Steric mechanism of auto-inhibitory regulation of specific and non-specific DNA binding by the ETS transcriptional repressor ETV6

Abstract

DNA binding by the ETS transcriptional repressor ETV6 (or TEL) is auto-inhibited ~ 50-fold due to an α-helix that sterically blocks its ETS domain binding interface. Using NMR spectroscopy, we demonstrate that this marginally-stable helix is unfolded, and not displaced to a non-inhibitory position, when ETV6 is bound to DNA containing a consensus ^5’GGAA^3’ recognition site. Although significantly lower in affinity, binding to non-specific DNA is auto-inhibited ~ 5-fold and also accompanied by helix unfolding. Based on NMR chemical shift perturbations, both specific and non-specific DNA are bound via the same canonical ETS domain interface. However, spectral perturbations are smaller for the non-specific complex, suggesting weaker and less well-defined interactions than in the specific complex. In parallel, the crystal structure of ETV6 bound to a specific DNA duplex was determined. The structure of this complex reveals that a non-conserved histidine residue in the ETS domain recognition helix helps establish the specificity of ETV6 for DNA-binding sites containing ^5’GGAA^3’ versus ^5’GGAT^3’. These studies provide a unified steric mechanism for attenuating ETV6 binding to both specific and non-specific DNA and expand the repertoire of characterized auto-inhibitory strategies utilized to regulate ETS factors.

Introduction

Auto-inhibition is a powerful ‘on-site’ regulatory mechanism used to modulate a wide variety of biomolecular interactions.¹ The ETS family of transcription factors displays several examples of DNA-binding auto-inhibition.² These proteins are defined by the conserved ETS domain, which recognizes very similar DNA sequences containing a core ^5’GGA(A/T)^3’ motif.³ One role of controlling DNA binding by auto-inhibition is to provide added specificity for targeting distinct ETS proteins to appropriate transcriptional regulatory sequences. The relief of ETS-1 auto-inhibition though cooperative DNA binding with RUNX1, a frequently occurring partnership in T cell enhancers, exemplifies such added specificity.^{2, 4}

We seek to understand the common and distinct mechanisms of auto-inhibition within the ETS family in order to gain a deeper insight into the evolution of DNA-binding control to achieve biological specificity. In the case of the prototypic member ETS-1, there is an extensive structural and dynamic understanding of its auto-inhibition. A combination of steric and allosteric mechanisms have been uncovered that integrate various signaling events, such as post-translational modifications and partner-protein interactions, resulting in both negative and positive control of ETS-1, respectively.^4-6 Briefly, ETS-1 auto-inhibition is mediated by a set of helices that flank its ETS domain. This inhibitory module is distal from the DNA-binding interface, and upon DNA binding, undergoes an allosteric conformational change highlighted by the unfolding of a marginally stable helix.^{7, 8} Additionally, an adjacent intrinsically disordered serine rich region (SRR) plays a critical role in the Ca⁺²-dependent phosphorylation-enhanced auto-inhibition of ETS-1 by stabilizing the inhibitory module and by transiently masking its DNA-binding interface.^{6, 9} A second ETS member, ERG, has also been recently shown to be modestly auto-inhibited by a flexible sequence N-terminal to its ETS domain.¹⁰ In contrast, the mechanisms by which most other ETS proteins are auto-inhibited are not well understood.

In this study, we extended our investigation of regulation by auto-inhibition to ETV6 (or TEL, translocation ETS leukemia). Unlike the transcriptional activators ETS-1 and ERG, ETV6 is a repressor.¹¹ Furthermore, it also self-associates due to the presence of a PNT (or SAM) domain, thereby facilitating cooperative binding on tandem ETS DNA-binding sites for repressive activity.^12-15 Previously, we mapped an ETV6 inhibitory region C-terminal to its ETS domain.¹⁵ This CID (C-terminal inhibitory domain) contains two helices, of which helix H5 sterically blocks the canonical ETS DNA-binding interface and thereby reduces its affinity for specific sequences by ~ 50-fold.^{15, 16} Similar to ETS-1, the inhibitory helices are marginally stable and their presence dampens dynamics of the ETS domain.¹⁶ Preliminary evidence suggesting a conformational change accompanying DNA contact was provided by the relief of ETV6 auto-inhibition due to mutations that potentially disrupt the CID.¹⁵ Thus, the central goal of this current study was to determine the conformational changes occurring in the ETV6 ETS domain and the CID upon DNA binding. We also investigated how the CID impacts both specific and non-specific DNA binding, because binding a limited number of specific target sites in the cell occurs against a very high background of non-specific interactions.

We used a set of complementary NMR experiments to demonstrate that residues forming the inhibitory helix H5 are unfolded when ETV6 is bound to DNA. In parallel, with isothermal titration calorimetry (ITC) and NMR spectroscopy, we investigated the impact of CID on the interaction of ETV6 with non-specific DNA lacking the ETS consensus motif. Non-specific binding, which is substantially lower in affinity, is also auto-inhibited and is accompanied by helical unfolding. We further show that ETV6 utilizes a similar binding interface for both specific and non-specific DNA sequences. However, the non-specific complex appears relatively “loose” in comparison to the tight specific complex. To better define this interface, we determined the free and DNA-bound structures of the uninhibited ETV6 ETS domain with NMR spectroscopy and X-ray crystallography, respectively. The DNA-bound structure also helps explain the role of a non-conserved histidine in the preferential binding of ETV6 to the core sequence ^5’GGAA^3’. Collectively, our studies uncover a unified steric mechanism of auto-inhibition that impacts both non-specific and specific DNA binding by an ETS protein.

Results

Helix H5 in CID unfolds when ETV6 binds to specific DNA

Structures of several ETS proteins show a highly conserved mode of DNA binding whereby the recognition helix H3 inserts into the major groove and makes extensive contacts with the ^5’GGA(A/T)^3’ motif.² A similar canonical mode of binding by ETV6 would require that the CID must be either unfolded or otherwise be displaced to remove steric blockage of the ETS domain. We tested this hypothesis by several complementary NMR experiments. Initially, we utilized ¹⁵N-HSQC spectroscopy to monitor the interaction of ¹⁵N-labeled ETV6^D446 with a 15 bp oligonucleotide (DNA^sp) containing the ETS consensus motif ^5’GGAA^3’ (see Table 1 for nomenclature). Upon titration, peaks corresponding to the free protein diminished and a new set of signals corresponding to the bound state emerged (Fig. 1A and Supplemental Fig. S1). Such binding in the slow exchange regime is consistent with the relatively high nanomolar affinity of inhibited ETV6 for the consensus DNA sequence.¹⁶ The chemical shifts of the ETV6^D446–DNA^sp complex were used to predict its backbone structure (Fig. 2A). This indicated that the ETS domain and the CID helix H4 retained the same secondary structure as in the unbound protein.¹⁶ In contrast, residues corresponding to helix H5 displayed large ¹⁵N-HSQC chemical shift perturbations (CSP's) and resulting chemical shifts diagnostic of a disordered polypeptide chain (Fig. 1B and 2A). In addition, these residues exhibited low order parameters (S²) based on the “random coil index”.¹⁷

Table 1.

