Functional Analysis of Missense Mutations in Kv8.2 Causing Cone Dystrophy with Supernormal Rod Electroretinogram*

Background: Homozygosity mapping linked mutations in KCNV2, encoding Kv8.2, to an inherited retinal disorder. Results: Mutant Kv8.2 subunits render heteromeric Kv2.1/Kv8.2 channels nonfunctional or fail to form heteromers, resulting in homomeric Kv2.1 channels. Conclusion: Although clinically indistinguishable, different missense mutations impair channel function via two distinct molecular mechanisms. Significance: Loss of channel function arising from KCNV2 mutations is confirmed. Mutations in KCNV2 have been proposed as the molecular basis for cone dystrophy with supernormal rod electroretinogram. KCNV2 codes for the modulatory voltage-gated potassium channel α-subunit, Kv8.2, which is incapable of forming functional channels on its own. Functional heteromeric channels are however formed with Kv2.1 in heterologous expression systems, with both α-subunit genes expressed in rod and cone photoreceptors. Of the 30 mutations identified in the KCNV2 gene, we have selected three missense mutations localized in the potassium channel pore and two missense mutations localized in the tetramerization domain for analysis. We characterized the differences between homomeric Kv2.1 and heteromeric Kv2.1/Kv8.2 channels and investigated the influence of the selected mutations on the function of heteromeric channels. We found that two pore mutations (W467G and G478R) led to the formation of nonconducting heteromeric Kv2.1/Kv8.2 channels, whereas the mutations localized in the tetramerization domain prevented heteromer generation and resulted in the formation of homomeric Kv2.1 channels only. Consequently, our study suggests the existence of two distinct molecular mechanisms involved in the disease pathology.


Mutations in KCNV2
have been proposed as the molecular basis for cone dystrophy with supernormal rod electroretinogram. KCNV2 codes for the modulatory voltage-gated potassium channel ␣-subunit, Kv8.2, which is incapable of forming functional channels on its own. Functional heteromeric channels are however formed with Kv2.1 in heterologous expression systems, with both ␣-subunit genes expressed in rod and cone photoreceptors. Of the 30 mutations identified in the KCNV2 gene, we have selected three missense mutations localized in the potassium channel pore and two missense mutations localized in the tetramerization domain for analysis. We characterized the differences between homomeric Kv2.1 and heteromeric Kv2.1/Kv8.2 channels and investigated the influence of the selected mutations on the function of heteromeric channels. We found that two pore mutations (W467G and G478R) led to the formation of nonconducting heteromeric Kv2.1/Kv8.2 channels, whereas the mutations localized in the tetramerization domain prevented heteromer generation and resulted in the formation of homomeric Kv2.1 channels only. Consequently, our study suggests the existence of two distinct molecular mechanisms involved in the disease pathology.
Cone dystrophy with supernormal rod electroretinogram (CDSRE) 4 is a progressive retinal disorder belonging to a genet-ically diverse group of photoreceptor dystrophies. The condition is diagnosed upon presentation of a characteristic electroretinogram (ERG) used to assess the functional integrity of the retina. Scotopic and photopic responses are reduced and delayed, indicating impairment of rod and cone photoreceptor pathways, respectively. However, at higher light intensities, scotopic b-wave responses are supernormal in amplitude (1).
More than 30 unique mutations in KCNV2 have been identified in patients with CDSRE (2, 26 -33). These include missense, nonsense, frameshift, nonstop, and deletion mutations. The purpose of this study is to elucidate the functional outcome of several missense mutations located in different parts of the Kv8.2 ␣-subunit.

EXPERIMENTAL PROCEDURES
Molecular Biology-The human (h) and mouse (m) cDNAs of Kv2.1 and Kv8.2 were cloned into a modified version of pcDNA3 containing the 5ЈUTR of the Xenopus ␤-globin gene (construct name and GenBank TM accession number: hKv2.1 NM_004975; mKv2.1 BC031776; hKv8.2 BC101353; and mKv8.2 BC039042). Mutations were introduced into Kv8.2 with the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). EGFP was fused to the COOH terminus of the indicated constructs by in-frame cloning into pEGFP-N1 (Clontech). Splicing by overlap extension was used to introduce mutations into Kv2.1 (34). All constructs were verified by sequencing.
