Engineered Knottin Peptides: A New Class of Agents for Imaging Integrin Expression in Living Subjects

There is a critical need for molecular imaging agents to detect cell surface integrin receptors that are present in human cancers. Previously, we used directed evolution to engineer knottin peptides that bind with high affinity (∼10 to 30 nmol/L) to integrin receptors that are overexpressed on the surface of tumor cells and the tumor neovasculature. To evaluate these peptides as molecular imaging agents, we site-specifically conjugated Cy5.5 or ⁶⁴Cu-1,4,7,10-tetra-azacyclododecane-N,N′,N″,N‴-tetraacetic acid (DOTA) to their N termini, and used optical and positron emission tomography (PET) imaging to measure their uptake and biodistribution in U87MG glioblastoma murine xenograft models. NIR fluorescence and microPET imaging both showed that integrin binding affinity plays a strong role in the tumor uptake of knottin peptides. Tumor uptake at 1 hour postinjection for two high-affinity (IC₅₀, ∼20 nmol/L) ⁶⁴Cu-DOTA–conjugated knottin peptides was 4.47% ± 1.21% and 4.56% ± 0.64% injected dose/gram (%ID/g), compared with a low-affinity knottin peptide (IC₅₀, ∼0.4 μmol/L; 1.48 ± 0.53%ID/g) and c(RGDyK) (IC₅₀, ∼1 μmol/L; 2.32 ± 0.55%ID/g), a low-affinity cyclic pentapeptide under clinical development. Furthermore, ⁶⁴Cu-DOTA–conjugated knottin peptides generated lower levels of nonspecific liver uptake (∼2%ID/g) compared with c(RGDyK) (∼4%ID/g) 1 hour postinjection. MicroPET imaging results were confirmed by in vivo biodistribution studies. ⁶⁴Cu-DOTA–conjugated knottin peptides were stable in mouse serum, and in vivo metabolite analysis showed minimal degradation in the blood or tumor upon injection. Thus, engineered integrin-binding knottin peptides show great potential as clinical diagnostics for a variety of cancers. [Cancer Res 2009;69(6):2435–42]

Cell surface receptors that are selectively expressed in human malignancies have generated much interest as potential targets for molecular therapeutics and diagnostic agents. Integrins are a family of extracellular matrix adhesion receptors that noncovalently associate into α/β heterodimers with distinct ligand binding specificities (1). Several integrins, including α_vβ₃, α_vβ₅, and α₅β₁, have been shown to be expressed on the surface of cancer cells and the tumor neovasculature (2–5). These integrins have been proposed to mediate angiogenesis, tumor growth, and metastasis (6–8), making them attractive targets for therapeutic intervention. One strategy for inhibiting tumor angiogenesis is to administer agents that will bind to these specific integrin receptors with high affinity and block their function (9–11). This approach highlights a need for noninvasive in vivo imaging probes that can be used to identify patients who will best respond to these integrin-targeted therapies and to monitor disease progression (12–14). In addition, the use of integrins as biomarkers in molecular imaging applications will be important for the early detection of cancer.

Many integrins, including those containing the α_v subunit, as well as α₅β₁ and α_iibβ₃ integrins, recognize an Arg-Gly-Asp (RGD) motif found in extracellular matrix protein ligands (15). In these ligands, the RGD motif is typically found in solvent-exposed loops, and the structural context of this loop, dictated by the residues flanking the RGD sequence, determines integrin binding affinity and specificity (16). Rational drug design and phage display have generated peptides and peptidomimetics containing the RGD sequence that target α_vβ₃ (and α_vβ₅) integrins or α₅β₁ integrin (4, 10, 15). However, chemical modifications that can be made to these small peptides to improve their receptor binding affinity, tumor uptake, and in vivo pharmacokinetics are limited. Moreover, covalent attachment of imaging probes have affected their integrin binding properties (14). Therefore, despite the prevalence of integrin-binding peptides and peptidomimetics in the literature, suboptimal tumor targeting efficacy and pharmacokinetics have limited their clinical translation as molecular imaging agents (13). Only one compound, a glycosylated cyclin RGD pentapeptide [¹⁸F-Galacto-c(RGDfK)], has advanced to the clinical level for molecular imaging in human subjects (17, 18). Although this agent was able to identify integrin-positive lesions in human subjects and image intensity correlated with α_vβ₃ integrin expression, its relatively poor tumor uptake and higher background (e.g., liver) indicates there is room for substantial improvements. Moreover, in preclinical studies using a M21 human melanoma mouse xenograft model, ¹⁸F-Galacto-c(RGDfK) exhibited low tumor uptake values of 1.6 ± 0.2 injected dose/gram (%ID/g; ref. 19), suggesting that imaging agents that have relatively low tumor contrast in small animal models will likely correlate to low imaging contrast in humans.

