Structural characterization of the PIT-1/ETS-1 interaction: PIT-1 phosphorylation regulates PIT-1/ETS-1 binding

Abstract

The POU-domain transcription factor Pit-1 and Ets-1, a member of the ETS family of transcription factors, can associate in solution and synergistically activate the prolactin promoter by binding to a composite response element in the prolactin promoter. We mapped the minimal region of Ets-1 required for the interaction with the Pit-1 POU-homeodomain. Here, we describe a detailed NMR study of the interaction between the POU-homeodomain of Pit-1 and the minimal interacting region of Ets-1. By using heteronuclear single quantum coherence titration experiments, we were able to map exact residues on the POU-homeodomain that are involved in the interaction with this minimal Ets-1 interaction domain. By using our NMR data, we generated point mutants in the POU-homeodomain and tested their effect on the interaction with Ets-1. Our results show that phosphorylation of Pit-1 can regulate the interaction with Ets-1.

The POU-domain transcription factor Pit-1 is one of the key regulators of gene expression in the anterior pituitary gland. The presence of Pit-1 is an essential and determining factor in the differentiation of the somatotrope, lactotrope, and thyrotrope cell lineages. Furthermore, in these cell lineages, Pit-1 is required for the expression of the growth hormone, prolactin (PRL), and thyrotropin β (TSHβ) genes, respectively (1–3). The expression of these genes is restricted to their respective cell lineages. However, because Pit-1 is present in all three cell types, cooperation with other transcription factors is required to achieve selective expression.

One example of a promoter-specific Pit-1 interaction is the Ras-activated expression of the PRL promoter. Ras-dependent activation of the PRL promoter is achieved through Ets-1 and requires the interaction of Pit-1 with Ets-1. Ets-1 is one of the founding members of the ETS family of transcription factors, which is defined by a highly conserved winged helix-turn-helix DNA-binding domain (4). Previously, it was found that Ras activation of the PRL promoter required the presence of a Pit-1/Ets-1 composite element in the PRL promoter (5, 6). Furthermore, we have shown that Pit-1 and Ets-1 form a DNA-independent complex in solution (7). The physical association of these proteins involved the POU-homeodomain of Pit-1 (Pithd) and a part of the region III activation domain of Ets-1 located between the N-terminal Pointed domain and the C-terminal ETS DNA-binding domain (7, 8).

In the present study, we further define the minimal region of Ets-1 required for the in-solution interaction with the Pithd. Both this minimal interacting region on Ets-1 and the Pithd were capable of interacting with one another outside the context of the full-length proteins. By using NMR spectroscopy, coupled with the known crystal structure of Pit-1 (9), we performed a fine mapping of the exact residues on the surface of the Pithd that are in contact with the minimal interacting region of Ets-1. Our results imply that the posttranslational state of Pit-1 is an important determinant for the physical association with Ets-1. Furthermore, we define a previously uncharacterized interaction domain that may apply for other POU homeodomain interactions. Our approach presents an interesting possibility to probe weaker protein–protein interactions.

Materials and Methods

Plasmid Construction.