Fluorescence Imaging-COS7L cells were cultured (in a humidified atmosphere, 5% CO 2 , 95% air) at 37°C in DMEM with high glucose, supplemented with L-glutamine, penicillin/ streptomycin, and 10% FBS. For transfection, cells were cultured on 13-mm glass coverslips and transfected at low density (ϳ30%) using 3 l of GeneJuice (Novagen, Merck) and 1 g of total cDNA. Kv8.2-EGFP was cotransfected with Kv2.1 in a 1:3 ratio. Cells were fixed with 4% paraformaldehyde in PBS 48 h post-transfection. For Kv2.1 antibody staining, cells were permeabilized (0.1% Triton X-100), washed, pre-blocked (5% FBS in PBS, 30 min), and then incubated for 1 h with the K89/34 antibody (1:400 in blocking solution; University of California at Davis/National Institutes of Health Neuromab Facility). After repeated washes with PBS, the cells were incubated again in blocking solution (30 min), followed by application of the secondary antibody (1:600; 1 h), and either TRITC-conjugated goat anti-mouse (Sigma) or Cy3-conjugated goat anti-mouse antibody (The Jackson Laboratory, Newmarket, UK). After further washes in PBS, the coverslips were mounted with VECTASHIELD (Vector Laboratories, Burlingame, CA). Confocal images were acquired using a Leica TCS SPE with a ϫ40 oil immersion objective (1.15 NA). Solid state lasers at 488 nm (EGFP) and 532 nm (TRITC) were used for excitation. Emission ranges of 500 -525 nm for EGFP and 545-650 nm for TRITC were set via the spectrophotometer detection system. Sequential scanning was used to image cotransfected cells, and bleed through was not observed. Imaging for hKv2.1 and hKv2.1 mutants was performed on a Leica SP5 confocal microscope using a ϫ40 oil immersion objective (1.25 NA). Excitation of Cy3 was performed using an argon laser (514 nm), and emission was collected between 548 and 631 nm.
Electrophysiology-HEK293 cells were cultured in DMEM/ F-12 with the same supplements and under the same conditions as for COS7L cells. Cells were transfected at 80 -90% confluency with 2.5 g of DNA using a Lipofectamine 2000 (Invitrogen) to DNA ratio of 2:1. For studies of Kv2.1/Kv8.2, 90 ng of Kv2.1-pcDNA3 was used with an equal amount of Kv8.2-pcDNA3. For studies of Kv2.1, 50 or 90 ng of Kv2.1-pcDNA3 was transfected. To visualize transfected cells, 100 -150 ng of pEGFP-C2 was included together with pcDNA3 to transfect constant amounts of DNA. 18 -22 h post-transfection, cells were plated on glass coverslips and allowed to recover for 3-4 h before recording. Cell recordings were performed at room temperature (20 -25°C). Currents were measured in the whole-cell patch clamp configuration using Pulse version 8.8 acquisition and stimulation software and an EPC10 amplifier (HEKA Elektronik, Lambrecht, Germany). Patch pipettes were pulled from borosilicate glass (KIMAX, Kimble Kontes, Mexico) and had resistances of 1.3-2.2 megohms when filled with intracellular solution (130 mM KCl, 1 mM MgCl 2 , 5 mM EGTA, and 10 mM HEPES, pH 7.2). Cells were continuously superfused with extracellular solution containing 140 mM NaCl, 4 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 10 mM glucose, and 10 mM HEPES, pH 7.35. Series resistance was compensated by 70%. A P/4 protocol was used for current density, activation, and long pulse protocols. Pulse protocols are described in the figure legends and text. Data were analyzed with Pulse/PulseFit (HEKA Elektronik, Lambrecht, Germany) and IGOR PRO (WaveMetric, Lake Oswego). Activation and inactivation curves were fitted with the Boltzmann function P o /P o, max ϭ offset ϩ 1/(1 ϩ exp((V m Ϫ V1 ⁄ 2 )/a h ). A single exponential function, y ϭ offset ϩ A exp (Ϫx/) was used to calculate the activation time constants () and to fit the cumulative frequency data. A double exponential function, y ϭ offset ϩ A 1, exp (Ϫx/ 1 ) ϩ A 2, exp (Ϫx/ 2 ), was used to fit closed-state inactivation and the recovery from inactivation data.
