The Structural and Dynamic Basis of Ets-1 DNA Binding Autoinhibition*

The transcription factor Ets-1 is regulated by the allosteric coupling of DNA binding with the unfolding of an α-helix (HI-1) within an autoinhibitory module. To understand the structural and dynamic basis for this autoinhibition, we have used NMR spectroscopy to characterize Ets-1ΔN301, a partially inhibited fragment of Ets-1. The NMR-derived Ets-1ΔN301 structure reveals that the autoinhibitory module is formed predominantly by the hydrophobic packing of helices from the N-terminal (HI-1, HI-2) and C-terminal (H4, H5) inhibitory sequences, along with H1 of the intervening DNA binding ETS domain. The intramolecular interactions made by HI-1 in Ets-1ΔN301 are similar to the intermolecular contacts observed in the crystal structure of an Ets-1ΔN300 dimer, confirming that the latter represents a domain-swapped species. 15N relaxation studies demonstrate that the backbone of the N-terminal inhibitory sequence is mobile on the nanosecond-picosecond and millisecond-microsecond time scales. Furthermore, hydrogen exchange measurements reveal that amide protons in helices HI-1 and HI-2 exchange with water at rates only ∼15- and ∼75-fold slower, respectively, than predicted for an unfolded polypeptide. These findings indicate that inhibitory helices are only marginally stable even in the absence of DNA. The energetic coupling of DNA binding with the facile unfolding of the labile HI-1 provides a mechanism for modulating Ets-1 DNA binding activity via protein partnerships, post-translational modifications, or mutations. Ets-1 autoinhibition illustrates how conformational equilibria within structural domains can regulate macromolecular interactions.

Gene expression can be controlled by modulating the DNA binding affinity of sequence specific transcription factors. Similar to several other transcription factors, the DNA binding of Ets-1 is modulated by an autoinhibitory module that provides a route to biological regulation (1). The Ets-1 inhibitory module is composed of sequences flanking the winged helix-turn-helix (HTH) 1 DNA binding ETS domain (2,3). When these sequences are deleted, as in an alternatively spliced isoform of Ets-1, or when their structural elements are disrupted by mutations, as in the case of the oncogenic v-Ets, the affinity of Ets-1 for its target DNA sites is enhanced by 10-to 20-fold (4 -6). In a cellular context, this module is essential for response to different regulatory signals. DNA binding of Ets-1 is enhanced 10-to 20-fold through a partnership with the transcription factor RUNX1 (CBF␣2/AML1) (7). Conversely, in activated T-cells, phosphorylation of a serine-rich region (residues 244 -300) inhibits the DNA binding of Ets-1 by another ϳ50-fold (8). Importantly, these two effects require an intact inhibitory module.
Mechanistic insight into autoinhibition has come from the observation that the Ets-1 inhibitory module changes conformation upon binding to DNA. Initial secondary structural studies performed in our laboratories demonstrated that this module is composed of four coupled ␣-helices, located N-terminal (HI-1, HI-2) and C-terminal (H4, H5) to the ETS domain (Fig.  1A). Chemical shift data from NMR spectra comparing an inhibited fragment, Ets-1⌬N280, 2 to one missing the N-terminal inhibitory sequences, Ets-1⌬N331, mapped the contacts of HI-1 and HI-2 to the first helix of the ETS domain (H1) and to the C-terminal inhibitory region near H4 and H5 (9,10). This implicated helix H1 in the autoinhibitory mechanism due to its role in both directly contacting DNA and in packing of the inhibitory helices (11). Importantly, these contacts also lie on a contiguous surface of the protein opposite the DNA recognition helix (H3), indicating that autoinhibition does not result from a simple steric blockage of DNA binding. Instead, limited proteolysis and CD spectroscopic studies revealed that HI-1 unfolds upon complex formation with DNA (2,3). Together, these data led to the proposal that autoinhibition results from the allosteric coupling of a helix-coil transition with DNA binding, and that the affinity of Ets-1 for DNA is inversely correlated with the stability of the inhibitory module.
X-ray crystallographic studies have extended the framework for understanding the autoinhibition phenomenon. A crystal structure of Ets-1⌬N280 in complex with DNA demonstrated that helix HI-1 is disordered, whereas HI-2, H4, and H5 are coupled and contact the core ETS domain via H1 (12). The crystal structure of a second, partially inhibited fragment, Ets-1⌬N300, solved in the absence of DNA, showed a conformation similar to bound Ets-1⌬N280, but with a folded HI-1 (12). Ets-1⌬N300, however, crystallized as a dimer, with HI-1 of one subunit interfaced with HI-2, H4, and H5 of an adjacent subunit. 3 This dimeric state was not expected, and indeed, subsequent mutagenic analyses suggested that the in trans position of HI-1 in this crystal approximated its in cis position in the monomeric protein in solution. Thus, Ets-1⌬N300 appeared to crystallize as a three-dimensional domain-swapped dimer (13). Both of these structures also displayed features of the ETS domain-DNA interface similar to those observed in several investigations of related Ets proteins (14 -17), including the role of H3 in major groove binding and direct sequence recognition, along with the involvement of H1 in a hydrogen-bonded phosphate contact (11,18,19).
Although these extensive structural and functional studies have laid a foundation for understanding the mechanism of autoinhibition, several key questions remained unanswered. What is the structure of the inhibitory module in the unbound state of the native Ets-1 monomer? What are the features of HI-1 that allow a facile helix-coil transition, and does this transition occur even in the absence of DNA? Finally, what couples DNA binding to the inhibitory action of helices that lie distant from the recognition helix of the ETS domain?
