Control of Cardiac Repolarization by Phosphoinositide 3-kinase Signaling to Ion Channels

Abstract

Upregulation of phosphoinositide 3-kinase (PI3K) signaling is a common alteration in human cancer, and numerous drugs that target this pathway have been developed for cancer treatment. However, recent studies have implicated inhibition of the PI3K signaling pathway as the cause of a drug-induced long QT syndrome in which alterations in several ion currents contribute to arrhythmogenic drug activity. Surprisingly, some drugs that were thought to induce long QT syndrome by direct block of the rapid delayed rectifier (I_Kr) also appear to inhibit PI3K signaling, an effect that may contribute to their arrhythmogenicity. The importance of PI3K in regulating cardiac repolarization is underscored by evidence that QT interval prolongation in diabetes also may result from changes in multiple currents due to decreased insulin activation of PI3K in the heart. How PI3K signaling regulates ion channels to control the cardiac action potential is poorly understood. Hence, this review summarizes what is known about the impact of PI3K and its downstream effectors including Akt on sodium, potassium and calcium currents in cardiac myocytes. We also refer to some studies in non-cardiac cells that provide insight into potential mechanisms of ion channel regulation by this signaling pathway in the heart. Drug development and safety could be improved with a better understanding of the mechanisms by which PI3K regulates cardiac ion channels and the extent to which PI3K inhibition contributes to arrhythmogenic susceptibility.

Signaling by Class I PI3Ks

Phosphoinositide 3-kinases (PI3Ks) are lipid kinases that phosphorylate the 3-hydroxyl position of the inositol head group of phosphoinositides. PI3Ks are grouped into three classes according to their primary structures, subunit composition and substrate specificity.^{1, 2} Class 1 PI3Ks phosphorylate phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] to produce the second messenger phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P₃] (Fig. 1). Class 1A PI3Ks are heterodimers consisting of a catalytic subunit (p110α, p110β or p110δ) tightly bound to a regulatory subunit (p85α, p85β, p55α, p50α or p55γ). The class 1B catalytic subunit p110γ binds to distinct p101 or p87 regulatory subunits. The mammalian heart expresses the p110α, p110β and p110γ catalytic isoforms along with several different regulatory subunits. Class 2 PI3Ks are monomers (PI3K-C2α, PI3K-C2β and PI3K-C2γ) that preferentially phosphorylate phosphatidylinositol (PI) or phosphatidylinositol 4-phosphate [PI(4)P]. The class 3 PI3K catalytic subunit Vps34 binds to a Vps15 regulatory subunit and preferentially phosphorylates PI. PI3K-C2α and Vps34 are also expressed in the heart, but most of the studies that have sought to define a role for PI3Ks in cardiac electrophysiology have centered on the class 1 PI3Ks. Our discussion will focus on these enzymes and will refer to the heterodimers as PI3Kα, PI3Kβ, etc.

Figure 1. PI3K signaling pathways regulating cardiac ion channels.

Receptor tyrosine kinases (RTKs) such as the insulin receptor activate PI3Kα to produce PI(3,4,5)P₃, which recruits Akt and PDK1 to the plasma membrane, resulting in Akt activation. RTKs can also activate atypical PKCs (aPKC) and SGK via PDK1. Gβγ subunits released from G protein-coupled receptors (GPCRs) activate PI3Kγ to increase PI(3,4,5)P₃ production and activate Akt, but the Gα_q subunits inhibit PI3Kα. Akt, PDK1, aPKC, SGK and possibly other downstream effectors of PI3K regulate ion channels that conduct potassium, sodium and calcium currents. PTEN dephosphorylates PI(3,4,5)P₃ to antagonize PI3K signaling. PI3Kγ also binds to and activates phosphodiesterases (PDE) to decrease cAMP, a second messenger that regulates many cardiac ion channels. This function of PI3Kγ is independent of its kinase activity.

Class 1 PI3Ks are tightly regulated by extracellular stimuli that alter the enzyme activity and reposition the PI3Ks to membranes where their lipid substrate is located.^{1, 2} A common feature of the class 1A regulatory subunits is the presence of two Src homology 2 (SH2) domains that mediate binding to specific phosphotyrosyl residues in proteins. Exposure of cells to insulin or other growth factors that activate receptor tyrosine kinases leads to phosphorylation of substrate proteins on tyrosyl residues. Class 1A PI3Ks are then recruited to the receptor complexes at the plasma membrane, where the activated enzymes produce PI(3,4,5)P₃ to initiate signaling cascades (Fig. 1). Experiments using mouse ventricular myocytes that express a dominant-negative mutant of p110α or that lack p110α, p110β or p110γ indicated that PI3Kα is the major PI3K isoform that couples to insulin or insulin-like growth factor-1 (IGF-1) receptors in cardiomyocytes.^{3, 4} By contrast, PI3Kγ is activated by binding to Gβγ subunits that are released upon hormone stimulation of G protein-coupled receptors (Fig. 1).^{5, 6} PI3Kβ is also activated by Gβγ subunits, but this mode of regulation has not been demonstrated to occur in cardiomyocytes.^{7, 8} On the other hand, PI3Kα is inhibited by Gα_q subunits that are released upon hormone stimulation of some G protein-coupled receptors (Fig. 1).^{9, 10} PI3Kγ has a second function in the heart that is independent of its kinase activity: it binds and activates phosphodiesterases (PDE) to decrease cAMP (Fig. 1).¹¹

