- Email: [email protected]
Development of Novel Tumor Imaging Agents with Phage-Display Combinatorial Peptide Libraries
Preliminary Investigations
Development of Novel Tumor Imaging Agents with Phage-Display Combinatorial Peptide Libraries1 Michael J. Campa, PhD, Scott B. Serlin, BA, Edward F. Patz, Jr, MD
Rationale and Objectives. Current radiologic methods do not provide sufficient information for unambiguous diagnosis and prognosis of cancer. The present investigation sought to address this deficiency by developing a system for designing novel small molecules targeted against tumor-specific molecules for use as radionuclide imaging agents. Materials and Methods. Part of a tumor-specific receptor, purified recombinant epidermal growth factor receptor (EGFR), variant III, extracellular domain (rEGFRvIII-ecd), was used as the target in the selection of EGFRvIII-specific peptide ligands from random peptide bacteriophage (phage) display libraries. After three rounds of screening, phage isolates were tested for binding affinity with an enzyme-linked immunosorbent assay. Positive phage were sequenced, and the peptides were synthesized and tested for binding affinity with a surface plasmon resonance assay. Results. Affinity screening identified 49 peptide-expressing phage that showed enhanced binding to the variant receptor compared with wild-type EGFR. Free peptides from the two phage isolates exhibiting the most favorable binding were tested for target binding. One of these demonstrated a binding affinity for rEGFRvIII-ecd in the 30-nmol/L range. Conclusion. These data suggest that phage display libraries may be very useful in the design of novel, high-affinity tumor imaging agents. Key Words. Neoplasms, diagnosis. ©
AUR, 2002
Radiologic studies play an integral role in the evaluation of cancer patients, but conventional modalities in current use have clear limitations. Most imaging studies provide a tremendous amount of anatomic and morphologic information, but abnormalities often remain indeterminate after standard evaluation, and patients must often proceed to an invasive procedure for tissue diagnosis. In addition, conventional imaging offers minimal prognostic information. Some tumors have an indolent course, while others are aggressive, and there is usually no way to differentiate these lesions that are radiologically and Acad Radiol 2002; 9:927–932 1 From the Department of Radiology, Duke University Medical Center, Box 3808, Durham, NC 27710. Received April 9, 2002; revision requested April 15; revision received April 16; accepted April 17. Supported in part by grant R21 CA82899-01 from the National Cancer Institute. Address correspondence to E.F.P.
©
AUR, 2002
often pathologically similar but biologically different. If the biologic behavior could be distinguished noninvasively, then patients could be stratified into the most appropriate treatment protocols, with possible improvements in outcome. Over the past several years [18F]-2-fluoro-2-deoxy-Dglucose positron emission tomography has emerged as an additional study used to address some limitations of conventional imaging. While it is an interesting model that provides biologic information (glucose metabolism), the specificity and characterization capabilities of fluorodeoxyglucose are less than optimal (1–3). Thus, the search continues for more accurate tumor imaging agents. We have taken a new approach in the effort to design novel tumor-specific molecular imaging agents. For the current investigation we chose to use random peptide phage display libraries as a ligand discovery technique, while targeting a genomically mutated, tumor-specific
927
CAMPA ET AL
epidermal growth factor receptor (EGFR), EGFR-variant III (EGFRvIII) (4 – 6). Once such a model system has been developed, new molecular imaging agents can also be produced for a variety of tumor targets. MATERIALS AND METHODS Affinity-Selection Target We began the initial phage library screens with targeting against Escherichia coli– expressed, purified recombinant EGFRvIII, extracellular domain (rEGFRvIII-ecd). We chose this recombinant protein as a target for our phage library screens because it was shown by immunologic methods to contain structural features consistent with authentic tumor-expressed EGFRvIII (7). Phage Display Libraries We used random peptide phage display libraries representing several different random peptide structural formats. Eight libraries were of the form (X)6-Z-(X)6, where X represents a random amino acid residue and Z a fixed residue. Each library possesses one fixed central residue that provides a common structural constraint across all sequences in the library. Each library, therefore, displays random sequences in different structural contexts. We used libraries containing all classes of amino acid side chains (G, D, F, H, K, L, P, and W). We also employed a random 16-mer library, as well as two different cysteinecontaining 16-mer libraries. Expressed peptides containing a minimum of two cysteine residues are held in a relatively rigid confirmation by disulfide bridges between the cysteine side chains. These constrained libraries have been shown to allow the identification of higher-affinity peptides compared with peptide libraries containing no cysteine residues. Affinity-Selection Screening A purified affinity-selection target protein (2.5 g of EGFRvIII, extracellular domain [rEGFRvIII-ecd] in 100 L of 100 mmol/L sodium bicarbonate, pH 8.5) is immobilized in microtiter wells with overnight incubation at 4°C. Additional binding sites are then blocked by incubation for 1 hour at room temperature in 10 mg/mL bovine serum albumin in a sodium bicarbonate buffer. The wells are then washed four times with phosphate-buffered saline containing 0.1% Tween 20 (PBST), and approximately 1010 plaque-forming units (pfu) from each phage library are added to the target-coated well. After a 2-hour incubation, nonbinding phage are removed by washing the
928
Academic Radiology, Vol 9, No 8, August 2002
