Relative Expression of a Dominant Mutated ABCC8 Allele Determines the Clinical Manifestation of Congenital Hyperinsulinism

A 1-day-old female (III-2 [individuals are identified by Roman numerals indicating the generation and Arabic numerals indicating the specific individual]) (Fig. 1A), born large for gestational age (4.8 kg) to nonconsanguineous Palestinian parents, developed severe hyperinsulinemic (10.1 mU/L) hypoglycemia (20 mg/dL) without ketonuria (ammonia and lactate were normal). After intravenous glucose infusion (20 mg/kg/min), gastric tube feeding and somatostatin analog (Octreotide; Novartis, Basel, Switzerland) (dose 18 μg/kg/day) were necessary to achieve stable normoglycemia during 96 h of subcutaneous glucose monitoring (Minimed, Sylmar, CA) and thereafter. At age 2 years, she was gradually weaned off octreotide (<5 μg/kg/day), but minimal continuous cornstarch intake was still required. The mother’s sister (II-6) and paternal cousin (II-2) had similar, albeit milder, clinical presentations. Because a heterozygous six–base pair (bp) in-frame insertion mutation in ABCC8 exon 37 was found in the cousin (II-2) and her unaffected haploidentical brother (II-3) but not in their mother (I-1) (Fig. 1B and C), it was thought that patient II-2 had focal CHI as a result of paternal uniparental disomy or paternal mutation and LOH (8). Pancreatectomy was not clinically required, and the hypoglycemia resolved spontaneously.

Leukocyte DNA from family members was sequenced for the relevant segment. For the proband and her father, all KCNJ11 and ABCC8 coding exons and the ABCC8 promoter were sequenced as previously described (9,10).

The ABCC8 insertion mutation (CGGCTT) was introduced into hamster ABCC8 cDNA in the pECE plasmid using the QuikChange mutagenesis kit (Stratagene, Santa Clara, CA) and confirmed by DNA sequencing (11). Rat KCNJ11 cDNA is in the pCDNAI vector. COSm6 cells were transfected with recombinant SUR1 and Kir6.2 cDNAs using FuGene6 (Promega Corporation, Madison, WI).

Cell surface expression of channels was determined by a chemiluminescence assay, channel activity in intact cells in response to metabolic inhibition (30-min incubation in 2.5μg/mL oligomycin and 1 mmol/L 2-deoxy-d-glucose) was assessed using an ⁸⁶Rb⁺ efflux assay, and channel response to ATP, MgADP, and diazoxide were monitored using inside-out patch-clamp electrophysiological recordings as previously described (12). Statistical analysis was performed using an independent two-population two-tailed Student t test.

RNA samples (1 μg) extracted from two to three different blood samples of each family member using the Tempus Spin RNA Isolation Kit (Applied Biosystems, Foster City, CA) were reverse transcribed.

cDNA samples (and cDNA from normal human β-cells as control) were amplified at the ABCC8 gene region flanking the mutation (exon 37–39). Purified PCR products were cloned into pGEM vectors, and 20–30 different clones were sequenced for each original RNA sample (pGEM-T Vector System I kit; Promega). Each cDNA sample was also amplified with 6FAM-labeled primer, separated by high-resolution sequencing gel electrophoresis (the mutant was six bp longer), and analyzed using Biosystems Inc. sequencing analyzer.

After bisulfite treatment (13) of DNA samples, the KCNQ1 and ABCC8 promoter regions were PCR amplified. Purified products were cloned into pGEM vectors and sequenced.

cDNA samples of proband and parents were amplified using primers flanking a synonymous single nucleotide polymorphism rs1799857(C/T) in ABCC8 exon 12. PCR products were digested (MaeII) and separated based on the MaeII restriction site present only on the C allele. The two visible gel bands represent the T allele (186 bp) and the C allele (120 bp). Difference in band intensity was a measure of allelic expression.

Primer sequence and PCR conditions are provided in the Supplementary Data.

Increased birth weight, persistent hypoglycemia, and inappropriately high insulin levels suggested CHI. Screening of family members for the exon 37 insertion mutation c4525insCGGCTT, previously identified in an affected maternal cousin, found that the proband (III-2) (Fig. 1A), her affected aunt (II-6), her unaffected mother (II-7), and her reportedly unaffected maternal grandfather (I-3) are also heterozygous for this mutation. This mutation predicts an insertion of AlaSer residues after codon 1508 (Fig. 1B and C).

