Bioorthogonal Small Molecule Imaging Agents Allow Single-Cell Imaging of MET

The hepatocyte growth factor receptor (MET) is a receptor tyrosine kinase (RTK) that has emerged as an important cancer target. Consequently, a number of different inhibitors varying in specificity are currently in clinical development. However, to date, it has been difficult to visualize MET expression, intracellular drug distribution and small molecule MET inhibition. Using a bioorthogonal approach, we have developed two companion imaging drugs based on both mono- and polypharmacological MET inhibitors. We show exquisite drug and target co-localization that can be visualized at single-cell resolution. The developed agents may be useful chemical biology tools to investigate single-cell pharmacokinetics and pharmacodynamics of MET inhibitors.


Introduction
The most dominant paradigm in drug discovery over the last two decades has been the design of exquisitely selective inhibitors that act on a single target within a disease pathway. However, lack of durable efficacy has challenged this 'one gene, one drug, one disease' hypothesis [1]. This is not entirely surprising given the robustness of many biological systems and their ability to utilize redundant networks to overcome inhibition of a single protein [2]. For these reasons, multi-targeting has gained renewed interest and indeed many clinically successful drugs have proven to be less selective than originally thought [3] [4] [5]. This observation, together with a systems understanding of cancer pathways has led to the concept of polypharmacology, i.e. the inhibition of multiple targets within a cell [2]. While combination therapies are an obvious first step towards multi-target inhibition, the deliberate design of a single kinase inhibitor that binds to multiple targets is a newer development [2] [6].
Receptor tyrosine kinases (RTKs) are key regulators of critical cellular processes in mammalian development, cell function and tissue homeostasis [7]. Dysregulation of RTKs has been implicated as causative factors in the development and progression of numerous human cancers [7]. Blockbuster drugs, Gleevec (Bcr-Abl and c-Kit), Herceptin (HER2), and Iressa (EGFR) have spawned intense investigation of other RTKs [8]. One of the emerging kinases of interest is the hepatocyte growth factor receptor (MET), which is widely expressed in epithelial and endothelial cells. MET is a central mediator of cell growth, survival, motility, and morphogenesis during development [9]. Consequently, MET overexpression relative to normal tissue has been detected in various types of cancers [10]. In addition, overexpression of MET is indicative of increased tumor aggressiveness and poor prognosis in cancer patients [11] [12] [13] [14]. A number of different MET inhibitors with varying levels of specificity are currently in clinical trials. These include the monospecific inhibitor, PF04217903, and the broad-spectrum inhibitor, Foretinib (GSK13630898; inhibits MET, AXL, RON, PDGFRα, and KDR) [15]. Despite the growing number of different MET inhibitors and peptide based whole body imaging agents [16], it has been difficult to visualize MET expression, intracellular drug distribution and small molecule MET inhibition.
It is generally believed that imaging is an invaluable tool in the drug development process. Imaging has been used to better understand the biology and pathophysiology of human cancer, enable earlier diagnosis and allow monitoring of therapeutic drug efficacy. Here we set out to develop a bioorthogonal imaging agent for high resolution imaging in live cells, based on clinical small molecule MET inhibitors. Specifically, we developed a mono-specific MET imaging agent based on PF04217903 [17] and compared its imaging characteristics to an imaging agent based on Foretinib [18], a polypharmacological MET inhibitor in phase III clinical development. Using this technique we were able to perform either very specific MET imaging or single-cell multi-target imaging of different proteins inside living cells. Companion imaging drug (CID) development with mono-and polypharmacologic inhibitors of MET would enable not only specific visualization of MET but also visualization of multiple RTKs at single-cell resolution. Such information can potentially provide new insight for biological understanding of MET and RTKs and, therefore, could help in the development of new drug candidates.

General experimental procedures
Unless otherwise noted, chemical reactions were carried out under an atmosphere of nitrogen or argon in air-dried glassware with magnetic stirring. Air-and/or moisture-sensitive liquids were transferred via syringe. Organic solutions were concentrated by rotary evaporation at 25 -60 °C at 15-30 torr. Analytical thin layer chromatography (TLC) was performed using plates cut from glass sheets (silica gel 60 F-254 from Silicycle). Visualization was achieved under a 254 or 365 nm UV light and by immersion in an ethanolic solution of cerium sulfate, followed by treatment with a heat gun. Column chromatography was carried out as "Flash Chromatography" using silica gel G-25 (40-63 μM).

