole-genome sequencing of five individuals from family 1 (Fig. 1A, asterisks) on the Complete Genomics platform (Mountain View, CA). The Complete Genomics pipeline was used for read mapping and allele calling. On average, both alleles were called for ∼96% genomic and ∼98% exonic positions, whereas ∼97% and ∼99% of the called ones were covered by at least 10 reads, respectively (Supplementary Table 1). We found ∼4 million small variations (small indels and single nucleotide polymorphisms [SNPs]) per individual genome including more than 23,000 variants in each exome.

To distinguish relevant SNPs and small indels from other variations, we used Complete Genomics Analysis Tools, ANNOVAR, and custom scripts. We first removed intergenic variants and anticipated recessive inheritance (Supplementary Fig. 2). Therefore, homozygous variants were required to be present in index case III-2 and heterozygous in II-1, II-2, and I-4, but not in I-3, who is not related to I-4. As monogenic diabetes is uncommon, we predicted the disease-causing variant to be rare and likely not yet identified. Thus, we removed all the variants present in dbSNP 137; 1000 Genomes Project; National Heart, Lung, and Blood InstituteExome Sequencing Project; 69 sequenced individuals from Complete Genomics; and our in-house database of nondiabetic individuals. We focused on nonsynonymous variants, splice site mutations, and small indels within protein-coding regions. We further selected conserved alterations, as defined by phastCons, and variants predicted to be deleterious by at least two tools: SIFT, PolyPhen-2, or MutationTaster. This reasoning left six genes (Supplementary Table 1), from which we selected PCBD1, encoding pterin-4 α-carbinolamine dehydratase also known as dimerization cofactor of hepatocyte nuclear factor 1 α (DCoH) (5), for further analysis. Our decision was based on the strong expression of PCBD1 in pancreatic islets (T1DBase and ref. 6), mouse pancreatic progenitors (7), and its interaction with HNF1A and HNF1B transcription factors, essential for proper pancreatic β-cell function (5,8,9). Moreover, PCBD1 lay within the suggestive linkage region (Fig. 1C) and was the only gene containing a frameshift deletion. Subsequently, we validated the deletion in all diabetic members of family 1 by Sanger sequencing (Fig. 1B), using exon-intron spanning primers (Supplementary Table 1).

PCBD1 is a bifunctional protein that acts as an enzyme in the regeneration of cofactor tetrahydrobiopterin (BH₄) (10), crucial for the function of aromatic amino acid hydroxylases, and as a dimerization cofactor of transcription factors HNF1A and HNF1B (5), which are important in liver and pancreas development and function. The enzymatic function of PCBD1 is defective in newborns with mild transient hyperphenylalaninemia (HPA) and high urinary levels of primapterin caused by recessive mutations (11,12). Recently, a PCBD1 defect was suggested to cause hypomagnesemia and diabetes (13). The novel homozygous deletion c.46del in family 1 results in a premature stop codon p.[(Lys16Cys*5)];[(Lys16Cys*5)] that abolishes the transcription factor–binding and enzymatic functions of PCBD1 (Fig. 1G).

As III-2 developed diabetes in puberty, we reevaluated patients with neonatal HPA caused by PCBD1 who appeared in the BIODEF database (14). Three out of seven children from six families, exhibiting three different biallelic PCBD1 defects of p.[(Glu87*)];[(Glu87*)], p.[(Glu97Lys)];[(Gln98*)], and p.[(Gln98*)];[(Gln98*)], had already developed antibody-negative diabetes with normal pancreatic morphology (Fig. 1D–F and Table 1). Taken together, our analyses of seven independent families provide strong genetic evidence that mutations in PCBD1 cause puberty-onset diabetes.

We tested the in vitro effects of a transient Pcbd1 knockdown in glucose-sensitive mouse insulinoma cells using small interfering RNA and found no obvious defects in insulin production or glucose-stimulated insulin secretion after an 80% inactivation of Pcbd1 (data not shown). Endocrine pancreas dysfunction not only may arise from β-cell inability to produce and/or secrete insulin but also from impaired β-cell development, proliferation, and adaptation during fetal, neonatal, and pubertal age. To address whether Pcbd1 controls early pancreas development and/or pancreatic β-cell fate specification, we first examined its expression pattern in the developing pancreas of both mouse and Xenopus embryos (Supplementary Data). In the mouse embryo, Pcbd1 transcript has been reported in the foregut endoderm, which contains liver and pancreas progenitors at embryonic stage (E) 8.5, and in E10.5 liver and ventral and dorsal pancreatic buds (7). We found an abundant expression of Pcbd1 in the embryonic pancreas at E12.5 and E14.5 (Fig. 2A–D). Interestingly, Pcbd1 accumulated in endocrine progenitors that had started to delaminate from E-cadherin–positive pancreatic epithelium and expressed insulin (Fig. 2A’ and C’). At E14.5, Pcbd1 expression was maintained in endocrine progenitors and was visible throughout the pancreatic epithelium (Fig. 2D). To analyze if this pcbd1 expression pattern was conserved in Xenopus endoderm, we performed reverse transcription-quantitative PCR (RT-qPCR) on microdissected endoderm cells. Fate map experiments in Xenopus have previously shown that pancreatic progenitors arise from anterior endoderm (AE), whereas posterior endoderm (PE) forms mainly intestine (15). We found that pcbd1 along with the pancreatic genes pdx1 and ptf1a marks future pancreatic endoderm, being expressed at higher levels in AE than in PE from the gastrula stage onward (Fig. 2E and data not shown). The pcbd1 binding partners, hnf1a and hnf1b, displayed similar expression profiles, whereas the close homolog pcbd2 was almost absent from AE cells (Fig. 2E). Furthermore, an in situ hybridization in Xenopus embryos showed pcbd1 expression in pancreatic rudiments, overlapping with the expression pattern of insulin (Fig. 2F–I). Overall, these results indicate that Pcbd1 is expressed in the developing pancreas of both mouse and Xenopus embryos, suggesting a potential evolutionarily conserved function.

