Tumour-associated macrophages act as a slow-release reservoir of nano-therapeutic Pt(IV) pro-drug

Dual imaging shows congruent vehicle and payload kinetics

For clinical applicability and generalizability, TNPs in this study were designed to incorporate desirable properties of previously described polymeric nano-formulations that have entered clinical trials⁵, especially those of materials that have received Food and Drug Adminstration approval. TNPs were formulated by combining three compounds (compounds 1–3, Fig. 1a) via nano-precipitation, using spectrally complementary derivatives of BODIPYs as ideal fluorophores for efficient nano-encapsulation that convey lipophilicity and robust in vivo imaging arising from in vivo structural stability, environmentally robust fluorescence, high brightness and high photostability¹⁶,¹⁸,¹⁹,²⁰. Compound 3 (poly(D,L-lactic-co-glycolic acid)-b-poly(ethylene glycol); PLGA-b-PEG) self-assembles to form the hydrophobic PLGA core of the NP and hydrophilic PEG outer shell. For the cytotoxic payload, we used a novel Pt(IV) pro-drug (c,c,t-Pt(NH₃)₂Cl₂-(hexadecylcarbamate)[4-((2-(4,4-difluoro-1,3,5,7-tetramethyl-4-dora-3a,4a-diaza-s-Indacene)ethyl)amino)-4-oxobutanoate]) with a C₁₆ aliphatic carbon chain and a BODIPY (λ_ex=498 nm, λ_em=543 nm; Supplementary Fig. 1a) as the two lipophilic ligands axially coordinated to the Pt(IV) centre. NPs were similar in size (135±1 nm; standard error of the mean (s.e.m.), across n=12; Supplementary Fig. 1b-c), polydispersity (PDI=0.18±0.01; s.e.m. across n=12) and drug release rates (t_{1/2 release}=15 h; n=2; Supplementary Fig. 1d) compared with previously published Pt(IV) nano-formulations⁶,⁹, and were stable in physiological saline (Supplementary Fig. 1e). Free carboxylic acid terminal groups on the PEG gave TNPs a slight negative charge (zeta-potential=−23±1.4 mV; s.e.m. across n=4), which has been shown in some cases to benefit tumoral accumulation²¹,²²,²³,²⁴ and mirrors TNPs in the clinic⁵. C₁₆-Pt(IV) remains relatively stable in whole blood²⁵. However, upon reduction to the active Pt(II) compound cisplatin, dissociated BODIPY increases fluorescence by sevenfold owing to loss of quenching by Pt(IV) (Supplementary Fig. 1f-g). BODIPY fluorescence thus serves as an effective indicator of both pro-drug delivery and activation. Overall, these properties enable clearly distinguishable imaging of both the TNP vehicle and its payload, in a stable and clinically relevant nano-formulation platform.

We performed a series of in vitro experiments to characterize TNP behaviour in tumour cell culture (Supplementary Fig. 2a). Co-localization studies with fluorescent compartmentally localized proteins showed that the TNP vehicle moved from early to late endosomes over the course of 24 h (Supplementary Fig. 2b). Compared with the TNP vehicle, the TNP payload exhibited more diffuse intracellular localization, and co-localization peaked in the late endosome 24 h post treatment (Supplementary Fig. 2c). Payload fluorescence increased over time owing to Pt reduction and BODIPY de-quenching (Supplementary Fig. 2d). Dose–response TNP treatment indicated that intracellular fluorescence correlates well with overall Pt uptake (R²>0.99) and resulting DNA damage (Supplementary Fig. 3), thereby validating the use of fluorescence as a surrogate marker of cellular payload content. Together, these results provide evidence that, over the course of 24 h, TNPs accumulate in endosomes and lysosomal compartments, where the Pt pro-drug is activated through reduction to form cisplatin.

TNPs exhibit long TMV half-life and perivascular cell uptake

We first used time-lapse intravital imaging to characterize PK within a xenograft tumour model often used for intravital studies (Fig. 1)¹⁶,²⁶,²⁷,²⁸. Tumours were generated from subcutaneously implanted HT1080 cells expressing a fluorescently tagged DNA damage response protein, 53BP1, which localizes to the nucleus²⁹,³⁰. TNPs reached the TMV within 10–15 min of intravenous injection, and at early time points (t<30 min) the vehicle and its payload exhibited strong co-localization within vessels (Fig. 2a; intravascular Pearson's correlation ρ=0.8±0.02, s.e.m. across n=4; Mander's M1 and M2>0.95). Both the vehicle and its payload showed initial vascular PK similar to each other (Fig. 2b; t_{1/2 vehicle}=55±5 min; t_{1/2 payload}=61±6 min; s.e.m. across n=6), to previous studies using similar nano-formulations entering the clinic⁵,⁶,³¹, and to clinical NP formulations such as PEGylated liposomal doxorubicin (DOXIL; t_{1/2 initial}<5.2 h in humans and 2 h in rats³²,³³). Immediately as the TNPs reached the TMV, roughly 10% of the payload diffused into the surrounding tumour tissue. Such an initial burst of drug release was also observed in vitro (Supplementary Fig. 1d) and with similar previously described nano-formulations³⁴,³⁵, characterized by significantly faster release kinetics at initial phases followed by a later phase of slower release. Notably, PK imaging of BODIPY-labelled Pt compounds that were not nano-encapsulated revealed much more rapid vascular PK (t_{1/2 unencapsulated}=10±5 min²⁹). Therefore, despite an initial burst of Pt release seen both in vitro (Supplementary Fig. 1d) and in vivo (Fig. 2), nano-encapsulation conferred a relatively long circulation half-life.

