Neonatal Diabetes: An Expanding List of Genes Allows for Improved Diagnosis and Treatment

Abstract

There has been major progress in recent years uncovering the genetic causes of diabetes presenting in the first year of life. Twenty genes have been identified to date. The most common causes accounting for the majority of cases are mutations in the genes encoding the two subunits of the ATP-sensitive potassium channel (K_ATP), KCNJ11 and ABCC8, and the insulin gene (INS), as well as abnormalities in chromosome 6q24. Patients with activating mutations in KCNJ11 and ABCC8 can be treated with oral sulfonylureas in lieu of insulin injections. This compelling example of personalized genetic medicine leading to improved glucose regulation and quality of life may—with continued research—be repeated for other forms of neonatal diabetes in the future.

Introduction

Diabetes mellitus is an etiologically heterogeneous disorder. The most common forms, type 1 and type 2 diabetes, are multifactorial with variation in a number of genes (polygenic) together with nongenetic factors leading to disease. However, 1% to 2% of all cases result from mutations in a single gene (monogenic) that are sufficient to cause diabetes. In particular, diabetes occurring under 6 months of age—usually termed neonatal diabetes—appears to be predominantly monogenic and the fraction of cases without a known cause is diminishing rapidly as new genes are discovered (Table 1). Because many of these monogenic causes have also been found to cause diabetes after 6 months of age, we prefer the term congenital diabetes, which emphasizes the genetic nature of this group of disorders rather than the age of onset; however, for consistency with existing literature, we continue the use of the term neonatal diabetes.

Table 1.

Monogenic causes of congenital diabetes known to occur within the first year of life