No other mutations were identified in ABCC8 or KCNJ11 exons (or exon/intron boundaries) or in the ABCC8 promoter in the proband or her father.

To test the mutation functionality, wild-type (WT) KCNJ11 and mutant ABCC8 were cotransfected in COSm6 cells. The resulting channels demonstrated nearly normal abundance of the mature and immature forms of SUR1 and ∼66 and 87% of normal cell membrane channel expression in the homozygous and heterozygous states, respectively (Fig. 2A and B). ⁸⁶Rb⁺ efflux assays showed rapid activation of WT channels, giving rise to >90% cumulative efflux after 40 min of incubation. By contrast, the 1508insAS mutant homozygously expressed in COSm6 cells showed no channel activity (Fig. 2C).

Inside-out patch-clamp recordings revealed that in contrast to the stimulatory effect seen in WT channels, the homozygote mutant channels and the simulated heterozygous channels completely lacked response to either MgADP or diazoxide, indicating a strong dominant-negative effect of the 1508insAS mutation (Fig. 2D and E).

The strong dominant-negative effect of the mutant SUR1 predicted that all carriers should be severely symptomatic. However, four heterozygous individuals were clinically asymptomatic. To address this unexpected finding, we tested the possibility that the mutant transcript may not be expressed at the same level as the WT transcript in all patients. Lymphocyte-derived cDNAs from five symptomatic and asymptomatic family members were amplified for the fragment of the ABCC8 gene flanking the mutation site (exon 37–39). PCR products were cloned. Sequencing of >20 clones from each individual showed that although the majority of the clones from the healthy mother (77%) and the healthy maternal grandfather (73%) carried the normal allele, the affected proband showed the opposite, with 66% of the clones being mutated (Fig. 3A). It is interesting that the affected aunt, who had a milder clinical presentation (survived without significant medical therapy) and has been in remission for 10 years with no hypoglycemia, currently expressed 66% normal cDNA alleles. To confirm this finding, the lymphocyte-derived cDNAs were amplified with a 6FAM-labeled primer. Because the mutation is an insertion of six bp, it could be separated by high-resolution sequencing gel electrophoresis. In three independent samples, withdrawn at 12, 18, and 24 months of age (while clinical improvement was taking place), the affected subject predominantly expressed the mutant mRNA (Fig. 3B). In contrast, her unaffected mother showed high relative expression of normal mRNA compared with the mutant mRNA. The aunt with the milder disease, and currently in clinical remission, showed an intermediate expression pattern, similar to the previous cloning studies. The father, who does not carry the mutation, expressed only normal mRNA as expected.

To exclude the possibility of an intronic or promoter mutation affecting expression of the paternally inherited allele, we amplified exon 12 of the parental and patient cDNA, which includes a heterozygous MaeII restriction enzyme polymorphic site. MaeII digestion (Fig. 3C) showed that the father transcribes both alleles at comparable levels. The patient and mother preferentially express one opposing allele (38 and 64%, respectively), confirming the findings of differential expression of the mutation itself.

In an effort to explain the differential allelic expression found in our patient and her unaffected family, we searched for epigenetic differences. DNA bisulfite methylation analysis of the KCNQ1 imprinting control region (ICR) and of the ABCC8 promoter was performed in lymphocytes of all five family members. The methylation patterns of the ABCC8 promoter region and the KCNQ1 ICR (50%) were normal, regardless of the patient’s mutation or disease status (Fig. 3D and E) (some data not shown).

The unusual finding of a severe, autosomal-dominant ABCC8 mutation causing CHI in only 50% of the carriers suggested that variable mutant allele expression may determine the disease phenotype. Mutations in the ABCC8 gene are responsible for 40–50% of CHI cases, focal or diffuse (4,14,15). Previous studies of dominant SUR1 mutations show that mutant channels were expressed on the cell surface, but their response to MgADP was defective; in the heterozygous state, channels had intermediate responses to MgADP (16–18). The mutated protein found in our family, with its predicted two amino acid insertion, is expressed on the cell surface, but the channel does not open in response to MgADP or diazoxide even in the heterozygous state. Thus, to our knowledge, the 1508insAS mutation is the first CHI mutation exhibiting a nearly complete dominant-negative gating effect. Because the 1510–11 AlaSer residues lie in the interface of the two nucleotide-binding domains (NBDs), duplication of these two residues may disrupt interactions between the NBDs and affect the octameric channel complex during MgADP stimulation.

