Drosophila functional screening of de novo variants in autism uncovers damaging variants and facilitates discovery of rare neurodevelopmental diseases

SUMMARY

Individuals with autism spectrum disorder (ASD) exhibit an increased burden of de novo mutations (DNMs) in a broadening range of genes. While these studies have implicated hundreds of genes in ASD pathogenesis, which DNMs cause functional consequences in vivo remains unclear. We functionally test the effects of ASD missense DNMs using Drosophila through “humanization” rescue and overexpression-based strategies. We examine 79 ASD variants in 74 genes identified in the Simons Simplex Collection and find 38% of them to cause functional alterations. Moreover, we identify GLRA2 as the cause of a spectrum of neurodevelopmental phenotypes beyond ASD in 13 previously undiagnosed subjects. Functional characterization of variants in ASD candidate genes points to conserved neurobiological mechanisms and facilitates gene discovery for rare neurodevelopmental diseases.

Graphical Abstract

In brief

Marcogliese et al. generate >300 Drosophila mutants and use complementary rescue-based and overexpression approaches to study the function of de novo missense variants found in autism. They find that 38% of missense changes have functional consequences and identify variants in GLRA2 that cause a variable neurological disorder.

INTRODUCTION

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition with impairments in social interaction, communication, and restricted interests or repetitive behaviors (APA, 2013). Individuals affected by ASD, particularly in severe cases, exhibit a higher burden of de novo mutations (DNMs) in an expanding list of genes (Coe et al., 2019; Fischbach and Lord, 2010; Iossifov et al., 2014). The genetic burden of DNMs in ASD subjects has been estimated to account for ~30% of disease causation (Iossifov et al., 2014; Rubeis et al., 2014; Sanders et al., 2012; Satterstrom et al., 2020; Takata et al., 2018; Yuen et al., 2017). While these studies have implicated hundreds of genes in ASD pathogenesis, which of these genes and variants causally contribute to this disease remains unknown. Missense DNMs in particular present a unique challenge because most genes lack established functional assays. Drosophila melanogaster is a genetically tractable system that is widely used to study human diseases (Bellen et al., 2019; Link and Bellen, 2020; Marcogliese and Wangler, 2001). In addition to studying disease mechanisms by establishing preclinical models, flies can be used as a “living test tube” to study functional consequences of variants of unknown significance found in subjects. Here, we integrate a number of state-of-the-art technologies in the fly field to establish an in vivo pipeline to effectively study the functional impact of DNMs identified in a large cohort of ASD subjects.

RESULTS

Prioritization of ASD variants to study in Drosophila

We prioritized genes with coding DNMs identified in ASD probands from the Simons Simplex Collection (SSC) (Iossifov et al., 2014) that are conserved in Drosophila. We primarily focused on missense variants and in-frame indels because functional consequences of these variants are more difficult to predict compared with nonsense and frameshift alleles. However, we also tested a few truncating variants in single-exon genes because these transcripts escape nonsense-mediated decay. In this cohort, 1,708 ASD proband-specific de novo missense or in-frame indels were identified through whole-exome sequencing (WES) (Figure 1A), corresponding to 1,519 unique human genes. Of these, 920 fly genes corresponding to 1,032 human genes were identified. Based on multiple ortholog prediction algorithms scores, 487 human genes had no or weak ortholog candidates in Drosophila (cut off: DIOPT <4/16; Hu et al., 2011; Table S1). By overlapping these 920 Drosophila genes with available fly lines containing Minos-mediated integration cassette (MiMIC) transposons within coding introns that permit targeting of all annotated protein isoforms (“gold”; 1,732 insertions; Nagarkar-Jaiswal et al., 2015; Venken et al., 2011), we identified reagents for 122 fly genes corresponding to 143 human genes and 179 ASD proband variants from the SSC. Compared with the entire genome, the SCC subset and our study subset showed enrichment for constrained genes by assessing gene-level metrics, such as probability of loss of function intolerance (pLI) (Figure 1B; Lek et al., 2016), loss of function observed or expected (o/e) upper bound fraction (LOEUF) (Figure 1C; Karczewski et al., 2020), or missense o/e (Figure 1D; Karczewski et al., 2020).

Figure 1. Gene and variant prioritization, resource generation, and screening outline.

(A) Criteria to prioritize ASD candidate genes and variants for this study.

(B–D) Gene level constraints from control individuals (gnomAD) for (B) probability of loss of function intolerance (pLI), (C) loss of function observed or expected (o/e) upper bound fraction (LOEUF), and (D) missense o/e. **p < 0.01, ***p < 0.001, and ****p < 0.0001.

(E) Schematic depicting generation and effect of TG4 lines on gene function.

(F) Schematic illustrating generation of UAS-human cDNA constructs.

(G) Total number of Drosophila reagents generated for this study.

(H) Screening paradigms using both humanization and overexpression strategies to assess SSC-DNM function.

Of the 122 fly genes that met our selection criterion, we were able to successfully generate 108 T2A-GAL4 (TG4) lines via recombinase-mediated cassette exchange (Diao et al., 2015; Gnerer et al., 2015; Lee et al., 2018). These 108 TG4 lines correspond to 128 SSC genes (some fly genes correspond to multiple human genes with variants in SSC; Figure S1). These fly lines behave as loss-of-function (LoF) alleles that simultaneously produce a GAL4 transactivator in the same temporal and spatial pattern as the gene of interest (Figure 1E; Tang et al., 2009).

To generate upstream activating sequence (UAS) human reference transgenic (Ref-Tg) and human SSC candidate variant transgenic (SSC-Tg) fly lines, we obtained human open reading frame (ORF) collections from the Mammalian Gene Collection (MGC Project Team et al., 2009) or commercial sources (Figure 1F). Out of the 143 human genes and 179 SSC-DNMs of interest that we attempted to generate, we were successful in generating 194 UAS-cDNA (106 Ref-Tgs; 88 SSC-Tgs) flies (Figures 1G and S1; Tables S2 and S3). The UAS-Ref-Tg and UAS-SSC-Tg were inserted into the same genomic docking site in the fly genome and are generated with the same construct, only differing in the variant point mutation, allowing for direct comparison of function by controlling for positional effects.

We were able to make a complete set of TG4, UAS-Ref-Tgs, and UAS-SSC-Tgs lines for 65 fly genes corresponding to 74 human genes and 79 variants (again, some fly genes correspond to multiple human genes and multiple SSC variants are found for a small subset of human genes), which were critical to test variant function using a rescue-based humanization strategy. In summary, 302 Drosophila stocks were generated for this project as a resource for the community, and these stocks are available from the Bloomington Drosophila Stock Center (BDSC) or in the process of being transferred and registered at BDSC (please see Figure S1 for detailed screen pipeline data).

In order to determine whether there are functional differences between the Ref-Tg and SSC-Tg, we used them in combination with TG4 lines to “humanize” Drosophila genes or crossed these lines to ubiquitous and tissue-specific drivers to ectopically overexpress reference or variant human proteins and assessed them for phenotypic differences (Bellen et al., 2019; Figure 1H). If the SSC variant lacked a function a reference allele possessed, we classified it as a LoF allele (e.g., amorph and hypomorph). If the variant had some function that the reference allele did not possess, we classified it as a gain-of-function (GoF) allele (e.g., hypermorph, antimorph, and neomorph) in this study.

Humanization of essential Drosophila genes reveal loss-of-function ASD variants

We identified 47 of 65 lethal TG4 mutants that remained lethal when placed in trans with a corresponding deficiency line. These 47 TG4 lines correspond to 60 human ASD candidate genes for both reference and variant human cDNA transgenic fly lines (Figure S1; Table S4). To assess whether the human homolog can replace the corresponding fly genes, we determined whether UAS-Ref-Tg can rescue the lethality of lethal TG4 mutants. We assessed rescue at four temperatures (18°C, 22°C, 25°C, and 29°C) as the GAL4/UAS system is temperature dependent (Nagarkar-Jaiswal et al., 2015). We found that lethality was suppressed in 17 of 37 genes tested (46%; Figure 2A; Table S5). We next tested whether SSC-DNMs have functional consequences by comparing the rescue efficiency of UAS-Ref-Tg and UAS-SSC-Tg. We observed significant functional differences in the ability for SSC variants to rescue lethality for ABL2, CAT, CHST2, TRPM6 (two variants), and TRIP12 (Figures 2B–2D). For ABL2 and CAT, we further found that humanized flies carrying the SSC-DNM (ABL2^A1099T or CAT^G204E) had significantly decreased lifespan compared with reference animals (Figures 2E and 2F). Overall, we found that 32% (6/19) of the tested SSC-DNMs functionally differed from the reference in vivo, all behaving as LoF alleles.

Figure 2. Assessment of SSC-DNM function through humanization of essential fly genes.

(A) Rescue of lethality to adult stage by TG4-driven UAS-reference human cDNA and subsequent comparison of reference and variant cDNA.

(B–D) Observed/expected Mendelian ratios for rescue of humanized TG4 mutants across different temperatures. Three independent crosses were set per genotype, and n > 50 flies were quantified for each cross. Statistical analyses were performed by ANOVA followed by Sidak’s multiple comparisons test. Error bars are +SEM (standard error of mean).

(E and F) Lifespan analysis of humanized TG4 lines at 25°C. Survival comparisons obtained by log rank (Mantel-Cox) test with a minimum of 11–49 flies for each genotype from three independent crosses.

(G) Single focal confocal images showing expression pattern of UAS-nls.mCherry driven by TG4 (red) in the anterior of the central adult brain. Bottom two rows depict the 5× zoom from dotted white box of co-localization between TG4 reporter and Elav (neurons) or Repo (glia) in cyan. Co-localization is depicted in white. Scale bars represent 50 μm. Dotted magenta lines outline of the brain. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

To assess whether the fly homologs of human ASD candidate genes from SSC are expressed in the central nervous system (CNS), we crossed each TG4 line to UAS-nls.mCherry (red fluorescent protein with a nuclear localization signal) and performed co-staining with neuronal (Elav) and glial (Repo) nuclear markers. We chose the anterior central brain of the fly CNS to image as it is enriched for neuronal nuclei. All five genes associated with deleterious LoF DNMs were expressed in the adult central brain (Figures 2G and S2A). Examining neuronal and glia expression throughout the central brain revealed that all five genes were found in subsets of neurons and some glia, where Abl (corresponding to ABL2) and ctrip (TRIP12) are expressed in neurons and a wider set of glia (Figure S2B). All five genes are also expressed in the third-instar larval CNS (Figure S2C).

Humanization of viable Drosophila TG4s reveals ASD variants with altered function

While we were able to test the function of 37 human genes based on rescue of lethality as mentioned above, 61 TG4 lines corresponding to 68 SSC-ASD candidate genes were viable and did not exhibit any obvious morphological phenotypes that can be used for variant functional studies. Note that 18 of 61 TG4 lines were homozygous lethal but were viable as compound heterozygotes over a molecularly defined deficiency that covers the locus, indicating that these lines carry second site lethal mutations (Figure S1), which has been previously reported in a subset of MiMIC strains (Nagarkar-Jaiswal et al., 2015). For the 43 of 61 remaining viable TG4 lines, we attempted to generate humanized TG4; UAS-Ref-Tg and TG4; UAS-SSC-Tg lines in an appropriate genetic background. Before carrying out behavioral studies, we had to replace the × chromosome of the TG4 lines with the × chromosome from a Canton-S strain to eliminate the effect of yellow (y) and white (w) alleles that were present in the original stocks (Figure S3). We were successful in generating all the necessary strains for eight genes. Using these humanized flies in a y⁺ w⁺ background, we performed courtship assays to assess social interactions in mutant and humanized flies considering that ASD patients exhibit social deficits (Figure 3A; Liu, 2013). Fly courtship involves a complex set of neurological components involving sensory input, processing, and motor output (Guo et al., 2019). We measured the amount of time a TG4 male fly spent performing wing extensions as a proxy for courtship, as well as the amount of time spent copulating with a wild-type (Canton-S) female. In addition to quantifying these two parameters, we also measured the time flies spent moving within the test chamber to assess their locomotion and tracked grooming, a stereotypic behavior in flies that involves a complex neurocircuit (Seeds et al., 2014).

Figure 3. Assessment of SSC-DNM function through humanization of viable TG4 lines and behavioral analysis.

(A) Analysis pipeline used to evaluate Drosophila behavior.

(B) SSC-DNMs in which variants display significant differences in time spent performing a specific behavior (courtship, copulation, movement, or grooming) when compared with reference humanized flies.

(C–G) The number of frames male flies spent performing courtship (single-wing extensions), copulating, moving within the chamber, or grooming during a 30-min test period. The red line represents the average number of frames a Canton-S (control) male spends performing the same task. n = 10–40 flies were used per genotype. Statistical analysis was performed by non-parametric Kruskal-Wallis one-way ANOVA and the Dunn’s multiple comparison test.

(H) Single focal confocal images showing expression pattern of UAS-nls.mCherry driven by TG4 (red) in the anterior of the central adult brain. Bottom two rows depict the 5× zoom from dotted white box of co-localization between TG4 reporter and Elav (neurons) or Repo (glia) in cyan. Co-localization is depicted in white. Scale bars represent 50 μm. *p < 0.05, ***p < 0.001, ****p < 0.0001, and ns, not significant. Error bars are ± SEM.

Of the eight SSC-DNMs tested, we found five variants that showed functional alterations from the reference allele in at least one of four behavioral paradigms (Figure 3B). Humanized KCND3^R86P flies displayed increased movement and decreased grooming behavior when compared with the humanized reference flies (Figures 3C, S4A, and S4B). The humanized KDM2A^R449K flies showed decreased time copulating compared with the humanized reference (Figures 3D and S4C–S4E). Humanized USP30^P200S flies displayed decreased grooming behavior when compared with humanized reference flies (Figures 3E and S4F–S4H). While these variants have clear functional differences from the reference allele, it was difficult to classify them as clean LoF or GoF alleles. Humanized ALDH1L1^N900H flies displayed a significant reduction in courtship and an increase in grooming behavior when compared with the humanized reference fly or the TG4 mutant alone, potentially indicating the variant acts as some kind of GoF allele (Figures 3F, S4I, and S4J). Finally, humanized reference GLRA2 flies failed to copulate at all but still exhibited normal movement while the humanized GLRA2^N136S flies were capable of copulating within the trial period similar to the TG4 mutant alone, suggesting it behaves as a LoF allele (Figures 3G, S4K, and S4L). Courtship assessment of humanized SSC-DNMs in HTR1D, SLC23A1, and MADD did not show altered function to reference (Figures S4M–S4P). In summary, 63% (5/8) SSC-DNMs act functionally different from reference alleles using quantitative behavioral measurements in flies.