N-(3-fluoro-4-hydroxyphenyl)-N-(4fluorophenyl)cyclopropane-1,1-dicarboxamide (2).
To a solution of 1 (1 g, 4.48 mmol) in DMF (8 drop) and THF (2 mL), stirred at 0 °C, a solution of oxalyl chloride (2.24 mL, 4.48 mmol) in DCM (1 mL) was added dropwise. The reaction mixture was stirred at ambient temperature for 2hr. Then, the solution of 1 and oxalyl chloride in DMF and DCM, was added to a solution of 4-Hydroxy-3-fluorophenol (626 mg, 4.93 mmol) and 2,6-Lutidine (519 µL, 4.48 mmol) in THF (2 mL). After additional stirring at 0 °C, the reaction mixture was gradually warmed up to ambient temperature. After completion of the reaction, monitored by TLC and LC-MS, it was quenched by addition of water. Organic material was extracted with ethyl acetate 3 times and the combined organic layer was washed with 1N HCl twice and washed once with NaHCO 3 (sat). The combined organic layer was dried over MgSO 4 and concentrated in vacuo. The resulting crude product was diluted with ethyl acetate and a brown solid was obtained by filtration and washed with EA:Hex = 1:3, 1:2 and to 1:1 to give compound 2 (899 mg, 60%). 1

Kinase assays
The IC 50 of Foretinib, Foretinib-TCO (11), Foretinib-BODIPY-FL (12), PF04217903, and PF04217903-TCO (15) were determined using the z´-LYTE assay kit. The Tyr6 peptide kit was used to measure MET, RON, and AXL activity. The Tyr4 peptide kit was used for PDGFRα and the Tyr1 peptide kit was used to measure KDR activity. All assays were run in accordance with the manufacturer's instructions, with minor modification. Briefly, MET, RON, and KDR were used at 0.2 μg/ml, 0.8 μg/ml, and 0.7 μg/ml, respectively. The kinase reaction buffer for MET, RON, and KDR was supplemented with 0.67% DMSO. AXL was used at 1.5 μg/ml and the reaction buffer was supplemented with 0.01% sodium azide and 0.67% DMSO. PDGFRα was used at 2.5 μg/ml and the reaction buffer was supplemented with 1mM DTT, 2mM MnCl 2 , and 0.67% DMSO. Inhibitors were prepared by 4-fold serial dilution at 75X the final concentration in 100% DMSO (8 μM to 0.1 nM). This stock was then diluted to 3X the final concentration in kinase reaction buffer. 5 μl of the 3X stock was added to the final 15 μl reaction volume. Both the z´-LYTE control phosho-peptide and the z´-LYTE peptide were used at a final concentration of 2 μM. ATP was added at a final concentration of 50 μM for MET and AXL, 10 μM for RON and PDGFRα and 75 μM for KDR to initiate the reaction, which proceeded for 1 hour at room temperature. Development reagent was then incubated for 1 hour at room temperature. After terminating the reaction with stop reagent, the coumarin and FRET based fluorescein emission was measured on a TECAN Saffire 2 plate reader (ex: 400 nm, em: 445 and 520 nm). IC 50 values were obtained by fitting the dose-response curves using Prism 6 (GraphPad).
For analysis of MET phosphorylation following Foretinib inhibition, OVCA429 cells were plated in 6-well plates 48 hrs before the assay. Cells were serum-starved in media containing 0.1% fetal bovine serum ~16 hrs before the assay. On the day of the assay, 0, 0.1, 0.5, or 1 μM Foretinib, Foretinib-TCO (11), or Foretinib-BODIPY-FL (12) were diluted in media containing 1 mg/ml BSA. Cells were incubated with Foretinib or its derivatives for 1 hr at 37°C. Phosphorylation was then stimulated with 100 ng/ml HGF diluted in RPMI with 1 mg/ml BSA for 5 min at 37°C. Cell lysates were then prepared and run on a gel as described above. Western blot analysis was performed as described above using the MET and phospho-MET primary antibodies (1:1000). Total MET and phospho-MET expression was quantified using densitometry (ImageJ).

