Structural insights into TAZ2 domain–mediated CBP/p300 recruitment by transactivation domain 1 of the lymphopoietic transcription factor E2A

The E-protein transcription factors guide immune cell differentiation, with E12 and E47 (hereafter called E2A) being essential for B-cell specification and maturation. E2A and the oncogenic chimera E2A-PBX1 contain three transactivation domains (ADs), with AD1 and AD2 having redundant, independent, and cooperative functions in a cell-dependent manner. AD1 and AD2 both mediate their functions by binding to the KIX domain of the histone acetyltransferase paralogues CREB-binding protein (CBP) and E1A-binding protein P300 (p300). This interaction is necessary for B-cell maturation and oncogenesis by E2A-PBX1 and occurs through conserved ΦXXΦΦ motifs (with Φ denoting a hydrophobic amino acid) in AD1 and AD2. However, disruption of this interaction via mutation of the KIX domain in CBP/p300 does not completely abrogate binding of E2A and E2A-PBX1. Here, we determined that E2A-AD1 and E2A-AD2 also interact with the TAZ2 domain of CBP/p300. Characterization of the TAZ2:E2A-AD1(1–37) complex indicated that E2A-AD1 adopts an α-helical structure and uses its ΦXXΦΦ motif to bind TAZ2. Whereas this region overlapped with the KIX recognition region, key KIX-interacting E2A-AD1 residues were exposed, suggesting that E2A-AD1 could simultaneously bind both the KIX and TAZ2 domains. However, we did not detect a ternary complex involving E2A-AD1, KIX, and TAZ2 and found that E2A containing both intact AD1 and AD2 is required to bind to CBP/p300. Our findings highlight the structural plasticity and promiscuity of E2A-AD1 and suggest that E2A binds both the TAZ2 and KIX domains of CBP/p300 through AD1 and AD2.

TAZ2 domain of CBP/p300. Characterization of the TAZ2:E2AAD1(1-37) complex indicated that E2A-AD1 adopts an α-helical structure and uses its ϕ-x-x-ϕ-ϕ motif to bind TAZ2. While this region overlapped with the KIX recognition region, key KIX-interacting E2A-AD1 residues were exposed, suggesting that E2A-AD1 could simultaneously bind both the KIX and TAZ2 domains. However, we did not detect a ternary complex involving E2A-AD1, KIX, and TAZ2 and found that E2A containing both intact AD1 and AD2 is required to bind to CBP/p300. Our findings highlight the structural plasticity and promiscuity of E2A-AD1 and suggest that E2A binds both the TAZ2 and KIX domains of CBP/p300 through AD1 and AD2.
The innate and adaptive immune systems rely on the development and differentiation of hematopoietic stem cells to mature blood cells, including B and T lymphocytes, natural killer cells, and plasmacytoid dendritic cells (1)(2)(3). The generation of these immune cells, or lymphopoiesis, involves numerous intermediates, a progressive limitation of differentiation potential, and a coincident loss of the ability to self-renew (4). This process of lineage commitment is tightly regulated at the transcriptional level by complex regulatory networks comprising both lineagespecific and ubiquitous transcription factors (2,(5)(6)(7). One such set of transcription factors is the Eprotein family, which comprise class I basic helixloop-helix (bHLH) transcription factors that play essential roles in the development and specification of B-and T-lymphocytes (8)(9)(10).
