Modulation of Kv4.2/KChIP3 interaction by the ceroid lipofuscinosis neuronal 3 protein CLN3

Voltage-gated potassium (Kv) channels of the Kv4 subfamily associate with Kv channel – interacting proteins (KChIPs), which leads to enhanced surface expression and shapes the inactivation gating of these channels. KChIP3 has been reported to also interact with the late endosomal/lysosomal membrane glycoprotein CLN3 (ceroid lipofuscinosis neuronal 3), which is modified because of gene mutation in juvenile neuronal ceroid lipofuscinosis (JNCL). The present study was undertaken to find out whether and how CLN3, by its interaction with KChIP3, may indirectly modulate Kv4.2 channel expression and function. To this end, we expressed KChIP3 and CLN3, either indi-vidually or simultaneously, together with Kv4.2 in HEK 293 cells. We performed co-immunoprecipitation experiments and found a lower amount of KChIP3 bound to Kv4.2 in the presence of CLN3. In whole-cell patch-clamp experiments, we examined the effects of CLN3 co-expression on the KChIP3-mediated modulation of Kv4.2 channels. Simultaneous co-expression of CLN3 and KChIP3 with Kv4.2 resulted in a suppression of the typical KChIP3-mediated modulation; i.e. we observed less increase in current density, less slowing of macroscopic current decay, less acceleration of recovery from inactivation, a less positively shifted voltage dependence of steady-state inactivation. The suppression of the KChIP3-medi-ated modulation of Kv4.2 channels was weaker for the JNCL-related missense mutant CLN3 R334C and for a JNCL-related C-terminal deletion mutant (CLN3 D C). Our data support the notion that CLN3 is involved in Kv4.2/KChIP3 somatodendritic and a lost in


Edited by Mike Shipston
Voltage-gated potassium (Kv) channels of the Kv4 subfamily associate with Kv channel-interacting proteins (KChIPs), which leads to enhanced surface expression and shapes the inactivation gating of these channels. KChIP3 has been reported to also interact with the late endosomal/lysosomal membrane glycoprotein CLN3 (ceroid lipofuscinosis neuronal 3), which is modified because of gene mutation in juvenile neuronal ceroid lipofuscinosis (JNCL). The present study was undertaken to find out whether and how CLN3, by its interaction with KChIP3, may indirectly modulate Kv4.2 channel expression and function. To this end, we expressed KChIP3 and CLN3, either individually or simultaneously, together with Kv4.2 in HEK 293 cells. We performed co-immunoprecipitation experiments and found a lower amount of KChIP3 bound to Kv4.2 in the presence of CLN3. In whole-cell patch-clamp experiments, we examined the effects of CLN3 co-expression on the KChIP3mediated modulation of Kv4.2 channels. Simultaneous coexpression of CLN3 and KChIP3 with Kv4.2 resulted in a suppression of the typical KChIP3-mediated modulation; i.e. we observed less increase in current density, less slowing of macroscopic current decay, less acceleration of recovery from inactivation, and a less positively shifted voltage dependence of steady-state inactivation. The suppression of the KChIP3-mediated modulation of Kv4.2 channels was weaker for the JNCLrelated missense mutant CLN3 R334C and for a JNCL-related Cterminal deletion mutant (CLN3DC). Our data support the notion that CLN3 is involved in Kv4.2/KChIP3 somatodendritic A-type channel formation, trafficking, and function, a feature that may be lost in JNCL.
Kv4 channels may form ternary complexes with auxiliary Kv channel-interacting proteins (KChIPs) (12) and dipeptidylaminopeptidase-related proteins (DPPs) (13). In heterologous expression systems, both auxiliary subunits cause an increase in Kv4 channel surface expression (12)(13)(14)(15)(16). Moreover, KChIPs and DPPs modulate Kv4 channel gating in a specific manner: KChIPs cause a slowing of macroscopic current decay and a positive shift in the voltage dependence of steady-state inactivation (12,14,15,17), whereas DPPs cause an acceleration of macroscopic current decay and a negative shift of the voltage dependence of both activation and steady-state inactivation (13,16,18). Both auxiliary subunits cause an acceleration of recovery from inactivation (12)(13)(14)(15)(16)(17)(18). Among the four different known KChIP subtypes (19), KChIP3 seems to be special, because it is known to interact not only with Kv4 channels but also with DNA to act as a transcription repressor (KChIP3 is then usually referred to as DREAM, for downstream regulatory element antagonist modulator) (20). Moreover, KChIP3 may interact with the Ca 21 -signaling protein presenilin (KChIP3 is then usually referred to as calsenilin) (Ref. 21; reviewed in Ref. 22). More recently, it has been reported that KChIP3 (calsenilin, DREAM) may also interact with the late endosomal/lysosomal membrane glycoprotein ceroid lipofuscinosis neuronal 3 (CLN3; Fig. S1) (23). The functional role of CLN3 is still poorly defined, but numerous loss-of-function mutations in the CLN3 gene are associated with juvenile neuronal ceroid lipofuscinosis (JNCL), an autosomal recessively inherited lysosomal storage disorder often referred to as juvenile CLN3 disease (24)(25)(26). JNCL is a childhood-onset neurodegenerative disease with first symptoms starting at the age of 4-8 years. More than 70 mutations have been identified in the CLN3 gene with a genomic 1.02-kb deletion most frequently found (;85%) in JNCL patients ( Fig. S1) (24). Here we asked whether CLN3, by its interaction with KChIP3 (23), may indirectly influence Kv4.2 channel surface expression and/or inactivation gating and whether these effects may differ for WT CLN3 and JNCLrelated CLN3 mutants. The results of our experiments support the intriguing notion that the Kv4.2/KChIP3 somatodendritic A-type channel formation, trafficking, and function involve CLN3. This involvement may be reduced or completely lost in JNCL.
