Regulation of Cardiac ATP-sensitive Potassium Channel Surface Expression by Calcium/Calmodulin-dependent Protein Kinase II*

Background: Surface expression of cardiac ATP-sensitive potassium (KATP) channels impacts cellular energy homeostasis. Results: Activation of calcium/calmodulin-dependent protein kinase II (CaMKII) results in KATP channel internalization, requiring specific motifs on the Kir6.2 channel subunit. Conclusion: CaMKII phosphorylation of Kir6.2 promotes endocytosis of cardiac KATP channels. Significance: This mechanism reveals new targets to improve cardiac energy efficiency and stress resistance. Cardiac ATP-sensitive potassium (KATP) channels are key sensors and effectors of the metabolic status of cardiomyocytes. Alteration in their expression impacts their effectiveness in maintaining cellular energy homeostasis and resistance to injury. We sought to determine how activation of calcium/calmodulin-dependent protein kinase II (CaMKII), a central regulator of calcium signaling, translates into reduced membrane expression and current capacity of cardiac KATP channels. We used real-time monitoring of KATP channel current density, immunohistochemistry, and biotinylation studies in isolated hearts and cardiomyocytes from wild-type and transgenic mice as well as HEK cells expressing wild-type and mutant KATP channel subunits to track the dynamics of KATP channel surface expression. Results showed that activation of CaMKII triggered dynamin-dependent internalization of KATP channels. This process required phosphorylation of threonine at 180 and 224 and an intact 330YSKF333 endocytosis motif of the KATP channel Kir6.2 pore-forming subunit. A molecular model of the μ2 subunit of the endocytosis adaptor protein, AP2, complexed with Kir6.2 predicted that μ2 docks by interaction with 330YSKF333 and Thr-180 on one and Thr-224 on the adjacent Kir6.2 subunit. Phosphorylation of Thr-180 and Thr-224 would favor interactions with the corresponding arginine- and lysine-rich loops on μ2. We concluded that calcium-dependent activation of CaMKII results in phosphorylation of Kir6.2, which promotes endocytosis of cardiac KATP channel subunits. This mechanism couples the surface expression of cardiac KATP channels with calcium signaling and reveals new targets to improve cardiac energy efficiency and stress resistance.

sumption in the compartmentalized cellular environment (17,22) allows K ATP channel opening to be regulated not only by severe metabolic stress, such as during ischemia, but also by heart rate acceleration within the range induced by normal physical activity (23,24). When activated by a reduced ATP/ ADP ratio, reflecting either increased cellular metabolic demand or reduced cellular ATP generation, K ATP channel-dependent potassium efflux shortens cardiac action potential duration, allowing for a longer diastolic interval that supports myocardial relaxation and restoration of ion gradients and energetic resources as well as limits sodium and calcium entry into the cell and thus reduces energy requirements for ion transport/exchange and contraction (1,9,10,(23)(24)(25)(26)(27)(28).
There is a growing body of evidence that the ability of cardiac K ATP channels to affect cellular excitability and function depends on their abundance at the membrane surface (24, 25, 27, 29 -31). For instance, in the heart an increase in functional sarcolemmal K ATP channel presence enhances the speed and degree of shortening of action potentials and limits cardiac energy consumption in response to escalating workloads, whereas a decrease in their presence has the opposite effect (24). Additionally, although even a small fraction of the normal population of K ATP channels can dramatically shorten the action potential duration in response to severe metabolic stress, the speed at which this myocardially protective action occurs is K ATP channel expression-dependent (25). Defining the mechanisms that control cardiac membrane K ATP channel expression and associated metabolic signaling is an important step toward understanding how to promote myocardial energy efficiency and stress resistance to prevent or treat cardiac disease.
Myocardial effects of K ATP channel current up-regulation are mirrored in many ways by inhibition of calcium/calmodulin-dependent protein kinase II (CaMKII), including shortened action potentials, improved resistance to cell death under metabolic stress, and normalized intracellular calcium homeostasis (32)(33)(34), suggesting a potential interaction between K ATP channels and CaMKII in a common calcium-related regulatory pathway. CaMKII is a multifunctional kinase, densely expressed in cardiomyocytes, that targets numerous proteins involved in excitation-contraction coupling and excitability to support short-term enhancement of cardiac performance, whereas persistent activation results in adverse cardiac remodeling and dysfunction (35)(36)(37)(38). Indeed, inhibition of CaMKII has been shown to increase cardioprotective K ATP channel surface presence and current density (29). However, the molecular mechanism underlying this regulation is unknown.
Here we identify CaMKII as a critical regulator of the surface expression and consequent whole cell current capacity of cardiac K ATP channels. We find that CaMKII activation triggers rapid internalization of cardiac K ATP channels by endocytosis. This process requires phosphorylation of threonine at positions 180 and 224 and an intact YXXØ endocytosis motif, 330 YSKF 333 of Kir6.2. A molecular model of Kir6.2 predicts that the 2 subunit of the endocytosis adaptor protein, AP2, docks with the K ATP channel by interacting with 330 YSKF 333 and Thr-180 on one and Thr-224 on an adjacent Kir6.2 subunit. These findings provide new insight into the regulation of cardiac K ATP channels and provide valuable new targets for the promotion of cardiac energy efficiency and stress resistance.

EXPERIMENTAL PROCEDURES
Mouse Models-Transgenic mice expressing a specific peptide inhibitor of CaMKII (AC3-I, which targets a conserved region of the CaMKII regulatory domain) or a scrambled noninhibiting peptide (AC3-C), both under control of the cardiacspecific Myh6 promoter (32), were compared with wild-type littermates. The AC3-I mice have been shown to exhibit increased membrane K ATP channels not caused by increased transcription or elevated total protein expression (29).
