Structural Requirements for Signal Transducer and Activator of Transcription 3 Binding to Phosphotyrosine Ligands Containing the Y XX Q Motif*

, Stat3 is an Src homology (SH)2-containing protein constitutively activated in a wide variety of human cancers following its recruitment to Y XX Q-containing mo-tifs, which results in resistance to apoptosis. Despite resolution of the crystal structure of Stat3 homodimer bound to DNA, the structural basis for the unique specificity of Stat3 SH2 for Y XX Q-containing phosphopeptides remains unresolved. We tested three models of this interaction based on computational analysis of available structures and sequence alignments, two of which assumed an extended peptide configuration and one in which the peptide had a (cid:1) -turn. By using peptide immunoblot affinity assays and mirror resonance affinity analysis, we demonstrated that only phosphotyrosine (Tyr(P)) peptides containing (cid:2) 3 Gln (not Leu, Met, Glu, or Arg) bound to wild type Stat3. Examination of a series of wild type and mutant Stat3 proteins demonstrated loss of binding to pY XX Q-containing peptides only in Stat3 mutated at Lys-591 or Arg-609, whose side chains interact with the Tyr(P) residue, and Stat3 mutated at Glu-638, whose amide hydrogen bonds with oxygen within the (cid:2) 3 Gln

Stat3 is an Src homology (SH)2-containing protein constitutively activated in a wide variety of human cancers following its recruitment to YXXQ-containing motifs, which results in resistance to apoptosis. Despite resolution of the crystal structure of Stat3 homodimer bound to DNA, the structural basis for the unique specificity of Stat3 SH2 for YXXQ-containing phosphopeptides remains unresolved. We tested three models of this interaction based on computational analysis of available structures and sequence alignments, two of which assumed an extended peptide configuration and one in which the peptide had a ␤-turn. By using peptide immunoblot affinity assays and mirror resonance affinity analysis, we demonstrated that only phosphotyrosine (Tyr(P)) peptides containing ؉3 Gln (not Leu, Met, Glu, or Arg) bound to wild type Stat3. Examination of a series of wild type and mutant Stat3 proteins demonstrated loss of binding to pYXXQ-containing peptides only in Stat3 mutated at Lys-591 or Arg-609, whose side chains interact with the Tyr(P) residue, and Stat3 mutated at Glu-638, whose amide hydrogen bonds with oxygen within the ؉3 Gln side chain when the peptide ligand assumes a ␤-turn. These findings support a model for Stat3 SH2 interactions that could form the basis for anticancer drugs that specifically target Stat3.
Signal transducer and activator of transcription 3 is a latent transcription factor activated by cytokine and growth factor receptors including interleukin-6 and EGFR 1 (1)(2)(3). Stat3 is recruited to the cytoplasmic domain of receptors via its SH2 domain and phosphorylated on tyrosine 705 by either intrinsic or receptor-associated tyrosine kinases, most notably members of the Janus family. Phosphorylation of Stat3 leads to dimerization mediated by reciprocal SH2-Tyr(P)-705 motif interactions, followed by nuclear translocation, binding to specific DNA elements, and up-regulation of target genes.
Stat3 has been demonstrated to be required for transformation of fibroblasts by v-Src (4,5) and for autocrine growth of squamous cell carcinoma of the head and neck (6) where it is activated by an autocrine loop involving TGF-␤ and EGFR (7). Expression of a constitutively activated form of Stat3 alone in fibroblasts was oncogenic (8). Constitutive activation of Stat3 occurs in a wide variety of cancers in addition to squamous cell carcinoma of the head and neck including breast, prostate, renal cell, melanoma, ovarian, lung, leukemia, lymphoma, and multiple myeloma (9) as a result of autocrine or paracrine activation of the EGFR and the interleukin-6 receptor or secondary to as yet unidentified mechanisms.
