Layer‐by‐layer nanoparticles for novel delivery of cisplatin and PARP inhibitors for platinum‐based drug resistance therapy in ovarian cancer

Abstract Advanced staged high‐grade serous ovarian cancer (HGSOC) is the leading cause of gynecological cancer death in the developed world, with 5‐year survival rates of only 25–30% due to late‐stage diagnosis and the shortcomings of platinum‐based therapies. A Phase I clinical trial of a combination of free cisplatin and poly(ADP‐ribose) polymerase inhibitors (PARPis) showed therapeutic benefit for HGSOC. In this study, we address the challenge of resistance to platinum‐based therapy by developing a targeted delivery approach. Novel electrostatic layer‐by‐layer (LbL) liposomal nanoparticles (NPs) with a terminal hyaluronic acid layer that facilitates CD44 receptor targeting are designed for selective targeting of HGSOC cells; the liposomes can be formulated to contain both cisplatin and the PARPi drug within the liposomal core and bilayer. The therapeutic effectiveness of LbL NP‐encapsulated cisplatin and PARPi alone and in combination was compared with the corresponding free drugs in luciferase and CD44‐expressing OVCAR8 orthotopic xenografts in female nude mice. The NPs exhibited prolonged blood circulation half‐life, mechanistic staged drug release and targeted codelivery of the therapeutic agents to HGSOC cells. Moreover, compared to the free drugs, the NPs resulted in significantly reduced tumor metastasis, extended survival, and moderated systemic toxicity.


