Nano-palladium is a cellular catalyst for in vivo chemistry

Palladium catalysts have been widely adopted for organic synthesis and diverse industrial applications given their efficacy and safety, yet their biological in vivo use has been limited to date. Here we show that nanoencapsulated palladium is an effective means to target and treat disease through in vivo catalysis. Palladium nanoparticles (Pd-NPs) were created by screening different Pd compounds and then encapsulating bis[tri(2-furyl)phosphine]palladium(II) dichloride in a biocompatible poly(lactic-co-glycolic acid)-b-polyethyleneglycol platform. Using mouse models of cancer, the NPs efficiently accumulated in tumours, where the Pd-NP activated different model prodrugs. Longitudinal studies confirmed that prodrug activation by Pd-NP inhibits tumour growth, extends survival in tumour-bearing mice and mitigates toxicity compared to standard doxorubicin formulations. Thus, here we demonstrate safe and efficacious in vivo catalytic activity of a Pd compound in mammals.


Supplementary Figure 6. Time-staggered administration of Pd-NP and its substrate leads to tumorselective activity. a-d)
In the fibrosarcoma tumor model, Pd-NP was first injected followed by NIR-labeled Alloc 2 R110 NP at the indicated time points, and 24 h later organs were excised and analyzed by fluorescence for R110 uncaging (a-b) and accumulation of the Alloc 2 R110 NP vehicle (c-d). Data are shown relative to fluorescence in the tumor for each animal and across all replicates and time-points for each organ, such that each dot represents one organ measurement, lines denote averages across n=3 replicates for each organ, and bars denote mean ± SEM (n = 15) across all organs (*p=0.003, two-tailed t-test). e) 5 h time-staggering and tissue analysis was performed as described above (see a-d), using i.p. injections in the ES2 orthotopic OVCA model and i.v. injections in the HT1080 fibrosarcoma model; n.d., not detected (mean ± SEM, n=3).

Supplementary Figure 7. Fluorescent properties of Pd-NP. a)
Fluorescence excitation and emission spectra of PdCl 2 (TFP) 2 in HBSS/DMSO. b) Pd nano-encapsulation quenches fluorescence, which increases 36-fold upon dissolution of the nanoparticle via DMF addition. c) Fluorescent microscopy reveals intracellular fluorescence generated by the 25 μM Pd-NP within cancer cells following 24 h treatment. Inset: untreated control. Scale bar, 100 μm. d) Intracellular fluorescence from 24 h Pd-NP treatment increases with dose, measured by microscopy in HT1080 cells (mean ± SEM, n = 9). Figure 9. Local Pd catalyst activity depends on proximity to tumor microvasculature. HT1080 tumors were treated with 25 mg kg -1 Alloc 2 R110 NP 1 h following treatment with 50 mg kg -1 Pd-NP by i.v. injection. After 2 h post-injection with Alloc 2 R110, tumors were imaged for co-localization between microvasculature (still visible by circulating Alloc 2 R110 NP), Pd complex, and uncaging of its substrate. b) Profiles of fluorescence intensity were measured as a function of distance to the nearest vasculature, focusing on local tumor regions that lacked substantial vasculature identified by 3D z-stack image datasets. Thick line and shading denote mean ± SEM, respectively (n=7). Scale bar = 100 μm. Figure 10. Tumor-associated host phagocytes accumulate Pd-NP. a-b) Pd-NP treatment followed by injection of Alloc 2 R110 NP with NIR-labeled NP vehicle (via PLGA-BODIPY630 co-encapsulation) leads to local Pd-NP activity in both tumor cells and tumor-associated host phagocytes 3 h (a) and 24 h (b) following injection (n ≥ 3). Scale bar, 100 μm. c) Immunofluorescence of NIR-labeled NP vehicle 24 h postinjection shows high NP accumulation in tumor cells and F4/80+ host phagocytes in HT1080-53BP1-mApple xenograft tumors (n = 4). Scale bar, 50 μm. d) FACS analysis (n ≥ 4) of NIR-labeled NP vehicle 24 h postinjection shows NP uptake primarily in tumor cells (HT1080), MΦ (CD45 + CD11b + Lin -Ly6C -F4/80 + CD11c + ), with little additional accumulation in monocytes and other cell populations (including lymphocytes, neutrophils, along with CD45fibroblasts, pericytes, endothelium, and other stromal populations). Graph shows distribution of total NP within the bulk tumor as a function of cell-type. Figure 11. Characterization of proDOX activation mediated by Pd-NP. a-b) 10 μM Pd-NP was incubated with 10 μM proDOX NP for 24 h in HBSS at 37 o C, and was analyzed by LC/MS before (a) and after (b) incubation. Plots at left show fluorescence detection as a function of elution time, while plots at right show corresponding HR-ESI-MS (negative mode) mass fragmentation patterns of the indicated peaks for proDOX (C 31 H 32 NO 13 10 μM PdCl 2 (TFP) 2 was incubated with 5 μM proDOX in HBSS at 37 o C, and proDOX conversion to DOX was monitored by HPLC fluorescence chromatography (means ± SEM, n = 2). d) DLSmeasured distribution of proDOX-encapsulated PLGA-PEG NP according to diameter. e) ProDOX NP is stable in HBSS at 37 o C over time (mean ± SD, n = 3). f) In vitro release of proDOX from NP encapsulation, measured in HBSS at 37 o C. g) In vitro cell count (measured by resazurin-based assay) of HT1080 after 72 h treatment with varying doses of Pd-NP and DOX-NP (as compared to proDOX-NP co-treatment, Fig. 4). Data are means ± SEM, n = 2.