ETV6 constructs and DNA sequences

Name	Sequence	Comment
ETV6R426	G329-R426	Uninhibited ETV6a
ETV6D446	G329-D446	Inhibited ETV6a
ETV6D446	R335-D446	Inhibited ETV6a
ETV6R458	R335-R458	Inhibited ETV6a
DNAsp	5′-CAAGCCGGAAGTGAG-3′	Specific DNAb
	3′-GTTCGGCCTTCACTC-5′
DNAsp-cryst	5′-AAAGCCGGAAGTGAG-3′	Specific DNAb
	3′-TTCGGCCTTCACTCT-5′
DNAnonsp	5′-GATGCAGTGTAGTCG-3′	Non-specific DNA
	3′-CTACGTCACATCAGC-5′
DNAsp-emsa-1	5′-CAAGCCGGAAGTGAG-3′	Specific DNAb
	3′-GTTCGGCCTTCACTC-5′
DNAsp-emsa-2	5′-CAAGCCGGATGTGAG-3′	Specific DNAb
	3′-GTTCGGCCTACACTC-5′

^aThe core ETS domain spans Leu337 to Phe415, and Pro419 to Ser424 form helix H4.

^bThe core ^5′GGAA^3′ or ^5′GGAT^3′ binding motif is in bold.

Fig. 1. NMR monitored titrations of ETV6 fragments with DNA.

(A) Overlaid regions of the ¹⁵N-HSQC spectra of ETV6^D446 recorded in the absence (red) or presence of a 1.1 molar ratio of DNA^sp (green). High affinity binding occurs the slow exchange regime and only signals from the free or bound protein are detected at intermediate molar ratios (not shown). Also shown are the same spectral regions when ETV6^D446 is titrated with DNA^nonsp in molar ratios of 0 (red), 0.1, 0.25, 0.5, 0.75, 1, 2, and 3 (magenta). With weaker affinity, binding occurs in the fast exchange regime and signals progressively change from the chemical shift of the free to the DNA-bound state. Some broadening also occurs at intermediate saturation. See Supplemental Figure S1 for the full spectra. (B) Amide CSP's (Δδ = {(Δδ_H² + (0.154Δδ_N)²}^1/2) for ETV6^D446–DNA^sp (green) and ETV6^D446–DNA^nonsp (magenta) with respect to free ETV6^D446. The unfolding of helix H5 leads to large CSP's of similar magnitude for residues 430 - 440 in both complexes. In contrast, the CSP's for the ETS domain result from the displacement of helix H5 and the binding of DNA. In (C), the effects of helix H5 unfolding are removed by calculating CSP's relative to unbound ETV6^R426, which lacks this helix. Note that these CSP's were calculated using the chemical shifts of ETV6^D446 in the presence of a 1.1-fold molar excess of DNA^sp and 3-fold molar excess of DNA^nonsp, and thus correspond to 99% and 98% saturation, respectively, based on their respective K_D values and binding site sizes. The secondary structure (helix, cylinder; strand, arrow) is displayed as cartoon on top with the core ETS domain in red and helix H4 of the CID in cyan. Helix H5, which is unfolded in both complexes, is not colored.

Fig. 2. Helix H5 is unstructured in both specific and non-specific ETV6–DNA complexes.

The normalized secondary structure propensities (helices, blue; strands, orange) and RCI order parameters (S², black line) for the (A) ETV6^D446–DNA^sp and (C) ETV6^D446–DNA^nonsp complexes were calculated from ¹³C^α, ¹³C^β, ¹⁵N and ¹H^N chemical shifts using the program MICS ⁸¹. The ¹⁵N-NOE data for (B) ETV6^D446–DNA^sp and (D) ETV6^D446–DNA^nonsp are shown for the core ETS domain (red) and CID (cyan) residues. Decreasing NOE and S² values indicate increasing amide mobility on the sub-nsec timescale and thus show that helix H5 is unfolded in both complexes (see also Supplemental Fig. S2). The histogram bars for the two C-terminal residues are truncated, and missing data corresponds to prolines and residues with overlapping or unassigned signals. The top cartoon shows the secondary structural elements (helix, cylinder; strand, arrow) of the free inhibited ETS domain (red) and CID (cyan or not colored).

To more directly characterize the dynamic properties of the ETV6^D446–DNA^sp complex, we also collected amide ¹⁵N T₁, T₂ and heteronuclear NOE relaxation data (Supplemental Fig. S2). From T₁/T₂ ratios, the global isotropic tumbling correlation time was determined to be 12.5 ± 0.2 ns. This is consistent with the 23 kDa molecular mass of the ETV6^D446–DNA^sp complex, ¹⁸ and confirms that it is monomeric under the experimental conditions. More importantly, the ¹⁵N-NOE values, which are very sensitive to the sub-nsec timescale motions of the amide ¹⁵N-¹H^N bond vector, provide a measure of the fast local backbone dynamics of a protein. These data revealed a well-folded core ETS domain with high ¹⁵N-NOE values, yet a highly flexible CID with substantially lower values indicative of pronounced conformational mobility (Fig. 2B). An analysis of the full set of amide ¹⁵N relaxation data according to the model-free formalism^{19, 20} similarly showed that the CID residues are dynamic with substantially lower conventional order parameters than those of the ETS domain (Supplemental Fig. S2).

