Domain–domain interactions in full-length p53 and a specific DNA complex probed by methyl NMR spectroscopy

Abstract

The tumor suppressor p53 is a homotetramer of 4 × 393 residues. Its core domain and tetramerization domain are linked and flanked by intrinsically disordered sequences, which hinder its full structural characterization. There is an outstanding problem of the state of the tetramerization domain. Structural studies on the isolated tetramerization domain show it is in a folded tetrameric conformation, but there are conflicting models from electron microscopy of the full-length protein, one of which proposes that the domain is not tetramerically folded and the tetrameric protein is stabilized by interactions between the N and C termini. Here, we present methyl-transverse relaxation optimized NMR spectroscopy (methyl-TROSY) investigations on the full-length and separate domains of the protein with its methionine residues enriched with ¹³C to probe its quaternary structure. We obtained high-quality spectra of both the full-length tetrameric p53 and its DNA complex, observing the environment at 11 specific methyl sites. The tetramerization domain was as tetramerically folded in the full-length constructs as in the isolated domain. The N and C termini were intrinsically disordered in both the full-length protein and its complex with a 20-residue specific DNA sequence. Additionally, we detected in the interface of the core (DNA-binding) and N-terminal parts of the protein a slow conformational exchange process that was modulated by specific recognition of DNA, indicating allosteric processes.

Intrinsically disordered domains are crucial, functional components of many proteins, especially those involved in cell signaling and regulation of the cell cycle (1, 2), p53 being an archetypical example of such a protein. p53 is a homotetramer of 4 × 393 residues (Fig. S1): Residues 1–93 form the intrinsically disordered N-terminal domain (TAD), with a proline-rich region (PRR) 61–93; 94–292 form the folded core domain (p53C); 293–324 form a disordered linker; 325–353 associate to form the folded tetramerization domain (TET); and 354–393 are intrinsically disordered (CT). The large size of this protein and presence of disordered regions have so far prevented the full-length p53 from being crystallized or its high-resolution structure solved by NMR. The structures of isolated p53 domains have been extensively studied by high-resolution methods (for an overview see ref. 3; Fig. S1); however, their arrangement in the full-length protein has been a subject of controversy, and alternative models of quaternary structural arrangement have been postulated. One model of the human protein obtained from electron microscopy and small-angle X-ray scattering (SAXS) is consistent with structural studies on the isolated TET domain whereby the full-length tetramer and truncated constructs associate via the TET domain (4–6). But in a model of the murine protein, it is proposed that oligomerization occurs via contacts between N and C termini of the protein, as well as the core domains, and the subunits of the TET domain do not oligomerize (7, 8). The latter model assumes large discrepancies between the interactions observed for isolated protein domains in vitro and their arrangement in the full-length context, yielding a model that is inconsistent with the small-angle X-ray–scattering data (4, 6) as well as FRET experiments on the separation of N and C termini of the human protein (9) and NMR studies of an engineered p53 dimer (10, 11). Both models are also contradictory from a functional point of view; they provide different explanations of posttranslational regulation of p53 activity or dominant negative effect of mutations within the protein.

Here, we map domain–domain interactions in full-length tetrameric p53 and its DNA complex by solution-state NMR and assess the conformational dynamics of its domains in the full-length protein. Owing to its large molecular mass, full-length tetrameric p53 is problematic to study by conventional backbone-based NMR experiments and we have simplified the problem by labeling isotopically only the methyl groups of methionines. This approach benefits from favorable NMR properties of methyls (which empower methyl-TROSY signal enhancement, ref. 12) and fortunate location of methionines that are abundant in the interaction interfaces of isolated domain constructs (4, 13–15). By employing point mutagenesis, we were able to assign and compare the NMR spectra of p53 domains with the full-length protein and its DNA complex.

From those experiments, an unbiased and model-free picture of domain–domain interactions emerged. Fundamental for the domain organization of the full-length protein is that the TET residues are present in a fully tetramerized structure. p53C domain interactions further define the global arrangement of p53, which is retained also in the complex with specific DNA duplex. It is interesting to note that in the interface of the p53C and PRR domains, we detected a slow conformational exchange process that is modulated by specific recognition of DNA, suggesting DNA binding might influence the position of N termini comprising the TAD regions.

Results and Discussion

Isotope Labeling and Methyl NMR Spectroscopy.