| INTRODUCTION
Epithelial ovarian cancer (EOC) is the second most common gynecological cancer and the leading cause of death from gynecologic malignancies in the developed world. 1,2 High-grade serous ovarian cancer (HGSOC) accounts for more than 70-80% of ovarian cancerassociated mortalities. 2,3 This high mortality rate is attributable to an aggressive phenotype, diagnosis at advanced stages, and the development of resistance against mainstay platinum-based therapies. 4 Cisplatin and other platinum-based chemotherapies efficiently bind and induce DNA double-strand breaks (DSBs) and apoptosis in cancer cells 5 and are the current cytotoxic drugs of choice for ovarian cancer and other carcinomas. 6,7 However, the rapid development of resistance often limits the effectiveness of platinum-based drugs alone against solid tumors such as HGSOC. 8,9 Among solid tumors, 15-46% of HGSOC, 40-66% of triplenegative breast cancer, and 2-9% of non-small cell lung carcinoma are estimated to carry mutations in the p53, BRCA1, BRCA2, and PTEN genes, which are required for DNA damage repair via homologous recombination (HR). 5 Consequently, using DNA-damaging agents in combination with inhibitors of DNA damage repair proteins is a very attractive strategy. In the past 5 years, new classes of inhibitors have emerged against poly(ADP-ribose) polymerases (PARPs), a family of nuclear DNA damage repair enzymes with a role in the maintenance of genomic stability. 10 PARPs perform this function by initiating base excision repair and nucleotide excision repair of DNA single-strand breaks (SSBs). 11 The inhibition of SSB repair by PARP inhibitors (PARPis) induces and confers sensitivity and synthetic lethality to cells with defective HR-directed DSB repair. 11,12 PARPis exhibit synergistic activity when combined with a DNA-damaging agent by interfering with DNA repair and potentiating the activity of the chemotherapeutic agent. The potentiation effect is achieved via inhibition of the catalytic activity of PARP by PARPis, or by trapping PARP at SSB sites, thereby stalling the replication fork and DNA transcription 10,11 and eventually leading to apoptosis. Different classes of PARPis of varying toxicity and efficacy have been developed. 13,14 Of the five most clinically relevant PARPis, three of them: AZD2281 (olaparib, Lynparza; AstraZeneca, UK), 15 niraparib (Zejula, MK4827 Tesaro, Waltham, MA), 16 and rucaparib (Rubraca; Clovis Oncology, Boulder, CO) 17 are FDA approved for the treatment of recurrent EOC. BMN 673 and veliparib are under investigation in different phases of clinical trials. 10,11 The ability of DNA damaging agents to enhance apoptosis and reduce drug resistance in HR-deficient cells in tumors has led to a number of preclinical investigations. Rottenberg et al. 18 and Hay et al. 19 showed that the free-drug combination of AZD2281 with cisplatin or carboplatin significantly reduced resistance to platinum-based agents in BRCA1 mutated ovarian and breast cancer tumor-bearing mice and prolonged overall survival compared with either monotherapy. Others studies have shown high tolerance for AZD2281 alone but not in combination with other chemotherapies. 18 Several Phase I-III clinical trials have been conducted to evaluate AZD2281 in combination with cisplatin and other chemotherapies in advanced breast and ovarian cancers in patients with BRCA mutation. 20,21 Overall, the data indicated that the high-dose combination of cisplatin with AZD2281 was not tolerable in most patients. However, a moderate dose of cisplatin (60 mg/m 2 ) and AZD2281 (50 mg/twice daily) was better tolerated in most patients. In addition, the AZD2281 and cisplatin combination prolonged progression-free survival in patients compared to monotherapy, with tolerable side effects. 20,22 BMN 673 (talazoparib) remains one of the most promising PARP1/2 inhibitors, and we have also tested BMN 673 alongside AZD2281 as monotherapies or in combination with cisplatin. 23 Preclinical testing has shown that BMN 673 exhibits superior PARP inhibition and antitumor activity in vitro [24][25][26][27] and in vivo. 28 A number of completed Phase I and II clinical trials of BMN 673 have evaluated its tolerability, efficacy, pharmacokinetics, and safety in both ovarian and metastatic breast cancer 24,29 and Phase III clinical trials are currently underway. 11 In Phase I and II clinical trials, the combination of BMN 673 with carboplatin showed synergy and significant therapeutic effects. However, hematologic toxicity was pronounced, particularly in gBRCA patients. 29 The clinical benefit of BMN 673 was 56-86% in both breast and ovarian cancer patients, with higher efficacy for the combination with carboplatin. 29 Although combinations of PARPis with cisplatin are efficacious, these preclinical and clinical trials of AZD2281 and BMN 673 alone or in combination with chemotherapies have revealed a number of hurdles that remain to be overcome to harness their full antitumor potential in the clinical setting. First, PARPis are highly hydrophobic, with limited bioavailability and a relatively rapid plasma clearance rate.
Rothenberg et al. 18 described rapid plasma clearance of AZD2281 when delivered in free form in tumor-bearing mouse models. Second, cisplatin, which remains a key platinum agent for ovarian cancer therapy, is subject to the development of resistance in tumors and therefore is typically administered at a high dose in the clinic, leading to its well-known systemic toxicity. 25,26 Third, the therapeutic combination of cisplatin and AZD2281 is poorly tolerated in patients due to the overlapping toxicities of the two drugs 27 ; hence, only reduced doses have been evaluated in clinical trials. Fourth, the infusion and oral routes of administration of cisplatin and PARPis, respectively, can reduce medication compliance, leading to a less effective therapeutic response in patients. Moreover, to obtain the highest therapeutic index, both cisplatin and PARPi should be codelivered at their highest doses and at an appropriate therapeutic ratio to tumors, which is difficult to achieve via conventional free drug delivery approaches. Finally, the two drugs have different biodistribution profiles when administered via different routes by traditional approach. These factors affect the time it takes each drug to reach the tumor and the drug concentration delivered and can significantly affect treatment outcomes.
Codelivery of these drug combinations via a nanocarrier approach could significantly reduce or eliminate these hurdles. 30,31 Advances in nanotechnology and nanomedicine have provided new opportunities for synergistic combinations of therapeutic agents via single multicompartment nanoparticles (NPs). [32][33][34][35] The three main goals of this study are to (a) address the unmet clinical need for an effective and safe platform for the delivery of combination therapies to ovarian cancer; (b) design safe, full-dose delivery of cisplatin, and PARPis to overcome cisplatin drug resistance in ovarian cancer therapy; (c) evaluate potential systemic toxicity associated with this treatment platform. We describe a novel approach to provide safe therapeutic delivery of cisplatin and PARPis to tumors using the layerby-layer (LbL) polymeric liposomal NPs approach. These NPs achieve synergistic drug delivery while inherently addressing many of the challenges associated with the conventional delivery of cocktails of free drugs, such as lack of targeted mechanistic delivery, reduced drug blood circulation, and the use of dual routes. 32,36,37 The HAterminated outer layer of the LbL NPs enables CD44 receptor targeting on HGSOC tumors, while the pH-responsive poly( L -lysine) (PLL) layer facilitates tunable intracellular release of the therapeutic cargo in tumor cells. 36,38 We report the novel packaging of cisplatin with AZD2281 or BMN 673 in LbL NPs for orthotopic HGSOC therapy. We also perform a head-to-head comparison of the therapeutic efficacy of both free and nano-encapsulated delivery of AZD2281 and BMN 673 in vivo as a single maintenance agent in orthotopic HGSOC tumor-bearing mice. In summary, we observed an overall increase in survival and improved treatment outcomes in mice treated with the LbLencapsulated drug combination compared with free drug combination therapy. 39,40 2 | RESULTS 2.1 | LbL polymeric liposomal NPs exhibit controlled drug release and inhibit HR We designed single modular NPs for the efficient encapsulation of cisplatin and PARPi for synergistic dual-drug delivery (Figure 1a). Two PARPis were used: AZD2281 (olaparib), which is FDA approved for germline BRCA-mutated (gBRCAm) advanced ovarian cancer, 15 and BMN 673, which is currently in clinical trials for BRCA-deficient ovarian cancer patients. 28 Both AZD2281 and BMN 673 are very hydrophobic, with poor solubility of 0.1 mg/mL in water. 14 We successfully formulated liposomal NPs by self-assembly of the lipids DSPC (1,2-distearoyl-sn-glycero-3-phosphocholine, POPG (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1 0 -rac-glycerol)) (sodium salt), and cholesterol in a mass ratio of 56:39:5. The PARPis were introduced F I G U R E 1 Physiochemical characterization of the polymeric liposomal nanoparticles. (a) Illustration of the polymeric liposomal nanoparticle formulated by lipid self-assembly and layered with a polycation, poly( L -lysine) (PLL), and a terminal hyaluronic acid (HA) layer for CD44 targeting. The PARPi AZD2281 or BMN 673 was loaded into the liposomes lipid bilayer, and cisplatin was loaded into the core. (b) Electron micrographs of the nanoparticles. Left panel, liposomes alone. Middle panel, liposomes layered with PLL and HA. Right panel, magnification of layers ×4. (c) The hydrodynamic size of the nanoparticles increased by approximately 10 nm per layer. (d) The zeta potential confirms the transformation of the negatively charged (−41 ± 8 mV) liposomes surface to positive (31 ± 6 mV) upon layering of the polycation PLL, followed by net charge reversal to (−27 ± 8 mV) upon deposition of the polyanion HA. (e) The polydispersity index (PDI) revealed that the overall size of the nanoparticles remained homogeneous, with no second-degree aggregation formation during formulation. Combination release of (f) AZD2281 and cisplatin, (g) BMN 673 and cisplatin from the polymeric nanoparticles at 37 C in an excess volume of buffered citrate PBS in a time-dependent manner at pH 5.0 (**p < .001, ***p < .0001) than at pH 5.0. (h) release of cisplatin from polymeric nanoparticles at 37 C in an excess volume of PBS at pH 5.0 and pH 7.4 (***p < 0.0001) in a chloroform-ethanol mixture during liposome preparation. Due to their hydrophobicity, the PARPis partitioned into the liposomal bilayer ( Figure 1a). Cholesterol was added to stabilize and compact the liposomes. The liposome film was hydrated under sonication with 300 mM citric acid solution (pH 4). Cisplatin was dissolved in 0.9% sodium chloride solution (1 mg/mL; pH 7.4) under sonication at 65 C for 10 min to ensure complete drug dissolution prior to addition to the liposomal suspension under sonication for 15-30 min. The encapsulated NPs were washed with tangential flow filtration (TFF), and the amount of loaded cisplatin was determined using inductively coupled plasma mass spectrometry (ICP-MS). The encapsulation efficiency of cisplatin was 64%, with a drug/lipid ratio of 9.7% (w/w). The net encapsulation efficiency was 26% for AZD2281, corresponding to a drug/lipid ratio of 2.5% (w/w), and 21% for BMN 673, corresponding to a drug/lipid ratio of 2.4% (w/w), as determined by high performance liquid chromatograph mass spectrometry (HPLC/MS). The free drugs were initially purified by filtration through a sterile 0.2-μm filter membrane, followed by TFF. Final liposome purification and concentration were performed using TFF. 41 To enhance the structural stability, cell-targeting capability, and dual-drug release mechanism of the NP platform, we employed an LbL polyelectrolyte deposition approach. 32,36 Polycation PLL (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) was deposited on the negatively charged liposomes to provide stability and pH sensitivity, followed by deposition of the polyanion, hyaluronic acid (HA, 40 kDa), to form the terminal layer. HA is a ligand for the receptor CD44, which is highly over-expressed on most ovarian cancer cells, 42,43 and hence functions as a targeting layer on the NPs  (Figure 2a,b), we elected to further study OVCAR8 and COV362, which carry due to promoter hyper methylation and BRCA1 mutations 47 and have high and moderate CD44 expression levels, respectively. These cell lines are of interest to us because of the CD44-targeting ability of our HA-coated LbL NPs. Cell viability assays were performed to assess the efficacy of the free and encapsulated drugs in OVCAR8 cells (Figure 2c,d) and COV362 cells (Figure 2e,f) and to determine their apparent IC50s (Figure 2g). Previous preclinical studies have shown that BMN 673 is significantly more potent than other PARPis, 28,48 and in our cell viability assay (Figure 2g) we observed a significant difference (p < .05) in potency between BMN 673 and AZD2281. Interestingly, BMN 673 was also significantly more potent than cisplatin (p < .05) based on IC50 values, consistent with the literature. 28,48 Notably, OVCAR8 cells were more sensitive than COV362 cells to PARPi treatment, and similar dose-dependent cytotoxicity was observed in OVCAR4, Kuramochi, and OVISE cells ( Figure S1a-c). We believe that the effectiveness of the BMN 673-containing liposomes is due to its broad targeting and tight inhibition of PARP1 catalytic activity, as reported by Shen et al. 28 and Murai et al. 48 To determine if the cytotoxicity observed in the cell viability  Figure S1d). This observation is consistent with the status of BRCA1 in these cell lines. OVCAR8 cells carry residual BRCA1 expression, whereas COV362 carry BRCA1 mutation. 47,51 Overall, in both cell lines, the free drugs showed greater potency than the encapsulated drugs, most likely due to the faster delivery of the free drugs to cells; the encapsulated drugs require an initial step of CD44 receptor-mediated endocytosis of the LbL NPs as well as controlled release from the carrier. Finally, these findings are also consistent with the significant elevation of γH2AX(Ser139) expression compared to RAD51 in BRCA-mutated cell lines after exposure to DNA-damaging agents. 49,52 In addition to their role in DNA repair, PARPs are also substrates for Caspase 3 cleavage to initiate apoptosis. During apoptosis, PARP proteins undergo proteolytic cleavage, which can be monitored by western blot. The main PARP protein cleavage site is Asp214/Gly215, and the two resulting protein fragments can be used as markers of cleaved Caspase 3-mediated apoptosis activation. 53 We therefore

