Recombinant human heat shock protein 72 was either expressed in E.coli and purified as described previously [29Jindal S, Murray P, Rosenberg S, Young R, Williams KP. Human stress protein hsp 70: Overexpression in E.coli, purification and characterization Biotechnology 1995; 13: 1105-9.] or purchased from Sigma (H7283). DnaK was obtained from Stressgen (SPP-630). Labeled and unlabeled synthetic peptides were synthesized using conventional Fmoc chemistry (Applied Biosystems Corp. and Anaspec Inc.). Prior to deprotection the resin for each peptide was divided into two portions. One portion of the resin was deprotected and cleaved as normal. The remainder of the resin was reacted with carboxyfluorescein succinimidyl ester (Invitogen) using a 20-fold molar excess. Unreacted NHS-fluorescein was washed away and the labeled peptide deprotected and cleaved as normal. Masses of peptides were confirmed by mass spectrum and HPLC. A model unfolded polypeptide, reduced and carboxymethylated lactabumin (RCMLA; Sigma) was labeled in solution with NHS-fluorescein (Invitrogen) as described previously [29Jindal S, Murray P, Rosenberg S, Young R, Williams KP. Human stress protein hsp 70: Overexpression in E.coli, purification and characterization Biotechnology 1995; 13: 1105-9.]. MALDI-TOF MS analysis of the fluorescein labeled RCMLA indicated that there were on average three fluorescein groups per RCMLA molecule (data not shown).

All assay and fluorescence polarization (FP) measurements were performed at room temperature and with a 10 min incubation. The assay buffer used consisted of 25 mM HEPES pH 7.2, 150 mM KCl. Binding was assessed by incubating increasing concentrations of a Hsp70 protein with a fixed concentration of fluorescently-labeled substrate and FP values measured. FP measurements were performed either on a Beacon Fluorescence Polarization system (Invitrogen) or PHERAstar microreader (BMG Labtech) using the FP filter set (emission 450 and excitation 520 nm). FP values are expressed in millipolarization (mP) units. All statistical analyses were performed with GraphPad Prism 5.0 software. Binding data was analyzed using a non-linear regression analysis method (single site binding model) in Prizm 5.0.

We are interested in developing assays for a range of Hsp70 protein family members. As the best characterized, we began with mammalian Hsp72 and bacterial DnaK. Fluorescence polarization (FP) was chosen as the assay format as it has been generally applicable to measuring protein-protein or protein-ligand interactions [33Jameson D, Croney J. Fluorescence polarization: Past, present and future Comb Chem High Through Screen 2003; 6: 167-76.] including other heat shock proteins [31Kang Y, Taldone T, Clement C, et al. Design of a fluorescence polarization assay platform for the study of human Hsp70 Bioorg Med Chem Lett 2008; 18: 3749-51., 34Du Y, Moulick K, Rodina A, et al. High-throughput screening fluorescence Polarization assay for tumor-specific Hsp90 J Biomol Screen 2007; 12: 915-24.]. The advantages of FP include that it is a homogeneous format, utilizes small substrates and the assays are generally insensitive to fluorescence interference.

To provide a valuable readout, the FP substrate should bind specifically and with high affinity to the target protein. Substrate binding was first assessed using the permanently unfolded model substrate, reduced and carboxymethylated lactalbumin (RCMLA). We have previously shown in a gel shift assay format that both DnaK and Hsp72 can bind to RCMLA [29Jindal S, Murray P, Rosenberg S, Young R, Williams KP. Human stress protein hsp 70: Overexpression in E.coli, purification and characterization Biotechnology 1995; 13: 1105-9.]. RCMLA was fluorescently labeled with fluorescein (Flu-RCMLA) and binding to Hsp72 and DnaK assessed as described in materials and methods. The data for Hsp72 binding Flu-RCMLA is shown in Fig. (1).

The data in Fig. (1) yielded an apparent K_d value of 1.53 µM. A comparable value (apparent K_d value of 2.0 µM) was obtained for DnaK (data not shown). Although Flu-RCMLA could potentially fulfill the requirement for a universal Hsp70 substrate, the relatively low affinity, limited assay window of ~ 20 mP and the challenge in consistently preparing the substrate precluded its use. For these reasons we decided to look to small peptide substrates to facilitate Hsp70 assays.