Patch Clamp-Studies were performed using an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA) integrated with a Nikon TE2000-U microscope. Experiments were performed at 33-35°C using a temperature controller TC2r (Cell MicroControls, Norfolk, VA).
All recordings were monitored, sampled, and analyzed using pCLAMP software (Molecular Devices). K ATP channel activity in membrane patches (inside-out configuration) was calculated relative to activity in the absence of ATP and fitted by the Hill equation ( 50 ] is the half-maximal inhibitory ATP concentration. Single channel activity was analyzed as previously described (39).
Whole Heart Biotinylation-Hearts were excised from anesthetized mice and retrogradely perfused at 90 mm Hg with Krebs-Henseleit buffer bubbled with 95% O 2 , 5% CO 2 at 37°C and pH 7.4. The AV node was mechanically ablated, and hearts were paced at 150 or 100 ms cycle length (i.e. 400 or 600 beats/ min, Bloom Electrophysiology, Fischer Imaging Corp., Denver, CO) using a platinum pacing catheter positioned in the right ventricle (NuMed; Hopkinton, NY). Coronary flow was measured in series with the aortic cannula (T402, Transonic Systems, Ithica, NY). After 25 min the hearts and perfusion system were cooled to 4°C (Isotemp 3006D, Fisher Scientific, Inc., Pittsburgh, PA) and perfusion buffer was switched from Krebs-Henseleit buffer to PBS containing 1.0 mg/ml Sulfo-NHS-biotin (ThermoScientific) and supplemented with glucose 5 mmol/liter. During the cooled stage, hearts were asystolic, pacing was discontinued, and perfusion was changed from constant pressure to constant flow to maintain coronary flow at the level measured immediately before cooling. After 25 min of perfusion with biotin, hearts were perfused with PBS containing 15 mmol/liter glycine for 5 min at 4°C to quench unbound biotin. Immediately after this ventricles were freeze-clamped in liquid nitrogen. The tissue was lysed in buffer containing 10 mmol/liter Tris-Base, 50 mmol/liter NaCl, 5 mmol/liter EDTA, 50 mmol/liter NaF, 30 mmol/liter Na 4 P 2 O 7 , 1.25% Triton X-100, and protease inhibitor mixture (Roche Applied Science) at pH 7.4. Biotinylated (cell surface) proteins were isolated by incubating the tissue lysate with immobilized NeutrAvidin beads (ThermoScientific) for 12 h at 4°C. After separation by 3-8% gradient Nu-Page Tris acetate gel (Invitrogen), biotinylated Kir6.2 was detected by immunoblot with anti-Kir6.2 antibody (Novus) and HRP-conjugated secondary antibody (Santa Cruz Biotechnology).
Cell Culture and cDNA Expression-HEK293T/17 cells (ATCC, CRL-11248) were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum, 2 mmol/liter glutamine, and penicillin-streptomycin. Cells were grown on coverslips for immunofluorescence in a 24-well plate and in 100 ml of tissue culture dishes for immunoprecipitation and phosphorylation assays. Cells were transfected with pcDNA constructs for WT or mutant K ATP channel subunits and CaMKII (see below) using Lipofectamine 2000 (Invitrogen).
Confocal Fluorescence Imaging-Cells were imaged on a Zeiss LSM 510META laser scanning confocal microscope under a 63ϫ oil immersion lens. Middle cuts through cells were analyzed using ImageJ Software (National Institutes of Health, public domain). The ratio of intracellular:total intensity was calculated.
Immunoprecipitation-The left ventricle was isolated from WT mice and freeze-clamped. The tissue was ground in liquid N 2 until powder. Tissue was added to lysis buffer containing protease and phosphatase inhibitors (Roche Applied Science), vortexed, and placed on ice for 1 h with sonication for 15 s every 15 min. After centrifugation at 14,000 rpm for 1 h, protein was quantified (Bio-Rad). The lysed ventricle tissue sample (100 g/100 l of total protein), and lysis buffer with protease inhibitors (100 l) was added to Dynabeads (Bio-Rad) conjugated with anti-Kir6.2 (Santa Cruz Biotechnology) or anti-CaMKII (Cell Signaling) antibodies. After elution, Western blot with anti-CaMKII or anti-Kir6.2 antibodies was performed.
Phosphorylation-HEK cells transfected with WT and mutant Kir6.2-HA and SUR were lysed (see method under whole heart biotinylation), and the HA-tagged proteins were isolated using immunoprecipitation with agarose complexed with anti-HA antibody (ThermoScientific). Agarose beads complexed with channel proteins were resuspended in Hepes/ magnesium acetate buffer: 50 mM Hepes, 10 mM magnesium acetate, 0. The peptides (10 g each) were incubated with [␥-32 ]ATP and activated CaMKII. Phosphorylation was assessed by quantifying 32 P incorporation using densitometric analysis of Western blots after separation on a 4 -12% Bis-tris gel (Invitrogen) by electrophoresis at 200 V for 30 min.
