Assessing the utility of Magneto to control neuronal excitability in the somatosensory cortex

Article metrics

arising from Wheeler et al. Nature Neuroscience https://doi.org/10.1038/nn.4265 (2016)

Magnetic neuromodulation has outstanding promise for the development of novel neural interfaces without direct physical intervention in the brain. Here we tested the utility of Magneto in the adult somatosensory cortex by performing whole-cell intracellular recordings in vitro and extracellular recordings in freely moving mice. The results show that magnetic stimulation does not alter subthreshold membrane excitability or contribute to the generation of action potentials in virally transduced neurons expressing Magneto.

Recently, Magneto1 has been suggested to provide the highly sought after neuromagnetic actuation in a cell-targeted manner. Some of the excitement about Magneto originates from its design, which comprises a calcium-permeable non-selective cation channel (transient receptor potential cation channel subfamily V member 4 (TRPV4)) fused to the paramagnetic protein ferritin1. This single-construct approach provides a simplified means for magnetic intervention in neuronal activity. Here we used lentiviral delivery of Magneto linked to mCherry (Magneto2.0-P2A-mCherry), expressed under the control of the ubiquitin promoter for more than 2 weeks (Fig. 1a) before observing and interfering with neural activity (see Supplementary Methods), and, after confirming successful cleavage of Magneto from mCherry (Supplementary Figs. 2 and 3), we did subcellular analysis of the localization of the expressed protein (Supplementary Fig. 2b) in a neuronal cell line. Chronic extracellular recordings in freely moving mice2,3,4 with 15 tetrodes enabled high-density sampling of neural activity in the vicinity of transduced cells, and yielded well-isolated (Supplementary Fig. 4), stable units (Fig.1b). Comparison of firing rates within cells across magnetic stimulus conditions (off versus on) showed that magnetic stimulation does not alter the rate of action potentials (Fig.1c), modulate the interspike interval within cells, or modulate spike timing across single units recorded from the same tetrode (Supplementary Figs. 5 and 6). The lack of spiking was not because neurons could not respond to synaptic depolarization; deflection of magnetized whiskers (with nano iron particles, see Supplementary Methods) using an electromagnet4 induced stimulus-coupled spiking (Fig.1d). Considering that direct neuromagnetic stimulation failed to trigger any change in low-frequency local field potential oscillations (Fig.1e,f), we argue that belief in the utility of Magneto to control neural activity in vivo is not warranted.

Fig. 1: Electrical characterization of the consequences of magnetic neural stimulation.
figure1

Neurons in the primary somatosensory cortex (barrel field) were transduced using a lentiviral vector encoding Magneto-p2A-mCherry for at least 2 weeks before extracellular recordings with chronically implanted distributed microelectrode arrays, whole-cell current and voltage-clamp experiments were performed. Magnetic stimulation was provided with permanent magnets with an estimated magnetic field intensity >50 mT in the vicinity of the cells recorded (see Supplementary Fig. 1 for magnetic field strength measurements; see Supplementary Methods for magnet placement). a, Example of viral transduction throughout a cortical column of interest. Right: a higher magnification view from the cortical layer 2/3. L1, cortical layer 1; CC, corpus callosum. b, Rate of action potentials (bottom) and unit stability (top), as quantified by peak-to-peak (P2P) amplitude, during a 20 min-long recording session for a single-unit recording before and during magnetic stimulation in freely behaving mice. c, Firing rate of 353 single units recorded across five sessions in two mice. The two axes denote the firing rates (FRs) for each neuron with or without magnetic stimulation. Magnetic stimulation did not alter the probability of action potential generation (P = 0.64; paired Student’s t-test). See Supplementary Fig. 7 for the standard deviation across stimulus conditions for each unit. d, Raster plot and peristimulus time histogram of action potentials evoked by whisker stimulation delivered using an electromagnet after coating a whisker with iron nanoparticles. e, Power spectrogram of the LFP with or without magnetic stimulation (n = 24 per condition). f, Relative LFP power of across δ (1–4 Hz), θ (4–8 Hz), α (8–12 Hz), β (13–30 Hz) and γ (30–70 Hz) bands. Magnetic stimulation did not change the LFP frequency or power (n = 24 per condition; P > 0.99 for interaction between magnetic stimulation and LFP power across frequencies, two-way ANOVA). g–i, Sample traces showing spiking. g, Subthreshold activity. h, Isolated excitatory and inhibitory postsynaptic potentials. i, With (red) and without magnetic stimulation (gray), recorded in current-clamp configuration. Note that initiation of (bursts of) action potentials and subthreshold potentials lack temporal correlations with initiation and termination of magnetic stimulation, suggestive of spontaneous events. j,k, Frequency, amplitude and duration of the excitatory (j) or inhibitory (k) postsynaptic potentials are not modulated by magnetic stimulation (P values calculated by paired Student’s t-test). For action potential statistics, refer to the main text. l,m, Spontaneous inward and outward currents were unaffected by magnetic stimulation (paired Student’s t-test). n, Analysis of action potential (AP) statistics in vitro. All statistical comparisons were performed using paired Student’s t-test within cells across magnetic stimulation conditions, that is, ‘off’ (magnet off) versus ‘on’ (magnet on) in Magneto-expressing cells. Data from control cells that did not express Magneto are labeled as ‘non’. AP thr, AP threshold. o, Voltage-gated conductances as visualized using triangular voltage sweeps (see Supplementary Methods for details). Rate of adaptation is quantified as the fractional change in the half-width of inward rectifications (see Supplementary Methods for details). In all boxplots, center lines represent the distribution median, and box limits are upper and lower quartiles; whiskers are 1.5× the interquartile range, and outliers are shown as crosses. These results suggest that Magneto does not alter neuronal excitability or control action potential generation in vivo or in vitro.

