Intact RNA-binding Domains Are Necessary for Structure-specific DNA Binding and Transcription Control by CBTF 122 during Xenopus Development*

CBTF 122 is a subunit of the Xenopus CCAAT box transcription factor complex and a member of a family of double-stranded RNA-binding proteins that function in both transcriptional and post-transcriptional control. Here we identify a region of CBTF 122 containing the double-stranded RNA-binding domains that is capable of binding either RNA or DNA. We show that these domains bind A-form DNA in preference to B-form DNA and that the (cid:1) 59 to (cid:1) 31 region of the GATA-2 promoter (an in vivo target of CCAAT box transcription factor) adopts a partial A-form structure. Mutations in the RNA-binding domains that inhibit RNA binding also affect DNA binding in vitro . In addition, these mutations alter the ability of CBTF 122 fusions with engrailed transcription repressor and VP16 transcription activator domains to regulate transcription of the GATA-2 gene in vivo . These data support the hypothesis that the double-stranded RNA-binding domains of this family of proteins are important for their DNA binding both in vitro and in vivo .

The 122-kDa subunit of the Xenopus CCAAT box transcription factor CBTF 122 (also known as 4F and ubp4) belongs to a family of double-stranded RNA-binding domain (dsRBD) 1 containing proteins implicated in a variety of cellular processes. CBTF 122 and its splice variant CBTF 98 have been identified as subunits of CBTF, a transcription factor that activates the GATA-2 gene both in oocytes and zygotically at the start of gastrulation (1,2). Homologues of CBTF 122 have been identified mainly through their affinity for RNA and include NF90 (also known as DRBP76 and NFAR1), NF110, TCP80, ILF3 and MPP4 in humans (3), SPNR and ILF3 in mice (4), and p74 and ILF3 in rats (5). It is now clear that many of the human homologues arise from alternative splicing of transcripts from the ILF3 locus (6,7).
The functions assigned to these related proteins include the regulation of both transcription and mRNA stability. For example, the transcript levels of more than 90 genes, most of which are known to be regulated by type I interferons, were altered by expression of NF90, suggesting that NF90 has a role in mediating the antiviral response (8). However, it is unclear whether these changes are a result of transcriptional or posttranscriptional mechanisms. By contrast, specific regulation of interleukin-2 levels by NF90 is clearly dependent upon its stabilization of interleukin-2 mRNA (9). In transcriptional terms, the human homologues of CBTF 98 and CBTF 122 , NF90 and NF110, have been shown to be capable of both transcription activation and repression depending upon at which promoter they are acting (3,10). Deletion analysis showed that transcription activation required the nuclear localization signal and two dsRBDs (3). Mutations within the dsRBDs that abrogated RNA binding also inhibited transcription activation (10).
During Xenopus development, CBTF 122 has been shown to be critical for activation of GATA-2 transcription both by in vitro (2) and in vivo methods. 2 GATA-2 is a transcription factor necessary for correct hematopoietic, neural, and urogenital development in mice (11,12) and is also implicated in the formation of ventral mesoderm in Xenopus (13). Although CBTF is maternal, it does not activate GATA-2 transcription in embryos until the start of gastrulation and this is, at least in part, due to regulation of its nuclear translocation. CBTF is present in the oocyte nucleus where GATA-2 is initially transcribed. However, after oocyte maturation, CBTF activity is cytoplasmic and relocates to the nucleus only at the gastrula stage when zygotic GATA-2 transcription commences (1). CBTF 122 is anchored in the cytoplasm by RNA binding (14), and its movement from the cytoplasm to the nucleus is synchronous with the mass degradation of maternal RNA at the mid-blastula transition, suggesting a mechanism for the developmental stage-specific control of CBTF activity based on RNA binding.
