|Home | About | Journals | Submit | Contact Us | Français|
Voltage-gated A-type K+ channel Kv4.2 subunits are highly expressed in the dendrites of hippocampal CA1 neurons and are thought to regulate dendritic integration. However, little is known about the subcellular distribution and trafficking of Kv4.2 subunits. Here we provide evidence for activity-dependent trafficking of Kv4.2 in spines and shafts of hippocampal dendrites. Live imaging and electrophysiological recordings showed that Kv4.2 internalization is induced rapidly upon glutamate receptor stimulation. Kv4.2 internalization was clathrin-mediated and required NMDA receptor activation and Ca2+ influx. In dissociated hippocampal neurons, miniature excitatory postsynaptic current amplitude depended on functional Kv4.2 expression level and was enhanced by stimuli that induced Kv4.2 internalization. Long-term potentiation (LTP) induced by brief glycine application resulted in synaptic insertion of GluR1-containing AMPA receptors along with Kv4.2 internalization. We also found evidence of Kv4.2 internalization upon synaptically evoked LTP in CA1 neurons of hippocampal organotypic slice cultures. These results present a novel mechanism for synaptic integration and plasticity by the activity-dependent regulation of Kv4.2 channel surface expression.
A number of recent studies investigating the molecular mechanisms of synaptic plasticity have focused on glutamate receptor (AMPAR and NMDAR) trafficking (Bredt and Nicoll, 2003; Groc et al., 2004; Lu et al., 2001; Perez-Otano and Ehlers, 2005; Shi et al., 1999; Snyder et al., 2001). AMPAR subunits in particular display a high degree of mobility, undergoing insertion and removal from the synaptic membrane via endo- and exocytosis, respectively. Changes in the level of synaptic AMPAR expression lead, at least in part, to the degree and direction of synaptic plasticity (Beattie et al., 2000; Brown et al., 2005; Collingridge et al., 2004; Ehlers, 2000; Malinow and Malenka, 2002; Park et al., 2004; Passafaro et al., 2001). It has been shown that rapid insertion of the AMPAR subunit GluR1 mediates an increase in spontaneous miniature excitatory postsynaptic current (mEPSC) amplitude in hippocampal cultured neurons (Lu et al., 2001), while removal of AMPAR subunits GluR1 and GluR2 from the synapses results in LTD of mEPSC amplitudes (Brown et al., 2005). Although these results clearly demonstrate a role for AMPAR trafficking in synaptic plasticity, synaptic events are susceptible to subsequent processing by passive dendritic filtering and dendritic voltage-gated ion channel activity (Cai et al., 2004; Faber et al., 2005; Hoffman et al., 1997; Lipowsky et al., 1996; Magee, 1999; Magee and Johnston, 1995; Ngo-Anh et al., 2005; Ramakers and Storm, 2002). We questioned whether the regulated trafficking of ion channels might also be employed by neurons to control dendritic excitability and synaptic integration.
We focused on the K+ channel subunit Kv4.2, which underlies the large A-type current found in the dendrites of CA1 pyramidal neurons of the hippocampus. In these neurons, A-type currents have been shown to regulate both sub- and suprathreshold dendritic signals (Cai et al., 2004; Cash and Yuste, 1998; Hoffman et al., 1997; Kim et al., 2005; Ramakers and Storm, 2002). Although, Kv4.2 has previously been shown to be enriched in spines (Alonso and Widmer, 1997; Kim et al., 2005), the subcellular trafficking of Kv4.2 has yet to be been investigated. We report here an activity-dependent redistribution of Kv4.2 out of dendritic spines in cultured hippocampal neurons. This redistribution requires NMDAR activation and Ca2+ influx and occurs through clathrin-mediated endocytosis. Glycine-induced long term potentiation (chemLTP) resulted in concurrent synaptic GluR1 insertion and Kv4.2 internalization. Whole-cell recordings supported these results showing that endogenous A-type, but not sustained, K+ currents are reduced after chemLTP induction. Expression of a Kv4.2 dominant negative mutant was found increase miniature excitatory postsynaptic current (mEPSC) amplitudes and stimuli that induce Kv4.2 internalization resulted in enhanced synaptic currents. Finally, in hippocampal organotypic slice cultures we show that synaptically induced LTP is reduced by hyperpolarization, suggesting that Kv4.2 internalization contributes to LTP. Activity-dependent regulation of Kv4.2 surface expression thus presents a novel, synapse-specific molecular mechanism for the fine-tuning of dendritic signals.
To monitor the subcellular distribution and activity dependent trafficking of Kv4.2, enhanced green fluorescent protein (EGFP) was attached to the C-terminal (Kv4.2g). This construct mimics the somatodendritic expression pattern of endogenous Kv4.2 in hippocampal CA1 pyramidal neurons, with nearly identical biophysical properties (Kim et al., 2005). Using immunogold labeling of endogenous Kv4.2, electron micrographs showed Kv4.2 subunits located in hippocampal CA1 spines, supporting our previous observation of Kv4.2g spine localization (Figure S1). In mature, cultured hippocampal neurons, application of a selective AMPAR agonist (100 μM AMPA, 15 min) caused a redistribution of Kv4.2g out of spines, accumulating in puncta both in the dendritic shaft and soma (Figure 1A). This AMPA-induced redistribution of Kv4.2g was reversible, indicating that the treatment was not excitotoxic and showing that Kv4.2g returns to spines within 6 h (Figure 1A). Redistribution of Kv4.2g was also observed with KCl (25mM) depolarization, glutamate (50 μM), and NMDA (20 μM) treatment (Figure S2).
