Introduction

Congenital hyperinsulinism (HI) is the most common cause of persistent hypoglycemiain infants and children, which if unrecognized may lead to development delays andpermanent neurologic damage. In general, the high risk of brain damageappears to be due to delays in diagnosis and treatment rather than a consequence ofthe genetic defects and, thus, is potentially preventable [1].

Within the last two decades, major advances have been made in understanding themolecular and genetic basis of HI. This work has helped elucidate the physiology ofbeta cell function and insulin regulation, as well as advanced clinical care forchildren with HI. Among the nine known genetic causes of HI, mutations in the genesencoding the ATP-sensitive potassium channel represent the most common defectaccounting for the majority of the cases [2]. At least 50% of the children carrying these mutations have a focal formof the disease and can be cured by surgery, thanks to the introduction of novelimaging techniques. In this article, we will review the genetic basis of HI with afocus on the recently identified genetic causes and mechanisms of disease, describethe recommended diagnostic work-up and discuss recent advances in the management ofthis challenging disorder.

Molecular genetics

Nine genes expressed in the ß-cell have been implicated in thepathophysiology of HI [Figure 1. They includeABCC8 and KCNJ11 encoding SUR-1 and Kir6.2, the twosubunits of the ATP-sensitive potassium channel (KATP channel);GLUD1 encoding glutamate dehydrogenase (GDH); GCK encodingglucokinase (GK); HADH encoding short-chain 3-hydroxyacyl-CoA (SCHAD);UCP2 encoding uncoupling protein 2 (UCP2); HNF4A andHNF1A encoding the transcription factors hepato-cyte nuclear factor4-alpha (HNF-4α) and hepatocyte nuclear factor 1-alpha (HNF-1α),respectively; and ectopic expression of monocarboxylate transporter 1 (MCT-1)encoded by SLC16A1. Many of these genes are also involved in thepathogenesis of monogenic diabetes, including ABCC8, KCNJ11,GCK, HNF4A, and HNF1A. There remain approximately50% of diazoxide-responsive cases and 10% of diazoxide-unresponsive cases of HIwith unknown genetic etiology, suggesting that additional, yet to be identifiedgenes are implicated in the pathogenesis of HI [3].

Figure 1
figure 1

Genetic defects in the beta cell leading to hyperinsulinism. Inthe pancreatic beta cell, ATP production from fuel metabolism leads toinhibition and closure of ATP-sensitive potassium channels, whichtriggers membrane depolarization and opening of voltage-dependentcalcium channels. The resulting increase in cytosolic calcium triggersinsulin secretion. Defects in this pathway can result inhyperinsulinism. The known protein defects are depicted in bold italics.Five are inactivating mutations: SUR-1 (sulfonylurea receptor), Kir6.2(potassium channel), SCHAD (short-chain 3-OH acyl-CoA dehydrogenase),UCP2 (uncoupling protein 2), HNF-4α (hepatic nuclear transcriptionfactor 4α), and HNF-1α (hepatic nuclear transcription factor1α). The last 2 are transcription factors and are not depicted inthe figure. Three are activating mutations: GK (glucokinase), GDH(glutamate dehydrogenase), MCT-1 (monocarboxylate transporter 1).Positive effects are shown by a plus arrow; negative effects by a minusarrow. Dashed arrows denote multiple steps in a pathway. G6P =glucose-6-phosphate, ATP = adenosine triphosphate, ADP = adenosinediphosphate.

Full size image

KATP-hyperinsulinism

Inactivating mutations in the genes encoding the two subunits of theß-cell ATP-sensitive potassium channel (KATP channel),ABCC8 and KCNJ11 (encoding SUR-1 and Kir6.2,respectively), cause the most common and severe form of hyperinsulinism,although mutations in ABCC8 are more common [4, 5]. The effect of these mutations on channel expression and functiondetermines the clinical phenotype, particularly the response to diazoxide, aKATP channel opener used in the treatment of HI [2]. Thus, KATPHI can been classified into three subtypes:(1) recessive diazoxide-unresponsive, (2) dominant diazoxide-unresponsiveand (3) dominant diazoxide-responsive [6–9]. Clinically, the first two groups are undistinguishable. Thesechildren present as neonates with large for gestational age birth weightsand severe hypoglycemia that requires high glucose infusion rates and failsto respond to therapy with diazoxide, because there are no functionalKATP channels (recessive defects) or their function isseverely impaired (dominant defects) [8]. Therefore, children with recessive or dominantdiazoxide-unresponsive KATPHI, frequently require pancreatectomyto control their hypoglycemia. The third group (dominantdiazoxide-responsive) typically has less severe hypoglycemia, which may notbe noted at birth.

