⁶⁴Cu-MM-302 Positron Emission Tomography Quantifies Variability of Enhanced Permeability and Retention of Nanoparticles in Relation to Treatment Response in Patients with Metastatic Breast Cancer

Nanoparticle drug delivery systems provide a means to alter the biodistribution and pharmacokinetics of small-molecule drugs. Such systems are of particular importance in oncology, where there is a need to improve the toxicity profiles and therapeutic windows for small-molecule chemotherapies. Therapeutic nanoparticles can enable long-circulating pharmacokinetics and tunable sustained release and improved deposition in solid tumors through leaky vasculature. This phenomenon is referred to as the enhanced permeability and retention (EPR) effect and is well characterized in animal models (1–3). The extent to which the EPR effect is present in human tumor lesions remains controversial but has been proposed as an explanation for variable responses to nanotherapeutics and has important implications for the development and design of future nanomedicines (3, 4).

Liposomes are a class of nanomedicines that have been proven to be clinically useful drug delivery vehicles, with several approved agents for cancer treatment (Doxil/Caelyx, Myocet, DaunoXome, Marqibo, and ONIVYDE). The large size of liposomes, typically about 100 nm in diameter, prevents extravasation from normal vasculature and results in deposition and retention in areas of functionally porous vasculature, such as the liver and spleen, or leaky vasculature in some tumor lesions and areas of inflammation (5–7). Deposition of liposomes via the EPR effect is a nonspecific phenomenon governed primarily by their size and surface characteristics (8, 9).

MM-302 (HER2-targeted PEGylated liposomal doxorubicin) is a nanoparticle in clinical development for patients with HER2-positive metastatic breast cancer (NCT01304797, NCT02213744). Targeting of liposomes to tumor antigens, such as HER2, serves as a means to alter the microdistribution of liposomes within tumor lesions and to direct deposited liposomes into tumors cells rather than macrophages (10). MM-302 was specifically designed to maximize doxorubicin uptake into tumor cells while minimizing uptake into nontarget cells and tissues (11) and to enable combination with trastuzumab (12). Although many have reported that antigen-targeted nanoparticles can lead to increased tumor uptake, the extent to which targeted nanoparticles can enhance tumor uptake is determined by many factors, such as choice of targeting ligand and particle size (13–15). For some nanoparticles, tumor accumulation is independent of targeting ligand and, therefore, may be dictated primarily by the EPR effect. Previous work in preclinical models has demonstrated that the HER2 targeting did not alter overall deposition of MM-302 into tumors (Supplementary Fig. S1), but only altered the cellular fate within the tumors (10).

Effective drug delivery is a necessary step for antitumor activity, and poor penetration of anticancer drugs is believed to limit the effectiveness of chemotherapy in solid tumors (16–18). As molecular size of a therapeutic agent increases, effective delivery becomes increasingly important (19). In preclinical models, effective tumor deposition of nanoparticles has also been shown to be a potentially rate-limiting step for effective drug delivery to tumor cells and the resulting antitumor activity (20–23).

The goal of this study was to understand the biodistribution and evaluate the potential role of the EPR effect on a nanotherapeutic in patients with metastatic breast cancer to determine whether tumor lesion delivery could ultimately be used to predict therapeutic efficacy. We transformed an HER2-targeted liposomal doxorubicin (MM-302) into a tracer for PET through labeling with ⁶⁴Cu to enable quantitative characterization of tumor delivery kinetics. ⁶⁴Cu is a positron-emitting radionuclide with a 12.7-hour half-life, well matched to the pharmacokinetics of MM-302. We demonstrated the clinical feasibility and safety of ⁶⁴Cu-liposome PET in patients. ⁶⁴Cu retention within liposomes was shown to be stable in patients within the image acquisition time frame. Computational modeling enabled detailed elucidation of the EPR kinetics, establishing both the deposition and washout components of tumor delivery. Variability of the EPR effect was established across lesions within a patient and across patients, as well as in nontarget tissues including the liver, spleen, and bone marrow. Despite heterogeneity at the individual lesion level, we described a method of patient classification based on minimum lesion deposition and demonstrated association of tumor deposition with response to treatment.

Patients imaged with ⁶⁴Cu-MM-302 PET were part of a multi-site phase I MM-302 study (NCT01304797). A companion imaging protocol was included after the dose escalation phase in which all patients who consented underwent ⁶⁴Cu-MM-302 PET. The primary objectives of the ⁶⁴Cu-MM-302 PET study were to determine the radiation dosimetry and biodistribution of ⁶⁴Cu-MM-302. The study was conducted following International Conference on Harmonization guidelines in accordance with Good Clinical Practice and the ethical principles based on the Declaration of Helsinki. The study was reviewed and approved by the local Institutional Review Board at each center prior to the start of the study, and each subject gave written informed consent.