Dephosphorylation of the 3-hydroxyl group of PI(3,4,5)P₃ by the lipid phosphatase PTEN regenerates PI(4,5)P₂ and terminates PI3K signaling (Fig. 1). PTEN is a tumor suppressor whose loss or inactivation leads to upregulation of PI3K signaling in a wide variety of tumors. PI(3,4,5)P₃ can also be sequentially dephosphorylated to yield phosphatidylinositol 3,4-bisphosphate [PI(3,4)P₂] and phosphatidylinositol 3-phosphate [PI(3)P].^{1, 2} These 3-phosphoinositides bind to specific domains on PI3K effector proteins to modulate their localization and activity. There are a wide variety of PI3K effectors, including kinases, adaptor proteins, and regulators of small GTPases.^{1, 2} We will limit our discussion to protein kinases that have been shown to regulate cardiac ion channels (Fig. 1). PDK1 (3-phosphoinositide-dependent protein kinase 1) and the protein kinase Akt (also known as PKB) are key PI3K effectors. Binding of PI(3,4,5)P₃ to the pleckstrin homology (PH) domain of PDK1 causes the enzyme to translocate to the plasma membrane. Binding of PI(3,4,5)P₃ or PI(3,4)P₂ to the PH domain of Akt causes it to colocalize with PDK1, enabling PDK1 to phosphorylate Akt at a site that partially activates the enzyme. Maximal activation of Akt occurs after phosphorylation of a second site by mTORC2, which itself is controlled by PI3K in an as yet undetermined way.¹² Serum- and glucocorticoid-regulated kinases (SGK) and atypical PKC isoforms are also activated by phosphorylation by PDK1 and mTORC2, and some isoforms also possess 3-phosphoinositide binding domains that contribute to their regulation.¹³

Long QT Syndromes

Long QT syndromes are a family of diseases of multiple etiologies whose common outcome is prolongation of the QT interval on the electrocardiogram (ECG). Activation of the ventricular myocardium is reflected on the ECG by the onset of the Q wave, and final repolarization is defined by the end of the T wave. Although the QT interval normally varies with cardiac rate, a pathological increase in the QT interval corrected for rate using various algorithms (i.e., QTc) indicates heightened risk for torsades de pointes, an arrhythmia which can cause sudden death.¹⁴ Primary prolongation of the QT interval (i.e., that which is independent of an altered QRS complex on the ECG) results from lengthening of the action potential duration (APD) in ventricular myocytes. An increase in depolarizing currents (sodium and calcium) or a decrease in repolarizing currents (potassium) that are major determinants of the action potential waveform can cause an increase in the myocyte APD that is manifested clinically as QT interval prolongation.

Most congenital long QT syndromes arise from mutations that cause a reduction in the outward delayed rectifier currents I_Kr or I_Ks.^15–18 Gain-of-function mutations that cause an increase in the persistent (late) sodium current (I_Na) or L-type calcium current (I_Ca,L) are found in a smaller number of patients.^19–22 The mutations can affect trafficking, gating, binding to other proteins, or other channel functions.²³ Acquired long QT syndromes are far more prevalent than the congenital forms and can arise from drug exposure,²⁴ diabetes,^25–27 or other conditions. The prevailing view regarding drug-induced long QT syndrome is that it is mainly an I_Kr disease resulting from direct blockade of the channel pore or disruption of channel trafficking to the cell surface.^28–32 We know now that some forms of drug-induced long QT syndrome, in particular those caused by tyrosine kinase inhibitors, are due to inhibition of PI3K signaling.³³

PI3K and Drug-induced Long QT Syndrome

Small molecule inhibitors of tyrosine kinases and PI3Ks have entered clinical use or are in clinical trials as anti-cancer drugs. The package inserts for the tyrosine kinase inhibitors Tasigna^® (nilotinib) and Sprycel^® (dasatinib) contain warnings about the risks of QT prolongation, cardiac arrhythmia and sudden death (www.FDA.gov/drugs/). We showed that nilotinib and dasatinib significantly prolonged the time to 90% repolarization (APD₉₀) in canine ventricular myocytes.³³ This APD₉₀ prolongation was reversed by intracellular infusion of the PI3K second messenger PI(3,4,5)P₃, whereas the control phospholipids PI(3,5)P₂ and PI(4,5)P₂ were not effective.

Tyrosine kinase inhibitors that caused APD₉₀ prolongation also reduced PI3K activity, so we next tested two nonselective PI3K inhibitors, one of which (PI-103) is widely used in vitro and the other (BEZ235) which has entered clinical trials for cancer therapy. Both of these compounds also prolonged the APD₉₀.³³ Furthermore, isoproterenol in the presence of PI-103 or BEZ235 induced early after-depolarizations (EADs) that may trigger arrhythmias. The effects of both types of inhibitor on APD₉₀ were seen after a two-hour incubation but not after acute application of 1 to 5 minutes, indicating that the drugs were probably not acting as direct channel blockers.

Nilotinib and PI-103 also caused an increase in QTc in the perfused mouse heart.³³ In searching for a PI3K whose inhibition might mediate these effects, we found that p110α-null hearts exhibited QTc prolongation. In addition, myocytes lacking p110α exhibited prolonged APD₉₀ that was reversed by intracellular application of PI(3,4,5)P₃. Deletion of PI3Kβ had little or no effect on APD₉₀.³³ Deletion of p110γ did not affect APD in the absence of calcium transients, but APD was prolonged when calcium transients were present.³⁴ Gain-of-function mutations in p110α are often found in human cancers, and development of inhibitors to target this enzyme has been a major priority of the pharmaceutical industry.³⁵ If PI3Kα plays a major role in regulating the human cardiac action potential as it does in the mouse, then we predict that PI3K inhibitors such as GDC-0941 (Genentech) and BEZ235 (Novartis) that have entered clinical trials will prolong the QT interval in patients.