Members of the E-protein family include the alternatively spliced isoforms E12 and E47 (also referred to collectively as E2A), HEB, and E2-2. Each family member contains a C-terminal bHLH domain responsible for E-protein dimerization and binding DNA at E-box CANNTG consensus sites in gene enhancer and/or promoter regions (11)(12)(13)(14)(15)(16)(17). In addition, the E-proteins possess three conserved activation domains, one of which (AD1) is positioned at the extreme N-terminus while the other two (AD2 and AD3) are more centrally located (Fig. 1A) (14,15,18,19). These activation domains have been shown to display cooperative, or independent transcriptional regulatory functions in a cell-specific manner (19)(20)(21)(22)(23)(24)(25). For example, AD1, AD2, and AD3 independently induce transcriptional activation but bind the same site of co-activators in a redundant manner, and when combined AD1 and AD2 cooperate to display greater than additive gene induction (19,21,22,(26)(27)(28). Deletion of AD1 or AD2 in E2A abolishes B-lymphoid differentiation in a pre-B cell line, which is consistent with the observation that E2A plays a critical regulatory role at the earliest stages of B-lymphoid specification (21). The transcriptional regulatory role of these activation domains resides in their ability to recruit general transcriptional factors and coactivators, such as TFIID and the histone acetyltransferases SAGA, GCN5, PCAF, the paralogs CBP and p300, and corepressors such as ETO (19,22,23,26,(29)(30)(31). Whereas AD3 allows the E-proteins to recruit TFIID to the core promoter by binding the TAFH domain of TAF4 (19), both AD1 and AD2 interact with CBP/p300 to enhance the acetyltransferase activity of this cofactor (26)(27)(28)32). A conserved region within AD1 called the 'p300/CBP and ETO target in E proteins' (PCET) motif, is also the target of the transcriptional repressor ETO, and competition between ETO and CBP/p300 for binding to the PCET motif has been proposed to be the mechanism underlying E-protein mediated transcriptional silencing (31).
CBP/p300 are multimodular proteins in which several protein-protein interaction domains allow recruitment to enhancers and promoters via interactions with the activation domains of an array of transcription factors (Fig. 1A) (33)(34)(35). AD1 and AD2 of the E-proteins have been reported to bind the same surface of the KIX domain of CBP/p300 via their Φ-x-x-Φ-Φ sequences (where Φ represents a hydrophobic amino acid and x any other amino acid), which comprise the core PCET motif (Fig.  1B). This interaction has been shown to be particularly important for leukemia induction by the oncogenic protein E2A-PBX1 that arises from a t(1;19) chromosomal translocation (20,27,28). In addition to the KIX domain, the TAZ1 and TAZ2 domains of CBP/p300 have been shown to bind activation domains of several transcription factors, including B-Myb, C/EPBε, FOXO3a, HIF-1α, p53, p63, p73, STAT1, and STAT2; many of which also contain Φ-x-x-Φ-Φ sequences (Fig. 1B) (36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47). Furthermore, several activation domains have displayed binding promiscuity to the KIX, TAZ1, and TAZ2 domains presenting the opportunity for a multivalent mode of binding with transcription factors comprising multiple activation domains; with p53 representing the archetypical example (33,38,48,49). Despite functional roles for the three activation domains being attributed to the transcriptional activity of E-proteins, their interactions with CBP/p300 have not been fully explored.
Here, we characterize a direct interaction between E2A-AD1 and the TAZ2 domain of CBP/p300, and elucidate the structure of a E2A-AD1(1-37):TAZ2 complex by NMR spectroscopy. The structure shows residues throughout and adjacent to the helical PCET motif of E2A-AD1 interact with the TAZ2 domain in a manner reminiscent of activation subdomain 2 (AD2) of p53. Peptide microarray and mutagenesis revealed the requirement of both hydrophobic and electrostatic interactions in complex formation. These results provide a mechanistic rational for the cooperative manner in which the AD1 and AD2 domains of E2A induce gene expression.
The interaction of E2A-AD1(1-37) with TAZ2 was further characterized by NMR spectroscopy because it is more feasible to characterize smaller proteins. Similar to what was observed for E2A-AD1, addition of saturating amounts of unlabeled TAZ2 domain to 15 N labeled E2A-AD1(1-37) caused large chemical shift changes and increased spectral dispersion for resonances corresponding to the backbone amide groups of Lys14-Phe26 ( Fig.  3B; red vs. black spectra). Chemical shift analysis of backbone 1 H, 13 C, and 15 N chemical shifts of free and TAZ2-bound E2A-AD1(1-37) suggest that Asp13-Met25 undergo a random coil to α-helix structural change upon TAZ2 binding (Fig. S5). Steady-state { 1 H}-15 N heteronuclear NOE analysis further indicated that when bound to the TAZ2 domain, E2A-AD1(1-37) adopted a more ordered structure between Lys14 and Leu28 (i.e., { 1 H}-15 N NOE values > 0.4) while the preceding N-terminal residues and the seven C-terminal residues displayed reduced { 1 H}-15 N NOE values suggestive of a disordered conformation in solution (Fig. 3C).