Given the impairment of Kv4.2/KChIP3 binding by CLN3, we tested whether CLN3 may interfere with the typical KChIP3-mediated modulation of Kv4.2 channels. For this purpose we performed whole-cell patch-clamp experiments on transfected HEK 293 cells (see "Experimental procedures"). Kv4.2-mediated currents were recorded on the first (d1) and on the second day (d2) after transfection ( Fig. 2A). There was an increase in the current density mediated by homomeric Kv4.2 channels from 125 pA/pF on d1 (n = 7) to 216 pA/pF on d2 (n = 12), but co-expression of CLN3 had no effect on current density on either day (n = 9 and 12, respectively; Fig. 2B). KChIP3 coexpression, on the other hand, caused a strong increase in current density on both d1 and d2 (n = 15 and 9, respectively). If Kv4.2 was simultaneously co-expressed with KChIP3 and CLN3, the KChIP3-mediated increase in current density was strongly suppressed on d1 (n = 17) and virtually absent on d2 (n = 24; Fig. 2B; see also Table S1).
We asked whether the presence of CLN3 may also suppress the KChIP3-mediated modulation of the kinetics and voltage dependence of Kv4.2 channel inactivation and measured the . Surprisingly, KChIP3-Myc could not be precipitated from extracts of cells co-expressing GFP-CLN3 and KChIP3-Myc in the absence of Kv4.2-GFP. B, HEK 293 cells were co-transfected with Kv4.2-GFP, KChIP3-Myc, and either empty pcDNA3.1 vector or HA-CLN3. The lysis, co-precipitation, and washing steps were performed in nominal Ca 21 -free buffers (2) or in the presence of different Ca 21 concentrations (15 nM and 50 mM; see "Experimental procedures"). Prior to co-IP analyses, aliquots of the input fractions were probed with antibodies against GFP and HA. Equal loading was confirmed by glyceraldehyde-3-phosphate dehydrogenase (GAPDH) immunoblotting. After immunoprecipitation of Kv4.2-GFP, bound KChIP3-Myc was detected by Myc immunoblotting. Co-IP of Kv4.2-GFP and KChIP3-Myc was tested in the absence (lanes 1-3) and presence of HA-CLN3 (lanes 4-6) and with different Ca 21 concentrations in the lysis and washing buffers. relevant parameters on d2. Kv4.2-mediated currents showed a multiphasic decay, which was best described by the sum of three exponential functions (Fig. 3, A and B). The mean time constants obtained for homomeric Kv4.2 channels were 13, 58, and 1480 ms with relative amplitudes of 74, 22, and 4%, respectively (n = 12). Surprisingly, CLN3 co-expression caused the Kv4.2-mediated currents to decay faster (n = 12; see "Discussion"), whereas KChIP3 co-expression caused the well-known slowing of initial current decay (n = 9) (12,14,17). If Kv4.2 was simultaneously co-expressed with both KChIP3 and CLN3, the KChIP3-mediated slowing of initial current decay was attenuated (n = 22; Fig. 3, A and B; see also Table S1). Next we studied the recovery from inactivation. The recovery kinetics were described by a single-exponential function with a mean time constant of 400 ms for homomeric Kv4.2 channels (n = 11). The recovery kinetics were not influenced by CLN3 co-expression (n = 8), but KChIP3 co-expression caused the well-known acceleration of recovery kinetics (n = 6; Fig. 3, C and D) (12,14,17). If Kv4.2 was simultaneously co-expressed with KChIP3 and CLN3, the recovery kinetics were in some cases biphasic, resulting in fast and a slow recovery components. This may be explained by different channel populations in the plasma membrane: one with fast kinetics (modified by KChIP3) and one with slow kinetics (not modified by KChIP3). A weighted recovery time constant, which was calculated for all Kv4.2 1 KChIP3 1 CLN3-expressing cells studied (n = 15), lay closer to the time constant of the slow component ( Fig. 3D; see also Table S2). This indicates that CLN3 attenuated the KChIP3mediated acceleration of recovery kinetics. Finally, we studied the voltage dependence of steady-state inactivation (Fig. 4). Inactivation was half-maximal at 264 mV with a slope factor of 7 mV for homomeric Kv4.2 channels (n = 9). The voltage dependence of steady-state inactivation was not influenced by CLN3 co-expression (n = 5) but strongly shifted in the positive direction and steepened by KChIP3 co-expression (n = 6). If Kv4.2 was simultaneously co-expressed with both KChIP3 and CLN3, the positive shift and the steepening of the voltage dependence were attenuated (n = 13; Fig. 4; see also Table S2).
