On-demand cell-autonomous gene therapy for brain circuit disorders

Summary Several neurodevelopmental and neuropsychiatric disorders are characterized by intermittent episodes of pathological activity. Although genetic therapies offer the ability to modulate neuronal excitability, a limiting factor is that they do not discriminate between neurons involved in circuit pathologies and “healthy” surrounding or intermingled neurons. We describe a gene therapy strategy that down-regulates the excitability of overactive neurons in closed loop, which we tested in models of epilepsy. We used an immediate early gene promoter to drive the expression of Kv1.1 potassium channels specifically in hyperactive neurons, and only for as long as they exhibit abnormal activity. Neuronal excitability was reduced by seizure-related activity, leading to a persistent antiepileptic effect without interfering with normal behaviors. Activity-dependent gene therapy is a promising on- demand cell-autonomous treatment for brain circuit disorders.

experiments, neurons were plated at a density of 120,000 -150,000 cells/well on poly-L-lysine (10 mg/ml, in borate buffer) treated coverslips (13mm, VWR, #631-0148) in 24-well cell culture plates. Neurons were maintained at 37 C in a humidified 5% CO2 environment, and 50% of the media was replaced each week.

Lentiviral transduction of primary cultures
Lentiviral transduction of cortical neurons was performed on DIV 1. Neurons were incubated with the lentivirus for 24 hours and then all of the virus-containing media was removed and replaced with preconditioned media (56).

AAV transduction of primary cultures
AAV transduction of cortical cultures was performed between DIV 6 and DIV 8. Cells were transduced by directly adding the virus into the culture media at a MOI (multiplicity of infection) of >5000. Cultures were monitored for 24 hours posttransduction for signs of toxicity. If no cytotoxicity was observed, the AAV was left in the media. If some toxicity was apparent, a 50% media change was performed to reduce the number of viral particles. pyramidal neurons identified by their shape as previously described (58). Prior to patch clamp recordings, 50% of the medium was removed, and conserved, and cells were incubated for 30 minutes with 30µM PTX. The cells were then treated by exchanging the reserved 50% medium containing 1 μM tetrodotoxin (TTX) for 2 hours to prevent further dsGFP expression. The coverslips were washed at least twice in extracellular solution to wash out TTX just before the experiment. Patch clamp pipettes were pulled from thin wall capillary (1.5OD, 1.17ID) (Harvard Apparatus) with a two-step vertical puller. Action potentials were counted only if the voltage crossed 0 mV. All the recordings were carried out from neurons held at -70mV at room temperature with continuous perfusion of the extracellular solution. Electrophysiological recordings were made with a Multiclamp 700A amplifier (Axon Instruments, Molecular Devices). The amplifier was used in combination with Power 1401 (CED) and Signal 6.0 software (Cambridge Electronic Design, Ltd). The data were filtered at 10 kHz and digitized at 50 kHz (BNC-2090, NI-6221, National Instruments). Bridge balance was applied. The current step protocol consisted of 500ms 10pA current steps from -20pA to 250pA. If no action potential was evoked, the neuron was injected with a higher current step of 350pA.

