ATF-1 and CREB show homology of 70% at the prim ary sequence level, in the region downstream of the PKA phosphorylation site, where the two proteins share an identical domain structure. ATF-1 has a truncated N-terminal Q-rich domain and lacks an equivalent of the a-peptide. The Q-domain nomenclature is as in Brindle et al., 1993.
Seri 33
01 a k id 0 2 bZIP
A T F -1
Ser63
C REB complex (Hurst at a!., 1991). Thus, although this property of ATF- 1 is not
accounted for in studies of GALATF-1 fusions (Liu et al., 1993) it may be of relevance when considering how the PKA response is attenuated under more natural conditions.
Although the reasons given above explain why CREB-ATF-1 dimers should be less active than CREB-CREB in mediating a PKA response, this fails to explain why the CRE should show no activity at all in UF9 cells. Either additional mechanisms operate to tighten the repression which is achieved by A TF-1, or indeed ATF-1 itself has unsuspected properties which manifest them selves uniquely in the heterodimeric form. In support of the latter possibility, transfection studies with ATF-1 and co-transfected CREB suggest that the heterodimer is less active than the ATF-1 homodimer (Ellis et al.,
1994).
3.6.ii GALCREB fusion proteins as a tool for studying
properties of CREB
It seems that the study of CREB activity by m eans of G ALCREB fusion proteins is less simple than might be expected from the observed similarity in the behaviour of these proteins in cAMP non-responsive cell lines.
The data shown in figure 3.2 are difficult to interpret given the information in figure 3.6, which indicates that GALCREB may heterodimerize extensively with cellular factors in a manner predicted to interfere with its binding to a G A L UAS. It is likely that the data generated by expressing GALCREB in any cell line is in part a measure of the ratio of transfected and endogenous factors produced in the experiment. For instance, UF9 cells may display the 'inhibitory' phenotype in this experiment because this cell line has a poor transfection efficiency. The levels of G A LC R E B produced are consequently low and all of the GALCREB present may be in the form of heterodimers with endogenous CREB or ATF1 (expressed at high levels in these cells). Upon mutation of the leucine zipper, G ALCREB can no longer heterodimerize in this way, and exists as a homodimer which is able to activate transcription of the reporter gene through a G A L UAS. D F9 cells, being transfected at higher efficiency, contain higher levels of the G A LC R EB homodimer such that a significant proportion of the wild-type fusion is able to homodimerize and to exhibit transcriptional activity, which is not increased by mutating the zipper.
However, this hypothetical interpretation of the results presented in section 3.1 does not consider the elevation of GALCREB wild-type activity over that of GALCREB mutants in DF9 cells, which clearly suggests that there is recruitment of positively acting cellular factors by wild-type G A LC R E B
homodimers occurring in these cells. It is thus reasonable to expect that recruitment of negatively acting factors might be responsible for some of the inactivity of GALCREB seen in cell lines where such factors are known to occur. In addition, such explanations fail to account for the phenotype shown by C 0 S 7 cells. The use of the replicating pCOTG vector to achieve very high levels of transfected GALCREB in these cells should give rise to a non- inhibitory phenotype, if titration of cellular factors by G A LC R EB w ere an important contributor to the overall phenotype, but the C O S 7 phenotype in this experiment is identical to that of UF9 cells.
T h e re is no single explanation app aren t to us which is comprehensive enough to cover all of the observed behaviour of GALCREB fusion proteins. Whilst it is certain that CREB does interact with several other factors in the cell lines used for these experiments (see Chapter 4), and possible that one or more of these could have a negative effect on DNA binding by CREB, the data described in this chapter were not able to provide conclusive evidence for the existence of such a factor.
The difficulties in interpretation of data gathered by use of GALCREB fusion proteins prompted us to find an alternative method for investigating the interactions of CREB with ATF-1 and other factors present in F9 cells: This work is described in Chapter 4.
