Identification and Structural Characterization of a CBP/p300-Binding Domain from the ETS Family Transcription Factor GABPα

Summary

Using NMR spectroscopy we identified and characterized a previously unrecognized structured domain near the N-terminus (residues 35−121) of the ETS family transcription factor GABPα. The monomeric domain folds as a 5-stranded β-sheet crossed by a distorted helix. Although globally resembling ubiquitin, the GABPαfragment differs in its secondary structure topology and thus appears to represent a new protein fold that we term the OST (On-SighT) domain. The surface of the GABPα OST domain contains two predominant clusters of negatively-charged residues, suggestive of electrostatically-driven interactions with positively-charged partner proteins. Following a best candidate approach to identify such a partner, we demonstrated through NMR-monitored titrations and GST-pulldown assays that the OST domain binds to the CH1 and CH3 domains of the co-activator histone acetyltransferase CBP/p300. This provides a direct structural link between GABP and a central component of the transcriptional machinery.

GA-binding protein (GABP), also known as E4 transcription factor 1 (E4TF-1) and nuclear respiratory factor 2 (NRF-2), is a widely-expressed member of the ETS transcription factor family with roles in gene regulation related to embryogenesis, immune system development, cell cycle progression, neuromuscular junctions, protein synthesis, cellular respiration, and viral pathogenicity.^{1,2,3,4,5,6,7} A distinguishing feature of GABP amongst ETS family members is that it recognizes tandem promoter sites as a heterotetrameric αβ₂αcomplex with the DNA-binding and transactivation activities provided by separate polypeptide chains (Fig. 1(a)). The GABPα subunit is composed of a C-terminal winged helix-turn-helix ETS domain, which binds preferentially to purine-rich regions based on the sequence 5’-GGAA/T-3’,⁸ and a central PNT domain. The helical bundle PNT domain⁹ appears to mediate protein-protein interactions with the co-activator histone acetyltransferase CBP/p300.¹⁰ Full length GABPβ1 and closely related GABPβ2 are composed of N-terminal ankyrin repeats (ANK), a transactivation domain (TAD), and a C-terminal leucine-zipper domain (LZ).¹ The LZ domain, which serves for homo/heterodimerzation of the GABPβhomologs, is absent in two of four alternatively spliced GABPβ1 isoforms.¹ As revealed by the crystal structure of an ETS domain/ANK/DNA ternary complex, GABPβ ANK binds GABPα via the ETS domain and a flanking C-terminal sequence, and in doing so indirectly increases core promoter affinity.^8,11

Figure 1.

(a) Cartoon of GABP domain organization. Boundaries are predicted or correspond to fragments used for studies of the individual domains. (b) The close superimposition of corresponding peaks in the ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled GABPα³⁵⁻¹²¹ (black) and GABPα¹⁻³²⁰ (red) demonstrates that the OST domain adopts an independently folded structure. The additional well-dispersed peaks in the spectrum of GABPα¹⁻³²⁰ arise from the PNT domain,⁹ whereas those clustered near ∼8.0 to 8.6 ppm in ¹H correspond to residues with predominantly random coil conformations. Assignments are given for GABPα³⁵⁻¹²¹, and aliased peaks are denoted with an asterisk. (c) “Beads-on-a-string” model of full-length GABPαshowing the 3 structured domains (PNT, 1SXD.pdb; ETS, 1AWC.pdb) along with all remaining residues in extended conformations.

