A ⁶⁴Cu-labeled Gp2 Domain for PET Imaging of Epidermal Growth Factor Receptor

Abstract

Purpose

Determine the efficacy of a 45-amino acid Gp2 domain, engineered to bind to epidermal growth factor receptor (EGFR), as a positron emission tomography (PET) probe of EGFR in a xenograft mouse model.

Methods

The EGFR-targeted Gp2 (Gp2-EGFR) and a non-binding control were site-specifically labeled with 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) chelator. Binding affinity was tested towards human EGFR and mouse EGFR. Biological activity on downstream EGFR signaling was examined in cell culture. DOTA-Gp2 molecules were labeled with ⁶⁴Cu and intravenously injected (0.6–2.3 MBq) into mice bearing EGFR^high (n=7) and EGFR^low (n=4) xenografted tumors. PET/computed tomography (CT) images were acquired at 45 min, 2 h, and 24 h. Dynamic PET (25 min) was also acquired. Tomography results were verified with gamma counting of resected tissues. Two-tailed t tests with unequal variances provided statistical comparison.

Results

DOTA-Gp2-EGFR bound strongly to human (K_D = 7 ± 5 nM) and murine (K_D = 29 ± 6 nM) EGFR, and non-targeted Gp2 had no detectable binding. Gp2-EGFR did not agonize EGFR nor antagonize EGF-EGFR. ⁶⁴Cu-Gp2-EGFR tracer effectively localized to EGFR^high tumors at 45 minutes (3.2 ± 0.5 %ID/g). High specificity was observed with significantly lower uptake in EGFR^low tumors (0.9 ± 0.3 %ID/g, p < 0.001), high tumor-to-background ratios (11 ± 6 tumor:muscle, p < 0.001). Non-targeted Gp2 tracer had low uptake in EGFR^high tumors (0.5 ± 0.3 %ID/g, p < 0.001). Similar data was observed at 2 h and tumor signal was retained at 24 h (2.9 ± 0.3 %ID/g).

Conclusion

An engineered Gp2 PET imaging probe exhibited low background and target-specific EGFR^high tumor uptake at 45 min, with tumor signal retained at 24 h post-injection, and compared favorably with published EGFR PET probes for alternative protein scaffolds. These beneficial in vivo characteristics, combined with thermal stability, efficient evolution, and small size of the Gp2 domain validate its use as a future class of molecular imaging agents.

Graphical Abstract

Introduction

Molecular cancer therapeutics have provided effective treatments for many cancers, yet are typically characterized by efficacy on only a subset of patients, even within a type of cancer as defined by tissue^{1, 2}. Personalized or precision medicine via molecular characterization to differentiate responders from non-responders can aid patient outcomes³. Epidermal growth factor receptor (EGFR) overexpression is present in many cancer types^4–10, correlates with differentiation, reduced disease-free and overall survival, and is an independent prognostic indicator of poor survival in colorectal cancer patients^{11, 12}. EGFR amplification is predictive of response to cetuximab in wild-type KRAS metastatic colorectal cancer patients^13–15. In HER2-positive primary breast cancer, EGFR overexpression – but not copy number – is a poor prognostic factor and predictive of response to trastuzumab¹⁶. The current biopsy/immunohistochemistry approach to EGFR characterization is invasive and does not account for spatiotemporal heterogeneity, most notably differential expression between primary tumors and metastases^17–19. Positron emission tomography (PET) targeting EGFR could inform personalized treatment plans by enabling identification, localization, and characterization of primary tumors and metastases, while being non-invasive, quantitative, and sensitive to picomolar quantities. PET based imaging has been clinically useful for other receptors, such as imaging estrogen receptor for breast cancer^{20, 21}.

Numerous scaffolds have been explored as molecular PET tracers of EGFR. Therapeutic monoclonal antibodies (~150 kDa) have been radiolabeled to visualize EGFR in vivo but slow clearance results in high background and liver signal and necessitates late imaging times that elevate patient dose^22–27. 94-residue fibronectin domains^{28, 29}, 58-residue affibodies^30–34, 120-residue nanobodies^35–37 and 400-residue Fab fragments³⁸ have provided good tumor-to-background ratios at early time points (?4 h) via nuclear imaging due to their fast clearance, better extravasation, and increased tissue penetration compared to antibodies^39–42. Additional scaffolds have been used for other targets⁴³. Small molecule inhibitors^44–49 and natural EGF ligand⁵⁰ provide molecular characterization but are not biologically passive.

