Paradoxical Maturity-Onset Diabetes of the Young Arising From Loss-of-Function Mutations in ATP-Sensitive Potassium Channels

Abstract

Pancreatic β-cell ATP-sensitive K⁺ (K_ATP) channel closure underlies electrical excitability and insulin release, but loss or inhibition of K_ATP channels can lead to paradoxical crossover from hyperinsulinism plus hypoglycemia, to glucose intolerance or diabetes. We report genotype-phenotype information from a set of patients clinically diagnosed with maturity-onset diabetes of the young (MODY) and carrying coding variants in the K_ATP regulatory subunit gene ABCC8. In contrast to the naive prediction that diabetes should be associated with K_ATP gain-of-function (GOF; as in K_ATP-dependent neonatal diabetes), each mutation caused mild to severe loss-of-function (LOF), through distinct molecular mechanisms, suggesting the affected individuals may have crossed over to glucose intolerance from K_ATP channel LOF-dependent congenital hyperinsulinism. Our data provide definitive support for a paradoxical form of MODY associated with K_ATP channel LOF that is genetically and mechanistically distinct from a late diagnosis of neonatal diabetes resulting from K_ATP GOF. To avoid confusion and inappropriate treatment efforts, we argue that diabetes driven by K_ATP-GOF and K_ATP-LOF mutations should be officially recognized as distinct diseases.

Article Highlights

• Gain-of-function (GOF) ATP-sensitive K⁺ (K_ATP) mutations cause neonatal diabetes mellitus. K_ATP mutations are also associated with maturity-onset diabetes of the young (MODY), but the molecular cause is not clear.

• What are the molecular consequences of MODY-associated K_ATP mutations?

• K_ATP mutations from a large group of patients with MODY were all associated with loss-of-function (LOF) of different molecular etiologies.

• There are two distinct forms of K_ATP-associated MODY resulting from GOF and LOF, respectively, with very different clinical and therapeutic implications.

Introduction

Glucose-induced insulin secretion is controlled by β-cell electrical activity, and ATP-sensitive K⁺ (K_ATP) channels play a central role: nutrient oxidation causes an increased intracellular ratio of the concentrations of ATP to ADP ([ATP]/[ADP]), which reduces K_ATP channel activity, resulting in membrane depolarization, activation of voltage-dependent Ca²⁺ channels, increased intracellular concentration of Ca²⁺ [Ca²⁺], and triggering of insulin release. Therefore, a fine balance between open K_ATP channels at low glucose levels and closed K_ATP channels when glucose levels are elevated is critical to the normal switch from basal to glucose-stimulated insulin secretion. As a consequence, mutations in genes affecting β-cell K_ATP channel activity lead to dysregulated insulin secretion and altered blood glucose levels (1). Gain-of-function (GOF) mutations in the ABCC8 or KCNJ11 genes, which encode the SUR1 and Kir6.2 subunits, respectively, of the β-cell K_ATP channel, cause electrical inexcitability and loss of secretory response, resulting in neonatal diabetes mellitus (NDM), characterized by loss of insulin secretion and persistent hyperglycemia (2–5). Conversely, K_ATP loss-of-function (LOF) mutations cause β-cell hyperexcitability and increased secretory response, resulting in congenital hyperinsulinism (CHI) and hypoglycemia (6–9). Life-changing therapies for NDM and CHI have been developed by targeting K_ATP channels in β-cells with inhibitory sulfonylureas (SUs) for NDM, and the activator diazoxide for CHI.

Although the paradigmatic role of K_ATP activity in β-cell excitability and insulin secretory response appears straightforward, it is complicated by the observation that K_ATP channel loss or inhibition can lead to a paradoxical “crossover” from hyper- to hypo-secretion of insulin. There are now multiple clinical reports of patients with CHI, beginning with an extensive Finnish pedigree carrying a K_ATP mutation (10), who gradually progress, through unknown mechanisms, from canonical hyperinsulinism-induced hypoglycemia to hyperglycemia and diabetes (11–14). Animal studies reveal a similar crossover when K_ATP channels are genetically knocked out (15) or permanently inhibited pharmacologically (16), but again, the mechanistic basis remains unclear.

India now ranks second in the world in terms of individuals living with diabetes, with >100 million diagnosed individuals (17), representing an important and relatively untapped well from which to identify patients with novel and previously uncharacterized disease-associated mutations and follow their clinical evolution and therapeutic management. Over the past 10 years, the Madras Diabetes Research Foundation has developed a network across the whole of India that brings together patients, caregivers, and clinicians treating patients with pancreatic disorders of glucose homeostasis. The network provides a large cohort in which to carry out correlative analysis of diseases mechanism and progression. We have evaluated the relationship between patient genotype, molecular phenotype, and long-term clinical evolution, and identified and characterized several novel disease-associated K_ATP variants (18). Among these, we now report multiple unrelated patients who were clinically diagnosed as having a form of maturity-onset diabetes of the young (MODY) and who carry distinct coding variants in ABCC8. Two of these variants cause K_ATP GOF associated with NDM. However, as we show here, most of these variants cause mild to severe LOF, and not GOF, in recombinant K_ATP channels. This group of patients, who may all have crossed over from hyperinsulinemia/hypoglycemia to diabetes, thus further define a paradoxical form of diabetes that is genetically, mechanistically, and clinically distinct from early or late-onset NDM, with important implications for clinical management.