The wild-type pGex Pit-1 (199–291) expression vector has been described (8). The mutant Pithd fragments of rat Pit-1 (T220D, K226A, K226D, E254A, W261F, R268A, K270A, and L288A) were generated by overlap extension polymerase chain amplification by using primers purchased from GIBCO. Mutant primers were used in combination with 5′ and 3′ Pit-1 (199–291) primers that incorporate a NotI site to facilitate subcloning. The amplified DNA was initially cloned into pCR 2.1 (Invitrogen) and then subcloned into the NotI site of pGexDFGK. The sequences of the primers are as follows: 5′Pit-1 (199–291)-GCGGCCGCCAGGTCGGAGCTTTGTACAAT, 3′Pit-1 (199–291)-GCGGCCGCTTATCTGCACTCAAGATGCTC. 5′T220D-GAGGACAGATATCAGTATCGC, 3′T220D-GATACTGATATCTGTCCTCCGT, 5′K226A-GCCGCTGCCGATGCTTTGGAGA, 3′K226A-AAAGCATCGGCAGCGGCGATAC, 5′K226D-GCCGCTGACGATGCTTTGGAG, 3′K226D-AAGCATCGTCAGCGGCGATAC, 5′E254A-TGAATCTCGCCAAAGAAGTAGTAAG, 3′E254A-TACTTCTTTGGCGAGATTCAATTC, 5′W261F-GTAAGAGTGTTCTTTTGCAACCG, 3′W261F-GTTGCAAAAGAACACTCTTACTAC, 5′R268A-AAGGCAGGCCGAAAAACGGGT, 3′R268A-CGTTTTTCGGCCTGCCTTCGG, 5′K270A-AGAGAGAAGCCCGGGTGAAAAC, 3′K270A-CACCCGGGCTTCTCTCTGCC, 3′L288A-GAGCGGCCGCTTATCTGCACTCGGCATGCTCCTTTG.

Plasmid pSG5c-Ets-1 encodes the p68 chicken Ets-1 under control of the SV40 early promoter. This construct as well as the series of C-terminal deletions of Ets-1, pSG5c-Ets-1Δ3-1 through pSG5c-Ets-1Δ3-10 (10), and the glutathione S-transferase (GST)-Ets-1 fusion construct, pGex p68 Ets-1 full length, were provided by Bohdan Wasylyk (Institut de Genetique et de Biologie Moleculaire et Cellulaire, Illkirch, France). pGex Ets-1 (190–257) was generated by combining the amino terminus of pSG5 Ets-1Δ5-4 (190–485) with the C terminus of pSG5 Ets-1Δ3-8 (1–257) by using a common restriction site. The resulting construct was subcloned into pGexDFGK or pRSETA.

The fragment encoding residues 213–289 of the human Pit-1 ORF was cloned into an NdeI/BamHI-digested pET15b expression vector (Novagen), yielding His-6-tagged Pithd. The fragment encoding residues 190–257 or 235–304 of the human Ets-1 protein was cloned into an NdeI/BamHI-digested pET15b-derived vector. This vector was obtained by using the primers 5′GGAAATACTTACCCATGGGCGATAAAATT3′ and 5′ATATGAGATCCCATGGGTACCTTGTCATCGTC3′ purchased from Amersham Pharmacia to clone the ORF for thioredoxin from the pTrxFus vector (Invitrogen) into the NcoI site of pET-15b. The resulting vector codes for an N-terminal thioredoxin tag in addition to the His-6-tag. All DNA sequences were checked by cycle sequencing on an automated sequencer (Prism, Applied Biosystems).

GST Pulldown Assays.

Recombinant fusion proteins were prepared from bacterial extracts and bound to glutathione Sepharose CL-4B (Amersham Pharmacia) as described (8). Ets-1 and Pit-1 fusions also were supplemented with 10 and 1 mM DTT, respectively. Protein concentration was measured by the Bio-Rad assay. ³⁵S-labeled proteins were synthesized by using [³⁵S]methionine (NEN) and the TnT-coupled transcription-translation reticulocyte lysate system with T7 polymerase (Promega). Pulldown assays were performed in dilute 500 μl solution by using equal amounts (20 μg) of GST fusion proteins bound to 20 μl of glutathione Sepharose beads in the presence of 50 μg/ml ethidium bromide, as described (8). The saturation curves were basically done by the method of Darimont et al. (11). Increasing concentrations of GST fusion proteins (0.35 to 100 mM, 7 points) bound to 25 μl of glutathione Sepharose beads were suspended in a total volume of 0.05 ml binding buffer with 20 mg/ml BSA and incubated on a rotator at 4°C for 30 min. ³⁵S-labeled Ets-1 (0.5 nM, 2–3 μl) was added to the GST-fusion samples and rotated at room temperature for 1 h. The beads were pelleted by centrifugation at 1000 × g for 1 min and washed 4× with 0.5 ml of binding buffer with 5 mM NaFl/0.2 mM Na Vanadate/0.1% Triton X-100. The ³⁵S-labeled proteins were eluted from the beads by boiling in SDS sample buffer and analyzed by SDS-PAGE and autoradiography. Bands were quantified by using a Molecular Dynamics PhosphorImager with IMAGEQUANT software and normalized for input of equal amounts of fusion protein based on densitometry of the Coomassie blue-stained gels. K_D values were determined by nonlinear regression analysis by using a one-site hyperbolic curve fit and holding B_max at 100% of input.