To develop a new class of agents to image integrin expression in vivo, we engineered small (∼3 kDa), conformationally constrained peptides that bind to α_vβ₃/α_vβ₅ or α_vβ₃/α_vβ₅/α₅β₁ integrins with high affinity.³

We incorporated RGD-based integrin binding motifs into a surface-exposed loop of the Ecballium elaterium trypsin inhibitor (20), a cystine knot peptide from the squash family of protease inhibitors (Fig. 1A). This provided a rigid structural framework that enabled us to engineer peptides with high integrin binding affinities (IC₅₀, ∼10–30 nmol/L) through directed evolution. Cystine knot peptides (also known as knottins) are ideal for in vivo tumor-targeting applications as they possess a disulfide-bonded core that confers outstanding proteolytic resistance and thermal stability; moreover, they have been shown to be nonimmunogenic (21–24). Due to their small size, knottin peptides can be produced by chemical synthesis, which allows for site-specific conjugation of imaging probes, radioisotopes, or chemotherapeutic agents for direct delivery to cancer cells. In addition, knottin peptides can be chemically modified to tailor their in vivo pharmacokinetic properties for diverse clinical applications.

In this study, we conjugated optical and positron emission tomography (PET) imaging probes to our engineered integrin-binding knottin peptides to evaluate their potential as in vivo integrin imaging agents. We tested the ability of knottin peptides with varying integrin binding affinities to target tumors in small animal xenograft models using NIR fluorescence and microPET imaging. We compared these results to those generated with a knottin peptide containing a scrambled RGD sequence, and a cyclic RGD pentapeptide [c(RGDyK)], another monomeric, unmodified peptide currently under clinical development.

Materials, cell lines, and reagents. The U87MG human glioblastoma cell line was obtained from American Type Culture Collection. Detergent-solubilized α_vβ₃, α_vβ₅ integrin receptors (both octyl-β-d-glucopyranoside formulations) and α₅β₁ integrin (Triton X-100 formulation) were purchased from Millipore, and α_iibβ₃ (Triton X-100 formulation) was purchased from Enzyme Research Laboratories. ¹²⁵I-labeled echistatin and c(RGDyK) were purchased from Amersham Biosciences, and Peptides International, respectively. PBS was from Invitrogen. All other chemicals were purchased from Fisher Scientific unless otherwise specified. Integrin binding buffer (IBB) was composed of 25 mmol/L Tris (pH 7.4), 150 mmol/L NaCl, 2 mmol/L CaCl₂, 1 mmol/L MgCl₂, 1 mmol/L MnCl₂, and 0.1% bovine serum albumin (BSA).

Cell surface integrin receptor competition binding assay. Cell surface competition binding assays were performed as previously described (25). Briefly, 2 × 10⁵ U87MG cells were incubated with 0.06 nmol/L ¹²⁵I-labeled echistatin and varying concentrations of peptides in integrin binding buffer at room temperature for 3 h. The cell-bound radioactivity remaining after washing was determined by γ-counting. Half-maximal inhibitory concentration (IC₅₀) values were determined by nonlinear regression analysis using Kaleidagraph (Synergy Software), and are reported as the average of experiments performed on three separate days.

