Regulation of Ion Channel Expression

An Ideal Model of Homeostatic Regulation

Before discussing homeostatic regulation, it is useful to first consider a simple model of possible homeostatic mechanisms to help organize the presently fragmented experimental data.

Voltage-gated ion channels are multimeric membrane proteins with a complicated biosynthetic pathway,²⁵ and, as such, the expression of functional channels in the plasma membrane potentially can be regulated at multiple levels. An ideal physiological regulatory system for the control of ion channel expression in heart would look something like that shown in Figure 1. In such a system, there are multiple levels of control and multiple fail-safe systems. Channel expression can be modified at several levels along the biosynthetic pathway. As a consequence, if a partial failure occurs at an early step in the pathway, there can be changes in biosynthesis rates at downstream levels that help to maintain the final phenotype. Information about the final product, in this case the electrophysiological phenotype, would feed back into the biosynthesis pathway via both cognate and compensatory feedback pathways to maintain the phenotype within limits appropriate to the function of a particular cell type and the physiological stresses on that cell.

Regulated Parameters and Biological Feedback

The first issue to address in trying to understand the homeostatic regulation of ion channel expression is to determine exactly what is regulated. In the idealized model, the primary regulated parameter is the electrophysiological phenotype, or the number of each type of functional ion channel expressed in the cell membrane. Myocytes do not seem to have any mechanism that keeps track of the absolute numbers of a particular ion channel in the cell membrane. This implies that any feedback signal is likely to be a complicated reflection of the actions of many different ion channels and auxiliary proteins. Of the electrophysiological properties listed in Table 1, none are thought to be directly sensed, in large part because there is presently no established mechanism by which the myocyte can directly sense the trajectory of its membrane potential. Although voltage-gated ion channels are membrane potential sensors par excellence, to date there is little evidence that the voltage sensor function of ion channels has any feedback function beyond the immediate regulation of ion flows.

A compounding problem is that the syncytial properties of the myocardium will limit the value of the membrane potential as a source of information for the regulation of ion channel expression at the cellular level. An individual myocyte could express no ion channels and still have a membrane voltage profile similar to that of its neighbors, although there may be some blunting of the fastest components of the membrane voltage signal because of cable properties.

This seems to leave changes in intracellular ion concentrations as the only source of feedback regarding the electrophysiological behavior of individual myocytes. However, ion concentrations are an even more indirect measure of ion channel expression than membrane voltage because they are also strongly dependent on the concentrations and function of buffers, pumps, and transporters.

Steady-state calcium concentrations will represent weighted averages with surrounding cells, but dynamic changes in cell calcium levels will be cell specific. The relatively large size of the myocytes combined with the slow rate of intracellular calcium ion diffusion means that dynamic changes in calcium levels will be transmitted poorly from cell to cell. Because of this, transient calcium concentration changes are likely to reflect, albeit indirectly, the electrophysiological behavior of individual cells. If this is the only physiological property that can provide feedback to the regulatory apparatus, it is likely that there is a considerable loss of information.

It is widely accepted, for many different electrically excitable cells, that voltage-gated calcium channels provide an indirect measure of membrane voltage changes and that the subsequent influx of calcium ions can trigger changes in gene expression, including changes in ion channel expression.^16,26,27 Calcium ions influence gene expression through multiple transcription factors; some well-studied examples include members of the CREB, NFAT, and MEF2 families,^28–31 although many more transcription factors may be sensitive to changing calcium levels.³² A broadly held assumption is that calcium channels represent the most important feedback link between electrical activity and the regulation of ion channel gene expression, in large part because there is no other well-established pathway. It has been proposed that the flux of other ions, including Na⁺ and Zn²⁺, can also affect either gene expression or ion channel expression in excitable cells.^33–35

Further removed from the electrical activity of the cell is the contractile activity induced by electrical excitation. Mechanical stress can affect gene expression in myocytes via intrinsic mechanoreceptors or via autocrine/paracrine pathways.^36,37 Whether it has a direct role in the regulation of ion channel expression in vivo is not clearly established.