EGFR contains an extracellular ligand-binding domain, a single transmembrane region, and an intracellular domain harboring intrinsic tyrosine kinase activity (10). Ligand-induced dimerization of EGFR allows reciprocal transphosphorylation of residues within the catalytic domain of the kinase leading to its enzymatic activation and autophosphorylation of C-terminal cytoplasmic tyrosine residues. Five autophosphorylation sites have been identified in EGFR as follows: Tyr-992, Tyr-1068, Tyr-1086, Tyr-1148, and Tyr-1173 (11,12). These phosphorylated tyrosine residues serve as docking sites for signal proteins containing Src homology (SH2) domains, including phospholipase C-␥ (13,14), Grb-2 (15,16), Shc (17), SHP-1 (18), and most recently Stat3 (19), which was shown by us to bind to EGFR at Tyr(P) sites located at Tyr-1068 and Tyr-1086. Both of these tyrosine residues are followed at the Tyr(P) ϩ3 position by Gln thereby conforming to the consensus Stat3 SH2-binding motif, YXXQ (20,21). The preference of Stat3 SH2 for Tyr(P) peptide ligands containing Gln (or the polar residues Thr or Cys) at the ϩ3 position is unique among SH2 domains. The structural basis for this is unknown but could be exploited to specifically target Stat3 activation in cancers such as squamous cell carcinoma of the head and neck in which Stat3 activation occurs downstream of activated EGFR.
Although the structure of Stat3 SH2 bound to Tyr(P) ligand has not been solved, the structure of Stat3␤ bound to DNA has been encompassing the domains of Stat3␤ from residues 127 to 722 including the SH2 domain (22). The authors concluded that Stat3 SH2 shares structural features of other SH2 domains having a central, three-stranded anti-parallel ␤-pleated sheet (strands B-D) flanked by helix ␣A and strands ␤A and ␤G. However, since the electron density was not well defined for the SH2 domain and the Tyr(P)705-containing phosphopeptide region, the structure obtained did not clarify the preference of Stat3 SH2 for binding to phosphopeptide ligands with Tyr(P) ϩ3 Gln (or ϩ3 Thr because Thr-708 is located at the ϩ3 position downstream of Tyr(P)-705). Two models have been proposed to explain this preference (23,24); both assume an extended configuration for the Tyr(P) peptide ligand and two pockets as follows: one, a positively charged pocket that interacts with the Tyr(P) residue; and the other, a hydrophilic pocket that interacts with the ϩ3 Gln; but neither model has been tested and verified.
By using wild type and mutated EGFR Tyr-1068 PDPs in peptide immunoblot and mirror resonance affinity analyses, we demonstrated the following. 1) Tyr(P) binding requires interaction of the phosphate group with the side chains of Lys-591 and Arg-609 within the Stat3 SH2. 2) The ϩ3 Gln is required for binding of Stat3 to pYXXQ-containing peptides. 3) Binding of Stat3 SH2 to pYXXQ-containing peptides does not require the side chains of Glu-638, Tyr-640, and Tyr-657 or Tyr-657, Cys-687, Ser-691, and Glu-692 proposed to form pocket 2 in the Chakraborty et al. (24) and Hemmann et al. (23) models, respectively. Rather, our affinity analysis coupled with computer modeling supports a model in which the Tyr(P) ligand has a ␤-turn and the oxygen on the side chain of the ϩ3 Gln forms a bond with the amide hydrogen within the peptide backbone of Stat3 at Glu-638. These findings have important implications for design of peptidomimetics to specifically target Stat3 recruitment and activation in cancer cells.

EXPERIMENT PROCEDURES
Site-directed Mutagenesis of Stat3 and EGFR-Human Stat3␣ cDNA was a gift from Dr. Rolf Van de Groot (25). A HindIII/XhoI DNA fragment containing Stat3␣ was cloned into the baculovirus expression vector, pFastBac1 (Invitrogen), with a 6-histidine tag engineered onto the N terminus of human Stat3. Single or combination mutations were generated by using Quikchange site-directed mutagenesis kit (Stratagene) to target amino acid residues within the Stat3 SH2 domain implicated in models of Stat3 SH2-phosphotyrosine binding (K591L, R609L, E638P, E638L, Y640F, Y657F, C687A, S691A, and Q692L; Fig.  1). The sequence of each constructs was verified by sequencing analysis.
