DEND Mutation in Kir6.2 (KCNJ11) Reveals a Flexible N-Terminal Region Critical for ATP-Sensing of the K_ATP Channel

Abstract

ATP-sensitive K⁺-channels link metabolism and excitability in neurons, myocytes, and pancreatic islets. Mutations in the pore-forming subunit (Kir6.2; KCNJ11) cause neonatal diabetes, developmental delay, and epilepsy by decreasing sensitivity to ATP inhibition and suppressing electrical activity. Mutations of residue G53 underlie both mild (G53R,S) and severe (G53D) forms of the disease. All examined substitutions (G53D,R,S,A,C,F) reduced ATP-sensitivity, indicating an intolerance of any amino acid other than glycine. Surprisingly, each mutation reduces ATP affinity, rather than intrinsic gating, although structural modeling places G53 at a significant distance from the ATP-binding pocket. We propose that glycine is required in this location for flexibility of the distal N-terminus, and for an induced fit of ATP at the binding site. Consistent with this hypothesis, glycine substitution of the adjacent residue (Q52G) partially rescues ATP affinity of reconstituted Q52G/G53D channels. The results reveal an important feature of the noncanonical ATP-sensing mechanism of K_ATP channels.

INTRODUCTION

The ATP-sensitive K⁺-channel (K_ATP) controls cellular function in neurons, β-cells, and other tissues by translating changes in cellular metabolism (represented by the [ATP]/[ADP] ratio) into changes in electrical activity. Aberrantly elevated K_ATP activity is predicted to decrease excitability, and thereby compromise cell function. An increase in K_ATP activity in the pancreatic β-cell impairs insulin release at stimulatory blood-glucose concentrations (1–3). In addition, channel overactivity in the central nervous system is predicted to underlie seizure events (4–7), and it appears that ketogenic diets may actually protect against seizure by enhancing the damping effect of K_ATP channels on electrical activity (8). Activating mutations in K_ATP (both the Kir6.2 (KCNJ11) and SUR1 (ABCC8) subunits) represent a common cause of neonatal diabetes mellitus (NDM) in humans, a disease characterized as an insulin-dependent hyperglycemia diagnosed before 6 months of age (9). The diabetes may be either permanent (PNDM) or transient (TNDM) in form (10,11). In the most severe cases, K_ATP mutations are associated with neurological features that include developmental delay (both motor and intellectual), epilepsy, and neonatal diabetes (referred to as DEND syndrome) (12), and these extrapancreatic symptoms likely reflect K_ATP overactivity in muscle, peripheral neurons, and/or brain. Importantly, sulfonylureas, which specifically block the K_ATP channel, can reverse the diabetes and ameliorate the associated neurological features (6,13,14).

Structurally, the K_ATP channel is a hetero-octomer comprised of four pore-forming Kir6.2 subunits bordered by four regulatory SUR1 subunits. Whereas SUR1 confers sensitivity of the channel to stimulatory MgADP and the inhibitory sulfonylureas, the Kir6.2 subunit confers the characteristic ATP-inhibition of the channel. The predicted ATP-binding pocket in Kir6.2 is formed by residues located in both the cytoplasmic N- and C-termini such that each Kir6.2 subunit in the tetramer is capable of binding a molecule of ATP, and binding of ATP to one subunit is sufficient to cause channel closure (15,16).

To date, more than 40 heterozygous activating mutations in Kir6.2 (KCNJ11) have been reported worldwide and shown to be variably associated with both the transient and permanent forms of the disease (17–19). Without exception, these diabetes-causing mutations activate K_ATP by increasing the half-inhibitory ATP concentration (K_1/2,ATP) (20–24). Mechanistically, these mutations decrease ATP-sensitivity either 1), by directly decreasing ATP-binding to Kir6.2 (e.g., R50P), or 2), by altering intrinsic channel-gating (i.e., increasing open probability in the absence of ATP (p_o,zero), e.g., Q52R). In the latter case, ATP-sensitivity is reduced allosterically, because ATP predominantly interacts with closed states of the channel, to which access is reduced when the open state is favored (25,26).