Statistical Analysis-Data were logarithmically transformed when not normally distributed or showed heteroscedasticity. Statistical analysis was performed using the Student's t test or one-way ANOVA followed by the Dunnett's or Bonferroni's multicomparison test. Data are presented as mean Ϯ S.E. p Ͻ 0.05 was considered statistically significant. p Ͻ 0.05 and p Ͻ 0.01 are indicated in figures by * or **, respectively, and the number of experiments (n) is indicated above each bar.

RESULTS
Coexpression of Kv8.2 and Kv2.1 Subunits-The observation that ␣-subunits of the Kv5, Kv6, Kv8, and Kv9 subfamilies only form functional heteromeric channels with members of the Kv2 subfamily in heterologous expression systems led to their denomination as modulatory ␣-subunits. We examined the localization of mKv8.2 and hKv8.2 ␣-subunits, both tagged with EGFP at the COOH terminus, in COS7L cells. An intracellular localization with no apparent membrane expression was observed when either mKv8.2 or hKv8.2 was expressed (Fig. 1, A and E). Not surprisingly, the expression of hKv2.1 (see Fig.  6A) and mKv2.1 (data not shown) resulted instead in a clear signal at the plasma membrane. The coexpression of Kv8.2 with Kv2.1 also led to a clear signal at the plasma membrane, as shown by EGFP fluorescence for Kv8.2 (Fig. 1, B and F), antibody staining for Kv2.1 (Fig. 1, C and G), and the overlay of signals (Fig. 1, D and H). The most straightforward explanation for this observation was that as a consequence of the interaction between Kv2.1 and Kv8.2, the intracellular retention of Kv8.2 was overcome, and this resulted in an efficient trafficking of these ␣-subunits to the plasma membrane.
Electrophysiological Properties of the hKv2.1/hKv8.2 Channel-It has been previously shown that the expression of hKv8.2 in Ltk Ϫ cells (12) and mKv8.2 in Xenopus oocytes (17) does not lead to the formation of functional channels at the plasma membrane. To test whether expression of hKv8.2 could elicit functional currents in HEK293, transfected cells were stimulated with a range of protocols, but the resulting currents did not differ from mock-transfected cells (data not shown). The absence of Kv8.2 current is in agreement with the intracellular retention of Kv8.2 subunits (Fig. 1). Currents mediated by functional heteromeric hKv2.1/hKv8.2 channels were observed on coexpression of hKv8.2 and hKv2.1 (Fig. 2, A-K). Families of current traces elicited by 200-ms voltage steps, from Ϫ80 to ϩ80 mV with 10-mV increments ( Fig. 2A), showed current amplitudes clearly above background but revealed also a 6-fold reduction in current density at a voltage of ϩ30 mV for hKv2.1/ hKv8.2 (0.32 Ϯ 0.06 nA/pF, n ϭ 17) compared with hKv2.1 alone (2.04 Ϯ 0.24 nA/pF, n ϭ 20) (Fig. 2B). The open probability (P o /P o, max ) voltage curve of hKv2.1 and hKv2.1/hKv8.2 channels showed similar values for half-maximal activation but significantly different slopes (a n ) ( Fig. 2C and Table 1). In addition, the time course of activation was slightly faster at more negative potentials for hKv2.1/hKv8.2 channels ( Fig. 2D and Table 1).
Repetitive depolarizations can lead to a successive decline in current amplitude. This cumulative inactivation has been described for the rKv2.1 channel (36) and is a consequence of its preferential inactivation from intermediate closed states and slow recovery from inactivation. Because hKv2.1/hKv8.2 channels show a reduced inactivation from intermediate closed states and an accelerated recovery from inactivation, we hypothesized that these heteromeric channels may be less susceptible to cumulative inactivation. In particular, the faster time course of recovery from inactivation should enable a greater proportion of hKv2.1/hKv8.2 channels to recover during short hyperpolarizing steps between high frequency depolarizations. To test this hypothesis, repetitive depolarizations to ϩ60 mV were applied at a frequency of 1 Hz (Fig. 2I), 4 Hz ( Fig.  2J), or 8 Hz (Fig. 2K). As predicted, at all three frequencies tested, hKv2.1/hKv8.2 channels showed less cumulative inactivation than hKv2.1 channels.