Taken together, these data support the notion that CLN3 interferes with functional Kv4.2/KChIP3 interaction. As a consequence the pronounced KChIP3-mediated augmentation of Kv4.2 current density was completely abolished in the presence of CLN3 ( Fig. 2B; d2). The typical KChIP3-mediated modulation of Kv4.2 gating parameters, on the other hand, was left intact to some degree in the presence of CLN3 (see Figs. 3 and 4 and "Discussion"). Therefore, we examined the dose dependence of the CLN3 effects in a stable Kv4.2-expressing cell line (62), which was transiently transfected with different amounts of KChIP3 and CLN3 cDNA (see "Experimental procedures" and Fig. S3). In a steady background of Kv4.2 expression (mean current density of 167 pA/pF, n = 6, when transfected only with empty pcDNA3.1 vector) either KChIP3 alone or KChIP3 together with CLN3 were expressed. The cDNA amounts for the transient transfection of the stable cell line were chosen to yield a KChIP3:CLN3 cDNA mass ratio of either 1:10 or 1:20 (see "Experimental procedures"). The results of these experiments (3 n 7) confirmed our findings obtained with transient transfection of Kv4.2 and indicated a dose dependence of the CLN3 effect on the functional Kv4.2/KChIP3 interaction. However, in the stable Kv4.2-expressing cell line, none of the CLN3 effects on the KChIP3-mediated channel modulation observed with a 1:10 ratio proved to be significant, and a 1:20 ratio significance was only reached for current density, for amp1 and amp2 of the macroscopic current decay kinetics, and for the voltage dependence of steady-state inactivation ( Fig. S2 and Tables S1 and S2).
The CLN3 effect on functional Kv4.2/KChIP3 interaction is Ca 21 -dependent Based on our finding that the Kv4.2/KChIP3 complex is stabilized by Ca 21 both in the absence and in the presence of CLN3 (Fig. 1B), we conducted whole-cell patch-clamp experiments with different amounts of free Ca 21 (nominal Ca 21 -free, A, currents were recorded on d1 and d2. Channels were activated by voltage jumps from 2100 to 140 mV, and current traces were leak-subtracted with a 235 mV prepulse-inactivation-subtraction protocol (64). B, current densities were calculated based on the measured peak current amplitude and the corresponding whole-cell capacitance. The data are presented as means 6 S.D., the number of observations (n) is indicated for each group, and individual data points are shown (gray circles). The typical increase in Kv4.2 current densities caused by KChIP3 co-expression is strongly attenuated in the presence of CLN3. Statistical analyses were done with one-way ANOVA and Dunnett's post hoc testing. Asterisks indicate values significantly different in the presence of CLN3 compared with Kv4.2 1 KChIP3: *, p , 0.05; **, p , 0.0001 (see also Table S1).
15 nM Ca 21 , and 50 mM Ca 21 ) in the pipette solution to manipulate in a controlled manner the cytoplasmic Ca 21 concentration (see "Experimental procedures"). The relevant electrophysiological parameters were measured on d2 for Kv4.2/ KChIP3 channel complexes in the absence and in the presence of CLN3, and the data obtained with 15 nM Ca 21 (n 9 and n 24, respectively) in the patch-pipette served as control (Fig.  5). We first tested the Ca 21 dependence of Kv4.2/KChIP3 channels in the absence of CLN3 and found no significant differences for the well-known KChIP3-mediated modulation if Ca 21 was increased to 50 mM (n 14). However, the KChIP3-mediated modulation in the absence of CLN3 was less pronounced in recordings with a nominal Ca 21 -free solution in the patch-pipette (n 20), especially for the recovery kinetics and the voltage dependence of steady-state inactivation (Fig. 5, D and E). Intriguingly, for Kv4.2 1 KChIP3 1 CLN3, we observed a weaker CLN3 effect on the KChIP3-mediated channel modulation in 50 mM Ca 21 (n 14; i.e. larger current density, slower current decay, faster recovery kinetics. and more positive voltage dependence of steady-state inactivation as compared with 15 nM Ca 21 ; Fig. 5, B-E). For the inactivation kinetics (i.e. macroscopic current decay and recovery from inactivation; Fig. 5, C and D), but not for other parameters, the CLN3 effect on the KChIP3-mediated modulation was also weaker under nominal Ca 21 -free conditions (n 9) as compared with 15 nM Ca 21 . The results of these experiments are summarized in Tables S3a and S3b. Taken together, the data suggest a Ca 21 dependence for both the KChIP3-mediated channel modulation per se and the CLN3-mediated attenuation of KChIP3-mediated channel modulation but apparently within different concentration ranges (see "Discussion").