Finally, we determined the tertiary structure of ETV6^D446 bound to DNA^sp using NMR spectroscopy. Extensive ¹H, ¹³C and ¹⁵N chemical shift assignments, along with backbone dihedral angle and inter-proton NOE distance restraints for the protein component of the ETV6^D446–DNA^sp complex, were obtained by TROSY-based heteronuclear NMR experiments (Table 2). Although detected, the ¹H signals from the unlabeled DNA were not assigned. Using these data, the structural ensemble of the bound protein was calculated (Fig. 3A). Overall, the protein adopts the winged helix-turn-helix (wHTH) fold characteristic of ETS domains, and closely resembles the structure of free auto-inhibited ETV6 ETS domain.¹⁶ However, in marked contrast to a well-folded helix H5 blocking the binding interface of the unbound protein, the corresponding CID residues are disordered in the complex with high rmsd values. This is consistent with their dynamic nature and hence lack of any structural restraints. Collectively, chemical shift, ¹⁵N relaxation, and structural data clearly demonstrate that helix H5 unfolds upon specific DNA binding, and is not simply displaced as an intact helix to an alternative position that is no longer auto-inhibitory.

Table 2.

NMR refinement statistics for protein structures

	ETV6R426	ETV6D446 - DNAsp
NMR distance and dihedral restraints
Distance restraints
Total NOE	1620	1540
Intra-residue	467	464
Inter-residue	1153	1076
Sequential (|i – j| = 1)	382	459
Medium-range (|i – j| <= 4)	285	271
Long-range (|i – j| >= 5)	486	346
Dihedral angle restraints
ϕ, ψ	85, 88	82, 84
Structure statistics
Violations (mean ± std. dev,)
Distance restraints (Å)	0.06 ± 0.02	0.16 ± 0.13
Dihedral angle restraints (°)	4.72 ± 1.92	5.50 ± 2.08
Max. dihedral angle violation (°)	6.09	10.25
Max. distance restraint violation (Å)	0.42	0.32
Deviations from idealized geometry
Bond lengths (Å)	0.005	0.005
Bond angles (°)	0.742	0.774
Impropers (°)	2.13	2.51
Average pairwise r.m.s. deviationa (Å)
All heavy atoms	1.03 ± 0.09	1.21 ± 0.12
Backbone only	0.59 ± 0.08	0.74 ± 0.11

^aPairwise r.m.s. deviation was calculated among 10 refined structures for the residues 337 to 415.

Fig. 3. Structural ensembles of free and bound ETV6 are similar except for the unfolding of helix H5.

The NMR-derived structural ensembles of (A) the ETV6^D446–DNA^sp complex (ETS domain helices and strands, red; CID helix H4, cyan) and (B) uninhibited ETV6^R426 align closely to the lowest energy structure of inhibited ETV6^R458 (green).¹⁶ CID helix H5, which blocks the DNA-binding interface of ETV6^R458, is absent in ETV6^R426 and unfolded in the ETV6^D446-DNA^sp complex. Although present in the latter complex, DNA^sp was not included in the structure calculations. The N-terminal Gly-Ser-His-Met and unstructured residues (329-335 and 444-446) are not shown for clarity. Arrows point to the DNA-binding interface along helix H3.

ETV6 binding to non-specific DNA is also auto-inhibited

Using isothermal titration calorimetry (ITC) and NMR spectroscopy, we also investigated the interaction of ETV6^D446 and ETV6^R426 (lacking helix H5) with a 15 bp non-specific DNA oligonucleotide (DNA^nonsp, Table 1). In contrast to the case of specific binding, for which there is one single high-affinity site within DNA^sp, ETV6 can potentially bind DNA^nonsp in numerous positions and in either orientation. Therefore, the measured ITC binding isotherms were fit to a variation of the McGhee-von Hippel model²¹ developed by Record and co-workers²² for non-specific binding with neighbor exclusion to a finite lattice. As summarized in Figure 4, this yielded the average microscopic dissociation constant (K_D), binding enthalpy (ΔH) and the effective binding site size (n). Both ETV6^D446 and ETV6^R426 bound DNA^nonsp with K_D values in the micromolar range, which is 10³ - 10⁴ fold weaker than their affinities to a specific oligonucleotide.^{15, 16} Most importantly, the K_D value for ETV6^D446 is ~ 5-fold higher than that of ETV6^R426. Thus, binding to non-specific DNA by the ETV6 ETS domain is also auto-inhibited, albeit to a lesser extent than to specific DNA (~ 50-fold).

Fig. 4. Binding to non-specific DNA is also auto-inhibited.

(A) Summary of the fit ITC data for DNA^nonsp, along with the binding isotherms for (B) inhibited ETV6^D446 and (C) uninhibited ETV6^R426. Raw data are open squares and the best fits to the Record model²² are solid lines.

It is interesting to compare the differences in the binding site size between the specific and non-specific complexes. As described below, the crystal structure of ETV6 bound to specific DNA revealed that the protein contacts ~ 10 nucleotide pairs. In contrast, fitting of the ITC isotherms indicated that ETV6^R426 and ETV6^D446 effectively bind only 5.2 and 6.5 nucleotide pairs, respectively, in DNA^nonsp. These n values reflect the average site size (~ 6 bp) occluded by one ETV6 along the non-specific DNA duplex. Therefore, the first ETV6 molecule could potentially bind in either orientation to the 15 bp oligonucleotide at 10 (= 15-6+1) overlapping sites. However, due to neighbor exclusion, only 2 to 3 protein molecules bind one oligonucleotide under saturating conditions. This could be accommodated by adopting staggered rotated positions along the double helix. Parenthetically, these non-integral values likely arise from the violation of the key premises in the Record model that the DNA lattice consists of identical binding sites and that the protein has only one mode of binding.²³ Also, since DNA^nonsp contains many potential binding sites, fitting these ITC data to a simple isotherm yielded apparent, overall macroscopic K_D values ~ 10-fold lower than the microscopic values presented in Figure 4 (not shown). Importantly, regardless of model, auto-inhibited ETV6^D446 binds to DNA^nonsp with ~ 5-fold weaker affinity than uninhibited ETV6^R426.

Helix H5 also unfolds when ETV6 interacts with non-specific DNA

Complementing ITC, we used NMR spectroscopy to obtain structural details of the ETV6–DNA^nonsp interaction. Upon titration with DNA^nonsp, numerous amide ¹⁵N-¹H^N signals of ETV6^D446 showed progressive chemical shift changes (Fig. 1A and Supplemental Fig. S1). This is diagnostic of fast exchange between the free and bound forms of the protein and is consistent with a micromolar dissociation constant. Furthermore, only one signal per amide was observed in the ¹⁵N-HSQC spectrum of the saturated ETV6^D446–DNA^nonsp complex (Supplemental Fig. S1 and S3). Thus, translocation of the bound protein between the many possible non-specific sites along this oligonucleotide is also fast on the chemical shift timescale.

