Astrocytes Regulate Neuronal Network Burst Frequency Through NMDA Receptors in a Species- and Donor-Specific Manner

Background Development of synaptic activity is a key neuronal characteristic that relies largely on interactions between neurons and astrocytes. Although astrocytes have known roles in regulating synaptic function and malfunction, the use of human- or donor-specific astrocytes in disease models is still rare. Rodent astrocytes are routinely used to enhance neuronal activity in cell cultures, but less is known about how human astrocytes influence neuronal activity. Methods We established human induced pluripotent stem cell–derived neuron-astrocyte cocultures and studied their functional development on microelectrode array. We used cell lines from 5 neurotypical control individuals and 3 pairs of monozygotic twins discordant for schizophrenia. A method combining NGN2 overexpression and dual SMAD inhibition was used for neuronal differentiation. The neurons were cocultured with human induced pluripotent stem cell–derived astrocytes differentiated from 6-month-old astrospheres or rat astrocytes. Results We found that the human induced pluripotent stem cell–derived cocultures developed complex network bursting activity similar to neuronal cocultures with rat astrocytes. However, the effect of NMDA receptors on neuronal network burst frequency (NBF) differed between cocultures containing human or rat astrocytes. By using cocultures derived from patients with schizophrenia and unaffected individuals, we found lowered NBF in the affected cells. We continued by demonstrating how astrocytes from an unaffected individual rescued the lowered NBF in the affected neurons by increasing NMDA receptor activity. Conclusions Our results indicate that astrocytes participate in the regulation of neuronal NBF through a mechanism that involves NMDA receptors. These findings shed light on the importance of using human and donor-specific astrocytes in disease modeling.


Astrocyte differentiation
The astrocyte differentiation was done as previously described (2,3).Briefly, neural induction was performed by culturing hiPSCs with 10 μM SB and 200 nM LDN.After 10 days of induction, NPCs arranged in rosettes were manually picked and transferred to ultra-low attachment plates, where the NPCs formed spheres.These spheres were cultured and expanded with bFGF (100-18B, Peprotech) for 6-9 months, after which they were dissociated and plated in co-cultures with neurons at 1:1 density.Astrocytes from different cell lines were cultured concurrently and represented the same age when plated.The astrocytes used in this study have been fully characterized before (3).

Image analysis
The image analysis for neuronal characterization was done with ImageJ (NIH).The neurons were labeled with GFP to separate them from astrocytes during the analysis, and the astrocytes were identified based on S100β expression.The images were thresholded using the default option to select for the cells expressing the target protein.The image calculator was used to co-localize the selected cells with a reference marker.DAPI was used as a reference for all cells and GFP was used as a reference for neurons.The analyze particles feature was used to quantify the number of cells expressing the proteins of interest.

MEA recordings
The electrophysiological activity was recorded with Maestro Edge MEA system using AxIS Navigator software and 24-well CytoView plates containing 16 electrodes (Axion Biosystems).The recordings were performed at 37 °C temperature in a 5% CO2 atmosphere.To start the measurements, the well plate was placed in the MEA system and the temperature and CO2 were allowed to stabilize for 10 min.The baseline activity was measured for 10 min.For pharmacological tests, 10 μM NBQX (N138, Sigma), 25 μM D-AP5 (A8054, Sigma), 100 μM GABA (Sigma), 10 μM Ifenprodil (I2892, Sigma) and 3 μM TCN-201 (SML0416, Sigma) were used according to previous studies (1,(6)(7)(8).The baseline activity was recorded for 10 min prior to treatment.After this, the pharmacological treatments were carried out by pipetting 5 μl of the compound into wells containing 500 μl of culture media.The plate was then placed back in the MEA device and incubated for 10 min before a 10 min recording was started.