To our surprise, we found that terfenadine, the non-sedating antihistamine on which the “I_Kr hypothesis” was based,^{36, 37} also appeared to inhibit PI3K signaling, as most of its effects on APD₉₀ were reversed by PI(3,4,5)P₃.³³ It will be important to learn how prevalent PI3K inhibition is among proarrhythmic drugs, including those that have been classified as I_Kr blockers. In answer to this question, a recent study showed that chronic exposure of adult mouse myocytes, which lack I_Kr, to the prototypical I_Kr blocker dofetilide caused APD prolongation that was reversed by PI(3,4,5)P₃.³⁸ In tests using I_Kr blockers from multiple therapeutic classes, six (including dofetilide) had effects consistent with PI3K inhibition while two others (including moxifloxacin) did not. Dofetilide also caused a reduction in Akt phosphorylation that was not seen with moxifloxacin.³⁸ Thus, some I_Kr blockers also inhibit PI3K/Akt signaling, and a drug with both activities might have increased potential to cause QT prolongation. Future studies are needed to characterize how I_Kr blockers such as terfenadine and dofetilide inhibit PI3K signaling and to determine the extent to which PI3K inhibition contributes to their arrythmogenic activity.

Long QT Syndrome in Diabetes

The association of diabetes with prolonged QTc and cardiovascular death is strongly supported by multiple epidemiological studies.^25–27 Experiments in animal models showed that streptozotocin-induced diabetic rats,³⁹ type 2 diabetic db/db mice⁴⁰ and alloxan-induced diabetic dogs⁴¹ and rabbits⁴² exhibited QTc prolongation and/or APD lengthening in ventricular myocytes. Hyperglycemia per se might not be the cause of these repolarization defects: mice lacking the insulin receptor only in cardiac myocytes also exhibited APD prolongation, even though the animals were euglycemic.⁴³ Since reduced production of or sensitivity to insulin in diabetes results in decreased activation of PI3K, we hypothesized that downregulation of cardiac insulin/PI3K signaling plays a role in QT interval prolongation in diabetes.⁴⁴ In support of this hypothesis, we found that APD₉₀ prolongation in ventricular myocytes of diabetic db/db mice and insulin-deficient Ins2^Akita mice was reversed by intracellular delivery of PI(3,4,5)P₃ but not control phospholipids. Adenoviral expression of constitutively active p110α also corrected APD₉₀ in cultured myocytes from both types of animals. In addition, perfused hearts from db/db and Ins2^Akita mice exhibited QTc prolongation. Circulation of insulin through the Ins2^Akita heart corrected the abnormal QTc, and this effect was blocked by PI-103.⁴⁴

Thus, acquired long QT syndromes due to diabetes and some drugs can arise from a common mechanism of suppressed PI3K signaling. It is not surprising that these two syndromes are caused by changes in multiple ion channels, some of which have been shown to be PI3K dependent. At least five cardiac currents—I_Na (peak and persistent), I_Kr, I_Ks and I_Ca,L—are affected by PI3K signaling, and the following sections will discuss these currents in more detail.

Sodium Channel

In considering the cardiac sodium current I_Na, it is important to separate effects on peak I_Na which generates the action potential upstroke and conduction speed from effects on persistent I_Na which sustains the plateau. Since the predominant evidence suggests that both are generated by the same Na_v1.5 channel protein,^{19, 45} effects on channel number or trafficking are likely to alter both currents in the same direction, while effects on gating could differentially increase one while decreasing the other. Indeed, treatment of canine ventricular myocytes for 2 hours with nilotinib or PI-103 caused a decrease in peak I_Na and an increase in persistent I_Na.³³ Effects of nilotinib on both of the sodium currents were reversed by intracellular delivery of PI(3,4,5)P₃ through the patch pipette or by extended washout of the drug. These results suggested that I_Na is regulated by PI(3,4,5)P₃, which is slowly depleted during incubation of myocytes with nilotinib or PI-103 and gradually replenished following drug washout. Two lines of evidence demonstrated that elevated persistent I_Na contributes to the drug effects on repolarization. First, treatment of myocytes with the sodium channel blocker mexiletine at a concentration selective for persistent I_Na prevented the PI3K inhibitor-induced prolongation of APD₉₀ and EAD generation.³³ Second, computer simulations of the canine ventricular action potential indicated a role of elevated persistent I_Na in the lengthening of APD₉₀.³³ Terfenadine, dofetilide and several other I_Kr blockers also caused a time-dependent increase in persistent I_Na that was reversed by PI(3,4,5)P₃.^{33, 38} These results suggest that screening drug candidates for chronic effects on persistent I_Na as well as for acute I_Kr block might improve drug safety.

To identify the PI3K isoform that regulates I_Na, the current was studied in mouse myocytes lacking p110α or p110β.³³ I_Na was not altered in p110β-null myocytes, but the p110α knockout recapitulated the effects of drugs on I_Na in canine ventricular myocytes. The changes in peak and persistent I_Na in p110α knockout myocytes were eliminated by intracellular perfusion of PI(3,4,5)P₃, and QTc prolongation was reversed after mexiletine treatment of the p110α knockout hearts.³³ This study suggests that the tyrosine kinase inhibitors and PI3K inhibitors exert their effects on I_Na through the inhibition of PI3Kα. Whether or not I_Kr blockers also target PI3Kα remains to be determined using biochemical assays.

PI3K signaling has been reported to upregulate gene expression of sodium channel subunits. Induction of constitutively active p110α in the heart of adult mice increased Akt phosphorylation and mRNA levels of both the alpha (Scn5a) and beta (Scn1b) subunits of the cardiac sodium channel.⁴⁶ Treatment of the mice with an Akt inhibitor did not block increases in channel expression, suggesting that p110α regulates sodium channel mRNA levels independently of Akt.⁴⁶ By contrast, increased transcription of Scn5a in rat ventricular myocytes exposed to transforming growth factor β-1 was attributed to activation of PI3K and subsequent phosphorylation of the transcription factor Foxo1 by Akt at a site that eliminates its ability to suppress Scn5a expression (Fig. 2).⁴⁷

Figure 2. Hypothetical modes of regulation of cardiac sodium currents by PI3Kα.

