Differential mechanisms of Cantú syndrome–associated gain of function mutations in the ABCC9 (SUR2) subunit of the K_ATP channel

ATP-sensitive potassium (K_ATP) channels, found in many excitable tissues, including brain (Ashford et al., 1988), heart (Noma, 1983), pancreatic islets (Cook and Hales, 1984), and skeletal and smooth muscle (Standen et al., 1989; Winquist et al., 1989) are hetero-octameric protein complexes composed of four pore-forming inwardly rectifying potassium channel Kir6.X subunits, each associated with one sulfonylurea receptor subunit (SURX) (Shyng and Nichols, 1997). K_ATP channels link cell metabolism with membrane excitability as a result of intracellular nucleotide regulation of channel activity. ATP binds directly to the pore-forming Kir6.X subunits (Tucker et al., 1997) to inhibit the channel, and MgADP binds to SURX to activate the channel (Nichols et al., 1996). Channels formed from different combinations of Kir6.1 (KCNJ8), Kir6.2 (KCNJ11) (Inagaki et al., 1995a,b), SUR1 (ABCC8), and SUR2 (ABCC9) subunit isoforms vary in nucleotide sensitivity and pharmacology (Masia et al., 2005), as well as tissue localization (Aguilar-Bryan et al., 1995; Inagaki et al., 1996), generating additional functional diversity of K_ATP channels (Nichols, 2006; Akrouh et al., 2009).

Cantú syndrome (CS) was first recognized as a unique disorder in 1982 (Cantú et al., 1982). Clinical hallmarks of CS include hypertrichosis, macrosomia, macrocephaly, coarse facial appearance, cardiomegaly, and skeletal abnormalities (Harakalova et al., 2012; van Bon et al., 2012; Cooper et al., 2014). Recent studies have shown that the majority of known CS patients have missense mutations in ABCC9 (Harakalova et al., 2012; van Bon et al., 2012) or KCNJ8 (Brownstein et al., 2013; Cooper et al., 2014) that result in a gain of function (GOF) in K_ATP channel activity. CS resulting from mutations in either subunit confirms that functional changes in the K_ATP channels formed from them underlie the disease.

Conversely, KCNJ11 and ABCC8, which encode Kir6.2 and SUR1, respectively, generate K_ATP channels in pancreatic β cells (Inagaki et al., 1995a), and GOF mutations in either of these subunits give rise to neonatal diabetes mellitus (NDM) (Gloyn et al., 2004; Babenko et al., 2006), characterized by decreased insulin secretion and elevated blood glucose levels. Detailed analyses of NDM mutations in ABCC8 demonstrate that distinct molecular mechanisms can underlie the increase in channel activity, including increased open probability (P_o) (Koster et al., 2005; Babenko et al., 2006; Proks et al., 2007) and increased MgADP activation (de Wet et al., 2007b; Masia et al., 2007). To understand the molecular mechanisms of CS-associated ABCC9 mutations, we have engineered and analyzed channel properties for three CS mutations that localize to different regions of the protein: P429L, A475V, and C1039Y. Given the experimental difficulty of recording Kir6.1/SUR2 currents (Cooper et al., 2014), we studied Kir6.2+SUR2A channels in the present experiments, and by means of ⁸⁶Rb⁺ efflux assays in intact cells transiently transfected with WT or mutant subunits, we confirm that all three mutations result in overactive K_ATP channels. Consistent with the diversity of mechanisms found in NDM patients with ABCC8 mutations, we further show that CS-associated ABCC9 GOF mutations can result from at least two different mechanisms: decreased sensitivity to inhibition by ATP or enhanced channel activation by MgADP.

Models of human SUR2A (Fig. 1) were built using Modeller v9.8 from the template of the multidrug ABC transporter Sav1866 (2ONJ, to model the “open” conformation) and heterodimeric ABC transporter TM287-TM288 (3QF4, to model the “closed” conformation), respectively. The TM0 domain and L0 loop (1–281) were omitted because of lack of sequence homology to any proteins of known structure. Two extended loops unique to SUR2A were also omitted: one connecting TM1 and NBD1 (614–672) and a second connecting NBD1 and TM2 (920–966). A multiple sequence alignment (MSA) was performed using ClustalW2 between SUR2 and the two template sequences in conjunction with other proteins of the human ABCC family (ABCC1, 2, 3, 4, 5, 6, 8, and 10) for TM1 and NBD1 because of low sequence identity. High sequence identity enabled a reliable sequence alignment of TMD2 and NBD2 by MSA between SUR2 and SUR1, 2ONJ, and 3QF4.

