Pharmacodynamic Response of the MET/HGF Receptor to Small-Molecule Tyrosine Kinase Inhibitors Examined with Validated, Fit-for-Clinic Immunoassays

The receptor tyrosine kinase (RTK) MET [MET proto-oncogene, RTK; hepatocyte growth factor receptor (HGFR)] is an important drug target because of its roles in cancer progression, metastasis, and acquired resistance to EGFR inhibitors (1, 2). Aberrant MET signaling can occur through ligand-dependent as well as ligand-independent mechanisms (3). Consistent with multiple mechanisms of MET dysregulation, there are multiple therapeutic strategies to target MET, including HGF or MET-neutralizing antibodies, decoy receptors, small-molecule tyrosine kinase inhibitors (TKI), and allosteric inhibitors of MET activation (4–7). Drugs from each mechanistic class are currently under clinical investigation, both as single agents and in combination with other treatments (8–11). Many of these clinical trials require patient selection based on immunohistochemical and/or gene copy number assessment of MET (11, 12). However, these diagnostic assays cannot provide quantitative pharmacodynamic information (i.e., magnitude and duration of target modulation) to guide clinical development, nor can they distinguish between phosphorylated epitopes of full-length MET and its degradation products (13).

MET receptor activity is regulated by phosphorylation of a number of sites including Y1234 and Y1235 in the activation loop, which are crucial for the regulation of kinase activity; the carboxy-terminal Y1349 and Y1356 in the multifunctional docking site required to recruit cytoplasmic signal transducers and adaptors, and S975 and Y1003 in the juxta-membrane region, phosphorylation of which causes MET ubiquitylation and degradation (5, 8, 14, 15). On the basis of this understanding of MET signaling, we designed MET immunoassays to measure full-length, transmembrane MET protein and three key phosphospecies involved in its signal transduction: pY1234/1235-MET, pY1235-MET, and pY1356-MET. These assays were designed for the measurement of wild-type MET, because mutations account for only 10% to 30% of all cancer subtypes where MET acts as a driver (compilation at http://www.vai.org/Met/Index.aspx; ref.16). After validating the immunoassays and specimen collection and processing methods in preclinical models, the possible clinical suitability of the immunoassays for studying MET in diseases without gene amplification was evaluated in a case of MET-driven hereditary papillary renal cell carcinoma (HPRC).

Purchased antibodies were qualified for use, including antigen affinity–purified goat polyclonal anti-human MET extracellular domain (catalog number AF276, R&D Systems); mouse monoclonal anti-full-length MET (clone Met4, Van Andel Research Institute, Grand Rapids, MI); mouse monoclonal anti-C-terminal MET (clone L41G3); and rabbit monoclonal anti-pY1234/pY1235-MET (clone D26, Cell Signaling Technology, Inc.). Two rabbit mAbs specific to pY1235-MET (clone 23111) and pY1356-MET (clone 7334) were developed under contract with Epitomics Inc. using phosphorylated peptide antigens corresponding to amino acid sequences surrounding these tyrosine residues (Supplementary Material). Recombinant MET (rMET) calibrator protein was produced in HEK293 cells (Supplementary Material). Antibodies were biotinylated using Sulfo-NHS-LC-biotin (Thermo Fisher Scientific). Details of commercially available key assay reagents are described in SOP341203 and SOP341206, available at http://dctd.cancer.gov/ResearchResources/ResearchResources-biomarkers.htm.

Athymic nude mice (nu/nu NCr; Animal Production Program, NCI-Frederick, Frederick, MD) were implanted with the human cancer cell lines U87 (glioblastoma); A549 (lung carcinoma); MDA-MB-231 (breast carcinoma); HT-29 (colon carcinoma); or with GTL-16, MKN45, or SNU5 (all gastric carcinomas, MET-amplified) as described in ref.17. All cell lines were obtained from the Division of Cancer Treatment and Diagnosis Repository, NCI-Frederick and authenticated using AmpFLSTR Identifiler (Applied Biosystems).

