Recognition of the disordered p53 transactivation domain by the transcriptional adapter zinc finger domains of CREB-binding protein

Significance

The tumor suppressor p53 regulates the cellular response to genomic damage by recruiting the transcriptional coactivator cyclic-AMP response element-binding protein (CREB)-binding protein (CBP) and its paralog p300 to activate stress response genes. We report NMR structures of the complexes formed between the full-length, intrinsically disordered N-terminal transactivation domain of p53 and the transcriptional adapter zinc finger domains (TAZ1 and TAZ2) of CBP. Exchange broadening of NMR spectra of the complexes was ameliorated by using fusion proteins and segmental isotope labeling. The structures show how the p53 transactivation domain uses bipartite binding motifs to recognize diverse partners, reveal the critical interactions required for high affinity binding, and provide insights into the mechanism by which phosphorylation enhances the ability of p53 to recruit CBP and p300.

Abstract

An important component of the activity of p53 as a tumor suppressor is its interaction with the transcriptional coactivators cyclic-AMP response element-binding protein (CREB)-binding protein (CBP) and p300, which activate transcription of p53-regulated stress response genes and stabilize p53 against ubiquitin-mediated degradation. The highest affinity interactions are between the intrinsically disordered N-terminal transactivation domain (TAD) of p53 and the TAZ1 and TAZ2 domains of CBP/p300. The NMR spectra of simple binary complexes of the TAZ1 and TAZ2 domains with the p53TAD suffer from exchange broadening, but innovations in construct design and isotopic labeling have enabled us to obtain high-resolution structures using fusion proteins, uniformly labeled in the case of the TAZ2–p53TAD fusion and segmentally labeled through transintein splicing for the TAZ1–p53TAD fusion. The p53TAD is bipartite, with two interaction motifs, termed AD1 and AD2, which fold to form short amphipathic helices upon binding to TAZ1 and TAZ2 whereas intervening regions of the p53TAD remain flexible. Both the AD1 and AD2 motifs bind to hydrophobic surfaces of the TAZ domains, with AD2 making more extensive hydrophobic contacts consistent with its greater contribution to the binding affinity. Binding of AD1 and AD2 is synergistic, and structural studies performed with isolated motifs can be misleading. The present structures of the full-length p53TAD complexes demonstrate the versatility of the interactions available to an intrinsically disordered domain containing bipartite interaction motifs and provide valuable insights into the structural basis of the affinity changes that occur upon stress-related posttranslational modification.

The tumor suppressor p53 plays a central role in the cellular response to stress, functioning as an important signaling hub for the cellular response to various degrees of genomic damage and instability (1, 2). p53 is a multidomain protein that contains an N-terminal transactivation domain (TAD), a proline rich region, a core DNA-binding domain, a tetramerization domain, and a C-terminal regulatory domain (Fig. 1A). In unstressed cells, p53 binds to the E3 ubiquitin ligase mouse double minute protein 2 (MDM2), which mediates ubiquitination and degradation of p53 (3–6). In response to stress, p53 can be phosphorylated at more than 20 sites, 9 of which are located within the TAD (7). These modifications lower the affinity of the p53TAD for MDM2 and promote binding to the transcriptional coactivator cyclic-AMP response element-binding protein (CREB)-binding protein (CBP) and its paralog p300 (8–13). Binding of p53 to CBP/p300 facilitates acetylation of the C-terminal domain of p53 and further inhibits its degradation (14, 15). The interaction between p53 and CBP/p300 is also required for p53-mediated transcription and stabilization of the p53–DNA interaction (16–18). Four domains of CBP/p300 are involved in the interaction with the p53TAD: the folded transcriptional adapter zinc finger (TAZ) 1, KIX, and TAZ2 domains, and the molten-globular nuclear coactivator-binding domain (NCBD) (Fig. 1B) (16–25). In addition, the CBP/p300 bromodomain binds to the C-terminal regulatory domain after acetylation at K382 by the CBP HAT domain in response to DNA damage (26). The unphosphorylated p53TAD binds to the CBP/p300 domains with a broad range of affinities, with the strongest binding to TAZ2 (0.026 μM) (24, 25). It has been hypothesized that each of the four TADs of the p53 tetramer may bind to a separate domain of a single CBP/p300 molecule, resulting in an avidity effect that further stabilizes the p53–CBP/p300 complex and enhances p53-mediated transcription (24, 25). The p53TAD, which is disordered throughout its length in the absence of binding partners (27, 28), is bipartite and contains two activation subdomains, between residues 1–39 and 40–61, which function both synergistically and differentially in mediating p53 function (29–31). The activation subdomains contain short amphipathic interaction motifs (32–34), termed AD1 (residues 18–26) and AD2 (residues 44–54) (Fig. 1A), that have a weak propensity for transient helical secondary structure in the unbound state (32–34) and are frequently observed to fold into stable amphipathic helices upon binding to their partners (35–39).

Fig. 1.

Schematic diagrams showing the domain structure of (A) p53 and (B) CBP. (C) Construct design for the TAZ2–p53TAD fusion protein. (D) Intein ligation scheme and final construct for the p53TAD–TAZ1 fusion protein. Blue colors denote the p53TAD, with the AD1 and AD2 consensus regions outlined in blue in A. Interaction and catalytic domains identified in CBP are shown in green for the nuclear receptor interaction domain (NRID), the KIX domain, the bromodomain (bromo), the Cys-His rich domain-2 (CH2), the histone acetyl transferase domain (HAT), the ZZ domain, and the nuclear coactivator binding domain (NCBD) and in red for TAZ1 and in yellow for TAZ2. Intein and other tag and linker sequences are shown in gray.

The transcriptional adapter zinc finger (TAZ) (40) domains of CBP and p300 frequently act as scaffolds for the binding-related folding of the transactivation domains of numerous transcription factors (41, 42). The TAZ1 and TAZ2 domains adopt similar folds that consist of four α-helices (α₁–α₄) stabilized by binding to three zinc atoms (43, 44). The largest structural differences between the two TAZ domains are the length of the α₁ helix, which is extended in TAZ1, and the orientation of the α₄ helix, which is rotated 180° between the two domains. Despite their topological similarity, TAZ1 and TAZ2 differ significantly in their amino acid sequence and are highly selective in their interactions, preferentially binding different sets of transactivation domains. Several structures of TAZ1 and TAZ2 complexes with transcriptional activation domains have been reported (45–52) and have yielded insights into modes of binding: TAZ1 primarily binds long, extended activation domains in deep grooves whereas TAZ2 binds tightly to discrete amphipathic helices through an exposed hydrophobic surface (49, 50).

Whereas many intrinsically disordered transactivation domains bind with high specificity to either TAZ1 or TAZ2, the p53TAD binds to both TAZ domains, with differing affinities (24, 25). Although the isolated AD2 region of p53 binds with much higher affinity than the isolated AD1, the AD2 and AD1 motifs synergize in the context of the full-length p53TAD to enhance the binding affinity (25). Moreover, the p53TAD can bind simultaneously to TAZ1 or TAZ2 (through AD2) and to MDM2 (through AD1) to form a ternary complex that promotes polyubiquitination and degradation of p53 (25). Structures have been reported for complexes of the isolated AD1 and AD2 motifs bound to TAZ2 (38, 39), but no structural data are available for complexes of the full-length p53TAD or for p53 complexes with TAZ1. To gain insights into the structural basis for CBP/p300 recruitment by p53, we have determined solution structures of the full-length p53TAD bound to the TAZ1 and TAZ2 domains of CBP, as well as the structure of the isolated p53AD2 domain bound to TAZ2.

Results

Design of Protein Constructs.