Recently we described the phenotype and therapy of an adult patient with intermediate DEND (developmental delay and neonatal diabetes, but not epilepsy) attributable to the G53D mutation in KCNJ11 (14). Functional analysis of reconstituted G53D channels (Kir6.2[G53D] + SUR1) demonstrated a significant rightward shift in the ATP dose-response curve. Consistent with channel overactivity in the cellular environment, whole-cell K_ATP currents were significantly increased in transfected mammalian cells expressing G53D channels, compared with the wild-type (WT) (14). Here, we characterize the effects of mutations in this region in greater detail. Although G53 resides outside the predicted ATP-binding pocket, the G53D mutation alters ATP-sensitivity without altering the intrinsic gating behavior of the channel (as reflected by the p_o,zero and kinetic behavior of single channels). This contrasts sharply with all other reported non-ATP-binding mutations that indirectly alter ATP-sensitivity by changing gating kinetics (20,21,24). Furthermore, we show that the G53 position is highly intolerant of amino-acid substitutions with respect to ATP-sensitivity, but the loss of ATP-sensitivity can be partially rescued by glycine substitution at the adjacent residue (Q52G). These results demonstrate that flexibility of this N-terminus region is necessary for efficient ATP binding, and this leads us to speculate that the rotational flexibility of G53 constitutes an interfacial gating hinge, separate from the gate involved in ion conduction, which allows for movement of the N-terminal region and an induced fit of ATP at its binding pocket.

METHODS

Molecular biology

We cloned Kir6.2 into the EcoRI/ClaI sites of pCMV6b and the parental plasmid DNA was used to generate Kir6.2 mutations using the QuickChange Site Directed Mutagenesis Kit (Stratagene, La Jolla, CA). The SUR1 was cloned into the pECE expression vector. The nucleotide sequences of mutant Kir6.2 constructs were verified by fluorescence-based cycle sequencing, using AmpliTaq DNA polymerase, FS (Perkin-Elmer, Foster City, CA), and an ABI Prism DNA sequencer (Perkin-Elmer).

Expression of K_ATP channels in COSm6 cells

COSm6 cells were plated at a density of ∼2.5 × 10⁵ cells per well (30-mm six-well dishes) and cultured in Dulbecco's modified Eagle's medium plus 10 mM glucose (DMEM-HG), supplemented with fetal calf serum (FCS, 10%). The COSm6 cells were transfected with cDNA using FuGENE 6 Transfection Reagent (Roche Diagnostics, Indianapolis, IN). Total DNA (0.3 μg Kir6.2 (27) plus 0.5 μg SUR1 (28) plus 0.3 μg green fluorescent protein as a marker for transfection) was mixed with FuGENE 6 and preincubated for 1 h. Cells were incubated in the presence of the transfection mixture for 12–24 h, and plated on sterile glass coverslips overnight before patch-clamp experiments.

Electrophysiological methods

Patch-clamp experiments were performed at room temperature in an oil-gate chamber that allowed the solution bathing the exposed surface of the isolated patch to be changed rapidly. The COSm6 cells that fluoresced green under ultraviolet illumination were selected for patch-clamping 3–5 days after transfection. Membrane patches were voltage-clamped using a CV-4 head stage, an Axopatch 1-D amplifier, a Digidata 1322A digitizer board (all from Axon Instruments, Union City, CA), and an MP-225 micromanipulator (Sutter Instrument, Novato, CA). All currents were measured at a membrane potential of −50 mV (pipette voltage, +50 mV). Data were collected using the pClamp 8.2 software suite (Axon Instruments) and Microsoft Excel (Microsoft, Redmond, WA). Bath and pipette control solutions (K_INT) contained 150 mM KCl, 10 mM HEPES, and 1 mM EGTA, pH 7.4. Where indicated, ATP was added to the bathing solution as dipotassium salts. Phosphotidyinositol-4,5-bisphosphate (PIP₂; Avanti Polar Lipids, Alabaster, AL) was dissolved in K_INT to a stock concentration of 1 mg/mL, and diluted to a working concentration of 5 μg/mL.