wells five times with PBST, and bound phage are eluted with 50 mmol/L glycine– hydrochloric acid (pH 2.0). Eluted phage are neutralized with 200 mmol/L sodium phosphate buffer (pH 8.5), titered, and amplified in TG-1 E coli. Amplified phage (1010 pfu) are then used for the second and third rounds of affinity selection, performed as in the initial round. After the third round, eluted phage are plated out, and individual plaques are amplified in TG-1 E coli and tested for target protein binding by means of enzyme-linked immunosorbent assay (ELISA). ELISA Target-binding Assay Amplified phage from each plaque are first tested for binding to rEGFRvIII-ecd and to the blocking agent, BSA. Phage demonstrating enhanced binding to the target protein are then tested against rEGFRvIII-ecd and wildtype recombinant EGFR-ecd to gauge the specificity of binding. The assay is carried out by incubating phage in microtiter wells containing immobilized target protein, removing nonbinding phage by washing, and detecting bound phage with a horseradish peroxidase-conjugated anti-M13 antibody and ABTS (2,2⬘-azinobis [3-ethylbenzthiazoline-6-sulfonic acid]) and hydrogen peroxide. The colorimetric reaction is quantitated by measuring the absorbance at 405 nm. BIAcore Assay Phage isolates and free peptides were also tested for binding to rEGFRvIII-ecd by surface plasmon resonance (SPR) in a BIAcore analysis unit (Pharmacia, Piscataway, NJ). For this analysis, purified target proteins are covalently attached to a gold-coated chip located inside a BIAcore flow cell. A solution containing phage at a concentration of 1012 pfu/mL in 150 mmol/L sodium chloride, 3.4 mmol/L ethylenediaminetetraacetic acid, 0.005% (vol/vol) Triton X-100, and 10 mmol/L HEPES (pH 7.4) is then directed across the chip at a flow rate of 10 L/ min. Any phage binding to the immobilized target will increase the total mass of protein attached to the chip, which, in turn, alters the refraction angle of light directed at the nontarget side of the chip. The change in light refraction is quantitated, allowing phage binding to be detected in real time. RESULTS After three rounds of affinity selection against rEGFRvIII-ecd with seven phage libraries of the form (X)6-Z(X)6, 48 phage isolates from each library were tested for
DEVELOPMENT OF NOVEL TUMOR IMAGING AGENTS
Academic Radiology, Vol 9, No 8, August 2002
Table 1 Results of Affinity Selection Library
Positive Isolates
Proline Aspartic acid Phenylalanine Histidine Lysine Leucine Tryptophan
3 (6.1) 12 (24.5) 6 (12.2) 6 (12.2) 10 (20.4) 7 (14.3) 5 (10.2)
Figure 1. Phage isolates were tested against equimolar amounts of immobilized rEGFRvIII-ecd, rEGFR-ecd, and BSA by means of ELISA. MR-1 is an EGFRvIII-specific phagemid and is included as a positive control. Values shown are the means of duplicate determinations.
target binding with ELISA. The results (Table 1) demonstrated that 49 isolates representing all seven libraries bound the target protein in preference to BSA. These isolates were subsequently tested by means of ELISA against rEGFRvIII-ecd and recombinant wild-type EGFR-ecd to determine binding specificity. All 49 isolates exhibited varying degrees of binding specificity for rEGFRvIII-ecd. The results for 10 isolates are shown in Figure 1. We used the EGFRvIII-specific single-chain Fv phagemid MR1 as a positive control in these assays (8). To verify the ELISA data and to gauge the relative avidities of the various isolates for the target protein, we also tested all isolates by SPR against immobilized rEGFRvIII-ecd. The phenomenon of SPR occurs when surface plasmon waves are excited at a metal-liquid interface. Light is directed at, and reflected from, the side of the surface not in contact with the sample, and SPR causes a reduction in reflected light intensity at a specific combination of angle and wavelength. Biomolecular bind-
ing events cause changes in the refractive index at the surface layer, which are detected as changes in the SPR signal. The magnitude of these SPR changes, measured in resonance units, is proportional to the quantity of bound molecules. Relative avidity for the rEGFRvIII-ecd target was determined by comparing the number of resonance units bound among all phage isolates. Using this method, we observed phage binding to immobilized rEGFRvIIIecd that varied over a 20-fold range. Representative SPR plots for a high-binding and a low-binding phage are shown in Figure 2. Random inserts of the phage plasmids from the two best binders were then sequenced, and peptides were synthesized with an amino-terminal biotin moiety attached to a spacer of six glycine residues. The peptides were then tested for binding to immobilized wildtype or variant receptor in the BIAcore assay. The results of the BIAcore studies demonstrated a difference in binding preferences for the two peptides tested. Peptide 1 exhibited a slightly faster association rate with the wild-type receptor (ka ⫽ 5.44 ⫻ 104 vs 3.61 ⫻ 104 M⫺1 䡠 sec⫺1), but it dissociated nearly three times more slowly (kd ⫽ 3.08 ⫻ 10⫺3 vs 1.10 ⫻ 10⫺3 sec⫺1), resulting in an overall higher binding affinity for the variant receptor (KD ⫽ 30.5 nM for EGFRvIII vs 56.6 nM for wild-type EGFR). In contrast, peptide 2 exhibited both a faster on-rate (ka ⫽ 2.53 ⫻ 104 vs 1.73 ⫻ 104 M⫺1 䡠 sec⫺1) and a slower off-rate (kd ⫽ 1.41 ⫻ 10⫺3 vs 2.51 ⫻ 10⫺3 sec⫺1) with the wild-type receptor. The result is that peptide 2 bound wild-type EGFR more tightly than it bound EGFRvIII (KD ⫽ 145 nM for EGFRvIII vs 55.5 nM for wild-type EGFR). In contrast to the SPR results obtained with rEGFRvIII-ecd and rEGFR-ecd used as targets, neither peptide showed binding above background to solubilized cellexpressed receptor in a pull-down assay (9) or to the receptor on intact cells by using fluorescence-activated cell sorting, or FACS. We hypothesized that the bacterial expressed rEGFRvIII-ecd did not faithfully reproduce structural characteristics of the cell-expressed protein crucial for peptide recognition. Therefore, we elected to carry out additional phage library screens by using a baculovirusbased eukaryotic expression system for production of the target protein. We expected that glycosylation and folding of the purified protein in this expression system would be more representative of the authentic tertiary structure of EGFRvIII in tumor cells. We used a phage library of the form C-(X)16 for our initial screens with the baculovirus-expressed rEGFRvIIIecd (bvrEGFRvIII-ecd). The cysteine residue at the amino
929
CAMPA ET AL
Academic Radiology, Vol 9, No 8, August 2002
Figure 2. Phage isolates diluted to 1012 pfu/mL and tested in the SPR assay against immobilized rEGFRvIII-ecd or negative control protein (mouse immunoglobulin G). All determinations were carried out in 10 mmol/L HEPES, pH 7.4, 150 mmol/L sodium chloride, 3.4 mmol/L ethylenediaminetetraacetic acid, 0.005% (vol/vol) Tween 20 with an analyte flow rate of 10 L/min. In all the above plots, analyte (phage) injection begins approximately 115 seconds into the run and continues until 350 seconds, with the ligand-analyte association phase occurring between these two points. The dissociation phase is from approximately 350 to 700 seconds. The plots show results from two phage isolates, F6 and W46, representing high-binding and low-binding phage. The y axis (resonance units) is a measure of the quantity of phage associated with immobilized rEGFRvIII-ecd or immunoglobulin G. Solid line ⫽ EBFRvIII, dashed line ⫽ immunoglobulin G.