Peptide Synthesis-The peptides listed in Table I were synthesized in the Baylor College of Medicine Protein Core Facility on an Applied Biosystems (Foster City, CA) model 433A peptide synthesizer using standard 9-fluorenylmethoxycarbonyl amino acid chemistry. Seventy percent of the peptide reaction mix was biotinylated at the N terminus while the peptide remained on the resin using d-Biotin-LC (AnaSpec, Inc.). All peptides were purified using reverse-phase high performance liquid chromatography and were Ն95% pure.
Mirror Resonance Affinity Assay-Kinetics experiments were performed using an Iasys Autoϩ resonant mirror biosensor (Affinity Sen-sor, Paramus, NJ) as described (26). Briefly, two-welled cuvettes coated on the bottom of each well with biotin were purchased from Affinity Sensor and prepared for immobilization of biotinylated peptides by coating each surface with 0.04 mg/ml NeutrAvidin (Pierce) and washing with PBS-T (20 mM sodium phosphate, 0.05% Tween 20). Biotinylated peptide (5 g) was added into each well, experimental peptide to one well and control peptide to the other, and change in arc seconds was monitored simultaneously in both wells by using the biosensor until stable followed by washing with PBS-T. Real time binding of Stat3 was conducted at 25°C at a stir speed of 70 for 10 min starting at the lowest concentration of Stat3. The wells were washed out with three changes of 60 l of PBS-T, and dissociation was allowed to proceed for 5 min. Each well bottom was regenerated by washing with 50 l of 100 mM formic acid for 2 min and equilibrated with PBS-T for the next round of association assay. Data were collected automatically and analyzed with the FASTplot and GraFit software (27).
CD-CD spectra of the wild type and E638P mutants of Stat3 were recorded between the 280 and 190 nm range in 10 mM phosphatebuffered saline on an Olis DSM 1000 CD spectrophotometer. Measurements were performed at a protein concentration of 1.8 and 1.6 M for the wild type and mutant Stat3, respectively, using a 1-mm cuvette. Spectra were acquired at 10°C with a 2-s integration time and repeated three times for each sample.

Requirement for ϩ3 Glu within the Tyr-1068 Phosphopeptide
Ligand for Stat3 Binding-Our previous studies indicated that sites of autophosphorylation within the C terminus of EGFR at Tyr-1068 and Tyr-1086, which are each followed at the Tyr(P) ϩ3 position by Gln, mediated the recruitment of Stat3 leading to its activation (19). Peptide affinity immunoblot analysis and mirror resonance imaging studies using phosphorylated and non-phosphorylated dodecapeptides based on the amino acid sequence within the region of the EGFR containing Tyr-1068 and Tyr-1086 demonstrated the requirement for their phosphorylation on tyrosine to achieve measurable binding of native and recombinant Stat3. These studies also revealed that Tyr-1068 phosphododecapeptide bound with 2-fold higher affinity than Tyr-1086 PDP.
