Spatiotemporal 7q11.23 protein network implicates the GTF2I-PRKDC-DDR 3 pathway during early-fetal brain development in psychiatric diseases

19 Background: Recurrent deletions and duplications of chromosome 7q11.23 copy 20 number variants (CNVs) are associated with several psychiatric disorders. Previous 21 works showed GTF2I associated with Williams-Beuren syndrome, but pathways 22 affected by GTF2I are poorly defined. Although phenotypic abnormalities have been 23 observed in patients and animal models, the targeted human brain regions, 24 developmental stages, protein networks, and signaling pathways, influenced by this 25 CNV remain unclear. 26 Results: Topological changes were observed in protein-protein interaction (PPI) 27 networks throughout different stages of brain development. Early and late fetal 28 periods of development in the cortex, striatum, hippocampus, and amygdale were 29 observed as the vital periods and regions for 7q11.23 CNV proteins. As a driver gene, 30 GTF2I interacted with PRKDC and BRCA1 to involve in DNA Damaging Response 31 ( DDR) pathway. The physical interaction between GTF2I with PRKDC was confirmed 32 experimentally by the liquid chromatography-tandem mass spectrometry 33 (LC-MS/MS). 34 Conclusion: We identified that striatum, hippocampus, and amygdala are crucial regions for 35 establishing connectivity between 7q11.23 proteins and their partners in early and late fetal 36 periods. Our results suggested that GTF2I-PRKDC-DDR and GTF2I-BRCA1-DDR pathway is 37 crucial for the 7q11.23 CNV genes to contribute to the pathogenesis of psychiatric diseases. Our study demonstrated that 7q11.23 proteins interact with their partners mainly in three spatiotemporal intervals, and that the interaction patterns change across these intervals. We identified that striatum, hippocampus, and amygdala are crucial regions for the interactions between 7q11.23 proteins and their partners in early and late fetal periods. In addition, our results indicated that GTF2I plays a key role in a dynamic network by interacting with DNA-dependent protein kinase catalytic subunit (PRKDC) Breast type susceptibility experiments spectrometry to identify the interactions between GTF2I


169
Next, we compared connectivity of co-expressed interacting protein pairs either 170 within the same developmental period (early fetal, P1) or within the same brain region 171 (R3) to identify both topological and functional differences between the 172 spatiotemporal 7q11.23 networks. As noted, we have identified three spatiotemporal 173 networks with significantly enriched co-expressed PPI pairs across different brain 174 regions (R1 and R3) within the same developmental period (early fetal, P1) and also 175 across different developmental periods (early fetal (P1) and late fetal (P4)) within the 176 same region (R3). Network changes were assessed by calculating the fractions of 177 co-expressed PPI pairs that were unique to one network against the fractions of 178 co-expressed PPI pairs that were shared among networks (Fig. 4). We noted a 179 significant difference between identical region within different developmental periods 180 (P1R3 and P4R3；ANOVA; p = 2.95×10 −16 ) (Supplementary Table S3). In contrast, 181 no significant difference was observed between same period within different regions 182 (P1R1 and P1R3, ANOVA; p = 0.349) (Supplementary able S4). terms for the biological process were "DNA repair", "regulation of cell cycle process", 192 and "double-strand break repair" (Fig. 5). The top-three terms for the biological 193 processes involving 7q11.23 proteins and partners from the P1R3 network were 194 "DNA-dependent DNA replication", "RNA splicing via transesterification reactions 195 with bulged adenosine as nucleophile", and "regulation of cell cycle process" (Fig. 5). 196 For 7q11.23 proteins and their partners from the P4R3 network, the top-three 197 significant terms were "regulation of DNA metabolic process", "RNA splicing", and 198 "regulation of cellular protein localization" (Fig. 5). Next, we observed that 71 199 co-expressed and interacting partners of CNV proteins were exclusively from the 200 P4R3 network, not from P1R1 and P1R3 networks, and associated with "signaling 201 pathways regulating pluripotency of stem cells (PSCs)", "EGFR tyrosine kinase 202 inhibitor resistance", and "cell cycle", and the "hippo signaling pathway". 203 (Supplementary Fig. S1). Of the three significant networks that we identified above, GTF2I possesses the 208 highest radiality value, thus indicating that GTF2I is a driver gene and adopts a 209 central position within these networks (Supplementary table S5). It has been 210 reported that GTF2I heterozygotes exhibit microcephaly and retarded growth [14].

211
Recent studies have also demonstrated that GTF2I is involved in neurodevelopment 212 [33,34]. Importantly, the phenotype observed in mice mirrors that observed in 213 humans [35]. As a driver gene within networks, GTF2I is a crucial contributor to 214 neuropsychiatric disorders [36][37][38]. Hence, we investigated the interaction pattern of 215 GTF2I across three significant spatiotemporal networks.

217
Seventy-three proteins interacted physically and were co-expressed with GTF2I 218 across three spatiotemporal intervals (Fig. 4). Several of these partners were hub 219 proteins which interacted with several CNV proteins. These hub proteins also 220 interacted physically and were co-expressed with each other, thereby forming one 221 functional module (Fig. 6A)   There was no significant difference between the entire 7q11.23 network and genes for 271 genes associated with neurodegenerative disease (supplementary

274
We constructed a dynamic network for the 7q11.23 CNV, which is a crucial risk factor 275 for psychiatric disorders. Importantly, the spatiotemporal network indicated that 276 7q11.23 CNV genes played a crucial role in three intervals: P1R1 (early fetal, frontal 277 lobe), P1R3 (early fetal, striatum, hippocampus and amygdale) and P4R3 (late fetal, 278 striatum, hippocampus and amygdale). The early-fetal and late-fetal periods were the 279 vital periods for 7q11.23 CNV proteins to affect human brain development. These 280 results are in agreement with studies showing that GTF2I haplo-insufficient mice 281 exhibit a small brain and neural defects during embryonic development [14,50].

282
Sanders SJ and colleagues showed that the 7q11.23 CNV is involved in the 283 pathogenesis of autism [28]. We observed that mutations in the proteins within the 284 spatiotemporal 7q11.23 network were significantly enriched in ASD genes and the 285 genes associated with developmental delay. Our results are in agreement with 286 previous studies that discovered the fetal period was crucial for the pathogenesis of 287 ASD [28,51]. Our study suggests that the hippocampus, amygdale, striatum, and 288 frontal lobe, are crucial regions affected by CNV genes. This result is in accordance 289 with a previous report that showed the amygdala, cortex, and hippocampus to be  Our data also suggest that GTF2I is a candidate driver gene within the significant 295 networks. Deurloo MHS and colleagues previously showed that GTF2I plays a 296 pathological role in WBS [34]. Microcephaly and retarded growth were observed in 297 mice that were heterozygous for GTF2I [14]. Importantly, one of the pathways our 298 study suggests as being most likely impacted by the 7q11.23 CNV is DNA repair 299 pathway. We observed that GTF2I and RFC2 interacted with PRKDC, a vital hub 300 partner which possesses the highest radiality value. Through proteomic analyses, we 301 identified that GTF2I interacted with PRKDC. As a significant hub partner, PRKDC