Recording and Analyzing Multimodal Large-Scale Neuronal Ensemble Dynamics on CMOS-Integrated High-Density Microelectrode Array

All experiments were performed in accordance with the applicable European and national regulations (Tierschutzgesetz) and were approved by the local authority (Landesdirektion Sachsen; 25-5131/476/14).

1. Ex-vivo brain slices from hippocampal-cortical and olfactory bulb circuits on HD-MEA

1. Preparation of experimental cutting and recording solutions (Figure 2A)
   1. On the experimental day, prepare 0.5 L of high-sucrose cutting solution and 1 L of artificial cerebrospinal fluid (aCSF) recording solution (Table 1A,B).
      1. Add all solid chemicals to a dry volumetric flask, then fill part of the way with double distilled (dd) water.
      2. Add MgCl₂ and CaCl₂ from 1 M stock solutions, then fill the remainder with dd water. Begin constantly stirring with a magnetic stirrer until visible solids have dissolved ~5 min.
   2. Use a freezing point osmometer to validate the osmolarity between 350-360 mOsm for the high-sucrose cutting solution and 315-325 mOsm for the aCSF recording solution.
   3. Use a pH meter to validate the pH between 7.3-7.4 for the high-sucrose cutting solution and 7.25-7.35 for the aCSF recording solution. Begin continuously bubbling with 95% O₂ and 5% CO₂.
   4. Place the high-sucrose cutting solution on ice for at least 30 minutes prior to slicing and begin continuously bubbling with 95% O₂ and 5% CO₂.
   5. After 10 min of carbogenation, fill a 50 mL beaker with 30 mL cutting solution and store it in the freezer (-20 °C) for 20-30 min or until partially frozen.
      NOTE: All solutions should be prepared fresh for each experiment. The dd water used here is autoclaved ultrapure water stored at room temperature (RT). The amount of solution prepared should be tailored to the specific study question.
2. Preparation of brain slice workspace areas (Figure 2A)
   1. Bring the animal into the experimental room.
      NOTE: In this protocol, C57BL/J6 female mice aged 8-16 weeks were used as previously described^26,29,32. The animal should be allowed to acclimate for at least 30 min after transport. Long-distance transfers (i.e., inter-institute) should be avoided on the same day as the experiment. The age, sex, and strain of the animal have to be determined based on the specific study question.
   2. While the animal is acclimating and the high-sucrose solution is cooling down, place the required tools in each designated workspace (see Table of Materials).
   3. Prepare brain slice recovery and maintenance workspace. Fill the slice recovery chamber with carbogenated aCSF recording solution and place the chamber in the water bath set to 32 °C. Maintain continuous carbogenation throughout the experiment.
   4. Prepare brain slice preparation workspace. Setup the vibratome – place the blade in the vibratome blade holder and calibrate the vibratome to the correct settings (blade travel speed: 0.20 mm/s, height amplitude: 95 µm, blade angle: 45°). Fill the vibratome ice tray with ice and the buffer tray with a high-sucrose cutting solution and begin carbogenating the solution in the buffer tray.
   5. Prepare the brain preparation workspace. Fill the 150 mm glass Petri dish with ice and place a 90 mm plastic culture dish with filter paper inside. Fill the plastic culture dish with a high-sucrose cutting solution and begin carbogenating. Add a drop of super glue to the chilled specimen plate and attach the agarose mold.
      NOTE: Agarose mold is prepared at least the day before with 3% agarose in water in a custom mouse brain mold.
   6. Lastly, prepare the brain extraction workspace. Cover aluminum foil with tissue paper, retrieve a 50 mL beaker containing high-sucrose cutting solution slush, and add isoflurane to the anesthesia chamber.
      NOTE: Anesthesia will be added to the anesthesia chamber ~1 min prior to animal placement. A 50 mL beaker with 30 mL of high-sucrose cutting solution slush will be removed from -20 °C freezer ~2 min before decapitation.
3. Extraction and slicing of mouse brain
   NOTE: This entire procedure should be performed as quickly as possible to avoid a lack of oxygenation to the brain. Brain removal should only take 1-2 min from decapitation to immersion in the high-sucrose cutting solution slush.
   1. Anesthetize the animal with the appropriate dosage of isoflurane (0.5 mL/1 L anesthesia chamber). Determine the depth of anesthesia through a paw pinch; confirm the lack of paw withdrawal reflex before proceeding.
   2. Transfer the animal to the tissue paper in the brain extraction workspace and decapitate it with surgical scissors.
   3. Insert iris scissors into the brain stem and keep the bottom scissors flush with the calvaria. Cut along the sagittal suture until the coronal suture is reached. Place iris scissors in eye sockets and cut through the metopic suture. Use curved forceps to move the sides of the calvaria down, exposing the whole brain.
      NOTE: Be careful with both the iris scissors and forceps to not puncture the brain while cutting through the sutures.
   4. Slide the brain with the blunt edge of the curved forceps into the 50 mL beaker with 30 mL of high-sucrose cutting solution slush. Let it remain for 1 min.
   5. Transfer the brain to the 90 mm plastic culture dish with chilled carbogenated cutting solution in the brain preparation workspace. Orient the brain for positioning in the agarose mold.
   6. Add a small dot of super glue to the rostral end of the agarose mold. Place the brain in the mold with the spatula. Ensure that the brain is placed with the dorsal side down for horizontal slicing.
      NOTE: The location of glue in the mold will change depending on the region of interest (ROI). For hippocampal-cortical (HC) and olfactory bulb (OB) slices, ensure that the OB is stabilized and the sides of the brain remain free from glue. Too much glue will affect slicing quality and cause tears during vibratome slicing.
   7. Move the specimen plate into the buffer tray, move the blade into position with the correct angle, and increase the buffer tray height to bring the blade as close as possible to the brain.
   8. Slice at 0.20 mm/s speed 300 µm intervals of HC and OB tissues, then collect them after each slicing round with a glass Pasteur pipette.
   9. Leave the slices in the aCSF-filled recovery chamber in a 32 °C water bath for 45 min, followed by 1 h at RT. Ensure slices do not overlap and are fully exposed to the carbogenated solution.
      NOTE: Be sure to maintain continuous carbogenation of all solutions and any chambers mentioned containing the solution. A pressure regulator may be used to sustain consistent carbogenation.