MET inhibitors PHA665752 (NSC 748798-T), PF02341066 (NSC 749769-Y, crizotinib), and tivantinib (NSC 758242); VEGFR inhibitor pazopanib (NSC 737754); and multikinase inhibitor sorafenib (NSC 747971, lot #747971-U/3) were provided by the Developmental Therapeutics Program, NCI (Bethesda, MD). Purity was established by proton–carbon NMR, HPLC, and mass spectrometry. Sorafenib was dissolved in DMSO for in vitro studies. PF02341066 and pazopanib were administered by oral gavage in a saline vehicle and PHA665752 by intraperitoneal injections in a vehicle composed of 10% DMSO in saline. Tivantinib was administered orally in a PEG 400:20% vitamin E tocopheryl polyethylene glycol succinate solution (60:40) vehicle.

The NCI Animal Production Program, NCI-Frederick, is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International and follows Public Health Service policy on the humane care and use of laboratory animals. All studies were conducted according to approved NIH Animal Care and Use Committee protocols.

Specimen collection and handling conditions were adaptations of those achievable in past NCI clinical trials (18, 19). Briefly, 18-gauge Temno Trucut needle biopsies were immediately flash frozen in O-ring–sealed, conical-bottomed, screw-cap, 1.5-mL Sarstedt cryovials. Tubes were sealed, returned to liquid nitrogen, and stored at −80°C until use. Whole xenograft tumors were collected on the same schedule as tumor biopsies by standard dissection methods and cut into 2 to 4 equal pieces with fine-point scissors before flash freezing. All preclinical samples were frozen within 2 minutes of excision.

Tissue samples were processed by adding ice-cold Cell Extraction Buffer (Invitrogen) and supplemented with PhosSTOP (Roche) and protease inhibitor tablets (Roche) to the frozen tissue (0.35 mL buffer/biopsy and 0.75 mL buffer/tumor quarter). Tissue was immediately homogenized with a PRO200 homogenizer with a Multi-Gen adaptor (PRO Scientific) and a 5-mm generator at the maximum setting for 10 seconds at 2°C to 8°C. The extract was vortexed and homogenization was repeated. Extracts were incubated at 2°C to 8°C for 60 minutes with orbital shaking and clarified by centrifugation at 12,000 × g for 5 minutes at 2°C to 8°C. Cleared supernatant was aspirated and aliquoted. Total protein was measured by Bradford protein assay procedure (Bio-Rad).

Mice were inoculated bilaterally with human tumor line cells (1 × 10⁷) and tumor growth monitored daily. One cohort of mice was grouped when tumors reached 100, 200, 400, 500, 800, 1,000, 1,500, or 2,000 mg; the mean tumor size for each weight-bin was determined and the tumors removed for analysis. A second cohort was euthanized 10, 14, 18, 22, 26, and 30 days postimplantation, irrespective of tumor size, and retrospectively grouped into 100, 200, 400, and 600 mg weight-bins. DNA from one tumor quarter from each animal was analyzed for mouse and human DNA content (20).

SNU5 tumor xenografts were staged to approximately 200 mg (n = 5/group). Needle biopsies were collected under anesthesia and immediately flash frozen as controls. Tumors were excised and quarters transferred to sterile normal saline maintained at 25°C ± 3°C (cold ischemia) or 37°C (warm ischemia) for 1, 2, 3, 4, 5, 8, 10, 15, 30, or 60 minutes in temperature-controlled saline before being flash frozen. All flash-frozen samples were stored at −80°C, and lysates were processed within 2 weeks of collection.

Fit-for-purpose studies result in validated pharmacodynamic assays and provide drug mechanism of action information. To demonstrate assay fit-for-purpose in preclinical models, mice bearing GTL-16 and SNU5 xenografts were randomized when tumors reached 200 ± 25 mm³ in size and dosed daily for 8 to 10 days with vehicle, oral PF02341066 (6.25, 12.5, or 25 mg/kg), or intraperitoneal PHA665752 (25 or 50 mg/kg). In a separate study, mice with SNU5 xenografts were staged to 200 mm³ and treated for 8 days with vehicle (daily), pazopanib (100 mg/kg daily), tivantinib (200 mg/kg daily), or combinations of pazopanib (daily) with 2 dose levels of tivantinib (200 mg/kg daily or 200 mg/kg twice daily). Extracts of core needle biopsies collected 4 hours after tivantinib administration on day 8 were analyzed for MET, pMET, and hypoxia-inducible factor 1, α subunit (HIF1α; ref. 21).

Multiple 18-gauge needle biopsies were collected from a surgically resected tumor from a patient with HPRC (Urologic Oncology Branch, NCI, Bethesda, MD) within 2 minutes of resection and immediately flash frozen. The patient gave written informed consent for study inclusion and was enrolled in an NCI Institutional Review Board–approved protocol. Study design and conduct complied with all applicable regulations, guidelines, and local policies.