To determine whether or not the polar residue Gln at the ϩ3 position of Tyr(P)-1068 peptide is essential for Stat3 SH2 binding, we synthesized a panel of tyrosine-phosphorylated dodecapeptides based on Tyr-1068 in which ϩ3 Gln was left unchanged or replaced by a residue with a non-polar side chain Leu or Met, an acidic side chain Glu, or a basic side chain Arg (Table I). Each peptide was incubated with equal amounts of purified wild type Stat3 protein in peptide pull-down assays (Fig. 1). Immunoblotting for Stat3 demonstrated a prominent Stat3 band in pull-down assays using wild type Tyr-1068 PDP. In contrast, little to no Stat3 was detected in pull-down assays using PDPs in which the Gln was mutated to Leu, Met, Glu, or Arg similar to results using unphosphorylated Tyr-1068 dodecapeptide. Thus, Gln at the ϩ3 position of Tyr-1068 phosphopeptide is required for Stat3 binding and appears to be as important for Stat3 binding as phosphorylation on tyrosine.  (24) proposed previously two distinct but overlapping two-pocket models for the binding of YXXQ-containing PDP ligands by the Stat3 SH2 domain; both models assumed the peptide ligand was in an extended configuration (Fig. 2, A and B). In our model, the phosphotyrosine residue interacts with a positively charged pocket (pocket 1) within the SH2 domain formed primarily by the side chains of Lys-591 and Arg-609 and secondarily by the side chains of Ser-611, Glu-612, and Ser-613. The Tyr(P) ϩ3 Gln was predicted to interact with a hydrophilic pocket (pocket 2) formed by the side chains of Glu-638, Tyr-640, and Tyr-657. In the Hemmann model, the phosphotyrosine was predicted to interact with the side chain of Arg-609 (pocket 1) and the ϩ3 Gln with the side chains of Tyr-657, Cys-687, Ser-691, and Glu-692 (pocket 2).
In order to test each of the two models proposed, we generated Stat3 mutants in which mutations were introduced to change charged or polar side chains to non-polar within amino acid residues predicted in each model to be critical for Stat3 binding (Fig. 2C). His tags were added at the N terminus of each protein to aid in purification; we demonstrated previously that this modification did not interfere with binding of wild type Stat3 to native full-length, activated EGFR, or to PDPs Tyr-1068 and Tyr-1086. The recombinant Stat3 proteins were expressed in Sf9 insect cells and purified to equivalent levels using Ni-NTA resin (Fig. 2D).
Peptide affinity immunoblot studies using Stat3-3M to test the pocket 2 component of the Chakraborty model demonstrated levels of Stat3-3M bound to Tyr-1068 and Tyr-1086 PDPs similar to wild type Stat3 (Fig. 3A). Peptide affinity immunoblot studies using Stat3-4M to test the pocket 2 component of the Hemmann model demonstrated levels of binding of Stat3-4M bound to Tyr-1068 and Tyr-1086 phosphopeptides equal to or greater than wild type Stat3 (Fig. 3A). Stat3-6M, in which all six amino acid residues predicted by both models to form pocket 2 were mutated, also bound both PDPs at levels similar to wild type Stat3 as did Stat3-2M and Stat3-3MϩC687A. These results do not support either model for Stat3 SH2 binding to ϩ3 Glu within phosphopeptide ligands.
To test the pocket 1 component of the two models and to ensure that our peptide pull-down system was sufficiently sensitive to detect reduced binding of Stat3 containing mutations in pocket 2, we added either K591L or R609L to the 3M mutant to generate Stat3-3MϩK591L and Stat3-3MϩR609L. Addition of either mutation resulted in elimination of binding to both Tyr-1068 and Tyr-1086 PDPs indicating that each of the side chains of Lys-591 and Arg-609 make important contributions to binding of the phosphotyrosine.