Supplementary Figure 12. Characterization of proDOX activation by Pd-NP in cell culture. a)
Endogenous DOX and proDOX (1 μM) fluorescence was imaged in ES2 OVCA cells after 16 h incubation, in the presence or absence of 30 μM Pd-NP co-treatment, after washing and DNA counter-staining with Hoechst 33342. b) Corresponding to images in a, average cell profiles of hoechst and drug fluorescence were quantified to determine selective nuclear accumulation of proDOX only after Pd-NP co-treatment (lines and shading denote means ± SEM, n>10). c) Co-treatment of proDOX (1 μM) with Pd-NP (30 μM) leads to 51% drug activation in HT1080 cells after 12 h treatment, determined after washing, lysing, and analyzing using MeCN extraction and HPLC fluorometric chromatography. Figure 13. Nano-palladium efficacy in fibrosarcoma and OVCA tumor models. a) Representative images of excised HT1080 tumors after 8 days treatment (see Fig. 5d). b-c) Human ES2 ovarian cancer cells were administered intraperitoneally and 3 days later were treated with either the drug-free vehicle or the combination of Pd-NP followed by proDOX NP. b) Combination Pd-NP and proDOX treatment (purple line) extends survival compared to the vehicle control (black line) (p = 0.002, log-rank test, n ≥ 4). c) 11 days post-treatment in the ES2 intraperitoneal xenograft, the control group developed significant ascites while the treatment groups did not (data are means ± SEM; two-tailed t-test, n = 4). d) proDOX and activated DOX were quantified by HPLC from two different xenograft models, 24 h post-treatment (mean ± 95% C.I.; n ≥ 3; n.d. = not detected). e) Pre-treatment with Pd-NP followed by proDOX leads to drug accumulation in tumor nuclei of orthotopic A2780CP ovarian tumors. 24 h post-treatment, tumors were excised and imaged for colocalization between intrinsic doxorubicin fluorescence and DNA (via Hoechst 33342 counterstain); lectin marks tumor vasculature. f) Traces show mean (thick line) ± SEM (shading) in single-cell profiles of drug fluorescence across n ≥ 4 tumors. Figure 14. Selective proDOX activation minimizes systemic drug exposure. a-b) 24 h following q4dx2 i.v. treatments in the fibrosarcoma mouse model, organs were excised and analyzed for DOX (a; n = 3) and proDOX (b; n ≥ 2) concentrations using HPLC (mean ± SEM; n.d., not detected). c-d) 24 h after i.p. treatment, organs were excised from OVCA-bearing mice and imaged for co-localization between intrinsic doxorubicin fluorescence (red) and DNA via Hoechst 33342 counterstain (n ≥ 4); lectin-DyLight649 labels vasculature (light blue). Data show organs relevant to drug toxicity. d) Using data as in c, drug accumulation in nuclei of key organs related to toxicity was quantified based on Hoechst co-localization. Traces show mean (thick line) ± SEM (shading) across populations of single cells in n ≥ 4 tumors (corresponds to Fig. 6b-c). e) No treatment caused significant loss in body weight in nu/nu mice bearing HT1080 tumors (α = 0.05; n ≥ 5; twotailed t-test). Purple arrows mark treatment days. Figure 15. Imaging suggests mechanisms of mitigated off-target prodrug activation in clearance organs. a) 24 hr following Pd-NP treatment and 19 hr following treatment with NIR-labeled proDOX NP vehicle, nu/nu mice were injected with lectin-rhodamine to label vasculature. 5 min later, livers were excised and imaged for patterns of Pd complex accumulation (imaged by its intrinsic fluorescence), proDOX NP vehicle, and vasculature. Results show accumulation in perivascular phagocytes (which can include resident liver macrophage, Kupffer cells) rather than larger, morphologically distinct hepatocytes. b) The same experiment was performed as in a, but with imaging intrinsic fluorescence of proDOX-encapsulated NP (10 mg kg -1 ) rather than lectin-rhodamine. Results show co-accumulation of proDOX, the proDOX NP vehicle, and Pd complex in similar phagocyte populations. However, the intra-phagocyte distribution of Pd complex appears distinct from proDOX at the subcellular level (Pearson's correlation coefficient after thresholding = -0.12 ± 0.3, mean ± SD, n=3). c) ProDOX accumulates more in perivascular phagocytes of the liver compared to morphologically distinct hepatocytes. d) Kidneys were imaged similarly to the liver in b, showing distinct localization of Pd complex and its substrate proDOX (Pearson's correlation coefficient after thresholding = -0.2 ± 0.02, mean ± SD, n=3). Table 2. Pd-NP impact on complete blood counts. Complete blood counts were measured in HT1080 tumor-bearing mice after 50 mg kg -1 treatment with Pd-NP or PLGA-PEG NP vehicle control (q4dx2) 24 h after last treatment (data are means, n=3, p-value according to two-tailed t-test).