The full-length MET immunoassay measures the total levels of full-length MET in tumor tissue samples, irrespective of phosphorylation status. An affinity-purified goat polyclonal antibody against human MET extracellular domain (catalog number AF276) was used to coat 8-well Nunc MaxiSorp strips overnight at 2°C to 8°C with 1 μg/mL buffer (0.1 mol/L sodium carbonate/bicarbonate buffer pH 9.6, 100 μL/well). Wells were blocked for 2 hours with 1× PBS, 0.5% mannitol, 0.2% glycine, and 0.2% BSA, and strips were assembled on 96-well plate frames, freeze dried, and sealed in desiccated pouches (6-month stability). For the immunoassay, 100 μL each of purified rMET calibrator, controls, and tumor lysates (at protein concentrations of 10–50 μg/mL) prepared in assay buffer [1× PBS/casein (BioFx) containing PhosSTOP, protease inhibitors, and 0.1% Triton X-100 (Roche)] were added to the AF276-coated wells. The rMET assay calibrator was prepared for the dynamic range of the assay. Plates were incubated at ambient temperature (25°C ± 3°C) on an orbital shaker (600 rpm) for 1 hour to capture MET. After three washes with assay buffer, 100 μL/well 200 ng/mL biotin-conjugated mouse monoclonal MET antibody (clone L41G3) was added to the plates and incubated for 1 hour at 25°C ± 3°C, followed by a second wash and the addition of 100 μL/well 200 ng/mL streptavidin poly-HRP conjugate (Pierce). After 30-minute incubation with the HRP-conjugate, plates were washed 4 times and 100 μL SuperSignal ELISA Pico Chemiluminescent Substrate (Pierce) was added to each well. Plates were read within 10 minutes using an Infinite 200 Microplate ELISA Reader (Tecan). MET immunoassay calibration curves were analyzed on GraphPad Prism, and unknown values were calculated using 4-parameter curve fitting.

Three different pMET immunoassays were developed for the pharmacodynamic studies: pY1234/1235-, pY1235-, and pY1356-MET (Supplementary Material). The pMET immunoassay procedures were similar to the full-length MET immunoassay, except that rabbit mAbs specific to pY1234/1235-MET (clone D26, Cell Signaling Technologiy), pY1235-MET (clone 23111), or pY1356-MET (clone 7334) were conjugated to biotin and used as reporter antibodies for the immunoassay. The pMET assay calibrators were prepared for the dynamic range of each assay. Assay control samples were prepared by combining GTL-16, SNU5, A549, MKN45, or U87 xenograft tumor lysates to achieve low, medium, and high MET concentrations at 10 to 50 μg/mL total protein concentrations.

HT-29, GTL-16, and A549 cancer cell lines were treated with DMSO vehicle (0.1% w/v), 100 nmol/L PF02341066, or 100 nmol/L sorafenib for 24 hours at 37°C. Cells were fixed in 10% neutral-buffered formalin (Sigma-Aldrich) for 24 hours, pelleted, and then embedded in paraffin. After 5-μm sectioning, antigen retrieval with Bond Epitope Retrieval Solution 2 (at 100°C for 10 minutes) and immunofluorescent staining were performed on the BOND-MAX Autostainer (Leica Biosystems). For antigen detection, 10 μg/mL primary antibody (anti-pY1235–MET, clone 23111) was followed by 10 μg/mL goat anti-rabbit AF488 (Life Technologies). Immunofluorescence microscopy was performed on core needle biopsies of a resected human HPRC tumor stained with 5 μg/mL anti-full-length MET (clone Met4), followed by 10 μg/mL anti-mouse AF660 (Life Technologies). Image acquisition and analysis were performed on a wide-field fluorescent, confocal microscope (Nikon 90i, Andor Camera, NIS Elements Software).

Regression analysis and descriptive statistics, including means, SDs, coefficients of variation (CV), one-way ANOVA analyses, and Student t tests were conducted with Microsoft Excel and GraphPad Prism (v3.04). The 95% confidence interval was significant at α = 0.05 for a two-sided t test.