MEA data analysis
The AxIS Navigator software was used for spike sorting during the recordings.The spike threshold was set to 5 x the standard deviation of the estimated noise.The burst detection was done using Neural Metric Tool (Axion biosystems).The minimum number of spikes per burst was set to 5 and the maximum inter-spike interval (ISI) within a burst was set to 100 ms.For NB detection, the Envelope algorithm was used due to its ability to merge repetitive sub-bursts within a network event into a single burst.A threshold factor value 2 and minimum inter-burst interval (IBI) 100 ms were selected for the analysis.The minimum number of electrodes in NB was set to 25 % and a burst inclusion value was set to 75%.The Envelope algorithm was not suitable for the analysis of NBQX-treated samples that contained a great amount of non-synchronous bursting activity.Instead, we used Adapted algorithm that more efficiently separated the NBs from non-synchronous bursts.The minimum number of spikes per NB was set to 50 and the minimum number of electrodes in NB was set to 25 %.The NBs in the patient lines were analyzed using the max ISI algorithm that performed the best in terms of defining NB duration in samples with short bursts and bursting outside the NBs.
The max ISI value was set to 50 ms and the minimum number of spikes per NB was set to 50.
NeuroExplorer (Plexon) software was used to analyze high frequency bursting activity in the samples.First, spike-sorted .spkfiles generated by the AxIS Navigator software were converted to rate histograms using a 0.025 s bin width in NeuroExplorer.Electrodes with fewer than 5 spikes per minute were removed from the files.Power spectral densities (PSD) were drawn for each electrode using Welch's windowing function.The same function was used for drawing the spectrograms.For visualization of spectrograms and PSDs, Log of PSD (dB) normalization was used, and for the analysis of high frequency bursting, raw PSD values were used.The power of the signal at frequencies 0.5-4 Hz, 4-8 Hz and 8-12 Hz was acquired from the PSDs for each electrode by averaging values within the defined frequency range.Finally, the values for each sample were acquired by averaging the values from the electrodes.The results for the power of high frequency bursting were presented as a ratio to the network burst frequency (NBF) in each well.

qRT-PCR
RNA was extracted from neuron-astrocyte co-cultures at 5 weeks using RNeasy Mini kit (74104, Qiagen) following the manufacturer's instructions.The extracted RNA was eluted in nuclease-free water.The cDNA conversion was performed with Maxima reverse transcriptase enzyme approach using Random hexamer primer (S0142, Fermentas), 10 mM dNTP (R0192, Fermentas)

Supplemental Figure 4 :
Characterization of neuronal D-AP5 response in co-cultures with rat astrocytes at 4 weeks of differentiation.A-E.The NMDA receptor blockage had a significant effect on NBF and mean ISI within NB in the mixed-species cultures.F-G.The hiPSC-derived cultures displayed a stronger response to D-AP5 than the mixed-species co-cultures.(n = 5 cell lines, data was collected from 1 experiment.For A-E: Paired t test was used for the statistical comparisons, normal distribution of the data was verified with Kolmogorov-Smirnov test.For F-G: Mann-Whitney U test was used for the statistical comparisons.* signifies p < 0.05, ** signifies p < 0.01, ns = nonsignificant) rat-specific Grin1 subunits in hiPSC-derived and mixed-species co-cultures.E-F.ICC quantification of GRIN1 subunit expression in hiPSC-derived and mixed-species co-cultures revealed higher NMDA receptor expression in neurons compared to astrocytes in both culture conditions.(n=5 cell lines, data was collected from 1 experiment, Mann-Whitney U test was used for the statistical comparisons, * signifies p < 0.05, ** signifies p < 0.01, ns = non-significant) twins discordant for treatment-resistant schizophrenia.A-B.Neurons in cultures derived from affected twin (AT), unaffected twin (UT) and control cultures (CTR) were CUX1 and MAP2 positive.C-D.PAX6 and PRPH expressing cells were not detected in the cultures.E-F.Approximately 50% of the cells in all co-culture conditions expressed astroglial marker S100β and on average 20 % of the cells expressed GFAP. (n = 3 cell lines per group, data was collected from 2 independent experiments, each data point represents 2-4 replicate samples) subunit-dominant NMDA receptors in human induced pluripotent stem cell-derived neurons.Sci Rep 6. https://doi.org/10.1038/srep238379. Muller PY, Janovjak H, Miserez AR, Dobbie Z (2002): Processing of gene expression data generated by quantitative real-time RT-PCR.Biotechniques 32.