Suppression of Neuronal and Cardiac Transient Outward Currents by Viral Gene Transfer of Dominant-Negative Kv4.2 Constructs*

To probe the molecular identity of transient outward (A-type) potassium currents, we expressed a truncated version of Kv4.2 in heart cells and neurons. The rat Kv4.2-coding sequence was truncated at a position just past the first transmembrane segment and subcloned into an adenoviral shuttle vector downstream of a cytomegalovirus promoter (pE1Kv4.2ST). We hypothesized that this construct would act as a dominant-negative suppressor of currents encoded by the Kv4 family by analogy to Kv1 channels. Cotransfection of wild-type Kv4.2 with a β-galactosidase expression vector in Chinese hamster ovary (CHO)-K1 cells produced robust transient outward currents (Ito) after two days (14.0 pA/pF at 50 mV,n = 5). Cotransfection with pE1Kv4.2ST markedly suppressed the Kv4.2 currents (0.8 pA/pF, n = 6,p < 0.02; cDNA ratio of 2:1 Kv4.2ST:wild type), but in parallel experiments, it did not alter the current density of coexpressed Kv1.4 or Kv1.5 channels. Kv4.2ST also effectively suppressed rat Kv4.3 current when coexpressed in CHO-K1 cells. We then engineered a recombinant adenovirus (AdKv4.2ST) designed to overexpress Kv4.2ST in infected cells. A-type currents in rat cerebellar granule cells were decreased two days after AdKv4.2ST infection as compared with those infected by a β-galactosidase reporter virus (116.0 pA/pFversus 281.4 pA/pF in Ad β-galactosidase cells,n = 8 each group, p < 0.001). Likewise, Ito in adult rat ventricular myocytes was suppressed by AdKv4.2ST but not by Adβ-galactosidase (8.8 pA/pFversus 21.4 pA/pF in β-galactosidase cells,n = 6 each group, p < 0.05). Expression of a GFP-Kv4.2ST fusion construct enabled imaging of subcellular protein localization by confocal microscopy. The protein was distributed throughout the surface membrane and intracellular membrane systems. We conclude that genes from the Kv4 family are the predominant contributors to the A-type currents in cerebellar granule cells and Ito in rat ventricle. Overexpression of dominant-negative constructs may be of general utility in dissecting the contributions of various ion channel genes to excitability.

Eukaryotic cells express a rich tapestry of potassium channel genes. Indeed, the expressed genes often outnumber the recognized ionic currents in any given cell type (1,2). Because different genes can produce channels with very similar phenotypic properties and overlapping pharmacologic sensitivities, the functional role of each gene often has proven difficult to assign. The conundrum is epitomized by the transient outward potassium current. This current (abbreviated I to , 1 but commonly known as the A-type current (I A ) in neurons) has attracted considerable attention given its dynamic regulation in disease states including epilepsy and heart failure (3)(4)(5)(6).
Various members of two separate potassium channel gene families have been implicated in the formation of I to . The first was Kv1.4, a member of the Shaker family which, when expressed in oocytes, exhibits roughly appropriate kinetics (fast activation and fast inactivation) during single depolarizing voltage stimuli (7,8). However, the recovery kinetics of the heterologously expressed Kv1.4 channel are much slower than those of I to in native heart cells, and there is no correlation between the amount of current and the message levels for Kv1.4 (9). More recently, Kv4.2 and Kv4.3 have emerged as stronger candidates (10,11). When expressed alone, these members of the Shal family exhibit plausible kinetics (although again, the match with native cells is imperfect) (12). These genes are richly expressed in ventricular myocytes and in neurons that express I A (4,10,(12)(13)(14); furthermore, antisense oligonucleotides targeting Kv4.2 suppress I to in rat ventricular myocytes (15). Nevertheless, there is a paucity of evidence at the protein level that Kv4 genes underlie I to . The picture is further confused by the fact that many tissues express both Kv4.2 and Kv4.3 (9,11,12), which differ when expressed individually but are thought to be capable of forming heteromultimers that obscure such differences.