Phosphorylation.

Pithd fusion protein bound to glutathione Sepharose was washed with 10 bed volumes of 1× HMK buffer (20 mM Tris, pH 7.5/100 mM NaCl/12 mM MgCl₂). The bead slurry containing 1 mg of protein was resuspended in 3.7 ml of 1× HMK buffer containing 1 mM DTT, 1 mM ATP, 5 mM NaF, and 500 units of catalytic subunit of protein kinase A from bovine heart (Sigma) and rotated at 30°C for 6 h. The extent of phosphorylation was determined to be greater than 70% for T220 and the protein kinase A (PKA) site located at the carboxyl terminus of the GST fusion by incubating 1/10 of this reaction with [γ-³²P]ATP and measuring the extent of ATP incorporation. Previous studies have shown that T219 is phosphorylated only under extreme conditions in the absence of the T220 phosphorylation site (12). The reaction was stopped and the glutathione Sepharose was washed 5× with 10 bed volumes of PBS containing 1× Complete protease inhibitors (Roche Molecular Biochemicals) with EDTA/5 mM NaFl/0.2 mM Na vanadate.

Protein Purification.

Pithd was expressed in BL21 (DE3) bacteria grown on synthetic medium (10.38 g/liter K₂HPO₄⋅3H₂O/4.38 g/liter KH₂PO₄/50 mg/liter MgSO₄⋅7H₂O/7 mg/liter (NH₄)₂FeSO₄⋅6H₂O/10 mg/liter thiamin/0.05 μg/ml ampicillin) containing 0.5 g/liter ¹⁵NH₄Cl as the sole nitrogen source and 5 g/liter ¹²C-glucose or 1 g/liter ¹³C-glucose as the sole carbon source. Bacteria were grown to an OD₆₀₀ of 1.0 and induced by the addition of isopropyl β-d-thiogalactoside to a final concentration of 1 mM. After 2.5 h, the cells were harvested and lysed in 50 mM Na_xH_yPO₄ (pH 6.5)/500 mM NaCl/10% (vol/vol) glycerol/0.01% Nonidet P-40/0.5 mM PMSF/1 μg/ml aprotinin/10 mM β-mercaptoethanol/10 μg/ml lysozyme. The lysate was cleared by centrifugation at 30,000 × g for 45 min at 4°C and loaded onto an Ni-NTA column (Qiagen) equilibrated in 50 mM Na_xH_yPO₄, pH 6.5/200 mM NaCl/10% (vol/vol) glycerol/20 mM imidazole/10 mM β-mercaptoethanol. After washing, the protein was eluted in a linear gradient from 20–400 mM imidazole. Peak fractions were pooled, diluted to 100 mM NaCl, and loaded onto an SP-Sepharose column (Amersham Pharmacia) equilibrated in 50 mM Na_xH_yPO₄, pH 6.5/100 mM NaCl/10% (vol/vol) glycerol and 10 mM β-mercaptoethanol. The protein was eluted in a linear gradient from 100–1,000 mM NaCl. Peak fractions were pooled and concentrated to ≈1 mM in an Amicon stirred ultrafiltration cell (Millipore) by using a 3 kDa cut-off filter. The buffer was exchanged to a final 50 mM Na_xH_yPO₄, pH 5.5/100 mM NaCl/1 mM DTT. The thioredoxin-his₆-Ets-1 (190–257) or (235–304) were expressed in BL21 (DE3) bacteria grown on LB medium containing 0.05 μg/ml ampicillin. Expression and lysis were done as above, with the exception that induction took place at an OD₆₀₀ of 0.6. The cleared lysate was bound to Ni-NTA material (Qiagen) equilibrated in 30 mM Tris⋅HCl, pH 8.0/200 mM NaCl/10% (vol/vol) glycerol/20 mM imidazole/10 mM β-mercaptoethanol by rotating for 1 h at 4°C. The material was washed and block eluted with 400 mM imidazole. Eluted protein was further purified on a Superdex 75 gel filtration column (Amersham Pharmacia) equilibrated in 30 mM Tris⋅HCl, pH 8.0/200 mM NaCl/10% (vol/vol) glycerol/20 mM imidazole/10 mM β-mercaptoethanol. Peak fractions were subjected to thrombin cleavage by adding 2.5 mM CaCl₂ and 100 units of thrombin from bovine plasma (Sigma) and incubating at room temperature for 2 h. Cleaved protein was applied to an Ni-NTA column (Qiagen). Ets-1 (190–257) or Ets-1 (235–304) were present in the flow through, while thioredoxin-his₆ remained bound to the column. After a second run on the Superdex 75 gel filtration column (Amersham Pharmacia), the sample was concentrated and washed as described above.

