The POZ domain is a conserved protein–protein interaction motif present in a variety of transcription factors involved in development, chromatin remodelling and human cancers. Here, we study the role of the POZ domain of the GAGA transcription factor in promoter recognition. Natural target promoters for GAGA typically contain multiple GAGA‐binding elements. Our results show that the POZ domain mediates strong co‐operative binding to multiple sites but inhibits binding to single sites. Protein cross‐linking and gel filtration chromatography experiments established that the POZ domain is required for GAGA oligomerization into higher order complexes. Thus, GAGA oligomerization increases binding specificity by selecting only promoters with multiple sites. Electron microscopy revealed that GAGA binds to multiple sites as a large oligomer and induces bending of the promoter DNA. Our results indicate a novel mode of DNA binding by GAGA, in which a large GAGA complex binds multiple GAGA elements that are spread out over a region of a few hundred base pairs. We suggest a model in which the promoter DNA is wrapped around a GAGA multimer in a conformation that may exclude normal nucleosome formation.
The Drosophila sequence‐specific DNA‐binding protein GAGA is involved in several distinct aspects of chromosome dynamics (Granok et al., 1995; Read and Driscoll, 1997; Wilkins and Lis, 1997). GAGA is encoded by the essential Trithorax‐like (Trl) gene (Farkas et al., 1994) and was first discovered as an activator of the Ubx and en genes and by its association with heat shock and histone gene promoters (Biggin and Tjian, 1988; Soeller et al., 1988; Gilmour et al., 1989). Normal expression of several developmental genes, including homeotic genes, requires GAGA (Farkas et al., 1994; Bhat et al., 1996). Immunofluorescent staining of polytene chromosomes revealed that many euchromatic genes include binding sites for GAGA, suggesting a general role in transcription control (Tsukiyama et al., 1994; Benyajati et al., 1997).
How does GAGA activate transcription? A number of studies have shown that GAGA‐binding elements coincide with DNase I‐hypersensitive sites at promoters, indicative of in vivo chromatin remodelling (Wu, 1980; Costlow and Lis, 1984; Cartwright and Elgin, 1986; Lis and Wu, 1993; Lu et al., 1993; Shopland et al., 1995). Biochemical studies indicate that GAGA activates RNA polymerase II transcription by counteracting chromatin repression (Croston et al., 1991; Tsukiyama et al., 1994; Okada and Hirose, 1998). This is likely to involve the disruption of nucleosomes harbouring GAGA‐binding sites by GAGA acting together with the energy‐dependent chromatin remodelling factor NURF (Tsukiyama et al., 1994; Tsukiyama and Wu, 1995, 1996).
The function of GAGA is not restricted to that of a gene‐specific transcriptional activator. Trl mutations are dominant enhancers of position‐effect variegation, indicating that GAGA counteracts heterochromatic silencing (Farkas et al., 1994). GAGA has also been implicated in the functioning of the polycomb response elements (Strutt et al., 1997). Immunolocalization studies revealed a strong association of GAGA with the GA‐rich centric heterochromatin throughout the cell cycle in early embryos (Raff et al., 1994). More recent studies suggested a mitosis‐specific association of GAGA with GA‐rich satellite DNA (Platero et al., 1998). This observation might be related to a variety of nuclear cleavage cycle defects, displayed by Trl mutants, that include asynchrony and failure in chromosome condensation and segregation (Bhat et al., 1996). Thus, GAGA is a multipurpose protein that mediates gene‐specific regulation but also plays a global role in chromosome function.
A number of distinct cDNAs have been described that encode different GAGA isoforms (Soeller et al., 1993; Benyajati et al., 1997). The major GAGA species is a 519 amino acid polypeptide (Soeller et al., 1993) and this is the form discussed herein. The GAGA protein contains two well‐conserved structural domains: a zinc finger DNA‐binding domain (DBD) located within the C‐terminal half of the protein, and a so‐called BTB/POZ (broad complex tramtrack bric‐a‐brac/poxvirus and zinc finger) domain at the N‐terminus (Figure 1A). Additionally, the C‐terminus of GAGA is glutamine rich, a feature characteristic of a subclass of transcriptional activators. The solution structure of the DBD in complex with DNA recently has been determined (Omichinski et al., 1997). The DBD of GAGA comprises a single classical C2–H2 zinc finger preceded by a basic helix and tail (Pedone et al., 1996; Omichinski et al., 1997). The GAGA DBD binds DNA as a monomer and does not introduce any major changes in the DNA structure. Although there is considerable variability in GAGA sites, optimal binding sequences contain a minimal GAGAG pentanucleotide. DNA contacts are made in the major as well as the minor groove and extend over one turn of the DNA helix.