The pathophysiology of KATPHI is characterized by a failure tosuppress insulin secretion as glucose concentration falls, manifesting assevere fasting hypoglycemia, and a failure to increase insulin secretion inresponse to a glucose load [2]. In contrast to the impaired glucose-stimulated insulinsecretion, amino acids trigger insulin release in some individuals withKATPHI, causing severe protein-induced hypoglycemia [10].

In the 1980s, it was recognized that some patients with severehyperinsulinism were cured after a partial pancreatectomy. On histologicinspection, although the majority of their pancreas appeared normal, a focal“tumor-like” area of abnormal beta cell proliferation wasidentified. In contrast, children who were not cured with surgery had anormal number of beta cells, but these cells showed signs of hyperactivitythroughout the pancreas [11]. These findings lead to the recognition of two distincthistologic forms of HI: diffuse and focal. The molecular mechanismresponsible for these focal adenomatous lesions in which abnormal beta cellproliferation occurs in a discrete region of the pancreas was laterdescribed [12]. The pathophysiology of focal KATPHI includes a“two hits” mechanism: first, a paternally inherited recessivemutation in ABCC8 or KCNJ11; and second, a deletion of thematernally inherited 11p15 chromosomal region, compensated by paternaluniparental disomy [13]. The loss of maternally expressed genes involved in tumorsuppression explains the histological findings of focal KATPHI [12, 14].

Clinically, children with the focal form are indistinguishable from thosewith the recessive diffuse form, presenting with severe hypoglycemia andhigh glucose requirements. Unlike the diffuse form, focal KATPHImay be cured with surgical resection of the discrete lesion. Thus, therecognition of these cases prior to surgery is critical as explainedlater.

Recently, a novel “atypical” form of focal hyperinsulinismconsisting of morphologic mosaicism of pancreatic islets has been reported,which similar to focal lesions, involves only a portion of the pancreas andmay be cured with partial pancreatectomy [15]. The histology in this form shows co-existence of two abnormalislet types (large islets with occasional enlarged nuclei and shrunkenislets with small nuclei) in a limited region of the pancreas. No mutationsin ABCC8, KCNJ11 or GCK were identified in thesecases and the molecular mechanism is unknown at this time.

GDH-hyperinsulinism

The second most common form of HI is due to activating mutations of glutamatedehydrogenase (GDH), encoded by GLUD1, leading to thehyperinsulinism/hyperammonemia (HI/HA) syndrome [16]. In the beta cell, GDH is involved in amino acid-stimulatedinsulin secretion and loss of inhibitory control of GDH in HI/HA leads todysregulated insulin secretion. The majority of mutations in GLUD1occur de novo (70%) with the reminder inherited in an autosomal dominantpattern [17]. Individuals with HI/HA have fasting and protein-inducedhypoglycemia, which is easily controlled with diazoxide. Ammonia levels aretypically elevated 3–5 times the normal range, but these individualsdo not exhibit the classical symptoms associated with hyperammonia due toother causes. Children with HI/HA have increased rates of seizures, mostcommonly atypical absence, and learning disabilities [18]. These neurologic abnormalities appear to be unrelated tohypoglycemia or elevated ammonia levels.

GK-hyperinsulinism

Activating mutations in glucokinase (GK) cause an autosomal dominant form ofHI. Glucokinase, encoded by GCK, is the “glucosesensor” of the beta cell, triggering insulin secretion in response torising glucose concentration [19]. Activating mutations lower the threshold for insulin release;thus, the glucose set point for these individuals is lower. GK-HI presentswith fasting hypoglycemia of variable degrees of severity and, although theinitial reported case responded well to diazoxide, less than a third of thereported cases have been treated successfully with diazoxide [2, 20].