Multi-electrode arrays (MEAs)
6-well-MEA chambers with 60 electrodes were purchased from MultiChannel Systems (60-6wellMEA200/30iR-Ti). The bottoms of wells were coated with Bovine Serum (TFS, #261700430) or HI-Fetal Bovine Serum (TFS) for 2-24 hours, washed with sterile dH20, air-dried and then coated with a fresh solution of Poly-L-lysine (1mg/ml in borate buffer, Sigma-Aldrich P2636) and laminin (laminin from Engelbreth-Holm-Swarm murine sarcoma, #L2020-1MG, and PBS (Phosphate-Buffered Saline, pH = 7.4, TFS, #10010023) in a ratio of 4:1:55. Cortical neurons were plated directly onto the electrodes with a high density of 90,000 -100,000 cells/well. Neurons were transduced with AAV on DIV6-8 with MOI > 5000. For the recordings, MEA chambers were placed on the MEA recording setup, which was grounded, fixed on an air-supported platform and enclosed in a Faraday cage. The recording chamber was maintained at 37°C with a 1-channel temperature controller (Multichannel systems, #TC01). For experiments that required repeated recordings from the same chamber, the top was tightly covered with sterile Parafilm. Raw data were collected using 'MC_Rack' (V 4.6.2, MultiChannel Systems). The sampling frequency was 25kHz and the recording time was 5-10 minutes per session. Data processing and analysis were performed with MatLab based software, 'SPYCODE', which was a generous gift from Drs Ilaria Colombi and Michela Chiappalone (60).
The data analysis pipeline consisted of data conversion, data filtering, baseline thresholding, spike detection and burst detection. The data filtering used a cut-off frequency of 300 Hz to select Multi-Unit-Activity components of the signal. The threshold for spike detection was set as 10x the standard deviation. Channels with high levels of noise were removed from the dataset. Bursts were defined as clusters of spikes that were simultaneously detected in at least 50% of the electrodes. The algorithm to detect and analyze all the parameters has been described in Pasquale et al. 2010 (61).
The brain was then excised from the skull and cooled for 1 minute in a beaker containing a semi-frozen 'slush' of carbogenated sucrose-ACSF. The brain was placed on filter paper (Whatman) in a petri-dish and surrounded by sucrose-ACSF: the cerebellum and the rostral third of the frontal cortices were removed, and the brain was hemisected along the sagittal fissure. The right hemisphere was placed on its medial surface and the dorsal surface of the cortex was removed at an angle that best preserves the integrity of the CA1 hippocampal subregion, the so-called 'magic cut' (blade tangential to the dorsal surface and 10° acute to the sagittal plane) (61).
The right hemisphere was mounted on the freshly cut dorsal surface to a vibrating microtome stage (Leica VT1200S, Leica Microsystems) that was coated with a shallow strip (~4 cm x 2 cm) of cyanoacrylate glue. The stage was then placed into the slicing chamber, submerged in sucrose-ACSF, and oriented so that the lateral aspect of the brain faced the vibratome blade. 400 µm brain slices were prepared at a horizontal oscillation amplitude of 1.2 mm and a forward velocity of 0.05 mm.s -1 .
Cut slices were transferred individually using a disposable wide mouth Pasteur pipette into a submerged brain slice holding chamber that contained carbogenated sucrose ACSF maintained at 33°C. After 30 minutes the holding chamber was removed from the water bath and stored at room temperature until needed.
Individual slices were transferred to a submerged recording chamber that was continually perfused with room temperature, carbogenated ACSF containing (in mM): 125 NaCl, 2.5 KCl, 25 NaHCO3, 1.25 NaH2PO4, 25 glucose, 1 MgCl2, 2 CaCl2). The ACSF flow rate was 6-8 mL.min -1 and slices were held in place with a 'harp' fabricated from an O-shaped platinum wire strung with parallel nylon threads. CA1 pyramidal neurons were visualized using an upright microscope (Scientifica SliceScope) equipped with infrared differential interference contrast illumination and a water immersion objective (Olympus XLUMPLFLN water-immersion objective, 1.00 NA). Fluorescently labelled neurons were identified under epifluorescence using a metal-halide lamp epifluorescence illumination system (X-cite 120Q) and GFP filter sets (Chroma), and a CMOS camera (Hamamatsu C11440-36U ORCA-spark).
Electrophysiological recordings were made with a Multiclamp 700B amplifier (Molecular Devices), filtered online at 10 kHz with the built-in 4-pole Bessel Filter, and digitized at 62.5 kHz (NI-6221, National Instruments) using WinWCP software (courtesy of John Dempster, University of Strathclyde, Glasgow, UK). Bridge balance was applied. Neuronal excitability was assessed from the input-output relationship of the neuron. Action potentials were elicited using square-wave depolarizing current steps (25 pA steps, 500 ms; 0 -525 pA) from a holding potential of -60 mV and then from RMP. If no action potentials were observed using this protocol, 50 pA steps (500 ms; 0 -1 nA) were applied from a holding potential of -60 mV to try to obtain the action potential current threshold. The number of action potentials during each 500 ms step was obtained using the threshold (0mV) event detection function in Action potential waveform analysis. Action potential (AP) kinetics were analyzed for the first AP that was elicited from a holding potential of -60 mV using a custom MATLAB script. AP threshold was defined as the first point at which dV/dt > 20 mV/ms. Maximal rising and repolarizing slopes were also obtained from the first derivative. AP amplitude and AHP amplitude (the minimum point occurring 5ms after AP peak) were measured relative to AP threshold. AP half-width was the time window between the AP reaching the half-maximal amplitude on the upstroke and downstroke.