The Quick Change II Site-Directed Mutagenesis kit (Agilent Technologies) was used to engineer P429L, A475V, and C1039Y mutations (equivalent to CS-associated P432L, A478V, and C1043Y mutations in human SUR2; Harakalova et al., 2012; van Bon et al., 2012) into rat SUR2A-pCMV6. Mutations were confirmed by direct sequencing of the entire SUR2A coding region. For channel expression, COSm6 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 10⁵ U/L penicillin, and 100 mg/L streptomycin. At 60–70% confluence, cells were transfected with the relevant plasmids using FuGENE6 transfection reagent (Promega). For experiments with homomeric CS mutant channels, cells were cotransfected with pCDNA3.1-mKir6.2-WT (0.6 µg) and WT or mutant pCMV6-ratSUR2A (SUR2) (1 µg). For experiments using heteromeric channels, cells were cotransfected with Kir6.2-WT, SUR2-WT, and mutant SUR2 at ratios of 0.6:0.5:0.5 (w/w/w). Cells transfected with GFP-pcDNA3.1 served as controls. A small amount of GFP DNA was coexpressed for identification of transfected cells in electrophysiology experiments.

Cells were incubated overnight at 37°C in DMEM containing 1 µCi/ml ⁸⁶RbCl (PerkinElmer), and then incubated in Ringer’s solution (mM: 118 NaCl, 10 HEPES, 25 NaHCO₃, 4.7 KCl, 1.2 KH₂PO₄, 2.5 CaCl₂, and 1.2 MgSO₄, adjusted to pH 7.4 with NaOH) in (a) the absence (basal) or (b) the presence of 2.5 mg/ml oligomycin and 1 mM 2-deoxy-d-glucose (metabolic inhibition [MI]), (c) the SUR2-specific K⁺ channel opener pinacidil (PIN; 250 µM), or (d) a combination of MI and PIN to achieve maximal channel activation. Subsequently, at selected time points (2.5, 5, 7.5, 15, 25, and 40 min) the solution was collected and replaced with fresh solution. Upon completion of the assay, cells were lysed with 2% SDS and collected, and radioactivity in these samples was measured by liquid scintillation. ⁸⁶Rb⁺ efflux is expressed as a fraction of total content. A nonspecific efflux that inactivated with time was assumed to be present in all ⁸⁶Rb⁺ efflux experiments, and apparent rate constants for the K_ATP-independent ⁸⁶Rb⁺ efflux were obtained from GFP-transfected cells using the equation:

where k₁ and k₋₁ are the apparent activation and inactivation rate constants. K_ATP-dependent ⁸⁶Rb⁺ efflux was also assumed to activate and inactivate with time and was obtained using the equation:

where k₁ and k₋₁ are the rate constants for K_ATP-independent pathways (obtained from GFP-transfected cells by Eq. 1), and k₂ and k₋₂ are the activation (k₂) and inactivation (k₋₂) rate constants for K_ATP-specific K⁺ conductance. k₂ is then assumed to be proportional to the number of active channels. To account for potential day-to-day variability, for every efflux experiment, a k₁ value was determined for untransfected cells and then incorporated as a fixed parameter in the equation to estimate the K_ATP-dependent k₂ for each construct. All k values are listed in Tables 1–3.

After 24–48 h, transfected cells that fluoresced green under UV light were selected for analysis by excised patch-clamp experiments at room temperature in a perfusion chamber that allowed for the rapid switching of solutions. K_int solution (mM: 140 KCl, 10 HEPES, and 1 EGTA, pH 7.4) was used as the standard pipette (extracellular) and bath (cytoplasmic) solution in these experiments. ATP or ADP was added as indicated. The appropriate amounts of MgCl₂ to be added in each Mg²⁺-nucleotide–containing solution to attain 0.5 mM of free Mg²⁺ were calculated by means of the CaBuf program (Katholieke Universiteit Leuven). Membrane patches were voltage clamped using an Axopatch 1D amplifier (Molecular Devices), and currents were recorded at −50 mV (pipette voltage, 50 mV) in on-cell and inside-out excised patch configuration. Data were typically filtered at 1 kHz and digitized at 5 kHz with a Digidata 1322A (Molecular Devices) A-D converter. pClamp and Axoscope software (Molecular Devices) were used for data acquisition. For ATP inhibition, [ATP]-response relationships (Fig. 4) were fitted by Eq. 3:

where I_rel (relative current) is the current in the presence of a given concentration of ATP relative to current in zero ATP, K_i is the apparent ATP inhibition constant, and H is the Hill coefficient.