Tyrosines 1234, 1235, and 1356 mediate MET signal transduction via phosphorylation; so many MET-targeted agents are designed to reduce MET phosphorylation at one or more of these sites. To study the pharmacodynamics of such drugs, a sandwich immunoassay was designed to capture MET via its extracellular domain and then probe the captured MET for a particular phospho-epitope (pY1234/1235, pY1235, or pY1356) using specific mAbs (Fig. 1A). The capture antibody essentially did not crossreact with recombinant extracellular domains of the two most likely cross-reactive human tyrosine kinases, RON (macrophage-stimulating 1 receptor; MST1R) and EGFR. At 10-fold excess concentration (4–400 pmol/L) over rMET (0.4–40 pmol/L), cross-reactivity with RON was <5%, and EGFR cross-reactivity was not detectable (data not shown). The pY1235-MET species was measured using a new rabbit mAb (clone 23111) that recognizes the pY1235 epitope independent of Y1234 phosphorylation status, and its high specificity was demonstrated by Western blot analysis using synthetic peptides, rMET, cell lysates, and mouse tissues (Supplementary Fig. S1A and S1D). Unlike the specificity of D26 antibody for the dual-phosphorylated pY1234/pY1235 epitope, the mAb produced by clone 23111 showed no cross-reactivity with the cytoplasmic domain of RON or with this MET epitope when phosphorylated only at Y1234 or when tyrosine 1235 was replaced by aspartic acid (Supplementary Fig. S1A and S1B). It stained the plasma membrane of MET-expressing cell lines in an HGF-dependent manner that was sensitive to MET inhibitors, but not sorafenib (Fig. 1B). A new rabbit mAb recognizing the pY1356 epitope of MET (clone 7334) was also generated for use in the immunoassay (Supplementary Fig. S1C and S1D).

Using full-length, purified rMET as calibrator, the dynamic ranges of the calibration curves for each assay were established: 0.3 to 40 pmol/L for full-length and pY1234/1235-MET assays and 3.125 to 200 pmol/L for the pY1235- and pY1356-MET assays (Fig. 1C). Assay sensitivities were 1.5, 1.5, 7.8, and 15.6 fmol/μg protein, respectively, and the assays did not significantly crossreact with mouse MET species (Supplementary Materials).

The MET immunoassays were subjected to a rigorous validation protocol for analytic performance using clinically relevant tumor sampling procedures (core needle biopsy of tumors in mice) and specimen preparation procedures (Supplementary Material). Dilution linearity studies showed that tumor extracts could be diluted up to 8-fold without affecting assay performance (Supplementary Table S1). Spike-recovery experiments of rMET in xenograft extracts established immunoassay accuracy. Recovery ranged from 78% to 116% for full-length MET, pY1234/1235-MET, and pY1235-MET assays and 86% ± 29% for pY1356-MET (mean ± SD; Supplementary Table S1). Intraplate (n = 20) and interday (n = 5) variation were evaluated using three xenograft tumor extracts assayed at minimum of 5 different days by two operators; intra-assay CV was <10% and interassay CV was <14% (Supplementary Table S1). After reducing the procedures and operating parameters of the validated assay to standard operating procedures (SOP), assay robustness was formally demonstrated by SOP-driven assay transfer from the development to the clinical testing laboratory as described previously (18, 22).

The validated assay was capable of quantifying full-length MET in extracts of GTL-16, A549, MDA-MB-231, and U87 xenograft tumors over a wide range of protein loads per well (Fig. 1D). Full-length MET was detected over a range of 0.05 to 1.79 fmol/μg protein, and as expected, the MET-amplified model (GTL-16) contained the highest levels (1.31–1.79 fmol/μg protein). A least significant change (LSC) calculation, which combines both technical variation of the assay and biologic variation of the biomarker (determined by analyzing multiple quadrants of xenografted SNU5 tumors), established a 45% change as the minimal effect level required to demonstrate a drug effect (Supplementary Material).