We recently developed the 45-residue Gp2 domain as a small, stable protein scaffold that has been successfully evolved towards multiple targets with high affinity (0.2–18 nM K_d) while retaining thermal stability (65–80°C)⁵¹. The Gp2 scaffold contains a framework of a single alpha-helix and three beta-strands, and two solvent-exposed loops that form the diversified paratope. Thermal stability, lack of cysteine, and presence of a single lysine residue distant from the proposed paratope provide ease of chemical conjugation of imaging moieties through amine or thiol chemistry. Additionally, the small size and straightforward structure enable direct chemical synthesis. The two Gp2 variants used here are Gp2-EGFR_2.2.3, which was previously evolved to bind to EGFR with 18 ± 8 nM affinity, and EGFR non-binding control, Gp2-rIgG_3.2.3, which previously evolved to bind to an irrelevant control (rabbit IgG; notably the molecule does not cross-react with murine IgG) (herein referred to as Gp2-EGFR and Gp2-nb). These variants share 70% sequence identity (Table S1).

We hypothesize that the small size of Gp2 will provide high tumor uptake with fast blood clearance enabling high contrast images at early time points. The ease of evolution and synthesis combined with high thermal stability and different paratope topology may provide a useful tool as an alternative imaging agent to available molecules. In particular, variant Gp2-EGFR_2.2.3 has 18±8 nM affinity for cell-surface EGFR with a midpoint of thermal denaturation of 71°C. The current study evaluates the ability of this scaffold to function as molecular PET agent in a small animal model.

Experimental Section

Protein production and DOTA conjugation

Gp2 domains were produced recombinantly in E. coli as described previously⁵¹. Briefly, one liter of LB medium with 50 mg/L kanamycin was inoculated with 5 mL of overnight BL21(DE3) E. coli culture carrying the pET-Gp2-His₆ plasmid, grown at 37 °C to an optical density (600 nm) of 0.6-1.5 units, and induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside for 20-24 hours at 30 °C. Cells were pelleted, resuspended in 10 mL of lysis buffer (50 mM sodium phosphate (pH 8.0), 0.5 M NaCl, 5% glycerol, 5 mM 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate, and 25 mM imidazole), and underwent four freeze-thaw cycles. The soluble fraction was isolated by centrifugation at 12,000 g for 10 min. Gp2 was purified by metal affinity chromatography on a HisPur resin (Pierce, Thermo Fisher Scientific). Purified Gp2, 30-60 µM, in PBS containing 150 mM imidazole was mixed with 25 to 50-fold molar excess 10 mg/mL DOTA-NHS-ester (Macrocyclics) in dimethyl sulfoxide and allowed to react at room temperature for 1 h. The reaction was quenched with excess 1 M Tris pH 8.0, purified on a PD-10 column (GE Healthcare), and evaluated by matrix-assisted laser desorption ionization mass spectrometry.

Size Exclusion Chromatography

Protein solutions in 100 mM sodium acetate at pH 5.0 were filtered with a 0.2 µM filter to remove any particulates. 200 µL of 40 µM DOTA-Gp2 was loaded onto an AKTA primeplus (GE Healthcare Bio-Sciences) and with a Superdex 75 10/300 GL column. The mobile phase was 100 mM sodium acetate at pH 5.0 flowing at 0.5 mL/min.

Cell growth

A431 epidermoid carcinoma were kindly provided by Dr. Daniel Vallera (University of Minnesota). MDA-MB-435 cells, which have similarities to a melanoma cell line but also show evidence of breast cancer lineage⁵², were kindly provided by Dr. Tim Starr (University of Minnesota). Cells were cultured in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum at 37 °C in humidified air with 5% CO₂.

Affinity measurement

Cells to be used in flow cytometry were detached using trypsin for a shorter time (3–5 min) than recommended. Detached cells were washed and labeled with Gp2 at varying concentrations for 15–30 min at 4 °C. Cells were pelleted and washed with PBSA (PBS + 0.1% w/v BSA), then labeled with fluorescein-conjugated rabbit anti-His₆ antibody (Abcam ab1206) for 15 min at 4 °C. Fluorescence was analyzed on a C6 Accuri flow cytometer (BD Biosciences). The equilibrium dissociation constant, K_D, was identified by minimizing the sum of squared errors assuming a 1:1 binding interaction.

Affinity of Gp2 towards soluble murine EGFR ectodomain (Sino Biological) was determined using Gp2 displayed on the yeast surface as described previously⁵¹.