NMR structural studies of p53/CBP complexes are frequently hindered by conformational averaging and exchange broadening, making structural restraints difficult to obtain (25, 38, 53). Binding of the p53TAD to the TAZ1 (K_d 0.9 μM) and TAZ2 (K_d 0.026 μM) domains occurs in fast to intermediate exchange on the chemical shift timescale, resulting in weak or missing resonances as a result of exchange broadening (25). Additionally, the presence of secondary binding sites prevents the use of a large excess of one binding partner in an effort to drive the interaction into the slow-exchange regime (54). These difficulties have limited previous structural characterization of p53TAD:TAZ2 complexes to the isolated AD1 or AD2 subdomains, for which the resonances are less broadened because they exchange on a faster timescale (25, 38, 39).

To obtain insights into the molecular basis for recognition of TAZ1 and TAZ2 by the bipartite p53TAD, we attempted to determine NMR structures of p53(13–61) in a 1:1 complex with each of the TAZ domains. Because of exchange broadening and missing resonances, we were able to obtain only preliminary, low-resolution structures that revealed the location and N-to-C orientation of the AD1 and AD2 helices but could not provide atomic details of the interactions with TAZ1 and TAZ2. To circumvent this problem, we created fusion proteins, in which the two interacting partners are joined by a flexible linker. Fusion proteins enforce close proximity of the components of the complex, increasing the effective concentration and association rate and driving the binding process into the slow-exchange regime, thereby reducing line broadening and enhancing spectral quality (55, 56). On the basis of the preliminary structure of the nontethered complex, we designed a fusion protein containing the TAZ2 domain of mouse CBP (residues 1764–1855) joined to the p53TAD (residues 2–61) by a six-residue Gly-Ser repeat sequence (Fig. 1C). The improvement in the spectral quality of both the TAZ2 and p53TAD spectra is illustrated in Fig. 2, with complete spectra shown in Figs. S1 and S2. The backbone NH cross-peaks in the spectrum of the free p53TAD are poorly dispersed (Fig. S1), consistent with the disordered nature of the free domain, but are sharp and intense (cyan cross-peaks in Fig. 2 A and B and Fig. S1). The cross-peaks corresponding to the p53TAD in the TAZ2–p53TAD fusion protein are better dispersed (labeled black cross-peaks in Fig. 2 A and B and Fig. S1), showing that the p53TAD is folded in the complex. In contrast, many cross-peaks are broadened or missing from the spectrum of the complex obtained by mixing equimolar amounts of p53TAD and TAZ2 (red cross-peaks in Fig. 2 A and B and Fig. S1), and those cross-peaks that are visible are frequently not shifted to the same extent as the corresponding cross-peaks for the TAZ2–p53TAD fusion protein. Due to the presence of a weak (K_d 10 μM) secondary AD2-binding site on TAZ2, the cross-peaks continue to shift beyond a 1:1 mole ratio upon titration of p53TAD with TAZ2 (25, 54). Because the cross-peaks for a given residue in the various complexes shift in a linear fashion, we infer that the fusion protein construct does not cause a change in binding mode but stabilizes the complex with the p53 AD2 motif bound in its primary, high affinity site on TAZ2. An increase in spectral quality is also observed for the heteronuclear single-quantum coherence (HSQC) spectra of ¹⁵N-labeled TAZ2 in the fusion protein (black cross-peaks in Fig. 2C), compared with in the 1:1 complex with unlabeled p53TAD (magenta cross-peaks in Fig. 2C). The NMR spectra of the isotope-labeled TAZ2–p53(2–61) fusion protein are of sufficient quality for structure determination.

Fig. 2.

(A and B) Overlay of sections of the ¹⁵N-¹H HSQC spectra of free p53TAD (cyan), p53TAD with addition of one molar equivalent of TAZ2 (red), and of the TAZ2–p53TAD fusion protein (black). Cross-peaks belonging to the TAZ2 portion of the (uniformly labeled) TAZ2–p53TAD fusion protein are labeled with a red asterisk. (C) Overlay of ¹H-¹⁵N HSQC spectra of free TAZ2 (green), TAZ2 with addition of one molar equivalent of p53TAD (magenta), and of the TAZ2–p53TAD fusion (black). Arrows indicate chemical shift changes for individual cross-peaks. Complete spectra are shown in Figs. S1 and S2.

Fig. S1.

Overlay of the ¹H-¹⁵N HSQC spectra of free ¹⁵N-labeled p53(1–61) (blue), ¹⁵N-labeled p53(1–61) in complex with 1× unlabeled TAZ2 (red), and the TAZ2–p53(2–61) fusion (black). Weak peaks in the spectra arise from isomerism of several of the prolines in the p53TAD.

Fig. S2.

Overlay of the ¹H-¹⁵N HSQC spectra of free ¹⁵N-labeled TAZ2 (green), ¹⁵N-labeled TAZ2 in complex with 1× unlabeled p53(1–61) (magenta), and the TAZ2–p53 fusion (black).

To obtain additional evidence that our TAZ2–p53(2–61) fusion protein represented an accurate binding model, we also determined the solution structure of a protein containing TAZ2 (residues 1764–1855) fused to only the AD2 region of the p53TAD (residues 37–61) through a GSGSG linker. The ¹H-¹⁵N HSQC spectra obtained with this construct were of sufficiently high quality (Fig. S3) to allow us to determine the solution structure of this complex.

Fig. S3.

¹H-¹⁵N HSQC spectrum of the TAZ2–p53(37–61) fusion protein.

A uniformly labeled fusion protein with p53(13–61) tethered to the C terminus of TAZ1(340–439) gave similarly enhanced spectral quality compared with the complex formed between the isolated proteins. However, the TAZ1 and p53 resonances in the ¹H-¹⁵N HSQC spectrum of this construct (orange cross-peaks in Fig. 3A) were not sufficiently well-resolved for detailed structural analysis. To increase resolution and enable the use of isotope-edited NMR techniques, we used a system of split inteins to prepare a selectively labeled fusion protein (57). The constructs used to obtain selectively labeled p53TAD–TAZ1 fusion proteins are shown in Fig. 1D. Preliminary structures calculated using the limited nuclear Overhauser effect (NOE) constraints available for a nontethered 1:1 TAZ1:p53(13–61) complex indicated that the C terminus of the p53TAD is located close to the N terminus of TAZ1. The intein-mediated fusion proteins were therefore designed to place the p53 sequence at the N terminus and the TAZ1 sequence at the C terminus of the fusion protein; p53(1–61) was fused to the N-terminal intein and TAZ1 to the C-terminal intein. To obtain a selectively labeled complex in the form of a fusion protein, one intein fusion (isotopically labeled) was mixed with the other intein fusion (unlabeled). Splicing occurs as the two portions of the split intein coalesce, producing a p53TAD–TAZ1 fusion protein with a GSCFNGT linker sequence and with either the p53 or the TAZ1 component labeled with isotope. The p53TAD–TAZ1 fusion was then purified from the folded intein by-product, as described in Materials and Methods.

Fig. 3.

(A) Overlay of the ¹H-¹⁵N HSQC spectra of the uniformly labeled TAZ1–p53TAD (nonintein) fusion (orange), and the p53TAD–TAZ1 intein fusion construct segmentally labeled in the p53TAD portion (blue) and in the TAZ1 portion (black). (B and C) Overlay of HSQC spectra for free p53TAD (cyan), with one molar equivalent of TAZ1 (red) and of the p53-labeled p53TAD–TAZ1 fusion protein (black). Arrows indicate chemical shift changes for individual cross-peaks. Complete spectra are shown in Figs. S4 and S5.