Supplementary Methods:
All reagents were purchased from Sigma-Aldrich and used without further purification unless otherwise noted.
Deuterated solvents were obtained from Cambridge Isotope Laboratories, Inc. NMR spectra were recorded on a Bruker Avance III 400 spectrometer and calibrated to the residual proton resonance and the natural abundance 13  Palladium catalysts were initially selected based on previous publications describing their synthesis and activity under relatively mild reaction conditions 1-11 . PdCl 2 (TPPTS) 2 , Pd(TPPTS) 4 and Pd(TFP) 4 were synthesized according to published procedures 12,13 . Bis-allyloxycarbonyl-and bis-propargyloxycarbonyl-protected rhodamine 110, and coumarin precursor were all synthesized according to published procedures 14-17 . All catalysts were characterized by NMR for purity and successful synthesis following previously described protocols. In all cases purity was found to be >99%. For the most important compound 1, prepared as Preparation of alloc-doxorubicin (proDOX): Following a previously described procedure 21 we observed formation of bis-allyloxycarbonyl-substituted doxorubicin as an inseparable by-product of proDOX. For that reason we modified this procedure as follows: doxorubicin hydrochloride (580 mg, 1.0 mmol) was suspended in anhydrous DMF (100 mL) and 4-(dimethylamino)pyridine (366.5 mg, 3.0 mmol, 3 eq) was added. After stirring at room temperature for 15 min the solution was cooled to -78°C and allyl chloroformate (106.3 µL, 1.0 mmol, 1 eq) dissolved in anhydrous DMF (20 mL) was dropwise added within 10 min. The mixture was slowly warmed to room temperature and stirring was continued for 5 h. HPLC analysis showed ~50% conversion, which is why the solution was again cooled to -78°C followed by addition of further allyl chloroformate (53.2 µL, 0.5 mmol, 0.5 eq) in anhydrous DMF (10 mL). The mixture was slowly warmed to room temperature and stirring was continued for 2 h. The solvent was evaporated and the residue was purified by reverse phase chromatography (60 g C 18 silica, water/acetonitrile gradient elution, 0.1% formic acid added) to obtain pure proDOX as a red solid (490 mg, 78%). NMR data matched that reported previously 21 Fig. 11a).
In vitro NP characterization: For cytotoxicity assays, 5,000 cells per well were added to 96-well plates, treated the next day with compound or the appropriate buffer control (DMF or drug-free PLGA-PEG nanoparticles, as appropriate), and assessed for viability 72 hr later using PrestoBlue (Life Technologies) following the manufacturer's protocol. IC 50 values for each compound were calculated by interpolating from an 11-point dose-response curve. For in vitro NP payload release measurements, samples were incubated in PBS at 37 o C and filtered using 30 kDa molecular weight cutoff filters (Millipore Amicon) at the indicated time-points. Flow-through was measured for drug content by absorbance using a Nanodrop spectrophotometer. At the end of the time-course, NP was dissolved in DMF and measured for drug content by absorbance. Additionally, flow-through and remaining NP was assayed for activity by incubation with 10 μg mL -1 Alloc 2 R110 in PBS + 10% DMF for 2 h. Initial concentrations were 500 μM. For tested release across multiple buffers ( Supplementary Fig. 2f), samples were incubated at 37 o C and filtered using 100 kDa molecular weight cutoff filters (Millipore Amicon) at a single time-point, with an initial concentration of 1mM (12 h). Flow-through was measured for drug content by absorbance using a Nanodrop spectrophotometer, after accounting for buffer-dependent absorbance properties. DMEM + 10% FBS + L-glutamine and Pen/Strep was used as the growth media. Lysis buffer consisted of 150 mM NaCl, 0.1% Triton X-100, 50 mM Tris-HCl, pH 8.0. In the figure, T-X denotes Triton X-