Patients enrolled in this study were ≥18 years of age and had advanced HER2-positive (i.e., HER2 3+ by IHC, or HER2 2+ by IHC and FISH or CISH-positive disease) breast cancer, measureable disease by RECIST v1.1, an Eastern Cooperative Oncology Group performance status of 0 to 1, adequate bone marrow reserves (absolute neutrophil count ≥ 1,500/μL, platelet count ≥ 100,000/μL, hemoglobin ≥ 9 g/dL), adequate hepatic function [serum total bilirubin within normal limits, aspartate aminotransferase, alanine aminotransferase up to 2× upper limit of normal (ULN)], adequate renal function (serum creatinine ≤ 1.5× ULN), and adequate cardiac function (left ventricular ejection fraction of ≥50% by echocardiography or radionuclide ventriculography). The main exclusion criteria were total cumulative doxorubicin >300 mg/m², active infection or unexplained fever, symptomatic brain metastases, congestive heart failure, coronary artery disease, history of myocardial infarction, hypertension, angina pectoris, valvular heart disease, severe and/or uncontrolled ventricular arrhythmia, prolonged QTc interval, history of allogeneic transplant, and infection of HIV, hepatitis B or C.

Patients imaged with ⁶⁴Cu-MM-302 PET were in Arms 3 and 4 of the phase I study (NCT01304797) and received the following treatment: trastuzumab at 6 mg/kg (with an 8 mg/kg loading dose on cycle 1) administered as a 90-minute intravenous infusion, MM-302 at 30 mg/m² administered as a 60-minute intravenous infusion, as well as a tracer dose of ⁶⁴Cu-MM-302 as described below. The first 3 patients of Arm 3 received trastuzumab starting on cycle 2 to enable initial safety assessment of ⁶⁴Cu-MM-302. Patients in Arm 4 also received cyclophosphamide at 450 mg/m² 5 days prior to ⁶⁴Cu-MM-302 based on a preclinical study showing that tumor priming with cyclophosphamide enhanced tumor delivery of MM-302 in xenografts (24). All patients received prophylactic premedication with diphenhydramine or equivalent prior to dosing of MM-302.

⁶⁴Cu-MM-302 was prepared by a commercial radiopharmacy using ⁶⁴CuCl₂ obtained from Washington University (St. Louis, MO), a chelating/loading agent (diacetyl 4,4′-bis(3-(N,N-diethylamino)propyl)thiosemicarbazone, or 4-DEAP-ATSC), and MM-302. 2.5 mL of 0.06 mg/mL 4-DEAP-ATSC was added to ⁶⁴CuCl₂ and incubated for 1 minute at room temperature. The chelated ⁶⁴Cu mixture was passed through a 0.2-μm filter into a vial containing 5 mL of 2 mg/mL MM-302. The mixture was heated to 65°C for 10 minutes and then cooled to room temperature. The efficiency of loading was measured by size exclusion chromatography as described previously (25).

Quality control testing for ⁶⁴Cu-MM-302 consisted of total activity (400 MBq ± 10%), radiochemical purity (⁶⁴Cu loading efficiency ≥ 85%), appearance (red opalescent solution), endotoxin (≤35 EU/mL), and sterility. The effect of ⁶⁴Cu labeling on MM-302 was found to have negligible effects on key MM-302 characteristics (HER2-binding, doxorubicin content, phospholipid concentration) and maintain an acceptable pH for administration. Preclinical characterization and stability of ⁶⁴Cu-MM-302 was reported elsewhere (25).

For all patient doses, radiochemical purity was observed in the range of 95% to 99%. Within 3 hours after administration of 30 mg/m² MM-302 (doxorubicin equivalent), approximately 400 MBq of ⁶⁴Cu-MM-302 (∼3–5 mg/m² doxorubicin) in 3 to 7 mL (depending on the ⁶⁴Cu-labeling–specific activity at the time of administration) was administered as an intravenous infusion over 10 minutes using an infusion pump, followed by a saline flush.

Each patient underwent PET/CT 0 to 3 hours after administration of ⁶⁴Cu-MM-302 on day 1 and additional scan(s) on day 2 and/or day 3 depending on scan group assignment upon enrollment. Five of the 19 patients consented to all three scan times. Image acquisition times on days 1, 2, and 3 were 3 to 5, 7 to 9, and 12 to 15 minutes per bed position, respectively. Approximately 6 bed positions were obtained for each patient, spanning from mid-brain to mid-thigh axially.

