Members of the POU and SOX transcription factor families exemplify the partnerships established between various transcriptional regulators during early embryonic development. Although functional cooperativity between key regulator proteins is pivotal for milestone decisions in mammalian development, little is known about the underlying molecular mechanisms. In this study, we focus on two transcription factors, Oct4 and Sox2, as their combination on DNA is considered to direct the establishment of the first three lineages in the mammalian embryo. Using experimental high-resolution structure determination, followed by model building and experimental validation, we found that Oct4 and Sox2 were able to dimerize onto DNA in distinct conformational arrangements. We demonstrate that the DNA enhancer region of their target genes is responsible for the correct spatial alignment of glue-like interaction domains on their surface. Interestingly, these surfaces frequently have redundant functions and are instrumental in recruiting various interacting protein partners.[Keywords: Oct4; Sox2; POU domain; HMG domain; FGF4 and UTF1 enhancers; crystal structure] Supplemental material is available online at http://www.genesdev.org.

Two crystal structures of Oct-1 POU domain bound to DNA provide a rationale for differential, conformation-dependent recruitment of transcription cofactors. The POU-homeo and POU-specific subdomains of Oct-1 contain two different nonoverlapping pairs of surface patches that are capable of forming unrelated protein-protein interfaces. Members of the POU factor family contain one or two conserved sequence motifs in the interface that are known to be phosphorylated, as noted for Oct-1 and Pit-1. Modeling of Oct-4 reveals the unique case where the same conserved sequence is located in both interfaces. Our studies provide the basis for two distinct dimeric POU factor arrangements that are dictated by the architecture of each DNA response element. We suggest interface swapping in dimers could be a general mechanism of modulating the activity of transcription factors.

POU domain proteins contain a bipartite DNA binding domain divided by a flexible linker that enables them to adopt various monomer configurations on DNA. The versatility of POU protein operation is additionally conferred at the dimerization level. The POU dimer formed on the PORE (ATTTGAAATGCAAAT) can recruit the transcriptional coactivator OBF-1, whereas POU dimers formed on the consensus MORE (ATGCATATGCAT) or on MOREs from immunoglobulin heavy chain promoters (AT[G/A][C/A]ATATGCAA) fail to interact. An interaction with OBF-1 is precluded since the same Oct-1 residues that form the MORE dimerization interface are also used for OBF-1/Oct-1 interactions on the PORE. Our findings provide a paradigm of how specific POU dimer assemblies can differentially recruit a coregulatory activity with distinct transcriptional readouts.

The POU transcription factors Oct1 and Oct2 bind to DNA in various monomer and dimer configurations. Depending on the DNA sequence to which they bind, the dimers are arranged in configurations that are either accessible (PORE sequence) or inaccessible (MORE sequence) to the B-cell-specific cofactor OBF1 (OcaB, Bob1). As shown previously, the MORE and related sequences (such as the heptamer/octamer motif) are found in immunoglobulin heavy chain promoters. Here we show that the expression of Osteopontin, which contains a PORE sequence in its enhancer region, depends on the presence of OBF1 in B cells. OBF1 alleviates DNA sequence requirements of the Oct1 dimer on PORE-related sequences in vitro. Furthermore, OBF1 stabilizes POU dimer-DNA interactions and overrides Oct1 interface mutations, which abolish PORE-mediated dimerization without OBF1. Our data indicate that the PORE-type Oct1 or Oct2 dimer, rather than the monomer, is the primary target of the cofactor OBF1. Based on our biochemical data, we propose a mode of OBF1-Oct1 dimer interaction, suggesting a novel arrangement of the subdomain connectivities.