Cadherin-13 is a critical regulator of GABAergic modulation in human stem-cell-derived neuronal networks

Activity in the healthy brain relies on a concerted interplay of excitation (E) and inhibition (I) via balanced synaptic communication between glutamatergic and GABAergic neurons. A growing number of studies imply that disruption of this E/I balance is a commonality in many brain disorders; however, obtaining mechanistic insight into these disruptions, with translational value for the patient, has typically been hampered by methodological limitations. Cadherin-13 (CDH13) has been associated with autism and attention-deficit/hyperactivity disorder. CDH13 localizes at inhibitory presynapses, specifically of parvalbumin (PV) and somatostatin (SST) expressing GABAergic neurons. However, the mechanism by which CDH13 regulates the function of inhibitory synapses in human neurons remains unknown. Starting from human-induced pluripotent stem cells, we established a robust method to generate a homogenous population of SST and MEF2C (PV-precursor marker protein) expressing GABAergic neurons (iGABA) in vitro, and co-cultured these with glutamatergic neurons at defined E/I ratios on micro-electrode arrays. We identified functional network parameters that are most reliably affected by GABAergic modulation as such, and through alterations of E/I balance by reduced expression of CDH13 in iGABAs. We found that CDH13 deficiency in iGABAs decreased E/I balance by means of increased inhibition. Moreover, CDH13 interacts with Integrin-β1 and Integrin-β3, which play opposite roles in the regulation of inhibitory synaptic strength via this interaction. Taken together, this model allows for standardized investigation of the E/I balance in a human neuronal background and can be deployed to dissect the cell-type-specific contribution of disease genes to the E/I balance.


Cadherin-13 is a critical regulator of GABAergic modulation in human stem cell derived neuronal networks
week using an enzyme-free dissociation reagent (ReLeSR, Stem Cell Technologies).

RNA interference
For RNAi knockdown experiments, DNA fragments encoding shRNAs directed against human CDH13 mRNA (Sigma, Supplementary table 8 and 9) were cloned into the pTRIPΔU3-EF1α-EGFP lentiviral vector. Empty vector expressing GFP only was used as control vector. Lentiviral particles were prepared from both shRNA expressing vectors and empty vector, and tittered as described previously 10 .

Optogenetics
Optogenetic activation of E/I networks were performed at DIV 49 using the MW24-opto-stim LED cap for the Multiwell-MEA system (Multichannel Systems). Stimulation of cultures was conducted as follows: 200ms, 30.00mA, 470nm LED light pulses were delivered to each well. Inter stimulus interval was set at 5 seconds (onset to onset), and the stimulation protocol was repeated 24 times for a total duration of 2 minutes. Pre stimulation condition in Supplementary figure 4i and j represents the MFR normalized to 50 ms pre-stimulation baseline activity. Post stimulation condition represents the activity in a window between 10-30 ms after stimulus onset. Both pre and post stimulation responses on the level of the MFR were normalized to pre-stimulation condition.

Animals
The rodent astrocytes presented in this study were harvested embryonic (E18) rat brains (Wistar Wu) as previously described 2,11,12 . All experiments on animals were carried out in accordance with the approved animal care and use guidelines of the Animal Care Committee, Radboud University Medical Centre, the Netherlands, (RU-DEC-2011-021, protocol number: 77073).

Compound application
Picrotoxin (PTX, Tocris Cat No 1128) and Bicuculline (BIC, Sigma Cat No B6889) were prepared fresh into concentrated stocks and stored frozen at −20°C (PTX 50 mM in ETOH (MEA) or 100 mM in DMSO (single-cell recordings); BIC 20 mM in DMSO). For all experiments on MEAs an aliquot of the concentrated stock PTX or BIC was first diluted 1:2 in room temperature DPBS and vortexed briefly.
Then, 2.5 µl working dilution was added directly to the cell culture medium (500 µl) to reach a 100 µM concentration for PTX, and 40 µM concentration for BIC. ETOH or pre-diluted DMSO were used as vehicle. For all single cell experiments, PTX was directly diluted 1000 x in artificial cerebrospinal fluid (ACSF). The DMSO concentration in the ACSF was always ≤0.05% v/v. All experiments were performed at 37°C.
The GABA reversal was measured using a cesium-based intracellular solution containing (in mM) 115 CsMeSO3, 20 CsCl, 10 HEPES, 2.5 MgCl2, 4 Na2ATP, 0.4 Na3GTP, 10 sodium phosphocreatine, 0.6 EGTA (pH 7.2, mOsmol 290). For sucrose application, cells were recorded with the KCl based solution described before. GABA (10 mM dissolved in ACSF) was applied locally at a distance of 10-20 µm from the soma of the patched excitatory neuron using a PDES-2DX pressure ejection system (NPI, Tamm, Germany). Micropipettes used for compound application had a resistance of 3-5 MΩ. Injection pressure was set at 7psi/0.5 bar and injection duration was set to 100 ms. Analysis of peak response and reversal potential was conducted using Clampfit 10.7.