Quantification of images was performed using MIM software (MIM Software Inc., Version 6.2 or higher). Rather than maximum standardized uptake value (SUV_max), tracer concentration was reported as the percent injected dose per kg of tissue (%ID/kg) within a region of interest (ROI), derived from SUV_median normalized to body weight (BW). %ID/kg is related to SUV in the following manner: %ID/kg tissue = SUV/BW × 100%. This measure provides a quantitative measure of drug concentration in the ROI. It is hypothesized that drug activity will be closely related to absolute drug concentration, and for this reason, SUV_max, which is conventionally used to correlate with tumor metabolic activity [in the context of ¹⁸F-fludeoxyglucose (FDG)], is not appropriate. The median tissue uptake was determined by creating a 3D ROI in metastatic lesions and in normal tissues (heart, lung, liver, spleen, kidneys). Tumor lesion ROIs were selected by an independent radiologist board-certified in nuclear medicine. The median radioactivity for all tumor lesions and tissues was exported from the MIM software into an electronic case report form (eCRF) for further analysis. As an exploratory analysis, bone marrow uptake was quantified using an 8-mm spherical ROI in the lumbar vertebrae (average of L1-L5). Vertebrae with bone metastases or sclerotic lesions based on CT were excluded from normal bone marrow analyses.

Safety of the underlying chemotherapy regimen is reported elsewhere (26, 27). Briefly, vital signs were monitored at multiple time points, both before and after the MM-302 and ⁶⁴Cu-MM-302 administrations. Hematology, liver, and kidney functions were assessed throughout the study. Continual follow-up assessment for adverse events at each study visit was in place to monitor for any possible sequelae of the radiopharmaceutical imaging study. Safety of ⁶⁴Cu-MM-302 as a PET tracer was established through radiation dosimetry as described below.

Data were presented as PET image files to an independent medical physicist with expertise in radiopharmaceutical dosimetry. Counts in tissues were extracted from the images using the MIPAV software (28). Activity in each visualized organ and the total body were expressed as fractions of injected activity, normalizing the activity in the whole body at the earliest time point to be 100% of the administered activity. Resultant values of percent of injected activity per organ were fit using the SAAM II software (29). Time integrals of activity (30) were entered into the OLINDA/EXM software (31), using the adult male model. The number of disintegrations in the “remainder of body” was assumed to be that in total body within the field of view minus the values in other organs of uptake.

The tracer kinetic model used follows a general form previously presented for describing liposome transport into and out of tumors (20, 32). The pharmacokinetics of ⁶⁴Cu-MM-302 was represented with a single blood compartment with volume V_c and clearance characterized by elimination rate constant k_el. The tumor (volume = V_t, assuming tissue density ρ = 1 kg/L) was described in a semiphysiologic manner with a vascular portion and tissue portion consisting of cellular and interstitial space. The fractional volume of the tumor occupied by vasculature was described by a vascular volume fraction (VVF). Blood flow rate into and out of the tumor (Q) was assumed constant at 0.0282 L/kg/minute (33). Washout of particles from the tumor, either back into the blood or via lymphatic drainage, is lumped into a single process for simplicity. Deposition and washout of ⁶⁴Cu-MM-302 into and out of the tumor tissue space were assumed to follow first-order kinetics and were characterized by rate constants k₁ and k₋₁, respectively.

Equations are listed below:

where C_b, C_tv, C_tt, and C_t,total are the concentrations of ⁶⁴Cu-MM-302 in the blood, tumor vasculature, tumor tissue, and total tumor, respectively. The following parameters were estimated directly from the kinetic data: k_el, VVF, k₁, k₋₁ using median values extracted from ROI analysis from the images and associated tumor volume measurements. Additional information on the model parameters are provided in Supplementary Table S1. The model was implemented in MATLAB (The Mathworks).

Treatment outcome was not assessed on the companion imaging protocol, but was documented under the parent protocol per RECIST v1.1 criteria. Image-based ROI analysis data were captured using an eCRF developed in compliant with 21 CFR Part 11, with the exception of the bone marrow uptake data that were obtained for additional exploratory analysis. Uptake data were then exported for further computational analyses. Statistical and pharmacokinetic analyses were performed using MATLAB (The Mathworks) or GraphPad Prism version 6.0 (GraphPad Software). Only nonparametric statistical tests were utilized, including Mann–Whitney, Kruskal–Wallis, or Spearman correlation, wherever applicable. Data are shown as median (interquartile range) unless otherwise indicated.

