Evaluation of the Interaction between Phosphohistidine Analogues and Phosphotyrosine Binding Domains

We have investigated the interaction of peptides containing phosphohistidine analogues and their homologues with the prototypical phosphotyrosine binding SH2 domain from the eukaryotic cell signalling protein Grb2 by using a combination of isothermal titration calorimetry and a fluorescence anisotropy competition assay. These investigations demonstrated that the triazole class of phosphohistidine analogues are capable of binding too, suggesting that phosphohistidine could potentially be detected by this class of proteins in vivo.

Tom E. McAllister, Katherine A. Horner, and Michael E. Webb* [a] We have investigated the interaction of peptides containing phosphohistidine analogues and their homologues with the prototypical phosphotyrosine binding SH2 domain from the eukaryotic cell signalling protein Grb2 by using a combination of isothermal titration calorimetry and a fluorescence anisotropy competition assay. These investigations demonstrated that the triazole class of phosphohistidine analogues are capable of binding too, suggesting that phosphohistidine could potentially be detected by this class of proteins in vivo.
Phosphorylation of proteins is well established as a dominant mechanism of signal transduction in eukaryotic cells. [1] Phosphorylation occurs in response to extracellular and intracellular signals, and the modified protein can then be sensed by sequence-specific modular binding domains. [2] These include SH2 and PTB domains, which bind phosphotyrosine, [3] and 14-3-3 and WW domains, which bind phosphothreonine and phosphoserine. [4] In contrast, it is still unknown how, or indeed whether, other phosphorylated amino acids such as phosphohistidine are sensed. The lability of this latter modification means that it has only been directly observed in a limited subset of eukaryotic proteins, including the cystic fibrosis transmembrane receptor (CFTR), [5] histone H4 [6] and the potassium-activated calcium channel KCa3.1. [7] In the first and last cases, histidine phosphorylation directly activates the channel; however, in the case of histone H4 and many others, the function has not been determined. Although proteomic methods [8] are increasingly able to identify this modification in proteins, no general function for it has been identified.
We recently reported the synthesis of a triazolyl analogue, 3 a, of t-phosphohistidine (1, Scheme 1), which is fully compatible with the Fmoc-protection strategy for solid-phase synthesis. [9] Independently, Kee et al. have demonstrated that, following Boc-peptide synthesis, it was possible to use analogue 2 as a hapten to identify antibodies capable of identifying phosphohistidine in a site-specific fashion. [10] More recently a trun-cated form of 2 has been used to generate a polyclonal antibody that is able to identify 1 in a variety of proteins. [11] The study of phosphoprotein interactions in vitro requires a source of phosphorylated protein. There are well known techniques for isolating phosphate ester-containing proteins, but maintaining phosphorylated histidine residues during protein isolation is notoriously difficult, [12] even though histidine can be chemically phosphorylated in vitro. [13] An alternative approach is the incorporation of phosphoamino acids (or precursors) into recombinant proteins. This approach was used by Serwa et al. to generate an analogue of phosphotyrosine 4 by Staudinger phosphite reaction on a genetically incorporated pazidophenylalanine to form phosphoramidate 5 (Scheme 2). [14] A further phosphotyrosine analogue, p-carboxymethylphenylalanine (6), [15] has also been incorporated by an amber-suppression approach; direct genetic incorporation of phosphoserine has also been achieved. [16] We hypothesised that we could generate a protein containing our phosphohistidine analogue by first producing a protein with an azidoalanine residue and subsequent reaction with a suitably protected alkynyl phosphonate 8. Unfortunately lazidoalanine cannot be incorporated by amber suppression and is only poorly incorporated as a methionine surrogate. [17] Scheme 1. Structures of t-phosphohistidine (1) and the corresponding triazole-phosphonate mimic, 2, which can be incorporated into peptides by using precursors 3 a and 3 b. Scheme 2. Structures of phosphotyrosine 4 and analogues 5 and 6, which have previously been incorporated into recombinant proteins.
[a] Dr. T However, the homologous compound, l-azidohomoalanine (7), is well accepted as a methionine surrogate [18] and reaction of this residue with alkynyl phosphonate 8 would ultimately produce homotriazole 10 (Scheme 3). Though this triazole could perhaps no longer mimic phosphohistidine, it occurred to us that 10 could present the phosphoryl group in a similar orientation to phosphotyrosine (due to the extra flexibility bestowed by the g-methylene group) and thus constitute a novel analogue of phosphotyrosine 4. Though not our primary goal, investigation of this seemed prudent as, if a suitable mimic, 10 would be a more accessible phosphotyrosine analogue than the current alternatives.
We envisaged producing suitable peptides by Fmoc-SPPS and investigating the interaction with proteins known to bind to phosphotyrosine. Our initial test system was the prototypical SH2 phosphotyrosine binding domain from Grb2. [19] This domain has been well characterised, binding with a high affinity (K d = 200 nm) to the Shc-derived phosphotyrosine-containing hexapeptide 11 (Scheme 4). [20] We generated the required Fmoc-/benzyl-protected amino acid 14 from Fmoc-protected homoazidoalanine 12 and dibenzyl phosphonylacetylene 13 through a [3+2] cycloaddition in an analogous manner to our previously reported synthesis of Fmoc-/benzyl-protected 3 a (Scheme 4). [9] Fmoc-SPPS was then used to generate peptides 11 and 15, which contain phosphotyrosine (pY) 4 and homotriazole (phTz) 10, respectively. We also investigated using an Fmoc-/ethyl-protected derivative of 14 but encountered deprotection difficulties (as reported previously [21] ), so this approach was not pursued further.