100.
Computational image analysis: Unless otherwise stated, all experiments were independently performed at least twice as biological replicates, and t-tests assumed normal or log-normal distribution of data with equal variance. Where appropriate, equal variance was confirmed by F-test and normality was confirmed by the D'Agostino and Pearson omnibus test (Prism GraphPad); when tests showed non-Gaussian distribution, nonparametric Mann-Whitney U test was used (Fig. 5c-d). Intravital microscopy images were analyzed using either Matlab (Mathworks) or ImageJ and were pre-processed using background subtraction based on data acquired immediately prior to NP injection. For vascular half-life calculations (Fig. 3), fluorescence time-lapse values from multiple vessels across multiple animals were recorded, averaged, and fit to an exponential decay  Fig. 8) was calculating using the ratio of fluorescence, but using the fluorescence quantile yielding max turn-on (97 th percentile) rather than the simple mean. Average 53BP1 puncta (Fig. 5b) was determined by counting foci per individual cell, averaging values across the entire cell population (n > 100), and normalizing these values to the untreated control values. Nuclear drug accumulation (Fig. 6c) was determined by integrating fluorescence intensity that co-localized with Hoechst DNA counterstain compared to elsewhere in the cell, averaged across a population of single-cells across multiple (n = 4) tumors.
Nuclear doxorubicin imaging: 24 h following drug treatment, mice bearing intrabursal A2780CP ovarian cancer tumors were treated with lectin-DyLight649 Lycopersicon Esculentum (Tomato) Lectin (Vector Labs) for 30 min to label vasculature, anesthetized, perfused with formalin via cardiac puncture, and organs were removed for 6 h formalin fixation at r.t. To reduce autofluorescence, tissue was then cleared under rocking agitation at 37 o C for 12 h according to previously published methods 23 in the presence of 0.1% Hoechst 33342, and immediately imaged with an Olympus FV1000 confocal-multiphoton imaging system. ProDOX was dosed at 32 μmol kg -1 while DOX and DOX-NP was dosed at the MTD of 16 μmol kg -1 . For Supplementary Fig. 15a, lectin-Rhodamine Griffonia Simplicifolia Lectin I (Vector Labs) was used. 16 μmol kg -1 proDOX encapsulated in PLGA-BODIPY630 labeled PLGA-PEG NP was used in Supplementary Fig. 15b-d, and organs were imaged immediately following sacrifice and excision, without fixation.
Bioluminescence imaging: To assess tumor burden in the A2780CP and ID8 orthotopic tumor models, bioluminescence imaging was performed by stable transgenic overexpression of firefly luciferase 25 , and imaging was performed using the SPECTRAL Ami-X (Spectral Instruments Imaging). Animals were anesthetized using isoflurane, injected with 7 mg D-Luciferin i.p., imaged every 5 minutes until maximum radiance was achieved, and concordantly x-ray imaged for anatomical reference.
Biodistribution: Palladium uptake into tissues was assessed by ICP-MS. Briefly, organs were harvested 24 h post-treatment after blood draw via cardiac puncture and cardiac perfusion with 10 mL PBS under isoflurane anesthesia. As previously described 26 , 10-150 mg tissue was added to 1.5 mL 70% nitric acid, digested at Toxicity: Complete blood count and blood chemistry panels were performed by the MGH Center for Comparative Medicine Veterinary Clinical Pathology Laboratory. Blood draws were performed by cardiac puncture into EDTA coated BD microtainers for blood count (which included manual reticulocyte count), and into lithium-heparin BD microtainers for blood chemistry analysis. Murine IgE was measured according to manufacturer's guidelines (eBioscience Ready-SET-Go ELISA), with reference to a 6-point standard curve (R 2 > 0.97).

Solvent Pd administration:
In order to safely administer un-encapsulated PdCl 2 (TFP) 2 into tumor-bearing mice, complex was dissolved in DMSO at 40 mg ml -1 , added to 7 µl DMAC/Solutol and 100 µl PBS, vortexed and sonicated, and slowly administered via tail-vein catheter and syringe pump while the animal was under isoflurane anesthesia. DMSO was confirmed to not significantly impact Alloc 2 R110 activation, compared to DMF solvent, in an in vitro assay in HBSS buffer as in Fig. 1 (p>0.05; n=3).