When compared to the structure of HEB-AD1 bound to the KIX domain (27), E2A-AD1  was extended by one helical turn at the N-terminus (residues Gly11-Lys14).

E2A-AD1:TAZ2 interface
The E2A-AD1 interactive surface on the TAZ2 domain involved a hydrophobic groove bounded by basic residues and covered ~1260 Å 2 of solvent accessible surface area (Fig. 5A,B). The hydrophobic groove comprised the aliphatic region of Arg1732, Ile1735 and Ala1738 of α1, the aliphatic region of Lys1760, Met1761 and Val1764 of α2, and Pro1780, Ile1781, Gln1784, Leu1785, Leu1788 and Tyr1791 of α3 (Fig. 5C). Residues throughout the helical region of E2A-AD1, including Leu19, Phe22, Met25 and Phe26, and the residues Pro27 and Leu28 C-terminal to the helix made extensive non-polar contacts with the TAZ2 domain ( Fig. 5D). Within the E2A-AD1 helix, the side chain of Leu19 inserted into a cavity formed by Val1764, Pro1780, Ile1781, the aliphatic region of Gln1784, and Leu1785. The aromatic side chain of Phe22 from E2A-AD1 participated in van der Waals contacts with the aliphatic portion of Lys1760, Met1761, and Val1764. Met25 interacted with Ala1738 and Leu1788 while Phe26 associated with Ile1735, Ala1738, Ala1787, Leu1788, and the aliphatic region of Gln1784. Beyond the helical region of E2A-AD1, Pro27 formed non-polar contacts with Ile1735 while Leu28 resided in hydrophobic pocket created by aliphatic region of Arg1732, Ile1735, Leu1788, and Tyr1791 of the TAZ2 domain.

KIX and TAZ2 of CBP/p300 compete for E2A-AD1
The orientation of E2A-AD1 on the TAZ2 surface presented the possibility of a higher-order interaction with the KIX domain of CBP/p300 as several of the KIX-interactive residues, including the critical Leu20 (20), were solvent exposed; a scenario supported by the recent observation that full-length CBP/p300 displays intrinsic conformational flexibility (33,55). To directly assess the formation of such a complex or whether KIX and TAZ2 displayed exclusive E2A-AD1 binding, an NMR-based displacement experiment was performed. Addition of unlabeled E2A-AD1(1-37) to uniformly 15 N-labeled KIX resulted in significant chemical shift changes for a subset of backbone amide resonances in 15 N-KIX spectra (Fig. 7A), consistent with the site-specific binding of E2A-AD1(1-37) to KIX (27). Subsequent addition of unlabeled TAZ2 to the sample caused those backbone resonances of uniformly 15 Nlabeled KIX to revert back towards chemical shift values consistent with the free form of the KIX domain (Fig. 7B). These results indicate that the TAZ2 and KIX domains of CBP/p300 can only interact with E2A-AD1 exclusive of one another.

Discussion
The E2A activation domains are essential for lymphopoiesis and are involved in the onset of ALL through the oncogenic fusion protein E2A-PBX1 (56). In either situation, the recruitment of CBP/p300 is essential for E2A or E2A-PBX1 function. Here we investigate recruitment of CBP/p300 by E2A-AD1, which provides insight into the molecular mechanisms underlying lymphopoiesis and ALL.
Depending on the context, the activation domains of E2A function independently, redundantly, or cooperatively with each other (19,(21)(22)(23)(24)(25)(26)(27)(28)57). The KIX domain of CBP/p300 binds both E2A-AD1 and E2A-AD2 in a functionally redundant manner, with both E2A-AD1 and E2A-AD2 competing for the same site of KIX (27,28). The additional interactions of TAZ2 with E2A-AD1 or E2A-AD2 that we describe ( Fig. 2 and S2) raise the possibility that both AD1 and AD2 could simultaneously interact with the KIX and TAZ2 domains to allow for higher affinity association between CBP/p300 and E2A. This possibility is supported by our pulldowns of full-length CBP (Fig. 8), where intact AD1 and AD2 were required for maximum CBP/p300 pulldown and mutations in either the AD1 or AD2 region lessened the interaction by ~50%. It is unclear what the binding preferences of E2A-AD1 and E2A-AD2 are to full length CBP since E2A-AD1 and E2A-AD2 are able to bind both the KIX and TAZ2 domains. The ability of E2A to form a tight yet dynamic complex with CBP through multiple weaker interactions is likely essential for its transcriptional activity and provides rationale for how deletion of AD1 or AD2 abolishes B-lymphoid differentiation (21). This is reminiscent of p53, which also associates with multiple domains of CBP/p300 to form a tight, yet dynamic, complex (49).