The BTB/POZ domain, from here on referred to as the POZ domain, is an evolutionarily conserved region of ∼120 amino acids present in a family of metazoan zinc finger transcription factors as well as in a restricted number of actin‐binding and poxvirus proteins (Godt et al., 1993; Bardwell and Treisman, 1994; Zollman et al., 1994). In humans, some POZ domain transcription factors have been implicated in a number of cancers. These include BCL‐6 (Ye et al., 1993; Chen et al., 1998), PLZF (Chen et al., 1993; Dong et al., 1996; X.Li et al., 1997) and APM‐1 (Reuter et al., 1998). The PLZF POZ domain is involved in the recruitment of histone deacetylase co‐repressor complexes (Grignani et al., 1998; Guidez et al., 1998; Lin et al., 1998). The POZ domain provides a protein–protein interaction interface implicated in transcriptional activation as well as repression by POZ transcription factors (Deweindt et al., 1995; Chang et al., 1996; Seyfert et al., 1996; Kaplan and Calame, 1997). POZ domains can associate with each other, most likely via a coiled‐coil type of interaction. Binding is specific and restricted to certain combinations of POZ‐containing proteins (Bardwell and Treisman, 1994; Dong et al., 1996). Previously, it has been shown that the presence of the POZ domain inhibits DNA binding (Bardwell and Treisman, 1994; Kaplan and Calame, 1997). This observation raises the question of how POZ transcription factors bind to their target promoters.
In this study, we address the function of the GAGA POZ domain in promoter recognition. Natural target promoters for GAGA typically contain multiple GAGA‐binding elements. DNase I footprinting and bandshift experiments revealed that GAGA binding to these multiple sites is highly co‐operative and critically depends on the POZ domain. The effects on DNA binding correlate with POZ domain‐mediated multimerization of GAGA as revealed by protein cross‐linking and size exclusion chromatography. We used electron microscopy (EM) to examine the effects of GAGA binding on promoter topology. Based on our DNA binding, protein cross‐linking and EM experiments, we propose a model for a GAGA–promoter DNA complex. The implications of this model for GAGA‐mediated chromatin remodelling are discussed.
The POZ domain inhibits GAGA binding to single sites
To examine the role of the GAGA POZ domain in promoter recognition, we expressed hemagglutinin (HA) epitope‐tagged full‐length GAGA, a deletion mutant lacking the POZ domain (ΔPOZ) and the isolated DBD in Sf9 cells infected with recombinant baculoviruses. These proteins were immunopurified to near homogeneity utilizing the HA epitope (Figure 1). We compared the DNA‐binding activities of about equal molar amounts of these different polypeptides to a consensus GAGA‐binding site (1×GAGA). The bandshift assay revealed that the DBD and ΔPOZ efficiently bind to a canonical single site whereas full‐length GAGA binds only very weakly (Figure 2A).
DNase I footprinting analysis on a DNA fragment harbouring a single GAGA site confirmed that the presence of the POZ domain has an inhibitory effect on DNA binding (Figure 2B). However, DNA binding is not completely blocked since at the highest protein concentration, full‐length GAGA does protect the canonical GAGA site against digestion (Figure 2B, lane 4). The footprinting borders of all three GAGA polypeptides are similar, confirming that the previously identified DNA‐binding domain is responsible for contacting the DNA (Pedone et al., 1996). From these bandshift and DNase I footprinting experiments, we estimated an ∼6‐ to 9‐fold reduction in DNA‐binding affinity due to the presence of the POZ domain.
In addition, we tested the DNA‐binding activity of a number of other GAGA deletion mutants. However, these experiments showed that the inhibition of DNA binding is caused solely by the POZ domain (data not shown). These observations for GAGA agree well with a previous study which demonstated inhibition of DNA binding by the ZID POZ domain (Bardwell and Treisman, 1994). Why would a transcription factor contain a domain that inhibits DNA binding? To address this issue further, we studied the binding of GAGA to several of its natural target promoters.