It is well established that there is a dose–activity relationship for liposomal anticancer agents in preclinical tumor models (34). We therefore sought to identify the minimum tumor delivery of MM-302 required for antitumor activity in preclinical models (details described in Supplementary Material). As expected, increased tumor delivery was observed with increasing dose (Supplementary Fig. S2A; range: 1–13 μg doxorubicin per gram of tumor tissue, or μg doxorubicin/g). The 3 mg/kg dose level, roughly corresponding to the dose level at which MM-302 can effectively inhibit tumor growth (Supplementary Fig. S2B), resulted in a minimum tumor concentration of 2.3 μg doxorubicin/g. On the basis of these findings, we sought to determine the concentrations of liposomal drug achieved in tumor lesions in patients and hypothesized that a minimum effective dose would be necessary for effective treatment with liposomal therapies such as MM-302.

Between July 2013 and July 2014, 25 patients were dosed with MM-302 as part of a phase I study (NCT01304797); 19 of the patients underwent ⁶⁴Cu-MM-302 PET. Patient demographics are shown in Supplementary Table S2. All 19 patients were imaged on day 1 between 0.2 and 2 hours (median 0.7 hour) postadministration of ⁶⁴Cu-MM-302; 16 and 8 patients also underwent scans on day 2 (median of 21.2 hours) and day 3 (median of 44.6 hours), respectively. A total of 5 patients had three PET scans on days 1, 2, and 3 postadministration. The median age of imaged patients was 54 (range, 41–71) and the median body weight was 71.7 kg (range, 50.5–97.3 kg). Patients were enrolled into two different arms where Arms 3 and 4 include treatment with trastuzumab + MM-302 and cyclophosphamide + trastuzumab + MM-302, respectively.

The mean activity of ⁶⁴Cu-MM-302 administered was 389 MBq (range, 337–432 MBq). The administration of ⁶⁴Cu-MM-302 was well tolerated by all subjects. Any study drug-related adverse events recorded could not be uniquely attributed to ⁶⁴Cu-MM-302 because ⁶⁴Cu-MM-302 is administered within a few hours of MM-302, and the exposures to the two molecules overlap during the period of treatment-emergent adverse event reporting.

Whole-body distribution of ⁶⁴Cu-MM-302 is shown for 2 patients on days 1 to 3 postadministration in Fig. 1. Immediately following administration, ⁶⁴Cu-MM-302 is almost exclusively localized in the blood pool, as expected for a PEGylated liposome. On days 2 to 3 (19–47 hours), deposition of ⁶⁴Cu-MM-302 was seen in the liver and spleen, but there was relatively slow clearance from the blood. Images were consistent with hepatic clearance of MM-302, and no significant uptake of the tracer was seen in the kidneys or bladder of any patients.

Decay-corrected time–activity plots for ⁶⁴Cu-MM-302 in selected normal tissues are shown in Fig. 2 for all patients. Because of the relatively long circulation time of ⁶⁴Cu-MM-302, the presence of large blood vessels in any ROI has the potential to skew the average values within an ROI. For this reason, median activity within each ROI was used to provide a more robust measure of tissue uptake. There was very little uptake in normal muscle and normal lung tissue (<6 %ID/kg). Median normal liver uptake was approximately 6.5 (5.2–7.3) %ID/kg postadministration and increased with time to 15.7 (12.4–21.8) %ID/kg on day 3 (decay-corrected). Uptake into normal spleen appeared to be the most variable with 10.5 (8.6–13.3) %ID/kg postadministration, with a mild increase to 17.0 (13.5–18.3) and 13.2 (7.8–19.6) %ID/kg on days 2 and 3, respectively. ⁶⁴Cu-MM-302 uptake was also observed in the bone marrow, peaking between days 2 and 3 at 4.2 (3.6–4.7) and 3.6 (2.0–4.0) %ID/kg, with similar uptake kinetics as normal spleen; this is consistent with the expected uptake of nanoparticles by the reticuloendothelial system (35). Although the level of uptake in bone marrow seems to be on par with that of lung tissue, the kinetics of bone marrow uptake resemble that seen in the liver/spleen instead of the constant decrease observed for muscle and lung (i.e., blood clearance).