Preanalytic variables (specimen handling, shipping, and storage procedures) can have a significant impact on the reliability of biomarker measurements in the laboratory, and phosphoproteins involved in dynamic responses of signaling pathways are notoriously labile during specimen collection due to ischemia and other factors (23–25). The stability of full-length MET and its phosphospecies was characterized in biopsies of SNU5 xenografts subjected to increasing ischemia time, which was defined as the total time needed for core needle biopsy sampling, specimen handling, and flash freezing. MET levels in core needle biopsy samples frozen immediately after collection from anesthetized animals (defined as the zero time point) were set as baseline (100%). Both pY1235- and pY1356-MET levels decreased by >60% during 15 minutes of cold ischemia (25°C ± 3°C) and continued decreasing over the next 15 minutes (P < 0.05; Fig. 2A). During 15 minutes of warm ischemia (37°C), both pY1235- and pY1356-MET levels decreased by >80% (P < 0.05; Fig. 2B). The pY1234/p1235-MET species exhibited similar degradation under these conditions (data not shown). In contrast, full-length MET levels were relatively stable for up to 60 minutes of cold ischemia but decreased significantly at 37°C (Fig. 2A and B).

A follow-up study evaluated the stabilities of pMET and full-length MET during 8 minutes of cold ischemia. After just 3 minutes of cold ischemia, pY1234/1235-MET levels decreased by >50% (P < 0.05; Fig. 2C). On the basis of Western blot analysis results, loss of full-length pY1235-MET signal during cold ischemia was predominantly due to loss of the pY1235-MET epitope, plus a small amount of degradation of full-length MET (the appearance of N- and C-terminal fragments; Fig. 2D). Stabilities of the pY1234/1235-MET and pY1356-MET analytes were similar (data not shown). Therefore, core needle biopsies frozen within 2 minutes of collection will yield valid pMET assay results.

Mice with GTL-16 tumor xenografts, a MET-amplified model, were treated daily with the MET inhibitors, PHA665752 or PF02341066, and sampled at clinically relevant time points (4 or 24 hours) after drug administration. The 25 and 50 mg/kg/day dose levels of PHA665752 slowed the growth of GTL-16 xenograft tumors and achieved tumor stasis after 8 to 10 days of treatment (Fig. 3A, study days 17–19). Four hours after the first dose, there was a dose-dependent reduction in the pY1235-MET:MET ratio of 62% and 80%, respectively (Fig. 3B), effect sizes that exceeded the LSC threshold of 45%. Over the next 20 hours, the pY1235-MET:MET ratio partially to fully recovered, depending upon dosage. Additional daily doses of 50 mg/kg PHA665752 significantly decreased the pY1235-MET:MET ratio, whereas the 25 mg/kg dose produced inconsistent changes (Fig. 3B). The ratio of pY1356-MET:MET also declined after the first dose, although it recovered to normal levels within 20 hours (Supplementary Fig. S3A). Therefore, the tumor stasis observed with this agent and dosage regimen was associated with intervals of modest reductions in the pY1235 and pY1356 targets, followed by target recovery. The absolute level of full-length MET also decreased by 40% to 50% within 10 days of treatment with either vehicle or drug (Supplementary Fig. S3B).

As reported previously (10), treatment of GTL-16 xenografts with daily oral PF02341066 at the 25 mg/kg dose level achieved stasis of GTL-16 xenografts by the seventh day of treatment (Fig. 3C; study day 11). Only this dose level reduced the pY1235-MET:MET ratio (by 65%) 4 hours after treatment, and this ratio recovered completely during the next 20 hours (Fig. 3D). Additional daily doses of PF02341066 further reduced the pY1235-MET:MET ratio, so molecular target control 4 hours after each dose was achievable throughout the daily treatment period. The SNU5 model exhibited a similar dose-dependent pharmacodynamic response to PF02341066 and again pointed to greater pharmacodynamic responsiveness of pY1235 than pY1356 following TKI therapy (Supplementary Fig. S4).

A second study of GTL-16 xenografts treated daily with PF02341066, but with denser tumor sampling to more fully characterize the pharmacodynamic response, confirmed that only the 25 mg/kg/day dose level achieved a molecular target response after the first dose. However, there was a cumulative, dose-dependent pharmacodynamic effect on the pY1235-MET:MET ratio such that by day 4, both the 25 and 12.5 mg/kg dose levels produced pharmacodynamic responses 4 hours and 12 hours after drug administration, although only the higher dose maintained this pharmacodynamic response for the entire 24-hour dosing interval (Fig. 4A). The pY1356-MET:MET ratio decreased after 4 and 12 hours but fully recovered by 24 hours at both dose levels (Fig. 4B). Again, full-length MET levels decreased by 40% to 50% by treatment day 10 in both drug- and vehicle-treated groups (Fig. 4C).

