Anti-MET ImmunoPET for Non–Small Cell Lung Cancer Using Novel Fully Human Antibody Fragments

Since its discovery in the mid-1980s, MET, the receptor of hepatocyte growth factor (HGF), has been found to be very important in embryonic development, cell migration, cell growth, cell survival, epithelial–mesenchymal transition, wound healing, and tumorigenesis (1–3). Activation of MET has been found in various cancers, including bladder, breast, cervical, colorectal, gastric, kidney, liver, lung, ovarian, and prostate (1). MET amplification has also been found to be an important mechanism for acquired resistance to anti-EGFR therapies in non–small cell lung cancer (NSCLC; refs. 4, 5). Because of the important roles of HGF–MET signaling in various cancers, many inhibitors targeting this pathway are currently being developed for clinical applications, including both small-molecule inhibitors and monoclonal antibodies (3). A humanized one-armed anti-MET antibody, onartuzumab (MetMAb), has been evaluated in clinical trials for advanced NSCLC in combination with erlotinib. Although patients with MET-positive tumor benefited from such combination treatment, the MET-negative patients actually had worse overall survival when treated with onartuzumab plus erlotinib, compared with erlotinib plus placebo (6). Such results emphasize the importance to evaluate MET expression level for patient stratification to improve these anti-MET therapies.

Compared with traditional biopsy and immunohistochemistry, antibody-based positron emission tomography, or immunoPET, offers a unique opportunity for noninvasive evaluation of the in vivo expression levels of various biomarkers. The whole body information provided by immunoPET scans can help illuminate the heterogeneity of the primary tumor and metastatic lesions, and the evolving molecular status of tumors can be easily monitored via serial immunoPET scans to aid treatment planning and follow-up (7). Previously, anti-MET immunoPET imaging has been successfully demonstrated in preclinical mouse models using the intact monoclonal mouse antibody DN-30 or the humanized one-armed antibody onartuzumab (8, 9). However, these antibodies with full Fc domains require relatively long imaging delays (3 days to 1 week) to clear from the circulation to produce high-contrast images.

By using smaller antibody fragments with shorter serum half-lives, such as diabodies and minibodies (described in greater detail below), immunoPET can be performed at earlier time points with similar or even higher contrast, highly desired for clinical imaging applications (7, 10–12). Compared with an intact antibody (150 kDa) with heavy and light chain variable and constant domains, a single-chain variable fragment (scFv; ∼27 kDa) is a small monovalent fragment consisting of the antibody V_H and V_L domains linked by a flexible linker. A diabody is related to a scFv, comprised of only the V_H and V_L domains, but with a shorter linker that induces dimerization, resulting in a bivalent fragment (∼55 kDa). The bivalent minibody fragment is formed by fusion of the scFv to the immunoglobulin C_H3 constant domain for dimerization. Their higher molecular weight (∼80 kDa) promotes longer serum persistence, facilitating higher uptake levels in target tissues. Cys-diabodies are modified diabodies with engineered cysteines at their C-termini to allow site-specific conjugation and labeling (13–18). Figure 1 shows sizes and structures of these antibody fragments in comparison with an intact antibody. The availability of engineered antibody fragments allows selection of the optimal format for an imaging probe based on the target and application.

For clinical use, fully human antibodies and their respective fragments are preferred because of the lower risk of immunogenicity, and phage display technology offers a reliable source of high affinity fully human antibodies (19–21). With the FDA-approved Humira (adalimumab) and Benlysta (belimumab), and dozens more in clinical trials (22), phage display technology is now one of the major sources of fully human monoclonal antibodies for therapeutic applications. Furthermore, because the human antibody gene sequences are already in hand, reformatting selected scFvs into diabodies or minibodies is very fast and simple. This approach has been used to generate several previous antibody fragment imaging probes from phage libraries (16, 23, 24). As immunoPET is playing an increasingly important role in diagnosis, and in the development and implementation of targeted therapies (25, 26), phage display can be an effective path for the rapid generation of antibody fragment–based imaging probes. Instead of the months required to generate and screen conventional hybridomas, isolate the antibody genes, humanize, and conduct affinity maturation, by using phage display technology, fully human antibodies can be identified and reformatted in just a few weeks. This can lead to significantly faster discovery and development of novel antibodies and fragments for various applications.