The behavior of ⁶⁴Cu-MM-302 in the blood pool was assessed from PET/CT images using an ROI in the aorta and is shown in Fig. 2F and G. The clearance of ⁶⁴Cu-MM-302 was highly reproducible and consistent with monoexponential clearance kinetics previously reported for MM-302 based on doxorubicin content (26, 27). Monoexponential clearance kinetics made it possible to determine estimates of the pharmacokinetic parameters from the 2 to 3 images obtained per patient. Estimation of pharmacokinetic parameters (compartment model) from the imaging data (mean t_1/2 = 33 hours) compared well with the 39-hour (95% confidence interval, 29–53 hours; n = 10) half-life reported for MM-302 at the 30 mg/m² dose level (26). One patient had significantly faster blood clearance (t_1/2 = 9.53 hours) with significant sequestration of ⁶⁴Cu-MM-302 at 0.5 hour in the liver and spleen (Supplementary Fig. S3). This patient was in Arm 4 (with cyclophosphamide treatment), but rapid clearance does not seem to be attributed to cyclophosphamide, as the blood t_1/2 was not significantly different in Arm 3 without cyclophosphamide [median t_1/2 = 36.5 (23.6–44.3 hours] compared with Arm 4 [median t_1/2 = 35.4 (25.2–40.9) hours; P = 0.97]. Most normal tissue uptake between the two groups were also comparable, with the exception of muscle on day 1 [Arm 3 = 1.9 (0.6–1.3) %ID/kg vs. Arm 4 = 0.6 (0.3–0.8) hour; P = 0.027] and bone marrow on day 2 [Arm 3 = 4.3 (4.1–5.4) %ID/kg vs. Arm 4 = 3.9 (2.9–4.4) hours; P = 0.042] where Arm 3 patients have higher uptake.

The stability of ⁶⁴Cu labeling of MM-302 was assessed in 3 patients. Free ⁶⁴Cu was not detectable above background any time point (Supplementary Fig. S4), indicating highly stable labeling of MM-302 with ⁶⁴Cu. In addition, almost all the ⁶⁴Cu activity remained in the serum component, with <4% ⁶⁴Cu detected in blood cells. These results were consistent with preclinical development of the ⁶⁴Cu-MM-302 labeling method (25).

Internal radiation dosimetry was evaluated in a subset of patients (n = 11). The target radioactivity was 400 MBq, and the median radioactivity administered to the 11 patients for radiation dosimetry analysis was 389 (range, 337–432) MBq. The mean estimated radiation absorbed dose for each organ, effective dose, and total body dose are shown in Supplementary Table S3. The highest absorbed radiation dose was in the heart wall, followed by the spleen, and liver receiving 0.25 (0.20–0.31), 0.12 (0.09–0.21), and 0.09 (0.09–0.12) mGy/MBq, respectively. The median effective dose was 0.028 mSv/MBq.

Tumor lesions could be visualized on days 2 and 3 of ⁶⁴Cu-MM-302 PET scans (Fig. 3). Similar to normal tissue uptake, there was no significant difference in lesion uptake for patients treated versus not treated with cyclophosphamide (Arm 3 vs. 4, P ≥ 0.67 for all 3 scans; Fig. 4A). On the basis of these findings, subsequent analyses were performed by combining data from the two treatment arms. Increasing activity in tumor lesions was noted from the initial scan on day 1 to days 2 to 3 scans at varying degrees (Fig. 4B), while the activity in the blood decreased over this time frame (Fig. 2F). In the initial scan, median ⁶⁴Cu-MM-302 signal in lesions was 2.3 %ID/kg (range, 0.24–7.1 %ID/kg, n = 68 lesions), reflecting predominantly tumor blood pool. Comparison with blood measurements yield a ratio of 0.13, which is consistent with the expected tumor vascular volume fraction determined by other methods (36–38). The median tumor-to-blood ratios were 0.080 (0.045–0.120), 0.208 (0.082–0.188), and 0.468 (0.111–0.451) on days 1, 2, and 3, respectively. There were a few lesions that showed either no change or a decrease in signal over time, but did not decrease as rapidly as the blood pool signal, still suggesting a small degree of ⁶⁴Cu-MM-302 deposition. Median ⁶⁴Cu-MM-302 depositions on days 2 and 3 were 3.7 (2.8–7.5) and 4.0 (2.3–10.6) %ID/kg, indicating no significant differences in lesion uptake between day 2 and day 3.

On the basis of the known ⁶⁴Cu-to-doxorubicin ratio, it is estimated that the tumor deposition of MM-302 on days 2 to 3 ranges from 0.22 to 11 μg doxorubicin/g. Ninety percent of the patient lesions had deposition that are consistent with the range of MM-302 tumor uptake in preclinical xenograft models (Supplementary Fig. S2A, range, 1–13 μg doxorubicin/g). ⁶⁴Cu-MM-302 uptake was heterogeneous within each patient among multiple lesions (Fig. 4C). The extent to which ⁶⁴Cu-MM-302 accumulates in individual lesions is therefore unique to the lesion characteristics rather than the pharmacokinetics of ⁶⁴Cu-MM-302, as lesion uptake on days 2 or 3 was not correlated with exposure (P > 0.08, Pearson correlation with blood AUC_0→∞; Fig. 4D). There was also no significant correlation between lesion signal and lesion volume (P > 0.27) for scan 2 or 3 (Fig. 4E).

