Update of variants identified in the pancreatic beta-cell KATP channel genes KCNJ11 and ABCC8 in individuals with congenital hyperinsulinism and diabetes.

The most common genetic cause of neonatal diabetes and hyperinsulinism are pathogenic variants in ABCC8 and KCNJ11. These genes encode the subunits of the beta-cell ATP sensitive potassium channel, a key component of the glucose-stimulated insulin secretion pathway. Mutations in the two genes cause dysregulated insulin secretion; inactivating mutations cause an over-secretion of insulin leading to congenital hyperinsulinism, whilst activating mutations cause the opposing phenotype, diabetes. This review focuses on variants identified in ABCC8 and KCNJ11, the phenotypic spectrum and the treatment implications for individuals with pathogenic variants. This article is protected by copyright. All rights reserved.

Identifying these mutations is important for informing prognosis, medical management, and recurrence risk.
Over recent years, the number of variants identified in these two genes has expanded tremendously. In 2006, 124 disease-causing mutations were reported, which increased to 265 pathogenic variants 3 years later Gloyn, Siddiqui, & Ellard, 2006). By combining published reports together with data from five international molecular genetic screening laboratories in the UK, Denmark, France, and the United States of America, we now report 953 pathogenic ABCC8 and KCNJ11 variants (Tables S1-S6) and discuss the role of these genes in congenital hyperinsulinism (CHI) and monogenic diabetes.

| CONGENITAL HYPERINSULINISM
CHI is characterized by the inappropriate secretion of insulin despite low blood glucose, which can result in irreversible brain damage if not promptly treated (Helleskov et al., 2017). The condition has a variable phenotype usually presenting during the neonatal period or infancy with seizures and/or coma and a large birth weight due to high levels of insulin acting as a growth factor in utero.
Although most cases of CHI are sporadic, rare familial forms have been well documented. Sporadic CHI has an estimated incidence of between 1 in 27,000 and 1 in 50,000 live births (Glaser, Thornton, Otonkoski, & Junien, 2000;Otonkoski et al., 1999).
However, in some isolated populations or in countries with high rates of consanguineous unions, the incidence is higher (i.e., 1 in 2,675 to 1 in 3,200; Mathew et al., 1988;Otonkoski et al., 1999).

| CHI due to K ATP channel mutations
Loss-of-function ABCC8 mutations were first described in 1995 (Thomas et al., 1995). These mutations either prevent trafficking of the channel to the membrane surface or are associated with channels that reach the surface but are not fully responsive to MgADP activation (Figure 1; Ashcroft, 2005;Nichols et al., 1996;Taschenberger et al., 2002). The majority of ABCC8 loss-offunction mutations are recessively acting with a small number of dominant missense mutations reported that produce channels that traffic to the membrane but have impaired mgADP activation.
Fewer loss-of-function mutations have been reported in KCNJ11 in keeping with the gene being much smaller (1173 vs. 4749 bases, respectively; Thomas et al., 1996). Similar to ABCC8, both dominant and recessively acting KCNJ11 mutations have been described (Pinney et al., 2013). Mutations in these two genes together account for 36-70% of CHI cases (Kapoor et al., 2013;Snider et al., 2013).
There exist mouse models for K ATP channel CHI; however, their inability to fully recapitulate the human phenotype means that they have a limited value for studying specific disease mechanisms. For example, mice generated with a deletion of ABCC8 or KCNJ11, or the homozygous recessive KCNJ11 mutation p.(Tyr12Ter), do not have the sustained neonatal hypoglycemia observed in humans with homozygous null mutations. Instead the blood glucose levels normalize in the mouse within a few days of birth with glucose intolerance developing in later life (Hugill, Shimomura, Ashcroft, & Cox, 2010;Miki et al., 1998;Seghers, Nakazaki, DeMayo, Aguilar-Bryan, & Bryan, 2000). The differences in the phenotype between mice and humans are not fully understood, but they highlight the need to develop human-specific models for studying disease mechanisms.