Analysis of the amino acid sequence of CBTF 122 indicates a number of conserved domains (see Fig. 1a). The N-terminal domain is predicted to contain ZFR type zinc fingers (15). CBTF 122 also contains two dsRNA-binding domains (dsRBDs) of the type found in Staufen and Xlrbpa (16,17). C-terminal to this is an arginine-glycine-rich region (RGG domain), similar to those found in hnRNP A and U1, which are implicated in nucleic acid binding (18,19). There is also a poly-Q region that has been suggested to be involved in transcriptional activation * This work was supported by a Biotechnology and Biological Sciences Research Council studentship (to S. J. E.) and a Wellcome Trust project grant. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
in the context of other proteins (20). Finally, there is a nuclear localization signal located N-terminal to the dsRBDs (14,21) with a bipartite arrangement similar to the nuclear localization signal of nucleoplasmin (22).
The RNA binding activity of CBTF 122 and NF90, as well as of other dsRBD-containing proteins, has been well characterized. A CBTF 122 -maltose-binding protein fusion binds to dsRNA and A-form DNA-RNA hybrids without detectable sequence specificity (21). The NMR structure of the Staufen dsRBD has been solved (23), as has the crystal structure of the Xlrbpa dsRBD complexed with dsRNA (24). Saturation mutagenesis studies on the dsRBD of Xlrbpa demonstrated that the substitution of a conserved central Phe residue by Ala in both Staufen and Xlrbpa dsRBDs leads to abolition of RNA binding activity (23,25). Reference to the published structure of the dsRBD-RNA complex demonstrates that this residue is important in stabilizing a lysine side chain that makes important backbone contacts adjacent to the major groove (24). It is this mutation that we have previously used to inactivate the dsRBDs of CBTF 122 (14).
The human orthologue of CBTF 122 has been shown to require intact RNA-binding domains for transcription regulation but on promoters that are not known to be its in vivo targets (10). We therefore tested whether an intact RBD was similarly required for transcription regulation at the GATA-2 promoter in Xenopus, a well characterized in vivo target of CBTF 122 .

EXPERIMENTAL PROCEDURES
Embryo Manipulation and Western Blotting-The embryos were obtained as described by Smith and Slack (26) and staged according to Nieuwkoop and Faber (27). In vitro transcription and microinjection of mRNA was carried out as described previously (28,29). Injections were carried out at the 8-cell stage with 4 nl of RNA solution (37 pg of RNA total). One embryo equivalent of freon extracted whole embryo lysate was assayed for CBTF 122 fusion expression by SDS-PAGE and Western blotting (30) using CBTF 122 antibody (provided by B. Bass).
Analysis of mRNA Levels by Real Time Reverse Transcriptase-PCR-cDNA was synthesized from RNA extracted from dissected explants of injected embryos cultured until stage 11 using the method of Steinbach and Rupp (31). RNA expression was assayed as cDNA by PCR conducted on an ABI 7900HT sequence detection system using TAQMAN fluorescently labeled probes. The data were analyzed using the ⌬⌬Ct method (ABI) using ODC as a control gene and uninjected marginal zones as control tissue. The probe and primer sequences are available by request.
Protein Expression and Purification-The RBD and RBDϩRGG regions of CBTF 122 as well as the mutant RBD (mRBD) region were subcloned into pGex2T (Amersham Biosciences). The GST fusion proteins were expressed in Escherichia coli and purified by glutathione-Sepharose and DNA-cellulose column chromatography as previously described (14). The purified proteins were assayed and quantified by SDS-PAGE and UV spectroscopy using the calculated extinction coefficient of each protein. Embryo extract containing the multi-subunit CBTF complex was prepared as described previously (2).
Electrophoretic Mobility Shift Assays of Recombinant CBTF 122 and Endogenous CBTF-Either RNA (Cruachem) or DNA (Invitrogen) oligonucleotides were annealed to form duplexes and end-labeled by T4 polynucleotide kinase (New England Biolabs) using [␣-32 P]ATP (32). The proteins were incubated with the appropriate nucleic acid probe for 15 min on ice in EMSA buffer (2) with the addition of 500 ng of poly(dI-dC) in the case of embryo extract, prior to separation of DNAprotein complexes on a 4% native polyacrylamide gel in 0.25ϫ TBE. The gels were dried, and the free DNA and DNA-protein complexes were quantified using a phosphorimaging device (Molecular Dynamics). The K d values were estimated from the protein concentration required to bind 50% of the DNA.