NMR Spectroscopy.

NMR experiments were carried out at 305 K (32°C) on Varian UnityInova 500 and 750 spectrometers. The protein concentration was 1 mM in all experiments; 10% (vol/vol) D₂O was added to obtain a lock signal. Spectra were processed by using NMRPIPE (13) and analyzed by using NMRVIEW (14). Protein–protein titrations were carried out by the repeated addition of Ets-1 (235–304) to a ¹⁵N-labeled Pithd and the recording of a 750 MHz ¹H¹⁵N-HSQC spectrum after each addition. The sample was concentrated in an Amicon stirred ultrafiltration cell to keep the total volume around 500 μl.

Results

Mapping of the Minimal Interacting Region on Ets-1.

To identify the minimal region of Ets-1 required for interaction with Pit-1, we used 3′ deletions of p68 chicken Ets-1 and tested their ability to interact with GST-bound Pithd (residues 199–291). As shown in Fig. 1B, ³⁵S-labeled in vitro-translated full-length Ets-1 was capable of binding to the Pithd. Deletion of the C-terminal auto inhibitory region caused a 50% decrease in binding. This was likely caused by changes in the overall structure of the ETS domain, because subsequent deletions up to Δ3-4 increased binding affinity almost to wild-type level. The most C-terminal Ets-1 deletion that retained Pit-1 binding was truncated at amino acid 257 (Δ3-8), indicating that amino acids 257–485 do not contribute to binding. However, deletion of 68 amino acids to position amino acid 189 (Δ3-9) resulted in almost complete loss of Pit-1 binding, and further deletion to amino acid 71 resulted in complete loss of binding activity. Thus, binding affinity for Pit-1 dropped drastically with deletions in the Ets-1 RIII activation domain.

Figure 1.

Residues 190–257 of the Ets-1 RIII are necessary and sufficient to bind Pit-1. Equal amounts of in vitro-translated ³⁵S-labeled proteins were bound to 5 (B) or 20 μg (C) of GST control or fusion proteins, as described in Materials and Methods. Nonspecific binding to GST alone was subtracted from total binding, and specific binding is expressed as % of input. Data are representative of three separate experiments. (A) Schematic representation of Ets-1 and the constructs used in this study. (B) Deletion mapping of binding of ³⁵S-p68 Ets-1 and Ets-1 truncations Δ3-1 (AA450), Δ3-2 (AA412), Δ3-3 (AA390), Δ3-4 (AA368), Δ3-5 (AA358), Δ3-6 (AA314), Δ3-7 (AA275), Δ3-8 (AA257), Δ3-9 (AA189), and Δ3-10 (AA71) to the Pithd; 10% of input, nonspecific and specific binding of ³⁵S-Ets-1 and deletion constructs are shown. (C) The 190–257 region of Ets-1 is sufficient to bind in vitro-translated ³⁵S-Pit-1.