Data analysis

Quantitative analysis of ATP inhibition

The ATP dose-response was quantified by fitting the raw data with a Hill equation:

(1)

where I_rel is the current relative to that in the absence of ATP, [ATP] is the ATP concentration, K_1/2,ATP is the half-maximal inhibitory ATP concentration, and H is the Hill coefficient, which was fixed at 1.3.

Single-channel analysis

Unconditional mean lifetimes in the open and closed states were determined from idealized (using pClamp 8.2 software) records of channel activity from patches containing 1–4 channels.

Estimation of p_o,zero

Two approaches were used to estimate p_o,zero, the initial open probability (in zero ATP), after the excision of isolated membrane patches containing multiple channels. In method I, PIP₂ was added to the patch until the current reached a saturating level (I_PIP2). This was assumed to represent a maximum p_o,zero of ∼0.97 (26). The fold increase in current was calculated (fold increase = I_initial/I_PIP2; see Fig. 2), and p_o,zero was estimated from:

(2)

FIGURE 2.

Reduced ATP-sensitivity of reconstituted K_ATP channels containing G53 mutant subunits. Representative currents were recorded from inside-out membrane patches from COS cells expressing (A) wild-type (Kir6.2(WT) + SUR1) or (B) mutant (Kir6.2(G53D) + SUR1) K_ATP channels at −50 mV in K_int solution (see Methods). Patches were exposed to differing [ATP], and baseline current was determined by exposure to ATP (5 mM). (C) Steady-state dependence of membrane current on [ATP] (relative to current in zero ATP (I_rel)) for wild-type and G53X-containing channels. K_1/2,ATP = 10 μM (WT), 190 μM (G53D), 124 μM (G53F), 87 μM (G53S), 66 μM (G53A), 62 μM (G53C), and 52 μM (G53R). Data points represent mean ± SE (n = 4–30 patches). Fitted lines correspond to least-squares fits of a Hill equation (see Methods).

Method II involved noise analysis (NA). Mean p_o,zero was estimated from stationary fluctuation analysis of macroscopic currents (29,30) on short (<1 s) recordings of currents in zero [ATP] and in 5 mM [ATP]. Currents (4 pA < mean current < 4 nA, at −50 mV, corresponding to ∼1–1000 channels) were filtered at 1 kHz and digitized at 3 kHz, with 12-bit resolution. Mean patch current (I) and variance (α²) in the absence of ATP were obtained by subtraction of the mean current and variance in 5 mM ATP (i.e., assuming all channels were closed), respectively. Single-channel current (i) was assumed to be −3.75 pA, corresponding to a WT single-channel conductance of 75 pS. The p_o,zero was then estimated according to

(3)

Model simulation of p_o,zero versus K_1/2,ATP relationship

Although complex models of K_ATP gating can describe essentially all of the gating behavior of the channel (16,31,32), a simplified four-state model quantitatively describes channel behavior over a wide range of conditions, including changing levels of ATP and PIP₂, as well all mutational effects on ATP affinity (K_D) or the intrinsic open-closed equilibrium (K_CO) (model II of Enkvetchakul et al. (26)):

In this kinetic scheme, C represents the ATP-dependent closed state, C_in the ATP-independent closed state, C_F the fast closed state, and O the open state. K_D is the dissociation constant for the ATP-binding step, and K_CO is the equilibrium constant for the open-to-closed (C_in) transition. The key feature of this model is that ATP acts by binding to a closed state and, in consequence, the apparent ATP-sensitivity can be reduced either by shifting the equilibrium between the C_in and O states toward the open state (increasing K_CO), or by reducing ATP-binding (increasing K_D). Model simulations were generated using Microsoft Excel.

Calculations of free energy for ATP-binding

The free energy of ATP-binding was calculated for individual membrane patches using

(4)

where R is the gas constant (1.98 × 10⁻³ kcal mol⁻¹deg⁻¹), T is the absolute temperature (298 K), and K_D is the dissociation constant for ATP in the above model. For mutant K_ATP channels (G53D, Q52G, and G53D/Q52G), the free-energy effect of the mutation (ΔΔG) was obtained by subtracting the averaged value of ΔG for wild-type channels (see Fig. 6 F).