Extracellular recordings with multi-electrode arrays enable high-throughput observation of neural activity without selectively targeting virally transduced neurons. If Magneto-mediated neuroactuation or viral transduction were limited in efficacy, the spatiotemporally correlated membrane depolarization across local populations triggered by magnetic stimulation would be insufficient to induce spiking in a large population of neurons, potentially explaining the negative results described above. Therefore we next performed intracellular recordings in visualized neurons that express the reporter fluorescence protein mCherry (Fig.1g–o). Whole-cell single-unit recordings showed that magnetic stimulation does not change the probability of action potential generation (magnet off, 0.7 ± 1.4; magnet on, 0.7 ± 1.8 spikes (mean ± s.d.); n = 15 cells, P > 0.8, paired t-test). Current-clamp and voltage-clamp recordings showed that magnetic stimulation is also ineffective in modulating the frequency, amplitude or duration of the subthreshold postsynaptic potentials (Fig.1j,k) and inward–outward currents (Fig.1l,m), suggesting that magnetic stimulation alone is not sufficient to control neural activity in Magneto-expressing neurons. Analysis of the action potential threshold and frequency of action potentials showed that magnetic stimulation does not change the basic statistics of spiking in vitro (Fig.1n). To address whether voltage-gated conductances might be independently modulated during magnetic stimulation, we performed current-clamp recordings with triangular (sawtooth) voltage sweeps. The results showed that the membrane voltage at which the channels open, the amplitude and latency of inward rectifications and the rate of amplitude adaptation were comparable across magnetic stimulation conditions (magnet off versus magnet on) (Fig.1o). These results suggest that magnetic stimulation of neurons that express Magneto does not result in subthreshold or suprathreshold modulation of neural activity.

Neuromodulation via extracranial stimulation has outstanding promise for sensory and motor prosthetics applications. Considering the negligible expression of native TRPV channels in the barrel cortex5,6 (for data visualization visit http://barrelomics.science.ru.nl), heterologous expression of TRPV4–ferritin might be considered a suitable tool for controlled reproducible activation of somatosensory neurons. However, we failed to trigger action potentials or subthreshold depolarization by magnetic stimulation in vitro and in vivo. These findings replicate observations made by others in the cerebellum7, hippocampus, entorhinal cortex and barrel cortex8, and in different cell types (neuro-2a (N2A) in this study and HEK2939) using multiple expression systems (including transfection, lentivirus, Sindbis, adeno-associated virus (AAV) and Fujinami sarcoma virus (FSV)); some of these experiments were performed in the same neuronal classes, using a viral expression system and stimulation magnet identical to that described in the paper by Wheeler et al.1. In support of these findings, immunohistochemical observations (Supplementary Fig. 2b) and biochemical measurements7,9 show that Magneto2.0, following expression, primarily remains in reticular structures in the cytoplasm and is not efficiently transported to the plasma membrane, although TRPV4 (that is, Magneto2.0 primogenitor) is inserted into the membrane as expected (Supplementary Fig. 2b). A critical re-evaluation of the Magneto2.0 design might be necessary.

Methods

Experiments that involved animals were conducted in accordance with the European Directive 2010/63/EU, national regulations in the Netherlands and international guidelines on animal care and use of animals. Details on the materials and methods can be found in the Supplementary Methods.

All experimental procedures were performed in accordance with the European Directive 2010/63/EU, guidelines of the Federation of European Laboratory Animal Science Associations and the NIH Guide for the Care and Use of Laboratory Animals. Experiments were approved by the Animal Ethical Committee of the Radboud University Nijmegen, the Netherlands (permit numbers DEC-2013-172-001 and DEC-2014-275-001).

Animals

Adult transgenic mice, B6;129P2-Pvalbtm1(cr)Arbr/J (n = 6) or Ssttm2.1(cre)Zjh/J (n = 1), were obtained from local breeding colonies, and maintained with ad libitum access to food and water, under a 12–12 h light–dark cycle. Animals were housed together with their littermates until the day of viral injection, after which they were housed individually to reduce the risk of postoperative injury. Animals that received drive implantation for chronic electrophysiological recordings were placed in larger cages with higher ceilings to reduce the risk of mechanical damage. In total seven animals (3–7 months old) were used for the experiments described herein. Except two B6;129P2-Pvalbtm1(cr)Arbr/J mice used for chronic recordings, all mice were female.

Lentiviral vector preparation

For in vivo gene delivery, pcDNA3.0-Magneto2.0-p2A-mCherry (Addgene, 74308) was subcloned into a lentiviral vector, pFUGW-V4trunc-fer-traffick-p2A-mCherry. Lentiviruses were produced as described before8 with modifications. In brief, HEK293T/17 (ATCC, CRL-11268) cells were seeded onto 8 × 10 cm dishes and, per dish, transfected with a total of 12 µg endotoxin-free plasmid DNA containing the helper plasmids pPL1 (3.4 µg), pPL2 (1.7 µg), pPL-VSVg (2.6 µg) and pFUGW-V4trunc-fer-traffick-p2A-mCherry (4.2 µg) using jetPRIME transfection reagent (Polyplus Transfection, 114-15) according to the manufacturer’s instructions. After 36 h, culture medium (DMEM with 10% FCS, 1 mM Na-pyruvate, 100 units per ml penicillin and 100 µg ml−1 streptomycin) was replaced with fresh medium containing 4 mM valproic acid (VPA). The next day (24 h later), medium was removed from culture dishes, supplied with 20 mM HEPES and filtered using a prewashed 45 µm filter, and fresh medium (containing VPA) was added to the dishes. Filtered medium was slowly transferred onto a cushion of 20% sucrose in Hank’s Balanced Salt Solution (HBSS) and ultracentrifuged for 5 h at 12,500 r.p.m. in a Sorvall Discovery 100 ultracentrifuge (Thermo Scientific) using an AH 629 rotor. After centrifugation, medium and sucrose in HBSS were discarded and the pellet was supplied with 50 µl HBSS and placed at 4 °C overnight. This process was repeated with fresh medium 24 h later, after which the second pellet was stored overnight at 4 °C. The two pellets were then resuspended and combined to constitute a total of 200 µl suspension in HBSS, which was aliquoted and stored at −80 °C until further use.