Solid phase integrin receptor competition binding assay. Integrin receptor competition binding assays were performed as previously described (26). Briefly, detergent-solubilized α_vβ₃, α_vβ₅, α₅β₁, and α_iibβ₃ integrin receptors were diluted to a final concentration of 1 μg/mL in integrin binding buffer. Aliquots (100 μL) were used to coat wells of Maxisorb plates (NalgeNunc; Fisher Scientific), overnight at 4°C. The wells were washed and blocked with integrin binding buffer containing 1% bovine serum albumin for 2 h at room temperature. ¹²⁵I-labeled echistatin (0.06 nmol/L) and varying concentrations of unlabeled peptides were incubated in the wells for 3 h at room temperature with gentle rocking, and washed three times in integrin binding buffer. Plate-bound radioactivity was solubilized with 200 μL of boiling 2N NaOH followed by γ-counting. Each data point represents the average value of triplicate wells.

Cy5.5 chemical conjugation. Cy5.5 monofunctional N-hydroxysuccinimide ester (Amersham Biosciences) was dissolved in a solution of 1 mL of dimethlysulfoxide and 15 μL triethylamine. Purified peptide was added to this Cy5.5 solution, and the reaction was mixed at room temperature in the dark. Cy5.5 conjugation reactions were monitored by absorbance at 675 nm by analytical scale reversed-phase high performance liquid chromatography (HPLC). Upon completion, the reaction mixtures were purified by reversed-phase HPLC. Fractions containing Cy5.5-peptide conjugates were collected, lyophilized, and redissolved in water. Peptide purity was assessed by analytical scale reversed-phase HPLC and concentrations were determined by amino acid analysis (AAA Service Laboratory). Molecular masses were confirmed with electrospray or matrix-assisted laser desorption/ionization time-of-flight mass spectrometry using Stanford core facilities (Supplementary Table S1).

1,4,7,10-tetra-azacyclododecane-N,N′,N″,N‴-tetraacetic acid chemical conjugation and ⁶⁴Cu radiolabeling. 1,4,7,10-tetra-azacyclododecane-N,N′,N′′,N′′′-tetraacetic acid (DOTA; Sigma Aldrich) was activated with 1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide (Pierce) and N-hydroxysulfonosuccinimide (Pierce) in water (pH 5.5) for 40 min at room temperature using a 1:1:1 molar ratio of DOTA/1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide/N-hydroxysulfonosuccinimide. Peptides were dissolved in 300 μL of sodium phosphate buffer [30 mmol/L (pH 8.5)], and added to the above in situ prepared sulfosuccinimidyl ester of DOTA (DOTA-OSSu). A molar excess of DOTA-OSSu was used to drive the conjugation reaction to completion. The reaction was allowed to proceed at room temperature for 1 h and mixed at 4°C overnight. The resulting DOTA-peptide conjugates were purified by reversed-phase HPLC and stored as a lyophilized solid. The product masses were verified by electrospray mass spectrometry and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Supplementary Table S1) and peptide concentrations were determined by amino acid analysis.

The DOTA-conjugated peptides (25 μg) were radiolabeled with ⁶⁴Cu by incubating with 2 to 3 mCi ⁶⁴CuCl₂ (University of Wisconsin-Madison) in 0.1 N sodium acetate (pH 6.3) for 1 h at 45°C. The reaction was terminated with the addition of EDTA. The radiolabeled complexes were purified using a PD-10 column (Amersham) or by radio-HPLC using a γ detector, dried by rotary evaporation, reconstituted in PBS, and passed through a 0.22-μm filter for animal experiments. The radiochemical purity, determined as the ratio of the main product peak to other peaks, was determined by HPLC to be >95%. The radiochemical yield, determined as the ratio of final activity of the product over the starting activity used for the reaction, was usually over 80%. At least seven radiolabeling reactions were performed for experiments run on different days.

