Abstract
The Kv4.2 potassium channel plays established roles in neuronal excitability, while also being implicated in plasticity. Current means to study the roles of Kv4.2 are limited, motivating us to design a genetically encoded membrane tethered Heteropodatoxin-2 (MetaPoda). We find that MetaPoda is an ultrapotent and selective gating-modifier of Kv4.2. We narrow its site of contact with the channel to two adjacent residues within the voltage sensitive domain (VSD) and, with docking simulations, suggest that the toxin binds the VSD from within the membrane. We also show that MetaPoda does not require an external linker of the channel for its activity. In neurons (obtained from female and male rat neonates), MetaPoda specifically, and potently, inhibits all Kv4 currents, leaving all other A-type currents unaffected. Inhibition of Kv4 in hippocampal neurons does not promote excessive excitability, as is expected from a simple potassium channel blocker. We do find that MetaPoda’s prolonged expression (1 week) increases expression levels of the immediate early gene cFos and prevents potentiation. These findings argue for a major role of Kv4.2 in facilitating plasticity of hippocampal neurons. Lastly, we show that our engineering strategy is suitable for the swift engineering of another potent Kv4.2-selective membrane-tethered toxin, Phrixotoxin-1, denoted MetaPhix. Together, we provide two uniquely potent genetic tools to study Kv4.2 in neuronal excitability and plasticity.
Significance Statement
Inhibition of the Kv4.2 potassium channel in neurons via two unique and potent membrane tethered toxins reveals a major role for the channel in plasticity, without increasing neural excitability.
Introduction
Voltage-gated potassium channels play prototypical roles in shaping neuronal excitability, but it is emerging that some channels may also play active roles in plasticity (Pongs, 1999; Remy et al., 2010; Jan and Jan, 2012). One potassium channel family that is particularly implicated in plasticity is the Kv4-family. Kv4 channels can be found in various tissues, notably the brain and the heart (Yuan et al., 2007). Of the three different Kv4 subtypes (Kv4.1–4.3), Kv4.2 (Kcnd2/KCND2) and Kv4.3 (Kcnd3/KCND3) are primarily expressed in the brain, with Kv4.2 showing predominant expression in cortical and hippocampal pyramidal neurons, whereas Kv4.3 is mainly expressed in inhibitory interneurons (Serôdio and Rudy, 1998; Dabrowska and Rainnie, 2010; Carrasquillo et al., 2012; Duménieu et al., 2017; Alfaro-Ruíz et al., 2019). In excitatory neurons, Kv4.2 displays exclusive expression in dendrites (including tufts) and spines, with increased densities at more distant sites from the soma (Kim et al., 2007; Alfaro-Ruíz et al., 2019). Functionally, Kv4.2 evokes rapidly activating and inactivating A-type currents (Covarrubias et al., 2008), and these lay behind its ability to control, for instance, neuronal excitability by preventing backpropagating action potential (bAP) (Covarrubias et al., 2008; Zemel et al., 2018). Not surprisingly, mutations in the Kv4.2 gene are associated with diseases such as epilepsy (Cercós et al., 2021). Importantly, the synaptic localization of Kv4.2 places it in an ideal position to directly modulate synaptic plasticity (Kim et al., 2007; Lugo et al., 2012; Labno et al., 2014). Indeed, several mutations within the channel are intimately associated with severe cognitive deficits, ataxia, and neurodevelopmental delays in human patients (Lee et al., 2014; Lin et al., 2018; Budisteanu et al., 2020; Cheng et al., 2021; Zhang et al., 2021).
Despite its importance in neuronal physio- and pathophysiology, the study of Kv4 channels, Kv4.2 in particular, has largely been stagnant for over a decade, with many unanswered questions specifically pertaining to the means by which the channel may gate plasticity [e.g., (Sandler et al., 2016)]. One key reason for this gap is the limited toolbox of pharmacological and genetic methods to isolate the channel’s activity in select cells. More specifically, there are no subtype-specific pharmacophores (Diochot et al., 1999; Jung and Eun, 2012; Yunoki et al., 2014; Sandler et al., 2016; Cai et al., n.d.), soluble drugs are difficult to restrict to defined cells or subcellular regions, and genetic manipulations [e.g., knockout (KO)-mice (Barnwell and Hrachovy, 2009; Lugo et al., 2012; Smith et al., 2016)] are chronic and may promote compensations and alterations in the neural transcriptome/proteome (Lowenstein and Castro, 2001; Barnwell and Hrachovy, 2009; Smith et al., 2016).
To try to address some of these limitations, we set out to develop genetically encodable peptides (Wu et al., 2008; Auer et al., 2010; Stürzebecher et al., 2010; Choi and Nitabach, 2013; Zhao et al., 2020, 2022) to control the activity of Kv4.2, specifically in excitatory neurons. Here, we present the unique development of a membrane-tethered Kv4-selective HeteroPodatoxin-2, denoted MetaPoda. We find that MetaPoda is an ultrapotent and selective inhibitor of Kv4.2 channels. We combine electrophysiology, docking models, and immunohistochemistry to determine the site of interaction of MetaPoda with the channel and its gating mechanism. In primary hippocampal pyramidal neurons, we find that short (7 d) expression of MetaPoda completely and selectively suppresses Kv4 currents only. This is complemented by elevation in cFos protein levels and hinders potentiation. Lastly, we employ the same strategy to produce another tethered toxin, Phrixotoxin-1, with similar potency. Together, our study provides two new genetic tools to study Kv4.2 and describes a novel relationship between Kv4.2 and neural plasticity.
Materials and Methods
Molecular biology
MetaPoda and T-Snap are based on pCMV(MinDis).iGluSnFR (Addgene, #41732). HpTx-2 (amino acid seq. DDCGKLFSGCDTNADCCEGYVCRLWCKLDW) was amplified from the NosTag fusion HpTx-2 (DNA construct generously provided by Alomone Labs ltd. Israel) and replaced iGluSnFR; for T-Snap, Snap was integrated by ligation. NoMeta was generated by removing HpTx-2 from MetaPoda using inverse PCR. MetaMut was created by three point mutations in MetaPoda (in reference to the HpTx-2 sequence): T27G, T45G, T48C turning C10G, C16G, C17G. For all the above constructs, membrane trafficking (KSRITSEGEYIPLDQIDINV) and ER export signal (FCYENEV) were added to the C-Terminus (CT). pCDNA-Kv4.2 was generously provided by Alomone Labs. Kv4.2GFP was produced by ligating EGFP to the CT of the channel. pCDNA-Kv2.1 was generously provided by Prof. Ilana Lotan of Tel Aviv University. GABAB1 and GABAB2 were generously provided by David J. Adams of University of Wollongong. GIRK1/2 was generously provided by Nathan Dascal of Tel Aviv University. pCDNA-cFos was generously provided by Ami Aronheim of Technion. Kv4.2 S3S4 was created by inverse PCR inserting ferret Kv1.4 S3S4 sequence, as previously described (Zarayskiy et al., 2005). MetaLOV was constructed by inserting LOV-jα into MetaPoda between HpTx-2 and PDGFR, and LOV-jalpha was amplified from pcDNA3.1-CONK1-lumitoxin (Addgene, #51689). Kv4.2GFPAA was created by two point mutations, L277A and V278A, based on MetaPoda. MetaPhix was made by using inverse PCR based on MetaPoda. AAV-CAG-Cre,AAV-hSyn-GCaMP7b, pAAV2/1 (capsid - denoted 2/1) and rAAV2-retro (also known as SL1 capsid) (Tervo et al., 2016) were generously provided by Dr. Kimberly Rittola of Janelia Research Campus. AAV-CAG-tdTomato was made using AAV-CAG as backbone and inserted tdTomato. To create a MetaPoda virus, pAAV-hSynapsin1-FLEx-axon-GCaMP6s-P2A-mRuby3 (Addgene, #112008) was used as the backbone. MetaPoda was then amplified and ligated to replace the axon-GCaMP6s-P2A-mRuby3 insert, leading to a MetaPoda in the aforementioned backbone but with dTomato and not with tdTomato due to a PCR error which works well since plasmid size is an important factor in AAV production. All three AAV plasmids were then used to produce AAV viruses (see below AAV production). All primers used in this report are listed in Table 1.
Primers used for all PCR Reactions
Electrophysiology
Patch-clamp recordings were performed 24–48 h after transfection of HEK293t cells or at 16 DIV (days in vitro) for dissociated hippocampal neurons with MetaPoda infection at day 9, as previously described (Berlin and Isacoff, 2018; Kellner et al., 2021). Briefly, patch-clamp recordings were performed with an Axon MultiClamp 700B amplifier and Axon Digidata 1440A acquisition system, mounted on an OLYMPUS BX51WI microscope (equipped with 5× and 40× objectives). Illumination was performed using a TH4-200 microscope light power supply (OLYMPUS) and X-Cite 110LED (EXCELITAS Technologies) with 435, 475, and 560 nm LEDs. Imaging during patch-clamp recordings was performed using µManager (Edelstein et al., 2010, 2014) with a Pixelink camera sensor. All cells were recorded in the whole-cell mode with a low-pass filter at 10 kHz. Pipette resistance ranged between 5 and 11 MΩ and were filled with an internal solution containing the following (in mM): 135 K-gluconate, 10 NaCl, 10 HEPES, 2 MgCl2(H2O)6, 1 EGTA, 2 Mg-ATP, and pH 7.3 with KOH. HEK293t cells were clamped at −70 mV, and full IV curves were typically obtained by a series of 200–400 ms long voltage steps, ranging from −100 or −50 mV to +60 mV in 10 mV increments, for 200 ms, unless mentioned otherwise. For HEK293t cell recordings, the extracellular recoding solution contained the following (in mM): 135 NaCl, 4 KCl, 10 HEPES, 10 D-glucose, 1.8 CaCl2(H2O)2, 1 MgCl2(H2O)6, and pH 7.4 with NaOH.