Selective gene suppression has become an attractive method for elucidating protein function experimentally. The two most commonly used strategies to accomplish this include antisense methodologies (16) and dominant-negative constructs (17). Both have been used to manipulate functional expression of ion channels (15,18), but the latter has proven to be particularly useful in manipulating potassium channels due to the fact that they form multimers in the cell membrane (19 -27). One dysfunctional K channel subunit (e.g. with a missense mutation in the pore region) can suffice to cripple an otherwise normal tetrameric complex; indeed, such dominant-negative interactions underlie various forms of the inherited long QT syndrome (28,29). The introduction of a dominant-negative ion channel construct into native cells also has been used to map the levels of specific channel families in neurons from Xenopus embryos (30). Unfortunately, such a technique generally is not applicable to mammalian systems without manipulating the germline.
In  (34). The portion of the sequence coding for the first 206 amino acids (Fig. 1A) was amplified by polymerase chain reaction using primers that contained unique restriction sites on the 5Ј and 3Ј ends. The product (Kv4.2ST) was cloned into pSL301 (Invitrogen, San Diego, CA) and sequenced to confirm the absence of polymerase chain reactioninduced mutations. The Kv4.2 ST sequence was then cloned into the adenovirus shuttle vector pE1CMV (Fig. 1B) (31). Constructs were made in both the sense and antisense orientations with respect to the promoter, designated pE1Kv4.2ST and pE1CKv4.2AS respectively. An additional vector, pE1RKv4.2AS, contains the Rous sarcoma virus long terminal repeat as the promoter and was used to generate AdRKv4.2AS. The 4.2ST sequence was also fused in frame to the enhanced green fluorescent protein (EGFP) sequence in the vector pEGFP-C3 (CLONTECH, Palo Alto, CA) to make the construct pEGFP-4.2ST. The full-length Kv4.2 sequence was subcloned into pREP4 (Invitrogen, Carlsbad, CA) and into the pEGFP-C3 backbone to make pRep4.2FL and pE-Kv4.2FL respectively. The full-length rat Kv4.3 sequence was obtained from B. Rudy (New York University Medical Center) and cloned into the expression vector pGFPIRS. This vector is a modified version of pEGFP-C3, which contains the polio virus internal ribosomal entry site, obtained from G. Ketner (Johns Hopkins School of Public Health) (34) and cloned between the EGFP sequence on the 5Ј side and the polycloning site on the 3Ј site. This vector (pGFPIrKv4.3) produces a single transcript, encoding both the EGFP protein and the Kv4.3 protein. The human inward rectifier cDNA (Kir2.1) containing the inactivating mutation of GYG to AAA (27) was provided by G. Tomaselli (Johns Hopkins University) and was fused to the EGFP sequence in pEGFP-C3 (pGFPKir2.1-AAA). The CMV-␤ galactosidase plasmid contains the Escherichia coli lacZ gene under the control of the CMV immediate early promoter. The mitochondrial targeted GFP (mt-GFP) sequence was provided by M. Rizzuto (University of Padova) (35). The mtGFP sequence was similarly subcloned into the shuttle vector pE1CMV to make pE1CMVmtGFP. The humanized GFP was subcloned from the vector pGreenLantern (Life Technologies, Inc.) into pE1CMV to create pE1CMVhGFP. Mammalian expression vectors containing full-length Kv1.4 and Kv1.5 sequences were supplied by E. Levitan (University of Pittsburgh) and L. Philipson (University of Chicago) (36), respectively.
Adenovirus Vector Preparation-The strategy is outlined in Fig. 1B. The various adenovirus shuttle plasmids were cotransfected with pJM17, containing the full human adenovirus serotype 5 genome (37) into HEK293 cells using LipofectAMINE (Life Technologies Inc.). As described previously, homologous recombination between the shuttle vector and pJM17 replaces the region of the adenovirus between map units 1.0 and 9.8 with the expression cassette containing the desired cDNA. Successful recombinations were screened either by direct visualization (AdmtGFP and AdhGFP) or by Southern blot analysis of small scale infections (Ad4.2ST, AdC4.2AS, and AdR4.2AS) followed by RNase protection assays of RNA made from infected cells as described below. AdCMV␤-Gal contains the ␤-galactosidase gene driven by the CMV promoter and was provided by G. Wilkinson (University of Wales College of Medicine, Cardiff, U.K.).