Analogous to mechanical stress, ion channel expression can be influenced by both intrinsic pathways (Ca²⁺ ions) and extrinsic pathways (neurohumoral). The extrinsic pathways are potentially numerous. A range of signaling molecules has been implicated in regulating long-term changes in ion channel expression (a partial list is shown in Table 2), with delivery to myocytes mediated through either neuronal release or autocrine/paracrine/endocrine systems. Many of these molecules have pleiotropic effects, however, and the changes in ion channel expression seen in vivo may be secondary or tertiary responses. The broader the action of a particular molecule, the less likely it is to form part of a feedback pathway specifically regulating ion channel expression.

Some of these neurohumoral pathways are active in normal adult hearts and could therefore act as part of homeostatic feedback loops. For example, the sympathetic nervous system through the actions of the signaling molecules norepinephrine and neuropeptide Y (NPY) regulates ion channel expression in normal hearts.^38,39 Whether it simply provides a chronic, unregulated input or can be activated in such a way as to provide meaningful feedback in response to changes in the state of the electrophysiological phenotype of the cardiac myocytes that it innervates is not established. What the source of sensory information would be and whether there is sufficient specificity in the connectivity of the sympathetic system to produce targeted delivery of signaling molecules to different regions in the heart is also unknown.

Several of the molecules listed in Table 2 are thought to be released via autocrine/paracrine pathways,^40,41 and these are the molecules most likely to produce specific actions on cardiac ion channel expression. Whether these molecules exert a tonic influence on channel expression or participate in physiological feedback loops in normal adult hearts has not been definitively established.

The changes in ion channel expression that are induced in several pathophysiological conditions, including cardiac hypertrophy,⁴² myocardial infarction,⁴³ thyroid dysfunction,⁴⁴ and diabetes,⁴⁵ reflect in part the activation of neurohumoral pathways. In every case, however, these are complex responses that can also involve the reactivation of developmental pathways⁴⁶ and tissue remodeling pathways.⁴⁷ Although the changes in ion channel expression seen in these pathologies may reflect appropriate responses to changing demands on the electrical function of the heart, at least in some cases these changes seem to be epiphenomena that may actually contribute to the pathology.⁴²

Limitations of Cardiac Ion Channel Regulation

The myocardium could still approach the performance of an idealized system if it could make unexpectedly effective use of the feedback information that is available. Alternatively, we may not understand the full nature of the feedback information available. To date, neither of these possibilities seems to be the case, and, if anything, the system seems more insensitive than might be expected.

The most basic fail-safe system in both the regulatory model (Figure 2A) and in biological systems is the two alleles for each ion channel gene. One potential response to a loss-of-function mutation in one allele would be to increase the expression of the other allele at the level of transcription. Alternatively, the efficiency of posttranscriptional biosynthesis could be increased through mechanisms such as increased channel subunit synthesis, increased transport of assembled channels to the cell surface, or reduced degradation rates. Any of these mechanisms could, in principle, be activated through a cognate feedback pathway (Figure 1) so that the final phenotype would be affected imperceptibly.

In practice, loss-of-function mutations in a single allele often produce no detectable phenotype.⁴⁸ Even a significant fraction, possibly a majority, of homozygous gene knockouts have limited or no phenotypic expression under laboratory conditions, indicating the activation of compensatory pathways or the presence of functional redundancies in the system.^49–51 One potential contributing factor to this genetic robustness, which is seen in both heterozygous and homozygous knockouts, is likely to be homeostatic regulation.^52,53 In the case of heterozygous knockouts, homeostatic feedback from an expressed phenotype could result in either an increase in transcription from the second allele or increased production of functional proteins from a reduced level of mRNA (Figures 2B and 2C). Activation of either mechanism would result in the maintenance of a constant level of channel production and a maintained electrophysiological phenotype.

The phenomenon of haploinsufficiency (Figure 2D) demonstrates that the fail-safe systems that might be expected in an ideal regulatory system are lacking for at least some cardiac ion channels. Loss-of-function mutations in either the KCNQ1^54–56 or KCNH2 (HERG)^57,58 potassium channel genes can produce long-QT syndrome, even when the gene product encoded by the mutant allele is functionally silent and has little or no effect on the subunits produced from the normal allele. Despite the continued presence of one wild-type allele and at least eight potential levels of regulation downstream from transcriptional regulation (Figure 1), there is not enough flexibility within any one of these levels to compensate adequately for the loss.