For phosphatidylinositol (4,5)-bisphosphate (PIP₂) activation experiments, an ammonium salt of l-α-phosphatidylinositol-4,5-bisphosphate from porcine brain (Brain PI(4,5)P2; Avanti Polar Lipids, Inc.) was dissolved in K_INT to make a 5-µg/µl working solution. For each patch, we estimated the relative P_o by dividing the maximum steady-state current in zero ATP by the maximum steady-state current in PIP₂.

CS-associated mutations have been found throughout the coding sequence (Fig. 1). For the present study, we focused on two previously unexamined mutations (human A478V and C1043Y corresponding to rat A475V and C1039Y, respectively), located in the TMD1 and TMD2 segments, and P432L (corresponding to rat P429L), also located in the TMD1 region. To examine K_ATP channel activity in intact cells, we performed ⁸⁶Rb⁺ efflux assays under four different conditions: basal, MI, in the presence of PIN, and MI and PIN combined (MI+PIN). As shown in Figs. 2 and 3, homomeric expression of SUR2A-P429L, A475V, or C1039Y channels, results in a significantly higher basal ⁸⁶Rb⁺ efflux rate compared with WT (Figs. 2 A and 3 A). P429L- and A475V-based channels, but not those composed of C1039Y, also showed a significantly higher rate of efflux compared with WT channels under MI conditions and in the presence of PIN (Figs. 2, B and C, and 3, B and C), consistent with the GOF observed in the basal condition. Maximal efflux rates (estimated using simultaneous exposure to MI and PIN) of P429L and A475V were comparable to SUR2A-WT, implying similar channel densities at the cell surface (Figs. 2 D and 3 D). C1039Y showed significantly lower absolute fluxes in all stimulatory conditions (Figs. 2, B–D, and 3, B–D), implying a lower channel density at the membrane (Fig. 4).

The ratio of K_ATP-dependent (k₂) rate constants in basal, MI, and PIN to the maximal K_ATP-dependent rate constant (in MI+PIN) should reflect the relative activation of each channel in these specific conditions. In the basal condition, the ratio was higher for all three mutations than for WT, markedly so for C1039Y (Fig. 4 A). In PIN alone, although only P429L and A475V reach significance, the ratio of active channels in all three mutations are also increased (Fig. 4 C). Collectively, the results suggest that each mutant will result in higher channel activation under any stimulatory conditions, particularly for C1039Y.

⁸⁶Rb⁺ flux assays provide evidence for overactivity in CS mutants, but do not provide any indication of underlying molecular mechanisms. We assessed the details of channel properties in inside-out excised patch-clamp electrophysiology experiments. Intrinsic sensitivity to ATP inhibition (in zero Mg²⁺) was similar for WT, P429L, and A475V channels (K_i = 7–9 µM). However, K_ATP channels expressing C1039Y exhibited a significant right shift in ATP sensitivity (K_i = 21.3 µM; Fig. 5 B). A diminished sensitivity to intracellular ATP may account, at least in part, for the increased activity of C1039Y-based K_ATP channels in basal conditions in the intact cell (Figs. 2 A and 3 A). Consistent with reduced channel density, the maximal current in zero ATP was significantly lower in C1039Y channels than in WT, P429L, or A475V channels (Fig. 5 C). The P_o in the absence of inhibitory ATP (zero) was estimated by the application of PIP₂ (Koster et al., 2005). After the application of PIP₂, both WT and C1039Y channels lost ATP sensitivity; however, the current in zero ATP increased for WT channels but not for C1039Y channels (Fig. 6). The ratio of the initial maximum current in zero ATP to that in PIP₂ (WT: 0.70 ± 0.11 vs. C1039Y: 1.26 ± 0.21) reflects marked increase in WT maximum P_o but no increase for C1039Y.

We investigated the current response to intracellular MgADP in the presence of 0.1 mM ATP (Fig. 7 A). MgADP-dependent activation was estimated as the ratio between the steady-state activated current in MgADP+ATP and the maximal current in zero ATP, immediately after excision (Fig. 7 B). The relative current in MgADP was markedly higher than WT for P429L and A475V channels. Using this analysis, the current was also statistically higher for C1039Y channels. However, this analysis ignores the intrinsically lower sensitivity of C1039Y channels to ATP inhibition (Fig. 5 C). An alternative estimation for the stimulatory action of Mg nucleotides is to calculate the ratio of current in MgADP+ATP to that in Mg-free ATP alone (Fig. 7 B, inset). This analysis implies no difference in the MgADP activation of C1039Y and WT channels. In addition, although the MgADP activation for P429L and A475V channels is more rapid than WT, activation of the C1039Y channel is even slower than WT (Fig. 7 C).