U87MG glioblastoma xenograft mouse model. U87MG cells were maintained at 37°C in a humidified atmosphere containing 5% CO₂ in Dulbecco's modified eagle medium, 10% heat-inactivated fetal bovine serum, and penicillin-streptomycin (all from Invitrogen). Animal procedures were carried out according to a protocol by Stanford University Administrative Panels on Laboratory Animal Care. Female athymic nude mice (nu/nu), obtained at ages 4 to 6 wk (Charles River Laboratories, Inc.), were injected s.c. in the right or left shoulder with 2 × 10⁷ U87MG glioblastoma cells suspended in 100 μL of PBS. Mice were used for in vivo imaging studies when their tumors reached ∼8 to 10 millimeters in diameter.

Near-IR fluorescence and microPET imaging of U87MG xenograft tumors. Near-IR (NIR) fluorescence imaging was performed with an IVIS 200 (Xenogen) as previously described (25). Briefly, a Cy5.5 filter set was used and fluorescence emission was normalized to photons per second per centimeter squared per steradian (p/s/cm²/sr). Images were acquired and analyzed using Living Image 2.5 (Xenogen). For all experiments, mice (n = 3 for each probe) were injected via tail vein with 1.5 nmol of Cy5.5-labeled peptides in 100 μL PBS and imaged at various times postinjection. Tumor contrast was quantified by drawing identically sized regions of interest around the tumor (T) and normal (N) tissue located in the mouse's flank (average radiance, p/s/cm²/sr). Data points represent the average tumor-to-normal tissue ratio (T/N) for a group of three animals.

U87MG tumor-bearing mice (n = 3 or more for each probe) were injected with ∼100 μCi of ⁶⁴Cu-DOTA–conjugated peptides via the tail vein and imaged with a microPET R4 rodent model scanner (Siemens Medical) using 3- or 5-min static scans. For blocking experiments, mice were coinjected with 330 μg (∼0.5 μmol) of unlabeled c(RGDyK). Images were reconstructed by a two dimensional ordered expectation maximum subset algorithm and calibrated as previously described (27). Regions of interests were drawn over the tumor on decay-corrected whole body images using ASIPro VM software (Siemens Medical). The mean counts per pixel per minute were obtained from the regions of interest and converted to counts per milliliter per minute with a calibration constant. Regions of interests were converted to counts/gram/min, and %ID/g values were determined assuming a tissue density of 1 gram/mL. No attenuation correction was performed.

In vivo biodistribution studies. Anesthetized nude (nu/nu) mice bearing U87MG tumor xenografts were injected with ∼100 μCi of ⁶⁴Cu-DOTA–labeled knottin peptides via the tail vein, and were euthanized at 0.5, 1, and 24 h. Blood, muscle, heart, liver, lungs, kidneys, spleen, brain, intestine, skin, stomach, pancreas, and tumor tissue were removed and weighed, and their radioactivity levels were measured by γ counting. Results are expressed as the %ID/g of tissue and represent the mean and SD of experiments performed on at least three mice. For each mouse, the activity of tissue samples was calibrated against a known aliquot of the radio-tracer and normalized to the whole bodyweight and to the residual activity present in the tail.

In vivo metabolite analysis. For metabolite analysis, anesthetized nude (nu/nu) mice bearing U87MG tumor xenografts were tail-vein injected with 200 to 400 μCi of ⁶⁴Cu-DOTA-labeled knottin peptides, and were euthanized at 1, 4, or 24 h. Blood, kidney, and tumor tissue were removed and suspended in ∼500 μL PBS. Tissues were homogenized with a mortar and pestle and the homogenate was extensively filtered using a Nanosep 10K device (Pall Corporation) to isolate low molecular weight metabolites. The filtrates were analyzed by reversed-phase HPLC under identical conditions used for analyzing the original radiolabeled compound. Eluted fractions were collected in 30 s intervals and a γ counter was used to determine counts per minute.

Statistical analysis. All data are presented as the average value ± the SD of n independent measurements. Statistical analysis for animal studies was performed by t test using Microsoft Excel or Matlab, and significance was assigned for P values of <0.05.

