Preclinical Efficacy of Bevacizumab with CRLX101, an Investigational Nanoparticle–Drug Conjugate, in Treatment of Metastatic Triple-Negative Breast Cancer

Patients with triple-negative breast cancer (TNBC) have the highest risk of recurrence and metastasis (1). Various targeted therapies have been investigated, but as yet, none are currently approved. One notable example is bevacizumab, a VEGF-targeting antibody, which was granted accelerated FDA approval with weekly paclitaxel for first-line treatment of metastatic breast cancer, but this approval was later revoked after follow-up phase III clinical trials with different chemotherapy backbones showed less impressive benefits in progression-free survival (PFS; ref. 2). More recently, however, there has been renewed interest in reconsidering bevacizumab for the treatment of breast cancer based on several phase III clinical trial results, including one in the neoadjuvant and adjuvant setting (NSABP-B40) and two in the maintenance metastatic setting (IMELDA and TANIA; refs. 3–5). Importantly, a number of trial results and meta-analyses suggest the extent of beneficial effect (and associated toxicities) of adding bevacizumab to chemotherapy may depend on the concurrent chemotherapy regimen used (2, 4, 6, 7). Thus, an appropriate chemotherapy partner, one with better efficacy, manageable toxicity, and complementary modes of action, may be critical to gaining optimal benefit out of adding bevacizumab (and vice versa). One promising investigational chemotherapy drug is a nanoparticle–drug conjugate (NDC) known as CRLX101, which contains the payload camptothecin, a highly potent cytotoxic agent that inhibits topoisomerase-I (8, 9).

Camptothecin showed impressive preclinical antitumor activity but had low solubility, metabolic instability of the active lactone form, rapid renal clearance, and severe toxicities, which resulted in disappointing results in early-phase clinical trials (10, 11). A number of analogues were thus developed to improve the solubility of camptothecin, including the subsequently approved drugs, topotecan (Hycamtin, GlaxoSmithKline) and irinotecan (Camptosar, Pfizer; refs. 9, 12). Currently, there is no topoisomerase-I inhibitor approved for breast cancer treatment. Although both irinotecan and topotecan have been evaluated in phase II trials for metastatic breast cancer, they have not been approved as treatments for that indication given the associated high-grade toxicities experienced by patients for a modest benefit gain (13–17). Nonetheless, lessons learned from early trials using camptothecin (18), topotecan, and irinotecan suggest camptothecin analogues are indeed active in the treatment of breast cancer, but further clinical development requires better drug solubility, improved lactone ring stabilization, and less systemic toxicity. Notably, a recent phase III trial (BEACON) evaluated etirinotecan pegol, a new formulation of irinotecan, in patients with recurrent or metastatic breast cancer (19). It showed single-agent PFS benefits similar to treatment of physician's choice. Moreover, subgroup analyses showed etirinotecan pegol significantly prolonged overall survival in patients with a history of brain or liver metastases, and with two or more sites of disease (19). As an NDC of camptothecin, CRLX101 was designed to have superior solubility and stabilization of the lactone ring, as well as favorable safety and pharmacokinetics in patients (20, 21).

CRLX101 is composed of a cyclodextrin-containing polymer conjugated to camptothecin. In mice, the cyclodextrin polymer itself has no observable side effects or antitumor efficacy when tested up to 240 mg/kg (22). CRLX101 was designed to have improved accumulation within tumors by the enhanced permeability and retention (EPR) effect and thus reduced systemic exposure and toxicity (8, 20). The drug has been administered to more than 300 patients to date and appears to be generally well tolerated, achieving an overall response rate of 16% in 19 patients in a phase II clinical trial of platinum-resistant ovarian cancer as a monotherapy (23) and 23% in 22 patients in a phase I/II study of metastatic renal cell carcinoma in combination with bevacizumab (21).

We recently reported that CRLX101 plus bevacizumab resulted in synergistic antitumor efficacy in a preclinical model of advanced, metastatic ovarian cancer (23). CRLX101 was also shown to effectively and durably suppress elevated hypoxia-induced upregulation of hypoxia-inducible factor-1α (HIF1α) following therapy with bevacizumab, thus downregulating expression of downstream HIF1α-regulated markers, such as carbonic anhydrase IX (CAIX; ref. 23), and blocking the induction of cancer stem cells (24).