The ¹H-¹⁵N HSQC spectra of the intein-mediated fusion proteins selectively enriched with ¹⁵N in TAZ1 (Fig. 3A, black) or in the p53TAD (Fig. 3A, blue) are shown in Fig. 3A, superimposed on the spectrum of the fully labeled nonintein fusion protein (Fig. 3A, orange). These spectra show clearly that the intein labeling method gives improved spectral quality and decreased spectral complexity. Fig. 3 B and C shows a comparison of regions of the ¹H-¹⁵N HSQC spectrum of free ¹⁵N-labeled p53TAD(1–61) (cyan) with the corresponding regions of ¹⁵N-labeled p53TAD(1–61) in a 1:1 complex with unlabeled TAZ1 (red) and of the selectively labeled ¹⁵N p53TAD fused to unlabeled TAZ1 (black) (complete spectra are shown in Figs. S4 and S5).

Fig. S4.

Overlay of the ¹H-¹⁵N HSQC spectra of free ¹⁵N-labeled p53(1–61) (blue), ¹⁵N-labeled p53(1–61) in complex with 1× unlabeled TAZ1 (red), and the p53(1–61)–TAZ1 intein fusion construct segmentally labeled in the p53TAD portion (black).

Fig. S5.

Overlay of the ¹H-¹⁵N HSQC spectra of free ¹⁵N-labeled TAZ1 (green), ¹⁵N-labeled TAZ1 in complex with 1× unlabeled p53(1–61) (orange), and the p53TAD–TAZ1 intein fusion construct segmentally labeled in the TAZ1 portion (black).

Structure Determination.

The structures of the TAZ2–p53(2–61), TAZ2–p53(37–61), and p53(1–61)–TAZ1 fusion proteins were determined using restraints derived from ¹⁵N-NOESY HSQC and ¹³C-NOESY HSQC spectra. NOEs between p53TAD and TAZ1 were identified using isotope-edited ¹³C-NOESY-HSQC spectra (58). Dihedral angle restraints were derived from ${}^{13}\text{C}_{\alpha },{}^{13}\text{C}_{\beta },\hspace{0.17em}{}^1\text{H}_{\alpha },\hspace{0.17em}{}^1\text{H}_{\mathrm{N}}$ and ¹⁵N chemical shifts using TALOS+ (59). Initial structures were calculated using CYANA with CANDID and were refined by restrained molecular dynamics, first in vacuo and finally using a generalized Born solvent model (60, 61). The NMR restraints and structure statistics are summarized in Table S1.

Table S1.

NMR statistics for TAZ2–p53TAD, TAZ2–p53AD2, and p53TAD–TAZ1 generalized Born AMBER structures

	TAZ2–p53TAD	TAZ2–p53AD2	p53TAD–TAZ1
NMR constraints
Distance constraints
Total NOE	2,292	2,330	2,359
Intraresidue	600	487	735
Sequential (|I − j| = 1)	539	552	534
Medium range (|I − j| < 4)	596	652	616
Long range (|I − j| > 4)	557	639	474
Interdomain	243	161	218
Dihedral angle restraints
ϕ	147	93	162
Ψ	101	77	96
χ1	38	0	42
Structural statistics (20 structures)
AMBER restraint violations
Maximal NOE violation, Å	0.19	0.17	0.18
Maximal torsion angle violation, °	0.00	0.00	0.00
Deviations from ideal geometry
Bond lengths, Å	0.0013	0.013	0.0013
Bond angles, °	2.1	2.2	2.1
AMBER energies, kcal⋅mol−1
Mean restraint energy	7.59	8.80	4.71
Mean AMBER energy	−5,172	−4,501	−5,439
Rmsd from mean*
Backbone heavy atom, Å	0.33	0.28	0.45
Heavy atom, Å	0.81	0.76	0.85
PROCHECK statistics
Most favored region, %	91.2	88.8	90.8
Additionally allowed region, %	8.8	9.4	9.2
Generously allowed region, %	0.1	1.3	0.0
Disallowed region, %	0.0	0.5	0.0
PSVS output
Verify3D Z-score	−0.32	0.64	−1.93
MolProbity Clash Z-score	1.28	1.33	1.32

^*For selected residues. TAZ2–p53: all of TAZ2; E17–P27 and M44–T55 of p53. TAZ2–p53AD2: all of TAZ2; L45–T55 of p53. p53–TAZ1: F19–P27 and D42–T55 of p53; E348–A372 and C384–N434 of TAZ1.

Structures of TAZ2 and TAZ1.

The structures of the TAZ2 and TAZ1 domains in the TAZ2–p53(2–61), TAZ2–p53(37–61), and p53(1–61)–TAZ1 fusion constructs are well-defined (Fig. 4). Similar to the structures in the unligated state (43, 44), each TAZ domain consists of a bundle of four helices (α₁ – α₄) with zinc atoms bound in the interhelical loops. The backbone atom RMSDs between free TAZ2 (43) and TAZ2 fused to the p53TAD and p53AD2 are 1.13Å and 1.01Å respectively, with the largest differences being in the relative position of helix α₁ in the free protein. There is also a substantial shift in the position of the α₂–α₃ loop of TAZ2, which forms part of the binding surface for the AD1 motif, in the TAZ2–p53(2–61) complex. The overall structure of TAZ1 in the TAZ1–p53TAD fusion resembles that of the free domain, with a backbone RMSD between free TAZ1 (44) and TAZ1 in the p53TAD–TAZ1 fusion of 1.57Å.

Fig. 4.

Backbone representation of ensembles of 20 solution structures (Left) and ribbon representation of lowest AMBER energy models (Right) of (A) TAZ2–p53TAD, (B) TAZ2–p53AD2, and (C) and p53TAD–TAZ1 fusion proteins. TAZ1 is colored pink, TAZ2 is colored gold, and p53 is colored blue, with the AD1 helix shown in turquoise and the AD2 region in purple. The fusion linkers between the p53 and TAZ domains are shown in gray. Zinc atoms are shown as gray spheres. The TAZ1 and TAZ2 helices α₁–α₄ are labeled.

Structure and Dynamics of p53TAD in Complex with the TAZ Domains.

In the absence of binding partners, the p53TAD is intrinsically disordered. Upon binding to TAZ2 and TAZ1, the AD1 and AD2 regions fold into amphipathic helices (Fig. 4) that span residues Q16–L25 and P47–T55 in the complex with TAZ2 and Q16–L25 and P47–W53 when bound to TAZ1. The AD1 and AD2 helices in the fusion proteins are bound to the same surfaces and in the same orientation as in preliminary structures of the nontethered complexes; we are therefore confident that the p53–TAZ1/2 interactions have not been altered by fusing the binding partners into a single polypeptide chain. Residues N-terminal to AD1 are unstructured and do not seem to be involved in the binding process, as their ¹H-¹⁵N HSQC cross-peaks are minimally perturbed upon addition of the TAZ domains. Residues beyond the C-terminal end of the AD2 helix are also unstructured in the complex.

The backbone conformation for residues between E28 and D42 of the p53TAD has a relatively large RMSD between structures in both the TAZ1 and TAZ2 complexes (central blue region, Fig. 4 A and C). This disorder might be due to local conformational flexibility or merely reflect the low density of intermolecular NOE restraints in this region. Heteronuclear {¹H}-¹⁵N NOE experiments were performed to provide a direct measure of the backbone flexibility of ¹⁵N-labeled p53 in the p53TAD–TAZ1 fusion, the (untethered) TAZ2:p53(1–61) complex, and free p53(1–61) (Fig. 5). Except at the N- and C terminus, the NOE values for the complexes are substantially higher than for the free protein, consistent with the transition from the disordered free p53TAD to a more ordered state in the complexes. The NOE values for the p53TAD in both complexes are very similar at most positions, with two local maxima corresponding to the positions of the AD1 and AD2 helices. The dip in the NOE values for residues 28–42 confirms that the backbone between the AD1 and AD2 motifs is flexible on the ns timescale, even when the p53TAD is bound to the surface of TAZ1 or TAZ2.