The likelihood that heterozygosity for a homozygous-null mutation will produce a haploinsufficiency defect depends strongly on the nature of the gene that is mutated.⁴⁸ Null mutations in transcription factor genes seem to have a high probability of producing haploinsufficiency defects,^48,59 unless they contribute a feedback signal to their own regulatory pathway.⁶⁰ This sensitivity is not entirely surprising, because transcription factors act at the top of a hierarchy of genes and a reduction in transcription factor production will produce pleiotropic effects that will be difficult to sense in a way that could provide intelligible feedback to the regulatory apparatus. In contrast, ion channels are final effector proteins whose expression controls, in a relatively simple way, the final phenotype of electrically excitable cells. In this case, there is a direct relationship between the product, the electrophysiological phenotype, and its primary constituent components, the ion channels. It would seem to be a much simpler problem to devise quality controls for the final product. Yet the loss of one allele of one potassium channel gene can result in a significantly increased mortality rate.⁶¹ One might not have expected a priori that the cardiac system would be quite so vulnerable to the loss of a single copy of a single potassium channel gene.

Even in the case of mutations in the KCNQ1 and KCNH2 genes that produce nonfunctional, dominant-negative subunits, expression of wild-type channels is not entirely eliminated,^54,57 and increased transcriptional activity or posttranscriptional processing would reduce the physiological effect of the mutation. In addition, compensatory mechanisms, such as the upregulation of expression of other potassium channels, could be activated to largely eliminate the functional impact of the mutation.

These results suggest that there are constraints on the system over and above those of good engineering design principles. Although genetic robustness can compensate for a certain level of genetic variability within a species and the resultant variability may actually be selectively favored,⁵² there may be limits to the adaptive advantage that this process confers. If mechanisms for the compensation of genetic defects become too effective, the load of defective genes within a population may become too high, with negative long-term consequences for individual reproductive fitness. The need to prune out the more extreme genetic defects may be a competing process, resulting in limited evolutionary pressure to create a perfectly robust system.

Is There Any Feedback System Capable of Sensing Changes in Electrophysiological Phenotype?

The schema shown in Figure 1 assumes that there is a feedback system that is sensitive to electrophysiological phenotype. The observation that a 50% reduction in KCNQ1 or KCNH2 gene dosage can disrupt normal cardiac electrical function calls this assumption into question. There are, however, examples of homeostatic regulation of ion channel expression in the heart that support the idea that feedback mechanisms exist.

In mice that have a dominant-negative knockout of the Kv4.2 and Kv4.3 channels, which underlie I_to,f, expression of the Kv1.4 channel is significantly and selectively upregulated.^62,63 Presumably, the cardiac system senses that there is something wrong with the electrical function of the heart, possibly because of an abnormal prolongation of the ventricular action potential duration in the knockout animals,⁶⁴ and acts to compensate for this defect, in this case via a compensatory feedback pathway. This result suggests that there is a mechanism that can sense and initiate a response to prolongation of the ventricular action potential duration in mouse heart.

Similarly, treatment with the sodium channel blocker mexiletine results in the upregulation of sodium channel expression⁶⁵ attributable to increased sodium channel gene transcription.⁶⁶ Treatment with the calcium channel blocker verapamil produces a similar effect to that seen with sodium channel blockade, suggesting that the response to sodium channel blockade is mediated by changes in calcium ion fluxes.⁶⁶ In this case, homeostatic regulation seems to act in response to a reduction in the number of functional sodium channels, suggesting the presence of a feedback pathway in rat heart.

An indirect argument favoring the possibility that a feedback system also exists in humans is the relatively tight control of ventricular action potential duration in the normal adult population. The hallmark clinical manifestation of LQT syndrome (lengthening of the QT interval) is only detectable because the normal range of variation is maintained within relatively narrow limits.⁶⁷ It is hard to imagine that the trajectory of development could be so reliably predetermined as to arrive at such a consistent final end point in the absence of at least some feedback.

Compensatory Pathways

Compensation for the loss of an allele is not limited to modulation of the output of the other allele via a cognate feedback system (Figure 1). Changes in the expression of genes that encode channels with similar functions could be initiated by compensatory feedback pathways, and this could partially or completely compensate for the loss of one or even both alleles.