Cerebellar Granule Cell Isolation-Cerebellar granule cells were isolated as described previously (38). Cells were plated on dishes that had been coated with poly-L-lysine (5 g/ml for 10 min). If coverslips were needed they were sterilized by immersion in 95% EtOH followed by flaming in a bunsen burner. Coverslips were then placed in the dishes prior to coating with polylysine. Cells were cultured in Basal Eagle's medium with Earle's salts free of glutamine (Life Technologies Inc.) that had been supplemented with 10% heat-inactivated fetal bovine serum, 100 g/ml gentamicin, 25 mM KCl, and 2 mM L-glutamine. The medium was removed after one day and replaced with fresh medium containing 10 M cytosine arabinoside. Under these culture conditions, expression of the A-type current is stable throughout the course of the experiment (39).
Cell Transfections-CHO-K1 cells (CCL-61 American Type Culture Collection, Rockville, MD) were grown in Ham's F-12 medium (CellGro, Mediatech) supplemented with 10% fetal bovine serum (Life Technologies, Inc.) at 37°C in a 5% CO 2 humidified incubator. Cells were split and plated at 30 -40% confluency on coverslips in 6-well plates 24 h before transfection. Cells were transfected with plasmid DNA (1 g/well total) using LipofectAMINE (Life Technologies, Inc.). After 4 h of exposure, cells were washed once with normal growth medium and then incubated for 1 to 2 days in normal growth medium under normal growth conditions.
Cell Infections-Cultured rat cerebellar granule cells were coinfected with a 3:1 ratio of Ad4.2ST or Ad␤-Gal:AdhGFP on day 3 in culture. The multiplicity of infection that gave the best results was in the range of 100 -500. The AdhGFP allowed visual inspection of the percent infected cells. Infections were carried out in a minimal volume of culture medium supplemented with heat-inactivated 2% fetal bovine serum for 2 h at 37°C. The best results were obtained by removing a coverslip of cells from the dish and placing it into a fresh dish where it was covered with 100 -200 l of infection medium. Surface tension was sufficient to hold the infection medium in place. Following the 2-h infection time the coverslip was placed back in the original culture plate and monitored for the appearance of GFP in granule cells (24 -48 h.). If the cells had been coinfected with ␤-galactosidase, they were further processed by fixing in 4% paraformaldehyde and stained for ␤-galactosidase activity.
Rat ventricular myocytes were similarly infected except that the multiplicities of infection used were in the range of 5 to 20 and it was not necessary to infect a single coverslip at a time as there were no surrounding cells as with neuronal culture. AdmtGFP was used to monitor infection efficiency for myocytes rather than AdhGFP.
Electrophysiology-Experiments were performed at 21-23°C. Whole To reduce experimental variability in cotransfections, cells were routinely studied between 42 and 50 h after the introduction of the transgene. Viral infections of primary cells were also studied at this time for consistency with the cotransfection and for the following practical reasons. In the case of adult cardiac myocytes, variable changes in ion channel expression with increasing time in culture were too great to allow statistical evaluation beyond 2-3 days. The cerebellar granule neurons began to exhibit cytopathic effects at time points past 72 h of infection in both control and test infections.
Molecular Analysis-Viral DNA was isolated as described previously (42). DNA was size-separated by agarose gel electrophoresis and blotted onto Nytran nylon membranes using a Turboblotter apparatus (Schleicher & Schuell). Blots were prehybridized and hybridized in Rapid-hyb buffer (Amersham Life Science, Inc.). Probes were labeled using a random primed labeling kit (Boehringer Mannheim) and [␣-32 P]dCTP (NEN Life Science Products).
Total RNA was isolated from HEK293 cells using Trizol reagent (Life Technologies, Inc.). Ribonuclease protection assays were performed using the RPAII kit (Ambion, Austin, TX). 10 g of total RNA was hybridized to both sense and antisense RNA probes for the Kv4.2ST sequence. Protected fragments were separated on a 5% denaturing polyacrylamide gel.