Dimethyl Sulfate Interference Footprinting-Four picomoles of oligonucleotide were labeled at the 5Ј end and annealed with the complementary sequence to form a double-stranded 55-mer probe. The probe was then gel-purified to remove residual single stranded oligonucleotide. The purified dsDNA was treated with 10% dimethyl sulfate (33), and the partially methylated probe was used in an EMSA. Separated species were transferred onto DE81 paper. After autoradiography the bound and free species were excised and eluted in 500 l of buffer (1 M NaCl, 20 mM Tris, pH 8.0, 1 mM EDTA). 10 g of tRNA was added, and the DNA was ethanol-precipitated. Piperidine cleavage and further treatment was performed as described by Maxam and Gilbert (33); DNA fragments were separated on a denaturing 16% polyacrylamide gel and visualized by autoradiography.
Circular Dichroism-An Applied Photophysics *-180 instrument was flushed with nitrogen gas for all CD experiments. Cell path lengths of 1, 2, 4, and 10 mm were used with sample concentrations of 15-60 g/ml. All 29-bp duplexes were dissolved in 100 mM KF, 5 mM NaPO 4 buffer, pH 7.6, at 22°C. The concentrations were determined by UV measurements at 260 nm coupled with snake venom time course digestions to correct for hypochromic differences. The data were collected over the wavelength range 180 -360 nm using adaptive sampling in conjunction with signal averaging in all cases. The instrument wavelength accuracy was Ϯ0.1 nm, determined from the Xenon lines, and the CD was calibrated from camphor sulfonic acid at 290.5 nm.
A-form DNA Predictive Methods-A-DNA propensity energy values for the duplex oligonucleotides were calculated using the method of Basham et al. (38). The method was modified to determine the percentage of each duplex in either A-or B-form using the following classifications and rules. Class 1 is strong A or B formers are base pairs that have the same sign and whose sum is greater than 0.5. Class 2 is neutral A or B formers are base pairs that are of opposite sign and whose sum is zero. Class 3 is weak A or B formers are base pairs whose sum is less than 0.5 but not zero. Class 4 is base pairs unclassified by Basham et al. (38) because of lack of data. These are defined by their base composition as GC, CG or TA, AT. The following rules were applied to the above classifications: 1) all Class 1 base pairs remain as they are under all situations, conserved type; 2) if one or two consecutive class 2 or 3 are between two or more class 1 base pairs of the same type (A or B), then they both take on strong base pair type; 3) class 2 neutrals adjacent to class 1 base pairs take on that type if the base pair on the other side to the class 2 neutral is of class 2 or class 3; and 4) class 4 base pairs between class 1 base pairs of the same type can take on the same character as the class 1 type. The predictions shown (see Table III) show only the A and B percentage forms; there is 7% contribution that cannot be determined using the three base moving window model of Basham et al. (38), because these involve the ends of the molecule. Also, regions where no bias for each form can be determined add to the percentage of uncertainty.

Mutations in the dsRBDs of CBTF 122 EN Lead to a Loss of
Transcriptional Regulation-The engrailed transcription repression domain (EN) when fused to a transcription activator leads to transcription repression of the target gene (34). We have recently shown that the transcription inhibiting (CBTF 122 EN) fusion form of CBTF 122 gives rise to a double axis phenotype (Fig. 1b, arrow) when expressed in Xenopus embryos. We confirmed that this effect was a result of changes at a transcriptional level and not a failure of CBTF 122 EN to be incorporated into the full CBTF complex, or an effect on mRNA turnover, by testing the strong transcription up-regulating CBTF 122 form, CBTF 122 VP16. This protein gave the opposite, ventralized microcephalic phenotype when expressed in embryos dorsally (Fig. 1b, note the lack of eyes and head morphology), whereas CBTF 122 expression alone gave rise to no obvious change in phenotype. 2 Transcription activation by CBTF 122 is regulated in part by its localization in the cytoplasm through anchoring to RNA. The mutation F435A/F559A in the dsRBDs leads to a loss of RNA binding. Such mutants localize early to the nucleus and thus arrive at their site of action earlier than native CBTF 122 (2). We investigated how this would affect transcriptional control by mutating the RBDs in the context of full-length CBTF 122 EN and CBTF 122 VP16 fusions. These fusion proteins (CBTF 122 mRBDEN and CBTF 122 mRBDVP16) were expressed in embryos by injection of in vitro transcribed mRNA alongside those injected with CBTF 122 EN and CBTF 122 VP16 mRNA. Although constructs with active dsRBDs resulted in the expected double axis and headless phenotypes, those containing a mutant form of the domains did not (Table I), suggesting a role for the dsRBDs in gene regulation.