To address these questions, we have studied the partiallyinhibited Ets-1⌬N301 in its DNA-free form using NMR spectroscopy. Our findings yielded the first high-resolution structure of a monomeric Ets-1 fragment that displays the intramolecular packing of all inhibitory helices. In parallel, backbone amide 15 N relaxation and hydrogen exchange data for Ets-1⌬N301 demonstrate that HI-1 is only marginally stable and thus poised to unfold. Additional regions of this protein, including the DNA binding interface, also display conformational mobility suggestive of a role in the allosteric mechanism of autoinhibition.
Assignments of the Ets-1⌬N301 backbone and side chain 1 H, 13 C, and 15 N resonances were obtained using an extensive set of multidimensional NMR experiments (25) acquired on protein samples uniformly 15 N-, 15 N/ 13 C-, or 2 H/ 15 N/ 13 C-labeled, as well as selectively labeled with 13 C ␤ -Ala or with ring-deuterated Tyr, Phe, or Trp. Full assignments of only nine residues, located within the HI-1/HI-2 loop (Lys 318 , Pro 319 , Ile 321 , Pro 322 ), HI-2 (Ala 323 , Thr 330 ), H3 (Tyr 397 , Asp 398 ), and between H3 and S3 (Lys 399 ) were not obtained because of exchange broadening and/or spectral overlap. Stereochemical assignments of the signals from non-degenerate ␤-methylene protons (37 of 104 possible) were based upon HNHB and short mixing time 15 N-TOCSY-HSQC ( m ϭ 38 ms) experiments, while those of side chain Gln and Asn amide protons (7 of 8 pairs) were obtained from an EZ-HMQC-NH 2 spectrum (26). Complete stereochemical assignments of the Val and Leu methyl group resonances were determined from a constant time 1 H-13 C HSQC spectrum acquired on a non-randomly 10% 13 C-labeled sample (21).
Structure Calculations-Structural calculations were performed using ARIA/CNS version 1.2 with torsional angle dynamics (27,28). NOE restraints were determined from a 1 H- 15 (29). Most NOE cross-peaks were manually picked and assigned prior to intensity calibration in ARIA.
Backbone dihedral angles were determined from C ␣ , C ␤ , CЈ, H ␣ , and H N chemical shifts using TALOS (30). The 1 side chain dihedral angles for residues with ␤-methylene protons were determined from a conservative examination of short mixing time 15 N TOCSY-HSQC ( m ϭ 38 ms) and HNHB spectra, according to a staggered rotomer model. Side chain 1 dihedral angles for Ile, Val, and Thr residues were derived on the basis of 3 J NC␥ and 3 J C'C␥ coupling constants measured from 13 C-{ 15 N} and 13 C-{ 13 CЈ} spin echo difference CT-HSQC spectra (31). All Xaa-Pro amides were constrained to the trans conformation, based upon their proline 13 C ␣ chemical shifts (32), as well as the absence of (Xaa)H ␣ -(Pro)H ␣ interactions in NOESY spectra. The charge states of the His imidazoles were not experimentally determined and thus were set to the default fully protonated form. A limited set of hydrogen bond restraints were included for amides located in helices or strands, as determined via NOE and chemical shift information and protection from hydrogen-deuterium exchange.
Residual dipolar couplings (RDCs) were acquired on a uniformly 15 N-labeled protein sample diffused into a 5% acrylamide gel (29:1 acrylamide:bisacrylamide stock) (33). The gel was cast at a 6 mm diameter and ϳ20 mm length, dialyzed against NMR sample buffer, and soaked in a solution containing labeled protein overnight, giving a calculated final Ets-1⌬N301 concentration of ϳ0.1-0.2 mM. Prior to spectral acquisition, the pellets were loaded into a 5-mm silanized bottomless NMR tube using a gel stretching apparatus (34) purchased from New Era Enterprises, Inc. The gel stretched ϳ2.1 times its original length, yielding splittings of ϳ5 Hz in a 2 H-NMR spectrum of the 1 HO 2 H lock solvent. Backbone 1 H-15 N RDCs were measured from the 1 H-15 N-IPAP-HSQC spectrum (35) of the partially aligned protein, relative to that of a reference spectrum recorded for unaligned Ets-1⌬N301. 49 RDC restraints, ranging from Ϫ17.8 to ϩ11.5 Hz for amides in the helical and ␤-sheet regions of Ets-1⌬N301 with well resolved NMR signals, were used for the structural calculations. Axial and rhombic parameters were estimated using a histogram method, followed by a grid search to determine the optimal values of D a (Ϫ8.9 Hz) and R (0.3) resulting in minimal SANI violations during the final energy minimization step of ARIA/CNS (36).
The Ets-1⌬N301 structure ensemble was calculated using a two-step ARIA protocol. An initial round of ARIA (it0 3 it8), starting with an unfolded polypeptide, was completed with only dihedral angle and hydrogen bond restraints and uncalibrated NOE data. A second full ARIA calculation (it1 3 it8) followed, with RDC restraints included, using the prefolded structures and the unambiguous and ambiguous NOE restraint sets from the previous round as the starting parameters. A total of 50 structures were calculated per iteration during each round, of which the energetically top 25 from the final iteration were further refined in a water box using Lennard-Jones potentials. The ARIAderived unambiguous and ambiguous NOE distance restraints were evaluated using VMD-Xplor (37). Analyses of the structures were performed using MolMol (38), InsightII (Accelrys, Inc.), and Procheck-NMR (39). Secondary structure boundaries for the ensemble were defined according to Promotif (40) and Vadar (41).