The promiscuous interactions of intrinsically disordered proteins are recognized as key to their roles as protein interaction hubs (33,58). This appears to be the case with E2A since E2A-AD1 and its nearly identical homologue HEB-AD1 (Fig.  1B) can complement and bind a variety of molecular surfaces, with high-resolution structures available of the KIX:HEB-AD1 and eTAFH:HEB-AD1 complexes. HEB-AD1 adopts an amphipathic α-helix from Lys14-Phe26 (renumbered according to the E2A sequence) when bound to KIX, a kinked α-helix spanning residues Asp13-Leu20 and Ser23-Phe26 when bound to eTAFH, and E2A-AD1 adopts an amphipathic α-helix from Asp13-Met25 when bound to TAZ2. Leu16, Leu19, and Leu20 form essential hydrophobic contacts with both eTAFH and KIX (27,59), while Phe22 forms additional hydrophobic contacts with KIX and TAZ2 but not eTAFH. Although the entire PCET motif is involved in recognition of eTAFH, KIX, and TAZ2, Phe26, Pro27 and Leu28 have additional hydrophobic contacts to TAZ2. Finally eTAFH, KIX, and TAZ2 have basic residues situated nearby the E2A-AD1 binding site, providing unique potential electrostatic contacts for Glu15, Asp18 and Asp21. The different modes of E2A-AD1 binding to eTAFH, KIX, and TAZ2 are highlighted by the importance of Leu20, which is essential for the HEB:KIX, and HEB:eTAFH interactions (27,59), but dispensable for binding to TAZ2 as it is facing solvent and alanine substitutions are non-perturbing to a pulldown experiment ( Fig. 5 and 6).
The TAZ2 domain of CBP/p300 is promiscuous and interacts with many intrinsically disordered proteins, with structures available for TAZ2 in complex with activation domains from E1A, C/EBPε, p53, p63, p73, and STAT 1 (36,38,39,41,42,47). Highlighting the promiscuity of the TAZ2 domain, E1A, p73, and STAT 1 bind TAZ2 in part to a hydrophobic groove formed by α1, α2, and α3, while p53, p63, and E2A-AD1 adopt an α-helical structure when they bind to this same region of TAZ2 (Fig. 9). The adenoviral protein E1A has been shown to decrease E2A transcriptional activity (60) and the observation that E2A-AD1 and E1A bind overlapping sites of TAZ2 suggests that E1A inhibits E2A function through direct competition for TAZ2. Interestingly, although E2A-AD1, p53, and p63 bind the same region of TAZ2 as an αhelix, the peptides have differing orientations ( Fig.  9 and 10). Despite the opposite orientations of p53 and E2A-AD1, Phe22 and Phe26 of E2A-AD1 occupy similar positions as Phe54 and Ile50, respectively (Fig. 10). These residues all face TAZ2, suggesting that an amphipathic helix is key for binding to the α1, α2 and α3 binding surface of TAZ2. The importance of these hydrophobic residues is confirmed by our mutagenesis studies in which mutation of Phe22 or Phe26 to alanine decreased the signal observed in a microarray, while mutation of I50 and Phe54 of p53 weakens binding of p53-AD2 to TAZ2 by approximately 3 fold and 2 fold, respectively (44). Other than the presence of a Φ-x-x-Φ-Φ sequence and being acidic with some hydrophobic residues, the sequences of TAZ2-binding activation domains are quite divergent and TAZ2 appears to be able to accommodate a variety of peptide sequences, orientations, and secondary structure content.
