Drosophila models of pathogenic copy-number variant genes show global and non-neuronal defects during development

While rare pathogenic copy-number variants (CNVs) are associated with both neuronal and non-neuronal phenotypes, functional studies evaluating these regions have focused on the molecular basis of neuronal defects. We report a systematic functional analysis of non-neuronal defects for homologs of 59 genes within ten pathogenic CNVs and 20 neurodevelopmental genes in Drosophila melanogaster. Using wing-specific knockdown of 136 RNA interference lines, we identified qualitative and quantitative phenotypes in 72/79 homologs, including 21 lines with severe wing defects and six lines with lethality. In fact, we found that 10/31 homologs of CNV genes also showed complete or partial lethality at larval or pupal stages with ubiquitous knockdown. Comparisons between eye and wing-specific knockdown of 37/45 homologs showed both neuronal and non-neuronal defects, but with no correlation in the severity of defects. We further observed disruptions in cell proliferation and apoptosis in larval wing discs for 23/27 homologs, and altered Wnt, Hedgehog and Notch signaling for 9/14 homologs, including AATF/Aatf, PPP4C/Pp4-19C, and KIF11/Klp61F. These findings were further supported by tissue-specific differences in expression patterns of human CNV genes, as well as connectivity of CNV genes to signaling pathway genes in brain, heart and kidney-specific networks. Our findings suggest that multiple genes within each CNV differentially affect both global and tissue-specific developmental processes within conserved pathways, and that their roles are not restricted to neuronal functions.


