Setd5 haploinsufficiency alters neuronal network connectivity and leads to autistic-like behaviors in mice

SETD5, a gene linked to intellectual disability (ID) and autism spectrum disorder (ASD), is a member of the SET-domain family and encodes a putative histone methyltransferase (HMT). To date, the mechanism by which SETD5 haploinsufficiency causes ASD/ID remains an unanswered question. Setd5 is the highly conserved mouse homolog, and although the Setd5 null mouse is embryonic lethal, the heterozygote is viable. Morphological tracing and multielectrode array was used on cultured cortical neurons. MRI was conducted of adult mouse brains and immunohistochemistry of juvenile mouse brains. RNA-Seq was used to investigate gene expression in the developing cortex. Behavioral assays were conducted on adult mice. Setd5+/− cortical neurons displayed significantly reduced synaptic density and neuritic outgrowth in vitro, with corresponding decreases in network activity and synchrony by electrophysiology. A specific subpopulation of fetal Setd5+/− cortical neurons showed altered gene expression of neurodevelopment-related genes. Setd5+/− animals manifested several autism-like behaviors, including hyperactivity, cognitive deficit, and altered social interactions. Anatomical differences were observed in Setd5+/− adult brains, accompanied by a deficit of deep-layer cortical neurons in the developing brain. Our data converge on a picture of abnormal neurodevelopment driven by Setd5 haploinsufficiency, consistent with a highly penetrant risk factor.

for dilutions in 1% BSA in PBS, overnight, 4°C, followed by 3x PBS washes), secondary antibody, and nuclear staining (4',6-Diamidino-2-Phenylindole (DAPI), 0.1µg/mL in PBS, 10 minutes, room temperature). Cortical layer thickness measurements were performed as described previously 2 , as the average between measurements perpendicular to the cortical surface taken at 30% and 70% from the dorsal midline. For cells stained by immunocytochemistry, for morphological analysis neurons were analyzed by Neurolucida Explorer (MBF Bioscience, Williston, VT, USA) for the metrics of soma size, total neurite length and number, and Sholl analysis. Synaptic puncta were counted as colocalized overlapping HOMER1 + and VGLUT1 + puncta along randomly selected MAP2 + neurites as described previously 3 . To ensure standardization of analysis, only cells with a pyramidal soma shape, with at least 2 neurites emanating from the cell body were included. As a quantification of cell death, cells were also processed by terminal deoxynucleotidul transferase dUTP nick end labeling (TUNEL, Click-iT TUNEL Imaging Assay, Life Technologies, Carlsbad, CA, USA). Briefly, fixed permeabilized cells were exposed to the terminal deoxynucleotidyl transferase enzyme, which can incorporate a modified dUTP nucleotide into damaged cellular DNA. Cells were then exposed to the click copper-catalyzed detector and imaged by fluorescence microscopy, with labeled cells considered to be apoptotic. Statistical comparisons were conducted by unpaired, 2-tailed t-test. To avoid confounding from the GFP transgene in Setd5 +/tissues or primary cells, no-secondary antibody negative controls were employed in the 488 nm channel in all experiments. While GFP could be detected by staining with an anti-GFP primary antibody, no endogenous GFP expression was observed. Stained sections were mounted (Fluoromount) for fluorescence microscopy (Zeiss Apotome Microscope, Oberkochen, Germany). Sample size was selected from an initial pilot study of neurons from n=4 animals per genotype.
Electrophysiology MEA: Cells analyzed by MEA electrophysiology were plated at 250k cells/MEA of 12w plate (Axion Biosystems, Atlanta, GA, USA) and cultured up to 32 DIV. Beginning at 8 DIV, recordings were conducted every 1-3 days through 32 DIV on the Axion Maestro system (Axion). Recordings measured spontaneous electrical spiking activity, 5.5 standard deviations (s) above noise through a 0.1-5000 Hz filter, proceeded 10min, and discarded the initial 60 seconds (s) from the analysis. Raw data were analyzed by the NeuralMetrics tool (Axion). Weighted mean firing represented the mean firing rates from only active electrodes (5 spikes/min); normalized burst frequency defined as 5 spikes with maximum inter-spike interval (ISI) of 100 µs; electrodes per burst defined as the number of electrodes participating in a single bursting event; and synchrony index as a unitless measure of firing synchrony, between 0-1, with 1=greatest synchrony, as a metric of inter-electrode cross correlogram of firing events (Supplemental Table 2). For local field potential (LFP) computation, raw 64-channel MEA recordings (12500 Hz) were low-pass filtered (300 Hz) and down-sampled to 1000 Hz (MATLAB resample.m).
Power spectral density (PSD) were computed using Welch's method (pwelch.m) with window length of 2-seconds and step length of 1-second. For each culture, a single recording yields a 64-by-1000 power matrix, where 64 represents the number of MEA channels and 1000 represents the number of frequencies (0-500Hz at 0.5Hz resolution).
For power comparison over time, baseline spectral power was taken to be the first day of recording (3 DIV), and all subsequent PSDs were normalized per frequency to 3 DIV and logged to compute log 10 power ratio. First recording, by definition, has log power ratio of 0 at every frequency. Band-specific log 10 power ratio were computed as the following: first take the normalized median log 10 power ratio over the specified frequencies (1-10 Hz for low frequency, 100-150 Hz for high frequency), then compute the median over all 64 channels. Finally, compute the mean and standard deviation over samples from the same condition. 2-way ANOVA (anova2.m) was run on the above genotype-pooled power ratios over time. At the end of the experiment, since typical action potentials are triggered by voltage-gated sodium channels, tetrodotoxin (TTX,10nM)  minutes to achieve a complete de-esterification of the dye. After the incubation with the dye, neurons were washed 3 times with the conventional media and were stored in the CO 2 incubator for additional 10 min to recover. During experiments, we recorded the spontaneous network activity of wt or het neurons for the same period of time, using a live-cell imaging chamber with CO 2 and 37°C incubation. Acquisition and analysis were performed using Zen Pro software (Zeiss) and FIJI 2 software. To extract the data from raw movies, a user manually determined regions of interest (ROIs) that exhibited considerable activity-driven intensity changes, and plotted changes in intensity vs. time. Caulk ® , Milford, DE, USA). Mice were left unrestrained for 5 days to allow for recovery from surgery before further manipulation, and prolonged EEG recording were performed. Continuous EEG recording and video monitoring was performed on each mouse. During EEG monitoring, mice had free access to food and water ad libitum. EEG (CEEG) was monitored for 16 hours/day for 3 consecutive days. EEG data were acquired and analyzed by using a Vangard system (Vangard, Cleveland, OH, USA).
EEG sampling rate was 100 Hz, and the recording filter was set at 5.3 Hz and 30 HZ for low-and high-frequency, respectively. The correlations between EEG events and overt behavioral responses were assessed by using split screen video monitoring system.
Epileptic spikes and seizure activities were analyzed from EEG data. Epileptic spikes (ictal spikes) were defined as paroxysmal electrical activity: they last 20-150 ms and their amplitude is greater than two times of average background EEG activity. Seizure activity is scored when epileptic spikes persist longer than 10 seconds. Seizure frequency and duration was measured. Once EEG recording was complete, mice were sacrificed and the location of electrodes was confirmed with H&E staining. Sample size was chosen as an initial pilot experiment to detect any genotype-dependent effects.

