SUB CAPITULO Nº 02

The second part of the DNA binding domain is the basic region, which is responsible for contacting the DNA. It contains 8-10 amino acids as the contact face (Weiss et al., 1990). Kouzarides and Ziff (1988) showed that the basic motifs of Fos, Jun and GCN4 have comparable affinities for an oligonucleotide containing a TRE.

GCN4 does not distinguish between the ATF/CREB DNA site (CRE) and its shorter consensus site (TRE), either in vivo or in vitro (Sellers et a i, 1990). The central base pairs of the CRE were shown to be important and that GCN4 specifically recognised

the central base pair of its consensus binding site (5'-ATGA[C/o]TCAT-3'). The mechanism of binding was proposed to be through two overlapping and nonequivalent half sites: 5 -ATGAC-3' and 5-ATGAG-3'. Data from Park etal. (1992) supported the half site binding mechanism for LCC proteins to bind DNA. V-Jun basic regions were linked together by the amino termini. The orientation of the basic region in the resultant dimer was reversed. The altered peptide bound to a DNA sequence in which the position of the two half sites was reversed: 5 -TC A TX A T G A -3W here 'X' was zero to two additional nucleotides added to the sequence. Binding was estimated to occur with nanomolar affinity.

Models for DNA binding.

Two models for DNA binding by LCC proteins have been proposed, both have a bifurcating Y shape and differ only in the flexing of the basic region as it follows the DNA site.

Saudek et at. (1991) attempted to use NMR to obtain a structure of the basic region of a GCN4 peptide in the absence of DNA. The LCC was highly structured, but the flexibility of the basic region prevented the successful application of distance geometry calculations. There was some evidence of helicity in the absence of DNA.

Scissor's Grip model.

Vinson et al. (1989) proposed a model for DNA binding based on the dyad symmetry of the DNA sites of the bZIP family of proteins. A comparison of the amino acid sequences of eleven bZIP proteins, including C/EBP, Jun, Fos, GCN4 and CREB reveals a number of conserved amino acids in the basic region.

Assuming that the a-helix of the LCC continues into the basic region then in a coiled coil the seventh residues are opposed, in this region there is always a lysine or an arginine (at position 7), suggesting that these helices would repel one another leading to a bifurcating "Y” shaped structure. This explains the dyad symmetric binding site for these proteins. It was proposed from these data that the DNA binding region would bind DNA as a fork shaped structure. Each basic region would track along the DNA in opposite

directions. In other words one basic region of the dimer for each half site. Alteration of the spacing of the basic region from the LCC by even a single residue eliminated specific DNA binding in a disulphide bonded dimer (Smeal et al., 1989).

The amino termini of a truncated GCN4 homodimer lie in adjacent major grooves 9-10 basepairs apart and symmetrically displaced 4 - 5 basepairs from the central cytosine of the recognition site of the sequences 5'-CTGACTAAT-3' and 5'-ATCGACTCTT-3' (Tullius and Dombrowski, 1986; Oakley and Dervan, 1990). The data from both groups suggests that the GCN4 homodimer binds the TRE and CRE sites in much the same fashion.

However, a continuous a-helix tracking the DNA would tend to protrude beyond the major groove. Mutagenesis studies suggested that both the region distal to the leucine zipper (which would protrude) and that proximal were equally important. Thus there must be a bend in the alpha helix to allow continuous tracking of the major groove. Due to its properties as a helix capping residue, an asparagine was proposed for this role.

Mutation of the conserved asparagine (Asn 235 of GCN4) was used to evaluate the scissor's grip model. Tryptophan 235 variants bind to an alternative DNA site: 5'- TTGACTCAA-3' with similar affinity to the consensus GCN4 site: 5 -ATGACTCAT-3'. Asn 235-GCN4 discriminates against the mutant site.