Full-length p53 as well as isolated p53 domains were selectively labeled with protonated -methionine. We found that the number of cross-peaks in ¹H–¹³C correlation spectra of p53 domains generally corresponded to the number of methionines in the primary sequence (with exceptions of multiple conformations, discussed below) and cross-labeling was not observed. All cross-peaks were unambiguously assigned by mutating methionines to alanine (all but M160) or valine (M160). Assignment of the signals in longer protein constructs was achieved by comparison with spectra of isolated p53 domains.

¹H–¹³C methyl correlations were obtained for all samples in ¹H–¹³C-heteronuclear multiple quantum correlation (HMQC) experiments. Each ¹³C-labeled methionine methyl group gives rise to NMR cross-peaks that have two characteristic features: (i) the chemical shift, which reflects the local environment of the nuclei, as characterized by the ¹H, ¹³C resonance frequencies; and (ii) the line shape that depends on the local and global dynamics. The average spectral line width is an indicator of the global rotational dynamics of a molecule (or a domain); large, slowly tumbling proteins give broad, weak NMR cross-peaks, whereas peptides or highly flexible domains give rise to narrow, intense signals. Consequently, the line widths of individual cross-peaks indicate the site-specific, local dynamics of their corresponding nuclei. Finally, the shape of spectral lines is sensitive to slower fluctuations in the chemical environment: When the frequency of fluctuations is comparable with difference in resonance frequencies of exchanging states (the condition is usually met with μs-ms time scale dynamics), NMR signals become broadened and acquire distorted line shapes. Slower conformational exchange is indicated by appearance of multiple cross-peaks that exhibit resonance frequencies of long-lived conformations (τ_ex usually in range ≥ms) (16).

Methyl NMR Spectra Indicate TET Domain Oligomerization Status.

Structural studies of the isolated TET domain indicate that residues 323–353 associate in vitro to form a dimer of dimers with D2 symmetry and a dissociation constant in the low nM range (17, 18). Primary dimers are stabilized by an antiparallel arrangement of β-strands (residues 327–333), and the dimer–dimer interface is formed by large hydrophobic patches at the orthogonal intersection of α-helical bundles (residues 335–355) (15, 19–21). The oligomerization region contains a single methionine residue (M340) that is located in the middle of the helix and stably trapped inside the tetramerization interface (see Fig. 1A; the wild-type TET domain construct we used also had an additional N-terminal Met, giving rise to an additional signal). Owing to the symmetry of the complex, all four protein chains of the tetramer are chemically equivalent, and thus M340 gives rise to a single methyl cross-peak in a ¹H–¹³C-HMQC spectrum. The hydrophobic environment of the tetramer interior induces a strong upfield shift of the signal (δ_H = 1.4 ppm) (Fig. 1B).

Fig. 1.

Structure and methionine methyl NMR spectra of the TET domain. (A) Isolated p53 peptides comprising residues 323–353 form a tetramer with D2 symmetry comprising of four β-strands and α-helices and dissociating into dimers with K_d = 20 nM (for the native sequence) or 10 μM (for the destabilized L348A mutant). M340 (shown in stick representation) is located in the hydrophobic interior of the tetramer. Dissociation of the domain leads to solvent exposure of the tetramerization interface. (B) ¹H–¹³C-HMQC spectra of the isolated TET domain (red) and the TET L348A mutant (blue). M340 signal appears as a single, upfield-shifted cross-peak in the native TET peptide spectra (cross-peak marked with asterisk originates from additional N-terminal methionine), and as two signals in the TET L384A spectra (the mutant is devoid of the N-terminal methionine). (C) Projections of the TET L384A ¹H–¹³C-HMQC spectra on the ¹H axis: Dilution-induced dissociation of the TET L384A peptide into dimers results in increase of the relative intensity of the “dimeric” M340 signal at 2.0 ppm. Concentrations expressed in terms of monomeric peptide, peak intensities normalized for the 1.6-ppm signal.