RNA-Seq/scRNA-Seq
Sorting: Primary cortical neurons from E18.5 fetuses were harvested into suspension All sequencing was conducted using an Illumina Hi-Seq 4000 sequencer using singleend 75bp reads. All data were aligned to the mm10 assembly of the mouse genome, and all subsequent data analysis was performed using HOMER, and detailed instructions for analysis can be found at http://homer.salk.edu/homer/. Each sequencing experiment was normalized to total of 10 7 uniquely mapped tags by adjusting the number of tags at each position in the genome to the correct fractional amount given the total tags mapped. Sequence experiments were visualized by preparing custom tracks for the UCSC genome browser. Differentially expressed genes were identified using HOMER as described previously 5 . Gene ontology was performed using the web based functional annotation tool Metascape 6 .
sc-RNA-Seq: Library construction and sequencing: Primary cortical neurons from E18.5 fetuses were isolated as described above. Single cell suspensions were loaded onto the 10X Genomics Chromium Controller instrument to generate single cell GEMs, targeting 5000 cells per sample. GEM-RT and library construction were performed following the 10X Genomics Protocol (Single Cell 3' Reagent Kits V2). Library fragment size distributions was determined using an Agilent Bioanalyzer High Sensitivity chip, and library DNA concentrations were determined using a Qubit 2.0 Fluorometer (Invitrogen).
Libraries were sequenced to a depth of approximately 50,000 paired end reads per cell using an Illumina HiSeq4000.
Data Preprocessing and Quality Control: Data was mapped to the mouse mm10 genome and aggregated using Cell Ranger V2.1 (10X Genomics) 7 . Data was filtered, processed, and analyzed using the Seurat R toolkit version 2.3.0 for single cell genomics 8 . Initial pre-processing thresholds were set to include genes detected in at least 3 cells and cells with at least 200 genes. Additional pre-processing was performed after plotting distributions of number of genes (nGenes), number of unique molecular identifiers (nUMI), and percent mitochondrial transcripts (percent.mito) per cell. The following thresholds were then applied: nGene > 600 and < 5000, nUMI > 500 and < 15000, and percent.mito < 0.06. A total of 6562 cells passed these thresholds and were used for downstream analyses. Data was then normalized using the NormalizeData function with parameters normalization.method = "LogNormalize", scale.factor = 10000.
Clustering and Analysis: Variable genes were identified with the FindVariableGenes function with parameters mean.function = ExpMean, dispersion.function = LogVMR, x.low.cutoff = 0.0125, x.high.cutoff = 3, y.cutoff = 0.5, which identified 1548 variable genes. Data was then scaled using the ScaleData function, with the "vars.to.regress" argument including nUMI and percent.mito to remove these unwanted sources of variation. A total of 100 principal components were calculated from these variable genes using the function RunPCA. Selection of significant PCs to include in further analyses was determined using the JackStraw method with the num.replicate parameter set to 200, along with visualization of each PC using the PCHeatmap function. 61 total PCs with p < 1e-5 were included. Clusters were then identified using the FindClusters function with arguments reduction.type = "pca", dims.use = sigPCs, resolution = 0.6, which identified 14 clusters of cells. Quality control of clusters and samples was performed by plotting nGene and nUMI, which confirmed no major differences based on these parameters. The TSNEPlot function was used to generate t-SNE 9 plots to visualize data by cluster and by sample of origin. Marker genes for each cluster was fold change values prior to plotting, and X-and Y-linked transcripts were excluded from plots.

