I. Introduction

Peptide ligands have attracted attention as powerful molecular tools in the treatment of cancer patients^［1,2］. Current chemotherapy using standard anticancer drugs often leads to severe side effects, causing extreme distress to the patient. Development of methodologies that allow for cancer cell-specific delivery of therapeutic agents offers a potential strategy in decreasing these harmful and undesirable side effects^［3,4］.
Recent studies have shown that peptide ligands bound to glucose-regulated protein 78 （GRP78）can successfully be used for peptide-based drug delivery^［5］. GRP78, a member of the HSP70 family, is an endoplasmic reticulum （ER）protein that possesses a carboxy（C）-terminal KDEL sequence that constitutes a feedback signal from the Golgi to the ER^［6］. GRP78 is principally located in the ER; however, reports also indicate its existence on the cell surface. Expression levels in the ER and on the cell surface are both increased under stress conditions^［7,8］, and high expression levels on the cell surface are linked to a greater degree of cancer malignancy ^［9,10］. GRP94, a paralog of HSP90, is also an endoplasmic reticulum（ER）protein that possesses peptide and adenine nucleotide-binding domains. Like other proteins residing in the ER, GRP94 contains a C-terminal KDEL sequence. GRP94 also exists on the cell surface and functions as an antigen-representing chaperon protein^［11］. Increased expression of GRP94 on the cell surface has also been demonstrated to be associated with tumorigenicity^［12,13］. In this study, we focused on the discovery of GRP94-binding peptides. Since the amino （N）-terminal fragment of the GRP94 protein is involved in peptide presentation via professional antigen presenting cells^［14］, it is possible that any peptide can bind to this fragment sequence-independently. We therefore searched for peptides that could bind to the C-terminal fragment of the GRP94 protein.

II. Materials and Methods

Vector Construction

Full-length GRP94 cDNA（accession no. X15187） was amplified using a forward primer（5'-AAA GGA TCC ATG AGG GCC CTG TGG GTG CT-3'） and a reverse primer （5'-AAA GAA TTC TTA TTC AGC TGT AGA TTC CT-3'） and inserted into the pGEX-6P-1-GST fusion protein expression vector （GE Healthcare UK Limited, Buckinghamshire, UK） to yield pGEX-GRP94. The plasmid was further digested with BamH1 and HindIII, and a 3.7-kb fragment containing the C-terminal domain of GRP94（amino acids 493 to 799） was re-inserted into the empty expression vector pGEX-6P-1 to yield pGEX-GRP94-C.

Purification of recombinant protein

BL21 competent cells were transformed with pGEX-GRP94-C. Transformants were cultured in 2×YT medium containing 100 ug/ml ampicillin at 30℃, and fusion protein expression was induced using isopropyl-1-thio-Β-D-galactopyranoside （IPTG）. Glutathione S-transferase （GST） and the C-terminal domain of the GRP94 fusion protein （GST-GRP94-C） was bound to a GSTrap FF column （GE Healthcare UK Limited） and further digested overnight with PreScission Protease （GE Healthcare Limited, UK） at 4℃. The C-terminal domain of GRP94 （～37.2 kDa-peptide; GRP94-C） was subsequently eluted with a buffer without glutathione.

Phage biopanning

Peptide sequences were selected from a commercially available disulfide constrained 7-mer random library displayed on phage M13 via N-terminal fusion to the minor coat protein, g3p, with a diversity of 1.3×10⁹ （Ph. D.-C7C™ phage display peptide library kit, New England Biolabs, Beverly, MA, USA）, as described previously^［15］. Initially, recombinant GRP94-C protein was biotinylated using a sulfo-NHS-LC-biotin （Pierce, USA） according to the manufacturer’s instructions. Biotinylated protein （0.1 ug） and 2×10¹¹ phage particles from the Ph. D.-C7CTM phage display peptide library kit were mixed in 400 ul of Tris-buffered saline （50mM Tris pH 7.5, 150mM NaCl） supplemented with 1% （v/v） Tween-20 （TBS-T） to prepare the phage-target complex. Phage-target complex was then panned against a streptavidin-coated plastic plate for affinity capture. After washing, bound phages were eluted with 1mM dithiothreitol （DTT） and amplified for next-round panning. After a third round of amplification and panning for binding GRP94-C protein, individual phages were isolated and the displayed peptide sequences were deduced after DNA sequencing.