| Evaluation of the hematological toxicity of NPs versus free drug delivery in vivo
Overall, in vivo preclinical testing and clinical trials have shown that the significant treatment response of cisplatin and PARPI combination therapy is accompanied by intolerable overlapping systemic toxicity, which requires a significant reduction of the dosing of either cisplatin or PARPI or their combination. Here, we administered escalating doses of cisplatin, AZD2281, and BMN 673 as monotherapies or combination therapies. 61 Two different dose studies were performed.
In the first dose study, the mice received the selected dose of each drug or the combination for three consecutive days, followed by thrice-daily monitoring for 2 weeks. 61,62 In the second dose study, the mice received only a single dose and were again monitored for a 2-week period. In both studies, 6-to 8-week-old healthy immunocompetent BALB/c female mice were dosed with cisplatin, AZD2281, or BMN 673 individually or in combination as free drugs or encapsulated in NPs using the dosages previously described with moderate adjustment. 28 The standard procedure used is to determine the core NP drug dosing based on the cisplatin concentration.
In the first high-dose study ( Figure S2a  (e and f) The body weight distribution of the treatment groups was measured and graphed as scatter plots. Statistical significance was determined by one-way ANOVA with Bonferroni's multiple comparison tests. Data are presented as the mean ± SEM; *p < .05, **p < .01, ***p < .001. (g) Schematic illustration of the design of the polymeric liposomal nanoparticle assembly with loaded therapeutic cargo and the treatment mechanism. FD, denote free and NP-encapsulated nanoparticles. HGSOC, high-grade serous ovarian cancer; IV, intravenous increased survival to 71 days, compared to 63 days for the corresponding free drug treatments (p < .0002). Treatment with encapsulated BMN 673 increased survival to 64 days, compared to 58 days for the free drug treatment (p < .0002) (Figure 5c,d and Figure S4c,d). Overall, compared with free drugs, treatment with encapsulated drugs improved survival by 10.34-12.85% among the various treatment groups.
In untreated mice, tumors grew rapidly, and the mice were  (Figure 6b,c). We also performed immunohistochemical staining of tumor cells for the Paired-box gene 8 (PAX8) protein, which is used as a differential marker of EOC. 65 H&E staining was performed to examine and confirm tumor infiltration into key organs such as the intestine, spleen, and liver (Figure 6d). The increase in volume of peritoneal ascites also correlated with the mesenteric tumor burden across most of the treatment groups. Mice treated with PBS were used as a positive control, and tumor-free mice were used as a negative control without ascites ( Figure 6e). Taken together, these findings suggest that the application of targeted delivery of cisplatin and other therapeutics may be sufficient to mitigate platinum resistance particularly in BRCA-deficient cancers.