Phosphorylation of Na_v1.5 by PI3Kα/Akt suppresses the persistent sodium current; this effect is reversed by inhibitors of PI3K/Akt signaling and diabetes. Phosphorylation of Na_v1.5 by SGK1 on a separate site might increase the persistent current. Phosphorylation of NEDD4-2 by kinases downstream of PI3Kα increases peak sodium current by maintaining Na_v1.5 on the cell surface. In the presence of PI3K inhibitors, NEDD4-2 is dephosphorylated, it binds to and ubiquitinates Na_v1.5, and the channel is internalized. PI3Kα upregulates transcription of the sodium channel gene SCN5A by Akt-dependent phosphorylation and inactivation of the transcription repressor FOXO1 and by an unknown Akt-independent mechanism.

SGK1 and SGK3 (the major SGK isoforms in the heart), like Akt, can be activated by insulin, IGF-1 or constitutively active p110α in some cell types.^{48, 49} Because Akt and SGKs have similar substrate specificities and phosphorylate some proteins on the same physiologically important sites,⁵⁰ it seems possible that some effects of PI3K on persistent I_Na may be mediated by SGKs. Interest in the regulation of the cardiac sodium channel by SGKs was prompted by experiments with the amiloride-sensitive epithelial sodium channel (ENaC) that mediates sodium transport in tissues such as kidney and lung.⁵¹ ENaC binds to the ubiquitin-protein ligase NEDD4-2, leading to ubiquitination and internalization of the sodium channel.⁵² SGK1 increases channel activity at least in part by phosphorylating NEDD4-2 and blocking its interaction with ENaC, resulting in an increased number of ENaC channels on the plasma membrane.⁵³

Similar studies in Xenopus oocytes expressing the cardiac sodium channel showed that NEDD4-2 decreased the peak sodium current with no alteration in voltage dependence of activation or inactivation, consistent with a loss of Na_v1.5 due to ubiquitination.⁵⁴ Regulation of NEDD4-2 and Na_v1.5 currents by SGK was studied in Xenopus oocytes⁵⁵ and in mice with cardiac-specific expression of constitutively active SGK1.⁵⁶ Ventricular myocytes from these mice exhibited increases in peak I_Na density and cell surface localization of Na_v1.5, with no change in the total amount of channel protein.⁵⁶ There was a marked decrease in NEDD4-2 bound to Na_v1.5 in the transgenic hearts, suggesting that active SGK1 increased peak I_Na by blocking ubiquitination of Na_v1.5 and increasing its abundance on the cell surface (Fig. 2). Peak I_Na in myocytes from the transgenic mice also showed a −10 mV shift in voltage dependence of activation and a −5 mV shift in steady-state inactivation. These gating changes should increase and shift the window current to more hyperpolarized potentials, lengthening the action potential at more negative potentials and allowing for greater recovery from inactivation of the calcium channel. Persistent I_Na was also increased 3.6-fold in transgenic vs. wild type myocytes.⁵⁶ Not surprisingly, the transgenic SGK1 myocytes exhibited APD₉₀ prolongation, EADs and delayed after-depolarizations, all of which were reversed by treatment with the sodium channel blocker ranolazine at a concentration selective for persistent I_Na. QTc prolongation and lethal ventricular arrhythmias were also improved by treating the mice with ranolazine. Five candidate SGK1 phosphorylation sites were identified in Na_v1.5, one of which (T1590) is located in a region important for channel inactivation, suggesting that phosphorylation of this site by SGK1 might be involved in upregulating persistent I_Na (Fig. 2).⁵⁶ No peptide containing T1590 was detected in a study that identified 11 “basal” phosphorylated sites in Na_v1.5 from mouse ventricular tissue, but the other putative SGK1 sites were found to be phosphorylated.⁵⁷ It is interesting that inhibition of PI3Kα/Akt signaling and constitutive activation of SGK1 seem to have the same effects on persistent I_Na. This apparent discrepancy might be due to differential regulation of the current by different effectors downstream of PI3K (Fig. 2).

Few studies have examined I_Na in diabetes. One noted that peak I_Na was not altered in alloxan-induced diabetic rabbits.⁴² We found that mexiletine treatment reversed APD₉₀ prolongation in ventricular myocytes from diabetic Ins2^Akita and db/db mice, suggesting that persistent I_Na was increased.⁴⁴ Measurement of persistent I_Na confirmed that this was the case. No difference in the amount of Na_v1.5 protein was detected in diabetic vs. wild type hearts.⁴⁴ Intracellular delivery of PI(3,4,5)P₃ or adenoviral expression of constitutively active p110α reduced persistent I_Na in Ins2^Akita or db/db myocytes to wild type levels. Conversely, treatment of wild type myocytes with an Akt inhibitor increased persistent I_Na, but not to the levels seen in diabetic⁴⁴ or p110α-null³³ myocytes. These results suggested that the repolarization defect in diabetes is due in large part to an increase in persistent I_Na, which is caused by suppression of PI3Kα signaling to Akt and perhaps other effectors.⁴⁴

In summary, current evidence suggests that PI3Kα/Akt signaling leads to phosphorylation of Na_v1.5 on a site that regulates its gating properties, thus suppressing persistent I_Na (Fig. 2). Therefore, suppression of PI3Kα signaling due to either diabetes or drug inhibition contributes to QT prolongation by decreasing the phosphorylation of Na_v1.5 and increasing persistent I_Na. On the other hand, the decrease in peak I_Na following inhibition of PI3Kα is likely due to decreased abundance of Na_v1.5 on the cell surface and could, if large enough, slow action potential conduction (Fig. 2).