Homology Modeling of Kir6.2-The amino acid sequence of mouse Kir6.2 (UniProt ID Q61743) was employed as a query for homology modeling. The N-terminal segment Met-1-Arg-25 and the C-terminal segment Ler-355-Ser390 were deleted before any calculation. The mouse Kir3.2 channel in closed state co-crystallized with K ϩ , Na ϩ , and a phosphatidylinositol diphosphate analog (PDB coda 3SYA) was used as a template for the homology modeling. The sequence of Kir6.2 was aligned to that of Kir3.2 using T-Coffee (43). The alignment was manually refined with VMD1.9.1 and the Multiseq plugin. Homology modeling was performed with Modeler 9.10. One hundred models of the Kir6.2 homotetramer were generated and ranked according to their DOPE score. The model with the lowest score was selected. The best model was minimized in five successive stages using the CHARMM27 force field and NAMD2.9: 1) minimization of the hydrogen atoms (1,000 steps); the position of non-hydrogen atoms was restrained by a force constant of 2 kcal/mol/Å 2 ; 2) minimization of the amino acid side chains (2,000 steps) with position restraints applied on carbons C␤ plus the backbone; 3) minimization of the whole amino acid side chains (4,000 steps) with position restraints applied on the backbone; 4) minimization (10,000 steps) with position restraints applied on carbons C␣; 5) minimization of the whole Kir6.2 homotetramer (15,000 steps). The phosphorylation (PO 4 2Ϫ ) of Thr-180 and/or Thr-224 was introduced before stage 1. The final models were checked with the What If and the JCSG protein structure validation servers. One disulfide bridge between Cys110 and Cys142 was defined during the calculation. The coordinates of the four K ϩ ions and the four phosphatidylinositol diphosphate analogues initially present in the x-ray structures of Kir3.2 were preserved in the Kir6.2 model. The position of all their non-hydrogen heteroatoms was restrained over all the minimization process. The phosphatidylinositol diphosphate analog was modeled using the CHARMM36 parameters. Analysis of the results was performed with VMD 1.9.1.
The model of 2 is based on the published structure of 2 co-crystallized with the TGN38 internalization signal peptide DYQRN (PDB code 1BXX). The peptide YQ was superimposed with the YS of YSQKF and 2 positioned by trial and error. The position with the fewest steric conflicts was selected.
Statistical Analysis-Results are expressed as the mean Ϯ S.E. Comparisons between two groups were made using the 2-sided Student's t test and appropriate post hoc tests. A p value Ͻ 0.05 was considered statistically significant.

RESULTS
CaMKII Activation Reduces Whole Cell K ATP Channel Current and Cell Membrane Expression-Transgenic mice expressing a specific peptide inhibitor of CaMKII (AC3-I) exhibit an increased presence of membrane K ATP channels not caused by increased transcription or elevated total protein expression (29). CaMKII is activated by signals that increase intracellular calcium, including catecholamine stimulation (32). Using real-time patch clamp monitoring of the K ATP channel, whole cell current induced by the opener pinacidil, and metabolic inhibition by DNP, we found that application of isoproterenol significantly reduces K ATP current in isolated ventricular cardiomyocytes within 2-3 min (Fig. 1, A and D, fractional reduction ϭ 0.66 Ϯ 0.1, p Ͻ 0.05). A similar significant and prompt reduction was found in ventricular cardiomyocytes from transgenic hearts expressing a non-CaMKII-inhibiting control peptide, AC3-C (Fig. 1, B and D, fractional reduction ϭ 0.58 Ϯ 0.08, p Ͻ 0.05, p ϭ NS versus WT). In contrast, very little K ATP channel current reduction was observed in cardiomyocytes isolated from transgenic mice expressing the CaMKII inhibitory peptide, AC3-I (Fig. 1, C and D, fractional reduction ϭ 0.14 Ϯ 0.06, p Ͻ 0.05 versus WT), supporting the hypothesis that the isoproterenol effect on reducing whole cell K ATP channel current in the WT cardiomyocytes occurred through CaMKII activation. In WT and AC3-I cardiomyocytes, the reduction of stimulated K ATP channel whole cell current in response to isoproterenol was reversed within ϳ5 min upon washout of isoproterenol, indicating that the phenomenon is not related to channel rundown (Fig. 1, A and B). K ATP channel current may also be regulated by calcium activated forms of PKC (30, 44 -49). To test for a potential role of PKC in the witnessed reduction of K ATP channel current, we examined the isoproterenol effect in the presence of the PKC inhibitor, chelerythrine. Despite pretreatment with chelerythrine, isoproterenol induced a significant decrease in whole cell K ATP channel current in isolated cardiomyocytes (Fig. 1, E-G, additional fractional reduction with isoproterenol ϭ 51.2 Ϯ 7.6, p Ͻ 0.05 versus reduction by chelerythrine alone).
The interaction of CaMKII and K ATP channels in cardiomyocytes was further confirmed by co-immunoprecipitation. Anti-CaMKII Ig co-immunoprecipitated Kir6.2 and anti-Kir6.2 antibodies co-immunoprecipitated CaMKII from detergentsoluble lysates generated from left ventricles of WT mice (Fig. 1,  H and I).
In the presence of increased pacing rates, previously linked to CaMKII activation (50), the membrane fraction of ventricular lysates from biotinylated isolated hearts (WT and AC3-C) exhibited lower Kir6.2 membrane expression (fractional difference in ratio of Kir6.2 expression normalized to Na ϩ /K ϩ ATPase expression at 100 compared with 150-ms pacing cycle length ϭ Ϫ0.76 Ϯ 0.09 for WT and Ϫ0.66 Ϯ 0.18 for AC3-C, p Ͻ 0.05). This difference is similar in magnitude to the whole cell current change observed with isoproterenol treatment of isolated WT and AC3-C cardiomyocytes (Fig. 1D), indicating that the current change can be mostly accounted for by a reduction in sarcolemmal expression of K ATP channels. A reduction in surface channel expression at higher pacing rates as assessed by whole heart biotinylation was not seen in hearts expressing a peptide inhibitor of CaMKII (AC3-I, Fig. 1, J and K, fractional difference .01Ϯ.13, p ϭ NS), similar to the absence of whole cell current reduction observed in isolated cardiomyocytes from AC3-I mice (Fig. 1D). The expression of Na ϩ /K ϩ ATPase and GAPDH in membrane and total fractions, respectively, did not show significant changes with faster pacing (p ϭ NS).
