PDZD8 Disruption Causes Cognitive Impairment in Humans, Mice, and Fruit Flies

Background The discovery of coding variants in genes that confer risk of intellectual disability (ID) is an important step toward understanding the pathophysiology of this common developmental disability. Methods Homozygosity mapping, whole-exome sequencing, and cosegregation analyses were used to identify gene variants responsible for syndromic ID with autistic features in two independent consanguineous families from the Arabian Peninsula. For in vivo functional studies of the implicated gene’s function in cognition, Drosophila melanogaster and mice with targeted interference of the orthologous gene were used. Behavioral, electrophysiological, and structural magnetic resonance imaging analyses were conducted for phenotypic testing. Results Homozygous premature termination codons in PDZD8, encoding an endoplasmic reticulum–anchored lipid transfer protein, showed cosegregation with syndromic ID in both families. Drosophila melanogaster with knockdown of the PDZD8 ortholog exhibited impaired long-term courtship-based memory. Mice homozygous for a premature termination codon in Pdzd8 exhibited brain structural, hippocampal spatial memory, and synaptic plasticity deficits. Conclusions These data demonstrate the involvement of homozygous loss-of-function mutations in PDZD8 in a neurodevelopmental cognitive disorder. Model organisms with manipulation of the orthologous gene replicate aspects of the human phenotype and suggest plausible pathophysiological mechanisms centered on disrupted brain development and synaptic function. These findings are thus consistent with accruing evidence that synaptic defects are a common denominator of ID and other neurodevelopmental conditions.


SUPPLEMENTAL VIDEOS
Video S1 Provided as a separate file (Video S1.mp4). Spontaneous repetitive hindlimb jumping by group housed Pdzd8 tm1b mouse in the home cage (length: 19 s).

Video S2
Provided as a separate file (Video S2.mp4). Spontaneous repetitive hindlimb jumping by singly housed Pdzd8 tm1b mouse in the home cage (length: 13 s).

Ethical Approvals
The human study was approved by the Sultan Qaboos University Ethical Committee.
Informed consent was obtained from the parents of the affected individuals using a process that adhered to the tenets of the Declaration of Helsinki. Ethnically matched Omani controls were recruited from members of the Leeds Omani Society, using a process approved by the  (7).

Bulk Human Brain Gene Expression Analysis
The Allen Human Brain Atlas publishes a rich dataset of gene expression across brain regions, from age 8 wpc to adult ages (2). BrainSpan data analysis of PDZD8 and the CYC1 reference gene (3) was performed. RNA-seq expression measured in RPKM was obtained from the BrainSpan project data and summarized to Gencode v10 exons for selected annotated brain tissues aged 8 wpc to 23 years. To obtain a moving average across ages, a polynomial was fitted to the data.

Fly Stocks and Husbandry
Fruit flies (D. melanogaster) were raised in a 25°C humidified room, with a 12:12 light:dark cycle. Flies were maintained in plastic vials containing 7 ml sugar-yeast-agar medium (8).
Larvae were raised 100 per vial and supplemented with live yeast. Upon eclosion, flies were separated using ice anaesthesia into single-sex groups of ten, and the females analysis; only males were assessed because CG10362 is located on the X chromosome.

CG10362 Knockdown
To inhibit expression of CG10362, UAS-Gal4-mediated RNA interference (RNAi) was used After an interval of 30 seconds with room air, flies tested for a preference between the two odours over 2 minutes. Flies were counted as having learnt if they chose the odour that was not simultaneous with mechanical shock. To calculate a learning index, the number of flies that did not learn was subtracted from the number of flies that learnt, and then divided by the total number of flies tested. An independent cohort of 50 flies would then be trained against the opposite odorant combination and these learning indices averaged for one data point.

Courtship Conditioning
Courtship conditioning is a well-established assay of Drosophila associative learning and memory, in which male flies that have had their copulation attempts rejected by pre-mated non-receptive females respond by suppressing their subsequent courtship behaviour. This learned suppression of courtship behavior is measured over time, thereby serving as an indicator of learning and memory (14,15). After exposure to a pre-mated CS female over a training period (1 hour or 5 hours), the male was isolated for a period of time (30 minutes or 48 hours) before being exposed to a virgin CS female for 10 minutes, during which the proportion of time the male performed stereotypic courtship behaviors was measured (courtship index). To establish whether each trained male had learnt, the courtship index was compared with that of untrained males of the same genotype with no prior exposure to pre-mated females. A memory index was calculated by dividing the courtship index of each trained male by the average courtship index of all untrained males of the same genotype.
This creates a scale whereby 0 represents total suppression, and 1 represents no suppression (16).
To assess short-term (30-minute) memory, males were conditioning by being placed individually into a courtship chamber (0.3 cm 3 ) with a pre-mated female for 1 hour. Males were then removed and isolated in food vials for 30 minutes before being exposed to a virgin female in another, fresh, courtship chamber. Pairs were videoed for 10 minutes or until successful copulation to establish courtship indices.
To assess long-term (48-hour) memory, males were conditioned by exposure to a pre-

Mouse Behavioral Testing
All behavioral experiments were conducted using early adults over 8 weeks of age. Mice were handled for one week prior to behavioral testing. Mice were transferred to the experimental room 30 mins prior to the start of testing. All apparatus was cleaned with 70% ethanol between each mouse. All experiments were conducted between 9am and 5pm. A webcam was positioned directly above the apparatus to record the trials using ANY-maze Al-Amri et al.

Supplement 24
Video Tracking Software (Stoelting, Dublin, Ireland). Experimenters were blinded to genotype during behavioral testing.