Fig. 5.

Per-residue plot of {¹H}–¹⁵N heteronuclear NOEs measured at 600 MHz for ¹⁵N-labeled p53(1–61) in the free state (black), in the untethered TAZ2:p53(1–61) complex (blue), and in the p53(1–61)–TAZ1 fusion protein (red). The NOE values for backbone amides are shown as circles whereas NOEs for Trp indole Nε resonances are shown as triangles. The AD1 and AD2 sequences are shown as boxes colored turquoise and purple, respectively.

Interactions Between the p53TAD and TAZ2.

In the TAZ2–p53(2–61) and TAZ2–p53(37–61) constructs, residues 43–54 of the p53TAD adopt very similar backbone structures and the side chains make similar contacts with the TAZ2 surface (Fig. 4 A and B). The AD2 helix and two hydrophobic residues (M44, L45) that precede it bind to a hydrophobic patch, formed by the side chains of I1773, M1799, V1802, V1819, L1823, A1825, L1826, C1828, and Y1829, at the interface between the α₁, α₂, and α₃ helices of TAZ2 (Fig. 6A). Binding of the AD2 helix is mediated by the side chains of I50, W53, and F54, with the phenylalanine ring docked into a concave pocket on the surface of TAZ2. The M44 side chain of p53 packs against the side chains of A1825, C1828, and Y1829 on helix α₃ whereas that of L45 participates in an extensive network of hydrophobic interactions involving residues located on the α₁ and α₃ helices of TAZ2 as well as side chains of p53 itself (M44, I50, and W53). This extensive hydrophobic cluster probably plays a major role in stabilization of the complex. The importance of hydrophobic interactions in this region has been previously demonstrated, as a W53Q/F54S mutant greatly diminishes AD2 binding to TAZ2 (24). The heteronuclear NOE for the indole NεH of W53 is similar to the backbone amide NOEs in the AD2 helix (Fig. 5), consistent with tight packing of the W53 side chain in the molecular interface. The regions of TAZ2 surrounding the α₁–α₂–α₃ hydrophobic binding surface are strongly electropositive and acidic side chains on p53, namely D41, D42, and E51, make complementary electrostatic interactions (Fig. S6A).

Fig. 6.

Interaction between p53TAD and (A) TAZ2 and (B) TAZ1. (Top) Overview of the lowest AMBER energy structures, showing the positions of AD1 (turquoise) and AD2 (purple). (Bottom) Local side chain interactions for AD2 (purple) and AD1 (turquoise).

Fig. S6.

Electrostatic surfaces of (A) the TAZ2 and (B) TAZ1 domains. The bound p53TAD is shown in a schematic representation (green), and the locations of negatively charged residues (orange) and phosphorylation sites (magenta) are displayed as spheres.

The p53 binding mode is very similar in the structures of the TAZ2–p53(2–61) and TAZ2–p53(37–61) fusion proteins (Fig. 7A), with a backbone RMSD of 0.85Å between the two structures for p53 residues L43 to T55. The conformations of the hydrophobic side chains of I50, W53, and F54 are almost identical in the two structures while the preceding residues, M44 and L45, vary only slightly. The similarities in the mode of binding of the p53 AD2 region in the independently determined TAZ2–p53AD2 and TAZ2–p53TAD fusion protein structures provide strong validation of our structural model.

Fig. 7.

(A) Comparison of the AD2 region of p53 in the structures of the TAZ2–p53(2–61) (purple) and TAZ2–p53(37–61) (lavender) fusion proteins. The TAZ2 surface is shown in pale yellow. (B) Superposition of the structure of the TAZ2–p53(37–61) fusion protein (p53 AD2 region, purple; TAZ2 surface) and that of the complex between TAZ2 and an isolated p53AD1 peptide (brown) (2K8F, ref. 38). (C) Superposition of the structure of the TAZ2-p53(37-61) fusion protein (p53 AD2 region, purple; TAZ2 surface) and that of the complex between TAZ2 and an isolated p53AD2 peptide (red) (2MZD, ref. 39).

In the absence of AD2, the isolated AD1 motif (residues 13–37) binds preferentially, but with much lower affinity than AD2 (24 μM vs. 32 nM), to the AD2 binding site at the TAZ2 α₁–α₂–α₃ interface; there is an even weaker (160 μM) secondary binding site for the isolated AD1 in the vicinity of the α₃ and α₄ helices (54). In the context of the full-length p53TAD, AD1 occupies its secondary binding site on TAZ2, a deep hydrophobic groove between the α₃ and α₄ helices and the α₂–α₃ loop, formed by residues T1812, L1823, I1824, A1825, F1843, I1847, and L1851 (Fig. 6A). The hydrophobic side chains of F19, L22, W23, L25, and L26 of p53TAD all project into this groove to make favorable intermolecular contacts. The proximity of the K1850 side chain to the indole ring of W23 suggests the possibility for pi–cation interactions, which would be consistent with the NOE cross-peaks observed between these residues. Similar to W53, the ps–ns timescale dynamics of the W23 side chain are restricted in the TAZ2 complex (Fig. 5). Residues N-terminal to the AD1 helix are largely unstructured and lack long-range NOEs. The hydrophobic side chains of V31 and L32 of p53 contact L1851 and the aliphatic portion of K1832; however, the side chain conformations are not well defined and the interaction is likely transient given the decreased heteronuclear NOE values for these p53 residues (Fig. 5). The region of TAZ2 to which AD1 binds presents less exposed hydrophobic surface area than does the AD2 binding region, and there is less charge complementarity in the AD1 binding site. This difference is mainly due to the presence of several polar residues on the flexible loop between α₂ and α₃ of TAZ2. These differences may explain why binding at this site represents the “weaker” interaction of p53TAD with TAZ2.

Interactions Between the p53TAD and TAZ1.

Binding of p53TAD to TAZ1 is also mediated by hydrophobic interactions. Similar to its interaction with TAZ2, the AD2 helix binds at the interface of the α₁, α₂, and α₃ helices of TAZ1. Residues N-terminal to the AD2 helix pass through a surface groove between α₁ and α₂ and wrap around α₁; in the TAZ2 complex this part of the p53 chain passes between the α₁ and α₃ helices. Residues L43, M44, L45, P47, I50, W53, and F54 of the AD2 region make hydrophobic contacts with L352, I353, Q356, L359, L360, M387, V390, S410, S411, I414, and V428 of TAZ1 (Fig. 6B). The F54 side chain reaches to the opposite side of α₁ to make a hydrophobic contact with V428 at the beginning of α₄. M40, L43 and L45 stack against α₁ whereas M44 contacts α₁ and the start of α₂. I50 and W53 protrude into the groove between α₁ and α₃. As in the complex with TAZ2, the W53 side chain is tightly packed in the molecular interface and the heteronuclear NOE shows that its ps–ns timescale motions are restricted (Fig. 5). The p53 structure between AD1 and AD2 is not well-defined, with increased backbone RMSD and decreased heteronuclear NOEs.

Compared with the binding pocket of AD2 on TAZ2, the AD2 binding site on TAZ1 presents a less continuous hydrophobic surface due to the presence of Q355, Q356, S410, S411, and Q413, which are all either in or proximal to the binding interface. Previous structures of TAZ1 in complexes with transactivation domains have demonstrated that TAZ1 prefers to bind to relatively long intrinsically disordered regions containing multiple amphipathic motifs (45–50). By contrast, TAZ2 mostly interacts with short amphipathic helices. As such, isolated subdomains of p53TAD may not represent an ‘ideal’ TAZ1 binding protein. These differences may help explain the two orders of magnitude lower K_d for binding of an AD2 peptide to TAZ2 compared with TAZ1 (0.055 μM vs. 4.9 μM) (25).