Knottin peptide synthesis and conjugation to Cy5.5 and DOTA. Previously, we used combinatorial methods to engineer Ecballium elaterium trypsin inhibitor mutants that bound with high affinity (IC₅₀, ∼10–30 nmol/L) to integrin receptors that are overexpressed on the tumor vasculature.⁴

The sequences of these mutants are shown in Fig. 1B. In addition, we also engineered a knottin peptide that binds with weaker integrin binding affinity (FN-RGD2; IC₅₀, ∼0.4 μmol/L) by grafting a loop from the extracellular matrix protein fibronectin into the Ecballium elaterium trypsin inhibitor. A knottin peptide that contains the scrambled sequence RDG and does not bind to integrins was used as a negative control in all studies (FN-RDG2). Knottin peptides were synthesized using Fmoc-based solid phase peptide synthesis and were purified and folded as described (Supplementary Methods). The NH₂-terminal amine of the folded, purified knottin peptides was used for site-specific attachment of Cy5.5, a NIR fluorochrome, or a DOTA chelating agent and the radionuclide ⁶⁴Cu. Modified knottin peptides were purified by reversed-phase HPLC and were characterized by mass spectrometry (Supplementary Table S1) and amino acid analysis. Analogous bioconjugation reactions were performed through the ε-amino group of the lysine residue in c(RGDyK), a backbone cyclized pentapeptide that has been extensively characterized for in vivo molecular imaging applications (reviewed in refs. 12, 28), and is included here for comparison to our knottin peptides.

Binding of Cy5.5 and DOTA-labeled knottin peptides to integrin-expressing tumor cells. The relative binding affinities of DOTA- and Cy5.5-conjugated peptides was tested and compared with unmodified peptides to determine if these modifications disrupt integrin binding interactions. Echistatin is a RGD-containing protein from snake venom that binds to α_vβ₃ integrin with a K_D of 0.36 nmol/L (29). Peptides were tested for their ability to compete for cell surface integrin binding with ¹²⁵I-labeled echistatin. U87MG glioblastoma cells, which express ∼10⁵ α_vβ₃ integrin receptors per cell (30), were used for these studies. Relative binding affinities for modified and unmodified peptides are reported as IC₅₀ values (Table 1). Knottin peptides 2.5D and 2.5F, which were obtained by directed evolution, were shown to bind to U87MG cells with a significantly stronger affinity (IC₅₀, 19 ± 6 nmol/L and 26 ± 5 nmol/L, respectively) than both the loop-grafted FN-RGD2 (IC₅₀, 370 ± 150 nmol/L) and c(RGDyK) (IC₅₀, 860 ± 400 nmol/L) peptides (Table 1). FN-RDG2 was not able to compete for ¹²⁵I-echistatin binding to U87MG cells, as expected. Next, DOTA-conjugated peptides were shown to bind to U87MG cells in a dose-dependent manner with affinities that were comparable with the unmodified peptides (Supplementary Fig. S1A; Table 1). In contrast, Cy5.5-labeled peptides showed stronger affinities (∼4- to 8-fold) to U87MG cells compared with the corresponding unmodified peptides (Supplementary Fig. S1B; Table 1). This increase in binding could be due to interactions between the hydrophobic dye molecule and cells, however, only occurs with peptides that have affinity for integrin receptors, as the Cy5.5-FN-RDG2–negative control does not exhibit nonspecific binding.