Insights into the structure of the bound protein could be obtained from its assigned NMR spectra. Based on its mainchain chemical shifts, the ETV6^D446 ETS domain retained its wHTH secondary structure when bound to DNA^nonsp (Fig. 2C). In contrast, large CSP's occurred for residues corresponding to helix H5, suggesting a significant conformational change (Fig. 1B). Indeed, both chemical shift and ¹⁵N relaxation measurements demonstrated that, as with DNA^sp, the CID adopted a dynamic random coil conformation in the ETV6–DNA^nonsp complex (Fig. 2C,D).

ETV6 binds non-specific DNA via the canonical ETS domain interface

The CSP's also revealed that the ETV6 ETS domain binds non-specific and specific DNA sequences via the same general interface. For this analysis, it is important to recognize that the spectral perturbations reflect both the unfolding of helix H5, which sterically blocks this interface, and interactions with DNA (Fig. 1B and Supplemental Fig. S4). Accordingly, to focus only on the latter, we calculated the CSP's for the ETV6^D446–DNA^sp and ETV6^D446–DNA^nonsp complexes relative to free ETV6^R426, which lacks the CID helix H5. These CSP's are presented as the magnitude of the combined amide ¹H^N and ¹⁵N changes in Figure 1C, and separately (with upfield or downfield “direction”) for the two nuclei in Supplemental Figure S5. For both complexes, residues in helix H3, the turn between helices H2 and H3, and the wing between S3 and S4 experienced the largest CSP's. When mapped onto the structure of ETV6 (Fig. 5), these residues cluster to the DNA-binding interface that has been well characterized in many ETS domain complexes.²⁴ Amide chemical shifts are exquisitely sensitive to their environment. Thus, CSP's may arise due to proximity to the charged and aromatic moieties in DNA, as well as from local or propagated conformational changes, such as those influencing hydrogen-bonding networks. Regardless of exact cause, the similar patterns of CSP's demonstrate that ETV6^D446 uses the same canonical interface to bind both specific and non-specific DNA.

Fig. 5. ETV6 ETS domain binds specific and non-specific DNA via the same canonical interface.

The amide chemical shift perturbations (CSP's, Δδ = {(Δδ_H² + (0.154Δδ_N)²}^1/2) for (A) ETV6^D446–DNA^nonsp and (B) ETV6^D446–DNA^sp with respect to unbound ETV6^R426 are mapped onto the crystal structure of ETV6^R426-DNA^sp-cryst (DNA not shown). Residues (backbone cartoon) are color-coded in the indicated CSP ranges. Prolines and unassigned residues are in grey. Side chains are shown for residues in (A) with CSP > 0.2 ppm and in (B) with CSP > 0.4 ppm. See Figure 1 and Supplemental Figures S1 and S4 for the original data.

Paralleling their relative binding affinities, the CSP's for many residues were in the same approximate direction,²⁵ yet smaller in magnitude, for the DNA^nonsp complex than for the DNA^sp complex (Figs. 1 and 5, and Supplemental Fig. S5). Comparable patterns of relative NMR spectral changes have been reported for the HMG-box²⁶ and ZNF217 zinc finger²⁷ proteins bound to non-specific versus specific DNA oligonucleotides. This is not a trivial result of incomplete saturation of the ETV6^D446–DNA^nonsp complex, as the protein was ~ 98 % bound (Supplemental Fig. S3). Rather, this indicates that ETV6^D446 forms generally similar, albeit less well-defined, time-averaged interactions with DNA^nonsp than with DNA^sp. Such interactions likely involve electrostatic contacts between the positively-charged DNA-binding interface of ETV6^D446 and the negatively-charged phosphodiester backbone of DNA^nonsp, rather than base-specific hydrogen bonds. Also, rapid exchange between binding sites along DNA^nonsp should lead to smaller net CSP's due to averaging of potential positive and negative chemical shift changes.

Residue-wise comparison of the CSP's revealed potentially important structural differences between the two complexes (Figs. 1 and 5, and Supplemental Fig. S5). For example, Arg392 and His396, whose side chains interact with DNA bases in the ETV6-DNA crystal structure (see below), experienced substantial CSP's in the specific complex, yet smaller perturbations with DNA^nonsp. Five additional lysine and arginine side chains, which interact with the phosphodiester backbone of DNA, also showed large amide CSP's in the specific complex, whereas only two (Lys380 and Arg382) exhibited substantial perturbations in the non-specific complex. In contrast, the ¹⁵N chemical shift of Ala393, a residue in helix H3, changed with opposite sign upon binding DNA^nonsp relative to DNA^sp. We hypothesize that these spectral differences reflect looser, dynamic electrostatic interactions and the lack of direct base contacts in the non-specific DNA complex versus the specific complex. Consistent with this notion, the ¹⁵N-NOE values for residues in the turn of the helix(H2)-turn-helix(H3) are slightly lower for ETV6^D446 bound to DNA^nonsp than to DNA^sp, indicating greater fast timescale mobility (Figs. 2C,D).

Insights into the binding interface from structures of free and DNA-bound ETV6^R426

To relate the above findings with the atomic details of the ETV6 binding interface, we used X-ray crystallography to determine the structure of its ETS domain bound to DNA. Since disordered regions are not conducive to crystallization and our NMR measurements clearly demonstrated that residues following Arg426 are unstructured in the ETV6–DNA^sp complex, we used ETV6^R426 for these studies. The crystal structure of this construct bound to a 14bp specific DNA^sp-cryst sequence (Table 1) was solved at 2.2 Å resolution (Table 3, Fig. 6A). In parallel, we also used NMR spectroscopy to determine the structural ensemble of free ETV6^R426 (Table 2, Fig. 3B). Other than the absence of the folded inhibitory helix H5, the structures of ETV6^R426 in its free and DNA^sp-cryst-bound states closely resemble that determined previously¹⁶ for free inhibited ETV6^R458 with an average rmsd of 1.1 Å and 0.9 Å between all corresponding main chain atoms for ordered residues, respectively. Therefore, neither the inhibitory helix H5 nor the DNA measurably alter the average structure of the ETV6 ETS domain. A similar lack of any significant backbone structural changes has been observed for other ETS proteins in their free versus bound states.²

Table 3.