Gene	Protein/function	Phenotypes/ syndromes	Inheritance	Age of diabetes onset	Pancreas appearance/ exocrine function	Other features	References
PLAGL1   HYMAI   (6q24)	Overexpression of paternally expressed   genes PLAGL1 (zinc finger protein or   ZAC tumor suppressor) and HYMAI   (non-protein coding) within the   imprinted region of chromosome   6q24/unknown function	TNDM	UPD6 (40%; de novo,   nonrecurrent), paternal   duplication (40%, may be   inherited) or maternal   methylation defect (20%;   autosomal recessive; e.g.,   ZFP57)	Within days;   remission   within months;   relapse during   adolescence	Normal/   normal	Very SGA; macroglossia and/or   umbilical hernia often present;   other features may be seen in   those with HIL, especially ZFP57   mutations (see below)	[30]
ZFP57	Zinc finger protein 57/transcription   factor with a role in maintenance of   imprinted DNA methylation	TNDM	Autosomal recessive	Similar to 6q24	Normal/   normal	Very SGA; HlL (9/9); macroglossia   (6/9); variable developmental delay   (6/9); umbilical defect (3/9); CHD   (3/9); visual impairment (3/9);   epilepsy (2/9)	[31]
KCNJ11	Inward rectifier K(+) channel   (Kir6.2) subunit of KATP	PNDM (more often) or   TNDM (less often);   DEND	Spontaneous (80%) or   autosomal dominant	<6 months;   rarely   later	Normal/   normal	Often SGA; possible developmental   delay; usually responsive to   sulfonylurea therapy	[1–3]
ABCC8	Sulfonylurea receptor 1 (SUR1)   subunit of KATP	PNDM (less often) or   TNDM (more often);   DEND	Spontaneous (80%) or   autosomal dominant	<6 months;   rarely   later	Normal/   normal	Often SGA; usually responsive to   sulfonylurea therapy	[3, 5]
INS	Insulin/hormone	PNDM (more often)   or TNDM (rarely)	Spontaneous (80%),   autosomal dominant   or recessive (rarely)	<6 months; less   often later	Normal/   normal	Often SGA; rare later-onset patients   with a MODY or antibody-negative   phenotype	[10–15,   22–25,   27, 28]
EIF2AK3	Eukaryotic translation initiation factor   2-α kinase 3 (EIF2AK3)/kinase   involved in regulation of translation	Wolcott-Rallison   syndrome (WRS)	Autosomal recessive	Most cases:   within weeks   (2–28 weeks);   1 case:   30 months	Rare   hypoplasia/   often   reduced   (25%)	Mild SGA or normal, rarely very SGA;   epiphyseal dysplasia (90% to 100%);   acute liver failure (60% to 75%);   developmental delay (60% to 80%);   hypothyroidism (~ 25%); exocrine   pancreatic dysfunction (~ 25%)	[89, 90,   92, 93]
FOXP3	Forkhead box protein P3 (FoxP3)/   transcription factor	Immunodysregulation   polyendocrinopathy,   enteropathy, X-   linked   (IPEX) syndrome	X-linked recessive	Days-3.5 months	Normal/   normal	Only males affected; severe immune   dysregulation; chronic diarrhea   with villus atrophy (95%); pancreatic and   thyroid autoantibodies (75%);   thyroiditis (20%); eczema (50%);   anemia (30%); often die before 1 year	[88]
GCK	Glucokinase/glycolytic enzyme	PNDM; MODY2	Autosomal recessive   (PNDM); or autosomal   dominant (MODY2)	PNDM: days of   life; MODY2:   present from   birth but not   usually   detected   until later	Normal/   normal	PNDM: very SGA, 12/14   homozygous, 1/14 compound   heterozygous, 1/14 heterozygous   (unclear mechanism), parents have   impaired fasting glucose with   GCX-MODY (MODY2)	[94–98]
PDX1	Pancreas/duodenum homeobox   protein 1 (PDX1 or IPF1)/   transcription factor	PNDM with   pancreatic agenesis/   hypoplasia	Autosomal recessive	3/5 cases: within   days; 2/5 cases:   12–15 days	Absent (1/5),   small (3/5)   or normal   (l/5)/absent   (3/5) or   reduced   (2/5)	SGA; diarrhea; malnutrition; parents   have PDX1 MODY (MODY4)	[54–59,   60•, 61•]
PTF1A	Pancreas transcription factor 1,   subunit α (PTFlA)/bHLH   transcription factor	PNDM with cerebellar   and pancreatic   agenesis	Autosomal recessive	5 cases: within   days	Absent/   absent	Very SGA; cerebellar agenesis; flexion   contractures; poor subcutaneous fat;   optic nerve hypoplasia; detectable   C-peptide/insulin	[47–49]
NEUROD1	(NeuroDl or BETA2)/bHLH   transcription factor	PNDM with cerebellar   (but not pancreatic)   hypoplasia	Autosomal recessive	2 cases: by   2 months	Normal/   normal	SGA; severe cerebellar hypoplasia;   moderate to severe developmental   delay; sensorineural deafness;   visual impairment	[46•]
NEUROG3	(NeuroG3 or NGN3)/bHLH   transcription factor	PNDM with severe   congenital diarrhea	Autosomal recessive	2/5 cases: within   days; 2/5 cases:   by 9 years	Small/4/5   normal	Very SGA; severe intractable congenital   diarrhea unresponsive to pancreatic   enzyme replacement with absent   intestinal enteroendocrine cells	[42•, 43•,   44]
RFX6	DNA-binding protein (RFX6)/   winged-helix transcription factor	PNDM with intestinal   atresia, gall bladder   hypoplasia	Autosomal recessive	5 cases: within   days	Small/   normal	Very SGA; intestinal atresias; gall   bladder hypoplasia/aplasia; diarrhea	[37–39,   40•]
IER3IP1	Immediate early response 3   interacting protein 1 (IER3IP1)	PNDM with   microcephaly	Autosomal recessive	2 cases: birth-   2 months	Normal/   normal	Microcephaly with simplified gyral   pattern; severe infantile epileptic   encephalopathy	[41•]
HNF1B	Hepatocyte nuclear factor 1β   (HNF-1 β)/transcription factor	TNDM/PNDM;   (RCAD; MODY5)	Spontaneous or   autosomal dominant	2 cases: 15-   17 days	Hypoplastic/   reduced	Very SGA; renal abnormalities;   relapsing/remitting DM (RCAD:   renal cysts, urogenital abnormalities)	[62, 63]
GLIS3	Glioma-associated oncogene-similar   family zinc finger 3 (GLIS3)/   Kriippel-like transcription factor	Neonatal diabetes   with congenital   hypothyroidism   (NDH)	Autosomal recessive	8 cases: within   days	Small,   normal or   cystic/   normal or   reduced   (2/8)	SGA; congenital primary   hypothyroidism; glaucoma (4/8);   liver fibrosis (5/8); cystic kidney   disease (4/8); osteopenia (1/8);   deafness (1/8)	[50, 53]
PAX6	Paired box 6/paired box and   homeo domain box containing   transcription factor	PNDM with severe   microcephaly and   eye defects	Autosomal recessive	2 brothers: within   days (other   case: DM not   reported)	Normal/not   reported	Brain malformations; microcephaly;   micropthalmia (eye defects in   parents); panhypopituitarism	[65–67]
WFS1	Wolframin/membrane glycoprotein	Wolfram syndrome;   DIDMOAD	Autosomal recessive	Median 6 years   (3 weeks–   14 years)	Normal/   normal	Optic atrophy (earliest feature);   diabetes insipidus; deafness	[68, 69,   74, 75]
SLC19A2	Thiamine transporter 1/transports   thiamine across the plasma   membrane	TRMA syndrome	Autosomal recessive	12 cases: 2–   13 months;   Others: DM   later	Normal/   normal	TRMA; sensorineural deafness;   occasional CHD (conduction   defects); short stature	[76–80]
SLC2A2	GLUT2/facilitative glucose   transporter	Fanconi-Bickel   syndrome (FBS)	Autosomal recessive	1 case: 6 days;   Others: IGT or   DM in   infancy/   childhood	Normal/   normal	Hepatomegaly related to hepatorenal   glycogen accumulation; proximal   tubular nephropathy with   glucosuria and hypophosphatemic   rickets; glucose intolerance or   diabetes; galactosemia	[81–87]

bHLH basic helix-loop-helix; CHD congenital heart defect; DEND developmental delay, epilepsy, neonatal diabetes; DIDMOAD diabetes insipidus, diabetes mellitus, optic atrophy, and deafness; DM diabetes mellitus; HIL hypomethylation of multiple imprinted loci; IGT impaired glucose tolerance; K_ATP ATP-sensitive potassium channel; MODY maturity-onset diabetes of the young; PNDM permanent neonatal diabetes; RCAD renal cysts and diabetes; SGA small for gestational age; TNDM transient neonatal diabetes; TRMA thiamine-responsive megaloblastic anemia