We have shown CBTF 122 EN to be a powerful, specific repressor of transcription from the GATA-2 promoter. 2 We therefore tested the ability of CBTF 122 mRBDEN to repress GATA-2 transcription in a Xenopus animal cap assay (35). In this assay, injection of CBTF 122 EN mRNA reduces GATA-2 mRNA levels by over 1000-fold, whereas injection of CBTF 122 mRBDEN mRNA reduces GATA-2 mRNA levels by only 10-fold (Table II). We confirmed by Western blotting of whole embryo extract using CBTF 122 antibody that the two proteins were expressed at comparable levels (Fig. 1c). Thus a single point mutation in each of the dsRBDs is almost sufficient to abolish repression by CBTF 122 EN at the GATA-2 promoter. To understand how the RNA-binding mutations affected the ability of CBTF 122 to act at the transcriptional level, we then analyzed how the binding of CBTF 122 dsRBDs to nucleic acids was affected by these mutations in the dsRBDs in vitro.
The dsRBD Domains of CBTF 122 Bind Nucleic Acids Preferentially in the Order dsRNA Ͼ dsDNA Ͼ ssRNA-CBTF 122 contains a single RGG domain of the type found in hnRNP A1 and U, which is implicated in single-stranded nucleic acid binding (18,19), and twin dsRBDs of the type found in Staufen and Xlrbpa (16). To determine the contribution of these different domains to the interaction with nucleic acids, we expressed and purified CBTF 122 fragments corresponding either to the dsRBD and RGG domains together, or the dsRBDs alone, as GST fusions (Fig. 1a). The predicted sizes of the expressed protein fragments were 41 and 28.5 kDa, respectively (CBTF 122 (41) and CBTF 122 (28.5)); both contained the nuclear localization signal. We purified these proteins to 95% homogeneity and then tested their ability to bind a dsRNA homologue of the Ϫ69 to Ϫ33 CCAAT box-containing region of the GATA-2 promoter by EMSA (Fig. 2a). The binding curves for CBTF 122 (41)GST and CBTF 122 (28.5)GST show that their affinities for RNA are comparable, with K d values of ϳ0.3 nM. Consequently the smaller 28.5-kDa fragment was chosen for subsequent EMSA studies.
To estimate the relative affinity of the dsRBDs of CBTF 122 for ssRNA, dsRNA, and dsDNA, data from EMSA experiments using CBTF 122 (28.5)GST were quantitated, and the percentage of bound probe was plotted as a function of protein concentra-  . c, sets of 20 embryos were injected with 37 pg of the RNA shown and allowed to develop to stage 11. Embryo lysate was prepared, and yolk proteins were removed by freon extraction prior to analysis of the exogenous protein by Western blotting using an antibody recognizing CBTF 122 . tion (Fig. 2b). The estimated binding constants show that the dsRBDs have the highest affinity for dsRNA, much lower affinity for dsDNA, and the lowest affinity for ssRNA with estimated K d value s of 0.3, 6.0, and 30 nM, respectively. Previously a full-length CBTF 122 -maltose-binding protein fusion has been reported to have a K d of 0.27 nM for dsRNA (21). Therefore, our data suggest that dsRNA binding activity resides entirely within the region of the dsRBDs represented in the CBTF 122 (28.5)GST protein. Because this fragment bound to DNA, we then tested whether the RNA binding mutations that affected the ability of CBTF 122 EN to control GATA-2 transcription also affected DNA binding.