Western Blot Analysis

A431 cells were grown to approximately 60% confluency, washed with PBS and incubated in serum-free medium overnight at 37 °C in humidified air with 5% CO₂. The next day, cells were washed with PBS and exposed to four different conditions at 37 °C: (1) PBS for 20 min; (2) 5 nM DOTA-Gp2-EGFR for 20 min; (3) 5 nM epidermal growth factor (Gemini Bio Products) for 20 min; or (4) 5 nM DOTA-Gp2-EGFR for 30 min, washed with PBS, followed by 5 nM epidermal growth factor for 20 min. Cells were detached from the plate by mechanical shearing in RIPA buffer (PBS with 1% v/v Triton X-100, 0.5% w/v sodium deoxycholate, 0.1% w/v sodium dodecyl sulfate). Cells were lysed through rotation at 4 °C for 30 min in RIPA buffer. After centrifuging at 15,000g for 15 min at 4 °C the supernatant was collected and protein concentration was determined with a Pierce BCA assay kit (Thermo Scientific).

Whole-cell lysates (60 µg) were boiled in 5X Laemmli loading buffer at 95°C for 5 minutes, separated by 8% SDS-PAGE, transferred to PVDF membrane and subjected to indicated immunoblotting analyses according to manufacturer guidelines. The primary antibodies bind phosphorylated AKT serine 473 (#9271 Cell Signaling Technology), total AKT (#9272), phosphorylated EGFR tyrosine 1068 (#2234), total EGFR (#2232S) and actin (#A3853 Sigma-Aldrich) were incubated overnight at 4 °C. After washing with Tris-buffered saline with Tween-20 (50mM Tris, 150mM sodium chloride and 0.05% Tween-20), the membrane was further immunoblotted with either anti-rabbit horseradish peroxidase-conjugated antibody (#NA934V GE Healthcare Life Science) or anti-mouse horseradish peroxidase-conjugated (#170-6516 Biorad) secondary antibody for 1 h at 37 °C.

Internalization

Gp2-EGFR and Gp2-nb in PBS with 150 mM imidazole was allowed to react with fluorescein isothiocyanate in DMSO (3 mg/mL) at 100x molar excess at room temperature for 1 h. The reaction was quenched with excess 1 M Tris buffer pH 8, purified on a Zeba Spin Desalting Column 7K molecular weight cutoff (ThermoFisher). Fluorescein conjugation was verified by matrix-assisted laser desorption ionization mass spectrometry.

A431 and MDA-MB-435 cells were grown and detached as above. Cells were labeled with 100 nM fluorescein conjugated Gp2 at 37 °C for 0.5 and 1 h, followed by incubation with 0.2 M acetic acid, 0.5 M NaCl pH 2 for 5 min to strip extracellular binding. Fluorescence was detected by flow cytometry. Internalization was calculated by normalizing the change in fluorescence signal over time to fluorescence signal of A431 cells labeled with 100 nM fluorescein-Gp2-EGFR at 4 °C for 0.5 h.

Copper chelation and purification

⁶⁴CuCl₂ (UW-Madison) was diluted into 150 µL of 100 mM sodium acetate pH 5.0 and pH adjusted to pH 5.0. Approximately 50 MBq of the ⁶⁴CuCl₂ was added to 100 µL DOTA-Gp2 in 100 mM sodium acetate pH 5.0 at 30-60 µM. The mixture was allowed to incubate at 47 °C for 1 h and purified by PD-10 column equilibrated with 10 mM sodium acetate pH 5.0 in order to remove unchelated copper.

Radio TLC

1 µL of ⁶⁴Cu-Gp2 was spotted on filter paper and a mobile phase of PBS was applied for 20 minutes. An AR-2000 radio-thin layer chromatography scanner (Eckert & Ziegler) scanned and analyzed the filter paper for migration of radioactive peaks. Comparison of scans before and after PD-10 purification showed removal of the peak near the solvent front (the unconjugated ⁶⁴Cu) while retaining the less mobile peak (⁶⁴Cu-Gp2), which cold PD-10 purifications along with SDS-PAGE and binding assays have shown to contain highly pure Gp2.

Tumor inoculation

Eight week old female (Foxn1^nu/Foxn1^nu) mice (Jackson Laboratory) were anesthetized with 1.5% isoflurane in 1 mL/min O₂ and subcutaneously injected with 10 million MDA-MB-435 cells in 50% v/v Matrigel Matrix (Corning) in one shoulder. After 4 weeks, the mice were injected with two million A431 cells in 50% Matrigel Matrix into the opposite shoulder. Xenografted tumors were grown to 5-10 mm in diameter (approximately two weeks for A431 and six weeks for MDA-MB-435).