JNCL-related CLN3 mutants exert weaker effects on functional Kv4.2/KChIP3 interaction
We tested two JNCL-related CLN3 mutants (the missense mutant CLN3 R334C and the C-terminal deletion mutant CLN3DC; see Fig. S1 and "Experimental procedures") with our standard pipette solution (15 nM Ca 21 ; Fig. 6). The relevant electrophysiological parameters were measured on d2 for Kv4.2 1 KChIP3 1 CLN3 (WT, n 24), Kv4.2 1 KChIP3 1 CLN3 R334C (n 16), and Kv4.2 1 KChIP3 1 CLN3DC (n 17). We asked, first, whether the electrophysiological parameters obtained with the CLN3 mutants differ from the ones obtained with WT CLN3, and second, whether the functional Kv4.2/KChIP3 interaction is still significantly affected by the CLN3 mutants (i.e. whether the electrophysiological parameters significantly differ from the ones obtained with maximal KChIP3-mediated modulation in the absence of CLN3). There was a trend toward weaker effects on Kv4.2/KChIP3 current densities, which proved to be significant for CLN3DC; however, similar to WT CLN3, both mutants still significantly affected the KChIP3-mediated modulation of current densities (Fig.  6B). The effects on Kv4.2/KChIP3 inactivation kinetics (macroscopic onset and recovery) were significantly weaker than the CLN3 WT effects for both CLN3 R334C and CLN3DC, and the data showed no significant differences from the ones obtained with maximal KChIP3-mediated modulation (Fig. 6, C and D). Intriguingly, the effects of the mutants on the voltage dependence of steady-state inactivation were not significantly different from CLN3 WT effects, and in all cases the data differed significantly from the ones obtained with maximal KChIP3-mediated modulation ( Fig. 6E; see also Tables S1 and S2). Taken together, our experimental results showed that, except for the voltage dependence of steady-state inactivation, the effects on Kv4.2/ KChIP3 functional interaction were less pronounced for the tested CLN3 mutants than for WT CLN3. For the current density data, a trend is seen suggesting that functional surface expression of Kv4.2/KChIP3 channels is still strongly . Current traces were leak-subtracted and normalized to peak. Insets, initial current decay kinetics shown on an expanded time scale. Note that CLN3 accelerated the decay kinetics of Kv4.2-mediated currents in the absence of KChIP3 (upper traces). The cross-over of normalized current traces, typically caused by KChIP co-expression, is still seen in the presence of CLN3, but the KChIP3-mediated slowing of the initial current decay is strongly attenuated by CLN3 (lower traces; the Kv4.2 trace in the absence of KChIP3 and CLN3 is shown as a reference). B, time constants of current decay (t 1 , t 2 , and t 3 ) and their relative amplitudes (amp 1 , amp 2 , and amp 3 in %) obtained by triple-exponential fitting of the macroscopic current decays. C, the kinetics of recovery from inactivation were obtained by plotting relative current amplitudes (I test /I control , no leak subtraction) against the interpulse duration and fitting the data with a single-exponential or, if necessary, with a double-exponential function. D, recovery time constants obtained from single-exponential fitting (circles) and weighted time constant based on double-exponential fitting (diamond). The data in B and D are presented as means 6 S.D., and the number of observations (n) is indicated for each group; the data in C are presented as means 6 S.E. Statistical analyses were done with one-way ANOVA and Dunnett's post hoc testing. Asterisks indicate values significantly different in the presence of CLN3 compared with Kv4.2 alone or Kv4.2 1 KChIP3. *, p , 0.05; **, p , 0.0001 (see also Tables S1 and S2). suppressed by CLN3 R334C , similar to WT CLN3, whereas the suppression is moderate, albeit still significant, for CLN3DC.

Discussion
We tested whether CLN3, via its putative interaction with KChIP3, may influence Kv4.2/KChIP3 interaction and, thus, the expression level and inactivation gating of A-type potassium channels. The combined results of our co-IP and wholecell patch-clamp experiments support this hypothesis and suggest that JNCL-related mutant CLN3 proteins have less influence on Kv4.2/KChIP3 interaction.

KChIP3 is a multifunctional neuronal calcium sensor involved in apoptosis
KChIPs belong to the neuronal calcium sensor (NCS) superfamily of Ca 21 -binding EF-hand proteins, and it was postulated that these auxiliary Kv4 channel b-subunits may regulate Atype currents and, thus, neuronal excitability in response to changes in cytoplasmic Ca 21 (12,22). The data directly supporting this idea exist for cerebellar neurons, in which Kv4.2/ KChIP3 complexes make a major contribution to I SA (27). Ca 21 entering these cells through low voltage-activated Ca 21 channels is thought to bind to KChIP3, thereby promoting its modulatory functions including a positive shift in the voltage dependence of Kv4.2 channel steady-state inactivation. This guarantees a high I SA availability at physiological membrane potentials (27). Our patch-clamp data obtained with Kv4.2/KChIP3 in the absence of CLN3 with a nominal Ca 21 -free and a 15 nM Ca 21 pipette solution, respectively, reflect this form of Ca 21 -dependent modulation of A-type potassium channels (Fig. 5, D and E). The observed Ca 21 effects on the KChIP3-mediated modulation of Kv4.2 channel gating may involve a stabilization of the Kv4.2/ KChIP3 complex, as suggested by our co-IP data obtained in different Ca 21 concentrations (Figs. 2B and 6).