We next developed a heterologous cell expression system to dissect the components of the CaMKII-K ATP channel pathway. We examined HEK cells with and without exogenous overexpression of CaMKII and found that HEK cells had very little native CaMKII expression ( Fig. 2A). We then performed a densitometric analysis of immunoblots from HEK cells with exogenous CaMKII overexpression and found that their exposure to the calcium ionophore A23187 resulted in a significant increase in phosphorylated (activated) CaMKII (0.75 Ϯ 0.09 versus 1.29 Ϯ 0.07, n ϭ 3 each, p Ͻ 0.05).
To confirm the effect of CaMKII activation on the surface expression of K ATP channels in this model, we used a cell surface biotinylation assay. Extracted total and biotinylated proteins from HEK cells transfected with CaMKII and the cardiac K ATP channel subunits Kir6.2 and SUR2A with and without CaMKII activation by A23187 were assayed by Western blot. This indicated that CaMKII activation reduced the fraction of surface versus total HA-tagged Kir6.2 subunits in a time-dependent manner (Fig. 2, B and C, % of expression at time 0 ϭ 67 Ϯ 8 at 10 min and 17 Ϯ 11 at 20 min, p Ͻ 0.05).
These changes in the presence of K ATP channels at the cell membrane were further confirmed by measuring whole cell K ATP channel current by patch clamp. A23187 dramatically reduced K ATP channel current (Fig. 2, D and E, n ϭ 4, fractional reduction ϭ 0.49 Ϯ 0.04, p Ͻ 0.05). Similar to the finding in cardiomyocytes, the effect was reversible with washout of the ionophore (Fig. 2D). The dependence of this phenomenon on CaMKII activation was confirmed in HEK cells expressing K ATP channel subunits but without exogenous CaMKII overexpression. In these cells very little K ATP channel current reduction by A23187 was observed (Fig. 2E, n ϭ 5, fractional reduction ϭ 0.06 Ϯ 0.03, p Ͻ 0.05 versus cells with CaMKII). These biochemical and biophysical data indicate that CaMKII directly interacts with cardiac K ATP channels and that CaMKII activation promotes a reversible reduction in the presence of K ATP channels at the cell membrane.
CaMKII Activation Reduces K ATP Channel Membrane Expression through Endocytosis-The above whole cell patch clamp experiments and biotinylation experiments in HEK cells expressing K ATP channel subunits and CaMKII indicate a reversible down-regulation of membrane K ATP channel subunits and K ATP channel current without a change in total cellular Kir6.2, suggesting endocytosis as a possible mechanism. To test this, we tracked whole cell K ATP channel current when endocytosis was prevented by the addition of a dominant negative dynamin peptide (30) to the patch pipette solution. We found that K ATP current reduction in response to CaMKII acti-vation by A23187 was eliminated in the presence of dominant negative (DN) dynamin (Fig. 2, F and G). In contrast, an inactive scrambled peptide had no effect on the K ATP channel current response to A23187 (Fig. 2G, n ϭ 5 each, fractional current reduction ϭ 0.40 Ϯ 0.05 for scrambled, p Ͻ 0.05, and 0.03 Ϯ 0.04 for dominant negative dynamin, p ϭ NS versus base line). These data implicate enhanced endocytosis of channel subunits as the dominant mechanism for the witnessed CaMKII activa-tion-dependent decrease in whole cell K ATP channel current capacity.
We confirmed these data using immunofluorescence to detect internalization of labeled surface Kir6.2 subunits in response to A23187. HEK cells transfected with Kir6.2-HA, SUR2A, and CaMKII-GFP were incubated with an anti-HA antibody to allow the antibodies to bind the surface channels and undergo internalization. We then treated cells with FIGURE 1. CaMKII effect on native K ATP current in isolated ventricular cardiomyocytes. Shown is representative whole cell K ATP channel current, stimulated by 100 mol/liter pinacidil (pin) and 50 mol/liter DNP, measured during application and wash-out of the CaMKII activator 500 nmol/liter isoproterenol (iso) in isolated cardiomyocytes from WT mice (A), trangenic mice expressing a scrambled, control peptide AC3-C (B), and transgenic mice expressing the CaMKII inhibiting peptide, AC3-I (C). pF, picofarads. D, shown are summary data: AC3-I (n ϭ 12) versus WT (n ϭ 4) and AC3-C (n ϭ 3); *, p Ͻ 0.05. E, representative whole cell K ATP channel current, stimulated by 100 mol/liter pinacidil and 50 mol/liter DNP, was measured during application of the PKC inhibitor, 5 mol/liter chelerythrine (che), and 500 nmol/liter isoproterenol in an isolated WT murine cardiomyocyte. F, shown are summary data for chelerythrine and isoproterenol effects on pinacidil-and DNP-stimulated K ATP channel current in isolated WT murine cardiomyocytes (n ϭ 5; *, p Ͻ 0.05 versus base line, the isoproterenol bar represents additional reduction beyond that attributed to chelerythrine application). G, shown is an example of whole cell current tracings measured at points marked from the graph in E. H, On left ventricular lysates from WT mice, immunoprecipitation was performed with the anti-Kir6.2 antibody and probed for CaMKII by Western blot (IB) with the anti-CaMKII Ab. I, immunoprecipitation was performed with the anti-CaMKII Ab and probed for Kir6.2 by Western blot with the anti-Kir6.2 Ab. J, shown are Western blots of biotin-labeled membrane and whole cell fractions of ventricular lysates from isolated hearts paced at 150 or 100 ms cycle length (C.L.). Membrane and total protein fractions are probed with anti-Kir6.2 antibody and with antibodies for Na ϩ /K ϩ ATPase and GAPDH as controls. K, shown are summary data for the fractional change in Kir6.2 expression with pacing cycle length of 100 versus 150 ms (*, p Ͻ 0.05, n ϭ 2 hearts under each pacing condition for WT and AC3-I and n ϭ 3 each for AC3-C).