Immunocytochemistry
Cells were fixed and stained as described before 2 . All antibodies are listed in Supplementary table 13. Neurons were generally fixated at DIV 49, and at DIV 35 and DIV 49 for membrane expression of NKCC1 and KCC2. When membrane expression of NKCC1 and KCC2 was examined, coverslips were not permeabilized. We imaged at a 20x magnification to count the number of GABAergic subtypes and at a 63x magnification for all other measures using the Zeiss Axio Imager Z1 equipped with apotome. Images in figure 5 c and e were taken with the Zeiss AxioObserver Z1 with AryScan. All conditions within a batch were acquired with the same settings in order to compare signal intensities between different experimental conditions. Fluorescent signals were quantified using FIJI software. The intensity of NKCC1 or KCC2 expression on the cell membrane was calculated by: integrated density -(Area of selected cell X Mean fluorescence of background readings). The number of synaptic puncta was determined per individual cell via manual counting and divided by the dendritic length of the dendrite.
VGAT puncta intensity was determined using particle analysis in the FIJI software. All analysis was performed blinded for genotype using an open source random names application.

Quantification of mRNA by RT-qPCR
RNA samples were isolated using Nucleospin RNA isolation kit 740955.250) according to the manufacturer's instructions. RNA samples were converted into cDNA by iScript cDNA synthesis kit (BIO-RAD, 1708891). CDNA products were cleaned up using the Nucleospin Gel and PCR clean-up kit (Macherey-Nagel, 740609.250). Human-specific primers were designed with Primer3plus (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi) and IDT PrimerQuest (https://eu.idtdna.com) tools, respectively. Primer sequences are given in supplementary table 9. QPCRs were performed in the Quantstudio 3 apparatus (Thermo Fisher Scientific) with GoTaq qPCR master mix 2X with SYBR Green (Promega, A600A) according to the manufacturer's protocol. The qPCR program was designed as following: After an initial denaturation step at 95ᵒC for 10 min, PCR amplifications proceeded for 40 cycles of 95ᵒC for 15 s and 60ᵒC for 30 s and followed by a melting curve. All samples were analyzed in duplicate in the same run, placed in adjacent wells. The arithmetic mean of the Ct values of the technical replicas was used for calculations. Relative mRNA expression levels were calculated using the 2^-ΔΔCt method with standardization to housekeeping genes 13 .

RNA-sequencing
RNA was isolated from three biological replicates of E/I networks composed of iGLU#2 and iGABA#1 index]TTTTTTTTTTTTTTTTTTTTTTTTTTTTTTVN-3ʹ, where "N" is any base and "V" is either "A", "C" or "G"; IDT) in a tube containing 7 µL Vapor-Lock (Qiagen, 981611) to prevent evaporation.

Pre-processing of RNA-seq data
Base calls were converted to fastq format and demultiplexed using Illumina's bcl2fastq conversion software (v2.16.0.10) tolerating one mismatch per library barcode. Reads were filtered for valid unique molecular identifier (UMI) and sample barcode, tolerating one mismatch per barcode. Trimming of adapter sequences and over-represented sequences was performed using Trimmomatic (version 0.33) 15 .
Trimmed reads were mapped to a combined human (GRCh38.p12) and rat (Rnor_6.0) reference genome to separate reads belonging to the human iNeurons from reads originating from the rat astrocytes.

Supplementary tables
Supplementary    Fig. 2g, i, t: Sample size for DIV 28 n=42, DIV 35 n=40, DIV 49 n=44 cells from 3 batches. All data represent means ± SEM. Two-way ANOVA with Tukey correction for multiple testing was used to compare between DIVs. DIV= Days in vitro.  Figure 3g, j and k and Supplementary figure 4b, f-j. All data represent means ± SEM. * p < 0.05; ** p < 0.01; *** p < 0.001. Paired T-test or Wilcoxon matchedpairs signed rank test was performed between network activity pre, and post treatment in. Basal = before treatment, PTX= Picrotoxin, BIC= Bicuculline, NBR= Network burst rate, NBD= Network burst duration, MFR= Mean firing rate, n= number of wells, DIV= Days in vitro.  of supplementary figure 5c. All data represent means ± SEM. Mann-Whitney test was performed between control and CDH13-shRNA#1+2 transduced networks. Significance was corrected for multiple comparisons using Bonferroni. Paired T-test was performed between CDH13-shRNA#1, CDH13-shRNA#2 and controls. N represents the number neuronal preparations from which one sample is isolated for analysis.

Gene
Forward ( Figure 4d, f, j and Supplementary figure 5e, i, j. All data represent means ± SEM. * p < 0.05; ** p < 0.01; *** p < 0.001. Total VGAT intensity, MEA parameters and sIPSCs from iGABA#5 vs iGABA#5-KO were compared using Mann-Whitney ranked sum test with post hoc Bonferroni correction. Ordinary one-way ANOVA with Dunnett correction for multiple testing or Kruskal Wallis ANOVA with Dunn's correction for multiple testing was used to compare between CDH13-shRNA transduced wells and controls. NBD= Network burst duration, MFR= Mean firing rate, PRS= percentage of random spikes, NBR= Network Burst Rate, n= number of wells, DIV= Days in vitro.  Figure 5r and Supplementary figure 5m. All data represent means ± SEM. * p < 0.05; ** p < 0.01; *** p < 0.001. Paired T-test was performed between pre and post-treatment conditions at DIV 49. Basal = before treatment, DIV= Days in vitro.