Before the identification of KChIPs as Kv4 channel b-subunits, proteins identical to KChIP3 had been identified in two different contexts as "DREAM" (a Ca 21 -dependent transcription repressor) (20) and "calsenilin" (a Ca 21 -and presenilinbinding protein) (21). KChIP3 (DREAM) binds to the downstream regulatory element of the prodynorphin gene (20), which relates KChIP3 to pain sensing. The other previously identified function relates KChIP3 (calsenilin) to intracellular Ca 21 homeostasis and brain pathophysiology, because mutations in the KChIP3-binding partner presenilin are associated with familial Alzheimer's disease (28). It is thought that, via its interaction with presenilin, KChIP3 controls other presenilinbinding partners (22). Because presenilin is tightly associated with endoplasmic reticulum (ER) Ca 21 release channels, KChIP3 is involved in the control of intracellular Ca 21 release (reviewed in Ref. 22). In fact, overexpression of KChIP3 modifies ER Ca 21 release and may lead to apoptosis (29,30). Presenilin is also part of the g-secretase enzyme complex, which mediates the g-cleavage of amyloid precursor protein to produce Ab peptides (reviewed in Ref. 22). Because of its interaction with presenilin, KChIP3 functions as a Ca 21 sensor for the g-secretase, which mediates enhanced enzyme activity at elevated Ca 21 levels (31). The presenilin/g-secretase enzyme complex is integral to the proapoptotic activity of KChIP3 (32).
The intriguing finding that KChIP3 can also interact with CLN3 (23) may add another aspect to the involvement of KChIP3 in brain pathophysiology, because mutated CLN3 is found in JNCL (24). On the other hand, it may shed more light on the hitherto not well-defined function of CLN3 (25). KChIP3 may represent a Ca 21 sensor for CLN3, similar to what is postulated for the other known KChIP3-binding partners. Chang et al. (23) found that the KChIP3/CLN3 interaction is disturbed in the presence of high Ca 21 (50 mM), possibly reflecting a regulatory mechanism similar to the Ca 21 -dependent dissociation of KChIP3 (DREAM) from DNA (20). Notably, KChIP3 and CLN3 seem to be opponents in the control of cell survival, because in contrast to KChIP3, CLN3 is anti-apoptotic. Overexpression of CLN3 suppresses Ca 21 -induced neuronal cell death, whereas down-regulation of CLN3 has the opposite effect, and additional down-regulation of KChIP3 in the same cells again prevents cell death (23). It is possible that the proapoptotic activity of KChIP3 can be neutralized by an interaction with CLN3. Moreover, KChIP3 has been reported to be up-regulated in CLN3 knockdown cells and in the brains of CLN3 knockout mice but down-regulated by CLN3 overexpression (23). From these data it has been concluded that CLN3 can negatively regulate cellular levels of KChIP3 expression.  Table S2).

Mode of CLN3 action on A-type potassium channels
The KChIP3/CLN3 interaction put forward by Chang et al. (23) represented the starting point for the present study. We hypothesized that, similar to the suppression of the proapoptotic activity of KChIP3, CLN3 may also suppress the KChIP3mediated modulation of Kv4.2 channel surface expression, the KChIP3-mediated modulation of Kv4.2 channel inactivation gating, or both. Surprisingly, we found that CLN3 significantly accelerated the decay kinetics of Kv4.2-mediated currents even if KChIP3 was not co-expressed. Thus, CLN3 may either be able to directly interact with Kv4.2 channels, or it may activate other KChIP3-unrelated regulatory mechanisms, which influence Kv4.2 channel macroscopic inactivation. Although numerous CLN3 interaction partners have been identified (25), it is not known whether other KChIP or Kv4 isoforms interact with CLN3. This was not studied further because neither any of the other gating parameters examined nor current densities were influenced by CLN3 in the absence of exogenous KChIP3. On the other hand, we were able to show that CLN3 co-expression suppressed all aspects of the KChIP3-mediated Kv4.2 channel modulation. Thus, our experimental results fully confirm the formulated working hypothesis; however, they provide no direct information on the mode of CLN3 action. There are different not mutually exclusive possibilities: similar to the findings of Chang et al. (23), CLN3 may negatively regulate KChIP3 expression levels in HEK 293 cells. Although not systematically investigated, our Western blot data do not support this notion but rather suggest higher KChIP3-Myc levels in the presence of GFP-tagged and untagged CLN3 as compared with the KChIP3-Myc 1 GFP control (Fig. 1A). Higher KChIP3-Myc levels were also seen with Kv4.2-GFP co-expression. We think that both CLN3 (or GFP-CLN3) and Kv4.2-GFP coexpression exerted a stabilizing effect on KChIP3-Myc in our experiments. Such a stabilizing effect was indirectly shown by the finding that KChIP expression levels, especially KChIP3, are actually down-regulated in Kv4.2 and Kv4.3 knockout mice (33,34).
Our combined results support the notion that CLN3, instead of decreasing exogenous KChIP3 expression levels, directly influences Kv4.2/KChIP3 complexes or their formation. The absence of a co-IP signal for KChIP3-Myc 1 GFP-CLN3 was unexpected and apparently contradicts the findings of Chang et al. (23). However, these authors used a glutathione S-transferase-KChIP3 fusion protein and untagged CLN3 instead of KChIP3-Myc and GFP-CLN3, respectively, as in our study. Because our data demonstrated undisturbed complex formation between KChIP3-Myc and Kv4.2-GFP (Fig. 1A), we suspect that the GFP tagging of CLN3 interfered with KChIP3-Myc association. On the other hand, our co-IP data clearly show that both untagged and HA-tagged CLN3 interfere with Kv4.2-GFP/KChIP3-Myc complex formation (Fig. 1A and Fig. S2).