The POZ domain mediates co‐operative binding of GAGA to natural target promoters
Promoters regulated by GAGA have been identified by in vivo as well as in vitro studies. We have chosen the previously well characterized Ubx, ftz, hsp70 and eve promoters to compare the binding of GAGA polypeptides. We note that all these promoters are characterized by the presence of multiple GAGA‐binding sites (Biggin and Tjian, 1988; Read et al., 1990; Topol et al., 1991; Weber et al., 1997). DNase I footprinting experiments revealed a dramatic difference in DNA‐binding properties between full‐length GAGA and the polypeptides lacking the POZ domain (Figure 3). The GAGA elements on the natural promoters are bound efficiently by full‐length GAGA but not by equal molar amounts of either ΔPOZ or DBD. The amount of GAGA required to bind the multiple promoter elements is significantly lower (>4‐ to 12‐fold, depending on the promoter) than that required to bind a single site, indicative of co‐operative DNA binding. The spacing of the GAGA elements in these different promoters varies considerably (Figure 3B). However, GAGA appears to be quite flexible and able to bind co‐operatively to GAGA sites located at variable distances from each other. The hsp70 promoter is generally GA rich and, at increasing GAGA concentrations, the footprints start to spread and most of the promoter DNA is protected against digestion, as has been observed previously (Weber et al., 1997).
In contrast to full‐length GAGA, equal molar amounts of the ΔPOZ or DBD polypeptides fail to bind the GAGA target promoters significantly. On the Ubx, ftz and eve promoters, protection of a single GAGA site by ΔPOZ and DBD can be observed. As expected, these sites are the ones that most closely resemble the optimal GAGA‐binding sequence. In these experiments, ΔPOZ and DBD fail to bind to the weaker GAGA sites. This indicates that POZ‐mediated co‐operativity increases the binding affinity for these sites by at least one order of magnitude. Together, these DNase I footprinting experiments demonstrate that efficient binding of GAGA to its natural target promoters depends critically on the presence of the POZ domain, in addition to the DBD.
The promoters studied above contain weak as well as strong GAGA sites. To characterize the co‐operative nature of GAGA–DNA interactions further, we synthesized a 64 bp oligonucleotide (5×GAGA) containing five optimal GAGA sites. DNase I footprinting analysis of GAGA binding to these artificial sites demonstrates strong co‐operative binding that is dependent on the POZ domain (Figure 4A). In spite of the optimized binding sites in the 5×GAGA probe, we estimate that the binding affinity of the ΔPOZ and DBD is >9‐fold lower than that of full‐length GAGA (Figure 4A, compare lanes 2, 9 and 14).
The close spacing of the GAGA‐binding sites allowed us to use this fragment in bandshift assays. Labelled 5×GAGA oligonucleotide was incubated with increasing amounts of either full‐length GAGA, ΔPOZ or DBD and analysed by native PAGE (Figure 4B). Again, the DNA‐binding properties of GAGA and the POZ domain‐lacking polypeptides are strikingly different. At higher concentrations of ΔPOZ or DBD, distinct slower migrating complexes gradually appear which correspond to increased occupancy of binding sites.
In contrast, in reactions containing GAGA, the transition from mostly free DNA to complete occupancy is quite sharp. When bound, the vast majority of probe immediately shifts to a complex of very slow mobility, indicative of binding by multiple GAGA molecules. Together, the footprinting and the bandshift results indicate that GAGA binds multiple sites simultaneously (Figure 4A and B). It should also be noted that the amount of GAGA that produces a full shift with the 5×GAGA probe fails to bind the 1×GAGA probe significantly (compare Figure 2A, lane 2 and Figure 4B, lane 4). These results, using optimal binding sites, again demonstrate that GAGA binds DNA co‐operatively.
Since the binding sites within the 5×GAGA probe are about equivalent, a shift in mobility can result from binding to any of these sites. Although the majority of DNA molecules are bound at the higher concentrations of ΔPOZ or DBD polypeptides, two observations indicate that a full site occupancy has not been reached in these reactions. First, not all the probe DNA is in the slowest mobility complexes, and secondly, there is no clear DNase I protection. Thus, deletion of the POZ domain results in less efficient binding to multiple binding elements even if these sites are optimal. It is pertinent to note that on natural promoters the majority of binding sites are suboptimal (Figure 3). GAGA binding to these physiological sites is even more dependent on synergism.