Integrin binding specificities of Cy5.5 and DOTA-labeled knottin peptides. Because U87MG cells have been shown to express α_vβ₅ and α₅β₁ integrins in addition to α_vβ₃ integrin (31), we measured integrin binding specificity by competition of ¹²⁵I-echistatin to detergent-solubilized α_vβ₃, α_vβ₅, α₅β₁, and α_iibβ₃ integrin receptors coated onto microtiter plates. Unlabeled echistatin, our positive control, bound strongly to all of the tested integrins, in agreement with previous reports (32). All RGD-containing peptides bound to α_vβ₃ and α_vβ₅ integrins to some degree, with the knottin peptides 2.5D and 2.5F showing the strongest levels of binding compared with FN-RGD2 and c(RGDyK) (Fig. 2). DOTA-conjugated FN-RDG2, our negative control, did not bind to any of the integrins used in this study (Fig. 2A). The DOTA-conjugated knottin peptide 2.5F bound with strong affinity to α₅β₁ integrin, whereas knottin 2.5D exhibited only minimal binding to this receptor. DOTA-labeled peptides did not bind to the α_iibβ₃ integrin receptor, which is important for in vivo imaging applications, because the α_iibβ₃ integrin is widely expressed on platelet cells and is involved in mediating the blood clotting process (33). Collectively, the integrin binding specificity of the DOTA-labeled peptides is identical to results obtained with unmodified peptides (data not shown). In contrast, Cy5.5-labeled peptides showed increased binding to all integrin receptors compared with unmodified and DOTA-conjugated peptides (data not shown; Fig. 2B). This was likely due to nonspecific binding of the hydrophobic Cy5.5 molecule (i.e., Cy5.5-FN-RGD2 binding to α_iibβ₃ integrins) but could also be due to increased interactions of Cy5.5-labeled peptides with integrin receptors.

Knottin peptides as in vivo optical imaging probes. We tested the ability of Cy5.5-labeled knottin peptides to target tumors in small living animals to begin to evaluate their potential as in vivo molecular imaging agents. Whole-body NIR fluorescence imaging of subcutaneous human tumor mouse xenografts was performed and the fluorescence intensity of T/N was measured as a function of time. Figure 3A shows typical NIR fluorescent images of athymic nude mice bearing subcutaneous U87MG glioblastoma tumors after tail vein injection of 1.5 nmol of Cy5.5-labeled peptides. Cy5.5-labeled knottins 2.5D and 2.5F showed increased T/N ratios compared with signals generated by both FN-RGD2 knottin peptide and c(RGDyK) peptide, which were only slightly higher than the background signal of the FN-RDG2–negative control (Fig. 3B). Because Cy5.5-FN-RDG2 does not seem to bind to U87MG cells (Table 1), the NIR fluorescence signal observed for this peptide in the tumor likely results from the extravasation of the probe from leaky tumor vasculature. Cy5.5-labeled knottins were taken up and retained by the kidneys at all time points tested (data not shown; Fig. 3A).

Knottin peptides as in vivo PET imaging probes. Next, we tested the potential of engineered knottin peptides for use as PET imaging probes in mice bearing U87MG human tumor xenografts. Higher tumor uptake was observed with ⁶⁴Cu-DOTA-2.5D and 2.5F knottins relative to ⁶⁴Cu-DOTA-FN-RGD2 and ⁶⁴Cu-DOTA-c(RGDyK) (Fig. 4A and B). In microPET imaging, tumor uptake at 1 hour postinjection for ⁶⁴Cu-DOTA-2.5D and ⁶⁴Cu-DOTA-2.5F was 4.47 ± 1.21%ID/g and 4.56 ± 0.64%ID/g, respectively, compared with ⁶⁴Cu-DOTA-FN-RGD2 (1.48 ± 0.53%ID/g) and c(RGDyK) (2.32 ± 0.55%ID/g). The knottin-based PET probes exhibited reduced liver uptake (∼2%ID/g) compared with ⁶⁴Cu-DOTA-c(RGDyK), which showed significantly higher accumulation in the liver (4.19 ± 0.78%ID/g and 3.59 ± 0.87%ID/g at 1 and 4 hours postinjection, respectively). Tumor uptake was blocked by coinjection of ⁶⁴Cu-DOTA-2.5D with a molar excess (0.5 μmol) of unlabeled c(RGDyK) (1.67 ± 0.28%ID/g at 1 hour postinjection; Fig. 4A and B). The PET signal generated by ⁶⁴Cu-DOTA-FN-RDG2 reflects the rate of extravasation from the tumor vasculature and subsequent washout, and was found to be 1.09 ± 0.48%ID/g and 0.76 ± 0.33%ID/g at 1 and 4 hours postinjection, respectively.