The genomic titers were determined using quantitative PCR (qPCR). Amplicons were generated using primers against WPRE (forward, 5′-ggcactgacaattccgtggt-3′; reverse, 5′-agggacgtagcagaaggacg-3′; Sigma-Aldrich). To control for unpackaged plasmid DNA, viral suspensions were treated with DNAse I (Invitrogen, 18068015). Each sample contained 5.5 µl milli-Q, 12.5 µl SensiFast SYBR-Green Mastermix (Bioline, BIO-98005), 5 µM forward primer, 5 µM reverse primer and 1 µl sample. An initial 2 min at 95 °C were followed by 40 cycles of melting (at 95 °C for 5 s), annealing (at 60 °C for 10 s) and extension (at 72 °C for 20 s). After cycling was complete, a melting protocol was performed, measuring fluorescence intensity from 60 °C to 95 °C with a step size of 0.5 °C to control for amplicon specificity. To determine the physical titer, a standard curve was generated based on the plasmid DNA. The calculated titer was 5.6 × 108 particles per ml.

The nucleotide sequence of the Magneto2.0-p2A-mCherry open reading frame in the lentiviral vector was verified by Sanger sequencing (Lightrun (GATC Biotech)) using the following primers (Sigma-Aldrich):

  • Magneto-2FW (5′-caaggcacttctgaacttaagc-3′)Magneto-3FW (5′-ctggtttacaacagcaagatc-3′), Magneto-4FW (5′-ctggacctcttcaagctcac-3′)Magneto-5FW (5′-acttcctggagactcacttc-3′), Magneto-6FW (5′-tctttgacaagcacaccctg-3′)Magneto-7FW (5′-tcctccgagcggatgtac-3′), Magneto-1RV (5′-tagccaccctcatccttg-3′)Magneto-2RV (5′-ggagctccacgtaatgc-3′).

  • The sequence was compared with pcDNA3.0-Magneto2.0-p2A-mCherry (Addgene, 74308) using ContigExpress and AlignX (Vector NTI Advance 10 (Invitrogen)), and had 100% sequence similarity (see Supplementary Table 1).

Viral gene delivery in vivo

Lentiviral particles were pressure injected as described before8,10,11. For in vitro slice experiments (n = 5 mice) in vivo viral injections were performed under isoflurane anesthesia. Body temperature was monitored and maintained at 37 ± 0.5 °C. A glass capillary with an approximately 20 µm tip containing lentiviral particles was used to deliver the virus to the barrel subfield of the primary somatosensory cortex bilaterally (anteroposterior −1.5 mm from bregma, and mediolateral 3.0 mm from midline) after skin incision and retraction. Injections were initiated at a depth of approximately 500 µm while slowly retracting the capillary until a depth of approximately 200 µm was reached. Approximately 300 nl was injected over 15 min, after which pneumatic pressure was removed and the capillary was left for an additional 2 min to allow viral particles to spread before full retraction of the injection needle. After the injections were completed, carprofen (8–10 mg kg−1) was administered subcutaneously. The skin was sutured and the animal was returned to its home cage after recovery. Animals were typically awake and mobile within 20–30 min.

For in vivo chronic recording experiments (n = 2 mice), viral injections were performed through a polyimide tube (outer diameter = 105 µm, inner diameter = 40 µm) that was positioned at the center of 15 tetrodes carried by a ‘FlexDrive’, secured permanently to the skull. This ensured the spatial alignment of the virally transduced neurons and recording electrodes. Each tetrode was placed in a polyimide tube and connected to a fine screw that allowed axial positioning of the tetrodes in the barrel cortex. The drive preparation was carried out as described previously12 with the exception of the inclusion of the aforementioned access port for viral injections. Viral injection was performed under isoflurane anesthesia as approximately 300 nl viral vectors delivered in around 15 min approximately 200 μm below the cortical surface.

Chronic extracellular recordings

FlexDrive implantation was performed under isoflurane anesthesia while body temperature was maintained at 37 ± 0.5 °C. The surgery started after subcutaneous injection of cefazolin (20 mg kg−1), dexamethasone (2 mg kg−1), carprofen (5 mg kg−1) and saline (15 ml kg−1) and intramuscular injection of buprenorphine (0.03 mg kg−1). After skin incision, a window was prepared above the barrel cortex as described before3,13. In short, a 4 mm2 area of skull (0 to −2 mm from bregma and 2 to 4 mm from midline) was thinned while intermittently cooling the skull using saline drops. The space between the skull and the pia was buffered with saline, after which thinned skull was removed using a fine pair of forceps. The dura mater was incised using a 29 gauge insulin syringe needle. The angle of the FlexDrive was adjusted to ensure perpendicular penetration of the electrodes into the cortex before the FlexDrive was fixed on the skull using dental cement (Super-Bond C&B, Sun Medical Co.). The animal was returned to a warm cage for recovery and then returned to home cage, where HydroGel (DietGel Recovery, Clear H2O) and wet food were provided for ease of water and food consumption. Animals were awake within 30 min and regained motor activity shortly after.