Open Field
The open field test was conducted according to a protocol adapted from a previously described method (18). The open field arena had semi-transparent (frosted) Perspex walls, with an internal diameter of 40 x 40 cm (1,600 cm 2 ). The floor of the arena was white plastic.
A circular white outer wall (30 cm away from the arena wall) prevented the testing room from being visible during the trial. Lighting of the maze was measured using a luminance meter at ~200 lux at the center of the arena.
Each mouse was placed into the arena facing the same wall and left to explore undisturbed for 60 minutes. The tracking software divided the arena into the following zones, using the following dimensions: outer zone, 5 cm from the outer wall (700 cm 2 , 43.75% total area); inner zone (100 cm 2 ; 6.25% total area); intermediate zone (800 cm 2 ; 50% total area).
Time spent in and entries to each of the zones and distance travelled were measured. Entry into a zone was recorded when the centre point of the animal crossed into the specific zone.
Rearing, jumping and grooming were recorded during the trial by the experimenter.

Elevated Plus Maze
The elevated plus maze test was conducted according to a protocol adapted from a previously described method (18).

Y-maze
The Y-Maze spontaneous alternation test was conducted according to a protocol adapted from a previously described method (19). The Y-maze was constructed from three identical arms (35 x 5 x 15 cm) made of matt white acrylic, placed at an angle of 120° from each other. Mice were placed into the end of one of the arms (start arm was counterbalanced between genotype and sex) and allowed to roam freely through the maze for 5 minutes. The Mice could repeat entries into a single arm in this variant of the Y-maze procedure, resulting in a chance performance level of 22% alternation (2/9), relative to the 50% alternation chance performance level in versions of this task that preclude multiple entries into the same arm.

Barnes Maze
The Barnes maze test was conducted according to a protocol adapted from a previously described method (20). The maze consisted of a 9mm thick satin white PVC circular arena (122 cm diameter) with 20 equidistant holes (5 cm diameter) around the perimeter (7.5 cm from the edge). The maze had no perimeter walls and was raised 40 cm from the ground. During acquisition trials, the following parameters were used to assess spatial learning: latency (s) to enter the target hole; latency (s) of the mouse's head to enter the target hole for the first time; and distance travelled (m) before the mouse made its first head entry into the target hole (primary path length).
Twenty-four hours after the last acquisition day, a probe trial was conducted, during which the escape box was removed and the mouse was allowed to explore the arena for 80 s. Three measures were used to assess spatial memory during the probe trial: time spent in the target quadrant (a geometric area covering 25% of the arena with the escape hole in the center of five holes); time spent in the target sector (a geometric area covering 5% of the arena for each of the 20 holes); and time spent in and number of entries to the target hole annulus (a 7 cm diameter circle centered on the target hole).
Search strategies in the Barnes maze were classified using the BUNS (Barnes-maze unbiased strategy) algorithm (21). Fisher's exact test was used to analyse the spatial strategy data.

Statistical Analysis
Statistical analysis was performed using SPSS and GraphPad Prism. Data were tested for normality using the Shapiro-Wilk test, and for and homoscedasticity using Levene's test. We used Student's t tests (with Welch corrections for heteroscedasticity), and two-way repeated-measures analysis of variance (ANOVA) when data passed normality and homoscedasticity assumptions. If applicable, the Greenhouse-Geisser adjustment was used to correct for violations of sphericity. Significant ANOVA interactions were analysed further using simple main effect analysis with Fisher's least significant difference (LSD). If the data violated normality, it was square root transformed (SQRT) or Mann-Whitney tests and Friedman's ANOVA were used. Unless otherwise stated, α was set at .05. Graphs were prepared using GraphPad Prism version 6.

Slice Preparation
Transverse hippocampal slices were prepared from 4-6-week-old male and female

Data Analysis
All data are expressed as mean ± SEM. Example traces shown are averages of 10-20 consecutive sweeps. Statistical significance was assessed using paired or unpaired Student's t-tests as appropriate and the level of significance set at p < .05. Synaptic strength for extracellular recordings was measured as the initial slope (20-50%) of the field potential response.
After perfusion, the brain and remaining skull structures were incubated in 4% PFA + 2 mM gadoterate meglumine overnight at 4°C and transferred to 0.1 M PBS containing 2 mM gadoterate meglumine and 0.02% sodium azide for at least 1 month prior to MRI scanning.

Supplement 28
A multi-channel 7.0 Tesla MRI scanner (Agilent Technologies, Palo Alto, USA) was used to image the brains within their skulls. Sixteen custom-built solenoid coils were used to image the brains in parallel. The parameters for the MRI scan were as follows: T2weighted 3D fast spin-echo sequence, with a cylindrical acquisition of k-space, and with a TR of 350 ms, and TEs of 12 ms per echo for 6 echoes, 2 averages, field-of-view of 20 x 20 x 25 mm 3 and matrix size = 504 x 504 x 630 giving an image with 0.040 mm isotropic voxels. The total scanning time was ∼14 hr.
To visualize and compare any changes in the mouse brains, the images were linearly Volumetric changes were calculated on a regional and a voxel-wise basis. The Jacobian determinants of the deformation fields were computed and analyzed to measure the volume differences between subjects at every voxel. To compute the volume of brain regions in all the input images, a pre-existing classified MRI atlas encompassing 159 different segmented structures throughout the brain was warped onto the population atlas.
To assess differences in specific brain regions, we normalized the volume of each region to the absolute brain volume, using the formula [individual absolute volume region / individual absolute volume whole brain * mean absolute volume whole brain], and reported the relative volume as % total brain volume. A linear model with genotype and sex as predictors was fitted to the absolute and relative volume of every region independently and to every voxel independently (voxel-wise statistics) in the brains of Pdzd8 tm1b and WT mice, with a false discovery rate (FDR) threshold of 1%. Multiple comparisons were controlled for using the FDR within the RMINC package for R.