Over the course of several hours, TNPs moved from the TMV (t<1 h) specifically to peritumoral host cells (1–3 h) adjacent to or even extending processes into the TMV. Uptake of TNPs by tumour cells themselves occurred more gradually and to a lesser degree compared with these perivascular host cells. By 24 h, TNPs were undetectable in vessels and had accumulated primarily in a heterogeneous mix of tumour cells and host cells, with host-cell uptake characterized by high local accumulation within cell-sized (7±1 μm; s.e.m. across n=4 tumours) diameters. At 24 h post-injection, the TNP payload also co-localized with the TNP vehicle in these host cells (Pearson's correlation within tumour tissue: ρ=0.7±0.08; Mander's M1 and M2>0.6±0.05; s.e.m. across n=4). Compared with the TNP vehicle, however, the payload accumulated more diffusely throughout the bulk tumour tissue, suggesting some payload release from the TNP vehicle (Supplementary Fig. 4a). We quantified these observations by measuring the coefficient of variation (CV) across cells within the tumour tissue (Supplementary Fig. 4b). The payloads exhibited >40% decrease in CV compared with the TNP vehicle, indicating more homogeneous distribution. In contrast to TNPs, naked BODIPY-labelled Pt that was injected in the absence of an encapsulating NP vehicle showed a substantially more homogeneous distribution within tumour tissue, with little preferential accumulation in any particular cell type (Supplementary Fig. 4; 60% reduction in CV). In summary, these data show that (i) nano-encapsulation enables a relatively long circulating half-life of the Pt(IV) pro-drug; (ii) nano-encapsulation influences spatial distribution within the bulk tumour tissue and (iii) TNPs primarily accumulate in perivascular host cells and, to a lesser extent, tumour cells.

TNPs cause mitotic arrest and localized DNA damage

In addition to PK analysis, high-resolution intravital imaging allowed for simultaneous monitoring of the cellular response to TNPs at the molecular level. Fluorescently tagged 53BP1 is a key member of the non-homologous end-joining pathway that forms visible puncta at sites of DNA double-strand breaks²⁹,³⁶. Based on the hypothesis that treatment with a cisplatin pro-drug would lead to increased DNA damage in tumour cells, we imaged 53BP1 puncta in response to TNP and quantitated the DNA damage response across populations of individual tumour cells at various time points (Fig. 3a). The results show a dose- and time-dependent increase in the fraction of tumour cells exhibiting high levels of DNA damage (Fig. 3b). We next tested whether there was an observable relationship between local payload accumulation and resulting single-cell DNA damage response. We used automated three-dimensional (3D) segmentation software to quantify TNP vehicle and payload accumulation within individual tumour nuclei, and correlated this information with DNA damage response in those cells. The results showed that cells exhibiting high levels of DNA damage also had a roughly 300% higher local payload concentration (Fig. 3c). Compared to TNP payload, local concentration of TNP vehicle also correlated with DNA damage, but to a lesser degree (Fig. 3c). DNA damage elicited by Pt treatment generally induces mitotic arrest in vitro²⁵, but it has been difficult to directly observe this response in vivo, where anti-mitotics often behave much differently²⁶. To determine TNP effects on cell growth, we used time-lapse intravital imaging to track individual tumour cells and mitotic events over the first 3 h following TNP injection in the HT1080 model (Supplementary Fig. 5) and found that TNP indeed caused a substantial reduction in tumour cell division (Fig. 3d). To test generalizability in another tumour model, we next examined local drug accumulation and DNA damage in an orthotopic model of disseminated ovarian cancer (OVCA)³⁷. Human A2780CP OVCA cells were intraperitoneally injected in nu/nu mice, and 6 weeks later (once palpable tumours and/or ascites had developed), animals received TNP treatment. The following day, tumours were excised and imaged for TNP accumulation and DNA damage. Imaging shows similar TNP localization patterns as observed with the HT1080 model (Fig. 4a). We also found similar statistically significant correlation between local TNP payload levels and tumour cell DNA damage (Fig. 4b). Taken together, these results demonstrate that TNPs elicit substantial DNA damage response activity and corresponding growth arrest in vivo. In both the subcutaneous and orthotopic OVCA tumour models, the correlation between local drug accumulation and DNA damage response implies that NP-mediated drug delivery plays a significant role in governing such PD.

Identification of cell populations with drug in steady state

We next performed a combination of intravital imaging, flow cytometry and histology studies to determine in which immunologically defined cell populations the TNP vehicle and its payload accumulated within the bulk tumour mass at 24 h, after TNPs had largely cleared the circulation. With the same HT1080 xenograft model used in the previous imaging experiments, we labelled TAMs with a fluorescent dextran-coated NP¹⁶ and found significant TNP uptake in TAM-rich regions of the bulk tumour mass (Fig. 5a). For more detailed immunological examination, we next performed flow cytometry analysis of HT1080 tumours. Results show that tumour cells made up 61% of the bulk tumour; 31% of the cells within the bulk tumour were CD45⁻ host cells (erythrocytes, tumour associated fibroblasts, endothelium and others); and the remaining 8% were CD45⁺ leukocytes (Fig. 5b). TAMs made up the largest population of leukocytes within the bulk tumour and also accumulated the highest levels of both TNP vehicle and payload on a per-cell basis (Fig. 5c). Because tumour cells far outnumbered TAMs in the HT1080 xenograft model, the majority of TNP vehicle and payload within the bulk tumour still resided in tumour cells themselves (Fig. 5d). Nonetheless, even though TAMs comprised only 4% of cells within the total tumour mass, they accumulated more than 40% of the total injected TNPs relative to tumour cells, and 30% of the total TNP overall.