The electrodes were gradually inserted into the brain 1 week after surgery, moving the electrodes approximately 50–200 μm in every session (the speed was determined by the predicted distance to the brain and electrical recording characteristics in a given location). Data acquisition was performed using Open Ephys interface14 after the signals were filtered (0.1–6,000 Hz) and digitized (at 30 kHz per channel) using a 64 channel amplifier (Intan Technologies, RHD2164). The data were processed offline (see “Data analysis” below). Animals were habituated to the behavioral chamber for 15 min before magnetic stimulation during chronic recordings was provided either by a permanent magnet block, or a custom electromagnet, as described previously15. The permanent magnet block consisted of three block (20 × 10 × 2 mm) and 14 ring (8 mm in diameter) neodymium magnets. It was manually placed within 7–9 mm of the skull. Animals were not aware of the upcoming magnetic stimulation, as the experimental chamber was surrounded with translucent dark gray plexiglas walls (thickness = 3 mm). The intensity of magnetic stimulation at the recording site was calculated to be >50 mT (see Supplementary Fig. 1 for magnetic intensity measurements). The results were comparable across magnetic stimulation conditions, and were therefore combined and presented together. Whisker deflections were delivered using the electromagnet after select individual whiskers in the C-row were coated with iron nanoparticles4.

Acute intracellular recordings

Acute brain slices were prepared, 5–8 weeks afterpost injection, as described before3,13,15 with modifications for the adult brain. In brief, ice cold slicing medium (choline chloride 108 mM, KCl 3 mM, NaHCO3 26 mM, 6 MgSO4, NaHPO4 1.25 mM, d-glucose 25 mM, Na-pyruvate 3 mM, and CaCl2 2 mM) was carbogenated (95% O2, 5% CO2) for ≥ 30 min before brain slice preparation. Animals were anesthetized using isoflurane, after which the brain was perfused by clamping the aorta and injecting the slicing medium into the left atrium of the heart, and the right atrium was rapidly cut. Perfusion was maintained for approximately 1 min, after which the brain was quickly removed. The brain was embedded in 2% agarose and coronal sections (300 µm) were made using a VF-300 compresstome (Precisionary Instruments). Slices were immediately placed in carbogenated artificial cerebrospinal fluid (ACSF, NaCl 120 mM, KCl 3.5 mM, MgSO4 1.3 mM, CaCl2 2.5 mM, d-glucose 10 mM, NaHCO3 25 mM, NaHPO4 1.25 mM), which was kept at 30 °C. After placing the slices in ACSF, heating of ACSF was stopped and its temperature was allowed to drop to room temperature (20–25 °C) gradually for 1 h before intracellular recordings where it was kept in the incubation chamber. All chemicals were obtained from Sigma-Aldrich unless otherwise specified.

Slice recordings were performed at room temperature in carbogenated ACSF (flow rate 1 ml min−1) under a Nikon Eclipse FN1 upright microscope. Glass capillaries (Sutter, GC150-15 F; i.d. 0.5 mm, o.d. 1.0 mm) were pulled using a PP2000 pipette puller (Sutter) to prepare recording pipettes with an impedance of 8 ± 2 MΩ. During all experiments, pipettes contained the same intracellular solution (k-Gluconate 130 mM, KCl 5 mM, HEPES 10 mM, MgCl2 2.5 mM, Mg-ATP 4 mM, Na-GTP 0.4 mM, Na-phosphocreatine 10 mM and EGTA 0.6 mM). Fluorescence-guided (mCherry) targeting was performed using a bandpass filter (590 nm, Nikon G-2A) and a white LED light source (CoolLED). Data were acquired using a HEKA EPC9 amplifier controlled using HEKA PatchMaster software (v2 × 90.2). After entering the whole-cell configuration, cells were kept stable at around −70 mV. The permanent magnet (cylindrical N42 neodymium magnet, diameter 1.6 mm) was positioned using a precision micromanipulator (Sensapex) within approximately 200 ± 100 μm. The intensity of magnetic stimulation at the recording site was calculated to be > 140 mT (see Supplementary Fig. 1 for magnetic intensity measurements).

Current-clamp or voltage-clamp protocols were performed with and without magnetic stimulation in every cell; the order was pseudo-random. Excitatory and inhibitory postsynaptic potentials were recorded for 2–5 min after clamping each neuron 10 mV below its spiking threshold, which was determined using a step-and-hold protocol (t = 500 ms, inter-sweep interval = 5 s). If needed, current injection was adjusted to maintain the target membrane potential. In voltage-clamp recordings, the membrane potential was clamped 10 mV below the spiking threshold. To determine the membrane potentials at which voltage-gated conductances are initiated a triangular (sawtooth) stimulus protocol was used. The cell was clamped 10 mV below spiking threshold (Vstart) for 50 ms, after which the membrane potential was ramped linearly to –Vstart over a 100 ms period before the membrane was depolarized back to Vstart, again over a 100 ms period. Each triangular pulse was repeated an additional four times in every given sweep. Each sweep was repeated three times with an inter-sweep interval of 10 s. Spontaneous excitatory and inhibitory postsynaptic currents were recorded while clamping neurons 10 mV below their spiking threshold for a duration of 2–5 min.