One example of compensatory homeostatic regulation is seen in calcium channel Cav1.2 knockout mice. Expression of the Cav1.3 gene is significantly upregulated during embryonic development in response to knockout of the Cav1.2 gene.⁶⁸ Activation of this compensatory pathway seems to extend the embryonic life of the knockout mice, although the Cav1.2 knockout mutation remains embryonic lethal, indicating the limits of this compensation pathway. In a complementary experiment using Cav1.3 knockout mice, there seems to be compensatory upregulation of Cav1.2 channel expression.⁶⁹

Expression of a Kv1 dominant-negative construct in transgenic mice largely eliminates I_K,slow1.⁷⁰ In these mice, there is a specific compensatory response involving increased expression of the I_K,slow2 component of the delayed rectifier, which seems to be mediated by increased Kv2.1 mRNA expression.⁷⁰

In Kv4 dominant-negative knockout mice, Kv1.4 channel expression is selectively upregulated to compensate for the loss of I_to,f.⁶³ The limits of this compensation pathway are shown in mice in which both the Kv4 and Kv1.4 channels are knocked out.⁶³ In this case there are only minor subsequent changes in ion channel expression, suggesting that even though the system is aware of the deficit, as shown by the response to the loss of the Kv4 channels alone, it cannot mount an appropriate response.

This last example highlights some of the limitations of the compensatory pathway. In the case of the Kv4 channel knockouts, channels that might better compensate for the loss of Kv4 channel expression are not significantly upregulated. Similarly, when both the Kv4 and Kv1.4 channels are eliminated, expression of other transient potassium channels, not normally expressed in heart, is not activated. Both of these failures reflect the fact that the developmental process creates a network of transcription factors in the heart that significantly limits the universe of possible homeostatic responses. In other words, the cells are differentiated and no longer pluripotent. This limitation applies whether the response is transcriptional or posttranscriptional. If the compensatory response involves transcriptional mechanisms, it is dependent on the preexisting network of transcription factors expressed by the differentiation program active in the cell, which places significant limits on which genes can be expressed. If the response involves posttranscriptional mechanisms, these mechanisms can only act on those channels whose genes have been transcriptionally activated during development, such as the Kv1.4 channel, which is expressed at low but finite levels in the adult.

Cognate Pathways

There is a more limited set of data on cognate homeostatic regulatory pathways. As noted above, if the cognate pathway was always effectively activated after loss of a single allele, simple null mutations in the KCNQ1 and KCNH2 genes would not produce arrhythmias in long-QT syndrome.

One example where the cognate pathway seems to be activated is in SA nodal cells of heterozygous Cav1.3 knockout mice. In these cells, calcium channel function is unaffected by the loss of one Cav1.3 allele, suggesting that there is upregulation at some level of the output from the remaining functional allele.⁶⁹ A similar finding has been described for the heterozygous knockout of the clcn1 chloride channel in skeletal muscle.⁷¹ However, in both cases the data are incomplete, and it is difficult to be absolutely certain that there has been compensation via a cognate pathway. As a consequence, whether a transcriptional or posttranscriptional response can be activated by the heterozygous loss of an ion channel gene remains an open question. Transcriptional compensation activated by a cognate feedback pathway has been described for genes other than ion channels,^60,72 although these may be special cases.

Heterozygous knockout of the Scn5a or KChIP2 genes produces a 50% reduction in I_Na or I_to expression, respectively,^73,74 suggesting that no cognate homeostatic regulatory pathway can compensate for the loss of these genes in mice. The results from the Scn5a mice are at variance with the results obtained using pharmacological blockade of sodium channels^65,66 described above. There is one marked complication with the Scn5a knockout mice. Tonic sympathetic drive to the heart is apparently much greater in the heterozygotes than in wild-type mice to compensate for the much slower intrinsic heart rate of the heterozygote hearts. As a consequence, the increased sympathetic activity by itself may have modified channel expression. Intriguingly, this seems to be an example where physiological mechanisms, as opposed to regulatory mechanisms, are activated to maintain homeostasis. This example highlights the importance of obtaining temporally restricted knockouts for the study of homeostatic regulation.

Is Transcriptional Regulation a Suitable Mechanism for the Control of Ion Channel Expression?