Pcbd1 knockout mouse showed mild glucose intolerance (16), although no pancreas-specific function had yet been assigned to Pcbd1. To determine if Pcbd1 influences early pancreas fate specification, we undertook a loss-of-function approach in Xenopus. A specific morpholino oligonucleotide (pcbd1-MO) was designed to block pcbd1 pre-mRNA splicing. The injection of pcbd1-MO into AE cells of the eight-cell stage Xenopus embryos resulted in a dose-dependent downregulation of pcbd1 mRNA (Fig. 3A and data not shown), accompanied by a significant reduction in the expression of pancreatic progenitor genes pdx1, ptf1a, sox9, and insulin (Fig. 3A). Moreover, hnf1a and hnf1b and the hnf1 target gene fibrinogen were downregulated upon pcbd1-MO injection, whereas pcbd2 mRNA levels remained unchanged (data not shown). In situ hybridization of pcbd1-MO injected Xenopus embryos corroborated these observations, showing a strong reduction or complete loss of ptf1a expression in both dorsal and ventral pancreatic buds, but not in the eye and hindbrain (Fig. 3B–D ”). Taken together, these results indicate that pcbd1 activity within the endoderm is required for the proper establishment of the pancreatic region in vertebrates.

We examined eight patients with inherited biallelic PCBD1 mutations from seven families: one case by whole-genome sequencing combined with linkage analysis, which identified a novel deletion in the PCBD1 gene and seven cases by recalling the patients from the BIODEF database (14). Interestingly, two diabetic patients (family 1, III-2 and family 3, II-2) had normal phenylalanine levels, suggesting that the enzymatic function of PCBD1 had been compensated for by PCBD2, as has been proposed for mouse (16).

Insulin was the first option of treatment of PCBD1-diabetes, subsequently replaced by sulfonylureas or glinides. This brisk response to oral drugs resembles the patients with HNF1A-diabetes (17). PCBD1-diabetes manifests earliest in puberty as no younger cases that were investigated were diabetic (Table 1). Remarkably, mid-adulthood–onset T2D developed in four out of seven families (families 1, 3, 4, and 6; Table 1 and Supplementary Table 1). We confirmed heterozygous PCBD1 defects in affected individuals of families 1 and 3. Therefore, it is likely that monoallelic variants in PCBD1 increase the risk of T2D, as has been shown for other monogenic diabetes genes (2). Moreover, it is suggestive that PCBD1 mutations increase T2D susceptibility specifically when combined with other risk factors such as excess weight and age. This is implied by the fact that only overweight/obese heterozygotes developed T2D, whereas none of those with a normal BMI did (Supplementary Table 1). Furthermore, the two overweight/obese parents who did not develop diabetes were 32 and 34 years old and thus had not yet reached the age of the onset of diabetes in the other heterozygotes.

We investigated two major mechanisms of diabetes: the production and secretion of insulin in insulinoma cells and the early regulation of pancreatic and β-cell specification in vertebrates. As PCBD1 has been described to enhance HNF1A activity on some promoters (5), we examined whether a transient inactivation of PCBD1 decreases insulin production and secretion in vitro, but failed to see any relevant effect (data not shown). Previous studies have shown that Pcbd1 is abundantly expressed throughout embryonic development in both Xenopus and mammalian embryos (18,19). Interestingly, we found that PCBD1 is expressed in mouse pancreatic progenitors (7) and delaminating endocrine cells from very early stages onward (Fig. 2A–D). This spatiotemporal expression of pcbd1 is also conserved in Xenopus embryos (Fig. 2G and I), suggesting conserved regulatory functions. In line with this, Xenopus pcbd1 morphants exhibited a reduced expression of endodermal and pancreatic transcription factors, indicating defects in early pancreas specification. Notably, both dorsal and ventral pancreatic rudiments fail to be established in pcbd1-depleted embryos, as judged by the absence of ptf1a expression (Fig. 3B–D). Altogether, these findings suggest an early role of pcbd1 in establishing the pancreatic progenitor pool during embryogenesis, which might lead to a reduced pancreatic β-cell mass in the adult. The human PCBD1-diabetes phenotype with developing insulin deficiency in the face of somatic growth and weight gain is in line with the notion that an intrinsic program established early in development is critical in determining the final size of the pancreas and is not subject to growth compensation (20).