EGFR expression quantification

To quantify EGFR expression within in vivo xenografted tumor cells, GentleMACS dissociator C Tubes (Miltenyi Biotec) were used to generate single cell suspensions from excised tumors. Receptor expression was quantified by flow cytometry with Quantum Simply Cellular anti-mouse IgG calibration beads (Bang’s Laboratories), using Gp2-EGFR and/or mouse anti-EGFR antibody (Abcam ab30) at 1 µM, followed by secondary labeling with fluorescein conjugated rabbit anti-His6 (Abcam ab1206) or AlexaFluor 647 conjugated goat anti-mouse IgG (ThermoFisher), respectively. The cell population from the A431 tumor was approximated as two normally distributed subpopulations.

PET imaging – static and dynamic

All procedures performed in studies involving animals were in accordance with the ethical standards of the University of Minnesota and approved by the Institutional Animal Care and Use Committee. Mice were anesthetized with 1.5% isoflurane in 1 mL/min O₂ and tail vein injected with approximately 0.6 to 2.3 MBq of ⁶⁴Cu-Gp2 as measured by a Atomlab 100 dosimeter with a setting of 50.2. Five-minute static PET scans were performed at 45 min, 2 h and 24 h after injection using an Inveon micro-PET/CT (Siemens). The PET energy cutoffs were 350-650 keV with a timing window of 3.438 ns. The PET images were reconstructed with an OSEM2D method using 4 iterations of Fourier rebinning. PET images were smoothed with a 1×1×1 voxel Guassian filter. The CT used 340 projections of 80 kV at 500 µA with 200 ms exposure over 384 s of total scan time with an effective pixel size of 98.3 µm. The CT was reconstructed using the Feldkamp algorithm with a Shepp-Logan filter. The preceding methods are included in the Inveon Acquisition Workplace software (Siemens). A second batch of independently produced, DOTA-conjugated, ⁶⁴Cu chelated, and purified ⁶⁴Cu-Gp2-EGFR injected into another set of tumor inoculated mice validated the results of the other 45 min and 2 h PET/CT scans.

PET images were quantified using the Inveon Research Workplace software (Siemens). Using the CT as an anatomical guide, the volume of 10-20 mm³ that resulted in the maximum average PET signal for that tissue was selected. The anterior end of the liver was selected to avoid noise from kidney signal. The posterior leg furthest from any bladder signal was chosen to represent muscle background.

Tissue gamma counting

After imaging, mice were euthanized by cervical dislocation under isoflurane anesthesia. Blood, bone, brain, heart, large intestine, kidneys, liver, lungs, muscle, pancreas, skin, spleen, stomach, and tumors were resected, weighed, and had their activity measured by a CRC-25W (Capintec) gamma counter averaged over 45 seconds. The CRC-25W collected counts from all windows and was calibrated through serial dilutions based on the dose reported by the Atomlab 100 dosimeter used to measure injected dose. Renal radiation dose was calculated with the Medical Internal Radiation dose method.

Statistics

Comparisons between two samples were determined using a two-tailed student’s t-test for unequal variances. P-values are stated where relevant. Data were presented as average ± standard deviation.

Results

Gp2 Production and Conjugation

EGFR-binding Gp2-EGFR and non-binding control Gp2-nb, both containing a C-terminal His-6 tag, were produced in the soluble fraction of E. coli and purified by immobilized metal affinity chromatography. Purity was verified with SDS-PAGE and molecular weight was verified by matrix assisted laser desorption ionization mass spectrometry (Gp2-EGFR expected: 6873, actual: 6869; Gp2-nb expected: 6228, actual: 6226). The copper chelator 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) was conjugated to the N-terminal lysine residue distal from the proposed paratope in the Gp2 scaffold framework (Fig. 1). Mass spectrometry was used to verify an average labeling of 0.83 DOTA per molecule for Gp2-EGFR and 1.1 DOTA per molecule for Gp2-nb (Fig. S1). Size exclusion chromatography verified that DOTA-Gp2-EGFR (7.4 kDa) is dominantly monomeric, eluting at a comparable time to control proteins of a similar size (6.5 kDa aprotinin and 7.5 kDa affibody; Fig. S2).

Fig. 1. Gp2 conjugation.