Mutations in the dsRBDs Affect DNA Binding Activity-The previously described mutated protein CBTF 122 mRBD (2) contains two point mutations, F435A and F559A, known to be critical for RNA binding (23,36). However, it is possible that residues Phe435 and Phe559 are also important for the less understood DNA binding activity of this protein. We expressed and purified recombinant forms of both the native and mutant RNA-binding domains, again as GST fusions. These proteins (CBTF 122 (28.5)GST and CBTF 122 mRBD(28.5)GST) were tested for their ability to bind to either DNA or RNA homologues of the Ϫ59 to Ϫ31 GATA-2 promoter sequence. In agreement with our previous data (37) the CBTF 122 mRBD(28.5)GST protein has a much reduced affinity for RNA duplex compared with CBTF 122 (28.5)GST (Fig. 3a). However, its DNA binding activity is also inhibited (Fig. 3b). Because these data show that dsRNA and dsDNA binding by CBTF 122 involves at least some of the same critical amino acids, we investigated the interaction of CBTF 122 (28.5) with DNA further.
The dsRBDs of CBTF 122 Bind DNA in a Structure-dependent Manner-CBTF 122 is known to bind dsRNA in a sequenceindependent manner (21). To test whether CBTF 122 (28.5)GST showed any sequence specificity in binding DNA, we designed five oligonucleotides, based around the Ϫ59 to Ϫ31 CCAAT box region of the GATA-2 promoter (P1 to P5, Table III). These oligonucleotides were then used in competition EMSA experiments against the P1 DNA probe. The results showed that P3, P4, and P5 competition was comparable with P1 self-competition, but P2 was found to compete very poorly (Fig. 3c).
Binding of P2 was disrupted despite the retention of the core CCAAT sequence to which CBTF binds. One possible explana- injection into Xenopus embryos Two cell embryos were injected in the animal pole with 37 pg of synthetic RNA encoding either CBTF 122 EN or CBTF 122 mRBDEN. The embryos were allowed to develop to stage 8, and the animal caps were removed and cultured until sibling embryos reached stage 11. Relative amounts of GATA-2 mRNA were measured using real time reverse transcriptase-PCR. The data were analyzed by the ⌬⌬C t method (ABI), where C t is the threshold cycle, and ⌬⌬C t is normalized for amount of input cDNA by comparison with the control gene (ODC) and levels relative to a calibrator sample (unijected embryos).  (41)GST forms of the CBTF 122 RNA-binding domains were expressed in E. coli as fusions with glutathione S-transferase. The purified proteins were used in EMSA assays with 0.3 nM of either the dsRNA, dsDNA, or ssRNA version of the CBTF 122 -binding sequence as a probe. a, EMSA data using the 36-bp synthetic dsRNA probe for CBTF 122 (41)GST and CBTF 122 (28.5)GST. b, EMSA data for CBTF 122 (28.5)GST using the dsRNA, ssRNA, or dsDNA 36-bp probes. All of the data were analyzed using a Molecular Dynamics phosphorimaging device. The proportion of bound probe at each protein concentration was calculated using: % bound ϭ bound/(free ϩ bound) ϫ 100 and plotted against protein concentration. c, the sequences of the probes used in the assays; the CCAAT sequence is boxed.
tion for this was that a structural rather than sequence-dependent binding mechanism was disrupted by the mutation. Because dsRBDs may be expected to have a preference for binding A-form structure, we initially used the method of Basham et al. (38) to calculate the average A-DNA forming potential for each of the oligonucleotides used in this study (Table III). We refined this method to predict the percentage Aand B-forms for each duplex (see "Experimental Procedures"); this is shown as "predicted" in Table III. The P1 sequence was predicted to be strongly A-DNA forming. In particular, the bases mutated in P2 mapped to a region of high A-DNA forming potential (Ϫ57 to Ϫ47). We therefore tested the hypothesis that A-DNA elements in the P1 probe are important for interactions of the dsRBDs with dsDNA by extending our study to the structure of the Ϫ59 to Ϫ31 region of the GATA-2 promoter.