Intuitively, the position of the M340 methyl cross-peak should be sensitive to the oligomerization state of the domain. It is known from numerous biophysical studies (17, 18) that isolated TET domain tetramerizes with a K_d in the low nM range, which is far lower than the mM concentrations used in NMR experiments. Accordingly, the dimeric TET domain form is not expected to be detectable by NMR. Indeed, the M340 chemical shift was the same in all native constructs, indicating that M340 was always in a tetrameric state. To test chemical shift perturbations in M340 in the dimeric oligomerization state, we introduced a point mutation L348A, which destabilizes the dimer–dimer interface and increases the K_d of tetramerization to ca. 10 μM for the TET L348A dimer–tetramer equilibrium (22). At high concentrations of the mutant protein, the M340 methyl cross-peak appeared at the upfield-shifted “tetrameric” position (δ_H = 1.6 ppm). Lowering protein concentration induced TET L348A dissociation and resulted in the appearance of a second cross-peak at δ_H = 2.0 ppm, originating from a deshielded M340 methyl in the dimer (Fig. 1C). The coexistence of separate NMR cross-peaks from the tetramer and dimer is an indication that, like in the native TET domain (23), oligomerization kinetics of the L348A mutant is slow (τ_ex≥ms). This observation strongly suggested that similar NMR effects would be associated with dissociation of the native TET peptide. The results clearly indicate no TET dimer is detectably present.

p53C Domain Exists in Two Slowly Interconverting States.

The p53 core domain as defined in the seminal work by Pavletich and coworkers is a globular β-barrel domain formed by residues 94–293 (24). The wild-type core domain of p53 is thermodynamically weakly stable. We routinely use in structural studies a stabilized quadruple mutant of the domain (M133L/V203A/N239Y/N268D) as a pseudo-wild type that has the same DNA-binding properties and activity as the native protein (25, 26).

The five methionines of this p53 core (94–296) construct gave rise to five signals in the ¹H–¹³C methyl spectrum, and variations in their line widths suggest differences in side-chain mobility (Table S1 and Fig. 2; however, we have not found evidence for chemical exchange-related signal broadening in relaxation dispersion NMR experiments).

Fig. 2.

Core domain structure and NMR spectra. (A) Overlay of ¹H–¹⁵N-HSQC spectra of p53C (94–296) (red) and p53C (89–296) (blue). W91 gives rise to two cross-peaks separated by over 1 ppm in ¹⁵N dimension; positions of other cross-peaks are also affected by presence of residues 89–93. (B) ¹H–¹³C-HMQC methionine methyl spectrum of p53C (94–296) (red) and p53C (89–296) (blue). Inset depicts cross-section through the M169 cross-peak, exhibiting signs of slow chemical exchange. (C) X-ray structure of p53C domain (residues 91–289, Protein Data Bank ID: 2XWR); atoms exhibiting in NMR spectra slow chemical exchange are labeled with spheres.

The boundaries of the p53 core domain were recently questioned by the finding that cation-π interaction of R174 and W91 stabilizes the domain (27). Methyl NMR spectra of the N-terminally extended core domain (89–296) spectra are significantly different from the shorter 94–296 construct (Fig. 2B). Interestingly, the presence of residues 89–93 not only changed the chemical environment of M160 and M169, but also induced slow chemical exchange (τ_ex≥ms), indicated by the presence of multiple NMR cross-peaks. The same type of slow dynamics was observed also for the amide group of W91 in the ¹H–¹⁵N-correlation spectrum (Fig. 2A). At 20 °C, the minor set of peaks had ca. 15% of the intensity of the major set. Results of concentration- and temperature-dependence experiments (Fig. S2A) indicate that chemical exchange observed in the extended core domain is caused by an internal slow-conformational dynamics within the domain. The recent X-ray studies of an extended p53C domain (27) and the magnitudes of observed chemical shift perturbations suggest that the second set of NMR signals observed in the methyl NMR spectra results from breaking of the R174–W91 cation-π interaction, which allows the “opening” of the p53C–PRR hinge and increases the solvent exposure of W91 residue. This mechanism is corroborated by methyl NMR spectrum of a p53C (89–296) W91A mutant, which exhibits single set of cross-peaks resembling the “minor” conformation of the native p53C (89–296) (Fig. S2B).

NMR spectra of a longer p53 construct (1–312) comprising intrinsically disordered N termini (TAD, PRR) and p53C reveal, similar to p53C (89–296), multiple appearance of M169 and M160 cross-peaks (Fig. S3 A–C). As the PRR domain forms a relatively stiff polyproline II helix (6), such dynamics is likely to yield significant rearrangements of the N-terminal domains of the protein. This hypothesis is consistent with recent results of single-molecule FRET measurements, indicating the presence of multiple slowly interconverting configurations of the N terminus characterized by different, discrete TAD–p53C distances (9).

Tetramerization of Full-Length p53 Occurs via TET Domain.