Methods: Genes encoding GABPα³⁵⁻¹²¹, GABPα¹⁻¹⁶⁹, and GABPα¹⁻³²⁰ were PCR-amplified from the full-length murine GABPαDNA (Genbank 34328119) and inserted in the vector pET28a (Novagen) for expression with an N-terminal His₆-tag and thrombin cleavage site. A codon for non-native Trp122 was added to the GABPα³⁵⁻¹²¹ gene to facilitate concentration measurements by absorbance spectroscopy. His₆-GABPα^168−256 was previously described.⁹ The GABP constructs were expressed in E. coli BL21(λDE3) cells grown at 37 °C in LB or minimal M9T medium containing 1 g/l ¹⁵NH₄Cl for uniform ¹⁵N-labeling, 1 g/l ¹⁵NH₄Cl and 3 g/l ¹³C₆-glucose for uniform ¹³C/¹⁵N-labeling, and 0.3 g/l ¹³C₆-glucose and 2.7 g/l ¹²C₆-glucose for non-random 10% ¹³C-labeling (Stable Spectral Isotopes). After induction with 1 mM IPTG at OD₆₀₀ ∼ 1 and growth for 4 hours at 37 °C, the cells were harvested by centrifugation. The cell pellets were resuspended in buffer A (5 mM imidazole, 50 mM HEPES (pH 7.5), 500 mM NaCl, 5 % glycerol) and lysed by passage through a French press three times at 10,000 psi, followed by 15 min of sonication on ice. The lysate was cleared by centrifugation, and the supernatant loaded onto a Ni-NTA column (GE Healthcare), pre-equilibrated with buffer A. The column was washed with 10 column volumes of buffer A plus 60 mM imidazole, and then the bound proteins were eluted with buffer A plus 250 mM imidazole. Following SDS-PAGE analysis, the appropriate fractions were pooled and dialyzed overnight in 20 mM sodium phosphate (pH 7.2) and 20 mM NaCl with a few crystals of thrombin (Roche) for the cleavage of the N-terminal His₆-tag. Thrombin and the cleaved His₆-tag were removed by incubation with ρ-aminobenzamidine beads (Sigma) and TALON metal affinity resin (BD Biosciences) at room temperature with mild shaking for 15 min, followed by centrifugation. The purified proteins, with an N-terminal Gly-Ser-His remaining from the cleavage site, were concentrated to ∼ 1 to 1.5 mM in NMR sample buffer (20 mM sodium phosphate (pH 7.0), 50 mM NaCl, 2 mM DTT, ∼10 % D₂O) using a 5 kDa cut-off Amicon Ultrafiltration device (Millipore). Spectra were recorded at 30 °C using Varian 500 MHz Unity, 600 MHz Inova, or 800 MHz Inova NMR spectrometers equipped with a triple resonance gradient probes, and analyzed using NMRpipe ³⁷ and Sparky.³⁸ Main-chain and aliphatic side-chain ¹H, ¹³C, and ¹⁵N resonances were assigned using standard triple resonance correlation experiments.³⁹ Resonances from aromatic side-chains were assigned using ¹H-¹³C HSQC, CβHδand CβHεexperiments.⁴⁰ Stereospecific assignments of prochiral H^β,β signals were obtained from HNHB and short mixing time (30 msec) ¹⁵N-TOCSY-HSQC spectra,⁴¹ of Asn and Gln ¹⁵NH₂ groups from a EZ-HMQC-NH₂ spectrum,⁴² and of Val and Leu methyl groups from a constant time ¹H-¹³C HSQC spectrum of the non-randomly 10% ¹³C-labeled protein.⁴³

GABPα has been reported to interact with several partner transcription factors, including ATF1,¹² Sp1 and Sp3,¹³ and C/EBPα,¹⁴ as well as CBP/p300^10,15,16 and histone deacetylase HDAC1.⁷ Although the molecular bases for these interactions remain unknown, those involving ATF1 and CBP/p300 have been mapped to the region of this protein preceding its ETS domain. In addition to the PNT domain, this region also includes a poorly characterized ∼170 amino acid sequence that is conserved only among GABPα orthologs, and to a lesser extent, Elg homologs from the ETS family (Supplemental Figure S1). To provide a structural framework for establishing the functional role(s) of this sequence, we investigated by NMR spectroscopy a series of GABPα N-terminal truncation fragments, thereby discovering that residues 35−121 encompass a previously unrecognized folded domain. We determined that this monomeric domain is composed of a five-stranded β-sheet crossed by a distorted helix. Although globally similar to ubiquitin, the GABPα fragment has a distinct secondary structure topology, and thus represents a new fold that we term the OST (On-SighT) domain. Through in vitro binding studies, we also demonstrated that the isolated GABPα OST domain interacts with the cysteine-histidine rich CH1 and CH3 domains of CBP/p300, whereas the PNT domain only binds the latter.