Purified Gp2 was conjugated with the N-hydroxysuccinimidyl ester of the chelator DOTA then radiolabeled with ⁶⁴Cu

EGFR Binding

Gp2-EGFR binding affinity towards cellular EGFR was previously found to be 18 ± 8 nM⁵¹. The effect of DOTA conjugation on binding affinity was examined by labeling EGFR-expressing A431 epidermoid carcinoma cells with varying levels of DOTA-Gp2-EGFR. DOTA conjugation did not significantly change ligand affinity (7 ± 5 nM; Fig. 2). The non-binding control DOTA-Gp2-nb showed no detectable binding up to 300 nM on A431 cells (Fig. 2). Preclinical imaging experiments with EGFR targeted Gp2 were carried out in mice, so the affinity of Gp2-EGFR towards murine EGFR was examined. Yeast displaying Gp2-EGFR were labeled with varying levels of recombinantly-produced murine EGFR extracellular domain, which revealed an affinity of 29 ± 6 nM (Fig. S3).

Fig. 2. Affinity titration.

A431 cells were labeled with DOTA-Gp2 domains (squares, DOTA-Gp2-EGFR; triangles, DOTA-Gp2-nb) at the indicated concentrations. Binding was detected by fluorescein-conjugated anti-His₆ antibody via flow cytometry. Fluorescence signal is normalized between minimal and maximal fluorescence. One representative titration of triplicate experiments is presented. A representative Gp2-EGFR titration curve is also included for comparison (circles, dotted line). The equilibrium dissociation constant for DOTA-Gp2-EGFR, assuming a 1:1 binding model, is 7±5 nM

Biological Activity

The effect of DOTA-Gp2-EGFR binding on the EGFR signaling pathway was determined by Western blot to detect phosphorylated AKT (p-AKT at S473), a downstream protein kinase, and phosphorylated EGFR (p-EGFR at Y1068) (Fig. 3a). A431 cells labeled with 5 nM DOTA-Gp2-EGFR show no change in p-AKT or p-EGFR compared to PBS only control suggesting that DOTA-Gp2-EGFR is not agonistic to EGFR, DOTA-Gp2-EGFR is also not antagonistic, as blocking the A431 cells with 5 nM Gp2 before addition of 5 nM epidermal growth factor showed no change in level of p-AKT or p-EGFR compared to EGF only control.

Fig. 3. Biological Activity.

(a) A431 cells were labeled with four different conditions in triplicate: PBS only, 5 nM DOTA-Gp2-EGFR in PBS, 5 nM epidermal growth factor (EGF) in PBS, or 5 nM DOTA-Gp2-EGFR followed by 5 nM EGF (Gp2 block). Cells were lysed and separated by SDS-PAGE. Blotting was done to detect phosphorylated AKT (S473), a protein kinase in the EGFR signaling pathway, and phosphorylated EGFR (Y1068), as well as total amounts of the two proteins and actin to verify similar total protein concentration. DOTA-Gp2-EGFR is neither agonistic, since it does not activate the EGFR pathway, nor antagonistic, since it does not block activation when EGF is present. (b) A431 and MDA-MB-435 cells were labeled with 100 nM fluorescein conjugated Gp2 at 37 °C for 0.5 and 1 h, followed by incubation with acid for 5 min to strip extracellular binding. Internalization was calculated by normalizing the change in fluorescence signal over time to fluorescence signal of A431 cells labeled with 100 nM fluorescein-Gp2-EGFR at 4 °C for 0.5 h. Error bars represent standard deviation for n = 3 biological replicates. P < 0.001 is indicated by *.

The ability of A431 cells to internalize Gp2 was examined through flow cytometry of cells grown in tissue plate culture. Fluorescein was conjugated to Gp2-EGFR (0.45 fluorescein/protein) and Gp2-nb (0.61 fluorescein/protein) through amine chemistry and used to label A431 and MDA-MB-435 cells at 37 °C for up to 1 h. At 0.5 and 1 h, cells were acid stripped and the increase in signal over time was used to calculate internalization rate (Fig 3b). Fluorescein-Gp2-EGFR rapidly internalized into A431 cells (2.2±0.3-fold of saturated surface EGFR per hour) compared to control cells (MDA-MB-435, p<0.001) and control non-binder (fluorescein-Gp2-nb, p<0.001).

Copper chelation and purification

Radioactive ⁶⁴Cu was incubated with DOTA-Gp2 for 1 h at 47°C. Free ⁶⁴Cu was separated by size exclusion chromatography resulting in, on average, 93% purity. Labeling efficiency was 29%, perhaps due to the low number of DOTA per Gp2 (to assure site specific conjugation) or modest protein concentration. Based on historical yields from non-radioactive DOTA-Gp2 purifications, the specific activity of chelated protein was 0.6-1.1 MBq/nmol. Radiolabeled DOTA-Gp2 variants are referred to as ⁶⁴Cu-Gp2-EGFR or ⁶⁴Cu-Gp2-nb.