Due to their function in binding a wide range of protein sequences, defining the peptide specificities of Hsp70 chaperones has been challenging. A number of groups have used synthetic peptides to map such motifs [12Rüdiger S, Germeroth L, Schneider-Mergener J, Bukau B. Substrate specificity of the DnaK chaperone determined by screening cellulose-bound peptide libraries EMBO J 1997; 16: 1501-7.-14Takenaka I, Leung S, McAndrew S, Brown J, Hightower L. Hsc70-binding peptides selected from a phage display peptide library that resemble organellar targeting sequences J Biol Chem 1995; 270: 19839-44., 16Hightower L, Sadis S, Takenaka I. Interactions of vertebrate hsc70 and hsp70 with unfolded proteins and peptides The Biology of Heat Shock Proteins Mol Chaperones 1994; 26: 179-207.-20Williams KP, Evans DM, Rosenberg S, Jindal S. Use of a peptide library to characterize differential peptide binding specificities of bacterial and mammalian Hsp70 In: Marshak DR, Ed. Techniques in Protein Chemistry VII. New York: Academic Press 1996; pp. 57-64., 22Crouy-Chane A, Kohiyama M, Richarme G. Specificity of DnaK for arginine/lysine and effect of DnaJ on the amino acid specificity of DnaK J Biol Chem 1996; 271: 15486-90., 35Gragerov A, Zeng L, Zhao X, Burkholder W, Gottesman M. Specificity of DnaK-peptide binding J Mol Biol 1994; 235: 848-54.]. For DnaK the bacterial form of Hsp70, a 21-mer peptide comprising the targeting sequence of the precursor of mitochondrial aspartate aminotransferase (ALLQSRLLLSAPRRAAATARA; peptide name ALLQ, Table 1) had been previously identified [11Schmid D, Baici A, Gehring H, Christen P. Kinetics of molecular chaperone action Science 1994; 263: 971-73.] as binding tightly. It was also shown [11Schmid D, Baici A, Gehring H, Christen P. Kinetics of molecular chaperone action Science 1994; 263: 971-73.] that this peptide could be fluorescently tagged on the N-terminus and used to follow real time binding by monitoring changes in fluorescence upon peptide binding.

Because of the reported relatively high affinity of this peptide for DnaK (63 nM) [11Schmid D, Baici A, Gehring H, Christen P. Kinetics of molecular chaperone action Science 1994; 263: 971-73.] we first used this sequence (Table 1) as the basis for developing a fluorescence polarization assay for measuring peptide binding to DnaK and potentially to other Hsp70 proteins. The ALLQ peptide was labeled with NHS-fluorescein on the N-terminus on resin (Flu-ALLQ). Because labeling is done prior to deprotection only the N-terminal amine is labeled. This is especially advantageous for labeling of lysine/arginine containing peptides.

A second peptide containing predominantly lysine and arginine residues (NIVRKKK; peptide name KKK) previously shown to bind Hsp70 proteins [17Fourie A, Sambrook J, Gething M. Common and divergent peptide binding specificities of hsp70 molecular chaperones J Biol Chem 1994; 269: 30470-8.] was also labeled on-resin using NHS-fluorescein. Peptide substrates used in these studies are shown in Table 1.

The optimal trace concentration was determined for each fluorescent-peptide. Each peptide was titrated and the fluorescence intensity measured. The ratio of fluorescence intensity to background was plotted versus tracer concentration. For optimal sensitivity, the fluorescently labeled peptide tracer concentration was set to produce a total fluorescence signal-to-background ratio greater than 50. For example, for Flu-ALLQ this required a concentration of 50 nM (data not shown) and 10 nM for Flu-KKK (Fig. 2).

Binding of DnaK to a fixed concentration of Flu-ALLQ (50 nM) was determined. For this experiment it was determined that an incubation time of 10 min at room temperature was sufficient to reach equilibrium. A binding isotherm constructed by plotting DnaK concentration against millipolarization (mP) values shows that DnaK binds in a dose-dependent manner. As the concentration of DnaK is increased mP values increase until saturation is reached (Fig. 3). The assay window reached approximately 70 mP.