FIGURE 6.

ATP affinity is partially restored in Q52G/G53D double-mutant channels. (A) Representative currents recorded from inside-out membrane patches from COS cells expressing WT, G53D, Q52G, or Q52G/G53D channels. Patches were exposed to various [ATP], as indicated. (B) Steady-state dependence of membrane current on [ATP] (relative to current in zero ATP (I_rel)) for wild-type and G53x mutant channels. Data points represent mean ± SE (n = 10–30 patches). Fitted lines correspond to least-squares fits of a Hill equation (see Methods). (C) Average of K_1/2,ATP values from individual patches, calculated using Hill equation: **p < 0.01 by unpaired Student's t-test. (F) Maximum open probability in absence of ATP (p_o,zero) was calculated from individual membrane patches, as in Fig. 3. Data points represent mean ± SE (n = 9–17 patches, method I; n = 9–24 patches, method II). *p < 0.05 by unpaired Student's t-test, compared with WT and G53D channels. (E) Relationship between p_o,zero and K_1/2,ATP (mM). Solid lines represent modeled data calculated from model II of Enkvetchakul et al. (26). Symbols represent data points calculated from individual membrane patches. Modeled data were fit using nonlinear regression analysis. For statistical analysis, the Kruskal-Wallis test was performed on equilibrium constants for ATP-binding (K_D derived from model fits, Eq. 3) from individual patches shown: Q52G/G53D versus G53D, p < 0.01; WT versus Q52G/G53D, p < 0.001; and WT versus G53D, p < 0.001. (D) Free energy of ATP binding for mutant K_ATP channels. The Gibbs free energy (ΔG(kJ mole⁻¹)) was calculated from Eq. 3 for individual patches, using the fitted dissociation constant (K_D) (see Methods). To calculate ΔΔG (kcal mole⁻¹) values for mutant channels, the averaged free energy value from WT channels was subtracted from the averaged values for mutant K_ATP channels. Diamonds represent predicted ΔΔG (kcal mole⁻¹) value for combined mutant channel (ΔΔG[G53D] + ΔΔG[G53D/Q52G]), assuming an additive effect. Data points represent mean ± SE (n = 9–37 patches).

RESULTS

Molecular model of Kir6.2 and location of G53

Structural modeling of the Kir6.2 tetrameric pore places the G53 residue distal to the ATP-binding pocket and adjacent to the slide helix, which previously has been implicated in Kir channel gating (Fig. 1, A and B) (15,33). Mutations of nearby residues Q52 and V59 (Q52R and V59M) also underlie NDM, and were shown to reduce ATP-sensitivity indirectly by increasing the open probability of the channel (20,21). Just distal to G53 in the primary sequence is residue R50, which forms part of the binding pocket and is predicted to interact electrostatically with the γ-phosphate of ATP. Consistent with this role, the diabetes-causing R50P mutation significantly reduces the ATP-sensing of reconstituted channels, but without altering intrinsic gating behavior (34).

FIGURE 1.

Location of G53 in predicted Kir6.2 structure. (A) Ribbon diagram of two of four Kir6.2 subunits that form K⁺-selective pore in K_ATP. Dashed box corresponds to region that was magnified to create B. (B) Residues (highlighted) in Kir6.2 that are mutated in neonatal diabetes. Residues are classified as those that form part of the ATP-binding pocket (red), or are involved in channel-gating, and therefore indirectly in ATP-sensitivity (green). The G53 residue that neither forms part of the ATP-binding pocket nor alters gating behavior when mutated is shown in orange. The bound ATP molecule is shown in yellow, and the interfacial slide-helix region is shown as α-helical. (C) Conformational angles around α-carbons in amino-acid backbone of Kir6.2 are shown in a Ramachandran plot (ψ versus φ). Bond angles associated with α-helical and β-sheet structures are indicated. Glycine residues are shown as open symbols, and all other residues are shown as solid symbols. Ramachandran data analysis was performed using Visual Molecular Dynamics software (http://www.ks.uiuc.edu/Research/vmd/).