Illumination of cells expressing MetaLOV was performed by illuminating either 475 or 435 LED lights with a 460–495 filter during the recording. NaV1.5 expressing HEK293 T cells were recorded with a low resistance pipette (1–3 MΩ) filled with an internal solution containing the following (in mM): 100 CsF, 25 NaF, 10 HEPES, 10 NaCl, 2 ATP-Mg, and 10 EGTA. All cells were Whole cell and Rs compensated (40%). Isolated cells (cells with no or one neighboring cell) were preferred for recording NaV currents, as these were less connected to other cells. For the GluN recordings, cells were recorded in neuronal recording solution (in mM): 138 NaCl, 1.5 KCl, 5 HEPES, 10 D-glucose, 2.5 CaCl2(H2O)2, 2 glycine, and pH 7.4 with NaOH, and the exact same solution containing 100 µM glutamate was perfused during the recording as seen in Figure 2a. GABA/GIRK recordings were done using three solutions perfused as seen in Figure 2b. The initial solution was low K+ (in mM), in 56 NaCl, 4 KCl, 5.5 HEPES, 183 D-glucose, 1.8 CaCl2(H2O)2, 1.2 MgCl2, pH 7.4 with NaOH; then for high K+ (in mM) it is the same as low K+ but with 90 KCL and 11 Glucose with pH 7.4 with KOH. The third solution is the GABA solution in which 10 µM of GABA was added to a high K+ solution.
MetaPoda-positive pyramidal neurons were identified by fluorescence of dTomato (using 560 nm excitation). Neurons were clamped at −70 mV and recorded under gap-free mode. Neuronal recording solution consisted of the following (in mM): 138 NaCl, 1.5 KCl, 5 HEPES, 10 D-glucose, 2.5 CaCl2(H2O)2, 2 glycine, and pH 7.4 with NaOH. For intrinsic excitability, neurons were current clamped at −60 mV resting membrane potential and injected with currents ranging from −10 to 140 pA (in 5 pA increments) and the external solution was supplemented with (in µM) 10 gabazine, 20 CNQX, and 50 AP5 to silence synaptic transmission. Transient A-type currents measurements were done as previously described (Aceto et al., 2020). Briefly, neurons were recorded in regular extracellular solution (above) supplemented with 1 µM TTX, 20 µM CNQX, 10 µM Gabazine, 20 mM TEA, and 300 µM Cd2+. Neurons were patched and compensated for whole-cell resistance and series resistance (40%) and were held at −70 mV. The “maximum K+-conductance” protocol (e.g., Fig. 1h) consisted of an initial hyperpolarization step to −110 mV for 100 ms), followed by 400 ms long series of depolarizing voltage steps ranging from −110 to 60 mV. Then, the same cell was recorded again under a “delayed K+-currents protocol” consisting of an initial depolarization pulse from −70 to −10 mV (for 100 ms) followed by 400 ms long voltage steps ranging from −110 to 60 mV.
For chemical potentiation of neurons (cLTP), neurons were initially recorded under basal conditions in an extracellular recording solution (denoted wash) consisting of the following (in mM): 138 NaCl, 1.5 KCl, 5 HEPES, 10 D-glucose, 2.5 CaCl2(H2O)2, 1.2 MgCl2, and pH 7.4 with NaOH. Cells were then recorded for 5 min in current clamped mode followed by an additional 5 min under voltage clamp (at −70 mV). Then, neurons were stimulated for 5 min by a potentiation solution consisting of the “wash” solution supplemented with 100 µM glycine. After potentiation, the potentiation solution was replaced with wash solution, and neurons were recorded as under basal conditions.
For the native NaV recordings the extracellular recording solution was used which was supplemented with 100 nM TTX, 20 µM CNQX, 10 µM Gabazine, 10 µM APV, 20 mM TEA, 300 µM Cd2+, and 5 mM Ba2+ as was previous shown (Milescu et al., 2010). TTX was added to reduce the sodium current amplitude. The protocol consisted of voltage steps starting at −70 to +20 mV with 10 mV intervals. All cells were Whole cell and Rs compensated (80%). Electrophysiological data were recorded with Clampex 11.0.3 and analyzed by Clampfit 11.0.3 software (Molecular Devices).
Electrophysiology analysis
All analyses described were performed with dedicated software (Clampfit, Molecular Devices). Action potential (AP) features were analyzed with current clamp intrinsic excitability protocols (Szabó et al., 2021). The firing threshold was determined by the peak of the second derivative of the AP. The half-width was calculated as the time across the AP at 50% of the action potential, calculated by the peak amplitude and the firing threshold. The negative peak is reached after the AP before returning to baseline. Sodium and potassium currents were measured from the intrinsic excitability protocol. The negative peak occurring immediately after the voltage step and the positive peak occurring thereafter was considered the sodium and potassium currents, respectively. Input resistance was calculated using Clampfit at nonfiring current stimulations.
The IV curve for all Kv4.2 channel recordings was calculated as follows: the peak current after the initial stimuli (5–30 ms) was subtracted by the leak current toward the end of the stimuli. Then, plotted as a current-voltage plot since Kv4.2 current should not be present at the end of the stimuli (200 ms). Kv2.1 currents were calculated as the peak current at each voltage.
Transient A-type currents were calculated by subtracting the delayed K currents from the maximum K conductance; then, peak currents (at 5–30 ms) were subtracted from the leak current found at the end of the 400 ms trace.
For glycine potentiation experiments, action potentials (i.e., frequency) were counted by using threshold analysis with Clampfit. sEPSCs were collected by a template-based search (based on 20, manually curated, sEPSCs) (Jang et al., 2020; Richardson et al., 2022), with a bottom threshold, followed by automatic detection and vizualization by the software of all selected events. These were then inspected and manually curated, to ensure that only sEPSC were selected (e.g., unclamped action currents as seen in Fig. 8b were not included in the analysis). sEPSCs were fitted and collected before and after glycine treatment.
Conductance (G) was calculated as previously described (Lipinsky et al., 2020). Briefly, using G = I/(V−Vrev) conductance was calculated where I is the measured current amplitude, V is the voltage, and Vrev (reversal potential) was calculated to be −90.41 mV in HEK293t solutions and −115.606 mV in the neurons. GV curves were fit using Boltzmann sigmoidal equation constrained to 1 (GraphPad Prism 8). For the GABA/GIRK experiments, Ibasal was the difference between the steady state of low k+ and high k+ and Ievoked was the difference between the steady state of high k+ with and without 10 µM GABA. For the GluN experiments, the difference was calculated after the trace has stabilized (in its steady state).
Confocal imaging and FRET
All confocal imaging was performed 24–72 h (live imaging) post-cell transfection on a ZEISS laser scanning microscope (LSM) 880/900 using a 20× water immersion objective equipped with 488 and 561 nm lasers. ZEISS LSM 880 was equipped with a META detector. The fluorescence intensity was measured manually with ZEN 3.5. FRET imaging and bleaching were performed on the ZEISS LSM 880. Multiple images before and after bleaching were taken with a 488 nm laser in lambda mode (a 488 nm laser could excite both EGFP and tdTomato). Bleaching was performed with a full power 561 nm laser with a 0.64–0.77 µs/pixel dwell time for 30 s, which led to 90% bleaching of the initial red florescence of tdTomato and had no effect on green fluorescence by itself. We calculated the FRET efficiency (E) by using
AAV virus production
HEK293T cells were seeded on six 100 mm plates in Dulbecco’s modified Eagle’s medium with nutrient mixture F-12 (DMEM-F12) supplemented with 10% fetal bovine serum (FBS) and 1% glutamine. Approximately 24 h after plating, transfection was performed using PEI (1 mg/ml polyethyleneimine in PBSx1) and three plasmids: pAAV Helper, which encodes adenoviral proteins necessary for replication; pAAV Rep-Cap (rAAV2-retro) or pAAV2/1, which encodes the viral replication and capsid proteins; and the genome of interest at a 3:2:5 molar ratio, respectively (∼27 µg DNA/plate). Approximately 6 to 8 h post-transfection, old media was changed to new DMEM-F12 media supplemented only with 1% glutamine. 48 and 72 h post transfection, media containing AAV viruses was collected and filtered through a 0.45 µm filter. Viruses were also collected from lysed HEK293t cells in gradient buffer (described below) and underwent three rounds of freeze/thaw cycles in liquid nitrogen and a 37°C water bath, followed by triturating through an 18G needle 8–10 times. Virus cell lysates were centrifuged at 3,724 rcf and 4°C for 10 min and filtered through a 0.45 µm filter.