Identification of the OST domain

A new structured domain in the previously uncharacterized N-terminal region of GABPα was evident from the ¹H-¹⁵N HSQC spectrum of GABPα¹⁻³²⁰ (Figure 1(b)). In addition to the expected ¹H^N-¹⁵N signals corresponding to the PNT domain,⁹ numerous additional well-dispersed peaks were also observed. The same peaks were present in the spectrum GABPα¹⁻¹⁶⁹, which lacks the PNT domain (not shown). Through partial backbone assignments of this latter truncation fragment, the new domain was localized approximately to residues 35−121, whereas the flanking residues 1−34 and 122−169 exhibited chemical shifts and ¹⁵N relaxation properties diagnostic of a random coil polypeptide. Allowing for a few additional N- and C-terminal residues, GABPα³⁵⁻¹²¹ was expressed and found to yield an excellent quality ¹H-¹⁵N HSQC spectrum. This spectrum, indicative of a well-folded domain termed OST, accounted for all of the additional peaks first observed with GABPα¹⁻³²⁰.

Structure and dynamics of the OST domain

The structural ensemble of the OST domain was determined using an extensive set of NMR-derived restraints recorded for GABPα³⁵⁻¹²¹ (Figure 2, Table 1). The GABPα fragment adopts a compact monomeric fold comprised of a five-stranded mixed parallel/anti-parallel β-sheet (S₁; 36−43, S₂; 67−70, S₃; 73−74, S₄; 91−99, S₅; 107−115) crossed by a distorted helix (48−59). Strand S₅ contains a classic β-bulge centred at Leu111-Glu112. Chemical shift, NOE, and ³J_NH-Hα coupling measurements demonstrate that the single helix in the OST domain begins with 3₁₀ conformation and extends to Leu59 as an α-helix, despite being kinked due to Pro57 (Supplemental Figure S2).

Figure 2.

(a) Ribbon diagram of the lowest-energy NMR-derived structure of the GABPα OST domain (“front” view), with the β-sheet, helix, and two distinct loops (S₃/S₄ and S₄/S₅) shown in yellow, red, and cyan, respectively. Superimposed backbone (b) and hydrophobic sidechain (c) atoms from the structural ensemble of GABPα³⁵⁻¹²¹, with selected residues labelled for reference. (d) Two clusters of negatively-charged Asp and Glu residues (red) are observed on the “front” (left) and “back” (right) of the OST domain. In contrast, no extended clusters of (d) positively-charged Arg, His, or Lys residues (blue) or (e) exposed hydrophobic residues (green) are found on its surface (polar residues in grey).

Methods: The tertiary structure of OST was calculated with ARIA(v1.2)/CNS utilizing distance, dihedral angle, and orientation restraints (Table 1). Interproton distance restraints were obtained from 3D ¹⁵N- NOESY-HSQC (600 MHz, 150 msec mixing time),⁴⁴ simultaneous aliphatic ¹³C- and amide ¹⁵N-NOESY-HSQC (800 MHz, 125 msec),⁴⁵ aromatic ¹³C- NOESY-HSQC (500 MHz, 125 msec),⁴⁶ and constant time methyl-methyl and amide-methyl-NOESY spectra (600 MHz, 140 msec),⁴⁷ followed by the automatic and iterative NOE assignments in ARIA. Mainchain dihedral angles were derived from ¹H, ¹³C, and ¹⁵N chemical shifts using TALOS.⁴⁸ The predicted dihedral angles were only used when the Φ values agreed with ³J_NH-Hα coupling constants obtained from a HNHA spectrum.⁴⁹ The χ₁ angle restraints for residues with prochiral H^β protons were determined in concert with the stereospecific assignments, and those of the Val, Thr, and Ile residues were determined from ³J_NCα and ³J_C’Cα coupling constants measured using N-Cα and C'-Cαspin echo experiments.⁴⁹ For residues not showing ³J couplings indicative of rotamer averaging, the χ₁ angles were set to ± 60° or 180° based on a staggered rotamer model. ¹H^N-¹⁵N RDC orientational restraints were measured from ¹H-¹⁵N IPAP-HSQC spectra using protein aligned with stretched gels⁵⁰, ⁵¹ and incorporated at iteration 4 with the SANI routine in ARIA, with default energy constants of 0.2 and 1 kcal mol⁻¹ Hz⁻² for the first and second simulated annealing cooling stages, respectively. Values of the alignment tensor (R = 0.3, D_a = 5.3 Hz) were estimated by the histogram method, followed by a grid search for minimal SANI violations.⁵² All His imidazole sidechains were determined to be in the neutral N^ε2 tautomeric state from long-range ¹H-¹⁵N correlation measurements.⁵³ All Xaa-Pro amides were constrained to the trans conformation based on a chemical shift analysis with POP.⁵⁴ In the final set of 10 water-refined structures, there were no violations greater than 0.5 Å and 5° for distance and angle restraints, respectively. The structural ensemble was analyzed using Procheck-NMR⁵⁵ and secondary structures determined according to Promotif⁵⁶ and Vadar.⁵⁷ Structural figures were generated using MOLMOL⁵⁸ and PyMOL.⁵⁹