The orientation of an amphipathic α-helix bound to TAZ2 likely depends on the specific interactions of polar residues and neighbouring sequences with TAZ2. For E2A-AD1 Asp18 and Asp21 are likely involved in a salt bridge to the sidechain of Lys1760 (N ζ -O δ distance of 3.1 ± 0.53 Å and 3.6 ± 0.9 Å, respectively), while Phe26, Pro27 and Leu28 all participate in further hydrophobic contacts to TAZ2. For p53 a similar situation occurs where Glu11 and Glu17 contact the guanidinium group of Arg1731 and multiple other residues engage in specific polar contacts to TAZ2 (41). In support of this, substitution of Asp18 and Asp21 with alanine decreases TAZ2 binding in a peptide microarray (Fig. 6). Given that acidic residues influence E2A-AD1 affinity for TAZ2, phosphorylation of E1A (e.g. at Thr12, Ser17 and Ser23) may enhance the affinity of E2A-AD1 to TAZ2 by forming salt bridges to nearby basic residues of TAZ2. A similar situation has been observed with p53, which has a graded enhancement of affinity for CBP/p300 upon phosphorylation (61). Overall, TAZ2 is able to accommodate a wide variety of peptides through a complex interplay of hydrophobic and electrostatic interactions, and phosphorylation may be a common method to regulate protein association with TAZ2.
Protein expression and purification. The isolated KIX domain (residues 586-673 of CBP) was expressed and purified as previously described (27). E. coli BL21(DE3) cells harbouring the His6-GB1-TAZ1 and His6-GB1-TAZ2 encoding plasmids were grown in LB or 15 N-or 13 C/ 15 Nenriched M9 media at 37°C supplemented with 100 μM ZnCl2. Protein expression was induced at optical density of 0.6 at 600 nm by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM. Growth was continued overnight at 23°C with shaking. Harvested cell pellets were lysed by sonication in denaturing binding buffer (20 mM Tris-HCl pH 8.0, 250 mM NaCl, 8 M urea, 10 mM β-mercaptoethanol, 10 µM ZnCl2), clarified by centrifugation, and applied to Ni 2+ -affinity resin (GE Healthcare). Upon extensive washes with denaturing binding buffer containing 10 mM imidazole, protein constructs were refolded on-column by application of native Ni 2+ binding buffer (20 mM Tris-HCl pH 8.0, 250 mM NaCl, 10 mM β-mercaptoethanol, 10 µM ZnCl2), and subsequently eluted in the same buffer containing 300 mM imidazole. The elution fractions were pooled, and β-mercaptoethanol and ZnCl2 were added to final concentrations of 40 mM and fourfold excess relative protein, respectively. Thrombin, at 1 unit per 50 nanomoles of protein, was added and the samples were dialyzed overnight in buffer A (20 mM HEPES pH 8.0, 50 mM NaCl, 5 mM β-mercaptoethanol, 10 µM ZnCl2). The cleaved, refolded TAZ1 or TAZ2 constructs were separated from the His6-GB1 fragments via fast flow SP sepharose cation chromatography (GE Healthcare) using buffer A as a wash buffer, and eluted with buffer A containing 500 mM NaCl.
Expression of all His6-GB1-E2A constructs were induced in transformed E. coli BL21(DE3) cells with 0.5 mM IPTG at an optical density of ~1 at 600 nm. Growth was continued for an additional 4 hours at 37°C with shaking. Expression of His 6 -GB1-E2A(1-483) L397A/I401P mutants was performed as described except that after induction with IPTG, growth was continued for an additional 16 hours at 23°C with shaking. Purification of His6-GB1-E2A(1-483), E2A-AD1, and E2A-AD2 was performed as described above for the TAZ domains, with the exception that ZnCl2 was excluded from all buffers. Purification of wild-type and mutant E2A-AD1(1-37) constructs was performed as previously described (28).
For uniformly 13 C-and/or 15 N-labeled NMR samples high performance liquid chromatography was used to purify TAZ2 or E2A-AD1(1-37) on a C18 reverse phase column with a water:acetonitrile gradient with 0.05% trifluoroacetic acid. Fractions containing protein were pooled, lyophilized, and stored at -20°C. The integrity of each protein sample was verified by SDS-PAGE analysis, mass spectrometry, and NMR spectroscopy.