Finally, we determined the CNS expression of TG4 lines corresponding to all eight lines we were able to humanize. Surprisingly, we only detected expression of 4/8 genes in the adult (Figures 3H, S5A, and S5B) and larval CNS (Figure S5C). All are expressed in a subset of neurons, whereas GluClα (GLRA2) is also expressed in some glia (Figure S5B).

Overexpression assays reveal ASD variants with diverse functional consequences

We complemented our rescue-based assays by overexpressing Ref-Tg and SSC-Tg in a wild-type background using ubiquitous (tub-GAL4), eye-specific (GMR-GAL4), and wing-specific (nub-GAL4) drivers (Figure 4A). This approach has routinely been employed to discern functional differences between reference and disease-associated variant proteins in vivo, regardless of whether the phenotypic readout has similarities to the patient’s conditions (Ansar et al., 2019; Chung et al., 2020; Goodman et al., 2021; Harel et al., 2016; Huang et al., 2020; Kanca et al., 2019; Liu et al., 2018; Marcogliese et al., 2018; Post et al., 2020; Ravenscroft et al., 2021; Splinter et al., 2018). Critically, UAS-Ref-Tg and UAS-SSC-Tg are inserted into the same genomic landing site in the fly genome, and the construct only differs by the point mutation, allowing for direct functional comparison. Across the three drivers and testing 66 human genes (73 SSC variants), we found 21/73 SSC-DNMs (in 19 fly genes) showed functional alteration in phenotypic assays, 17/73 displayed phenotypes comparable to reference, and 35/73 did not produce a scorable phenotype (Table S5).

Figure 4. Variant assessment by overexpression of reference and SSC-DNMs.

(A) Phenotypes observed upon overexpressing the reference and variant cDNAs using a ubiquitous driver (tub-GAL4) at 25°C, an eye-specific driver (GMR-GAL4) at 29°C, or a wing-specific driver (nub-GAL4) at 25°C. Black denotes there was no phenotype (NP), purple there was a comparable phenotype (CP), or red there was a functional difference (FD).

(B) Quantification of viability upon overexpression of reference or variant human cDNAs using a ubiquitous driver for genes where the variants showed a functional difference. Minimum of three independent crosses were set with two independent UAS-transgenic lines. We scored 50–100 flies (a minimum of 10 if overexpression caused survival defects). Statistical analyses were performed by unpaired t test.

(C and D) Representative optical sections of eyes and wings for variants with a functional difference using eye-specific (GMR-GAL4) and wing-specific (nub-GAL4) drivers, respectively, at 25°C.

*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Error bars are +SEM.

Twelve variants in eleven human genes (ATP2B2^T818M, EPHA1^V567I, GLRA2^N136S, GRK4^P385A, ITGA8^R748G, IRF2BPL^F30L, IRF2BPL^N701fs66*, KCND3^R86P, MINK1^C269R, NPFFR2^M163I, PDK2^R120Q, and TSC2^R1557W) behaved as LoF alleles. GLRA2^N136S, GRK4^P385A, ITGA8^R748G, KCND3^R86P, MINK1^C269R, NPFFR2^M163I, PDK2^R120Q, and TSC2^R1557W were annotated as LoF alleles using a ubiquitous driver because they failed to reduce the expected viability to the extent of the corresponding reference alleles upon overexpression (Figure 4B). Notably, GLRA2^N136S and KCND3^R86P were also annotated as LoF alleles in our rescued based assay, showing consistency (Figures 3C and 3G). Moreover, KCND3^R86P and IRF2BPL^N701fs66* variants behaved as LoF alleles when assessed with multiple drivers. KCND3^R86P abolished the activity of the reference transgene, which caused lethality when overexpressed with a ubiquitous driver (Figure 4B). In the wing, KCND3^R86P failed to produce a severe notching phenotype that is observed by expression of the reference transgene (Figure 4D). Ubiquitous or wing-specific overexpression of reference IRF2BPL caused lethality, whereas the IRF2BPL^N701Tfs66* frameshift allele (note that IRF2BPL is a single-exon gene) does not cause any phenotype (Figures 4B and 4D). Interestingly, the missense variant, IRF2BPL^F30L, behaved as a LoF using the wing driver but was indistinguishable using the ubiquitous driver, indicating it is likely to be a partial LoF allele(Figures 4B and 4D). In addition to IRF2BPL DNMs, variants in two genes (ATP2B2 and EPHA1) were identified as LoF alleles based on the wing-specific driver and assay. Wing-specific overexpression of reference ATP2B2 in the developing wing disc causes a curled wing phenotype while the ATP2B2^T818M variant fails to do so (Figure 4D). Expression of reference EPHA1 caused a wing-size reduction and wing-margin serration, whereas the EPHA1^V567I variant caused serrated wings of normal size (Figure 4D), indicating partial LoF.

Seven variants (ACE^Y818C, GPC5^M133T, MYH9^R1571Q, PC^P1024R, SLC23A1^L465M, HTR1D^T99N, and BAIAP2L1^A481V) behaved as GoF alleles. Flies overexpressing variant forms of ACE, GPC5, MYH9, PC, and SLC23A1 exhibited enhanced lethality when compared with reference protein (Figure 4B). The GoF nature of MYH9^R1571Q was also observed in the wing-size-based assay (Figure 4D). HTR1D^T99N displayed consistent stronger phenotypes compared with reference when expressed in the eye or the wing, resulting in eye size reduction and absent wing phenotype, respectively (Figures 4C and 4D). BAIAP2L1^A481V caused a smaller, more crumpled wing phenotype compared with its reference allele (Figure 4D).

Intriguingly, EPHB1^V916M and MAP4K1^M725T exhibited conflicting results in the eye and wing; therefore, they could not be categorized as simple LoF or GoF variants. While overexpression of reference EPHB1 or MAP4K1 in the developing eye causes eye-size-reduction phenotype, SSC variant forms of either gene result in normal eyes, indicating they behave as LoF alleles in this tissue. However, the same variant transgenes for these two genes expressed in the wing result in blistered or crumpled wings, respectively, that are phenotypically stronger than the reference alleles (Figures 4C and 4D), indicating they behave as GoF alleles in this tissue. In summary, when a scorable phenotype was present, 48% (21/44) of the SSC-DNMs tested with an overexpression strategy impact function. Furthermore, we found diverse SSC variant consequences, including 12 LoF, 7 GoF, and 2 with complex phenotypes.

While this overexpression-based screening approach was not directly investigating the function of genes in the nervous system, expression analysis revealed that most (15/19) fly genes that correspond to SSC-DNMs with a functional difference identified through our overexpression assay are expressed in the adult (Figures 3H, S6, S7, and S8A) and larval (Figure S8B) CNS (see Marcogliese et al., 2018 for Pits [corresponding to human IRF2BPL]). The deleterious nature of variants in three of these genes was identified in our behavioral screen (GluClalpha, 5-HT1B, and Usp30; Figure 3H). Four genes (CG6293, Mhc, Shal, and Ance) were not detected in the brain in our analysis, which could be because they are expressed at very low levels or may primarily function in non-neural tissues. While most genes are enriched in neuronal subpopulations, Pdk (PDK2) and hppy (MAP4K1) are enriched in glia as well. Interestingly, if ITGA8 is not detected in either neurons or glia but revealed a unique expression pattern, that may reflect its expression in tracheal cells. In addition, based on imaging with an nls.GFP that leaks into the cytoplasm, it may also be present within cells that wrap around neurons reminiscent of cells in pars intercerebralis, a neuroendocrine organ analogous to the mammalian hypothalamus (de Velasco et al., 2007; Figure S8A). Taken together, most of the fly genes corresponding to SSC-DNMs in which we found in vivo alterations using overexpression-based assays are expressed in the fly CNS, similar to hits from rescue-based studies.

Identification of 30 deleterious SSC variants by merging all functional data

In total, we found 29 missense and 1 frameshift SSC-DNMs that displayed functional differences when compared with their respective reference allele (for a total of 28 genes: one variant for 26 genes and two variants for two genes; Tables 1 and S5). Approximately 53% (30/57) of the SSC-DNMs exhibited functional differences compared with the reference. Intriguingly, in our study, we only found GoF variants for genes corresponding to viable TG4 fly mutants based on both rescue-based and overexpression-based assays (Figure S9A). Interestingly, while we were able to classify the variants into LoF or GoF based for most genes, we found in two cases where different assays gave different results.

Table 1.

Identification of 30 SSC-DNMs with functional consequences

H. sap gene	pLI (LOEUF)	Missense O/E	OMIM disease	SSC variant	CADD	D. mel gene	TG4 lethality	Functional assay	SSC-CV consequence
ABL2	0 (0.58)	0.81	-	p.A1099T	28.2	Abl	yes	RB	LoF
ACE	0 (1.08)	1.07	267430 (AR)	P.Y818C	7.5	Ance	no	OE	GoF
ALDH1L1	0 (0.78)	0.93	-	P.N900H	9.8	CG8665	no	RB	GoF?
ATP2B2	1 (0.15)	0.54	601386 (AR)a	P.T818M	33.0	PMCA	yes	OE	LoF
BAIAP2L1	0 (0.65)	0.90	-	P.A481V	17.8	IRSp53	no	OE	GoF
CAT	0 (1.05)	1.01	614097 (AR)	P.G204E	28.1	Cat	yes	RB	LoF
CHST2	0.02 (0.81)	0.66	-	p.R52P	12.8	CG31637	yes	RB	LoF
EPHA1	0 (1.04)	0.97	-	p.V567I	1.3	Eph	no	OE	LoF
EPHB1	1 (0.26)	0.73	-	P.V916M	34.0	Eph	no	OE	complex
GLRA2	0.97 (0.30)	0.43	-	P.N136S	25.1	GluClα	no	RB, OE	LoF
GPC5	0 (1.08)	1.09	-	P.M133T	24.3	dally	nob	OE	GoF
GRK4	0 (1.09)	1.09	-	P.P385A	26.0	Gprk2	yes	OE	LoF
HTR1D	0 (1.30)	0.98	-	p.T99N	19.4	5-HT1B	no	OE	GoF
IRF2BPL	0.84 (0.41)	0.90	618088 (AD)a	p.F30L p.N701fs	24.8	Pits	yes	OE OE	LoF LoF
ITGA8	0 (0.66)	1.03	191830 (AR)	P.R748C	35.0	if	yes	OE	LoF
KCND3	0.99 (0.28)	0.48	607346 (AD)a	p.R86P	32.0	Shal	no	RB, OE	LoF?
KDM2A	1 (0.04)	0.43	-	P.R449K	5.7	Kdm2	no	RB	?
MAP4K1	0.99 (0.29)	0.59	-	P.M725T	21.3	hppy	no	OE	complex
MINK1	1 (0.13)	0.60	-	P.C269R	26.8	msn	yes	OE	LoF
MYH9	1 (0.09)	0.71	603622 (AD)a	p.R1571Q	35.0	Mhc	nob	OE	GoF
NPFFR2	0 (1.13)	1.20	-	P.M163I	13.3	SIFaR	yes	OE	LoF
PC	0.01 (0.43)	0.69	266150 (AR)a	P.P1042R	24.6	PCB	no	OE	GoF
PDK2	0 (0.92)	0.63	-	P.R120Q	25.3	Pdk	yes	OE	LoF
SLC23A1	0.02 (0.54)	0.71	-	P.L465M	17.9	CG6293	no	OE	GoF
TRIP12	1 (0.06)	0.60	617752 (AD)a	P.R1643Q	36.0	ctrip	yes	RB	LoF
TRPM6	0 (0.45)	0.86	602014 (AR)a	p.T2011P p.A641 E	12.2 17.8	Trpm	yes	RB RB	LoF LoF
TSC2	1 (0.07)	1.03	613254 (AD)a	p.R1557W	16.0	gig	yes	OE	LoF
USP30	0 (0.66)	0.76	-	P.P200S	14.7	Usp30	no	RB	?

List of all human genes and corresponding SSC variants determined to have a functional difference across all assays in this study. GoF, gain of function; LoF, loss of function; OE, overexpression; RB, rescue-based.

^aNervous system disorder

^bKnown lethal mutants

When we informatically surveyed the genes and variants with functional consequences identified through our screen in comparison to other genes included in our study (variants with comparable function or those lacking a phenotype to assess) using the MARRVEL tool (Wang et al., 2017), we did not find any significant differences in gene-level metrics, such as pLI (Lek et al., 2016); LOEUF (Karczewski et al., 2020); missense o/e (Karczewski et al., 2020); pathogenicity prediction scores based on several in silico algorithms, including sorting intolerant from tolerant (SIFT) (Vaser et al., 2016), PolyPhen-2 (Adzhubei et al., 2010), and combined annotation-dependent depletion (CADD) (Kircher et al., 2014); or absence or presence of identical variant in gnomAD (Lek et al., 2016) (Tables S5 and S6; Figures S9B–S9Z). By analyzing Gene Ontology (GO) by PantherDB (Ashburner et al., 2000) and visualizing with reduce visualize gene ontology (REVIGO) (Supek et al., 2011), ASD candidate genes from SSC with deleterious variants in vivo were compared with all protein-coding genes. We found most significant enrichment for genes with GO terms for “synapse (cellular component)” (Figure S10A) and “ion binding (function)” (Figure S10B). Finally, we systematically imaged the expression pattern of 41 additional TG4 lines generated through our study to document their expression in the adult (Figure S11A) and larval (Figure S11B) CNS as a resource for the community.

Loss- or gain-of-function alleles in GLRA2 cause X-linked neurodevelopmental disorders

Many genes implicated in ASD are also associated with neurodevelopmental disorders beyond autism (Levitt and Campbell, 2009; Sullivan and Geschwind, 2019). Therefore, we asked whether additional variants in genes with disruptive SSC-DNMs could also be responsible for neurological diseases beyond ASD by identifying human subjects with rare, potentially deleterious variants that have not previously been associated with neurological disease (Chong et al., 2015; Gahl et al., 2016; Sobreira et al., 2015). Out of 28 genes in which we identified damaging SSC-DNMs, eight are associated with Mendelian diseases that have neurological presentations documented in OMIM (Amberger et al., 2019; Table 1). For one of these genes (IRF2BPL), we recently reported de novo truncating variants as the cause of a severe neurodevelopmental disorder that presents with abnormal movements, loss of speech, and seizures. This work was done in collaboration with the Undiagnosed Diseases Network (Gahl et al., 2016; Marcogliese et al., 2018). Here, aided by using the online matchmaking software, GeneMatcher (Sobreira et al., 2015), internal human genetics databases, and re-analysis of clinical exome sequencing data, we report variants in GLRA2 as a cause of a neurodevelopmental syndrome with developmental delay (DD), intellectual disability (ID), ASD, and epilepsy (Figure 1H; Table 2).