INTRODUCTION
8 phenotypes (Fig. 3A, Supp. Data 3). For example, RNAi lines for both UQCRC2/UQCR-C2 and 151 POLR3E/Sin within 16p12.1 showed lethality. Within the 3q29 region, NCBP2/Cbp20 and 152 MFI2/Tsf2 showed severe phenotypes while DLG1/dlg1 showed lethality. In contrast, 12/20 153 known neurodevelopmental genes showed no observable wing phenotypes, suggesting that these 154 genes could be responsible for neuronal-specific phenotypes (Fig. 3B, Supp. Data 3). We note 155 that 18/79 fly homologs showed discordant phenotypes between two or more RNAi lines for the 156 same gene, which could be due to differences in expression of the RNAi construct among these 157 lines (Supp. Data 3). 158 Certain qualitative phenotypes exhibited higher frequency in males compared to females. 159 For example, discoloration (87 lines in males compared to 56 lines in females; p=1.315×10 -4 , 160 two-tailed Fisher's exact test) and missing vein phenotypes (92 lines in males compared to 29 161 lines in females; p=2.848×10 -16 , two-tailed Fisher's exact test) at any degree of severity were 162 more commonly observed in males than females (Supp. Data 2). In particular, 25/92 lines in 163 males (compared to 1/29 in females) showed a total loss of the anterior crossvein (ACV) (Supp. 164 Data 2). We further identified 17 RNAi lines that were lethal in males with wing-specific 165 knockdown of fly homologs. While higher frequencies of wing phenotypes in males could be 166 due to a sex-specific bias of developmental phenotypes, the increased severity we observed in 167 males is most likely due to a stronger RNAi knockdown caused by an X-linked dosage 168 compensation, as the bx MS1096 -GAL4 driver is inserted on the fly X chromosome 54,55 . 169 Next, we measured the total adult wing area and the lengths of six veins (longitudinal L2, 170 L3, L4, L5, ACV, and posterior crossvein or PCV) in the adult wing for each of the tested RNAi 171 lines that did not show lethality (or severe wrinkled phenotypes for vein length measurements) 172 (Fig. 4A). Overall, we identified significant wing measurement changes for 89 RNAi lines compared to controls, which included lines that did not have an observable qualitative wing 174 phenotype (Fig. 2D). A summary of L3 vein lengths is presented in Fig. 4B, and the 175 measurements for the remaining five veins are presented in Supp. Figure 1 and Supp. Data. 2. 176 We found that 33/61 of the homologs (54%) showed significant changes in L3 vein length, 177 including 20 homologs with longer vein lengths and 13 homologs with shorter vein lengths  (Fig. 4B-C). In addition, PAK2/Pak within 3q29, TBX1/org-1 within 22q11.2, 184 autism-associated CHD8/kis, and microcephaly-associated ASPM/asp also showed smaller wing 185 areas and vein lengths ( Fig. 4B-C). In contrast, TRPM1/Trpm within 15q13.3 and the cell 186 proliferation gene PTEN/Pten 57 both showed larger wing areas and vein lengths ( Fig. 4B-C). 187 Furthermore, we identified eight homologs that showed no qualitative wing phenotypes but had   197 We previously showed that many of the same fly homologs of CNV genes that showed wing 198 defects in the current study also contributed towards neuronal phenotypes in the fly eye 41,42 , 199 suggesting a role for these genes in global development. We therefore performed ubiquitous and including Ppp4C -/and Kif22 -/-, showed embryonic lethality 25,58, 59 . In our study, the DLG1/dlg1 209 line that showed lethality with wing-specific knockdown also exhibited larval lethality with 210 ubiquitous knockdown, indicating its role in global development (Fig. 5A). In addition, six 211 homologs that showed severe wing phenotypes also showed larval or pupal lethality with 212 ubiquitous knockdown, including ALDOA/Ald and PPP4C/Pp4-19C within 16p11.2 and 213 ATXN2L/Atx2 and TUFM/mEFTu1 within distal 16p11.2 (Fig. 5A). The remaining homologs 214 that showed lethality with ubiquitous knockdown showed at least a mild qualitative or 215 quantitative wing phenotype. 216 We next compared the phenotypes observed with wing-specific knockdown of fly 217 homologs to their corresponding eye-specific knockdowns to evaluate neuronal versus non-218 neuronal effects. To quantitatively assess the phenotypic severity of cellular defects with eye-219 specific knockdown of fly homologs, we developed a tool called Flynotyper 60 that determines the 220 degree of disorganization among the ommatidia in the adult eye. We analyzed phenotypic scores 221 obtained from Flynotyper for 66 RNAi lines of 45 fly homologs, including from previously-222 published datasets 41,42,60 . We found that 37/45 homologs (82.2%) exhibited both eye and wing-223 specific defects (Fig. 5B, Supp. Fig. 2 OSTalpha/CG6836 within 3q29, and UQCRC2/UQCR-C2 (Fig. 5B, Supp. Fig. 2). In particular, 228 UQCRC2/UQCR-C2 showed lethality with wing-specific knockdown, suggesting potential 229 tissue-specific effects of this gene in non-neuronal cells (Fig. 5B). While most homologs 230 contributed towards both eye and wing-specific phenotypes, we observed a wide range of 231 severity in eye phenotypes that did not correlate with the severity of quantitative or qualitative 232 wing phenotypes (Fig. 5C). For example, TUFM/mEFTu1 showed a severe wing phenotype but 233 only a mild increase in eye phenotypic score, while SH2B1/Lnk, also within the distal 16p11.2 234 region, showed severe rough eye phenotypes but only a mild increase in wing size (Fig. 5D). 235 Similarly, BCL9/lgs also showed opposing tissue-specific effects with mild qualitative wing 236 phenotype and severe eye phenotype, suggesting that the role of these homologs towards 237 development differs across tissue types. To assess how expression levels of CNV genes vary across different tissues, we first examined 241 the expression patterns of fly homologs in larval and adult tissues using the FlyAtlas Anatomical 242 Microarray dataset 62 . We found that 76/77 homologs with available data were expressed in at 243 least one larval and adult tissue (Supp. Fig. 3, Supp. Data 6). In general, we did not observe a 244 correlation between wing phenotype severity and expression patterns of homologs in larval or 245 adult tissues (Fig. 6A) such as PPP4C/Pp4-19C and NCBP2/Cbp20 (Fig. 6A, Supp. Fig. 3). Furthermore, 30/39 249 homologs (76.9%) that showed eye phenotypes also had ubiquitous larval expression, providing 250 further support to the observation that genes causing neuronal phenotypes may also contribute to 251 developmental phenotypes in other tissues (Supp. Data 5). Of note, 9/77 homologs (11.7%) did 252 not have any expression in the larval central nervous system, including FMO5/Fmo-2, 253 BDH1/CG8888 within 3q29, and TBX6/Doc2 within 16p11.2 (Fig. 6A, Supp. Fig. 3). However, 254 we observed wing phenotypes for 8/9 of these homologs, suggesting that they may contribute to 255 tissue-specific phenotypes outside of the nervous system. Except for the epilepsy-associated 256 SCN1A/para 63 , which was exclusively expressed in both the larval central nervous system (CNS) 257 and adult brain tissues, other tested neurodevelopmental genes were also expressed in non-258 neuronal tissues (Fig. 6A). 259 We further used the GTEx Consortium dataset 64 to examine tissue-specific expression of 260 150 human CNV and known neurodevelopmental genes across six tissues including brain, heart, 261 kidney, lung, liver, and muscle. We found 121 genes that were expressed in at least one adult 262 tissue, including 49 genes (32.7%) that showed ubiquitous expression across all six tissues 263 (Supp. Data 6). Of the 112 genes expressed in non-neuronal tissues, 34 did not have any 264 neuronal expression, including TBX1, FMO5 and GJA5 within 1q21.1, and ATP2A1 within distal 265 16p11.2 (Fig. 6B, Supp. Data 6). FMO5 and TBX1 also showed non-neuronal expression in 266 Drosophila tissues, suggesting that their tissue-specific expression is highly conserved (Fig. 6A). 267 Other genes showing ubiquitous expression also had preferentially high expression for specific 268 non-neuronal tissues, including ALDOA and UQCRC2 for muscle and heart for (Fig. 6B). In 269 contrast, we found nine genes that were expressed only in the adult brain, including FAM57B 270 and DOC2A within 16p11.2, as well as SCN1A, which showed similar CNS-only expression in 271 Drosophila tissues (Fig. 6B, Supp. Data 6). Histone H3 Ser10 (pH3) and anti-Drosophila caspase-1 (dcp1), respectively, in the third instar 282 larval wing discs. We identified 23/27 homologs that showed significant increases in apoptotic 283 cells compared to controls, including seven homologs, such as PPP4C/Pp4-19C, ATXN2L/Atx2, 284 and AATF/Aatf , which showed dcp1 staining across the entire larval wing pouch ( Fig. 7A- 290 Similarly, 3/4 homologs of genes showing lethality with wing-specific knockdown also showed 291 defects in apoptosis or proliferation, with the exception of ACACA/ACC (Supp. Figure 4, Supp. 292 Data 7). As bx MS1096-GAL4 is located on the X-chromosome, we expected to see more severe 293 defects in males compared with females with knockdown of homologs due to the X-linked 294 dosage compensation 54,55 . However, knockdown of 3/11 tested homologs with sex-specific 295 differences in adult wing phenotypes, including BCL9/lgs, CYFIP1/Sra-1, and 296 DNAJC30/CG11035 within 7q11.23, showed significantly decreased levels of cell proliferation 297 in females but no change for males compared to their respective controls, suggesting a sex-298 specific effect of these genes for cell proliferation (Supp. Fig. 5, Supp. Data 7). Overall, our 299 results suggest that cell proliferation and apoptosis play an important role towards development 300 in both neuronal and non-neuronal tissues.