To directly visualize genetically defined tumour-associated host leukocytes, we imaged TNP vehicle uptake in tumour-bearing fractalkine Cx3cr1^GFP/+ reporter mice, which are known to contain GFP⁺ macrophages³⁸. In this model, KRAS mutant p53^−/− (KP) cells derived from autochthonous lung tumours³⁹ were subcutaneously implanted into immunocompetent Cx3cr1^GFP/+ reporter mice, and TNP vehicle (without the green fluorescent payload) was intravenously administered. The vast majority of TNP vehicle accumulated within these GFP⁺ host leukocytes, especially near TMV (Fig. 6a,b). Using the above described KP lung cancer and intraperitoneal OVCA xenografts, we confirmed by histology that high levels of TNP vehicle accumulate in F4/80⁺ host phagocytes in these disease models as well (Fig. 6c and Supplementary Fig. 6). These analyses ultimately corroborate results from the intravital imaging data showing uptake of TNPs in tumour cells but also host cells, which are confirmed here as TAMs.

TNP payload spatially redistributes from TAMs to tumour cells

Based on intravital imaging evidence showing that the Pt(IV)-payload was more diffusely distributed within tumour tissue compared with its TNP vehicle (Supplementary Fig. 4), we next quantified the extent of payload redistribution and its effect on drug response. Using flow-cytometry data, we confirmed that tumour cells exhibited more than twice the amount of payload accumulation as a ratio of vehicle accumulation, compared with TAMs (Fig. 7a), thereby suggesting that payload may transfer from TAMs to tumour cells. We hypothesized that local payload release from TAMs into neighbouring tumour cells would lead to elevated payload concentration in these tumour cells, which would in turn enhance DNA damage. By histology we also measured greater than threefold elevated payload concentrations within roughly one cell-length (15 μm) of TAMs (Fig. 7b,c), and analysis of intravital imaging data provided further evidence that relevant payload concentration gradients occur over length-scales of<25 μm (Supplementary Fig. 7a). One explanation for these observations is that TAMs induce neighbouring tumour cells to take up more TNP directly; however, the TNP vehicle itself was not similarly elevated in the same neighbouring region (Fig. 7c). Tumour cells within the same phagocyte-neighbouring region also exhibited higher DNA damage on average, as determined by histology (Fig. 7d) and supported by corroborating intravital imaging analysis (Supplementary Fig. 7b). These analyses cumulatively show local payload redistribution from TAMs to neighbouring tumour cells exhibiting elevated DNA damage.

We next performed a series of in vitro experiments to directly test whether macrophages could release functional, DNA-damaging TNP payload into their surrounding environment. Similar to in vivo observations, we found that in vitro macrophages accumulated TNPs more rapidly and to a greater degree than tumour cells (Supplementary Fig. 8a). Once taken up by macrophages, Pt-payload was released at high levels into the extracellular supernatant. In fact, over the course of 24 h, more Pt was found in the supernatant than in the macrophages themselves (Supplementary Fig. 8b,c). The Pt-containing macrophage supernatant was both DNA damaging (Supplementary Fig. 8d) and cytotoxic (Supplementary Fig. 8e). In contrast, supernatant from macrophages treated with unencapsulated cisplatin at concentrations ranging from 1 to 300 μM was not cytotoxic to tumour cells (Supplementary Fig. 8e). For supernatant from TNP-treated macrophages, filtration with molecular-weight cutoff filters (3 and 100 kDa) did not reduce its potency, which indicated that cytotoxic effects were independent of macrophage-derived cytokines or other signalling proteins (Supplementary Fig. 8f). In addition, Pt-containing macrophage supernatant exhibited 50% inhibitory effect on tumour cell viability/cytotoxicity at Pt concentrations of 1.5±0.5 μM (s.e.m. across n=3) within the macrophage supernatant (Supplementary Fig. 7e), which is comparable to the potency of the Pt-payload when TNP was directly applied to tumour cells (IC₅₀=1.7±0.02 μM; s.e.m. across n=2; Supplementary Table 1). Taken together, these combined in vitro and in vivo data further confirmed that TAMs can serve as drug depots that accumulate high levels of TNPs and then release their DNA-damaging cytotoxic payload to the surrounding tumour tissue.

TAM depletion reduces TNP accumulation and efficacy

Finally, we tested whether TAM depletion would reduce the therapeutic efficacy of TNP in cancers. Macrophages were systemically depleted by pre-treatment with clodronate liposomes (clod-lip)⁴⁰ before TNP injection. In this tumour model, macrophage depletion itself had no detectable impact on tumour growth (P=0.2; two-tailed two-way analysis of variance; n=38 mice). However, it substantially reduced the degree to which TNP treatment was able to slow tumour growth (Fig. 8a). Direct measurement of intratumoral TNP accumulation confirmed that macrophages play a key role in mediating drug delivery: macrophage-depleted tumours exhibited both lower total intratumoral Pt (Fig. 8b) and a lower concentration of Pt (Fig. 8c) compared with macrophage-containing tumours, 24 h post administration of TNP, at which time no TNP remained circulating in the TMV of any tumour. The 24% decrease in Pt concentration (from 5 to 3.8 μM; Fig. 8c after converting to molarity) correlates well with the flow cytometry-based estimate that TAMs accumulate 30% of the total intratumoral TNP. It also correlates with the observation that unencapsulated Pt exhibits 37% lower intratumoral accumulation than nano-encapsulated Pt (Supplementary Table 1)⁹. This decrease disproportionately impacted TNP effects on tumour growth, probably because intratumoral Pt concentrations were near the threshold at which HT1080 tumour cells respond (in vitro IC₅₀=3.7±0.1 μM for unencapsulated Pt(IV) payload; s.e.m. across n=2). TNP clearance kinetics, TMV leakiness and other EPR factors all may influence drug accumulation within the tumour; nonetheless, the experiments here show that all things being equal, macrophage uptake substantially contributes both to the delivery of TNP to the tumour and, consequently, to the therapeutic impact on tumour growth.