Table 2.

Salient features of subjects with GLRA2 variants

Subject	1	2	3	4	5	6	7	8	9	10	11	12	13
GLRA2 variant (hg19, NM_001118886.1)	c.887C>T, p.Thr296Met	c.887C>T, p.Thr296Met	c.887C>T, p.Thr296Met	c.887C>T, p.Thr296Met	c.887C>T, p.Thr296Met	c.887C>T, p.Thr296Met	c.140T>C, p.Phe47Ser	c.777C>G, p.Ile259Met	c.754C>T, p.Arg252Cys	c.862G>A, p.Ala288Thr	c.1186C>A, p.Pro396Thr	c.1199C>T, p.Pro400Leu	c.1334G>A, p.Arg445Gln
Inheritance	de novo	de novo	de novo	de novo	de novo	de novo	de novo	de novo	maternal	maternal	maternal	maternal	maternal
CADD score	27	27	27	27	27	27	27.8	23.4	31	27.2	20.9	22.3	33
Gender	female	female	female	female	female	female	female	female	male	male	male	male	male
Age at most recent evaluation (years)	6.7	6.5	5.5	0.5	5.4	0.8	6.7	5	0.9	7	34	15	5.9
Developmental delay/intellectual disability	yes	yes	yes	yes	yes	yes	yes	yes	yes	yes, with regression	yes	yes	yes
Hypotonia/incoordination	no	no	yes, ataxic gait	yes	no	no	no	yes, incoordination	yes	yes, ataxia	yes	no	yes, impaired fine motor coordination
Autism spectrum disorder	no	yes	no	NA	no	NA	no	no, diagnosis of child psychosis	NA	yes	yes	yes	suspected
Inattention/hyperactivity	yes	yes	no	NA	yes	NA	yes	no	NA	no	no	yes	yes
Sleep disturbance	no	yes	no	no	no	NA	yes	yes	yes	no	no	no	yes
Microcephaly	no	yes	yes	yes	yes	borderline	no	no	no	no	no	no	no
Ocular features	myopia, astigmatism, and nystagmus (improved with age)	nystagmus (improved with age)	alternating exotropia, borderline opsoclonus	none	oculomotor apraxia, ptosis	upbeat nystagmus (starting 6 weeks after birth)	strabismus, nystagmus (improved with age)	none	strabismus	myopia	myopia, astigmatism	none	reduced visual acuity
Epilepsy	no	yes	no	yes	no	yes	yes	no	no	yes	no	yes	yes
EEG findings	slow background suggestive of mild encephalopathy	bilateral synchronized high-amplitude spikes, epileptic potentials	normal	slow background, infantile spasms, multifocal spikes during sleep	not performed	intermittent hypsarrythmic pattern, infantile spasms	infantile spasms and then normal interictal EEG	left fronto-temporal spike waves focus, which diffuses in the right frontal region, activated by sleep	not performed	generalized slowing and generalized epileptiform discharges associated with myoclonic jerks	not performed	right temporal focus of high and polymorphic alfa spike-wave complexes, with ipsilateral propagation	left and right posterior and right frontal intermittent slowing, bilateral polyspikes during sleep, and excessive beta-activity (with medications)
Brain MRI findings	normal	delayed myelination, a small arachnoid cyst	mild cortical atrophy, thinning of corpus callosum	normal	normal	normal	cortical and white-matter atrophy, including vermian atrophy	increased signal intensity in FLAIR of the subcortical white matter of the frontal region	not performed	minimally increased T2 signal intensity on the occipital lobes	normal	increased signal intensity in FLAIR of the cortical matter of the parietal	normal

CADD, combined annotation-dependent depletion; EEG, electroencephalography; MRI, magnetic resonance imaging.

We identified rare GLRA2 variants in 13 unrelated subjects with or without autistic features. In addition to developmental and cognitive delay of variable severity, 4/13 subjects have microcephaly, 6/13 subjects have a history of epilepsy, and 10/13 subjects have ocular manifestations, including congenital nystagmus that improved with age in three of the cases (Table 2; see GLRA2 subject case histories). Glycine receptor alpha 2 (GLRA2) is an X-linked gene that encodes a subunit of a glycine-gated chloride channel (Zeilhofer et al., 2018). All female subjects harbored DNMs, including a recurring GLRA2^T296M de novo variant found in 6/8 female subjects. The GLRA2^T296M variant was also identified in a female subject in previous large-scale developmental disorder study (Deciphering Developmental Disorders Study, 2017). The five male subjects had inherited GLRA2 variants from unaffected mothers. The mother of subject 12 has a history of learning problems. The CADD scores for all five male subjects are predicted to be damaging (Table 2). Four of the five male subjects had diagnosed or suspected ASD. A maternally inherited microdeletion of GLRA2 was previously reported in a single male patient with ASD (Pinto et al., 2010), indicating that hemizygous LoF allele of this gene in males may cause ASD. Indeed, the de novo GLRA2^N136S variant present in the SSC in a male subject acts as a LoF allele based on overexpression studies (Figure 4B), which is supported by our behavioral assay on humanized flies, as the GLRA2^N136S variant loses the toxic effect on copulation caused by expression of the humanized reference protein (Figure 3G).

To better understand the functional consequences of variants found in our GLRA2 cohort, we generated additional transgenic flies to assay the function of p.T296M (found in six female subjects) and p.R252C (found in a male subject) variants (Figures 5A and S12A). By overexpressing reference and variant GLRA2 using a ubiquitous driver, we found that GLRA2^R252C behaves as a LoF allele (Figure 5B), similar to GLRA2^N136S (Figure 4B). In contrast, this assay did not distinguish GLRA2^T296M from the reference (Figure 5B). Given the recurrent nature of this variant, as well as structural prediction that the residue has a potential role in obstruction of the ion pore in the closed conformation (Figure S12G; Du et al., 2015; Moraga-Cid et al., 2015), we further tested GLRA2^T296M and other alleles using additional GAL4 drivers. Using a pnr-GAL4 that is expressed in the dorsocentral stripe in the notum, we found that GLRA2^T296M, but not the reference or any other GLRA2 variant tested, causes lethality when expressed at high levels (Figure S12B). When we expressed GLRA2^T296M at lower levels by manipulating the temperature, we found that this variant induces the formation of melanized nodules in the thorax, a phenotype that we never observe when the reference or other GLRA2 variants are overexpressed (Figures 5C, 5D, and S12B).

Figure 5. GLRA2^T296M found in female probands acts as a GoF allele while GLRA2^R252C and GLRA2^N136S found in male probands behave as LoF alleles.

(A) Schematic diagram of domain structure of GLRA2 and the relative positions of subject variants functionally assessed in Drosophila.

(B) Mendelian ratios upon overexpression of the GLRA2 reference or variant human cDNAs using a ubiquitous driver (tub-GAL4). A minimum of three independent crosses were set.

(C and D) Representative images and quantification of melanized nodules formed on the notum of flies expressing GLRA2^T296M driven by a dorsocentral thorax-specific (pnr-GAL4) driver at 25°C.

(E–H) Representative traces of ERG and quantification of “OFF”-transient amplitude (blue bracket) in animals expressing GLRA2 pan-neuronally (both pre-synaptic photoreceptors and post-synaptic laminar neurons; nSyb-GAL4) or only in the pre-synaptic photoreceptors (Rh1-GAL4). ERG was analyzed by ANOVA followed by Bonferroni’s multiple comparison test. Four to ten flies were examined for each genotype. Recording was repeated at least three times per fly. **p <0.01, ***p < 0.001, and ****p < 0.0001. Error bars are +SEM.

To further examine the functional consequences of overexpression of reference and variant GLRA2 in the nervous system, we performed electroretinogram (ERG) recordings on the fly eye expressing human GLRA2 using two distinct drivers. Pan-neuronal driver (nSyb-GAL4; Pauli et al., 2008) allows one to express GLRA2 in both pre-synaptic photoreceptors and post-synaptic neurons in the nervous system. Using this driver, we found a significant increase in amplitudes of “OFF” transients with GLRA2^T296M (Figures 5E, 5F, S12C, and S12D). This indicates an increase in synaptic transmission (Deal and Yamamoto, 2018), supporting the finding in the notum that p.T296M behaves as a GoF allele. Interestingly, when we limited the expression of GLRA2 to pre-synaptic photoreceptors using Rh1-GAL4 (Xiong et al., 2012), we did not observe any functional difference between GLRA2^T296M and the reference allele. However, with this driver, we were able to discern that both GLRA2^R252C seen in subject 9 and GLRA2^N136S found in an SSC subject behave as LoF alleles based on observing a decrease in amplitude of “OFF” transients, indicating a decrease in synaptic transmission (Figures 5G, 5H, S12E, and S12F). Hence, we have identified a cohort of subjects with deleterious variants in an X-linked gene GLRA2 and have shown that a recurrent missense DNM in females acts as a GoF allele, whereas rare variants found in male subjects behave as LoF alleles.

DISCUSSION

In this study, we generated >300 fly strains that allow functional studies of human variants and homologous fly genes in vivo. These reagents can be used to study many coding variants that are being identified through next-generation sequencing efforts in the human genomics field in diverse disease cohorts beyond ASD. Our screen elucidated 30 SSC variants with functional differences compared with reference, which was over half (~53%) of the genes in which we were able to perform a comparative functional assay.

Our screen was part of a larger effort to characterize the functional consequences of missense de novo changes from the SSC dataset using different strategies. One approach was based on proteomics by performing yeast two-hybrid assays on 109 SSC-DNMs found in subjects, showing 20% of protein-protein interactions that are found in reference proteins are disrupted by variants (Chen et al., 2018). Another study reported that ~70% of 37 SSC-DNMs knocked into homologous C. elegans genes caused scorable phenotypes (Wong et al., 2019). These studies are complementary to each other because, while some variants have been identified as deleterious by more than one approach (e.g., GLRA2^N136S identified in both worm and fly screens), others are uniquely identified in one study, which could be due to technical limitations. For example, our approach utilizing human cDNA transgenes allowed us to test variant function, regardless of residue conservation in Drosophila. Of the 29 disruptive missense SSC-DNMs identified through our study, 14 affect residues that are conserved in flies and 10 in worms.

To take an unbiased approach, our gene prioritization was only based on gene-level conservation and tool availability (e.g., intronic MiMIC lines and full-length human cDNA) rather than based on gene level constraints and variant-level pathogenicity prediction scores. Hence, our study subset, although somewhat limited, can be considered a random sample of ASD-implicated genes and variants. Interestingly, we could not find any significant difference in pathogenicity prediction for disruptive variants in vivo. It should also be noted that we were limited by the availability of full-length human cDNA, which could select against genes encoding larger transcripts, for which reagents are often harder to obtain. We were able to generate 13 additional TG4 lines that were homozygous viable and successfully crossed back to a Canton-S × chromosome. We assessed their behavior phenotypes in comparison to the reference Canton-S files, which is presented in Figures S4Q–S4T as an additional resource for the community.

Of the 29 missense SSC-DNMs that had functional consequences in our assays, four were not predicted as damaging variants (CADD < 10), nine had moderate scores (CADD: 10–20), and 16 were predicted to be disruptive (CADD > 20). Understanding how variants that are not predicted to be damaging based on state-of-the-art informatics programs impact protein function may provide guidance to improve the accuracy of in silico tools.

To study the functional consequences of SSC-DNMs, we took two conceptually different approaches (rescue-based humanization strategy and overexpression-based strategy). Indeed, the two approaches were complementary, as only two variants (GLRA2^N136S and KCND3^R86P) were detected in both screens, showing consistent LoF effects using both approaches. However, it should be noted that variant interpretation is not always straightforward for behavioral analysis. Interestingly, GRK4, NPFFR2, and PDK2 SSC-DNMs were found to be LoF variants when overexpressed ubiquitously, yet these variants were able to rescue lethality in a similar manner to their respective reference alleles (Figure 4; Table S5). This suggests that these variants are partial LoF alleles and different drivers and assays have different sensitivity. Moreover, two disruptive SSC-DNMs, EPHB1^V916M and MAP4K1^M725T, behaved as complex alleles, displaying discordant phenotypes in the eye and wing. This suggests that these variants may behave in a context-dependent manner, acting as a GoF allele in one tissue while behaving as a LoF allele in another. Thus, one functional assay may not be enough to reveal the full nature of pathogenic mechanisms, and some disease-associated variants may act differently in different tissues or cell types. It is also important to note that a variety of factors may explain why functional differences are not observed across assays, as different GAL4 lines may have variable developmental timing, strength, and context dependency. In addition, the presence of the endogenous protein in cells and the cell-type-specific effect of exogenous protein may also contribute to our functional readout. Therefore, variant annotation in Drosophila should be supplemented by deeper characterization of the loss-of-function mutant and gene-expression studies to provide the clearest supportive evidence to a molecular diagnosis.

Starting from a single de novo hemizygous missense variant that we identified as a LoF allele in GLRA2 (p.N136S), we identified a cohort of subjects with overlapping neurodevelopmental phenotypes carrying LoF or GoF variants in this gene. Interestingly, female subjects harbored DNMs and male subjects carried maternally inherited variants in this X-linked gene, which undergoes random × inactivation in females, but not in males (Barakat and Gribnau, 2012). The X-linked status of GLRA2 may mean that variants causing reduced GLRA2 activity lead to disease in males but can be tolerated in heterozygous females. This is supported by asymptomatic mothers of male probands who had maternally inherited alleles (subjects 9–11 and 13). In contrast, GoF variants in this channel could be overrepresented in females since hyperactivation of this channel may cause neurological defects (Zhang et al., 2017). Of the eight female subjects, six carried the identical recurring DNM, p.T296M. None of the variants observed in females are present in control databases, arguing strongly that these variants are pathogenic. While the exact mechanism of how the p.T296M variant affects GLRA2 function remains unclear, the presence of melanized nodules in flies expressing this variant are indicative of an innate immune response (Dudzic et al., 2019), potentially as a result of leaky ion channel function (Feske et al., 2015). Fittingly, our structural analysis revealed that the p.T296 residue is adjacent to a critical amino acid that is likely important for keeping the ion pore in a conformationally closed state (Figure S12G).