Combined with previous results using 5′ deletions to map the interaction (8), this study defines the region of Ets-1 required for Pit-1 binding as part of the RIII activation domain, located between residues 190–257. To confirm that this region is sufficient, we expressed it as a GST fusion. As shown in Fig. 1C, GST-Ets (190–257) could indeed bind Pit-1. Because neither the N-terminal Pointed domain, nor the C-terminal Ets-1 DNA-binding domain are required for the interaction, we can consider the 190–257 region as an independent binding domain within Ets-1. Therefore, we chose to study the interaction of the Pithd with this minimal Ets-1 region, outside the context of the full-length protein.

NMR Studies on the Interacting Residues in Pithd.

The chemical shifts of amide protons are highly sensitive to the binding of an interaction partner or ligand (15), or conformational changes in the protein (16). By using this sensitivity, we investigated which amide protons of the ¹⁵N-labeled Pithd (residues 213–289) show changes in chemical shift in a ¹H-¹⁵N correlated heteronuclear single quantum coherence (HSQC) experiment upon addition of the unlabeled interaction domain of Ets-1 (residues 190–257).

The ¹H-¹⁵N HSQC spectrum of the Pithd (residues 213–289) recorded at 750 MHz showed well dispersed signals, which is indicative of an ordered structure (Fig. 2A). Unfortunately, addition of any amount of Ets-1 (190–257) resulted in heavy precipitation and only minor changes in the Pithd spectrum. Attempts to solve this problem by changing the pH and salt concentration did not enhance solubility (not shown). Epitope-scanning experiments along the Ets-1 (190–257) region, mapped a persistent interaction with Pit-1 to the (235–257) region (not shown). Therefore, we chose to use an Ets-1 construct containing this (235–257) region, extended toward the C terminus of Ets-1. The resulting (235–304) is less hydrophobic in nature and did not precipitate when added to the Pithd at pH 5.5 and 100 mM NaCl. The ¹H-¹⁵N HSQC of Ets-1 (235–304) showed a modest dispersion of signals (Fig. 2B), which indicates that the peptide is not completely unfolded. Indeed, circular dichroism measurements showed a modest helical content, which could be increased by the addition of trifluoroethanol (not shown).

Figure 2.

NMR analysis of the Ets-1 (235–304) interaction with the Pithd. (A) The ¹H-¹⁵N-correlated HSQC spectrum of Pithd (213–289) recorded at 750 MHz. The selected region shows superimposed spectra of free Pithd (blue) and after the addition of 0.5 (green) and 1 (red) molar equivalent of Ets-1 (235–304). The chemical shift values of some N-H signals (like T220) change drastically upon addition of Ets-1, whereas others are moderately (R268) affected. (B) The ¹H-¹⁵N-correlated HSQC spectrum of Ets-1 (235–304) recorded at 750 MHz. (C) Combined ¹⁵N-¹H chemical shift perturbations mapped against the secondary structure of the Pithd. The values are the means of two independent experiments and were calculated as Euclidian distances between peaks: ΔδNH = [(Δδ¹⁵N)² + (Δδ¹H)²]^1/2, where Δδ¹⁵N and Δδ¹H denote the nitrogen and proton shifts in Hz, respectively.

A number of backbone amide protons display a change in chemical shift that depends on the amount of Ets-1 (235–304) added, whereas others are unaffected (Fig. 2A, enlarged region). These changes were specific for Ets-1, as an equimolar amount of BSA was unable to affect the spectrum (not shown). Because these backbone amides were the same as the ones displaying very minor shifts after the addition of Ets-1 (190–257), we are confident that the majority of the interaction indeed resides in the overlapping region between our two constructs. The gradual, concentration-dependent changes in chemical shift in the presence of Ets-1 indicate that the two proteins are in fast exchange on the NMR timescale. We were able to add Ets-1 (235–304) up to about equimolar level to Pithd. Further addition of protein resulted in heavy precipitation without any further change in chemical shifts. SDS/PAGE analysis of the precipitate showed that it contained mostly Ets-1 (235–304) (data not shown). Given the moderate changes in chemical shift, the two peptides may not yet have reached saturation state of binding when precipitation occurs. This finding indicates that, at the conditions used, the K_D for the minimal domain complex is in the micro- to millimolar range.