Liver TNP uptake is well tolerated and blocks metastasis

Given our evidence for TAMs in mediating TNP delivery, we next investigated TNP effects in the liver, which is a prominent site of the mononuclear phagocyte system and is known to accumulate NPs⁶. Despite substantial liver accumulation (Supplementary Fig. 9a), TNPs did not significantly impact animal body weight (Supplementary Fig. 9b) and did not cause changes outside of the normal range in blood markers of renal or liver toxicity at the therapeutic dose (Supplementary Fig. 9c). Within the bone marrow, another important site of the mononuclear phagocyte system, TNP payload concentrations were much lower (Supplementary Fig. 9d,e). Although clinically relevant doses of un-encapsulated cisplatin caused a significant reduction in the number of nucleated cells in the bone marrow, TNP effects were not statistically significant (Supplementary Fig. 9f,g). Overall, these results demonstrate a positive safety profile despite liver accumulation.

We next tested whether TNP treatment could block the development of liver micro-metastases in a model of haematogenous breast cancer dissemination. 4T1 murine breast cancer cells were intravenously injected into immunocompetent balb/c mice, thus forming tumours in the liver, lung and other organs. To primarily deliver TNPs to host phagocytes rather than tumour cells, we pre-treated naive mice with TNPs, waited 6 h for a majority of the drug to clear circulation, and then injected a single-cell suspension of 4T1 cells. Two weeks later, livers were excised and analysed by histology for tumour burden. Results show that pre-treatment with TNPs reduced the development of micro-metastases by over 65%, and repeated TNP doses caused a slightly greater reduction of 80% (Fig. 9). No treatment impacted tumour burden in the lung (two-tailed t-test; n≥5). Taken together, these results demonstrate that TNPs can safely accumulate in host cells of the liver, thus allowing TNP payload to block the development of local metastases from haematogenously disseminated cancer cells.

C₁₆-Pt(IV)-BODIPY synthesis

A 0.2-ml N,N-Dimethylformamide (DMF) solution of cis,cis,trans-[Pt(NH₃)₂Cl₂(O₂CCH₂CH₂COOH)(OCONH(CH₂)₁₅CH₃)] (15 mg, 21 μmol) and 1-[Bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxid hexafluorophosphate (HATU) (8 mg, 21 μmol) was stirred at room temperature (RT) for 15 min. To the clear solution was added a 0.4 ml DMF suspension of the amino BODIPY (4,4-difluoro-5,7-di- methyl-4-bora-3a,4a-daiza-s-indacene; 6 mg, 21 μmol), and the orange suspension was stirred at RT for 30 min. Finally, a 4 μl portion of N,N-Diisopropylethylamine (DIPEA) was added to the mixture and the resulting orange solution was stirred at RT for 5 h. A 5-ml volume of saturated aqueous NaCl was added to the reaction mixture and a red precipitate was obtained. The precipitate was washed with 3 × 3 ml water. The resulting orange solid was dried in the vacuum overnight and the product was isolated by silica chromatography (acetone, R_f=0.5). Yield: 7 mg (30%). ¹H NMR (400 MHz, acetone-d₆): δ: 7.73 (s, 1H, NHCH₂CH₂-BODIPY), 6.54 (m, 6H, NH3), 6.18 (s, 1H, BODIPY), 5.80 (s, 1H, NH_carbamate), 3.42 (m, 2H, NHCH₂CH₂-BODIPY), 3.29 (m, 2H, NHCH₂CH₂-BODIPY), 3.01 (m, 2H, CH₂(CH₂)₁₄CH₃), 2.45 (s, 6H, BODIPY), 2.21 (m, 10H, BODIPY, succinate), 1.20 (m, 28H, CH₂(CH₂)₁₄CH₃), 0.8(t, 3H, CH₂(CH₂)₁₄CH₃). ¹³C NMR (100 MHz, DMSO-d₆): δ: 180.4, 172.1, 164.4, 153.9, 143.0, 142.0, 131.5, 122.3, 41.5, 40.9, 31.9, 31.8, 30.3, 29.5, 29.4, 29.2, 29.1, 22.6, 16.5, 14.6, 14.4. ¹⁹⁵Pt NMR (86 MHz, DMSO-d₆): δ: 1,240.1. High resolution electrospray ionization mass spectrometry (HR-ESI-MS) (positive mode) for C₃₆H₆₃BCl₂F₂N₆O₅Pt: m/z [M−H]⁺ calculated: 975.4020; observed: 975.4001.