The data in Fig. (3) was analyzed using a non-linear regression analysis method (single site binding model) in Prizm 5.0 and yielded an apparent K_d value of 64.4 nM, very close to that reported previously [11Schmid D, Baici A, Gehring H, Christen P. Kinetics of molecular chaperone action Science 1994; 263: 971-73.]. A Scatchard plot transformation of the same data (not shown) yielded a value of 53.7 nM for “target concentration” (x intercept) indicating stoichiometric 1:1 binding of peptide substrate ligand and DnaK in this FP assay.

It has been shown that Hsp70 members associate with polypeptide substrates or peptides in the presence of ADP whereas the ATP bound form of Hsp70 proteins have poor affinity for substrates [10Wei J, Gaut J, Hendershot L. In vitro dissociation of BiP-peptide complexes requires a conformational change in BiP after ATP binding but does not require ATP hydrolysis J Biol Chem 1995; 270: 26677-82.]. We tested the effect of ADP and ATP on the FP assay. The addition of ADP to the assay did not affect the binding of labeled substrates (data not shown) suggesting that the Hsp72 proteins as tested exist in a predominantly ADP-bound form as a result of their purification using ATP affinity chromatography [29Jindal S, Murray P, Rosenberg S, Young R, Williams KP. Human stress protein hsp 70: Overexpression in E.coli, purification and characterization Biotechnology 1995; 13: 1105-9.]. For DnaK binding to Flu-ALLQ, no peptide binding was measurable in the presence of 3 mM ATP/1 mM MgCl2 (data not shown) suggesting that this FP assay can detect conformational changes in DnaK.

Using the information obtained in the binding isotherm experiments, we next developed a competition binding assay in which tracer and protein concentrations are fixed and increasing concentrations of compounds are added and assessed for inhibition. DnaK and Flu-ALLQ concentrations were fixed at 200 nM and 50 nM respectively to give ~90% binding. Compounds can then be added over a range of concentrations and mP values measured after incubation.

To further validate the competitive DnaK FP assay, the IC₅₀ for a known DnaK binding peptide was to be determined. In the absence of any known DnaK binding small molecules, another peptide previously identified as a DnaK binder was assessed in this assay. The peptide with sequence NRLLLTG (Table 1; termed LLL) was previously identified as a DnaK binder from phage display screening [35Gragerov A, Zeng L, Zhao X, Burkholder W, Gottesman M. Specificity of DnaK-peptide binding J Mol Biol 1994; 235: 848-54.] and has been well characterized as a DnaK binder [37Stevens S, Cai S, Pellecchia M, Zuiderweg E. The solution structure of the bacterial HSP70 chaperone protein domain DnaK (393-507) in complex with the peptide NRLLLTG Protein Sci 2003; 12: 2588-96.-39Swain J, Schulz E, Gierasch L. Direct Comparison of a stable isolated Hsp70 substrate-binding domain in the empty and substrate-bound states J Biol Chem 2006; 281: 1605-1.]. The K_d for LLL mediated stimulation of DnaK ATPase was previously reported as 10 µM [35Gragerov A, Zeng L, Zhao X, Burkholder W, Gottesman M. Specificity of DnaK-peptide binding J Mol Biol 1994; 235: 848-54.]. LLL was analyzed in our FP competition assay. Results are shown in Fig. (4).

The LLL peptide inhibited Flu-ALLQ binding to DnaK, with an IC₅₀ value of 2.4 µM comparable to that reported previously [35Gragerov A, Zeng L, Zhao X, Burkholder W, Gottesman M. Specificity of DnaK-peptide binding J Mol Biol 1994; 235: 848-54.], and an R-squared value of 0.97 for curve fit. A preliminary analysis of the applicability of the assay for high throughput screening (HTS) was assessed using the maximum assay window from the data in Fig. (4). A Z’ value [36Zhang J, Thomas D, Chung Y, Oldenburg KR. A Simple statistical parameter for use in evaluation and validation of high throughput screening assays J Biomol Screen 1999; 4: 67-74.] of 0.64 was calculated from this data suggesting that this assay can be applied to HTS.

Next we assessed the ability of another Hsp70 protein, Hsp72 to also bind to the peptide substrate Flu-ALLQ. A binding isotherm of the data (Fig. 5) demonstrates that Hsp72 also binds Flu-ALLQ in a dose-dependent manner.