The biodistribution of ⁶⁴Cu-DOTA-2.5D and ⁶⁴Cu-DOTA-2.5F in various tissues and organs was determined at 0.5 and 4 hours postinjection (Fig. 4C). Both knottin peptides accumulated rapidly in tumors 0.5 hour postinjection (4.2 ± 1.1%ID/g for 2.5D and 5.3 ± 0.7%ID/g for 2.5F). Knottin 2.5F cleared from the tumor at a slower rate than knottin 2.5D (3.4 ± 0.4%ID/g versus 1.51 ± 0.02%ID/g 4 hours postinjection) in agreement with the microPET data. High tumor uptake and rapid blood clearance led to tumor-to-blood ratios of 42.3 ± 8.65 for ⁶⁴Cu-DOTA-2.5F and 24.93 ± 2.62 for ⁶⁴Cu-DOTA-2.5D at 4 hours postinjection (Fig. 4D). In contrast to Cy5.5-labeled knottin peptides, which exhibited high kidney retention, minimal amounts of radioactivity remained in the kidneys after 24 hours (1.25 ± 0.11%ID/g for ⁶⁴Cu-DOTA-2.5D and 1.09 ± 0.15%ID/g for ⁶⁴Cu-DOTA-2.5F; data not shown), indicating that conjugation of different chemical moieties to knottin peptides affected their pharmacokinetic properties. Moderate amounts of radioactivity were observed in the other major organs, including the lungs, skin, spleen, stomach, intestines (1 – 2%ID/g), and also the liver (2.3 ± 0.8%ID/g and 2.5 ± 0.2%ID/g for 2.5D and 2.5F, respectively, 0.5 hour postinjection). Lower levels of activity (<1%ID/g) were present in the blood, heart, bone, brain, and pancreas (Fig. 4C). In addition, there was minimal background signal from muscle tissue (Fig. 4C and D), further demonstrating the potential of knottin peptides as diagnostic agents to detect lesions throughout the body.

Serum stability and metabolite analysis. Finally, we tested the stability of ⁶⁴Cu-DOTA-2.5D in mouse serum and in the whole mouse blood, tumor, and kidney. First, radio-HPLC analysis was performed 1, 4, and 24 hours after peptide incubation in mouse serum at 37°C. Minimal breakdown products were observed at each time point (Supplementary Fig. S2 and Table S2). Next, the in vivo metabolic stability of ⁶⁴Cu-DOTA-2.5D in whole mouse blood, tumor tissue, and kidney tissue were determined 1 and 4 hours postinjection (Fig. 5A–C). Radio-HPLC analysis of solubilized tissue homogenates showed a major elution peak between 19.5 and 20 minutes, corresponding to the intact radiotracer. Metabolites with retention times between 4 to 6 minutes could be seen at significant levels in the kidneys after 4 hours (Fig. 5C), indicating either breakdown of the probe in the kidneys or metabolites that are generated by various organs in the animal and cleared through renal excretion. Values of % intact tracer isolated from the serum, tumor, and kidneys are summarized in Supplementary Table S2.

There is a critical need for molecular imaging probes that will specifically target integrin receptors and allow noninvasive characterization of tumors for patient-specific cancer treatment and disease management (12, 14, 34). Here, we developed engineered knottin peptides as a new class of agents for imaging integrin expression in living subjects. We determined that conjugation of Cy5.5 to the knottin peptides slightly increased their integrin binding affinity and decreased their integrin binding specificity, whereas conjugation of DOTA to the knottin peptides had no effect on integrin binding affinity or specificity (Supplementary Fig. S1; Table 1; Fig. 2). We also showed that Cy5.5- and ⁶⁴Cu-DOTA–conjugated FN-RGD2 knottin peptides, which bind to integrins with affinities in the low micromolar range, generated significantly weaker imaging signals compared with knottin peptides 2.5D and 2.5F (Figs. 3 and 4). These results strongly suggest that integrin binding affinity influences tumor uptake of knottin peptides, although other factors such as hydrophobicity can also affect tissue biodistribution. Interestingly, in PET studies knottin peptide 2.5F exhibited slower tumor washout compared with 2.5D, resulting in much higher tumor/blood ratios 4 hours postinjection (Fig. 4D). This could be due to the ability of knottin 2.5F to bind more tightly to α₅β₁ integrins compared with knottin 2.5D (Fig. 2), or potential differences in peptide hydrophobicity, charge, or off-rates of integrin receptor binding. Finally, we showed that knottin peptides were stable in vitro upon prolonged serum incubation, and in vivo in the tumor and blood during the timeframe in which imaging experiments were performed.