A23187, permeabilized, and stained them with red fluorescence-conjugated secondary antibody. Thus, only channels that were initially at the cell membrane and either remained there or became internalized were labeled. When CaMKII was not overexpressed (Fig. 3, A and B) or when CaMKII was overexpressed but cells were not treated with A23187 (Fig. 3, C and  D), we detected bright red fluorescence only on the cell surface (n ϭ 16 cells from 1 transfection, fluorescence ratio 0.52 Ϯ 0.04, p Ͻ 0.05 versus control and n ϭ 16 cells from 1 transfection, fluorescence ratio .52Ϯ.03, p Ͻ 0.05 versus control, respectively). However, when cells had CaMKII overexpression and were treated with A23187 (Fig. 3, E and F), we detected distinct puncta of red fluorescence in the cytoplasm, consistent with K ATP channel subunit internalization (n ϭ 34 cells from 2 transfections, cytosolic/total cellular intensity ratio ϭ 0.71 Ϯ 0.02). This effect was eliminated in CaMKII-overexpressing cells that had treatment with A23187 and the membrane-permeable endocytosis inhibitor dynasore (Fig. 3, G and H, n ϭ 18 cells from 1 transfection, fluorescence ratio 0.48 Ϯ 0.01, p Ͻ 0.05 versus control). Taken together, measurement of membrane currents, biotinylation of surface proteins, and immunodetection of K ATP channel subunits indicates that CaMKII activation FIGURE 2. CaMKII effect on recombinant K ATP channels in HEK cells. A, HEK cells with/without exogenous overexpression of CaMKII and with/without exposure to A23187 5 mol/liter were lysed and probed by immunoblot. GAPDH expression served as a control. B, HEK cells expressing Kir6.2-HA, SUR2A, and CaMKII-GFP underwent exposure to 10 mol/liter A23187 for 0, 10, or 30 min 37°C. At each time point, groups of cells were cooled to 4°C and biotinylated. Surface and total protein from lysed cells were assayed for Kir6.2 by anti-HA immunoblot. Cells expressing only CaMKII-GFP without K ATP channel subunits served as the negative control. C, shown are summary data of biotinylation experiments in HEK cells, indicating a time-dependent reduction in Kir6.2 surface expression with A23187 exposure (n ϭ 3 each; *, p Ͻ 0.05). D, shown is representative whole cell K ATP channel current, stimulated by 50 mol/liter pinacidil (pin) and 50 mol/liter DNP, measured during application and wash-out of the 5 mol/liter calcium ionophore A23187 in HEK cells engineered to express Kir6.2, SUR2A, and CaMKII. pF, picofarads. E, shown are summary data from these experiments performed in the presence and absence of CaMKII overexpression (*, p Ͻ 0.05). F, the same experiment was performed as in D but with dialysis of the endocytosis inhibitor (DN) dynamin 100mol/liter through the patch pipette. G, shown are summary data from the experiments performed the presence of dominant negative dynamin or a non-inhibiting, scrambled peptide (100 mol/liter each, n ϭ 5 each; *, p Ͻ 0.05).
reduces the cardiac K ATP channel current capacity of cells by enhancing endocytosis.
CaMKII-dependent Endocytosis of K ATP Channels Does Not Require the SUR Subunit-Because cardiac K ATP channels are composed of pore-forming and regulatory subunits, we sought to determine whether both were required for CaMKII-triggered endocytosis. WT Kir6.2 requires SUR for normal assembly and trafficking to the cell membrane. To circumvent the interdependence of Kir6.2 and SUR, we transfected HEK cells with an HA-tagged Kir6.2 truncation mutant (Kir6.2⌬36) that can traffic to the membrane without the SUR subunit (51). Application of A23187 resulted in internalization of the Kir6.2⌬36-HA subunit to a similar degree as HA-tagged WT K ATP channels as demonstrated by immunofluorescence (Fig.  3, I and J, n ϭ 36 cells from 3 transfections, cytosolic/total cellular intensity ratio ϭ 0.58 Ϯ 0.02, p ϭ NS versus WT) with a significant reduction in DNP-stimulated K ATP channel current as demonstrated by whole cell patch clamp (Fig. 3, K and L, n ϭ 5, fractional reduction ϭ 0.89 Ϯ 0.04, p Ͻ 0.05). Similar findings were also seen for Kir6.2⌬26, which can also traffic to the membrane without SUR (51) (Fig. 3L, n ϭ 4, fractional reduction ϭ 0.95 Ϯ 0.02, p Ͻ 0.05, immunofluorescence not shown, n ϭ 24 cells from 3 transfections, cytosolic/total cellular intensity ratio ϭ 0.60 Ϯ 0.02, p ϭ NS versus WT channels). These findings support that CaMKII-induced internalization of membrane cardiac K ATP channels is mediated through Kir6.2 and does not require SUR.
Phosphorylation of Thr-180 and Thr-224 on Kir6.2 Are Required for K ATP Channel Endocytosis-CaMKII regulates the function of proteins by targeting specific serines and threonines embedded within consensus sequences. To further investigate regions on Kir6.2 that may be required for CaMKII-mediated K ATP channel endocytosis, we investigated four sites identified through scanning the Kir6.2 sequence for the RXX(S/T) motif and through literature reports of sites with other serine/threonine kinase interactions (44,45,47,52,53). We identified candidate sites at Ser-37, Thr-180, Thr-224, and Ser-372. In experiments to determine the impact of these sites on endocytosis of cardiac K ATP channels, we found no effect of S37A and S372A mutations on K ATP channel current or membrane expression (data not shown), and thus further results and discussion focus only on the Thr-180 and Thr-224 sites.