Initially, the mother’s cousin was thought to have focal CHI because only one mutation was found and her haploidentical brother was asymptomatic. In a similar manner, the clinical findings in the aunt, with a paternally inherited mutation, were also consistent with focal CHI. In contrast, our proband, with severe CHI, inherited her mutated allele from her mother, which is not consistent with the two-hit mechanism proposed to cause focal CHI (19,20). Furthermore, three case subjects with focal CHI, usually described as a rare, sporadic event in one family, seemed extremely unlikely (21).

We hypothesized that the marked phenotypic variability in mutation-carrying family members was related to the relative expression of the mutant allele. Lymphoblastic mRNA was previously used to study the ratio between HNF1A mRNA levels and the risk of type 2 diabetes (22); however, lymphoblast cell lines tend to clonalyze over time, limiting their use in predicting human pathophysiology (23). Because no pancreatic tissue was available, and given the limitations of lymphoblast-derived studies, we amplified ABCC8 cDNA fragments derived from lymphocytic ABCC8 mRNA.

Although our patient consistently expressed mostly the mutant allele in independent mRNA samples, her asymptomatic mother and grandfather primarily expressed the normal transcript. Of interest, the more mildly affected aunt (II-6) (Fig. 1), currently in long remission, had an intermediate expression pattern, suggesting that differential expression of mutated and WT alleles may correlate with disease severity. Given the severe dominant-negative nature of this mutation, the hetero-octameric channel structure and the assumption that mutant and normal SUR1 protein are randomly recruited into the channels, dysfunction is already expected when one of the four SUR1 molecules is mutated (18). This means that if both the mutant and the normal transcript are equally expressed, the chance of having all four subunits made from the normal transcript is only one-sixteenth (6%). Therefore, if 70% (the normal mother) of the expressed transcript is normal, ∼25% of the channels are expected to function normally, enabling normal β-cell function. In marked contrast, if only 30% (our patient) of the transcripts are normal, then only ∼1% of the channels are expected to be normal, abolishing β-cell function. The intermediate expression of the mutant allele in patient II-6 also raises the intriguing possibility that the relative expression of the mutant allele may change over time, thus explaining the clinical remission seen in this patient and in patient II-2. Prospective studies are needed to fully address this question.

In an attempt to understand the mechanism causing interindividual variability in allele expression, we studied two possible epigenetic etiologies. Although ABCC8 gene’s promoter is known to be a CpG island and, thus, not expected to be regulated by methylation, given its de novo methylation in colon cancer (H. Cedar, unpublished observations), we tested its methylation pattern and that of the adjacent 11p15 ICR (Fig. 1D). Although LOH is thought to be a somatic event restricted to the pancreatic lesion, systemic aberrant methylation of the short arm of chromosome 11 is seen in Wilms tumor and Beckwith-Wiedemann syndrome (24). However, no difference in methylation pattern, either in chromosome 11p15 (ICR) or in the ABCC8 promoter region, was apparent in any of the family members tested.

Another possibility was that the mutation, located 15 bp from the 37 exon splice donor site, could be variably affected by splice-regulating factors and result in variable splicing and phenotype, as described for cystic fibrosis (25). Splice prediction programs (http://sroogle/tau/ac/il and RESearch) failed to predict any effect on splicing efficiency by the 1508insAS mutation.

In conclusion, we demonstrate that variability in transcript expression may determine the clinical phenotype of CHI. A mechanism that regulates the transcript expression of a mutant allele may protect from or enable a severe phenotype in subjects with the same genotype. It is important that the use of peripheral lymphocytes as a window to the β-cell transcription profile may help explain other unexpected β-cell phenotypes. It may be especially significant given the epigenetic regulation at the telomeric end of chromosome 11p, where several genes of glucose metabolism are encoded. Finally, the unique severe dominant-negative nature of the 1508insAS mutation may provide new insight into the structure-functional relationship of the SUR1 NBD regions on channel gating.

S.-L.S. is supported by National Institutes of Health Grants R01-DK-57699 and R01-DK-66485.

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

R.S., S.-L.S., B.G., and D.H.Z. researched data and wrote, reviewed, and edited the manuscript. C.A.Z. researched data and wrote the manuscript. E.L., Q.Z., J.G., and S.T.-B. researched data. D.H.Z. is the guarantor of this article.

Parts of this study were presented in abstract form at the 70th Scientific Sessions of the American Diabetes Association, Orlando, Florida, 25–29 June 2010.