TABLE III
Oligonucleotides used in competition EMSA Double-stranded oligonucleotides were annealed and used in competition EMSA. Wild-type sequence of the Ϫ59 to Ϫ31 region of the GATA-2 promoter and mutants based around it are designated P1 to P5, and the mutated bases are highlighted in bold type. Control oligonucleotides are designated C1 to C3, APEs (38) are also shown. The percentages of each duplex predicted to be A-and B-form are also shown; this does not add to 100% because of the inability to calculate these data for the duplex ends and for certain sequences.
in solution. A-form DNA is characterized by a CD spectrum that has a strong positive band centered between 267-273 nm and a negative band centered between 209 -211. B-form DNA is characterized by a strong positive band centered between 273-278 nm, and two negative bands centered at ϳ206 and 250 nm. A spectrum of previously described B-form (C1) and A-form (C2) oligonucleotides (38) is shown in Fig. 4a. Each base has different transition dipoles; hence the band center positions and intensities vary for the same structures depending on composition. The spectrum (Fig. 4a) of the P1 oligonucleotide exhibits a negative band centered at 211 nm and a positive band centered at 268 nm. The CD results therefore show that oligonucleotide P1 has significant A-form structure. The CD spectrum of P1 could be the result of mixed regions of A-form and B-form DNA or a continuous structure sharing the properties of both. The A-and B-form components were deconvoluted by measuring the signal intensity at 278 nm for C1 (100% B-form) and C2 (0% B-form) and plotted against percent B-form, assuming a linear change between the two points (Fig. 4b).
The equation of the resulting graph was used to calculate the percentage of B-form for P1 and P2. The data suggest that 35% of P1 and 51% of P2 is in the canonical B-form, assuming a two component model where the oligonucleotide duplex is composed solely of A-and B-forms. To test the validity of using a single wavelength to estimate the amount of B-form, we reconstituted the P1 spectrum using the calculated 65% A-form and 35% B-form (Fig. 4c). The two curves fitted well, particularly over the important 250 -310-nm range that reflects the contribution from the bases.
In light of the partial A-form nature of P1, we used oligonucleotides C1 and C2, as well as designing a further control FIG. 4. Circular dichroism of P1, P2, C1, and C2 oligonucleotides. a, CD spectra from 205 nm to 360 nm for: 1) oligonucleotide C1 (B-form), 2) oligonucleotide P2, 3) oligonucleotide C2 (Aform), and 4) oligonucleotide P1. All of the samples were in 100 mM KF, 5 mM NaH 2 PO 4 buffer, pH 7.6, at 22°C. b, calibration curve derived from CD signal intensity at 278 nm and assuming C1 to be 100% B-form and C2 to be 0% B-form. The equation defining the line was used to predict percentage of B-form for P1 and P2. c, comparison of the observed P1 CD spectrum with a calculated spectrum based on a duplex predicted to contain 65% A-form and 35% B-form. oligonucleotide (C3) to test whether CBTF 122 (28.5)GST bound preferentially to A-DNA (Table III). C2 is a known A-form oligonucleotide (38), whereas C3 is an alternating poly(AT) tract predicted to form a B-form structure. We then used these oligonucleotides for competition EMSA analysis. The data showed that the A-form DNA oligonucleotide C2 competed better than any of the oligonucleotides P1 to P5. In contrast, the control B-form oligonucleotides C1 and C3 competed poorly (Fig. 3c).
Having shown a correlation between A-form structure and the DNA binding affinity of CBTF 122 , we investigated the multisubunit CBTF complex-DNA interaction and the relative importance of sequence and structure elements in the specific binding of CBTF to elements of the GATA-2 promoter.