In conclusion, although GAGA contacts the DNA via the DBD, efficient and co‐operative promoter binding depends critically on the presence of the POZ domain. How does the POZ domain mediate synergistic DNA binding?
The POZ domain mediates GAGA oligomerization
Since it seemed likely that co‐operative DNA binding of GAGA involves protein–protein interactions, we next investigated the oligomeric status of GAGA. Pure GAGA or ΔPOZ polypeptides were treated with the covalent cross‐linkers glutaraldehyde (GA) or, the more stringent, ethylene glycol bis(succinimidylsuccinate) (EGS) and resolved by SDS–PAGE (Figure 5A). In both GA‐ and EGS‐treated samples, higher order oligomeric species were readily detected for GAGA whereas the ΔPOZ polypetide did not form oligomers. Likewise, the DBD does not form higher order complexes (data not shown). The mobilities of the main cross‐linked GAGA complexes suggest that they may correspond to dimers, trimers, tetramers and hexamers. Addition of a DNA fragment harbouring multiple GAGA sites did not alter the oligomeric status of GAGA, suggesting that protein–protein interactions occur in solution and do not require DNA binding.
To determine the apparent native molecular weight of GAGA, we performed size exclusion chromatography with pure recombinant GAGA (Figure 5B). The elution profile of GAGA, as detected by protein immunoblotting, is surprisingly broad. GAGA migrates with an apparent molecular mass between ∼60 and 600 kDa. This result agrees well with the cross‐linking studies and suggests that GAGA is present in various oligomeric complexes ranging from monomers to very large multimers. GAGA does not seem to aggregate in an uncontrolled fashion since no GAGA eluted in the void of the column and the elution profile has distinctive borders. Together, these results suggest that POZ domain‐mediated oligomerization of GAGA underlies co‐operative DNA binding to multiple binding sites.
Consequences of GAGA binding for the trajectory of the Ubx promoter DNA
We next considered the consequences of GAGA oligomer binding to a natural promoter harbouring multiple binding elements. In particular, we were interested in the possibility that promoter binding by the GAGA complex might alter the DNA trajectory. To visualize the GAGA–promoter complex, we have employed EM. For these experiments, we have used the Ubx promoter that contains four GAGA‐binding sites between positions −193 and −30 relative to the transcription start site (Biggin and Tjian, 1988) (Figure 3B). We used a 4 kbp DNA fragment containing a centrally located 1 kbp Ubx promoter sequence, with the 163 bp region containing the four GAGA sites 2.3 and 1.5 kbp from the DNA ends. This DNA fragment was incubated with either GAGA or ΔPOZ purified to near homogeneity. Electron micrographs of GAGA–DNA complexes reveal a roughly spherical protein mass located centrally on the DNA fragment (Figure 6C and D). This protein mass has a diameter of ∼150–300 Å, consistent with the presence of a large protein complex. In contrast, DNA fragments incubated with ΔPOZ were indistinguishable from naked DNA (Figure 6A and B). This observation again highlights the critical role of the POZ domain in oligomerization and promoter recognition. The specificity of GAGA binding was established further by the generation of DNA fragments in which the Ubx promoter sequence is positioned near the end. As expected, on these fragments, the protein mass associates with the end of the fragments (data not shown). These EM results reinforce the notion that GAGA binds multiple sites, within ∼200 bp of DNA, as a single large oligomeric complex.
In order to analyse possible changes in the DNA trajectory in more detail, we examined GAGA in complex with shorter DNA fragments (Figure 6E–H). In these 1055 bp fragments, the Ubx promoter is located slightly off centre, with the 163 bp region containing the four GAGA sites 466 and 426 bp from the DNA ends. Visualization clearly indicated that in the majority of complexes the promoter DNA is bent. The observed centre of DNA bending is located close to the centre of the protein mass, and we estimate a mean bend angle of ∼80–90°. Interestingly, in ∼19% of the complexes we observed, the GAGA oligomer binds two DNA molecules simultaneously (Figure 6F). Our biochemical and EM experiments suggest that GAGA can indeed form sufficiently large multimers to occupy the GAGA sites present in these two DNA fragments. These results point to the possibility that GAGA acts to bring distant DNA elements together.