Research Design and Methods

Molecular Biology

The cDNA sequence of human Kir6.2 is identical to NG_012446.1, and the cDNA sequence of human SUR1 is identical to NM_001287174.3. Mutations were annotated based on NM_000352 (unless otherwise indicated), cloned into cDNA, and stably expressed in HEK293 cells using the landing pad system (19). Briefly, the attB-containing plasmid with mCherry and a puromycin resistance gene (19) was used as vector. Restriction endonuclease sites for MluI, BglII, NheI, and EcoRI were used to sequentially insert hKir6.2, an internal ribosome entry sites sequence, and hSUR1, into the vector. Site-directed mutations were first introduced to hSUR1 in pcDNA3.1, using Q5 polymerase and confirmed by Sanger sequencing. The wild-type (WT) or mutant hSUR1 sequence was amplified with PCR, introducing NheI and EcoRI sequence to the 5′ and 3′ terminals, respectively, and then subcloned into the final attB plasmid by restriction digestion and ligation, before transfection and selection, as previously described (20). Landing pad HEK293 cells were cultured in 25 mmol/L glucose DMEM, with 25 mmol/L glucose supplemented with 10% FBS, 100 units/mL penicillin, and 0.1 mg/mL streptomycin, and plated into 6-well plates 1 day before transfection. In each well, 1.5 μg of pCAG-NLS-Bxb1 and 1.5 μg of attB recombination plasmid (containing the K_ATP channel subunits) were cotransfected with 6 μL of Fugene 6. Two days after transfection, the medium was replaced by culture medium supplemented with doxycycline (4 μg/mL) and puromycin (10 µmol/L). Death of cells without DNA recombination was observed within 2 days. The cells were then cultured for 2–3 weeks with occasional replacement of the culture medium. The established stable cell lines lost blue fluorescence and uniformly acquired red fluorescence when viewed under a fluorescence microscope.

Membrane Potential Assessments by Voltage-Sensitive Fluorescent Dye

Transparent, 96-well cell-culture plates were coated with poly-l-lysine (0.1 mg/mL) 1 day before stably transfected cells were plated at an appropriate density to ensure an even distribution without layering. Cells were cultured with 100 μL of 25 mmol/L glucose DMEM containing 10% FBS, 100 units/mL penicillin, 0.1 mg/mL streptomycin, and 4 μg/mL doxycycline. After 1 day in culture, the medium was discarded and the cells were washed twice with low concentration potassium [K⁺] buffer (in mmol/L: 139 NaCl, 1 KCl, 2 CaCl₂, 1 MgCl₂, 10 HEPES, 10 glucose, pH 7.4). Cells were then incubated with 3 μmol/L DiBAC4(3) in low [K⁺] buffer, with or without additional channel modulators, for at least 30 min. Separate images of green (intensity 0.005) and red (intensity 0.379) fluorescence were generated with an EVOS M5000 imaging system. Exported .tiff images were analyzed by CellProfiler software with custom designed programs for landing pad cells, and stable cell lines were generated, as previously described (20). Fluorescence intensities were used to describe the membrane potential for each specific cell line with a certain treatment in one image. The average intensity value of duplicate images was calculated as one data point for the statistical analysis and data presentation.

Electrophysiological Methods

The consequences of ABCC8 variants on K_ATP channel properties were evaluated by performing patch-clamp recordings in inside-out configuration on HEK293 cells stably expressing Kir6.2/SUR1 channels. Mutant channel densities and properties were compared with WT channels on the same experimental day to control for day-to-day variation in cellular behavior. Experiments were performed at room temperature (20–22°C) in an oil-gate chamber that allowed rapid exchange of the solution facing the exposed surface of the isolated patch (21). Bath and pipette control solutions (KINT) contained (in mmol/L): 148 KCl, 10 HEPES, 1 EDTA, and 1 EGTA (pH 7.3–7.4 with KOH). Recording electrodes were obtained from soda-lime hematocrit glass (Kimble-Chase catalog no. 2502) using a P–97 puller (Sutter Instruments); pipette tips had resistances of 1–2 MΩ when filled with KINT solution. All currents were measured at a membrane potential of +50 mV, analog filtered at 1 kHz, and sampled at 3 kHz, using a CV-4 head stage, an Axopatch 1–D amplifier, a Digidata 1322A digitizer board (Axon Instruments), and an MP-225 micromanipulator (Sutter Instruments). Where indicated, ATP and ADP were added to the bath solution as dipotassium salts. Free Mg²⁺ concentrations were maintained at 0.5 mmol/L by supplementation of MgCl₂, as calculated by using Maxchelator (WEBMAXC Standard; University of California, Davis, Health). Data were collected using pClamp 8.2 software suite (Axon Instruments) and analyzed using Clampfit (Molecular Devices).

Data Analyses

All statistical analyses were performed using Microsoft Excel and GraphPad Prism 9 (GraphPad Software). Significance values were calculated using one-way ANOVA, with subsequent post hoc Dunnett multiple comparisons test or t test, as appropriate. Values are expressed as mean ± SEM, unless otherwise specified.

Data and Resource Availability

The original data will be made available to any interested parties upon reasonable request.