The neonatal diabetes phenotype encapsulates numerous subtypes, wherein most etiologies involve a severe disruption in β-cell function. Neonatal diabetes can be permanent and require lifelong treatment, or may be transient, in which case the diabetes may spontaneously remit (or be so mild as not to require treatment), but will often relapse, usually during adolescence. Although the majority of cases of neonatal diabetes involve isolated diabetes, many of the known monogenic causes are characterized by a variety of syndromic features. Increased attention to the primarily genetic nature of early-onset diabetes has resulted in an expanding list of causal genes, as well as an expansion of phenotypic characteristics in syndromic forms. This list includes some genes that more often cause diabetes that is diagnosed beyond the neonatal period, thus supporting the consideration of all monogenic diabetes gene causes in the differential diagnosis of early-onset diabetes, guided when possible by other clinical features (Table 2).

Table 2.

Clinical features associated with multiple monogenic causes of neonatal diabetes

Clinical feature	Genes to consider testing
Neurodevelopmental disability	KCNJ11, ABCC8, EIF2AK3, GLIS3, NEUROD1, PTF1A, PAX6, IER3IP1
Diarrhea and/or exocrine pancreatic insufficiency	NEUROG3, FOXP3, PDX1, PTF1A, RFX6, GLIS3
Thyroid dysfunction	GLIS3, FOXP3, EIF2AK3
Transient or relapsing/remitting diabetes	6q24 abnormalities, ZFP57, KCNJ11, ABCC8, INS, HNF1B, SLC2A2, SLC19A2
Family history of diabetes	Both parents or in their families: GCK, PDX1, NEUROD1, PTF1A, RFX6, WFS1, INS    (recessive mutations upstream of coding region or deletions)
	One parent (neonatal or adult onset): ABCC8, KCNJ11, INS, HNF1B, GCK, 6q24    duplications (paternally inherited)
Kidney structural or functional defects	HNF1B (structural anomalies and/or cysts), GLIS3 (cysts), EIF2AK3 (acute renal failure),    SLC2A2 (tubular dysfunction with glucosuria and phosphaturia)
Liver dysfunction	EIF2AK3 (episodic liver failure), SLC2A2 (hepatomegaly without liver failure), RFX6    (gall bladder hypoplasia with intestinal atresia and malformations), GLIS3 (liver fibrosis    in some cases)
Skeletal abnormalities	EIF2AK3 (epiphyseal dysplasia will be apparent radiographically if not clinically), GLIS3    (osteopenia with elevated alkaline phosphatase)
Visual impairment	PAX6 (aniridia, microphthalmia, also in parents), NEUROD1, WFS1 (optic atrophy outside    neonatal period), PTF1A (optic nerve hypoplasia), GLIS3 (congenital glaucoma)
Deafness	WS1, SLC19A2, NEUROD1, GLIS3
Megaloblastic anemia	SLC19A2

We briefly review here the known genetic causes of diabetes occurring in the first year of life. In addition, we focus on recent reports to highlight newer aspects of the most common forms of neonatal diabetes caused by mutations in KCNJ11, INS, and ABCC8, as well as some general considerations in the diagnosis, treatment, and prognosis of these disorders.

KCNJ11/ABCC8: Neonatal Diabetes due to Activating Mutations of the K_ATP Channel

Nearly half of all cases of permanent neonatal diabetes are due to activating mutations in KCNJ11 or ABCC8; their diabetes can be remarkably well controlled by oral sulfonylurea tablets instead of injected insulin [1, 2]. These discoveries have led to a renewed interest in neonatal diabetes, not only in the scientific community, but also among individual clinicians, families, and even the general public, who have been struck by the compelling stories told about lives changed by a remarkable simplification of treatment that no longer requires multiple daily insulin injections. Although this realization of the promise of personalized medicine may not affect a large number of patients, it has had a large impact on the relatively small number of patients concerned. In most such cases, transition from insulin to sulfonylurea therapy is best done during an inpatient admission using published guidelines [2], with advice sought from recognized centers with extensive experience. Although the resulting treatment is clearly easier and the level of diabetes control better, many questions still remain to be clarified:

1. How sustainable will treatment with oral sulfonylureas be, and will age, obesity, or other factors eventually require supplemental insulin or other medications? In this regard, a single case has been fortuitously treated for over 50 years [1]. Furthermore, our cases [3] and those reported by others [4] all continue to do well and give reason for optimism, but only time and careful tracking of long-term outcome will tell.

2. How often will patients have hypoglycemia, and what happens to their blood sugar levels during illness, procedures, or hospitalization, especially if oral medications cannot be taken? The incidence of hypoglycemia seems to be remarkably low despite the high levels of sulfonylureas that stimulate closure of the ATP-sensitive potassium channel (K_ATP) channel. Only one case has been reported to have had an episode of severe hypoglycemia, which has also been very rare among our patients [5]. Glucose levels must be monitored carefully in these patients during illness or when not eating, because blood glucose levels will be difficult to predict. Anecdotally, subjects within our Registry have variably reported mild hypoglycemia or hyperglycemia during illness, although none has reported requiring supplemental insulin (Greeley et al., Unpublished data).