Binding of the Multisubunit CBTF Complex to DNA Is Mainly Sequence Not Structure-dependent-The binding of the CBTF 122 (28.5)GST fusion protein to DNA was unaffected by mutations in or around the CBTF consensus sequence but dependent on A-form structure. To test the relative importance of the sequence and structure of the DNA in the binding of the intact CBTF complex, we used the same oligonucleotides we had previously used for binding studies on the dsRBDs, as competitors in EMSAs, not of purified CBTF 122 domains as before but of CBTF from whole cell extract. Binding of the full CBTF complex requires the presence of a CCAAT box in the sequence (1, 2); we therefore included a new oligonucleotide (P6) lacking the CCAAT box (Fig. 5). The results show that oligonucleotides based around the CCAAT box compete well for the CBTF complex and that sequences that lack the CCAAT box compete poorly for the complex, although an A-form oligonucleotide (C2) competed best of those oligonucleotides containing no CCAAT box (Fig. 5).
The binding of the intact CBTF complex to DNA was thus found to be dependent on specific sequences within the P1 region. To determine further those bases involved in these interactions, the binding was analyzed by dimethyl sulfate interference footprinting. In this assay, the unbound fraction is considered to contain every species of methylated DNA, whereas the retarded band is enriched in species of methylated DNA that do not interfere with binding (39). It therefore identifies the extent of protein contacts with specific bases that are important for complex stability. Our results show that modification of purines between Ϫ56 to Ϫ40 (inclusive) reduces binding of CBTF to its cognate sequence (Fig. 6). Thus the multisubunit CBTF makes extensive contacts over a sequence spanning 17 bases, including nucleotides well beyond the core CCAAT sequence. DISCUSSION Here we report that single point mutations in both dsRBDs of the protein CBTF 122 prevent the formation of phenotypes we have previously found associated with the dominant negative and up-regulating fusion forms of this protein. Further, these mutations severely inhibit the GATA-2 transcription regulation activity of CBTF 122 EN in embryos. Our in vitro binding studies show that the dsRBDs can bind both RNA and DNA and that the same point mutations inactivate both of these functions. The RBDs show a preference for binding to A-form DNA. This is likely to be important for the regulation of GATA-2 because an oligonucleotide corresponding to the Ϫ59 to Ϫ31 CBTF-binding region of the GATA-2 promoter adopts a FIG. 5. Competition EMSA of the multisubunit CBTF-DNA complex by mutant probes. a, in all cases end-labeled wild-type CCAAT probe (4 fmol, *) was used in a standard binding reaction to assay CBTF from crude embryo extracts. Competitor oligonucleotides (Table III) were in the binding reactions at a 5-, 10-, or 50-fold molar excess prior to the addition of embryo extracts and the subsequent separation of DNA-protein complexes using standard EMSA conditions. The samples were run alongside free probe (lane F) and in the absence of competitor (lane S). The specific CBTF complex is marked with an arrow. Other complexes observed are nonspecific and are a consequence of using crude embryo extracts, the only current source of the intact CBTF complex. The gel shown in the lower right panel has been run less far than the other three. b, the percentage of inhibition of CBTF binding to the native probe was calculated when the EMSA reactions contained competitor at a 10-fold molar excess over probe. The data are the means from three experiments.
FIG. 6. Dimethyl sulfate interference footprinting of the CBTF-DNA complex. A DNA probe corresponding to bases Ϫ66 to Ϫ22 from the GATA-2 promoter with either radiolabeled ϩ or Ϫ strand underwent exposure to dimethyl sulfate prior to preparative EMSA using crude embryo extract. Free and bound species were recovered from the EMSA gel and cleaved; the resulting DNA fragments were resolved on a 16% denaturing polyacrylamide gel. Methylated bases that interfere strongly with binding are marked by •, and those that interfere partially by E. The region corresponding to the CCAAT sequence is boxed. partial A-form structure. However, sequence and not structure is the prime requisite for binding of the complete CBTF complex to the GATA-2 promoter in vitro.