The AD1 motif of p53 binds in a deep hydrophobic pocket, surrounded by positively charged side chains, formed by all four of the TAZ1 helices. The side chains of F19, L22, and W23 in AD1 pack against L357, V358, L360, and L361 (α₁), L391, M394, and T395 (α₂), I415 and W418 (α₃), L432 and A435 (α₄). The N-terminal region of AD1 makes contact with α₄, and the C-terminal region, starting with L26, begins to wrap around the α₁ helix of TAZ1 (Fig. 6B). L14 makes contact with W418, although the structure in this region is poorly defined and the heteronuclear NOE data suggest that the interaction is likely to be transient (Fig. 5). Similarly, L32 binds in the groove between helices α₁ and α₄; as with L14, heteronuclear NOE data suggest this contact is transient. The side chain of F19, on the first turn of the AD1 helix, is deeply buried in the p53:TAZ1 interface, closely packed against the plane of the indole ring of W418 and contacting hydrophobic residues on helices α₁, α₃, and α₄. The side chain of W23 at the C-terminal end of the AD1 helix packs against L391, M394, and I415 on α₂ and α₃. The ¹⁵Nε resonance of W23 has a much smaller heteronuclear NOE (0.39 ± 0.01) than the backbone NHs in the AD1 helix, suggesting significant side chain fluctuations on a picosecond to nanosecond timescale. This flexibility is fully consistent with AD1 being the weaker binding subdomain of the bipartite p53TAD. In contrast to the TAZ2 complex, where L25 and L26 form an integral part of the molecular interface, the L25 side chain projects into solvent where it makes minimal contact with TAZ1. L26 contacts the aliphatic portions of K365 and R368, two of the many positively charged residues (the others being H364, K419, K433, K438, and R439) that ring the AD1 binding pocket.

Although the AD2 motif of the p53TAD binds to both TAZ1 and TAZ2 in a hydrophobic groove at the interface between the α₁, α₂, and α₃ helices (Fig. S7), the orientation of the AD2 helix is opposite in the two structures. Although L45 and the hydrophobic side chains on the AD2 helix adopt similar conformations in the two complexes, the intermolecular interactions are different. For example, the W53 indole ring packs against hydrophobic side chains in helix α₃ of TAZ2, whereas it packs primarily against helix α₁ in the TAZ1 complex. As a result of differences in the orientation of helix α₄ (Fig. S7A), the AD1 motif binds to different surfaces of the TAZ 1 and TAZ2 domains. In TAZ2, the α₄ helix extends away from the core of the protein and creates a hydrophobic groove between α₃, α₄, and the α₂–α₃ loop that accommodates the AD1 helix. In contrast, the α₄ helix of TAZ1 packs tightly against α₁, creating an extensive hydrophobic surface, comprising side chains from the α₁, α₂, α₃ and α₄ helices, that binds the AD1 motif (Fig. S7C).

Fig. S7.

(A) Overlay of the lowest-energy structures of TAZ2 (gold) and TAZ1 (pink). (B) Position of p53AD1 (aquamarine) on TAZ2 (gold) and TAZ1 (pink). (C) Position of p53AD2 (purple) on TAZ2 (gold) and TAZ1 (pink).

Discussion

The present work provides valuable insights into the molecular interactions through which p53 recruits CBP and p300 to activate transcription of stress response genes and underscores the complex nature of molecular recognition by intrinsically disordered proteins containing multipartite binding motifs. Three structures have been reported previously for complexes of the p53TAD with domains of CBP/p300: p53(13–61) bound to the CBP nuclear coactivator binding domain (NCBD) (62), p53(1–39) bound to the p300 TAZ2 domain (38), and p53(35–59) bound to p300 TAZ2 (39). The NCBD:p53TAD complex includes both the AD1 and AD2 subdomains of p53TAD whereas the TAZ2 complexes contain either the AD1 or AD2 motif.

Here we report, to our knowledge, the first structures of the full-length p53TAD bound to the TAZ1 and TAZ2 domains of CBP/p300. The structure of the TAZ2 complex, which is likely to be representative of the complex in vivo where intact AD1 and AD2 motifs are required for efficient transactivation of the majority of p53-regulated genes (31, 63), differs significantly from previously reported structures of the isolated AD1 and AD2 peptides bound to the p300 TAZ2 domain. The isolated p53AD1 and p53AD2 peptides bind preferentially to the same site on TAZ2, at the interface between the α₁, α₂, and α₃ helices; however, AD2 binds with nearly 500-fold greater affinity than AD1 (0.055 μM vs. 27 μM) (25). Both motifs also bind with lower affinity to a common secondary site on the opposite surface of TAZ2, but again the strongest interaction is with AD2 (10 μM vs. 164 μM) (54). The affinity of the full-length TAD for TAZ2 (0.026 μM) is increased only slightly over that of AD2 alone (24, 25). On the basis of these observations, it is not surprising that our structure of the full-length p53TAD complex, with both the AD1 and AD2 motifs bound to different sites on TAZ2, differs from that of the complex formed by AD1 alone (38). The structure shows the AD2 motif of the full-length p53TAD bound to the hydrophobic surface that was previously identified through chemical shift titrations as the primary binding site for isolated AD2 (54). The very high affinity of AD2 for its primary site forces AD1 to bind in its weak (164 μM) secondary binding site, where it contributes very little to the overall binding free energy.

Comparison of our structures of the TAZ2–p53(2–61) and TAZ2–p53(37–61) fusions with that of the TAZ2:AD1 complex of Feng et al. (38) suggests a molecular basis for the difference in AD1 and AD2 affinities. Although the AD1 and AD2 helices bind in a similar location and orientation on TAZ2 (Fig. 7B), the AD2 motif makes significantly more hydrophobic contacts with the surface of TAZ2. The side chains of W53 and F54 of AD2 are deeply buried in the hydrophobic groove formed by the α₁, α₂, and α₃ helices and additional hydrophobic contacts are made by M40, L41, and I50. In contrast, AD1 side chains are not well packed in this pocket; only T18, F19, and L22 make hydrophobic contacts with TAZ2 while the W23, L25, and L26 side chains are on the surface of the complex and are almost fully solvent exposed (38).

An NMR structure of the complex between an isolated p53(35–59) peptide and the p300 TAZ2 domain was recently reported (PDB accession code: 2MZD) (39), which differs substantially from the structures determined in the present work. The AD2 motif binds to the same hydrophobic surface on TAZ2 in both of the structures, but the orientation of the AD2 helix differs by ∼90° (Fig. 7C). The p53(35–59) peptide in structure 2MZD uses the same hydrophobic residues (I50, W53, and F54) to bind to TAZ2; however, these side chains pack very differently compared with our TAZ2–p53(2–61) and TAZ2–p53(37–61) structures. In structure 2MZD, the only residue fully buried in the hydrophobic groove of TAZ2 is I50 whereas W53 packs along the surface of helices α₂ and α₃, and F54 packs along the aliphatic portion of R1737 (CBP R1775) located on helix α₁. By contrast, in our TAZ2–p53(2–61) and TAZ2(37–61) structures, I50, W53, and F54 are all deeply buried in the hydrophobic interface between the AD2 helix and TAZ2. The hydrophobic residues preceding the AD2 helix (M44 and L45, and to a lesser extent L43) pack along the surface of the TAZ2 α₃ helix in all of the structural models. The observed differences between 2MZD and the present structures are highly unlikely to be a consequence of fusing the p53 peptides to the TAZ2 domain. Firstly, the orientation of the AD2 helix is identical in the structures of the TAZ2–p53(2–61) and TAZ2–p53(37–61) fusions (Fig. 7A), where the “linkers” between the two proteins are 52 and 16 residues in length, respectively. Secondly, the AD2 helix is bound in a very similar position and orientation in the structures of the two fusion constructs and in preliminary structures of the nonfused 1:1 p53(13–61):TAZ2 complex. We are therefore confident that our structural models accurately represent the molecular interface between the CBP TAZ2 domain and the p53 AD2 motif. We suggest two possible reasons for the observed differences relative to 2MZD. The side chains that form the binding surface for the AD2 motif are 100% identical in the CBP and p300 TAZ2 domains. However, the p300 TAZ2 construct used for structure determination was mutated to replace four Cys residues (C1738A, C1746A, C1789A, C1790A) by Ala (39) whereas the WT CBP sequence, retaining all four Cys residues, was used in the present work. One of the mutated cysteines (C1738A in p300, C1776 in CBP) lies in the binding pocket for the AD2 helix, and it is possible that mutation of this residue influences the binding interface. Alternatively, the structure might represent an alternate binding conformation that is stabilized under the different buffer and solution conditions used to collect the p300 NMR data.