Immunohistochemistry

After in vivo or ex vivo experiments, brain tissue (whole brains or acutely prepared slices) was fixed in 4% paraformaldehyde (PFA) at least overnight. The tissue was then transferred to a solution of 40% sucrose in PBS until saturation, after which it was frozen using dry ice and cut to 40 µm sections using a Microm HM-430 sliding microtome (Thermo Fisher Scientific). Sections were individually stored in antifreeze at −30°C until further use or transferred to PBS and immediately processed. For immunolabeling, sections were treated with the following solutions: 3 × 20 min PBS; 1 × 30 min 0.5% Triton in PBS; and 2 × 15 min PBS. Sections were then pre-incubated in blocking solution (PBS + 2% normal donkey serum (Jackson, 017-000-001) + 0.5% TSA blocking reagent (PerkinElmer, FP1020) for 1 h. Primary antibody (rabbit anti-mCherry (Abcam, 167453)) was diluted 1:200 in blocking solution, and sections were incubated overnight. After primary antibody incubation, sections were washed three times for 15 min in PBS, followed by secondary antibody incubation. Alexa488 donkey anti-rabbit (Jackson, 711-545-152) was diluted 1:200 in blocking solution and sections were incubated for 3 h. Secondary incubation was followed by 2 × 15 min washing in PBS, after which the sections were mounted onto glass slides and allowed to air dry for 2 h. Finally, FluorSave (Millipore, 345789) was applied and sections were covered with a coverslip, and allowed to harden for at least 24 h before confocal imaging. The imaging was done using a Leica SP8 inverted scanning confocal microscope with LAS X software at the General Instrumentation Department of the Faculty of Science, Radboud University, Nijmegen, the Netherlands. Image processing was performed using Fiji software (v1.52n).

Tetrode locations were confirmed using Nissl staining. After sectioning (coronal, 40 µm), sections were mounted onto gelatin-coated slides and left to air dry. Slides were then transferred to 96% ethanol for 10 min, followed by sequential steps (2 min each) of 90%, 80%, 70% and 50% ethanol and finally rinsed in demineralized water. Tissue was then stained using 0.1% cresyl fast violet in demineralized water for 30 min, followed by 5 min washing in demineralized water. Sections were then dehydrated by rinsing in sequential steps of 50%, 70%, 80%, 90% and 96% ethanol (2 min each). Differentiation was done in acidified 100% ethanol for 1 min. For mounting, slides were transferred to 100% ethanol for 2 min, followed by 2 × 5 min xylene, after which mounting medium (Entellan) was applied and slides were covered with a glass coverslip (Thermo Fischer Scientific).

Culturing and transfection of N2a cells

N2a cells were obtained from ATCC (CCL-131) and maintained in culture medium (MEM (Gibco, 41090-028) supplemented with 10% FCS (Gibco, 10270), Na-pyruvate (Gibco, 11360-039) and penicillin–streptomycin (Gibco, 15140-122)) at 37 °C and 5.5 % CO2 atmosphere. 1 day before transfection, 25,000 cells were seeded onto 14 mm coverslips in 24-well plates (for immunofluorescence assay) or 50,000 cells per well in a 12-well plate (for western blot analysis). For transfection jetPRIME (Polyplus Transfection, 114-15) transfection reagent and endotoxin-free plasmid DNA (Nucleobond (Macherey-Nagel, 740422.10)) were used. Recombinant protein was expressed for 48 h before biochemical and immunohistochemical assays were performed.

Immunofluorescence assay for N2a cells

Cells were washed with PBS, fixed for 30 min with 4% PFA in PBS at 4 °C and re-washed with 50 mM NH4Cl in PBS to block any residual PFA. Cells were permeabilized using PBS-T (0.1% Triton-X100 in PBS), blocked for 30 min in blocking buffer (PBS-T with 2% bovine serum albumen (BSA)) at room temperature and incubated overnight at 4 °C with mouse anti-Flag tag antibody (Thermo Fisher Scientific, MA1-91878; 1:100 in blocking buffer). After washing with PBS-T, cells were incubated with secondary antibody goat-anti-mouse IgG H&L-Alexa488 (Invitrogen, A11001; 1:500 in blocking buffer) for 1 h at room temperature, washed with PBS-T, washed with PBS and incubated in 300 nM DAPI (Invitrogen, D-1306) in PBS for 5 min to stain the nuclei. Following a final washing step with PBS, cells were embedded using FluorSave (Calbiochem, 345789). Imaging was performed using a Leica SP8x confocal laser scanning microscope.

Western blot analysis

Cells were lysed in 200 µl lysis buffer (150 mM NaCl, 1.0% Triton-X100, 0.5% sodium deoxycholate, 0.1% SDS, and 50 mM Tris, at pH8.0, supplemented with protein inhibitor mix (cOmplete (Roche, 11873580001)) and incubated on ice for 10 min. The lysates were cleared by centrifugation (12,000 g, for 10 min at 4 °C) and 20 µl of cell lysate was denatured using Laemmli sample buffer for 4 min at 100 °C, separated on 10% SDS-polyacrylamide gel electrophoresis and subsequently transferred onto polyvinylidenedifluoride (PVDF) membrane (Hybond (Amersham)). After blocking, membranes were incubated with anti-Flag tag (1:500), anti-mCherry (1:1,000) and anti-tubulin (E7, 1:100) primary antibodies overnight at 4 °C. Secondary HRP-conjugated goat-anti-rabbit (Santa Cruz, sc-2004; 1:5,000) or goat-anti-mouse (Santa Cruz, sc-2005; 1:5,000) antibodies were used, followed by chemiluminescence (ECL plus (Pierce)). Signal detection was done using an ImageQuant LAS 4000 (GE Healthcare) imaging system.

Data analysis

All data analyses were performed offline in MATLAB (2017b) using custom-written software16, unless stated otherwise.

In vivo recordings

To facilitate spike detection, the extracellular signal was zero-phase bandpass filtered between 600 and 6,000 Hz. Outliers in the frequency domain were removed via an adaptive filter17 before spike detection and sorting were performed using KiloSort18. If necessary, the resulting clusters were merged. The normalized distance in principal component space between any two clusters projected on the axis joining their two centroids was used as a merging criterion as described previously19. To obtain more accurate merging results, high-frequency oscillatory noise that is sometimes present (for example, due to electrical interference) was substituted with Gaussian noise based on the background signal. Additionally, spike clusters with a mean spike amplitude of less than four times the standard deviation of the background noise were removed before doing principal component analysis and cluster merging. Note that unlike other clustering algorithms KiloSort does not include the thresholding pre-processing step, so spikes with smaller amplitudes are also detected (although they tended to be multi-unit clusters).