Here, we evaluated the combination of CRLX101 and bevacizumab in long-term therapy experiments of orthotopic primary TNBC xenografts either derived from tumor tissue fragment implantation of a patient-derived xenograft (PDX) called HCI-002 (25, 26) or cell injection of LM2-4, a luciferase-tagged variant of the established cell line MDA-MB-231 serially selected in vivo for aggressive spontaneous metastatic properties (27). We also evaluated long-term efficacy of CRLX101 and bevacizumab in a preclinical model of postsurgical, advanced metastatic TNBC (28), a model that recapitulates the more challenging clinical treatment of systemic metastatic disease (28, 29). We report that CRLX101 alone and in combination with bevacizumab is an effective treatment for advanced metastatic TNBC in these mouse models and provide new mechanistic results to help explain this encouraging antitumor activity.

MDA-MB-231/LM2-4luc16+ (LM2-4) is a highly aggressive variant of MDA-MB-231 [parental line obtained from Jeff Lemontt (Genzyme Corp., Boston, MA) in 2000] and maintained in RPMI1640 supplemented with 5% FBS (27). LM2-4 was last authenticated in 2013 by Genetica DNA Laboratories (a LabCorp Specialty Testing Group) using analytical procedures for DNA extraction, PCR, and capillary electrophoresis on a 3130xl Genetic Analyzer (Applied Biosystems). The 13 core CODIS short tandem repeat (STR) loci plus PENTA E and PENTA D, and the gender-determining locus, amelogenin, were analyzed using the commercially available PowerPlex 16 HS Amplification Kit (Promega Corporation) and GeneMapper ID v3.2.1 software (Applied Biosystems) with appropriate positive and negative controls. Authentication of cell lines is confirmed by entering the STR DNA profile of each tested cell line into known repository cell line databases; authentication is defined as having a percent match with the reference STR profile ≥80% when using the ANSI/ATCC guidelines (ASN-0002-2011) or having a “unique” STR DNA profile (no matches found) for “in-house” cell lines not distributed by any cell line repository.

PDX HCI-002 tumor fragments were generously provided by Dr. Alana Welm (University of Utah, Salt Lake City, UT). Tumor fragments were originally obtained from a patient with TNBC and verified by histology to have retained morphology and characteristics as matched patient samples (25, 26). HCI-002 tumor fragments were serially passaged in SCID mice, and only early passages were used for all experiments.

Eight-week-old female YFP-SCID mice were bred in-house. Procedures involving animals and their care were conducted in strict conformity with the guidelines of Sunnybrook Health Science Centre (Toronto, Ontario, Canada) and the Canadian Council of Animal Care. Mammary fat pad injections of LM2-4 (5 × 10⁵ cells) and tumor fragment implantations (pieces of 2–5 mm³) of HCI-002 were carried out as described previously (26, 27). Mice were randomized by tumor volume prior to treatment initiation. All doses of CRLX101 are reported as camptothecin-equivalent doses.

In vivo imaging of primary tumors was conducted using a commercially available high-frequency laser-integrated ultrasound system (VevoLAZR, VisualSonics), which allowed for both photoacoustic and contrast-enhanced imaging (30). Contrast data were quantified with VevoCQ using time intensity curves (TIC) generated from the wash-in of microbubbles. Two parameters are taken from the TIC: peak enhancement and wash-in rate. Dual-wavelength PA imaging was used for real-time monitoring and calculation of oxygen saturation with the VevoLAB software.

LM2-4 cells were orthotopically implanted into the right mammary fat pad and resected when tumors reached 400 to 500 mm³, as described in ref. 27. Following primary tumor resection, distant visceral metastases can be detected by total body bioluminescence imaging using an IVIS200 Xenogen. Mice with metastatic disease were randomized on the basis of metastatic load and location before therapies were initiated (27, 29).

Results were reported as mean ± SD or SEM, as indicated. Tumor growth curves were reported as mean ± SD. Survival curves were plotted by the method of Kaplan and Meier and tested for survival differences with the log-rank test. Statistical significance was assessed by one-way ANOVA (Kruskal–Wallis test with Dunn post hoc) or t test (Mann–Whitney, two-tailed) using GraphPad Prism 4 (P < 0.05 was used as the threshold of statistical significance).