Comparison with Other p53 Transactivation Domain Structures.

Structures have now been reported for complexes of the p53 AD2 motif with several target proteins. The motif adopts a similar structure, with helix between residues P47 and F54 and very similar conformations of the hydrophobic side chains, when bound to TAZ1, TAZ2 (this work), replication protein A (36), and the PH domain of the Tfb1 subunit of TFIIH (37) (Fig. 8A). It thus seems that the local amino acid sequence predisposes AD2, which is disordered in the unbound state, toward a well conserved structure when bound to its targets. A similar AD2 structure is observed in the complex with HMGB1, although the helix is somewhat shorter (64). Somewhat surprisingly, given the strong propensity of AD2 to fold into a helix, a doubly phosphorylated p53 AD2 motif binds in a more extended conformation to a highly electropositive region of the PH domain of human TFIIH p62 (65).

Fig. 8.

(A) Superposition of residues P47–F54 in the AD2 helix in complexes with TAZ1 (pink, this work), TAZ2 (green, this work), replication protein A [red (36)], and the Tfb1 subunit of TFIIH [blue (37)]. A schematic view of the backbone is shown, and the hydrophobic side chains are shown as sticks. (B) Superposition of residues F19–L25 in the AD1 helix in complexes with TAZ1 (pink, this work), TAZ2 (green, this work), TAZ2 [blue (38)], and MDM2 [red (35)]. The hydrophobic side chains are shown as sticks.

The AD1 region adopts a helical backbone conformation between T18 or F19 and L25 in its complexes with TAZ1, TAZ2, and MDM2 (this work and refs. 35 and 38) (Fig. 8B). However, in contrast to hydrophobic side chains in AD2, the side chains of F19 and W23 occupy different rotamers in the different AD1 complexes, reflecting differences in the binding surfaces of the target proteins.

Enhancement of CBP Binding by p53 Phosphorylation.

Phosphorylation of the p53TAD has been shown to greatly enhance binding affinity for the TAZ domains (12, 13). The p53TAD can be phosphorylated at nine sites, many of which are well positioned to make favorable electrostatic interactions with TAZ1 and TAZ2. Phosphoryl groups can be accommodated at all of these sites without steric clashes with the TAZ domains. In the TAZ1 complex, the AD1 motif binds in a region ringed with basic residues (Fig. S6B) with S15 positioned close to R439 and K433, T18 close to K438 and R439, S20 close to K418, and S33 close to H362, K365, and R369. Although single site phosphorylation in this region leads to only a modest (∼twofold) increase in affinity, two- or three-site phosphorylation increases the TAZ1 binding affinity by as much as 10-fold (12). T55, in the AD2 region of p53, is proximal to K349. S37 and S46 are more distant from basic residues in the family of NMR structures, but still contribute to enhancement of the binding affinity upon phosphorylation (12, 13). It is of note that S46 is located at the N-cap position of the AD2 helix, and its phosphorylation may lead to substantial stabilization of the helical state (66). Similar electrostatic interactions are observed in the TAZ2–p53TAD structure (Fig. S6A). For p53AD1 bound to TAZ2, T18 and S20 are positioned close to K1850 and K1812 respectively. Within AD2, phosphorylation of S46 could serve to N-cap and stabilize the AD2 helix as well as promote interactions with the nearby side chain of R1769 whereas phosphorylation of T55 could stabilize interactions with K1798. Although S15, S33, and S37 in the p53TAD do not make specific contacts with basic side chains of TAZ2 in the NMR structural ensemble, all are located close to electropositive regions of the TAZ2 surface such that phosphorylation could enhance binding through bulk electrostatic effects. Indeed, bulk electrostatics seem to play a dominant role since the degree of enhancement of TAZ1 or TAZ2 binding upon phosphorylation of the p53TAD is determined primarily by the number of phosphoryl groups and not by their location in the p53 sequence (12). Similar to TAZ1, two-site and three-site phosphorylation of the AD1 region of the p53TAD increases the TAZ2 binding affinity by up to 20-fold (12).

Implications of the Bipartite Activation Domain.

The bipartite p53 N-terminal activation domain interacts with its binding partners via two short amphipathic motifs that bind simultaneously to the target protein and fold into helical structure upon binding. In both the TAZ1 and TAZ2 complexes, the AD1 and AD2 motifs bind to opposite surfaces of the target protein. Heteronuclear NOE measurements reveal enhanced backbone flexibility of the p53 residues between the motifs, and most of the side chains in this region are disordered and make few or no contacts with the TAZ domains (Fig. 5). Despite the flexibility of the intervening sequence, binding is synergistic and the overall binding affinity of the full-length p53TAD is two- to fivefold higher than for the AD2 motif in isolation and up to 1,000-fold higher than for the isolated AD1 motif (25). Similar behavior was observed for the complex of the p53TAD with the nuclear coactivator binding domain of CBP (62). The exposed and flexible nature of the peptide between the AD1 and AD2 subdomains suggests the possibility of posttranslational modification in this region while the bipartite p53TAD remains bound to TAZ2 or TAZ1.

In addition to binding synergistically to a single target protein in a binary complex, the AD1 and AD2 interaction motifs of the p53 transactivation domain can also bind independently to two different target proteins to form a ternary complex. In the complexes with the TAZ2 and TAZ1 domains, the p53 AD2 motif occupies a more favorable binding pocket than does AD1 and makes the dominant contribution to the binding free energy (25). Because many of the residues between AD1 and AD2 remain flexible in the complex and contribute little to the binding interactions, ternary complexes can be formed with AD2 bound to TAZ1/2 and AD1 bound to a different target, such as MDM2, leading to cross talk between different signaling pathways (25). The ternary complex between p53, MDM2, and CBP/p300 seems to play a central role in regulation of p53 stability, determining in a phosphorylation-dependent manner whether p53 becomes polyubiquitinated and degraded or is activated to initiate transcription of p53-regulated stress response genes (25, 67). The ability of multipartite systems to promiscuously interact with binding partners has been shown with other CBP/p300 binding proteins and may represent a widely used mechanism by which proteins can use disparate motifs to regulate transcription (53, 68, 69).

The present work clearly illustrates the dangers of a reductionist approach to characterize target binding by intrinsically disordered proteins containing two or more interaction motifs. The isolated AD1 and AD2 regions of the p53 transactivation domain interact promiscuously with their targets (25, 53, 54). For example, in the absence of AD2, the AD1 motif binds preferentially in the AD2 binding site at the interface of the α₁, α₂, and α₃ helices, as was observed in the structure of the complex of TAZ2 with an AD1 peptide (38), and only weakly in a secondary site between the α₃ and α₄ helices. When both motifs are present in the same polypeptide chain, as in the present work, the AD2 motif out-competes AD1 for binding in the preferred high affinity site, forcing AD1 to occupy its secondary binding site on TAZ2. Studies using isolated subdomains of disordered proteins that contain bipartite or multipartite interaction motifs can potentially yield misleading results.