For the first round of concentration, a 10× gradient buffer (described in Materials and Methods) was added to the viral media followed by centrifugation using Centric on Plus-70 (UFC703008, Millipore) for ending-up with ∼4–6 ml of viral media (at this stage, the virus could be kept at 4°C overnight or at −80°C for longer storage). To destroy the leftover plasmid-DNA from the transfection, @@@viral media was treated with Benzonase (E1014–25KU 250 units/µl, Sigma, at 1 µl/5 ml viral media, for 1 h at 37°C water bath), followed by a second round of concentration by loading on iodixanol-density gradient media (OptiPrep 60% w/v D1556, Sigma, prepared as described below loading is done from lightest to heaviest gradient solutions in an 11.2 ml tube (Ultra-Clear 344059, Beckman) and centrifuged in a Beckman WX Ultra90 centrifuge at 41,000 rpm and 18°C for 2.5 h. After centrifugation, the virus was collected from the 40% iodixanol layer, washed (at 3,724 × g at 4°C for 3–5 min) three times with PBSx1 using Amicon Ultra 15 ml (UFC910024, Millipore), followed by concentration in storage buffer (described in Materials and Methods) using Amicon Ultra 0.5 ml (UFC510024 - Millipore) and stored at −80°C. Virus titer was tested by qPCR using the forward 5′-GCTGTTGGGCACTGACAAT-3′ and reverse 5′-CCGAAGGGACGTAGCAGAAG-3′ -WPRE primers.
Viral production reagents
Gradient buffer ×10: 100 mM Tris (pH 7.6), 1.5 M NaCl, 100 mM MgCl2, filter sterilized using a 0.22 µm vacuum filter. Storage buffer (for 1 L): 50 g D-sorbitol, 212 mM NaCl - diluted in PBSx1.
Iodixanol (OptiPrep 60% w/v D1556) solutions for gradient preparation:
Cell culture and transfection/infection
HEK293t (and a stable HEK293t-line expressing NaV1.5) were maintained in DMEM with 10% FBS and 1% L-glutamine on 10 cm plates with up to ∼1 × 107 cells per plate. Cells (5 × 103 to 1 × 104) were then transferred to poly-D-lysine-coated (PDL) 12 mm glass coverslips and incubated at 37°C and 5% CO2 for 24 h. Cells were then transfected or co-transfected with 1 µg of each plasmid using ViaFect (Promega) at a ratio of 1:2 or 1:3. In the case of Kv4.2-day transfection with MetaPoda 10:1, Kv4.2GFP was transfected with 1 µg DNA on day 2 post-incubation and left for another 24 h and then transfected with 0.1 µg DNA of MetaPoda. In the dilution experiments 1 µg of Kv4.2GFP DNA was used and the amount of MetaPoda DNA was adjusted accordingly (1:1, 1:10 and 1:100). In the GluN and GABA/GIRK experiments, cells were first plated and transfected 1:3 with ViaFect on a 35 mm dish. In the GluN transfection, each subunit was transfected with 2 µg of DNA and 1 µg of either soluble GFP or MetaPoda. In the GABA/GIRK experiment, each GABA subunit was transfected with 0.5 µg, and each GIRK subunit was transfected with 1 µg, and MetaPoda was transfected with 1 µg. The 35 mm were then washed, and the cells were transferred to coated 12 mm glass coverslips in a 24 well plate as described above for recording. For the GluN experiment, 200 µM APV was added in the 24 well plate for blocking the channels during incubation. Patch-clamp recordings were performed 24–48 h after transfection in all cases and, for 2-day transfection, 24 h post MetaPoda transfection. Dissociated postnatal hippocampal neurons (P0–P2) were prepared and grown from Sprague–Dawley neonate male and female rats (Charles River) at 50–100 K cells/coverslip as described previously (Berlin and Isacoff, 2018). Each neuronal coverslip was then co-infected with 1 µl each of AAV-hSyn-flexed-MetaPoda (1 × 1012 viral genomes/ml) and CAG-Cre recombinase (1 × 1013 viral genomes/ml) at 9 d in vitro (DIV). For GCaMP imaging, neuronal coverslips were infected with 2 µl of AAV-hSyn-GCaMP7b on 9 DIV. Patch-clamp recordings and confocal imaging of spontaneous events were performed on day 16 DIV, that is, 7 d after infection. GCaMP7b events (representing activity of voltage-gated calcium channels) were spontaneous events.
Protein docking simulation
To simulate protein docking between Kv4.2 and soluble HpTx-2 or soluble PaTx-1, we used the HDOCK server (Yan et al., 2020). For the input receptor molecule, we used the human Kv4.2-DPP6S-KChIP1 complex (PDB 7E8H), and for the input ligand molecule, we used a solution structure of HpTx-2 (PDB 1EMX) or the solution structure of PaTx-1 (PDB 1V7F). Out of 100 results, 4 results were found that matched our criterion of having the lowest docking score and lowest ligand rmsd (a lower docking score representing higher certainty) and near the voltage sensor. One of them was selected and is shown. Only Kv4.2 and HpTx-2 or PaTx-1 are shown for a better visualization of the binding site.
Immunofluorescence
MetaPoda infected and uninfected neurons (16 DIV., as described here) were fixed by 4% paraformaldehyde (PFA) in PBS for 15 min at room temperature (RT). After three washes with PBS for 15 min, with gentle shaking at RT, cells were permeabilized with 1% Triton in PBS for 10 min. After three washes, a blocking solution containing 5% FBS in PBS was incubated overnight at 4°C. After three washes, the wash was replaced with the primary rabbit antibody anti-c-Fos (1:1,000, Calbiochem, PC38 or 1:1,000 EnCor, RPCA-c-FOS) in 3% FBS and 0.1% Triton in PBS and incubated overnight at 4°C. After three washes, we incubated the cells with the secondary antibody donkey anti-rabbit IgG Alexa Fluor 488 (1:200, Jackson ImmunoResearch, AB_2313584 for c-Fos or 1:400, Southern Biotech) in the same buffer as the primary antibody for 2 h at RT. Map2 stained neurons were obtained as described above with the primary MAP2 Monoclonal Antibody (AP18) (1:200, Invitrogen, MA5-12826) and the secondary Goat anti mouse 488 (1:500, Jackson ImmunoResearch, AB_2338840).
Neurons were selected manually for cFos analysis based on the brightfield images and fluorescence was quantified based on soma staining. All neurons selected for MetaPoda and tdTomato analyses were confirmed to be infected. Secondary-only staining was subtracted from the results if this staining was significant (>3 AU) and inconsistent between coverslips and treatments.
Cell viability assay
To verify viability of the cells we used CellTiter-Glo Luminescent Cell Viability Assay (Promega). Briefly, 10,000 HEK293 cells (transfected with MetaPoda and untransfected) were plated in opaque 24 well plates and then an equal amount of CellTiter-Glo reagent was to the medium in each well. After 2 min of shaking and another 15 min of room temperature incubation, the luminescent signal was measured on Infinite 200 PRO (TECAN) and the background signal was subtracted from each measurement.
Animal usage
This study was carried out in strict accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Committee on the Ethics of Animal Experiments of the Technion (permit number, IL-130-09-2017, IL-161-10-21).
Statistical analysis
All results are shown as mean ± SEM. Multiple-group comparisons were performed using one-way analysis of variance (ANOVA) with a post hoc Tukey test, two-group comparisons were performed using a two-tailed t test, and pairwise comparisons were performed using a paired t test (GraphPad Prism 8). Slope comparisons were performed using analysis of covariance (ANCOVA), and nonlinear regression comparisons (GV curves and Frequency distribution) were performed using the Kolmogorov-Smirnov test, IV curves’ comparisons were performed using multiple t tests with FDR correction (5%, q values were used instead) or two-way ANOVA for multiple (>2) groups. Significance is indicated by *p < 0.05; **p < 0.01; ***p < 0.001; n.s., nonsignificant. Histograms necessarily represent normally distributed data, and violin plots depict nonnormal distributions. The sample size included at least six cells for electrophysiology experiments and tens to hundreds for imaging experiments (as noted in the legends). HEK293t cells with less than 300 pA peak current in control cells (without MetaPoda) were discarded, as well as cells in which the GV curve showing 50% conductance at −60 or −50 mV. Neurons with input resistance >1,000 MΩ were discarded. Experiments with light activation of MetaLOV with Kv4.2GFP included four cells (each cell was measured 25 consecutive times), though no visible light responses were obtained.
Results
To engineer a Kv4.2-selective genetically-encoded tool, we focused our attention on a peptide toxin derived from the Heteropoda venatoria spider, Heteropodatoxin-2 (HpTx-2) (Sanguinetti et al., 1997). HpTx-2 is a small peptide (30 a.a., see sequence in Materials and Methods), with a unique selectivity and nanomolar affinity toward the Kv4 subunits, notably Kv4.2 and Kv4.3 (Sanguinetti et al., 1997). HpTx-2 is not a pore blocker or competitive antagonist but rather is thought to bind the voltage sensor of the channel and to negatively modify its gating; shifting channel opening by ∼20 mV (Sanguinetti et al., 1997; Bernard et al., 2000; Zarayskiy et al., 2005; DeSimone et al., 2009).