| Correlation of apoptotic markers and biochemical metabolites with treatment response
In this report, we describe the packaging of cisplatin, AZD2281, or BMN 673 monotherapies and cisplatin in combination with AZD2281 or BMN 673 in an LbL NP drug delivery platform for administration to orthotopic OVCAR8 HGSOC-xenograft tumor-bearing mice. This nano-delivery platform employs liposomes modified with polyelectrolyte nano-layers to provide improved biodistribution and a terminal functionality that enhances NP trafficking for drug delivery and colocalization into tumors. 36,37,66,67 The liposomal NPs are layered by electrostatic deposition of a polycation, PLL, to provide structural stability and pH-responsiveness in the tumor microenvironment, whereas the terminal polyanion layer of the NPs, HA, is critical for selective CD44 targeting of HGSOC tumor cells. 68 The extravasation of LbL NPs from the blood circulation into the tumor microenvironment (~pH 6.8) facilitates intracellular uptake and trafficking of NPs into the endosomal compartment. The prevailing low endosomal pH of 5.5 enhances the swelling and disassembly of the LbL architecture, resulting in mechanistic release of the PARPi, followed by cisplatin. This LbL NP drug delivery platform enables ratiometric, synergistic, and modular delivery of combinations of hydrophilic drugs, such as cisplatin, and highly hydrophobic drugs, such as PARPis, that conventional approaches cannot codeliver simultaneously. The mechanistic release of the PARPi followed by cisplatin ensures complete blockade of DNA repair by HR and effective induction of apoptosis, leading to synthetic lethality and a level of effectiveness not achievable by conventional anticancer drug delivery approaches. were used as received from the vendor. All solutions were sterile filtered with a 0.2-μM filter prior to use. The liposomes suspension was diluted to 1 mg/mL and added dropwise to 45 mL of 500 μM PLL with rapid stirring, followed by TFF purification. The concentrated PLLlayered liposomes were diluted to 1 mg/mL and added dropwise to a rapidly stirring 45-mL solution of HA (10 μM), followed by stirring for 30 min at 4 C. The HA-layered liposomes were recovered by TFF wash as described previously. 41
The zeta potential was determined using laser Doppler electrophoresis