terminus of the random sequence has the potential to form a disulfide bond with a second cysteine residue located at any position in the random segment, imparting a high degree of flexibility to the overall three-dimensional structure of the selected peptides. We performed three rounds of affinity selection against bvrEGFRvIII-ecd and tested individual isolates against immobilized bvrEGFRvIII-ecd and bvrEGFR-ecd. Of 192 phage isolates tested, five showed a binding preference for bvrEGFRvIII-ecd over bvrEGFR-ecd, as judged by ELISA. Table 2 shows the optical density at 405 nm resulting from phage binding to bvrEGFR-ecd or bvrEGFRvIII-ecd. We are continuing to evaluate additional phage isolates for target-specific binding. Phage isolates demonstrating bvrEGFRvIII-ecd binding will then be tested for EGFRvIII binding in whole cells. Since EGFRvIII-binding peptide ligands are intended for in vivo use as tumor targeting agents, it is critical that phage binding in whole cells be verified before synthesis of free peptides and affinity determinations. Peptides exhibiting high-affinity (KD ⱕ 10 nM) binding to cell-expressed EGFRvIII will be evaluated further for their suitability as targeting agents, whereas EGFRvIII-binding peptides of lower affinity will serve as the starting point for the synthesis and screening of focused phage display libraries.
930
Table 2 Optical Densities Resulting from Phage Binding Optical Density at 405 nm Phage
bvrEGFR-ecd
bvrEGFRvIII-ecd
e12 g10 d8 a1 d3
0.212 0.094 0.105 0.251 0.238
0.733 0.447 0.269 0.666 0.422
DISCUSSION Development of tumor targeting agents has often focused on monoclonal antibodies, but these molecules are relatively large and have poor tumor penetration and diffusion, which limits their utility as imaging agents (10 – 12). In addition, monoclonal antibodies can be immunogenic, and their specificity is usually suboptimal (10,13– 16). A different approach is to design novel small molecules (⬍1,500 Da) with higher specificity, more rapid tumor penetration, and less immunogenicity than larger molecules (15,17). A powerful technique for ligand discovery involves the use of phage display libraries. These
Academic Radiology, Vol 9, No 8, August 2002
libraries contain a vast array of random-sequence small peptides (approximately 107–109) that can be rapidly screened in an effort to select high-affinity molecules targeted against a protein. Specific peptides can be partitioned by means of affinity selection, amplified in bacteria, characterized with ELISA techniques, and then sequenced. In addition, given the size of peptides isolated with phage display (seven to 16 amino acid residues), it is possible to identify the specific peptide amino acid side-chain functional groups required for binding. For the current study, our work has focused on a tumor-specific target, EGFRvIII (4 – 6). This cell surfaceexpressed protein has been shown to be prevalent in gliomas, in breast and ovarian cancer, and in non–small cell lung cancer (5,18,19). Notably, EGFRvIII has never been detected in normal tissues, making it truly tumor specific. EGFRvIII is the result of an in-frame deletion of exons 2–7 from the native EGFR, which results in a gene product lacking 268 amino acid residues of the extracellular domain (8) and produces a unique tumor-specific receptor epitope located at the deletion junction. The in vivo expression of EGFRvIII in tumors has been estimated at 105–106 receptors per cell. This number should be sufficient for the receptor to be used as a diagnostic imaging target (20). The preliminary results presented here demonstrate that such a systematic screening process can be used as a first step to isolate potential peptide ligands against a tumor target. Although we recognize that the binding affinity of peptides discovered with this technique may initially be less than optimal for tumor imaging, the approach permits the identification of motifs that may be useful in constructing a more focused library. These focused libraries can then be used to screen the target again rapidly, with considerable improvements in affinity. The feasibility of using peptides in this size range as imaging agents is substantiated by octreotide, a widely used imaging agent that exhibits very favorable affinity, stability, and internalization (21–23). While we are using EGFRvIII as a model system, our goal is to develop a rapid and efficient system for designing novel small peptide imaging agents for any tumor target. Research over the past several years has helped optimize a number of different methods of phage library screening. This flexibility gives the phage display platform the ability to be adapted to any target. Such translational research could have a important effect on patient care and provide a more efficient, cost-effective way to manage patients with cancer.