Enzyme-linked immunosorbent assay （ELISA） for binding of selected phages to GRP94-C

Candidates were selected for the ELISA assay based on amino acid similarity. Ninety-six-well microtiter plates （Nunc Maxisorp, Denmark） were coated with 200 ul per well of GRP94-C protein （100 ug/ml）. 0.1 M NaHCO₃ （pH 8.6） and incubated overnight at 4℃. After washing with TBS-T, wells were incubated with filter-sterilized blocking buffer （0.1 M NaHCO₃ pH 8.6, 5mg/ml bovine serum albumin, 0.02% NaN₃） to reduce any non-specific interaction of phages to the plate. Fourfold-serially diluted samples of selected phages were then placed in ELISA microplates and incubated for 2 h at room temperature. After the unbound phages were removed by washing with TBS-T, a horseradish peroxidase （HRP）-conjugated anti-M13 antibody （GE Healthcare Limited） diluted with blocking buffer at a ratio of 1:5000 was added to the plates and incubated for 1h at room temperature. Immune complexes were then detected using 2, 2'-azino-di-［3-ethylbenzthiazoline sulfonate］ diammonium salt （Roche, Berlin, Germany） as a substrate for HRP. After a 20-min incubation at room temperature, the absorbance at 405 nm was determined using an automated ELISA reader （Model E-max, Molecular Devices, CA, USA）. From the pool of candidate peptides ^{(Table 1)}, a single randomly selected peptide （peptide No.34） with a lower or no amino acid similarity to the above selected peptides was used as a negative control under identical experimental conditions.

Peptide biotinylation

Cysteine-constrained peptides with biotinylated amino termini were purchased from Scrum Co. Ltd. （Tokyo, Japan）. Intra-molecular disulfide bridge formation of the peptides was characterized using matrix-assisted laser desorption/ionization time-of-flight mass spectroscopy before and after reduction using 1mM DTT and also by their unreactivity to Ellman’s reagent, a sulfhydryl assay reagent （Pierce, IL, USA）.

Quartz Crystal Microbalance （QCM） analysis

Binding analysis of the interaction between GRP94-C and synthetic peptides was performed with a 27 MHz QCM machine （Affinix Q4, Tokyo, Japan）. Briefly, the sensor chip was initially cleaned twice with 1% SDS solution and rinsed with phosphate-buffered saline （PBS, 150 mM NaCl in 10 mM phosphate buffer, pH 7.2）. Neutravidin （10 ug/ml） was then absorbed onto the chip for immobilization of biotinylated peptides. Biotinylated peptide solution （10 uM） was further added to the neutravidin-absorbed sensor chip for one hour at room temperature. Finally, GRP94-C protein was added to the peptide-immobilized chip at different concentrations for binding measurements and the K_d value was calculated. Purified GST protein was used as a negative control under identical experimental conditions.

III. Results

Selection of GRP94-binding Phage by Phage display:

To screen peptide ligands for the GRP94-C protein, we used a phage display peptide library displaying 7-mer cysteine constrained random peptides （C-X7-C）. During the course of selection, the relative yield of phage recovery improved from 0.000005 （first round） to 0.0001 （third round）, which is consistent with enrichment in binding clones. After three rounds of panning using purified GRP94-C protein, 80 phage clones were randomly selected, amplified and sequenced to identify the displayed peptides. Primarily from the pool of 80 candidate peptides ^{(Table 1)}, 16 different phage peptide sequences were selected and arranged into four groups based on amino acid sequence or consensus motif similarity ^{(Fig. 1)}.

Confirmation of phage binding to GRP94-C by ELISA

Phage ELISA experiments were performed to test the specific binding of peptides to the target GRP94-C protein. The 16 phages selected^{(Fig. 1)} were individually amplified, serially diluted and incubated with purified recombinant GRP94-C target protein, and a randomly selected phage was investigated in parallel as a negative control. Fourteen phages （excluding phage No. 21 and 35） showed increased absorbance values in the presence of the target protein when compared with the negative control ^{(Fig. 2)}. We chose eight clones （phage Nos. 4, 6, 18, 58, 60, 66, 72 and 79） that showed the greatest increase in reactivity using the first ELISA. We further performed phage ELISA analyses to confirm the binding strength of phage-borne peptides to GRP94-C protein ^{(Fig. 3)}. Four of the eight phage clones （phage Nos. 6, 58, 60 and 72） were selected based on their binding activity to GRP94-C and ranked on the basis of the final ELISA results. The best three candidates （phage Nos. 6, 58 and 72, ^{Fig. 4}） were chosen for downstream experiments. All three phages showed a similar absorbance range and thus, were expected to bind to GRP94-C with a similar affinity range.