3. To what extent will neurodevelopmental effects be improved or even prevented by sulfonylurea treatment? In this regard a notable study recently confirmed that mice with brain expression of a mutant form of the protein encoded by KCNJ11 (Kir6.2) have a neurologic pheno-type reminiscent of patients with the same mutation (V59M) [6•]. These results support anecdotal evidence that pharmacologic blockade may improve neurodevelopmental outcomes but will likely depend on passage of the drug across the blood-brain barrier. Our experience has been in line with that of others, suggesting a modest improvement in neurologic sequelae after treatment with sulfonylureas; however, long-term studies using standardized measures are needed to substantiate the presumed benefit, especially in those starting treatment at very early ages. Such potential benefit should be considered in individuals who continue to require insulin despite sulfonylurea treatment.

Such questions and many others are being addressed through efforts such as our Neonatal Diabetes Registry (http://NeonatalDiabetes.org) to track long-term outcomes in as many patients as possible [7]. Over 70 subjects from North America are now included within the Registry who have mutations in KCNJ11 or ABCC8, and more are added each month. Rare cases (usually older patients) who have not had complete success with sulfonylureas have shown benefit from other oral agents, such as dipeptidyl peptidase 4 inhibitors [3]; nevertheless, in the absence of a randomized controlled trial it will be difficult to make firm recommendations regarding their use in the treatment of this form of diabetes. The observation that older patients may not as easily achieve insulin independence is corroborated by mouse models suggesting a loss of β-cell mass over time [8, 9]. Thus, even if the cells that remain are able to regain full function, there may not be enough β cells in some patients to achieve insulin independence. Further research is needed to establish the dependence upon age or other factors, such as the particular mutation involved, in determining the likelihood of sulfonylurea responsiveness. Fortunately, the majority of the patients with activating mutations in KCNJ11 and ABCC8, regardless of age at genetic diagnosis and years of insulin therapy, can be switched to oral agents.

INS: Diabetes Caused by Mutations in the Insulin Gene

Heterozygous autosomal dominantly inherited mutations in the insulin gene (INS) are the second most common cause of permanent neonatal diabetes (after mutations in KCNJ11), with diagnosis of diabetes sometimes occurring after 6 months of age [10–15]. The original suggested mechanism involving β-cell endoplasmic reticulum stress due to misfolding of the protein encoded by the mutated allele has been upheld in experiments utilizing in vitro cell expression as well as mouse models [16–21]. Further study of populations diagnosed outside the neonatal period has revealed that INS mutations may also be found in patients with a diagnosis of type 1B (antibody-negative) diabetes [13]. Although these studies consistently revealed that approximately 10% of patients diagnosed clinically as having type 1 diabetes are autoantibody-negative, they varied in the incidence of INS mutations in these patients: 2/7, 2/25, and 0/24 [22–24]. Although the overall proportion of INS mutations in those clinically diagnosed with type 1 diabetes was quite low (4/812 or 0.5%), they represented a significant proportion of the antibody-negative subset (7%), further justifying antibody testing as a routine part of clinical diabetes diagnosis and management. It is unclear whether INS mutations ever cause antibody-positive type 1 diabetes.

Similar studies have suggested that INS mutations may also be a rare cause of maturity-onset diabetes of the young (MODY) [13, 18, 25]. The location of mutations causing later-onset diabetes in the signal peptide of preproinsulin is distinct from the location of those causing neonatal diabetes (often of cysteine residues involved in proinsulin folding), and suggests that different pathophysiologic mechanisms may be involved; however, other mutations causing later-onset diabetes are in similar locations as those causing neonatal diabetes but may cause more mild dysfunction [26].

Recessively acting mutations that impair INS gene expression through various mechanisms also cause permanent or transient neonatal diabetes [27]. Finally, large deletions in INS were recently described to cause neonatal diabetes characterized by complete insulin deficiency [28]. Of note, a high rate of adult-onset diabetes not associated with obesity was seen in family members carrying one copy of an INS deletion [28]. Another interesting observation is that children who inherit INS mutations from their fathers seem to have an earlier onset of diabetes than those who inherited such mutations from their mothers, which the authors suggest may arise from preferential silencing of maternally inherited mutant allele by imprinting effects at the INS-IGF2 locus [29].

6q24: Transient Neonatal Diabetes Related to Overexpression of Imprinted Genes

Transient neonatal diabetes characterized by severe intrauterine growth restriction and diagnosis of diabetes within days of life is most often related to overexpression of associated paternally imprinted genes at chromosome 6q24 [30]. This is due to uniparental paternal disomy, a paternally inherited duplication, or a maternal methylation defect, such as from recessive mutations in ZFP57 [31]. Regardless of mechanism, these cases exhibit diabetes that resolves spontaneously within a few months, only to return later in life (usually adolescence). The treatment of these patients has typically relied on insulin, but some suggestion of a benefit of oral sulfonylureas has been made [32].