SCHAD-hyperinsulinism

Short-chain 3-hydroxyacyl-CoA (SCHAD) deficiency leads to an autosomalrecessive form of HI. SCHAD, encoded by HADH, catalyzes a step inthe fatty acid oxidation (FAO) cycle [21]. Although FAO defects are well known to cause hypoglycemia, theconnection between an enzyme in the FAO cycle and HI was unclear.Subsequently, it was shown that SCHAD is an inhibitory regulator of GDH, theenzyme involved in amino-acid stimulated insulin secretion, and loss ofGDH’s inhibition due to SCHAD deficiency results in insulindysregulation [22]. Children with SCHAD-HI have fasting and protein-inducedhypoglycemia and similar to patients with HI/HA, they respond well todiazoxide therapy. Biochemical markers of SCHAD-HI include increasedconcentration of 3-hydroxybutyrylcarnitine in plasma and 3-hydroxyglutaricacid in urine. These children do not exhibit the cardiac, skeletal orhepatic dysfunction associated with FAO disorders.

UCP2-Hyperinsulinism

Uncoupling protein 2 (UCP2), a membranous mitochondrial carrier, acts as anegative regulator of insulin secretion in the beta cells. Recently, loss offunction mutations in UCP2 have been described that result inhyperinsulinism [23]. UCP2 mutations are inherited in an autosomal dominant manner andhave been identified in children responding to diazoxide. The reported caseshave resolution of HI by 7 months to 6 years of age, which suggests thatUCP2-HI is a transient disorder [2].

HNF4A and HNF1A-hyperinsulinism

Hepatocyte nuclear factor 4-alpha (HNF-4α), a transcription factorinvolved in pancreatic development and function, has been classically linkedto a monogenic form of early-onset diabetes, MODY1 [24]. Mutations in HNF4A, which encodes HNF-4α, have anautosomal dominant pattern of inheritance. A study of families with MODY1found that carriers of HNF4A mutations were born macrosomic and 8carriers had transient neonatal hypoglycemia with hyperinsulinism identifiedin 3 of the 8 [25]. Further studies demonstrated HNF4A mutations inchildren presenting with diazoxide-responsive HI [26]. Patients with HNF4A mutations respond well to diazoxideand the HI resolves within the first year of life in the majority of cases,although cases of persistent HI caused by HNF4A mutations have beenreported [27]. The phenotype in these cases is complex and may involve theliver and kidney [28]. Recently, mutations in the transcription factor, hepatocytenuclear factor 1-alpha (HNF-1α), encoded by HNF1A and known tocause MODY3, have been shown to also present with hyperinsulinism in infancy [28]. The mechanism by which loss of function mutations inHNF4A and HNF1A can lead to this dual phenotype withhypoglycemia in early life and diabetes later, has not been elucidated, butlikely implies a changing pattern of gene expression regulation by thesetranscription factors throughout the life of an individual.

MCT1-Hyperinsulinism

Aberrant expression of monocarboxylate transporter 1 (MCT-1) leads to a veryrare and unusual form of hyperinsulinism, which is triggered by exercise.Identified in two Finnish families, exercise-induced hyperinsulinism (EIHI)is characterized by episodes of hypoglycemia associated with elevatedinsulin levels at the time of anaerobic exercise [29]. Autosomal dominant mutations in the regulatory regions of theSLC16A1 gene, which encodes MCT-1, have been identified [30]. In normal individuals, MCT-1, a transporter of pyruvate andlactate, is not expressed on beta cells, but in EIHI, mutations in theregulatory regions of SLC16A1 lead to expression of MCT-1 on betacells. The presence of MCT-1 allows pyruvate, elevated during anaerobicexercise, to enter the beta cell and through the triggering pathway(KATP-mediated), increase insulin release resulting inhypoglycemia [31]. The degree of hypoglycemia associated with exercise is variableand is only partially responsive to diazoxide.

