RBFOX1 regulates both splicing and transcriptional networks in human neuronal development

Abstract

RNA splicing plays a critical role in the programming of neuronal differentiation and, consequently, normal human neurodevelopment, and its disruption may underlie neurodevelopmental and neuropsychiatric disorders. The RNA-binding protein, fox-1 homolog (RBFOX1; also termed A2BP1 or FOX1), is a neuron-specific splicing factor predicted to regulate neuronal splicing networks clinically implicated in neurodevelopmental disease, including autism spectrum disorder (ASD), but only a few targets have been experimentally identified. We used RNA sequencing to identify the RBFOX1 splicing network at a genome-wide level in primary human neural stem cells during differentiation. We observe that RBFOX1 regulates a wide range of alternative splicing events implicated in neuronal development and maturation, including transcription factors, other splicing factors and synaptic proteins. Downstream alterations in gene expression define an additional transcriptional network regulated by RBFOX1 involved in neurodevelopmental pathways remarkably parallel to those affected by splicing. Several of these differentially expressed genes are further implicated in ASD and related neurodevelopmental diseases. Weighted gene co-expression network analysis demonstrates a high degree of connectivity among these disease-related genes, highlighting RBFOX1 as a key factor coordinating the regulation of both neurodevelopmentally important alternative splicing events and clinically relevant neuronal transcriptional programs in the development of human neurons.

Figure 1.

Characterization of RBFOX1 expression in human brain and fetal-derived neural progenitor cells. (A) A schematic illustration of the RBFOX1 genomic organization is shown. Figure adapted from Underwood et al. (9). Untranslated exons are shown in light gray, translated exons are in white. The brain-specific exon 16 is shown in dark gray, whereas the muscle-specific exon 17 is shown in black. The location of the RNA-binding domain (RRM) is indicated. (B) The expression level of RBFOX1 was assessed by qRT-PCR, with mRNA from PHNP cells differentiated for the indicated times. Primers were directed against exons 8–9 to detect all RBFOX1 isoforms (light gray). Autoregulatory alternative splicing of exon 11 eliminates RNA binding, so primers against exons 9–11 were utilized to detect isoforms with an active RNA-binding domain (dark gray). (C) In situ hybridization was performed with a human fetal brain, age of 19 weeks, using an S³⁵-labeled antisense riboprobe directed against exons 8–13 of RBFOX1. Two representative coronal and sagittal sections are shown. The sense probe is used as a control (lower panels). (D) To quantitate the pattern of RBFOX1 isoforms expressed in the indicated tissues and cell lines, RT-PCR was performed using primers to amplify exons 15–20, which represent the largest region of alternative splicing diversity in the gene. Amplified products were subcloned and sequenced. Total clones are indicated with the representative counts and percentages of the various alternative spliced isoforms. The most highly expressed patterns are highlighted in gray. c, caudate; cp, cortical plate; gz, germinal zone; p, putamen; t, thalamus.

Figure 2.

RNA sequencing detects altered alternative splicing patterns in PHNP cells with RBFOX1 knockdown. (A) Heat map showing clustering of gene expression among five biological replicates representing three experimental conditions; wild-type (black), RBFOX1 knockdown (shRBFOX1, red) and non-targeting RNA interference (shGFP, green). Analysis is of the top 250 most significant genes, using the Bayes method with a Spearman correction. (B) Analysis of the intronic regions 400 nucleotides upstream and downstream of the alternative exons whose splicing was most significantly affected by RBFOX1 knockdown for the presence of the binding sites for RBFOX1, NOVA1 or PTBP1. Observed sites are shown as well as the number predicted by iterative analysis of an equivalent number of random introns culled from all human genes. Enrichment of the various sites is indicated by gray boxes and arrows. Significance is based on the normal distribution. ns, not significant. A schematic illustration of the predicted effects on alternative splicing based on the location of the RBFOX1-binding site is shown, with downstream sites enhancing and upstream sites repressing exon inclusion. The correlation between RBFOX1-binding site location and splicing changes identified by RNA sequencing in this study is shown. (C) Validation of splicing changes detected by RNA sequencing. Exons are labeled using a sequential annotation based on location within the gene. Genomic coordinates can be found in Supplementary Material, File S1. qRT-PCR or semi-qRT-PCR was used to calculate the ratio of exon inclusion in the RBFOX1 knockdown cells lines when compared with the shGFP control line. A selection of 25 genes is shown with the differential fold change (log base 2) in exon inclusion detected by RNA sequencing shown in red and the observed fold change by RT-PCR shown in blue. Standard error of the mean is indicated by black bars. (D) Comparison of the RBFOX1 gene set with published gene lists. The number of overlapping genes is indicated along with the percentage they represent from each list. Lists were derived from the references indicated and are also shown in Supplementary Material, File S3. Online sources for gene lists include the Genes to Cognition (G2C) database (http://www.genes2cognition.org/), the Organelle DB (http://organelledb.lsi.umich.edu/), the Online Mendelian Inheritance in Man database (http://www.omim.org/), the GeneTests database (http://www.ncbi.nlm.nih.gov/sites/GeneTests/) and the Simons Foundation Autism Research Initiative database (https://sfari.org/). Lists referenced as supplemental are composites of multiple lists derived from the above sources. P-values were determined based on hypergeometric probability. ER, endoplasmic reticulum.

Figure 3.

Characterization of differential gene expression in RBFOX1 knockdown cells. (A) A selection of 44 genes is shown with the differential fold change (log base 2) in gene expression detected by RNA sequencing shown in red and the observed fold change by qRT-PCR shown in blue. Standard error of the mean is indicated by black bars. (B) Comparison of the RBFOX1 differentially expressed gene set with published gene lists. The number of overlapping genes is indicated along with the percentage they represent from each list. Lists were derived from the references indicated and are also shown in Supplementary Material, File S6. Online sources for gene lists are as described for Figure 2. Lists referenced as supplemental are composites of multiple lists derived from the above sources. P-values were determined based on hypergeometric probability. ER, endoplasmic reticulum.

Figure 4.

WGCNA in the RBFOX1 knockdown cell line reflects pathways important to neurodevelopment and to autism. For clarity, only the most highly connected module members are shown. Genes with the highest connectivity (i.e. hubs) are indicated in red. (A) Blue module. (B). Yellow module. Differentially expressed ASD genes are in purple.

Figure 5.

A model of RBFOX1 function in PHNPs. During neuronal differentiation, RBFOX1 is induced and directly modulates (solid arrows) an RNA splicing network (black box) which in turns coordinately regulates a transcriptional network of additional genes (black box). Downstream genes present within these networks can further modulate either RNA splicing or transcription to generate additional layers of control (dashed arrows). RBFOX1 can further alter its own splicing and downregulate the activity of these networks as indicated. Additional regulatory factors and/or programs can also contribute to the coordinate regulation of these networks (dotted arrows). The number of factors identified in this study at each of the steps is indicated with selected proteins of interest as mediators noted with their predicted regulatory contribution indicated, see text for complete details. The genes within these networks affect a number of key cellular developmental processes that promote neural development and synaptogenesis, leading to the formation of mature neurons. Disruption of these pathways, particularly genes in the transcriptional network, can lead to neurodevelopmental and/or neuropsychiatric phenotypes in humans. The asterisk indicates factors involved in splicing and/or other aspects of RNA-processing.