| Cell viability and immunofluorescence
Cells were cultured at 5,000 cells/well in a 96-well plate in triplicate for 24 hr, followed by treatment with varying doses of free or encapsulated drugs alone or in combination. Cell viability was determined at 96 hr using the CellTitre Glo assay (Promega) on a Tecan microplate reader.
Immunofluorescence was performed as previously described. 75 Briefly, to detect HR, COV362, OVCAR4, and OVCAR8 cells were plated at

| Western blotting
Western blotting was performed using a standard protocol. 76 The fol-

| Pharmacokinetics and targeting of ovarian tumor cells
Four-to six-week-old female immunocompetent BALB/c mice (Taconic) were IV and IP injected with 100 μL of blank Cy5.5-PLLconjugated and HA terminal-layered liposomal NPs (2.5 mg/mL). Fluorescence signals were obtained for the whole body and for harvested organs on a Xenogen IVIS imaging system (Caliper) at various time points (n = 3). For blood half-life determination, retro-orbital bleeds were performed on a set of animals (n = 3). All signals were normalized to the Cy5.5 autofluorescence signal, and the fluorescence signal from the NP content in the blood circulation was used to calculate and fit a two-compartmental pharmacokinetics model in the PRISM GraphPad software v5. To target ovarian tumors in vivo, COV362-mCherry cells (3 × 10 6 cells) were IP implanted in female NCR nude mice. Three weeks after tumor induction, xenograft tumor-and nontumor-bearing mice were injected with Cy5.5-PLL-conjugated and HA terminal-layered NPs. D-luciferin (100uL of 15mg/mL) was IP administered, and IVIS was used to simultaneously image the Cy5.5-NP biodistribution and luciferase signal at various time points (excitation at 675 nm for Cy5.5 and 720 nm for luciferase).