Amide Hydrogen Exchange-Slow amide proton-deuterium exchange rates (28°C, pH* 6.5) were measured from a series of sensitivityenhanced 1 H-15 N HSQC spectra recorded after rapid transfer of wild type 15 N-labeled Ets-1⌬N301 through a Sephadex G-25 spin column equilibrated with NMR sample buffer prepared in 99% D 2 O (50). The first spectrum, started ϳ8 min. after transfer, was acquired in 11 min, while subsequent spectra were recorded with increasing numbers of transients per t1 increment for improved signal-to-noise. After scaling according to the number of transients, pseudo-first order rate constants for exchange, k ex , were obtained by least squares fitting of peak intensity, I t , versus the mid-point time of each spectrum to the equation I t ϭ I o exp(Ϫk ex t) ϩ I ϱ , where I ϱ accounts for the effect of residual protonated water (ϳ5%).
Rapid amide proton-proton exchange rates at 28°C and pH 6.86 to 8.25 were determined for wild-type 15 N-labeled Ets-1⌬N301 by the CLEANEX-PM method (51) using a series of 6 spectra with transfer periods ranging from 10 to 60 msec. Each spectrum was recorded in ϳ5 h using a recycle delay of 2.05 s. Reference spectra were recorded in ϳ5 h using a recycle delay of 12.1 s to ensure complete relaxation. For improved signal-to-noise, duplicated data sets were measured and merged at both pH 7.50 and 7.74. Pseudo-first order rate constants for chemical exchange, k ex , were obtained by least squares fitting of peak intensities to published equations (51), using a Matlab module provided by W.-Y. Choy (University of Toronto) and assuming a negligible contribution from water relaxation.
Predicted exchange rates, k pred , for an unstructured polypeptide with the sequence of Ets-1⌬N301 were calculated with the program Sphere (52) using poly-DL-alanine reference data corrected for amino acid type, pH, temperature, and isotope effects (53,54).

Spectral Assignments of Ets-1⌬N301-This
NMR study was performed on Ets-1⌬N301, a truncation fragment of Ets-1 spanning residue 301 to its native C terminus. The affinity of Ets-1⌬N301 for a consensus DNA sequence is ϳ2-fold inhibited relative to Ets-1⌬N331, the fully de-repressed fragment lacking the entire N-terminal inhibitory sequence, as compared with ϳ10fold for both Ets-1⌬N280 and full-length Ets-1 (data not shown). Thus, although retaining all of the inhibitory helices (Fig. 1A), Ets-1⌬N301 is only partially-inhibited due to the absence of residues 280 -300. While essential for full repression (2), these residues are conformationally disordered (9) and their deletion substantially improved the amenability of the resulting Ets-1⌬N301 toward detailed NMR spectroscopic studies.
Nearly complete backbone and side chain 1 H, 13 C, and 15 N resonance assignments of Ets-1⌬N301 were obtained through a suite of 1 H/ 13 C/ 15 N scalar and NOE-correlated experiments (Fig. 1B). Critical to this process was uniform 2 H isotopic labeling, which helped to enhance spectral sensitivity and resolution. Selective 2 H isotopic labeling of aromatic residues also aided in the resonance assignments of their side chain nuclei, while selective 13 C ␤ -alanine labeling of Ets-1⌬N301 facilitated analysis of the poorly resolved signals from 323 AAALA 328 within the N-terminal inhibitory region. These extensive assignments permitted the detailed structural and dynamic analyses of Ets-1⌬N301.
NMR-based Structure Determination of Ets-1⌬N301-The tertiary structure of Ets-1⌬N301 was calculated with ARIA/ CNS (27, 28) using manually-picked NOE peak lists from an extensive set of three-dimensional 15 N-and 13 C-resolved NOESY spectra. Calculations also included backbone and side chain dihedral angles, along with a limited set of hydrogen bond and 1 H-15 N RDC restraints corresponding to residues within elements of regular secondary structure. A summary of the refinement statistics for the energetically best 25 final structures is listed in Table I, and a plot of backbone and side chain r.m.s.d. values is given in Supplemental Fig. S1.
The Ets-1⌬N301 structural ensemble extends previous studies by providing the first picture of the intramolecular packing between the autoinhibitory module and the ETS domain. The N-and C-terminal inhibitory helices (HI-1, HI-2, H4, and H5) are structurally coupled to one another and pack with the first helix (H1) of the ETS domain (Fig. 2, A and B). The N-terminal inhibitory helix, HI-1, lies approximately anti-parallel to HI-2, parallel to H1 and H5, and abuts H4 in a tail-to-head fashion. This tertiary packing is consistent with comparisons of the HSQC spectra of Ets-1⌬N280 and Ets-1⌬N331 (9) and of Ets-1⌬N301 and Ets-1⌬N331 (data not shown), which reveal significant amide chemical shift perturbations (i.e. ⌬␦ Ͼ 0.1 ppm for 1 H) for residues within H1 and the C-terminal inhibitory sequence because of the presence of the N-terminal inhibitory sequence. In contrast, the chemical shifts of amides within the remainder of the ETS domain correspond more closely between the two fragments, indicating that the inhibitory module, located on the opposite face of the molecule, does not markedly perturb the structure of the winged HTH DNA binding motif.