Table 1.

NMR restraints and structural statistics for the GABPα³⁵⁻¹²¹ ensemble

Summary of restraints
NOEs
Unambiguous		2745
Ambiguous		923
Total		3668
Dihedral angles
ϕ, ψ, χ1		51, 51, 34
Residual dipolar couplings (1HN-15N)		52
Deviation from restraints
NOEs (Å)		0.027±0.001
Dihedral angles (deg.)		0.85±0.06
Residual dipolar couplings (Hz)		1.81±0.02
Deviation from idealized geometry
Bond lengths (Å)		0.005±0.000
Bond angles (deg.)		0.73±0.01
Improper angles (deg.)		2.29±0.08
Residues in allowed region of Ramachandran plot		98.1%
Mean energiesa, kcal·mol−1
Evdw		−168±20
Ebonds		35±2
Eangles		487±4
Eimpr		138±11
ENOE		121±7
Ecdih		7±1
Esani		171±6
Residual dipolar couplingsc
Q value		0.186 ± 0.013
Q′ value		0.194 ± 0.015
RMSD from average structure, Å
	Backbone	Heavy atoms
Structured residuesb	0.15±0.02	0.41±0.02
All residues	0.59±0.09	0.97±0.06

^aFinal ARIA/CNS energies for van der Waals (vdw), bonds, angles, NOE restraints, dihedral restraints (cdih), and residual dipolar coupling restraints (sani).

^bStructured residues: 36−43, 48−59, 67−70, 73−74, 91−99, 107−115

^cQ = rms(D^obs-D^calc)/rms(D^obs) and Q′ = rms(D^obs-D^calc)/(D_a²[4 + 3R²]/5)^1/2, as defined in ref 60.

Complementing the structural analysis, the global and internal backbone dynamics of GABPα³⁵⁻¹²¹ were studied by ¹⁵N heteronuclear relaxation experiments. The isotropic tumbling time, τ_c, of the fragment was calculated to be 5.9 ± 0.04 ns at 30 °C from the T₁ and T₂ relaxation values of well-ordered amides in the OST domain.¹⁷ This is consistent with that expected for a globular, monomeric 10.2 kDa protein.¹⁸ The internal backbone dynamics of GABPα³⁵⁻¹²¹ were fit using the anisotropic Lipari-Szabo model-free approach (Supplemental Figure S3).^19,20,21 The structured core of the protein, corresponding to the minimal OST domain (residues 36−115), is well ordered, and no significant changes in backbone mobility were found for the bulge in S₅ or within the distorted helix. In contrast, besides the N- and C-terminal tails, there are three relatively flexible loop regions: residues 61−63 between the helix and strand S₂, residues 87−91 between strands S₃ and S₄, and residues 101−104 between strands S₄ and S₅. Among these, the exposed S₄/S₅ loop has the highest degree of flexibility on the psec-nsec timescale, as revealed by reduced heteronuclear NOE values and model free order parameters. This region also has relatively high backbone RMS deviations in the structural ensemble (Figure 2 and Supplemental Figure S3).