Voltage-dependent Potassium Channels

I_Kr and I_Ks are major regulators of cardiac repolarization in humans and other large mammals but contribute little to the action potential in adult mice. For that reason, the native currents are not usually studied in genetically modified mice with altered PI3K signaling. Another complication in studying PI3K as a regulator of potassium currents is that some PI3K inhibitors can block I_Kr or other channels in a PI3K-independent manner. For example, direct block of two slowly inactivating potassium currents [I_k,slow1 (K_v1.5) and I_k,slow2 (K_v2.1)] and I_Kur (K_v1.5) by LY294002, a commonly used nonselective PI3K inhibitor, can result in APD prolongation.^{58, 59} We measured reductions in I_Kr and I_Ks in canine ventricular myocytes treated for 2 hours with nilotonib or PI-103.³³ Inhibition of I_Kr by nilotinib was reversed by adding PI(3,4,5)P₃ to the patch pipette solution, suggesting that it was PI3K-dependent. I_Kr required an extended washout period to recover from nilotinib inhibition, indicating that the drug was not acting as a channel blocker. Computer simulations showed that alterations in both I_Kr and I_Ks contributed to nilotinib-induced APD₉₀ prolongation. The 60% decrease in I_Kr alone accounted for less than half of the change in APD₉₀ induced by nilotinib or PI-103, whereas the combined alterations in I_Kr and persistent I_Na accounted for about 80% of APD₉₀ prolongation.³³

Many of the studies on PI3K regulation of the channel that mediates I_Kr (K_v11.1, also known as hERG) were carried out using heterologous expression systems in which signaling, channel-interacting proteins and trafficking may differ from the heart. Constitutively active p110α or Akt enhanced the function of K_v11.1 stably overexpressed in HEK293 cells, while dominant-negative mutants of p110α or Akt or treatment with wortmannin (a nonselective PI3K inhibitor) reduced the current.⁶⁰ Potentiation of K_v11.1 current by SGK3 or constitutively active Akt was also demonstrated in the Xenopus oocyte expression system.⁶¹ Mutation of two putative SGK/Akt phosphorylation sites in K_v11.1 to alanine decreased basal channel function but did not abolish current activation by SGK3, suggesting that SGK3 does not regulate K_v11.1 by direct phosphorylation. A subsequent study in HEK293 cells found that SGK1 and SGK3 increased the level of K_v11.1 on the cell surface by phosphorylating and inhibiting NEDD4-2 and by promoting Rab11-mediated channel recycling (Fig. 3).⁶² The Rab11-dependent process is likely regulated by PIKfyve, a lipid kinase that converts PI(3)P to PI(3,5)P₂. PIKfyve is phosphorylated and activated by isoforms of Akt and SGK and was shown to increase the number of K_v11.1 channels on the surface of Xenopus oocytes.⁶³ Treatment of neonatal rat cardiomyocytes with dexamethasone increased the expression levels of SGK1 and K_v11.1 and increased I_Kr, but whether this is a PI3K-dependent process remains an open question.⁶²

Figure 3. Hypothetical regulation of cardiac delayed rectifier currents by PI3Kα.

Phosphorylation of NEDD4-2 by kinases downstream of PI3Kα increases the currents by preventing the ubiquitination and internalization of the two channels. In the presence of PI3K inhibitors or diabetes, NEDD4-2 is dephosphorylated, it binds to and ubiquitinates K_v7.1 and K_v11.1, and the channels are internalized. PI3Kα may also increase cell surface expression of the channels by a second mechanism that involves phosphorylation of PIKfyve and activation of Rab11-mediated trafficking of channel subunits located in intracellular vesicles (circle) to the cell surface. PI3Kα also upregulates transcription of the potassium channel genes KCNH2 and KCNQ1 by unknown mechanisms.

Thyroid hormone increases K_v11.1 currents in rat pituitary cells by inducing the dephosphorylation of T895 in K_v11.1 by protein phosphatase PP5, which is activated by the small GTPase Rac1 downstream of PI3K.^64–67 A polymorphism in human K_v11.1 (T897, as opposed to the most common K897) was found to disrupt the kinase recognition site surrounding T895 and create a new Akt phosphorylation site at T897.⁶⁵ The K_v11.1 T897 variant is associated with a shorter QT interval at baseline.⁶⁸ However, because thyroid hormone inhibited the K_v11.1 T897 current in a PI3K/Akt-dependent manner, Gentile and coworkers⁶⁵ predicted that increased PI3K/Akt signaling in people with K_v11.1 T897 would cause QT prolongation. Indeed, a recent study found that 9 of 13 patients who presented with QT prolongation and torsades de pointes during the subacute phase of myocardial infarction (during which Akt is activated) carried the K_v11.1 T897 polymorphism.⁶⁹ A limitation of this study is that the number of patients who developed torsades de pointes was very small and the enrichment for K_v11.1 T897 could be coincidental.