Diagnosis

For children presenting with hypoglycemia (plasma glucose < 70 mg/dL), promptdiagnosis and establishment of effective treatment is essential to avoidneurologic sequelae. Clinical clues to the diagnosis of the HI include large forgestational age birth weight and severe, persistent hypoglycemia requiring highglucose infusion rates (> 10 mg/kg/min). However, the clinical phenotype ofhyperinsulinism is a spectrum and infants with HI can also present with normalbirth weights and lower glucose requirements.

The diagnosis of hyperinsulinism is made based on the critical sample obtainedduring a spontaneous or provoked hypoglycemic event. The threshold blood glucoseto obtain the critical sample by convention is set low at < 50 mg/dL todecrease the likelihood of false positive results. If a diagnostic fast isneeded to obtain the critical sample, close monitoring of blood glucose, vitalsigns and mental status is essential to ensure the patient’s safety.Parental dextrose, as well as all appropriate specimen collection tubes, shouldbe at the bedside prior to the start of the fast. Upon completion of diagnostictesting, blood glucoses should be monitored every 10–15 min until they arereliably above 70 mg/dL.

In addition to obtaining the critical sample, the glycemic response to glucagonshould be evaluated [Table 1. A detectable insulin levelis inappropriate at the time of hypoglycemia and is consistent with insulinexcess. A common pitfall in the diagnosis of HI is that insulin concentration isnot always elevated, even at the time of hypoglycemia, thus the diagnosis shouldbe based on other indicators of increased insulin action [32]. Laboratories consistent with excess insulin action includesuppressed beta-hydroxybutyrate and free fatty acid concentrations as well as aninappropriate glycemic response to glucagon of 30 mg/dL or more at the time ofhypoglycemia [33].

Table 1 Criteria for diagnosing hyperinsulinism
Full size table

In cases with overgrowth and failure to respond to diazoxide and octreotide,activating mutations in AKT2 should be considered in the differential diagnosis [34]. Beckwith-Wiedemann syndrome, neonatal panhypopituitarism andcongenital disorders of glycosylation should also be considered andappropriately evaluated if warranted by clinical features. Beckwith-Wiedemannsyndrome has significant clinical heterogeneity and is characterized byhemihypertrophy, macrosomia, macroglossia and predisposition to embryonaltumors. Neonates with panhypopituitarism may have diagnostic findings identicalto HI with suppressed ketones and free fatty acids and a glycemic response toglucagon. Clinical features suggestive of panhypopituitarism include midlinedefects and micropenis. Congenital disorders of glycosylation are a highlyvariable group of disorders caused by abnormal glycosylation of N-linkedoligosaccharides and hypoglycemia may be found with failure to thrive and liverdysfunction. Low growth hormone and cortisol at the time of hypoglycemia are notdiagnostic of growth hormone deficiency or adrenal insufficiency and theappropriate stimulation tests should be performed to confirm those diagnoses [35]. Insulinomas should be considered in the differential diagnosis ofchildren presenting with hyperinsulinemic hypoglycemia beyond infancy,particularly during the second decade of life.

A failure to respond to maximum dose of diazoxide (15 mg/kg/day) after at least 5days of treatment, suggests a KATP channel defect as the most likelycause of hyperinsulinism. Such children are potential surgical candidates andrequire referral to a specialized HI center with 18-fluoroL-3,4-dihydroxyphenylalanine positron emission tomography(18 F-DOPA PET) scan availability [Figure 2. Commercial genetic testing is available for the four most commonHI genes (ABCC8, KCNJ11, GLUD1 and GCK) andfor the HNFs defects. As a cost-reducing measure, genetic testing should betargeted based on the clinical phenotype; for example sending for GLUD1in children who are responsive to diazoxide and have elevated ammonias. Ourrecommendation is to send genetic testing as soon as possible for the child andhis/her parents, especially for the diazoxide-unresponsive cases as thedetection of a single recessive paternal KATP mutation(ABCC8 or KCNJ11) has a positive predictive value of 94%for focal hyperinsulinism [3].