DEVELOPMENT OF NOVEL TUMOR IMAGING AGENTS
The innovative model system, targeting a tumor-specific epitope with phage display libraries, may offer a powerful method for designing a series of new tumor imaging agents. It should lead eventually to the production of a panel of tumor-specific ligands that could be combined to create a noninvasive molecular imaging “tumor profile.” This profile will reflect the biologic behavior of the tumor and replace traditional morphologic classification. Such molecular analysis will have an important effect on the diagnosis, staging, and stratification of patients for treatment protocols while providing prognostic information. The new direction in noninvasive imaging should lead to better treatment strategies, while decreasing morbidity and improving survival. REFERENCES 1. Cook GJR, Maisey MN. The current status of clinical PET imaging. Clin Radiol 1996; 51:603– 613. 2. Patz EF, Lowe VJ, Hoffman JM, et al. Focal pulmonary abnormalities: evaluation with F-18 fluorodeoxyflucose PET scanning. Radiology 1993; 188:487– 490. 3. Sazon DAD, Santiago SM, Hoo GWS, et al. Fluorodeoxyglucosepositron emission tomography in the detection and staging of lung cancer. Am J Respir Crit Care Med 1996; 153:417– 421. 4. Ekstrand AJ, Sugawa N, James CD, Collins VP. Amplified and rearranged epidermal growth factor receptor genes in human glioblastomas reveal deletions of sequences encoding portions of the Nand/or C-terminal tails. Proc Natl Acad Sci U S A 1992; 89:4309 – 4313. 5. Moscatello DK, Holgado-Madruga M, Godwin AK, et al. Frequent expression of a mutant epidermal growth factor receptor in multiple human tumors. Cancer Res 1995; 55:5536 –5539. 6. Wong AJ, Ruppert JM, Bigner SH, et al. Structural alterations of the epidermal growth factor receptor gene in human gliomas. Proc Natl Acad Sci U S A 1992; 89:2965–2969. 7. Reist CJ, Garg PK, Alston KL, Bigner DD, Zalutsky MR. Radioiodination of internalizing monoclonal antibodies using n-succinimidyl 5-iodo-3-pyridinecarboxylate. Cancer Res 1996; 56:4970 – 4977. 8. Lorimer IA, Keppler-Hafkemeyer A, Beers RA, Pegram CN, Bigner DD, Pastan I. Recombinant immunotoxins specific for a mutant epidermal growth factor receptor: targeting with a single chain antibody variable domain isolated by phage display. Proc Natl Acad Sci U S A 1996; 93:14815–14820. 9. Campa MJ, Kuan CT, O’Connor-McCourt MD, Bigner DD, Patz EF. Design of a novel small peptide targeted against a tumor-specific receptor. Biochem Biophys Res Commun 2000; 275:631– 636. 10. Berkower I. The promise and pitfalls of monoclonal antibody therapeutics. Curr Opin Biotech 1996; 7:622– 628. 11. Jain RK. Vascular and interstitial physiology of tumours: role in cancer detection and treatment. In: Bicknell R, Lewis CE, Ferrara N, eds. Tumour angiogenesis. Oxford, England: Oxford University Press, 1997; 45–59. 12. Rapley R. The biotechnology and applications of antibody engineering. Molec Biotech 1995; 3:139 –154. 13. Jain RK. Vascular and interstitial barriers to delivery of therapeutic agents in tumors. Cancer Metast Rev 1990; 9:253–266. 14. Kuus-Reichel K, Grauer LS, Karavodin LM, Knott C, Krusemeier M, Kay NE. Will immunogenicity limit the use, efficacy, and future development of therapeutic monoclonal antibodies? Clin Diagn Lab Immunol 1994; 1:365–372. 15. Teicher BA, ed. Physiological resistance to the treatment of solid tumors. In: Drug resistance in oncology. New York, NY: Dekker, 1993.
931
CAMPA ET AL
16. Winter G, Harris WJ. Humanized antibodies. Trends Pharmacol Sci 1993; 14:139 –143. 17. Reubi JC, Lamberts SJW, Krenning EP. Receptor imaging of human diseases using radiolabeled peptides. J Recept Signal Transduction Res 1995; 15:379 –392. 18. Wikstrand CJ, Hale LP, Batra SK, et al. Monoclonal antibodies against egfrviii are tumor specific and react with breast and lung carcinomas and malignant gliomas. Cancer Res 1995; 55:3140 –3148. 19. Garcia de Palazzo IE, Adams GP, Sundareshan P, et al. Expression of mutated epidermal growth factor receptor by non-small cell lung carcinomas. Cancer Res 1993; 53:3217–3220. 20. Wikstrand CJ, McLendon RE, Friedman AH, Bigner DD. Cell surface localization and density of the tumor-associated variant of the epider-
932
Academic Radiology, Vol 9, No 8, August 2002
mal growth factor receptor, EGFRvIII. Cancer Res 1997; 57:4130 – 4140. 21. Krenning EP, Kwekkeboom DJ, Reubi JC, et al. 111In-octreotide scintigraphy in oncology. Cancer 1993; 54(suppl 1):84 – 87. 22. Shi W, Johnston CF, Buchanan KD, et al. Localization of neuroendocrine tumours with [111In] DTPA-octreotide scintigraphy (Octreoscan): a comparative study with CT and MR imaging. QJM 1998; 91:295– 301. 23. Oomen SP, Hofland LJ, Lamberts SW, Lowenberg B, Touw IP. Internalization-defective mutants of somatostatin receptor subtype 2 exert normal signaling functions in hematopoietic cells. FEBS Lett 2001; 503:163–167.