Chapter 4 Expression of recombinant
CREB and its use in probing protein-
protein interactions
CREB is ubiquitous and yet functionally distinct in different cell-types. Th e activity of CREB may be controlled to some extent at the level of expression and in a few tissues by the generation of alternative splice products (as discussed in Chapter 1). However most control of C R EB behaviour is exerted at a post-translational level. S e ri 33 phosphorylation by PKA is necessary and sufficient to achieve activated transcription through a CRE, and there are several other sites for phosphorylation which may support alternative states of activity by CREB or which may facilitate the critical phosphorylation event at Ser133. Additional types of post-translational control are postulated to explain both the inability of CREB to activate transcription at all in certain cell lines, and its ability to activate tissue-specific transcription in others.
The maintenance and regulation of differentiated states commonly relies on the presence (or absence) of negatively- and positively-acting factors to repress unwanted functions and promote those characteristic of that tissue, and on the use of combinatorial activation of transcription, where ubiquitous factors can be harnessed to act on certain promoters only in concert with tissue-specific factors. Both of these mechanisms would appear to operate as additional levels of control on CREB, which is constrained by both negative and combinatorial regulation. In light of the results discussed in Chapter 3, and of the em ergence of protein-protein interaction as a common them e for modulation of transcription factor activity, we were interested in the possibility that CREB might be regulated in this way. At the time this project was started C REB was known to interact with homologous members of the A TF/C R EB family, ATF-1 and C REM , through leucine zipper dimerization. W e hoped to show that C REB could interact more extensively than had previously been shown. In particular we were interested in the significance of the interaction of CREB with ATF-1 in F9 cells, and the possibility that CREB might interact with a novel factor present in these cells.
Our chosen method for investigating these suggested interactions was the 'Far Western' blot, first described by (Hoeffler et al., 1991) and shortly thereafter used by (Blackwood and Eisenman, 1991) to clone Max as a Myc- interacting factor. This procedure involves expressing and labelling recombinant protein for use as a probe in Western-style hybridizations. This chapter will describe the strategies which were employed to generate CREB
suitable for this purpose, together with some of the Far W estern results obtained during the course of these studies, discussing the significance of the interactions shown. For ease of reference, the results are tabulated and presented at the end of the chapter (fig4.7).
4.1 Recombinant CREB expressed in bacteria
4.1.1 GSTCREB radiolabelled by iodination
A series of plasmid vectors directing the expression of cloned sequences as fusions to the C-terminus of a Glutathione-S-Transferase (GST) from Schistosoma japonicum (pGstEXpression, or pGEX vectors), (Smith and Johnson, 1988) have become widely used in recent years. The 26 kD G ST moiety contains a glutathione binding pocket which shows high-affinity binding to glutathione (GSH) immobilized on an agarose matrix; purified fusion protein can subsequently be removed by competition-elution with excess reduced glutathione. This property is conveniently retained by the G S T when expressed in E.ooli, thus giving rise to a ready system for single-step purification of bacterially expressed fusion proteins, under non-denaturing conditions.
pG E X -K G C R E B (197-341) (fig4.1a) was made as described in the figure legend and codes for G ST fused to Gly197-Asp341 of aC R E B . These CREB sequences contain all of the information required for dimerization and DNA-binding of CREB (Yun et al., 1990). Since we were primarily interested in leucine zipper-mediated interactions, we chose this small fragment to provide the added benefits of higher yield and solubility of the resulting fusion protein. Expression of the coding sequences in pGEX-KG is under the control of the IPTG-inducible tac promoter, negative control of this promoter by the la d repressor allele being alleviated in the presence of an excess of IPTG. Figure 4.1b shows a 4 hour time-course for induction of G STCREB in a late log phase culture of E.ooli strain SCS-1. The higher yield of G S T C R E B obtained with shorter time of induction indicates that expression of the induced protein is not well tolerated in SCS- 1 culture. Bacterial lysates were prepared as in Chapter 2 and treated with glutathione-beads as described by (Smith and Johnson,
1988). As can be seen from the left-hand panel, the recommended procedure for elution of purified GST-fusion protein from the beads (washing in 5mM reduced GSH, 50mM Tris-HCI pH8.0), which was effective in eluting 30-40% of bead-bound GST, was not stringent enough to remove any G STCREB. Boiling the beads directly into loading buffer allowed detection of the purified G STCR EB , in addition to a 30 kD band of unknown origin. Recovery of native G S T C R E B from its bead bound state was not achieved by increasing the concentration of GSH in the elution buffer, nor by elution in 8M urea followed