To confirm these findings and to determine whether introduction of the pocket 2 mutations resulted in subtle alterations in kinetics of binding undetectable using phosphopeptide affinity immunoblot analysis, we performed mirror resonance affinity assays using phosphorylated and non-phosphorylated Tyr-1068 dodecapeptide (Fig. 3, B and C, and Table II). Review of the real time mirror resonance affinity curves (Fig. 3, B and C) and kinetic analysis (Table II)  Computational Modeling of Stat3 SH2 Binding to ϩ3 Glu within YXXQ-containing Phosphopeptides-To generate a new and more accurate model for Stat3 SH2 binding to ϩ3 Gln, we used the structure of Tyr-1068 phosphopeptide (EpYINQ), available from its crystal structure bound by Grb2 (28) (Protein Data Bank code 1ZFP), and the structure of Stat3 from Trp-580 to Leu-670, obtained from the crystal structure of Stat3␤ bound to DNA (22) (Protein Data Bank code 1BG1), to computationally model the interaction with the lowest energy. All energy minimization calculations were carried out under AMBER force field by using the DISCOVER/Insight II program. A total of 300 steps of conjugate gradient energy minimization was performed following rigid hand-docking to fit the Tyr(P) of the EpYINQ peptide into the binding pocket composed of residues Lys-591 and Arg-609 by taking into consideration Van der Waals and Coulomb forces. The interaction between Stat3-SH2 and EpYINQ with the complex lowest energy (Fig. 4A) had a total binding energy of Ϫ478.8 kcal/mol. This computational result predicted that the major binding energy for this binding configuration comes from a hydrogen bond interaction involving oxygen within the Tyr(P) ϩ3 Gln side chain and the peptide amide hydrogen at Glu-638 located within a loop region of Stat3 SH2.
To test the contribution of the Glu-638 amide hydrogen, we generated Stat3-E638P by site-directed mutagenesis, which eliminated the amide hydrogen donor predicted to bind with oxygen within the ϩ3 Gln side chain. In consideration of the possible effect of this mutation on secondary structure, we modeled E638P within Stat3-SH2 using Biopolymer in the Insight II environment and carried out local energy minimization as follows: 1) with all residues fixed except for Val-637 to Pro-639 to assess the effect in the immediate vicinity of the E638P mutation; and 2) will all residues fixed except for residues from Ile-628 to Met-648 to assess the effect of E638P on structure further N-and C-terminal to E639P. When the resultant structures were overlaid onto the wild type Stat3, there was no physical differences between the two structures with the exception of a slight reduction in the angle of the backbone loop turn at residues Val-637, E638P, and Pro-639 (Fig. 4B). Recombinant Stat3-E638P was produced in Sf9 cells. It was expressed to levels similar to wild type Stat3 (Fig. 5A) and demonstrated solubility characteristics similar to wild type Stat3. Furthermore, CD analysis of Stat3-E638P (Fig. 5B) revealed a folded protein with a predominantly ␣-helical structure essentially identical to wild type Stat3 confirming and strengthening the conclusions reached from computational modeling that introduction of the E639P mutation does not FIG. 1. Gln at the ؉3 position within Tyr-1068 peptide is required for Stat3 SH2 binding of peptide. NeutrAvidin-agarose was incubated with the indicated biotinylated peptides (see Table I for sequence) or no peptide (CON) as control, washed thoroughly, and mixed with identical amount of wild type Stat3. Bound proteins were separated by SDS-PAGE and immunoblotted using Stat3 mAb. Lane ST represents purified wild type Stat3 (0.6 g) loaded directly onto the gel as positive control. result in unanticipated local or global secondary structural changes.