We probed isolated WT and mutant Kir6.2-HA-containing channels for in vitro phosphorylation by CaMKII. Mutation of Thr-180 or Thr-224 to a non-phosphorylatable alanine (T224A and T180A) resulted in only a faint signal for incorporation of the radio-labeled ␥-phosphate from ATP in the presence of activated CaMKII, and this signal was further attenuated in the presence of double T180A/T224A mutation in Kir6.2 compared with WT (Fig. 4A).
Next, we generated segments of the Kir6.2 subunit surrounding Thr-224 and Thr-180 to further demonstrate phosphorylation of these sites by CaMKII (Fig. 4B). Mutation to T224A and T180A in these fragments eliminated incorporation of the radiolabeled ␥-phosphate from ATP in the presence of activated CaMKII (Fig. 4B).
Similarly, in HEK cells expressing CaMKII, SUR2A, and Kir6.2T224A-HA, the whole cell K ATP channel current reduction after CaMKII activation was diminished compared with cells with HA-tagged WT Kir6.2 subunits (Fig. 4, J and K, n ϭ 6, FIGURE 4. CaMKII effect in Kir6.2T224A and Kir6.2T180A mutants. A, in isolated recombinant cardiac K ATP channels with WT or mutated Kir6.2, an in vitro phosphorylation assay was performed in the presence and absence of CaMKII. Kir6.2-HA subunits were mutated from threonine to alanine at Thr-224 or Thr-180 or both. Kir6.2-HA is identified as a double band at ϳ37 kDa. The prominent signal near 50 kDA level represents autophosphorylated CaMKII. B, in WT and mutated peptide fragments of Kir6.2, in vitro phosphorylation assay was performed in the presence and absence of CaMKII. Fragments were mutated from threonine to alanine at Thr-224 or Thr-180. Differences in band migration are attributed to different sizes and charge of the fragments. The WT Thr-180 and mutant T180A fragments are identified at ϳ3.3 kDa (best seen for Thr-180 ϩ CaMKII and faintly for T180A ϩ CaMKII), whereas the Thr-224 and T224A fragments are identified at ϳ4.5 kDa (best seen for Thr-224 ϩ CaMKII). C-H, representative immunofluorescence confocal imaging was performed in HEK cells following the same protocol as in Fig. 3. fractional reduction ϭ 0.05 Ϯ 0.03 for T224A versus n ϭ 4, fractional reduction ϭ 0.49 Ϯ 0.04 for WT, p Ͻ 0.05). However, although surface presence of K ATP channels in HEK cells expressing CaMKII, SUR2A, and Kir6.2T180A-HA was clearly seen by immunofluorescence (Fig. 4F), we were unable to identify significant pinacidil and DNP-stimulated whole cell K ATP channel current (n ϭ 20 cells). In rare cases (n ϭ 4 of 30 patches), a single K ATP channel was identified by inside-out patch clamp in the presence of low (0 -25 M) ATP in cells expressing Kir6.2T180A-HA (Fig. 5C). Similar electrophysiologic findings were obtained in cells transfected instead with Kir6.2T180A without the HA tag (data not shown).
We also examined whether there could be a contribution to whole cell K ATP channel current reduction by a CaMKII activation-dependent change in K ATP channel biophysical properties. We observed a rightward shift in the ATP sensitivity (decreased ATP sensitivity) of K ATP channels in inside-out patches from cells expressing WT K ATP channel subunits and constitutively active CaMKII (n ϭ 7, Hill ϭ 1, IC 50 ϭ 290 M) versus WT CaMKII (n ϭ 3, Hill ϭ 1.3, IC 50 ϭ 110 M, Fig. 5A). Expression of Kir6.2T224A with SUR2A and constitutively active CaMKII eliminated the rightward shift in ATP sensitivity (n ϭ 4, Hill ϭ 1.6, IC 50 ϭ 128 M, Fig. 5B), whereas expression of Kir6.2T224E (to mimic phosphorylation of the Kir6.2 subunit) with SUR2A and constitutively active CaMKII restored the rightward shift in ATP sensitivity (n ϭ 7, Hill ϭ 0.96, IC 50 ϭ 360 M, Fig. 5B). These data indicate that CaMKII-dependent phosphorylation of Kir6.2 at Thr-224 results in a reduction in K ATP channel ATP sensitivity. However, this decrease in ATP sensitivity of the channel would oppose, not contribute to or account for, the reduction in whole cell K ATP channel current in the presence of CaMKII activation.
Mutations of the Kir6.2 Thr-180 residue resulted in loss of normal channel gating. Overall activity of channels with the Kir6.2T180A mutation was rare, and these channels appeared to have dramatically increased ATP sensitivity (n ϭ 3, Hill ϭ 1.8, IC 50 ϭ 40 M). No K ATP channel current was detected in the presence of the Kir6.2T180E mutant. These data indicate that the Thr-180 site is essential for normal regulation of channel gating.
Single channel conductance was unchanged in channels comprised of WT Kir6.2 versus Kir6.2T224A or Kir6.2T180A (71.4 Ϯ 0.8 versus 70.9 Ϯ 0.6 versus 70.1 Ϯ 0.8 picosiemens, respectively, n ϭ 3 each, Fig. 5C). Overall, these data show that CaMKII-dependent phosphorylation of the Thr-224 and Thr-180 residues of Kir6.2 has a role in regulating channel gating and is critical to promote endocytosis of the K ATP channel subunits.