CLN3 may compete with Kv4.2 for KChIP3 binding already during co-expression and co-trafficking (35) and withdraw KChIP3 from complex formation with Kv4.2. Alternatively, CLN3 may interact with mature Kv4.2/KChIP3 channel complexes in the plasma membrane to exert its effects. CLN3 is mainly located in the late endosomal/lysosomal compartment but has also been reported to reside, among others, in the ER, trans-Golgi network and plasma membrane (reviewed in Refs. 25 and 35). Given the numerous subcellular locations reported for CLN3 (35,36), effects on immature Kv4.2/KChIP3 channels caused by co-expression and co-trafficking and effects on  Our data suggest that CLN3 may disturb Kv4.2/KChIP3 complex formation, thereby negatively influencing the trafficking of Kv4.2 channels to the cell surface. It should be noted in this context that current density, which is a good correlate of channel surface expression, showed the highest CLN3 sensitiv-ity. In fact, current density, unlike the other electrophysiological parameters examined, exhibited a complete neutralization of the KChIP3-mediated modulation when CLN3 was coexpressed (Fig. 2). Moreover, the moderate effects observed in the stable Kv4.2-expressing cell line (Fig. S3) support the idea that the CLN3 effects depend on co-expression and co-trafficking early in biogenesis.
Co-trafficking of CLN3 with Kv4.2 and KChIP3 to the cell periphery may render the KChIP3/CLN3 interaction Ca 21 -sensitive (35). The results obtained with glutathione S-transferase pulldown assays in 50 mM Ca 21 and with cells treated with the ER Ca 21 pump inhibitor thapsigargin have previously suggested that the KChIP3/CLN3 interaction is weaker in high Ca 21 (23). Thus, the negative influence of CLN3 on Kv4.2/ KChIP3 interaction demonstrated by our own experiments should be relieved in high Ca 21 . Such an indirect Ca 21 dependence of the Kv4.2/KChIP3 interaction was probably masked in our co-IP experiments, which suggested that the Kv4.2-GFP/ KChIP3-Myc interaction per se (i.e. in the absence of CLN3) was favored by Ca 21 (Fig. 1B). Aside from that, our quantified co-IP data suggest that even in the presence of 50 mM Ca 21 , HA-CLN3 co-expression still reduced Kv4.2-GFP/KChIP3-Myc interaction by 50% (Fig. S2). Only by performing wholecell patch-clamp experiments with different Ca 21 concentrations in the pipette solution, we were able to resolve different modes of Ca 21 -dependent channel modulation. Our results suggest that increasing Ca 21 from 0 to 15 nM favors the KChIP3-mediated modulation of Kv4.2 channels, although this concentration lies below the hitherto reported in vitro K d values for high affinity metal binding of KChIPs (10-50 mM) (22). On the other hand, differences between the current parameters measured with 15 nM and 50 mM Ca 21 , respectively, were only seen in the presence of CLN3 (Fig. 5). The latter results may be explained by a weakening of the KChIP3/CLN3 interaction in 50 mM Ca 21 , which favors the KChIP3-mediated modulation of Kv4.2 channels. Because all current parameters were measured within a time range between tens of seconds and a few minutes after establishing the whole-cell configuration, the observed Ca 21 effects can be classified as acute. Acute Ca 21 effects on the surface expression of Kv4.2/KChIP complexes in the whole-cell patch-clamp configuration have been studied previously by Murphy and Hoffman (37), who found increased Kv4.2 current densities with elevated Ca 21 in the pipette solution for certain KChIP isoforms including the KChIP3a splice variant used in the present study. The authors suggested a post-ER/Golgi pool of Kv4/KChIP channel complexes recruited to the plasma membrane in a Ca 21 -dependent manner. CLN3 may critically influence this short-distance surface membrane trafficking and, thus, the turnover of Kv4.2 channels in a KChIP3-dependent manner. It should be noted that CLN3 has been shown previously to be involved in the control of Na 1 / K 1 -ATPase turnover via the cytoskeletal protein fodrin (38). Taken together, CLN3 not only may influence Kv4.2/KChIP3 channel formation and trafficking early in biogenesis but may also play a role in the dynamic Ca 21 -dependent A-type channel trafficking and recycling at the plasma membrane.  Tables S1  and S2).
CLN3 effects on Kv4.2/KChIP3 channels A role for KChIP3 and somatodendritic A-type potassium channels in JNCL?
Knockout mice with targeted disruption of the Cln3 gene (Cln3 2/2 ) and knockin mice, which harbor the most common alteration of the human CLN3 gene (Cln3 Dex7/8 ), have been created to be used as preclinical disease models for JNCL (39,40). Both models exhibit the hallmarks of JNCL, including intracellular accumulation of autofluorescent storage material, astrocytosis, microglial activation, neuronal loss, and neurological deficits (39,40). Given the typical neurological phenotype of JNCL, the previously shown interaction between the CLN3 protein and the Kv4 channel b-subunit KChIP3 (23) combined with the results of the present study leads to the question of whether KChIP3 and somatodendritic A-type potassium channels may play a role in JNCL.