Animals
The Setd5 GFP transgenic mouse, hereafter referred to as the Setd5 +/heterozygous knockout animal, was obtained and crossed for 5 generations onto the C57/Bl6 background 10 . Animals were housed in same-sex, single-genotype cages, 5 mice per cage, on a light/12 hour dark cycle with access to chow and water ad libitum. Statistical analysis for behavioral experiments was conducted by t-test of pooled male-female group means, or 2-way ANOVA for experiments with several factors (eg, genotype and trial), with wild-type littermates serving as controls. Behavioral experiments and offline analyses were carried out by a blinded observer, and other than tests requiring consecutive days, animals were permitted at least 4 days rest between different behavioral tasks. All breeding, husbandry, behavioral, and euthanasia manipulations were carried out in accordance with the Institutional Animal Care and Use Committee (IACUC) at the University of California, San Diego. Sample sizes were determined by an initial pilot experiment of n=6 animals per genotype.

Neurologic and Metabolic Assays
Weight: animals were weighed once at 10 weeks age.
Grip strength: 10-week-old animals were assayed for grasping strength on the grip strength grid inserted into the Omnitech Animal Grip Strength Meter (Columbus, OH, USA). The mean of three measurements was recorded and normalized by body weight.
Neurological score: 10 week old animals were assayed on a composite neurologic scoring scale as described previously 11 . Briefly, each animal was observed in 4 separate tasks (ledge test, hindlimb clasping, gait, kyphosis) and scored from 0 (unaffected) to 3 (severely affected), with a maximum total of 12 per mouse.
Rotorod: 10-week-old animals were subject to the rotorod test of motor learning and coordination as described previously 12 . The trials were conducted on an accelerating rotorod, (Columbus Instruments, Columbus, OH, USA), with 4 trials per day over 2 consecutive days, with constant acceleration from 0-40rpm over 60 seconds, with 100 seconds total per trial.