In this study, we used a previously reported fully human scFv phage display library (27) to screen for novel human antibodies recognizing human MET. One of the antibody clones, H2, was reformatted into a cys-diabody and a minibody for MET-specific immunoPET studies in mouse models to evaluate the potential of assessing MET expression levels in vivo. In parallel, the potential biologic activity of the human anti-MET antibodies was assessed, because avoidance of agonistic biologic activity is desirable in an imaging agent.

Details of antibody reformatting, expression and purification, radiolabeling of antibody fragments, and the immunoreactivity assay can be found in the Supplementary Methods.

Hcc827 parental cells and the gefitinib-resistant Hcc827-GR6 cells (4) were obtained from the laboratory of Dr. Jeffrey A. Engelman (Massachusetts General Hospital, Boston, MA) in 2012. Hcc827^−/+ cells were derived from Hcc827 cells by 15 days of treatment with 50 ng/mL HGF and 1 μmol/L erlotinib (LC Laboratories; replenished every 3 days). H1975 and H1650 cells were obtained from the ATCC in 2011. All these cell lines were maintained with RPMI-1640 media (Cellgro) supplemented with 10% FBS (Gemini), 1% sodium pyruvate (Cellgro), and 1% penicillin–streptomycin (Gibco). The Hcc827-GR6 media were also supplemented with 100 nmol/L gefitinib (Santa Cruz Biotechnology), and Hcc827^−/+ cells were maintained with 1 μmol/L erlotinib. C6 rat glioma cells (from the laboratory of Sanjiv Gambhir, Stanford University, Stanford, CA, in 2002) were maintained in Deficient DME High Glucose media (Irvine Scientific) supplemented with 10% FBS, 1% penicillin-streptomycin and 2 mmol/L l-glutamine. The cell lines were not authenticated by our laboratory. Female SCID mice (The Jackson Laboratory) were injected s.c. into the left or right shoulders with cells (2–4 × 10⁶) in growth media: 50% Matrigel (BD Biosciences). The tumors were allowed to develop for 2 to 4 weeks before imaging. All animal studies were carried out under a protocol approved by the Chancellor Animal Research Committee of the University of California in Los Angeles.

A naïve human scFv phage display library from James D. Marks laboratory (UCSF; ref. 27) was used to perform three rounds of selection against recombinant human MET protein using the previously published method (28). Briefly, recombinant MET protein was used to coat an immunotube (Nunc). After washing and blocking, the immunotube was then incubated with 10¹²–10¹³ phage in 2 mL 1% milk-PBS for 2 hours at room temperature. The unbound phages were washed off with PBS and 0.05% PBS-Tween, and the bound phages were eluted with 1 mL of 100 mmol/L triethylamine for 8 minutes and neutralized with 0.5 mL of 1 mol/L Tris-HCl (pH 7.5). The eluted phages were used to infect exponentially growing E. coli TG1, and the infected bacteria were then used to produce phage for the next round of selection. Of note, 100 μg of MET protein from R&D Systems (358-MT/CF) was used for the first round. Reduced amounts of MET protein from another supplier (Sino Biological Inc.; 10692-H08H) were used for the following two rounds.

Phage ELISA was performed as previously described (28) to screen for binding clones. Briefly, antibody-displaying phages were expressed from 93 clones from the third round of selection to detect human MET protein immobilized on a 96-well plate. Horseradish peroxidase (HRP)–conjugated anti-M13 antibody (GE Healthcare; cat#: 27-9421-01) was used to detect the bound phage. 2,2′-Azino-bis(3-ethylbenzothiazoline-6-sulfonic acid; ABTS; Sigma, A9941) was used as the HRP substrate. The 64 clones with the highest positive signals (A₄₀₅) were subjected to DNA sequencing.

To determine cross-reactivity to mouse MET protein, purified anti-MET cys-diabodies were used to coat an ELISA plate (anti-EpCAM cys-diabody or 2% milk-PBS were used for negative controls). Of note, 0.1 μg human or mouse MET-hIgG Fc fusion protein (Sino Biological Inc.) was then applied to each well. The captured human or mouse MET-hIgG Fc was then detected with alkaline phosphatase–conjugated goat anti-human IgG. Each combination was done in triplicate.