The ¹H–¹³C-methyl spectrum of full-length protein was dominated by intense cross-peaks of five methionines from the N- and C-terminal segments of the protein (M1, M40, M44, M66, M384). Their narrow spectral line widths confirmed that those parts of p53 are rather flexible and lack secondary structure (11, 28) (Fig. 3A). A well-separated M340 methyl cross-peak from the tetrameric TET domain is clearly visible even in the spectrum of the full-length protein; however, baseline distortions and strong signals from unstructured regions partially overlapped signals from the core domain (Fig. 3B). We assume that core domains in tetrameric p53 form low-affinity dimers, where M243 has been mapped to the dimerization interface (4). The ¹H–¹³C-HMQC methyl spectrum indeed revealed that the M243 methyl cross-peak becomes broadened by chemical exchange, which is in agreement with previous NMR data obtained for the backbone amide groups in ¹H–¹⁵N-TROSY experiments (4).

Fig. 3.

Methionine methyl NMR spectra of the full-length p53. (A) Spectrum is dominated by signals from the unstructured N and C termini of the protein; M66 appears as double cross-peak because of cis-trans isomerization of neighboring Pro. (B) The same spectrum plotted with lower contour level (red) overlayed with a spectrum of the protein with unstructured methionines (M1, M40, M44, M66, M384) mutated to alanines (blue). (C) Overlay of the full-length p53 spectrum (M1, 40, 44, 66, 384A) (blue) with spectra of isolated p53C (89–296) (orange) and TET (323–353) (green); additional N-terminal Met from the TET construct marked with an asterisk. (D) Schematic representations of two quaternary organization models postulated for the full-length p53; NMR spectra confirm presence of tetramerized TET domain, loosely arranged p53C domain dimers, and unstructured N and C termini, therefore being in agreement with the models based on structures of isolated p53 domains (4, 6). Alternative model (7) implying existence of structured N/C nodes comprised of monomeric TET and TAD domains has no confirmation in NMR spectra.

Although structures of the TET and p53C domains remained the same, their mobility and global dynamic properties were different in the full-length protein. All the NMR signals from the p53C and TET domains were significantly broadened indicating their impaired mobility. The line widths of the cross-peaks from the p53C (molecular mass of the isolated domain 25 kDa) were broadened to a lower extent than the signal from the TET domain (molecular mass 12 kDa) (Table S1). Those changes imply that the TET domain is rotationally more constrained than the p53C domains, which retain a large degree of rotational independence. These results are supported by the SAXS-based model of p53, in which stably tetramerized TET domain is effectively constrained by eight polypeptide chains, whereas the only weakly dimerizing p53C domains are flanked by two polypeptide linkers (4).

TET Domain Remained Tetramerized in DNA Complex.

p53 specifically recognizes DNA sites containing two copies of the decameric motif RRRC(A/T)|(T/A)GYYY (R, purine; Y, pyrimidine) (29). The contiguous consensus sequence used in NMR measurements contains two half-sites in the form of four inverted 5-bp quarter sites (→ ←  → ←), giving a total length of 20 bp. Earlier NMR studies of the full-length p53–DNA complex based on amide backbone resonances were hampered by the fact that the slow tumbling of the complex and restricted rotational dynamics of core domains lead to a loss of NMR signals (30). The ¹H–¹³C-HMQC methyl spectrum of the p53–DNA complex was dominated by five residues from the N and C termini that retain their disordered character after DNA binding (Fig. 4A), but the remaining cross-peaks from p53C and TET domains were also visible. Despite the lower intensity signals, it was clear that the TET domain remained tetramerized (indicated by the upfield-shifted M340 methyl cross-peak) in the DNA complex (Fig. 4 B and C and Fig. S3E). Use of a full-length p53 construct devoid of flexible N- and C-terminal methionines enabled us to obtain good-quality spectra of the complex with well-resolved ¹H–¹³C methyl NMR cross-peaks from the p53C domains.

Fig. 4.

Methionine methyl NMR spectra of the p53–DNA complex. (A) Spectrum is dominated by the signals from unstructured N and C termini. Inset depicts overlay with a spectrum of free p53 (green) plotted on the same level. (B) Overlay of the full-length p53–DNA complex spectrum (M1, 40, 44, 66, 384A) (blue) with spectrum of p53 (M1, 40, 44, 66, 384A) (green); narrow signal marked with an asterisk comes from a residual impurity that is insufficiently removed by gel filtration; M237, M246, and M160 signals are broadened below the noise level. (C) Spectrum of p53 (M1, 40, 44, 66, 384A)–DNA complex (blue) is matched by spectra of isolated TET domain (green) and p53 (1–312; M1, 40, 44, 66A)–DNA complex (orange). (D) Schematic representations of two quaternary organization models postulated for the p53–DNA complex; NMR spectra confirm presence of tetramerized TET domain, DNA-bound p53C domains, and unstructured N and C termini, being in agreement with the models built from structures of isolated p53 domains (4–6). Model in which DNA recognition is mediated by a pair of p53C and a pair of N/C nodes (8) has no evidence in NMR spectra.