To identify the amide protons involved in the interaction, we purified ¹³C/¹⁵N double-labeled Pithd. By using triple resonance experiments [HNCA, HN(CO)CA, CBCANH and CBCA(CO)NH], coupled with 3D-¹⁵N-NOESY HSQC, we could assign the resonance frequencies of nearly all backbone amide protons (see Table 1, which is published as supporting information on the PNAS web site, www.pnas.org). Our assignment allowed us to plot the difference in chemical shift in the absence and presence of Ets-1 (235–304) against the primary structure (Fig. 2C).

The overall picture shows two apparent interacting regions, with a less affected “bridging” region. The most drastic changes in chemical shifts of backbone amides are located around Thr-219 and Thr-220 in the N-terminal part of the Pithd. Interestingly, these threonines can be phosphorylated in vivo (17). The amide protons at the C terminus of the DNA-binding helix of the Pithd form a moderately affected region. The flexible C terminus outside the Pithd was hardly affected, except for H287 and L288 at the very end of the protein. A surprisingly large change in chemical shift occurred on the amide proton of W261. In the crystal structure, the side chain of this residue is almost completely buried in the hydrophobic core of the Pithd and is, therefore, unlikely to be involved in Ets-1 binding. However, there is a hydrophobic pocket around the amide proton of W261, formed by V257, V258, I221, and part of the tryptophan side chain itself in which Ets-1 might dock. The only other amide protons, besides the W261 backbone amide, bordering this hydrophobic pocket are the backbone amides of V258, I221, and T219. The amide protons of all these residues shift drastically upon addition of Ets-1.

Mutational Analysis.

In previous studies using GST-pulldown assays, deletion of either the N- or C-terminal half of the Pithd resulted in a 50% decrease in pull-down efficiency (8). This finding suggests that there is more than one region involved in Ets-1 binding. The NMR results suggest binding of Ets-1 centers around T219/T220 and the hydrophobic pocket next to W261, possibly with stabilizing contacts along the DNA-binding helix and the C terminus. To test these possibilities, we constructed several Pithd mutants and tested their ability to interact with full-length Ets-1 in GST-pulldown assays. As shown in Fig. 3A, several mutants were affected in Ets-1 binding.

Figure 3.

Effect of Pithd point mutations on binding of p68 c-Ets-1. In vitro-transcribed and -translated ³⁵S-labeled proteins were bound to 20 μg of GST fusion proteins, as described in Materials and Methods. Nonspecific binding to GST alone was subtracted from total binding, and specific binding is expressed as a % of input normalized for equal loading of fusion constructs. The graphs contain the mean ± SEM of data from three experiments. (A) Binding of ³⁵S-Ets-1 to GST and GST-Pit-1 (199–291) wild type and mutants. (B) Saturation curves for wild type, T220D, phosphorylated wild type, W261F and K226D mutants (see Materials and Methods for details).

In the crystal structure, the hydrophobic pocket next to W261 is partially covered by K226. Modifying this residue might alter the accessibility of the hydrophobic pocket. Indeed, both the K226A (which removes the positive side chain) and K226D (which switches the charge from positive to negative) mutations strongly increased Ets-1 binding. These results suggest that both the presence and charge of K226 are important in regulating the binding of Ets-1. Because the hydrophobic pocket is formed by a number of structurally important residues, it is not possible to remove it without affecting the overall structure of the Pithd. However, mutating W261 to a phenylalanine should alter the pocket to some extent, without drastically affecting the folding of the Pithd. Indeed, the binding affinity of W261F Pithd is increased sevenfold (lane 6).