QCM Analysis

Binding affinity of the selected three peptides （peptide Nos. 6, 58 and 72） with GRP94-C protein was measured by QCM analysis. Protein binding to the immobilized peptide was calculated based on the Sauerbrey equation^［16］, which represents a linear dependence between frequency shift and the bound GRP94-C protein. Binding of the target GRP94-C protein to the peptides immobilized on the QCM sensor chip was observed as a decreased frequency ^{(Fig. 5A)}. Addition of purified GST protein did not produce a shift in frequency and suggested that the GST proteins did not bind to the immobilized peptides ^{(Fig. 5A)}. The affinity constant （K_d） value was determined by plotting absorbed GRP94-C protein in terms of frequency shift （ΔF） versus the concentration of injected GRP94-C in the buffer solution ^{(Fig. 5B)}. K_d values for all three peptides （peptide Nos. 6, 58 and 72） binding to target GRP94-C protein were within the range of 0.39-0.46 uM （0.45 uM, 0.46 uM and 0.39 uM, respectively）. No significant difference in the binding affinity of these peptides to GRP94-C was observed, although peptide No. 72 did show a slightly higher binding affinity.

IV. Discussion

Low bioavailability and nonspecific toxicity to healthy tissues present major obstacles in the development of new therapeutic anticancer drugs. Discovery of functional ligand-receptor systems is one of the critical steps toward the development of an efficient cancer therapy. Recently, several studies have suggested that proteins selectively expressed on the surface of cancer cells are potentially useful targets for the development of novel anticancer therapies^{［5, 17］}.
Previous studies have shown that the GRP94 chaperon protein is expressed on the cell surface in some cancer cells but not in non-transformed cells^［12］. Characteristics of cell surface expression of GRP94 have however received much less attention, since it is widely considered to be a chaperon in immune signaling and peptide presentation for antigen representing cells^［18］. In the present study, we focused on cell surface expression of GRP94 as a target for a peptide-drug conjugate. As a first step, we screened GRP94 protein-binding peptides using a phage display technique. The C-terminal domain of GRP94 protein was chosen to avoid non-specific, physiological peptide binding of the N-terminal domain^［14］. Initially, 16 peptides were selected from the phage display-derived pool of candidate peptides and divided into four groups based on their amino acid similarity ^{(Fig. 1)}. Threonine, proline and leucine amino acid residues were frequently found in the peptide sequences for each of the four groups and these may play a role in GRP94 protein binding.
Three peptides （peptide Nos. 6, 58 and 72） out of the above-mentioned 16 candidates were selected based on their higher absorbance at 405 nm using multiple ELISAs and gave values within a similar range （absorbance reading 1.3-1.6 at 405 nm）. QCM analysis revealed a high affinity of these peptides to GRP94-C protein. Kd values for the binding of all three peptides to GRP94 were within a similar range of 0.39-0.46 uM, which also corresponded well to the similar range of absorbance levels obtained using ELISA ^{(Fig. 4,Fig. 5)}. The GRP94-C protein did not bind to either the random peptide in ELISA assay or the GST protein in QCM analysis, which further confirmed the specificity of our peptides towards the C-terminal of the GRP94 protein.
In summary, we have reported for the first time the discovery of synthetic peptide ligands that bind to the C-terminal domain of the GRP94 protein with high affinity. Our results suggest that these novel peptide sequences could potentially be used as a part of a peptide-conjugated drug system, designed to selectively target cancer cells with surface expression of GRP94.

Acknowledgments

This work was supported in part by Grants-in-Aid for Scientific Research through the Industrial Technology Research Grant Program in ‘04 from the New Energy and Industrial Technology Development Organization, the Smoking Research Foundation, the Tokyu Foundation for a Better Environment, the Kieikai Research Foundation, and the Japan Society for the Promotion of Science. M. Zahed was a doctoral student supported financially by a MEXT Honors Scholarship for International Students.