Characterization of companion imaging drugs
To convert the above amine containing intermediates into different imaging agents, we pursued two approaches: direct conjugation to a fluorochrome (BODIPY-FL) and two-step bioorthogonal clickable reaction via TCO-drug and tetrazine(Tz)-BODIPY FL. The inhibitory effects of Foretinib, Foretinib-TCO (11) and Foretinib-BODIPY-FL (12) were evaluated using a FRET based MET kinase activity assay. Direct modification of Foretinib with the BODIPY-FL fluorochrome (12) led to a significant loss in drug activity, decreasing the IC 50 over 24-fold (from 28.7 nM [23] to 697.4 nM; Figure 3A and Table 1). Interestingly, the smaller TCO modification only modestly impacted the IC 50 value of Foretinib (88.6 nM; Figure 3A and Table 1). To further assess the impact of the BODIPY-FL (12) and the TCO (11) modifications on Foretinib activity, Western blot experiments were done for MET phosphorylation. OVCA429 ovarian cancer cells were pretreated with increasing concentrations of Foretinib, Foretinib-BODIPY-FL (12) or Foretinib-TCO (11). Cells were subsequently lysed and Western blot analysis was done for phosphorylated MET, as well as total MET expression ( Figure  3B and C). MET exhibited some basal phosphorylation, consistent with the high overexpression of MET in OVCA429 cells and with previous reports [24]. Stimulation with HGF led to a significant increase in MET phosphorylation. For total MET expression, we observed two different MET protein bands, the 145 kDa beta-chain and the unprocessed 170 kDa precursor [25] [26] [27]. At all concentrations tested, Foretinib treatment led to a complete decrease in MET phosphorylation. Both Foretinib-TCO (11) and Foretinib-BODIPY-FL (12) pretreatment led to a decrease in MET phosphorylation, with the TCO (11) modification exhibiting greater potency than the BODIPY-FL (12) version ( Figure 3B and C). This trend is consistent with the assay using purified MET, but the absolute inhibitory effect may be different in cell culture due to the abundance of other targets available for interaction with Foretinib and its derivatives. Similarly, TCO modification of PF04217903 (15) showed minimal reduction of the IC 50 value from 0.54 nM to 1.48 nM ( Figure S2C). Figure S3 compares imaging of Foretinib-TCO to Foretinib-BODIPY-FL in OVCA429 ovarian cancer cells. The former showed a dose-dependent increase in cellular fluorescence signal ( Figure S3) whereas the latter shows cellular staining only at the highest concentration tested (in addition to or because of its poor target binding). Collectively, therefore, we confirmed the efficiency of the two-step bioorthogonal imaging strategy for PCID. Focusing on the two-step bioorthogonal imaging strategy we next assessed the affinity of Foretinib-TCO against other target kinases: AXL, RON, PDGFRα, or KDR. Similar to the results obtained using purified MET, the TCO (11) modification only minimally impacted the IC 50 value for each of the protein targets tested ( Figure 3D and Table 1). On average, TCO modification still resulted in an affinity ligand in the low nM range (Table 1) suggesting its use as a model PCID.