In summary, we utilized a model organism-based in vivo functional genomics approach to study the functional consequence of rare genetic events in a common neurological disorder, ASD. In addition to garnering variant functional data for ASD subjects in the SSC, we leveraged this information to identify and document a rare neurological condition through matchmaking and collaboration. Such bidirectional communication and collaboration between bench scientists and clinicians greatly facilitates functional studies of human variants found in common diseases, such as ASD, and can also lead to novel discoveries that have an impact on rare-disease research.

Limitations of the study

Although this study revealed a number of rare variants found in ASD patients that have functional consequences, there are several caveats to recognize based on the design and assays used in this screen. First, it is not clear for the majority of the 30 hits found by the screen whether the disruptive variants are truly pathogenic and directly contribute to ASD. Identification of additional patients with similar genotypes and phenotypes will be necessary to establish a causal relationship between these variants and ASD pathogenesis. Second, we do not argue that this screen was able to identify all variants that had functional consequences, considering most variants were shown to be deleterious based on one phenotypic assay. Hence, if one performs additional assays in different biological contexts, more variants with altered function may be discovered. Third, it is not clear how many phenotypic assays are required to determine whether a variant has functional consequences and which phenotypic assays in flies are more relevant to complex human phenotypes, such as ASD. Additional studies of variants identified as having functional consequences in an invertebrate model organism should be followed up using mammalian models or human cells, tissues, and organoids to assess whether the deleterious variants affect biological processes that relate to ASD. Fourth, although the majority of the hits were alleles of genes in which the fly homolog is expressed in the nervous system, some were in genes in which we failed to detect any expression in this organ system in Drosophila. This could be because some of these genes are expressed at low levels that are beyond the detection limit of our assay systems; only expressed at a specific time point during neural development; expressed in the nervous system of humans, but not flies; or they may be contributing to neurological phenotypes through their function in non-neural organ systems. Additional gene expression and functional studies will be required to fully understand their mechanistic contributions to ASD. Fifth, although we were able to identify some deleterious variants that affect fly behavior, which is the most relevant phenotype to ASD out of all assays we performed in this screen, we have not been able to assess this for all variants due to technical limitations. In order to perform clean behavioral experiments, one must control the genetic background since this could be a significant confounding factor. Although we tried to eliminate some of the genetic variability by inserting the reference and variant transgenic constructs into the same genomic location and swapping out mutant chromosomes (y w) that are known to affect behavioral outcomes, we did not isogenize all chromosomes through multiple rounds of back crossing to facilitate the speed of our screen. Hence, additional behavioral studies performed on a more standardized genetic background (e.g., cantonized flies) will likely provide additional information regarding the role of these genes in fly behavior, which could provide additional insights into their links to ASD. Finally, we would like to emphasize that this work should be considered as a pilot screen rather than a comprehensive screen of de novo missense variants identified in the SSC. The genetic variants that we were able to study in our screen were limited by the availability of intronic MiMIC elements in fly genes because recombinase-mediated cassette exchange (RMCE) was the only efficient way to generate TG4 lines at the time of project initiation. With the advent of CRISPR-mediated integration cassette (CRIMIC) and KozakGAL4 insertion technologies (Lee et al., 2018; Kanca et al., 2021), virtually any fly gene is now targetable to generate a strong LoF allele with GAL4 expression that allows for subsequent humanization experiments. Therefore, future screening strategies could employ these and other emerging techniques to assess functional consequences of many rare missense variants found in ASD or other disease patients.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Shinya Yamamoto (yamamoto@bcm.edu).