The POZ domain is a structural motif present in a large number of zinc finger transcription factors involved in developmental regulation and cancer. Here, we have addressed the role of the GAGA POZ domain in promoter recognition. Natural promoters regulated by GAGA are characterized by the presence of multiple GAGA sites, which are typically spread over a region of a few hundred base pairs. Moreover, in vivo experiments have indicated that GAGA requires multiple binding sites to stimulate transcription efficiently (Soeller et al., 1993; Granok et al., 1995). We have used bandshift and DNase I footprinting experiments to establish that GAGA binding to natural target promoters is highly co‐operative. Synergistic DNA binding depends critically on the presence of the POZ domain. The POZ domain provides a protein–protein interaction interface that mediates GAGA oligomerization into higher order complexes as revealed by protein cross‐linking and gel filtration chromatography experiments. The EM experiments directly show that GAGA binds multiple DNA elements as a large oligomer. Taken together, our results suggest that promoter recognition by GAGA requires two distinct, well separated protein domains, the zinc finger DBD that contacts the DNA and the POZ domain that directs GAGA oligomerization.
A model for promoter binding by GAGA factor
What are the consequences of the binding of a GAGA oligomer for the topology of the promoter DNA? The association of a GAGA oligomer with multiple binding sites within a promoter is expected to constrain the DNA trajectory. Indeed, electron micrographs of the GAGA‐bound Ubx promoter indicate that the DNA in the complex is bent and might be wrapped around the surface of a GAGA multimer. Since the solution structure of a GAGA DBD–DNA complex did not reveal a major distortion of the DNA (Omichinski et al., 1997), we do not think that the DNA bending is caused directly by the DBD–DNA contacts. Rather, we favour the idea that DNA bending is the consequence of the DNA wrapping around the GAGA complex.
The reorganization of the Ubx promoter due to binding of the GAGA complex is depicted in a schematic model in Figure 7. To illustrate that the GAGA sites are spread over a distance of about one nucleosome repeat length (of ∼200 bp), we have included a cartoon of the DNA wrapped around a nucleosome. It should be noted that within a nucleosome, most of the DNA is in close contact with the histones. In contrast, GAGA only contacts the GAGA elements, resulting in a more accessible DNA–protein complex. This model also provides some clues as to why the binding of GAGA to the promoter involves changes in the local chromatin structure. Since the GAGA DBD clamps almost one turn of the DNA (Omichinski et al., 1997), GAGA binding to multiple sites within a nucleosome repeat length is expected to severely compromise histone–DNA contacts. These contacts might be hampered further by DNA bending and wrapping around a GAGA oligomer. However, it is not clear whether GAGA binding leads to complete displacement of the histone core or whether some histone–DNA contacts are preserved (Tsukiyama et al., 1994; Tsukiyama and Wu, 1995). In summary, after transient chromatin remodelling by NURF to allow for GAGA binding, GAGA may function as an architectural factor that reorganizes the promoter DNA and maintains it in an open conformation.
Within many GAGA target promoters, binding sites for distinct transcription factors are located between the GAGA elements. An attractive posibility is that after binding of the GAGA complex, these sites become accessible to other factors which are unable to overcome the nucleosome barrier by themselves. Indeed, there is in vivo evidence that such a mechanism occurs at the hsp70 promoter where GAGA binding facilitates subsequent binding by the heat shock factor (Lis and Wu, 1993; Shopland et al., 1995). POZ domain‐mediated complex formation might also be important for the association of GAGA with the highly repetitive DNA sequences that form the bulk of Drosophila melanogaster heterochromatin (Raff et al., 1994; Platero et al., 1998).
The POZ domain and DNA binding
The results described here show that the GAGA POZ domain plays a critical role in promoter selectivity. The functioning of the POZ domain differs in several aspects from that of other protein–protein interaction domains, such as Leu zipper or helix–loop–helix motifs, involved in DNA binding. The latter motifs are integral parts of the DBD and mediate the binding of protein dimers to dimeric DNA sites (McKnight and Yamamoto, 1992). In contrast, multiple, well separated GAGA elements are each contacted by a single GAGA molecule that is part of an oligomer. This distinction is also reflected in the protein structure of POZ transcription factors. Within these proteins, the POZ domain is well separated from the zinc finger DBD.
Zeste, a member of the Trithorax group of proteins, binds DNA in a manner that is reminiscent of that of GAGA. Althought lacking a POZ domain, it forms oligomers that bind co‐operatively to multiple sites (Chen and Pirrotta, 1993a, b). Thus, synergistic DNA binding by higher order complexes might be a more common mechanism of DNA binding.