Results

Study Participants

We focused our attention on a subset of patients referred to the Molecular Genetics Department of Madras Diabetes Research Foundation, and in the database at Dr. Mohan’s Diabetes Specialities Centre, in Chennai, India, all of whom were clinically diagnosed with MODY (i.e., diabetes not neonatally identified, no clinically type 2 diabetes currently; n = 22), and who carry SUR1 coding variants (n = 20) (Tables 1 and 3). Genetic testing was performed as part of clinical diagnosis. Ethical approval was not required.

Table 1.

List of the K_ATP mutations identified in our cohort of patients with MODY

Gene	Amino acid	Classification on ClinVar*	Reported associated diseases	Functional characterization	Literature information
ABCC8	His36Asp	Not found	–	–	–
ABCC8	Val37Ile	SNP variant	–	–	https://www.ncbi.nlm.nih.gov/snp/
ABCC8	Val84Ile	Likely pathogenic/ uncertain significance	HI, NDM, MODY	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Gly163Ser	Uncertain significance	HI, MODY	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Arg216Cys	Uncertain significance	NDM	–	https://www.ncbi.nlm.nih.gov/clinvar/
KCNJ11	Glu227Lys	Pathogenic	NDM, MODY	GOF	https://www.ncbi.nlm.nih.gov/clinvar/
KCNJ11	Arg347Ser	Not found	–	–	–
ABCC8	Arg519Cys	Uncertain significance	MODY	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Arg519His	SNP variant	–	–	https://www.ncbi.nlm.nih.gov/snp/
ABCC8	Val601Ile	Uncertain significance	CHI, MODY	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Ala758Val	SNP variant	–	–	https://www.ncbi.nlm.nih.gov/snp/
ABCC8	Val849Ile	Uncertain significance	T1D but without Ab testing	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Asp860His	Variant	CHI	–	Bellanne-Chantelet et al. (48)
ABCC8	Asp897Glu	Uncertain significance	CHI	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Glu971Val	Uncertain significance	MODY	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Ala977Thr	Uncertain significance/likely benign	HI, T2D	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Ala1007Thr	SNP variant	–	–	https://www.ncbi.nlm.nih.gov/snp/
ABCC8	Gly1008Ser	Uncertain significance	MODY	–	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Arg1182Trp	Pathogenic/likely pathogenic	HI, NDM, T2D	GOF	https://www.ncbi.nlm.nih.gov/clinvar/
ABCC8	Arg1352Cys	SNP variant	–	–	https://www.ncbi.nlm.nih.gov/snp/
ABCC8	Arg1436Pro	SNP variant	–	–	https://www.ncbi.nlm.nih.gov/snp/
ABCC8	Ala1472Thr	Pathogenic	MODY, T2D	–	https://www.ncbi.nlm.nih.gov/clinvar/

Bold type indicates the mutations that were further characterized with functional assays. –, not applicable; Ab, antibody; HI, hyperinsulinism; SNP, single nucleotide polymorphism; T1D, type 1 diabetes; T2D, type 2 diabetes.

*See https://www.ncbi.nlm.nih.gov/clinvar/.

In each of these participants, there were no other detected variants in our monogenic diabetes panel comprising 75 genes (HNF4A, BSCL2, GATA4, LPL, PTF1A, GCK, CACNA1D, GATA6, LRBA, RFX6, HNF1A, CAV1, GLIS3, MAFA, SALL1, IPF1, CISD2, GLUD1, MNX1, SIRT1, HNF1B, COQ2, GPC3, NEUROG3, SIX1, NEUROD1, COQ9, HADH, NKX2-2, SLC16A1, CEL, CTLA4, IER3IP1, NKX6-1, SLC19A2, INS, DCAF17, IL2RA, PAX2, SLC29A3, ABCC8, DNAJC3, INSR, PIK3R1, SLC2A2, KCNJ11, DYRK1B, ITCH, PLAGL1, STAT1, AGPAT2, EIF2AK3, JAK1, PLIN1, STAT3, AIRE, EIF2S3, KDM6A, PMM2, STAT5B, WFS1, EYA1, KMT2D, POLD1, TNFAIP3, AKT2, ZBTB20, LMNA, PPARG, TRMT10A, APPL1, FOXP3, ZFP57, PPP1R15B, and UCP2). Two of the variants (encoding Glu227Lys in KCNJ11, and Arg1182Trp in ABCC8) have been functionally characterized and recognized as responsible for a GOF of recombinant channels (22–24); thus, they are conclusively causative of NDM, which was presumably diagnosed late in the two patients carrying these variants. Notably, however, four variants (encoding Val84Ile, Gly163Ser, Val601Ile, and Ala977Thr in ABCC8) have been reported in association with hyperinsulinism (unlike the clinical phenotype of diabetes), as well as later-onset diabetes, but without functional characterization, and classified as variants of uncertain significance (Table 1). Val37Ile, Arg519His, Ala758Val, Ala1007Thr, Arg1352Cys, and Arg1436Pro have been reported as single nucleotide polymorphism variants but have not been associated with any pathological condition. Asp860His and Asp897Glu have been identified only in association with CHI (Table 1). Six variants (Arg216Cys, Arg519Cys, Val849Ile, Glu971Val, Gly1008Ser, and Ala1472Thr) have been reported in association with early or delayed onset of diabetic symptoms, but again, without functional characterization of their effect on K_ATP channel activity (Table 1). Two more variants (encoding Arg347Ser in KCNJ11 and His36Asp in ABCC8) have not yet been classified and are reported here, to our knowledge, for the first time (Table 1).