The slow delayed rectifier current I_Ks is conducted by the K_v7.1 channel (encoded by KCNQ1) and its accessory subunit KCNE1 (also called minK). All three SGK isoforms and Akt were shown to increase the K_v7.1/KCNE1 current expressed in Xenopus oocytes.^{70, 71} Later studies showed that, similar to Na_v1.5 and K_v11.1, PI3K/SGK1 signaling promoted cell surface localization of K_v7.1/KCNE1 by inhibiting NEDD4-2–mediated internalization (Fig. 3).⁷² NEDD4-2 was also shown to regulate native I_Ks in guinea pig ventricular myocytes.⁷³ An SGK1/PIKfyve/PI(3,5)P₂/Rab11-mediated pathway was also demonstrated to enhance insertion of the channel into the plasma membrane (Fig. 3).⁷⁴ Some mutant K_v7.1 channels that cause long QT syndrome were found to respond in the opposite manner to SGK1, with a decrease in current due to altered trafficking.⁷⁵ Conversely, a study of monozygotic and dizygotic twins found that polymorphisms in the SGK1 gene that are associated with increased blood pressure (presumably due to an increase in SGK1 activity) were associated with a shortened QT interval.⁷⁰

Treatments that activate PI3K signaling have also been reported to suppress some potassium currents in cardiac myocytes. Acute treatment of rat ventricular myocytes with IGF-1, induction of volume-overload cardiac hypertrophy (which is associated with increased IGF-1 signaling and Akt activation) or adenoviral expression of Akt or constitutively active PI3K in rat neonatal cardiomyocytes were all reported to decrease the delayed rectifier current I_K and the inward rectifier I_K1.^{76, 77} On the other hand, repolarizing potassium currents in ventricular myocytes were upregulated in two models of physiological hypertrophy, one produced by cardiac-specific expression of constitutively active p110α and the other by swim training.⁷⁸ Protein and/or mRNA expression of numerous potassium channel subunits was also increased (Fig. 3). Action potential waveform and QT interval were normal in hearts expressing constitutively active p110α because ion channel expression was increased in proportion to cell size to maintain normal current densities. Inhibition of Akt did not reverse the effects of enhanced cardiac PI3Kα signaling on the potassium channels.⁴⁶ The downstream effector of PI3Kα that mediates this response has not been identified.

Prior to our discovery that an increase in persistent I_Na contributes to long QT syndrome in diabetic mice,⁴⁴ other investigators described alterations in cardiac potassium currents in animal models of diabetes. Reductions in the transient outward current I_to were seen in streptozotocin-induced diabetic rats and db/db mice,⁴⁰ I_Kr, I_Ks and I_to were decreased in alloxan-induced diabetic rabbits,^{42, 79} and I_Ks and I_to were attenuated in alloxan-induced diabetic dogs.⁴¹ Lower abundance of the affected ion channel protein was observed in some cases. Computer simulations suggested that the decrease in I_Kr was the major driver of QT prolongation in the diabetic rabbit.⁴² Chronic treatment of alloxan-treated rabbits with insulin completely restored I_Kr function and increased the expression of K_v11.1 to above control levels, even though the animals were still hyperglycemic.⁷⁹ Chronic insulin treatment also prevented QTc prolongation, spontaneous ventricular tachycardias, and APD prolongation in ventricular myocytes.

Taken together, the current evidence suggests that PI3Kα signaling upregulates the cell surface expression of K_v11.1 and K_v7.1. Therefore, decreased PI3Kα signaling due to drug inhibition or diabetes leads to decreased I_Kr and I_Ks, contributing to QT prolongation (Fig. 3).

L-type Calcium Channel

Early work in neurons and smooth muscle myocytes established PI3Ks as mediators of I_Ca,L potentiation by hormones acting through tyrosine kinase or G protein-coupled receptors. In rat cerebellar granule neurons, IGF-1 signaling through a class 1A PI3K and Akt increased I_Ca,L and shifted the voltage dependence of activation, with a fourfold larger current at more hyperpolarized potentials.^{80, 81} In rat portal vein myocytes, angiotensin II acting through the G protein-coupled receptor AT1_A upregulated I_Ca,L through Gβγ activation of PI3Kγ and production of PI(3,4,5)P₃.^82–84

PI3K/Akt signaling also positively regulates I_Ca,L in ventricular myocytes, and PI3Kα plays a central role in this process. Sun and coworkers⁴ determined that IGF-1 signals through PI3Kα and Akt to increase I_Ca,L in mouse myocytes. Transgenic expression of constitutively active p110α also upregulated I_Ca,L and increased expression of the calcium channel pore subunit (Ca_v1.2) and accessory subunits (Ca_vβ2 and Ca_vα2δ1).^{78, 85} Transcriptional upregulation of Ca_v1.2 and Ca_vα2δ1 was not blocked by treatment with an Akt inhibitor (Fig. 4).⁴⁶ I_Ca,L density was also significantly larger and inactivation kinetics were faster in myocytes expressing constitutively active Akt.⁸⁶ Transgenic expression of nuclear-targeted Akt did not alter I_Ca,L,⁸⁷ indicating that Akt signaling at the sarcolemma is important for calcium channel regulation. PTEN-null cardiomyocytes, in which Akt is highly active due to the accumulation of PI(3,4,5)P₃, exhibited an increase in I_Ca,L and a negative shift in the voltage dependence of activation, but no change in Ca_v1.2 protein levels.^{4, 88} Treatment of PTEN-null myocytes with a PI3K inhibitor or expression of dominant-negative p110α reversed the increase in I_Ca,L density and the shift in voltage dependence of activation, whereas an Akt inhibitor reversed the increase in I_Ca,L density only.⁴ These results established PI3Kα as the mediator of changes in I_Ca,L seen in PTEN-null myocytes. Perhaps PTEN constrains PI(3,4,5)P₃ production by PI3Kα in a microdomain that also contains the calcium channel.

Figure 4. Hypothetical regulation of the cardiac L-type calcium current by PI3Ks.