Cell culture and cell lysis. HEK 293T cells were seeded at 0.8 x 10 6 cells per well in a 6-well culture plate. The following day, the cells were transfected with 2 ug of a flag-CBP plasmid using jetPRIME transfection reagent (Polyplus, 114-01) according to the manufacturers protocol. After 24 hours, the transfection medium was exchanged for cell growth medium. Following 48 hours after transfection, the cells were washed twice with PBS and lysed for 10 minutes at room temperature with gentle agitation in NP-40 buffer (50 mM Tris pH 8.0, 150 mM NaCl, 1.0% NP-40) plus protease inhibitors (Thermo Fisher Scientific, A32965). The lysates were sonicated briefly and clarified by centrifugation.
Immunoblotting. In vitro pulldown experiments were performed as described above with the following modifications: i) His6-GB1 and His6-GB1-E2A(1-483) wild-type and mutant constructs were incubated with 5 uL of 50% IgG agarose slurry (GE Healthcare) for 30 minutes in NP-40 buffer, ii) 250 uL of flag-CBP transfected HEK 293T cell lysate was added to the beads and left for 3 hours with gentle agitation, iii) all washes were done with NP-40 buffer. The pull-down experiments were loaded onto 6% Tris-glycine gels and separated by SDS-PAGE. The proteins were transferred to nitrocellulose membranes using Bio-Rad wet electroblotting system for 16 hours at 20 V at 4°C. Membranes were blocked for 1 hour at room temperature with 5% skimmed milk in Trisbuffered saline plus 0.1% Tween 20 (TBST) and subsequently incubated for 1 hour at room temperature with horse radish peroxidaseconjugated α-FLAG antibody (Sigma Aldrich, A8592) diluted 1:5000 in TBST. Membranes were then washed four times with TBST, stained with Immobilin Forte Western HRP substrate (EMD Millipore, WBLUF0100), and exposed to X-ray films.
The pull-down experiments and immunoblotting were performed in triplicate and ImageJ was used to quantify the X-ray films.
Peptide microarray. Peptide microarrays were synthesized through automatic SPOT synthesis with a MultiPep automated peptide synthesis system (Intavis) using 9fluorenylmethyloxycarbonyl chemistry (62). HEB-AD1 derived peptides were synthesized onto continuous cellulose membranes to generate strip arrays, which were hydrated with ethanol, washed with assay buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl) and blocked overnight in assay buffer containing 2% milk. Following washes with assay buffer containing 0.5% Tween 20, the arrays were incubated for 4 hours with 5-10 µg/mL His6-GB1-TAZ2 in 50 mM Tris-HCl pH 7.4, in assay buffer containing 2% skim milk, washed with assay buffer, and incubated for 1 hour with rabbit polyclonal antibody against the His6 tag conjugated to horse radish peroxidase (Abcam) diluted in assay buffer containing 2% skim milk. After a final wash with assay buffer, luminata horseradish peroxidase substrate (Millipore) was applied to the arrays and X-ray film was used to detect the signal.
NOESY peaks lists were exported from CCPNMR Analysis and used in CYANA for automatic NOE assignment and distance restraint calibration (52,53). Dihedral angle restraints based on chemical shifts were generated using TALOS+, with angle error set to two times the error output of TALOS+ (65). After confirming a correct TAZ2 fold using only NOE-based distance and dihedral angle restraints, hydrogen bond restraints (1.8 ≤ dOH ≤ 2.2 Å; 2.7 ≤ dON ≤ 3.2 Å) were applied to helical regions of the E2A-AD1(1-37):TAZ2 complex while zinc coordination restraints were applied to known Zn 2+ coordinating residues (54). Using CYANA, an initial ensemble of the lowest energy 20 models was retained from 100 generated models. These 20 models were further energy minimized in explicit water using fmcGUI and CNS (51). The final ensemble of 20 models was validated using PROCHECK-NMR (66) and the recall, precision, F-measure and discriminating power scores of the final ensemble of the E2A-AD1(1-37):TAZ2 complex were calculated using CCPNMR Analysis (64,67). The protein structure validation software suite was used to determine the ordered residues and evaluate the quality of the final E2A-AD1(1-37):TAZ2 ensemble (68). The ensemble of 20 lowest-energy structural models was deposited to the Protein Data Bank (accession no. 2MH0) while 1 H, 13 C, and 15 N chemical shifts and restraints were deposited to the BMRB (accession no: 19610).