Materials availability

The fly lines generated in this study have been deposited to the Bloomington Drosophila Stock Center (BDSC). Please see key resources table for unique identifiers. Please contact the Lead Contact for further information.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-elav	DSHB Cat# 7E8A10	RRID:AB_528218
anti-Repo	DSHB Cat# 8D12	RRID:AB_528448
anti-mouse-647	Jackson ImmunoResearch Labs Cat# 715-605-151	RRID:AB_2340863
anti-rat-Cy3	Jackson ImmunoResearch Labs Cat# 712-165-153	RRID:AB_2340667
anti-GFP	Abcam Cat# ab6662	RRID:AB_305635
anti-HA	BioLegend Cat# 902301	RRID:AB_2565018
anti-actin	Millipore Cat# MAB1501	RRID:AB_2223041
Experimental models: Organisms/strains
UAS-2xEGFP, hs-Cre, vas-dφC31	Diao et al., 2015	N/A
Tub-Gal4	BDSC	RRID: BDSC_5138
nubbin-GAL4	BDSC	RRID: BDSC_51635
GMR-GAL4	BDSC	RRID: BDSC_1104
UAS-nls.GFP	BDSC	RRID: BDSC_4775
UAS-nls.mCherry	BDSC	RRID: BDSC_38424
nSyb-GAL4	BDSC	RRID: BDSC_51635
Rh1-GAL4	BDSC	RRID: BDSC_8691
pnr-GAL4	BDSC	RRID: BDSC_3039
CG4562 TG4	BDSC	RRID:BDSC_76740
Abl TG4	BDSC	RRID:BDSC_67429
Adk1 TG4	Lee et al., 2018	N/A
CG7470 TG4	BDSC	RRID:BDSC_76749
CG8665 TG4	BDSC	RRID:BDSC_66811
RhoGAP19D TG4	BDSC	RRID:BDSC_76687
osa TG4	Lee et al., 2018	N/A
PMCA TG4	BDSC	RRID:BDSC_76741
Cat TG4	BDSC	RRID:BDSC_76660
Cep135 TG4	BDSC	RRID:BDSC_66853
CG31637 TG4	BDSC	RRID:BDSC_76647
CIC-a TG4	BDSC	RRID:BDSC_66801
CLIP-190 TG4	BDSC	RRID:BDSC_66834
Dh44-R2 TG4	BDSC	RRID:BDSC_66865
b TG4	BDSC	RRID:BDSC_76724
arm TG4	BDSC	RRID:BDSC_66903
mbc TG4	BDSC	RRID:BDSC_66840
spg TG4	BDSC	RRID:BDSC_76205
CG17684 TG4	Lee et al., 2018	N/A
shot TG4	BDSC	RRID:BDSC_76760
fne TG4	BDSC	RRID:BDSC_77796
dom TG4	BDSC	RRID:BDSC_76192
Eph TG4	BDSC	RRID:BDSC_66800
fry TG4	BDSC	RRID:BDSC_76736
Gclc TG4	BDSC	RRID:BDSC_76654
dally TG4	BDSC	RRID:BDSC_66830
GluRIB TG4	BDSC	RRID:BDSC_76135
Nmdar2 TG4	BDSC	RRID:BDSC_76705
Gprk2 TG4	BDSC	RRID:BDSC_66828
Lerp TG4	BDSC	RRID:BDSC_77798
Pits TG4	BDSC	RRID:BDSC_77731
if TG4	BDSC	RRID:BDSC_66867
Lpt TG4	BDSC	RRID:BDSC_76714
wb TG4	BDSC	RRID:BDSC_76189
Lrch TG4	BDSC	RRID:BDSC_77756
LRP1 TG4	BDSC	RRID:BDSC_76640
Rab3-GEF TG4	BDSC	RRID:BDSC_76623
hppy TG4	BDSC	RRID:BDSC_67447
mbl TG4	BDSC	RRID:BDSC_66779
msn TG4	BDSC	RRID:BDSC_76204
Mhc TG4	BDSC	RRID:BDSC_76653
ck TG4	BDSC	RRID:BDSC_76720
Nlg3 TG4	BDSC	RRID:BDSC_76134
Cad99C TG4	BDSC	RRID:BDSC_67483
a TG4	BDSC	RRID:BDSC_76725
pli TG4	BDSC	RRID:BDSC_77693
Piezo TG4	BDSC	RRID:BDSC_76658
ptx1 TG4	BDSC	RRID:BDSC_67497
l(1)G0289 TG4	BDSC	RRID:BDSC_67467
CG31211 TG4	BDSC	RRID:BDSC_76718
Spn TG4	This study	N/A
CG6767 TG4	BDSC	RRID:BDSC_77797
otk TG4	BDSC	RRID:BDSC_76759
Lar TG4	BDSC	RRID:BDSC_67451
Pxn TG4	BDSC	RRID:BDSC_66850
CG5521 TG4	BDSC	RRID:BDSC_76180
emp TG4	BDSC	RRID:BDSC_66904
sdk TG4	BDSC	RRID:BDSC_76628
retm TG4	BDSC	RRID:BDSC_66816
Sema5c TG4	BDSC	RRID:BDSC_77809
CG6293 TG4	BDSC	RRID:BDSC_76761
CG18304 TG4	BDSC	RRID:BDSC_76128
rols TG4	BDSC	RRID:BDSC_76150
CG7744 TG4	BDSC	RRID:BDSC_76662
Trpm TG4	BDSC	RRID:BDSC_77748
Nipped-A TG4	BDSC	RRID:BDSC_76723
gig TG4	BDSC	RRID:BDSC_67515
Tusp TG4	BDSC	RRID:BDSC_66798
unc80 TG4	BDSC	RRID:BDSC_76686
Usp30 TG4	BDSC	RRID:BDSC_76704
bchs TG4	BDSC	RRID:BDSC_76762
CG6225 TG4	BDSC	RRID:BDSC_76769
Yip1d1 TG4	BDSC	RRID:BDSC_67492
Ance TG4	BDSC	RRID:BDSC_76676
Ace TG4	BDSC	RRID:BDSC_76688
Aldh-III TG4	BDSC	RRID:BDSC_77692
CG33298 TG4	BDSC	RRID:BDSC_76700
IRSp53 TG4	BDSC	RRID:BDSC_67637
Best1 TG4	BDSC	RRID:BDSC_76671
tok TG4	BDSC	RRID:BDSC_76679
CASK TG4	BDSC	RRID:BDSC_76631
Ddr TG4	This study	N/A
cv-c TG4	This study	N/A
CRMP TG4	This study	N/A
CG11594 TG4	This study	N/A
GluClalpha TG4	BDSC	RRID:BDSC_77841
Galphao TG4	This study	N/A
5-HT1B TG4	BDSC	RRID:BDSC_76668
ltp-r83A TG4	This study	N/A
Shal TG4	This study	N/A
Kdm2 TG4	This study	N/A
Pvr TG4	BDSC	RRID:BDSC_76657
beta-Man TG4	This study	N/A
drpr TG4	This study	N/A
Smr TG4	BDSC	RRID:BDSC_76743
Ndg TG4	BDSC	RRID:BDSC_76768
Nos TG4	BDSC	RRID:BDSC_76766
SIFaR TG4	BDSC	RRID:BDSC_76670
NetB TG4	BDSC	RRID:BDSC_76730
PH4alphaEFB TG4	BDSC	RRID:BDSC_76678
PCB TG4	BDSC	RRID:BDSC_66832
Pdk TG4	BDSC	RRID:BDSC_77785
PKD TG4	BDSC	RRID:BDSC_76706
Rim TG4	This study	N/A
CG9098 TG4	BDSC	RRID:BDSC_76763
ctrip TG4	BDSC	RRID:BDSC_76764
CG1815 TG4	This study	N/A
UAS-ABCC4	BDSC	RRID:BDSC_78481
UAS-ABCC4.M276V	BDSC	RRID:BDSC_92726
UAS-ABCC5	BDSC	RRID:BDSC_78508
UAS-ABCC5.R697W	BDSC	RRID:BDSC_92728
UAS-ABCC5. T1046M	BDSC	RRID:BDSC_92727
UAS-ABL2.A1084T	BDSC	RRID:BDSC_92729
UAS-ABL2	BDSC	RRID:BDSC_78453
UAS-ACE.Y818C	BDSC	RRID:BDSC_92730
UAS-ACE	This study	N/A
UAS-ACHE	BDSC	RRID:BDSC_78466
UAS-ACHE.G151R	BDSC	RRID:BDSC_92732
UAS-ACHE.P548L	BDSC	RRID:BDSC_92731
UAS-AK1	BDSC	RRID:BDSC 78462
UAS-AK1.S58L	BDSC	RRID:BDSC_92733
UAS-ALDH18A1.D703H.HA	BDSC	RRID:BDSC_92734
UAS-ALDH18A1.HA	BDSC	RRID:BDSC_78488
UAS-ALDH1L1.N900H	BDSC	RRID:BDSC_92735
UAS-ALDH1L1	This study	N/A
UAS-ALDH3A1.F402L.HA	BDSC	RRID:BDSC_92926
UAS-ALDH3A1.HA	BDSC	RRID:BDSC_78497
UAS-ATP10A	This study	N/A
UAS-ATP2B2.T818M	BDSC	RRID:BDSC_92952
UAS-ATP2B2	This study	N/A
UAS-ATP2B4	BDSC	RRID:BDSC_78455
UAS-BAIAP2L1.A481V	BDSC	RRID:BDSC_92922
UAS-BAIAP2L1	BDSC	RRID:BDSC_78446
UAS-BCHE.HA	This study	N/A
UAS-BCHE.F433V.HA	This study	N/A
UAS-BEST3	This study	N/A
UAS-BEST3.R130S	This study	N/A
UAS-BMP1.G927S.HA	BDSC	RRID:BDSC_92931
UAS-BMP1.HA	BDSC	RRID:BDSC_77944
UAS-CAMK2A	This study	N/A
UAS-CAMK2A.E183V	This study	N/A
UAS-CARS.HA	BDSC	RRID:BDSC_79001
UAS-CARS.N348S.HA	BDSC	RRID:BDSC_92918
UAS-CASK	This study	N/A
UAS-CAT.HA	BDSC	RRID:BDSC_78471
UAS-CAT.G204E.HA	This study	N/A
UAS-CEP135.HA	BDSC	RRID:BDSC_78458
UAS-CEP135.S947P.HA	This study	N/A
UAS-CHST2.HA	BDSC	RRID:BDSC_78472
UAS-CHST2.R52P.HA	This study	N/A
UAS-CLCNKB.HA	BDSC	RRID:BDSC_77934
UAS-CLCNKB.M176I.HA	BDSC	RRID:BDSC_92910
UAS-CLIP2.G13W.HA	BDSC	RRID:BDSC_92934
UAS-CLIP2.HA	BDSC	RRID:BDSC_78461
UAS-CSAD.HA	BDSC	RRID:BDSC_78498
UAS-CSAD.A411V.HA	This study	N/A
UAS-CTNNB1	BDSC	RRID:BDSC_78496
UAS-CTNNB1.T551M	This study	N/A
UAS-DDR2.B	BDSC	RRID:BDSC_78483
UAS-DLC1.HA	BDSC	RRID:BDSC_78089
UAS-DPP6	This study	N/A
UAS-DPYSL2.HA	BDSC	RRID:BDSC_77938
UAS-DPYSL2.R496C.HA	BDSC	RRID:BDSC_92946
UAS-DPYSL3.HA	BDSC	RRID:BDSC_77939
UAS-DPYSL3.V139I.HA	BDSC	RRID:BDSC_92945
UAS-ELAVL3	BDSC	RRID:BDSC_78505
UAS-ELAVL3.L186P	BDSC	RRID:BDSC_92912
UAS-EP400.B	BDSC	RRID:BDSC_78493
UAS-EPHA1.HA	BDSC	RRID:BDSC_77931
UAS-EPHA1.V567I.HA	BDSC	RRID:BDSC_92963
UAS-EPHB1	BDSC	RRID:BDSC_78473
UAS-EPHB1.V916M	BDSC	RRID:BDSC_92932
UAS-EPT1	BDSC	RRID:BDSC_78485
UAS-EPT1.H82N	This study	N/A
UAS-EXD2.E513D.HA	BDSC	RRID:BDSC_92942
UAS-EXD2.HA	BDSC	RRID:BDSC_77933
UAS-FGGY	BDSC	RRID:BDSC_78091
UAS-GCLC.HA	This study	N/A
UAS-GCLC.R128W.HA	This study	N/A
UAS-GLRA2.HA	BDSC	RRID:BDSC_77954
UAS-GLRA2.N136S.HA	BDSC	RRID:BDSC_92915
UAS-GNAO1.HA	BDSC	RRID:BDSC_79003
UAS-GPC5.HA	BDSC	RRID:BDSC_77936
UAS-GPC5.M133T.HA	BDSC	RRID:BDSC_92913
UAS-GRIA1.HA	BDSC	RRID:BDSC_78474
UAS-GRIA1.R218H.HA	BDSC	RRID:BDSC_92930
UAS-GRIK5	BDSC	RRID:BDSC_77957
UAS-GRK4.HA	BDSC	RRID:BDSC_78503
UAS-GRK4.P385A.HA	BDSC	RRID:BDSC_92916
UAS-HTR1D.HA	This study	N/A
UAS-HTR1D.T99N.HA	This study	N/A
UAS-IGF2R.HA	BDSC	RRID:BDSC_78454
UAS-IRF2BPL	BDSC	RRID:BDSC_78509
UAS-IRF2BPL.F30L	This study	N/A
UAS-IRF2BPL.N701X	This study	N/A
UAS-ITGA8.HA	BDSC	RRID:BDSC_78501
UAS-ITGA8.R748C.HA	BDSC	RRID:BDSC_92924
UAS-JUP.HA	BDSC	RRID:BDSC_78500
UAS-JUP.N690S.HA	This study	N/A
UAS-KCND3.HA	BDSC	RRID:BDSC_78475
UAS-KCND3.R86P.HA	BDSC	RRID:BDSC_92921
UAS-KCNH8.HA	BDSC	RRID:BDSC_78470
UAS-KDM2A	BDSC	RRID:BDSC_78476
UAS-KDM2A.R449K	BDSC	RRID:BDSC_92927
UAS-KDR	BDSC	RRID:BDSC_78451
UAS-KDR.D1171N	BDSC	RRID:BDSC_92955
UAS-LAMA2	BDSC	RRID:BDSC_79000
UAS-LRCH4	BDSC	RRID:BDSC_78477
UAS-LRCH4.V42M	BDSC	RRID:BDSC_92928
UAS-MADD	BDSC	RRID:BDSC_78457
UAS-MADD.R514C	BDSC	RRID:BDSC_92933
UAS-MANBA	BDSC	RRID:BDSC_77960
UAS-MAP4K1.HA	BDSC	RRID:BDSC_77946
UAS-MAP4K1.M725T.HA	This study	N/A
UAS-MBNL1.HA	BDSC	RRID:BDSC_78467
UAS-MBNL1.V45A.HA	BDSC	RRID:BDSC 92908
UAS-MEGF11	BDSC	RRID:BDSC_78460
UAS-MEGF11.R911C	This study	N/A
UAS-MINK1.HA	BDSC	RRID:BDSC_78489
UAS-MINK1.C269R.HA	This study	N/A
UAS-MYH9.HA	BDSC	RRID:BDSC_79002
UAS-MYH9.R1571Q.HA	This study	N/A
UAS-NCOR1	BDSC	RRID:BDSC_78486
UAS-NCOR1.P569S	BDSC	RRID:BDSC_92953
UAS-NID2	This study	N/A
UAS-NLGN1	This study	N/A
UAS-NLGN1.H795Y	BDSC	RRID:BDSC_92936
UAS-NLGN3.R195W	This study	N/A
UAS-NOS3	This study	N/A
UAS-NPFFR2	BDSC	RRID:BDSC_78478
UAS-NPFFR2.M163I	BDSC	RRID:BDSC_92917
UAS-NR2F1	BDSC	RRID:BDSC_77959
UAS-NR2F1.R404H	This study	N/A
UAS-NTN1.A449D	BDSC	RRID:BDSC_92958
UAS-NTN1	BDSC	RRID:BDSC_78495
UAS-NTN5	BDSC	RRID:BDSC_78507
UAS-P4HA2.G153E.HA	BDSC	RRID:BDSC_92938
UAS-P4HA2.HA	BDSC	RRID:BDSC_77935
UAS-PC.HA	BDSC	RRID:BDSC_77928
UAS-PC.P1042R.HA	This study	N/A
UAS-PDGFRB.A366T.HA	BDSC	RRID:BDSC_92941
UAS-PDGFRB	This study	N/A
UAS-PDK2.HA	BDSC	RRID:BDSC_77949
UAS-PDK2.R120Q	BDSC	RRID:BDSC_92962
UAS-PEAR1	BDSC	RRID:BDSC_77956
UAS-PEAR1.T824I	BDSC	RRID:BDSC_92911
UAS-PELI1.A270V.HA	BDSC	RRID:BDSC_92919
UAS-PELI1.HA	BDSC	RRID:BDSC_78506
UAS-PITX1.HA	BDSC	RRID:BDSC_92043
UAS-PITX1.L242F.HA	BDSC	RRID:BDSC_92914
UAS-PLXDC1	BDSC	RRID:BDSC_78504
UAS-PLXDC1.R42Q	BDSC	RRID:BDSC_92909
UAS-PPP1R9A	BDSC	RRID:BDSC_78499
UAS-PRKD1	BDSC	RRID:BDSC_78494
UAS-PRKD1.R441W	BDSC	RRID:BDSC_92950
UAS-PRPS1L1	This study	N/A
UAS-PRPS1L1.G61D	BDSC	RRID:BDSC_92929
UAS-PTK7.R570Q	This study	N/A
UAS-PTPRF	BDSC	RRID:BDSC_92404
UAS-PTPRF.S334R	BDSC	RRID:BDSC_92937
UAS-PXDN.HA	BDSC	RRID:BDSC_77955
UAS-PXDN.R643Q	BDSC	RRID:BDSC_92951
UAS-RALGAPA1	BDSC	RRID:BDSC_78449
UAS-RALGAPA1.LL1769L	This study	N/A
UAS-RIMS2.HA	BDSC	RRID:BDSC_77937
UAS-SCARB2.HA	BDSC	RRID:BDSC_77929
UAS-SCARB2.V173A.HA	This study	N/A
UAS-SDK2.HA	BDSC	RRID:BDSC_78487
UAS-SEC14L5	BDSC	RRID:BDSC_78092
UAS-SH2D3C.HA	BDSC	RRID:BDSC_78492
UAS-SH2D3C.R227Q.HA	This study	N/A
UAS-SLC23A1	BDSC	RRID:BDSC_78459
UAS-SLC23A1. L465M	BDSC	RRID:BDSC_92943
UAS-SLC8A2	BDSC	RRID:BDSC_78490
UAS-SLC8A2.G792R	This study	N/A
UAS-SLCO4A1	BDSC	RRID:BDSC_78479
UAS-SLCO4A1.V679I	BDSC	RRID:BDSC_92920
UAS-SOGA3.HA	BDSC	RRID:BDSC_78469
UAS-SOGA3..R873P.HA	This study	N/A
UAS-SRCAP	BDSC	RRID:BDSC_78450
UAS-SRCAP.2137del	This study	N/A
UAS-SRCAP.G1937S	This study	N/A
UAS-TANC2	BDSC	RRID:BDSC_78452
UAS-TANC2.H1689R	BDSC	RRID:BDSC_92925
UAS-TRIP12	BDSC	RRID:BDSC_78518
UAS-TRIP12.R1643Q	BDSC	RRID:BDSC_92956
UAS-TRPM1	BDSC	RRID:BDSC_78517
UAS-TRPM1.F794L	BDSC	RRID:BDSC_92954
UAS-TRPM6.HA	BDSC	RRID:BDSC_77958
UAS-TRPM6.A641E.HA	BDSC	RRID:BDSC_92939
UAS-TRPM6.T2011P.HA	BDSC	RRID:BDSC_92940
UAS-TRPM7.HA	BDSC	RRID:BDSC_78447
UAS-TRPM7.T379A.HA	BDSC	RRID:BDSC_92923
UAS-TSC2.HA	BDSC	RRID:BDSC_78465
UAS-TSC2.R1557W.HA	BDSC	RRID:BDSC_92949
UAS-TSC2.R548M.HA	BDSC	RRID:BDSC_92947
UAS-TULP4	BDSC	RRID:BDSC_92408
UAS-USP30.HA	BDSC	RRID:BDSC_78480
UAS-USP30.P200S.HA	BDSC	RRID:BDSC_92957
UAS-YIPF5.HA	BDSC	RRID:BDSC_82197
UAS-ZMYND8.HA	BDSC	RRID:BDSC_77945
Deposited data
TG4 imaging videos	This paper: Mendeley Data	https://doi.org/10.1016/10.17632/64jrz799sb.1
Human GLRA2 variant p.Thr296Met	ClinVar	SCV002055997
Human GLRA2 variant p.Phe47Ser	ClinVar	SCV002056017
Human GLRA2 variant p.Ile259Met	ClinVar	SCV002056018
Human GLRA2 variant p.Arg252Cys	ClinVar	SCV002056022
Human GLRA2 variant p.Ala288Thr	ClinVar	SCV002056021
Human GLRA2 variant p.Pro396Thr	ClinVar	SCV002056020
Human GLRA2 variant p.Pro400Leu	ClinVar	SCV002056019
Human GLRA2 variant p.Arg445Gln	ClinVar	SCV002056023
Recombinant DNA
pUASg-HA.attB	Drosophila Genomics Resource Center,	DGRC_1423
Software and algorithms
Prism8	Graph Pad	https://www.graphpad.com
Zen Blue	Zeiss	https://www.zeiss.com/microscopy/us/products/microscope-software/zen-lite.html
Zen Black	Zeiss	https://www.zeiss.com/microscopy/us/products/microscope-software/zen-lite.html
Snapgene	Snapgene	https://www.snapgene.com
Imaris	Imaris	https://imaris.oxinst.com/
Lab Chart	Ad instruments	https://www.adinstruments.com/products/labchart

Data and code availability

• Confocal imaging movies of neuron/glia colocalization with TG4>UAS-nls.mCherry have been deposited at Mendeley Data and are publicly available as of the date of publication. The DOI is listed in the key resources table. De-identified GLRA2 variant data have been deposited at ClinVar. They are publicly available as of the date of publication. Accession numbers are listed in the key resources table.

• This study didn’t generate any code.

• Any additional information required to reanalyze the data reported in this work paper is available from the Lead Contact upon reasonable request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Generation of TG4 lines

All TG4 alleles in this study were generated by ϕC31-mediated recombination-mediated cassette exchange of MiMIC (Minos mediated integration cassette) insertion lines (Gnerer et al., 2015; Nagarkar-Jaiswal et al., 2015; Venken et al., 2011). Conversion of the original MiMIC element was performed via genetic by crossing UAS-2xEGFP, hs-Cre,vas-dϕC31, Trojan T2A-GAL4 triplet flies to each MiMIC strain and following a crossing scheme (Diao et al., 2015). 73 TG4 lines were described previously but not extensively characterized (Lee et al., 2018), while 35 lines were generated specifically for this study.

Generation of UAS-human cDNA lines

The majority of reference human cDNA clones were obtained in either pDONR221 or pDONR223 donor vectors. The LR clonase II (ThermoFisher) enzyme was used to shuttle ORFs into the p.UASg-HA.attB destination vector via Gateway^™ cloning. Some ORFs that were not Gateway compatible were obtained from additional sources (Table S2), amplified with flanking attB sites and cloned into pDONR223 plasmid using BP clonase II (ThermoFisher). Sequence-verified variants were generated in the DONR vectors by either site-directed mutagenesis (SDM) or High-Throughput Mutagenesis (HiTM) as previously described (Tsang et al., 2016). SDM was performed with primers generated using NEBaseChanger (Table S3) with the Q5® mutagenesis kit (NEB). Sequence-verified reference and variant ORFs in the pUASg-HA.attB destination plasmid were microinjected into ~200 embryos in one three attP docking sites (attP86Fb, VK00037 or VK00033) docking sites by ϕC31 mediated transgenesis (Bischof et al., 2007; Venken et al., 2006). The docking site of choice were selected based on the genomic locus of the corresponding fly gene. In principal, VK00037 docking site on the 2^nd chromosome was used for human genes that correspond to fly genes on the X, 3^rd or 4^th chromosome, whereas VK00033 or attP86Fb docking site on the 3^rd chromosome was used for human genes that correspond to fly genes on the 2^nd chromosome.

Fly husbandry

Unless otherwise noted, all flies used in experiments were grown in a temperature and humidity-controlled incubator at 25°C and 50% humidity on a 12-hour light/dark cycle. Some experiments were conducted at different temperatures that are specifically indicated in the text and figures. Stocks were reared on standard fly food (water, yeast, soy flour, cornmeal, agar, corn syrup, and propionic acid) at room temperature (~22°C) and routinely maintained.