Murine model micro-PET/CT and tissue biodistribution

The efficacy of the Gp2 domain was evaluated in a murine model with xenografted human tumor lines. To assess specificity EGFR^high A431 tumors (mean: 5.2×10⁵ EGFR/cell; 75^th percentile: 1.9×10⁶) and EGFR^low MDA-MB-435 tumors (mean and 75^th percentile: < 4×10³ EGFR/cell) (Fig. S4) were simultaneously evaluated. A non-binding Gp2 domain was tested in parallel. ⁶⁴Cu-Gp2 was injected via the tail vein into mice harboring dual tumors. PET/CT was performed at 45 min and 2 h. ⁶⁴Cu-Gp2-EGFR effectively localized to A431 tumors highly expressing EGFR (3.2 ± 0.5 %ID/g) and cleared from background (11 ± 6 tumor:background ratio, p < 0.001) as early as 45 minutes after injection (Fig. 4). Targeting was molecularly specific as EGFR^low MDA-MB-435 tumors had demonstrably lower signal (0.9 ± 0.3 %ID/g, p < 0.001). Moreover, the non-targeted control ⁶⁴Cu-Gp2-nb exhibited lower signal in EGFR^high tumors (0.5 ± 0.3 %ID/g, p < 0.001). As for most small protein imaging agents, high kidney signal is observed (78 ±16 %ID/g) resulting from renal processing. Similar imaging is observed at 2 h where ⁶⁴Cu-Gp2-EGFR uptake to EGFR^high tumors was 3.2 ± 0.6 %ID/g and 12 ± 4 tumor:background (p = 0.006). Specificity is retained at 2 h as EGFR^low tumors had low uptake (0.7 ± 0.2 %ID/g, p = 0.009) and the non-targeted control had lower signal in EGFR^high tumors (0.7 ± 0.3 %ID/g, p = 0.007). While early time point imaging is the preferred translational route, we acknowledge that for alternative applications, such as targeted therapy, and biological safety concerns the behavior of engineered proteins at later times is relevant. Even with the fast clearance, preferential EGFR^high tumor signal from ⁶⁴Cu-Gp2-EGFR is still evident at 24 h (2.9 ± 0.3 %ID/g) with high tumor:background (8 ± 6, p = 0.009).

Fig. 4. PET/CT imaging.

Coronal and axial micro-PET/CT images of anesthetized athymic nude mice bearing subcutaneously xenografted A431 tumors (EGFR^high) in the left shoulder and MDA-MB-435 tumors (EGFR^low) in the right shoulder. The mouse in the 24 h image lacks a MDA-MB-435 tumor. Mice were injected by tail vein with 0.6-2.6 MBq of either ⁶⁴Cu-Gp2-EGFR (top row) or, the non-targeted control, ⁶⁴Cu-Gp2-nb (bottom row). Five minute static PET scans followed by CT scans were acquired at 45 min (left), 2 h (middle), and 24 h (right, for targeted Gp2 only) post-injection. Image planes were selected such that both tumors appear in the image.

PET images were corroborated by ex vivo tissue gamma counting at 2 h and 24 h (Fig. 5). At 2 hours post injection, ⁶⁴Cu-Gp2-EGFR localized significantly more to xenografted EGFR^high tumors (7.0 ± 1.9 % ID/g) as compared to EGFR^low tumors (1.4 ± 0.3 % ID/g; p < 0.001). The targeted Gp2 had 14 ± 8 tumor-to-blood ratio and 23 ± 6 tumor-to-muscle at 2 h, compared to 1.8 ± 1 tumor-to-blood (p = 0.005) and 3.3 ± 3.1 tumor-to-muscle (p < 0.001) for the non-targeted Gp2. In addition, the non-targeted Gp2 showed significantly lower EGFR^high tumor uptake with 1.4 ± 0.4 % ID/g (p = 0.001). Renal retention was high for the targeted (244 ±66 %ID/g) and non-targeted (208 ± 19 %ID/g) probes. Liver signal was modest for both (4.8 ± 1.8 and 4.9 ± 1.9 %ID/g). At 24 h the fast clearance leads to lower signal in most tissues, including EGFR^high tumor (4.0 ± 0.3 %ID/g) and kidney (114 ± 20 %ID/g), with the exception of a notable increase in liver signal (10.1 ± 1.3 %ID/g). Tumor-to-blood and tumor-to-muscle ratios (3.4 ± 1.1, p = 0.002 and 8.1 ± 3.6, p < 0.001, respectively) indicate there is still preferential uptake to EGFR^high tumor.

Fig. 5. Resected tissue gamma counting.