Why does the POZ domain hinder single site recognition by GAGA? Two, not mutually exclusive models, might explain this inhibition. First, the POZ domain may fold back and block DBD–DNA contacts. Secondly, inhibition might be a consequence of multimerization. In the absence of multiple binding sites, GAGA molecules within a complex may inhibit binding to a single site due to steric hindrance. In other words, the GAGA complex may not fit on the DNA if only one binding site is available. Since the presence of the POZ domain can also inhibit DNA binding of heterologous DBDs (Bardswell and Treisman, 1994), the notion of it specifically contacting the DBD and hampering DNA contacts seems less likely. Alternatively, as a consequence of multimerization, the DBD might be less accessible within the GAGA complex. Multiple binding elements may be required to ‘unwrap’ the GAGA oligomer to allow for efficient DNA binding.
Another potential role for GAGA oligomerization might be to assist communication between distal enhancers and promoters. In our EM experiments, we frequently observed that a GAGA oligomer binds two DNA molecules simultaneously. This suggests that POZ–POZ interactions may lead to the association of enhancer‐ or PREs‐bound GAGA with promoter‐bound GAGA. Future experiments will address the potential role of GAGA in DNA looping and long‐distance effects within a chromosomal context.
We propose that the POZ domains in other POZ transcription factors perform similar functions to that which it has in GAGA, and mediate co‐operative DNA binding to multiple sites but inhibit single site recognition. It should be noted that POZ–POZ interactions might result in the formation of heteromeric complexes containing distinct POZ transcription factors. Several studies have demonstrated the functional importance of the POZ domain in transcriptional repression as well as activation (Deweindt et al., 1995; Seyfert et al., 1996; Dhordain et al., 1997; Kaplan and Calame, 1997). Part of such an involvement in opposite functions may simply reflect the need for the POZ domain in order to direct the POZ transcription factors to target promoters.
Deregulated expression of BCL‐6, a POZ domain transcriptional repressor, occurs in the majority of diffuse large cell lymphomas cases and a significant portion of follicular lymphomas (Ye et al., 1993; Chang et al., 1996; Chen et al., 1998). As we have shown here, due to co‐operative DNA binding mediated by the POZ domain, small changes in protein concentration can lead to dramatic differences in promoter occupancy. Likewise, it is tempting to speculate that the effects of misregulation of BCL‐6 might be amplified by synergistic DNA binding, causing changes in gene expression that may promote leukaemic transformations. Translocation of the POZ domaincontaining N‐terminal half of the zinc finger protein PLZF to the retinoic acid receptor α (RARα) is associated with a subset of acute promyelocytic leukaemias. Further studies have indicated that the PLZF POZ domain is responsible for the aberrant activities of the fusion protein (Chen et al., 1993, 1994; Dong et al., 1996; J.Y.Li et al., 1997; Ruthardt et al., 1997). The PLZF POZ domain is involved in the recruitment of histone deacetylase co‐repressor complexes (Grignani et al., 1998; Guidez et al., 1998; Lin et al., 1998). This function explains at least part of the dominant‐negative effects of the POZ domain in the PLZF–RARα fusion protein. Additionally, the presence of the POZ domain is likely to interfere with the binding to normal RARα target promoters and may divert the fusion protein to other genes, resulting in an aberrant pattern of gene expression.
In summary, our findings reveal a novel role for the POZ domain in mediating co‐operative DNA binding. The POZ domain is required for the formation of higher order GAGA oligomers. The GAGA complex binds target promoters by contacting multiple binding sites that are spread out over a region of a few hundred base pairs and induces bending of the promoter DNA. These results may provide a paradigm for DNA recognition by other POZ transcription factors.