Although the AD2 subdomain of p53 dominates the interaction with TAZ1 and TAZ2 and binds in a highly preferred site, interactions with other targets can be more promiscuous. The p53 TAD forms an ensemble of conformational states in complex with the KIX domain of CBP/p300, with each of the AD1 and AD2 helices binding simultaneously in two distinct orientations and in two distinct site on the KIX surface (53). Similar promiscuous interactions with both KIX sites have also been observed for the bipartite transactivation domain of the FOXO3a transcription factor and are required for full transcriptional activity (68).

Competition with Other Transactivation Domains.

p53 must compete with a multitude of cellular transcription factors for binding to limiting amounts of CBP/p300 in the cell (41, 70, 71). Structures have been determined for TAZ2 bound to the activation domains of STAT1 (49), and C/EBP (51), both of which bind through helical structures to the same hydrophobic surface as the p53 AD2 motif. The adenovirus oncoprotein E1A and oncoprotein E7 from high-risk strains of human papillomavirus also bind to the same surface on TAZ2 as the AD2 motif of p53, but with higher affinity (E1A, 3 nM; phosphorylated E7, 7 nM; p53, 26 nM) (25, 52, 72). By binding with higher affinity to the same region of TAZ2, the viral proteins can out-compete p53 for binding to CBP/p300 and thereby inhibit activation of p53-mediated stress response or apoptotic pathways.

Structures have been reported for TAZ1 bound to the activation domains of the hypoxia inducible factor 1α (HIF-1α) (45, 46), CITED2 (47, 48), STAT2 (49), and RelA (50), all of which bind in the same hydrophobic groove as the p53TAD. Upon binding to TAZ1, all of these activation domains form local elements of amphipathic helical structure, but for the most part are located in different regions of the hydrophobic groove. Thus, the p53 AD2 helix binds to the same surface as the C-terminal (αC) helix of HIF-1α but in a different orientation whereas the AD1 motif binds in a groove that is occupied by the LPQL motif and αB helix of HIF-1α (45, 46). Although all of the above activation domains contain ΦXXΦΦ or ΦΦXXΦ motifs, where Φ is a bulky hydrophobic residue, they share little in common in their interactions with TAZ1, which are clearly determined by the distribution of the amphipathic clusters within the peptide sequence and by the nature of the flanking amino acids.

The p53-Mediated Stress Response Requires the Full-Length TAD.

The interaction between p53 and CBP/p300 is essential for activation of p53-mediated transcriptional programs in response to cellular stress and genomic damage. High affinity binding is mediated primarily by interactions of the intrinsically disordered N-terminal transactivation domain of p53 with the TAZ domains of CBP/p300 and is enhanced by multisite phosphorylation of p53 in response to cellular stress (12, 13). Although the AD2 motif dominates binding of the p53TAD to the TAZ domains (24, 25), the AD2 and AD1 subdomains are both required for transactivation of most p53 target genes and L22Q/W23S (human numbering) mutant p53 is unable to induce cell cycle arrest or apoptosis in response to DNA damage in knock-in mice (31, 73). Activation of p21-like genes, which requires CBP/p300-mediated acetylation of K382 in the C-terminal regulatory domain of p53, is severely impaired by deletion or mutation of the AD1 subdomain (73, 74), suggesting that synergistic interactions involving both motifs are required for efficient recruitment of CBP/p300. Our results here demonstrate a structural basis for the importance of AD1 in regulating the p53-mediated cellular stress response. In contrast to previous reports describing interactions of TAZ2 with the isolated AD1 and AD2 subdomains (38, 39), the current work reveals the molecular basis for binding of the full-length p53 transactivation domain to both TAZ1 and TAZ2, providing insights into critical interactions required for activation of the p53-regulated damage response and showing how intrinsically disordered proteins can recognize disparate binding partners.

Materials and Methods

Protein Production.

The TAZ1 and TAZ2 domains of mouse CBP (residues 340–439 and 1764–1855, respectively) and p53TAD (residues 1–61 of human p53) were expressed and purified as previously described (25). Fusion proteins were expressed with the TAZ2 domain of mouse CBP (1764–1855) joined via a (Gly-Ser)₃ linker to the transactivation domain of p53. The design of the fusion constructs was guided by knowledge of the binding surfaces derived from chemical shift titrations with free p53TAD constructs and by preliminary low-resolution NMR structures determined for the 1:1 TAZ2:p53(13–61) complex, which showed the location and orientation of the AD1 and AD2 helices. The linker length and site of fusion were optimized to give high-quality HSQC spectra, free of exchange broadening. A TAZ2-(GS)₃–p53(2–61) fusion gave high-quality NMR spectra with more uniform cross-peak intensities. In contrast, a construct with a shorter linker between the C terminus of TAZ2 and the AD1 motif (TAZ2-(GS)₃–p53(13–61) resulted in exchange broadening of resonances associated with the AD1 binding site. The TAZ2-(GS)₃–p53(2–61) fusion was therefore used for NMR structure analysis. Data were also acquired for a TAZ2-GSGSG–p53AD2(37–61) fusion. The TAZ2–p53 fusions were expressed in Escherichia coli BL21 (DE3) [DNAY] cells in M9 minimal media for 12–16 h at 20 °C after induction with 0.8 mM isopropyl β-d-thiogalactopyranoside (IPTG) and 150 μM ZnSO₄. Cells were lysed via sonication in 20 mM Mes, pH 6.1, 40 mM NaCl, and 10 mM DTT. These constructs were purified via SP Sepharose on a gradient from 20 mM Mes, pH 6.1, 40 mM NaCl, 2 mM DTT to 20 mM Mes, pH 6.1, and 1 M NaCl. Further purification was achieved via Sephacryl S100HR size exclusion in 20 mM Tris, pH 6.8, 100 mM NaCl, and 2 mM DTT. Purity was verified by SDS/PAGE and analytical HPLC, and molecular weights were verified by MALDI. Purified product was exchanged into 20 mM Tris, pH 6.8, 50 mM NaCl, 2 mM DTT, and 10% (vol/vol) D₂O for NMR data collection.

A segmentally labeled fusion protein with the C terminus of the p53TAD joined to the N terminus of the TAZ1 domain was also constructed on the basis of preliminary low-resolution structures of a 1:1 TAZ1:p53(13–61) complex. Segmentally labeled p53TAD–TAZ1 constructs were generated using the Nostoc punctiforme PCC73102 (Npu) DNA polymerase III (DnaE) intein system (75–77). p53TAD was expressed with an N-terminal H₆GB1 tag and was fused to the N terminus of the N-split of the DnaE intein (Fig. 1D). The region between H₆GB1 and p53(1–61) contained a tobacco etch virus (TEV) protease cleavage site. This construct was expressed in E. coli BL21 (DE3) [DNAY] cells for 20 h at 16 °C in either isotopically enriched or unenriched media after induction with 1 mM IPTG. The TAZ1 domain of mouse CBP (residues 340–439) was expressed fused to the C terminus of the C-split of the DnaE intein. This construct had an H₆GB1 tag N-terminal to the intein fragment (Fig. 1D). This construct was expressed for 20 h at room temperature after induction with 0.1 mM IPTG. Cells for both constructs were lysed via sonication in 20 mM Tris, pH 8.0, and 200 mM NaCl and purified using His60 Ni SuperflowTM Resin (Clontech) with elution in 20 mM Tris, pH 8.0, 200 mM NaCl, and 150 mM imidazole. Intein ligation was performed at a ratio of 1:1.2 p53:TAZ1 in 20 mM Tris, pH 8.0, 200 mM NaCl, 0.5 mM tris(2-carboxyethyl)phosphine (TCEP) for 1 h at room temperature before addition of 1/200 molar equivalent of TEV protease. Intein ligation and TEV protease cleavage reactions were carried out for an additional 12–15 h at 4 °C.