For the standard anti-MET flow cytometry, murine anti-human MET antibody (Cell Signaling Technology; #5631) was used for primary incubation and FITC- or phycoerythrin (PE)-conjugated goat anti-mouse IgG was used for secondary incubation. To determine the copy number of MET protein per cell, quantitative flow was performed using QIFIKIT (Dako; K0078) according to the recommended protocol.

To estimate affinities, Hcc827-GR6 cells were incubated with various concentrations of anti-MET cys-diabodies at 4°C for 2 hours, then cells were quickly washed with flow buffer and incubated on ice with excess mouse anti-MYC IgG for 20 minutes, and washed again before a final incubation on ice with excess PE-conjugated goat anti-mouse IgG. Mean fluorescence intensities at different concentrations were calculated using FlowJo software, and the binding curves were analyzed using GraphPad Prism to calculate dissociation constants using the “one site-specific binding” model.

For phage flow cytometry, 30× concentrated phages were incubated with cells in flow buffer (PBS/1% FBS/2 mmol/L EDTA) with 4% milk on ice for 1 hour. After washing three times with flow buffer, FITC-conjugated anti-M13 antibody (Santa Cruz Biotechnology; sc-53005 FITC) was added to detect the phage.

Cell growth and inhibition were evaluated by MTS assay using the CellTiter96 AQueous Assay Kit from Promega (Cat# G5421), according to the recommended protocol. A total of 1 × 10⁴ cells were seeded in each well of the 96-well plates and treated with different combination of antibodies and drugs for 3 days before the MTS assay. Each combination of cell line and drug and antibody treatment was evaluated in at least triplicate wells. The results were normalized to the cells treated with media only.

MET-positive Hcc827-GR6 cells were seeded into a 24-well plate and cultured until confluent. The media were removed and the cells were incubated with ice-cold PBS/1% FBS with excess ⁸⁹Zr-labeled H2 cys-diabody for 1 hour to saturate cell surface MET. Following washing with ice-cold PBS-1% FBS three times, the cells were incubated at 37°C in growth media with excess ⁸⁹Zr-labeled cys-diabody. At each time point (0, 1, 2, 4, and 24 hours), cells were washed with ice-cold PBS and the membrane-bound cys-diabody was stripped with two 3 minutes treatment of 1 mL cold stripping buffer (0.1 mol/L glycine, 0.15 mol/L NaCl, and pH 2.5). The 2 mL stripping buffer was combined and considered as the membrane-bound fraction. The cells were then lysed with 1 mL lysis buffer (PBS/10 mmol/L Tris/0.5% SDS, pH 7.4) at room temperature for 3 minutes. Each well was washed one more time with 1 mL lysis buffer and the combined lysis buffer was considered as internalized fraction. The radioactivity of each fraction was measured by gamma counting. The assay was performed in triplicate for each time point.

Tumor bearing mice (tumor weight, 58–467 mg) were injected via tail vein with approximately 13 to 25 μg of ⁸⁹Zr-labeled cys-diabody or minibody in PBS (specific activity, 0.08–0.2 MBq/μg). Ten-minute microPET acquisitions were performed at 4 and 20 hours after injection for the cys-diabody, and 24 and 44 hours after injection for the minibody, with an Inveon microPET Scanner (Siemens). The mice were anesthetized with 2% isoflurane during the scanning process. MicroPET images were reconstructed using a filtered back projection algorithm, and analyzed with AMIDE software. Ex vivo biodistribution studies were performed after the final imaging time point. Organs, tumors, and blood were harvested and weighed, and the radioactivities were measured by gamma counting. The percentage of injected dose per gram of tissue (%ID/g) was calculated after decay correction to evaluate the uptake level in each organ. About half of the Hcc827-GR6 tumors were cystic, and the biodistribution of these tumors was calculated excluding the cystic fluid. The radioactivities of some kidney samples were too high to be accurately measured by gamma counting, so the uptake levels were estimated by image quantitation using the previously described method (29).

⁸⁹Zr-labeled H2 cys-diabody (22 μg, 1.15 MBq/mouse) or minibody (29 μg, 1.20 MBq/mouse) was injected into female SCID mice via tail vein. Of note, 5 μL of blood samples were taken at different time points (0.5, 1, 2, 4, 6, 12, and 24 hours after injection for the cys-diabody and 1, 2, 4, 6, 12, 24, 48, 72 hours after injection for the minibody) and the radioactivities were measured by a gamma counter. The blood curves were analyzed with GraphPad Prism, and fitted with the “one phase decay” model to calculate the terminal half-lives.

