DNA binding specificity for the bHLH-PAS transcription factor family

Transcription factor-DNA interactions and their specificities have been described for many different classes of transcription factor families. However, heterodimeric transcription factor complexes still remain poorly characterised. The basic-Helix-Loop-Helix (bHLH) transcription factor family is one of the largest transcription factor families that typically bind DNA though a degenerate CANNTG elements as heterodimers or homodimers. Here we characterise the DNA binding of the bHLH - Per-Arnt-Sim (PAS) (bHLH-PAS) domain containing transcription factor family using SELEX-high-throughput sequencing coupled with quantitative computational modelling analysis. We show that most dimeric bHLH-PAS transcription factors bind to distinct core NNCGTG response elements but bind over a much larger footprint than previously characterised. Modelled DNA-protein interactions were found to correlate with structural analysis, DNA shape predictions and in vivo transcription factor occupancy. Overall design: We determined the DNA binding specificity of the bHLH-PAS transcription factors by SELEX-seq and No-Read-Left Behind (NRLB) Protein-DNA Modelling. We purified full length human ARNT (1-774) or ARNT2 (1-717) homodimers from Expi293 cells, bHLH-PAS (PAS A and PAS B, N-terminus) heterdimers (human HIF1a(1-375)/ARNT(1-474), human HIF2a(1-353)/ARNT(1-474), human SIM1(1-348)/ARNT2(1-439) and mouse NPAS4(1-329)/ARNT2(1-481)) from baculovirus infected Sf9 insect cells or human AhR(1-287)/ARNT(1-364) from bacteria. Seven purified homo- or heterodimer complexes were incubated with SELEX libraries (Random 18mer or FixedCore (8N-CGTG-10N) 22mer) and bound DNA were separated by electrophoretic mobility assay (EMSA), followed by DNA isolation and PCR amplification. Random 18mer libraries were subjected to 3 rounds of SELEX, whereas FixedCore (8N-CGTG-10N) 22mer libraries were subjected to 1 or 2 rounds of SELEX. The Random 18mer initial library and round 3 were sequenced using a NextSeq500 1x75bp HighOutput mode yielding ~30million reads per sample. The FixedCore (8N-CGTG-10N) 22mer initial library, round 1 and round 2 were sequenced using a HighSeq2500 yielding ~30million reads per sample.