Development of Novel Tumor Imaging Agents with Phage-Display Combinatorial Peptide Libraries1 Michael J. Campa, PhD, Scott B. Serlin, BA, Edward F. Patz, Jr, MD
Rationale and Objectives. Current radiologic methods do not provide sufficient information for unambiguous diagnosis and prognosis of cancer. The present investigation sought to address this deficiency by developing a system for designing novel small molecules targeted against tumor-specific molecules for use as radionuclide imaging agents. Materials and Methods. Part of a tumor-specific receptor, purified recombinant epidermal growth factor receptor (EGFR), variant III, extracellular domain (rEGFRvIII-ecd), was used as the target in the selection of EGFRvIII-specific peptide ligands from random peptide bacteriophage (phage) display libraries. After three rounds of screening, phage isolates were tested for binding affinity with an enzyme-linked immunosorbent assay. Positive phage were sequenced, and the peptides were synthesized and tested for binding affinity with a surface plasmon resonance assay. Results. Affinity screening identified 49 peptide-expressing phage that showed enhanced binding to the variant receptor compared with wild-type EGFR. Free peptides from the two phage isolates exhibiting the most favorable binding were tested for target binding. One of these demonstrated a binding affinity for rEGFRvIII-ecd in the 30-nmol/L range. Conclusion. These data suggest that phage display libraries may be very useful in the design of novel, high-affinity tumor imaging agents. Key Words. Neoplasms, diagnosis. ©
AUR, 2002
Radiologic studies play an integral role in the evaluation of cancer patients, but conventional modalities in current use have clear limitations. Most imaging studies provide a tremendous amount of anatomic and morphologic information, but abnormalities often remain indeterminate after standard evaluation, and patients must often proceed to an invasive procedure for tissue diagnosis. In addition, conventional imaging offers minimal prognostic information. Some tumors have an indolent course, while others are aggressive, and there is usually no way to differentiate these lesions that are radiologically and Acad Radiol 2002; 9:927–932 1 From the Department of Radiology, Duke University Medical Center, Box 3808, Durham, NC 27710. Received April 9, 2002; revision requested April 15; revision received April 16; accepted April 17. Supported in part by grant R21 CA82899-01 from the National Cancer Institute. Address correspondence to E.F.P.
©
AUR, 2002
often pathologically similar but biologically different. If the biologic behavior could be distinguished noninvasively, then patients could be stratified into the most appropriate treatment protocols, with possible improvements in outcome. Over the past several years [18F]-2-fluoro-2-deoxy-Dglucose positron emission tomography has emerged as an additional study used to address some limitations of conventional imaging. While it is an interesting model that provides biologic information (glucose metabolism), the specificity and characterization capabilities of fluorodeoxyglucose are less than optimal (1–3). Thus, the search continues for more accurate tumor imaging agents. We have taken a new approach in the effort to design novel tumor-specific molecular imaging agents. For the current investigation we chose to use random peptide phage display libraries as a ligand discovery technique, while targeting a genomically mutated, tumor-specific
927
CAMPA ET AL
epidermal growth factor receptor (EGFR), EGFR-variant III (EGFRvIII) (4 – 6). Once such a model system has been developed, new molecular imaging agents can also be produced for a variety of tumor targets. MATERIALS AND METHODS Affinity-Selection Target We began the initial phage library screens with targeting against Escherichia coli– expressed, purified recombinant EGFRvIII, extracellular domain (rEGFRvIII-ecd). We chose this recombinant protein as a target for our phage library screens because it was shown by immunologic methods to contain structural features consistent with authentic tumor-expressed EGFRvIII (7). Phage Display Libraries We used random peptide phage display libraries representing several different random peptide structural formats. Eight libraries were of the form (X)6-Z-(X)6, where X represents a random amino acid residue and Z a fixed residue. Each library possesses one fixed central residue that provides a common structural constraint across all sequences in the library. Each library, therefore, displays random sequences in different structural contexts. We used libraries containing all classes of amino acid side chains (G, D, F, H, K, L, P, and W). We also employed a random 16-mer library, as well as two different cysteinecontaining 16-mer libraries. Expressed peptides containing a minimum of two cysteine residues are held in a relatively rigid confirmation by disulfide bridges between the cysteine side chains. These constrained libraries have been shown to allow the identification of higher-affinity peptides compared with peptide libraries containing no cysteine residues. Affinity-Selection Screening A purified affinity-selection target protein (2.5 g of EGFRvIII, extracellular domain [rEGFRvIII-ecd] in 100 L of 100 mmol/L sodium bicarbonate, pH 8.5) is immobilized in microtiter wells with overnight incubation at 4°C. Additional binding sites are then blocked by incubation for 1 hour at room temperature in 10 mg/mL bovine serum albumin in a sodium bicarbonate buffer. The wells are then washed four times with phosphate-buffered saline containing 0.1% Tween 20 (PBST), and approximately 1010 plaque-forming units (pfu) from each phage library are added to the target-coated well. After a 2-hour incubation, nonbinding phage are removed by washing the
928
Academic Radiology, Vol 9, No 8, August 2002