To evaluate the use of knottin peptides as molecular imaging agents compared with c(RGDyK), we performed in vivo experiments under identical conditions, as differences in amount of probe injected, image acquisition, and data analysis can influence tumor uptake values. Nevertheless, previously published biodistribution studies with ⁶⁴Cu-DOTA-c(RGDyK) in a similar U87MG xenograft model showed tumor uptake values of ∼2.5%ID/g at 1 hour postinjection (35), consistent with our results. In addition to increased tumor uptake, high affinity ⁶⁴Cu-DOTA–labeled knottin peptides 2.5D and 2.5F showed more favorable tissue distribution as indicated by lower liver uptake compared with ⁶⁴Cu-DOTA-c(RGDyK). Collectively, our data indicate that knottin peptides have potential as diagnostic imaging agents to monitor multiple regions of the body including the chest and abdomen.

Several strategies have been used to improve the in vivo performance of monomeric c(RGDyK) peptides as imaging agents and can also be applied in future studies to our engineered knottin peptides. Multivalent versions of c(RGDyK) peptides have been synthesized (28), and IC₅₀ values were measured for competition binding of ¹²⁵I-echistatin with monomeric (203 ± 32 nmol/L), dimeric (103 ± 14 nmol/L), tetrameric (34.6 ± 2.6 nmol/L), and octameric (10.0 ± 1.7 nmol/L) peptides on U87MG cells (36). In these studies, multivalent peptides exhibited increased integrin binding affinity, which resulted in much greater levels of tumor uptake compared to monomeric c(RGDyK). MicroPET imaging of these peptides in U87MG xenograft models indicated tumor uptake values of 9.6 ± 1.4%ID/g and 10.6 ± 0.7%ID/g for ⁶⁴Cu-DOTA-c(RGDyK) tetramer and octamer, respectively, 1 hour postinjection, with little tumor washout after 20 hours (36). One potential drawback of these monovalent or multivalent c(RGDyK) peptides is that they exhibit sustained liver uptake (∼2.5–3%ID/g; refs. 36–38). Kidney uptake was more severely affected by peptide multimerization, with the ⁶⁴Cu-DOTA–labeled octamer demonstrating >50%ID/g 1 hour postinjection (36). PEGylation has been used to improve the pharmacokinetics and tissue distribution of monomeric c(RGDyK) to address undesired liver and kidney uptake (35).

In this study, our goal was to compare the knottin peptides, which are monomeric and unmodified, with the first-generation monomeric, unmodified version of c(RGDyK). In future directions of this work, we are creating PEGylated versions of the knottin peptides, as well as oligomeric knottin proteins that present multiple integrin-binding RGD motifs. We expect these peptides will elicit enhanced tissue distribution and/or tumor uptake compared with unmodified knottin peptides, much like that observed with PEGylated and multivalent c(RGDyK) peptides, respectively. Here, we show that engineered integrin-binding knottin peptides are promising molecular imaging agents for clinical translation and future development, with potential applications in other imaging modalities including single photon emission computed tomography, targeted ultrasound, and magnetic resonance imaging.

No potential conflicts of interest were disclosed.

Grant support: NIH National Cancer Institute (NCI) Howard Temin Award 5K01 CA104706 and the Mallinckrodt Faculty Scholar Award (J.R. Cochran), NIH ICMIC P50 CA114747, NCI 5R25T CA118681, and the Canary Foundation (S.S. Gambhir), and a NCI Molecular Imaging Scholars postdoctoral fellowship (R.H. Kimura; R25T CA118681).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Frank V. Cochran for help with peptide synthesis and purification, and Zhe Liu and Zheng Miao for help with radiolabeling for PET experiments.