The loss of spine Kv4.2g fluorescence upon stimulation could be due to spine retraction, lateral mobilization of Kv4.2g within the membrane but out of spines, or internalization of Kv4.2g out of the membrane as has been found for AMPARs (Collingridge et al., 2004; Malinow and Malenka, 2002). To test for Kv4.2 internalization, we performed a biotinylation assay to detect surface Kv4.2g levels before and after stimulation. Surface Kv4.2g was markedly decreased in stimulated neurons without a significant change in total protein level (AMPA: 52.5±2.9%, n=3; total protein: 98.4±2.1%, n=3; Figure 1B). Again, the stimulation-induced decrease in surface Kv4.2g was almost completely recovered 6 h after withdrawal of AMPA (normalized surface Kv4.2g level: 53.1±2.7% of basal for 1 h washout, 59.4±5.1% for 2 h, 76.4±10.7% for 4 h, 96.3±1.1% for 6 h, n=3 ; Figure 1B, bottom).
Given the large number of synapses stimulated in these conditions we were able to directly measure internalization as a decrease in the endogenous whole-cell transient K+ current from young, uninfected hippocampal neurons (Figure 1C and D). After a control recording period, AMPA (50μM) was applied for 5 min. Endogenous transient K+ currents measured at +120 mV decreased by 24.5±5.0% after AMPA stimulation (from 2.96±0.19 nA to 2.24±0.19 nA, n=12, p<0.05, paired t-test) without a change in sustained or non-inactivating delayed rectifier-type voltage-gated K+ current amplitudes (0.96±0.05, n=12, p>0.05, paired t-test; Figure 1D). We also noted an increase in the rate of inactivation of the transient K+ currents after stimulation (from 57.8±7.3 ms to 41.8±8.8 ms, n=12, p<0.05, paired t-test, not shown). This change in average inactivation rate may indicate targeted internalization of a subset of Kv4.2-containing channels exhibiting slower inactivation (depending on their phosphorylation state or complement of auxiliary βsubunits) (Hoffman and Johnston, 1998; Jerng et al., 2004). Consistent with our imaging and biochemical analysis of Kv4.2g expression, as well as electrophysiological recordings of the endogenous transient K+ currents, the surface expression level of endogenous Kv4.2 is reduced by AMPA stimulation in acute hippocampal slices as well as in dissociated neurons (normalized surface Kv4.2g level: 67.8±6.1 % of basal n=4 for slices; Figure 1E, 71.8 8±5.7 % of basal n=3 for dissociated neurons, not shown). Thus, the surface expression level of Kv4.2 is regulated by membrane internalization and reinsertion in an activity-dependent manner.
Stimulation triggering Kv4.2g internalization might also lead to changes in synapse number or structure. To determine this, we visualized pre-and postsynaptic elements using antibodies against endogenous synaptophysin and the primary NMDAR subunit (NR1), respectively (Figure S3). In AMPA-stimulated dendrites, the number of Kv4.2g puncta co-labeled with synaptophysin decreased by approximately 84% (11.9±2.4%, n=552 puncta from 11 neurons after AMPA vs. 75±2.5%, n=539 puncta from 11 neurons for unstimulated sister cultures, “basal”). Likewise, the number of Kv4.2g puncta co-labeled with NR1 was considerably reduced (~ 82%) upon AMPA stimulation (15.4±3.2%, n=443 puncta from 12 neurons for AMPA vs. 87.7±1.7%, n=489 puncta from 12 neurons for basal). The number of NR1 clusters per 10 μm of dendrite was unchanged after stimulation (6.54±0.31, n=24 for AMPA stimulation vs. 6.52±0.41, n=19 for basal). Another spine marker, fluorophore-conjugated phalloidin used to label F-actin, showed similar results (Kv4.2g-positive spines: 17.9±4.5%, n=500 from 16 neurons for AMPA stimulation vs. 89.4±2.2%, n=1186 from 26 neurons for basal; Figure 2). Kv4.2g therefore appears to undergo activity-induced internalization, without a gross change in synaptic architecture or number.
Preventing synaptic release with tetanus toxin (TeTN) largely eliminated AMPA-induced Kv4.2g redistribution (Kv4.2g-positive spines: 65.6±2.0%, n=1402 from 27 neurons; Figure 2). In addition, the voltage-gated sodium channel blocker TTX (1 μM) reduced the AMPA-induced redistribution of Kv4.2g but not redistribution induced by KCl depolarization (Kv4.2g-positive spines: 55.7±7.3%, n=403 from 9 neurons for AMPA+TTX; Figure 2, 10.0±3.0%, n=264 from 7 neurons for KCl+TTX; Figure S2). These results indicate that, although synaptic release is necessary to trigger Kv4.2g redistribution, AMPAR activation alone is insufficient, suggesting a role for NMDARs. To test for NMDAR-dependence, the NMDAR antagonist D,L-APV (100 μM) was applied 15 min prior to AMPA stimulation. NMDAR block largely prevented activity-induced redistribution of Kv4.2g (Kv4.2g-positive spines: 74.9±3.6%, n=302 from 8 neurons for AMPA+APV; Figure 2 and 59.7±3.3%, n=270 from 7 neurons for KCl+APV, Figure S2). Consistent with the requirement of synaptic release and NMDAR activation, we found AMPA-induced Kv4.2g redistribution to be blocked by pre-incubating neurons with the membrane-permeable Ca2+ chelator BAPTA-AM (10 μM) (Kv4.2g-positive spines: 78.4±2.6%, n=544 from 16 neurons for AMPA+BAPTA-AM; Figure 2). Kv4.2g redistribution thus requires NMDAR activation and Ca2+ influx.
To estimate the time course of Kv4.2g redistribution we performed live-imaging experiments in cultured hippocampal neurons co-expressing Kv4.2g and the improved red fluorescent protein, tdTomato (Shaner et al., 2004). Time-lapse imaging showed that AMPA-driven Kv4.2g redistribution from spines begins rapidly upon stimulation and progresses over a period of about 15 min (ΔF/F0 of normalized Kv4.2g in spines after 15 min AMPA: −0.25±0.02, n=192 spines from 9 neurons; Figure 3 and Movie S1). Only a small change in ΔF/F0 was found in the adjacent dendritic shaft after 15 min AMPA-treatment (−0.08±0.03, not shown). Pre-incubation with APV almost completely blocked AMPA-induced Kv4.2g fluorescence changes in spines ( ΔF/F0 of normalized Kv4.2g at 15 min time point: −0.05±0.02, n=100 spines from 6 APV-treated neurons vs. −0.02±0.01, n=75 spines from non-treated 4 neurons; Figure 3 and Movie S2; non- or APV pre-treated dendritic shafts: −0.01±0.01, not shown).