Confocal Microscopy-CHO-K1 cells (CCL-61, American Type Culture Collection, Rockville, MD) plated on glass coverslips were transfected with pGFP-Kv4.2ST or pGFP-Kv4.2ST and pE-Kv4.2FL using LipofectAMINE. After 36 h, cells were washed with phosphate-buffered saline and placed upside down on a microscope slide over a drop of phosphate-buffered saline. The edges of the coverslip were sealed with rubber cement to prevent drying. Images were taken on a laser scanning confocal microscope (PCM 2000, Nikon Inc., Melville, NY) with a 60ϫ objective lens (numerical aperture 1.2).
Statistical Analysis-Pooled data are shown as the mean Ϯ S.E. Comparisons of means between groups were performed using one-way analysis of variance. P values less then 0.05 were deemed significant.

RESULTS AND DISCUSSION
Cotransfections-Kv4.2ST was designed to multimerize with and cripple full-length products of the Kv4 gene family. By analogy to the Shaker family, we reasoned that the truncated construct including the first transmembrane segment (S1) would contain the determinants of multimerization but would otherwise be nonfunctional. To assess the efficiency of the Kv4.2ST construct in suppressing currents expressed by Kv4 genes, we performed cotransfection experiments. When cotransfected into CHO-K1 cells, Kv4.2ST suppresses functional expression of a full-length Kv4.2 construct (Fig. 2, A, B, and C). These experiments were done at a 2:1 molar ratio of pE1Kv4.2ST:pRepKv4.2FL based on preliminary experiments in which the currents were found to be much more variable at molar ratios of 1:1 and 1:2. Suppression of Kv4 current is specific, since cotransfection of Kv4.2ST with a Kv1.5 expression vector (also at a 2:1 molar ratio) does not affect functional expression of this channel (Fig. 2, D, E, and F). These results encouraged the production of the recombinant adenovirus containing this construct (AdKv4.2ST) for infection of neurons and cardiac myocytes, which are notoriously difficult to transfect by conventional means.
AdKv4.2ST-Southern blot analysis of isolated viral DNA confirmed the presence of the Kv4.2ST sequence in the recombinant adenovirus (data not shown). RNase protection analysis of total RNA isolated from HEK293 cells, which had been infected with either AdKv4.2ST, AdCKv4.2AS, or AdRKv4.2AS, revealed that only sense and antisense (respectively) RNAs were expressed from these constructs in detectable amounts (Fig. 3). Based on preliminary experiments in isolated dog cardiac myocytes, AdKv4.2ST was expanded for further analysis.
Cerebellar granule neurons were chosen to test the idea that the A-type current is encoded by Shal family genes (12,39,43), because these cells are easily maintained in culture with minimal changes in A-type current density over time. Infection of cultures of granule neurons using low multiplicities of infection (range 10 -50) of AdhGFP resulted primarily in infection of surrounding glial cells, as had been previously observed in other neuronal cultures (32). Granule cells are easily distinguished from glial cells by their size and morphology. When multiplicities of infection were increased to 100 -500, infection of neuronal cells could be observed at 24 -48 h. To positively identify infected neurons for electrophysiologic study, AdhGFP was included in all infections and the duration of expression was calculated from the first appearance of GFP-positive neurons. Fig. 4A shows representative A-type currents elicited in an Ad␤-Gal-infected control cell upon depolarization to ϩ40 mV following prepulses to two different potentials (Ϫ90 mV, Ϫ40 mV, 500 ms). Infection with AdKv4.2ST suppresses the A-type current in the cerebellar granule neurons without affecting the maintained component (Fig. 4B). The pooled data from 8 control and 8 test cells (Fig. 4C) confirm that the reduction in current density by AdKv4.2ST is significant (p Ͻ 0.001). The virally mediated suppression of native A-type current in neurons is not complete at 42-50 h unlike the suppression of the expressed Kv4.2 current (Fig. 2). The remaining transient currents decay at the same rate as the control currents, suggesting that the residual current is comprised of channels that are functionally identical to those that were knocked out. Thus, the lack of complete knockout likely reflects competition between the time course of expression of the Kv4.2ST gene product and the turnover rate of the functional channel subunits. Alternatively, it is possible that other non-Kv4 family genes encode a minor fraction of the A-type current.