The MTD (8 mg/kg i.p. once weekly) of CRLX101 was previously established for SCID mice and is well tolerated even when combined with bevacizumab for long-term studies (23). In a primary tumor model derived from cell injection of LM2-4, CRLX101 caused rapid and durable tumor regressions (Fig. 1). Lower doses of CRLX101 monotherapy resulted in slower rates of tumor shrinkage compared with 8 mg/kg, resulting in small tumors that still eventually caused hind leg mobility impairment (smaller tumor volumes compared with vehicle but not statistically significant). Although the addition of bevacizumab to 2 mg/kg CRLX101 showed a trend for tumor growth delay, this was not statistically significant compared with 2 mg/kg CRLX101 monotherapy. Nonetheless, the addition of bevacizumab to 4 mg/kg CRLX101 significantly improved antitumor efficacy (P < 0.05 compared with 4 mg/kg CRLX101 alone), showing improved tumor growth suppression and shrinkage despite bevacizumab alone having no antitumor efficacy. However, despite early tumor suppression and continued therapy, bioluminescence imaging of mice treated with CRLX101 4 mg/kg plus bevacizumab showed these tumors eventually continued to grow. Notably, while mice were on-therapy continuously for 6 months, 8 mg/kg CRLX101 quickly and dramatically shrank established primary tumors, resulting in complete regressions (P < 0.001), with 3 of 5 mice essentially cured by 6 months of therapy (no observable bioluminescence signals). Similarly, 8 mg/kg CRLX101 plus bevacizumab also led to complete primary tumor regressions (P < 0.001 compared with vehicle, no statistically significant difference compared with 8 mg/kg CRLX101 monotherapy). Four of 5 mice treated with 8 mg/kg CRLX101 and bevacizumab did not have any residual disease or regrowth even after being off-therapy for an additional 4 months (Supplementary Fig. S1).

Compared with primary tumors derived from LM2-4, PDX HCI-002 tumors were highly vascular. In contrast to bevacizumab's absence of efficacy when treating LM2-4 primary tumors, bevacizumab led to significantly delayed HCI-002 tumor growth while mice were on-therapy continuously for 5 months (P < 0.05 compared with vehicle; Fig. 2). Nonetheless, both doses of CRLX101 durably suppressed tumor growth (P = 0.002 for 4 mg/kg and P < 0.001 for 8 mg/kg CRLX101). Notably, the combination of either dose with bevacizumab significantly improved the antitumor efficacy compared with either drug alone (P < 0.001 compared with monotherapy). In this PDX model, despite tumor growth suppression or regression while mice were on-therapy, when all therapies were stopped after 5 months, tumor regrowth did occur but was delayed longer if mice were originally treated with a higher dose of CRLX101 or with concurrent bevacizumab.

In separate experiments, tumors from mice treated for two weeks were evaluated for changes in microvessel density (MVD) and any corresponding changes in HIF1α (Fig. 3 and Supplementary Figs. S2 and S3). Bevacizumab reduced the number of tumor vessels by approximately 50% in LM2-4 tumors (Fig. 3A) and approximately 65% in HCI-002 tumors (Supplementary Fig. S2). This reduced MVD by bevacizumab monotherapy corresponded to an increase in HIF1α protein levels (Fig. 3B), which was confirmed by similar trends observed using immunohistochemical staining of HIF1α and CAIX, a downstream marker of HIF1α activity (Fig. 3C and D; ref. 23). In vitro, higher doses of CRLX101 showed a trend (although not statistically significant) of decreasing VEGF levels (Supplementary Fig. S3). However, in vivo, CRLX101 monotherapies did not significantly decrease MVD even though CRLX101 maintained low levels of HIF1α protein levels. Although CRLX101 monotherapies did not change the extent of tumor cell proliferation (Fig. 3E), there was a significant increase in apoptosis at all doses administered (Fig. 3F). Tumors from mice treated with concurrent bevacizumab and CRLX101 showed decreased MVD to levels similar to bevacizumab monotherapy. Nonetheless, concurrent CRLX101 was able to suppress any bevacizumab-induced increases in HIF1α. The reduced MVD observed for combination-treated tumors did not affect tumoral accumulation of CRLX101 (Supplementary Fig. S2). Interestingly, concurrent bevacizumab did not further increase the extent of apoptosis observed compared with CRLX101 monotherapies.