A cellular hallmark of the neuronal ceroid lipofuscinoses (NCL) including JNCL is the accumulation of autofluorescent ceroid lipopigments with subunit c of mitochondrial ATP synthase or sphingolipid activator proteins A and D as major protein components (41). Autophagy, the process in which such intracellular macromolecules are normally digested during organelle turnover, requires the fusion of autophagic vacuoles with late endosomes and lysosomes. This process is disrupted in CLN3 deficiency leading to the accumulation of autophagic vacuoles accompanied by disturbed Ca 21 homeostasis (42,43). The latter may represent an important mechanistic link between CLN3 and KChIP3 in JNCL, because the reported vulnerability of CLN3-deficient cells to Ca 21 -induced cytotoxicity following treatment with thapsigargin or with the Ca 21 ionophore A23187 is thought to be mediated by KChIP3 (23). CLN3 deficiency may lead to increased availability of KChIP3, which by its known interaction with presenilin may in an unphysiological manner influence ER Ca 21 channels (22,30). Apparently, despite its important functions as a neuronal calcium sensor, too-high levels of free KChIP3 may be detrimental to cell function. In accordance with this notion, the down-regulation of KChIP expression levels observed in Kv4.2 knockout mice has been previously interpreted as a feedback mechanism to ensure that free KChIPs do not accumulate (33). Thus, a sofar-unconsidered role of CLN3, which is reduced or lost in JNCL, may be to keep cellular levels of free KChIP3 below a critical proapoptotic level at physiological and moderately increased cytoplasmic Ca 21 concentrations. A strong increase in cytoplasmic Ca 21 may augment the proapoptotic activity of KChIP3, a mechanism expected to be exaggerated by CLN3 loss of function as in JNCL. It should be noted that targeting KChIP3 with small molecule inhibitors has recently been considered a possible strategy for the treatment of Huntington's disease, which is another neurodegenerative disorder (44).
To find out whether JNCL-related mutant CLN3 proteins still affect Kv4.2/KChIP3 interaction, we tested the missense mutant CLN3 R334C and the C-terminal deletion mutant CLN3DC, which is most common in JNCL (45) (Fig. S1). The effects of these mutants on the KChIP3-mediated modulation of Kv4.2 channels were weaker compared with WT CLN3, and concerning current density, the effect was weaker for CLN3DC than for CLN3 R334C (Fig. 6B). The variable impact of the mutants on Kv4.2/KChIP3 interaction may be caused by their different protein stabilities, altered intracellular localization, and/or binding affinities for KChIP3. The missense mutation R334C does not impair lysosomal targeting of CLN3 but compromises its function (46), whereas the truncated CLN3DC protein is thought to be retained in the ER because of misfolding (47). It is likely that a large amount of mutant CLN3DC RNA is degraded (48); however, the observation that RNA interference-mediated down-regulation of residual CLN3DCrelated transcripts had an effect on lysosomal size led to the conclusion that some residual function is still preserved in CLN3DC (49). The proposed residual function of CLN3DC may include a moderate effect on channel trafficking out of the ER. Notably, the CLN3DC mutant protein lacks the proposed binding site for KChIP3 encompassing C-terminal amino acids 314-438 (23) (Fig. S1) and is not expected to interact with KChIP3, unless additional as-yet-unidentified binding sites exist. The different degrees of suppression of the functional Kv4.2/ KChIP3 interaction mediated by CLN3DC and CLN3 R334C , respectively, suggest that this effect is important only in JNCL patients with the CLN3DC deletion mutant but not in JNCL patients carrying the CLN3 R334C mutation.
The results of our electrophysiological measurements obtained in the presence of WT and mutant CLN3 proteins support the notion that a physiological CLN3 effect on Kv4.2/ KChIP3 interaction is reduced in JNCL. This in turn raises the question of whether and how the reduced regulatory effect on KChIP3-mediated channel modulation can be pathogenic. It should be noted in this respect that the CLN3 R334C and CLN3DC-mediated Kv4.2/KChIP3 channel gain of function (compared with WT CLN3) observed in our experiments suggests increased I SA availability caused by CLN3 loss of function in JNCL, a scenario in which somatodendritic excitability is expected to be reduced.
Because JNCL patients exhibit deficits in motor coordination and cerebellar atrophy (50)(51)(52), cerebellar neurons, especially cerebellar granule cells, have moved into the focus of preclinical JNCL research. In both Cln3 2/2 and Cln3 Dex7/8 mice, cerebellar granule cells show a higher vulnerability to a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor-mediated excitotoxicity (53,54). Altered AMPA receptor trafficking and enhanced AMPA receptor function have been initially suggested to underlie the increased AMPA receptor-mediated excitotoxicity in Cln3 2/2 mice (54). However, a more recent detailed examination of the physiology of the mossy fiber-granule cell synapse performed by Studniarczyk et al. (55) found no differences in postsynaptic AMPA receptor expression or function but rather presynaptic alterations in Cln3 2/2 mice. In particular, the authors reported altered short-term plasticity under conditions of reduced extracellular Ca 21 , which may be associated with disturbed Ca 21 handling and sensing (55). Moreover, a reduced density of synaptic vesicles and decreased numbers of membrane adjacent synaptic vesicles in Cln3 2/2 mice were found in that study (55). In accordance with this, Grünewald et al. (56) found severely affected excitatory and inhibitory synaptic transmission, including the loss of GABAergic interneurons in the amygdala, hippocampus, and cerebellum of Cln3 2/2 mice. Cerebellar network activity depends critically on Kv4 channels and their fine-tuning via the NCS protein KChIP3 (27,(57)(58)(59)(60). In particular, long-term potentiation at the mossy fiber-granule cell synapse involves the modulation of postsynaptic Kv4 channels by shifting the voltage dependence of activation and inactivation (60). Moreover, cerebellar granule cell excitability in different subregions of the cerebellum is governed by the expression pattern of specific combinations of Kv4 and Ca 21 channels, thereby allowing inputspecific processing of information via the NCS protein KChIP3 (58). KChIP3-mediated regulation of Kv4 channels also controls the excitability of cerebellar stellate cells, which provide inhibition of Purkinje cells (27). The Ca 21 -dependent regulation of Atype potassium channels determines the spike latency in stellate cells (59) and maintains inhibitory charge transfer to the Purkinje cells especially during extracellular Ca 21 fluctuations (57). All of these aspects of Ca 21 -dependent KChIP3-mediated modulation of Kv4 channels may be strongly impaired in JNCL, in which CLN3 loss of function leads to disturbed Ca 21 homeostasis. It remains to be seen whether and how altered somatodendritic Atype channel expression and function caused by loss of CLN3 function is causally involved in JNCL.