The Kir6.2 Internalization Motif 330 YSKF 333 Is Required for CaMKII-induced K ATP Channel Down-regulation-Patients with a genetic form of neonatal diabetes have been found to harbor Y330C and F333I mutations of Kir6.2 that inhibit spontaneous endocytosis of K ATP channels at 37°C and result in a ϳ2-fold increase in surface channels (54). Here, we examined whether the 330 YSKF 333 sequence is required for internalization of cardiac K ATP channels driven by CaMKII activation. In HEK cells expressing Kir6.2Y330C-HA, SUR2A, and CaMKII, pinacidil-and DNP-stimulated KATP channel current was reduced very little by application of A23187 (fractional reduction 0.05 Ϯ 0.02, n ϭ 6, p Ͻ 0.05 versus 0.49 Ϯ 0.04, n ϭ 4 for WT Kir6.2, Fig. 6, A and B). Similarly, there was little cytosolic appearance of labeled Kir6.2Y330C-HA after stimulation of HEK cells with A23187 ( Fig. 6 C and D, n ϭ 44  and Thr-224 sites and the 330 YSKF 333 motif) has been shown to interact directly with the 2 subunit of the endocytosis adaptor protein, AP2 (54). Here, to investigate the effect of phosphorylation of Thr-180 and Thr-224 on this relationship, we formulated a molecular homology model of Kir6.2 based on the crystal structure of mouse Kir3.2. The model predicts that Thr-224 and Thr-180 are too distant from the 330 YSKF 333 sequence to directly interact and that phosphorylation of Thr-224 and Thr-180 is unlikely to induce large enough conformational changes to alter accessibility of the 330 YSKF 333 sequence (Fig. 7, A  and B).
However, docking of the known structure of the 2 subunit of the endocytosis adaptor protein, AP2, to the modeled Kir6.2 suggests that 2 could bind 330 YSKF 333 and interact with the nearby Thr-180 via a long loop (possibly the lysine-rich Asn-217-Asp-244, Fig. 7, C and D), assuming the ATP binding pocket was unoccupied. The 2 subunit could simultaneously interact with Thr-224 of the neighboring Kir6.2 subunit via a loop (possibly Ile-159 -Asn-171, Asn-217-Asp-244, or Val-406 -Ser-414, Fig. 7, C and D). The arginine-rich N terminus could be electrostatically compatible with a Thr-224 phosphate. Thus, the model suggests a possible role for the Thr-180 and Thr-224 residues in coordinating the docking of 2 with Kir6.2 at 330 YSKF 333 .

DISCUSSION
Cardiac K ATP channels are critical for cellular energetic homeostasis and protection from injury and arrhythmia (1, 6, 26). As such, deciphering their regulation is key to understand-ing both the normal physiology of cardiac adaptation to varying workloads and how functional/metabolic diseases such as heart failure occur and progress. This study establishes a previously unrecognized mechanism for regulation of the surface expression and whole cell current capacity of sarcolemmal K ATP channels by CaMKII.
CaMKII can be viewed as an ideal candidate for cardiac workload-dependent regulation of K ATP channels by endocytosis. First, CaMKII is a vital regulator of cardiac excitation-contraction coupling and, thus, cardiac energy consumption (35)(36)(37). Second, CaMKII activation underlies increased cardiac performance during heart rate acceleration (38,50), an established trigger of K ATP channel opening (23,24). Finally, CaMKII has been linked to endocytosis of membrane proteins in other cell types (55,56). Here, we reveal that activation of CaMKII induces endocytosis of cardiac K ATP channels. We propose that this phenomenon is critical for augmented myocardial performance in response to increased bodily demand. We previously demonstrated that a higher density of surface sarcolemmal K ATP channels results in a more rapid adjustment of ventricular action potential duration and thus higher fidelity matching of energy consumption to available metabolic resources under a sudden increase in cardiac workload or under metabolic insult (24,25). However, during persistent, intense cardiac performance and augmented energy consumption, a high surface density of K ATP channels could be a liability for the cell as the continued recruitment of K ATP channel current could render the membrane less excitable (1,5,6,57). A reduction in the overall surface abundance of K ATP channels through endocytosis promoted by CaMKII activation would effectively reduce the gain in the energetic feedback circuit, thereby supporting enhanced myocardial performance, whereas the rapid restoration of K ATP channel surface density when CaMKII activation is abated would permit the cell to "reset" to its base-line level of action potential responsiveness. This hypothesis is supported by the blunted contractility at high pacing rates of hearts expressing a CaMKII inhibiting peptide (50).
On the other hand, chronic CaMKII activation, as occurs in disease states such as heart failure (35,37,38), would prohibit resetting of the normal baseline K ATP channel surface density. This would blunt the protective shortening of action potentials under repeated acute stress, thereby predisposing myocytes to worsened energetic insult and injury that would promote disease progression. In this way the plasticity of K ATP channel surface presence can be viewed as vital for optimal cardiac function.
This series of experiments indicates that mutation of any one of Thr-180, Thr-224, or Tyr-330 is sufficient to interfere with CaMKII-mediated endocytosis. Based on our homology model of the Kir6.2 subunit and consistent with that of others (58,59), Thr-180 is likely located in or near the channel pore at the site of ATP binding and adjacent to the internalization motif 330 YSKF 333 . Our model suggests that Thr-180 is unlikely to directly interact with 330 YSKF 333 and that phosphorylation of Thr-180 is unlikely to alter accessibility to 330 YSKF 333 . However, docking of 2 may be influenced by Thr-180, assuming the ATP binding site is unoccupied. Because the non-phosphorylatable T180A mutation appears to augment ATP sensitivity of the channel, we can hypothesize that phosphorylation of Thr-180 by CaMKII would discourage ATP binding and thus allow 2 to dock, thereby permitting endocytosis. Similarly, our model predicts that Thr-224 is unlikely to interact directly with 330 YSKF 333 and that phosphorylation of Thr-224 is unlikely to affect accessibility of 330 YSKF 333 . However, it does appear that 2 at 330 YSKF 333 on one Kir6.2 subunit may interact with Thr-224 on a neighboring Kir6.2 subunit. Thus, our model suggests several potential mechanisms for how Thr-180, Thr-224, and 330 YSKF 333 on Kir6.2 may interface with the cellular endocytic machinery to permit CaMKII-triggered endocytosis of cardiac K ATP channels. These prospects will be the subject of future experimental studies.