We next evaluated whether the reduction in MVD by bevacizumab correlated with a change in functional tumor perfusion using in vivo contrast-enhanced ultrasound (CEUS) imaging, which uses gas-filled microbubbles to noninvasively monitor changes in blood flow and volume within tumors following therapy (Fig. 4; ref. 30). Tumors from mice treated with vehicle or bevacizumab alone showed decreased tumor blood volume and flow rate over the course of two weeks of therapy. In contrast, tumors from mice treated with CRLX101 monotherapy showed greatly improved perfusion, whereas perfusion in tumors treated with the combination remained unchanged. Parametric maps showing regions of high, low, or no perfusion confirm that tumors from mice treated with CRLX101 had smaller necrotic cores compared with vehicle and bevacizumab treatment. In the combination group, while a less perfused core was present, the surrounding “viable rim” remained well perfused. Similarly, CEUS imaging of PDX tumors showed CRLX101 maintained higher tumor perfusion and reduced tumor hypoxia, also confirmed by pimonidazole staining (Supplementary Fig. S4).

In addition to measuring perfusion using CEUS, photoacoustic imaging was used to measure oxygen saturation (Fig. 4; ref. 30). Short laser pulses are directed at the tumor, generating thermoelastic expansion to create acoustic waves detected by an ultrasound transducer. Different absorption spectra of deoxygenated and oxygenated hemoglobin are then used to noninvasively estimate the spatial distribution of oxygen saturation in vivo (30). Consistent with CEUS data showing improved perfusion in CRLX101-treated tumors, these tumors had relatively higher tissue oxygenation levels compared with vehicle- and bevacizumab-treated tumors (Fig. 4C).

Improved perfusion with CRLX101 treatment suggests a more functional tumor vasculature, as results from Fig. 3A show CRLX101 does not result in an increase in the number of vessels. Tumor cells rapidly growing within a confined space causes compression (or “solid stress”) of tumoral blood vessels (31, 32). One possible explanation for the overall improved perfusion following CRLX101 is the drug causes extensive tumor cell apoptosis, which may then relieve compression of tumor blood vessels. To assess whether this may be the case, we evaluated whether CRLX101 resulted in more vessels with an open lumen, which would allow better blood flow within the tumor (Fig. 4D). Indeed, we observed that CRLX101 monotherapy resulted in a higher proportion of tumor blood vessels with open lumen compared with both vehicle and bevacizumab treatment. It is worth noting that in tumors treated with the drug combination, overall number of tumor vessels was reduced, but more of the remaining vessels had open lumens. These results were similarly confirmed in PDX tumors (Supplementary Fig. S5).

To obtain confirmation that CRLX101 relieves compression of tumor blood vessels by killing tumor cells, we evaluated whether tumor cell-packing density, that is, the number of cells found within a given area, changed following therapy (Fig. 5A). Although the average number of cells present within an imaged field was similar between vehicle- and bevacizumab-treated tumors, all doses of CRLX101 caused a significant reduction in cell-packing density. Fewer tumor cells densely packed within a tumor thus may result in fewer compressed tumor blood vessels and hence improved perfusion and oxygenation.

Paradoxically, a more functional vasculature could result in a higher degree of tumor cell dissemination and metastasis to distant organs, such as the lungs. We therefore analyzed lungs from mice still bearing primary tumors for the presence of micrometastases or tumor cells that previously shed from the primary tumor and seeded in the lungs (Fig. 5B). Mice treated with bevacizumab monotherapy showed more micrometastases that were also larger in cluster size than those present in vehicle-treated mice. This is consistent with some previous preclinical reports that antiangiogenic agents used as monotherapies may cause an increase in metastasis and promote disease progression in mice despite an initial antitumor effect (33, 34). In stark contrast, mice treated with CRLX101 (particularly at higher doses) either showed no micrometastases or only singly dispersed tumor cells within the lungs, even when combined with bevacizumab. CRLX101 was therefore able to counteract potential prometastatic effects of bevacizumab. This effect is similar to published data where paclitaxel counteracted the proinvasive and prometastatic effects of DC101, the VEGFR-2 antibody (20), and metronomic topotecan reduced metastatic spread elicited by sunitinib (35).