To engineer a membrane-tethered HpTx-2, we cloned the entire sequence of the mature form of the toxin (100 base pairs, bps) downstream of a cleavable signal peptide, followed by a flexible linker, a single transmembrane domain (PDGFR), a red fluorescent protein reporter (tdTomato), along two trafficking motifs for proper delivery of the protein to the membrane, together denoted MetaPoda (Fig. 1a, cartoon). This design enabled strong expression and trafficking of MetaPoda to the membrane of cultured mammalian cells (Human Embryonic Kidney 293t cells; HEK cells in brief) (Fig. 1b, red arrowheads and see micrographs on right with transmitted light micrographs). HEK cells overexpressing MetaPoda appeared normal and expression of MetaPoda did not affect the health of the cells, for instance, by inhibiting other potassium channels in these cells (Zhang et al., 2022) (Fig. 1b, right). To examine the functionality of MetaPoda, we co-expressed MetaPoda and Kv4.2wt at equal DNA amounts and scrutinized the channel’s voltage-gated potassium currents by patch-clamp electrophysiology. Briefly, cells were clamped at −70 mV and the current–voltage relationship (IV curve) was determined by a series of incrementing voltage steps (Δ10 mV) ranging from −100 to +60 mV (Materials and Methods) (Han et al., 2006). Cells solely transfected with Kv4.2wt (and soluble GFP for visualization, Fig. 1c, inset) showed the prototypical outward A-type voltage-dependent currents (Fig. 1c, traces), yielding the expected IV and GV relationships (Fig. 1d, green) (Moise et al., 2010; Liu et al., 2014). However, cells co-transfected with Kv4.2wt and MetaPoda showed no visible voltage-gated currents (Fig. 1c,d; blue). In a small subset of cells expressing the channel (with or without MetaPoda), we observed substantial voltage-sensitive outward currents, as is typically observed in various types of HEK cells (Fig. 1e) (Zhang et al., 2022). These atypical currents did not share the signature features of Kv4.2 [fast activating, <10 ms, and fast inactivating (Bähring et al., 2001)]. To ensure expression of the channel in patched cells, we also examined the effect of MetaPoda over channels tagged by GFP, denoted Kv4.2GFP. Kv4.2GFP expressed robustly in HEK cells and all “green” cells yielded very large current amplitudes (Fig. 1f), on par with those of Kv4.2wt (Fig. 1f, IV curve). Importantly, Kv4.2GFP exhibited identical behavior to Kv4.2wt (Fig. 1f, inset- G/Gmax). We next co-expressed Kv4.2GFP with MetaPoda and find, yet again, no trace of Kv4.2-currrents even though we could easily monitor green fluorescence in these cells (Fig. 1g). We further examined the effect of MetaPoda over Kv4.2GFP by using a well-established recording protocol commonly used to isolate Kv4 currents in neurons (see (Aceto et al., 2020) and below), to bypass any potential artifacts that may stem from endogenous voltage-gated potassium channels in these cells (see Fig. 1e). Briefly, HEK cells underwent two consecutive voltage-step protocols (Fig. 1h, top protocol). The first part consisted of a rapid (100 ms) hyperpolarization step, from −70 to −110 mV, followed by a series of longer (400 ms) depolarization steps reaching 50 mv; revealing all voltage-gated potassium currents in the cell. Then, the same cell underwent an identical protocol except for the initial step which was switched to a rapid depolarization to −10 mV; to expose the remaining delayed potassium currents in the cell. The transient A-type currents are thereby isolated by subtracting currents obtained in protocol 2 from protocol 1 (Materials and Methods). In HEK cells expressing Kv4.2GFP alone, this two-step protocol revealed prototypical (and much cleaner) recordings of Kv4.2 currents (Fig. 1h, top traces, isolated currents in green), IV and GV curves (Fig. 1i,j-comparison with KV4.2wt, dashed light green). Cells co-expressing channel and MetaPoda showed no signs of the A-type currents (Fig. 1h,i; cyan), supporting our above results (see Fig. 1c,d,f,g). Thus, we concluded that MetaPoda completely inhibits all Kv4.2 currents in HEK cells. Of note, the persistence of voltage-gated potassium currents in the presence of MetaPoda demonstrates that MetaPoda is not a broad inhibitor of voltage-gated potassium channels. Indeed, co-expression of MetaPoda with a variety of channels and receptors had no effect over currents recorded in HEK cells (Fig. 2).
MetaPoda Potently Inhibits Kv4.2wt and Kv4.2GFP Currents. a, Cartoon representation of the action of MetaPoda (toxin domain in blue and transmembrane domain in yellow) on the Kv4.2 channel (green). b, Representative microscope images (micrographs) showing membrane expression of MetaPoda (red) highlighted by arrowheads. Nuclei are noted by dashed lines. Smaller micrographs show fluorescence and transmitted light images depicting red fluorescence at the circumference of cells, namely plasma membrane. (right) MetaPoda expression does not affect viability of MetaPoda-expressing cells (+MetaPoda, red bar), compared to nontransfected cells (black). c, Representative micrograph and recordings from HEK293t cells expressing Kv4.2 wt and GFP. Representative traces showing activity from cells expressing Kv4.2 wt alone (n = 18) or with MetaPoda (bottom, n = 10), summarized in (d). Scale bar- 10 µm and patch pipette- dashed white lines. d, Current-voltage relationship and conductance-voltage curves (IV- and GV-curves, respectively) of cells expressing Kv4.2wt alone (green) or with MetaPoda (blue). e, Representative recordings from HEK293t cells expressing Kv4.2wt and MetaPoda (n = 10, top), MetaPoda alone (n = 8, middle) and nontransfected cells (n = 9, bottom). Left traces show silent cells (leak current not affected by voltage), whereas traces on the right show atypical voltage-gated traces (leak affected by voltage). Pie charts depict the fraction of cells affected (or not) by voltage for each group. f, Micrograph showing cells expressing Kv4.2GFP (green) and its current recording. IV- and GV-curves are shown on the right, compared to behavior of Kv4.2wt (dashed green) (n = 29). g, Representative image and recordings from cells co-expressing Kv4.2GFP (green) and MetaPoda (red), and IV curve (bottom) (n = 10). h, Isolating Kv4.2 currents. (top) Two consecutives voltage-jumps are used for isolating the transient A-type current (by subtraction), and examples from cells expressing Kv4.2GFP alone (top trace) and with MetaPoda (bottom). Inset- Isolated currents (dashed rectangle showing the remaining A-type current). Summary of IV- (i) (n = 6) and GV-curves (j) (n = 6) are presented. *p < 0.05; ***p < 0.001; n.s., nonsignificant.
MetaPoda specifically inhibits Kv4 channels. a–d, Representative recordings and summaries of cells expressing: (a) GluN1a and GluN2B; (b) GABAB1, GABAB2 and GIRK1/2 (c) Kv2.1; (d) NaV1.5, co-expressed with (red) or without (black) MetaPoda. NMDARs (GluNRs) were activated by 100 µM glycine and glutamate (grey bar). Recordings of basal GIRK currents (Ibasal) were obtained by application of a high K+ solution, and evoked currents (Ievoked) by additional application of 10 µM GABA (dark and grey bars, respectively). GluNs (control, n = 8; with MetaPoda, n = 7), GABA/GIRK (control, n = 7; with MetaPoda, n = 9), Kv2.1 (control, n = 12; with MetaPoda, n = 10) and NaV1.5 (control, n = 14; with MetaPoda, n = 8). n.s., nonsignificant.
Metapoda is an ultrapotent inhibitor of Kv4.2
We then examined currents in HEK cells that were transfected with Kv4.2GFP a day prior to a subsequent transfection with MetaPoda. Of note, overnight expression of Kv4.2GFP is sufficient for channels to reach the membrane at very large amounts (producing very large currents, e.g., Fig. 1c,f). This procedure yielded three populations of cells, namely green- or red-only cells (expressing Kv4.2GFP or MetaPoda-RFP alone, respectively), and orange cells expressing both (Fig. 3a, micrographs). We patched differently “colored” cells from the same coverslip and noted that, whereas green-only cells yielded the expected Kv4.2 currents (Fig. 3a, top green cells and trace), orange cells were completely silent, as in all previous instances with MetaPoda (Fig. 3b, bottom trace, summary in yellow IV curve). To determine the potency of MetaPoda, we conducted a titration experiment in which we varied the DNA amounts of the channel and MetaPoda. We examined three groups with different channel/MetaPoda DNA ratios of 1:1, 10:1, and 100:1. Under the 1:1 condition, none of the fluorescent cells exhibited channel activity, consistent with our previous findings (Fig. 3c, cyan bar). When the ratio was 10:1, only two out of 10 cells (20%) exhibited Kv4.2-specific currents (Fig. 3d), whereas the highest ratio (100:1) enabled expression of Kv4.2 in a larger fraction of cells (in 5 out of 8 cells, ∼60%) (detailed current descriptions are shown in Fig. 3e,f). These results demonstrate the even when MetaPoda is expressed 100-fold lower than the channel, it still inhibits Kv4.2 currents in a significant proportion of cells (40%), in which we could not detect any Kv4.2-activity. Due to this digital performance (completely inhibiting or enabling current), its potency could not be simply described by a dose-response curve, instead is better described using “fraction of cells” to represent responses (y-axis, analogous to I0/Imax), and channel/MetaPoda ratio to represent dose (x-axis- analogous to concentration). This analysis reveals that the pseudo-IC50 of MetaPoda (i.e., the concentration at which there is a 50% chance of obtaining a current— analogous to 50% inhibition) is achieved at a channel/MetaPoda ratio of ∼50:1 (Fig. 3d).
