The Fos:Jun-IAF complex gave an increase in the emission maximum upon addition of oligodeoxynucleotides containing the AP I DNA binding site. Gel retardation studies had shown that the Fos-Jun-IAF complex was capable of binding to the TRE (Patel
et al., 1990).
In some members of the bZIP family, most notably GCN4, the cysteines are replaced by serines. Serine variants of wbFos and wbJun, gave functionally active
proteins as judged by DNA binding (Abate et al. 1990c; Glover and Harrison, 1995). In the reducing environment of the cell any sulphydryls would be reduced at all times. Previous workers considered the oxidation of the sulphydryls a possible control mechanism (Abate etal. 1990b). This is unlikely given the reducing conditions found in the cell.
Replacing the cysteine with a methionine interfered with the DNA binding of the heterodimer. Methionine is significantly larger than either cysteine or serine and is not a hydrogen donor (Abate et at., 1990c). These data would be consistent with the disruption of a base specific contact. However, compared to N-ethylmaleimide, the sidechain of methionine is fairly small.
Glover and Harrison (1995) recently solved a crystal structure to 3.05 Â for a FosiJun heterodimer bound to the TRE and CRE. The crystals of the serine variant were more stable. The DNA binding between both variants was the same. Serine 242, the equivalent of the conserved cysteine seen in other members of the bZIP family, was directly bonded to the phosphates. Thus the presence of an inactive (methionine) sidechain or bulky group (n-ethylmaleimide) would be predicted to interfere with DNA binding. The gel retardation data showed that the lAA modified heterodimer had reduced affinity for DNA.
In contrast to the experimental work in this thesis, previous studies have shown that the addition of DNA containing the TRE or CRE sites increased the alpha helicity of the heterodimer by 10 %. The DNA effect was concentration dependent and saturation occurred at a protein: DNA ratio of 1:1 (Patel et al.y 1990). Deletion mutants of Fos and Jun which lack the basic region showed no change in a helicity upon the addition of the TRE site (Patel et al.,1990). The 10 % increase is consistent with the effect of 50 % TEE seen in this experimental work. However, the wbFos-IAD:wbFos-IAD did not bind DNA. The absence of binding in the quantitative gel retardation studies was probably due to the low concentrations used. In other words there was no dimer. However the lack of any increase in the CD of the wbFos:wbFos dimer suggests that there was no DNA binding or binding occurred by an alternative mechanism.
yet no NMR structure of an LCC protein bound to DNA has been published. Thus there is a lack of dynamic information available. Two crystal structures of GCN4 have been reported, and two of the FosJun heterodimer.
EUenberger et al. (1992) deduced a structure for the GCN4 LCC and basic region bound to the TRE at 2.9 Â resolution. Konig and Richmond (1993) resolved the crystals of GCN4 LCC and basic region bound to the CRE to 3 Â. This was sufficient to trace the backbone and most of the sidechains. There were no major distortions of the protein and only a slight alteration in the DNA. Though no structure of the DNA alone was published.
The parallel coiled-coil of the dimérisation region ends after the amino-terminal leucine of the coiled-coil. At this point the helices of the basic region diverge in a smooth curve to contact the DNA. This supports the induced helical model.
Konig and Richmond (1993) identified a slight bend around the DNA from alanine 233 to alanine 239, similar to that reported by EUenberger et al. (1992). The C a distances over the whole of this region for asparagine 235 to alanine 238 were 4.7 Â and for asparagine 235 to alanine 239 were 5.9 Â. This compares to i - i+3 and i - i-h4 distances of 5 Â and 6.2 Â respectively for more typical a helices.
In both structures the conserved asparagine (235) binds to the edge of the bases T(-4), (the asparagine NÔ to the thymine 04) and C(3) (the asparagine OS to the thymine N4) as predicted in the induced helix model. It recognises the A T and C G base pair. This was part of the reason for the distortion reported for the DNA containing the CRE (Konig and Richmond 1993). Both alanine 238 and 239 contact the methyls of T (2) and T (-4). Thus the sequence TCA of each half site is recognised by these three amino acids. Threonine 236 directly contacts the C2 and C3 of G (1).