We next addressed the mechanism of Kv4.2g redistribution using a synthetic dynamin-derived peptide (DYN) to inhibit clathrin-mediated endocytosis by blocking the recruitment of dynamin to clathrin-coated pits (Damke, 1996; Nong et al., 2003). This peptide has been used previously to show functional AMPA receptor expression level is controlled by clathrin-dependent receptor internalization (Brebner et al., 2005). Imaging and electrophysiological experiments both show that activity-dependent Kv4.2g redistribution and/or endogenous A-current reduction is blocked by the DYN peptide (Figure 4). Pre-incubation of cultured neurons with membrane permeant myrs-DYN for 10 min almost completely eliminated AMPA-mediated Kv4.2g redistribution whereas a scrambled peptide (myrs-scramDYN) had no effect on Kv4.2g redistribution (Kv4.2g-positive spines: 82.3±1.6%, n=927 from 20 neurons for AMPA+myrs-DYN and 20.0±2.6%, n=455 from 13 neurons for AMPA+myrs-scramDYN, Figure 4A and B). Preventing endocytosis by pre-incubating neurons in hypertonic media (0.45 M sucrose) also abolished AMPA-induced Kv4.2g redistribution (Kv4.2g-positive spines: 92.1 ± 1.9%, n=685 from 19 neurons; not shown). AMPA-driven Kv4.2g redistribution thus occurs through clathrin-mediated endocytosis.
In support of the imaging results, inclusion of membrane impermeant DYN (but not scramDYN) in the patch pipette in whole-cell recordings prevented the AMPA-induced reduction of endogenous A-currents in cultured hippocampal neurons (normalized peak transient current after AMPA stimulation for DYN, 0.91±0.02, p>0.05, paired t-test, n=7; Figure 4C and D). In recordings using scramDYN, AMPA again reduced endogenous transient but not sustained K+ currents (normalized peak transient current after AMPA stimulation for scramDYN, 0.57±0.07, p<0.05, paired t-test, n=4; Figure 4C and D).
These experiments revealed a physiological consequence of Kv4.2 internalization: a reduced capacity for neurons to counter membrane depolarization. In these recordings, we monitored the peak AMPA-induced membrane depolarization (see Experimental Procedures) and the time it took to recover after washout. Upon AMPA stimulation, intracellular DYN reduced both peak membrane depolarization and recovery time compared to scramDYN (Figure 4E) and control recordings (no peptide, −36.42±1.69 mV peak depolarization and 305.83±16.01 sec for recovery, n=12). These data show that regulated Kv4.2 trafficking provides a potent mechanism for controlling membrane excitability.
Previous investigations demonstrated AMPAR internalization by a number of different stimuli (synaptic activity, insulin treatment, Ca2+ influx, and ligand-binding to AMPARs) (Brown et al., 2005; Carroll et al., 1999; Lin et al., 2000; Lissin et al., 1999; Park et al., 2004; Zhou et al., 2001), through distinct endosomal sorting pathways. Ligand-binding (i.e. AMPA) induced AMPAR internalization is NMDAR independent (Lin et al., 2000; Lissin et al., 1999), indicating a separate internalization mechanism from our results with Kv4.2. To compare the trafficking of Kv4.2 and AMPARs, Kv4.2g expressing neurons were stimulated by AMPA (50 μM, 15 min) and subsequently immunostained with anti-GluR1 to detect endogenous AMPARs. AMPA treatment triggered the internalization of both Kv4.2g and GluR1 away from spines into the dendritic shaft (Figure 5). For analysis, Kv4.2g and GluR1 fluorescent signal intensity was plotted for both the dendritic shaft and spines, as illustrated in Figure 5B. Line plots of spine fluorescence indicated almost complete co-localization of Kv4.2g and GluR1 in unstimulated dendritic spines, which was abolished with AMPA stimulation (Figure 5A and C). Within the dendritic shaft, after AMPA stimulation, fluorescent signals from internalized Kv4.2g and GluR1 were non-overlapping. Co-application of APV and AMPA resulted in GluR1 but not Kv4.2g redistribution (Figure 5A,C, and D). These results show that AMPA-dependent internalization of the two channels proceeds through distinct intracellular pathways.
We note that GluR1 internalization during AMPA stimulation would be expected to decrease neuronal excitability. Thus, the reduced depolarization observed upon AMPA stimulation in experiments where internalization is blocked using an inhibitory dynamin peptide (DYN, Figure 4E) may actually reflect an underestimate of the effect of Kv4.2 internalization on membrane excitability.
To determine if surface Kv4.2 expression affects synaptic integration, we altered the level of functional Kv4.2 by expressing Kv4.2g or the dominant negative mutant Kv4.2gW362F (Kim et al., 2005; Malin and Nerbonne, 2000) and compared basal mEPSC amplitudes (Figure 6). Whole-cell transient K+ currents were increased ~ 2-fold in Kv4.2g expressing neurons, and decreased ~ 60% in Kv4.2gW362F expressing neurons compared to control, with no change in the sustained K+ current amplitude (transient: 7.85±0.56 nA, n=6 for Kv4.2g expressing vs. 3.87±0.46 nA, n=7 for control and 1.97±0.22 nA, n=7 for Kv4.2gW362F, p<0.05, One-way ANOVA; sustained: 1.23±0.13 nA, n=6, 1.19±0.38 nA, n=6 and 1.56±0.86 nA, n=7 for Kv4.2g, control and Kv4.2gW362F, respectively, p>0.05, data not shown) (Kim et al., 2005).