Fly stocks used that were not generated here

tub-GAL4 (y¹ w*; P{w[+mC]=tubP-GAL4}LL7/TM3, Sb¹ Ser¹) BDSC_5138, GMR-GAL4 (w*; P{w[+mC]=GAL4-ninaE.GMR}12) BDSC_1104, nub-GAL4 (P{GawB}nubbin-AC-62) (Calleja et al., 2000), nSyb-GAL4 (y¹ w*; P{nSyb-GAL4.S}3) BDSC_51635, Rh1-GAL4 (P{ry[+t7.2]=rh1-GAL4}3, ry[506]) BDSC_8691, pnr-GAL4 (y¹ w¹¹¹⁸; P{w[+mW.hs] = GawB}pnr[MD237]/TM3, P{w[+mC]=UAS-y.C}MC2, Ser¹) BDSC_3039, UAS-nlsGFP (w¹¹¹⁸; P{w[+mC]=UAS-GFP.nls}14) BDSC_4775, and UAS-nls.mCherry (w*; P{w[+mC]=UAS-mCherry.NLS}3) BDSC_38424.

Patient recruitment and consent

Affected individuals were investigated by their referring physicians at local sites. Prior to research studies, informed written consent for testing and publication was obtained according to the institutional review boards (IRB) and ethnics committees of each institution. Individuals who were ascertained in diagnostic testing procedures (and/or their legal guardians) gave clinical written informed consent for testing, and their permission for inclusion of their anonymized data in this cohort series. This was obtained using standard forms at each local site by the responsible referring physicians.

GLRA2 subject case histories

Subject 1 is an 8 years old female with global developmental and cognitive delay. Pregnancy was naturally conceived and uncomplicated, other than decreased fetal movements noted by the mother. She was delivered at term (39 weeks gestational age) via c/section due to breech presentation. Birth weight was 3,600 grams. Neonatal period was uneventful. There were no feeding difficulties and her growth remained within the normal limits. She was delayed with all her milestones but most significantly for speech (walked at 18 months, scribbled with a crayon at 3.5 years, first words at 24 months and combined words to sentences at 4–5 years of age). She was diagnosed with mixed expressive-receptive speech delay and received speech therapy, occupational therapy and physical therapy interventions. In school she exhibits learning problems, inattention and is below her grade level. She has a modified curriculum and is receiving resources in reading and math. There is no history of developmental regression or seizures. The medical history is otherwise significant for nystagmus that was first noted in infancy and improved with age, as well as myopia and astigmatism requiring corrective glasses. Family ethnicity is Hispanic and the family history is non-contributory. The patient had a normal brain MRI (magnetic resonance imaging) at 6 months of age. EEG (electroencephalography) at 4 years of age showed a slow and poorly formed background, indicative of mild encephalopathy, but did not detect epileptiform activity. Genetic testing included: mitochondrial DNA sequencing which detected a pathogenic variant m.13042 G>A though at heteroplasmy level of 1.9%, which was felt unlikely to explain the phenotype. CMA (chromosomal microarray) was negative. Trio whole exome sequencing (WES) detected a de novo, heterozygous variant of unknown clinical significance in GLRA2, c.887C>T, p.Thr296Met (NC_000023.10, chrX: g.14627284C>T). This variant is absent in gnomAD. More recent clinical reanalysis of exome data did not detect any other candidates that may explain the phenotype.

Subject 2 is a 6 years old female with epilepsy, developmental delay (DD), mild intellectual disability (ID) and autism spectrum disorder (ASD). Pregnancy was uncomplicated and she was delivered at term (41 weeks gestational age) via vaginal delivery with vacuum extraction. The neonatal period was uneventful. At the age of 6 months, she developed a severe epileptic encephalopathy with myoclonic seizures. Seizure control was achieved with medications, and she has been seizure-free without medications since the age of about two years old. Delayed psychomotor development was noted, most significantly for her speech with a mixed expressive-receptive speech delay (non-verbal). Her ability to concentrate is poor and she displays mood swings. The medical history is otherwise significant for nystagmus that was first noted in infancy (6 weeks old) and improved with age, and sleep disturbance. She has mild microcephaly [< 1st centile: −2.84 standard deviation (SD)] and mild bilateral cutaneous 3rd-4th syndactyly, with no other congenital anomalies. Family ethnicity is European (German/Italian) and the family history is significant for a maternal aunt that had epilepsy in adulthood but her cognitive development was normal. Brain MRI showed delayed myelination at 7 months old and a small arachnoid cyst. EEG was abnormal for bilateral synchronized, sometimes high amplitude spike/polyspike-waves-complexes, and bitemporo-occipital hints for severe functional defects with epileptic potentials. Chromosomal analysis, Angelman syndrome methylation study, epilepsy next generation sequencing (NGS) gene panel and MECP2 sequencing were negative. Trio WES detected a de novo, heterozygous variant of unknown clinical significance in GLRA2, c.887C>T, p.Thr296Met (NC_000023.10, chrX: g.14627284C>T). This variant is absent in gnomAD.

Subject 3 is a 5 year and 6 months old female with DD, microcephaly, abnormal eye movements and ataxic gait. Pregnancy was uncomplicated and she was born at term via c/section. Abnormal eye movements were noticed two weeks after birth, during hospitalization due to a lower respiratory tract infection. At the age of 6 months, clinical examination revealed mildly delayed developmental milestones and erratic conjugate eye movements akin to opsoclonus. At age 4 years OFC (occipitofrontal circumference) was 43 cm (< 1st centile: −4.28 SD) and ophthalmological evaluation revealed alternating exotropia, for which patching therapy was initiated. Language was limited to a few words and neuropsychological evaluation documented moderate developmental delay (Bayley-III). The patient could walk unsupported with ataxic gait. At age 5 years 6 months, erratic eye movements were considerably reduced and she could walk independently but her expressive language was still limited to a few words, with delayed receptive speech and nonverbal communicative skills. Family ethnicity is European and the family history is unremarkable. Brain MRI at 6 months of age showed mild cortical atrophy with thinning of the corpus callosum. EEG, while awake and asleep, laboratory and metabolic investigations were unremarkable. Array-CGH (comparative genomic hybridization) highlighted a maternally inherited 3q25.32 duplication (chr3:157746089–158324659, hg19) that was interpreted as likely benign. Trio WES detected a de novo heterozygous variant in GLRA2, c.887C>T, p.Thr296Met (NC_000023.10, chrX: g.14627284C>T). This variant is absent in gnomAD. In addition, it detected a de novo variant in CACNA1B, c.5381C>T, p.Thr1794Met (NC_000009.11:g.141000212C>T), which is a variant of unknown significance in a gene that is linked to an autosomal recessive condition (Neurodevelopmental disorder with seizures and nonepileptic hyperkinetic movements, MIM #618497). Failure to identify a second allele in this gene reduces the likelihood that this variant is responsible for this patient’s phenotype.

Subject 4 was a female infant with seizures and severe developmental delay who passed away at 7 months of age secondary to complications of COVID-19 infection. Pregnancy was uneventful and she was born at term (40 weeks gestational age). She was noted to have focal seizures at 2–3 weeks of age, and was diagnosed with infantile spasms when she was 5 months old. At 6 months of age she was not reaching for objects, not sitting up and only making high-pitched sounds. She had borderline microcephaly with dysmorphic features including midface retrusion, apparent hypotelorism, deep set eyes, thick eyebrows, downturned corners of the mouth, and wide-spaced nipples. Family ethnicity is Hispanic and the family history is non-contributory. She had normal plasma and CSF (cerebrospinal fluid) lactate, pipecolic acid and piperideine-6-carboxylate, ammonia, urine organic acids, plasma amino acids, acylcarnitine profile, and CSF amino acids. An Epilepsy gene panel was non-diagnostic. Trio WES detected a de novo heterozygous variant in GLRA2, c.887C>T, p.Thr296Met (NC_000023.10, chrX: g.14627284C>T). This variant is absent in gnomAD.

Subject 5 is a 5 years and 4 months old female with global developmental delay. She was born in Afghanistan to consanguineous parents, and there is limited information available regarding her birth history. Pregnancy was uneventful. She walked at 2 years and is currently non-verbal. She has ptosis and oculomotor apraxia. On physical exam she is noted to have broad halluces. Brain MRI at 5 years old was normal. Metabolic testing was non-revealing. Array-CGH, molecular genetic testing of FMR1 and screening for congenital disorders of glycosylation resulted negative. WES detected a de novo, heterozygous variant of unknown clinical significance in GLRA2, c.887C>T, p.Thr296Met (NC_000023.10, chrX: g.14627284C>T). This variant is absent in gnomAD.

Subject 6 is a 9 months old female with developmental delay, West syndrome, microcephaly, up-beat nystagmus and myopia. She was born at term (38 gestational weeks, birth weight 2770 g (10^th percentile), OFC 34 cm (33^rd percentile)) via c/section due to breech presentation. She had postnatal respiratory distress. Abnormal eye movements were noticed at 6 weeks of age. Ophthalmologic assessment was normal with normal fixation apart from up-beat nystagmus. Clinical evaluation at 4 months of age revealed developmental delay and an OFC of 38.5 cm (3rd percentile). At approximately 7 months of age, she developed infantile spasms with an intermittent hypsarrhythmic pattern in EEG (West syndrome). Therapy with Sulthiame resulted in resolution of seizure activity and EEG normalisation. She has made developmental progress since starting treatment with antiepileptic medications. At 9 months of age she is showing mild gross motor delay. Family ethnicity is European, and the family history is non-contributory. Brain MRI, abdominal ultrasound, and metabolic screening labs (including plasma amino acids, acylcarnitine profile, CSF analysis and CSF neurotransmitters) were normal. Trio exome sequencing detected a heterozygous de novo missense variant in GLRA2 (NM_002063.4), c.887C>T, p. Thr296Met (NC_000023.10, chrX: g.14627284C>T, hg19). This variant is absent in gnomAD.

Subject 7 is a 6 years and 7 months old female with a history of infantile spasms, epilepsy and intellectual disability. She was born at term and first presented with infantile spasms at 3 months of age. This evolved to atonic and tonic-clonic seizures as she grew up. She was delayed with all milestones (walked at 4.5 years old and remains non-verbal). She had nystagmus that improved with age and strabismus. The medical history is otherwise significant for hyperactivity, inattention and sleep disturbance. Her ethnicity is African (Senegal). Brain MRI at 3 years of age showed cortical and white matter atrophy, including vermian atrophy. EEG showed hypsarrythmia at onset and she had normal interictal EEG afterwards. She had normal SNP (single nucleotide polymorphism)-array, negative targeted epilepsy panel and negative metabolic lab results. Trio WES identified a de novo heterozygous variant in GLRA2, c.140T>C, p.Phe47Ser (NC_000023.10, chrX: g.14550432T>C). This variant is absent in gnomAD.

Subject 8 is a 5 years and 3 months old female with mild developmental delay and learning disabilities. Pregnancy was naturally conceived and uncomplicated. She was born near term following premature rupture of membrane. Birth weight was 3,280g, birth length was 50cm and OFC was 33.5cm. Apgar scores was 8/9 at 1^st/5^th minutes of life. She had feeding difficulties during the neonatal period. With regards to her milestones, she was able to sit unsupported at 9 months and walked independently at 17 months. Speech was not delayed but she has difficulties in pronunciation and articulation. There is no history of developmental regression. She has a School aide for learning difficulties but is interacting well with other children. She is receiving therapy for fine motor difficulties and is followed by a psychiatrist with a diagnosis of infantile psychosis. She also has a history of sleep disturbance (short sleep duration). The family history is non-contributory, she is the only child to her parents as a couple and has two paternal half-siblings that are healthy. Her growth is normal, with weight at 22 kg (+1.1 SD), height at 117.5 cm (+1.4 SD) and OFC 51 cm (+0.2 SD). On exam she has mild dysmorphic features, including long face, pointed chin, and overlapping 1^st and 2^nd toes. Although she does not have clinical seizures, EEG was abnormal for left fronto-temporal spike-waves focus which diffuses in the right frontal region, activated by sleep but not meeting criteria for Epilepsy with continuous spike-wave during sleep (CSWS). Brain MRI showed nonspecific dot-like hypersignal in FLAIR of the subcortical white matter of the frontal region. Fragile × testing was negative. Exome sequencing detected a de novo heterozygous variant in GLRA2, c.777C>G, p.Ile259Met (NC_000023.10, chrX: g.14627174C>G). This variant is absent in gnomAD.

Subject 9 is an 11 months old male with hypotonia, DD and dysmorphic craniofacial features. Pregnancy was uncomplicated, he was delivered at term (38 and 3/7 weeks gestational age) and the neonatal period was uneventful. Soon after birth dysmorphic features were noted, including an elongated face, high anterior hairline, epicanthal folds, downslanting palpebral fissures and a bulbous nose. Growth remains within the normal limits. His medical history is otherwise significant for obstructive sleep apnea and strabismus. Family ethnicity is European (Dutch) and the family history is significant for the maternal grandfather who has not further specified unexplained neurological complaints, and which could not be further investigated. Investigations for metabolic disorders, Fragile × syndrome and a SNP-array were normal. Trio WES identified a rare variant in GLRA2, c.754C>T, p.Arg252Cys (NC_000023.10, chrX: g.14627151C>T), which was inherited from mother. No other possible disease explaining variant was identified. The mother displayed skewed × chromosome inactivation (82% on two measurements). The variant was absent in the maternal uncle and the maternal grandmother, but was inherited from the maternal grandfather, who was not available for clinical investigations. His level of functioning remains unknown. This variant is present in one heterozygous female in gnomAD.

Subject 10 is a 7 years old male with epilepsy, DD with regression, and ASD. Pregnancy was uncomplicated. He was born full term via uncomplicated delivery, and was meeting his early developmental milestones. He was speaking in sentences at 2.5 years old when he started having generalized tonic-clonic seizures. He developed staring spells, ataxia, and an increased frequency of myoclonic jerks, which around the age of 6 years old were occurring 20 times per day on average, with 5–6 atonic seizures per day each lasting less than 30 seconds. Following seizure onset, he experienced developmental regression. At 3 years of age he was diagnosed with ASD. At 6 years of age his vocabulary was about 20 words, with gains in development lost following significant seizures. His ethnicity is European, and the family history is significant for a younger brother with ASD, although he has not presented with seizures. Neither mother nor father have a history of seizures or delays. At age 3, EEG depicted generalized slowing and generalized epileptiform discharges associated with myoclonic jerks. MRI showed minimal increased T2 signal intensity on the occipital lobes that was thought to be within normal limits. Genetic testing for Fragile × syndrome, Prader-Willi/Angelman syndromes, and congenital disorders of glycosylation was normal. Additional tests, including plasma amino acids, lysosomal enzymes, and cerebral creatine deficiency were also normal. Array-CGH reported a maternally inherited 1p33 deletion of unknown significance (chr1: 48,688,391–49,922,153). The patient was enrolled to The Manton Center for Orphan Disease Gene Discovery Core protocol. Trio WES identified a maternally inherited variant in GLRA2, c.862G>A, p.Ala288Thr (NC_000023.10, chrX: g.14627259G>A). This variant is absent in gnomAD.