Structural comparison of the OST domain with ubiquitin

To obtain possible insights into the function of the OST domain, searches for proteins with similar three dimensional structures were performed using DALI and VAST.^22,23 Significant matches were found only with ubiquitin (1UBI.pdb, Z = 4.0 for 61 residues aligned) or ubiquitin-like proteins (Yeast uniquitin-like protein Hub1, 1M94.pdb, Z = 4.0 for 57 residues; hypothetical Arabidopsis thaliana protein At1g16640, 1YEL.pdb, Z = 3.0 for 64 residues; mouse D7Wsu128e protein, 1V86.pdb, Z = 3.0 for 62 residues). Indeed, despite their disparate sequences (18% identity using ClustalW²⁴), the global fold of ubiquitin is strikingly similar to that of the OST domain with five stranded β-sheet crossed by a helix (Figure 3). However, upon closer inspection, several key differences exist between the two proteins. First, the linear order of their secondary structure elements is permuted such that ubiquitin has a helix between β-strands S₂ and S₃, whereas in the OST domain, the helix lies between S₁ and S₂. Second, relative to the remainder of the β-sheet, the central β-strand S₅ in ubiquitin has the opposite orientation as that in GABPα. Third, unlike ubiquitin with a regular α-helix, the helix in the OST domain begins with a 3₁₀ conformation (Ile48 to Asn50), and ends at Leu59 as a distorted α-helix (Supplemental Figure S2). Finally, the OST domain has loops corresponding to the two short 3₁₀ helices in ubiquitin, and its flexible S₄/S₅ loop matches the C-terminal tail of the latter. Based on these differences, and the lack of similarity to any other proteins of known structure, the OST domain appears to represent a new fold.

Figure 3.

Structure comparisons of the GABPα OST domain and ubiquitin (1UBQ.pdb). (a) The global folds of the two proteins are clearly similar. (b) However, secondary structure topology diagrams show that the direction of β-strand S5 in the OST domain is opposite from that of the corresponding strand in ubiquitin. (c) The sequential orders of secondary structural elements are also different. In addition, ubiquitin has a regular α-helix and 2 short 3₁₀-helices, whereas the OST domain has a distorted α-helix preceded by a 3₁₀-helix.

Surface properties of the OST domain

The lack of any structural similarities beyond that of the ubiquitin-like proteins precluded obvious clues to the function of the GABPα OST domain. Furthermore, no additional insights were provided by the ProFunc server, which attempts to identify the biochemical role of a protein from its three-dimensional structure.²⁵ Following the hypothesis that the OST domain mediates macromolecular interactions necessary for the regulation of transcription by GABP, we examined its surface properties. The amino acid composition of GABPα³⁵⁻¹²¹ has a relatively low predicted pI ∼ 4.5, and thus at neutral pH, is net negatively-charged. Most surface exposed Asp and Glu residues localize in two clusters on the opposite sides of the OST domain (Figure 2(d)). The largest of these is formed by carboxylate sidechains from the loops between S₁/H (Asp43, Glu46), S₃/S₄ (Asp76, Asp78, Asp83, Asp89), and S₄/S₅ (Glu105), and at the start of S₂ (Glu67). Although less prominent, the second cluster arises from residues in S₅ (Glu112) and at the C-terminal end (Glu118 and Glu121) of GABPα³⁵⁻¹²¹. In contrast, no substantial regions of positively-charged or hydrophobic residues are present on the surface of the OST domain. This suggests that the GABPα OST domain may interact with positively-charged partner proteins, or other biological molecules, at least in part through electrostatic interactions.