Measurements of K_ATP Channel Activities by Voltage-Sensing Dye DiBAC4(3)

We proceeded to determine channel functionality of 10 uncharacterized mutations in HEK293 cells stably expressing pancreatic channel complexes made by WT human Kir6.2 and WT or mutant SUR1 subunits (see Research Design and Methods). To test the naive hypothesis that these mutations should result in K_ATP channel GOF, as NDM-associated mutations do, we first used a DiBAC4(3) fluorescence screening assay, which, maintaining intact cells and preserving their physiological intracellular environment, allows assessment of physiological channel activity by measuring the degree of membrane hyperpolarization (25). The novel MODY mutations were tested in comparison with WT channels and two previously characterized SUR1 NDM mutations, Phe132Leu and His1023Tyr, known to cause K_ATP channel GOF (26,27) (Fig. 1A).

Figure 1.

Characterization of Kir6.2/SUR1 WT and MODY mutants channel activities with the DiBAC4(3) assay. A: Fluorescence intensities of DiBAC4(3)-treated, untransfected HEK293, WT, and NDM SUR1 mutants Phe132Leu and His1023Tyr in basal physiological conditions. Fluorescence intensity was significantly lower in WT cells vs. untransfected (P = 0.0008), indicative of lower membrane potential vs. untransfected cells, and even more in NDM mutations Phe132Leu and His1023Tyr (P = 0.04), as an expected consequence of the cell membrane hyperpolarization associated with a GOF of the channel (n = 5 independent experiments). B: In the same conditions, the novel MODY mutations showed a tendency toward an increase in fluorescence signal compared with WT, significant for the mutations Gly163Ser (P = 0.002), Asp897Glu (P = 0.001), and Arg1436Pro (P = 0.0002) (n = 6–9 independent experiments). C: The K_ATP-opener diazoxide (100 µmol/L ) did not affect the fluorescence signal in untransfected cells but significantly reduced it in WT and in NDM Phe132Leu and His1023Tyr (n = 5 independent experiments). D: Similarly, diazoxide reduced the signal in most of the novel MODY mutations, significantly so in Gly163Ser (P < 0.0001 vs. diazoxide on WT) and Arg1436Pro (P < 0.0001 vs. diazoxide on WT) (n = 6–9 independent experiments). E: The K_ATP inhibitor glibenclamide (100 nmol/L) did not affect the fluorescence signal in untransfected cells, but restored it in WT cells. Glibenclamide did not markedly change the fluorescence level in His1023Tyr and especially Phe132Leu (P = 0.0079) (n = 5 independent experiments). F: As it did in WT, glibenclamide increased the fluorescence intensity in all the novel MODY mutations. Levels of fluorescence of Gly163Ser and Arg1436Pro were higher than WT in all the experimental conditions, even those with glibenclamide (P = 0.0299 and 0.0043, respectively) (n = 6–9 independent experiments). Data are expressed as mean ± SEM together with single experimental values, and significant differences were evaluated with one-way ANOVA with Dunnett multiple comparisons test. arb., arbitrary unit.

In basal conditions, we observed a marked reduction of the DiBAC4(3) signal for Phe132Leu and His1023Tyr in comparison with WT, reflecting hyperpolarization and, hence, greater channel activity, as expected. In contrast, none of the novel MODY mutations had a lower DiBAC4(3) signal than WT. Most of them had a tendency, some markedly, toward increased DiBAC4(3) signals, indicating a more depolarized cell membrane potential than WT, thus indicating a relative loss of channel function (Fig. 1B). In particular, a significant increase of the fluorescence intensity of ∼30% compared with WT was seen with the mutations Gly163Ser, Asp897Glu, and Arg1436Pro (Fig. 1B). The K_ATP channel opener diazoxide (100 µmol/L) did not affect the fluorescence level of untransfected HEK293 cells, but it caused a marked reduction of fluorescence in cells expressing WT channels and most of the novel MODY mutations. Significantly reduced response to diazoxide was seen for Gly163Ser and Arg1436Pro, suggesting the maximum activatable channel activity was lower than WT (Fig. 1C and 1D). Finally, the K_ATP channel inhibitor glibenclamide (100 nmol/L) had no effect on the DiBAC4(3) fluorescence signal of untransfected HEK cells (Fig. 1E), but it did increase the signal in WT and all MODY mutants to a similar level (Fig. 1F), indicating the hyperpolarizing baseline effects were all K_ATP dependent.

Taken together, these data indicate that most of the ABCC8 mutations carried by patients diagnosed with MODY, in particular, the mutations Gly163Ser, Asp897Glu, and Arg1436Pro, actually result in loss of K_ATP channel function, rather than causing K_ATP GOF, which underlies NDM.

Patch Clamp Data: Current Density and ATP Sensitivity

Mutational effects on channel properties were directly assessed by inside-out patch-clamp experiments (Fig. 2). Channel activity in the absence of cytoplasmic nucleotides reflects channel expression level and maximum open probability. Again, none of the mutants showed greater activity than WT; instead, they showed a trend toward lower activity levels, with significantly lower levels for His36Asp, and Gly163Ser, indicative of lower maximum open probability, and an essentially complete loss of channel activity for Arg1436Pro (Fig. 3A).