The classic model systems used for the study of physiologically controlled transcriptional regulation have been monomeric soluble proteins such as metabolic enzymes.¹² Ion channels, which are typically both heteromeric and multimeric membrane proteins, seem, at least at first glance, to be much poorer candidates for transcriptional regulation. Yet there are multiple examples where either developmentally determined patterns of ion channel expression or subsequent changes in expression induced by various cardiac disorders seem to be mediated primarily at the level of transcription.^{6,38,66,68,70,73–84}

Transcriptional regulation can provide both specificity and precision to the control of ion channel expression. Specificity is necessary because multiple, functionally distinct ion channels with nonoverlapping or only partially overlapping functions are expressed in the heart.⁸⁵ Recruitment of varying combinations of multiple different transcription factors can result, in principle, in the independent regulation of each ion channel gene.⁸⁶ Precise regulation of channel expression levels will be important for some channels, such as those active during the plateau phase of the action potential, where small changes in current levels can result either in premature repolarization or significant prolongation of action potential duration.^{22,24,87–89} Although less well studied, transcriptional regulation can produce precisely graded levels of gene expression.⁹⁰

It is important to note that even when transcriptional regulation of channel gene expression is the primary regulated process, it will not be the sole influence on channel expression levels. For example, posttranscriptional steps may control the final level of channel expression by contributing a rate-limiting step to the biosynthesis pathway. A possible example of this is KCNH2 channel expression in rat heart. KCNH2 mRNA and protein expression levels are high in rat ventricular myocytes, whereas the I_Kr current is very small.^91,92 One explanation for these results is that a step in the posttranslational biosynthetic pathway is rate limiting. Alternatively, however, a component of the I_Kr channel complex might be missing or expressed at low levels in ventricular myocytes.^93,94

One major complication in determining the relative contributions of transcriptional versus posttranscriptional regulatory mechanisms is the heteromeric nature of most ion channels. An example of this is I_to expression in the ventricular myocytes of canine and human heart. Although there is a large gradient of I_to expression across the ventricular walls, Kv4.3 mRNA is expressed at constant levels throughout the ventricle.^79,80 For many years this seemed to be an example of posttranscriptional regulation, because of the marked disjuncture between current levels and mRNA expression. It was only with the cloning and elucidation of the function of the KChIP2 ß subunit⁹⁵ that an alternative hypothesis became available, that regulation of ß subunit expression controls expression of the I_to channel.^{74,79,80,82,96} To definitively conclude that posttranscriptional mechanisms are predominant in the control of a particular ion channel requires a complete knowledge of all of the auxiliary subunits that are either necessary for expression or can affect the expression of a particular channel. This is a very high standard to reach at the moment, and candidate genes will not always be intuitively obvious. The difficulty of this problem is well illustrated by the unanticipated results obtained from the study of mutations of the cytoskeletal protein ankyrin B gene, which alters the expression or function of several ion channels and pumps.^97,98 The development of high-throughput proteomic assays that can be used to detect protein-protein interactions for membrane proteins may address this problem in the near future.⁹⁹

Most specific long-term changes in ion channel expression are likely to be mediated at the level of gene transcription, although there will undoubtedly be exceptions to this general principle. For example, an activated signal transduction pathway could, in principle, modify any step in the biosynthetic pathway (Figure 1). To be useful for physiological regulation, the pathway should selectively modify the biosynthesis of a small number of ion channels without producing global changes in channel biosynthesis. This criterion can be achieved, at least in certain circumstances. Signal transduction pathways triggered by the influx of calcium ions through NMDA receptors regulate the trafficking of AMPA receptors to the postsynaptic membrane at some excitatory synapses without significantly affecting NMDA receptor expression at those same synapses.¹⁰⁰ Whether there are similar examples of highly specific regulated trafficking of ion channel complexes in cardiac myocytes has yet to be established. The specificity seen at excitatory synapses is attributable in part to compartmentalization, and in particular to the unique subcellular anatomy of the dendritic spine. Microdomains for calcium signaling pathways do not necessarily require such an elaborate anatomical basis^101–103 and may exist in cardiac myocytes.

Future Directions

Cardiac organogenesis is mediated by the interactions of complex networks of transcription factors, and many of these networks remain active in the adult heart. As a consequence, virtually every aspect of the study of ion channel regulation in the heart will benefit greatly if it can be more closely linked to the rapidly increasing knowledge of transcription factor function during cardiac myocyte development and differentiation.^104–106 This important task has only just begun and will need some time to yield useful results.