2. In-vitro human iPSC-based neuronal network on HD-MEA

NOTE: All iPSC neurons used in this study are commercially obtained (see Table of Materials). These human cells differentiated from stable iPS cell lines that were derived from human peripheral blood or fibroblasts.

1. Coating of HD-MEA chips for in-vitro human iPSC cell cultures (Figure 2B)
   1. Place the HD-MEA chip on the acquisition recording platform, fill the reservoir with PBS, and test the chip prior to coating. Start Brainwave software. Select File > New Recording Session. Set the recording parameters to have a Recording Frequency of 50 Hz and a Sampling Frequency of 18 kHz/electrode. Change the Amplifier Offset to calibrate the chip. See Table 2 for troubleshooting tips.
      NOTE: The recording frequency and sampling frequency parameters will depend on data type and individual system requirements.
   2. Sterilize and pre-condition HD-MEAs.
      1. Under the hood, wipe the chip and the glass ring with tissue moistened with 96% ethanol (EtOH), then place each device into a sterile 100 mm x 20 mm Petri dish and fill the MEA reservoir with 70% EtOH for 20 min.
      2. Aspirate the EtOH and wash the reservoir with sterile, filtered dd-water 3 times. Add 1 mL of pre-conditioning media and incubate overnight at 37 °C and 5% CO₂.
         NOTE: Pre-conditioning media needs to be a salt-based solution to make the HD-MEA surface more hydrophilic. This can include previously prepared BrainPhys (BP) complete media (not >3 months old) (Tables 1C).
   3. Coat HD-MEAs. The next day, aspirate pre-conditioning media. Add 1 mL of 0.1 mg/mL poly-dl-ornithine (PDLO) to coat the entire active area. Incubate at 37 °C overnight in an incubator.
   4. Prepare and warm media to RT. The protocols here exploit functional human iPSC neurons from two commercial sources; thus, media components vary for each supplier. One protocol is described in (Tables 1C, D) .
   5. Aspirate PDLO, wash 3 times with dd-water and let the chips dry under the hood for 10 min.
   6. Fill a 35 mm x 10 mm Petri dish with sterile, filtered dd-water and place it beside the chip to maintain proper humidity and avoid evaporation of the seeded cells in the next steps.
2. Plating and maintenance of human iPSC neurons in HD-MEAs (Figure 2B)
   1. Thaw and dilute cells to desired cells per microliter concentration (i.e., 1000 cells/µL to obtain 50,000 cell density in a 50 µL drop on the HD-MEA) (Table 1C).
   2. Pipette the cell suspension on the surface of the chip active area using high laminin dotting media (Table 1D).
   3. Incubate at 37 °C with 5% CO₂ for 45-60 min.
   4. Gently fill 2 mL of media in the HD-MEA reservoir (Table 1C).
   5. Perform 100% media change on day 1 (DIV1) post-seeding using RT media (Table 1C). Change 50% of the media every 3-4 days. Keep HD-MEAs incubated at 37 °C with 5% CO₂ throughout the experiment.
      NOTE: Pipette gently to avoid dislodging the cells. Check the color of the media for any contamination. The interval and amount of media change can be determined by individual study questions or cell needs/specifications.
   6. Optional: Check the progress of cell culture growth between DIV4-DIV8 under an upright differential interference contrast (DIC) microscope after cleaning the stage with >70% EtOH.