The standard deviation of the background noise, σ, was estimated from the mode of the signal envelope, given by the magnitude of the analytical signal calculated through Hilbert transform20. High-amplitude stimulus artifacts contaminating the spike clusters were deleted from the cluster depending on the mean-squared error between the waveform of the potential artifact and the mean waveform of the cluster.

To assess the quality of each cluster, several quality metrics were computed (see Supplementary Fig. 2). Clusters were automatically classified on the basis of these metrics as being ‘noise’, a ‘multi-unit’, a ‘contaminated single unit’ or an ‘isolated single unit’. Only isolated single units were considered for the downstream analysis. To be labeled as an isolated single unit, the cluster must pass the criteria listed in Supplementary Table 2. These criteria include common spike metrics as well as numerical description of the temporal stability of the cluster and waveform correlations across channels. To determine the temporal stability, it was assumed that the neuron’s firing pattern could be roughly approximated by a homogeneous Poisson process. Strong deviation from this process is typically an indication that the cluster is highly contaminated with mechanical or stimulus artifacts that occur only at specific moments during the session, leading to a very irregular firing pattern. This instability was quantified by counting the number of spikes in 10 s intervals throughout the entire session and fitting the resulting distribution with a Gaussian (Supplementary Fig. 2d). The integral of the distribution predicted by the Gaussian was then computed as the measure for the temporal stability. A Gaussian with standard deviation λ3/2 (where λ is the event rate) was used for the fit. This was empirically determined as it better deals with typical inhomogeneities in the Poisson process at higher firing rates than a regular Poisson distribution. To determine the single unit isolation quality, a mixture of drifting t-distributions was fitted to the spiking data21, which returns the fraction of false positive and negative spikes in each cluster. A toolbox has been developed in MATLAB that carries out the aforementioned steps to process the extracellular data. The source code can be downloaded from https://github.com/DepartmentofNeurophysiology/Paser.

FieldTrip22 was used to construct interspike interval histograms (ISIHs), (joint) post-stimulus time histograms (jPSTHs) and correlograms, which were compared between magnet-on and magnet-off (control) conditions. The jPSTHs were normalized by subtracting the product of the individual firing rates from the joint firing rate and dividing by the product of the two standard deviations22. A bin size of 0.5 ms was selected for the ISIHs and 1 s for the jPSTHs.

To separate the low-frequency local field potential (LFP) activity from the high-frequency (spiking) components, the raw signal was low-pass filtered (<300 Hz) before a second order IIR notch filter was applied to eliminate the 50 Hz mains hum. The resulting data were downsampled to 1.2 kHz. Artifacts due to mechanical motion of the drive and electromagnetic interference were removed by locating uncharacteristically large peaks in the filtered signal amplitude, its derivative and power spectral density23 using twice the median absolute deviation as an automatic threshold, and then replacing the detected artifacts with a linear interpolation between neighboring sample points. Time–frequency analysis was subsequently carried out using FieldTrip24. Power spectra were estimated by fast Fourier transform after Hann tapering the signal, followed by spectral smoothing with a Gaussian kernel.

In vitro recordings

Intracellular recordings included both current-clamp and voltage-clamp recordings. In current-clamp configuration a step-and-hold protocol was used to determine the pattern and statistics (for example, rate, timing and spike threshold) of action potentials. Excitatory postsynaptic potentials (EPSPs) and inhibitory postsynaptic potentials (IPSPs) were recorded separately in the absence of any somatic current injection. In the voltage-clamp configuration triangular pulse injections were used to determine at which membrane voltage the conductances were observed. Excitatory postsynaptic currents (EPSCs) and inhibitory postsynaptic currents (IPSCs) were recorded in the absence of any voltage clamp without tetrodoxin in the bath. All recordings included within-cell controls (that is, comparisons were across magnet-on versus magnet-off conditions). The data were split into groups post-hoc before being treated as described below.

For detection of EPSC and IPSC events, the signal was denoised using a second order IIR notch filter at 50 Hz. To detect the inward currents, the filtered signal was first smoothed by convolving with a Gaussian kernel and then binned (1 s bin−1). The mean and standard deviation of the processed signal in every interval were computed separately and used to set a threshold for detection of candidate events. A second threshold, calculated after excluding the candidate events determined during first thresholding, allowed adaptation of the event detection threshold and ensured that the event threshold estimate was not confounded by the rate of activity in any given period. The candidate events identified after this second passage were classified as events if they satisfied selection criteria, including: the duration of the events was more than 2 ms and less than1,000 ms, the absolute peak appeared within the first third of the event duration, the peak duration was less than 1.5 ms, the current decayed with an exponential decay from the absolute peak, and the current decay was complete by the end of the event, as quantified by comparing the average current within the first 0.5 ms of the event with that in the last 0.5 ms. Events with a decay rate <0 ms−1, and integral <5 pA*ms were excluded from the analysis. The characteristics of each event (amplitude, duration, decay rate (estimated with fitting an exponential decay), total current and instantaneous frequency) were then computed using the raw, non-smoothed signal. The signal processing was identical for the inward and outward currents. In the case of inward current, the data were inverted before event detection. For detection of EPSP and IPSP events the signal was denoised using a second order IIR to remove the 50 Hz mains hum and detrended, before a polynomial was fit to the data. Events were detected using the ‘findpeaks’ function in the Signal Processing Toolbox of MATLAB with minimum peak amplitude of 1 mV. The interpeak interval was >2 ms. The signal processing was identical for the EPSP and IPSP events. In the case of IPSPs, the data were inverted before event detection. For stimulus evoked response detection, as for the ramp-and-hold stimulation protocol in the voltage-clamp and current-clamp configurations, and the triangular sweeps in the voltage-clamp configuration, evoked responses were detected using findpeaks. The minimum peak amplitude was defined as 30 mV in current-clamp and 0.2 nA in voltage-clamp configuration during the ramp-and-hold protocol. Otherwise it was set to 0.1 nA. In voltage-clamp configurations event detection was performed in the absolute signal.