Polymer materials

Unless otherwise stated, all chemical reagents were purchased from Sigma-Aldrich. PLGA(55:45 lactide/glycolide)_35.2kDa-PEG_3.5 kDa was purchased from Advanced Polymer Materials, Inc. (Cat. #13–01-CX-55/45–35.2k/3.5k; lot 14–02–169). From the vender, gel permeation chromatography indicated M_n=38.8 kDa, M_w=58.8 kDa and PI=1.52. PEG weight fraction (8.7%±0.3%) and molecular weight were confirmed by ¹H NMR peak integration values, which were also used to calculate the lactide/glycolide ratio. ¹H NMR in CDCl₃: δ (in p.p.m.) 5.12–5.28 (m, (-OCH(CH₃)CO)), 4.66–4.9 (m, (-OCH₂COO-)), 3.64 (s, -OCH₂CH₂O-), 1.52–1.62 (m, (-OCH(CH₃)CO-). Differential scanning calorimetry was performed (Q10; TA Instruments), heating from −60 °C to 100 °C at 10 °C min⁻¹, with no apparent crystallization or melting temperature (Tm); glass temperature (Tg) was observed at 39.5 °C. BODIPY-630-NH₂ (Life Technologies; 2.8 mg, 0.0046, mmol) was conjugated to PLGA(50:50 lactide:glycolide)_{30–60 kDa} (100 mg, ca 0.0023, mmol; Sigma) using previously published procedures²⁷, by stirring at RT under Ar for 10 min, adding 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDCI) (4.4 mg, 0.023 mmol) and 4-Dimethylaminopyridine (DMAP) (2.8 mg, 0.023 mmol) in CH₂Cl₂ (0.5 ml), and stirring overnight. Polymer was precipitated in cold 1:1 MeOH/Et₂O, centrifuged (2,700g, 10 min) and repeatedly washed in minimal CH₂Cl₂, followed by MeOH/Et₂O precipitation and centrifugation. The resulting blue precipitate was dried under vacuum.

NP synthesis and characterization

TNPs were prepared by nano-precipitation. Briefly, the following reagents were dissolved in 50:50 (v/v) DMF/acetonitrile (ACN) at the indicated concentrations: PLGA-BODIPY-630 (0.17 mg ml⁻¹), C₁₆-Pt(IV)-BODIPY (0.17 mg ml⁻¹) and PLGA-PEG (0.83 mg ml⁻¹). The resulting solution was then mixed and added drop-wise to nuclease-free H₂O while maintaining a ratio of 20:1 H₂O to organic solvent. TNPs were stirred overnight and filtered through sterile 0.45 μm syringe filters (regenerated cellulose, 17 mm, Cole Palmer). The TNPs were concentrated by centrifugation using centrifugal filter units (Amicon; Millipore; molecular weight cut-off, MWCO=100 kDa). The concentrated TNPs were washed twice with de-ionized H₂O and re-suspended in 1 ml of nuclease-free H₂O. TNPs were diluted ten times in H₂O and characterized by size using dynamic light scattering (DLS; Malvern Zetasizer). Zeta-potential was also measured by DLS in dI-H₂O (−25±2.5 mV) and PBS, pH 7.4 (−23±1.4 mV). Although the buffered solution produced a slightly higher zeta-potential, measurements were not statistically different from each other (two-tailed t-test; α=0.05). The Pt concentration was determined by graphite furnace flameless atomic absorption spectroscopy (AAS). BODIPY-630 concentration was determined by a microplate fluorimeter after diluting TNPs in DMF and using pure BODIPY-630 for a standard curve. Transmission electron microscopy (TEM) experiments were performed on a JEOL 1011 electron microscope (JEOL). The TEM sample was prepared by depositing 10 μl of TNPs (5 mg ml⁻¹) onto a carbon-coated copper grid. The excess solution was blotted, and the grids were immersed in a solution of phosphotungstic acid stain. The stain was blotted, and the dried grids were immediately used for imaging. To characterize payload release from TNPs (Supplementary Fig. 1d), they were incubated in PBS at 37 °C, filtered using 100 kDa MWCO centrifugal filters (Amicon; Millipore) at the indicated time points and resuspended in fresh PBS. Filtrate was frozen for later analysis by AAS. After 72 h, both the TNPs and the filtrate were analysed by AAS to quantify the remaining Pt. Based on the above procedure, <5% of total PLGA-BODIPY-630 was released from TNPs over the course of 48 h.

Animal models

All animal research was performed in accordance with guidelines from the Institutional Subcommittee on Research Animal Care (Massachusetts General Hospital). All experiments were performed using female mice that were 5- to 7-week old at the start of the experiment. For experiments with HT1080 tumours, 2 million cells in PBS were subcutaneously implanted into nu/nu mice; roughly 2–3 weeks later (once tumours reached ∼8 mm diameter), imaging experiments were initiated. For the experimental model of haematogenous metastatic dissemination, 0.25 million 4T1 breast cancer cells suspended in PBS were intravenously (i.v.) injected into balb/c mice; roughly 2 weeks later tumours were excised for histological quantification of liver and lung tumour burden in formalin-fixed paraffin-embedded (FFPE), haematoxylin and eosin-stained sections. For intraperitoneal OVCA imaging, 10 million A2780CP cells suspended in PBS were intraperitoneally injected into nu/nu mice. TNP injection and imaging were performed once ascites or tumour masses became evident (roughly 6 weeks later). For histology and imaging experiments using KP cells, 1 million cells were subcutaneously implanted into C57Bl/6 background animals (all JAX), including Cx3cr1^GFP/+R26^mT−mG/+ dual-reporter mice. For dual reporter mice generation, Cx3cr1^GFP/GFP animals were crossed with R26^{mT−mG/mT−mG} animals that ubiquitously express membrane-anchored tdTomato⁶⁸. The resulting Cx3cr1^GFP/+ R26^mT−mG/+ animals can be used to visualize GFP⁺ tdTomato⁺ macrophages, monocytes and dendritic cells, as well as tdTomato⁺ stroma and endothelium.