Intolerance of K_ATP to amino-acid substitutions at the G53 residue

Coexpression of WT Kir6.2 and SUR1 in COSm6 cells generates K_ATP channels that are inhibited by ATP in excised patch-clamp experiments, with an ATP-sensitivity similar to that of native channels (K_1/2,ATP = ∼10 μM; Fig. 2) (27). Conversely, homomeric mutant channels containing substitutions at residue G53 (Kir6.2[G53x] + SUR1) all exhibit a significant decrease in ATP-sensitivity relative to WT: K_1/2,ATP = 190 μM (G53D), 124 μM (G53F), 87 μM (G53S), 66 μM (G53A), 62 μM (G53C), and 52 μM (G53R). The G53R and G53S mutations underlie several cases of TNDM, and functional analyses of these mutants expressed in Xenopus oocytes demonstrated similar shifts in ATP-sensitivities (23).

Unaltered open probability in G53 mutated channels

Two approaches were used to estimate open probability in the absence of inhibitory ATP (p_o,zero) in macroscopic recordings: nonstationary NA (method II) (29), and the application of PIP₂ (method I) (20)(see Methods). Both methods gave similar values for p_o,zero for both WT and G53D mutant channels (Fig. 3; p_o,zero= 0.47 (WT) and 0.49 (G53D), averages from both methods I and II). Similarly, the TNDM-causing G53R mutation did not significantly affect p_o,zero (0.57), although the G53S mutation did slightly increase open probability relative to WT (p_o,zero = 0.67). This increase in p_o,zero was statistically significant by the NA method, but did not reach significance using the PIP₂ method. By comparison, the diabetes-linked Glu to Arg mutation at the adjacent residue 52 (Q52R) significantly increased the open probability (p_o,zero = ∼0.9), and this gating effect can fully account for the reported decrease in ATP-sensitivity of Q52R channels (20,21).

FIGURE 3.

Unaltered maximum open probabilities of reconstituted K_ATP channels containing G53 mutations. Open probability in absence of ATP (p_o,zero) was calculated from individual membrane patches expressing reconstituted wild-type and G53x mutant channels. Method I used PIP₂ stimulation to estimate p_o,zero, and method II used nonstationary noise analysis (NA; see Methods). For comparison, the p_o,zero of the gating mutation, Q52R, is also shown (20). Data points represent mean ± SE (n = 5–17 patches, method I; n = 10–27 patches, method II). *p < 0.05 and **p < 0.01 compared with WT by unpaired Student's t-test.

Altered relationship between ATP-sensitivity and open probability in G53D channels

As described above, multiple studies of channel-gating demonstrated a nonlinear relationship between p_o,zero and K_{1/2, ATP} (26,35,36). Wild-type channels have an intrinsic p_o of ∼0.5, and an intrinsic K_1/2,ATP of ∼10 μM (Fig. 4, A and C). Application of PIP₂ leads to a gradual increase in p_o,zero and K_1/2,ATP, in accordance with the prediction of the model (solid lines in Fig. 4 C indicate the prediction of the inset model with parameters from model II of Enkvetchakul et al. (26), with K_D = 5.8 μM). The gating mutant Q52R has a much higher intrinsic p_o,zero (∼0.9) and K_{1/2 ATP} (∼125 μM), reflecting a higher intrinsic K_CO, but the response to PIP₂ follows the same relationship, reflecting no change in ATP affinity (K_D). The model predicts that a Kir6.2 mutation that alters ATP binding alone will be shifted to a different relationship, with a higher K_1/2,ATP for any given p_o,zero. This is the case for the G53D channels (Fig.4 C, upper solid line). At similar open probabilities, a higher K_{1/2 ATP} was measured for G53D channels compared with WT, corresponding to a K_D value of 72 μM (i.e., ∼13-fold lower than WT). The finding that G53 mutations affect ATP affinity, but not intrinsic gating, is initially surprising, given that G53 is located outside the ATP-binding pocket (Fig. 1).

FIGURE 4.