Acute pilocarpine in visual cortex
1.5ul AAV cfos-dsGFP or AAV CamKII-dsGFP (15) was injected into all the V1 cortical layers (DV: 0.7/0.5/0.3). A cannula for guided substrate injection were implanted at the same coordinate after the viruses were injected. Mice were returned to the home cage for post-surgery recovery for two weeks before induction of acute seizures by pilocarpine. Pilocarpine (3M in saline) was injected 0.5 mm below the cannula (DV = -1.0) with a Hamilton 5 μl syringe in a volume of 300 nl. The mice were observed for 2 hours post-pilocarpine injection to evaluate seizure severity. The brains were then extracted and fixed with PFA for immunohistochemistry.

Acute pentylenetetrazole (PTZ)
Male and female mice wild-type C57BL/6J mice (3 months old) were placed in a stereotaxic frame and injected with 1.5 μl AAV of cfos-GFP/cfos-EKC into bilateral ventral hippocampi (AP: 3 mm, ML: 2/3 bregma-lambda, DV = 3.5/3/2.5) with 500 nl at each depth. Two weeks after the virus injection, mice were subjected to PTZ seizure induction. A single dose of PTZ (Sigma-Aldrich) dissolved in saline was administered intraperitoneally at 55mg/kg. The mice were observed closely for the first 30 minutes and their behavior was scored every 5 minutes according to the Racine scale (36). The latency to seizure onset was recorded. The mice were further monitored for 2 hours until they had completely returned to normal and then they were returned to the home cages. The PTZ-seizure induction was repeated, with the same dose and observation period, at two time-points: 24 hours and 14 days after the initial PTZ administration. The researcher who performed the experiments was blinded to the virus injected.

Intra-amygdala kainic acid model of chronic epilepsy
Male wild-type C57BL/6J mice (9-12 weeks old) were injected with kainic acid After two weeks of ECoG data acquisition, we performed bilateral hippocampal virus injections. 1.5μl AAV9 cfos-dsGFP, cfos-EKC or cfos-KCNJ2 was injected bilaterally via cannulas into the ventral hippocampus (DV = 3.5/3/2.5 mm, cannula length = 7.05) under isoflurane anesthesia. The mice were returned to home cages for recovery. Two weeks later, transmitters were turned back on, in some cases with video recordings. Animals were sacrificed after the experiments for brain collection.

Intra-hippocampal kainic acid model
Following a 2 week recovery period after viral injection, mice were briefly anaesthetized and kainic acid (KA) injected into the dorsal hippocampus to induce status epilepticus. The injection needle was inserted to a depth of -2.0mm relative to the dura. 50 nl KA (20mM) was injected into the dorsal hippocampus at a rate of 50 nl/minute. The needle was left in place for 5 minutes, withdrawn, and the cannula cap resealed. Successful induction of status epilepticus was confirmed via monitoring of animal behaviour. Status epilepticus was allowed to continue until spontaneous seizure termination. We have recorded EEG this model and the first seizure appears at 13.22.2 (mean  SD) with interictal spikes present in the hippocampus after SE (recorded with depth hippocampal electrodes).

ECoG data acquisition and PyEcoG event detection
The wireless ECoG data acquisition was previously described (14). A list of features were extracted from the recordings, from which coastline and power spectrum across different frequencies (1-120Hz) were calculated.

Interictal spike analysis
Interictal spikes were detected by locating peaks with an amplitude exceeding 6x the standard deviation of the recording. Seizures were manually excluded from the analysis and large amplitude artefacts were filtered from true interictal spikes by analysis of parameters derived from a library of both true positive and false positive events, constructed by visual inspection of a random sample of 750 identified events.
Spike half-width, maximal amplitude, maximal slope, peak power frequency, and kurtosis were calculated for each spike and the 90% confidence interval of the distribution of true positive and false positive parameter values used to filter the total dataset. Filter values were adjusted empirically. Spike frequency and amplitude were analyzed for each mouse during both baseline and post-treatment phases of recording.

Behavior assessments
All behavioral tests were performed in age-matched and littermate mice 1 hour after starting the light phase and 1 hour before starting the dark phase of their daily day/night cycle. When possible, the tests were performed under red light to reduce anxiety.

