Supporting information Single-cell and subcellular pharmacokinetic imaging allows insight into drug action in vivo Greg Thurber 1, Katy Yang 1, Thomas Reiner 1, Rainer Kohler 1, Peter Sorger 2, Tim Mitchison 2, Ralph Weissleder 1,2, * 1 Center for Systems Biology, Massachusetts General Hospital, 185 Cambridge St, CPZN 5206, Boston, MA 02114, USA. 2 Department of Systems Biology, Harvard Medical School, 200 Longwood Ave, Boston, MA 02115, USA. *R. Weissleder, MD, PhD Center for Systems Biology Massachusetts General Hospital 185 Cambridge St, CPZN 5206 Boston, MA, 02114 617-726-8226 rweissleder@mgh.harvard.edu Supplementary Fig. S1: Comparison of Conventional PK analysis and SCPKI Supplementary Fig. S2: Chemistry Supplementary Fig. S3: Characterization of companion imaging probe Supplementary Fig. S4: Histology of MDA-MB-231-apple xenograft Supplementary Fig. S5: Correlation with PARP Supplementary Fig. S6: Cellular kinetics Supplementary Fig. S7: Blood half-life and tissue distribution Supplementary Fig. S8: Real time distribution in mammary fat pad orthotopic tumor model Supplementary Fig. S9: Cellular imaging of an orthotopic tumor Supplementary Fig. S10: Real time distribution in normal tissue the mouse ear Supplementary Fig. S11: Additional models of PARPi distribution Supplementary Table S1: Modeling parameters Supplementary References
Supplementary Fig. S1: Comparison of Conventional PK analysis and SCPKI Whole body pharmacokinetics (top row) are relatively standardized measurements taken in drug development. The drug is studied in animal models and the clinic typically using mass spectrometry for drug detection. Data is analyzed using compartmental or non-compartmental analyses of plasma concentrations. Single cell pharmacokinetic imaging (SCPKI, bottom row) seeks to integrate cell signaling pathway analysis and drug perturbation with therapeutic efficacy (bottom right). Using intravital microscopy techniques, companion imaging agents can be used to track drugs in vivo in individual cells with variable genetic and epigenetic backgrounds. Various cell signaling reporters can then be integrated into the framework to provide real-time feedback on signaling and drug effects.
Supplementary Fig. S2: Chemistry A. PARP1i companion imaging drug synthesis. B. Mass spectrometry of purified compound. C. Evaporative Light Scattering Detector (ELSD) high-performance liquid chromatography (HPLC) chromatogram.
Supplementary Fig. S3: Characterization of companion imaging probe A.PARPi companion imaging drug fluorescence absorption and emission spectrum. B. Phantom imaging data used to generate a calibration curve for drug quantification. C. IC 50 curves of PARP inhibition with either the therapeutic drug or imaging drug. While there is some loss in IC 50, it is still in the low nm range. D. The addition of 100-fold higher parent drug completely blocked uptake of the companion imaging drug in cell culture.
Supplementary Fig. S4: Histology of MDA-MB-231-apple xenograft Staining of PARP-1 indicates high nuclear levels of the primary drug target (top left). Little to no PARP-2 is detected in these tumors (top middle). PARP-3 is localized in a perinuclear fashion in the tumor cells (top right). The affinity for the companion imaging drug is 1000-fold lower for PARP-3 than PARP-1, consistent with the in vivo localization of the drug (bottom). Scale bar = 100 µm.
Supplementary Fig. S5: Correlation with PARP A. Co-localization of an anti-parp antibody and its companion imaging drug. B. Pearson correlation coefficient and scatter plot. C. Linear correlation between PARP expression and drug binding. Scale bar = 20 µm.
Supplementary Fig. S6: Cellular kinetics A. Imaging time series of drug uptake into single cancer cells in vitro (MDA-MB-231 breast cancer cells). Cells were treated for 5 minutes with 1 µm of imaging drug and imaged in realtime (using a Deltavision epifluorescence microscope). The medium was removed, cells were washed once, and the medium replaced. This was followed by imaging of cellular wash-out for an additional 5 minutes. B. Temporal kinetics in a representative cell. To obtain cellular uptake and wash-out rates, the signal in the nucleus and cytoplasm was quantified following background subtraction. Scale bar = 20 µm.
Supplementary Fig. S7: Blood half-life and tissue distribution A. Systemic clearance of the companion imaging drug was measured by retro-orbital blood sampling. (T 1/2 = 18 minutes; 77% redistribution with a 5 minute half-life and a 23% clearance phase of 60 minutes). B. Signal in individual vessels was measured in a region of interest within tumor vessels to determine the difference between individual tumor vessels and systemic drug concentrations. The tumor vessel signal was found to closely mimic systemic drug concentration, indicating blood flow and high plasma protein binding resulted in little depletion along the length of the vessel. C. Regions of interest close to a tumor vessel (green circles) and 60 µm away (blue circles) exhibited high variability in drug concentration at early time periods post-injection (orange shaded box). Drug gradients largely disappeared by 1 hour post-injection (green shaded box).