All documented CS patients with mutations in ABCC9 are heterozygous. To mimic the predicted heteromeric composition of channels in such patients, we also expressed each mutation in a 1:1 ratio with SUR2A-WT plus Kir6.2, and assessed channel activity by ⁸⁶Rb⁺ efflux assays (Figs. 8 and 9). In all conditions, P429L, A475V, and C1039Y heteromeric channels display no significant increases in the rate of ⁸⁶Rb⁺ efflux compared with WT channels (Figs. 8 A and 9 A). When normalized to the maximal efflux rates, channels with heteromeric expression of C1039Y are still considerably less active than WT. In the heteromeric state, maximal C1039Y channel fluxes were markedly higher than in the homomeric state (Fig. 9 D), also implying a partial rescue of the surface expression of C1039Y subunits by WT subunits (Figs. 8, B–D, and 9, B–D).

Some NDM patients with GOF mutations in Kir6.2 or SUR1 have successfully been switched from insulin therapy to K_ATP inhibitors such as glibenclamide (Zung et al., 2004). However, a correlation between increased channel activity and diminished effectiveness of such drugs has been noted (Koster et al., 2005). Therefore, to test the effectiveness of glibenclamide on overactive P429L, A475V, or C1039Y channels, expressed both homomerically and heteromerically; ⁸⁶Rb⁺ efflux experiments were performed in the presence of MI plus 10 µM glibenclamide (Fig. 10). The time course of Rb⁺ effluxes shows an unusual behavior in that although initial fluxes are markedly lower than in MI alone, the inhibition is not maintained through the time course of the assay. It is well understood that the inhibitory action of glibenclamide is a complex function of the metabolic conditions and decreases under conditions of MI (Findlay, 1994; Koster et al., 1999). To account for this behavior, the efflux data in glibenclamide were fit by Eq. 2, where positive values for k₋₂ now result in the rate of efflux actually increasing with time for A475V and P429L channels (see Table 3). From this analysis, the ratio of the k₂ in MI plus glibenclamide to k₂ in MI provides an estimate of the relative sensitivity to inhibition by glibenclamide (Fig. 10 C). Although the glibenclamide sensitivity of A475V and P429L channels was not different from WT, C1039Y channels appeared almost insensitive to the drug. Qualitatively similar results were obtained with channels expressed in heteromeric mixture with WT subunits (Fig. 10, B–D). As discussed below, this glibenclamide insensitivity is consistent with similar findings for SUR1 mutations that produce increased activity of resultant K_ATP channels.

CS is a rare disease characterized by complex vascular and skeletal anomalies, the underlying cellular and molecular mechanisms of which we are only now beginning to understand (Nichols et al., 2013). The majority of genotyped CS patients are heterozygous for mutations in the genes encoding ABCC9 (SUR2) (Harakalova et al., 2012; van Bon et al., 2012) in most cases, and KCNJ8 (Kir6.1) (Brownstein et al., 2013; Cooper et al., 2014) in others. A select few of these mutations have thus far been shown to form overactive K_ATP channels (Harakalova et al., 2012; Cooper et al., 2014), and most remain unexplored. Here, we characterized three CS-associated ABCC9 mutations, namely P429L (hP432L), A475V (hA478V), and C1039Y (hC1043Y), that localize to distinct regions of the SUR2 protein (Fig. 1). Each mutation leads to K_ATP channel GOF (Figs. 2–7), consistent with previous findings on isolated CS-associated ABCC9 mutations (Harakalova et al., 2012).

In the only previous assessment of CS-associated ABCC9 mutations, Harakalova et al. (2012) measured channel sensitivity to ATP in the presence of Mg²⁺. Mg nucleotides have complex activatory/inhibitory effects on K_ATP channels: MgATP can both inhibit the Kir6.X subunit and be hydrolyzed to activatory MgADP at the SUR subunits (Nichols, 2006). Therefore, such an analysis cannot separate mutant effects on ATP inhibition from those on Mg nucleotide activation. By separately assessing channel activity in response to Mg-free ATP and to MgADP, we show that CS-associated ABCC9 mutations lead to increased channel activity via different mechanisms, as has also been observed with NDM-associated ABCC8 mutations (de Wet et al., 2007a, 2008; Masia et al., 2007; Zhou et al., 2010). Specifically, we demonstrate that increased activity of P429L and A475V channels results from increased MgADP activation, whereas increased activity of C1039Y channels results predominantly from decreased ATP sensitivity caused by higher intrinsic open-state stability.