| Clinical management of K ATP channel CHI
In 2015, the Pediatric Endocrine Society published recommendations for the evaluation and management of persistent hypoglycemia in neonates, infants, and children (Thornton et al., 2015). The main treatment for CHI is the K ATP channel-opener diazoxide; however, patients with ABCC8/KCNJ11 mutations that prevent trafficking to the membrane do not respond to the drug as diazoxide targets the SUR1 subunit of the K ATP channel. For approximately 50% of patients with mutations that do not prevent the channel from reaching the membrane, diazoxide is an effective treatment (Boodhansingh et al., 2019). For patients with diazoxide-unresponsive CHI, secondline treatment with somatostatin analogs may be helpful to control hypoglycemia; however, adverse effects on somatostatin analogs, and likewise diazoxide, have been reported (Demirbilek et al., 2014;Herrera et al., 2018).
The mode of inheritance of the K ATP channel mutation determines the pancreatic histological subtype (de Lonlay et al., 1997;de Lonlay et al., 2002;Jack, Walker, Thomsett, Cotterill, & Bell, 2000;Rahier et al., 1984). Inheritance of two recessively acting or one dominant ABCC8/KCNJ11 mutation results in diffuse disease affecting the entire pancreas. Focal disease is caused by somatic loss of the maternal chromosome 11p15.5 region by uniparental disomy that unmasks a paternally inherited K ATP channel mutation at 11p15.1. These focal lesions often appear histologically as small regions of islet adenomatosis that develop as a result of the imbalanced expression of maternally imprinted tumor suppressor genes H19 and p57 Kip2 , and the increased expression of the paternally derived insulin-like growth 886 | DE FRANCO ET AL. factor II gene (Craigie et al., 2018;Damaj et al., 2008;de Lonlay et al., 1997). Rarely, giant focal lesions have been described where virtually the whole of the pancreas is affected (Ismail et al., 2012). Atypical mosaic disease has also been reported in a small number of cases (Han et al., 2017;Houghton et al., 2019;Hussain et al., 2008;Sempoux et al., 2011).
The identification of a single recessively acting K ATP channel mutation in an individual with CHI predicts focal disease with 84-97% sensitivity, with a positive predictive value up to 94% (Mohnike et al., 2014;Snider et al., 2013). 18 F-DOPA PET/CT scanning can identify and localize a focal lesion before surgery (Otonkoski et al., 2006). Intraoperative ultrasound may further aid the surgeon to perform tissue-sparing pancreatic resection in focal CHI, which is potentially curative (Bendix et al., 2018).

| DIABETES MELLITUS
Diabetes is the opposing disorder to CHI and results from hyperrather than hypoglycemia. Current estimates suggest that approximately 0.4% of all diabetes (and up to 3.5% of those diagnosed under 30 years of age) has a monogenic cause (Shepherd et al., 2016;Shields et al., 2017). Individuals diagnosed with monogenic diabetes outside of infancy are generally classified as having maturity onset diabetes of the young, whereas neonatal diabetes (NDM) describes congenital diabetes. In individuals with NDM, impaired insulin secretion results in a low birth weight and hyperglycemia diagnosed before the age of 6 months (Hattersley & Ashcroft, 2005). The minimal incidence of NDM has been calculated to be between 1 in 89,000 and 1 in 160,949 live births (Grulich-Henn et al., 2010;Wiedemann et al., 2010). 3.1 | Later-onset diabetes due to K ATP channel mutations Dominantly acting mutations in the K ATP channel genes have been rarely described in individuals with later-onset diabetes in the absence of documented hyper-or hypoglycemia in the neonatal period (Bowman et al., 2012;Hartemann-Heurtier et al., 2009;Huopio et al., 2003;Koufakis et al., 2019;Tarasov et al., 2008). The mechanism(s) leading to this variable penetrance are not fully understood and may differ according to whether the mutation is causing a gain or loss of channel function. Interestingly, in one study, the generation of a mouse model harboring a homozygous dominantly acting loss-of-function ABCC8 mutation p.(Glu1507Lys) recapitulated the biphasic phenotype with the mice having increased insulin secretion in early life and reduced insulin secretion later on. This was shown to be resulting from a reduction in insulin content rather than a reduction of islet number and/or size. Heterozygosity for the mutation did, however, not result in a phenotype in the mouse, further highlighting differences between the mouse models and human disease (Shimomura et al., 2013).