Data collection and refinement statistics for ETV6^R426–DNA^cryst-sp

Data collection
Space group	P3121
Cell dimensions
a, b, c (Å)	57.58, 57.58, 130.48
α, β, γ (°)	90, 90, 120
Resolution (Å)	50.00-2.20 (2.24-2.20)
Rsym or Rmerge	0.085 (0.711)
I / σI	29.13 (4.34)
Completeness (%)	99.6 (100.0)
Redundancy	7.9 (8.1)
Refinement
Resolution (Å)	26.34-2.20
No. reflections	13290
Rwork / Rfree	0.175/0.221
No. atoms
Protein	813
DNA	590
Water	124
B-factors (Å2)
Overall (Wilson)	47.5 (37.9)
Protein	46.1
DNA	48.9
Water	49.6
R.m.s. deviations
Bond lengths (Å)	0.006
Bond angles (°)	1.32

Fig. 6. Binding interface of ETV6^R426–DNA^sp-cryst complex.

(A) Crystal structure of the ETV6^R426–DNA^sp-cryst complex. ETS domain residues contacting the DNA are shown in stick representation. Secondary structural elements in the core ETS domain are colored red, loops in grey, and helix H4 in cyan. Corresponding backbone atoms for ordered residues in the crystal ETV6^R426–DNA^sp-cryst and solution ETV6^R426–DNA^sp complexes superimposed with an average rmsd of 1.0 Å (not shown). (B) Interfacial interactions with ETV6 residues grouped according to their secondary structure. Water molecules are denoted as blue circles. (C) Direct readout due to the invariant Arg392 and Arg395 interacting with G₊₂ and G₊₁, respectively. (D) The methyl group on T_+3’ fits a small hydrophobic pocket created by the Arg392 and Lys389 side chains. (E) Glu388 interacts with the C_-1 and C_-2 bases. (F) Hydrogen bond between His396 and T_+4’ provides specificity towards the A₊₄ of the ^5’GGAA^3’ motif (Supplemental Fig. S7). (In C-F: carbon, green; oxygen, red; nitrogen, blue.)

The contacts observed in the ETV6^R426–DNA^sp-cryst interface, as summarized in Figures 6B-F, are highly conserved in the ETS family.²⁴ Direct major groove base readout is mediated by bidentate hydrogen bonding of the invariant Arg392 and Arg395 in the recognition helix H3 to the G₊₂ and G₊₁ of the core ^5’GGA^3’ motif, respectively. In most ETS factors, a tyrosine side chain in helix H3 hydrogen bonds to A₊₃. However, the corresponding residue in ETV6 is His396, which is unable to provide this interaction. Rather, the specificity towards A₊₃ is determined via the complementary base T_+3’. The methyl group of T_+3’ occupies a hydrophobic pocket formed by the side chains of Arg392 and Lys389 (Fig. 6D). This hydrophobic interaction, which is also observed in other ETS domain–DNA structures, allows the Arg392 side chain to adopt the correct rotameric state required to hydrogen bond with the base G₊₂ and thus influences the overall specificity towards the core ^5’GGA^3’ motif. The side chain of Glu388 also forms direct and water mediated hydrogen bonds with the amino groups of C_-2 and C_-1, respectively, thereby establishing the preference of ETV6 for cytosines at these positions.³ Several residues also interact with the phosphodiester backbone of the DNA. The amide NH of Leu337 in helix H1 and Trp376 in helix H2 form highly conserved hydrogen bonds with the phosphate backbone. Residues Lys380 and Arg382 in the turn, Lys389 in H3, and Lys405 and Arg410 in the wing interact electrostatically with phosphate groups. These interactions may provide some specificity in the form of indirect shape readout, ²⁸ or may be relatively non-specific and simply increase the net affinity of ETV6 for DNA.

Histidine 396 determines specificity towards Adenine at the +4 position

Most ETS factors can recognize either adenine or thymine at the +4 position in the core ^5’GGA(A/T)^3’ sequence.^{3, 29, 30} In contrast, based on qualitative high-throughput binding assays, ETV6 appears to have a distinct preference for adenine at this position.³ This specificity has been hypothesized to arise from the presence of a non-conserved histidine in helix H3 of ETV6, rather than the more commonly found tyrosine. To confirm these observations, we used an electrophoretic mobility shift assay (EMSA) to measure the K_D values of ETV6^D446 for DNA binding sites with either a ^5’GGAA^3’ or ^5’GGAT^3’ core sequence (Supplemental Fig. S6). Strikingly, ETV6^D446 displayed 500-fold higher affinity towards the ^5’GGAA^3’ –bearing site (K_D values of 1.6 ± 0.3 nM versus 800 ± 100 nM). Mutating His396 to a tyrosine reduced this selectivity to only 8-fold (1.9 ± 0.3 nM versus 16 ± 7 nM). Given that both WT- and H396YETV6^D446 bound the ^5’GGAA^3’ oligonucleotide with very similar affinities, the presence of His396 strongly disfavors interactions with a thymine base at the +4 position.

The crystal structure of the ETV6^R426–DNA^sp-cryst complex provides an explanation for this sequence specificity. As illustrated in Figure 6F, His396 N^ε2 is within hydrogen bonding distance (2.8 Å) of the O₄ of T_{4 +4’} on the DNA strand complementary to the ^5’GGAA^3’. To act as a hydrogen bond donor, the side chain of His396 must be in the positively-charged imidazolium or neutral N^ε2H imidazole tautomeric state. When we modeled the A-T nucleotide pair to T-A at position +4, neither base at this position (T₊₄ or A_+4’) can form alternative hydrogen bonds with His396, unless perhaps it adopts the less favored neutral N^δ1H tautomer (Supplemental Fig. S7). In contrast, as shown by crystallographic studies of the ETS protein ELK4, the more common tyrosine can hydrogen bond with an adenine in either the +4 or +4’ position.³¹ Selectivity for the +4 base has also been observed for the ETS factors PU.1 and PDEF, which have Asn and Gln at the His396 equivalent position, respectively.³ This arises as the Asn and Gln side chains of PU.1 and PDEF form water mediated hydrogen bond with the T_+4’ and A_+4’ bases, respectively.^32-34 In contrast, His396 in ETV6 forms a hydrogen bond directly with the T_+4’ base.