302
Homologs of candidate CNV genes disrupt conserved signaling pathways 303 Several conserved signaling pathways that are active in a spatial and temporal manner in the 304 larval wing disc, such as Wnt, Hedgehog, BMP, and Notch signaling, regulate the anterior-   Fig. 6). We conclude that a  Connectivity patterns of candidate genes vary across human tissue-specific networks 335 We examined patterns of connectivity for the nine candidate genes which showed disruptions of 336 signaling pathways within the context of human brain, heart, and kidney-specific gene 337 interaction networks 80 . These tissue-specific networks were constructed using Bayesian 338 classifier-generated probabilities for pairwise genetic interactions based on co-expression data 80 . 339 We calculated the lengths of the shortest paths between each candidate gene and 267 Wnt, 340 Notch, and Hedgehog pathway genes in each network as a proxy for connectivity (Supp. Data 341 8). In all three networks, each of the candidate genes were connected to a majority of the tested 342 signaling pathway genes, suggesting that our results have translational relevance towards human 343 developmental pathways (Fig. 9A, Supp. Fig. 7). Interestingly, we observed a higher 344 connectivity (i.e. shorter path distances) between candidate genes and Wnt and Hedgehog 345 pathway genes in the brain-specific network compared to the heart and kidney-specific networks 346 (Fig. 9B). We further identified enrichments for genes involved in specific biological processes 347 among the connector genes that were located in the shortest paths within neuronal and non-348 neuronal tissue-specific networks (Fig. 9C, Supp. Data 8). For example, axon-dendrite 349 transport, dopaminergic signaling, and signal transduction functions were enriched among 350 connector genes only for the brain-specific network, while organelle organization and protein 351 ubiquitination were enriched among connector genes only for kidney and heart networks ( Fig.   352 9C). However, several core biological processes, such as cell cycle, protein metabolism, 353 transcriptional regulation, and RNA processing/splicing, were enriched among connector genes 354 within all three tissue-specific networks (Fig. 9C). Our analysis highlights that human CNV 355 genes potentially interact with developmental signaling pathways in an ubiquitous manner, but 356 may affect different biological processes in neuronal and non-neuronal tissues. 358 We used the Drosophila wing as a model to identify key CNV genes involved in non-neuronal 359 phenotypes associated with CNV disorders. We tested fly homologs of 79 genes and identified  First, we found that homologs of CNV genes contribute towards developmental 365 phenotypes through ubiquitous roles in neuronal and non-neuronal tissues. Although we did not 366 study models for the entire CNV, nearly all individual fly homologs of CNV genes contribute to 367 wing-specific developmental phenotypes. It is likely that these genes may also contribute to 368 additional phenotypes in other tissues that we did not assess. In fact, a subset of these genes also 369 showed early lethality with ubiquitous knockdown in addition to severe or lethal wing-specific 370 phenotypes. However, we found no correlation between the severity of the eye and wing 371 phenotypes, suggesting tissue-specific effects of these homologs towards developmental 372 phenotypes. In contrast, fly homologs of known neurodevelopmental genes generally showed 373 milder wing phenotypes compared with eye phenotypes, indicating a more neuronal role for 374 these genes. While our study only examined a subset of CNV genes with Drosophila homologs, 375 phenotypic data from knockout mouse models also support a global developmental role for  suggesting an underlying common conserved cellular mechanism for the distal 16p11.2 deletion.