Our data also indicate the significance of the Kir6.2 residues Thr-180 and Thr-224 and their phosphorylation state in the regulation of K ATP channel gating. The effect of the T180A mutant on K ATP channel function and current capacity is particularly intriguing. The immunofluorescence images clearly demonstrate bright fluorescence at the cell membrane thus indicating that the effect of the T180A mutation on reduced NPo of single channel K ATP channel recordings does not primarily occur through interference with surface presence of the FIGURE 7. Homology model of Kir6.2. A, a three-dimensional model of Kir6.2 presented as a solvent-accessible surface (except the phosphatidylinositol diphosphate analog) focused on the ATP binding site indicates the position of Thr-180 (red) within the ATP binding pocket (blue) adjacent to the YSKF internalization motif (green). Thr-224 (also red) is seen in the bottom right corner of the image. B, the same model shows little change in the relationship of Thr-180 or Thr-224 to YSKF with phosphorylation of both Thr-180 and Thr-224. The calculated residue displacements (root mean square deviation) induced by phosphorylation of Kir6.2 are: Kir6.2 ϩ Thr-180(P) ϭ 0.39 Å, Kir6.2 ϩ Thr-224(P) ϭ 0.17 Å, and Kir6.2 ϩ Thr-180(P) ϩ Thr-224(P) ϭ 0.43 Å. The root mean square deviation is slightly larger when Thr-180 is phosphorylated as Thr-180 is embedded in the Kir6.2 homotetramer, poorly accessible to the solvent. As a consequence, the neighbor residues as well as the backbone of Thr-180 adapt their conformation due to introduction of the bulky phosphate group. In contrast, Thr-224 is well exposed on the surface of Kir6.2. The phosphorylation of this amino acid has little consequence on neighbor residues. Both Thr-180 and Thr-224 are remote from the YSKF signal. The closest residue of the YSKF motif to Thr-180 is Tyr-330. The distance between the phosphate/hydroxyl of Thr-180 and the hydroxyl of Tyr-330 is, respectively, 9.3, 9.6, 9.3, and 9.7 Å in Kir6.2 and Kir6.2 phosphorylated at Thr-180, Kir6.2 phosphorylated at Thr-224, and Kir6.2 phosphorylated at both Thr-180 and Thr-224. The closest YSKF residue to Thr-224 is Lys-332. The distance between the phosphate/hydroxyl of Thr-224 and the ammonium of Lys-332 is, respectively, 18.8, 19.0, 19.8, and 19.8 Å in Kir6.2, Kir6.2 phosphorylated at Thr-180, Kir6.2 phosphorylated at Thr-224, and Kir6.2 phosphorylated at both Thr-180 and Thr-224. C, a side view model of Kir6.2, with two subunits removed for clarity, indicates proposed docking with 2 (large, green). The ATP binding pocket (blue), YSKF (green), Thr-180 (upper red), and Thr-224 (lower red) are also shown. D, an end-on view of the Kir6.2 model shows the relationship of docked 2 (large, green) with adjacent subunits of Kir6.2. Thr-224 (red) is largely obscured by 2. Kir6.2 subunits A and C are white, and subunits B and D are gray. mutant channel. Rather, it appears the mutation alters channel gating and/or ATP sensitivity. Interestingly, a previous report of the Thr-180 mutation showed abnormal biophysical properties of the channel as well (47). In that study, whole cell K ATP channel activity in response to openers and/or DNP, as was done in the current study, was not tested nor was the full-length Kir6.2T180A mutant in the presence of SUR2A tested in the inside-out mode as was done here, and thus a direct comparison with our results is not possible. Because we only saw rare lone channels in the full-length Kir6.2T180A mutant with SUR2A, we were unable to assess the effect of this mutation on rundown. Nonetheless, our study is in line with the findings of other investigators that indicate the Thr-180 residue is a critical determinant of channel gating behavior (47,58).
The T224A and T224(D/E) mutations in Kir6.2 have also been previously described and found to alter protein kinase A-mediated regulation of K ATP channel gating (52,53). We confirmed that mimicking phosphorylation of Thr-224, by the T224E mutation, shifts the ATP sensitivity of Kir6.2/SUR2A complexes to the right. However, this would oppose the decrease in K ATP channel current in the presence of CaMKII activation witnessed here and thus can be excluded as contributor to the phenomenon.
Taken together with previous findings that changes in K ATP channel surface abundance have a significant impact on cardiac energetics (23)(24)(25), sarcolemmal K ATP channel down-regulation via CaMKII-mediated endocytosis reveals a new mechanism for control of myocardial performance and energy use. Indeed, CaMKII inhibition has been investigated as a promising treatment and prevention strategy for heart failure (35); however, diverse tissue expression and pleiotropic vital functions of CaMKII make it a problematic target. Similarly, targeting of K ATP channel gating for cardioprotection has proven problematic, as presently available K ATP channel openers do not have cardiac specificity, and their therapeutic use is limited due to potent effects on vasculature tone and related hypotension (60,61). Thus, in identifying an important interaction between CaMKII and cardiac K ATP channels, this study provides a new set of potential targets for the promotion of cardiac energy efficiency and stress resistance.