Engrailed and VP16 fusions of CBTF 122 proteins were used as a method of analyzing the transcriptional regulation activity of this dual function protein in isolation. Opposing phenotypes were found to arise from CBTF 122 EN and CBTF 122 VP16 expression when they were expressed in Xenopus embryos (dorsalizing and ventralizing respectively), strongly suggesting that the effects are a consequence of down-and up-regulation at the transcriptional level. The mode of action of these fusions only requires the protein to be at the promoter site for transcription regulating complexes to be recruited by the EN and VP16 domains. In particular, the engrailed transcription repression domain would inhibit GATA-2 transcription if it were present at the GATA-2 promoter. Hence we infer that the mutations in the dsRBDs that prevent GATA-2 repression by CBTF 122 EN most likely involve a failure of the protein to bind to the GATA-2 promoter in vivo.
These results are in agreement with a similar study performed in cultured cells using NF110 (a human orthologue of CBTF 122 (10)), in which intact dsRBDs were also required for transcription control. RNA binding has previously been associated with transcription activity function, for example; an RNAinduced conformational change was proposed to be necessary for activation of the transcriptional regulatory role of NF110 (10), and RNA is also a co-factor of the steroid receptor coactivator (40). We cannot formally exclude similar RNA-induced activation or co-activator function underlying the requirement for RNA binding for transcription regulation by CBTF 122 . However, the fact that CBTF 122 fusion forms are inactivated by mutations that affect RNA binding despite a dominant domain providing transcription regulation make a model in which RNA is required for transcription regulation per se unlikely.
Our data support an alternative hypothesis, that the dsRBDs act as both RNA and DNA binding motifs, because their mutation inhibits both RNA and DNA binding in vitro. Other proteins also contain RNA and DNA binding activities residing in the same amino acid sequence, for example the cold shock domain of YB-1 (41) and the homeodomain-like motif of jerky (42). Further support for the proposal that dsRBDs also have a DNA binding role comes from Tn916 integrase, which has a structural domain highly related to the dsRBD that is capable of sequence-specific DNA binding (43).
In the case of the dsRBDs of CBTF 122 , binding to dsDNA is structure-dependent, with a strong preference for A-form DNA. Although dsDNA is normally considered to exist in the B-form state, the CD data presented here suggest that the Ϫ59 to Ϫ31 region of the GATA-2 promoter used in this study adopts a partial A-form structure. Analysis of the CD spectrum of P1 indicates 65% of the DNA to be A-form. That much of this corresponds to the sequence encompassed by the Ϫ55 to Ϫ45 region of the promoter is suggested by predictive methods. This region is just upstream of the bulk of the specific contacts revealed by the footprinting data. Mutation of approximately half this region (Ϫ56 to Ϫ50) effectively abolishes binding of CBTF 122 (28.5)GST, whereas mutations outside this region have little effect upon binding affinity. Together, these data argue for a two-component model of the P1 duplex (distinct regions of A and B helix), with the dsRBDs binding to the A-form region.
The crystal structure of the dsRBD from Xlrbpa complexed with dsRNA shows the dsRBD binding in the minor groove of an A-form helix and specific contacts being made with a 2Ј-OH of the RNA by loop 2. It is possible that the adoption of an A-form structure by the P1 DNA sequence allows the dsRBDs of CBTF 122 to bind the minor groove of dsDNA, which is then similar to that of dsRNA. The lack of a 2Ј-OH in DNA is unlikely to abolish binding but may reduce the binding affinity, as observed in our data.
Investigation of the DNA binding by CBTF 122 dsRBDs showed no clear evidence of sequence specificity outside the requirement for a sequence that may adopt an A-form helix. This is in contrast to the complete CBTF complex, for which our EMSA studies show clear a sequence requirement for the conserved CCAAT box region. An A-form structure alone is insufficient for binding by the intact complex, but our competition studies suggest that this structure makes some contribution to the overall affinity. CBTF makes sequence specific base contacts over a 17-bp region that is highly conserved between GATA-2 promoters and distinct from those made by other CCAAT factors (44,45). It is likely that CBTF subunits other than CBTF 122 are required to contact the CCAAT core.
Overall, the data presented here add considerably to our knowledge of the CBTF-DNA interaction and the role of CBTF 122 within the multisubunit complex. Our results support a role for dsRBDs as DNA-binding domains of functional significance both in vitro and in vivo.