The GABPα OST domain interacts with the CBP CH1 and CH3 domains

Several previous studies have demonstrated that GABPα interacts with the general transcriptional co-activator CBP/p300.^7,10,15,16 Thus, following a best candidate approach, we sought to determine if the OST domain and/or PNT domain contribute to this interaction by using pulldown assays with GST-tagged fragments corresponding to various regions of mCBP. As shown in Figure 4(f), the OST domain interacted with mCBP fragments containing the CH1 (or TAZ1; residues 340 − 439) and CH3 (or ZZ-TAZ2; residues 1680 − 1850) domains. Binding was not observed to other tested regions containing the nuclear receptor (residues 1−101) or KIX (residues 576−679) domains. The isolated PNT domain only bound to the CH3 domain (Figure 4(g)). Consistent with these results, GABPα¹⁻³²⁰, containing both the OST and PNT domains, also bound to CH1 and CH3 containing fragments of mCBP (Figure 4(h)). Binding of the OST domain to the mCBP fragment containing the CH1 domain was significantly reduced upon increasing the concentration of KCl from 180 to 300 mM (not shown), indicating an electrostatic contribute towards their association.

Figure 4.

The GABPα OST domain binds to the mCBP CH1 and CH3 domains. (a,c) Superimposed ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled OST domain before (black, 0.1 mM) and after (red, 0.09 mM) addition of unlabeled CH1 (0.09 mM final). The top quartile of residues with backbone or sidechain amides showing intensity decreases due to CH1 domain binding are mapped in green on to the surface of the OST domain. (b,d) Superimposed ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled CH1 before (black, 0.1 mM) and after (red, 0.05 mM) addition of unlabeled OST domain (0.05 mM final). The top quartile of residues showing intensity decrease due to OST domain binding are mapped in green on to the surface of the CH1 domain (1U2N.pdb). (e) Superimposed ¹H-¹⁵N HSQC spectra of ¹⁵N-labeled PNT domain before (black, 0.1 mM) and after (red, 0.1 mM) addition of unlabeled CH1 (0.1 mM final). The absence of any spectral changes indicates that no measurable binding occurs between the PNT and CH1 domains. These conclusions are supported by pulldown assays, involving 200−300 pmol of the indicated GST-tagged mCBP fragments, pre-bound to glutahionine-sepharose beads (20 μL), and 5 μM of FLAG-HMK-tagged (f) GABPα³⁵⁻¹²¹ (OST domain), (g) GABPα^168−256 (PNT domain), and (h) GABPα¹⁻³²⁰ (OST and PNT domains). Following bead collection and washing, binding was analyzed by Western blotting using an anti-FLAG antibody. Approximate molecular weight marker positions on the 4−15% SDS-PAGE gels are shown. The slower migrating band in the GABPα³⁵⁻¹²¹ sample is attributed to disulfide-linked dimers. (i) Cartoon of the domain structure of CBP.

NMR Methods: The gene encoding the His₆-tagged CH1 domain (mCBP^340−439) was PCR cloned from the full length CBP gene (Genbank 70995311) into the pET28b (Novagen) vector for expression in E. coli BL21(λDE3) cells. Cells were grown at 30 °C in ZnSO₄ supplemented (150 μM) LB or ¹⁵N-minimal M9T medium to OD₆₀₀ ∼ 1, followed by induction with 1 mM IPTG and growth for overnight at 16 °C. Purification was as in Figure 1, except that buffer A was modified to 150 μM ZnSO₄ , 5 mM imidazole, 50 mM Tris (pH 7.5), 500 mM NaCl, and 5 % glycerol. The purified CHI (mCBP^340−439), OST (GABPα³⁵⁻¹²¹), and PNT (GABPα^168−256) domains were dialyzed into NMR buffer (20 mM Tris (pH 6.9), 50 mM NaCl, 2 mM DTT, ∼10 % D₂O) and concentrated by ultrafiltration. ¹H-¹⁵N HSQC spectra were recorded at 25 °C.

GST-pulldown Methods: A DNA cassette encoding the FLAG tag and heart muscle kinase (HMK) target sequences was inserted into pET28a NdeI site of the GABPα constructs to generate clones for His₆-FLAG-HMK-GABPα³⁵⁻¹²¹, His₆-FLAG-HMK-GABPα^168−256, and His₆-FLAG-HMK-GABPα¹⁻³²⁰. Tagged fragments were purified as described for GABPα³⁵⁻¹²¹ in Figure 1. GST pulldown assays were conducted in 20 mM Tris pH 7.9, 10% glycerol, 180 mM KCl, 0.2 mM EDTA, 10 μM ZnSO₄, 0.5% NP40, 0.5 mM PMSF, 2−10 mM DTT, 0.1 mg/mL BSA, with Glutathione-Sepharose 4B beads (GE Healthcare). After incubation for 16 hr at 4 °C with constant rotation, the beads were collected by centrifugation and washed with binding buffer. The bound proteins were eluted using 3x SDS sample buffer, separated by SDS-PAGE, transferred to PVDF membranes, blotted with a 1:20,000 dilution of FLAG M2 antibody (Sigma), and visualized by ECL Plus (GE Healthcare). Panels (f), (g), and (h) are representative immoblots from experiments repeated four times. An SDS-PAGE gel of the protein samples used in these assays is provided in Supplemental Figure S4.