Contextual Fear Conditioning (CFC). The fear conditioning was carried out in the
Near Infrared (NIR) Video Fear Conditioning package by Med Associates Inc. A glass chamber with a metal grid was placed inside a sound proof chamber. Near infrared (NIR) light was used for recording under dark conditions. Mice were randomized into two groups, each injected bilaterally with either cfos-dsGFP or cfos-EKC. Mice were housed in their home cage for at least 3 weeks before CFC was assessed. Mice were handled for 7 days before experiment, until signs of fear or anxiety (jumping, freezing, excretion) were extinguished. On the day of the experiment, mice were kept in the habituation room for 1 hour and were only introduced to the behavior room where the CFC apparatus was located when they were due to take the test. The CFC experiment consisted of three phases. On day one, the mice received harmless electrical foot shocks. The electrical stimuli (0.6 mA) were delivered by a metal grid via manual control. Mice were placed in the chamber for 2 minutes of habituation, then three-foot shocks were delivered for 2s each, with ITI (inter-trial-intervals) of 60 seconds. The test was performed in a dark environment and the mice were recorded with the NIR camera. The grid floor was cleaned with 70% EtOH after testing each mouse. Different grid floors were used for males and females and males were always tested before females. The mice were then returned to the home cage. 24 hours after the foot shock, mice were returned to the same chamber with the same conditions, for assessment of fear recall. The mice were placed in the chamber for 5 minutes. No foot shock was delivered. After 1 hour, in which the mice were back in the home cage, the chamber was modified by replacing the grid floor with a hard plastic board, and the walls were covered with colored paper. We also placed a filter paper in the chamber with a drop of almond essential oil. The mice were placed in the 'novel' context for 5 minutes. The freezing behavior of the mice was counted manually by an experimenter blinded to the treatment given.
Open field. Open field behavior was tested in a white open arena of dimension 50x50x50cm. Mice had no prior exposure to the behavior room or to the apparatus.
The mice were habituated for 30 minutes before being introduced into the behavior room. The test was performed under red light to reduce anxiety. At the start of the experiment, mice were gently placed in the center of the arena. The test lasted 30 minutes with the experimenter not present in the same room. The apparatus was cleaned with 70% EtOH between animals. The mice were recorded with an overhead Raspberry Pi4 camera (XL-RB-AluP4+07FAN) and VideoArchiver software. The animal tracking and data analysis were carried out semi-automatically using a script written in Bonsai (Open Ephys, https://open-ephys.org/bonsai). In brief, the outline of the animal was extracted and its body centroid was calculated. The software generated a location for the center of the mice as an x, y coordinate every 30 frames of the recording. The tracked locations were then used to analyze parameters including thigmotaxis and travel distance using an in-house Python script.
Spontaneous T-maze alternations. Mice were returned to the same behavior room 24 hours after the Open field test assessment for T-maze alternations. This test was also performed under red light. The apparatus was specially made with transparent Pyrex materials. The apparatus was cleaned with water after each animal had finished their test. The walls of the apparatus were 20cm high. Each of the arms was 25cm long, and the running track was 50 cm long. A central partition 10 cm long divided the end of the track into two parts facing the two arms. The protocol for Tmaze alternation was adapted from (62). The mice were habituated to the room for 5-10 minutes before the experiment start. They were then placed in the start area at the base of the T, facing away from the track. The mice were handled gently using a plastic tube to which they were habituated during the handling sessions prior to the experiment. The animals ran spontaneously towards the top of the T and entered one arm. An entry was scored once the whole body of the mouse has entered the arm. A guillotine door was then lowered to trap the mouse in the arm for 30 seconds.
Mice were then transported back to the start area, and allowed to run down the track again with all doors open and without the separation wall in the middle. The arm of the T maze entered on the second trial was recorded as either the same or different from the first entry, and the mouse was again left confined to the arm for 30 seconds.
Mice were then immediately returned to the start area with no delay for the next trial.
For each mouse, 10 consecutive trials were performed. In any of the trials, if the mice failed to leave the start area after 90s, they were taken out of the maze for a short break of 10 minutes and tried again. If the mice repeatedly failed to run, they were temporarily suspended from the test, and another trial was conducted on the next day. If the animal still failed to initiate directional movements, they were eliminated from the dataset.