After PET/CT imaging, mice were euthanized and tissues were collected, weighed, and measured for activity. (A) The targeted ⁶⁴Cu-Gp2-EGFR (dark gray) and non-targeted ⁶⁴Cu-Gp2-nb (light gray) distribution is shown for the selected tissues at 2 h post-injection. (B) Ratios of tumor signal to relevant background signals in blood and muscle. The data is combined over two separate experiments, n = 4 for mice containing EGFR^high and EGFR^low tumors and another n = 3 for mice containing only EGFR^high tumors. Significance for important comparisons (p < 0.005) is denoted by *. (C and D) Biodistribution and tumor-to-background ratios of ⁶⁴Cu-Gp2-EGFR in n = 3 mice at 24 h post-injection. Error bars represent standard deviation.

The rapid distribution and clearance of ⁶⁴Cu-Gp2 evident at the 45 minute scan was more thoroughly investigated by 25-minute dynamic PET scans (Fig. 6). Using heart signal as a surrogate for probe blood levels, clearance half-time was revealed to be 3.2 ± 1.0 min, supporting the low accumulation in muscle background seen at 45 minutes post-injection.

Fig. 6. Dynamic PET scans.

25 minute dynamic PET scans were acquired on anesthetized mice containing xenografted EGFR^high tumors. The average signal within ~15 mm³ regions, guided by an anatomical CT scan, is presented. Data were fit assuming exponential kinetics. The clearance half-time within the heart (predominantly blood pool) was t_1/2 = 3.2 ± 1.0 min (n=2 mice).

Discussion

Other small scaffolds have been successfully used for in vivo imaging previously but drawbacks, such as the relatively larger size of fibronectins (11 kDa)⁵³ and DARPins (20 kDa)⁵⁴, or the difficultly of broad evolution and presence of cysteines in knottins⁵⁵ and cyclic peptides⁵⁶, has driven the search for additional scaffolds. Cysteine-free Affibodies⁵⁷ have gone to smaller size (58 amino acids) and their helical paratope has yielded high affinity binders, however they are typically severely destabilized after mutation⁵⁸. Gp2 domains push the size even smaller (45-49 amino acids), have thus far remained highly thermally stable after mutation, and provide a vastly different paratope structure compared to Affibodies. Beyond its previous characterization for high-affinity, EGFR-specific binding⁵¹, further biophysical evaluation of Gp2-EGFR in the current study revealed that it is well-suited for use in molecular imaging. Though selected solely for EGFR ectodomain binding, the current Gp2 variant is neither agonistic nor antagonistic (Fig. 3a). This enables passive imaging – unlike radiolabeled EGF or bivalent, crosslinking-compatible antibodies – which is preferred to avoid impacting EGF signaling cascades. Additionally, Gp2-EGFR is internalized into A431 cells (Fib. 3b). Internalization potentially allows for an accumulation of signal in target tissues over time, but may not be highly relevant for Gp2 due to the rapid clearance of the small agent, which has the benefit of reducing background. Primary amine / N-hydroxysuccinimidyl chemistry was selected for conjugation at the N-terminal lysine distal to the evolved loops (Fig. 1). As hoped, DOTA conjugation did not hinder binding affinity (18 ± 8 nM as Gp2-EGFR to 7 ± 5 nM as DOTA-Gp2-EGFR). Importantly, Gp2-EGFR exhibits cross-reactive binding to murine EGFR, which aids the validity of the murine model to assess the probe’s tumor selectivity relative to lower levels of EGFR expression in healthy tissue including liver. Modest liver accumulation was observed (4.8 ± 1.8 %ID/g at 2 h), which was due to physiological processing, not EGFR targeting, as the non-binding control exhibited equivalent hepatic retention (4.9 ± 1.8 %ID/g). This liver signal remains below the EGFR^high tumor signal (1.5 ± 0.4 tumor:liver). Nevertheless, efforts are underway to mutate surface hydrophobic amino acids to increase Gp2 hydrophilicity, which effectively reduced liver signal for engineered fibronectin domains²⁹.