The specific binding of the OST and CH1 domains was confirmed by reciprocal ¹H-¹⁵N HSQC monitored titration experiments. In each case, addition of the unlabeled partner led to a progressive decrease in the signal intensities of selected backbone or sidechain amides in the ¹⁵N-labeled domain (Figure 4(a-b)). Such binding in the intermediate-slow exchange regime on the chemical shift timescale is generally indicative of a dissociation constant K_d in the micromolar or lower range. However, the absence of new signals from the resulting complex upon saturation is suggestive of conformational exchange broadening and/or formation of species exceeding a 1:1 stoichiometry. Mapping of the intensity perturbations onto the tertiary structures of the OST and CH1 domains provides a qualitative identification of their respective binding interfaces (Figure 4(c-d)). For the OST domain, this interface partially overlaps its predominant negatively-charged surface patch and includes the flexible S₄/S₅ loop. In the case of the CH1 domain, the interface abuts a positively-charged region of its surface and partially overlaps the known binding sites for the transactivation domains of HIF-1α and CITED2.^26,27,28,29 It is also noteworthy that, at least for the OST domain, residues showing spectral perturbations are located over a rather large region of its surface. This could result indirectly through conformational changes upon CH1 binding, or might reflect multiple binding sites leading to higher order complex formation. Further structural, thermodynamic, and mutational studies of the interactions of the OST and CH1 domains are needed to resolve these possibilities. In contrast to the OST domain, the ¹H-¹⁵N HSQC spectra of the PNT domain remained completely unperturbed upon titration with the CH1 domain, confirming the lack of any significant interaction between these species (Figure 4(e)). Also, neither the OST or PNT domain bound to the isolated ZZ domain³⁰ of mCBP in NMR-monitored titrations (not shown), suggesting that their association with the CH3 fragment is dependent upon the TAZ2 domain.³¹ Studies to test this hypothesis are underway.

Concluding Remarks

Through deletion studies followed by NMR spectroscopic analyses, we have identified and structurally characterized a previously unrecognized domain in GABPα. Similar to the ETS and PNT domains, the OST domain adopts a stable folded structure when isolated. Importantly, the near exact superimposition of signals from the corresponding amides in the ¹H-¹⁵N HSQC spectra of GABPα³⁵⁻¹²¹ and GABPα¹⁻³²⁰ (Figure 1(b)) demonstrates that the OST domain retains this independent structure within the context of larger GABPαfragments and that it does not interact intramolecularly with the PNT domain or any flanking residues. Similarly, no intermolecular interactions between the isolated OST and PNT domains were detectable through NMR-monitored titrations (not shown). Thus, at least in the context of GABPα¹⁻³²⁰, the OST domain behaves as a “bead-on-a-string” (Figure 1(c)). Although not tested explicitly due to challenges in obtaining suitable samples for NMR analysis, this conclusion likely extends to the full length protein with the independently folded ETS domain. Also, algorithms for identifying disordered protein regions, such as PrDOS,³² predict that the sequences flanking the three domains of GABPαare unstructured. Interestingly, it has been reported that, in the absence of DNA, GABPα/α exists as a stable heterodimer, and that deletion of the N-terminal two-thirds of GABPα, along with serine substitutions for four cysteines in the remaining ETS domain-containing C-terminal fragment, enhances heterotetramer formation.^33,34 This raises the possibility of additional GABPα/α interactions, beyond that between the ANK and ETS domains, which may involve the OST, PNT or other yet unidentified segments of these two proteins. Nevertheless, the “beads-on-a-string” model of GABPαprovides great flexibility for the assembly of a complex network of transcriptional regulatory proteins at target promoter sites.