Selection of Kv8.2 Mutations-With the exception of the S4 transmembrane region involved in voltage sensing, missense mutations are located throughout the channel protein in different cases of CDSRE (2, 26 -33). In this study, we have selected five different mutations, three in the pore region that is responsible for ion conduction and two in the T1 domain that has been shown to be important in subunit assembly. Fig. 3 shows the location of the disease-causing mutations in hKv8.2 in CDSRE patients examined in this study. Three of these, W450G, G459D, and G461R, are located in the pore region of the hKv8.2 ␣-subunit. The missense mutations G459D and G461R affect the first and second glycine, respec- 1-pcDNA3 or cotransfected with hKv2.1-pcDNA3 and hKv8.2-pcDNA3 in a 1:1 ratio. Currents were elicited by 200-ms voltage steps from Ϫ80 to ϩ80 mV (10-mV increments) followed by a step to Ϫ40 mV. B, bar diagram comparing the current densities obtained for homomeric hKv2.1 and heteromeric hKv2.1/ hKv8.2 at ϩ30 mV. C, voltage dependence of activation. Amplitudes of the tail currents obtained by the pulse protocol described in A were used. After normalization, the relative open probabilities derived from the initial currents were plotted against the voltage of the depolarizing step and fitted with a Boltzmann function (n ϭ 6 -8). D, voltage dependence of the time course of activation. act was obtained from mono-exponential fits to current traces elicited by the 200-ms voltage steps described in A and plotted as a function of the test potential. E, inactivation from the open state. Currents elicited by a 10-s voltage step from Ϫ80 mV to ϩ30 mV. Solid lines represent mean currents, and gray shading indicates the S.E. (n ϭ 7-8). F, voltage dependence of the steady-state inactivation. Channels were inactivated for 20 s (hKv2.1) or 30 s (hKv2.1/hKv8.2) at prepulse potentials ranging from Ϫ100 to ϩ40 mV, followed by a test pulse to ϩ60 mV to activate residual noninactivated channels. The normalized amplitudes were plotted against the prepulse potentials and fitted with a Boltzmann function (n ϭ 7-9). G, inactivation from intermediate closed states. After a 200-ms test pulse (P1) to ϩ60 mV, channels were inactivated for an increasing time at Ϫ40 mV (P2), after which another 200-ms test pulse (P3) to ϩ60 mV was applied. The proportion of current remaining (I P3 /I P1 ) was plotted against the time spent at (P2) (n ϭ 8). H, time dependence of recovery from inactivation. After a 200-ms test pulse to ϩ60 mV (P1), a 20-s (hKv2.1) or 30-s (hKv2.1/hKv8.2) conditioning pulse to Ϫ10 mV was applied to obtain maximal inactivation. Recovery was measured by another 200-ms test pulse to ϩ60 mV (P3) after increasing intervals at the recovery potential (Ϫ80 mV, P2). The proportion of recovered current (I P3 /I P1 ) was plotted against the time spent at the recovery potential (n ϭ 6). I-K, cumulative inactivation. Repeated depolarizing steps between Ϫ80 and ϩ60 mV were applied at a frequency of 1 Hz (I), 4 Hz (J), or 8 Hz (K) with equal time spent at Ϫ80 and ϩ60 mV. The normalized current amplitudes at each depolarization were plotted against time (n ϭ 6 -13). Ϫ14.6 Ϯ 3.1 (n ϭ 8) Ϫ9.5 Ϯ 1.6 (n ϭ 6) a n (mV) 8.8 Ϯ 0.9 (n ϭ 8) 13.9 Ϯ 1.0 (n ϭ 6) a act at Ϫ20 mV (ms) 60.0 Ϯ 6.2 (n ϭ 6) 40.6 Ϯ 3.7 (n ϭ 6) a DECEMBER 21, 2012 • VOLUME 287 • NUMBER 52