Uptake was observed in lesions of various anatomic locations, including the lung, brain, lymph nodes, and bone, as shown in Fig. 3 with deposition quantification presented in Fig. 4F. Hepatic lesions primarily appeared as hypoactive regions, because of the high background activity of the liver, but showed evidence of signal enhancement over time. There were also several instances where hepatic lesions had uptake similar to that of normal liver tissue, requiring reference to contrast-enhanced diagnostic CT to guide ROI selection (Supplementary Fig. S5). In all three scan time points, hepatic lesions were consistently higher than other anatomic locations (Fig. 4F; P < 0.0001 on days 2 and 3; only anatomic locations with more than three data points were included in statistical analyses).

In patients who consented to undergo three scans, a detailed kinetic analysis of ⁶⁴Cu-MM-302 in tumor lesions was possible. Figure 5 shows the ⁶⁴Cu-MM-302 lesion deposition kinetic profiles of a patient, who had 4 tumor lesions that were selected for analysis by the independent reviewer. Quantification of ⁶⁴Cu-MM-302 uptake for the individual lesions is shown in Fig. 5A. A chest wall lesion showed comparatively high uptake on day 3, whereas the two hepatic lesions and one neck lesion had comparatively lower uptake. Images corresponding to comparatively high (ROI 3) versus low (ROI 4) uptake lesions are shown in Fig. 5B.

Three imaging data points and a basic set of assumptions regarding ⁶⁴Cu-MM-302 transport (see Materials and Methods) enabled elucidation of the full kinetics of ⁶⁴Cu-MM-302 deposition into and out of the tumor lesions, including determination of the contribution of blood versus tissue-deposited ⁶⁴Cu-MM-302. A schematic of the tracer kinetic model used is shown in Fig. 5C. The model assumes first-order clearance of ⁶⁴Cu-MM-302 from the central blood pool and a fixed rate of convective transport (blood flow) from the central compartment into the vascular compartment of the tumor. The tumor is characterized by a vascular volume fraction and ⁶⁴Cu-MM-302 transport from the vascular compartment into the tissue, through a combination of convective and diffusive transport, is characterized by a first-order rate constant. Washout of ⁶⁴Cu-MM-302 is also assumed to follow first-order kinetics. Cellular uptake and processing of ⁶⁴Cu-MM-302 is also assumed to be minimal over the 3-day time scale of imaging. Shown in Fig. 5D, the kinetic model was fit to the blood and tumor data, demonstrating that this model framework was able to quantify the interlesion heterogeneity of uptake through variations in rates of deposition, washout, and vascular volume fraction (Supplementary Fig. S6). The model results indicate peak deposition of MM-302 in human tumors occurs on approximately days 2 to 3 and supports the selected imaging times used. The fraction of tumor ⁶⁴Cu-MM-302 signal on days 2 to 3 arising from the vascular versus deposited ⁶⁴Cu-MM-302 was minimal compared with the tumor tissue signal, suggesting that most ⁶⁴Cu tumor signal after day 1 is attributed to deposition of ⁶⁴Cu-MM-302 in the tumor tissue.

Inadequate drug delivery to a single tumor lesion within a patient may be sufficient to result in progressive disease when patients are evaluated on the basis of RECIST v1.1. For this reason, we classified patients into two groups based on their lowest uptake lesion. At approximately 2 %ID/kg, a plateau was observed based upon the distribution of lowest uptake lesion deposition of each patient (Fig. 6A). Tracer deposition of 2 %ID/kg corresponds to approximately 1.2 μg doxorubicin/g, which is comparable with the preclinical effective threshold established previously (2.3 μg doxorubicin/g; Supplementary Fig. S2).

Although not a formal objective of the study, a preliminary retrospective analysis was attempted to correlate ⁶⁴Cu-MM-302 lesion uptake and patient response to treatment. The best overall responses per RECIST v1.1 for the two groups classified on the basis of imaging data are shown in Fig. 6B–D. Seventy-five percent of subjects experienced a partial response (PR) and/or stable disease (SD) in the high deposition group, whereas 43% of patients experienced SD with no PR in the low deposition group. The patients in groups with the low and high ⁶⁴Cu-liposome deposition lesions had median progression-free survival (PFS) of 1.7 and 2.0 months, respectively, with an HR of 0.42 (log-rank method). ROC analysis of the minimum lesion uptake against PFS yielded a deposition threshold of 2.4 %ID/kg, which resulted in the same patient stratification. The limited number of imaged patients on each treatment arm with analyzable RECIST data (n = 9 and 10 for Arms 3 and 4, respectively) precluded a meaningful statistical comparison.