3. Ex-vivo and in-vitro large-scale neural recordings with HD-MEAs

1. Preparation of brain slice recording workspace (Figure 2A)
   1. While brain slices are recovering, place the required tools in each designated workspace (see Table of Materials).
      NOTE: The main system setup must be optimized and tested well before the brain slice experimental day. The perfusion system (inlet lines, pump outlet lines, tubing, and grounding) needs to be tested with PBS or aCSF and an HD-MEA on the recording platform to ensure a clean signal, increased signal-noise ratio, and lack of perfusion noise.
   2. Coat the HD-MEA chip with 0.1 mg/mL of PDLO to enhance tissue-chip coupling and incubate at 37 °C for 20 min.
   3. During chip incubation, fill the gravity-based perfusion system and lines with recording aCSF. Ensure continuous carbogenation of the perfusion system. Set a flow rate of 4.5 mL/min and a temperature of 37 °C.
   4. Place the HD-MEA chip on the acquisition recording platform, fill the reservoir with aCSF, test the perfusion system, and troubleshoot any remaining system noise.
      1. Start Brainwave software. Select File > New Recording Session. Set the recording parameters to have a Recording Frequency of 1 Hz and a Sampling Frequency of 14 kHz/electrode. Change the Amplifier Offset to calibrate the chip. See Table 2 for troubleshooting tips.
         NOTE: The recording frequency and sampling frequency parameters will depend on data type and individual system requirements.
   5. Ensure the recording area is dark through a room lighting system or a shaded cage on the optical table.
   6. Align the stereomicroscope with the HD-MEA chip reservoir and active area for image acquisition.
   7. Place the anchor in the chip reservoir to equilibrate.
      NOTE: Anchor is a custom-made platinum harp with minimal wires to promote oxygenation; however, some commercial ones are available.
   8. Add pharmacological compounds to the appropriate perfusion tubes.
      NOTE: In this protocol, both spontaneous and 100 µM 4-aminopyridine (4-AP) pharmacologically induced recordings were obtained as previously described. Pharmacological compounds can be tailored to the specific study question.
   9. In the brain slice preparation workspace, place a new 90 mm plastic culture dish in a 150 mm glass Petri dish. Add aCSF and begin carbogenating.
2. Circuit-wide recordings from HC and OB slices using HD-MEAs
   NOTE: Slice coupling should be performed as quickly as possible to avoid a lack of oxygenation to the slice. Coupling should only take ~1 min from the initial placement of the microdissected slice on the chip active area to the final perfusion system startup.
   1. Remove the slice from the brain slice recovery chamber with a glass pipette and place it in a 90 mm plastic culture dish with continuous carbogenation. Using a microdissection tool, isolate the HC or OB from the surrounding brain slice tissue.
   2. Move the isolated HC or OB acute slices with a glass pipette into the HD-MEA reservoir. Gently align the slice on the MEA active area with a fine brush. Suck all solutions from the HD-MEA chip well with an aspiration system.
   3. Place the anchor gently on top of the slice using forceps.
      NOTE: The anchor should be placed without slice movement to avoid the loss of coupling.
   4. Gently add solution to the chip reservoir and start the perfusion system.
      NOTE: Ensure laminar flow from the perfusion inlet and pump outlet for optimal recording parameters.
   5. Ensure the recording area is adequately dimmed through the room lighting system or with a shaded cage on an optical table setup.
   6. Allow the slice to acclimate for 10 min before beginning recordings or additional pharmacological modulation.
   7. Start Brainwave software. Select File > New Recording Session. Set the recording parameters to have a Recording Frequency of 1 Hz and a Sampling Frequency of 14 kHz/electrode. Change the Amplifier Offset to calibrate the chip.
      NOTE: As denoted earlier in section 3.1.4.1, while performing system tests, make sure to apply these same recording parameters.
   8. Press Record to begin the acquisition with the preset experimental conditions.
   9. Immediately following the final recording, capture light imaging of the acute brain slice. Move the slice back to the slice recovery chamber, remove any organic material coupled to the chip with a brush, and continue with the next slice. Clean the HD-MEAs as described in section 3.4.
3. Preparation of human iPSC recording workspace and network-wide recordings on HD-MEA (Figure 2B)
   NOTE: Change media either the day before recording or immediately following human iPSC recording (Table 1C). In studies using the functional neurons, media was changed every 4 days, and on 4, 8, 16, and 24 DIV media is changed immediately following iPSC recordings.
   1. Ensure a sterile work environment by cleaning the HD-MEA acquisition platform with >70% EtOH.
   2. Gently place a polydimethylsiloxane (PDMS)-base cap with reference on the HD-MEA ring under the hood. Move the HD-MEA chip to the iPSC recording workspace and attach the HD-MEA chip to the acquisition platform.
   3. Ensure the recording area is adequately dimmed through a room lighting system or a shaded cage on an optical table.
   4. Allow the HD-MEA chip to equilibrate for 10 min prior to beginning recordings or additional pharmacological modulation.
   5. Start Brainwave software. Select File > New Recording Session. Set the recording parameters to have a Recording Frequency of 50 Hz and a Sampling Frequency of 18 kHz/electrode. Change the Amplifier Offset to calibrate the chip.
      NOTE: As denoted earlier in section 2.1.1, while performing a system test prior to coating and plating, make sure to apply these same recording parameters.
   6. Record the spontaneous firing activity or pharmacologically induced responses from the human iPSC network on each day of the experimental plan (i.e., 4, 8, 16, 24 DIVs).
      ​NOTE: Do not let the chip remain outside the incubator for >30 min to maintain stable temperature and humidity and prevent any temperature shock to cells.
   7. Incubate HD-MEAs at 37 °C with 5% CO₂ over the course of the experiment.
   8. Following experiment completion, fix the neuronal network on chips and stain for further optical imaging or clean the HD-MEAs directly, as described in step 3.4.
4. Cleaning of HD-MEA chips
   1. After the experiment, discard the solution according to proper waste disposal and rinse with dd-water.
   2. Add the detergent of choice, clean the active area and entire reservoir with a Q-tip, and discard the detergent. Refill with detergent, incubate for 20 min, then discard detergent.
   3. Thoroughly rinse with laboratory-grade water. Then, rinse 3-4 times with dd-water.
   4. Use air pressure to dry the HD-MEA chip thoroughly.