If coexpression of Kv2.1 with Kv8.2 ␣-subunits carrying a pore mutation leads to the formation of heteromeric channels, Kv2.1 should be found at the plasma membrane. To quantify the amount of Kv2.1 ␣-subunits at the plasma membrane, we performed surface biotinylation experiments on COS7L cells coexpressing mKv2.1 with either mKv8.2, mKv8.2-W467G, mKv8.2-G476D, or mKv8.2-G478R. An effective biotinylation of mKv2.1 surface channels was achieved by transiently incubating transfected cells with the membrane-impermeable sulfo-N-hydroxysulfosuccinimide-LC biotin. After cell lysis, the biotin-labeled surface proteins were isolated with neutravidin resin, and mKv2.1 was quantified by Western blotting. The absence of ␣-tubulin from the biotinylated surface fraction demonstrated that no accidental permeabilization of the cells had occurred, and only membrane proteins had been labeled (data not shown). Comparable levels of mKv2.1 were found in the surface fraction upon coexpression with mKv8.2 (100%, n ϭ 3), mKv8.2-W467G (91.6 Ϯ 4.7%, n ϭ 3), and mKv8.2-G478R (97.9 Ϯ 0.1%, n ϭ 3) (Fig. 4, O and P). However, the coexpression of mKv2.1 with mKv8.2-G476D led to a 50.1 Ϯ 5.7% (n ϭ 3) reduction in the amount of mKv2.1 in the membrane when compared with the coexpression of mKv2.1 and mKv8.2 (Fig. 4,  O and P). The reported changes in the level of surface expression are supported by a one-way ANOVA (F 3 Mutations at the Intracellular Amino Terminus of Kv8.2-Two-thirds of the mutations identified in patients with CDSRE are located in the intracellular amino-terminal region of hKv8.2. Ten of these mutations result in the generation of a stop codon that would terminate translation before the first transmembrane spanning segment (S1) is reached. However, two of the initially identified missense mutations (2), L126Q and W188C, lie within the T1 domain (37). Because this domain is important for the assembly with the ␣-subunits of the Kv2 family, we examined how these mutations affect the properties of Kv2.1/Kv8.2 heteromeric channels. In particular, we asked whether these mutated ␣-subunits were still able to interact efficiently with Kv2.1 subunits, and if so, how this interaction affected their electrophysiological behavior. The W188C mutation was introduced at the corresponding position (W196C) in the mKv8.2 ␣-subunit, whereas L126Q was introduced into the hKv8.2 ␣-subunit.
The yeast two-hybrid system has been used by us and others to demonstrate interactions between T1 domains of modulatory and Kv2.1 ␣-subunits (11)(12)(13)(14)38). It was therefore employed to directly test the potential influence of the T1 domain mutations on the interaction between Kv2.1 and Kv8.2 ␣-subunits. The NH 2 termini, including the T1 domains of hKv2.1, hKv8.2, hKv8.2-L126Q, and mKv8.2-W196C, were cloned into pLexN to generate fusion proteins with the DNA binding domain LexA and also into pVP16 to produce fusion proteins with the VP16 DNA activation domain. The various constructs were then transformed pairwise into L40 yeast to identify interacting partners. Additionally the semiquantitative ␤-galactosidase assay was performed. As anticipated, the NH 2 terminus of hKv8.2 and mKv8.2 interacted with that of hKv2.1 in both configurations tested (Fig. 5N, 1st and 2nd   for hKv2.1 with hKv8.2-W196C (OE), channels were inactivated for 20 s at prepulse potentials ranging from Ϫ100 mV to ϩ40 mV, followed by a test pulse to ϩ60 mV to activate residual noninactivated channels. The Boltzmann curve obtained for hKv2.1 (Fig. 2F) is shown for comparison (dashed line). M, time dependence of the recovery from inactivation for the coexpressions of hKv2.1 with hKv8.2-L126Q (f) and hKv2.1 with mKv8.2-W196C (OE). The pulse protocol was identical to that described in Fig. 2H for the homomeric hKv2.1 channel. The exponential fit obtained for the recovery of hKv2.1 (from Fig. 2H) is shown for comparison (dashed line). N, yeast two-hybrid assay probing the homo-and heteromeric interactions mediated by the indicated NH 2 termini. Growth observed after pairwise transformation of yeast two-hybrid constructs pLexN-KvX.Y and pVP16-KvX.Y (indicated on the left) on UTL Ϫ medium indicates successful transformation and growth on THULL Ϫ medium indicates interaction of fusion proteins. The last column shows the mean values obtained for the semi-quantitative test of interaction using the ␤-gal reporter gene.