Statistical analysis was performed using Prism (GraphPad.com). Student t tests were used to compare the in vitro MTS assay results, and the tracer uptake levels. Data are presented as mean ± SD.

Following three rounds of selection on immobilized recombinant MET protein, 93 clones were picked and ranked on the basis of binding to MET protein via phage ELISA. DNA sequencing results of the top 64 clones reveal 19 distinct functional sequences. Figure 2A shows the ELISA results of the 19 representative clones, each demonstrating at least 2-fold increase above background signal. To further confirm binding to the native cell surface antigen, all 19 phage clones were incubated with MET-positive Hcc827-GR6 cells and analyzed by flow cytometry (determined by quantitative flow, the Hcc827-GR6 cell line has 2.6 × 10⁵ cell surface MET per cell whereas the parental Hcc827 cell line has 6.7 × 10⁴ per cell). Eight of the clones (A12, C2, E9, F1, F11, G1, H2, and H5) showed good binding profiles in flow analysis. Figure 2B shows the flow results of two representative clones. The sequences of these eight clones were analyzed using the abYsis online tool (http://www.bioinf.org.uk/abysis/) to identify the germline genes (based on IMGT definition and classification) and the complementarity determining regions (CDR). The V_H germline genes showed an overrepresentation of the IGHV3 subgroup, which is consistent with the original Sheets library (27). The eight clones use very different V_H CDR3 sequences, with lengths ranging from 6 to 19 residues (Fig. 2C).

These clones were reformatted into cys-diabodies by shortening the 15-aa linker to 6-aa linker and incorporation of C-terminal tags (see Materials and Methods). Three of the cys-diabody clones, C2, H2, and H5, were successfully expressed, purified (yields ranging from 1 to 3 mg/L, stable at 4°C for months), and characterized further. Other clones were dropped because of low expression level and purity.

The purified C2, H2, and H5 cys-diabodies retained the ability to bind to MET protein. Cross-reactivity analysis showed that the H2 cys-diabody clearly bound to both human and mouse MET, whereas C2 and H5 have no significant cross-reactivity to the mouse MET (Fig. 2D). Binding curves on Hcc827-GR6 showed that the H2 cys-diabody has the highest binding affinity (K_D = 0.73 ± 0.10 nmol/L), whereas the other two cys-diabodies also have low nanomolar affinities, as shown in Fig. 2E.

To assess off-target toxicity, first, the parental Hcc827 cells were treated with cys-diabodies alone for 3 days in culture media. No significant effects were observed overall, but the H5 cys-diabody appeared to have a small activating effect at very high concentration (180 nmol/L; Fig. 3A), which might also be due to experimental variation. Next, the cys-diabodies were used to treat two gefitinib-resistant cell lines lacking MET amplification, H1650 and H1975, in combination with gefitinib. No significant inhibitory effects were observed on these two cell lines (Fig. 3B). For the H1975 cells, although the C2 cysDb treated at 18 nmol/L appeared statistically different compared with the no antibody group, differences at either higher or lower concentrations were not significant, and the mean cell activities were all very similar to the no antibody group. C2 showed a slight activating effect on the H1650 cell line as the cys-diabody concentration was increased.

The anti-MET inhibitory effects were tested by applying these cys-diabodies to two cell lines with amplified MET signaling, Hcc827-GR6 and Hcc827^−/+. The Hcc827-GR6 cell line was generated through prolonged selection from the parental Hcc827 treated with gefitinib (4). The Hcc827^−/+ cells were obtained by short-term treatment of Hcc827 with HGF and erlotinib (see Materials and Methods). Figure 3C shows that the H2 cys-diabody has the strongest inhibitory effect at lowest concentration, whereas C2 and H5 only showed inhibitory effects at higher concentrations. Because of the high affinity and inhibitory effect of the H2 cys-diabody, we further reformatted H2 into a minibody and expressed the H2 minibody in 293F cells. The protein A–purified H2 minibody was then tested for inhibitory effects on Hcc827-GR6 in comparison with the cys-diabody and scFv. Similar inhibitory effects were observed for the minibody as shown in Fig. 3D.