MetaPoda potently inhibits Kv4.2GFP, even when expressed a day apart or diluted. a, Micrographs of cells that underwent a two-day transfection protocol, in which Kv4.2GFP (green) was transfected a day prior the transfection of MetaPoda (red). Micrographs depict several nearby cells, solely expressing the channel (green only cells) or channel and MetaPoda (yellow cells), and traces recorded from these cells, summarized in bottom IV curve (b) (n = 10 for each group). c–f, Assessing MetaPoda’s potency. c, Peak currents recorded from cells co-expressing Kv4.2GFP and MetaPoda at a 1:1 (teal, n = 6), 10:1 (grey, n = 10) or 100:1 (black, n = 8) DNA-stoichiometry. d, Fraction of cells displaying Kv4.2 currents for each stoichiometry.1:1- 0% (teal), 10:1- 20% (grey) and 100:1- 63% (black). e, IV- and GV-curves for the recorded currents from the various groups (color coded as in c and d), compared to cells expressing Kv4.2GFP alone (dashed plot). Mean current amplitude from the 10:1 group is significantly smaller than other groups, but current behavior (i.e., GV-curves) is identical across all groups. Note the lack of IV- and GV-curves for the 1:1 group (no currents recorded in all cells examined from this group). f, IV curves (and zoom, inset) for all cells from all groups in which we did not detect any Kv4.2GFP-current. *p < 0.05; **p < 0.01; ***, p < 0.001; n.s., nonsignificant.
To assess the specificity of the interaction between toxin and channel, we produced three additional membrane tethered clones in which we either removed the toxin (denoted NoMeta), replaced the toxin with a Kv4-unrelated protein [membrane tethered SNAP (Sun et al., 2011), denoted T-SNAP (Thankarajan et al., 2023)], or disrupted the structure of HpTx-2 by mutating three of its cysteines involved in disulfide bridges (denoted MetaMut) (Fig. 4a). These were co-expressed with the channel, though none had any effect over the current (Fig. 4b,c), aside a small decrease in mean amplitudes when co-expressed with MetaMut. We also produced a fourth clone in which the sequence of the toxin was followed by a light-sensitive domain [LOV domain as is found in Lumitoxins (Schmidt et al., 2014)], thereby denoted MetaLOV, in an attempt to produce a light-gated probe (Fig. 4d). However, this probe had no effect over the current (before, during or after illumination) (Fig. 4e,f). We reason that MetaLOV is not functional (despite its intact toxin domain) due to incorrect positioning or distancing of the toxin from the surface of the membrane, as noted for other Lumitoxins (Schmidt et al., 2014). These demonstrate the basic requirements for the inhibition of Kv4.2, which requires the presence of a correctly folded toxin to be near the channel.
HpTx-2 is necessary for MetaPoda’s inhibition of Kv4.2. a, Cartoon representation of NoMeta (yellow, n = 8), T-SNAP (pink, n = 8) and MetaMut (green, n = 8) co-expressed with Kv4.2GFP, and corresponding recordings and micrographs below in (b). c, Summary of IV and GV-curves. d, Cartoon representation of MetaLOV composed of TM (yellow), LOV domain (cyan), linker (black noodle), toxin (blue and purple, supposedly before and after light illumination and activation of LOV domain), Kv4.2 (green). Light illumination is hypothesized to bring the toxin closer to the channel to exert its inhibition (red dashed line). e, (top) Representative initial recording from a cell co-expressing MetaLOV and Kv4.2GFP. (bottom) A series of traces taken from the same cell before (dark), during (blue, 460 nm illumination) and after (black) illumination has ceased (n = 4), summarized in (f). The first recordings was taken after the cell has been kept in the dark for several minutes. Note the complete lack of inhibition by MetaLOV under all conditions. g, Micrographs showing cells co-expressing Kv4.2GFP (green) and MetaPoda (red) obtained by the two-day transfection protocol (Materials and Methods) before (left) and after (middle) photobleaching (by 100% 561 nm laser) of RFP at the membranes of cells (dashed yellow circle), summarized on the right. h, Representative micrographs of cells expressing tdTomato-only before (left) and after (middle) photobleaching at 561 nm. Note the increase in green fluorescence. (right) Emission spectra collected from cells expressing Kv4.2GFP-only (reference, green trace), Kv4.2GFP + MetaPoda (cyan and gray-blue traces), or tdTomato-alone (orange and beige traces), before and after photobleaching, respectively. Photobleaching of tdTomato gives rise to a GFP-like spectrum (after, from orange to beige trace around 500 nm, dashed arrows). i, Bona fide FRET between KV4.2GFP and MetaPoda. The regression plot shows that photobleaching of MetaPoda-tdTomato (cyan bullets) yields significantly higher GFP fluorescence than is obtained by photoconversion of tdTomato (orange bullets) under identical expression levels and imaging settings. j, No FRET is observed between Kv4.2GFP with NoMeta (pink) or T-Snap (yellow) to tdTomato, summarized in (k); n = 58 for every variant. *p < 0.05; ***p < 0.001; n.s., nonsignificant.
To further examine the direct interaction between MetaPoda-RFP and Kv4.2GFP at the membrane, we reverted to examine the Förster Resonance Energy Transfer (FRET) by acceptor photobleaching (Ruiz-Velasco and Ikeda, 2001; Cabedo et al., 2004). Briefly, we analyzed the emission spectra of donor (Kv4.2GFP) and acceptor (MetaPoda-RFP) molecules, before and after photobleaching (Materials and Methods). We performed photobleaching of MetaPoda located on the plasma membrane of cells by intense and rapid bouts of 561 nm illumination (Fig. 4g, lime arrow and dashed circle). This procedure yielded reduction in RFP fluorescence with the concomitant increase in green fluorescence (Fig. 4g, after and plot). We performed this procedure over a wide range of donor/acceptor expression levels and compared increases in GFP fluorescence to that of control cells expressing the RFP alone (i.e., tdTomato) (Fig. 4h, micrographs). Of note, we performed this additional analysis after having observed that tdTomato undergoes weak, albeit real, photoconversion (Fig. 4h, after and emission spectra on the right- orange and beige traces). Interestingly, photoconversion of tdTomato has not been previously described, but photoconversion of fluorescent proteins is a well-established phenomenon (Shaner et al., 2004, 2007; Miyawaki et al., 2012; Hussein and Berlin, 2020; Heinrich et al., 2021). We therefore did not follow this unique observation, rather ensured correction of the signal (see Materials and Methods). Importantly, we find that the extent of photoconversion of tdTomato does not overlap with the increases observed when KV4.2GFP and MetaPoda are co-expressed (Fig. 4i), whereas it matches the increases obtained in negative control groups (t-SNAP-tdTomato or NoMeta-tdTomato with Kv4.2GFP) (Fig. 4j). Thus, after correction for photoconversion, we find that the adjusted FRET efficiency between MetaPoda and Kv4.2GFP is significantly higher compared to control groups (Fig. 4k). These experiments support our above results by demonstrating a bona fide FRET signal between MetaPoda and Kv4.2 at the membrane.
Metapoda binds the voltage sensor of Kv4.2
The native (soluble) toxin is suggested to interact with the voltage sensor of Kv4.2 via two different sites: the external S3-S4 linker (Zarayskiy et al., 2005) and/or the S3b domain (DeSimone et al., 2009). To examine which of the sites is essential for the effect of MetaPoda, we engineered a chimeric channel in which we replaced the S3-S4 linker in Kv4.2GFP with a sequence corresponding to a HpTx-2-insensitive channel (Kv1.4)(Zarayskiy et al., 2005), denoted Kv4.2GFPS3S4 (Fig. 5a, green sequence replaced by pink sequence from Kv1.4, and b). Kv4.2GFPS3S4 expressed well in HEK cells and yielded the typical IV-relationship, albeit provided slightly lower current amplitudes (Fig. 5c, left trace and d, dark and light green plots). Unexpectedly, co-transfection with MetaPoda resulted in complete inhibition of Kv4.2GFPS3S4’s currents (Fig. 5, right trace and summary in 5d, purple). We produced another channel in which we mutated two residues in its S3b domain (L277A and V278A) (DeSimone et al., 2009), denoted Kv4.2GFPAA (Fig. 5a, dark yellow bar and residues, highlighted in b). Kv4.2GFPAA also expressed well in cells (assessed by bright green fluorescence), although this channel produced significantly smaller currents (Fig. 5e,f) with slightly disrupted behavior (Fig. 5g, yellow). This was somewhat expected owing to the location of the mutations within the voltage sensing domain (VSD) of the channel. Interestingly, MetaPoda did not inhibit the currents of Kv4.2GFPAA (Fig. 5e, compare blue and yellow traces), but it did modify its gating, shifting the IV curve by ∼10 mV (Fig. 5f,g). Together, MetaPoda’s inhibition of Kv4.2 essentially depends on these two residues within the S3b domain. These results further demonstrate that MetaPoda is a gating modifier, much like the soluble toxin, but much more potently (no currents are observed even past 70 mV, see below).