wells five times with PBST, and bound phage are eluted with 50 mmol/L glycine– hydrochloric acid (pH 2.0). Eluted phage are neutralized with 200 mmol/L sodium phosphate buffer (pH 8.5), titered, and amplified in TG-1 E coli. Amplified phage (1010 pfu) are then used for the second and third rounds of affinity selection, performed as in the initial round. After the third round, eluted phage are plated out, and individual plaques are amplified in TG-1 E coli and tested for target protein binding by means of enzyme-linked immunosorbent assay (ELISA). ELISA Target-binding Assay Amplified phage from each plaque are first tested for binding to rEGFRvIII-ecd and to the blocking agent, BSA. Phage demonstrating enhanced binding to the target protein are then tested against rEGFRvIII-ecd and wildtype recombinant EGFR-ecd to gauge the specificity of binding. The assay is carried out by incubating phage in microtiter wells containing immobilized target protein, removing nonbinding phage by washing, and detecting bound phage with a horseradish peroxidase-conjugated anti-M13 antibody and ABTS (2,2⬘-azinobis [3-ethylbenzthiazoline-6-sulfonic acid]) and hydrogen peroxide. The colorimetric reaction is quantitated by measuring the absorbance at 405 nm. BIAcore Assay Phage isolates and free peptides were also tested for binding to rEGFRvIII-ecd by surface plasmon resonance (SPR) in a BIAcore analysis unit (Pharmacia, Piscataway, NJ). For this analysis, purified target proteins are covalently attached to a gold-coated chip located inside a BIAcore flow cell. A solution containing phage at a concentration of 1012 pfu/mL in 150 mmol/L sodium chloride, 3.4 mmol/L ethylenediaminetetraacetic acid, 0.005% (vol/vol) Triton X-100, and 10 mmol/L HEPES (pH 7.4) is then directed across the chip at a flow rate of 10 L/ min. Any phage binding to the immobilized target will increase the total mass of protein attached to the chip, which, in turn, alters the refraction angle of light directed at the nontarget side of the chip. The change in light refraction is quantitated, allowing phage binding to be detected in real time. RESULTS After three rounds of affinity selection against rEGFRvIII-ecd with seven phage libraries of the form (X)6-Z(X)6, 48 phage isolates from each library were tested for
DEVELOPMENT OF NOVEL TUMOR IMAGING AGENTS
Academic Radiology, Vol 9, No 8, August 2002
Table 1 Results of Affinity Selection Library
Positive Isolates
Proline Aspartic acid Phenylalanine Histidine Lysine Leucine Tryptophan
3 (6.1) 12 (24.5) 6 (12.2) 6 (12.2) 10 (20.4) 7 (14.3) 5 (10.2)
Figure 1. Phage isolates were tested against equimolar amounts of immobilized rEGFRvIII-ecd, rEGFR-ecd, and BSA by means of ELISA. MR-1 is an EGFRvIII-specific phagemid and is included as a positive control. Values shown are the means of duplicate determinations.
target binding with ELISA. The results (Table 1) demonstrated that 49 isolates representing all seven libraries bound the target protein in preference to BSA. These isolates were subsequently tested by means of ELISA against rEGFRvIII-ecd and recombinant wild-type EGFR-ecd to determine binding specificity. All 49 isolates exhibited varying degrees of binding specificity for rEGFRvIII-ecd. The results for 10 isolates are shown in Figure 1. We used the EGFRvIII-specific single-chain Fv phagemid MR1 as a positive control in these assays (8). To verify the ELISA data and to gauge the relative avidities of the various isolates for the target protein, we also tested all isolates by SPR against immobilized rEGFRvIII-ecd. The phenomenon of SPR occurs when surface plasmon waves are excited at a metal-liquid interface. Light is directed at, and reflected from, the side of the surface not in contact with the sample, and SPR causes a reduction in reflected light intensity at a specific combination of angle and wavelength. Biomolecular bind-
ing events cause changes in the refractive index at the surface layer, which are detected as changes in the SPR signal. The magnitude of these SPR changes, measured in resonance units, is proportional to the quantity of bound molecules. Relative avidity for the rEGFRvIII-ecd target was determined by comparing the number of resonance units bound among all phage isolates. Using this method, we observed phage binding to immobilized rEGFRvIIIecd that varied over a 20-fold range. Representative SPR plots for a high-binding and a low-binding phage are shown in Figure 2. Random inserts of the phage plasmids from the two best binders were then sequenced, and peptides were synthesized with an amino-terminal biotin moiety attached to a spacer of six glycine residues. The peptides were then tested for binding to immobilized wildtype or variant receptor in the BIAcore assay. The results of the BIAcore studies demonstrated a difference in binding preferences for the two peptides tested. Peptide 1 exhibited a slightly faster association rate with the wild-type receptor (ka ⫽ 5.44 ⫻ 104 vs 3.61 ⫻ 104 M⫺1 䡠 sec⫺1), but it dissociated nearly three times more slowly (kd ⫽ 3.08 ⫻ 10⫺3 vs 1.10 ⫻ 10⫺3 sec⫺1), resulting in an overall higher binding affinity for the variant receptor (KD ⫽ 30.5 nM for EGFRvIII vs 56.6 nM for wild-type EGFR). In contrast, peptide 2 exhibited both a faster on-rate (ka ⫽ 2.53 ⫻ 104 vs 1.73 ⫻ 104 M⫺1 䡠 sec⫺1) and a slower off-rate (kd ⫽ 1.41 ⫻ 10⫺3 vs 2.51 ⫻ 10⫺3 sec⫺1) with the wild-type receptor. The result is that peptide 2 bound wild-type EGFR more tightly than it bound EGFRvIII (KD ⫽ 145 nM for EGFRvIII vs 55.5 nM for wild-type EGFR). In contrast to the SPR results obtained with rEGFRvIII-ecd and rEGFR-ecd used as targets, neither peptide showed binding above background to solubilized cellexpressed receptor in a pull-down assay (9) or to the receptor on intact cells by using fluorescence-activated cell sorting, or FACS. We hypothesized that the bacterial expressed rEGFRvIII-ecd did not faithfully reproduce structural characteristics of the cell-expressed protein crucial for peptide recognition. Therefore, we elected to carry out additional phage library screens by using a baculovirusbased eukaryotic expression system for production of the target protein. We expected that glycosylation and folding of the purified protein in this expression system would be more representative of the authentic tertiary structure of EGFRvIII in tumor cells. We used a phage library of the form C-(X)16 for our initial screens with the baculovirus-expressed rEGFRvIIIecd (bvrEGFRvIII-ecd). The cysteine residue at the amino
929
CAMPA ET AL
Academic Radiology, Vol 9, No 8, August 2002
Figure 2. Phage isolates diluted to 1012 pfu/mL and tested in the SPR assay against immobilized rEGFRvIII-ecd or negative control protein (mouse immunoglobulin G). All determinations were carried out in 10 mmol/L HEPES, pH 7.4, 150 mmol/L sodium chloride, 3.4 mmol/L ethylenediaminetetraacetic acid, 0.005% (vol/vol) Tween 20 with an analyte flow rate of 10 L/min. In all the above plots, analyte (phage) injection begins approximately 115 seconds into the run and continues until 350 seconds, with the ligand-analyte association phase occurring between these two points. The dissociation phase is from approximately 350 to 700 seconds. The plots show results from two phage isolates, F6 and W46, representing high-binding and low-binding phage. The y axis (resonance units) is a measure of the quantity of phage associated with immobilized rEGFRvIII-ecd or immunoglobulin G. Solid line ⫽ EBFRvIII, dashed line ⫽ immunoglobulin G.