Transcription factor HNF1B regulates early pancreatic development in mouse and human and is stabilized by PCBD1 (5,21). A lack of PCBD1 might impair the HNF1B-mediated establishment of pancreatic cell fate during embryogenesis. In Xenopus embryos, this hypothesis is supported by a reduced expression of hnf1b and its direct target genes, such as fibrinogen, upon pcbd1-MO injection. However, a Pcbd1 knockout mouse shows a relatively mild phenotype compared with Xenopus pcbd1 morphant or human HNF1A and HNF1B loss-of-function phenotypes (8,9). These differences might be due to a partial functional redundancy between Pcbd1 and the close homolog Pcbd2 in the mouse (16), which does not seem sufficient to prevent diabetes. In line with this, pcbd2 is expressed at low levels in the endoderm and is almost undetectable in the pancreatic territory of Xenopus embryos (Fig. 2E). Further experiments are required to dissect the mechanisms underlying the Pcbd1 loss-of-function phenotype and the role of Pcbd2 in its development.

Recent observations have suggested that a gradual loss of transcription factors Pdx1, Nkx6.1, and MafA, crucial for early pancreas development, leads to a destabilized adult β-cell state and an exhaustion of function, possibly contributing to the pathogenesis of T2D (22). The HNF1 family of transcription factors also controls both aspects. HNF1B is required to set up the early pancreatic transcriptional program (21), and HNF1A maintains a proper transcriptional network in mature β-cells (23). Thus, their binding partner PCBD1 might have different effects on the pancreas at different time points. Finally, PCBD1 mutations probably cause HPA and diabetes by affecting not one particular pathway but several genetic, metabolic, and signaling programs in different tissues, as shown for HNF1A gene (24). Future studies will aim to fully understand how PCBD1 regulates β-cell functions and the mechanisms leading to diabetes.

In summary, we provide the first genetic evidence that PCBD1 mutations can cause early-onset monogenic diabetes. We recommend a monitoring of neonatal HPA patients and their relatives with PCBD1 mutations in puberty and later in life for an occurrence of diabetes. We suggest screening HNF1A-like diabetes cases without mutations in HNF1A and HNF4A genes for recessive alterations in PCBD1, as they can be treated with oral antidiabetes drugs. Larger numbers of patients with PCBD1-diabetes will be needed to determine how commonly insulin can be replaced with sulfonylureas, as larger studies of individuals with HNF1A, KCNJ11, and ABCC8 defects show that this treatment is successful in most but not all the cases.

Acknowledgments. The authors thank the patients and their families. The authors also thank Nadine Wittstruck, Sabine Schmidt, and Mathias Gerhard (Cardiovascular and Metabolic Diseases, Max-Delbrück Center for Molecular Medicine, Berlin, Germany) for superb technical support; Dr. Patrick Ferreira (Division of Medical Genetics, Alberta Children’s Hospital, Calgary, Canada); Dr. Roger Germann (Children’s Hospital, Klinikum Karlsruhe, Karlsruhe, Germany); and Dr. Johannis B.C. de Klerk (Department of Pediatrics, Sophia Children’s Hospital, Erasmus Medical Center, Rotterdam, the Netherlands) for supporting the BIODEF recall and providing additional family information; and Friedrich C. Luft (Experimental and Clinical Research Center, Berlin, Germany) and Russel Hodge (Max-Delbrück Center for Molecular Medicine) for reviewing the manuscript.

Funding. The Experimental and Clinical Research Center supports K.R. The Deutsche Forschungsgemeinschaft supported M.G. (DFG: GO 1990/1-1). F.M.S. is supported by the Helmholtz Association (VH-NG-425), ERC-2009-St. Grant HEPATOPANCREATIC (243045), and FP7 IRG-ENDOPANC (239534). This work was funded (in part) by the Helmholtz Alliance ICEMED—Imaging and Curing Environmental Metabolic Diseases, through the Initiative and Networking Fund of the Helmholtz Association.

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

Author Contributions. D.S. designed, planned, and carried out experiments; analyzed the data; and wrote the manuscript. J.K. planned and carried out mouse and Xenopus experiments and analyzed the data. M.G. planned experiments, discussed the data, and reviewed the manuscript. F.R. performed linkage analysis and discussed the data. S.J. performed mouse cell culture experiments. P.A., K.B., C.E., P.K., and G.F.H. provided clinical data and contributed to the discussion. N.B. planned and carried out the recall of patients and their families from the BIODEF database, provided clinical data, and reviewed and edited the manuscript. F.M.S. designed mouse and Xenopus experiments, discussed the data, and reviewed and edited the manuscript. N.H. designed the study, discussed the data, and reviewed and edited the manuscript. K.R. designed the study, followed up the patients, performed clinical examinations, discussed the data, and reviewed and edited the manuscript. K.R. 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.