MetaPoda’s inhibition of Kv4.2GFP depends on two residues (L277A/V278A) in the S3b domain of the channel, but not on the S3-S4 linker. a, Sequence alignment between Kv4.2 and Kv1.4. Residues L277 and V278 are highlighted in yellow, and these residues were mutated to alanine residues (mutant channel denoted Kv4.2GFPAA). The S3-S4 linker is highlighted in green. This linker was replaced with that of Kv1.4 (highlighted in pink), creating a chimeric channel denoted Kv4.2GFPS3S4. b, Crystal structure of Kv4.2 [PDB: 7E8H (Kise et al., 2021)] with the two residues and linker color coded as in (a). c, Representative recordings and micrographs of HEK293t cells co-expressing Kv4.2GFPS3S4 alone (green), or Kv4.2GFPS3S4 with MetaPoda (purple), summarized in (d). Kv4.2GFPS3S4 (n = 8) and Kv4.2GFPS3S4 with MetaPoda (n = 11). e, Representative recordings and micrographs of cells co-expressing Kv4.2GFPAA (yellow) or Kv4.2GFPAA with MetaPoda (blue), summarized in (f). Kv4.2GFPAA alone (n = 6) and Kv4.2GFPAA with MetaPoda (n = 9). g, Conductance relationships of Kv4.2GFP, Kv4.2GFPS3S4, Kv4.2GFPAA and Kv4.2GFPAA with MetaPoda. h, Simulation of the interaction between HpTx-2 and Kv4.2. Shown is the highest scoring docking simulation. Residues and linker of the channel are highlighted as in (a), soluble HpTx-2 in blue. *p < 0.05; **p < 0.01; ***p < 0.001; n.s., nonsignificant.
The persistent inhibition of Kv4.2GFPS3S4 by MetaPoda contrasts a previous report that examined this linker with the soluble toxin (Zarayskiy et al., 2005). Our results suggest that this linker may not be the main site of interaction of the toxin and voltage sensor (DeSimone et al., 2011). To try to reconcile between these observations, we simulated the docking of the toxin to the channel. Notably, this was not previously possible as the structures of Kv4.2 have only recently been elucidated (Kise et al., 2021; Ye et al., 2022). We employed a web-based peptide docking approach (see Materials and Methods) (Yan et al., 2020). Briefly, we obtained 100 models, however limited our scrutiny of models with the highest prediction scores, and to models that necessarily placed the toxin near the voltage sensor. This filtering yielded four highly comparable models, in which the toxin was unbiasedly docked adjacent to the S3b site and the S3–S4 linker (Fig. 5h, yellow and green, respectively). Unexpectedly, the models suggest that the toxin binds the voltage sensor from within the membrane (Fig. 5h, inset), as opposed to capping of the domain from the extracellular side, as suggested for the sister channel, Kv4.3 (DeSimone et al., 2009). Membrane partitioning is considered an obligatory feature for many other spider toxins that bind VSDs of other channels [e.g., (Lee and MacKinnon, 2004; Milescu et al., 2007)] and therefore supports the predictions obtained. These help reconcile between the different reports by suggesting that the toxin may initially require the protruding S3–S4 linker for initial recognition of the site, but its effect on gating is obtained by its subsequent (and likely higher affinity) interaction with the S3b site from within the membrane. The immobilization of the toxin next to the membrane increases the effective concentration of the toxin and may thereby overcome the strong decrease in affinity between the modified linker and toxin (Fig. 5e,f).
Expression of MetaPoda in cultured hippocampal neurons
To examine the effect of MetaPoda on endogenous neuronal Kv4.2 channels, we expressed MetaPoda in cultured hippocampal neurons (via AAV infection; Materials and Methods) (Berlin and Isacoff, 2018). Neurons were functionally examined following seven days of expression in vitro, and MetaPoda-positive neurons (i.e., neurons in which we could detect red fluorescence, demonstrating expression of MetaPoda) were compared side by side to naïve (noninfected) neurons from the same culture. MetaPoda showed robust expression in dendrites and axons of neurons (Fig. 6a). We then recorded A-type currents measurements (see description above, and Materials and Methods) (Fig. 6b, top protocols) (Aceto et al., 2020). MetaPoda-expressing neurons (+MetaPoda) exhibited much smaller A-type current amplitudes (Fig. 6b, red traces, summary in 6c) and the remaining currents (occupying 41.6% of the total Kv-currents, Fig. 6c, top plot, red bar) displayed a significantly different behavior compared to those of naïve neurons (Fig. 6c, bottom GV-curve), likely contributed by different Kv-channels. Moreover, the inhibited fraction of the current (58.4%) coincides with the fact that Kv4 (mostly Kv4.2) channels occupy ∼60% of the total A-type currents in pyramidal neurons (Aceto et al., 2020). Thus, our results imply that MetaPoda completely abolished all endogenous Kv4 currents in neurons, likely all Kv4.2 (Kv4.3 is predominantly expressed in inhibitory neurons). These results further demonstrate the specificity of MetaPoda by demonstrating that it does not inhibit all Kv-channels expressed in pyramidal neurons, for instance subtypes of Kv1 and Kv3 (also see Fig. 2 and below).
Expression of MetaPoda in cultured hippocampal neurons instigates minor effects on Intrinsic Firing Properties of neurons. a, MetaPoda is expressed in both axons and dendrites of hippocampal neurons. Micrographs of cultured hippocampal neurons expressing MetaPoda (+MetaPoda, red) and immunostained for MAP2 (green). High power micrographs of a single neuron (right), showing overlapping (yellow arrowheads) and nonoverlapping (red arrowheads) expression of MAP2 and MetaPoda. b, Isolation of Kv4 currents. (top) Voltage clamp recording protocol. (bottom) Representative currents showing A-type currents in naïve (black) and MetaPoda-expressing (red) neurons. Isolated (subtracted) currents are shown on the right (dashed region showing remaining A-type Kv4 currents), summarized in (c), IV- (top) and GV-curves (bottom) for naïve (black) and +MetaPoda neurons (red). Naïve neurons- n = 8; +MetaPoda-neurons- n = 9. Histogram depicts proportion of A-type currents from total Kv-currents. d, (left) Representative current-clamp recordings of naïve- (black) and +MetaPoda-neurons (red) under I = 0 and (right) current-steps for establishing intrinsic excitability, summarized in (e). Summary of resting membrane potential (RMP) and firing threshold are shown on the right. f, g, Summaries of action potential amplitudes, after hyperpolarization (AHP) amplitudes and action potential (AP) half width are shown. Representative APs are shown above plots. h, Representative firing instigated by a current step of naïve (black) and +MetaPoda (red) neurons showing attenuation of trains of APs, summarized on the right. In naïve neurons, attenuation is minor (95.27 ± 2.6%, n = 7), whereas +MetaPoda-neurons showing pronounced attenuation (76.2 ± 4.78%, n = 10). *p < 0.05; **p < 0.01; ***p < 0.001; n.s., nonsignificant.
We next examined the functional outcome of the transient expression of MetaPoda (7 d after infection with MetaPoda) in neurons. MetaPoda-positive neurons did not exhibit significant differences in spontaneous activity, intrinsic excitability, resting membrane potential (RMP), firing threshold and action potential (AP) amplitudes compared to naïve neurons (Fig. 6d–g) (Balena et al., 2008; Segal, 2018). However, we did find changes in action potential attenuation during firing (adaptation) (Fig. 6h), as well as in other features of the AP itself and input resistance (Fig. 7a). Lastly, adaptation was complemented by a small (albeit nonsignificant) tendency in the increase of fast-acting sodium (but not potassium) currents (Fig. 7b). This tendency however did not result from any direct effect of the expression of MetaPoda over endogenous voltage-gated sodium channels in neurons (Fig. 7c, top), supporting the results obtained for Nav1.5 (see Fig. 2d), nor over the activity of calcium channels (assessed by imaging the Ca2+-reporter GCaMP7b) (Fig. 7c, bottom traces and summary). Together, some of the phenotypes observed following the expression of MetaPoda are to be expected from a negative modulator of potassium channels in neurons (Cameron et al., 2000; Feria Pliego and Pedroarena, 2020), however the collective changes diverge from observations obtained from Kv4.2-knockout (KO) animals (see Discussion and Summary in Table 2). These may be attributed to differences between transient expression of MetaPoda (7 d) compared to chronic genetic elimination of Kv4.2 in pyramidal neurons (Chen et al., 2006; Andrásfalvy et al., 2008; Nerbonne et al., 2008; Carrasquillo et al., 2012).
