RESULTS: iCLIP analysis found SAFB1 binding was enriched, specifically in exons, ncRNAs, 3' and 5' untranslated regions. SAFB1 was found to recognise a purine-rich GAAGA motif with the highest frequency and it is therefore likely to bind core AGA, GAA, or AAG motifs. Confirmatory RT-PCR experiments showed that the expression of coding and non-coding genes with SAFB1 cross-link sites was altered by SAFB1 knockdown. For example, we found that the isoform-specific expression of neural cell adhesion molecule (NCAM1) and ASTN2 was influenced by SAFB1 and that the processing of miR-19a from the miR-17-92 cluster was regulated by SAFB1. These data suggest SAFB1 may influence alternative splicing and, using an NCAM1 minigene, we showed that SAFB1 knockdown altered the expression of two of the three NCAM1 alternative spliced isoforms. However, when the AGA, GAA, and AAG motifs were mutated, SAFB1 knockdown no longer mediated a decrease in the NCAM1 9-10 alternative spliced form. To further investigate the association of SAFB1 with splicing we used exon array analysis and found SAFB1 knockdown mediated the statistically significant up- and downregulation of alternative exons. Further analysis using RNAmotifs to investigate the frequency of association between the motif pairs (AGA followed by AGA, GAA or AAG) and alternative spliced exons found there was a highly significant correlation with downregulated exons. Together, our data suggest SAFB1 will play an important physiological role in the central nervous system regulating synaptic function. We found that SAFB1 regulates dendritic spine density in hippocampal neurons and hence provide empirical evidence supporting this conclusion.
CONCLUSIONS: iCLIP showed that SAFB1 has previously uncharacterised specific RNA binding properties that help coordinate the isoform-specific expression of coding and non-coding genes. These genes regulate splicing, axonal and synaptic function, and are associated with neuropsychiatric disease, suggesting that SAFB1 is an important regulator of key neuronal processes.
RESULTS: We show that loss of Hmgn1 or Hmgn2 in pluripotent embryonal carcinoma cells leads to increased levels of spontaneous neuronal differentiation. This is accompanied by the loss of pluripotency markers Nanog and Ssea1, and increased expression of the pro-neural transcription factors Neurog1 and Ascl1. Neural stem cells derived from these Hmgn-knockout lines also show increased spontaneous neuronal differentiation and Neurog1 expression. The loss of HMGN2 leads to a global reduction in H3K9 acetylation, and disrupts the profile of H3K4me3, H3K9ac, H3K27ac and H3K122ac at the Nanog and Oct4 loci. At endodermal/mesodermal genes, Hmgn2-knockout cells show a switch from a bivalent to a repressive chromatin configuration. However, at neuronal lineage genes whose expression is increased, no epigenetic changes are observed and their bivalent states are retained following the loss of HMGN2.
CONCLUSIONS: We conclude that HMGN1 and HMGN2 maintain the identity of pluripotent embryonal carcinoma cells by optimising the pluripotency transcription factor network and protecting the cells from precocious differentiation. Our evidence suggests that HMGN2 regulates active and bivalent genes by promoting an epigenetic landscape of active histone modifications at promoters and enhancers.