However, Asn 235-GCN4 binds to the DNA site GTGACTCAC, whereas the Trp 235 variant does not. This suggested a direct interaction between Asn-235 and the *^/-4 position of the half site: 5’-A[-4] T[-3 ]G[-2] A[-l] C[0] T[+l] C[+2] A[-k3] T[-h4]-3'. In the scissor's grip model, if the asparagine were vital for DNA binding, then its substitution would be predicted to all but abolish DNA binding (Tzamarias et at., 1992). Pu and Struhl (1991) also mutated Asn-235 as well as the two conserved alanines (Ala-238, Ala- 239). The Asn-235 mutants bound DNA.

Induced Helical Fork model.

An alternative model was proposed to account for CD data. When GCN4 and the Fos:Jun heterodimer bound their specific DNA site a random coil to a helix transition was observed. The folding of the basic region was proposed as the cause of the transition.

Instead of kinking at the asparagine, it was suggested that there was a more continuous curve.

The induced helical fork model predicts that Ala 238, Ala 239, Asn 235 and Cys 242 of GCN4 make base specific contacts with DNA (O'Neil et a/., 1991). Surrounding this conserved quartet were positively charged amino acids, such as arginine and lysine. These could be involved in backbone interactions. Blatter et al. (1992) used bromouracil crosslinking to show that the thymine (-3) interacted with the sidechain of alanine 238.

Mutagenesis of the second conserved alanine (Ala 239) to an aromatic, acidic, hydrophobic or neutral amino acid altered binding to DNA. Of the variants which bound DNA, those containing serine, asparagine and cysteine changed the affinity of the basic region. Serine and cysteine variants would bind to DNA sites in which the last adenine of the half site was altered to a thymine or a guanine (GCN4 half site: 5'-ATGA^/g"3').

Whereas wild type GCN4 would not bind to these sites at all. Cysteine and asparagine variants changed the specificity of binding whilst serine variants broadened it. Glutamate and valine variants had similar affinities to the wild type basic region sequence. The breadth of the affinity of the wild type basic region presumably reflects the presence of a number of different sequences in vivo. These data suggested alanine 239 formed a base specific contact with the DNA (Suckow et al., 1993).

The respective partners in the DNA of the two conserved alanines were likely to be the thymine methyl groups. Replacing thymine with uracil would remove these interactions, as uracil lacks the methyl group of thymine. The DNA binding affinity for such a site would be predicted to be lower than for a sequence containing thymine.

Replacement of thymines had more effect upon the binding if the substitutions were of those on the noncoding strand. As predicted by the induced helical fork model the thymine at position -3 (furthest from the centre of the site) was relatively more important than the thymine at position +1. This suggests an inherent asymmetry in the way in which GCN4 binds DNA. These data supported observations made by Oliphaunt et al. (1989), who found that mutations on the 3' side of the central G-C basepair were more important than those to the 5' side. The Cys 238 variant, in fact, behaves indistinguishably from the wild type GCN4, as do the Val 238 and Gly 239 variants for uracil substitutions on the

non-coding strand(Pu and Struhl, 1991).

No NMR structure has been solved for a bZIP protein bound to DNA.

Four crystal structures have been solved for the DNA binding domain of bZIP proteins bound to DNA. Two of GCN4 complexed to each of the two DNA sites - either the symmetrical CRE (Konig and Richmond; 1993) or the API DNA site, the TRE (EUenberger etal.; 1992). The other two were of the FosJun heterodimer complexed to both sites (Glover and Harrison, 1995). In support of previous structural studies carried out on the dimérisation region of bZIP peptides, both groups reported that the peptide bound DNA as a dimer of uninterrupted a helices. All o f the reported structures were at a resolution of about 3.0 Â, sufficient for resolution of the backbone of the protein.

These structures are discussed in more detail together with the solution structure studies for the LCC in chapter 6. In none of these structures was there a significant kink in the peptide backbone, instead tracking of the DNA was reported to be through a more continuous curve coupled with a slight distortion of the DNA.