In contrast, transient forms of neonatal diabetes may also be caused by mildly activating mutations in ABCC8 and KCNJ11, but in these cases the clinical course is often less predictable and characterized by relapsing/remitting diabetes that may require treatment intermittently through childhood [33–35]. In many of these families, parents or other relatives have adult-onset diabetes associated with the same mutation but any hypothetical neonatal hyperglycemia was probably not assessed soon after birth. Furthermore, some patients within the same family have been described to have transient or permanent neonatal diabetes, suggesting the presence of other factors (genetic or environmental) that can modify the diabetes phenotype. They could also be considered a form of MODY and recent studies would support the notion of screening for ABCC8 mutations in patients with a mild adult-onset or antibody-negative diabetes [24, 36]. As described elsewhere, very rare mutations in HNF1B or recessive mutations in the promoter/enhancer region of INS may also cause a phenotype that could be described as transient neonatal diabetes. However, 6q24-related diabetes seems to be most consistently characterized by a long remission phase, whereas other genetic causes seem to involve inconsistent episodes during which treatment may not be required.

Recessive and/or Syndromic Causes of Congenital Diabetes

RFX6: Syndrome of Diabetes, Intestinal Atresia, Gall Bladder Hypoplasia, and Diarrhea

Rare patients have been described with a syndrome characterized by diabetes within the first few days of life: pancreatic hypoplasia, intestinal atresia, gall bladder agenesis/hypoplasia, and congenital diarrhea [37–39]. Five out of six such cases were subsequently reported to carry homozygous mutations in RFX6, highlighting the role of this transcription factor as a key regulator of β-cell differentiation [40•]. Human cases for which pathology specimens were available were strikingly similar to knockout mice in having chromogranin A-positive islet cells, but negative staining for insulin, glucagon, and somatostatin. Despite relative hypoplasia of the small-appearing pancreata, pancreatic exocrine function appeared to be intact in these patients, and their intractable diarrhea was unresponsive to pancreatic enzyme replacement.

IER3IP1: Syndrome of Neonatal Diabetes with Microcephaly and Infantile Seizures

Infants in two unrelated consanguineous families exhibited a similar phenotype of neonatal diabetes with simplified gyral pattern microcephaly and severe infantile-onset epileptic encephalopathy. Linkage analysis led to sequencing of IER3IP1, revealing two different missense mutations at conserved positions predicted to be damaging [41•]. Pathology specimens of one case revealed widespread apoptotic neurons within a very small brain and greatly reduced numbers of insulin-positive pancreatic islet cells. In vitro studies suggested a role for IER3IP1 in protecting cells from stress-induced apoptosis.

NEUROG3: Syndrome of Intractable Diarrhea from Birth with Early-Onset Diabetes

A total of five cases with diabetes and chronic intractable malabsorptive diarrhea that started soon after birth have now been reported to have recessive mutations in NEUROG3 [42•, 43•]. Two recently reported cases had moderate hyperglycemia starting soon after birth that was treated with insulin at 20 days and 5 months of age. These two carried nonsense mutations whereas three cases originally reported with isolated diarrhea due to missense mutations later also developed insulin-requiring diabetes by 9 years of age [44]. These cases confirm a key role for NEUROG3 not only in β cells, but also in the development and function of enteroendocrine cells, which appeared completely absent in biopsy specimens from these patients, similar to NEUROG3 −/− mice. Importantly, the pancreas in all human cases appeared normal by imaging and only one case exhibited any evidence of exocrine pancreatic dysfunction, with the severe diarrhea exhibiting only limited improvement with pancreatic enzyme replacement.

NEUROD1: Syndrome of Diabetes with Cerebellar Hypoplasia without Pancreatic Exocrine Dysfunction

NEUROD1 is a basic helix-loop-helix (bHLH) transcription factor that is highly expressed in both developing and mature β cells; very rare cases of MODY have been attributed to heterozygous mutations in this gene [45]. More recently, two small for gestational age cases with neonatal diabetes by 2 months of age were described to exhibit cerebellar hypoplasia, developmental delay, sensorineural deafness, and visual impairment. They were found to carry two different homozygous frameshift mutations resulting in a truncated NeuroD1 protein lacking the activation domain [46•]. These patients were reported to have a normally sized pancreas with no evidence for pancreatic exocrine dysfunction and were not described to have gastrointestinal problems.

PTF1A: Syndrome of Diabetes with Cerebellar and Pancreatic Hypoplasia with Exocrine Dysfunction

PTF1A encodes a bHLH transcription factor that is essential for specification of pancreatic endocrine, exocrine, and ductal cells. Five cases in two families were found to carry two different homozygous truncating mutations in PTF1A causing a syndrome of congenital diabetes involving flexion contractures of arms and legs, paucity of subcutaneous fat and optic nerve hypoplasia, in addition to complete agenesis of the cerebellum and complete absence of the pancreas, a finding which was confirmed by extensive postmortem macro- and microscopic examination [47, 48]. Despite this, circulating C-peptide and insulin levels were low but detectable; however, the location of the insulin-producing cells was not determined. Those with NEUROD1 mutations also have cerebellar agenesis but have not been reported to have pancreatic agenesis as in those with PTF1A mutations, who also exhibit clinically apparent pancreatic exocrine dysfunction. Another case from Turkey was recently reported with this very rare recessive syndrome due to PTF1A mutations and also had detectable C-peptide and insulin levels during severe hyperglycemia [49].