PI3Kα/Akt increases the current by upregulating the cell surface localization and increasing the stability of Ca_v1.2 as a result of Ca_vβ2 phosphorylation. The current is decreased in the presence of PI3K inhibitors, Gα_q and in diabetes due to channel internalization. PI3Kα activation of effectors including atypical PKC (aPKC) also affects gating, perhaps through phosphorylation of Ca_v1.2. PI3Kα also upregulates transcription of the gene that encodes Ca_v1.2 (CACNA1C) by an unknown mechanism. PI3Kγ activation of a phosphodiesterase (PDE) decreases cAMP to affect the current.

PI3Kα-mediated activation of I_Ca,L by receptor tyrosine kinases may be counterbalanced by PI3Kα-mediated inhibition of I_Ca,L by receptors that couple to Gα_q. Active Gα_q binds to PI3Kα and inhibits its activity.⁹ Expression of a Gα_q mutant that inhibits insulin/PI3Kα/Akt signaling but that does not activate phospholipase Cβ caused a marked suppression of I_Ca,L in mouse ventricular myocytes that was reversed by PI(3,4,5)P₃.¹⁰

In contrast to the finding that IGF-1 increases I_Ca,L in mouse ventricular myocytes,⁴ we found that insulin treatment or intracellular PI(3,4,5)P₃ infusion of wild type myocytes did not increase the current.^{10, 89} There was also a disparity in the requirement for basal PI3K/Akt signaling to maintain I_Ca,L in different myocyte preparations. In one case, reducing basal PI3K/Akt activity through application of nonselective inhibitors of PI3K or Akt or by expression of dominant-negative p110α did not cause a significant decrease in I_Ca,L density.⁴ By contrast, we found that deletion of p110α caused a 23% reduction in I_Ca,L density,³ and intracellular delivery of PTEN to dephosphorylate PI(3,4,5)P₃ or the PH domain of Grp1 to sequester PI(3,4,5)P₃ caused a ~30% decrease in I_Ca,L density in wild type mouse cardiomyocytes.¹⁰ In addition, treatment of canine ventricular myocytes with PI-103, nilotinib or an Akt inhibitor caused a reduction in I_Ca,L and, in the case of PI-103, a shift in the steady-state inactivation curve to the right.^{3, 33} PI(3,4,5)P₃ reversed current inhibition caused by PI-103, and intracellular delivery of PI(3,4,5)P₃, PI3Kα or activated Akt1 caused a rapid increase in I_Ca,L density in p110α-null cardiomyocytes.³ We found that the reduction in current in p110α-null myocytes was due to a marked decrease in the fraction of Ca_v1.2 on the cell surface,³ consistent with an earlier study that used transfected COS-7 cells to demonstrate that PI3K/Akt-dependent phosphorylation of a Ca_vβ₂ accessory subunit promotes trafficking of calcium channels to the plasma membrane (Fig. 4).⁹⁰ It is possible that variations in basal PI3K/Akt signaling among myocyte preparations affect Ca_v1.2 surface localization and therefore the response of I_Ca,L to PI3K inhibition or activation.

Akt-dependent phosphorylation of Ca_vβ₂ also regulates Ca_v1.2 protein stability, as revealed in experiments using PDK1-null myocytes. Deletion of PDK1 in the heart of adult mice caused a progressive decrease in Akt phosphorylation, loss of Ca_v1.2 protein and death by heart failure within 5–10 days.⁹¹ Expression of constitutively active Akt increased the amount of Ca_v1.2 in PDK1 knockout myocytes, and experiments in transfected COS-7 cells showed that Ca_vβ₂ caused Ca_v1.2 to become susceptible to degradation under conditions of low Akt signaling (Fig. 4). Several PEST (rich in proline, glutamic acid, serine and threonine) sequences that signal rapid protein degradation were identified in Ca_v1.2, as was an Akt phosphorylation site in Ca_vβ₂.⁹¹ It was proposed that Akt-mediated phosphorylation of Ca_vβ₂ protects Ca_v1.2 from PEST-dependent degradation, thus increasing I_Ca,L.

Unlike PI3Kα, PI3Kβ is not required to maintain basal I_Ca,L. Deletion of p110β in cardiac myocytes of adult mice or incubation of canine ventricular myocytes with a PI3Kβ-selective inhibitor (TGX-221) did not change current density.³ On the other hand, intracellular delivery of PI(4,5)P₂ plus PI3Kβ increased the current in mouse ventricular myocytes whose I_Ca,L was suppressed due to expression of a Gα_q protein.¹⁰ It is possible that injection of PI3Kβ into myocytes allows it to enter a subcellular compartment in which it is not normally present to regulate I_Ca,L.

PI3Kγ suppresses cardiac cAMP levels by activating phosphodiesterases (PDEs) in a manner that is independent of its lipid kinase activity.¹¹ Two p110γ-null mouse lines have been studied, both of which exhibit elevated cAMP levels in the heart.^{11, 88} In ventricular myocytes from one p110γ-null mouse line, basal I_Ca,L density and the response to isoproterenol were abnormally high due to increased cAMP.⁹² These effects were attributed to a decrease in PDE3 activity (Fig. 4).⁹³ In the second p110γ-null mouse line, basal I_Ca,L density was normal but the rate of calcium-induced current inactivation was increased due to increased sarcoplasmic reticulum (SR) calcium release and load.^{4, 34} Additional studies suggested that PI3Kγ, PDE3 and PDE4 regulate basal cAMP in microdomains in the vicinity of the SR that do not contain the L-type calcium channel, and that PI3Kγ is required for PDE4 activity.⁹⁴ Interestingly, I_Ca,L density in pacemaker cells from the sinoatrial node (SAN) of these p110γ knockout mice was increased, and the voltage dependence of activation was shifted negative as compared to controls.⁹⁵ These changes were reversed by treatment with a cAMP antagonist, suggesting that PI3Kγ regulates I_Ca,L in the SAN by suppressing cAMP (Fig. 4).