Discussion

Steric mechanism of ETV6 auto-inhibition

In this study, we have extended our structural and thermodynamic understanding of the molecular basis of ETV6 auto-inhibition. Initially, we discovered that the CID attenuates specific DNA binding by ~ 50-fold.¹⁵ Subsequently, we reported that the CID includes two helices, of which helix H5 packs along the canonical ETS domain DNA-binding interface.¹⁶ This immediately predicted that substantial conformational changes must occur in ETV6 to enable its association with DNA. Alternatively, ETV6 might utilize a non-canonical mode of DNA binding, distinct from that of most ETS proteins. Although wHTH proteins typically contact DNA via the recognition helix H3, several different binding mechanisms have been identified.³⁵ Using both NMR spectroscopy and X-ray crystallography, we clearly demonstrated that ETV6 binds specific DNA via the canonical ETS domain interface. Furthermore, multiple lines of experimental evidence, including random coil chemical shifts, low ¹⁵N-NOE values and high structural rms deviations, prove that residues forming the inhibitory helix H5 undergo a folded-to-unfolded conformational change in order to form the DNA-bound state. These data also exclude alternative models such as the displacement of the folded helix to a non-inhibitory position. Collectively, this leads to a simple steric mechanism of auto-inhibition in which helix H5 and DNA compete in a mutually exclusive manner for the DNA-binding interface of ETV6. Importantly, amide hydrogen exchange studies revealed that helix H5 is only marginally stable and thus poised to unfold.¹⁶ Through thermodynamic linkage, the modest energetic penalty of the requisite unfolding of this helix contributes to the net ~ 50-fold reduction of overall DNA affinity.

In addition to occluding the DNA-binding interface of ETV6, helix H5 also suppresses msec-μsec timescale dynamics in the ETS domain.¹⁶ Given that the structures of the ETS domain in ETV6 fragments with or without the inhibitory helix H5 are similar in both the free and DNA-bound states, these motions likely reflect small scale excursions about a common average conformation. This structural plasticity is hypothesized to be a central feature along the protein-DNA recognition pathway, as exemplified by the well-studied lac and CAP repressors.^36-38 Although experimentally challenging, delineating the precise roles played by ETV6 backbone and side chain mobility in both DNA binding and its auto-inhibition by the CID is an important future objective.

Non-specific DNA binding and its auto-inhibition

Sequence-specific DNA-binding proteins also have significant affinity for non-specific sequences. In a living cell, such non-specific sites outnumber the specific sites by many orders of magnitude. Although this poses a severe challenge for these proteins to locate their target sites, non-specific binding also greatly helps in this process by both buffering the concentration of free protein and by “facilitated target location”.^{39, 40} Experimental and theoretical studies indicated that, upon initially encountering a non-specific site, a protein will undergo sliding (one-dimensional diffusion) along the DNA, combined with direct transfer or jumping (three-dimensional diffusion after transient dissociation) between DNA segments in order to search for its specific, high affinity target sites.^41-46

A growing number of proteins bound to non-cognate DNA sequences have been characterized by X-ray crystallography (e.g., steroid receptors,^{47, 48} cro repressor,⁴⁹ MATα2 homeodomain,⁵⁰ several restriction enzymes,^51-53 and Dam methyltransferase⁵⁴) and NMR spectroscopy (e.g., lac repressor,^{36, 37} a HMG-box protein,²⁶ HoxD9 homoedomain,^{41, 55} the bipartite Oct1,⁴⁴ papillomavirus E2,⁵⁶ and the Egr-1⁴⁵ and ZNF217⁴⁵ zinc fingers). In each of these cases, the same approximate interface is utilized to bind both non-specific and specific DNA. However, at a more detailed level, there are necessarily differences in precise intermolecular orientation, protein-DNA contacts, buried surface area, and so forth, that account for the differences in affinity, and hence specificity, towards various DNA sequences. Broadly speaking, the interaction of sequence-specific DNA-binding proteins with non-specific DNA results primarily from electrostatic interaction between the positively-charged side chains of the protein and the negatively-charged DNA backbone.^{36, 52} It is also postulated that non-specific complexes are relatively dynamic with a high degree of hydration^{48, 57} and significant backbone and sidechain motions^{36, 37, 44, 45, 55} that aid in the efficient scanning for a target sequence. Once the cognate site is located, the protein and DNA undergo conformational changes leading to tight base-specific interactions.^58-60

We have characterized the interaction of the ETV6 ETS domain with a non-specific DNA oligonucleotide that lacks the ^5’GGAA^3’ motif. Consistent with their relative affinities, NMR-monitored titrations show that specific and non-specific interactions are in slow (k_ex < Δω) and fast (k_ex > Δω) exchange regimes, respectively, where k_ex is the exchange rate constant (k_ex = k_on[DNA] + k_off) and Δω is the difference in chemical shifts between the free and bound states of ETV6. The faster exchange with non-specific DNA likely reflects a shorter lifetime (1/k_off) of the bound state, as would be required for rapid target location. Conversely, the longer lifetime of the specific ETV6-DNA complex at regulatory sites would enable subsequent transcriptional repression. NMR chemical shift measurements confirm that ETV6 binds both specific and non-specific sequences using the same canonical ETS domain interface. However, the non-specific complex appears relatively loose, with weaker, dynamic time-averaged interactions. This conclusion is based on the three to four orders of magnitude higher K_D values of various ETV6 constructs for non-specific versus specific DNA's, along with the general pattern of amide ¹H^N and ¹⁵N chemical shifts changing in the same direction, yet with smaller magnitudes, upon binding DNA^nonsp versus DNA^sp. Since the ETV6 backbone structure is similar in both complexes, switching from a non-specific to specific complex likely involves conformational rearrangements of amino acid side chains in the binding interface without any significant global changes.

We also discovered that the non-specific DNA binding by ETV6 is auto-inhibited and accompanied by the unfolding of helix H5. This is consistent with the need to remove the steric blockage of the ETS domain interface. However, non-specific binding is only attenuated by ~ 5-fold, whereas specific DNA binding is auto-inhibited by ~ 50-fold. This may reflect a contribution of weak non-specific DNA binding via a non-canonical interface not occluded by the CID.⁵⁰ To the best of our knowledge, this is the first instance where auto-inhibition has been quantified for non-specific DNA binding. However, previous circular dichroism spectroscopic studies had indicated helical unfolding of ETS-1 upon binding to both specific and non-specific DNA.⁶¹ Thus, auto-inhibition of non-specific DNA binding is likely a general feature of sequence-specific regulatory proteins such as those of the ETS family.