| Maximum tolerated dose
To determine the therapeutic dose for in vivo treatment, we performed high and medium maximum tolerated dose studies with free or encapsulated cisplatin and BMN 673 or AZD2281 alone or in combination. Animals were doses with free and nano-formulated drugs were administrated weekly by IV injections of the following agents: PBS or blank liposomal NP control (5.0 mg/kg once weekly), free cisplatin (7.00 mg/kg in 0.9% NaCl solution, free AZD2281 (50 mg/kg), free BMN 673 (0.33 mg/kg), free cisplatin and AZD2281 (5.00/50 mg/kg), free cisplatin and BMN 673 (5.0/0.33 mg/kg), and nano-formulated drugs at: cisplatin (5.00 mg/kg), free AZD2281 (50 mg/kg), free BMN 673 (0.33 mg/kg), free cisplatin and AZD2281 (5.00/50 mg/kg), and free cisplatin and BMN 673 (5.00/0.33 mg/kg). Four-to six-week-old female NCR nude mice (nu/nu, Taconic) were weighed and randomized to 12 treatment arms (n = 3 per group) in both the free and encapsulated dosing cohorts. The therapeutic agents were IV administered via the tail vein. Weight loss and lethargy or morbidity were assessed daily for 15 days. Body weight was measured daily, and an approximately 60-μL retro-orbital blood sample was taken weekly in a 0.2-mL K3E K3EDTA minicollect tube (VWR) for complete blood count (cbc) on a Vetscan HM5 hematology analyzer (Abaxis) to assess treatment efficacy and bone marrow toxicity.

| Mouse xenograft studies
The number of mice required in each treatment group to achieve statistical significance was determined by a power calculation. 77 Female NCR (nu/nu, Taconic, NY) mice were intraperitoneally injected with luciferase-expressing OVCAR8 cells (2 × 10 6 cells/200 μL) suspended in sterile Hank balanced salt solution. After 2-3 weeks, tumor-bearing mice were randomized into 12 groups (n = 8-10 mice for the free and encapsulated drug treatment groups, n = 15 for the PBS and liposomes alone control). Free and encapsulated drugs were administered weekly by IV injection of the following agents: PBS, drug carriers, 10% 2-hydroxyl-propyl-beta-cyclodextrine, 6% solutol, and 84% PBS for free olaparib and BMN 673 28 or blank liposomal NP control (5.0 mg/kg) once weekly); free cisplatin (6 mg/kg); free AZD2281 (50 mg/kg); free BMN 673 (0.35 mg/kg); free cisplatin and AZD2281 (5/50 mg/kg); free cisplatin and BMN 673 (5/0.35 mg/kg); encapsulated cisplatin (5 mg/kg); encapsulated AZD2281 (50 mg/kg); encapsulated BMN 673 (0.33 mg/kg); encapsulated cisplatin and AZD2281 (5/46.8 mg/kg); and encapsulated cisplatin and BMN 673 (5/ 0.32 mg/kg). Once weekly, the mice were weighed, and bioluminescence signals of tumor growth kinetics were obtained by IP injection of 0.1 mL of D-luciferin (30 mg/kg for 10 min) followed by animal imaging on a Xenogen IVIS Imaging system (Caliper). All animal experimentations adhere to the National Institute of Health (NIH) guide for the care and use of laboratory animals and procedures were also conducted with the approval of the MIT Committee on Animal Care (CAC).

| Necropsy and immunohistochemistry
Animals were monitored daily and euthanized when they became moribund, when their body weight decreased by more than 15%, or when lethargy, ruffled fur, or severe ascites were observed. Upon euthanasia, an arterial blood sample was withdrawn, and the body weight and volume of ascites were recorded. Complete liver and kidney biochemistry tests were performed using serum by Charles River Laboratories (Shrewsbury, MA) to assess organ toxicity. Tissue specimens were immediately fixed in formalin for paraffin embedding or

| Statistical analysis
Cell analysis was based upon triplicate experiments, and the results are presented as the mean ± SEM of at least three independent experiments. Student's t test was used for comparisons between two groups, and one-way ANOVA was used for comparisons among three or more groups of in vitro and in vivo data. Differences between samples were considered statistically significant at p < .05. Two-way ANOVA was used to compare histoscore for foci after treatments.
Survival data were analyzed using the Kaplan-Meier method, and logrank statistics was used to analyze the survival distribution. All statistical analyses were performed using the GraphPad Prism6 software (GraphPad Prism 5.0, GraphPad Software, La Jolla, CA).