The networks of transcription factors that are created during development both limit and enable the capabilities of the homeostatic regulatory processes in the adult heart. For developmental regulation of ion channel expression, the problem is well defined. We need to know how ion channel genes are turned on or off during cardiac development. In contrast, for homeostatic regulation, the issues are not as clearly formulated. There have been very few systematic studies on homeostatic regulation, and as a consequence we only have a broad outline of the capabilities and limitations of this regulatory system. To date, most results come from genetics, both traditional and reverse. Traditional human genetic studies have provided considerable information about the limitations of these regulatory systems.¹⁰⁷ The use of reverse genetics in mouse to modify gene dosage is an important paradigm for future studies on the capabilities of the homeostatic regulatory system.^63,68,69,74 One issue that could benefit from more detailed analysis is channel expression in heterozygous knockouts, which are particularly useful for studying cognate feedback pathways. Overexpression and misexpression of ion channel genes are other ways to test the capabilities of the homeostatic regulatory mechanisms,¹⁰⁸ although overexpression can produce nonspecific effects.¹⁰⁹

One limitation of all the gene dosage experiments conducted to date is that modified gene function has been active during development, which could lead to the disruption of normal developmental processes and the misinterpretation of results. The development of reliable systems to produce changes in ion channel gene dosage that are restricted to adult heart would be an extremely important addition to the field.

The nature of the feedback mechanisms involved in homeostatic regulation remains puzzling. It is difficult to imagine that calcium transients are by themselves the sole source of information that a cardiac myocyte receives about the state of its electrophysiological phenotype, because these transients could be regulated almost solely by increases or decreases in the level of calcium channel expression and other proteins affecting calcium metabolism. Obviously the temporal pattern of the calcium transients is important, including both duration and frequency. But again, changes in calcium handling proteins will have a disproportionately large effect. It seems almost necessary to assume that other pathways are important, but to date only calcium signaling pathways have much experimental validation.

Our understanding of intracellular sensing may not be complete at the moment, and it is possible that important mechanisms have been overlooked. This is a difficult area, but it will be important to establish whether sensory systems exist in addition to calcium signaling. One simple approach will be to establish whether there are examples of homeostatic regulation that cannot be accounted for by calcium signaling pathways. In this context, neurohumoral signaling pathways may prove to be important effectors in homeostatic regulatory pathways.

With regard to intracellular calcium sensing, there is a large and rapidly increasing literature on transcription factors activated by calcium influxes^29–32 but relatively few studies that have attempted to link this knowledge to the regulation of cardiac ion channel expression.¹¹⁰ More efforts in this area are likely to yield important results.

In this review we have largely ignored the problem of coordinate regulation. In many cases there is a yin-yang nature to the regulation of ion channel expression, or a necessity to balance inward currents with outward currents. This is particularly evident in the plateau phase of the action potential, during which very small and closely balanced currents flow. One way to create and maintain this balance would be to coregulate the expression of antagonistic inward and outward currents by coordinate regulation of the relevant channel genes. Alternatively, functional expression of antagonist channels could be regulated by expression of a common auxiliary subunit⁹³ or by common recruitment to the cell membrane by a scaffolding protein. Whatever the mechanism, coordinate regulation may be an important contributor to the stability of the electrophysiological phenotype and should receive more direct experimental analysis than has been attempted to date.

Conclusions

Waddington¹¹¹ coined the term “canalization” to describe developmental robustness, implying that development moves down certain well-trodden pathways or canals. These canalized pathways were imagined to have relatively steep walls, explaining why the developmental process and the establishment of the final adult phenotype are relatively impervious to environmental or genetic variation. The topology that seems to best characterize the robustness of the cardiac electrophysiological phenotype is a broad valley with several shallow local minima. This is perhaps best illustrated by the phenomena of pacing-induced atrial fibrillation.¹¹² Repetitive pacing readily shifts the electrophysiological phenotype of atrial myocytes to a new semi-stable state, which, unfortunately, actually promotes the maintenance of atrial fibrillation.^77,112 Obviously a balance must be established between phenotypic stability and phenotypic plasticity. The sensitivity to genetic, pharmacological, or pathological insults that the cardiac electrophysiological phenotype displays seems surprisingly high, however, and creates a genuine challenge for the design of effective therapies. Understanding those homeostatic regulatory mechanisms that promote phenotypic robustness is an important first step toward understanding how to shift an inappropriate electrophysiological phenotype toward a normal, healthy phenotype.