Functional Kv4.2 expression level determined basal mEPSC amplitude (11.7±0.5 pA, n=31 for control vs. 10.8±0.8 pA, n=17 for Kv4.2g and 17.6±1.5 pA, n=15 for Kv4.2gW362F, p<0.05, One-way ANOVA; Figure 6A and B) and rise-time (11.7±0.5 pA, n=31 for control vs. 10.8±0.8 pA, n=17 for Kv4.2g and 17.6±1.5 pA, n=15, for Kv4.2gW362F, p<0.05, One-way ANOVA, not shown). Due to inadequate voltage control of the synapse, synaptic charge is a more accurate measure of synaptic efficacy than amplitude (Barrett and Crill, 1974; Carnevale and Johnston, 1982; Spruston et al., 1993). We found charge (measured as the total mEPSC area) also depended on level of functional Kv4.2 level (81.7± 5.4 pA*ms for control vs. 64.1±4.4 pA*ms for Kv4.2g and 117.8±11.0 pA*ms for Kv4.2gW362F, p<0.05, One-way ANOVA; Figure 6B). Figure 6C shows the normalized charge distribution for all recorded mEPSCs. Kv4.2gW362F expressing neurons displayed a broader charge distribution with a shift toward larger charges compared to control whereas Kv4.2g expressing neurons had a more narrow distribution range with a shift toward smaller charges. In control, uninfected neurons, bath application of 4-aminopyridine (4-AP, 7 mM) to block endogenous A-type K+ channels increased mEPSC charge in 3 of 5 neurons tested (Figure 6D, p< 0.05, paired t-test).
The effect of Kv4.2 on mEPSC amplitudes suggests a mechanism by which active synapses could increase their efficacy by locally decreasing Kv4.2 surface expression. In our imaging experiments, NMDARs were activated through global, AMPA-mediated depolarization and subsequent glutamate release. That TTX did not entirely block Kv4.2 redistribution (Figure 2) suggests that upon depolarization, spontaneous activity may provide sufficient NMDAR activation to induce Kv4.2 internalization. To test this possibility we monitored mEPSCs before and after bath applied AMPA (25–50 μM, 2–3 min). Since this treatment causes not only Kv4.2 internalization but also AMPAR internalization (Figure 5) (Lin et al., 2000; Lissin et al., 1999) we report the ratio of mEPSCs recorded at −60 mV to those recorded at −80 mV. The rationale for this analysis is that, whereas AMPA-mediated EPSCs show a linear I-V relationship between rest and hyperpolarized potentials, a holding potential of −80 mV brings Kv4.2 channels out of their activation range (Kim et al., 2005). AMPA stimulation increased the −60/−80 mV mEPSC amplitude ratio in 6 of 6 cells (0.81±0.02 to 0.91±0.02, p<0.05, Wilcoxon; Figure 6E and F). As with Kv4.2g internalization (Figure 3) this effect was blocked by pre-application of APV (0.87±0.01 to 0.83±0.03, p>0.05, Wilcoxon; Figure 6F). Blocking A-type K+ channels with 4-AP (7 mM) also increased the ratio (0.86±0.03, pre to 0.97±0.01 after 4-AP, p<0.05, Wilcoxon; Figure 6F), showing that K+ channels shape mEPSCs recorded at −60 mV.
Our results demonstrate activity-dependent internalization of Kv4.2 in active spines enhances their efficacy through an NMDAR and Ca2+-dependent mechanism. Because LTP shares these NMDAR and Ca2+ requirements for induction we wondered if LTP induction would stimulate Kv4.2 internalization. A brief application of the NMDAR co-agonist glycine has been shown previously to increase the amplitude and the frequency of spontaneous mEPSCs through synaptic GluR1 insertion (chemLTP) (Lu et al., 2001). We too found enhanced mEPSC amplitude (120.6±8.3% after 30 min; Figure S4) and surface GluR1 enhancement after chemLTP in both control and Kv4.2g expressing neurons (200 μM glycine, 3–5 min; Figure 7A and F).
However, chemLTP induction also drove Kv4.2g out of spines (Figure 7A–D). Similar to AMPA-stimulated internalization of Kv4.2g, the number of Kv4.2g puncta co-labeled with synaptophysin or NR1 was markedly decreased (Figure 7B and C). No significant change in NMDAR distribution after glycine treatment was detected by immunostaining (number of NR1 clusters per 10 μm dendrite: 6.76±0.46, n=17, pre-chemLTP vs. 6.80±0.22, n=21, post-chemLTP). ChemLTP also led to the reorganization of F-actin away from spines, forming tether-like filamentous structures in the dendritic shaft. Such reorganization was previously shown to be NMDAR-dependent (Hering and Sheng, 2003). It is also important to note that no significant differences in expression level of total or surface AMPARs were observed between Kv4.2g expressing and control neurons by immunostaining and blotting (Figure 7E-G). A biotinylation assay confirmed that surface GluR1 increased similarly between the two groups (151.6±11.8%, n=3 for Kv4.2g expressing vs. 158.6±14.7%, n=4 for control) and that ~18% of Kv4.2g was internalized during chemLTP (82.1±4.2% of basal, n=3; Figure 7G).
Consistent with these imaging results showing Kv4.2g internalization during chemLTP, endogenous transient K+ current amplitude decreased in whole-cell recordings (60.0±5.8%, n=6, p<0.05, paired t-test), with no significant change in sustained K+ current amplitudes (96.5±6.5%, n=6, p>0.05, paired t-test; Figure 7H and I). The time course of this current decrease is also similar to the time course of Kv4.2g internalization after AMPA stimulation measured in our live imaging experiments (Figures 3 and and7I,7I, and Movie S1). In addition, the chemLTP-induced decrease in endogenous transient K+ current was blocked by APV (transient; 100.9±2.6%, sustained; 99.7±3.3%, n=5; data not shown). Notably, although a larger fraction of A-type K+ channel density remains in Kv4.2g expressing neurons after chemLTP stimulation compared to control (Figure 7H and I), the absolute magnitude of decrease in peak transient K+ current density is constant in both Kv4.2g expressing and control neurons (1.78±0.09 nA for Kv4.2g expressing vs. 1.74±0.12 nA for control; Figure 7J). This fixed capacity for Kv4.2 internalization in response to synaptic activity could be due to limitations of the endogenous Kv4.2-related endocytosis machinery. Taken together, these results show that the cellular mechanism of chemLTP expression is associated with a specific decrease in functional A-type K+ channels, along with the increase in synaptic AMPARs.