Residue E254 showed a significant (>15 Hz) change in chemical shift upon addition of Ets-1 (235–304). This change in chemical shift might reflect the relative flexibility of this residue, which is located in the loop preceding the DNA-binding helix. However, in a previous study, mutation of this residue to alanine resulted in an increased binding affinity for N-CoR (18). Interestingly, this mutation also increased Ets-1 binding by about twofold (lane 5).

Moderate changes in chemical shift occurred along the C-terminal part of the DNA-binding helix. These changes may reflect stabilizing contacts with Ets-1 of the residues involved. However, mutations R268A and K270A hardly affected Ets-1 binding (Fig. 3A). Finally, we tested whether L288 at the C terminus is involved in Ets-1 binding. Surprisingly, L288A Pithd showed an increased binding affinity, which indicates that L288 may be in close proximity to a hydrophilic interaction domain.

Contrary to expectations based on the NMR results, we found only a slight decrease in Ets-1 binding with the T220D mutation, which mimics a phosphorylated Pithd. However, the fact that a Pit-1 phosphorylation mutant Pit-1-A₃ (S115A, T219A, T220A) showed a small increase in Ets-1 binding (data not shown) hinted that changes in this region can, in fact, influence Ets-1 binding. This result prompted us to do a more detailed analysis of the relative K_D, by measuring the saturation curve of the T220D mutant. In this procedure, weak interactions can be quantified and a relative K_D can be approximated. In addition we tested the binding curves of a PKA-phosphorylated Pithd and the K226D and W261F mutants. Consistent with the NMR results, the wild-type Pithd showed a relative K_D of 321.9 μM (Fig. 3B). Furthermore, both T220D and PKA-phosphorylated wild-type Pithd showed a clear increase in binding (twofold to threefold lowering of K_D) compared with the untreated wild-type Pithd. Only the K226D and W261F mutants were able to reach saturation of binding in this experiment. Thus, T220 is contacted by Ets-1 and the phosphorylation status of T220 is an important regulator of the Ets-1/Pit-1 interaction.

Discussion

There have been relatively few NMR studies on transcription factor protein–protein interactions to date (16, 19, 20). The interaction partners in these studies all formed high-affinity complexes. Here, we report an NMR study on a 75-aa polypeptide spanning the Pithd (213–289) and a 70-aa region of the Ets-1 RIII activation domain (235–304), which we have shown to be the essential domains for the interaction between the two full-length proteins. These two protein domains did not bind each other with high affinity (K_D ≈ 0.3 mM). Even with this rather weak interaction, we were still able to show clear changes in chemical shift of the amino protons of the Pithd after the addition of Ets-1 (235–304).

Because we were able to track all α-helices—they were resolved in the crystal structure by NH_{(i, i + 1)} contacts in the 3D-¹⁵N-NOESY HSQC—we presume that the DNA-bound crystal structure and the solution structure of the Pithd are highly similar. Therefore, we can directly correlate the amount of chemical shift difference per amide proton to the known crystal structure of the Pithd (Fig. 4A). All amide protons that displayed more than 10 Hz change in chemical shift are located on one side of the Pithd. One limitation of this approach is that changes in chemical shift for any given backbone amide are indirect evidence that its side chain is involved in the interaction. Therefore, we verified our results by interaction assays.

Figure 4.

Mapping of chemical shift perturbations to the Pithd structure. Surface density representations of the crystal structure of the Pithd bound to DNA. These fragments were taken from the crystal structure as resolved by Jacobson et al. (9). (A) The results from Fig. 2B indicated by color-coding. Chemical shift changes > 25 Hz are colored dark red, changes between 25 and 15 Hz are colored red and between 15 and 10 are colored pink. Unaffected residues are colored yellow. (B) Two orientations of K226 (blue) in the Pit-1 dimer with different accessibility of the hydrophobic pocket formed by V257, V258, I222, and W261 (colored in beige). The DNA has been deleted for clarity. (C) Model for association of Ets-1 to the Pithd. K226 is shown in blue, the hydrophobic pocket in beige and T220 in green. Because the crystal structure does not extend beyond residue K273, an extended tail was added to represent the C terminus of the Pithd. Ets-1 could, depending on the posttranslational state of the Pithd, dock in the hydrophobic pocket next to W261. Additional stabilizing contacts may be made at the C terminus of the DNA recognition helix (helix 3) and at the region around L288. These pictures were generated by using the Accelrys VIEWERLITE.