Expression of Kv2.1 ␣-subunits truncated at the NH 2 terminus also gave rise to low current densities in Xenopus oocytes. Furthermore, the voltage dependence of activation of these channels was shifted to more depolarized potentials (39). Similarly, hKv2.1-L60Q and hKv2.1-W122C channels displayed a shift in their voltage dependence of activation (Fig. 6F), suggesting that by preventing T1 domain interaction, the mutations introduce a change that is comparable with the lack of the T1 domain. We therefore tested the interaction capabilities of the mutated NH 2 termini directly using the yeast two-hybrid assay. Although the NH 2 termini of hKv2.1 interact with each other (Fig. 6G), the absence of growth on THULL Ϫ media for all other combinations tested demonstrates the inability of hKv2.1-L60Q and hKv2.1-W122C to interact with hKv2.1 (Fig. 6G) or with themselves (data not shown).

DISCUSSION
Modulatory subunits, also known as electrically silent subunits, form a large group of K ϩ channel proteins that structurally resemble voltage-gated ␣-subunits. Their inability to form functional homomeric channels in heterologous expression systems hampered their functional analysis until suitable partners were identified, allowing an electrophysiological characterization. So far, only coexpression with Kv2 ␣-subunits, predominantly Kv2.1, has enabled functional expression and detailed biophysical insights. Little is known about the physiological function of channels formed with modulatory subunits (25). Recently, however, mutations in KCNV2, the gene that encodes the Kv8.2 ␣-subunit, have been identified in patients with the distinctive retinal disorder CDSRE, implying that the functional changes induced by these mutations led to the disease (2, 26 -33). Here, we have characterized the effects of mutations in the pore region and T1 domain of the Kv8.2 ␣-subunit on the molecular physiology of Kv2.1/Kv8.2 heteromeric channels. Each of the five Kv8.2 mutations results in a loss of function of the Kv8.2 subunit leading to the functional abolition of the Kv2.1/Kv8.2 heteromer. For the T1 domain Kv8.2 mutants, this is accompanied by a related increase in Kv2.1 homomeric channels. Furthermore, our observations suggest physiologically distinct roles for Kv2.1/Kv8.2 heteromeric channels and homomeric Kv2.1 channels.
The expression of hKv2.1 and hKv2.1/hKv8.2 channels in HEK293 cells showed hyperpolarized shifts of the activation threshold and the voltages of half-maximal activation and inactivation in comparison with other studies (12,17). The Kv2.1 channel is dynamically regulated by phosphorylation. Sixteen phosphorylation sites have been identified, which modulate its voltage dependence (40). Differences in the voltage dependence  (Fig. 2C). G, results of the yeast two-hybrid assay. Growth observed after pairwise transformation of yeast two-hybrid constructs pLexN-KvX.Y and pVP16-KvX.Y (indicated on the left) on UTL Ϫ medium indicates successful transformation, and growth on THULL Ϫ medium indicates interaction of the fusion proteins. The last column shows the mean values obtained for the semiquantitative test of interaction using the ␤-gal reporter gene. DECEMBER 21, 2012 • VOLUME 287 • NUMBER 52 of Kv2.1 in COS1 cells compared with HEK293 cells have been attributed to a different degree of phosphorylation of the channels (41). In addition, sphingomyelin interacts with the gating machinery of Kv2.1 thereby influencing its voltage dependence (42,43). Different degrees of phosphorylation and different availabilities of sphingomyelin in distinct cell types might shift the voltage dependence of Kv2.1-containing channels to varying degrees. This may explain the differences in the voltagedependence of the Kv2. 1 other modulatory subunits to fine-tune their response to alterations in the membrane potential and thereby introduce functional diversity to the role of Kv2.1-containing channels.