We also found evidence for Kv4.2 internalization after synaptically evoked LTP induction in hippocampal organotypic slice cultures (Figure 8). Here, we measured evoked synaptic currents alternatively at −60 and −80 mV. The −60/−80 mV ratio was monitored before and after LTP induced by a pairing protocol (Barria and Malinow, 2005). Figure 8A shows an individual experiment where we plot the % change in EPSC amplitude at the two holding potentials after LTP. The amount of potentiation observed was greater at −60 mV than at −80 mV, resulting in an increase in −60/−80 mV ratio. Enhanced potentiation at −60 mV compared to −80 mV was observed in 7 of 7 neurons tested (average EPSC change 30–35 min post-LTP: VH-60 mV =88±15%, VH-80 mV =51±16%, p<0.05, Wilcoxon; Figure 8B). The average −60/−80 mV EPSC ratio increased from 0.57±0.04 pre-LTP induction to 0.79±0.07 thirty min post-LTP induction. Only synapses receiving stimulation during pairing displayed any potentiation at either holding potential showing that Kv4.2 internalization is specific to potentiated synapses (Figure 8B, “unpaired”). These results are consistent with synapse specific Kv4.2 internalization as a contributing factor to LTP.
Our primary finding is the activity-dependent redistribution of the dendritic voltage-gated K+ channel subunit, Kv4.2, out of spines of cultured hippocampal neurons. This internalization enhanced mEPSC efficacy and shares common requirements for induction with LTP, namely, NMDAR activation and Ca2+ influx. Whole-cell voltage-clamp recordings demonstrated that stimulations inducing Kv4.2 internalization result in a specific reduction of endogenous A-type K+ currents upon synaptic activation. LTP induced by brief glycine application resulted synaptic GluR1 insertion but also Kv4.2 internalization. In addition we provide evidence for Kv4.2 internalization as a contributing factor to synaptically evoked LTP. These results present a novel mechanism for controlling synaptic integration through regulation of the surface expression level of voltage-gated A-type K+ channels.
Although the large number of synapses stimulated in our imaging experiments resulted in Kv4.2g internalization both in spines and the dendritic shaft, computer modeling suggests the major affect of A-type channels on synaptic events occurs at the site of input and the relative amount of voltage attenuation from dendrite to soma is unchanged by 4-AP application (Hoffman et al., 1997). Our previous results showing enriched spine expression of Kv4.2g (Kim et al., 2005) along with the present demonstration of activity-dependent removal of Kv4.2g from spines (Figures 1 and S2), suggest a specific role of spine Kv4.2 in regulating synaptic activity during normal synaptic input. Such an arrangement may facilitate the synapse specificity of internalization and its effect on synaptic efficacy. Previously it was shown that A-type K+ channels act to prevent subthreshold Na+ channel activation, countering Na+ channel EPSP boosting (Hoffman et al., 1997). Kv4.2 internalization would thus be expected to enhance synaptic potentials by removing this block to Na+ channel activation.
It is important to question whether our results showing Kv4.2g internalization occur normally for endogenous Kv4.2. A number of observations provide strong evidence that the activity-dependent regulation of the Kv4.2 represents a physiological mechanism for controlling excitability: (1) In Figure 1, we show both Kv4.2g and endogenous Kv4.2 membrane fractions are reduced upon AMPA stimulation in biotinylation experiments. (2) Kv4.2g expression did not alter the expression levels or trafficking behavior of endogenous GluR1 and NMDARs. (3) We found no differences in synaptic structure or number in Kv4.2g expressing neurons compared with control. (4) We show here that both Kv4.2g and endogenous Kv4.2 are located in hippocampal spines. (5) AMPA stimulation and chemLTP induction both result in a specific decrease in endogenous transient (but not sustained) K+ currents as well as Kv4.2g internalization (assessed by fluorescent imaging and biotinylation assays). (6) Both activity-dependent Kv4.2g internalization and endogenous A-type K+ current decreases are blocked by APV. (7) Both AMPA-induced Kv4.2g internalization and endogenous A-type K+ current decreases are blocked by an active dynamin peptide. (8) Blocking endocytosis with the dynamin peptide reduced AMPA-induced depolarization. (9) In chemLTP experiments, control and Kv4.2g expressing neurons displayed nearly identical reductions (magnitude and time course) in transient K+ current amplitudes. And finally, (10) LTP was reduced by hyperpolarization, consistent with Kv4.2 internalization during synaptically evoked LTP.
Accumulating evidence shows that synaptic plasticity is at least partially accomplished through the trafficking of postsynaptic receptor proteins in hippocampal neurons (Barry and Ziff, 2002; Collingridge et al., 2004). In particular, the insertion or removal of synaptic AMPARs is thought to determine the direction of synaptic plasticity (potentiation or depression) (Brown et al., 2005; Malinow and Malenka, 2002; Man et al., 2000; Shi et al., 1999). Here, we provide evidence that the regulated surface expression of a dendritic voltage-gated transient K+ channel serves as an additional contributor to synaptic plasticity. In support of previous reports showing that A-type K+ currents regulate EPSP amplitude in hippocampal neurons, we found decreasing functional Kv4.2 expression levels enhanced the average amplitude and charge of mEPSCs recorded in the soma (Hoffman et al., 1997; Ramakers and Storm, 2002). Given the distinctive responses of Kv4.2 and GluR1, we can now say that postsynaptic protein trafficking during chemLTP is, in effect, a two-way street. This finding raises the question of how neurons distinguish and control movement of different types of channels. Whether surface expression of other dendritic voltage-gated ion channels is also regulated by synaptic activity and knowledge of the conditions under which channels are re-inserted back into the membrane calls for further study. Future studies will also elucidate the molecular pathways and mechanisms of activity-dependent Kv4.2 internalization and reinsertion including the potential role of posttranslational modifications and binding partners.