The inhibitory module structural ensemble also exhibits several notable polar interactions. First, HI-1 and H4 lie in a skewed tail-to-head arrangement at an angle of 50.3 o Ϯ 2.5 o , such that the N terminus of HI-1 directly abuts the C terminus of H4 (Fig. 2, A and B), likely allowing a favorable interaction between the two helical macrodipoles. Furthermore, based on FIG. 1. A, secondary structure of Ets-1⌬N301 in context of full-length Ets-1. The ETS domain (black) is composed of three ␣-helices (H1, H2, H3), and four ␤-strands (S1, S2, S3, S4), while the inhibitory module (gray) includes two Nterminal (HI-1, HI-2) and two C-terminal (H4, H5) helices. Multiple phosphorylation of a serine-rich region (residues 244 -300) reinforces autoinhibition (8). The PNT domain, which serves as an ERK2 docking site (74,75), has no apparent role in DNA binding. B, assigned 1 H-15 N HSQC spectrum of murine Ets-1⌬N301 (pH 6.5, 28°C). The inset corresponds to the crowded region at the center of the spectrum. Aliased peaks ( ‡), tentative assignments (*), and weak peaks not observed at this contour level (ϩ) are indicated. The signals from Asn and Gln side chain 15 NH 2 groups are suppressed because of the use of a refocused INEPT within this HSQC pulse sequence. NOE interactions involving surrounding residues, potential carbonyl oxygen-amide nitrogen hydrogen bonds between Leu 421 and Phe 304 and between Leu 422 and Lys 305 in these two helices are consistently observed in the calculated Ets-1⌬N301 structure ensemble. Strikingly, in the crystalline Ets-1⌬N300 dimer model, HI-1 of one monomer and H4 of the other appear to form a continuous helix, with a bend of 18.6 o Ϯ 0.8 o , that is also stabilized by hydrogen bonds between Leu 421 -Phe 304 and Leu 422 -Lys 305 . Second, in addition to this helix-helix stacking, the Lys 305 and Glu 428 side chains appear to link HI-1 and H5 via a potential salt bridge observed within many members of the NMR-derived Ets-1⌬N301 structure ensemble (Fig. 2C). A similar intermolecular electrostatic interaction is also found in the crystalline Ets-1⌬N300 dimer. Finally, Lys 316 and Lys 318 in the HI-1/HI-2 loop may form salt bridges with Glu 343 and Asp 347 , respectively, along helix H1. These later interactions are not observed in the crystal dimer due to the role of this loop as an intermolecular bridge.
In conclusion, the NMR-derived structure of Ets-1⌬N301 defines the fold and stabilizing interactions of the native monomeric form of Ets-1 inhibitory module. The intramolecular packing of helix HI-1 with the other inhibitory helices is consistent with experimental data regarding the architecture and function of the inhibitory module (2,3,9,11,18). Furthermore, the similarities between the intramolecular interactions in Ets-1⌬N301 and the intermolecular interactions in crystalline Ets-1⌬N300 validate the proposal that the latter represents a three-dimensional domain-swapped dimer with the in trans position of HI-1 recapitulating its in cis position in a monomeric state (12,13).
Helix HI-1 Unfolds in the Ets-1⌬N301/DNA Complex-The mechanism of DNA binding autoinhibition is proposed to involve the unfolding of inhibitory helix HI-1. The hypothesis was developed based on the observed increase in the susceptibility of Ets-1 to tryptic cleavage at the C terminus of HI-1 (Arg 309 and Arg 311 ) and a loss of ␣-helical signal in its CD spectrum upon binding to DNA (2,3). In support of this model, residues 280 -308 are disordered in the crystal structure of the Ets-1⌬N280/DNA complex, whereas residues 310 -318 have poorly defined electron density (12). To further investigate this conformational transition in solution, we prepared a complex of Ets-1⌬N301 and a 16-bp DNA duplex encoding a high affinity binding site. NMR-monitored titrations of Ets-1⌬N301 with this DNA confirmed that a tight 1:1 complex formed in the slow-exchange regime relative to the chemical shift time scale (data not shown).
DNA binding dramatically alters the NMR spectrum of Ets-1⌬N301. Although not assigned, the 1 H-15 N HSQC spectrum of the Ets-1⌬N301/DNA complex (Supplemental Fig. S2) reveals well-dispersed amide resonances that can be attributed to the folded ETS domain by way of comparison with data reported for a minimal-sized (residues 320 -415) human Ets-1 ETS domain/ DNA complex (55). However, the spectrum also contains resonances with poorly resolved amide chemical shifts characteristic of a random coil polypeptide that are not observed with uncomplexed Ets-1⌬N301. Furthermore, these amides exhibit low heteronuclear 1 H{ 15 N}-NOE values indicating that they are conformationally mobile on the sub-nanosecond time scale (44). In contrast, a study of selectively 15 N-Tyr-labeled Ets-1⌬N331 demonstrated that Tyr 424 , which lies in the C-terminal inhibitory region, is not perturbed upon DNA binding (56). These results support the model that HI-1 is predominantly unfolded in the Ets-1⌬N301/DNA complex, whereas the C-terminal inhibitory helices are intact. The susceptibility of HI-2 toward cleavage by chymotrypsin increases in the presence of DNA, suggesting that it may become at least transiently unfolded (57). However, because of the lack of spectral assignments for DNA-bound Ets-1⌬N301, we are unable to determine if the conformation of HI-2 is also perturbed or if it remains folded, as observed in the Ets-1⌬N280/DNA crystal complex.