PIP₂ stimulation of wild-type and mutant (G53D) K_ATP channels. Representative currents were recorded from membrane patches containing either (A) wild-type (Kir6.2(WT) + SUR1) or (B) mutant (Kir6.2[G53D] + SUR1) K_ATP channels at −50 mV. Bars indicate application of PIP₂ (5μg/mL) or different [ATP] (as shown). Lower dashed line indicates zero current. Insets: Expanded records from first 50–100 s after patch excision, from which ATP-sensitivity and open probability (NA method) were estimated. (C) Relationship between p_o,zero and K_{1/2, ATP} (mM). Solid lines represent modeled data calculated from model II of Enkvetchakul et al. (26). Symbols represent data points calculated from individual membrane patches: WT (Kir6.2(WT) + SUR1), G53D (Kir6.2(G53D) + SUR1), or Q52R (Kir6.2(Q52R) + SUR1). Modeled data were fit using nonlinear regression analysis.

Unaltered gating behavior in G53D channels

To examine the gating kinetics of G53D channels in detail, single-channel currents were recorded from cells expressing WT or mutant K_ATP channels (Fig. 5). Consistent with the macroscopic analysis of channel open probability, there was no significant difference in p_o,zero values between WT and G53D single channels, although the estimated open probabilities were lower than those calculated using the indirect methods (Table 1). The gating behavior was also not different between WT and G53D channels with groups of openings being clustered in similarly timed bursts in both channels (Fig. 5), and both channels displayed similar mean open (τ_open) and closed times (τ_closed), as well as single-channel current (Table 1). By contrast the disease-causing Q52R mutation predictably displayed an increase in p_o,zero in single channels with a significantly lower mean closed time (Fig. 5, Table 1). Together with the macroscopic analysis, these data strongly support the conclusion that gating behavior is not altered in G53D channels.

FIGURE 5.

Unaltered gating behavior in single G53D channels. Representative single-channel currents were recorded at −50 mV from inside-out membrane patches excised from COS cells expressing either WT (Kir6.2 + SUR1), G53D (Kir6.2[G53D] + SUR1), or Q52R (Kir6.2[Q52R] + SUR1) channels. Two different timescales (2 s and 0.1 s) are shown for currents from each channel.

TABLE 1.

Single-channel properties of wild-type and mutant Kir6.2 + SUR1 channels

Construct	i (nA)	τopen (ms)	τclosed (ms)	po,zero	n
WT	3.59 ± 0.2	3.5 ± 0.4	24.4 ± 5.4	0.145 ± 0.034	4
G53D	3.43 ± 0.1	3.4 ± 0.6	23.7 ± 8.1	0.198 ± 0.125	4
Q52R	3.58 ± 0.02	3.4 ± 0.5	0.9 ± 0.2*	0.796 ± 0.037*	3

i, single-channel current; τ_open and τ_closed, are unconditional mean open and closed times, determined from idealized records from n patches, each containing 1–4 active channels (mean number of channels per patch of 2.75, 2.5, and 1 for WT, G53D, and Q52R, respectively). p_o,zero = τ_open/(τ_open + τ_closed).

^*Significantly different from WT (p < 0.001).

Ramachandran plot of Kir6.2

The intolerance for amino-acid substitutions at the G53 position with respect to ATP-sensitivity (Fig. 2) implies the importance of rotational flexibility at this position. The φ and ψ torsion angles around the α-carbons in the Kir6.2 model (15) were analyzed in a Ramachandran plot (37) (Fig. 1 C). Glycine has no side chain, and therefore can adopt φ and ψ angles in all four quadrants, and frequently occurs in regions of proteins where any other residue would be sterically hindered. Indeed, in the Ramachandran plot of the modeled Kir6.2 structure, the G53 residue is found in the lower right quadrant, occupied primarily by glycines and very few other amino acids (Fig. 1 C).