Excitatory neuronal stimulation with either AMPA application or chem-LTP induction caused Kv4.2 internalization. In contrast, GluR1 has opposing responses to the two stimulations with AMPA stimulation causing internalization (Fig. 5) but chem-LTP enhancing surface expression (Fig. 7). Some resolution for this disparate molecular behavior comes from the observation that GluR1 internalization by ligand binding (AMPA stimulation) is mimicked by CNQX and does not require NMDAR activation and thus is not a specific response to activity (Lin et al., 2000).
A previous report has found LTP induction in acute hippocampal slices to result in a hyperpolarizing shift in dendritic transient K+ current inactivation curve, enhancing back-propagating action potential amplitude (Frick et al., 2004). While numerous methodological differences prevent a direct comparison between the two studies, our results are not necessarily mutually exclusive. Although Frick et al. did not report a specific decrease in peak transient K+ channel amplitude with LTP in cell-attached dendritic patches, their LTP induction may indeed cause internalization, which is not detected in the cell-attached patch because the machinery is compromised during patch formation and/or because the patch is physically too distant from the potentiated synapses. On the other hand, the inactivation curve shift seen in dendritic patches by Frick et al. may also occur with chemLTP. This curve shift would not be expected to contribute to the decrease in transient K+ currents we observed (Figure 7), given that their measurement followed a hyperpolarizing prepulse to −120 mV, removing inactivation even with the inactivation curve shift. In our pairing induced LTP in CA1 neurons from hippocampal organotypic slices, outside-out patches pulled 30 min after LTP induction revealed transient K+ current inactivation curves identical to control (S-C. J., unpublished observations). Thus, our reported difference between potentiation measured at −60 and −80 mV (Figure 8) cannot be explained by a K+ current inactivation curve shift at this time point. More recently the same group has shown enhanced induction of long-term potentiation in hippocampal CA1 pyramidal neurons of Kv4.2 knockout mice, suggesting a potential role for Kv4.2 trafficking in meta-plasticity (Abraham and Bear, 1996; Chen et al., 2006).
Transient K+ channels regulate a number of other measures of excitability including the back-propagation of action potentials in dendrites of hippocampal neurons, action potential initiation in dendrites, the propagation of dendritic Ca2+ plateau potentials, action potential half-width and frequency dependent action potential broadening (Cai et al., 2004; Frick et al., 2004; Hoffman et al., 1997; Kim et al., 2005; Losonczy and Magee, 2006). Recent evidence shows that Kv4.2 prevents plateau potentials from propagating past branch points (Cai et al., 2004), that transient K+ currents shaped fast Na+ spikes in oblique CA1 pyramidal neuron dendrites (Losonczy and Magee, 2006), and that Kv4.2 is clustered at GABAergic as well as glutamatergic synapses (Burkhalter et al., 2006; Jinno et al., 2005; Kollo et al., 2006). Kv4.2 endocytosis may then have different effects on integration for different populations of Kv4.2 containing channels depending on their location in the dendrite. Likewise, targeted insertion of Kv4.2 into specific dendritic compartments may be employed by neurons to shape dendritic integration and backpropagation.
Here we have shown that Kv4.2 internalization occurs after global synaptic activation with AMPA, mimicking a large, distributed barrage of synaptic activity. This activity would also be expected to affect suprathreshold dendritic signals controlled by Kv4.2, such as dendritic action potential back-propagation, furthering dendritic Ca2+ influx and excitability. Such highly synchronized and frequent discharges occur during epileptic seizures. Recently a Kv4.2 truncation mutation was found in a patient with temporal lobe epilepsy (Singh et al., 2006) and studies in epilepsy models have demonstrated an activity-dependent downregulation of Kv4.2 channels (Bernard et al., 2004; Tsaur et al., 1992). Kv4.2 internalization may then be a contributing factor to the hyperexcitability observed in temporal lobe epilepsy and/or may play an important role in susceptibility to seizure onset in the hippocampus.
Hippocampal primary cultures were prepared from embryonic day 18 Sprague-Dawley rats as per Osten et al., 1998 with the exception of coverslip coating. Here, hippocampi were triturated 15 min after trypsinization and plated on coverslips coated with poly D-lysine (50 μg/ml, Sigma) and laminin (5 μg/ml, Invitrogen). Dissociated neurons were cultured in neurobasal medium supplemented with B27 (Invitrogen). To eliminate proliferative glia cells, 5 μM cytosine arabinoside (AraC, Sigma), a specific inhibitor of DNA synthesis during meiosis and mitosis, was included after 8 DIV. Primary neurons were infected with a normalized infectious titer of modified Sindbis virus resulting in 10–20% infection of neurons for imaging and electrophysiological experiments one day before use and ~30% for biotinylation. Attenuated Sindbis viruses expressing EGFP-tagged Kv4.2 and EGFP alone were produced using the SINrep(nsP2S726) viral vector and DH-BB(tRNA/TE12) helper plasmid as described (Kim et al., 2004). The improved red fluorescent protein (tdTomato, provided by Dr. Tsien, (Shaner et al., 2004)), a tandem dimer variant of DsRed, was transfected by Nucleofection (Amaxa Biosystems) in accordance with the manufacturer’s protocol. Other constructs used in present experiments were described previously (Kim et al., 2005). Hippocampal organotypic slice cultures (400 μm thick) were prepared from postnatal day 7–8 Sprague-Dawley rats after (Shi et al., 1999). Hippocampal CA1 neurons were infected on 4 DIV by microinjection (Nanoliter 2000, World Precision Instruments). All animal procedures were conducted with accordance of the National Institutes of Health Guide for the Care and Use of Laboratory Animals under a protocol approved by the National Institutes of Child Health and Human Development’s Animal Care and Use Committee.