Subject 11 is a 35 years old male with a history of DD, learning disabilities and ASD. Pregnancy was uncomplicated and he was born at term (40 weeks gestational age). Since early childhood he showed slow movement and difficulties in motor coordination. He walked and said his first words at 24 months, and first sentences at 3 years of age. In school learning disabilities were noted, including difficulties in writing, reading, praxias, temporal orientation, calculation, drawing, and visuo-spatial organization. He graduated high school and continued to higher education, though he did not complete a degree. Neuropsychiatric assessment in adulthood was consistent with ASD and social and cognitive deficits. There is no history of seizures. The medical history is otherwise significant for environmental allergies, myopia, and astigmatism. The ethnicity is European, and the family history is non-contributory. The patient had a normal brain MRI at 29 years old. Trio WES identified a maternally inherited variant in GLRA2, c.1186C>A, p.Pro396Thr (NC_000023.10, chrX: g.14748434C>A). This variant is present in 3 heterozygous females and 1 hemizygous male in gnomAD.

Subject 12 is a 15 years old male with a history of epilepsy, DD, expressive language disorder, and ASD. Pregnancy was uncomplicated and he was born at term. He walked at 17 months and said his first words at 13 months. Following febrile convulsions at 18 months of age, more pronounced developmental regression was noted. At 7 years of age, he had focal (partial) motor seizures for which he still receives antiepileptic medications. In school he is enrolled in mainstream classes, but he has learning problems and inattention, and is behind his age-peers. Speech remains delayed, he has limited vocabulary and receives speech therapy. He is frequently agitated and has social anxiety. The medical history is otherwise significant for obesity. The ethnicity is European, and the family history is significant for learning problems in his mother. On exam, he is noted to have mild dysmorphic features, including broad face, Widow’s peak, horizontal and broad eyebrows, long prominent eyelashes, and a broad nasal tip. Brain MRI at 7 years of age showed increased signal intensity in FLAIR of the cortical matter of the parietal region. EEG at that age showed abnormal alpha waves with high polymorphic spikes with a focus in the right temporal region (consistent with the clinical presentation of partial simple seizures). Genetic testing for Fragile × syndrome, and metabolic labs were normal. Trio WES identified a maternally inherited variant in GLRA2, c.1199C>T, p.Pro400Leu (NC_000023.10, chrX: g.14748447C>T). This variant is absent in gnomAD. × chromosome inactivation study in the mother, who is mildly affected, showed moderately unbalanced × inactivation (60:40) in two independent experiments.

Subject 13 is a 6 years old male with intractable epilepsy, developmental delay, and suspected autism spectrum disorder. Pregnancy was uncomplicated and he was delivered at term (39 and 2/7 weeks gestational age, weight: 3440 g (−0.19 SD), length: 49 cm (−1.34 SD), OFC: 35 cm (−0,3 SD)). The neonatal period was uneventful. He learned to walk at the age of 21 months. He spoke his first words at 18 months, but his speech development was delayed. At the age of 4 years and 3 months he first presented with generalized tonic-clonic and tonic seizures. At most recent evaluation he was reported to have 4–5 generalized tonic-clonic or tonic seizures, and absence seizures daily (while being treated with five anti-epileptic medications simultaneously). EEG at age 5 years and 6 months showed excessive beta-activity most likely due to medication; left and right posterior and right frontal Intermittent slowing; and bilateral polyspikes during sleep with a frequency of < 1/minute three times associated with tonic-clonic seizures. Brain MRI at the same age showed no abnormalities. Concomitant to the epilepsy, he was diagnosed with DD with a focus on speech development and cognitive impairment. He is presenting with poor expressive speech (partly able to speak in sentences). The behavioral abnormalities include a short attention span, poor impulse control, sleep disturbances and recently also some compulsive traits have been reported by the mother. An early-childhood autism spectrum disorder is suspected, and an evaluation has been initiated. He visits an integrative Kindergarten with permanent one-on-one care and receives speech therapy, occupational therapy, and physiotherapy. Growth parameters at the last assessment (5 years and 11 months) were normal (weight at 0.51 SD, height at 1.12 SD and OFC at −0,72 SD). On clinical exam, he was noted to have downslanting palpebral fissures and a high forehead. There is no family history of developmental delay and epilepsy. Neonatal metabolic screening was unremarkable. Genetic testing for Fragile × syndrome as well as array-CGH, karyotype and proband exome sequencing were non-revealing. Trio exome sequencing on a research basis identified a maternally inherited hemizygous missense variant in GLRA2 c.1334G>A, p.Arg445Gln (NC_000023.10, chrX: g.14748582G>A). This variant is present in three heterozygous females in gnomAD and has not been observed in a hemizygous state. With Sanger sequencing, it was shown that the variant is absent in the unaffected maternal half-brother of subject 13.

METHOD DETAILS

Ortholog candidate identification

We utilized DIOPT (Hu et al., 2011) to determine the fly ortholog of the 1519 human genes from the SSC. DIOPT scores lower than 4/16 were excluded. If there were multiple fly paralogs with equal DIOPT scores we referred to the weighted score. If the weighted scores were equal, we chose the ortholog in which a Gold MiMIC present in the other paralogs of equal strength.

Electroretinograms (ERG)

ERG recordings on adult flies were performed on nSyb-GAL4 (Pauli et al., 2008) and Rh1-GAL4 (Xiong et al., 2012) driven UAS-GLRA2 at 5 days post-eclosion raised at 25°C in 12h light/12h dark cycle as previously described (Verstreken et al., 2003) using LabChart software (AD instruments). 4–10 flies were examined for each genotype. Recording was repeated at least 3 times per fly. Quantification and statistical analysis was performed using ANOVA followed by Bonferroni’s multiple comparison test using Prism 8.0.

Complementation test of lethality in TG4 lines

Out of the 108 TG4 mutants generated, 65 TG4 mutants were homozygous lethal. Because lethality can be caused by disruption of the gene of interest or due to second site lethal mutations carried on the same chromosome, we performed complementation test using standard methodology. For genes on the 2^nd and 3^rd chromosome, female heterozygous TG4 lines balanced with either SM6a or TM3, Sb, Ser, respectively, were crossed with male flies carrying a corresponding deficiency (Df) that covers the gene of interest (see Table S4). Three independent crosses were set at 25°C for each TG4 line and we determined if any TG4 flies survived to the adult stage in trans with their corresponding Df (TG4/Df). If viable, a second Df line covering the same gene was used to validate this finding to make sure the complementation is not due to some problematic Df lines. If TG4 was viable over two independent Df lines, we ascribed the lethality to a second site mutation on the TG4 chromosome. TG4 that remained lethal in trans with a Df line are be considered to be disrupting an essential gene in flies. For five genes on the X-chromosome of the fly, complementation was performed by first rescuing hemizygous TG4 males with a duplication (Dp) line obtained from BDSC (Table S4), and crossing these rescued flies to female TG4/FM7 flies. If TG4/Y; Dp/+ lines were viable, we ascribed the lethality of TG4 to the gene of interest. All Df and Dp lines were obtained from BDSC, and the specific stock used in our analysis are listed in Table S4.

Through this experiment, we found 65 TG4 mutant lines that were homozygous lethal, and 47 of 65 remained lethal when in trans with a corresponding deficiency line (Figure S1; Table S4). The 47 essential genes in D. mel corresponded to 60 SSC related human genes (Figure S1). The lethality of 18 TG4 lines corresponding to 19 human genes were due to a second site lethal mutation, potentially present in the original MiMIC line, or introduced during RMCE which has been reported previously (Nagarkar-Jaiswal et al., 2015). These TG4 lines together with viable TG4 lines are likely associated with non-essential genes in Drosophila.

Rescue of lethality in TG4 lines by UAS-human cDNA transgenes

In order to assess the ability of human reference or SSC variant cDNAs to rescue lethality observed in TG4 mutants in essential genes, we first double balanced all Df lines that fail to complement a lethal TG4 line with UAS-reference or variant cDNA lines. For genes on the 2^nd chromosome, we generated Df/CyO; UAS-cDNA/(TM3, Sb, Ser) stocks. For genes on the 3^rd chromosome, we generated UAS-cDNA/(CyO); Df/TM3, Sb, Ser. Heterozygous TG4/Balancer females were crossed to double balanced Df/Balancer, UAS-human cDNA males at multiple temperatures (18°C, 22°C, 25°C, 29°C) to determine rescue of lethality to adult stage. A minimum of two independent crosses were conducted at each temperature. For the five genes on the X-chromosome of the fly, we attempted rescue by crossing female TG4/FM7 flies to UAS-cDNA/(SM6a) males to generate hemizygous TG4 males that expresses human cDNA (TG4/Y; UAS-cDNA/+) to test their viability. Statistical analysis was performed using Two-way ANOVA followed by Sidak’s multiple comparison test across temperature and genotype.

Lifespan assays

Lifespan analysis was performed as previously described (Chung et al., 2020). Briefly, newly eclosed flies were separated by genotype and sex and incubated at 25°C. Flies were transferred into a fresh vial every two days and survival was determined once a day. 11–49 flies were tested per group. Statistical analysis was performed using Log-rank (Mantel-Cox) test.

Behavioral assays

Of 61 TG4 mutants that were viable when in trans with a corresponding deficiency, 18 lines exhibited lethality in homozygous states, indicating the presence of a second site lethal mutation. Out of 43 TG4 mutants that were homozygous viable, we prioritized to study 21 TG4 mutants based on reagent availability. Courtship assay was performed as previously described (Guo et al., 2019). Due to the reported effects of the y¹ w* mutations on behavior (Krstic et al., 2013; Massey et al., 2019), we replaced the X-chromosome containg y¹ w* with the y⁺ w⁺ X-chromosome from a Canton-S strain. This Canton-S strain was provided by Dr. Hugo Bellen and has been in his stock collection since the 1980s. Please see Figure S3 for the crossing strategy. Collection of socially naive adults was performed by isolating pupae in 16 × 100 polystyrene vials containing approximately 1 mL of fly food. After eclosion, flies were anesthetized briefly with CO₂ to ensure they were healthy and lacking wing damage. Anesthetized flies were returned to their vials and allowed 24 hours to recover before testing. Courtship assays were performed in a 6 well acrylic plate with 40mm circular wells, with a depth of 3 mm and a slope of 11 degrees, as per the chamber design (Simon and Dickinson, 2010). One Canton-S virgin female (6–10 days post-eclosion), and one TG4 mutant male fly (3–5 days post-eclosion) with or without UAS-human cDNAs were simultaneously introduced into the chamber via aspiration. Recordings were taken using a Basler 1920UM, 1.9MP, 165FPS, USB3 Monochromatic camera using the BASLER Pylon module, with an adjusted capturer rate of 33 fps (frames per second). Conversion of captured images into a movie file was performed via a custom MatLab script, and tracking of flies in the movie was performed using the Caltech Flytracker (Eyjolfsdottir et al., 2014). Machine learning assessment of courtship was performed using JAABA (Kabra et al., 2013) using classifiers that scored at 95% or higher accuracy during ground-truthing trials. At least 10 animals were tested per genotype. Analysis of data was performed using Excel (Microsoft) and Prism (GraphPad). A ROUT (Q=1%) test was performed in Prism to identify outliers. Determination of significance in behavior tests was performed using the Kruskal-Wallis one-way analysis of variance and the Dunn’s multiple comparison test. P-values of 0.05 or less were considered significant.

Overexpression assays to assess lethality and morphological phenotypes

To detect any differences in the phenotypes induced by overexpression of reference and variant human cDNA in order to assess variant function, we crossed UAS-human cDNAs with reference or variant alleles to ubiquitous (tub-GAL4) (Guelman et al., 2006), wing (nub-GAL4) (Calleja et al., 2000) or eye (GMR-GAL4) (Mehta et al., 2005) specific drivers. In the ubiquitous expression screen, 3–4 virgin females of tub-GAL4/TM3 Sb flies were crossed to 24 males of the UAS-cDNA reference and variant at 25°C. After 3–4 days, the parents were transferred into new vials, and the new vial was placed at 29°C while the old vial was kept at 25°C, allowing us to test two temperatures simultaneously. The parents were discarded after 3–5 days. Flies were collected after most of the pupae eclosed. The total number of flies were counted and scored with the genotype of interest (i.e. tub-GAL4>UAS-cDNA) as well as all other genotypes, (i.e. genotypes with balancers). A minimum of 10 flies were scored per experiment, though for the majority of crosses 50–100 flies were scored in this primary analysis. Viability was calculated by taking the % of observed/expected based on Mendelian ratio, and any UAS-cDNA with survival less than 70% was recorded as having scorable phenotype (lethal or semi-lethal). All of lines showing a phenotype at 29°C also showed phenotypes at 25°C, so subsequent experiments were performed at 25°C. To validate our hits, we performed the same viability assay, except each UAS-cDNA was tested at least three times to statistically validate that there is a difference between reference and variant. In addition, two independent UAS-cDNA transgenic lines established from the same construct were tested for each reference and variant. A variant was considered to have functional consequence (true hit) if both transgenic lines showed the same phenotype. In the cases where the difference is rather minor (e.g. <20% difference between survival), this was considered within the variation of the experiment paradigm, and the variant phenotype was documented. Functional study using wing or eye drivers were performed using similar strategies, but morphological phenotypes were scored instead of lethality.

Imaging of adult fly morphology

Drosophila eyes, wings and nota (dorsal thorax) were imaged after flies were frozen at −20°C for at least 24 hours. Wings for some flies were dissected in 70% EtOH and mounted onto slides for imaging. Images were obtained with the Leica MZ16 stereomicroscope equipped with Optronics MicroFire Camera and Image Pro Plus 7.0 software to extend the depth-of-field for Z-stack images.