Importance of rotational flexibility at the G53 position for ATP-sensing of the channel

As detailed in the Discussion, we propose that residue G53 uniquely controls ATP-binding through a long-range action on the proximal N-terminus. Specifically, we suggest that G53 acts as a hinge to allow a rigid body movement of the distal N-terminus that is necessary to fully accommodate ATP at its binding pocket (38). The existence of a hinge would explain the specific requirement for a glycine at this 53 position, but may also suggest that the hinge could be placed at an adjacent residue, to overcome the detrimental effect of substitutions at residue 53.

As shown in Fig. 6, A–C, the Q52G mutation (Kir6.2[Q52G] + SUR1) has no significant effect on ATP-sensitivity or on open probability, compared with the WT channel (Kir6.2WT] + SUR1). However, on the G53D background, the double-mutant channel (Kir6.2[Q52G,G53D] + SUR1) exhibits a significant left shift in ATP-sensitivity, compared with G53D channels alone (K_1/2,ATP = 16.4 ± 1.7 μM (WT), 22.7 ± 3 μM (Q52G), 64.9 ± 6.1 μM (Q52G/G53D), and 226 ± 21.2 μM (G53D)). This shift in ATP-sensitivity corresponds to a considerably smaller increase in the free energy of ATP binding (ΔΔG = 0.99 kcal mol⁻¹) for the combined mutant channel than is predicted from the additive effect of each mutant alone (ΔΔG[Q52G] + ΔΔG[Q52G/G53D] = 1.79 kcal mol⁻¹) (Fig. 6 D; see Methods). Although p_o,zero is unaffected by either the Q52G or G53D single mutations, the Q52G/G53D double-mutant channel exhibits a small decrease in open probability that reached significance using the NA method (method II) (Fig. 6 F). Given the coupled relationship between p_o,zero and K_1/2,ATP (Fig. 4), the possibility exists that the decrease in ATP-sensitivity observed for the combined Q52G/G53D mutant channels is simply a consequence of the reduced open probability. However, when the p_o,zero and K_1/2,ATP relationship for the Q52G/G53D mutant is modeled as above (Fig. 4), the fit data reside on a different curve than that of the G53D channels alone (Fig. 6 E) (Kruskal-Wallis test, p < 0.01). This indicates that the decrease in open probability alone is insufficient to account for the observed shift in ATP-sensitivity of the double-mutant channel, and that Q52G can partly restore ATP affinity on the G53D background. Taken together, the data strongly implicate rotational flexibility in the immediate G53 region as necessary for high ATP affinity.

DISCUSSION

K_ATP channel overactivity and disease severity

Previous studies demonstrated that distinct mutations of the G53 residue in Kir6.2 (G53S, G53R, and G53D) decrease ATP-sensitivity and underlie NDM or DEND (14,23). In the current study, a direct comparison indicates that the ATP-sensitivity of DEND-causing G53D channels is decreased ∼20-fold, compared with a ∼5–9-fold decrease for both the TNDM-associated G53S and G53R mutations. Because the relative ATP-sensitivity of the channel is an accurate predictor of K_ATP activity in vivo, these data are consistent with a genotype-phenotype relationship in which greater overactivity of K_ATP causes a more severe disease phenotype. Only with the most severe activating mutations (e.g., G53D) is overactivity associated with both diabetic and neurological symptoms, indicating that extrapancreatic tissues (e.g., muscle, nerve, and brain) are less sensitive to changes in K_ATP activity than the pancreatic β-cell, and that a threshold of channel overactivity must be reached before neurological features become evident.

Mechanistic basis of G53D-associated diabetes

Homology modeling and ligand-docking analyses of Kir6.2 led to a detailed structural view of the ATP-binding site formed by residues in the N-termini and C-termini of Kir6.2 (Fig. 1) (15). In support of the model, residues that comprise the ATP-binding pocket decrease ATP-sensitivity when mutated, but without altering open probability. Consistently, residues located in regions outside the ATP-binding pocket decrease ATP-sensitivity when mutated, but do so indirectly through an increase in p_o,zero. The G53 residue is located at the N-terminal end of the slide helix of Kir6.2, and at a predicted distance of ∼13 Å from the bound molecule of ATP (Fig. 1). Immediately adjacent to G53 in the primary structure (and located between G53 and the bound ATP molecule) is Q52, which when mutated to arginine also causes NDM (Q52R) (13). However, in this case, the Q52R mutation clearly reduces ATP-sensitivity indirectly through an increase in the intrinsic open probability (20). Both nonstationary NA and PIP₂ estimates demonstrate that p_o,zero does not differ significantly between WT and G53 channels. Thus, the effect of the G53D mutation on ATP-sensitivity is clearly on ATP affinity, yet the residue is not physically part of the ATP binding site.