Some NDM-causing GOF mutations in SUR1 have also been shown to reduce ATP sensitivity (Tarasov et al., 2008; Takagi et al., 2013), either by reducing inhibitory ATP-binding affinity or by increasing the intrinsic open-state stability of the channel (Shyng and Nichols, 1998; Koster et al., 2005). That PIP₂ decreases ATP sensitivity, but does not increase maximum currents, in the C1039Y channel implies that the intrinsic P_o of these channels is near 1 (relative P_o = 1.26 ± 0.21), significantly higher than for WT channels (relative P_o = 0.70 ± 0.11; Fig. 6), explaining the intrinsically lower ATP sensitivity of the C1039Y channel.

In addition to increasing channel activity, homomeric expression of C1039Y appears to decrease expression of channels in the membrane, detectable in both ⁸⁶Rb⁺ flux and patch-clamp analyses. A similar dual effect of GOF mutations in both Kir6.2 (Lin et al., 2013) and SUR1 (Zhou et al., 2010) has been shown previously. In these previous examples, decreased surface expression was demonstrated by Western blot analysis. Significant rescue of channel activity in heteromeric C1039Y/WT channels is consistent with the patterns from such mutations (Figs. 8 and 9); therefore, we suggest that in addition to reducing ATP sensitivity of expressed channels, C1039Y also decreases K_ATP surface expression.

Kir6.1 GOF mutations have also been found in two CS patients (Brownstein et al., 2013; Cooper et al., 2014), confirming Kir6.1/SUR2 as the key subunit combination generating the overactive channels. Given the experimental difficulty of recording Kir6.1/SUR2 currents (Cooper et al., 2014), we used Kir6.2+SUR2A channels in the present study. ABCC9 encodes two splice variants: SUR2A and SUR2B, which vary only in the last 40 amino acids. All CS-associated SUR2 mutations identified thus far are found within the core of the SUR2 protein, such that these mutations will be present in both SUR2A and SUR2B. SUR2A is predominant in skeletal muscle and heart, whereas SUR2B is expressed in the vasculature. 6.1/SUR2B may be the more dramatically affected combination, as vascular phenotypes seem to predominate in CS patients (Nichols et al., 2013): neither skeletal muscle weakness nor shortening of the QT interval on the electrocardiogram has been noted, as might be expected if Kir6.2/SUR2A were markedly activated.

The severity of disease varies between CS patients such that not all reported features are exhibited by all patients. Direct comparison of the phenotypes of patients carrying the A475V and C1039Y equivalent mutations (van Bon et al., 2012) revealed hypertrichosis, macrosomia, macrocephaly, coarse features, cardiac anomalies, and umbilical hernia to be common to both, but pulmonary hypertension, cardiomegaly, osteopenia, hyperextensibility and lymphedema were only reported for the A475V equivalent patient. Because the reduced sensitivity to ATP inhibition caused by the C1039Y mutation results in greater relative basal channel activity than does the enhanced Mg nucleotide activation caused by A475V, a less severe disease phenotype for C1039Y suggests that reduced expression caused by this mutation may also be a consequence in vivo, as also reported for SUR1 NDM mutations (Zhou et al., 2010).

Finally, sulfonylureas, which inhibit K_ATP channels via interaction with the SUR subunits, are effective treatments for NDM patients with ABCC8-activating mutations (Babenko et al., 2006), and may be promising potential therapies for CS (Nichols et al., 2013). However, GOF mutations in Kir6.2 (Koster et al., 2005) and SUR1 (Takagi et al., 2013) that increase channel open-state stability also typically decrease sensitivity to sulfonylureas, which may result in a lack of sulfonylurea therapeutic efficacy in patients with these mutations. Assessment of ⁸⁶Rb⁺ efflux in glibenclamide (Fig. 10) suggests that the same can be true of SUR2 GOF mutations, specifically C1039Y channels, which required a significantly higher k₂ value, indicating reduced glibenclamide sensitivity (see Table 3). However, in general, SUR2 sensitivity to sulfonylureas is much lower than SUR1 sensitivity to these drugs (Dörschner et al., 1999), and so, given the further reduction in drug sensitivity of GOF mutants, the introduction of sulfonylureas to treat CS patients is likely to require high doses, such that undesired inhibition of SUR1-based K_ATP channels, leading to hypoglycemic effects, will need to be carefully considered. Such concerns may ultimately necessitate the development of Kir6.1- or SUR2-specific inhibitors to provide safer and more effective treatment options for CS patients.

This work was supported by National Institutes of Health (NIH) grant HL95010 (to C.G. Nichols). P.E. Cooper was supported by NIH training grant HL007275.

The authors declare no competing financial interests.

Merritt C. Maduke served as editor.