Non-linear regression analysis yielded an apparent K_d value of 621.6 nM indicating that Hsp72 binds this peptide with ~10-fold weaker affinity than DnaK. Interestingly, using a smaller fragment of ALLQ, comparable values for binding to DnaK and Hsc70 were observed: 0.43 µM to Hsc70 [31Kang Y, Taldone T, Clement C, et al. Design of a fluorescence polarization assay platform for the study of human Hsp70 Bioorg Med Chem Lett 2008; 18: 3749-51.] and 0.4 µM to DnaK [40Han W, Christen P. cis-Effect of DnaJ on DnaK in ternary complexes with chimeric DnaK/DnaJ-binding peptides FEBS Lett 2004; 563: 146-50.]. Whether this reflects differences in Hsp72 and Hsc70 binding or differences in the lengths of the peptides in mediating binding requires further investigation.

For Hsp72, a seven residue peptide (NIVRKKK, termed KKK) was previously identified as a Hsp70 binder [14Takenaka I, Leung S, McAndrew S, Brown J, Hightower L. Hsc70-binding peptides selected from a phage display peptide library that resemble organellar targeting sequences J Biol Chem 1995; 270: 19839-44., 17Fourie A, Sambrook J, Gething M. Common and divergent peptide binding specificities of hsp70 molecular chaperones J Biol Chem 1994; 269: 30470-8.] and shown to compete for RCMLA binding to Hsp72 [29Jindal S, Murray P, Rosenberg S, Young R, Williams KP. Human stress protein hsp 70: Overexpression in E.coli, purification and characterization Biotechnology 1995; 13: 1105-9.]. KKK was labeled with fluorescein and also used to measure direct binding to Hsp72 in an FP assay (Fig. 6).

Binding isotherms for Hsp72 binding to Flu-KKK and with Flu-RCMLA as a control (Fig. 6) yield apparent K_d values of 13.7 µM and 1.97 µM respectively. An IC₅₀ value for KKK binding to Hsc70 of 2.8 µM has been reported previously [14Takenaka I, Leung S, McAndrew S, Brown J, Hightower L. Hsc70-binding peptides selected from a phage display peptide library that resemble organellar targeting sequences J Biol Chem 1995; 270: 19839-44.]. These differences in binding for the KKK peptide may reflect different selectivity’s of Hsp72 and Hsc70 or may suggest that fluorescein labeling of this minimal binding sequence of seven residues may have steric effects on binding. The low affinity of the “charged” Hsp70 binder makes it a poor tracer to monitor binding and challenging to use in developing an assay to identify chemical probes.

Another set of peptides have been identified as Hsp70 binders. Peptides derived from the MHC class I antigens have been shown to bind to Hsc70 and also have immunosuppressive properties [25Krensky A, Clayberger C. HLA-derived peptides as novel immunotherapeutics Clin Immunol Immunopath 1995; 75: 112-6.,26Woo J, Iyer S, Cornejo M, et al. Immunosuppression by D-isomers of HLA class I heavy chain (amino acid 75 To 84)-derived peptides is independent of binding to Hsc70 Transplantation 1997; 64: 1460-7.]. We tested one of these peptides from HLA-B2702 (residues 75-84; RENLRIALRY; Table 1) in an FP assay. This 10-mer peptide was labeled on the N-terminus with fluorescein (termed Flu-HLA) and used as a tracer to measure Hsp72 binding (Fig. 7).

Non-linear regression analysis of the data in Fig. (7) yielded an apparent K_d value of 0.37 µM for Flu-HLA binding to Hsp72. Using a biotinylated version of this peptide as tracer in an ELISA type competition binding assay, Woo and coworkers measured an IC₅₀ value of 7 µM for binding of this peptide to Hsc70 [26Woo J, Iyer S, Cornejo M, et al. Immunosuppression by D-isomers of HLA class I heavy chain (amino acid 75 To 84)-derived peptides is independent of binding to Hsc70 Transplantation 1997; 64: 1460-7.]. These differences in binding may reflect differences in assay format or that the HLA peptide has higher affinity for Hsp72 versus Hsc70.