Dynamic Properties of Ets-1 ⌬N301-The fully assigned HSQC spectrum of Ets-1⌬N301 provided an opportunity to further investigate the dynamic properties of the inhibitory helices in the absence of DNA. The modest ϳ10and ϳ2-fold inhibition of DNA binding by Ets-1 and Ets-1⌬N301, respectively, predicts that the helix-coil transition of the N-terminal inhibitory region is relatively facile. A corollary of this prediction is that helix HI-1 is only marginally stable, even in the absence of DNA. To test these hypotheses, 15 N heteronuclear relaxation, relaxation-dispersion, and backbone amide hydrogen exchange measurements were performed to investigate the dynamic properties of Ets-1⌬N301.
(a) 15 N Heteronuclear Relaxation: To investigate the global hydrodynamic properties and fast time scale internal dynamics of Ets-1⌬N301, 15 N T 1 and T 2 lifetimes and steady-state 1 H{ 15 N}-NOE values were measured for 125 of the 132 backbone amides in this protein (Fig. 3A) (42,44). Excluding amides exhibiting anomalous T 1 /T 2 ratios or 1 H{ 15 N}-NOE values indicative of internal mobility, the average T 1 and T 2 lifetimes measured at a 1 H frequency of 500 MHz were 598 Ϯ 31 ms and 82 Ϯ 8 ms, respectively. These values correspond to an effective correlation time of 9.7 Ϯ 0.05 ns for the global tumbling of Ets-1⌬N301, confirming that this 16.3 kDa protein is monomeric in solution (58). After consideration of anisotropic diffusion, the internal dynamic properties of the Ets-1⌬N301 backbone can be described by the extended Lipari-Szabo model-free formalism in terms of a generalized order parameter, S 2 , which decreases from 1 to 0 with decreasing spatial restriction of the NH bond vector, as well as by additional terms including those accounting for the effects of slower conformational exchange  The orientation of H5 with respect to the ETS domain is reversed from that initially calculated for the NMR-derived Ets-1⌬N331 structure (PDB accession code 1ETC) because of misinterpretations of the poorly resolved spectra of the latter protein recorded with a 500 MHz spectrometer (56). C, surface view of the average Ets-1⌬N301 water-refined structure, focusing on the inhibitory module. Residues from HI-1 are indicated in orange, HI-2 in cyan, H1 in red, H4 in dark blue, and H5 in light yellow. The loop between HI-1 and HI-2 is displayed in gold, whereas the loop linking H4 and H5 is colored blue. Note that as drawn, the DNA binding interface, centered on the H2-turn-H3 motif, is on the backside of Ets-1⌬N301. D, ribbon diagram showing primarily the side chains contributing to the hydrophobic core of the inhibitory module. The coloring corresponds to that of panel C, while the molecule is rotated left about the vertical axis by ϳ90 o . The figures were generated with MolMol (38). broadening (R ex ) (43,45,47). The fit S 2 and R ex values are presented in Fig. 3B, and all model-free terms tabulated within Supplemental Table SI. The relatively uniform relaxation parameters and S 2 values for the ␣-helices and ␤-strands in Ets-1⌬N301 indicate that the inhibitory module and ETS domain behave as a single, well-folded globular entity.
Local differences in the 15 N relaxation parameters measured for Ets-1⌬N301 reveal diversity in backbone mobility between different regions of the molecule. Within the N-terminal inhibitory sequence, two segments exhibit 15 N relaxation properties indicative of local dynamic properties that are distinct from the average behavior of Ets-1⌬N301 (Fig. 3, A and B). First, the amides of residues Arg 311 -Val 320 in the exposed loop between HI-1 and HI-2 display significantly longer T 2 times, smaller heteronuclear NOEs, and, hence, reduced S 2 values (average 0.57 Ϯ 0.10), than those of the protein core. These data indicate that this sequence exhibits fast backbone motions on the nanosecond to picosecond time scale. This loop also has higher relative backbone r.m.s.d. values (0.44 Ϯ 0.19 Å) relative to the more ordered regions (0.22 Ϯ 0.09 Å) within the Ets-1⌬N301 solution structure ensemble ( Fig. 2 and Supplemental Fig. S1). The flexibility of this exposed polypeptide segment is consistent with its ability to serve as an intermolecular bridge within the crystalline Ets-1⌬N300 domain-swapped dimer. Second, although the 1 H{ 15 N}-NOE and S 2 values (average 0.90 Ϯ 0.03) of amides within HI-1 indicate that this helix is well-ordered on a sub-nanosecond time scale, the T 2 lifetimes measured for Lys 305 -Asp 310 are anomalously short. This relaxation behavior, which typically arises from a contribution of conformational exchange broadening (R ex ) to the overall rate of nitrogen transverse relaxation, suggests that HI-1 undergoes intermediate time scale (millisecond-microsecond) motions. This mobility, characterized further by relaxation dispersion measurements discussed below, demonstrates that even in the absence of DNA, HI-1 is conformationally dynamic relative to the remaining elements of secondary structure forming the inhibitory module and ETS domain.
