scRNA-Sequencing uncovers a TCF-4-dependent transcription factor network regulating commissure development

Intercortical connectivity is important for higher cognitive brain functions by providing the basis for integrating information from both hemispheres. We show that ablation of the neurodevelopmental disorder associated bHLH factor Tcf4 results in complete loss of forebrain commissural systems in mice. Applying a new bioinformatic strategy integrating transcription factor expression levels and regulon activities from single cell RNA-sequencing data predicted a TCF-4 interacting transcription factor network in intercortical projection neurons regulating commissure formation. This network comprises a number of regulators previously linked to the pathogenesis of intellectual disability, autism-spectrum disorders and schizophrenia, e.g. Foxg1, Sox11 and Brg1. Furthermore, we demonstrate that TCF-4 and SOX11 biochemically interact and cooperatively control commissure formation in vivo, and regulate the transcription of genes implied in this process. Our study provides a regulatory transcriptional network for the development of interhemispheric connectivity with potential pathophysiological relevance in neurodevelopmental disorders.


Introduction 20
Cognitive abilities are highly dependent on the establishment of proper neuronal 21 connectivity between different brain regions and its cellular components 22 (Constantinidis and Klingberg 2016; Hedden and Gabrieli 2004). The corpus 23 callosum, the anterior commissure, and the hippocampal commissure, carry axons 24 across the midline and ensure information flow and coordination between the 25 cerebral hemispheres. Of these, the corpus callosum (CC) is the largest commissural 26 tract of the human brain (Tomasch 1954). Callosal connections serve to integrate and 27 coordinate of sensory-motor functions from the right and left side of the body and are 28 integral to high-level cognitive functions including language, abstract reasoning and 29 high-level associative function (Paul et al. 2007). Orchestration of the precise temporo-spatial execution of developmental programs is We next asked which genes may be common target genes of both TFs and thus may 207 be involved in commissure development. Hence, we compared the predicted regulon 208 targets of SOX11 from the GRN analysis and the list of DEGs in the Tcf4KO. The 209 intersection of the datasets yielded a list of 73 genes ( Figure 4B and Furthermore, GO term analysis revealed an enrichment for genes involved in 215 axonogenesis ( Figure 4D and Table S3). From these associated genes Plxna2, a 216 gene involved in semaphorin plexin signalling for axon guidance (Mah et al. 2006;217 Mitsogiannis et al. 2017;Rohm et al. 2000) and Dcx, a gene essential for proper 218 neuronal morphology, migration and axon guidance were selected for further 219 investigation (Deuel et al. 2006;Fu et al. 2013;Karl et al. 2005;Koizumi et al. 2006). 220 Evolutionary conserved regions upstream of or at the promotor, which contained 221 conserved binding sites for TCF-4 and SOX11, were cloned into luciferase reporter 222 plasmids and then transfected into HEK293T cells together with expression plasmids 223 for Tcf4 and Sox11. SOX11 alone induced robust Dcx and Plxna2 reporter activity.   supplemented with 10% of fetal bovine serum and 5 ml penicillin/Streptomycin at 373 37°C and 5% CO 2 . 374

Experimental Design 375
For the single-cell RNA-Sequencing (5 WT and 4 KO samples) only samples with 376 more than 500 cells after filtering were used to ensure a complete reproduction of cell 377 diversity in the neocortex. Therefore, 2 samples for the WT and 2 samples for the KO 378 were removed. We had to exclude one WT animal that displayed lower Tcf4 379 expression than the KO and also excluded one cluster that displayed a high 380 background transcript expression of blood related genes such as Hbb-a1, Hbb-a2. After additional washing with 1x PBS for 5 min, slides were mounted with 60 μl 414 Mowiol (Sigma Aldrich Chemie GmbH Munich, Germany) and stored at 4 C. 415

Cell counting 416
Cell counting was done blind to avoid bias. Numbers were randomly assigned to 417 slides before imaging. Genotypes were only revealed for statistical analysis. All promoter. The hDCX-Promotor plasmid has been described before (Karl et al. 2005). Fast Flow (Millipore-Merck, Darmstadt) in TEN-Buffer (1:1) were added and the 502 samples were rotated for another 3 h at 4°C. Samples were centrifuged for 5 min at 503 1200xg and the supernatant was discarded. Beads were then washed three times 504 with 500 µl of TEN-Buffer and frozen at -80°C. For Western Blot analysis 30 µl of 505 3xLaemmli buffer was added to the beads and incubated at 95°C for 5 min. 30 µl of 506 the samples were loaded on 10% 1,5 mm SDS gels. 507 For in vivo Co-Immunoprecipitation, neocortices of E18.5 WT embryos were 508 dissected and either used directly or stored at -80°C until further use. Two cortices 509 were merged and homogenized in 1 ml of Buffer A. Samples were treated as 510 were performed for genotypes and for each mouse individually. Quality control 571 thresholds were set to 1,000-5,000 genes per cells, 1800-10000 UMIs and <6% of 572 mitochondrial transcripts. Only samples with more than 500 cells after filtering were 573 used to ensure a complete reproduction of cell diversity in the neocortex. 3 samples 574 for WT and 2 samples for KO were used for further analysis. We had to exclude one 575 WT animal that displayed lower Tcf4 expression than the KO and we also excluded 576 cells that displayed a high background transcript expression of blood related genes 577 such as Hbb-a1, Hbb-a2. 578

scRNA-Seq clustering and differential gene expression analysis using Seurat 579
Clustering of the cells was performed using the Seurat packages for R following the binarized. To that end a threshold was set at the mean of the area under the curve. 595 In cells below the threshold the GRN was considered not active (OFF) whereas in 596 cells above it was considered active (ON). As cells clustered apart according to the 597 genotype, a list of GRNs was identified which where only active in one genotype. It 598 was hypothesised that TCF-4 interacts with the heads of the GRNs and modulates 599 their activity. To obtain a manageable list of candidate genes which might interact 600 with TCF-4 the list of GRNs heads were analysed using the Panther online tool (Mi et  DisGeNET database (https://disgenet.org) (Pinero et al. 2020). 608

Statistical analysis 609
To determine statistical significance Mann-Whitney-U test was performed using the 610 ggplot2 implementation of R (*, P ≤0.05; **, P ≤0.01, ***, P ≤0.001) if not otherwise 611 indicated. n is indicated in the figure legends. Data is depicted as mean ± SD. 612 To determine whether differences in luciferase activities ( Figure 4B and C) were 613 statistically significant, a two-tailed student´s t-test was performed using the ggplot2 614 implementation of R (*, P ≤0.05; **, P ≤0.01, ***, P ≤0.001). Data is depicted as mean 615 ± SD. Results from independent transfections were treated as biological replicates. 616

Data and code availability 617
The accession number for the single-cell RNA Sequencing of E18.5 neocortices is 618 GEO: GSE147247. 619

Acknowledgments 620
We thank Silvia Cappello, Michael Wegner and all members of the Institutes of 621 Human Genetics and Biochemistry for helpful discussions.

Declaration of Interests 636
The authors declare no competing interests. 637