After completion of ligation and protease cleavage, the mixture was run over a 5-mL His60 Ni Superflow column (Clontech). The flow-through was purified via a 2 × 5-mL HiTrap Q HP Column (GE) on a linear gradient from 20 mM Tris, pH 7.5, 40 mM NaCl, 1 mM DTT to 20 mM Tris, pH 7.5, 1 M NaCl. The p53TAD–TAZ1 fusion eluted at ∼150 mM NaCl. Lastly, the fusion was eluted from a 345-mL S100 size exclusion column with buffer containing 20 mM Tris, pH 8.0, 500 mM NaCl, and 1 mM DTT. Samples were exchanged into 20 mM Tris, pH 6.8, 50 mM NaCl, 1 mM DTT, and 5% D₂O for NMR data collection. The ligation reaction and purification were monitored by analytical HPLC and SDS/PAGE, with ligation efficiency consistently above 90%. The molecular mass of the product was confirmed by mass spectroscopy.

NMR Experiments.

¹H-¹⁵N HSQC spectra for ¹⁵N p53(1–61), ¹⁵N TAZ1, 1:1 ¹⁵N p53TAD:¹⁴N TAZ1, 1:1 ¹⁵N TAZ1:¹⁴N p53TAD, and the p53TAD(¹⁵N)–TAZ1(¹⁴N) and p53TAD(¹⁴N)–TAZ1(¹⁵N) fusion proteins were acquired with 100 μM protein concentrations in 20 mM Tris, pH 6.8, 50 mM NaCl, 1 mM DTT, and 5% D₂O (NMR buffer). ¹H^-15N HSQC spectra for ¹⁵N TAZ2, 1:1 ¹⁵N p53TAD:¹⁴N TAZ2, 1:1 ¹⁵N TAZ2:¹⁴N p53TAD, and the ¹⁵N TAZ2–p53TAD and ¹⁵N TAZ2–p53AD2 fusion proteins were acquired at 100-μM protein concentrations in NMR buffer.

Spectra for resonance assignment and for restraint generation were acquired on Bruker Avance 900-MHz, DRX 800-MHz, DRX 600-MHz, and Avance 500-MHz spectrometers and were processed using NMRPipe (78) and NMRView5.0 (79).

Backbone resonances for the p53TAD–TAZ1 construct were assigned using HNCACB, HN(CO)CA, and ¹⁵N TOCSY spectra (80, 81) collected on segmentally labeled p53TAD(¹⁵N/¹³C)–TAZ1 and p53TAD–TAZ1(¹⁵N/¹³C) samples. These samples were also used to make side chain assignments using 3D HCCH-COSY and HCCH-TOCSY spectra (82). Backbone and side chain resonance assignments for the TAZ2–p53TAD and TAZ2–p53AD2 constructs were derived from ¹⁵N NOESY-HSQC, ¹³C NOESY-HSQC, and HNCACB experiments. These assignments were validated by comparisons with assignments of TAZ2 complexes with multiple p53 constructs, including AD1(13–37), AD2(38–61), and p53(13–61). Assignments of these complexes were made through use of ¹H-¹⁵N and ¹H-¹³C HSQC titrations, HNCA and HNCO spectra, and HCCH and CCHC COSY spectra, and through ¹⁵N NOESY-HSQC and ¹³C NOESY-HSQC experiments. These assignments were further validated by comparison with previously published assignments of free TAZ2 (43), TAZ2 bound with STAT1 (49), and bound chemical shifts of p53 with the NCBD and KIX domains of CBP (53, 62).

Distance restraints were obtained from ¹⁵N NOESY-HSQC (τ_m = 100 ms) and ¹³C NOESY-HSQC (τ_m = 120 ms) spectra. For the p53TAD–TAZ1 fusion protein, a ¹³C-filter-edit NOESY-HSQC spectrum (τ_m = 200 ms) was used to derive intermolecular distance restraints (58). NMR data for the TAZ2–p53TAD and TAZ2–p53AD2 constructs were collected at 32 °C, and data for p53TAD–TAZ1 were acquired at 35 °C. {¹H}-¹⁵N heteronuclear NOE data were acquired at 600 MHz and 35 °C on free ¹⁵N p53(1–61) (100 μM), p53(¹⁵N)–TAZ1(¹⁴N) fusion protein (400 μM), and a sample of 300 μM ¹⁵N p53(1–61) with 1.2-fold excess of unlabeled TAZ2.

Structure Calculations and Analyses.

Initial structure generation and NOE assignment were performed through CYANA with CANDID (60), using a subset of manually assigned, unambiguous NOE restraints. Distance restraints were derived from NOE cross-peak volumes. Dihedral angle restraints were determined from ${}^{13}\text{C}_{\alpha },\hspace{0.17em}{}^{13}\text{C}_{\beta },\hspace{0.17em}{}^1\text{H}_{\alpha },\hspace{0.17em}{}^1\text{H}_{\mathrm{N}}$ and ¹⁵N chemical shifts using TALOS+ (59). Any nonglycine residues whose dihedral angles were poorly defined in the TALOS+ output were restrained to negative phi angles. Where possible, χ₂ angles for leucine and isoleucine were derived from ¹³C chemical shifts (83, 84), and all leucine and isoleucine side chains were restrained to sterically favorable regions of χ₁,χ₂ conformational space. Hydrogen bond restraints were used in the helices of TAZ2 and TAZ1 for initial structure generation, but all hydrogen bond restraints (with the exception of the α₁ and AD1 helices in TAZ2–p53TAD and α₁ and α₄ helices in TAZ2–p53AD2) were removed before further refinement. Helical regions were identified by comparison with previously determined TAZ2 and TAZ1 structures and using dihedral angle restraints derived from TALOS+ (43, 44). Hydrogen bond restraints for the α₁ helix of TAZ2 and the p53AD1 helix in TAZ2–p53TAD were retained in the final structures to correct bends in the helices that arose from the absence of long-range distance restraints in the N-terminal half of each helix. Hydrogen bond restraints were also maintained in α₄ of TAZ2 in TAZ2–p53AD2 because of a similar lack of restraints in the C-terminal half of the helix. Additionally, for the p53TAD–TAZ1 fusion protein, all distance restraints for residues 28–41 of p53 (region between AD1 and AD2) were loosened by 1 Å in an attempt to more realistically model the flexibility, indicated by the heteronuclear NOE, in this region. Sets of 200 initial structures were generated using CYANA and refined by restrained molecular dynamics with AMBER12 (85, 86). The structures underwent 1,000 steps of energy minimization with 20 ps of simulated annealing in vacuum or in a generalized Born solvent model (61). The system was heated to 1,000 K for the first 2 ps, kept at constant temperature for 4 ps, and then annealed by cooling to 0 K over 12 ps. Force constants for NOE restraints were maintained at 30 kcal⋅mol⁻¹⋅Å^-2, and force constants for angle restraints were 1,000 kcal⋅mol⁻¹⋅rad⁻². Structures and restraints were initially refined using successive cycles of simulated annealing in vacuum. After final refinement using a generalized Born solvent model, the 20 structures with the lowest AMBER energy were selected for analysis. Structure quality was assessed using PROCHECK-NMR (87) and the Protein Structure Validation Software suite (88). Figures were prepared using the PyMOL Molecular Graphics System (Schrödinger, LLC).