Detailed inspection of the model-free analysis results for the Ets-1⌬N301 ETS domain and the C-terminal inhibitory sequence (Fig. 4B) also reveals reduced order parameters for amides within turns or loops (average S 2 ϭ 0.79 Ϯ 0.14) relative to those in helices or ␤-strands (0.92 Ϯ 0.05). This expected behavior, indicative of enhanced backbone motions on the nanosecond-picosecond time scale, is particularly noteworthy for residues near the Ets-1 DNA binding interface, specifically those preceding H1, in the turn of the HTH motif, and in the H3/S3 and S3/S4 loops. A similar pattern of elevated amide nitrogen thermal factors (B values) is observed for these residues in the Ets-1⌬N300 crystallographic coordinates. Furthermore, a comparable profile of local backbone mobility was detected in an 15  Residues within the inhibitory module, as well as in helix HI and the DNA binding interface of the ETS domain, display R ex contributions to transverse relaxation, indicating that these regions are conformationally flexible on a millisecondmicrosecond time scale. As expected for conformational exchange broadening, these R ex terms are larger when measured at higher magnetic fields. field strengths. These measurements sensitively detect the contribution of conformational exchange to the effective decay of transverse 15 N signal (48,60). As summarized in Fig. 3C, signals corresponding to 115 backbone amides in Ets-1⌬N301 were collected and processed. Consistent with the anomalously short 15 N T 2 lifetimes measured for amides within HI-1, significant relaxation dispersion R ex contributions were detected for this segment of the inhibitory module. The agreement of these complementary experimental approaches strongly indicates that in DNA-free Ets-1⌬N301, the backbone of helix HI-1 is conformationally mobile within the millisecond-microsecond time regime. Importantly, conformational exchange broadening was also detected by the relaxation-dispersion measurements for amides within the remainder of the inhibitory module (HI-2, and to a lesser extent, H4 and H5), as well as in helix H1 of the ETS domain with which they are interfaced. Furthermore, residues forming the DNA binding surface of Ets-1, including the turn of the HTH motif, the recognition helix H3, the loop following H3, and the S3/S4 loop also exhibited R ex terms. These measurements reveal that the backbone of the Ets-1⌬N301 DNA binding interface, as well as the inhibitory module, undergo motions on a millisecond-microsecond time scale. FIG. 4. Summary of the amide HX kinetic data for wild-type Ets-1⌬N301 at 28°C. A, lifetimes (1/k ex ) for slow proton-deuterium exchange (pH* 6.5), plotted according to residue number. Up arrows indicate the most protected amides for which insufficient exchange occurred over the course of the 10-day experiment to allow a reliable determination of their k ex values; for these residues, 1/k ex Ͼ 10 6 s. Open bars mark amides for which signals were observed only in the first few spectra recorded after transfer to D 2 O buffer and thus 1/k ex ϳ 750 s. The exchange kinetics of amides with lifetimes shorter than this limit could not be determined using this approach. B, rate constants, k ex , for rapid proton-proton exchange measured by the CLEANEX approach at pH 6.86, 7.50, 7.74, and/or 8.25. The increase of k ex with pH indicates that exchange occurs via base catalysis in the EX2 regime (62). The exchange kinetics of amides with k ex Ͻ 0.5 s Ϫ1 could not be determined by this method. C, HX protection factors (k pred /k ex ) derived using k ex values measured by proton-deuterium (pH* 6.50, black bar) or proton-proton (pH 7.50, gray bar, and/or 8.25, white bar) exchange and predicted k pred values for an unstructured polypeptide with the sequence of Ets-1⌬N301 under the corresponding experimental conditions. The small variation of protection factors with pH indicates that exchange is not exactly first order in [OH Ϫ ], likely because of pHdependent changes in the stability or electrostatic properties of Ets-1⌬N301. Residues without protection factors are either prolines (P) or have weak or overlapping amide 1 H-15 N HSQC signals, thus precluding reliable HX measurements. resemble those made by the N-terminal inhibitory sequence of Ets-1⌬N301. However, in contrast to a role in the negative regulation of Ets-1, within the context of the GABP␣/␤ heterodimer, these contacts enhance DNA binding activity (64).
Dynamics and Stability of the Inhibitory Module-The dynamic measurements reported here describe a labile inhibitory module. The evidence for limited stability and flexibility of the N-terminal inhibitory helices is provided through three experimental avenues: 15 N relaxation, 15 N relaxation-dispersion, and amide HX studies. The findings are consistent with previous chymotrypsin and trypsin proteolysis studies of Ets-1⌬N280 (2, 3) and Ets-1⌬N301 (57) 4 demonstrating cleavage at residues in or adjacent to HI-1 (Phe 304 , Tyr 307 , Arg 309 , Arg 311 ) and HI-2 (Leu 326 , Tyr 329 ), even in the absence of DNA. Such cleavage is indicative of structural fluctuations, allowing access of the polypeptide chain to the active site cleft of the protease. These observations support the hypothesis that Ets-1 autoinhibition involves a marginally stable inhibitory module with helices HI-1, and possibly HI-2, undergoing a facile helix-coil transition.