Peptide affinity immunoblot assays demonstrated no binding of Stat3-E638P to any of the EGFR-derived peptides tested including Tyr-1068 and Tyr-1086 PDP (Fig. 5C); mirror resonance affinity studies (Fig. 3C) confirmed these findings. These results strongly supported an important role for the Glu-638 amide hydrogen of Stat3 in binding of the ϩ3 Gln within Tyr-1068 PDP. DISCUSSION To understand the structural basis for Stat3 SH2 binding preference for YXXQ-containing Tyr(P) ligands, we performed peptide immunoblot and mirror resonance affinity measurements using a series of mutated phosphopeptides and Stat3 proteins. Mutations of Gln within Tyr-1068 PDP to Leu, Met, Glu, or Arg eliminated binding of wild type Stat3 confirming the importance of Gln at the ϩ3 position. Two models for Stat3 SH2 binding to YXXQ-containing ligands were initially examined, and each assumed an extended phosphopeptide configuration and proposed that the Tyr(P) residue interacts at one site (pocket 1) and the ϩ3 Gln interacts at another site (pocket 2) formed by key residue side chains. Our mutational analysis revealed that although mutations in pocket 1 (K591L and R609L) eliminated binding to Tyr-1068 and Tyr-1086 PDP, mutation of all the residues proposed to form pocket 2 had no effect on binding, indicating that pocket 1 is required for YXXQ PDP but leaving unresolved the structural basis for specificity of Stat3 SH2 binding of PDP with ϩ3 Gln. Computational analysis using the known structures of Tyr-1068 phosphopeptide (EpYINQ), which contains a ␤-turn, and Stat3␤ suggested an alternative model for ϩGln binding in which the oxygen on the side chain of the Tyr(P) ϩ3 Gln forms a bond with the amide hydrogen within the peptide backbone of Stat3 at Glu-638. To test this model, we generated His-Stat3 containing mutation E638P (Stat3-E638P), which eliminated the donor hydrogen. Stat3-E638P demonstrated undetectable binding to Tyr-1068 and Tyr-1086 PDP in peptide pull down and mirror resonance affinity analyses. Most important, computational modeling and CD analysis did not reveal differences between Stat3-E638P and wild type Stat3 in either local or global secondary structure. Thus, our findings support a model of Stat3 SH2 binding to YXXQ in which the phosphopeptide ligand contains a ␤-turn, the side chains of Lys-591 and Arg-609 are required for Tyr(P) binding, and the amide hydrogen of Glu-638 is required for binding of the ϩ3 Gln. The ␤-turn conformation of the phosphopeptide ligand and the structural basis for the ϩ3 Gln interaction could be exploited to target Stat3-Tyr(P) ligand interactions and form the basis for drugs that specifically target Stat3 activity for cancer therapy.
SH2 domains are structurally conserved protein modules of about 100 amino acid residues in length first identified as non-catalytic regions of homology within Src and Fps kinases (29) et al. (30)). These structures revealed a globular protein with similar topology. The spine of the domain is an anti-parallel ␤-sheet formed by strands A-D and G, which divides the domain into two functionally distinct sides. One side, flanked by helix ␣A, is concerned primarily with binding Tyr(P), and the other side, flanked by helix ␣B and the EF and BG loops, provides residues that interact with the side chains of the peptide that are Cterminal to the Tyr(P) (31).
The binding site topology of SH2 domains excludes the binding of phosphoserine and phosphothreonine residues. The elements of SH2 involved in Tyr(P) recognition are provided by ␣A and ␤B, notably Arg or Lys at position ␣A2 and Arg at position ␤B5, along with ␤D and the loop connecting strands B and C. In Stat3, the Lys-591 aligns at the ␣A2 position and Arg-609 at the ␤B5 position. Our affinity studies demonstrated elimination of binding of Tyr-1069 PDP by recombinant Stat3 containing K591L or R609L mutations confirming the contribution of the side chains of Lys-591 and Arg-609 to binding Tyr(P) within the phosphopeptide ligand.
Amino acids C-terminal to Tyr(P) form the basis of specific SH2-Tyr(P) peptide recognition and especially depend on the nature of residue ϩ2 or ϩ3 relative to the Tyr(P) residue (32,33). Phosphopeptides with specificity at ϩ3 tend to interact in  Table I for sequence) or no peptide (CON) as control, washed thoroughly, and mixed with identical amounts of wild type or mutant Stat3 proteins as indicated. Bound proteins were separated by SDS-PAGE and immunoblotted using Stat3 mAb. Lane ST represents purified wild type Stat3 (0.6 g) loaded directly onto the gel as positive control. B and C, mirror resonance affinity assay. Two cells of a biotin-coated cuvette pretreated with saturating amounts of NeutrAvidin. One well of the cuvette was pretreated with biotinylated phosphopeptide based on Tyr-1068 (pY1068, left panel), whereas the other well was pretreated with biotinylated non-phosphorylated peptide Tyr-1068 (Y1068, right panel) as a control for nonspecific binding. Wild type or mutated Stat3 protein was added in the concentrations indicated to each of the two cells, and mirror resonance measurements were recorded continuously for 10 min as shown.