Cell lines

With some exceptions, cell lines were obtained from American Type Culture Collection and routinely cultured according to the manufacturer's guidelines. OVCA-429 were a kind gift from Dr Michael Birrer (MGH), and A2780CP were from Sigma. The murine lung adenocarcinoma cell line KP1.9 was generated from lung tumour cells of Kras^LSL−G12D/+;p53^fl/fl mice (C57/Bl6 background), which develop lung adenocarcinoma after infection with an adenovirus expressing Cre recombinase by intratracheal administration³⁹. KP cells were maintained in Iscove's DMEM media supplemented with 10% fetal bovine serum and 5% penicillin/streptomycin. All cell lines were routinely tested for contamination using mouse antibody production testing (VRL Laboratories) and assaying for mycoplasma (VRL Laboratories; Lonza MycoAlert). The fluorescent protein mApple was subcloned into the pLVX vector using restriction enzymes AfeI and NotI. We inserted a XhoI site at in the 5′ end of the pLVX-mApple multiple cloning site using QuikChange site-directed mutagenesis (Agilent). The pDsRed-Monomer-Mem membrane targeting sequence (Clontech) was PCR amplified for In-Fusion cloning to construct pLVX-Mem-mApple. PCR product was then ligated into pLVX-Apple after AfeI and XhoI digestion, and fully sequenced. Lentiviral particles were produced using Lenti-X HTX Packaging System (Clontech) for transduction into HT1080 and A2780CP cells, which were then selected using 3 μg ml⁻¹ puromycin. HT1080–53BP1-mApple and A2780CP-53BP1-mApple were developed as previously described²⁹.

Intravital microscopic imaging

TNPs were injected at the indicated dose (1 mg kg⁻¹ unless otherwise stated) via tail-vein catheter immediately after mixing to a final 1 × PBS solution, at a final volume of 100 μl. Intravital microscopy was performed on an Olympus FV1000 multiphoton imaging system using a XLUMPLFLN × 20 water immersion objective (numerical aperture=1.0; Olympus America). Images were scanned sequentially using 405-, 473-, 559- and 633-nm diode lasers in combination with a DM405/488/559/635-nm dichroic beam splitter. Emitted light was then separated and collected using appropriate combinations of beam splitters (SDM473, SDM560 and/or SDM 640) and emission filters BA490–540, BA575–620, BA575–675 and/or BA655–755 (all Olympus America). Dextran pacific blue (λ_ex=405 nm) was injected to initially image TMV as previously described²⁹. Briefly, 500-kDa amino-dextran (Thermo) was labelled with Pacific Blue succinimidyl ester (Thermo), purified using 30 kDa MWCO centrifugal filtration (Amicon), and 250 μg i.v. injected 10 min before imaging. Dorsal window chamber imaging was performed following previously described procedures²⁹,⁶⁹,⁷⁰. Briefly, 2 million HT1080–53BP1-mApple cells were suspended in 50 μl PBS and injected under the fascia of nu/nu mice (Cox7, MGH) 30 min following surgical chamber implantation, and imaged 2 weeks later.

Histology

Excised tumours were embedded in optimal cutting temperature (OCT) compound, flash-frozen, sectioned using a Leica CM1900 Rapid Sectioning Cryostat (Leica), stained using anti-mouse F4/80 antigen eFluor 450 (clone BM8; eBioscience), and imaged using an upright Olympus BX63 microscope and a × 100 oil-immersion objective. For immunohistochemistry, a biotinylated anti-rat IgG antibody was applied, and VECTASTAIN ABC kit (Vector Laboratories Inc.) along with a 3-amino-9-ethylcarbazole substrate (Dako) were used for colour development. The sections were counterstained with Harris haematoxylin solution (Sigma-Aldrich) and scanned by using Nanozoomer 2.0RS (Hamamatsu). For immunofluorescence, anti-rat IgG-Alexa Fluor 405 (Abcam) as a secondary antibody and anti-mouse CD326 (EpCAM)-PE antibody (clone: G8.8, eBioscience) were used.

In vitro NP characterization

For cytotoxicity assays (Supplementary Table 1), 5,000 cells per well were plated in 96-well plates, treated the next day with Pt-compound or the appropriate buffer control (DMF or drug-free PLGA-PEG NPs, as appropriate), and assessed for viability 72 h later using PrestoBlue (Life Technologies) following the manufacturer's protocol. IC₅₀ values for each compound were calculated by interpolating from an 11-point dose–response curve.

For live-cell in vitro microscopy of TNP uptake (Supplementary Fig. 2a), HT1080–53BP1-mApple cells were seeded on an optical-bottom 96-well plate (Ibidi; Applied Biophysics), treated the following day with 1 μM TNP for 24 h, washed in warm medium, and immediately imaged on a DeltaVision (Applied Precision) modified Olympus BX63 microscopy system with an environmental chamber. Dose–response analysis of TNP uptake (Supplementary Fig. 3) followed the same protocol. Corresponding measurements of intracellular Pt uptake (Supplementary Fig. 3d) were performed by treating 80% confluent cells in six-well plates for 24 h with TNP, rinsing with PBS, trypsinizing, centrifuging the cell pellet for 300g for 5 min, re-suspending the pellet in 50 μl fresh PBS, and freezing it for subsequent AAS analysis.