Our results show large changes in chemical shift of the backbone amides around T219 and T220, which indicates that some part of Ets-1 (235–304) is in close proximity to these residues. Regulated phosphorylation of these threonines occurs in vivo, either as targets of PKA or of a cell cycle-dependent kinase (17). There have been conflicting reports on the influence of phosphorylation of Thr-220 with regard to DNA binding (12, 17, 21). Pit-1-A₃, in which the three principal phosphorylation sites (S115, T219, T220) are mutated to alanines, was unaffected or slightly less efficient than wild type in activating transcription on the GH and PRL promoters in transient transfections (17, 22). Furthermore, Pit-1-A₃ was not affected in synergistic activation of the PRL promoter with Ets-1 (8). Based on our NMR results, we undertook a closer examination of the effect of phosphorylation of T220 on Ets-1 binding in GST-pulldown assays. Our results clearly show a two- to threefold lowering of the K_D for both the T220D mutant and the PKA phosphorylated wild type. Thus, regulated phosphorylation of Pit-1 represents an additional level of control on Pit-1/Ets-1 binding. Such an additional regulatory mechanism is very likely to affect subtle in vivo equilibria, which can easily go unnoticed in transient overexpression assays.

A second posttranslational regulation of Ets-1 binding affinity may occur at the hydrophobic pocket next to W261. The accessibility of this hydrophobic pocket seems to be modulated by the side chain of K226. In the Pit-1 dimer crystal structure, K226 is found in two orientations, one folding over the hydrophobic pocket and one pointing away from it (Fig. 4B). The presence of K226 is detrimental to the association of Ets-1, because truncating this side chain or switching its charge increases Ets-1 binding. These findings present an interesting opportunity for an additional level of control for Pit-1/Ets-1 complex formation, as Pit-1 has been shown to be acetylated by CREB-binding protein (CBP) in vitro (23) and both Pit-1 and Ets-1 are able to bind CBP/p300 (18, 24, 25). Based on our results, this acetylation could take place on K226, which would serve to increase association with Ets-1 by increasing accessibility of the hydrophobic pocket. By using our data, we can suggest a model for the association of Ets-1 with Pit-1 (Fig. 4C). Depending on the posttranslational modification state of the Pithd, Ets-1 will dock in the hydrophobic pocket next to W261 and contact the region around T220. Additional stability may be provided by residues at the C terminus of the DNA-binding helix and by the region around L288.

Several other transcription factors, such as Oct-1, vitamin D receptor, N-CoR, and GATA-2, interact with Pit-1 through the homeodomain (18, 26–28). Only in the case of the GATA-2 interaction, some of the essential residues in the Pithd have been identified by interaction assays. These residues included R215 and K216 at the N terminus and P239 and Q242 at the beginning of the second helix of the Pithd. None of the amide protons of these residues showed any significant change in chemical shift upon addition of Ets-1 (235–304), which indicates that GATA-2 employs a binding interface that is different from Ets-1. It would be intriguing to perform NMR studies similar to those we have described in this report on the interactions with these transcription factors and to compare them with the results for Ets-1. Apart from the high sensitivity, one of the advantages of our NMR approach over random mutagenesis or deletion experiments is that the entire protein can be probed for interaction in one experiment. This means that combinations of residues or structural motifs that may not be affected with a single point mutant can be identified. Even if the proteins of interest bind one another with moderate affinity, this approach can still produce valuable information on protein–protein interactions.

Abbreviations

PKA

protein kinase A

HSQC

heteronuclear single quantum coherence