4. Analysis of large-scale neural recordings from HD-MEAs

NOTE: While step 4.1 is Brainwave software specific, step 4.2 can be modified based on each user's commercially available HD-MEA device type.

1. Raw data preprocessing and event detection
   1. Open a recorded raw data file (.brw) in Brainwave software. Select Analysis > LFP Detection or Spike Detection.
      NOTE: LFP Detection employs IIR filtering with a low pass 4^th order Butterworth filter (1-100 Hz). Hard threshold algorithms include a high threshold of 150 µV, a low threshold of -150 µV, an energy window between 70-120 ms, a refractory period of 10 ms, and a maximum event duration of 1 s. Single and MUA Spike Detection employs IIR filtering with a high pass 4^th order Butterworth filter (300-3500 Hz). A PTSD algorithm is applied with a standard deviation factor of 8, a peak lifetime period of 2 ms, and a refractory period of 1 ms.
   2. For HC and OB circuit recordings, add the Advanced Workspace option in the detected event file (.bxr) to import the structural light image captured from the stereomicroscope. When examining large-scale HC circuitry, create structural layers containing the dentate gyrus (DG), hilus, Cornu Ammonis 1 (CA1), Cornu Ammonis 3 (CA3), entorhinal cortex (EC), and perirhinal cortex (PC). When examining large-scale OB circuitry, create structural layers containing the olfactory nerve layer (ONL), glomerular layer (GL), external plexiform layer (EPL), mitral cell layer (MCL), and granule cell layer (GCL). Consider the EPL and the MCL as the projection layer (PL), including the olfactory cortex (OCx).
2. Data processing with a custom Python computational pipeline
   1. Denoising
      1. Read .bxr file using a custom-written Python script^26,29,32 and h5py 3.6.0 python package.
      2. Extract spike trains pertaining to the iPSC network recordings and LFP event trains pertaining to the HC and OB brain slice circuit recordings.
      3. Characterize events with a total number of active electrodes less than 0.1% or 10% of the average active electrodes per average event or detected events falling outside a statistically reasonable firing rate range as random events and remove them. Additionally, apply amplitude and event duration threshold values.
         NOTE: For the firing rate range, 0.1-15 spikes/s and 0.1-60 LFP events/min are considered. These are example rate threshold values used for the analyzed datasets. The rate, amplitude, and duration thresholds will depend on individual data.
      4. Save the resultant event train data with the accompanying spatiotemporal information in .npy file format.
   2. Rastergrams
      1. Read filtered event .npy and .bxr files and generate a raster plot using the Matplotlib pyplot function (https://matplotlib.org/3.5.3/api/_as_gen/matplotlib.pyplot.html).
      2. Additionally, for brain slice recordings with layer specificity, sort and group the electrode IDs based on the layers produced in step 4.1.2.
   3. Mean firing activity
      1. Process the time series data from the .bxr file, calculating each electrode's average firing rate (number of events/recording time).
      2. Construct a data matrix where the rows and columns represent the coordinates of electrodes in the HD-MEA 64 x 64 array, where each matrix value signifies the mean firing rate.
      3. Employ a plotting library such as Matplotlib's imshow or Seaborn's heatmap functions in Python.
      4. Employ the 'hot' color map here, creating an informative heatmap that visually encapsulates the spatial distribution of mean firing rates across the electrode array.