GLIS3: Syndrome of Diabetes and Congenital Hypothyroidism

Six individuals in three families with a syndrome of neonatal diabetes within the first few days of life, low birth weight, mild facial dysmorphism, and congenital primary hypothyroidism were found to have homozygous abnormalities of GLIS3, encoding a Krüppel-like zinc finger protein [50]. Of note, the one family with a frameshift insertion mutation also exhibited liver fibrosis and polycystic kidneys, the latter of which is also seen in mouse models [51, 52]. The two other families carried large homozygous deletions upstream of GLIS3 that abrogated pancreatic but not kidney expression, and these patients did not have renal cysts. Recently, a second report described an additional two cases within two families carrying two different large homozygous deletions that included the first few exons of GLIS3, although the parents in one family were not known to be related [53]. Interestingly, both of these cases displayed a severe phenotype similar to those with missense mutations in the original report, including liver fibrosis and renal cysts. In addition, one of these cases exhibited osteopenia and deafness, and both were found to have pancreatic exocrine insufficiency that responded to enzyme replacement therapy.

PDX1: Syndrome of Congenital Diabetes with Pancreatic Hypoplasia and Exocrine Dysfunction

Homozygous mutations in PDX1 leading to pancreatic agenesis was the first discovered genetic cause of permanent neonatal diabetes, found through investigation of a proband whose parents were consanguineous and part of a very large pedigree exhibiting a high frequency of adult-onset diabetes/MODY [54–56]. Another proband carrying the same homozygous Pro63fsX60 mutation was recently described and may be related to the original family [57, 58]. The recent case underwent CT and multiple ultrasound examinations that were interpreted by the authors as consistent with the presence of only an underdeveloped head of the pancreas. In light of these findings, the ultrasound of the original case was also reinterpreted to have some pancreatic head tissue present as well. A separate case with a similar phenotype was found to have compound heterozygous missense mutations in PDX1 but the pancreas was described as absent [59]. All three of these cases had clinically apparent pancreatic exocrine insufficiency that responded well to enzyme replacement therapy. In contrast, two additional cases who were related to each other (first cousins) exhibited isolated neonatal diabetes without clinical evidence of exocrine dysfunction and were found to carry a novel homozygous hypomorphic mutation in PDX1 [60]. Biochemical testing revealed evidence for mild subclinical pancreatic exocrine dysfunction that did not require enzyme replacement. Ultrasound examinations revealed normal pancreatic tissue in one case and only the pancreatic head in the other case. It is worth mentioning that although PDX1 mutations are described as a cause of MODY4, sequencing of 910 subjects with type 2 diabetes and 878 controls revealed that 5% of both populations carried low-frequency PDX1 variants that were not associated with the diabetes phenotype, suggesting that mutations in PDX1 are a rare cause of diabetes [61•].

HNF1B: Neonatal Diabetes with Renal Anomalies

Only two cases of neonatal diabetes have been reported to be caused by heterozygous mutations in HNF1B, although such mutations, or large deletions, have long been described as a cause of later-onset diabetes with renal and/or genitourinary abnormalities (renal cysts and diabetes syndrome, RCAD, or MODY5). In the first report, a child who was very small for gestational age presented with severe diabetic ketoacidosis at 15 days of life but subsequently required insulin only intermittently until 6 years of age, after which she was insulin dependent [62]. Her younger brother had a normal birth weight but malformed kidneys necessitating transplant at 3 years of age, with only one transient episode of hyperglycemia during illness that resolved without treatment. Their clinically unaffected mother was found to carry the same HNF1B mutation (S148W) as her children at a low level, suggesting genetic mosaicism. The second report found mutation S148L in a case with very low birth weight requiring hospital admission with monitoring of blood glucose that became elevated only at 17 days of life and required insulin treatment for only 6 days [63]. Diabetes recurred at 8 years of age and was initially treated with tolbutamide but thereafter required insulin. He was also noted to have small dysplastic kidneys (without cysts or evidence of renal failure), pancreatic hypoplasia, and biochemical evidence of reduced pancreatic exocrine dysfunction. Interestingly, a cohort of five patients from two families with HNF1B mutations were all found to have mild pancreatic exocrine insufficiency and hypoplastic pancreata characterized by presence of only the head, with absence of any detectable body or tail [64]. In relation to these neonatal cases with a form of relapsing/remitting diabetes, it is interesting to note the incomplete penetrance of diabetes in family members known to carry an HNF1B deletion or mutation, in whom the most consistent feature is a renal anomaly, especially cysts.

PAX6: Syndrome of Neonatal Diabetes with Brain Malformations, Microcephaly, and Microphthalmia

Heterozygous mutations in PAX6, a paired domain-containing transcription factor involved in islet cell differentiation and function, have been described as a cause of ocular anomalies such as aniridia, and members of families carrying PAX6 mutations also exhibit impaired glucose tolerance and diabetes later in life [65]. In one study, mutation carriers in one large family were found to have high proinsulin/total insulin levels in response to glucose challenge even before the onset of diabetes, suggesting a mechanism corroborated in a similar mouse model related to the effect of PAX6 on proinsulin processing through regulation of PC1/3 expression [66]. Only two cases carrying biallelic mutations in PAX6 have been reported: the first case died at 1 week of life and exhibited severe brain malformations, microcephaly, and anopthalmia, without mention of hyperglycemia. Another surviving case had brain malformations, microcephaly, microphthalmia, and panhypopituitarism, along with neonatal-onset diabetes treated with insulin, despite normal appearing pancreas on MRI [67]. A brother of this case who died in infancy also had anophthalmia and neonatal diabetes.