Following site-specific conjugation of the H2 cys-diabody with maleimide-deferoxamine (DFO) and radiolabeling with ⁸⁹Zr (Fig. 4A), the labeling efficiency was 49%, the radiochemical purity was 95% after column purification, and the specific activity was 0.13 MBq/μg. When MET-expressing Hcc827-GR6 cells were cultured with excess radiolabeled H2 cys-diabody, the membrane-bound radioactivity decreased in the first 6 hours whereas the internalized activity increased, giving a constant total cell radioactivity in the first few hours. The membrane radioactivity remained low and the internalized ⁸⁹Zr activity continued to accumulate over the 24 hour time course (Fig. 4B). Such results indicate rapid internalization of the radiolabeled H2 cys-diabody, making it necessary to use residualizing isotope to achieve better imaging results.

The ⁸⁹Zr-labeled H2 cys-diabody (⁸⁹Zr-DFO-H2 cys-diabody) was initially used for immunoPET studies in mice bearing MET-positive tumors (Hcc827 or Hcc827-GR6) and MET-negative C6 tumors. The ⁸⁹Zr-DFO-H2 cys-diabody showed specific uptake in antigen-positive tumors (Hcc827 and Hcc827-GR6) and high-contrast images were obtained at as early as 4 hours after injection. As shown in Fig. 4C and D, tracer uptake in MET-positive tumors was higher than the MET-negative C6 tumors in the PET scans at both 4 and 20 hours after injection. The ex vivo biodistribution data confirmed these observations. In the Hcc827 versus C6 group, the MET-positive Hcc827 tumors had uptake levels of 1.1 ± 0.1%ID/g whereas the negative C6 tumors showed uptake levels of 0.47 ± 0.02%ID/g (positive tumor: negative tumor = 2.4 ± 0.2; positive tumor: blood = 18 ± 2; positive tumor: muscle = 12 ± 4). In the Hcc827-GR6 versus C6 group, the Hcc827-GR6 tumors had higher uptake levels of 1.8 ± 0.2%ID/g, and the negative C6 tumors had uptake levels of 0.65 ± 0.15%ID/g (positive tumor: negative tumor = 2.9 ± 0.8; positive tumor: blood = 23 ± 5; positive tumor: muscle = 13 ± 7). For better comparison of Hcc827-GR6 and Hcc827 tumors, the high-MET gefitinib-resistant tumor and the low-MET parental tumor were imaged in the same mouse as shown in Fig. 4E. The images show clear differences between high-MET Hcc827-GR6 tumor and the low-MET Hcc827 parental tumor, confirmed by the ex vivo biodistribution analysis. The high-MET Hcc827-GR6 tumors had uptake levels of 3.4 ± 0.3%ID/g (Hcc827-GR6 tumor: Hcc827 tumor = 1.9 ± 0.1; Hcc827-GR6 tumor: blood = 22 ± 2; Hcc827-GR6 tumor: muscle = 27 ± 14). Table 1 summarizes the ex vivo biodistribution study of the H2 cys-diabody imaging. The kidney uptakes were not shown in the biodistribution results because the signals were too high to be accurately measured by the gamma counter. However, we estimated the uptake levels to be about 83% to 104%ID/g (Hcc827 vs. C6 group: 85 ± 1%ID/g; Hcc827-GR6 vs. C6 group: 99 ± 2%ID/g; Hcc827-GR6 vs. Hcc827 group: 98 ± 9%ID/g) by image quantitation using a previously described method (29). The high kidney uptakes are due to the renal clearance of these low molecular weight cys-diabody proteins.

To further confirm the specificity of anti-MET targeting, the H2 minibody was also evaluated in immunoPET studies. For the minibody imaging, tumor bearing SCID mice were scanned at 24 and 44 hours after injection. Figure 4F shows clear differences in uptake levels between the high-MET Hcc827-GR6 tumor and the low-MET Hcc827 tumor, both at 24 and 44 hours after injection. Because of the longer serum persistence of the minibody, the ex vivo biodistribution analysis at 44 hours after injection revealed higher uptake levels of 8.6 ± 0.5%ID/g in Hcc827-GR6 tumors, and the Hcc827 tumors also showed higher uptake level of 4.5 ± 0.4%ID/g. However, the image contrast was not improved because of higher background signal in the blood (Hcc827-GR6 tumor: blood = 4.2 ± 0.5, details shown in Table 2). To further compare the pharmacokinetic properties, radiolabeled H2 cys-diabody and minibody were also injected into tumor-free female SCID mice, and their terminal half-lives were estimated to be 24 ± 3 minutes (n = 2) and 10 ± 1 hours (n = 4) respectively.