Even though none of methionines from the core domain is directly involved in nucleic acid recognition, previous X-ray and NMR studies of p53C–DNA complexes show that the formation of stable p53C dimers on DNA should affect the chemical environment of M243 (13, 14, 31, 32). The methyl NMR spectrum corroborated that change: The signal of M243 split into two broad cross-peaks, and the “monomeric” cross-peak that is prevalent in free p53 spectra disappeared completely (one M243 signal was overlayed with the “dimeric” M243 from free p53 spectra and the other was downfield-shifted). Although disappearance of the “monomeric” signal indicated that all four p53C domains interacted with DNA, the splitting pattern of M243 was difficult to interpret in the light of X-ray studies of p53 complexes with 20-bp nucleotides (31, 32). Symmetry of the crystal lattice in those structures implies chemical equivalence of M243 and L3 loops in the DNA complexes with contiguous p53-binding sites (Fig. S1D). This contrasts with relatively high mobility of M243 and its appearance as multiple chemically distinct populations, which might indicate presence of slow (τ_ex≥ms) conformational dynamics in the L3 loop.

Although we observed chemical shift perturbations associated with formation of primary p53C dimers for M243, we did not observe analogous effects for M169, which is located in proximity of the dimer–dimer interface (Fig. S1D). This suggests a more transient character of the dimer–dimer interaction, which is also reflected by high affinity of p53 to noncontinuous DNA sequences with up to 13 bp spacers between the half-sites (29).

As with the free full-length protein, the methyl NMR spectrum of the DNA–p53 complex was well-reproduced by superposition of the spectra of isolated p53C–DNA complex and tetramerized TET domain (Fig. 4C). The line widths of the signals indicated that binding of DNA immobilized the p53C domains and constrained them rotationally more than the TET domain (Table S1). Those observations combined with tetrameric state of the TET domain and unstructured character of TAD and CT support the quaternary structure model based on combination of crystal structures of p53C–DNA complex and TET domain (4, 5, 33). In our experiments, TET appeared loosely attached to the p53C–DNA conglomerate and, because of its flexibility, is unlikely to contribute directly structurally to the formation of the complex. The role of the domain is, rather, to act as a tether for p53C domains, enabling cooperative DNA binding by p53C domains. Such an arrangement would permit p53 to accommodate different spacers between the half-sites without breaking the tetramer.

In order to limit the increase in the hydrodynamic radius upon formation of the DNA–p53 complex, we used a short oligonucleotide (20 bp), limited to four p53C quarter-sites. This construct did not allow us to observe unspecific interactions of the CT domain with DNA segments flanking the p53 recognition site, as reported elsewhere (5, 34). It is, however, possible that in the presence of longer DNA, interaction with additional bases might act as an additional constraint and bring the TET domain closer to p53C domains.

Multiple cross-peaks for M169 define PRR–p53C hinge dynamics, which in free p53 support the presence of two asymmetrically populated states (Fig. S2A). Upon DNA binding, the intensities of M169 cross-peaks equalize (Fig. S2C). This observation is intriguing, because it indicates coupling of DNA binding with structural rearrangements in the remotely located p53C–PRR interface. The atomic details of this transition remain unknown, but X-ray structures of the p53C–DNA tetramers (31, 32) reveal that the symmetry of the complex places the N termini of two p53C domains close to the dimer–dimer interface (Fig. S1D) and tetramerization of p53C domains on DNA may account for the observed NMR effects.

Conclusions

Methionine methyl labeling allowed us to record high-quality NMR spectra of tetrameric p53 and gave us an unprecedented opportunity to observe site specifically the environment and assess dynamics at 11 sites simultaneously. Using this approach, we were able to integrate previous NMR and biophysical studies on isolated p53 domains and extend the NMR studies to the full-length p53–DNA complex. Our data resolve conflicting models of p53 quaternary structure in favor of those that have tetrameric TET domains, and also reveal a previously unknown p53C–PRR interaction that occurs during DNA recognition. There are no significant deviations in chemical shifts of individual domains, apart from effects associated with formation of p53C primary dimers in tetrameric p53 constructs, and spectra of full-length protein are well-reproduced by superposition of individual domains. Thanks to the small number of methionines in the primary sequence, the NMR spectra of the full-length protein and DNA complex are simple and straightforward to interpret, and so encourage similar investigations of other p53 complexes as well as employing other methyl probes that cover areas not populated by methionines.