Figure 2
figure 2

Algorithm for the treatment of hyperinsulinism. Assessing theresponse to diazoxide is a critical step in the management of HI.Patients who fail to respond to diazoxide will most likely haveKATP channel defect and require referral to a specializedcenter with 18 F DOPA PET scan capability. A safetyfast should be 8 to 18 h long depending on the age of the patient. Notethat octreotide is not recommended as pre-operative treatment inneonates with HI due to high rate of treatment failure and risk ofnecrotizing enterocolitis. KATP = ATP-sensitive potassiumchannel, 18 F DOPA PET = 18-fluoroL-3,4-dihydroxyphenylalanine positron emission tomography.

Full size image

Management

The therapeutic goal for hyperinsulinism as well as other hypoglycemic disordersis to achieve and maintain plasma glucoses greater than 70 mg/dL. In the 1960s,the antihypertensive, diazoxide with its known side effect of hyperglycemia, wasfirst used to treat hyperinsulinism [36]. The 1970s saw the introduction of octreotide as an HI treatment [37]. These drugs remain the mainstay of medical treatment for HI.Diazoxide acts to open the KATP channel, decreasing insulin secretionand is the first-line agent for HI, although most cases of KATPHI donot respond.

The therapeutic dose range of diazoxide is wide (5 to 15 mg/kg/day) and variesaccording to the severity of the phenotype. Patients with severe hypoglycemiaand high glucose requirements should be started on the maximum dose of diazoxideat 15 mg/kg/day. Patients with more mild disease can be started on doses of5–10 mg/kg/day, which should be increased if there is no response afterseveral days of treatment. The half-life of diazoxide in children is between9.5-24 h [38] and is unknown in neonates, leading to controversy as to whethertwice a day or three times a day dosing is appropriate. In general, we find thatfor diazoxide-responsive children, dosing twice daily is sufficient to provideappropriate control. To evaluate diazoxide’s efficacy after 5 days oftherapy, a safety fast should be performed with a duration lasting 8 to 18 hbased on the age of the patient. Continued hypoglycemia after at least 5 days ofthe maximum dose (15 mg/kg/day) is considered a treatment failure. The sideeffects of diazoxide include hypertrichosis and fluid retention. Hypertrichosisis often quite severe, but resolves after stopping the drug. Fluid retention,especially in neonates, may require the use of a diuretic, such aschlorothiazide, but stronger loop diuretics should be avoided.

The second-line agent, octreotide decreases insulin secretion throughhyperpolarization of the beta cells and inhibition of calcium channels.Octreotide is associated with frequent treatment failure due to the developmentof tachyphylaxis. More importantly, octreotide has recently been associated withthe occurrence of fatal necrotizing enterocolitis and therefore, should be usedwith caution in neonates [39]. Our center no longer recommends its use in neonates pre-operatively.It continues to be used post-operatively in children with diffuse disease whoremain hypoglycemic following subtotal pancreatectomy. Successful treatment withlong-acting formulations of octreotide has recently been reported [40, 41].

Glucagon can be used as a continuous intravenous infusion of 1 mg/day to lowerglucose infusion rate requirements in infants awaiting surgery. Trials ofglucagon as a subcutaneous infusion through a pump were largely unsuccessful dueto the drug’s lack of stability in solution [42, 43]. A new potential therapeutic approach to children withKATP defects involves inhibition of GLP-1 action, an incretinknown to increase insulin secretion and lower blood glucose. Recently, a GLP-1receptor antagonist, exendin-(9–39) has been shown to elevate fastingblood glucose in individuals with hyperinsulinism [44].

Surgical intervention is indicated in children who have a focal lesion that canbe cured with resection and in children with diffuse disease who fail medicaltherapy. Pursuing surgery in the latter group requires a careful considerationof risks and benefits. The benefit of surgery in this group is that theirhypoglycemia is often easier to manage following a pancreatectomy, but this mustbe weighed against the risks of a surgical procedure and long-termcomplications, such as diabetes. Those children with diffuse disease who havevery limited fasting tolerances (less than 2–3 h) and very high glucoseinfusion requirements will most likely require a pancreatectomy. However, somechildren with diffuse disease and longer fasting tolerances (6–8 h) may bemanaged with a combination of frequent feeds, enteral dextrose and/oroctreotide. The risks of this management approach include potentially morefrequent hypoglycemia and exposure to octreotide.