Given the potency of CRLX101 in both primary tumor models and evidence showing CRLX101 was able to prevent formation of micrometastases, CRLX101 alone and with bevacizumab were evaluated in a postsurgical model of advanced metastatic breast cancer. Following surgical resection of established primary LM2-4 tumors, distant metastases were allowed to develop before treatment was initiated. Metastases were monitored by bioluminescence imaging and can manifest as lymph node metastases (causing mobility issues), lung metastases (leading to labored breathing), primary tumor regrowth or local metastases at the surgical site, as well as liver metastases and ascites (causing distended abdomens). Given the inherent variability of when and where metastases appear, every treatment group had randomized cohorts of mice with apparent distant metastases, local metastases or regrowths, as well as mice with no signs of metastases at the start of therapy (Fig. 6). Within 1.5 months of therapy, all mice treated with vehicle and bevacizumab monotherapy succumbed to disease, even if no apparent metastases were observed at the start of treatment. In contrast, bioluminescence imaging showed that in mice that had apparent metastases at the start of therapy, treatment with CRLX101 or CRLX101 plus bevacizumab caused these existing metastases to regress. In mice with no apparent metastases at the start of treatment, CRLX101 alone or combined with bevacizumab was able to prevent the emergence of new metastases both while mice were on-therapy (for 7 months) and then off-therapy (for 2 months). It should be noted, however, that if mice had a very heavy metastatic load in the lungs or ascites at the start of therapy, CRLX101 therapy was not as effective; we hypothesize this may be due to the need for some initial time for CRLX101 to accumulate within tumor cells and release its drug payload.

CRLX101 thus significantly extended the survival of mice with metastatic disease (Fig. 6B). Although mice treated with vehicle or bevacizumab succumbed to metastases after 4 to 6 weeks, survival of mice treated with 4 mg/kg CRLX101 monotherapy was extended to 15 weeks. CRLX101 at 4 mg/kg plus bevacizumab was effective at durably suppressing metastasis. Notably, MTD CRLX101 was highly efficacious, so much so that any added benefit of combining it with bevacizumab was not detected. Upon necropsy, 3 of 9 mice (treated with 4 mg/kg CRLX101 plus bevacizumab), 5 of 8 mice (treated with 8 mg/kg CRLX101 alone), and 6 of 9 mice (treated with 8 mg/kg CRLX101 plus bevacizumab) were still alive and free of macroscopic metastases at the end of the experiment (7 months on-therapy and 2 months off-therapy).

We recently reported metronomic topotecan plus pazopanib delayed tumor growth of primary TNBC tumors and prolonged survival of mice with advanced metastatic disease (36). Given this promising therapy of combining a topoisomerase-I inhibitor (topotecan) and an antiangiogenic drug (pazopanib) for treating TNBC in mice, here, we evaluated CRLX101 and bevacizumab. There are several reasons why this alternative drug combination may be superior to that of topotecan and pazopanib. Single-agent topotecan had no antitumor efficacy and resulted in increased HIF1α. In contrast, single-agent CRLX101 is potently effective in not only delaying tumor growth, but causing marked and sustained tumor regressions. Furthermore, CRLX101 whether alone or combined with bevacizumab durably suppressed HIF1α. Finally, one important concern of combining topotecan and pazopanib is the tolerability and toxicity of both drugs, especially the combination, in the clinic (13, 14, 37). In contrast, current clinical data suggest that CRLX101 is tolerable in combination with standard doses of bevacizumab in patients (38, 39). The combination of CRLX101 and bevacizumab would thus appear to be more promising considering its lesser toxicity and better preclinical efficacy, an improved therapeutic index compared with the pazopanib/topotecan combination.

Limited efficacy successes of approved VEGF pathway–targeting antiangiogenic drugs may be due to several possible factors, including reduced intratumoral delivery of concurrently administered drugs, such as chemotherapy, as well as elevated hypoxia (and hence HIF1α), which contributes to resistance and may promote metastases (24, 40–42). Given these considerations, a complementary chemotherapy partner for combination with an antiangiogenic agent would be a highly potent agent that is able to improve tumor perfusion and reduce HIF1α without increasing tumor dissemination and/or metastases. Here, we showed that CRLX101 may be such a drug.