| Neonatal diabetes due to K ATP channel mutations
Strong support for the role of gain-of-function K ATP channel mutations in the etiology of diabetes came from the observation that mice overexpressing a mutant K ATP channel with reduced ATP sensitivity developed diabetes within 2 days (Koster, Marshall, Ensor, Corbett, & Nichols, 2000). In 2004, the first heterozygous activating KCNJ11 mutations causing NDM were described in humans with activating ABCC8 mutations reported 2 years later (Babenko et al., 2006;Gloyn, Pearson, et al., 2004;Proks et al., 2006). Mutations in these two genes together have now been shown to account for approximately 40% of NDM cases (De Franco et al., 2015;Stoy et al., 2008).
Both dominant and recessive activating mutations are frequently identified in ABCC8. Conversely for KCNJ11, all, but one of the mutations reported so far, p.(Gly324Arg), have been dominantly acting.
The majority (~60%) of dominant mutations arise "de novo," so there is often no family history of diabetes; however, germline mosaicism has been observed in some families (Edghill et al., 2007;Gloyn, Cummings, et al., 2004).
There is added complexity associated with ABCC8 mutations, as compound heterozygosity for both an activating and an inactivating mutation can cause diabetes . Furthermore, a recessively inherited ABCC8 nonsense variant has been reported in two cases with NDM, which leads to the deletion of the in-frame exon 17 likely resulting in enhanced sensitivity of the channel to intracellular MgADP/ATP (Flanagan et al., 2017).
The specific K ATP channel mutation identified determines whether the diabetes will cause permanent or transient NDM (Gloyn, Reimann, et al., 2005;Patch, Flanagan, Boustred, Hattersley, & Ellard, 2007). Variable penetrance within families with mutations leading to transient diabetes is observed with some individuals being diagnosed with diabetes at birth, yet others developing diabetes for the first time in adulthood (see previous section on adult-onset diabetes; Flanagan, Edghill, Gloyn, Ellard, & Hattersley, 2006).

| Spectrum of central nervous system features in K ATP channel NDM
Central nervous system (CNS) features are frequently reported in individuals with K ATP channel NDM due to the Kir6.2 and SUR1 proteins being expressed in the brain (Karschin, Ecke, Ashcroft, & Karschin, 1997;Liss, Bruns, & Roeper, 1999;Sakura, Ammala, Smith, Gribble, & Ashcroft, 1995;Schmahmann & Sherman, 1998). The most severe neurological phenotype is termed as developmental delay, epilepsy and neonatal diabetes (DEND) syndrome, which includes muscle weakness and hypotonia (Hattersley & Ashcroft, 2005). Intermediate DEND (iDEND) syndrome is diagnosed when epilepsy is absent or presents after the age of 12 months (Gloyn, Diatloff, et al. 2006 Since these initial reports, studies on larger cohorts of individuals affected with K ATP channel NDM have characterized the neurological features in more detail. Additional features reported include autism and attention deficit hyperactivity disorder (ADHD), anxiety and sleep disorders, dyspraxia, and learning difficulties, resulting in impaired attention, memory, visuospatial abilities, and executive function (Beltrand et al., 2015;Bowman et al., 2016;Bowman et al., 2017;Bowman, Day, et al., 2018;Busiah et al., 2013;Landmeier, Lanning, Carmody, Greeley, & Msall, 2017). More important, it is now recognized that some degree of impairment can be detected on neuropsychological testing in the majority of patients with K ATP channel mutations, even if there is no obvious CNS involvement (Busiah et al., 2013;Carmody et al., 2016).

| Clinical management of neonatal diabetes and CNS features due to K ATP channel mutations
The identification of a K ATP channel mutation can have an impact on the medical management of patients with NDM as approximately 90% can transfer from insulin injections to high-dose sulphonylurea tablets (Pearson et al., 2006;Zung, Glaser, Nimri, & Zadik, 2004).
Sulphonylureas bind to the SUR1 subunit of the K ATP channel and close it independently of ATP, resulting in excellent long-term glycemic control and improved quality of life for affected patients and their families (Babenko et al., 2006;Bowman, Sulen et al., 2018;Rafiq et al., 2008). Patients who are unable to transfer to sulphonylureas tend to have a longer duration of diabetes before attempting transfer or functionally severe mutations (Babiker et al., 2016;Thurber et al., 2015). Few side effects and no episodes of severe hypoglycemia involving seizures or loss of consciousness have been reported in individuals with sulphonylurea-treated neonatal diabetes (Bowman, Sulen, et al., 2018;Codner, Flanagan, Ellard, Garcia, & Hattersley, 2005;Kumaraguru et al., 2009;Lanning et al., 2018).
Sulphonylureas can improve the neurological features in people with K ATP channel NDM, particularly in the first year of treatment (Beltrand et al., 2015;Fendler et al., 2013;Stoy et al., 2008). However, these features do not fully resolve after sulphonylurea therapy and persist for a long term into adulthood (Bowman, Day et al., 2018;Bowmen, Sulen, et al., 2018). Higher doses of sulphonylureas are recommended for patients with severe neurological features in an attempt to mitigate this (https://www.diabetesgenes.org/). In addition, starting sulphonylurea therapy as early as possible after a genetic diagnosis is crucial as the largest improvements appear to occur in younger patients (Beltrand et al., 2015;Shah, Spruyt, Kragie, Greeley, & Msall, 2012).