Supplementary Fig. S8: Real time distribution in mammary fat pad orthotopic tumor model Imaging studies were carried out in an orthotopic tumor model of breast cancer. MDA-MB-231 apple cells were injected in the mammary fat pad, allowed to develop for 1 month, and imaged similar to the window chamber model (Fig. 2). No qualitative differences were seen between the orthotopic model and window chamber, and drug permeability was not statistically different than the window chamber model (see Supplementary Fig S9 and S10). Scale bar = 50 µm.
Supplementary Fig. S9: Cellular imaging of an orthotopic tumor Four hours after injection of the imaging drug, nuclear localization of the imaging drug was apparent. Both red fluorescent tumor cell nuclei (red, MDA-MB-231 cells) and host cells were labeled with the drug (green) with a relatively even distribution compared to the microvasculature (blue). Scale bar = 30 µm.
Supplementary Fig. S10: Real time distribution in normal tissue the mouse ear For some molecules, such as antibodies and nanoparticles, the macromolecular permeability has a large impact on uptake. This results in distribution differences based on the maturity of tumor blood vessels and tumor location. Theoretical models 29 indicate that small molecules are not strongly affected by this permeability but are more impacted by tumor blood flow or blood vessel distribution. To test this hypothesis, uptake in the mouse ear was imaged, and the permeability was not statistically different than the window chamber or orthotopic mouse model. A maximum intensity projection of the ear (left) 4 hrs after injection shows the even distribution within normal tissue and similar uptake in individual cells (right). Scale bar on left = 100 µm, scale bar on right = 10 µm.
Supplementary Fig. S11: Additional models of PARPi distribution A. A simulation showing the effect of intermittent blood flow. The vessel on the left side of the image has no blood flow for the first 15 minutes, then spontaneously becomes perfused. The vessel on the right of the image maintains adequate blood flow over the entire 60 minute simulation. B. Simulation of drug distribution in a poorly vascularized region of a human tumor after twice-daily oral dosing. Maximum gradients are shown (for clarity without individual nuclei) immediately after and just before an oral dose. The slower uptake and clearance in humans results in higher drug concentrations (2-3 µm), lower drug gradients, and a more ubiquitous distribution.
Supplementary Table S1: Modeling parameters Parameter Value Reference Permeability 4.6 µm/s Measured in vivo Diffusion Coefficient 300 µm 2 /s 55 Non-specific Uptake 18 mol non-specific uptake permol in Ratio solution Measured in vitro Binding Affinity 12.2 nm Measured in vitro Nuclear Uptake Rate 0.0105 /s Measured in vitro Plasma Free Fraction 0.77% Measured in vitro Target Concentration 3 µm Measured in vitro and in 56 Plasma Clearance 5.2 min redistribution half-life (77%) 59.4 min clearance half-life (23%) Measured in vivo Parameters used in FEM simulations. Individual parameters including drug permeability, specific and non-specific uptake rates, mouse plasma protein binding, target concentration, and plasma clearance were measured in separate experiments. Plasma protein binding was measured using rapid equilibrium dialysis (Pierce), and the non-specific uptake ratio and nuclear uptake rate were fit from live cell imaging experiments (Supplementary Fig S6). Permeability was measured using the early phase uptake from intravital microscopy experiments as published for macromolecules 53, the target concentration was measured using drug depletion from bulk media in cell culture while titrating drug concentrations, and the binding affinity was done using an enzyme inhibition assay (Trevigen, Gaithersburg, MD). While plasma protein binding often has opposing effects in therapeutic efficacy,the free fraction has a large impact on distribution after a single bolus dose 54-56.
Supplementary References 53. Gerlowski, L. E. & Jain, R. K. Microvascular permeability of normal and neoplastic tissues. Microvasc Res 31, 288-305 (1986). 54. Smith, D. A., Di, L. & Kerns, E. H. The effect of plasma protein binding on in vivo efficacy: misconceptions in drug discovery. Nat Rev Drug Discov 9, 929-939 (2010). 55. Pruijn, F. B. et al. Extravascular transport of drugs in tumor tissue: effect of lipophilicity on diffusion of tirapazamine analogues in multicellular layer cultures. Journal of Medicinal Chemistry 48, 1079-1087 (2005). 56. D Amours, D., Desnoyers, S., D Silva, I. & Poirier, G. G. Poly(ADP-ribosyl)ation reactions in the regulation of nuclear functions. Biochem J 342, 249-268 (1999).