Recent molecular modeling, together with functional studies, supports a putative structure of the PIP₂ binding site on Kir6.2 that includes several residues in the slide helix region of the protein (38–41). Of these, R54 is predicted to interact electrostatically with the negative charges of the anionic headgroup of PIP₂. Therefore, given the close proximity of the slide helix region to the PIP₂ binding site, mutating Q52 to a positively charged arginine might electrostatically increase the binding affinity of PIP₂, which could then underlie the observed increase in p_o,zero, and the concurrent decrease in ATP-sensitivity of the reconstituted channel. Consistent with this interpretation, replacing Q52 with an uncharged glycine residue is not predicted to alter PIP₂ binding, explaining the lack of effect of the Q52G mutation itself on channel activity (Fig. 5).

Residue G53 as an interfacial gating hinge

Several lines of evidence support a requirement for rotational flexibility at residue G53 for normal ATP-sensing. First, all amino-acid substitutions at G53 were poorly tolerated with respect to ATP-sensitivity, including alanine, which has the second simplest side chain (−CH₃). Second, the Ramachandran plot of the Kir6.2 structural model reveals torsion angles around the α-carbon of G53 that are not found in most other amino acids in the structure. Third, introducing Q52G on the G53D background, which is expected to restore flexibility of the region, results in a decrease in free energy of ATP-binding, reflected in a leftward shift in the ATP-dose response curve. Although the Q52G/G53D double mutation decreases the open probability of reconstituted channels, this alone cannot fully account for the observed decrease in K_1/2,ATP (Fig. 6).

Because the G53D mutation does not alter channel-gating, a logical interpretation of this data is that rotational flexibility is required for 1), directly regulating access of ATP to its binding pocket, 2), supporting a proper fit after ATP is bound, or 3), transducing the signal of ATP-binding to channel closure. Modeled simulations of PIP₂ binding to Kir6.2 demonstrate that the ATP-binding residue, R50, resides in a flexible loop that acts as a “gate” for ATP-binding (38). In the closed, locked position, R50 interacts electrostatically with the γ-phosphate of the bound ATP molecule. However, with PIP₂ stimulation, this flexible region moves, breaking the electrostatic interaction between the R50 residue and ATP, and carrying R50 away from the binding pocket, thereby decreasing ATP-binding affinity. Given the close connection of G53 to R50, and the observation that rotational flexibility is important at the G53 position, we speculate that G53 may act as “gating hinge”, separate from the gate to ion conductance, necessary for flexibility of the proximal N-terminal region containing R50 (Fig. 7). Thus, any mutations in G53 that restrict movement of the “hinge” would decrease ATP-affinity. The existence of a glycine “hinge” will, of course, require detailed structural information that is not available in the current data set, in which we examined only the functional consequences of G53 mutations. Nevertheless, it is interesting that the glycine at position 53 is also found in the Kir6.1 channel, but is poorly conserved among other Kir channels (42). This may not be unexpected, insofar as Kir6.1 and Kir6.2 channels are the only members of the Kir family that are regulated by nucleotide-binding directly to the Kir subunit (43).

FIGURE 7.

Proposed model for role of rotational flexibility at G53 position in control of ATP binding. Proposed movement of G53 residue that regulates ATP-binding is depicted. Simulations suggest that R50 resides in a region that forms a “gate” to the binding pocket (34,38). Reversible interaction of the channel with PIP₂ regulates movement of the gate, and therefore ATP-affinity. Residue G53 is hypothesized to represent a flexible “gating hinge” or, at the very least, a critical, flexible region necessary for high-affinity ATP-binding. Introduction of side chains at this position is predicted to alter ATP-binding efficacy.