Hippocampal neurons (18–24 DIV) or acute slices (postnatal day 14) were incubated with freshly made sulfosuccinimidyl-2-(biotinamido) ethyl-1,3-dithiopropionate (sulfo-NHS-SS-biotin, 1 mg/ml, Pierce) in ice-cold PBS for 30 min at 4°C, followed by 10 min incubation in ice-cold glycine (100 mM) at 4°C, and lysed with lysis buffer containing 25 mM HEPES (pH 7.4), 150 mM NaCl, 1% Triton X-100, 0.05% SDS and a protease inhibitor mixture tablet (Roche). The lysate was incubated at 4°C for 20 min and spun at 13,000 rpm for 10 min. Biotin-labeled surface proteins were precipitated with 20–30 μl of Immobilized Streptavidin beaded agarose (Pierce) at 4°C overnight. The beads were washed five times with lysis buffer and proteins were eluted in 2x SDS loading buffer. Immunoblotting was performed as described (Kim et al., 2005). Antibodies: anti-GFP (Molecular Probes, 1:5,000), anti-GluR1-C (Chemicon, 1:200), anti-Kv4.2 (K57/1, NeuroMab, 1:1,000), and anti-rab4 (BD Transduction Laboratories, 1:1,000).
Hippocampal neurons (18–24 DIV) were stimulated for imaging and biotinylation assays in culture media with either KCl (50 mM, Sigma), Glutamate (50 μM, Sigma), NMDA (20 μM, Tocris) or (S)-AMPA (100 μM, Tocris) at 37°C for 15 min. Due to activity-dependent filamentous structural change of F-actin in these experiments (Halpain et al., 1998; Hering and Sheng, 2003), a more mild stimulation was achieved by reducing the concentration of stimulants (25 mM KCl, 50 μM AMPA). All blockers were applied 15 min prior to stimulation. The blockers used in the experiments were 100 μM D,L-APV, 50 μM CNQX, 1 μM TTX, 20 nM Tetanus toxin (TeTN, Sigma) and 10 μM BAPTA-AM (Molecular Probes). Myristolated DYN peptide (myrs-QVPSRPNRAP) and scrambled DYN (myrs-QPPASNPRVR) were synthesized and purified by Sigma Genosys. Peptides (50 μM) were applied 10 min prior to AMPA stimulation. For the chemical LTP induction, 200 μM glycine was bath-applied in the Mg-free ACSF containing (in mM): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 11 D-glucose, 0.0005 TTX, 0.001 strychnine, and 0.02 bicuculine (pH 7.2) at 37°C for 3-5 min. Neurons were then incubated in the same ACSF without glycine at 37°C for 20 min. Treated neurons were always compared to simultaneously prepared sister culture controls.
Treated neurons were immediately fixed with 4% paraformaldehyde and 0.1% glutaraldehyde in PBS containing 0.12 M sucrose for 8 min on ice and permeabilized with 0.5% Triton X-100 in PBS for 5 min. Actin was labeled with tetramethylrhodamine isothiocyanate (TRITC)-conjugated phalloidin (0.5 μM, Sigma) in PBS containing 1% BSA for 5 min.
Neurons were fixed/permeabilized as described above for GluR1 and synaptophysin staining or fixed with –20°C methanol for NMDAR1. After preblocking with PBS containing 5% NGS, 0.05% Triton X-100, and 450 mM NaCl for 1 h at 4°C, neurons were incubated with antibodies in the blocking solution overnight at 4°C and followed by incubation with Alexa 546-conjugated secondary antibodies (Molecular Probes) for 2 h at RT. To label surface AMPARs, non-permeabilized neurons were incubated with anti-GluR1-N in PBS containing 1% BSA and 4% NGS at 4°C overnight. Antibodies: Alexa 488-labeled anti-GFP (Molecular Probes, 1:10,000), anti-GluR1-N (Calbiochem, 1:20), anti-GluR1-C (Chemicon, 1:200), anti-NMDAR1 (Chemicon, 1:100), and anti-Synaptophysin (Sigma, 1:200).
Fixed cell images were acquired with Leica TCS RS confocal microscope. The same instrument parameter settings were kept for each experiment. Every experiment was repeated a minimum of five times. The two fluorophores were excited with different wavelengths, 488 and 543 nm, and were separately imaged using dual sequential scanning to avoid overlapped emission from one to the other. Acquired images were analyzed using MetaMorph v6.3 (Universal Imaging Corporation) and ImageJ v1.36 (http://rsb.info.nih.gov/ij/) under the same analytic parameter settings for each channel. To determine the number of Kv4.2g-positive spines or co-labeled clusters, marker-labeled spines or puncta (clusters) were randomly selected on 6–27 neurons (260–1400 spines) using a 1.7 mm diameter circle region of interest (ROI). For dendritic shaft analysis, the ROI was moved to the shaft directly below the spine. Fluorescent signals of Kv4.2g within the ROI were then measured. Pixels above the threshold intensity were counted and logged into Excel (Microsoft). To compare the co-localization of Kv4.2g and GluR1, fluorescent intensity was line-plotted by manually drawing a line either inside or outside of dendrites (range 0–255, arbitrary units).
Live imaging was performed at the NICHD Microscopy & Imaging Core using a Zeiss LSM 510 Inverted Meta with 100 × Zeiss alpha plan-neofluar oil objective (http://mic.nichd.nih.gov/index.htm). Hippocampal primary neurons transfected with the tdTomato were plated on 25 mm coverslips and then infected with modified Sindbis virus expressing Kv4.2g on 18–24 DIV. Coverslips containing Kv4.2g and tdTomato co-expressed neurons were placed in an Attoflour cell chamber (Molecular Probes) and perfused with Mg-free ACSF containing (in mM): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 11 D-glucose (pH 7.2). The neurons, objective, and stage were heated in a custom-built incubation chamber supplemented with 5% CO2. Time-lapse images were captured every 1 min for 15–20 min at 37 °C using Zeiss LSM Image Browser software v3.2. The instrument parameter settings were optimized in unstimulated Kv4.2g expressing neurons to avoid photobleaching and image saturation. Each image was a maximal projection of 5–7 z-stacks obtained at 0.5 μm depth intervals. Projection images generated for each time points were analyzed using MetaMorph, ImageJ, and Zeiss software. Spines were identified by tdTomato signal and randomly selected using ROIs. All spines identified in the first time point were followed for all the sequent time images, and thus selection was blind regarding the consequence. Fluorescent changes (ΔF) in Kv4.2g intensity were calculated after normalizing by the tdTomato intensity value in each time point using the initial green fluorescence (F0) prior to stimulation as a baseline. Hence, fractional fluorescent change in each time trace was represented by Δ; F/F0 =(normalized F–normalized F0)/normalized F0. The data were averaged from 75–195 spines of 4–9 neurons from at least 6 independent experiments. Averaged values were presented as means±standard error of the mean (S.E.M.).