Analysis of Kv8.2 Mutations Linked to Retinal Disorder
The formation of Kv2.1/Kv8.2 heteromers is accompanied by a reduction in current density compared with Kv2.1 homomers, a consequence of Kv2.1 being retained by Kv8.2 ␣-subunits. Our analysis of the effect of the coexpression of Kv2.1 with three Kv8.2 disease mutations, W450G, G459D, and G478R (2,31), all located in the pore region, showed a lack of any current resembling that produced by heteromeric Kv2.1/Kv8.2 channels. Intracellular retention in a fine reticular network, most likely the endoplasmic reticulum, has been reported for hKv8.2 in Ltk Ϫ cells (12) and was shown in this study for hKv8.2 and mKv8.2 in HEK293 cells. Similarly, intracellular retention was observed for the three Kv8.2 pore mutations analyzed. The molecular mechanism of retention, observed also for other modulatory ␣-subunits, remains uncertain however (25).
In contrast, coexpression of Kv2.1 with mKv8.2-G476D, performed under conditions designed to avoid competition for protein synthesis or trafficking machinery, showed considerably fewer Kv2.1 subunits in the plasma membrane and a substantially reduced current density with the residual current resembling that of homomeric Kv2.1 channels. Kv8.2-G476D might therefore affect the assembly or trafficking of heteromeric channels and, under our heterologous conditions, may allow the escape of a limited number of Kv2.1 homomeric channels to the plasma membrane.
Recently, histidine 105, located in the T1 domain of Kv2.1, has been shown to selectively disrupt the heteromerization with the modulatory subunits Kv6.3 and Kv6.4 (38). Additionally, two negatively charged amino acids present in Kv2.1 and all modulatory ␣-subunits have been shown to be essential for the efficient assembly of homomeric Kv2.1 and heteromeric Kv2.1/Kv6.4 channels (46). Two disease mutations were examined, L126Q and W188C, that are located in the T1 domain of Kv8.2 at sites that are conserved in all modulatory subunits and in Kv2.1. Both mutants showed an intracellular distribution pattern indistinguishable from the one observed for Kv8.2 ␣-subunits, demonstrating that they do not result in general misfolding or fast degradation of the protein. However, the currents observed on coexpression of mutant Kv8.2 and Kv2.1 showed the same voltage dependence, kinetics, and density as homomeric Kv2.1 channels, indicating that these mutations efficiently prevent the interaction with Kv2.1 ␣-subunits. The inability of the mutated T1 domains of Kv8.2 to interact with the T1 domains of Kv2.1 was confirmed by yeast two-hybrid assay.
When L126Q and W188C were introduced at the corresponding positions in Kv2.1, they showed an intracellular retention resembling that of Kv8.2 ␣-subunits, and further experimentation supported the hypothesis of intracellular retention rather than an increase in protein degradation as a cause for reduced current densities. The absence of mutated ␣-subunits in the plasma membrane and the reduced currents can be viewed therefore as indications that the mutations prevent an efficient interaction of the Kv2.1 T1 domains. Kv channels assemble also in the absence of a functional T1 domain, albeit less efficiently (5,39,47,48); this would account for the small amount of current and Kv2.1 ␣-subunit detected in the plasma membrane after transfection with large amounts of mutant Kv2.1. Finally, the yeast two-hybrid assay showed an absence of any interaction between mutated and nonmutated Kv2.1 NH 2 termini. The shift in the voltage dependence of activation of mutated Kv2.1 channels is not surprising, because changes to both the NH 2 and COOH termini influence the gating of Kv2.1 (39).
Our studies of the several mutations located in the pore and T1 domain of Kv8.2 allow us to postulate the existence of two distinct mechanisms involved in the disease pathology. Pore mutations lead simply to the formation of nonconducting heteromeric Kv2.1/Kv8.2 channels, whereas T1 domain mutations lead to the formation of homomeric Kv2.1 channels only. The mutations permitting the formation of homomeric Kv2.1 chan-nels behave similarly to mutations that result in the absence of Kv8.2 ␣-subunits, whether by gene deletion, frameshift mutation, or premature termination (49). Interestingly, no clinical differences have been observed between patients harboring these different classes of mutations (30).