Amide 15 N relaxation measurements revealed local backbone motions in HI-1, as well as in regions of Ets-1⌬N301 beyond this marginally stable helix. The HI-1/HI-2 loop exhibits sub-nanosecond conformational mobility, while the remainder of the inhibitory module is relatively well ordered on this time scale. However, relaxation-dispersion measurements indicate that HI-1 is undergoing intermediate millisecond-microsecond time scale motions leading to amide chemical shift exchange broadening (R ex terms). Additional regions also show relaxation-dispersion properties indicative of intermediate time scale mobility, including HI-2, and to a lesser extent, H4 and H5 of the inhibitory module. Residues forming the DNA binding surface of ETS domain, specifically helix H1, H3, the HTH turn, and the S3/S4 loop, also display similar properties. We propose that the inhibitory module and helix H1, with which it directly interfaces, are in an equilibrium between different conformational states even in the absence of DNA. This conformational equilibrium may extend to the DNA binding interface of Ets-1 through van der Waals packing of residues in the ETS domain. Future studies will address whether these R ex terms reflect fluctuations within a conformational ensemble sampled by the intact inhibitory module and/or excursions of the N-terminal inhibitory helices HI-1 and HI-2 and possibly the DNA recognition helix H3 to unfolded states detectable by HX measurements. Nevertheless, these findings are consistent with a growing body of evidence that internal motions of proteins on the millisecond-microsecond time scale are important for their allosteric regulation (65,66).
Amide HX experiments complemented NMR relaxation measurements in providing a view of the local stability of the inhibitory module and ETS domain. Fast HX studies revealed that residues in helices HI-1 and HI-2 in Ets-1⌬N301 have amide proton exchange rates only ϳ15and ϳ75-fold slower, respectively, than expected for a random coil polypeptide. Even lower protection from exchange is observed for the exposed amides in the intervening HI-1/HI-2 loop. Thus, the N-terminal inhibitory helices readily undergo local conformational fluctuations, allowing base-catalyzed hydrogen exchange with the water. In contrast, slower proton-deuterium HX measurements yielded protection factors of ϳ10 3 to 10 4 for amides in the C-terminal inhibitory helices H4 and H5, and Ͼ10 6 for those in H1 of the ETS domain. Therefore, helices H1, H4, and H5 appear to be a stable scaffold upon which the more labile helices HI-1 and HI-2 pack to form the inhibitory module. Indeed, in fully activated Ets-1⌬N331, which lacks the entire N-terminal inhibitory sequence, helices H1, H4, and H5 are folded (56).
Several plausible explanations for the limited stability of HI-1 and HI-2 can be garnered from consideration of the structure of Ets-1⌬N301. The hydrophobic interface between amphipathic helix HI-1 and the remainder of the N-and C-terminal inhibitory sequences buries only a modest amount of nonpolar surface area (Ͻ800 Å 2 ). Furthermore, of the first 20 residues in Ets-1⌬N301, half have ionizable side chains, yielding a high charge density on HI-1 and the HI-1/HI-2 loop. In particular, this may lead to unfavorable (i,iϩ4) charge-charge interactions between Lys 305 -Arg 309 and Asp 306 -Asp 310 when these pairs of residues become aligned along helix HI-1 (67). The side chain carboxylate of Asp 310 at the HI-1 C terminus might also repel the negative end of the helix macrodipole (68). In contrast to HI-1, the stability of the non-polar alanine-rich helix HI-2 may be limited by its packing across the relatively polar surface of helix H1, including the side chain of Gln 339 .
Dynamics of the DNA Binding Interface-The dynamic studies of Ets-1⌬N301 also reveal a flexible DNA binding interface. As expected due to their surface exposure, the turn of the HTH motif and S3/S4 loop or wing exhibit high r.m.s.d. within the NMR-derived structure ensemble and elevated thermal B-values in the crystallographically derived structure of this protein, along with reduced S 2 order parameters and limited protection from amide HX. Consistent with this flexibility, crystallographic studies of Ets-1 show that the conformations of the HTH turn and S3/S4 loop change upon major groove binding by the recognition helix H3 in order to provide flanking contacts to the DNA phosphodiester backbone (12,19). Surprisingly, H3 itself also exhibits significant conformational mobility as evident by HX protection factors of only ϳ400 and by millisecondmicrosecond motions detected though relaxation-dispersion measurements. This flexibility, which is an emerging theme with many DNA-binding proteins (66,69), could facilitate specific DNA recognition since alternative conformations of contact residues have been shown to play a role in target site selection by ETS domain proteins (19,70).
Allosteric Mechanism of Ets-1 Autoinhibition-A combination of conformational and dynamic data extends the previously hypothesized allosteric model of Ets-1 autoinhibition. The NMR-derived structural ensemble of monomeric Ets-1⌬N301 confirms that the inhibitory module is distal from the HTH DNA binding interface of the ETS domain. This is consistent with previous thermodynamic and kinetic studies, which revealed that DNA binding by Ets-1 is neither sterically blocked nor modified in the number or nature of protein-DNA contacts made due to the presence of the inhibitory sequences (3,71). Together with proteolysis studies, CD spectroscopic measurements, as well as crystallographic and NMR spectroscopic structural studies of Ets-1/DNA complexes, these data lead to an allosteric model of Ets-1 autoinhibition in which unfolding of at least helix HI-1 in the N-terminal inhibitory sequence accompanies DNA binding, while refolding of this helix favors complex dissociation (Fig. 5). Dynamic measurements by NMR relaxation and amide hydrogen exchange, combined with previous proteolysis studies (2, 3), provide two additional dimensions to this model: (i) HI-1 has low stability and thus is poised to unfold; and (ii) regions of the inhibitory module, as well as of the ETS domain, appear to undergo exchange between conformational states on a millisecond-microsecond time scale, even in the absence of DNA. These data also provide insights into the energetic coupling of DNA binding with the HI-1 helix-coil transition. Based on HX protection factors of ϳ15, ⌬G o HX ϳ1.6 kcal/mol for the transient unfolding of HI-1 in DNA-free Ets-1⌬N301. This small energetic penalty is consist-ent with ⌬⌬G o bind ϳ 0.4 kcal/mol corresponding to the ϳ2-fold inhibition of DNA binding by Ets-1⌬N301 relative to the fully activated truncation fragment, Ets-1⌬N331.