Materials and methods
Expression and purification of recombinant proteins
An NdeI site was created at the initiating methionine of GAGA using a PCR‐based strategy. The complete GAGA‐coding sequence was cloned into a modified version of the pVL1392 (PharMingen) baculovirus expression vector, pVL1392HAX, containing an HA tag (YPYDVPDYA)‐encoding sequence followed by the amino acids IEGRHM (factor X cleavage site). Using a similar strategy, DNA fragments encoding the ΔPOZ mutant (amino acids 120–519) and the DBD mutant (amino acids 321–379) were cloned into pVL1392HAX introducing an in‐frame N‐terminal HA tag. All GAGA sequences derived from the pARGAGA construct (Soeller et al., 1993). The integrity of all constructs was verified by DNA sequencing of both strands. pVLHAX‐GAGA, pVLHAX‐ΔN120 (ΔPOZ) and pVLHAX‐Δ(N321/C379) (DBD) were co‐transfected with BaculoGold viral DNA (PharMingen) into Sf9 cells. All recombinant baculoviruses were plaque purified and amplified. Expression in Sf9 cells, extract preparation and immunopurification were all performed essentially as described previously (Chen and Tjian, 1996). Briefly, Sf9 cells were infected at an m.o.i. of ∼5 and harvested 48 h post‐infection. All protein procedures were carried out at 4°C using HEMG buffer (25 mM HEPES–KOH pH 7.6, 0.1 mM EDTA, 12.5 mM MgCl2, 10% glycerol) containing 1 mM dithiothreitol (DTT), 0.2 mM AEBSF, 1 μM pepstatin and varying amounts of KCl. Whole‐cell extracts were prepared by sonication in 0.4 M KCl‐HEMG containing 0.1% NP‐40. After centrifugation at 10 000 g, the HA‐tagged proteins were immunopurified from the supernatant using protein A–Sepharose beads (Pharmacia) covalently conjugated with anti‐HA (12CA5) monoclonal antibodies (Harlow and Lane, 1988). After extensive washes, bound proteins were eluted using a peptide corresponding to the HA epitope in a buffer consisting of HEMG containing 100 mM KCl, 0.01% NP‐40 and 0.2 mg/ml HA peptide (YPYDVPDYA). The protein concentrations of the final preparations were determined by Bradford assay (Bio‐Rad) and Coomassie staining of protein gels using bovine serum albumin (BSA) as a standard.
Protein cross‐linking and gel filtration analysis
Protein cross‐linking reactions were carried out in a 20 μl volume of HEMG containing 50 mM NaCl, 0.05% NP‐40, 1 mM DTT and purified GAGA or ΔPOZ to a concentration of ∼0.05 μM. Reactions were started with the addition of GA (Sigma, G‐5882) to a final concentration of 0.01% or EGS (Sigma, E‐3257) to a final concentration of 10 mM. Reactions were terminated after 4 or 8 min (EGS) or after 8 min (GA) by the addition of 6 μl of 2× SDS sample buffer (62.5 mM Tris–HCl pH 6.8, 2% SDS). After quenching of the reactions, samples were analysed by SDS–PAGE on 8 or 5% gels. Proteins were transferred to nitrocellulose, probed with the α‐HA mouse monoclonal antibody and visualized using the alkaline phophatase staining reaction following standard procedures (Harlow and Lane, 1988). Gel filtration analysis was performed on a Pharmacia HiPrep 16/60 S‐300 Sephacryl column equilibrated and developed with HEMG buffer containing 0.1 M NaCl, 0.01% NP‐40 and 1.5 mM DTT on a Biologic HR system (Bio‐Rad). The column was calibrated with native protein standards according to instructions provided by the suppliers (Pharmacia). About 1–1.5 μg of purified recombinant GAGA was applied to the column; 1 ml fractions were collected throughout the runs and analysed for the presence of protein by SDS–PAGE followed by Western immunoblotting with the α‐HA monoclonal antibody.