For our initial assays of ligand-protein interactions, we used full-length Grb2 that was purified as a glutathione S-transferase (GST) fusion. Titration of the pY control peptide 11 into the protein yielded a sigmoidal binding curve (see the Supporting Information) corresponding to an affinity of (196 AE 21) nm, consistent with literature reports. The titration of phTz peptide 15 into the protein was inconclusive; although the interaction was clearly weak it was not possible to rule out an interaction. Protein aggregation prevented experiments at higher concentrations, and attempts to cleave the GST-tag enzymatically were unsuccessful. To overcome this hurdle, the SH2 domain was subcloned into pET28a to obtain an N-terminally Histagged isolated SH2 domain. The protein overexpressed well in the insoluble fraction, which was solubilised in 8 m urea and purified by immobilised metal-affinity chromatography. Oncolumn refolding then yielded pure soluble protein, but sizeexclusion analysis of this protein revealed two distinct non-interconverting species with identical appearance when analysed by SDS-PAGE ( Figure S4 in the Supporting Information). ITC assay of peptide 11 binding to the two protein fractions revealed that the larger (earlier-eluting) species had a significantly reduced affinity (4.8 AE 1.2) mm versus (344 AE 42) nm for the smaller species ( Figure S8). These data suggested that the early peak represents a domain-swapped dimer (as previously observed by Benfield et al.) [22] and so the later peak, corresponding to monomeric His 6 -(Grb2-SH2) was used in all subsequent assays, this allowed experiments to be conducted at significantly higher (more than tenfold) protein concentrations.
The interaction of phTz peptide 15 with Grb2-SH2 was analysed by ITC. Initial experiments under our standard conditions (final ratio of ligand to protein 2:1) were inconclusive, thus suggesting a possible poorly defined weak binding (not shown). The experiment was therefore repeated under low cvalue conditions [23] with a final ratio of ligand to protein of 8:1 (Figure 1 B). This titration was indistinguishable from a titration of the peptide into buffer alone, conclusively confirming no binding. This was somewhat disappointing and surprising given that recent work by Hofmann et al. revealed that the Src kinase-SH2 domain can bind to phosphoarginine. [24] However, this led us to consider that the Grb2-SH2 domain might bind to other proteinogenic phosphoamino acids. To determine if tphosphohistidine 1 would bind, we synthesised a further peptide, 16, that contained our phosphotriazole (pTz) analogue 2. Again, an initial ITC experiment under standard conditions was inconclusive (data not shown), but titration to a final ligand to protein ratio of 8:1 (Figure 1 C) yielded a clear binding interaction. This was readily distinguishable from the buffer dilution and corresponded to 1:1 binding with K d = (719 AE 28) mm (Figure 1). It has been shown that a 20-fold change in affinity can be attributed to a change in protonation state. [25] The relatively low affinity of the peptide 16:Grb2-SH2 interaction ob- served in our study might be due to a difference in protonation states between the phosphate of 4 and the phosphonate of 2. To investigate this, we conducted NMR titrations (see the Supporting Information) that revealed the pK a of 4 to be 5.8 and 2 to be 5.95. Hence, under the conditions of our experiments (pH 7.4) both phosphoryl groups should be di-anionic, so it is unlikely that the difference in affinity is due to differing protonation states.
To investigate the binding of other phosphoamino acids and confirm that peptides 11 and 16 were binding at the same site, we developed a competition fluorescence anisotropy assay. Titration of a phosphotyrosinecontaining peptide with an Nterminal fluorescein, 17, with purified His 6 -(Grb2-SH2) (see Figure S11) yielded a binding constant of (283 AE 18) nm; this is consistent with the K d previously measured by ITC. We then conducted competition experiments on all three peptides used for ITC together with peptides 18-21 (AcHN-Ser-Xaa-Val-Asn-Val-Gln-NH 2 Xaa = Tyr, His, pSer or pThr, see the Supporting Information). Binding (displacement of 17) was observed only for peptides 11 and 16, which contain pTyr and pTz, respectively (Figure 1 D). A slightly lower than expected IC 50 of 360 mm was observed for the pTz peptide; this suggests that the ITC analysis might have underestimated the affinity of this peptide. No evidence for binding was observed for the other peptides (Figure S11); this suggested that the Grb2-SH2 domain will not simply bind to a phosphoryl group presented in the correct peptide context, or to an aromatic ring.
We have shown that a proven t-phosphohistidine mimic is capable of sequence-specific binding to a canonical phosphotyrosine-binding domain. Phosphotyrosine antibodies have previously been shown to bind to phosphohistidine, [8b] but our findings contradict previous studies of phosphohistidine-containing peptides binding to SH2 domains (however the conditions used in these studies precluded the identification of lowaffinity binding such as we have observed and used a different SH2 domain, from phospholipase C-g1). [26] Phosphotyrosine binding protein modules are a common feature of human signalling proteins and, in many cases, the sequence specificity of these domains has not been fully elucidated. The affinities of known peptides for their target proteins also vary hugely, ranging from low nanomolar to high micromolar. [27] Low affinity therefore does not mean low significance; it is possible that a subset of assigned phosphotyrosine binding modules could additionally/alternatively interact with phosphohistidine-con- taining proteins. Further study to establish the generality of this observation is on-going.

Experimental Section
Experimental details are given in the Supporting Information.