Effects of MetaPoda on endogenous currents and features of the action potentials. a, Representative action potentials recorded from naïve (black) and +MetaPoda-neurons (red), and summarizes of (from left to right): Time to peak, maximal rising slope, time from AP-maxima to AHP minima, maximal decay slope and time from firing threshold to max decay. Input resistance and phase plots are shown on the right. b, Fast sodium (top) and potassium (bottom) currents recorded from control (black) and +MetaPoda-neurons (red). c, MetaPoda has no effect over endogenous voltage gated sodium (Nav) and calcium channels (Cavs). (top) Representative recordings of native sodium currents in naïve (black, n = 10) and +MetaPoda-neurons (red, n = 11), summarized on the right (IV curves). (bottom) Representative fluorescent traces by imaging of GCaMP7b in naïve (black, n = 14) vs +MetaPoda-neurons (black, n = 27), showing no effect of MetaPoda of firing frequencies, summarized on the right. ***p < 0.001; n.s., nonsignificant.
Comparison of the effects observed by the expression of MetaPoda in cultured hippocampal neurons to effects observed in Kv4.2-KO animal models
Metapoda infected neurons are resistant to potentiation
To assess the functional outcome of MetaPoda expression in neurons, we assessed potentiation of hippocampal cultures via chemically-induced LTP (cLTP; Materials and Methods) (Lu et al., 2001; Kim et al., 2007; Cheyne and Montgomery, 2008; Molnár, 2011). We patched hippocampal neurons (7 d after infection and expression of MetaPoda) and recorded basal activity under voltage- and current-clamp for 10 min in a physiological solution. Then, neurons were potentiated for 5 min by a Mg2+-free and high glycine solution (100 µM), after which we reverted to recording neuronal activity under physiological conditions for 10 additional minutes during which peak potentiation is typically observed (Fig. 8a). This protocol induced potentiation of naïve neurons, portrayed by significant increases in the amplitudes of spontaneous excitatory postsynaptic currents (sEPSCs), as well as in firing frequency (+30%) (Fig. 8b–d, black and grey). In contrast, MetaPoda-positive neurons were highly resistant to potentiation, displaying no significant increases in sEPSC amplitude or firing frequency (Fig. 8b–d, red and pink). In fact, MetaPoda-positive neurons displayed larger sEPSC amplitudes prior to potentiation (peak sEPSC amplitude in naïve neurons: 2194 pA; MetaPoda-positive neurons: 3222 pA) (Fig. 8c, zoomed inset in yellow, black and red). These results show that prolonged (7 d) and potent inhibition of Kv4.2 in excitatory neurons by MetaPoda prevents neurons from undergoing potentiation.
MetaPoda+ Neurons are resistant to potentiation. a, Depiction of the chemical LTP (cLTP) protocol. Baseline recordings were acquired for 10 min (five minutes in each mode, current and voltage clamp, respectively) under physiological conditions (pre-potentiation, black bar). Then, potentiation was induced by a high glycine-solution (100 µm and see Materials and Methods) (dark grey bar) for five minutes, followed by 10 min voltage- and current-clamp recordings, respectively, under physiological conditions (post-potentiation, light grey bar). b, Representative recordings of naïve (top, black trace) and +MetaPoda-neurons (bottom, red). c, Raw and cumulative distributions of spontaneous EPSCs (sEPSC) frequencies from naïve (top) and +MetaPoda-neurons (bottom) before and after glycine application (dark and bright colors, respectively). Inset- zoom-in (dashed region; up to 2% frequencies, y-axis) for better scrutiny of the distributions of EPSCs. Yellow shaded area in the insets highlights the largest differences in sEPSC amplitude between naïve (n = 12) and +MetaPoda-neurons (n = 13) pre- and post-glycine application. Significance of cumulative distributions were tested using the Kolmogorov–Smirnov test. d, Mean sEPSC amplitudes (top), pre- and post-glycine application (dark and light colors, respectively). Significance was examined using paired T-test. (bottom) Summary of firing frequencies pre- and post-glycine application for naïve (black and gray, respectively) and +MetaPoda-neurons (red and pink, respectively). Analysis of *p < 0.05; **p < 0.01; n.s., nonsignificant.
Cfos immunostaining of MetaPoda infected neurons
Potentiation of neurons is tightly associated with elevation in the immediate early gene cFos (Sheng and Greenberg, 1990; Zhang et al., 2002; Jaworski et al., 2018). Our results show that neurons expressing MetaPoda exhibit an increase in EPSCs amplitude and resistance to undergo further potentiation, portraying a post potentiated state. We therefore decided to explore the protein levels of cFos following MetaPoda expression. We first confirmed the specificity of the cFos antibody by overexpressing cFos in HEK cells (Fig. 9a), as well as confirmed the elevation of nuclear cFos in naïve neurons following application of the cLTP protocol; increasing cFos levels ∼1.5-fold (Fig. 9b,c). We then proceeded to evaluate the effect of expression of MetaPoda on cFos levels, in which case MetaPoda-positive neurons (examined 7 d after infection) showed specific increases in the nuclear expression of cFos, compared to naïve or tdTomato-expressing neurons (the same RFP as found in MetaPoda) (Fig. 9d,e). In fact, overexpression of the red fluorescent protein actually reduced fluorescence levels of cFos, likely via FRET (see Fig. 4g). Together, levels of cFos are significantly increased whether by the cLTP protocol or by expression of MetaPoda.
Expression of MetaPoda increases cFos expression levels in hippocampal neurons. a, Representative micrographs showing nuclear immunostaining for cFos (green) in nontransfected (left) and cFos-transfected HEK293t cells (right). Middle micrograph shows lack of nonspecific staining by the secondary antibody. b, cLTP protocol increases expression of cFos in hippocampal neurons. Representative micrographs showing nuclear cFos-immunostaining (green) of naïve hippocampal neurons before (untreated) and after potentiation, summarized in (c). Potentiated neurons exhibit 1.49-fold increase in cFos expression. d, MetaPoda leads to expression of cFos. Representative micrographs showing nuclear cFos-immunostaining (green) of naïve (left) and +MetaPoda-neurons (red, right). Note the expression of cFos particularly in regions where MetaPoda is apparent (red regions). e, Summary of extent of expression of cFos in naïve neurons (black bar) compared to neurons expressing tdTomato (grey bar) or MetaPoda (red bar). +MetaPoda-neurons exhibit a 2-fold increase in cFos expression. ***p < 0.001.
Transferability of the engineering strategy toward another Kv4.2-selective toxin
We were curious whether our strategy (and clone) would also be suitable for designing other Kv4.2-selective tethered toxins. We therefore engineered a membrane tethered Phrixotoxin-1 (PaTx-1), denoted MetaPhix (Fig. 10a). PaTx-1 is also a spider toxin (from Phrixotrichus auratus) and a gating modifier like HpTx-2, with high affinity and selectivity for the channel (IC50- 5 nM) (Chagot et al., 2004). It shows very minimal sequence and structural homology with HpTx-2 (Fig. 10b); however, our docking simulations place it at a very similar location as MetaPoda (Fig. 10b, red). We designed a single clone by simply replacing the sequence of HpTx-2 with that of PaTx-1 and tested it in HEK cells. We find that this strategy proved highly successful and yielded a strong negative modulator (>90%) of Kv4.2GFP-currents, though not as strong as MetaPoda (residual Kv4.2 currents can be discerned) (Fig. 10c). The inhibitory activity of MetaPhix also proved dependent on the two residues found in the S3b domain (L277 and V278), attested by the lack of inhibition of the channel when these were mutated to alanines (in Kv4.2GFPAA) (Fig. 10d), suggesting that PaTx-1 is a gating-modifier, akin to HpTx-2 (also strengthening the docking simulations). Thus, our strategy proved suitable for the rapid engineering of an additional tethered toxin.
MetaPhix, a Potent tethered toxin for Kv4 Based on Phrixotoxin-1. a, Cartoon representation of the action of MetaPhix (toxin is depicted in red and transmembrane domain in yellow), Kv4.2 channel (green). b, Sequence and structure alignments between HpTx-2 (PDB 1EMX, blue) and Phrixotoxin-1 (Patx-1, PDB 1V7F). Cysteines are highlighted. Optimal docking simulation comparing docking of soluble HpTx-2 (as shown in Fig. 5h, blue) and PaTx-1 (right, red) onto Kv4.2. c, Representative recordings of HEK293t cells expressing Kv4.2GFP alone (left), Kv4.2GFP (n = 5) with MetaPhix (right) (n = 7), summarized in IV curve on the right. For comparison, we included Kv4.2GFP with or without MetaPoda from Figure 1 (dashed green and cyan plots). MetaPhix inhibits 91% of the Kv4.2 current (at +60 mv). d, Representative recordings of cells co-expressing Kv4.2GFPAA alone (left, n = 6) and with MetaPhix (middle, n = 6). Summarizes of IV- and GV-curves are shown on the right. Reference plots are denoted by dashed lines. *p < 0.05; **p < 0.01; n.s., nonsignificant.