Based on its structural role in both the ETS domain and the inhibitory module, we have proposed that helix H1 is central to the allosteric mechanism of autoinhibition (11). This proposal is supported by several lines of evidence. X-ray crystallographic analyses demonstrated that H1 must be positioned precisely to form a hydrogen bond between the amide of Leu 337 at its N terminus and the DNA phosphodiester backbone. This hydrogen bond is likely augmented by orienting the positive end of the macrodipole of this helix adjacent to the negatively charged nucleic acid backbone (12,19). The importance of the interactions involving HI toward DNA binding and autoinhibition was confirmed by mutational analyses of both Ets-1⌬N280 and its target DNA (11).
Flexibility in both the ETS domain and autoinhibitory module appears to accommodate the structural changes that accompany DNA binding. The crystal structures of free Ets-1⌬N300 and DNA bound Ets-1⌬N280 are very similar, with the dramatic exception of HI-1 being domain-swapped or unfolded, respectively (12). However, important conformational changes between these species are detectable using a C ␣ difference distance matrix approach that is independent of the choice of co-ordinate superimposition (72). These changes can be divided into two categories, those that accommodate DNA binding, and those that affect the conformation of the inhibitory module. Exemplifying the former, the turn of the HTH motif and the loop between ␤-strands S3 and S4 shift to facilitate nucleic acid contacts. In the latter category, the relative positions of helix HI-2 and the HI-2/H1 loop change by over 1 Å with respect to the ETS domain. This perturbation is accompanied by pronounced alterations in the backbone conformation of the HI-2/H1 loop, as well as the disruption of a potential hydrogen bond between the side chains of Ser 332 in this loop and Gln 339 in helix H1. These observations suggest that, in order for helix H1 to precisely interact with DNA, the preceding HI-2/H1 loop must be shifted relative to its position in the free protein. This in turn would require the displacement of the non-polar alanine-rich helix HI-2 away from the DNA and across the relatively polar surface of H1 that is defined in part by Gln 339 . Such a displacement would likely be coupled to a disruption of the hydrophobic packing of the N-terminal inhibitory region and thus linked with the unfolding of the marginally stable HI-1. Each of these categories of conformational change is consistent with regions of flexibility within Ets-1⌬N301 detected by 15 N relaxation and HX measurements NMR measurements complement the insights gained from static Ets-1 structures and provide new dynamic views of the mechanism of autoinhibition. In addition to the limited stability of HI-1 and HI-2 detected by fast HX measurements, motions on the millisecond-microsecond time scale, leading to R ex terms, were also detected for amides within both the inhibitory module and the ETS domain. A simple interpretation of these data is that DNA-free Ets-1⌬N301 is in conformational equilibrium that extends beyond fluctuations of the N-terminal inhibitory region. We speculate that this pre-existing equilibrium may be an additional feature of the allosteric coupling of DNA binding by the ETS domain with the helix-coil transition of the inhibitory module. For example, in addition to the postulated link between the unfolding of HI-1 and the precise positioning of H1 and the HI-2/H1 loop, it is possible that conformational mobility of the winged HTH motif of the ETS domain is required for high affinity DNA binding and that this mobility is coupled to the dynamic behavior of the inhibitory module. Consistent with studies in other systems, the millisecond-microsecond time scale of these motions appears to be a hallmark of protein regulatability (65,66,73). In the case of Ets-1, this regulation is versatile. Cellular conditions that trap an unfolded helix would result in activation of DNA binding, whereas those that limit the unfolding of this helix would result in increased inhibition. In this way, the N-terminal inhibitory elements, distinct from the more stable C-terminal helices, serve as the control point or effector of signals that regulate DNA binding affinity.
In summary, we have used NMR spectroscopy to characterize the structure and dynamics of a partially inhibited fragment of Ets-1. These studies provided the first description of the intramolecular packing of the inhibitory module, identified a potential conformational exchange equilibrium linked to the allosteric mechanism of autoinhibition, and defined the marginally stable nature of the N-terminal inhibitory helices. Together, these results establish a molecular framework necessary for understanding how protein partnerships, post- FIG. 5. A model of the allosteric mechanism of autoinhibition in which the facile unfolding of helix HI-1 is coupled to DNA binding by the adjacent ETS domain. Shown are the NMR-derived structure of free Ets-1⌬N301 and the crystallographically determined structure of Ets-1⌬N280 in complex with DNA (19). The helices of the ETS domain are indicated in dark gray, while those of the inhibitory module are displayed in light gray. Amide 15 N relaxation and HX data demonstrate that HI-1 is only marginally stable and thus poised to unfold. Formation of the intermolecular hydrogen bond between the amide of Leu 337 and the DNA phosphodiester backbone (black dot) appears to require a shift in the position of the HI-2/HI loop, which could be linked with the unfolding of this inhibitory helix. translational modifications, or mutations alter the stability of the inhibitory module and thereby regulate the affinity of Ets-1 for its target promoter sequences.