Five out of the six arginines in the basic region are directly hydrogen bonded to the phosphates. Arginine 232, lysine 231 and lysine 246 at the edge of the basic region are close enough to stabilise the complex. The C a of lysine 231 hes in the middle of the major groove (Konig and Richmond 1993).
Arginine 243 is hydrogen bonded to the N7 of the first guanine of the half site as well as to the phosphate between A(-2) and C (-1). In EUenberger's structure the two
arginines cannot bind to the central G C equivalently as they do in that determined by Konig and Richmond (1993). Thus only one arginine hydrogen bonds to the N7 and 0 6 of the guanine, whilst the other arginine binds to the phosphates. The equivalent arginine of R243 in GCN4 of Fos (R155, complex I) and Jun (R279, complex II) contacted the central guanine in the major groove. In GCN4 this is equivalent to R243.
The structures demonstrate the flexibility of the proteins. In complex II of the Fos:Jun:DNA heterodimer the LCC is not perpendicular to the DNA, but is instead bent
10 0 towards it. The LCC in both complexes is asymmetric, the Fos LCC distorting around the straighter Jun LCC.
The DNA.
GCN4 and the FosrJun heterodimer bound both the TRE and CRE in a similar manner.
Although the CRE and TRE sites differ by only a single base-pair (G-C), this shifts the position of the other bases in each of the half sites relative to each other. In B-DNA by a translation of 3.4 Â and a rotation of 34 The phosphates shift by as much as 7 Â. Such a shift would be expected to cause a major distortion of either the DNA or the protein.
In the FosJun crystals the DNA was bent towards the LCC by 10 Konig and Richmond (1993) reported that the DNA in the GCN4:TRE crystals was distorted by as much as 20 The bend was about the centre of the oligonucleotide towards the LCC axis. However, the bending of the DNA was not as great as previously reported for the Jun homodimer or the Fos-Jun heterodimer (Kerppola and Curran, 1991). The bending in that instance was estimated as being as great as 60 This could have been an artifact caused by the angle between the protein and the DNA. Thus a "T" shaped structure, when the protein is bound in the middle of the oligonucleotide, would migrate slower than one where the binding was sufficiently close to the edge of the DNA. Binding at the edge of the DNA would allow the complex to bend and flex as it migrated through the gel matrix. The structures of the two oligonucleotides alone were not published.
The two different binding sites in the experimental work of this thesis shared a similar global and local structure. The values for both % and P published by previous workers compare well with those in this study. Mean % values published for the local structure of oligonucleotides were -120 o for purines and -90 ° for pyrimidines. The backbone angle, 5 was 120 ° for purines and 123 ® for pyrimidines (Clore and Gronenbom, 1984). Clore and Gronenbom also compared data to that published for crystal structures. The mean purine %s were -110 o and -166 ° for B and A DNA respectively. For pyrimidines the angles were -124 o and -155 ® for B and A DNA. These data support the CD of the two DNA oligomers in this work that both are B DNA. % for pyrimidines in the tetradecamer was -114 ° on average, for purines the mean % was -119 These were within the limits from previously published data.
The DNA bound to the GCN4 homodimer in the crystals did not conform to the straight B-DNA of the unbound oligonucleotides in this study. GCN4 bound the TRE in an undistorted form, but the CRE had a different screw displacement and hence base-pair tilt. However the resultant structures placed the half sites in the two different oligonucleotides in similar positions, and Konig and Richmond (1993) estimate that the binding energy of the two forms is very similar. The data reported for the F osJun heterodimer bound to DNA suggest that there was little distortion of the DNA, and that instead the protein flexed to accommodate the extra base pair of the CRE. This suggests that the FosJun heterodimer may have a slightly different method of binding the DNA than the GCN4 homodimer.