Figure 2.

Representative traces from inside-out patch clamp recordings of Kir6.2/SUR1 WT and MODY mutations. Currents were recorded in inside-out configuration in symmetrical high K⁺ conditions (148 mmol/L on both side of the membrane patch) at +50 mV membrane voltage. Mutations are indicated above each figure. For each recorded trace, currents were acquired in total absence of nucleotide (labeled as 0), 1 mmol/L ATP (labeled as 1), 0.1 mmol/L ATP +0.5 mmol/L free Mg²⁺ (labeled as 0.1), and 0.1 mmol/L ATP +0.3 mmol/L ADP +0.5 mmol/L free Mg²⁺ (labeled as 0.1 + 0.3). In absence of nucleotide, a tendency toward reduced K_ATP currents was observed for all the MODY mutations compared with WT; currents were equally totally abolished in the presence of 1 mmol/L ATP for WT and MODY mutations. K_ATP currents were subsequently evoked by Mg nucleotides. No further effects of ADP on top of ATP were observed with the mutation Asp897Glu (highlighted in the inset). No currents were observed with the mutation Arg1436Pro in any of the experimental conditions.

Figure 3.

Current density and response to nucleotides in patch clamp experiments. A: The graph shows the values of K ⁺ current as recorded in inside-out patches in KINT solution in absence of additional nucleotides. For each reported mutant (beige bar and dots), currents were normalized to the dimension (resistance) of the electrode tip and to the current of WT channels recorded on the same experimental day, which are defined as 100% (superimposed black line and dots). A general tendency toward reduced currents was observed for each of the mutants His36Asp, Arg519Cys, Arg519His, Ala758Val, and Arg1352Cys. Significantly lower currents were observed for the mutation Gly163Ser (≈35% of current amplitude vs. corresponding WT cells; P = 0.001 as evaluated with t test). No currents were detected for the mutation Arg1436Pro (0% of current amplitude vs. corresponding WT cells; P < 0.001 as evaluated with t test). B: ATP (1 mmol/L) blocked all K_ATP currents for WT and mutants. C: ATP (0.1 mmol/L) + free Mg²⁺ (0.5 mmol/L) evoked ≈10% of the currents observed in absence of nucleotides for WT and several mutants. A slight but not significant reduction in the recorded currents was observed for the mutant Asp897Glu vs. corresponding WT. No currents were detected for the mutation Arg1436Pro (P < 0.001 as evaluated with t test vs. corresponding WT currents in the same recording conditions). D: ATP (0.1 mmol/L) + ADP (0.3 mmol/L) in the presence of 0.5 mmol/L free Mg²⁺ evoked ≈50% of the currents observed in absence of nucleotides for WT and several mutants. No additional current was evoked for the mutant Asp897Glu by adding ADP on top of ATP (0.1 mmol/L ATP +0.5 mmol/L free Mg²⁺: ≈6% of the currents observed in absence of nucleotides; 0.1 mmol/L ATP +0.3 mmol/L ADP +0.5 mmol/L free Mg²⁺: ≈10% currents observed in absence of nucleotides; P = 0.003 as evaluated with t test vs. corresponding WT currents in the same recording conditions). No currents were detected for the mutation Arg1436Pro (P < 0.001 as evaluated with t test vs. corresponding WT currents in the same recording conditions). Data are expressed as mean ± SEM together with single experimental values, and significant differences among mutant and correspondent controls were evaluated with a t test. E: Nonlinear regression analysis of absolute DiBac signal (arbitrary units [arb.]) with diazoxide (y-axes) and percentage of currents in 0 ATP (x-axes) for each characterized mutant (R² = 0.78). F: Nonlinear regression analysis of absolute DiBac signal (arb.) in basal conditions (y-axes) and Mg-nucleotide activated K_ATP currents (x-axes) for each characterized mutant (R² = 0.59).

Nonlinear regression analysis revealed a good correlation between the fluorescence values observed in the DiBAC4(3) assay in diazoxide (i.e., maximal channel activity) and maximal currents recorded in patch clamp experiments (Fig. 3D), with the exception of Asp897Glu. In contrast to the DiBAC4(3) assay, Asp897Glu had a similar maximum patch current level as WT, indicating normal open probability and expression level, suggesting this mutation might be responsible for impaired channel gating by nucleotides. K_ATP channel activity is finely regulated by intracellular nucleotide levels inside the cells; inhibition by ATP is essential for maintaining low activity at low blood glucose levels, and the [ATP]/[ADP] ratio is the key determinant of physiological activation (7).

We used rapid change of the solution facing the intracellular side of the isolated membrane patches (21) to consecutively test nucleotide responses. In contrast to the finding that many NDM GOF mutations reduce sensitivity to inhibitory ATP (28), none of the novel mutations had altered relative activity in 0.1 or 1 mmol/L ATP (Fig. 2, Fig. 3B and C). Sensitivity to MgADP activation was also unaffected for all mutants, except for Asp897Glu (Fig. 2, Fig. 3D), which showed no enhancement of activity by MgADP. This lack of MgADP activation resolves the discrepancy between diazoxide-activated hyperpolarization and channel density for this mutation. When the DiBAC4(3) signal, under basal conditions, was plotted against channel activity in isolated membranes with pseudo-physiological nucleotides, the variant now fell close to the same line as other mutants (Fig. 3F).