Single-cell distribution of mono-and polypharmacologic companion imaging drugs
Given the affinity and cellular uptake of the bioorthogonal agents we next determined target localization by fluorescence microscopy. We chose carboxyfluorescein diactate-tetrazine (Tz-CFDA) to reveal Foretinib-TCO (11) given prior results in live cell imaging [15]. These experiments were performed in OVCA429 ovarian cancer cells because of their high MET expression as determined by Western blot ( Figure S4). We first tested PF04217903-TCO (15), given its exquisite affinity and specificity for MET. Compared to antibody co-staining for MET, there was superb co-localization of the drug and its target ( Figure 4 and Figure S5, Manders correlation coefficient of colocalization test about selected image was 0.969). A control experiment with unlabeled Foretinib/Tz-CFDA confirmed the low background signal as well as specificity of the TCO/Tz based two-step bioorthogonal labeling technique ( Figure S5). When SK-BR-3 cells that lack MET were used for the same imaging experiments, PF04217903-TCO (15) showed no significant staining compared to Foretinib-TCO (11) (Figure 4). Interestingly, Foretinib-TCO showed a different pattern. In addition to some membrane staining, there was also significant other cellular staining within the cell, presumably the cytoplasm/perinuclear region ( Figure 4). Similar staining patterns were also observed by live cell imaging for both PF04217903-TCO and Foretinib-TCO ( Figure S6). We next investigated the phenotype of intracellular Foretinib-TCO distribution ( Figure 5). We wanted to better understand the poly-pharmacologic drug distribution to understand where the drug localizes. We were specifically interested in determining whether the observed cellular staining co-localized with any of the other known Foretinib targets. To further probe the polypharmacologic distribution of Foretinib, we used correlative antibody staining against AXL, RON, PDGFRα, KDR, and MET. Figure 5 summarizes these experiments showing colocalization of Foretinib-TCO (11) with some of the other known targets. Consistent with the enzymatic assay result, Foretinib-TCO (11)/Tz-CFDA shows good co-localization with the membrane signal from the MET, AXL, and RON antibody staining ( Figure 5). In addition the red fluorescent nuclear signal from immunostaining with the RON antibody [28] shows good co-localization with Foretinib-TCO (11)/Tz-CFDA staining. There was less co-localization with PDGFRα and KDR, likely due to the low expression of these proteins in OVCA429 cells ( Figure S4). Overall, we observed very good co-localization between the specific MET inhibitor (PF04217903-TCO) and MET. In contrast, Foretinib-TCO/Tz-CFDA staining not only appears in the membrane but also in the cytoplasm as well as the nucleus. Based on recent kinome studies [29], these patterns are most likely explained by the very broad inhibition of > 30 kinases throughout the cell by this compound. Therefore, our results shows that the Foretinib-TCO (11)/Tz-CFDA bioorthogonal two-step labeling procedure is useful for imaging polypharmacologic distribution of drug targets not only in fixed cells, but also in live cells. Moreover PCID, combining with the image based phenotypic screening system, will be a useful chemical tool for new polypharmacologic drug discovery.

Discussion
The ability of many drugs, often unintended, to interact with multiple proteins is commonly referred to as polypharmacology. Using Foretinib as a model system we show that bioorthogonal conjugates of the drug can be used to image drug binding to its multiple targets. Specifically, we i) investigated the IC 50 for most known binding partners and showed them to be similar to that of the parent drug, ii) showed that the drug distribution inside single cells can be visualized (important in confirming target localization) and iii) that cellular images reflect the distribution and abundance of multiple drug targets (as shown by Western and comparative selective MET imaging).
The approach shown here has two practical and potentially far reaching applications: i) synthesis of imageable drug analogs to study their distribution in vivo and ii) to develop new multi-target imaging agents that can be translated to whole body and clinical imaging. While we used a bioorthogonal twostep procedure for cellular imaging to optimize spatial resolution we anticipate that the development of small footprint fluorochromes will ultimately enable in vivo imaging at the whole body level [30]. This will be further facilitated by the concurrent development of red-shifted cell permeable fluorochromes [31], turn-on fluorochromes [32], fluorochromes with improved pharmacokinetics and fluorochromes that can be labeled with radioisotopes for PET imaging [33].   After fixation with 2% paraformaldehyde, MET was labeled using a MET primary antibody and AlexaFluor 647 labeled secondary antibody (i-l). After nuclear staining with Hoechst 33342 (a-d) for 10 min, 40X images were collected using a DeltaVision microscope. Note the excellent co-localization between the MET antibody and affinity ligands on the membrane of the cells (m-p). Scale bar: 10 μm.  PF04217903-TCO (15) (d, e and f) or Foretinib-TCO (11) (g, h and i), washed, and incubated for 30 min with 1 μM Tz-CFDA for bioorthogonal reaction inside living cells. After fixation with 2% paraformaldehyde, MET was labeled using a MET primary antibody and AlexaFluor 647 labeled secondary antibody (i-l). After nuclear staining with Hoechst 33342 (blue nuclei) for 10 min, 40X images were collected using a DeltaVision microscope. Note the striking co-localization between the selective MET imaging agent and the MET antibody stain. Foretinib-TCO shows a much broader intracellular distribution. File S1. NMR-spectra of all the products. (PDF)