The OST domain globally resembles ubiquitin, yet differs significantly in its secondary structural topology. Furthermore, based on sequence similarities, this domain can only be recognized in GABPα orthologs including Drosophila Elg. Thus, the OST domain defines a new protein fold, possibly with functions in mediating protein-protein interactions specific to this subset of the ETS transcription factor family. An inspection of the surface properties of the OST domain revealed two predominant clusters of negatively-charged Asp and Glu residues, suggestive of electrostatically-driven binding to positively-charged partner proteins.

Following a best-candidate approach, we demonstrated by GST-pulldown assays that the OST domain of GABPα binds to fragments of the co-activator CBP/p300 containing the CH1 and CH3 domains, whereas the PNT domain only interacts with the latter. NMR-monitored titrations confirmed that the OST domain, but not PNT domain, specifically binds the isolated CH1 domain of mCBP. In addition to serving as a scaffold for assembly of specific transcriptional factors, CBP/p300 recruits the basal transcription factors and functions as a histone acetyltransferase to direct chromatin remodelling. Consistent with the results of Figure 4, binding of full length GABPα to fragments of p300 spanning residues 1−596 and 1572−2370, but not 744−1571, was reported based on in vitro GST-pulldown studies; these contain the CH1, CH3, and CH2 domains, respectively.¹⁵ However, more recently published GST-pulldown experiments indicated that the GABPα PNT domain binds to the CH3 domain of p300, whereas residues 1−143 (encompassing the OST domain) bind weakly to both the CH2 and CH3 domains; no interactions with the CH1 domain were detected.¹⁰ Resolving the exact functions of these GABPα regions will require unbiased searches for interacting proteins and additional in vitro and in vivo analyses, carried out with the knowledge of the newly-discovered OST domain. For example, co-immunoprecipitation measurements revealed that upon neuregulin-activation, presumably though MAPK or JNK signalling cascades, GABP binds p300, whereas in the unactivated state, it binds HDAC1 at N box promoters.⁷

The CH1 and CH3 domains of CBP/p300 have been implicated in binding a remarkable array of diverse transcription factors, suggesting a significant degree of structural plasticity for intermolecular complex formation. Both of the net positively-charged CH1 (or TAZ1) and CH3 (or ZZ-TAZ2) domains fold as helical bundles with zinc ions chelated via cysteine and histidine residues.^30,31 To date, structures of the isolated CH1 domain of CBP/p300 in complex with peptides corresponding to the transactivation domains of CITED2 and HIF-1α have been determined using NMR spectroscopy.^{26,27,28,29,35} In each case, binding induces folding of the otherwise disordered peptides to predominantly helical structures that wrap around extended surface regions of the CH1 domain. In contrast, the OST domain is a β-sheet module with a well-defined structure. Similarly, phosphorylation-dependent binding of Ets-1 and Ets-2 to the CH1 and CH3 domains of CBP/p300 is mediated by their α-helical PNT domains.³⁶ We are currently undertaking NMR spectroscopic studies to determine the structural bases for the interactions of these ETS family members with this central component of the transcriptional machinery.

Data Bank accession codes

The atomic coordinates of the GABPα^(35−121) ensemble (accession code: 2JUO) have been deposited in the Research Collaboratory for Structural Bioinformatics Protein Databank (http://www.rcsb.org/pdb/), and the NMR chemical shift list (accession code: 15451) has been deposited in the BioMagResBank (http://www.bmrb.wisc.edu/).

Abbreviations

ANK

ankryin repeat

GST

glutathione S-transferase

HSQC

heteronuclear single quantum correlation

LZ

leucine zipper

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

NOESY

nuclear Overhauser effect spectroscopy

PNT

Pointed

RDC

residual dipolar coupling

RMS

root-mean-square

TAD

transactivation domain. GABPα and CBP fragments are indicated by the residues spanned, e.g. GABPα^(35−121) corresponds to residues 35−121 of the 454 amino acid protein.