Clinical Characteristics of Patients With ABCC8 LOF

Thus, functional characterization did not show a channel GOF for any of the MODY-associated mutations; instead, it indicated channel LOF of varying degrees, with varying mechanistic bases. Given these findings, it is important to consider the clinical status of these patients, particularly with regard to treatments and efficacy.

At the time of the investigation, all 20 patients were treated medically, either with SUs (n = 6 patients; 30%), metformin (n = 2 patients; 10%), insulin (n = 2 patients; 10%), other oral hypoglycemic agents (n = 2 patients; 10%), or a combination of two or more of these therapeutics (n = 7 patients; 35%); one patient (5%) was not receiving any pharmacological treatment. The response to SU is particularly important to consider. For K_ATP GOF, it can effectively reverse the primary defect (29), but for K_ATP LOF, if anything, it might be predicted to exacerbate or accelerate the diabetes by further reducing channel activity. Indeed, although treatment with the SU gliclazide was trialed in four patients in our cohort, it was stopped in one because of poor glucose control, and HbA_1c levels remained continuously elevated in the other three patients (Table 3). In the patient with the p.His36Asp mutation, HbA_1c levels were lowered after reversion to voglibose/metformin treatment (Table 3).

Table 3.

Clinical characteristics defining MODY in our patient cohort

Characteristic	ABCC8 mutation
	c.106 C > G	c.487 G > A	c.1556 G > A	c.2273 C > T	c.4054 C > T	c.4307 G > C
Amino acid substitution	His36Asp	Gly163Ser	Arg519His	Ala758Val	Arg1352Cys	Arg1436Pro
Protein region	N terminal	L0 cytoplasmic linker	T1D	T1D	T2D	NDM2
Sex	Male	Male	Female	Male	Male	Female
Age at onset (years)	36	32.6	33.1	24.6	11	14.9
Diabetes in immediate relatives	Yes	Yes	Yes	Yes	Yes	Yes
Mutation in the family	–	–	Yes	–	–	Yes
Antibodies	Negative	Negative	Negative	Negative	Negative	Negative
Diabetes duration (years)	30	8	4	9	4	20
BMI, kg/m2	26	27.3	23.6	22.5	25.8	21.3
Fasting C-peptide level (pmol/L)	0.7	1.1	0.8	0.4	1.1	0.6
Stimulated C-peptide level (pmol/L)	2	1.6	1.1	1.4	2.4	1.6
MODY probability using Exeter calculator, %	ND	4.6 (1/21.7)	15.1 (1/6.6)	75.5 (1/1.3)	75.5 (1/1.3)	75.5 (1/1.3)
Hypertension	No	No	No	No	No	No
Dyslipidemia	No	No	No	No	No	No
Cardiovascular disease	No	No	No	No	No	No
Complications	No	No	No	No	No	No
Years of follow-up, n	4	3	3	5	6	6
Pharmacological therapy	Voglibose and metformin	Metformin	Metformin	Vildagliptin	Gliclazide	Gliclazide
Trial with sulfonylurea			Yes	Yes, interrupted due to poor glycemic control	Yes	Yes
Last HbA1c value, %	6.8	8.7	10	6.5	10.3	10

ND, not determinable due to age of diagnosis.

Discussion

Since the 1980s, K_ATP channels have been recognized as key controllers of pancreatic insulin secretion in an essentially perfect paradigm (30): K_ATP closure induces depolarization and insulin secretion, whereas the opposite happens when channels open. The straightforward correlation was solidified by the finding that GOF and LOF mutations in the β-cell K_ATP channel subunit genes cause NDM (in which insulin secretion is essentially absent) and in CHI (in which patients hypersecrete insulin), respectively (4,8,31) (Fig. 4). At this juncture there are hundreds of recognized mutations in KCNJ11 and ABCC8 associated with each of these conditions.

Figure 4.

Distinct MODY forms dependent on K_ATP LOF or GOF. Schematic illustrating primary and secondary consequences of K_ATP LOF or GOF. K_ATP LOF causes β-cell hyperexcitability and chronic elevation of intracellular [Ca] ([Ca]i), resulting in hypersecretion and CHI. By activating K_ATP channels, treatment of CHI with diazoxide (DZX) can correct the primary defect. Conversely, K_ATP GOF causes β-cell underexcitability and low [Ca]i, resulting in undersecretion and NDM or MODY. By inhibiting overactive K_ATP channels, treatment of K_ATP GOF diabetes with SUs corrects the primary defect. Chronically elevated [Ca]i due to K_ATP LOF secondarliy leads to β-cell undersecretion and a delayed-onset glucose intolerance and diabetes (MODY). In this case, although treatment with SUs may acutely trigger insulin secretion and lower blood glucose, it will exacerbate the primary defect, potentially worsening long-term disease outcome. We propose that these two mechanistically distinct forms of MODY be recognized as such, using the terms ABCC8-NDM/MODY (and KCNJ11-NDM/MODY) for classic K_ATP GOF, and ABCC8-HI/MODY (and KCNJ11-HI/MODY) for K_ATP LOF-dependent MODY and progression from CHI to MODY, respectively.