Co-IP and Western blotting analysis
24 h after the start of transfection, the growth medium was aspirated, and the cells were washed with ice-cold PBS, scraped in 1.5 ml of ice-cold PBS, and centrifuged for 5 min at 1,000 3 g at 4°C. The cell pellets were lysed in 100 ml of ice-cold lysis buffer (10 mM Tris-Cl, pH 7.5, 0.5% Nonidet P-40, 150 mM NaCl, 0.5 mM EDTA, protease inhibitors) and placed on ice for 30 min. Where indicated, lysis, dilution and wash buffers lacking EDTA were supplemented with Ca 21 at a final concentration of 15 nM and 50 mM, respectively. The cell lysates were centrifuged at 20,000 3 g for 15 min at 4°C, and the supernatants transferred to a new tube. Then 150 ml of ice-cold dilution buffer (10 mM Tris-Cl, pH 7.5, 150 mM NaCl, 0.5 mM EDTA, protease inhibitors) was added to the lysate, and aliquots (input) were removed for Western blotting analysis. Based on calculations with the program WEBMAXC extended (RRID: SCR_018807), the free Ca 21 concentration in the dilution buffer was adjusted to 0 nM (nominal Ca 21 -free), 15 nM, or 50 mM. The lysates were mixed with 25 ml of GFP-TRAP bead slurry (ChromoTek) and incubated for 2 h at 4°C on a rotating wheel. After centrifugation at 2,500 3 g, the supernatants were removed, and the beads were washed three times with 500 ml of ice-cold dilution buffer. GFP-TRAP beads were resuspended in 75 ml of 23 SDS sample buffer and boiled for 10 min at 95°C. Aliquots of the input and the eluates were separated by SDS-PAGE and analyzed by Myc immunoblotting. Immunoreactive bands were visualized by enhanced chemiluminescence detection using a molecular imager (model ChemiDoc XRS system, Bio-Rad).

Electrophysiological recordings
All recordings were done at room temperature (;22°C) in the whole-cell configuration of the patch-clamp technique using an EPC9 amplifier and PULSE software (Heka Electronics). The cells were bathed in external solution containing 135 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 5 mM HEPES, 10 mM sucrose, pH 7.4 (NaOH). Patch-pipettes were pulled from thin-walled borosilicate glass and filled with standard intracellular solution containing 125 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 11 mM EGTA, 10 mM HEPES, 2 mM K 2 -ATP, 2 mM GSH, pH 7.2 (KOH). The free Ca 21 concentration of this pipette solution was ;15 nM. Alternatively, a nominal Ca 21 -free pipette solution and one with 50 mM Ca 21 were used in some experiments as described previously (63) (EBMAXC extended). Pipette-to-bath resistances ranged between 2.5 and 3 MV, and series resistance compensation was between 80 and 90%. Currents were activated by voltage pulses to 140 mV from different prepulse voltages and with different interpulse intervals. In the absence of significant endogenous currents, leak subtraction was done based on a P/5 protocol. Subtraction of both leak and endogenous currents was done with a prepulse-inactivation-subtraction protocol (64).

Data analysis
Quantification of immunoreactive band intensities was performed using the software QuantityOne 4.5.0 (Bio-Rad). The current traces were analyzed with PulseFit (Heka Electronics), and the obtained data were further processed with Kaleidagraph (Synergy Software). Macroscopic current decay kinetics were described by the sum of three exponential functions (64), the kinetics of recovery from inactivation by a single-exponential function or by a double-exponential function with a weighted time constant. The voltage dependence of steadystate inactivation was described by a Boltzmann function of the form I/I max = 1/(1 1 exp (V 2 V 1/2 inact )/k), where V is the prepulse voltage, V 1/2 inact is the prepulse voltage that causes halfmaximal inactivation, and k is the slope factor of the voltage dependence. Statistical analyses were done with Kaleidagraph and Prism (GraphPad Software). Comparison of band intensities for two groups and current densities for one group on d1 and d2 was done using unpaired Student's t test. Comparison of electrophysiological parameters for more than two groups on d2 was done using one-way analysis of variance (ANOVA) with Dunnett's post hoc testing.

Data availability
All data and statistical analyses are summarized in the supporting tables.