an extended conformation with the surface of the SH2 domain. The prototype of this interaction is that between the Src family SH2 domains and peptides containing the optimal pYEEI motif (34), which resembles a two-pronged plug (the peptide) engaging a two-holed socket (the SH2 domain). Phosphopeptides with specificity at ϩ2 adopt a ␤-turn conformation, which is stabilized by a hydrogen bond between the carbonyl oxygen of the phosphotyrosine and the backbone amide of the peptide residue at ϩ3. The prototype of this interaction is that between the Grb2 SH2 domain and peptides containing the Grb2 consensus motif pYXNX. The structural explanation for the peptide assuming a ␤-turn is Trp at residue 121 in the 1 position of the EF loop of Grb2, which closes off the binding site for the ϩ3 residue (35,36). The binding preference of Grb2 for peptides with a ␤-turn is also supported by results of peptide library screening, which demonstrated a consensus sequence of peptides bound by Grb2 of pY(L/V)N(V/P) (36,37); Val and Pro at the ϩ3 location promote class VIII and class VIb ␤-turns (38).
Review of the model of Stat3 SH2 bound to peptide EpYINQ (Fig. 4A) reveals that Trp-623 occupies a position in Stat3 SH2 that may serve to block binding of YXXQ peptide in the extended conformation and force a ␤-turn similar to Trp-121 in Grb2. Of note, two groups (39,40) have demonstrated recently preferential binding of Stat3 to phosphopeptide ligands with Pro in the ϩ2 position; Pro in the ϩ2 position favors formation of class VIa1, VIa2, and VIb ␤-turns (38). In addition, both Grb2 and Stat3 bind to EGFR at Tyr-1068 located within a site (YINQ) that contains a consensus-binding motif for both Grb2 (Asn at Tyr(P)ϩ2) and Stat3 (Gln at Tyr(P)ϩ3). Structural data of Tyr-1068 phosphopeptide crystallized with Grb2 strongly suggest that the region within the EGFR at Tyr-1068 forms a ␤-turn starting at the Tyr(P) residue. Primary sequence analysis Tyr-1068 PDP using the program PROF (41) and AGADIR (42) in the absence of phosphorylation does not suggest the presence of a ␤-turn; however, mutating the Tyr to Glu to mimic the negative charge of Tyr(P) enhances turn propensity.  4. Revised model of Stat3 SH2 binding to ؉3 Gln within YXXQ-containing phosphopeptide ligands. A, computational modeling using the Biopolymer program in the Insight II environment was used to perform local energy optimization of the interaction of Stat3 SH2 (shown as a gray ribbon) with phosphopeptide ligand (EpYINQ shown as a green ribbon) based upon the known structures of each. As indicated, the oxygen on the side chain of the Tyr(P) ϩ3 Gln within the EpYINQ peptide is predicted to form a hydrogen (H-bond) bond with the amide hydrogen at Glu-638 and to make a major contribution to the binding energy. The positions are shown for the side chains of Lys-591 and Arg-609 proposed to be major contributors to pocket 1, Glu-638, Tyr-640, and Tyr-657 proposed by Chakraborty to form pocket 2, and for the side chain of Trp-623 proposed to force a ␤-turn in the peptide ligand. The ϩ3 Gln and Glu-638 are shown as ball-and-stick models, the remaining side chains as stick models; oxygen atoms are shown in red, carbon in gray, nitrogen in blue, and phosphorus in orange. B, overlay of the known structure of wild type Stat3 (green) with the predicted structure of Stat3-E638P (gray). The positions of the side chains of relevant residues are indicated for wild type Stat3 (aqua stick models) and for Stat3E638 (gray stick models).  Table I for sequence) or no peptide (CON) as control, washed thoroughly, and mixed with wild type Stat3 protein. Bound proteins were separated by SDS-PAGE and immunoblotted using Stat3 mAb. Lane ST represents purified wild type Stat3 (0.6 g) loaded directly onto the gel as a positive control.