Some studies reported that I_Ca,L density is diminished in ventricular myocytes of rats⁹⁶ or rabbits⁴² with toxin-induced type 1 diabetes. In rabbits, the inactivation kinetics of I_Ca,L were slowed and the protein expression of Ca_v1.2 was reduced.⁴² We found that I_Ca,L density was reduced in ventricular myocytes of diabetic Ins2^Akita mice, with shifts in the voltage dependence of activation and inactivation to more positive potentials.⁸⁹ Intracellular delivery of PI(3,4,5)P₃ increased I_Ca,L density to control levels and normalized the defect in inactivation. Similar to PI3Kα-null myocytes,³ the amount of Ca_v1.2 on the surface of Ins2^Akita myocytes was reduced, but the total amount of protein was the same as in nondiabetic cells.⁸⁹ Incubation of Ins2^Akita myocytes with taxol to block microtubule-dependent trafficking did not affect basal I_Ca,L density, but it almost completely inhibited current activation provoked by PI(3,4,5)P₃ infusion or perfusion with insulin.⁸⁹ These results suggest that a major factor underlying the suppression of I_Ca,L in Ins2^Akita hearts is reduced trafficking of the channel to the cell surface due to reduced insulin/PI3Kα/Akt signaling (Fig. 4).

Defects in I_Ca,L have also been seen in ventricular myocytes from type 2 diabetic db/db mice. I_Ca,L density was decreased and steady-state activation was shifted toward more depolarized potentials as compared to nondiabetic controls.⁹⁷ The activity of single channels was unchanged, but expression of the Ca_v1.2 protein was reduced in db/db hearts.⁹⁷ Infusion of db/db myocytes with PI(3,4,5)P₃, Akt1 or Akt2 increased current density almost to the wild type level.⁹⁸ Infusion of atypical PKC-ι also increased current density, but to a smaller extent than Akt.⁹⁸ The positive shift in steady-state activation in db/db myocytes was completely reversed by infusion of PI(3,4,5)P₃ or PKC-ι, whereas Akt1 or Akt2 was without effect (Fig. 4).⁹⁸

In summary, decreased insulin/PI3Kα signaling to Akt and atypical PKCs in myocytes from diabetic mice causes several alterations in I_Ca,L that can be partially or completely reversed by supplementing the cells with the affected signaling molecules (Fig. 4). The current information does not permit a clear definition of the role I_Ca,L plays in QT prolongation following down-regulation of PI3Kα signaling. In contrast to PI3Kα, PI3Kγ regulates I_Ca,L by modulating cAMP levels (Fig. 4).

Conclusions

Accumulating evidence indicates that PI3K signaling is a key regulator of multiple cardiac ion channels and as a result defines the duration of the cardiac action potential. PI3K signaling affects at least four ion channels in cardiac myocytes—Na_v1.5, K_v11.1, K_v7.1 and Ca_V1.2—and regulates many aspects of channel function, including protein expression levels, trafficking and gating. The mechanistic details of these regulatory processes remain unclear, and many unanswered questions remain. For example, chronic changes in PI3K signaling seem to affect transcript levels of many ion channel subunits. Is transcriptional downregulation of these genes part of the long-term mechanism by which drugs that inhibit PI3K signaling cause QT prolongation? The role of PI3K in regulating the transcription of ion channel genes in the heart is a relatively unexplored area. Another question relates to how suppression of PI3Kα signaling modulates Na_v1.5 such that an increase in persistent I_Na occurs concurrently with a reduction in peak I_Na. Additional studies will be needed to identify the specific downstream effectors that modulate these ion channel functions. This information may lead to the development of clinically useful drugs that specifically block persistent I_Na without affecting peak I_Na. Going forward, it will be important to identify physiological and pathological conditions that alter cardiac PI3K signaling and to identify the affected PI3K isoforms. Even though all four class 1 PI3K catalytic isoforms produce the lipid second messenger PI(3,4,5)P₃, it is unclear why PI3Kα seems to be the predominant isoform involved in ion channel regulation. It is possible that PI3K signaling modules may be confined to microdomains that target certain extracellular inputs to particular channels. Lastly, it is already clear that the high-throughput assays currently used to test candidate drugs for the ability to block I_Kr are not sufficient to identify all drugs that might cause long QT syndrome. Strategies that incorporate screening of drug candidates for effects on PI3K signaling or persistent I_Na will be helpful for pharmaceutical development.

Non-standard Abbreviations and Acronyms

APD

Action potential duration

APD₉₀

Action potential duration at 90% repolarization

EAD

Early after-depolarization

ECG

Electrocardiogram

ENaC

Epithelial sodium channel

IGF-1

Insulin-like growth factor-1

I_Ca,L

L-type calcium current

I_Kr

Rapid delayed rectifier current

I_Ks

Slow delayed rectifier current

I_Na

Sodium current

mTORC2

Mechanistic target of rapamycin complex 2

NEDD4-2

Neural precursor cell-expressed, developmentally down-regulated 4-2

PDE

Phosphodiesterase

PDK1

3-Phosphoinositide-dependent protein kinase 1

PH

Pleckstrin homology

PIKfyve

Phosphoinositide kinase, FYVE finger-containing

PI3K

Phosphoinositide 3-kinase

PI(3

4,5)P₃, Phosphatidylinositol 3,4,5-trisphosphate

PI(3

4)P₂, Phosphatidylinositol 3,4-bisphosphate

PI(3)P

Phosphatidylinositol 3-phosphate

PI(4

5)P₂, Phosphatidylinositol 4,5-bisphosphate

PTEN

Phosphatase and tensin homolog

QTc

QT interval corrected for heart rate

SGK

Serum- and glucocorticoid-regulated kinase

SH2

Src homology 2