Here we report on multiple patients diagnosed with MODY and who carry previously unrecognized mutations in ABCC8 from our network across India (Madras Diabetes Foundation). Given the paradigmatic control of insulin secretion by K_ATP channels, the naive assumption would be that, if causally linked to the diabetes status of the patients, these mutations should cause a channel GOF. We assessed the mutational consequences for K_ATP channel activity by DiBAC4(3) and electrophysiological assays in multiple, distinct experimental conditions adopted to stimulate or inhibit channel function. Channel GOF was obvious for two NDM mutations (Phe132Leu and His1023Tyr) that were tested for comparison, but our assays did not identify any GOF in channels expressing any of the novel K_ATP mutations. Instead, both assays indicated that most of the mutations cause a weak loss of channel function and that three cause a severe or complete LOF through various molecular mechanisms: Gly163Ser, characterized by reduced channel density; Asp897Glu, characterized by reduced sensitivity toward Mg-nucleotides; and Arg1436Pro, characterized by apparently complete lack of activity or expression on the membrane.

K_ATP LOF could not be directly responsible for the hyperpolarization of the membrane that would normally prevent insulin release (i.e., as underlies NDM resulting from K_ATP GOF), indicating that the MODY phenotype of these patients is not a late diagnosis of NDM. Instead, by causing depolarization and aberrant triggering of insulin secretion, channel LOF causes the opposite problem: CHI. Clinically, the dangers of hypoglycemia in early childhood have focused attention on this period and on efforts to stem insulin secretion in patients with CHI with considerable success (32–34). Generally, there has been little attention paid to the lifelong evolution of the disease in these patients, but it is recognized that symptoms of CHI generally become less severe as the child develops, and there has been a steady stream of reports of isolated SUR1-dependent CHI cases and several more extensive pedigrees, gradually crossing over to a glucose-intolerant or diabetic phenotype (10,11,13,14). An apparently similar but more rapid crossover occurs in mice in which the Glu1506Lys mutation, found in one extensive pedigree in Finland, has been knocked in to the equivalent mouse locus (35), as well as in mice with complete loss of Kir6.2 (15) or SUR1 (36). In all these animals, hyperinsulinism is reported in the very early neonatal phase, followed by insulin deficiency and glucose intolerance or frank diabetes.

Survey of the available literature reveals at least 13 previous reports of functionally characterized LOF mutations in ABCC8 or KCNJ11 in patients who manifest features of diabetes in adulthood, often preceded by diagnosis of CHI at birth (Table 2), as well as several other K_ATP mutations identified in association with a clinical diagnosis of MODY that have not been functionally characterized. Together with the new cases presented here, such patients represent a distinct group sharing a similar clinical phenotype: actual or presumed hyperinsulinemia at birth, followed by hyperglycemia later in life. None of these patients satisfies the classical diagnosis of NDM, showing 1) onset in adulthood and not neonatally; 2) LOF (rather than GOF) consequences of the K_ATP channel mutation; and 3) sometimes with overall good glycemic control without need of therapeutics. These patients also do not have classical type 2 diabetes; they show early onset (mean age at onset of diabetes in this list of 23 patients with LOF was 28.1 years) and general lack of marked risk factors for the disease, such as insulin resistance or obesity (Table 3).

Table 2.

Characteristics of patients with K_ATP LOF MODY reported in literature

Gene	Mutation	Functional characterization	Phenotype at birth (no. of patients)	Type and onset of hyperglycemia (no. of patients)	Reference
ABCC8 NM_001287174	Arg168Cys + Arg1421Cys	LOF	Mild hypoglycemia (1)	27, 28 yo, hyperglycemia (2)	Matsutani et al. (49)
ABCC8 NM_001287174.1	Leu171Phe	—	HH (1)	9 yo, insulin-deficient diabetes (1)	Isik et al. (47)
ABCC8 NM_000352.6	Arg370Ser	LOF	CHI (1)	10 years 6 months, hyperglycemia (1)	Abdulhadi-Atwan et al. (50)
ABCC8 NM00352.2	Arg1353His	LOF	—	17 yo, MODY (1)	Koufakis et al. (44)
ABCC8 NM_001287174	Lys1385Gln	—	Hypoglycemic during adolescence (1), HH (2)	39 yo, hyperglycemia (1); 12 yo and 14 yo, impaired glucose tolerance (2)	Karatojima et al. (51)
ABCC8 NM_000352.6	Arg1418His	LOF	Hypoglycemic seizures (1)	1 year 4 months, postprandial hyperglycemia alternated with hypoglycemia (1)	Harel et al. (52)
ABCC8 NM_000352.6	Arg1420His	LOF	HH (1)	3.5 yo, diabetes (1); increased risk for T2DM	Baier et al. (53)
ABCC8 NM_000352.2	Leu1431Phe	LOF	—	20 yo, diabetes (1)	Kapoor et al. (54)
ABCC8 NM_000352.2	Gln1459Glu	LOF	HH (1)	21, 23, and 25 yo, gestational diabetes (1)	Kapoor et al. (54)
ABCC8 NM_000352.2	Gly1479Arg	LOF	HH (1)	30 yo, diabetes (1)	Kapoor et al. (54)
ABCC8 NM_000352.6	Glu1506Lys	LOF	CHI (11)	−39, 42, 44, 60 yo, diabetes (4); impaired glucose tolerance (5); impaired fasting glucose (1)	Huopio et al. (55)
ABCC8 NM_000352.2	Ala1508Pro	LOF	HH (1)	50 yo, diabetes (1); 38 yo, diabetes (1)	Kapoor et al. (54)
ABCC8 NM_000352.2	Ala1537Val	LOF	—	26 yo, gestational diabetes (1)	Kapoor et al. (54)
ABCC8 NM_000352.2	Arg1539Gln	LOF	—	47 yo, diabetes (1); 39 yo, diabetes (1)	Kapoor et al. (54)
KCNJ11 NM_000525.3	Ser118Leu	LOF	Not aware	31 yo, persistent hyperglycemia	Vedovato et al. (56)

HH, hyperinsulinemic hypoglycemia; T2DM, type 2 diabetes mellitus; yo, years old.