Further confirmation and clarification of our model await higher resolution structural information about Stat3 SH2 bound to Tyr(P) peptide ligand.
Strategies to target Stat3 by identifying phosphopeptide inhibitors of Stat3 SH2-Tyr(P) ligand binding have been pursued by several groups including our own (19,39,40,43). We demonstrated previously that Tyr-1068 PDP, in addition to directly binding non-phosphorylated Stat3 and inhibiting DNA binding of activated Stat3, inhibited ligand-stimulated Stat3 activation and TGF-␣/EGFR-mediated autocrine growth when delivered into cancer cells (19). Studies are underway to determine whether cyclic Tyr(P) peptides that incorporate a ␤-turn and that strengthen the hydrogen bond interaction with the Stat3 Glu-638 amide hydrogen will more efficiently bind Stat3, destabilize Stat3 dimers, and when delivered into cells, block Stat3 activation and Stat3-mediated cell growth and resistance to apoptosis.
Turkson et al. (43) demonstrated that Tyr(P) peptides based on the sequence PYLKTK surrounding Tyr-705 within Stat3 inhibited Stat3 DNA binding and pulled down Stat3 from lysates of unstimulated cells. Furthermore, addition of a membrane translocation sequence to the C terminus of PYLKTK revealed the ability of this peptide to inhibit Stat3 activation within cells and suppress v-Src-mediated transformation. Alanine-scanning mutagenesis of the peptide mapped the essential sequence for inhibition of Stat3 DNA binding to the tri-peptide sequence PpYL; the tri-peptide demonstrated potency in inhibiting Stat3 DNA binding equivalent to full-length peptide. However, the tri-peptide was not shown to pull down Stat3. We and others (19,44) have shown that Tyr(P) peptides based on Tyr-992 within the EGFR are able to inhibit Stat3 DNA binding; however, we could not demonstrate direct interaction of Tyr(P)-992 peptide with Stat3 monomer either in peptide immunoblot or mirror resonance affinity assays (19). Thus, we concluded that Tyr-992 PDP inhibited DNA binding through an alternative mechanism. This conclusion may also apply to the ability to inhibit Stat3 DNA binding observed for the PYL tri-peptide and perhaps PYLKTK, as well.
Ren et al. (39) examined a series of Tyr(P)-containing peptides shown to bind Stat3 for the ability to inhibit DNA binding of recombinant Stat3. The peptide sequences were located within Stat3 itself or within receptors shown to activate Stat3 (interleukin-6 receptor, leukemia inhibitory factor receptor, EGFR, murine interleukin-10 receptor and granulocyte colonystimulating factor receptor). These studies revealed a 1,000fold range of activity and that Tyr(P) peptide based on Gp130 Tyr-904 (pYLPQTV) demonstrated the greatest activity (IC 50 ϭ 0.15 M). The primary mode of inhibition of Stat3 DNA binding by pYLPQTV is likely destabilization of homodimers through competition with the Tyr-705 region for binding to Stat3 SH2. However, in the absence of evidence for direct binding of Stat3 by this peptide, this issue, as for PYL tri-peptide, remains unresolved.
Wiederkerhr-Adams et al. (40) performed phosphopeptide library screening with recombinant Stat3 followed by confirmation of selected peptides by using an indirect surface plasmon resonance inhibition assay. These studies revealed that Stat3 binds peptides with the core motif Tyr(P)-basic/hydrophobic-Pro/ basic-Gln. However, similar to Ren et al. (39), Wiederkehr-Adams et al. (40) have not examined the ability of their lead peptides to inhibit Stat3 within cells or the consequences of this inhibition on cell growth or resistance to apoptosis.