ETV6 exemplifies further diversity of ETS family auto-inhibitory mechanisms

Although auto-inhibition of DNA binding has been reported for several ETS proteins.², detailed molecular mechanisms have only been elucidated for ETS-1, ERG and now, ETV6. In the case of ETS-1, unfolding of its helical inhibitory module is linked allosterically to DNA binding. Furthermore, transient “fuzzy” interactions⁶² by an adjacent disordered SRR increases auto-inhibition in a phosphorylation-dependent manner.^{6, 9} Auto-inhibition by a flexible sequence in ERG appears akin to that of the SRR in ETS-1.¹⁰ In contrast, ETV6 lacks N-terminal inhibitory sequences and contains a helical CID that sterically blocks its ETS domain. Although structurally distinct, ETS-1 and ETV6 auto-inhibition also share several common features, including coupled unfolding of a marginally stable helix with DNA binding and dampening of ETS domain dynamics by the appended inhibitory sequences. The latter dynamic changes were also reported to occur with ERG.¹⁰

ETS-1 activity is regulated both positively and negatively via auto-inhibition. In particular, the effect is progressively reinforced due to increasing levels of SRR phosphorylation in response to Ca⁺² signaling.⁶ Conversely, cooperative DNA binding with partner transcription factors, such as RUNX1⁴ and Pax-5,⁵ relieves auto-inhibition. Although the cellular control of ETV6 is less well characterized, it is reasonable to hypothesize that the CID could also integrate signaling pathways and protein partnerships to modulate DNA binding. Also, the self-association of ETV6 via its PNT domain may compensate for auto-inhibition through cooperative DNA binding to tandem ETS sites.¹⁵ The mechanistic insights presented in this manuscript provide a foundation for understanding such potential roles of auto-inhibition in regulating transcriptional repression by ETV6.

Materials and Methods

Protein expression and purification

Three murine ETV6 fragments, with the sole cysteine (Cys334) mutated to serine, were used in this study: ETV6^R426 (Gly329-Arg426), ETV6^D446 (Gly329-Asp446), and for preliminary experiments, ETV6^D446’ (Arg335-Asp446) lacking the first six residues (Table 1). These proteins differ from the previously characterized ETV6^R458 (Arg335-Arg458) by the exact N-terminal sequence and by the deletion of unstructured residues C-terminal to helix H5.¹⁶ Genes encoding the ETV6 fragments were cloned in pET28b+ vectors and expressed in E. coli BL21 (λDE3) cells. For ¹⁵N/¹³C-labeling, cell cultures were grown in M9 minimal media supplemented with 1 g/L ¹⁵NH₄Cl and 3 g/L ¹³C₆-glucose (or ¹³C₆/D₇-glucose for perdeuterated protein) as the sole nitrogen and carbon source, respectively. For perdeuterated and fractionally deuterated samples, cultures were grown in 99% and 70% D₂O M9 media, respectively. After induction for 6 hours at 30 °C, cells were lysed by sonication in the presence of 4 M guanidinium-HCl (H₂O). This denaturation step improved net yield and allowed complete back-exchange of the amide protons in the deuterated samples. The His₆-tagged proteins were purified by Ni²⁺ affinity chromatography with on-column refolding before elution. After thrombin cleavage of the His₆-tag, the non-native N-terminal residues Gly-Ser-His-Met remained. Using gel filtration chromatography (Superdex 75), the proteins were further purified and exchanged into final sample buffer (20 mM sodium phosphate, 50 mM NaCl, pH 6.5). Protein concentrations were determined by UV absorption using predicted molar absorptivity ε₂₈₀ values.⁶³

DNA samples for NMR, ITC and crystallographic studies

DNA oligonucleotides were purchased from Integrated DNA Technologies and Sigma-Aldrich (without HPLC purification). Double-stranded DNA duplexes were generated by mixing the relevant single strands at equimolar ratio (determined from predicted molar absorptivity ε₂₆₀ values), heating to 100°C for 5 mins, and slowly cooling to room temperature. For NMR and ITC titration experiments, gel-filtration chromatography was used to purify and buffer exchange the resulting duplex DNA duplexes. For crystallographic studies, the preformed 1:1 protein-DNA complex was purified by gel-filtration chromatography.

NMR spectroscopy

NMR experiments were performed using TCI-cryoprobe equipped Bruker Avance III 500, 600 and 850 MHz spectrometers at 25 °C. The proteins were 0.3 – 0.6 mM in sample buffer (20 mM phosphate, 50 mM NaCl, pH 6.5) with 6% lock D₂O. The collected spectra were processed and analyzed using NMRPipe⁶⁴ and Sparky,⁶⁵ respectively. Signals from the ¹H, ¹³C, and ¹⁵N nuclei in the backbone of ¹³C/¹⁵N-labeled ETV6^D446 and the backbone and side chains of ¹³C/¹⁵N-labeled ETV6^R426 were assigned using standard heteronuclear scalar correlation experiments.⁶⁶ Protein signals from DNA^sp complexes of amide-protonated ¹³C/¹⁵N-labeled ETV6^D446 with uniformly deuterated, randomly factional (70%) deuterated, or fully protonated side chains were assigned using TROSY-based experiments.⁶⁷. To ensure saturation, the DNA^sp was present in a 1.1 fold molar excess. Protein signals for the ETV6^D446-DNA^nonsp complex were assigned using ¹⁵N-HSQC spectra to monitor the titration of the ¹⁵N-labeled ETV6^D446 (initial 0.2 mM, final 0.15 mM) with aliquots of a 1.8 mM stock solution of DNA^nonsp in sample buffer. The molar ratios of DNA^nonsp to ETV6^D446 in the titration set were 0, 0.1, 0.25, 0.5, 0.75, 1, 2 and 3, yielding a final protein saturation of ~ 98% based on the data of Figure 4. Binding occurred in the fast exchange limit; thus, amide ¹H^N and ¹⁵N assignments were obtained by tracking shifts relative to the initial free ETV6^D446, and ¹³C assignments were then determined from standard ¹H/¹³C/¹⁵N-correlation spectra of the saturated complex.