Statistical analysis

Paired Student’s t-test was used in comparisons between and within cells across stimulation comparisons, unless otherwise stated. When there was more than one independent variable, two-way analysis of variance (ANOVA) was used after testing for normality and homoscedasticity of the distributions.

Material, data and code availability

The viral vector, data and data analysis software are available upon request.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

References

  1. 1.

    Wheeler, M. A. et al. Genetically targeted magnetic control of the nervous system. Nat. Neurosci. 19, 756–761 (2016).

  2. 2.

    Allen, C. B., Celikel, T. & Feldman, D. E. Long-term depression induced by sensory deprivation during cortical map plasticity in vivo. Nat. Neurosci. 6, 291–299 (2003).

  3. 3.

    Celikel, T., Szostak, V. A. & Feldman, D. E. Modulation of spike timing by sensory deprivation during induction of cortical map plasticity. Nat. Neurosci. 7, 534–541 (2004).

  4. 4.

    Clem, R. L., Celikel, T. & Barth, A. L. Ongoing in vivo experience triggers synaptic metaplasticity in the neocortex. Science 319, 101–104 (2008).

  5. 5.

    Kole, K. et al. Transcriptional mapping of the primary somatosensory cortex upon sensory deprivation. Gigascience 6, 1–6 (2017).

  6. 6.

    Kole, K. et al. Proteomic landscape of the primary somatosensory cortex upon sensory deprivation. Gigascience 6, 1–10 (2017).

  7. 7.

    Xu, F.-X. et al. Magneto is ineffective in controlling electrical properties of cerebellar Purkinje cells. Nat. Neurosci. https://doi.org/10.1038/s41593-019-0475-3 (2019).

  8. 8.

    Celikel, T. et al. Select overexpression of homer1a in dorsal hippocampus impairs spatial working memory. Front. Neurosci. 1, 97–110 (2007).

  9. 9.

    Wang, G., et al. Revaluation of magnetic properties of Magneto. Nat. Neurosci. https://doi.org/10.1038/s41593-019-0473-5 (2019).

  10. 10.

    Freudenberg, F. et al. GluA1 and its PDZ-interaction: a role in experience-dependent behavioral plasticity in the forced swim test. Neurobiol. Dis. 52, 160–167 (2013).

  11. 11.

    Freudenberg, F., Marx, V., Seeburg, P. H., Sprengel, R. & Celikel, T. Circuit mechanisms of GluA1-dependent spatial working memory. Hippocampus 23, 1359–1366 (2013).

  12. 12.

    Voigts, J., Siegle, J. H., Pritchett, D. L. & Moore, C. I. The flexDrive: an ultra-light implant for optical control and highly parallel chronic recording of neuronal ensembles in freely moving mice. Front. Syst. Neurosci. 7, 8 (2013).

  13. 13.

    Allen, C. B., Celikel, T. & Feldman, D. E. Long-term depression induced by sensory deprivation during cortical map plasticity in vivo. Nat. Neurosci. 6, 291–299 (2003).

  14. 14.

    Siegle, J. H. et al. Open Ephys: an open-source, plugin-based platform for multichannel electrophysiology. J. Neural Eng. 14, 045003 (2017).

  15. 15.

    Clem, R. L., Celikel, T. & Barth, A. L. Ongoing in vivo experience triggers synaptic metaplasticity in the neocortex. Science 319, 101–104 (2008).

  16. 16.

    Brouns, T. & Celikel, T. PASER for automated analysis of neural signals recorded in pulsating magnetic fields. preprint at bioRxiv https://doi.org/10.1101/739409 (2019).

  17. 17.

    Jun, J. J. et al. Real-time spike sorting platform for high-density extracellular probes with ground-truth validation and drift correction. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/101030v2 (2017).

  18. 18.

    Pachitariu, M., Steinmetz, N. A., Kadir, S. N., Carandini, M. & Harris, K. D. Fast and accurate spike sorting of high-channel count probes with KiloSort. Adv. Neural Inf. Process. Syst. 30, 4448–4456 (2016).

  19. 19.

    Yger, P. et al. Fast and accurate spike sorting in vitro and in vivo for up to thousands of electrodes. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/067843v1 (2016).

  20. 20.

    Dolan, K., Martens, H. C., Schuurman, P. R. & Bour, L. J. Automatic noise-level detection for extra-cellular micro-electrode recordings. Med. Biol. Eng. Comput. 47, 791–800 (2009).

  21. 21.

    Shan, K. Q., Lubenov, E. V. & Siapas, A. G. Model-based spike sorting with a mixture of drifting t-distributions. Preprint at bioRxiv https://www.biorxiv.org/content/10.1101/109850v1 (2017).

  22. 22.

    Aertsen, A. M., Gerstein, G. L., Habib, M. K. & Palm, G. Dynamics of neuronal firing correlation: modulation of “effective connectivity”. J. Neurophysiol. 61, 900–917 (1989).

  23. 23.

    Bakštein, E. et al. Methods for automatic detection of artifacts in microelectrode recordings. J. Neurosci. Methods 290, 39–51 (2017).

  24. 24.

    Oostenveld, R., Fries, P., Maris, E. & Schoffelen, J.-M. FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput. Intell. Neurosci. 2011, 156869 (2011).