| GENETIC VARIATION IN ABCC8 AND KCNJ11
KCNJ11 (MIM# 600937) is located 4.5Kb from ABCC8 on chromosome 11p15.1 and has a single exon encoding for the 390-amino acid Kir6.2 protein (GenBank NM_000525.3). ABCC8 consists of 39 exons that encode for the 1,582 amino acids of SUR1 (NM_001287174.1; MIM# 600509). This gene has an alternatively spliced recognition site at the 5′ end of exon 17, which results in two different transcripts differing in length by a single amino acid (GenBank AH003589.2). This alternative splicing has led to discrepancies in the literature for nomenclature of variants present in 17-39, which differ by a single amino acid depending on the isoform used (1581 amino acids, NM_000249.3 and 1582 amino acids, NM_001287174.1). For the purpose of this review, we have described ABCC8 variants according to the longer isoform (NM_001287174.1).

| Disease-causing variants
A total of 748 ABCC8 and 205 KCNJ11 pathogenic or likely pathogenic variants have been identified in individuals with CHI or NDM (Table 1 and Table 3 and Tables S1 and S4)please note that these tables are meant to direct to the appropriate references and laboratories. They do not provide in-depth clinical information and variants that had been previously reported as pathogenic with a GnomAD frequency compatible with the disease frequency (as calculated by http://cardiodb.org/allelefrequencyapp/ using a biallelic mode of inheritance, a prevalence of 1/50,000, an allelic heterogeneity of 0.1, genetic heterogeneity of 0.5, and penetrance of 0.5) were not re-assessed.
Founder mutations have been identified in many populations with the best recognized example being the ABCC8 p.(Phe1388del) and c.3992-9G>A mutations present in >90% of cases from the Ashkenazi Jewish population Otonkoski et al., 1999). In the Irish population, a deep intronic ABCC8 founder mutation at position c.1333-1013G>A has been described that generates a cryptic splice site and causes pseudoexon activation . Founder mutations have also been reported in Hispanic (Aguilar-Bryan & Bryan, 1999), Bedouin (Tornovsky et al., 2004), Spanish (Fernandez-Marmiesse et al., 2006), Finnish (Otonkoski et al., 1999), and Turkish populations .

| Common variation in ABCC8 and KCNJ11
Three hundred and sixty-eight benign/likely benign variants and variants of uncertain significance have been observed in both genes (Tables 2-4, S2, S3, S5, and S6). Two common variants in linkage disequilibrium, p.(Glu23Lys) in KCNJ11 and p.(Ser1370Ala) in ABCC8, predispose to type 2 diabetes (Florez et al., 2004). Although their effect size is small (odds ratio~1.2), given that 58% of the population carry at least one lysine allele at residue 23 in KCNJ11, this equates to a sizeable population risk (Gloyn, Weedon, et al., 2003;Nielsen et al., 2003).

| Variant interpretation
Given the highly polymorphic nature of ABCC8 and KCNJ11, the occurrence of both activating and inactivating mutations, the multiple modes of inheritance of disease, and the variable penetrance associated with dominantly acting mutations, interpreting variants identified in these genes can be extremely challenging. Although the identification of a null ABCC8 or KCNJ11 variant(s) in an individual with CHI provides strong evidence for pathogenicity, finding a missense variant is insufficient to assign disease causality and, as such, additional support is required to achieve a "pathogenic" classification according to the guidelines set out by the American College of Medical Genetics (Richards et al., 2015).
Large variant databases such as GnomAD and LOVD are powerful tools that aid in variant interpretation and allow for reclassification of variants (Fokkema et al., 2011;Lek et al., 2016). As such, some variants previously reported as pathogenic in the literature have now been found to be too common to be causative of disease and have now be reassigned as a variant of uncertain significance or a benign variant (Tables S2, S3, S5, and S6).  | 899