Amide ¹⁵N relaxation

Amide ¹⁵N relaxation data (T₁, T₂, heteronuclear NOE)⁶⁸ were collected for the specific complex of perdeuterated ETV6^D446’ and DNA^sp at 25 °C with a 600 MHz NMR spectrometer. Relaxation rate constants were determined with Sparky⁶⁵ by fitting the peak heights to an exponential decay. Heteronuclear {¹H}-¹⁵N NOE data were also collected for ¹⁵N-labeled ETV6^D446 complexed with DNA^nonsp. The ¹⁵N-NOE value was determined from the ratio of the peaks heights versus a control reference spectrum without ¹H saturation. The global tumbling correlation time for the specific complex was calculated using Tensor2.²⁰

NMR structure calculations

NOE-derived distance restraints for ¹³C/¹⁵N-labeled ETV6^R426 and ETV6^D446–DNA^sp were obtained simultaneous 3D ¹H-¹⁵N/¹³C-¹H NOESY-HSQC (aliphatic/aromatic; τ_mix = 110 ms), and constant time methyl-methyl and amide-methyl ¹⁵N/¹³C-¹³C-¹H NOESY spectra (τ_mix = 100 ms).^{69, 70} The NMR-derived structure ensembles of ETV6^R426 and the ETV6^D446–DNA^sp complex (protein only) were calculated using CYANA 3.0⁷¹ with chemical shift assignments, dihedral angle restraints from TALOS+,⁷² NOESY crosspeaks, and manually assigned methyl-methyl distance restraints as input data (Table 2). Structure calculations, combined with automated NOESY spectra assignments, were performed in 7 iterative steps each yielding 100 structures. The 10 lowest energy structures from the final step were further refined with CNS using explicit solvent and molecular dynamics simulations.⁷³ Although present in the ETV6^D446–DNA^sp complex, the signals from nuclei in DNA^sp were not assigned and its co-ordinates were not included in these calculations. Given this lack of chemical shift assignments, any intermolecular NOEs between ETV6^D446 and DNA^sp in the input NOESY crosspeak list were thus discarded by CYANA. Secondary structure boundaries were determined using DSSP and figures rendered with PyMol.⁷⁴

Crystallization and structure determination

Low salt buffer containing 20 mM HEPES and 50 mM NaCl at pH 7.5 was used for crystallization. ETV6^R426 was mixed with DNA^cryst-sp in 1:1.1 molar ratio and the resulting complex was purified by gel filtration chromatography. The purified ETV6^R426–DNA^cryst-sp was concentrated to ~0.3 mM. Crystallization trials were carried out by the hanging drop vapor diffusion method using 1 ml reservoir solution of 50 mM sodium cacodylate, 100 mM ammonium acetate, 10 mM MgCl₂, and 22% PEG 8000 at pH 6.0, and a mixture of 2 μL complex and 2 μL of well solution. Crystals were obtained at room temperature within 5-7 days.

ETV6^R426–DNA^cryst-sp crystals were soaked stepwise for a few seconds each in mother liquor supplemented with 5% and 10% PEG 8000, and flash frozen using liquid nitrogen. A 2.2 Å resolution dataset was collected at 100 K using the Stanford Synchrotron Radiation Lightsource beam-line 7-1 with 0.9753 Å incident radiation. After data processing with HKL2000,⁷⁵ MolRep⁷⁶ was used for initial phase determination using the co-ordinates of the Elk-1 ETS domain (PDB ID: 1DUX) as a starting model. Cycles of structure refinement and building were performed using Phenix,⁷⁷ Refmac 5⁷⁸ from the CCP4 suite of programs,⁷⁹ and COOT.⁸⁰ Water molecules were automatically added using Phenix and manually corrected. The crystal structure of ETV6^R426–DNA^cryst-sp was determined in space group P3₁2₁ with one monomer in the asymmetric unit. Sufficient electron density was observed to build residues 335 to 424 of ETV6^R426 and the full duplex oligonucleotide. Ramachandran statistics indicated excellent stereochemistry with 98.8% of residues in the preferred region and no outliers. Data collection and refinement statistics are listed in Table 3.

Isothermal titration calorimetry

ITC measurements were performed at 25 °C with a Microcal ITC200. The protein and DNA samples were buffer-exchanged by gel filtration into 20 mM sodium phosphate and 50 mM NaCl at pH 6.5. Titrations were carried out with 20 μM DNA^nonsp in the 200 μL reaction cell and 2 μL injections of ETV6^D446 (0.58 mM) or ETV6^R426 (0.40 mM) at 3 min intervals for a total of 20 injections. Heat of dilutions, measured by titrating proteins into buffer and buffer into DNA, was subsequently subtracted from the respective titration experiments. The binding of ETV6 fragments to non-specific DNA was analyzed using the Record model²² as described by equations (1-3)

$$\frac{{\upsilon }_{\mathrm{NS}}}{\left[\mathrm{L}\right]}={\mathrm{K}}_{\mathrm{A}}(1-\mathrm{n}{\upsilon }_{\mathrm{NS}}){\left(\frac{1-\mathrm{n}{\upsilon }_{\mathrm{NS}}}{1-(\mathrm{n}-1){\upsilon }_{\mathrm{NS}}}\right)}^{\mathrm{n}-1}\left(\frac{\mathrm{N}-\mathrm{n}+1}{\mathrm{N}}\right)$$ (1)

The non-specific binding density ν_NS = [ETV6]_bound/[total DNA bp], [L] is [ETV6]_free, K_A is the average microscopic association constant to any potential DNA site, n is the number of DNA base-pairs bound by the protein, and N is the total base-pairs in the oligonucleotide. In the ITC experiments, the heat content Q_i after each injection ‘i’ is given by:

$${\mathrm{Q}}_{\mathrm{i}}={\upsilon }_{\mathrm{NS},\mathrm{i}}\left[\text{total DNA bp}\right]\Delta \mathrm{H}{}^{\mathrm{o}}\mathrm{V}_{\mathrm{o}}$$ (2)

The reaction cell volume is V_o, binding enthalpy is ΔH^o and binding density after injection ‘i’ is ν_NS,i. The differential heat ΔQ_i measured by the ITC instrument is given by:

$$\Delta {\mathrm{Q}}_{\mathrm{i}}={\mathrm{Q}}_{\mathrm{i}}+\frac{{\mathrm{V}}_{\mathrm{i}}}{{\mathrm{V}}_{\mathrm{o}}}\left[\frac{{\mathrm{Q}}_{\mathrm{i}}+{\mathrm{Q}}_{\mathrm{i}-1}}{2}\right]-{\mathrm{Q}}_{\mathrm{i}-1}$$ (3)

The ITC binding isotherms were fit to these equations with Matlab. Errors in each fitted parameter were estimated from Monte Carlo simulations.

Highlights.

• An inhibitory helix sterically blocks the ETV6 DNA-binding interface.

• The inhibitory helix unfolds when ETV6 binds specific, as well as, non-specific DNA.

• ETV6 binds specific and non-specific DNA via the same canonical ETS domain interface.

• A non-conserved histidine helps define sequence specificity of ETV6.

• A unified mechanism regulates specific and non-specific DNA-binding auto-inhibition.

Abbreviations

ITC

isothermal titration calorimetry

SRR

serine rich region

CID

C-terminal inhibitory domain

NMR

nuclear magnetic resonance

CSP

chemical shift perturbation

HSQC

heteronuclear single quantum correlation