The extent of therapeutic nanoparticle deposition in solid tumors is a vital component of establishing local drug concentrations and the overall therapeutic window. Understanding the extent and variability of the EPR effect in patient tumors is at the core of understanding whether local drug levels are limiting in patient responses to therapeutic nanoparticles (3). Although preclinical examples exist, this simple pharmacologic concept has been difficult to translate into the clinical setting (39).

In this study, we utilized preclinical models to identify a minimum critical concentration threshold of MM-302 tumor delivery required to control tumor growth and translated this concept into a clinical study. We hypothesized that tumor lesion delivery of therapeutic nanoparticles, such as MM-302, is highly variable, and locally achievable drug concentrations span critical thresholds that determine sensitivity versus lack of response. To test this hypothesis, we transformed MM-302 into a PET tracer through ⁶⁴Cu-labeling and incorporated imaging with ⁶⁴Cu-MM-302 PET/CT into a clinical trial of MM-302 plus trastuzumab with or without cyclophosphamide. Patient tumor concentrations of MM-302, as determined with ⁶⁴Cu-MM-302, were remarkably consistent with those determined from preclinical data and span a range of concentrations that potentially impact therapeutic responses. Our results have important implications for clinical development of therapeutic nanoparticles.

To date, patients with cancer have been imaged in several clinical studies with ¹¹¹In-liposomes or ^99mTc-liposomes via planar scintigraphy or single-photon emission CT (SPECT) to assess tumor uptake of the liposomes (40–43). However, these clinical studies were limited to visualization or semiquantitative analysis of liposome behavior because of the limitations of the imaging methods and/or the short half-life of the radionuclide tracer. Relative to SPECT, PET has the advantage of increased sensitivity and spatial resolution while allowing for straightforward data quantification and whole-body 3D imaging.

The kinetic nature of our studies with ⁶⁴Cu-MM-302, coupled with the accuracy afforded by PET/CT, clearly establish that the EPR effect is present in human metastatic tumors. The accumulation of ⁶⁴Cu-MM-302 in tumor lesions over time was not observed in normal tissues, such as muscle, and is consistent with the EPR effect of nanoparticles as reported in the literature (44, 45). The data indicate the variable nature of this process, not only in terms of deposition rates but washout rates as well. Rates of deposition and washout determined from human tumors compared well with rates estimated from mouse xenograft models (20). Furthermore, kinetic modeling indicated that at later times, the ⁶⁴Cu signal was predominantly the result of tissue-deposited liposomes. In contrast, early imaging times, as used in ^99mTc-liposomes studies, are largely a measurement of tumor lesion vascular volume fraction (41–43).

This study also highlights the interplay between normal liver and hepatic tumor lesions for therapeutic nanoparticles. As expected, the liver was a site of high background signal because of its role in metabolizing and clearing liposomes. In contrast, little to no signal was observed in the kidney or bladder, supporting the stability of the ⁶⁴Cu-MM-302 tracer, as well as the liver being the primary route of clearance. The liver is also a common site of metastatic disease in breast and other cancers. Hepatic lesions were able to be visualized in most cases, as they tend to appear as hypoactive regions relative to the high normal tissue uptake background. In some cases, where hepatic lesion uptake was similar to normal liver tissue, diagnostic CT was required to help delineate the liver–lesion boundary. The spatial resolution of PET and partial volume effects make accurate quantification of these lesions difficult. This may be particularly true for lesions with an actively growing rim and/or necrotic core. Similarly, lesions with significant fibrosis or scar tissue may also be difficult to resolve from actively growing lesions by CT. For these reasons, comparison of ⁶⁴Cu-MM-302 PET with ¹⁸F-FDG PET would be an interesting future direction for identifying metabolically active lesions to further the understanding of liposome tumor deposition.

The data presented herein provide evidence supporting a mechanism by which ⁶⁴Cu-MM-302 liposomes deposit and accumulate in human tumors. An interesting and serendipitous finding of this study was the accumulation of ⁶⁴Cu-MM-302 in brain lesions. The delivery of large molecules to brain lesions has been previously shown by imaging with liposomes (41) and ⁶⁴Cu-/⁸⁹Zr–labeled trastuzumab (46, 47). Similarly, Siegal and colleagues had reported a 14-fold increase in liposomal doxorubicin delivery to brain tumor in a mouse model compared with adjacent normal brain tissue (48). This almost certainly reflects disruption of the blood–tumor barrier in metastatic lesions, rather than large liposomes traversing the blood–brain barrier directly. Greater disruption would be predicted to lead to increased delivery and might determine the extent of response to therapy (49, 50).