terminus of the random sequence has the potential to form a disulfide bond with a second cysteine residue located at any position in the random segment, imparting a high degree of flexibility to the overall three-dimensional structure of the selected peptides. We performed three rounds of affinity selection against bvrEGFRvIII-ecd and tested individual isolates against immobilized bvrEGFRvIII-ecd and bvrEGFR-ecd. Of 192 phage isolates tested, five showed a binding preference for bvrEGFRvIII-ecd over bvrEGFR-ecd, as judged by ELISA. Table 2 shows the optical density at 405 nm resulting from phage binding to bvrEGFR-ecd or bvrEGFRvIII-ecd. We are continuing to evaluate additional phage isolates for target-specific binding. Phage isolates demonstrating bvrEGFRvIII-ecd binding will then be tested for EGFRvIII binding in whole cells. Since EGFRvIII-binding peptide ligands are intended for in vivo use as tumor targeting agents, it is critical that phage binding in whole cells be verified before synthesis of free peptides and affinity determinations. Peptides exhibiting high-affinity (KD ⱕ 10 nM) binding to cell-expressed EGFRvIII will be evaluated further for their suitability as targeting agents, whereas EGFRvIII-binding peptides of lower affinity will serve as the starting point for the synthesis and screening of focused phage display libraries.
930
Table 2 Optical Densities Resulting from Phage Binding Optical Density at 405 nm Phage
bvrEGFR-ecd
bvrEGFRvIII-ecd
e12 g10 d8 a1 d3
0.212 0.094 0.105 0.251 0.238
0.733 0.447 0.269 0.666 0.422
DISCUSSION Development of tumor targeting agents has often focused on monoclonal antibodies, but these molecules are relatively large and have poor tumor penetration and diffusion, which limits their utility as imaging agents (10 – 12). In addition, monoclonal antibodies can be immunogenic, and their specificity is usually suboptimal (10,13– 16). A different approach is to design novel small molecules (⬍1,500 Da) with higher specificity, more rapid tumor penetration, and less immunogenicity than larger molecules (15,17). A powerful technique for ligand discovery involves the use of phage display libraries. These
Academic Radiology, Vol 9, No 8, August 2002
libraries contain a vast array of random-sequence small peptides (approximately 107–109) that can be rapidly screened in an effort to select high-affinity molecules targeted against a protein. Specific peptides can be partitioned by means of affinity selection, amplified in bacteria, characterized with ELISA techniques, and then sequenced. In addition, given the size of peptides isolated with phage display (seven to 16 amino acid residues), it is possible to identify the specific peptide amino acid side-chain functional groups required for binding. For the current study, our work has focused on a tumor-specific target, EGFRvIII (4 – 6). This cell surfaceexpressed protein has been shown to be prevalent in gliomas, in breast and ovarian cancer, and in non–small cell lung cancer (5,18,19). Notably, EGFRvIII has never been detected in normal tissues, making it truly tumor specific. EGFRvIII is the result of an in-frame deletion of exons 2–7 from the native EGFR, which results in a gene product lacking 268 amino acid residues of the extracellular domain (8) and produces a unique tumor-specific receptor epitope located at the deletion junction. The in vivo expression of EGFRvIII in tumors has been estimated at 105–106 receptors per cell. This number should be sufficient for the receptor to be used as a diagnostic imaging target (20). The preliminary results presented here demonstrate that such a systematic screening process can be used as a first step to isolate potential peptide ligands against a tumor target. Although we recognize that the binding affinity of peptides discovered with this technique may initially be less than optimal for tumor imaging, the approach permits the identification of motifs that may be useful in constructing a more focused library. These focused libraries can then be used to screen the target again rapidly, with considerable improvements in affinity. The feasibility of using peptides in this size range as imaging agents is substantiated by octreotide, a widely used imaging agent that exhibits very favorable affinity, stability, and internalization (21–23). While we are using EGFRvIII as a model system, our goal is to develop a rapid and efficient system for designing novel small peptide imaging agents for any tumor target. Research over the past several years has helped optimize a number of different methods of phage library screening. This flexibility gives the phage display platform the ability to be adapted to any target. Such translational research could have a important effect on patient care and provide a more efficient, cost-effective way to manage patients with cancer.