The relevance of non-invasive EGFR detection in the clinic has led to development of many imaging probes, including a variety of small protein scaffolds. The increased extravasation and tissue penetration of protein scaffolds compared to larger proteins allows for high contrast early imaging resulting in lower patient dose. Multiple successes have been realized for EGFR previously. The beneficial properties of Gp2 domains as evolvable protein scaffolds, such as small size, lack of cysteines, and high thermal stability, do not guarantee successful translation to an imaging agent. However, these properties provide benefits during evolution, conjugation, administration, and biodistribution that are useful for imaging agents or therapeutics towards many targets. Due to the variations between labs, strict quantitative comparisons between scaffolds does not prove superiority. Moreover, comparisons across scaffolds must take care to acknowledge the context-dependent properties – affinity, charge, hydrophilicity – of individual protein variants. Nevertheless, the current data demonstrate that the Gp2 domain is a promising PET imaging agent for EGFR with potential benefits versus other probes, and further optimization of the affinity and biophysical properties of Gp2-EGFR could lead to a clinically effective PET imaging agent. ⁶⁴Cu-Gp2-EGFR exhibits tumor accumulation (3.2 ± 0.5 %ID/g at 0.75 h via PET; 7.0 ± 1.9 %ID/g at 2 h via excised tissue) comparable to other small protein PET probes including fibronectin domains (3.4 ± 1.0 and 2.4 ± 1.0 %ID/g at 1 h)²⁸ and affibodies (5.7 ± 0.6 and 9.7 ± 4.9 %ID/g at 1 h)³⁰ as well as nanobodies for single-photon emission computed tomography (4.6 ± 0.4 %ID/g at 1 h)³⁷. The dramatically lower uptake of ⁶⁴Cu-Gp2-EGFR in EGFR^low tumors and non-binding control in EGFR^high tumors was similarly observed for the fibronectin domain. For affibody, neither EGFR^low tumors nor non-targeted affibody were evaluated as controls. Blocking did yield a reduction, albeit incomplete (47%), in EGFR^high tumor uptake. ⁶⁴Cu-Gp2-EGFR exhibits high tumor:blood ratio (14 ± 8 at 2 h) because of rapid clearance (3.2 ± 1.0 min half-time). Conversely, affibody provides limited tumor:blood differentiation (1.2 ± 1.1 and 1.0 ± 0.1 at 1 h and 4 h) because of slower clearance (20 – 120 min half-time^{30, 59–62}) while fibronectin is intermediate (8.9 ± 4.7 and 6.4 ± 4.3 at 1 h and 4 h²⁸) with rapid clearance (2.1 ± 0.3 min half-time²⁸). Tumor:muscle specificity is also strong for ⁶⁴Cu-Gp2-EGFR (11 ± 6 at 0.75 h via PET; 23 ± 6 at 2 h via excised tissue), comparable to affibody (16 ± 7 at 1 h, 18 ± 4 at 4 h, both via excised tissue) and higher than fibronectin (8.6 ± 3.0 at 1 h via PET; 10 ± 4 and 4.2 ± 1.3 at 1 and 4 h via excised tissue).

The main disadvantage with Gp2 as an imaging agent is the high kidney signal due to partial renal retention during clearance, which is observed for most small protein scaffolds⁴³. Dosimetry calculations indicate 3.0 mGy/MBq renal dose, which is 3% of the maximum tolerated dose for a 185 MBq injection thereby rendering this a minor concern clinically for non-renal tumors. Yet strategies exist to lower kidney signal. Modulation of charge has been shown to reduce renal uptake in fibronectin domains²⁹, affibodies⁶³, and knottins⁶⁴. Preliminary data indicate an ability to modify charge on Gp2-EGFR while retaining activity. Additionally, alternative radiochemical conjugation has drastically reduced renal uptake of other small protein scaffolds^{62, 65–70}. Specifically, transchelation from the DOTA chelator⁷¹ may account for some signal in the liver and kidney, and other chelators such as NOTA or PCTA have shown higher stability in vivo⁷². Notably, ⁶⁴Cu (t_1/2 = 12.7 h) was used in the current study to enable examination of distribution kinetics over short and long time periods, which is important for initial physiological characterization of this new protein scaffold. Yet, clinical use may benefit from a radioisotope with decay kinetics that align with the rapid distribution of the small Gp2 domain to reduce patient dose. Future studies with ¹⁸F (t_1/2 = 110 min), ⁶⁸Ga (t_1/2 = 68 min), or ⁶¹Cu (t_1/2 = 3.3 h) will be valuable for clinical translation. Evaluation on cells with intermediate EGFR expression will also be informative. It should be noted that, as with any synthetically engineered protein with non-human sequence components, immunogenicity of evolved molecules will need to be evaluated.

Overall, the performance of these initial Gp2 domains in vivo gives promise to the potential of Gp2-EGFR, and other targeted Gp2 domains, as molecular imaging agents.

Abbreviations

EGFR

epidermal growth factor

PET

positron emission tomography

DOTA

1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid

CT

computed tomography

PBS

phosphate buffered saline

SDS-PAGE

sodium dodecyl sulfate – polyacrylamide gel electrophoresis