This preclinical modeling suggests that greater control of MET signaling than the intermittent 80% reduction in the pMET:MET ratio observed here will be needed to achieve regressions of MET-amplified tumor models. These data also indicate the importance of using the more consistent pMET:MET ratio instead of absolute levels of particular pMET species as pharmacodynamic endpoints, at least in preclinical models where mouse cell infiltration of human tumor xenografts over time diluted the absolute level of human MET per microgram–extracted protein (Supplementary Fig. S5).

The first clinical application of the validated assay is assessment of tumor pharmacodynamic response during a phase I clinical trial of the VEGFR/platelet-derived growth factor receptor (PDGFR) inhibitor pazopanib alone and in combination with the MET inhibitor tivantinib (ClinicalTrials.gov NCT01468922). The trial is testing two hypotheses: (i) MET will be upregulated as part of a compensatory response to VEGFR/PDGFR inhibition by pazopanib and (ii) tivantinib will blunt that upregulation by inducing degradation of total MET (26). To model this trial, mice bearing SNU5 xenografts were treated with vehicle or the drugs individually or in combination, and the tumors were sampled on day 8 using 18-gauge biopsy needles to replicate the sampling planned for the clinical trial. Neither single-agent pazopanib nor tivantinib affected full-length MET levels (Fig. 5A). However, treatment with pazopanib, but not tivantinib, increased both pY1235-MET:MET and pY1356-MET:MET ratios compared with vehicle (Fig. 5B and C). Twice daily tivantinib (200 mg/kg) combined with pazopanib returned the pMET:MET ratios to vehicle-treated levels, whereas daily tivantinib treatment (200 mg/kg) caused an intermediate effect (Fig. 5B and C). To determine whether the underlying changes in MET levels were associated with hypoxia induced by VEGFR/PDGFR inhibition, HIF1α levels were measured using a validated HIF1α immunoassay (21). Pazopanib-induced increases in pMET:MET ratios were accompanied by a 46% increase in HIF1α levels (P < 0.05), whereas the addition of tivantinib returned HIF1α to vehicle-treated levels (Fig. 5D).

The baseline levels of MET and pMET in human tumor models without MET amplification (Fig. 1D) are low, lying near the assay lower limit of quantitation; therefore, drug-induced decreases in pMET:MET ratios cannot be quantified in these models. To address the question whether MET and pMET levels in preclinical models and in clinical cancer without substantial MET amplification are similar, we evaluated core needle biopsies of five regions of a resected HPRC tumor harboring a germline MET mutation (H1112R). HPRC is characterized by germline trisomy of chromosome 7, frequent expression of two copies of mutated MET and one copy of wild-type MET (27, 28), and decreased activation threshold of the kinase activity and enhanced transforming activity (15); so, HPRC was a nonamplified disease in which pMET quantitation seemed likely. As expected, full-length MET levels ranged from 0.073 to 0.368 fmol/μg protein (Supplementary Table S2), similar to nonamplified models. However, pY1234/1235-MET was measurable in only two of five cores (0.037 and 0.041 fmol/μg), and neither pY1235-MET nor pY1356-MET were detectable in any core specimen.

Immunofluorescence staining of sections of the HPRC cores with the capture antibody from the immunoassay (AF276, anti-MET extracellular domain) confirmed the presence of MET in the HPRC cells and localized it primarily to the plasma membrane (Fig. 6A). Western blotting revealed that the undetectable level of pY1235-MET in the immunoassay was due in large part to the association of the epitope with a truncated C-terminal 50 kDa MET fragment (Fig. 6B), which is too small to contain the N-terminal domain required for capturing full-length MET in the validated sandwich immunoassay. A small amount of full-length MET containing the pY1235 epitope was detectable by Western blotting, but detection required loading the gel lane with 25 to 50 μg total protein, which exceeds the amount of protein that can be used in the immunoassay while maintaining assay linearity (Fig. 1D). Interestingly, the antibody against pY1234/1235-MET (clone D26) recognized mainly full-length MET (Fig. 6B), which is why it was detectable in the immunoassay.