CellLight BacMam (Life Technologies) fluorescent protein-signal peptide fusions, delivered using baculovirus, were used for subcellular compartment co-localization experiments (Supplementary Fig. 2b). Briefly, 10,000 HT1080 cells were treated overnight with 20 μl CellLight reagent in 12-well dishes, plated the next day on collagen-I-coated glass coverslips in six-well dishes, and the following day were treated with 0.5 μM TNP for either 3 or 24 h. Following treatment, cells were counterstained with Hoechst 33342 and immediately imaged. Collagen-coating was performed by treating coverslips with 100 μg ml⁻¹ rat-tail collagen-I (BD Biosciences) overnight and then rinsing with PBS.

To analyse de-quenching of the C₁₆-Pt(IV)-BODIPY compound (Supplementary Fig. 2d), TNPs were added at 1 μM to full serum medium (DMEM+10% fetal bovine serum+1% penicillin/streptomycin) in the presence or absence of 90% confluent HT1080 in 96-well plates. Fluorescence was monitored by a Tecan Safire 2 fluorescence plate reader. For monitoring TNP uptake kinetics (Supplementary Fig. 8a), equal numbers of RAW 264.7 macrophages or HT1080 cells were plated in 96-well plates and incubated with 1 μM TNP. At the indicated time points, medium was temporarily transferred to empty wells and replaced with fresh media on cells. Cells were then immediately measured using a plate reader for BODIPY-630 fluorescence, and the TNP-containing media were replaced onto the cells again.

To analyse Pt uptake and release into supernatant by macrophages (Supplementary Fig. 8c), confluent 10 cm tissue-culture dishes of RAW macrophages were treated with 62 μM TNP for 24 h, rinsed in PBS, treated with fresh media, incubated for another 24 h and then analysed for Pt content by AAS. The supernatant was clarified by centrifugation (300g for 5 min) before AAS analysis. Cells were harvested by trypsinization, centrifugation (300g for 5 min), resuspension of the cell pellet in 50 μl PBS and freezing for subsequent AAS analysis.

To analyse whether the Pt-containing macrophage supernatant could invoke a DNA damage response in tumour cells (Supplementary Fig. 8d), RAW macrophages were treated for 6 h with 15 μM TNP, then rinsed and incubated in fresh media. After 24 h, media were clarified by centrifugation (300g for 5 min), transferred to HT1080–53BP1-mApple cells plated on optical-bottom 96-well plates (Ibidi; Applied Biophysics), and incubated overnight. The following day cells were rinsed and immediately imaged.

To analyse whether Pt-containing macrophage supernatant cause cytotoxic effects on tumour cells (Supplementary Fig. 8e), RAW macrophages were treated for 24 h with either TNP or unencapsulated C₁₆-Pt(IV)-BODIPY (the latter to achieve high Pt concentrations), then rinsed and incubated in fresh media. After 24 h, media were clarified by centrifugation (300g for 5 min), then transferred to 5000 HT1080 cells per well of a 96-well plate. After 72 h, the cell count was determined by the PrestoBlue assay (Life Technologies).

To analyse whether the above cytotoxic effects depend on proteins in the Pt-containing macrophage supernatant (Supplementary Fig. 8f), RAW macrophages were treated for 16 h with 60 μM TNP, then rinsed and incubated in fresh media. After an additional 16 h, media were clarified by centrifugation (300g for 5 min), filtered using MWCO centrifugal filters (Amicon; Millipore), and transferred to 5000 HT1080 cells per well of a 96-well plate. After an additional 72 h, the cell count was determined by the PrestoBlue assay (Life Technologies).

Tumour growth

Subcutaneous tumours were implanted in 7-week-old female nu/nu mice (Cox7, MGH), injected in four quadrant animal flanks as 1 million HT1080 cells suspended in 50 μl PBS. Nine days later, animals were ranked and sorted into groups of four based on tumour size. According to predefined protocol guidelines, mice were killed when tumour burden reached more than 1 cm in diameter, or 2 cm in diameter if only one tumour was present. Tumour volumes were measured two to three times per week, estimated using the spherical tumour volume formula V=⁴/₃ πr³, where r is averaged from four caliper measurements performed by two blinded researchers. Each size-group was randomly distributed to the four treatment-groups such that final treatment-groups exhibited roughly equal distribution in tumour size, with 9–10 mice per group. Groups received 150 μl of clodronate liposomes (clod-lip) or PBS liposomes as a vehicle control, intraperitoneally, on the indicated days (Fig. 8a). Liposomes were from ClodLip BV (Haarlem), with clodronate at a concentration of 5 mg ml⁻¹. As separate intravenous injections on indicated days (Fig. 8a), groups received 1 mg kg⁻¹ Pt of either TNP or the TNP without Pt(IV) as a vehicle control. TNPs for this experiment were formulated as described above, using PLGA_8kDa-PEG_5kDa and non-fluorescent PLGA_30–60kDa rather than PLGA_30–60kDa-BODIPY-630. TNP formulations were made fresh before each injection, and batches were tested for size and monodispersity by DLS and TEM. Pt loading was measured by AAS, and appropriate drug release was measured as previously described. At 19 days post implantation, tumours were excised, rinsed in PBS and processed for histology by freezing in OCT compound, or alternatively processed for AAS analysis by blotting tissue to remove excess water and freezing. One of thirty-eight animals was prematurely killed owing to biting and cachexia.