DNA binding assays
The following probes were used in DNA binding experiments: for bandshift assays, double‐stranded oligonucleotides harbouring a single GAGA site (1×GAGA: 5′‐AAGAGAGAGCGCAAGAGC‐3′) or five artificial GAGA sites (5× GAGA: 5′‐GAGAGAGCAAAGGCCTCTCGTTCATTGCTCTCTVGTTCTAGAAACAGAGAGCTTAGCTCTCTC‐3′) were used. These double‐stranded oligonucleotides contained 5′ overhangs compatible with an EcoRI and BamHI site, respectively, and were cloned into the EcoRI–BamHI sites of pBluescript KS. For DNase I footprinting experiments, Asp718–SacII fragments of each of the above constructs were labelled on the top strand using the T4 polynucleotide kinase. The following probes were used for DNase I footprinting on the natural promoters: Ubx, a NruI–MluI fragment corresponding to promoter sequences −203 to +115 relative to the transcription start site of the Ubx promoter (Biggin and Tjian, 1988), derived from pAK100 (containing a 1 kbp Ubx EcoRI–HindIII fragment comprising the region −600 to +400; A.Kal, unpublished data); ftz, an EcoRI–BamHI fragment corresponding to promoter sequences −445 to +40 of the ftz promoter (Topol et al., 1991) derived from pBlue‐ftz (containing ftz promoter sequences −445 to +40 flanked by EcoRI and BamHI restriction sites introduced by using a PCR‐based strategy); hsp70, an EcoRI–HindIII fragment corresponding to promoter sequences −190 to +84 of the hsp70 promoter derived from hsp70‐190XBS (Weber et al., 1997); eve, an EcoRI–BamHI fragment corresponding to promoter sequences −449 to +135 of the eve promoter (Read et al., 1990) derived from pBlue‐eve (containing eve promoter sequences −449 to +135 flanked by EcoRI and BamHI restriction sites introduced using a PCR‐based strategy). The eve and ftz fragments in the pBlue‐eve and pBlue‐ftz plasmids were generated by PCR from the plasmids p48‐4.7 (G.Struhl, unpublished data) and pFZPacP8 (H.Francis‐Lang, unpublished data) respectively. All promoter fragments were end‐labelled on the bottom strand using T4 polynucleotide kinase. Probes were prepared and purified using standard procedures (Sambrook et al., 1989). Bandshift and DNase I footprinting experiments were performed essentially as described previously (Verrijzer et al., 1995). Briefly, for bandshift assays, double‐stranded oligonucleotides were end‐labelled with T4 polynucleotide kinase. Binding reactions were carried out in a reaction volume of 20 μl of 0.5× HEMG buffer containing 50 mM NaCl, 50 μg/ml BSA, 0.05% NP‐40, ∼60 fmol of double‐stranded labelled probe, 75 ng of poly(dIdC)–poly(dIdC) and the indicated amounts of polypeptide. All binding reactions were carried out for 30 min on ice and were analysed on 5% polyacrylamide gels run in 0.5× Tris–glycine/0.01% NP‐40 buffer at room temperature. For DNase I footprinting reactions, ∼50 ng of probe was used with the appropriate polypeptide in 60 μl reactions in 0.5× HEMG buffer containing 50 mM NaCl, 2% polyvinyl alcohol, 0.05% NP‐40, 30 μg/ml BSA and 50 ng of poly(dIdC)–poly(dIdC). After binding for 45 min at room temperature, samples were processed and analysed on a 7% sequencing gel.
DNA fragments used in EM analysis were derived from plasmid pAK100 generated by cloning a 1 kbp EcoRI–HindIII fragment from the Ubx gene into pBluescript KS. The 4 kbp fragment was generated by linearizing pAK100 with SacI digestion. The 163 bp promoter region containing the four GAGA sites (−193 to −30 relative to the transcription start site) is 2.3 and 1.5 kbp from the respective DNA ends. A 1055 bp DNA fragment encompassing the Ubx promoter was generated by digestion of pAK100 with HindIII and SacI. The 163 bp promoter region containing the four GAGA sites is 466 and 426 bp from the DNA ends. DNA was purified by Qiagen column chromatography according to instructions provided by the supplier. All DNA binding reactions were performed as described above. Immediately after binding, protein–DNA complexes were fixed in 0.2% GA for 20 min at room temperature. Next, a small (4 μl) drop of sample material was deposited upon a carbon‐filmed nickel grid made hydrophobic by high‐voltage glow discharge. After 30 s absorption followed by 3–4 washes with 2–5 mM magnesium acetate, the samples were positively stained with 2% aqueous uranyl acetate as described (Spiess et al., 1987). Electron micrographs were obtained with a Jeol 1200WEXII electron microscope at 50000× magnification.
We thank T.Kornberg, M.Biggin, D.Gilmour and D.Ish‐Horowicz for the gift of plasmids, R.Treisman, D.Ish‐Horowicz and P.Badenhorst for helpful discussions, A.Kal for advice on the sample preparation for EM, G.Chalkley for advice on column chromatography, and C.Hill, D.Ish‐Horowicz, M.Parker, R.Treisman, B.Linder, T.Mahmoudi, S.Vincent and A.Kal for critical comments on the manuscript. This work was supported by the Imperial Cancer Research Fund.
- Copyright © 1999 European Molecular Biology Organization