Binding determinants of high mobility group proteins in the mouse genome

In this work we investigate the determinants of recruitment to DNA and chromatin for HMGB1-2-3-4 and HMGA1-2 and a selection of transcription factors (TF). We adopt a mouse embryonic stem cell (ESC) model system for the generation of antibody independent ChIP-sequencing data. We first report successful recapitulation of Sox2 binding, our internal control, and then focus on HMGA and HMGB proteins, for which no exhaustive genome-wide data had been available.
In the nucleus HMG proteins are one of the major chromatin-associated non-histone proteins. As such they have been implicated in a wide range of nuclear processes from transcription, to nucleosome remodeling, DNA damage and apoptosis.
For HMGB proteins we show frequent contacts with active regulatory regions, which however are also sites of preferred interaction for sequence-unspecific DNA binders and inert proteins such as DNAseI or monomeric GFP. Upon mutation of the DNA binding domains of Hmgb1 no change in the localization pattern for this protein is observed. Additionally upon Hmgb1 knock out (KO), ESC do not show alterations in transcription, as one would expect for a protein involved in regulatory functions. Nevertheless we cannot formally exclude that the biotin tagging is causing a mislocalization of the HMGB proteins, nor that upon Hmgb1 KO HMGB2 may compensate for HMGB1 absence.
As far as HMGA1 and HMGA2 are concerned, on the contrary we show binding throughout the genome with a preference for AT-rich DNA. Mutation of key residues in the DNA binding domains of both proteins causes loss of the AT dependence and the residual signal is comparable to that of a freely diffusing protein (monomeric GFP). Importantly AT-rich dependence is independent of chromatin states, as exemplified by invariance upon neuronal differentiation. These results highlight the fact that the three DNA binding domains of HMGA1 and HMGA2 are the sole determinants of their genomic distribution.
At the chromosomal scale, we show that enriched regions are also generally positive for features of heterochromatin such as presence of Histone H3 Lysine9 methylation, their late replication in S phase and their association with the nuclear lamina. This data points to enrichment of HMGA1-2 at constitutive heterochromatin, which has a known compositional bias. Lastly, we show a limited role for HMGA1 in the regulation of transcription in ESC by profiling expression patterns of an isogenic KO cell line.
Taken together, the findings on HMGA proteins reveal a broad DNA-binding modality, which supports their known preference for AT-rich DNA. At the same time, our genomic and gene expression results are in contrasts with the often-mentioned roles in transcriptional regulation.