Thick-walled, filamented patch electrodes had tip resistances of 3–6 MΩ. Series resistance varied between 8–30 ΩM and recordings where series resistance varied by more than 10% were rejected. No electronic compensation for series resistance was employed. All electrophysiological data were recorded using an Axopatch200B amplifier (Molecular Devices Corp.). Recordings were filtered at 2 kHz for mEPSCs and at 5 kHz for K+ currents and evoked EPSCs. Command pulse generation, data acquisition and analysis were performed using IGOR Pro (Wavemetrics). MiniAnalysis (Synaptosoft), SPSS (SPSS Inc.) and Excel (Microsoft) software were used for further data and statistical analysis. One-way ANOVA, Wilcoxon Signed Rank, and Student’s t tests were used to examine statistical significance, set to p<0.05.
Transient K+ currents were measured in young hippocampal primary neurons (7–8 DIV). These smaller neurons, possessing fewer dendritic processes, were used to limit space-clamp and access resistance errors associated with whole-cell recordings of large (nA) currents. Electrodes were filled with a solution containing (mM): 140 KCl, 2 MgCl2, 10 HEPES, 5 EGTA, 5 ATP, 1 CaCl2, 10 D-glucose (pH 7.3 with KOH). 1 μM TTX was added to the external solution to block voltage-gated Na+ currents except during AMPA stimulation. After recording K+ currents, AMPA (50μM, 5 min) or glycine (200μM, 3 min) containing external solution was applied and currents were recorded every 5 min for 30 min. Transient and sustained K+ currents were digitally separated using a prepulse protocol after the subtraction of leak currents. Peak currents were measured at +120 mV after a 400 ms prepulse to either −120 mV or +30 mV. Non-myristolated DYN peptide (QVPSRPNRAP, 100 μg/ml) or scrambled DYN (QPPASNPRVR, 100 μg/ml) was co-applied in internal solution, 20–25 min prior to AMPA stimulation. During and after AMPA stimulation, resting membrane potential was monitored. Cells depolarizing beyond −30 mV during AMPA stimulation were not analyzed. Glycine application depolarized cells up to −45 mV. Cells that did not recover to −60 mV after 5 min of AMPA or glycine washout were also not analyzed.
For recording mEPSCs, primary dissociated culture neurons of 14–21 DIV were used. The patch electrode solution contained either the same internal solution listed above for K+ current recordings (mEPSC amplitude ratio experiments, in some cases without MgCl2) or, for chemLTP experiments (mM): 100 Cs-gluconate, 5 MgCl2, 0.6 EGTA, 8 NaCl, 40 HEPES, 2 NaATP, 0.3 TrisGTP (pH 7.2 with KOH). TTX (0.5–1 μM), strychnine (1μM) and bicuculine (20μM) were included in the external solution during all mEPSC recordings. After whole-cell formation a 5–10 min recovery period elapsed before data collection. Ten seconds of spontaneous activity were recorded every 30 sec for each holding potential. mEPSC data were then averaged over a 5 min period (total sampling duration of 100 sec/5 min period) for up to 60 min. Only 5 min periods exhibiting more than 10 events and average decay times of less than 10 ms were analyzed. Neurons not recovering to within 10 mV of their resting membrane potential within 20 min after AMPA stimulation (25 μM, 2–3 min) were not analyzed. For amplitude and ratio experiments 2–3 episodes (10–15 min of data) were averaged before and after each condition. For chemLTP analysis, the largest 20% of mEPSC amplitudes for each 5 min period were averaged to reduce the impact of non-specific frequency changes (Stell and Mody, 2002).
Hippocampal organotypic slice cultures were transferred to a submerged recording chamber with a continuous flow of ACSF containing (in mM): 125 NaCl, 2.5 KCl, 25 NaHCO3, 1.25 NaH2PO4, 25 Glucose, 2 CaCl2, 1 MgCl2, 0.005 bicuculine (pH 7.4) and bubbled with 5% CO2/95% O2. In all experiments, 5 μM 2-chloroadenosine was included in the external solution to block recurrent synaptic connections. The patch electrodes were filled with (in mM): 20 KCl, 125 Kglu, 10 HEPES, 4 NaCl, 0.5 EGTA, 4 APT, 0.3 TrisGTP, 10 Phosphocreatin (pH 7.2). To record EPSCs at several holding potentials, whole-cell CA1 pyramidal neuron recordings were made in voltage-clamp mode at 31–32 °C. The pathway between CA1 and CA3 was cut before each experiment. EPSCs were elicited by a test pulse (0.2 ms duration at 0.1 Hz, 30–600 μA amplitude), alternatively at −60 or −80 mV through a glass bipolar electrode (10 μm tip) located at the Schaffer collateral pathway. For LTP induction, 2 Hz stimulation was paired with depolarization to 0 mV for 1 min. In some experiments a second stimulating electrode was used to record EPSCs in a control (unpaired) pathway. Pathway independence was measured using a cross paired-pulse facilitation protocol. LTP-induced changes of EPSC amplitude were monitored for up to 1 hour (but at least 40 min).
We thank Drs. Rebecca Hammond, Pavel Osten, and Chris McBain for their critical review of versions of this manuscript, Arrash Yazdani for technical support and Dr. Ya-Xian Wang for help with the immunogold study. Live imaging was performed at the NICHD Microscopy & Imaging Core with the assistance of Dr. Vincent Schram and Chip Dye. This work was supported by the National Institute of Child Health and Human Development Intramural Research Program.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.