Download references

Acknowledgements

We thank the members of the Department of Neurophysiology for stimulating discussions and critical insight on the manuscript. This work was supported by grants from the European Commission (Horizon2020, 660328), European Regional Development Fund (MIND, 122035) and the Netherlands Organisation for Scientific Research (NWO-ALW Open Competition, 824.14.022) to T.C., as well as doctoral fellowships from the Chinese Scholarship Council to Y.Z. and X.Y., and the National Council for Scientific and Technological Development of Brazil (CNPQ) to A.S.L.

Author information

Correspondence to Tansu Celikel.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Integrated supplementary information

Supplementary Figure 1 Change in magnetic field strength as a function of distance from the permanent magnet.

Measurements were made using a HT-20 Gaussmeter (Hangzhou BST Magnet Co. Ltd., China) and repeated three times at each distance. The error bars are standard deviation. The distance between the target cells and magnet was ~200 ± 100 𝜇m in in vitro and ~8 ± 2 mm in in vivo experiments.

Supplementary Figure 2 Magneto2.0 is cleaved from mCherry but remains primarily in the cytoplasm.

(a) To confirm that Magneto was effectively released from its fusion with mCherry, N2A cells were transfected with 1: pFUGW-V4trunc-fer-traffick-p2A-mCherry, 2: pcDNA3-Magneto-p2A-mCherry, 3: pcDNA3-flag-TRPV4 and 4: pcDNA3 plasmids. TRPV4 and empty pcDNA3 served as controls; tubulin was used as a loading control. Magneto and TRPV4-expression were analyzed by Western blot using anti-flag and anti-mCherry antibodies. The major protein products identified by the anti-flag antibody were the ~135 kDa Magneto protein and the ~95-kDa flag-TRPV4 protein. The ~160-kDa Magneto-p2A-mCherry product was recognized by both anti-flag and anti-mCherry antibodies. However, the principal protein detected by the anti-mCherry antibody was mCherry. These findings indicate that Magneto was effectively cleaved from its fusion with mCherry. Note that both pFUGW-V4trunc-fer-traffick-p2A-cherry and pcDNA3-Magneto-p2A-mCherry produce the same protein products, indicating that sub-cloning of Magneto into the viral vector pFUGW did not affect the Open Reading Frame of Magneto-p2A-mCherry. (b) I: N2a cells were transfected with pcDNA3-flag-TRPV4 and immunostained using anti-flag antibodies. Note that flag-TRPV4 was observed in the plasma membrane. II–IV: N2a cells were transfected with Magneto-p2A-mCherry plasmid and stained using anti-flag antibodies (II, IV, green). mCherry fluorescence was directly imaged (II, red), DAPI (blue) was used to stain the nuclei (II–III). Note that mCherry was mostly localized to the nucleus, however, also a significant amount was present in the cytoplasm. In contrast to flag-TRPV4, most Magneto expression was found in reticular structures in the cytoplasm, most likely representing the ER (IV). These results indicate that, compared to flag-TRPV4, Magneto is less effectively transported to the plasma membrane of N2a cells. Scale bar, 10 µm.

Supplementary Figure 3

The Western Blot shown in Supplmentary Figure. 2a at different exposures.

Supplementary Figure 4 Isolation quality of single units.

(a) Concatenated spike shapes across the tetrode channels visualized as a density heatmap of time-voltage values. (b) The peak-to-peak amplitude (top) and firing rate (bottom) over the whole recording period in a session, showing the temporal stability of the cluster. (c) Histogram of normalized spike amplitudes. Any amplitude to the left of the dashed vertical line crosses the minimum threshold for spike detection, which is three times the average background signal (voltage). (d) Distribution of running average spike counts in 10 sec bins (with 5 sec overlap). The red line indicates the predicted distribution of spike counts by a Gaussian distribution given the observed variance in spike count. (e) Cross-correlation between two tetrode channels that are maximally dissimilar. The large peak at zero lag indicates that the spike waveforms are temporally aligned. (f) Autocorrelation of spike events.

Supplementary Figure 5 Magnetic stimulation does not alter the temporal correlations of action potentials.

(a) A representative example of spiking correlations between two simultaneously recorded neurons from the same tetrode. Joint peristimulus time histograms represent the temporal correlations across the binary states of magnetic stimulation. (b) Spiking pattern in single units (auto-correlations; left) and across simultaneously recorded units (cross-correlations; right). Gray traces: Correlation in the absence of magnetic stimulation (magnet off), red: during magnetic stimulation (magnet on), blue: the pairwise difference between magnet off–magnet on. Thick traces are population averages; color-coded shadows in the background represent the standard deviation within stimulation condition. Neither single cell spiking correlations (N= 353, P = 0.98, paired t-test), nor spiking correlations across neurons (N= 939, P = 1.00, paired t-test) were altered upon magnetic stimulation. See Supplmentary Figure. 5 for Poincaré analysis of spiking in single neurons across stimulus conditions.

Supplementary Figure 6 Spiking pattern does not change during magnetic stimulation in vivo.

(a) Poincaré plot of interspike intervals (ISI) from a representative neuron under control conditions (magnet off) and during magnetic stimulation (magnet on). Corresponding ISI histograms (bin size = 0.5 ms) are shown on the left and bottom of each plot. (b) Post-stimulus time histograms of a representative neuron for the magnet on and off conditions. (c) Mean normalized ISI Poincaré plot across all neurons (N = 235). Neurons that fired <50 spikes during the period of observation were excluded from the plot (N = 118).

Supplementary Figure 7 Firing rate of identified single neurons before and during magnetic stimulation.

Each dot represents the average firing rate of a neuron (N= 353) across the two conditions as in Fig. 1c. The color code represents firing rate. The error bars are standard deviation from the mean within session. See Fig. 1c for the results of statistical comparison.

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Further reading