   4. Representative waveform traces
      1. Read time series data from the .brw file and generate a waveform trace using Matplotlib pyplot function. (https://matplotlib.org/3.5.3/api/_as_gen/matplotlib.pyplot.html).
      2. Input desired electrode ID, time bin, and frequency band for a representative waveform trace. Frequency bands defined in these analyses include low-frequency LFP oscillations (1-100 Hz) with bandpass filtered δ, θ, β, and γ frequency bands; sharp wave ripples (SWR) (140-220 Hz); and high-frequency single and MUA (300-3500 Hz). The frequency bands δ, θ, β, and γ are 1-4 Hz, 5-12 Hz, 13-35 Hz, and 35-100 Hz respectively.
   5. Power spectral density
      1. Read time series data from the .brw file and compute the periodograms to discern the dominant frequencies underlying the oscillatory activity within each time series.
      2. Construct pseudo-color spectrograms of the frequency-time dynamics.
         NOTE: Spectra are calculated using Welch's method by utilizing the Fast Fourier Transform of recorded LFPs to estimate spectral power density⁴¹.
      3. Input desired electrode ID, time bin, and frequency band for a spectral density map. Frequency bands defined in these analyses include those described in step 4.2.4.
   6. Functional connectivity
      1. For brain slice circuit recordings, follow steps 4.2.6.2-4.2.6.4.
      2. Read time series data from the .brw file and calculate the cross-covariance between pairs of active electrodes in the 64 x 64 array employing Pearson's correlation coefficient (PCC)⁴².
      3. Fit a vector autoregressive model to the time series using Multivariate Granger causality to quantify the influence of one time series on another.
      4. Apply directed transfer function (DTF) to assess directional information flow within the correlated links.
         NOTE: Functional connectivity in the multilayered network is established by setting a correlation value threshold based on those over the mean and two standard deviations of all cross-covariance values^43,44.
      5. For iPSC recording, follow steps 4.2.6.6-4.2.6.8.
      6. Read spiketrains data from the .bxr file and compute a 64×64 matrix of the PCC correlation coefficients between all combinations of binned spike trains using spike_train_correlation functions (https://elephant.readthedocs.io/en/v0.7.0/reference/spike_train_correlation.html).
         NOTE: Functional connectivity in the multilayered network is established by setting a correlation value threshold based on those over the mean and two standard deviations of all cross-covariance values.
      7. Furthermore, implement spatial-temporal filters (STF) and distance-dependent latency thresholds (DdLT) filtering procedures on the connectivity matrix to eliminate potential paired connections exceeding the maximum propagation velocity (set at 400 mm/s)⁴⁵.
      8. Extract negative peaks from resultant cross-correlation matrices with filtering and thresholding operations to identify the inhibitory connections using filtered and normalized cross-correlation histogram (FNCCH) algorithm⁴⁵.
      9. Transform each connectivity matrix into a dynamic graph (.gexf) file.
   7. Network connectivity maps
      1. Open data laboratory in Gephi program 9.2 version (https://gephi.org) for the dynamic graph to plot specific time bins.
      2. Apply Geo Layout in the layout window for spatial mapping.
      3. Place parameter constraints on Degree Range and Edge Weight for comparison.
      4. Assign Nodal Color, Edge Size, and Degree Size for better visualization.