In this work, we isolated multiple anti-MET antibody clones from a large naïve human scFv phage display library. Three of these clones were successfully reformatted into cys-diabodies for in vitro characterization, and demonstrated high affinities ranging from 0.7 to 5.1 nmol/L. The H2 cys-diabody shows the highest affinity (K_D = 0.73 ± 0.10 nmol/L), best anti-MET inhibitory effect (significant inhibitory effects at 1.8 nmol/L on MET-dependent cells, P < 0.01), and cross-reacts with murine MET. ImmunoPET studies using ⁸⁹Zr-radiolabeled H2 cys-diabody and minibody were able to distinguish an MET-overexpressing gefitinib-resistant tumor from the parental tumor at as early as 4 hours after injection. The ex vivo biodistribution analysis also showed significant differences in tracer uptake levels in the MET-overexpressing Hcc827-GR6 tumors versus the parental tumors (P < 0.01). Using the H2 cys-diabody, very high positive tumor-to-blood ratios (18–22) were achieved at 20 hours after injection. These results indicate the potential for rapid and even same day imaging for patient stratification.

In this study, we sequenced 64 positive phage clones and found 19 distinct antibody sequences, 11 of which occurred only once, indicating significant diversity of the MET-binding phage in the library. Among the solitary isolates, the H2 scFv showed excellent affinity, and reformatted versions were used as immunoPET agents, producing high-contrast images of MET-positive tumor xenografts. It is likely that there are additional good candidates in the phage pool remaining after the third round of selection. During the phage selection process, high affinity candidates are not always the most enriched ones. Individual antibody sequences can vary in the ability to be expressed, or may encode proteins that are toxic or inhibit robust growth of the host bacterium. Rare clones may express antibodies that perform better in vivo due to unique pharmacokinetic properties and stability. Complementary to the more traditional medium-throughput approach we used in this study, high-throughput approaches empowered by deep sequencing, low-cost gene synthesis, and high-throughput protein production technologies can enable researchers to screen higher numbers of candidates. This will make the rare clones more accessible to researchers, including clones with unique properties. For example, although antitumor effects are desirable for therapeutic applications, it is typically preferable to use biologically inert binders for strictly imaging applications. The ability to screen and evaluate larger numbers of candidate clones would increase the odds of finding the best fit for specific applications.

Although the intent was to isolate antibody fragments for in vivo PET imaging, it was also important to avoid antibodies that may stimulate MET signaling. H2 and H5 cys-diabodies inhibited cell proliferation of two MET-activated cell lines (Hcc827-GR6 and Hcc827^−/+), but showed no effects on two other gefitinib-resistant cell lines lacking MET amplification (H1650 and H1975), indicating that the inhibitory effects were MET specific. Focusing on the H2 clone alone, the cys-diabody, minibody, and scFv formats were further compared. Interestingly, although the bivalent H2 cys-diabody and minibody showed similar inhibitory effects on Hcc827-GR6 cells, the monovalent H2 scFv failed to inhibit the growth of the MET-overexpressing cells even at very high concentration (180 nmol/L), implying that bivalent binding is required for the anti-MET inhibitory effect. Previously, many anti-MET antibodies that have been isolated exhibit a partial agonistic effect, limiting their potential utility as cancer therapeutics. Despite blocking HGF-induced activation, the bivalent engagement of anti-MET antibodies appeared to cause at least partial activation of MET signaling. To address this problem, DN-30 and MetMAb (onartuzumab) were engineered into monovalent fragments, to exert their inhibitory effects without activating MET kinase (30–32). Engelman and colleagues (4) showed that Hcc827-GR6 acquired gefitinib resistance by activating ERBB3 through MET overexpression. Given the requirement of bivalency, it is unlikely that H2 antibody fragments' inhibitory effects are due to competition with HGF or directly blocking of binding site for other cell surface receptors such as ERBB3. As the internalization assay shows, binding of the H2 cys-diabody can cause rapid internalization of cell surface MET, suggesting a mechanism for downregulation of MET signaling similar to that of recently reported nonagonistic bivalent murine antibodies (33). Alternatively, the bivalent binding of H2 antibody fragments may lock MET into a conformation that prohibits downstream signaling. Investigation of the detailed mechanism is ongoing.