WFS1: Syndrome of Diabetes with Optic Atrophy, Diabetes Insipidus, and/or Deafness

Several large case series have reported onset of diabetes as the earliest and most consistent feature of Wolfram syndrome, with subsequent development of optic atrophy, then later onset of diabetes insipidus and/or deafness (DIDMOAD syndrome). Recessive mutations in the large WFS1 gene were described as the cause of the neurodegenerative diabetes syndrome but such mutations are quite heterogeneous and do not clearly predict associated phenotypic features [68, 69]. In vitro studies and rodent models suggest that WFS1 downregulates endoplasmic reticulum stress and that a reduction in its activity leads to increased apoptosis in β cells and other tissues [70, 71]. Of note, some isolated features, such as diabetes, deafness, optic atrophy, and psychiatric problems, have been reported in heterozygous carriers of sequence variants, such as within families affected by Wolfram syndrome [72, 73]. Although diabetes is most often diagnosed in early childhood (median age, 4.3–6 years in European populations), several cases have been reported with diabetes recognized in the first year of life, as early as 3 weeks of age [74, 75].Thus, although it has not usually been described as a cause of neonatal diabetes, mutations in WFS1 should be considered in patients with early-onset diabetes mellitus, particularly if they develop optic atrophy, diabetes insipidus, and/or deafness.

SLC19A2: Diabetes as Part of Thiamine-Responsive Megaloblastic Anemia Syndrome

The gene SLC19A2 encodes a plasma membrane thiamine transporter, THTR1, and has been reported as the cause of thiamine-responsive megaloblastic anemia (Rogers syndrome), with diabetes often diagnosed well after the neonatal period. Although at least 10 previously reported cases with SLC19A2 mutations were diagnosed with early-onset diabetes between 2 and 12 months of age, it has not often been mentioned as a neonatal diabetes gene. Mutations in SLC19A2 should be considered in any patient with megaloblastic anemia and diabetes, or especially with the complete triad including sensorineural deafness. Interestingly, in addition to the anemia, the diabetes has also been described to exhibit a variable degree of improvement in response to thiamine treatment [76–79]. Although most reports describe the deafness as being irreversible, one case treated before 2 months had apparent amelioration of deafness, further highlighting the importance of an early genetic diagnosis to guide appropriate treatment [80].

SLC2A2/GLUT2: A Rare Cause of Early-Onset Diabetes as Part of Fanconi-Bickel Syndrome

Fanconi-Bickel syndrome (FBS) was originally described in 1949, and was later found to be due to homozygous or compound heterozygous mutations in the gene SLC2A2, encoding the facilitative glucose transporter, GLUT2 [81– 83]. FBS involves hepatomegaly related to hepatic and renal glycogen accumulation; renal proximal tubular dysfunction characterized by glucosuria, as well as phosphate wasting often leading to hypophosphatemic rickets; delay of puberty and short stature; hypergalactosemia; and mild fasting hypoglycemia but postprandial hyperglycemia and diabetes or impaired glucose tolerance [84]. As such, the elevated glucose levels have in some cases been detected in patients well under 1 year of age and FBS should thus be considered in the differential diagnosis of neonatal diabetes when any characteristic features are present [85]. The characteristic galactosemia may also be detected during the neonatal period by newborn screening programs, and when other features are present, genetic testing of SLC2A2 may lead to an early diagnosis [86, 87].

FOXP3: Immunodysregulation, Polyendocrinopathy, Enteropathy, X-Linked (IPEX) Syndrome

Mutations in the X-linked gene FOXP3 are a rare cause of neonatal monogenic autoimmune diabetes, along with numerous other features including enteropathy causing severe diarrhea and malnutrition, severe eczema, and autoimmune thyroid disease. Although patients with the classically described syndrome have a severe clinical course usually resulting in death within the first few years of life without stem cell transplant, other patients have a milder phenotype and a few have lived into childhood or beyond [88]. This phenotypic variability may be related to the specific mutations or other unknown factors and suggests consideration of FOXP3 sequencing in cases that may not fit all of the features of the immunodysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome.

EIF2AK3: Syndrome of Diabetes with Epiphyseal Dysplasia and Episodic Liver or Renal Dysfunction

Wolcott-Rallison syndrome (WRS) was originally described in three siblings in 1972, and later was found to be due to recessive mutations in EIF2AK3 [89, 90]. This gene encodes a translation-regulating kinase present in many tissues that plays an important role in trafficking of proinsulin through the secretory pathway in β cells [91]. The clinical details of past cases were recently reviewed [92], in addition to a report of the largest series of 29 cases within 25 families [93]. Although most cases are homozygous, compound heterozygous cases have been described and consanguinity may not be known. Thus, because epiphyseal dysplasia (the most consistent feature of WRS) may not be apparent clinically, radiographs could be considered to help guide genetic testing in any case of neonatal diabetes without a known cause.