The largest advance in the management of children with HI over the past decadewas the introduction of imaging with 18 F-DOPA PET todifferentiate focal from diffuse disease and to localize focal lesions [45]. As discussed earlier, children with focal KATPHI can becured with surgical resection of the lesion. In contrast, for children withdiffuse HI, surgery is palliative. Differentiating focal from diffuse diseaseand accurately identifying the location of focal lesions in the pancreas iscrucial for ensuring that children with focal KATPHI are successfullycured. Conventional imaging techniques, such as CT or MRI, cannot identify focallesions and interventional radiology techniques, such as transhepatic portalvenous sampling or arterial calcium stimulation are highly invasive and havepoor accuracy for differentiating diffuse from focal HI and for localizing thefocal lesion [46].

The uptake of 18 F-DOPA identifies neuroendocrine tissue, whichtakes up amino acid precursors of dopamine, including DOPA. In diffuse HI, theuptake of the tracer is uniform throughout the pancreas; in contrast, a focallesion will have greater uptake in a specific region compared to the surroundingtissue [Figure 3. Since 2003, 18 F-DOPAPET scans have been used to differentiate diffuse from focal HI and to localizefocal lesions prior to surgery [47]. In the largest series to date of 50 patients who underwent18 F-DOPA PET scans, followed by surgery, the sensitivityfor diagnosing focal disease was 75% and the location of the focal was correctlyidentified in 100% of cases [45]. Similar results have been reported in smaller series [48, 49]. A meta-analysis in 2012 showed superiority of the18 F-DOPA PET scan compared to interventional radiologytechniques for diagnosing and localizing focal lesions [50].

Figure 3
figure 3

A. Frontal view of a 3D maximum intensity projection (MIP) 18-fluoroL-3,4-dihydroxyphenylalanine positron emission tomography (18F-DOPA PET) image demonstrating a focal lesion in the tail ofthe pancreas (arrow). B. A frontal view 3D MIP18 F-DOPA PET image fused with a contrast-enhanced CTshows a focal lesion in the pancreatic head (white arrow). C.Frontal view of a 3D MIP showing non-uniform pattern of uptake withincreased activity throughout the pancreas consistent with diffusedisease. Note the normal liver, kidney and bladder uptake.

Full size image

Accurate localization of the focal lesion aids in pre-operative planning and inselect cases (lesions on the anterior surface of the body and tail), allows forthe use of laparoscopic techniques [2]. Intraoperative biopsies and frozen section evaluation by experiencedpathologists allows for confirmation of a focal lesion and guides the extent ofpancreatic resection. Children with diffuse HI who fail medical management willrequire a subtotal pancreatectomy and placement of a gastrostomy tube to helpwith the post-operative management since most of these children continue to havehypoglycemia, although less severe [51].

Of the surgical cases performed at the Children’s Hospital of Philadelphia,95% of patients with focal disease have been cured and the majority requiredless than a 50% pancreatectomy. In contrast, the majority of patients withdiffuse disease post-operatively required continued intervention to maintaineuglycemia. However, after surgery, their HI can be more easily medicallymanaged. For children with continued hypoglycemia, octreotide during the daycombined with continuous intragastric dextrose overnight is effective atpreventing octreotide tachyphylaxis and allows for stable glucose control. Forthe smaller subset of patients with hyperglycemia following subtotalpancreatectomy, insulin may be necessary. The long-term risk of developingdiabetes in children with diffuse disease depends on the extent of pancreaticresection [52]. In the largest published series, 91% of children who had undergone anear-total pancreatectomy in infancy required insulin therapy for diabetes bythe age of 14 years [53].

Conclusions

Congenital hyperinsulinism is one of the most complicated and challenging disordersfaced by pediatric endocrinologists. The potential for preventing permanent braindamage caused by persistent hypoglycemia, makes it extremely important to identifyand treat these children early. The past two decades has seen tremendous progress inunderstanding the genetic and molecular basis of HI. This understanding has in turnlead to advancements in management and improved outcomes, particularly through theuse of 18 F-DOPA PET scan to identify and cure focal lesions.