A number of strategies have been proposed, including vessel normalization (43) and “vascular promotion” therapy (31, 44), to improve intratumoral chemotherapy drug delivery. Alternatively, impaired perfusion, whether caused by antiangiogenic therapy or due to mechanical factors, such as tumor cells compressing vessels (31, 32), would be expected to reduce intratumoral delivery of chemotherapy drugs, such as CRLX101, in solid tumors (8, 45). However, CRLX101 maintained tumor perfusion despite concurrent antiangiogenic therapy. We report here that the potent antitumor activity of CRLX101 alleviated the solid stress by rapidly targeting tumor cells, in effect leading to major tumor regressions and decompressing tumor blood vessels to improve tumor perfusion [“tumor priming” (46)]. We and others have reported that metronomic dosing of a gemcitabine prodrug resulted in increased perfusion (47, 48). Paclitaxel-loaded tumor-penetrating microparticles have also been used to enhance siRNA delivery into solid tumors (46). Clinically, a similar effect was observed, where weekly paclitaxel reduced interstitial fluid pressure and improved tumor oxygenation (49). Interestingly, results from these studies and our work here with CRLX101 suggest chronic exposure of tumor cells to a cytotoxic drug, either by administering chemotherapy drugs in a frequent, metronomic fashion, or using drug formulations with intrinsically long half-lives, such as CRLX101, may be required to maintain sufficient durable tumor cell kill to persistently reduce compression of tumoral vessels and thus potentially improve perfusion. This improvement in tumor perfusion significantly reduces tumor hypoxia, which contributes to the ability of CRLX101 to effectively and durably suppress HIF1α. Even though HIF1α is elevated by tumor hypoxia, reduction in tumor hypoxia by improving tumor perfusion alone may not completely account for the HIF1α suppression by CRLX101. Rapisarda and colleagues have shown in vitro that metronomic topotecan was able to suppress HIF1α protein accumulation even when cells were maintained in normoxic conditions (50), suggesting HIF1α inhibition may occur even in the absence of elevated hypoxia. Taken together, we have shown that CRLX101 maintains or improves tumor perfusion and is able to durably suppress HIF1α protein levels even in the presence of bevacizumab, where tumors are more hypoxic, thus making CRLX101 an effective chemotherapy partner to pair with an antiangiogenic agent, such as bevacizumab.

One limitation of our results is the use of two primary tumor models and one metastatic model. Nevertheless, given our preclinical results of CRLX101 showing potent antitumor activity, particularly when treating advanced metastatic TNBC, and its complementary mode of action for combining with bevacizumab, CRLX101 plus bevacizumab may be a promising treatment strategy for breast cancer and other solid tumors.

C.G. Peters is a scientist at Cerulean Pharma. S. Eliasof has ownership interest (including patents) in Cerulean Pharma, Inc. F.S. Foster reports receiving a commercial research grant from and is a consultant/advisory board member for VisualSonics. R.S. Kerbel reports receiving other commercial research support from Cerulean and is a consultant/advisory board member for Cerulean and Triphase Accelerator. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Pham, M. Yin, L. Jayaraman, R.S. Kerbel

Development of methodology: E. Pham, M. Yin, L. Jayaraman, E. Rohde, D. Lazarus, F.S. Foster

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Pham, M. Yin, C.G. Peters, C.R. Lee, S. Man, E. Rohde, R.S. Kerbel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Pham, M. Yin, C.G. Peters, L. Jayaraman, R.S. Kerbel

Writing, review, and/or revision of the manuscript: E. Pham, M. Yin, C.G. Peters, P. Xu, L. Jayaraman, E. Rohde, D. Lazarus, S. Eliasof, F.S. Foster, R.S. Kerbel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Pham, C.R. Lee, S. Man, A. Chow

Study supervision: L. Jayaraman, S. Eliasof, F.S. Foster, R.S. Kerbel

Other (bioanalysis of study samples): D. Brown

The authors sincerely thank Dr. Marta Paez-Ribes for her assistance with animal surgeries and technical advice. The authors also thank Cassandra Cheng for outstanding administrative assistance and their student volunteer, Taylor VanVeen. Virtual slide scanning and image analysis consultation was provided by Taha Rashed at the Biomarker Imaging Research Laboratory (BIRL, Sunnybrook Research Institute) under the direction of Dr. Martin Yaffe. HCI-002 was generously provided by Dr. Alana Welm.

This work was supported by grants from the Canadian Institutes of Health Research (CIHR) and the Canadian Breast Cancer Foundation (R.S. Kerbel). Financial support was also provided by Cerulean Pharma Inc. R.S. Kerbel and F.S. Foster were recipients of Tier I Canada Research Chairs during the course of these studies. E. Pham is a recipient of a CIHR Post-Doctoral Fellowship Award.

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.