The presence of Kv8.2 mRNA in rod and cone photoreceptors (2,17), together with the distinctive alteration in photoreceptor function in CDSRE, strongly suggest a role for Kv8.2 in retinal signal transduction, although its exact mechanism of action remains uncertain. Cone photoreceptors within the foveal region appear to be relatively more susceptible to damage by Kv8.2 mutations, whereas rod photoreceptors appear less affected (30). The latter observation is in agreement with the normal amplitude of the a-wave of the rod ERG in CDSRE patients. Therefore, it is likely that Kv8.2 is involved in shaping the signal response to light, because Kv8.2 mutations lead to severely impaired visual responses.
Absorption of photons in the outer segments of photoreceptors results in the closure of cyclic nucleotide gated (CNG) channels and membrane hyperpolarization from dark resting levels. Two currents are responsible for shaping this response, an inward current, I h , which is activated by large hyperpolarizations, and a sustained outward current, I Kx , which is responsible for shaping responses to dim light (50). In addition, in the dark, I Kx opposes the inward current through CNG channels and so is responsible for setting the dark resting membrane potential. I Kx currents have been identified in the photoreceptors of several mammalian species (51)(52)(53)(54). In view of similar kinetic and pharmacological properties, it has been suggested that Kv2.1/ Kv8.2 heteromers may contribute to the I Kx current (17). Because of differences in channel kinetics and voltage dependences, Kv2.1/Kv8.2 heteromeric channels, but importantly not Kv2.1 homomeric channels, are able to maintain a sustained outward current in response to current injection, replicating the "dark current" of photoreceptors (17). Furthermore, cessation of this current injection, mimicking the onset of light, leads to a transient hyperpolarization that is required for the acceleration of the response (17). If this interpretation is correct and Kv2.1/Kv8.2 does underlie I Kx , then the loss of functional Kv2.1/Kv8.2 heteromers, as demonstrated here, would result in membrane depolarization and a slowing of the light response. Membrane depolarization would in turn result in an increase in the threshold required to activate I h and shape high intensity light responses. These predictions correlate well with the clinical findings. The characteristic ERG of CDSRE patients includes a severe reduction in the amplitude and a delay in the time course of responses to dim light, whereas responses to high intensity light are normal or even enhanced. A complicating issue is that the T1 domain Kv8.2 mutations result in a related increase in Kv2.1 homomers. However, because of the pronounced inactivation characteristics of the Kv2.1 channels, Kv2.1 homomers are unable to produce a sustained outward current (17) and so could not contribute to I Kx . Therefore, the presence or absence of Kv2.1 subunits in the photoreceptors may be irrelevant because the Kv2.1 channels will remain in an inactivated state. This may provide an explanation for the lack of clinical differences between patients harboring the different classes of Kv8.2 mutations.
More in-depth analysis of native photoreceptor currents is required to confirm whether Kv2.1/Kv8.2 heteromers are the sole molecular correlates of I Kx . Other channels, including Kv10.1 and Kv10.2, have also been proposed as candidates underlying the I Kx current (55). Furthermore, the kinetic behavior of Kv2.1, and most likely also of Kv2.1/Kv8.2, is influenced by phosphorylation and membrane composition, making direct comparisons between native I Kx and currents of recombinant channels rather difficult. Therefore, a specific Kv8.2 blocker or a genetic approach to suppress Kv8.2 is needed to reveal in vivo the currents generated by Kv2.1/Kv8.2 heteromers.
Mutations in genes encoding K ϩ channels cause a number of diverse disease phenotypes, including long QT syndrome, epilepsy (56,57), and snowflake vitreoretinal degeneration (58). Kv8.2 is the first modulatory Kv ␣-subunit to be linked to a human disease. Potassium channels are the obvious candidates for the development of treatments for potassium channelopathies. Moreover, for CDSRE, it would appear that structural damage within the retina is delayed, thereby providing a window of opportunity during which therapeutic intervention may prove successful (30). However, for this to be achieved, a more complete understanding of the role of Kv8.2 in phototransduction is required.