Crystal Structure of a Free κB DNA: Insights into DNA Recognition by Transcription Factor NF-κB

The dimeric NF-κB transcription factors regulate gene expression by recognizing specific DNA sequences located within the promoters of target genes. The DNA sequences, referred to as κB DNA, are divided into two broad classes. Class I κB DNA binds optimally to p50 and p52 NF-κB subunits, while class II κB DNAs are recognized specifically by the NF-κB subunits c-Rel and p65. We determined the X-ray crystal structure of a class II κB DNA sequence at 1.60 Å resolution. This structure provides a detailed picture of κB DNA hydration, counter ion binding, and conformation in the absence of NF-κB binding partner. X-ray structures of both class I and class II κB DNA bound to NF-κB dimers were determined previously. Additionally, the NMR solution structure of a class I κB DNA is known. Comparison of the protein-bound and unbound κB DNA structures reveals that the free form of both classes approximates ideal B-form DNA more closely. Local geometries about specific DNA bases differ significantly upon binding to NF-κB. This is particularly evident at the 5′-GG/CC base-pairs; a signature of NF-κB specific DNA binding sequences. Differential phosphate group conformations, minor groove widths, buckle, twist, and tilt angles are observed between bound and unbound κB DNA. We observe that the presence of an extra G:C base-pair, 5′- to the GGA sequence in class I κB DNA, alters the geometry of the two internal G:C base-pairs within the GGGA tetranucleotide, which explains, at least in part, the structural basis for distinct NF-κB dimer recruitment by the two different classes of κB DNA. Together, these observations suggest that NF-κB dimers recognize specific structural features of κB DNA in order to make sequence-specific complexes.

Introduction

The Rel/nuclear factor (NF)-κB family of inducible transcription factors are key regulators of a diverse array of cellular activities including inflammation, innate and adaptive immune responses, cell growth, proliferation, differentiation, and apoptosis.¹ The NF-κB family consists of five polypeptide subunits, p50 (NFκB1), p52 (NFκB2), p65 (RelA), c-Rel, and RelB. Each NF-κB family member bears an approximately 300 residues long Rel homology region (RHR) near its N terminus. The subunits associate combinatorially to form functional homo- and heterodimers.1, 2, 3 In uninduced cells, NF-κB dimers with transcription activation potential are maintained inactive through association with members of the inhibitor of κB (IκB) family of proteins. The X-ray crystal structures of IκB/NF-κB complexes have revealed the molecular mechanisms of NF-κB inactivation.4, 5, 6 Upon induction, IκBs are removed through a ubiquitin-dependent 26 S proteasome-mediated degradation pathway, allowing free NF-κB to concentrate within the nucleus and activate gene transcription by binding to sequence-specific target DNA in gene enhancers.2, 3, 7 These DNA target sites are referred to collectively as κB DNA. Most physiological κB DNA sites are 10 bp in length and exhibit the consensus sequence 5′-GGGRN W YYCC-3′, where R, N, W, and Y denote purine, any nucleotide, A or T, and pyrimidine bases, respectively. The κB DNA consensus sequence is degenerate and hundreds of variations exist in gene promoters. Different NF-κB dimers show significant overlap in binding to κB DNA variants in vitro.

X-ray structures of several NF-κB homo- and heterodimers bound to different κB DNA sequences have been determined.8, 9, 10, 11, 12, 13 These structures serve to explain the rules of preferential DNA target recognition by different NF-κB dimers. The NF-κB p50 subunit recognizes the 5′-GGGRN-3′ half-site, whereas the p65 subunit binds specifically to the 5′-YYCC-3′ half-site. This specificity arises, in part, from the fact that both p50 and p52 monomers contact the 5′-G with a histidine side-chain unique to these subunits. This histidine residue is replaced by alanine within the sequences of p65 and c-Rel. As a result, the length of the half-site that these two NF-κB subunits recognize is shortened by one base-pair. Using this model of binding, p50 and p52 homodimers appear to bind optimally to 11 nt κB sites, whereas heterodimers containing p50 or p52 as one of the subunits prefer 10 bp sites. NF-κB dimers composed of p65 and c-Rel subunits only, on the other hand, prefer 9 bp κB DNA.

Both p50 and p52 homodimers and NF-κB heterodimers containing one p50 or p52 subunit bind to most 10 nt and 11 nt κB DNA sites with reasonably high affinity.¹⁴ In contrast, these dimers do not bind with appreciable affinity to 9 bp κB DNA. Even more surprisingly, the 9 bp consensus site preferred by p65 and c-Rel (5′-GGAA W TTCC-3′) is embedded within the 10 bp consensus (GGGRN W TTCC). Although one might consequently expect that p65 and c-Rel homo- and heterodimers should bind to both the 9 bp and the 10 bp consensus sites with similar affinities, such is not the case. This suggests that, whereas the aforementioned 5′-G at the first position promotes binding by NF-κB p50 or p52 subunits, the same base-pair negatively affects c-Rel and p65 subunit binding by some indirect manner. We observe that p65 and c-Rel homo- and heterodimers bind preferably to κB sites that contain five or six A/T base-pairs at the center while p50 homo- and heterodimers do not require such a long A/T-rich segment for high binding affinity. Prior to these structural and biochemical works, Parry et al. demonstrated that p65 and c-Rel homo- and heterodimers preferred the 9 bp κB-like sites in vivo.¹⁵ On the basis of these observations, we propose to classify physiological κB DNA sites into two broad classes: the 10 bp or 11 bp class I κB sites preferred by p50 and p52 homo- and heterodimers; and the 9 bp class II κB sites preferred by p65 and c-Rel homodimer and p65/c-Rel heterodimer (Figure 1(a)).15, 16

It is known that even small changes in promoter-specific κB DNA that do not alter binding affinity for NF-κB, alter the gene expression profile.17, 18, 19 Although the X-ray structures have revealed a stereochemical mechanism of how NF-κB dimers bind κB DNA sequences, the role of DNA conformation in the complex formation remains underappreciated. The elucidation of such a mechanism requires a firm understanding of the structure and dynamics of κB DNA. While it is known that NF-κB subunits play a specific role in regulating gene expression, without a correlation between binding affinity, in vivo specificity of NF-κB binding to different κB DNA, and conformational changes directed by κB DNA target sequences, it remains unclear why so many different κB sites can be found within the genomes of higher eukaryotes. Here, we report the X-ray crystal structure of the duplex κB DNA sequence 5′-CGCTGGAAATTTCCAGC-3′ at 1.60 Å resolution. This DNA sequence encompasses a palindromic κB DNA, referred to as κB-33 DNA (5′-GGAAATTTCC-3′), which was designed on the basis of the p65 half-site binding sequence from the interferon-β (IFN-β) gene promoter.²⁰ This sequence retains NF-κB binding and transcriptional activity, as measured by a cell-based reporter expression assay.²⁰ The identical sequence was used earlier to form the complex with the NF-κB p65 homodimer, and the X-ray crystal structure of that complex was determined.¹¹ We obtained the crystals of the free DNA fortuitously while attempting crystallization of the complex between it and v-Rel, the oncogenic version of c-Rel. We discuss the structural features of this free class II κB DNA and compare it to both NF-κB bound κB DNA and the solution structure of a free class I κB DNA determined by NMR spectroscopy. The comparison reveals structural features that directly and indirectly play a role in the NF-κB/κB DNA complex formation.