Generation and assembly of hCS and hSS. hiPS cells were harvested using
Accumax TM (Sigma), counted, resuspended in Essential 8 medium supplemented with ROCK-inhibitor (StemCell technology) and seeded at 3,000,000 cells/well of an Aggrewell TM 800 (StemCell Technologies), corresponding to 10,000 cells/microwell, to form embryoid bodies (EBs). After 24h, EBs were dislodged from the microwells and transferred to ultralow-attachment 10cm dishes (Corning) in Essential 6 medium Otherwise, LFPs were recorded for a total of 6000 seconds. We used the standard deviation of the LFP as a measure of excitability of the assembloid. The standard deviation was sampled over 200 ms at 5 second intervals using Clampfit. The median SD over 500 seconds was used to assess changes in activity after the addition of KCl.

Viral labelling of spheroids and assembloids.
To specifically label hSS-derived neurons, 8-10 days before assembly hSSs were placed into a well of an ultra-low attachment 24-well plate in 400 l media plus 10 l lentivirus suspension (LV-mdlx5/6-GFP) and incubated for 48h. For transduction with AAV9-cfos-dsGFP, assembloids were placed into a well of an ultra-low attachment 24-well plate in 400 l media plus 20 l AAV9-cfos-dsGFP. 500 l of media was added the next day and a half media change was performed after 48h. Routine media changes were performed afterwards. Transduced spheroids were analyzed for transgene expression 11 days after AAV infection.  Table S1. Slices were washed in 1X PBS and incubated with fluorochromelabeled secondary antibodies diluted in blocking solution for 3h at room temperature.
Slices were mounted using ProLong TM Gold Antifade (Life Technologies) and images were acquired with ZEN software (Zeiss) on a LSM710 confocal microscope (Zeiss).

Statistical analysis
The statistical tests for significance are detailed in the figure legends. Data are plotted as box and whiskers, representing interquartile range (box), median (horizontal line), and max and min (whiskers), together with all the points. The mean is further shown as "+". Deviation from normal distributions was assessed using D'Agostino-Pearson's test, and the F-test was used to compare variances between two sample groups. Student's t-test (parametric) or Mann-Whitney (non-parametric) tests were used to compare two groups. Fisher's exact test were used for event occurrence frequency comparisons. One-sample t-test was used to compare normalized data. One-way ANOVA was used to compare multiple groups. To compare two groups at different time points we used two-way repeated measure ANOVA, followed by a Bonferroni post-hoc test for functional analysis. Three-way ANOVA was used to compare data with three variables. Statistical analysis was carried out using Prism (GraphPad Software, Inc., CA, USA) and SPSS (IBM SPSS statistics, NY, USA).  were largely consistent regardless of which IEG promoter was driving expression.

Supplementary Text
The small differences between the promoters can be ascribed to either the different time course of activation (see Table S2) or the degree of activation reached by each promoter. We hypothesise that the antiepileptic effect of the cfos-EKC construct is due to the multiple mechanisms of action of this potassium channel that regulates both intrinsic and synaptic excitability, and potentially contributes to an homeostatic rearrangement of the network (74,75) On the other hand, the inability of cfos-KCNJ2 to decrease seizures, even with hyperpolarisation of the RMP, may be due to a short half-life of this potassium channel in the cell membrane (12-24hrs), which is also affected by its own expression levels (76,77). Fig. S1. Network mean firing rate following disinhibition with PTX application.

Supplementary Figures
All data points are shown. Statistics in Fig.1. n=8 in 2 independent repeats.      Single action potential (AP) parameters: peak, amplitude and after hyperpolarization (AHP). C. AP shape: rising slope, repolarizing slope and AP half-width. Student's t tests.         Scale bar overview = 500μm; Scale bar zoomed in images: 250 μm. Weighted cumulative plot normalized by the total seizure count before treatment. D.
Left: Spike frequency normalized to baseline (before viral injection). Student's t test.
Right: Weighted cumulative plot normalized by the total interictal spike count before viral treatment (cfos-dsGFP re-plotted from Fig. 5D). Two-way ANOVA. E.
Percentage change in coastline normalized to baseline (before viral injection) (cfos-dsGFP re-plotted from Fig. 5F). Student's t test. F. Percentage power change vs baseline in different frequency bands (cfos-dsGFP re-plotted from Fig. 5D). Two-way ANOVA. Alternation rate before and after either cfos-dsGFP or cfos-EKC injection in both hippocampi in epileptic animals.