Such patients, diagnosed with MODY but carrying K_ATP LOF mutations, might all have had some degree of CHI, as expected by the effects of channel LOF mutation on β-cell function, but perhaps it was mild enough to escape diagnosis when the patients were infants, and then showed features of diabetes later in life in a crossover event (Fig. 4). Although it is not known what mechanisms underlie crossover in humans, the glucose-intolerant phenotype in mice with K_ATP LOF is associated with some decrease of β-cell mass and apoptosis, potentially a result of depolarization-driven abnormal Ca²⁺ intake (37) but also potentially due to downregulation of the [Ca²⁺] dependence of secretion itself (38).

K_ATP-dependent GOF mutations, although typically manifesting in NDM very early in life, have also been reported as associated with a later-onset MODY, potentially because of relapse from undiagnosed transient NDM, and this has sometimes been classified as MODY12 (or ABCC8 MODY; i.e., GOF mutations in ABCC8, SUR1) (39–41), and MODY13 (or KCNJ11 MODY; i.e., GOF mutations in KCNJ11, Kir6.2) (42). However, multiple studies that reported an association of either recognized SUR1 LOF mutations with MODY or progression of individuals carrying uncharacterized SUR1 mutations from CHI to MODY lead to confusion in the literature about whether or how this condition relates to K_ATP GOF-dependent MODY (43,44). As discussed, it is mechanistically unfeasible that K_ATP GOF and LOF mutations can underlie the same diabetic etiology, and they will each have very different response to K_ATP-directed therapies. One recent review highlighted the wide range of clinical manifestations in patients diagnosed with MODY in association with ABCC8 mutations, and the authors pointed out that although SU treatment may be a preferred treatment (based on presumption of K_ATP GOF as the origin), the efficacy can vary substantially (45). Distinguishing between K_ATP GOF-driven MODY and K_ATP LOF-driven CHI and crossover to MODY is critical for choosing an appropriate treatment approach. Moreover, although SUs have proven effective for treatment of the former, they may instead be detrimental in the latter. To avoid further confusion, we first urge diagnostic recognition and separation of the two forms of K_ATP-dependent MODY, and suggest using the term ABCC8-NDM/MODY (and KCNJ11-NDM/MODY) for classic K_ATP GOF, and the term ABCC8-HI/MODY (and KCNJ11-HI/MODY) for K_ATP LOF-dependent MODY and progression from CHI to MODY (Fig. 4).

Second, we urge that once a K_ATP coding variant is identified in a patient clinically diagnosed with MODY, functional tests of the variant are essential to determine the correct diagnosis to enable appropriate treatment and avoidance of potentially harmful interventions. As inhibitors of K_ATP channels, SUs have represented a first-choice life-saving treatment of several forms of diabetes, including type 2 diabetes, monogenic forms such as MODY1 (HNF4A MODY) and MODY3 (HNF1A MODY), and especially for NDM (4). For NDM patients with K_ATP GOF mutations, SUs have proven to be a magic bullet, providing full and lasting remission by exactly targeting the molecular defect (29). In our cohort of patients carrying K_ATP LOF mutations, several were undergoing treatment with SUs, either alone or in association with additional therapeutics. The choice of SU therapy might have been simplistically motivated by the standard of care for K_ATP GOF-driven NDM. However, recognition of the LOF nature of the mutations clearly raises the question as to how appropriate this treatment is (Fig. 4). Because SU action is immediate, treatment might cause acute lowering of blood glucose levels even in incomplete LOF cases by inhibiting residual channel activity and acutely triggering secretion of insulin, but it may ultimately be detrimental. In this regard, it is interesting that in the two patients carrying SUR2 LOF mutations who were using SU monotherapy, HbA_1c levels were unresponsive to SU treatment (Table 3), which is consistent with the additional pharmacological inhibition leading to chronically worsened symptoms. Review of additional case literature on patients diagnosed with MODY and carrying known K_ATP LOF mutations suggests that such an outcome may be typical, with SU treatment providing poor or only temporary control of blood glucose levels and subsequently requiring supplementation or replacement by non-K_ATP treatments (43,44,46,47).

Therefore, we further urge that, although MODY resulting from K_ATP GOF mutations should be treated with SUs, patients diagnosed with MODY resulting from K_ATP LOF mutations should probably be treated with other glucose-lowering agents that do not target K_ATP channel inhibition. Randomized clinical trials with a larger series of individuals with K_ATP LOF mutations will be needed to identify the most appropriate treatment options.

Funding Statement

These studies were supported by the National Institutes of Health (grant R01 DK133838 to C.G.N. and Maria S. Remedi) and by the Indian Council of Medical Research (grant 5/4/5-2/Diab/2020-NCD-III to V.R.).