Discussion
We have engineered two genetically encoded tethered toxins to target Kv4, denoted MetaPoda and MetaPhix (Figs. 1, 10). We attempt this despite the presence of multiple cysteine residues (here three pairs, Materials and Methods) within the sequence of the toxins, owing to numerous successes in developing other genetically encoded toxins (Choi and Nitabach, 2013), particularly spider toxins [e.g., (Wu et al., 2008; Auer et al., 2010; Zhao et al., 2022)]. We present multiple lines of evidence that show that MetaPoda is a functional, ultrapotent and selective inhibitor (i.e., gating modifier) of Kv4.2 channels (Figs. 1–3). Potent inhibition is maintained even when the channel is expressed 1 d prior MetaPoda (Fig. 3a), or when MetaPoda is expressed at 100-fold lower DNA quantities (Fig. 3c–f). Furthermore, disruption of the cysteine residues by mutagenesis (in MetaPoda) renders the clone no longer active, as well as when the toxin sequence is completely removed (NoMeta) or switched to another protein domain (T-SNAP) (Fig. 4). Lastly, when the toxin is greatly distanced from the membrane (by LOV domain), it loses its effects (Fig. 4d–f). Together, these demonstrate that the presence of the sequence of the toxin, its cysteines (i.e., correct folding) and its close proximity to the channel are all necessary for the bona fide and potent inhibitory mode of action of MetaPoda.
We next explored the mechanism of action of the probe over Kv4.2. We show that, whereas the native soluble toxin shifts the activation curve by ∼20 mV [and is thereby a gating modifier, (Bernard et al., 2000; Zarayskiy et al., 2005; DeSimone et al., 2009)], no currents (and thus no shift) are observed when MetaPoda is co-expressed— even 60 mV past the activation threshold of the channel (Fig. 1g). We do not think these represent another mode of action by the toxin; rather that the toxin is at a very high local concentration adjacent to the channel (see FRET analysis, Fig. 4g–k). In support, MetaPoda is highly sensitive to disruption of two consecutive residues in the S3b domain, akin to the soluble toxin [L277 and V278 (DeSimone et al., 2009, 2011)] (Fig. 5a,e,f). However, and in contrast to the soluble toxin, MetaPoda’s activity is completely insensitive to changes in the external S3–S4 linker (Fig. 5c,d) (Zarayskiy et al., 2005). These lead us to suggest that MetaPoda’s high local concentration at the membrane enables it to reach its main binding site of activity (Fig. 5h).
Despite the latter assumptions, the exact binding site of the soluble toxin is unknown. We therefore modeled the interaction between the toxin domain and the recently published structures of Kv4.2 [e.g., (Ye et al., 2022)] and the models suggest that the toxin does not cap the VSD, rather partitions within a membrane pocket adjacent to both the S3b and S3–S4 domains (Fig. 5h). Thus, while the reduction in affinity of the S3–S4 linker, by its replacement with that of Kv1.4, is too significant for the efficient binding of the soluble toxin, it may be overcome by the high concentration of MetaPoda at the near-membrane space (Schmidt et al., 2014). Our hypothesis is in line with previous reports suggesting that the mechanism of action of spider toxins (e.g., HaTx and ProTx2) requires the initial recognition step to partition into the plasma membrane, but the gating mechanism itself is obtained by the subsequent and direct interaction of the toxin with the VSD (Milescu et al., 2007; Männikkö et al., 2018; Xu et al., 2019). The high local concentration and the lack of dependence on the S3-S4 linker result in the excessive potency of MetaPoda (i.e., ultrapotent) compared to the soluble toxin (∼20 mV shift). Together, our results demonstrate that MetaPoda remains a gating-modifier of Kv4.2, albeit ultrapotently, and it does so through interaction with the VSD (Fig. 5).
Kv4.2 limits neuronal potentiation
In neurons, MetaPoda completely abolished endogenous Kv4 currents, likely Kv4.2 since we only recorded from excitatory pyramidal neurons (Fig. 6b). This finding demonstrates that MetaPoda folds correctly in neurons and that the toxin domain can reach its site of action (i.e., the VSD), even when the channel is expected to be complexed with added proteins, such as DPP6 (Kise et al., 2021).
MetaPoda exerts several effects on features of the AP, such as slowing down of repolarization and increasing input resistance (Fig. 7), as well as enhancing attenuation of firing (Fig. 6h). These are expected outcomes from inhibitors of potassium channels (Cameron et al., 2000; Feria Pliego and Pedroarena, 2020), and thereby suggest a potential role for Kv4.2 in shaping the AP and regulating other voltage-gated currents (for instance, accelerating recovery from inactivation of NaV channels), in strong support of previous reports using Kv4.2-KO mice [Table 2 and (Chen et al., 2006; Andrásfalvy et al., 2008; Nerbonne et al., 2008; Carrasquillo et al., 2012; Granados-Fuentes et al., 2012)]. Nevertheless, unlike previous reports using genetic models, we observed very subtle changes in neuronal physiology following MetaPoda expression, showing that very minor compensation mechanisms may have ensued following expression of MetaPoda (for 1 week), such as lack of changes in the expression of other potassium or sodium currents (Fig. 7b,c).
The most prominent effect observed following expression of MetaPoda was prevention of potentiation by cLTP (Fig. 8), and this finding was supported by elevations in the protein levels of cFos (Fig. 9a–e). cFos is a proxy for neuronal activation (Fleischmann et al., 2003; Korb and Finkbeiner, 2011; Cruz et al., 2013; Jaworski et al., 2018). More specifically, cFos undergoes strong upregulation following intense neural activity, as can be obtained during processes of plasticity (Cruz et al., 2013), but also following neuronal hyperexcitability (Morgan et al., 1987). We ruled-out that the expression of MetaPoda instigated hyperexcitability (Fig. 6e–g), leading us to suggest a role for Kv4.2 in plasticity instead (Kim et al., 2007; Kim and Hoffman, 2008; Yunoki et al., 2014; Sandler et al., 2016; Oulé et al., 2021; Zhao et al., 2023).
The means by which Kv4.2 supports plasticity is poorly understood, counterintuitive and at times opposite. For instance, transient inhibition of Kv4.2 (e.g., by soluble drugs) promotes better propagation of back action potential (bAP) to dendrites and thereby enhances potentiation of distant synapses (Chen et al., 2006), while others show that transient inhibition of Kv4.2 prevents short term plasticity (lasting ∼1 h) of layer 5 pyramidal neurons (Sandler et al., 2016). The latter is in-line with our observations whereby the inhibition of the channel—specifically over the course of 7 d—results in the neuron’s inability to undergo further potentiation (Fig. 8c). We postulated that these opposing observations could be reconciled by the fact that the persistent inhibition of the channel may instigate progressive potentiation, under which conditions no further potentiation can be obtained. This notion is supported by recent reports that show that insufficient amounts of Kv4.2 at the membrane (prompted by knockout of Cav2.3) lead to enhanced synaptic inputs (neuronal excitability was not examined) (Murphy et al., 2022) and the persistence of enhanced inputs leads to saturation of potentiation (Lang et al., 2004; Navakkode et al., 2017; Zylberberg and Strowbridge, 2017). Notably, these observations diverge from observations made using Kv4.2-KO mice, in which neurons chronically lacking Kv4.2 (throughout the entire course of development to adulthood), are more easily potentiated (Chen et al., 2006). The differences between these observations may stem from differences in neuron types and potentiation protocols, but perhaps the most striking difference between these is the duration of inhibition of Kv4.2 (soluble HpTx-2- minutes, MetaPoda- days, KO-mice- weeks/months) (Frick et al., 2004; Kim et al., 2007; Sandler et al., 2016). An emerging explanation for the facilitation of potentiation by Kv4.2 (which would be abolished during inhibition of channel) is that Kv4.2 might associate with added proteins at the synapse so that, when persistently inhibited (and perhaps even internalized) leads to disruption of plasticity. Indeed, potassium channels, Kv4 in particular, are prototypical macromolecular signaling hubs—signalosomes— assembling multiple proteins, notably plasticity related proteins such as AMPARs and PKA, to name a few (Kim et al., 2007; El-Haou et al., 2009; Berlin et al., 2010; Norris et al., 2010; Sandler et al., 2016; Yao et al., 2016; Bissen et al., 2019). It should therefore be interesting to explore the added effects of the expression of MetaPoda in different cell types and on different Kv4.2-molecular partners.
In summary, we present MetaPoda, an ultrapotent genetically encoded gating-modifier of Kv4.2. Our systematic characterization leads us to suggest a link between inhibition of Kv4.2, upregulation of cFos and subsequent prevention of potentiation (at least by cLTP) of excitatory hippocampal neurons. Collectively, our results complement previous observations and propose novel roles for Kv4.2 in gating potentiation (i.e, plasticity).
Footnotes
We thank Prof. Jackie Schiller for helpful discussions and comments on the manuscript. The research submitted is in partial fulfillment for a doctoral degree for MA. This work is supported by Templeton World Charity Foundation (TWCF) (62056).
The authors declare no competing financial interests.
- Correspondence should be addressed to Shai Berlin at shai.berlin{at}technion.ac.il.