Behavioral Assays
Open field: The open field test was conducted as described previously 12 . The task itself is conducted in an open field apparatus (Omnitech Electronics, Columbus, OH, USA) plexiglass box, 48cm x 48cm containing photo beams that detect motion. The 11-week old naïve animal is placed in the box and allowed to move freely for the 1 hour test period. Automated tracking of locomotion in the 3 consecutive 20-minute periods is plotted as a metric of animal activity. For purposes of measuring thigmotaxis, or the animal's preference for the perimeter, the chamber was divided into the central 25% and peripheral 75% regions.
Barnes maze: Assays were conducted as described previously 13 , at 16 weeks age.
Specifically, the Barnes maze was a 1m diameter, flat white plastic disk. Around the perimeter were 20 symmetrical holes, one of which (the target) contained a closed plastic escape box filled with fresh animal cage bedding for each trial. The maze itself was elevated above the ground 1m. The position of the escape chamber relative to the room was randomized for each mouse but remained the same across trials per animal.
In each 3-minute trial, the animal was placed in the center of the maze in the dark, the room was illuminated, and latency to solve the maze and enter the escape chamber was measured. Following an initial 1 minute acquisition trial to acquaint the mouse with the maze and escape chamber, 4 trials per mouse per day were conducted for 4 consecutive days. A probe trial was conducted on day 5 and day 12 in which the hole covering the escape chamber was closed, to confirm that the animal indeed used spatial cues from the room. The maze was cleaned and disinfected to remove any olfactory cues between all trials.
Elevated plus maze: The elevated plus maze was conducted as described previously 14 .
The plus maze was constructed of opaque plastic and elevated 1m above the ground; it consists of 4 33x5cm arms, two of which are enclosed with 25cm high walls. During the single 10-minute test period, the 11 week old animal was placed in the center of the maze and allowed to move freely 12  Nest building: In this single overnight (8 hour) test 13 , the 12-week-old animal was singly housed in a clean cage with fresh bedding with a pressed cotton Nestlet (Ancare, Bellmore, NY, USA). In the morning, the nest formed was visually inspected and assigned a score (1=untouched; 2=partially torn; 3=mostly shredded, but no single nest site; 4=single but flat nest; 5=single, 3-dimensional, complex nest).

3-chamber social interaction:
The 3-chamber test was conducted on 13-14-week-old animals as described previously 13,15 . The box itself was 20x40x22cm, with clear plexiglass walls and 2 plexiglass interior dividers to separate the 3 separate chambers; a hole in the dividers permits free movement between them. In all tests, the subject animal was placed in the empty center chamber and allowed to move freely through the 10 min test period, with the time spent in each chamber (left or right) measured by offline video analysis, with the contents of each chamber (left vs. right) randomized for each experimental mouse. In the first test, for sociability and social approach, one chamber contained a sex-and age-matched but unfamiliar mouse (stranger 1) contained within a smaller wire mesh container permitting reciprocal snout-snout interactions; the other chamber contained a similar container, but empty. In the second test, for social-novelty preference, the initial unfamiliar mouse remained contained in the wire mesh container and now served as an investigated unfamiliar mouse (stranger 1).
The second, previously empty chamber was now filled with a new, age-and sexmatched but completely unfamiliar mouse (stranger 2) to the experimental mouse. In the third test, for social-preference, one chamber was filled with an unfamiliar sex-and age-matched mouse (stranger 3), while the other was filled with a cagemate animal.

Microscopy
Fluorescence microscopic images were acquired using Zen software on an inverted microscope (Zeiss Apotome) with 3-dimensional reconstruction of 10-20 µm composite z-stack images. The objective ranged from 10x (brain tissue immunohistochemistry) to 40x/63x (neuronal morphology and synaptic puncta). Illumination was provided by fluorescent bulb with appropriate filters for DAPI, AlexaFluor 488, AlexaFluor 555, or AlexaFluor 647 fluorophores. Zen Apotome deconvolution was applied to all raw images prior to analysis.

Statistics
Detailed statistical methods are included in the description of individual experiments.
Statistical analysis was conducted with GraphPad Prism 7 software. The a threshold for statistical significance was set at 0.05. In general, for comparison of 2 group means, 2tailed, unpaired t-test was used (Mann-Whitney test for ordinal data). For comparison of 2 more or more factors (eg, genotype and sex), 2-way ANOVA was used. For MRI data, a FDR P value of 0.05 was used to evaluate significance of P values obtained by t-test.
Randomization was not performed prior to experiments, and group assignment was based on genotype. In all cases, the experimenter was blinded during data acquisition and analysis. Significance level was defined as *P<0.05, **P<0.01, ***P<0.001, ****P<10 -4 .

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Supplemental Figure 9 Absolute differences in brain region volume. Coronal series highlighting regional size differences with significance exceeding FDR (blue=smaller, red=larger). Statistics: n=12 (wt) or n=11 (het) adult brains as independent replicates. Figure 10 Significantly reduced thickness of CTIP2 + cortical neuron layer in Setd5 +/postnatal brain. A-B) Representative images of fluorescent P1 cortical sections, scale bar=50 μm. C,E,G) Immunohistochemical quantification of P1 cortical layers. At P1 no differences observed in thickness of TBR1 + layer (P=0.9090,