Biochemical signaling from activated MET drives both proliferation and migration of malignant cells (1, 2), and MET can be activated not only by conventional ligand binding of paracrine and autocrine HGF, but also by several ligand-independent mechanisms. Transphosphorylation of Y1235 in the kinase domain initiates signaling, and subsequent phosphorylation of Y1349 and Y1356 in the SH2-binding motif transduces signaling, with phosphorylation of Y1356 critical for growth factor receptor–bound protein 2 docking that may link MET to the RAS pathway (29, 30), analogous to EGFR signaling. Several pharmacologic classes of MET inhibitors target these key phosphorylation events, so the extent of phosphorylation of the critical tyrosine residues should be a highly informative pharmacodynamic biomarker of molecular drug action. Importantly, we demonstrated that the phosphorylation status of the key tyrosine residues in full-length ligand-responsive MET is quantifiable using a dual-epitope sandwich immunoassay platform, with newly generated mAbs specific for pY1356 and for pY1235 (independent of phosphorylation status of the adjacent Y1234 residue), that will be made available to the research community via the Developmental Studies Hybridoma Bank (http://dshb.biology.uiowa.edu/). Also, to obtain valid assay results, cold ischemia time must be limited to 2 minutes to preserve the phosphorylated analytes. For preclinical pharmacodynamic studies in xenograft models, a human pMET:MET ratio normalizes pMET to total MET and mitigates the confounding effect of mouse cell infiltration as tumors grow (20).

Daily dosage regimens of two MET TKIs achieved tumor control and pharmacodynamic responses in GTL-16 and SNU5 human tumor xenograft models of poorly differentiated gastric carcinoma driven by MET amplification (31, 32). Tumor control of GTL-16 with PHA665752, a highly selective, ATP-competitive inhibitor of MET RTK activity (33), was associated with a 60% to 80% reduction in the pY1235-MET:MET ratio 4 hours after the first dose, but the ratio fully recovered by 24 hours and was less responsive to subsequent days of therapy. Treating GTL-16 with PF02341066, a potent, orally bioavailable, ATP-competitive inhibitor of MET catalytic activity (34), achieved tumor control at the only dose level that elicited a pharmacodynamic response (a consistent 65%–90% reduction in the pY1235-MET:MET ratio throughout the 24-hour dosing interval). The pY1356-MET:MET ratio was consistently less responsive to drug therapy than the pY1235-MET:MET ratio in both tumor models and with both MET TKIs, suggesting GRB2 signaling is more difficult to inhibit than MET kinase activity.

Importantly, 60% to 90% reductions in the pY1235-MET:MET ratio were associated only with tumor stasis, but not regression. Therefore, a greater than commonly seen MET pharmacodynamic response, that is, the magnitude and/or duration of pY1235-MET inhibition, or the addition of pY1356-MET inhibition (also linked to tumor regression; ref.35) must be needed to induce tumor regression in MET-amplified models. Clinical benefit from MET-directed therapy has been limited to patients with high baseline MET expression (36–39), but these assessments were made without a validated pharmacodynamic biomarker for activated MET or knowledge of the effect of warm ischemia on full-length MET degradation. Therefore, the trials to date unfortunately represent missed opportunities to understand clinical pharmacodynamic responses elicited by current MET inhibitors that could guide their future clinical development (40). The newly developed and validated assays described here are suitable for defining the pharmacodynamic response associated with tumor regression both in preclinical models and human patients with MET-amplified cancers. The analytic performance of the assays is sufficient for conducting small, fast phase 0 clinical trials of various dosage regimens of investigational MET inhibitors to discover how to maintain over 90% target suppression during dosing intervals (41, 42). From our error modeling, considering both biologic variability and assay precision, clinical pharmacodynamic study designs based on longitudinal patient sampling should require decreases in pMET:MET ratios of at least 45% (the LSC value) to attribute them to drug effect.

In addition to MET amplification with its relatively high baseline pMET:MET ratios, assay fitness-for-purpose was also demonstrated in a MET-induction setting during compensatory responses to tumor hypoxia. Induction of MET signaling converts a tumor from low to measurably high MET and pMET levels, which become targets of MET TKI therapy. We found that pharmacologic targeting of tumor vasculature in SNU5 xenografts with pazopanib, a multikinase inhibitor, induced the hypoxia biomarker Hif1α and increased pY1235-MET:MET and pY1356-MET:MET ratios. Although twice daily treatment with tivantinib, an allosteric inhibitor of MET (6, 7, 26) with some effects on microtubules (43–45), did not reduce baseline MET or pMET levels, it nevertheless blocked pazopanib-induced HIF1α MET signaling. Several molecular mechanisms could explain why tivantinib blocked pMET increases induced by pazopanib, including blocking the microtubule dependent coclustering of MET and VEGFR2 or prolonging PTP1B (protein tyrosine phosphatase, non-receptor type I) inhibition of MET (43, 44, 46). Clinical trials of MET/VEGF inhibitor combinations [pazopanib/tivantinib (NCT01468922) and bevacizumab/tivantinib (NCT01749384)] provide an opportunity for correlative pharmacodynamic studies to determine whether this compensatory mechanism operates in clinical cancers, and the preclinical modeling of one of these trials demonstrated that the validated pMET immunoassay is a suitable tool for these studies.