GCK: Isolated Congenital Diabetes due to Recessive Mutations

Recessive mutations in the gene encoding the glycolytic enzyme glucokinase (GCK) were the second described cause of permanent neonatal diabetes [94]. A total of 14 cases have been reported and none were noted to have any evidence for pancreatic hypoplasia or exocrine deficiency, nor any extrapancreatic features, aside from being very small for gestational age (only one case 10th percentile, all others <3rd percentile). One recently reported case was highly unusual in carrying only one heterozygous mutation, with no deletions detected, nor any mutations in other neonatal diabetes genes [95]. Despite the typical mild presentation of family members from her large pedigree, she presented with severe hyperglycemia and ketoacidosis on the 2nd day of life and has since had hemoglobin A_1c levels greater than 8% despite replacement doses of insulin (0.6–0.8 U/kg/day). The mechanism for the severity of phenotype in this case remains unclear but could involve mosaic inactivation of the normal allele in β cells. Of note, three of the cases have been reported to exhibit partial responsiveness to repaglinide [96] or the sulfonylurea glibenclamide (glyburide, doses 0.85–1 mg/kg/day) [97, 98], suggesting that sulfonylurea inhibition of the K_ATP channel may allow for insulin secretion despite the lack of glucose responsiveness related to their GCK deficiency. Although mutations in GCK are a rare cause of neonatal diabetes, it should be considered in cases with isolated diabetes, especially if consanguinity is suspected and/or family members have a GCK-MODY (MODY2) phenotype.

General Considerations regarding Diagnosis and Etiology of Neonatal Diabetes

Cost-Effectiveness of Genetic Testing in K_ATP-Related Neonatal Diabetes

We recently published an analysis of the cost-effectiveness of genetic testing in patients diagnosed with diabetes under 6 months of age. Given the high likelihood of KCNJ11 or ABCC8 mutations in this population, and the resulting improvements in diabetes control, quality of life, and lower costs of treatment, we found that there would be significant cost-savings resulting from a policy of testing compared with a policy of no genetic testing in this population of patients [99••]. This is a highly unusual situation for cost-effectiveness analyses, in that not only does this advance in medical diagnosis and treatment result in an improved outcome, but it does so at a decreased cost. Furthermore, a threshold analysis allowed us to estimate that for the testing policy to remain cost-effective for any population being tested, the prevalence of KCNJ11 or ABCC8 mutations would need be at least 0.3%. Since a few such cases have now been reported to have been diagnosed between 6 and 12 months of age, it is quite possible that testing this group may also be appropriate [15, 100]. Because such cases are often selected for being negative for diabetes-related autoantibodies, further research is required to more definitively determine the prevalence of mutations in later age groups, especially as a few cases have been reported to have positive antibodies [3, 101]. Inevitable decrease in the cost of genetic testing will also favor a broadening of the screening for mutations in these genes.

The Future of Genetic Testing in Neonatal Diabetes

Given the long list of genes now known to cause congenital diabetes, it has become increasingly time consuming, labor intensive, and expensive to sequence all possible genes using traditional methods. Even sequencing the three most commonly mutated genes is costly, given that ABCC8 is very large, with 39 exons. Furthermore, the choice of genes to be tested using this approach relies on the availability of reliable and comprehensive phenotypic information, although such features may be subclinical (Tables 1 and 2). At least 60% to 75% of probands diagnosed under 6 months of age will have an identifiable genetic cause using the standard approach, with this proportion being influenced by factors such as the frequency of consanguinity within the population being tested and whether all possible genes were considered even in the absence of clear syndromic features [3, 11, 15]. In this regard, next-generation sequencing and informatics approaches are likely to quickly become the standard for screening for sequence variants in any number of genes. We have successfully begun this approach, and others have also demonstrated that standard commercial platforms for whole-exome sequencing have sufficient coverage of the genes of interest for neonatal diabetes [102•].

In using high-throughput sequencing as a screening tool, it will be important to utilize informatics strategies that will minimize the chance that any possible causative variant could be missed. Although doing so may increase the likelihood of false-positive results, any variants found should still be confirmed by traditional methods. In addition to being a more cost-effective and efficient approach for screening of known genes, it is quite likely that new genetic causes for patients with unknown etiologies will result, particularly with the advent of affordable whole-genome sequencing. As the cost of these techniques lowers, our cost-effectiveness study would support the expansion of testing into other populations of diabetes patients who may be much less likely to have mutations that are treatable with sulfonylureas but may still benefit from uncovering a genetic diagnosis.

Conclusions

Mutations in 20 genes are now known to cause diabetes presenting in the first year of life. However, we and others have been able to find a genetic cause in only 60% to 75% of patients with permanent neonatal diabetes diagnosed under 6 months, suggesting that continuing research will identify new genes. Anyone diagnosed under 6 months of age should have comprehensive genetic testing guided by clinical features (Tables 1 and 2); however, the optimal cost-effective algorithm for consideration of testing in those diagnosed after 6 months of age remains unclear but should be considered in any case with negative antibody testing, syndromic features, or a strong family history of diabetes. Concurrent with the expansion in the number of known genes has been an extension of the known phenotypic spectrum of various syndromic etiologies as well as overlap between genes associated with neonatal diabetes and MODY. Decreasing costs of DNA sequencing will allow genetic testing to become routine in the diagnosis of neonatal diabetes. Furthermore, future advances in understanding the pathophysiology of other causes of congenital diabetes will hopefully lead to more effective cause-specific treatments.