Variable ⁶⁴Cu-MM-302 uptake occurred both across lesions within a patient and across patients. In general, in patients with multiple lesions, not all of the lesions had the same level of uptake. This suggests that patient classification based on nanoparticle delivery is potentially complex and cannot be solely determined by systemic exposure. However, patient lesion data can be classified into two groups: (i) variable uptake including low-uptake lesions, and (ii) variable uptake with only high-uptake lesions, based on the rationale that poor delivery to a single lesion could be sufficient to presage disease progression. Classification based on imaging data selected about one third of patients as “low uptake” and enabled a subsequent exploratory analysis of patient outcomes. Although our retrospective analysis illustrates that patient level classification based on lesion delivery is possible and was consistent with our delivery-based hypothesis, these results are limited by the small sample size of this phase I study among other factors, including the inherent chemosensitivity. For instance, patient classification was performed on the basis of the imaging data only; however, some patients received cyclophosphamide and others did not. Nonetheless, we did not find a systematic difference in lesion uptake between lesions that were treated with cyclophosphamide versus those that were not, suggesting this was not a primary factor in dictating outcome.

The significance of this work also extends to an improved understanding of safety. EPR effect also occurs in bone marrow and explains well-known hematologic toxicities of therapeutic nanoparticles, such as liposomal doxorubicin. Interestingly, less variability in uptake was observed in bone marrow than in tumor lesions, in general. This implies that the primary opportunity for patient selection lies in identifying patients based on uptake in tumor lesions rather than safety.

Together, these data suggest that it may be possible to use pretreatment imaging of nanoparticle deposition in tumors as a potential means to identify patients most likely to benefit from treatment with therapeutic nanoparticles. Future directions include development of potential diagnostic imaging agents specifically designed to assess tumor delivery of therapeutic nanoparticles. These agents would enable a comprehensive understanding of the delivery of nanomedicines to tumor lesions as a function of indication and anatomic location. Identification of patient characteristics correlating with effective nanoparticle delivery has potential to greatly benefit patients and dramatically influence clinical development decisions. Prospectively defined clinical trials will be needed to formally establish the relationships between therapeutic nanoparticle tumor delivery and treatment outcome.

B.A. Siegel reports receiving commercial research grants from and is a consultant/advisory board member for Merrimack Pharmaceuticals. I. Krop reports receiving other commercial research support and speakers bureau honoraria from Genentech. K. Campbell and T.J. Wickham hold ownership interest (including patents) in Merrimack Pharmaceuticals. D.B. Kirpotin reports receiving other commercial research support from and holds ownership interest (including patents) in Merrimack Pharmaceuticals. V. Moyo is an employee of and holds ownership interest (including patents) in L.E.A.F. Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H. Lee, A.F. Shields, B.A. Siegel, K.D. Miller, P.M. LoRusso, P.N. Munster, D.B. Kirpotin, V. Moyo, T.J. Wickham, B.S. Hendriks

Development of methodology: H. Lee, A.F. Shields, B.A. Siegel, P.N. Munster, E. Geretti, D.B. Kirpotin, T.J. Wickham, B.S. Hendriks

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Lee, A.F. Shields, B.A. Siegel, K.D. Miller, I. Krop, C.X. Ma, P.M. LoRusso, P.N. Munster, S.C. Leonard, E. Geretti, S.J. Blocker

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Lee, A.F. Shields, B.A. Siegel, K.D. Miller, C.X. Ma, P.N. Munster, S.C. Leonard, T.J. Wickham, B.S. Hendriks

Writing, review, and/or revision of the manuscript: H. Lee, A.F. Shields, B.A. Siegel, K.D. Miller, I. Krop, C.X. Ma, P.M. LoRusso, D.F. Gaddy, E. Geretti, D.B. Kirpotin, V. Moyo, T.J. Wickham, B.S. Hendriks

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Lee, D.F. Gaddy

Study supervision: H. Lee, A.F. Shields, K.D. Miller, P.N. Munster, K. Campbell, E. Geretti, V. Moyo, T.J. Wickham

We thank the patients and the clinical staff for their participation and support of the study. We also thank Shari Rabenstine and Paul Galette from ICON Medical Imaging for their excellent study support, image interpretation, and fruitful discussions. Tom Voller from Washington University has been tremendously helpful in coordinating logistics and supply of ⁶⁴Cu for the clinical study. We would like to acknowledge Cardinal Health for their role in preparation and distribution of ⁶⁴Cu-MM-302 to clinical sites. We also appreciate Dr. Michael Stabin's participation in the radiation dosimetry analysis. Lastly, we would like to thank Jonathan Fitzgerald, Istvan Molnar, Ty McClure, and Ulrik Nielsen from Merrimack Pharmaceuticals for their helpful input.

This study was solely funded by Merrimack Pharmaceuticals.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.