Flow cytometry

Subcutaneous tumours (HT1080 cells) were harvested from nu/nu mice 3 weeks post implantation, cut into small pieces and incubated in RPMI 1640 medium containing 0.2 mg ml⁻¹ collagenase type I (Worthington Biochemical Corporation) for 1 h at 37 °C while shaking (250 r.p.m.). Digested tumours were filtered through a 70-μm cell strainer (BD Falcon). The resulting single-cell suspensions were washed and resuspended in PBS with 0.5% BSA and 2 mM EDTA. Cell labelling was performed with appropriate antibodies as indicated below for 45 min at 4 °C. The following cell types were identified by flow cytometry (LSRII, BD Biosciences) depending on cell marker expression: tumour cells (hCD29⁺), macrophages (CD45⁺CD11b⁺Ly6C⁻Lin⁻CD11c⁺F4/80⁺), neutrophils (CD45⁺CD11b⁺Ly6C⁺Lin⁺) and lymphocyte-like cells (CD45⁺Lin⁺CD11b⁻Ly6G⁻). The anti-human CD29 antibody (clone MAR4) was from BD, and the following anti-mouse antibodies used were from BioLegend: CD45 (clone 30-F11), F4/80 (clone BM8), CD11c (clone N418), Ly6C (clone HK1.4). CD11b (clone M1/70) was from BD Biosciences, along with all antibodies used in the lineage (Lin) mixture: anti-CD90.2 (clone 53–2.1), anti-B220 (clone RA3–6B2), anti-NK1.1 (clone PK136), anti-CD49b (clone DX5), anti-Ter119 (cloneTER-119) and anti-Ly6G (clone 1A8). To exclude dead cells, 7-aminoactinomycin D (Sigma-Aldrich) was used. Data were analysed using FlowJo v.8.8.7 (Tree Star, Inc.) and MATLAB. BODIPY-630 and amino-BODIPY fluorescence was assessed by using an LSRII flow cytometer and the mean fluorescence intensity (MFI) of the different cell populations was determined. Background autofluorescence for each immunologically defined cell population was measured from control-treated animals that were injected with non-fluorescent PLGA-PEG vehicle; this value was subtracted from the data for the fluorescent TNP-treated animals.

Computational image analysis

Intravital microscopy images were analysed using either Matlab (Mathworks) or ImageJ and were pre-processed using background subtraction based on data acquired immediately before TNP injection. For vascular half-life calculations, fluorescence time-lapse values from multiple vessels across multiple animals were recorded, averaged and fit to an exponential decay model using nonlinear regression in MATLAB. Single-cell 3D segmentation for image cytometry analysis was performed using a custom MATLAB pipeline. Briefly, a previously described 3D nuclear-segmentation algorithm was modified to measure the MFI of TNP vehicle and its payload within a given radius of each nucleus. These measurements were then correlated with observed 53BP1-puncta in each individual nucleus, validated by eye (Fig. 3c). Throughout the manuscript, data falling >2 standard deviations from the mean were excluded as outliers from the statistical calculations. For calculation of spatial heterogeneity, measured by the CV (Supplementary Fig. 4), cell-sized regions of interest within images were quantified for MFI. CV for TNP and unencapsulated Pt (CP-11 (ref. ²⁹)) were measured >20 half-lives post injection as an approximation of steady-state tissue distribution (5 and 24 h, respectively).

Liver and BM analysis

Pt concentrations in the liver and bone marrow were determined by AAS and fluorescence imaging. Briefly, naive female nu/nu mice 8-week old were i.v. injected with TNP at a dose of 1 mg kg⁻¹ Pt. 24 h later, animals were perfused with PBS by intracardiac injection before dissection. For fluorescence quantification of TNP payload, livers were excised and frozen in OCT compound for later sectioning. Femoral BM was flushed using PBS and smeared on a glass slide for immediate imaging. Un-injected control animals were used to measure background signal for each organ, which was baseline-subtracted in the quantification. For AAS analysis of Pt content, livers were excised, patted dry and frozen. BM was flushed using PBS, centrifuged at 16,000g to pellet cells and frozen. Tissue was then dissolved in 70% nitric acid and analysed as with the HT1080 tumours (Fig. 8). For quantification of BM cellularity, femurs were excised 24 h post injection with TNP, formalin-fixed overnight, decalcified in EDTA and processed for FFPE sectioning. Immunohistochemistry was performed for apoptosis analysis by terminal deoxynucleotidyl transferase dUTP nick-end labelling, using Apoptag Peroxidase In Situ Apoptosis Detection Kit (Millipore), counterstaining with Harris haematoxylin solution (Sigma-Aldrich), and scanning with a Nanozoomer 2.0RS (Hamamatsu).

Metastasis burden in the lung and liver was quantified by morphological assessment in haematoxylin and eosin-stained liver FFPE sections that had been imaged using Nanozoomer 2.0RS, with organ excision at time of animal euthanasia according to the institutional guidelines (body condition score of 2) or 20 days post cancer injection, whichever came first. No statistically significant difference was observed in animal survival, body weight loss or lung tumour burden across the groups. One mouse from the group receiving repeated TNP dosing was excluded from analysis since no lung tumours (and very few liver tumours) were found by histology, presumably due to failed initial intravenous injection.

Clinical chemistry measurements in Swiss albino mice were performed using previously described TNP and methods⁶. Graphically re-plotted here for context (Supplementary Fig. 9a), bio-distribution of Pt in male rats was measured as described previously⁶.