Expression analysis of TG4 lines in larval and adult brains

All TG4 lines are crossed with UAS-nls.mCherry (3^rd chromosome) or UAS-nls.GFP (3^rd chromosome) at room temperature. Note that the nls.GFP that is being used here is prone to leak outside the nucleus while nls.mCherry is retained in the nucleus. The brains of mCherry/GFP positive third instar larvae and 3–5 days old adult flies were dissected in 1X phosphate-buffered saline (PBS). Adult brains were fixed immediately in 4% paraformaldehyde (PFA) and incubated at 4°C overnight (o/n) on a shaker. Next day these brains were post-fixed with 4% PFA with 2% Triton-X in PBS (PBST), kept in a vacuum container for an hour to get rid of the air from the tracheal tissue also make the tissue more permissive. Fixative was replaced every 10 minutes during this post-fixation step. Larval brains were fixed for 50 minutes on a rotator at room temperature. After thorough washing with PBS with 0.2% Triton (PBTX) both adult and larval brains were incubated with primary antibodies overnight (o/n) at 4°C on a shaker. The sample were extensively washed with 0.2% PBTX before secondary antibodies were applied at room temperature for 2 hours. Samples were thoroughly washed with PBST and mounted on a glass slide using Vectasheild (Vector Labs, H-1000–10). Primary antibodies used: Mouse anti-repo (DSHB: 8D12) 1:50, Rat anti-elav (DSHB: 7E8A10) 1:100, Goat anti-GFP (abcam: ab6662) 1:500. Secondary antibodies used: Anti-mouse-647 (Jackson ImmunoResearch: 715-605-151) 1:250, Anti-rat-Cy3 (Jackson ImmunoResearch: 712-165-153) 1:500. The samples were scanned using a laser confocal microscope (Zeiss LSM 880) with a 20X objective, and images were processed using ZEN (Zeiss) and Imaris (Oxford Instruments) software. Co-localization between mCherry and Elav or mCherry and Repo was performed with default thresholds in Imaris.

GeneOntology (GO) analysis

GO analysis was performed based on the PANTHER (Protein Analysis Through Evolutionary Relationships) system (http://www.pantherdb.org; date last accessed October 31, 2020 (Ashburner et al., 2000). Statistical analysis was performed by using the default PANTHER Overrepresentation Test (Released 20200728), Annotation Version and Release Date: GO Ontology database https://doi.org/10.1016/10.5281/zenodo.4033054 Released 2020-09-10 which used the Fisher’s Exact test with a false discovery rate p < 0.05.

Exome sequencing and identification of GLRA2 variants

Subjects 1, 2, 4, 5 and 8 had clinical exome sequencing at GeneDx (Gaithersburg, MD, United States), at the Praxis für Humangenetik Tubingen (Tubingen, Germany), at Baylor Genetics (Houston, TX, United States), at Pitié-Salpêtrière Hospital Genetics lab (Paris, France) and at Integragen (Evry, France), respectively. Subject 3 WES was performed at the Meyer Children’s Hospital, University of Florence, in the context of the DESIRE program and as previously described (Vetro et al., 2020). Briefly, the SureSelectXT Clinical Research Exome kit (Agilent Technologies, Santa Clara, CA) was used for library preparation and target enrichment, and paired-end sequencing was performed using Illumina sequencer (NextSeq550, Illumina, San Diego, CA, USA) to obtain an average coverage of above 80x, with 97.6% of target bases covered at least 10x. Reads were aligned to the GRCh37/hg19 human genome reference assembly by the BWA software package, and the GATK suite was used for base quality score recalibration, realignment of insertion/deletions (InDels), and variant calling (DePristo et al., 2011; McKenna et al., 2010). Variant annotation and filtering pipeline included available software (VarSeq, Golden Helix, Inc v1.4.6), focusing on non-synonymous/splice site variants with minor allele frequency (MAF) lower than 0.01 in the GnomAD database (Karczewski et al., 2020) (http://gnomad.broadinstitute.org/), an internal healthy control database and pre-computed genomic variants score from dbNSFP (Liu et al., 2011). Subject 6 WES was performed at the Institute of Human Genetics, Technical University of Munich, Germany as described previously (Brunet et al., 2021). DNA was extracted from blood samples by the Gentra Puregene Blood Kit (Qiagen, Hilden, Germany). WES was performed using the Sure Select Human All Exon 60 Mb V6 Kit (Agilent, Santa Clara, CA, USA) for exomic enrichent according to the manufacturer’s protocol. Sequencing was performed on a NovaSeq 6000 sequencer (Illumina, San Diego, CA, USA) to an average coverage of >94X. Reads were aligned to the UCSC human reference assembly (hg19) with the BWA algorithm v.0.5.9. For detection of single-nucleotide variants (SNVs) as well as small insertions and deletions SAMtools v.0.1.19 was applied. Copy number variations (CNVs) were called with the software ExomeDepth. In-house custom Perl scripts were used for variant annotation. Variant analysis was performed using I) a recessive filter for homozygous or compound-heterozygous variants with a minor allele frequency (MAF) of <1% in our in-house database of 22,000 exome datasets, II) a filter for X-chromosomal variants with a MAF<0,1%, III) a filter for de novo variants (MAF <0.01%) and IV) a phenotype-based filter using search terms of characteristic phenotypic traits and a MAF <0,1%. CNVs were assessed using a MAF <0,01%. Current criteria for variant classification according to the American College of Human Genetics (ACMG) were used for variant interpretation (Richards et al., 2015). Subject 7 had exome sequencing at Lyon Universiy Hospital (Lyon, France). The SeqCap EZ Medexome kit (Roche, Pleasanton, CA, USA) was used for library preparation and target enrichment before paired-end sequencing using an Illumina instrument (NextSeq500, Illulina, San Diego, CA, USA). A mean depth of coverage of 133x was obtained with 99.0% of target bases covered at least 10x. Reads were aligned to the GRCh37/hg19 human genome reference assembly by the BWA software package, and the GATK suite was used for base quality score recalibration, realignment of insertion/deletions (InDels), and variant calling (DePristo et al., 2011; McKenna et al., 2010). Variant annotation was performed with SnpEFF and filtering pipeline focused on non-synonymous/splice site variants with minor allele frequency (MAF) lower than 0.01 in the GnomAD database (Karczewski et al., 2020) (http://gnomad.broadinstitute.org/). Subject 9 WES was performed at the Erasmus MC as previously described (Perenthaler et al., 2020). In brief, exome-coding DNA was captured with the Agilent SureSelect Clinical Research Exome (CRE) kit (v2). Sequencing was performed on an Illumina HiSeq 4000 platform with 150-bp paired-end reads. Reads were aligned to hg19 using BWA (BWA-MEM v0.7.13) and variants were called using the GATK Haplotype Caller (McKenna et al., 2010) v3.7 (https://www.broadinstitute.org/gatk/). Detected variants were annotated, filtered and prioritized using the Bench lab NGS v5.0.2 platform (Agilent technologies). Subject 10 WES and data processing were performed by the Genomics Platorm at the Broad Institute of MIT and Harvard with an Illumina Nextera or Twist exome capture (~38 Mb target), and sequenced (150 bp paired reads) to cover >80% of targets at 20x and a mean target coverage of >100x. WES data was processed through a pipeline based on Picard and mapping done using the BWA aligner to the human genome build 38. Variants were called using Genome Analysis Toolkit (GATK) HaplotypeCaller package version 3.5 (McKenna et al., 2010) (https://www.broadinstitute.org/gatk/). WES for subjects 11 and 12 was performed in collaboration with the Autism Sequencing Consortium (ASC) at the Broad Institute on Illumina HiSeq sequencers using the Illumina Nextera exome capture kit. Exome sequencing data was processed through a pipeline based on Picard and mapping done using the BWA aligner to the human genome build 37 (hg19). Variants were called using Genome Analysis Toolkit (GATK) HaplotypeCaller package version 3.4 (McKenna et al., 2010) (https://www.broadinstitute.org/gatk/). Variant call accuracy was estimated using the GATK Variant Quality Score Recalibration (VQSR) approach. High-quality variants with an effect on the coding sequence or affecting splice site regions were filtered against public databases (dbSNP150 and gnomAD V.2.0) to retain (i) private and clinically associated variants; and (ii) annotated variants with an unknown frequency or having minor allele frequency <0.1%, and occurring with a frequency <2% in an in-house database including frequency data from > 1,500 population-matched WES. The functional impact of variants was analyzed by CADD V.1.3, Mendelian Clinically Applicable Pathogenicity V.1.0 (Jagadeesh et al., 2016; Kircher et al., 2014), and using InterVar V.0.1.6 to obtain clinical interpretation according to American College of Medical Genetics and Genomics/Association for Molecular Pathology 2015 guidelines (Li and Wang, 2017).

Trio exome sequencing of subject 13 and his parents was performed at the Institute of Human Genetics, University of Leipzig Medical Center, in the context of the research project “Genetics of rare disorders”. Testing in a research setting was approved by the ethics committee of the University of Leipzig (224/16-ek and 402/16-ek). Library preparation was done using the Nextera DNA Flex Pre-Enrichment LibraryPrep with Illumina Nextera DNA UD Indexes by Illumina. Target enrichment was achieved by using the Human Core Exome hybridization probes from Twist Bioscience. Paired-end Next-Generation-Sequencing was then performed on a NovaSeq 6000 Instrument using an S1 Reagent Kit (300 cycles) by Illumina. Analysis of the raw data was performed using the software Varfeed (Limbus, Rostock). Variants were annotated and prioritized using the software Varvis (Limbus, Rostock). Rare variants (minor allele frequency below 1 % in the general population) were prioritized based on inheritance mode, impact on protein, clinical relevance in variant databases and in silico prediction. GeneMatcher (Sobreira et al., 2015) (https://genematcher.org/) assisted in the recruitment of Subjects 2, 3 and 5–13.

SDS-PAGE/Western blot

Five heads of nSyb-GAL4 UAS-GLRA2 reference and variant flies aged for 5 days post eclosion were lysed in 30μL NETN buffer (50mM Tris pH 7.5, 150mM NaCl, 0.5% NP-40, 1 mM EDTA) with an electric douncer for 10 seconds for three times on ice. 30μL of 2x Laemmli Sample Buffer (Bio-Rad) with 10% 2-mercaptoethanol was added to the lysis and incubated on ice for 10 min. Samples were boiled at 95°C and spun at 14,000 RPM for 5 minutes at 4 °C. The soluble fraction was loaded onto a standard SDS-PAGE gel. PVDF (polyvinylidene difluoride) membrane activated for 1 minute with 100% methanol. After running and wet transfer, the membrane was blocked in 5% skim milk for 1 hour. The membrane was incubated (overnight, shaking, at 4°C) with mouse anti-HA (HA.11, 1:1,000, 901501, BioLegend) and mouse anti-Actin (C4) (1:50,000, MAB1501, EMD Millipore) primary antibodies in 3% BSA (bovine serum albumin), followed by 10 minute washes (3 times) with 1% Triton-X in Tris-buffered saline (TBST). We incubated this with goat anti-mouse HRP-conjugated (1:15000, 115-035-146, Jackson ImmunoResearch) secondary antibody in skim milk. The membrane was washed three times with 1% TBST and detected with Western Lightning^™ Chemiluminescence Reagent Plus (perkinelmerNEL104001EA) ECL solution using the Bio-Rad ChemiDoc MP imaging system.

Structural biological analysis of GLRA2 patient variants

Protein residues that corresponds to GLRA2 patient variants were mapped onto the crystal protein structure of GLRA1 protein in Protein Data Bank (PBD, ID: 4X5T) (Burley et al., 2019) using the PyMOL (https://pymol.org/) (Yuan et al., 2016) because GLRA1 and GLRA2 are highly homologous proteins (85% similarity, 78% identity and 3% gaps) based on DIOPT (Hu et al., 2011).

Image generation

Cartoon images in Figure 1H were generated with BioRender.com.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analysis was performed with GraphPad Prism version 8. Significance was defined as *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Gene and variant level statistics

Gene level constraints for Figures 1B–1D (pLI, LOEUF, missense O/E) for all genes (Global, n=19704), the SSC subset (n=1493), and the study subset (n=143) were based on metric availability from gnomAD and analyzed by ANOVA followed by Dunn’s multiple comparison test. The same test was used in Figure S9B–S9Y for gene and variant analysis. The contingency graph for variant consequences (Figure S9A) grouped by variants corresponding to lethal (essential genes) or viable (non-essential genes) TG4 mutants was analyzed by Chi square, df (20, 3, p<0.0002).

Rescue-based and overexpression-based survival

For humanized rescue of TG4 lethality (Figures 2B–2D), 3 independent crosses were set per genotype and a minimum of n>50 flies were quantified for each cross. Statistical analyses were performed by ANOVA followed by Sidak’s multiple comparisons test. For lethality caused by overexpression with Tub-GAL4 (Figures 4B, 5B, and S12B), a minimum of 3 independent crosses were set with two independent UAS-transgenic lines. 50–100 flies (a minimum of 10 if overexpression caused survival defects) were scored. Statistical analyses were performed by unpaired t test.

Lifespan analysis

Lifespan analysis of humanized TG4 lines at 25°C (Figures 2E and 2F) were analyzed by Log rank (Mantel-Cox) test with a minimum of 11–49 flies for each genotype from three independent crosses.

Behavior

Fore courtship analysis (Figures 3C–3G and S4A–S4T), at least 10 animals were tested per genotype. Analysis of data was performed using Excel (Microsoft) and Prism (GraphPad). A ROUT (Q=1%) test was performed in Prism to identify outliers. Data was analyzed with Kruskal-Wallis ANOVA followed by Dunn’s multiple comparison test.

Assessment of melanized nodules

Melanized nodules formed on the notum of flies expressing GLRA2^T296M driven by pnr-GAL4 (Figure 5D) were analyzed by ANOVA followed by Tukey’s multiple comparison test. 3 independent crosses were set and minimum of 50 flies examined for presence of melanized nodules per genotype.

Electroretinograms

Electrophysiological field potential recordings (ERG) of nSyb-GAL4 or Rh1-GAL4 flies expressing UAS-GLRA2 constructs (Figures 5E–5H and S12C–S12F) were analyzed by ANOVA followed by Bonferroni’s multiple comparison test. 4–10 flies were examined for each genotype. Recording was repeated at least 3 times per fly.

Highlights.

• We generate and characterize >300 (TG4 and cDNA) Drosophila mutants and transgenics

• Humanization and overexpression strategies to functionally assess ASD variants in vivo

• ASD variant data in flies help identify GLRA2-related neurodevelopmental disorders

• Basic and clinical collaboration facilitates variant testing and disease gene discovery