Materials and Methods

Isotope Labeling.

All proteins were produced in Escherichia coli C41(DE3) pRARE2 transformed with pET24a-derived vectors. Isotope labeling was achieved by growing the cells in M9 medium with defined isotope composition. Specific methyl labeling was achieved by supplementation of the medium with 100 mg/L of protonated -Met (35). Perdeuteration was performed by growing the cells in M9 medium prepared using 99% D₂O and protonated d-glucose, leading to a protein in which ca. 80% of protons were replaced by deuterons (as estimated from a 1D ¹H NMR spectrum). Expression of the recombinant protein was normally induced at A₆₀₀ = 0.8 using 1 mM IPTG, and cells were additionally supplemented with 100 μM ZnCl₂ and cultivated at 25 °C for 4–16 h.

Protein Purification.

Proteins were expressed as His-lipoyl domain fusions [full-length p53, p53C (94–296) and (89–296), TET L348A (325–355), p53 TAD–PRR–C (1–312), p53C–TET (96–355), and alanine point mutants of the proteins] were purified by combination of Ni²⁺-, heparin-affinity, and gel-filtration chromatography (6, 22, 27, 36). Briefly: Cell lysis was performed using high-pressure homogenizer in a lysis buffer containing 25 mM K-phosphate, 300 mM NaCl, 10 mM imidazole, 5 mM β-mercaptoethanol, pH = 8.0. Clarified lysate was loaded on HiTrap crude column and equilibrated with the lysis buffer, after which proteins were eluted with linear gradient of 10–300 mM imidazole. Fractions containing recombinant protein were pooled; mixed with tobacco etch virus (TEV) protease; dialyzed overnight against 20 mM Tris (pH 7.4) and 5 mM β-mercaptoethanol; loaded onto HiTrap Heparin HP (all except TET L348A) or HiTrapQ FF (TET L348A) column; and eluted with a linear gradient of 0–1 M NaCl. Fractions containing protein of interest were concentrated and subjected to final size-exclusion chromatography on Sephadex S200 column equilibrated with 25 mM Na-phosphate, 150 mM NaCl, and 10 mM DTT, pH = 7.2.

The TET domain (residues 323–353) was expressed without additional tags and purified from bacterial lysate by anion-exchange chromatography (using a HiTrap-Q column) followed by gel filtration on Sephadex S200. Because the protein did not contain N-terminal tags, no TEV cleavage was performed and, consequently, an extra N-terminal methionine was present in the final product.

All the protein constructs encompassing the p53C domain contained four stabilizing mutations: M133L/V203A/N239Y/N268D (25).

NMR Sample Preparation.

The final step of preparation of all NMR samples was buffer exchange using 2-mL Zeba Spin Desalting Columns (Pierce) to a standard NMR buffer consisting of 18 mM Na₂HPO₄, 7 mM NaH₂PO₄, 150 mM NaCl, 0.1% NaN₃, and 10 mM DTT (for protein samples containing free cysteines) prepared with 99.9% D₂O. DNA complexes were preformed by addition of palindromic 20-mer DNA oligonucleotide containing four pentameric DNA-binding sites (sequence GAACATGTTCGAACATGTTC) and subsequent gel-filtration chromatography on Sephadex S200.

NMR Methods.

All spectra were recorded at 293 K on Bruker Avance II+ 700-MHz, Bruker Avance III 600-MHz, or Bruker Avance 800-MHz NMR spectrometers equipped with TXI cryoprobes. ¹H–¹⁵N and ¹H–¹³C correlations were obtained with ¹H–¹⁵N-fHSQC (37) and ¹H–¹³C-HMQC (38) experiments, respectively. For the full-length protein constructs, typically a 1,024- by 256-point matrix was recorded with 512 transients per row, giving a total experimental time of 72 h per spectrum. Relaxation dispersion spectra of a single-quantum– coherence were obtained according to Lundstrom et al. (39). Spectra were processed with Topspin 3.0 (Bruker) and plotted with Sparky 3.1 (University of California, San Francisco).