This newly developed and validated immunoassay for pharmacodynamic biomarkers of MET molecular response is a robust tool for understanding and optimizing pharmacologic control of both amplified and induced MET signaling, and it is ready for training-based transfer to the research community by the NCI (http://dctd.cancer.gov/ResearchResources/ResearchResources-biomarkers.htm). Phosphorylation of Y1235 is the proposed first step of MET signaling upon ligand binding, and the specificity of the new mAb for phosphorylated Y1235 independent of the phosphorylation status of the adjacent Y1234 residue will be useful for teasing apart the role of these neighboring tyrosines during initiation of MET signaling and molecular response to MET TKIs. Importantly, the anti-pY1235 mAb we generated (clone 23111) exhibited much greater specificity than the anti-pY1234/1235 mAb (clone D26), which even recognized a recombinant GST-MET protein harboring a Y1235D substitution. Our studies found that pharmacodynamic responses of the pY1235 and pY1234/1235 biomarkers did not always correspond. Furthermore, both pY1235 and pY1234/1235 epitopes were identified in full-length MET extracted from an HPRC tissue sample, whereas the pY1235 epitope was mostly present in a 50 kDa MET fragment. Similarly, complex phosphorylation profiles have been found in non–small cell lung cancer, in which only 25% of MET+ cases were also positive for pY1234/1235-MET (47). The pY1235-specific antibody could be a key reagent in a new diagnostic test for selecting the right patients to receive MET-targeted therapies because it substitutes the phosphorylated MET kinase domain (48) in place of elevated MET, mRNA, gene copy number, colocalized proteins, or multiply phosphorylated epitopes (49, 50). However, unlike the validated immunoassay, cross-reactivity to MET fragments with undefined biologic significance (27) will confound the reported immunohistochemical tests of MET signaling (26, 37, 51–53).

No potential conflicts of interest were disclosed.

The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Conception and design: A.K. Srivastava, M.G. Hollingshead, J.J. Ji, R.J. Kinders, D.P. Bottaro, J.E. Tomaszewski, J.H. Doroshow, R.E. Parchment

Development of methodology: A.K. Srivastava, T. Navas, S.A. Khin, J.J. Ji

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Srivastava, M.G. Hollingshead, J. Weiner, T. Navas, S.A. Khin, J.J. Ji, T.D. Pfister, D.P. Bottaro, W.M. Linehan, J.H. Doroshow

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Srivastava, J. Weiner, T. Navas, Y.A. Evrard, J.J. Ji, Y. Zhang, T.D. Pfister, D.P. Bottaro, J.E. Tomaszewski, R.E. Parchment

Writing, review, and/or revision of the manuscript: A.K. Srivastava, M.G. Hollingshead, T. Navas, Y.A. Evrard, J.J. Ji, Y. Zhang, T.D. Pfister, R.J. Kinders, D.P. Bottaro, W.M. Linehan, J.H. Doroshow, R.E. Parchment

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.K. Srivastava, Y.A. Evrard, S.A. Khin, J.J. Ji, Y. Zhang, S. Borgel, T.D. Pfister, W.M. Linehan, J.E. Tomaszewski, J.H. Doroshow

Study supervision: A.K. Srivastava, M.G. Hollingshead, Y. Zhang, J.E. Tomaszewski, R.E. Parchment

Other (conceptualized the project with others in DCTD, assigned to the DCTD Pharmacodynamic Group at NCI-Frederick, and managed on a weekly to monthly basis during project team meetings): J.E. Tomaszewski

The authors thank Drs. Andrea Regier Voth (Leidos Biomedical Research, Inc.) and Joseph Shaw (Kelly Services) for knowledgeable and skilled medical writing support in the preparation of this manuscript.

This project has been funded in whole or in part with federal funds from the NCI, NIH, under contract no. HHSN261200800001E.

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.