Given the fact that the H2 cys-diabody is rapidly internalized upon binding to cell surface MET, a residualizing radiolabeling approach using chelated ⁸⁹Zr was selected, the same as two previous preclinical anti-MET immunoPET studies (8, 9). It is difficult to directly compare among these studies because of the differences in tumor models and mouse strains. However, it is known that radiolabeled diabodies usually have lower tumor uptake levels compared with larger fragments or intact antibodies, but high-contrast images can be produced at earlier time points, due to the rapid blood clearance (34). With the high tumor-to-blood ratio (∼20:1) at only 20 hours after injection, these results once more demonstrated the unique advantage of diabodies for rapid and high-contrast immunoPET imaging. Thus far, anti-MET imaging studies using similar small antibody fragments in patients have not been described, motivating the current efforts to develop these fully human antibody fragments for in vivo PET imaging. Previously, Santimaria and colleagues (35) had demonstrated diabody-based imaging of the ED-B fibronectin domain in patients. Because of the fast clearance of the diabody fragment, they were able to demonstrate tumor targeting at as early as 6 hours after injection instead of the 4- to 7-day delay typically required when imaging with intact antibodies (7). With the short serum half-lives of the H2 antibody fragments, similar rapid imaging in patients can be expected in the future.

In these cys-diabody imaging studies, some variations were observed between experiments. For example, the cys-diabody imaging of mice bearing dual Hcc827-GR6 and Hcc827 tumors showed higher tumor uptake levels relative to the other experiments. This may reflect the effects of variations in radiolabeling, tail vein injection, or perhaps the ages of the mice, and variations in physiology. Despite difference in absolute uptake levels, the tumor-to-blood ratios remain similar among the three groups, suggesting that the variations in uptake levels are likely caused by differences in availability of the radiotracer in the blood.

Compared with cys-diabodies, minibodies have longer serum persistence, allowing higher tracer uptake in the tumors. The imaging and biodistribution results indeed showed improved tumor uptake levels when using the H2 minibody, but the tumor-to-blood contrast was less impressive (∼4:1, at 44 hours after injection). Overall, the cys-diabody is the preferred choice for rapid immunoPET applications. A further potential advantage of the H2 antibody fragments is their cross-reactivity to mouse MET, allowing more representative imaging studies in the presence of endogenous MET expression in normal tissues of the mouse.

In conclusion, the present study demonstrates a successful example of rapid development of fully human antibody fragment–based immunoPET probes from a human antibody phage display library. Compared with previous anti-MET immunoPET studies using intact or one-armed antibodies, imaging using the H2 cys-diabody provides high-contrast images at early time points, with clear potential for same day imaging. Considering the growing interest in MET-targeted therapies, a noninvasive imaging tool for detecting and assessing MET expression in tumors could be useful for diagnosis and patient selection. Of interest, the H2 antibody fragments also show anti-MET inhibitory effects, making these antibody fragments candidates for potential therapeutic applications as well.

J.D. Marks and A.M. Wu have ownership interest (including patents) in and are consultant/advisory board members for Imaginab, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Li, P. Mallick, A.M. Wu

Development of methodology: K. Li, Y. Zhou, A.M. Wu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Li, R. Tavaré, K.A. Zettlitz, S.M. Mumenthaler

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Li, R. Tavaré, K.A. Zettlitz, A.M. Wu

Writing, review, and/or revision of the manuscript: K. Li, R. Tavaré, K.A. Zettlitz, Y. Zhou, J.D. Marks, A.M. Wu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.D. Marks

The authors thank past and present members of the Wu laboratory for helpful discussions and contributions. The authors thank Dr. Jeffrey A. Engelman for providing cell lines used in these studies. The authors also thank Waldemar Ladno, Darin Williams, and Dr. David Stout, of the Crump Institute for Molecular Imaging, for assistance with the small animal PET scans.

This work was supported by the Stanford Center for Cancer Nanotechnology Excellence and Translation (CCNE-T; NIH U54 CA151459; to A.M. Wu, K. Li, K. Zettlitz, S. Mumenthaler, and P. Mallick). Additional support was provided by an unrestricted gift from the Ralph and Marjorie Crump Foundation (to A.M. Wu). Support for the Small Animal Imaging Facility and Flow Cytometry Core was provided by the UCLA Jonsson Comprehensive Cancer Center (NIH CA016042).

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.