399
Furthermore, a subset of these genes also disrupted multiple signaling pathways, indicating a 400 potential role for these genes as key regulators of developmental processes. We specifically 401 identified five genes whose knockdown caused disruptions of Wnt, Notch, and hedgehog we also observed certain biological processes enriched among connector genes that were specific 414 to neuronal or non-neuronal tissues, indicating that haploinsufficiency of genes within CNV 415 regions may disrupt different biological processes in a tissue-specific manner.

416
Overall, we show that fly homologs of most CNV genes contribute towards global 417 developmental phenotypes, although exactly how they contribute toward such phenotypes varies 418 between neuronal and non-neuronal tissues. Previous functional studies for CNV disorders have 419 focused primarily on identifying candidate genes for the observed neuronal phenotypes. In this 420 study, we identified several homologs of CNV genes that are responsible for non-neuronal 421 phenotypes, as well as novel associations between these genes and conserved biological 422 processes and pathways. We therefore propose that multiple genes within each CNV disrupt 423 global and tissue-specific processes during development and contribute to the wide range of non-424 neuronal phenotypes associated with CNV disorders (Fig. 10). This multigenic model for non-

435
Fly stocks and genetics 436 We tested 59 Drosophila homologs for 130 human genes that span across 10 pathogenic CNV

444
To knockdown individual genes in specific tissues, we used RNA interference (RNAi) 445 and the UAS-GAL4 system ( Fig. 2A), a well-established tool that allows for tissue-specific Adult progeny were isolated from crosses between RNAi lines and bx MS1096 -GAL4 driver shortly 462 after eclosion, and kept at 25°C until day 2-5 ( Fig. 2A). At that point, the progeny were frozen at  For each non-lethal RNAi line, we scored the adult wing images for five qualitative 470 phenotypes, including wrinkled wing, discoloration, missing veins, ectopic veins, and bristle 471 planar polarity defects, on a scale of 1 (no phenotype) to 5 (lethal) (Fig. 3C). Lines showing 472 severely wrinkled wings or lethality were scored as 4 (severe) or 5 (lethal) for all five 473 phenotypes. We calculated the frequency of each phenotypic score (i.e. mild bristle polarity, 474 moderate discoloration) across all of the wing images for each line (Fig. 3A-B), and then 475 performed k-means clustering of these values to generate five clusters for overall wing 476 phenotypes (Fig. 2C). For quantitative analysis of wing phenotypes, we used the Fiji ImageJ  Wing imaginal discs from third instar larvae were dissected in 1X PBS. The tissues were fixed 505 using 4% paraformaldehyde and blocked using 1% bovine serum albumin (BSA). The wing discs 506 were incubated with primary antibodies using appropriate dilutions overnight at 4°C. We used 507 the following primary antibodies: mouse monoclonal anti-pHistone3 (S10) (1:100 dilutions, Cell median expression among all brain and heart sub-tissues was used to represent brain and heart 549 expression, while the median expression among all colon, esophagus, small intestine, and 550 stomach sub-tissues was used to represent digestive tract expression. Preferential gene 551 expression for a particular tissue within the GTEx dataset was determined if the expression 552 values for that tissue were greater than the third quartile of all tissue expression values for that 553 gene, plus 1.5 times the interquartile range. Venn diagrams were generated using the Venny 554 webtool (http://bioinfogp.cnb.csic.es/tools/venny) (Supp. Fig. 3).

556
Network analysis 557 We obtained human tissue-specific gene interaction networks for brain, heart, and kidney tissues 558 from the GIANT network database 80 within HumanBase (https://hb.flatironinstitute.org). These

Mild
No pheno.