Rat ventricular myocytes were also chosen due to the established presence of both Kv4.2 and Kv4.3 mRNA in the heart and the relative ease with which these cells could be maintained for several days in primary culture (9, 11). Fig. 5 shows the fully primed and prepulse-inactivated (Ϫ90 mV and Ϫ40 mV holding potentials, respectively) transient outward cur- rents elicited by test pulses to ϩ40 mV in a myocyte infected with Ad␤-galactosidase (panel A) and in another infected with AdKv4.2ST (panel B). The pooled data in Fig. 5C (n ϭ 6 in each group) confirm the significant suppression of native rat cardiac I to by infection with AdKv4.2ST. As was the case with the cerebellar granule cells, the suppression of current was substantial but not complete at 42-50 h. Nevertheless, the results indicate that Kv4 genes constitute the major contributors to I to in heart cells and to A-type currents in cerebellar granule cells.
GFP-Kv4.2ST Constructs-To probe the mechanism of action of Kv4.2ST, fusion constructs were generated with EGFP so that the expressed truncated protein could be localized within living cells with confocal imaging. We first confirmed that the fusion protein GFP-Kv4.2ST acted similarly to Kv4.2ST. Fig. 6, A and B, shows currents recorded in a CHO-K1 cell cotransfected with pRCCMVKv1.4 and pCMV␤-Gal (A) or pGFPKv4.2ST (B). As was the case for Kv4.2ST and Kv1.5, there is no suppression of Kv1.4 by GFP-Kv4.2 (Fig. 6C). Nevertheless, the fusion protein could suppress Shal family currents (either Kv4.2, Fig. 6, D-F, or Kv4.3, Fig. 6, G-I). The summary data in panels F and I show comparable levels of suppression for Kv4.2 and Kv4.3 (p Ͻ 0.005). Kir2.1-AAA is an unrelated inwardly rectifying potassium channel with a pore mutation designed to suppress Kir2.1 channels (27) and was added as a control in these experiments to ensure that suppression of Kv4.2 current was not simply attributable to coexpression with another membrane protein. This control is particularly apt because GFPKir2.1-AAA does suppress the functional expression of inwardly rectifying currents encoded by wild-type Kir2.1 (44). Confocal imaging of CHO-K1 cells transfected with pGFP-Kv4.2ST (Fig. 7A) or cotransfected with pGFP-Kv4.2ST and pE-Kv4.2FL (Fig. 7B) reveals that the fusion construct is richly concentrated in the perinuclear region of the cells. In families of Z-plane images, fluorescence intensity was also detected on the surface of the cells (not shown). These findings suggest that at least some of the suppression of functional current may be due to premature degradation of heteromeric channel complexes, and/or to effects on the processing of the mature protein prior to externalization. However, this apparently abnormal localization of a membrane protein does not seem to be restricted to the truncated version of this protein, as fusion constructs containing full-length channels also have similar localization patterns despite the fact that robust membrane currents can be readily detected in such cells (45). Therefore, it is not possible at this time to state unequivocally whether the suppression of current occurs as a result of pre-FIG. 5. AdKv4.2ST suppresses I to in adult rat ventricular myocytes. Ca 2ϩ -independent transient outward currents at ϩ40 mV recorded in AdCMV␤-Gal-infected myocytes (A) are compared with currents in AdKv4.2ST infected myocytes (B). Peak I to current was measured as the difference between prepulse inactivated (0 mV) and fully primed currents. A marked suppression in the peak current density is indicated in the summary data (C). mature degradation, as the result of the formation of nonfunctional tetramers in the surface membrane, or as a combination of the two effects.
Summary-The use of a dominant-negative Kv4.2 ion channel construct specifically suppresses the transient outward current of rat ventricular myocytes as well as the A-type current of cerebellar granule neurons. These two cell types have previously been shown by other methods to express Kv4 family genes (9 -12, 15). We show here that Kv4.2ST specifically suppresses members of the Shal family. In addition, this is the first experimental demonstration that Shal family members can form heteromultimers with each other. This strategy provides a unique way of determining the molecular identity of macroscopic ionic currents in native cells and may provide a useful tool in understanding the exact role these currents play in cellular physiology. The use of the adenovirus vector also allows the potential use of this strategy in vivo as well as in vitro. Whereas the introduction of dominant-negative constructs may also be achieved by transgenic approaches, developmental adaptation or possible lethal effects may complicate the interpretation of such experiments.