DEVELOPMENT OF NOVEL TUMOR IMAGING AGENTS
The innovative model system, targeting a tumor-specific epitope with phage display libraries, may offer a powerful method for designing a series of new tumor imaging agents. It should lead eventually to the production of a panel of tumor-specific ligands that could be combined to create a noninvasive molecular imaging “tumor profile.” This profile will reflect the biologic behavior of the tumor and replace traditional morphologic classification. Such molecular analysis will have an important effect on the diagnosis, staging, and stratification of patients for treatment protocols while providing prognostic information. The new direction in noninvasive imaging should lead to better treatment strategies, while decreasing morbidity and improving survival. REFERENCES 1. Cook GJR, Maisey MN. The current status of clinical PET imaging. Clin Radiol 1996; 51:603– 613. 2. Patz EF, Lowe VJ, Hoffman JM, et al. Focal pulmonary abnormalities: evaluation with F-18 fluorodeoxyflucose PET scanning. Radiology 1993; 188:487– 490. 3. Sazon DAD, Santiago SM, Hoo GWS, et al. Fluorodeoxyglucosepositron emission tomography in the detection and staging of lung cancer. Am J Respir Crit Care Med 1996; 153:417– 421. 4. Ekstrand AJ, Sugawa N, James CD, Collins VP. Amplified and rearranged epidermal growth factor receptor genes in human glioblastomas reveal deletions of sequences encoding portions of the Nand/or C-terminal tails. Proc Natl Acad Sci U S A 1992; 89:4309 – 4313. 5. Moscatello DK, Holgado-Madruga M, Godwin AK, et al. Frequent expression of a mutant epidermal growth factor receptor in multiple human tumors. Cancer Res 1995; 55:5536 –5539. 6. Wong AJ, Ruppert JM, Bigner SH, et al. Structural alterations of the epidermal growth factor receptor gene in human gliomas. Proc Natl Acad Sci U S A 1992; 89:2965–2969. 7. Reist CJ, Garg PK, Alston KL, Bigner DD, Zalutsky MR. Radioiodination of internalizing monoclonal antibodies using n-succinimidyl 5-iodo-3-pyridinecarboxylate. Cancer Res 1996; 56:4970 – 4977. 8. Lorimer IA, Keppler-Hafkemeyer A, Beers RA, Pegram CN, Bigner DD, Pastan I. Recombinant immunotoxins specific for a mutant epidermal growth factor receptor: targeting with a single chain antibody variable domain isolated by phage display. Proc Natl Acad Sci U S A 1996; 93:14815–14820. 9. Campa MJ, Kuan CT, O’Connor-McCourt MD, Bigner DD, Patz EF. Design of a novel small peptide targeted against a tumor-specific receptor. Biochem Biophys Res Commun 2000; 275:631– 636. 10. Berkower I. The promise and pitfalls of monoclonal antibody therapeutics. Curr Opin Biotech 1996; 7:622– 628. 11. Jain RK. Vascular and interstitial physiology of tumours: role in cancer detection and treatment. In: Bicknell R, Lewis CE, Ferrara N, eds. Tumour angiogenesis. Oxford, England: Oxford University Press, 1997; 45–59. 12. Rapley R. The biotechnology and applications of antibody engineering. Molec Biotech 1995; 3:139 –154. 13. Jain RK. Vascular and interstitial barriers to delivery of therapeutic agents in tumors. Cancer Metast Rev 1990; 9:253–266. 14. Kuus-Reichel K, Grauer LS, Karavodin LM, Knott C, Krusemeier M, Kay NE. Will immunogenicity limit the use, efficacy, and future development of therapeutic monoclonal antibodies? Clin Diagn Lab Immunol 1994; 1:365–372. 15. Teicher BA, ed. Physiological resistance to the treatment of solid tumors. In: Drug resistance in oncology. New York, NY: Dekker, 1993.
931
CAMPA ET AL
16. Winter G, Harris WJ. Humanized antibodies. Trends Pharmacol Sci 1993; 14:139 –143. 17. Reubi JC, Lamberts SJW, Krenning EP. Receptor imaging of human diseases using radiolabeled peptides. J Recept Signal Transduction Res 1995; 15:379 –392. 18. Wikstrand CJ, Hale LP, Batra SK, et al. Monoclonal antibodies against egfrviii are tumor specific and react with breast and lung carcinomas and malignant gliomas. Cancer Res 1995; 55:3140 –3148. 19. Garcia de Palazzo IE, Adams GP, Sundareshan P, et al. Expression of mutated epidermal growth factor receptor by non-small cell lung carcinomas. Cancer Res 1993; 53:3217–3220. 20. Wikstrand CJ, McLendon RE, Friedman AH, Bigner DD. Cell surface localization and density of the tumor-associated variant of the epider-
932
Academic Radiology, Vol 9, No 8, August 2002
mal growth factor receptor, EGFRvIII. Cancer Res 1997; 57:4130 – 4140. 21. Krenning EP, Kwekkeboom DJ, Reubi JC, et al. 111In-octreotide scintigraphy in oncology. Cancer 1993; 54(suppl 1):84 – 87. 22. Shi W, Johnston CF, Buchanan KD, et al. Localization of neuroendocrine tumours with [111In] DTPA-octreotide scintigraphy (Octreoscan): a comparative study with CT and MR imaging. QJM 1998; 91:295– 301. 23. Oomen SP, Hofland LJ, Lamberts SW, Lowenberg B, Touw IP. Internalization-defective mutants of somatostatin receptor subtype 2 exert normal signaling functions in hematopoietic cells. FEBS Lett 2001; 503:163–167.