Real-Time Imaging Reveals Local, Transient Vascular Permeability, and Tumor Cell Intravasation Stimulated by TIE2^hi Macrophage–Derived VEGFA

For almost two decades, tumor vasculature has been described as abnormal with increased vascular permeability (1, 2). VEGFA is known to promote vascular permeability, and inhibition of VEGFA results in the normalization of tumor vasculature and a decrease in permeability (3, 4). Due to the significant effects of VEGFA on tumor angiogenesis and vascular permeability, inhibitors of VEGF signaling have become an important research focus in the development of antitumor therapies.

Tumor-associated macrophages (TAM) have been implicated in tumor progression, angiogenesis, and metastasis (5, 6). A subpopulation of perivascular TAMs that have features of protumorigenic macrophages, promoting tumor angiogenesis and metastasis, have been identified as TIE2-expressing macrophages (TEM; ref. 7). Perivascular macrophages are also an essential component of the microanatomical sites termed “Tumor MicroEnvironment of Metastasis” (TMEM) that consist of a TAM in direct contact with a mammalian-enabled (MENA) overexpressing tumor cell and endothelial cell (8). TMEMs have been associated with tumor cell intravasation (9, 10), and TMEM density predicts distant metastatic recurrence in breast cancer patients independently of other clinical prognostic indicators (8, 11). However, the mechanistic link between perivascular macrophages and tumor cell intravasation has remained unclear. Further, hyperpermeability in tumor vasculature is not uniform, but rather is spatially and temporally heterogeneous (12). In a VEGFA overexpression model inducing vascular permeability, the presence of macrophages at vascular branch points was observed at hotspots of vascular permeability (4). Although hyperpermeability of tumor vasculature is widely accepted, a mechanistic understanding of the heterogeneity of vascular permeability, the contribution of TAMs, and the link with tumor cell intravasation have not been described.

Here we show, using intravital high-resolution two-photon microscopy, that transient vascular permeability events are restricted to TMEM sites of TIE2^hi/VEGFA^hi perivascular macrophages. Local loss of vascular junctions at the TMEM results in transient vascular permeability and tumor cell intravasation in the spontaneous autochthonous mouse mammary cancer model where the mouse mammary tumor virus long terminal repeat drives the polyoma middle T antigen (MMTV-PyMT), the human patient–derived xenograft model TN1, and human metastatic breast cancer.

To examine the functional role of the TMEM in tumor cell dissemination, we used the spontaneous autochthonous mouse mammary cancer model where the mouse mammary tumor virus long terminal repeat drives the polyoma middle T antigen (MMTV-PyMT), in which tumors exhibit histology similar to human luminal breast cancer and progress to metastasis (13). IHC revealed that TMEM structures in mouse tumors have the same microanatomical structure as identified in humans (Supplementary Fig. S1A; ref. 11). TMEM density and circulating tumor cells (CTC) increase with tumor progression with elevated TMEM scores in late carcinoma as compared with early carcinoma as seen by IHC (Supplementary Fig. S1A–S1C), though total perivascular macrophage (including macrophages not associated with tumor cells) density is not significantly different (13). High-resolution imaging demonstrates that in TMEM structures tumor cells and macrophages extend protrusions but are relatively nonmigratory and stay in direct contact over time (Supplementary Fig. S1D).

To directly observe TMEM function in vivo, we used extended time-lapse intravital microscopy (IVM) with high spatial and temporal resolution. To visualize blood flow, vessels were labeled with a high–molecular weight compound (155-kDa dextran or quantum dots; refs. 1, 14; Figs. 1, 2, 3; and Supplementary Fig. S2). In PyMT late carcinoma, migratory tumor cells and macrophages stream toward TMEM at sites with vascular permeability, whereupon tumor cells undergo transendothelial migration at the TMEM (Fig. 1A–E; Supplementary Fig. S2A–S2E). In late carcinoma, transient, local blood vessel permeability was observed at TMEM sites by the extravasation of quantum dots (Supplementary Fig. S2A and S2B) or 155-kDa dextran–tetramethylrhodamine (TMR; Figs. 2A and B, 3C; Supplementary Fig. S2C–S2E; and Supplementary Movie S1). Further, tumor cell intravasation occurs at TMEM sites concurrently with transient permeability (Fig. 2A–H and Supplementary Fig. S2C–S2E). Transient vascular permeability at the TMEM is spatially and temporally heterogeneous (Supplementary Fig. S2F), with events of permeability and tumor cell intravasation at the TMEM occurring predominantly at vascular branch points (Supplementary Fig. S2G). Transendothelial crossing of tumor cells is visualized by the hourglass shape of tumor cells, as they are partially in the vessel lumen and partially in the tissue (Fig. 1C, 2A, C–E; Supplementary Fig. S2E). During transendothelial migration of tumor cells, the TMEM tumor cell and macrophage neither migrate nor intravasate, indicating that tumor cells entering the blood vessel at the TMEM are supplied by the migratory stream of cells (Fig. 1A, B, and D). The stationary phenotype of these cells is consistent with previous results showing macrophage contact–initiated invadopodium formation uniquely in the TMEM tumor cell (9) and that perivascular invadopodium-containing tumor cells are relatively nonmotile in vivo (15).

The peak of extravascular dextran intensity and the appearance of circulating tumor cells coincide temporally and spatially (Fig. 2A–E, H; Supplementary Fig. S2; Supplementary Movie S1), demonstrating a direct link between localized blood vessel permeability and tumor cell intravasation at the TMEM. The coincidence of spontaneous, transient vascular permeability with tumor cell intravasation at the TMEM also has been observed in a patient-derived xenograft model of triple-negative breast cancer, TN1 (Supplementary Fig. S3).

To confirm that the TMEM is associated with transient vascular permeability and tumor cell intravasation, a 100-μm window, the approximate width of a TMEM site, was consecutively slid along all blood vessels (window measurement) to quantify the frequency of tumor cell intravasation and vascular permeability events in the presence or absence of the TMEM (Supplementary Fig. S4). Vascular permeability and tumor cell intravasation occur exclusively within the 100-μm window when it contains a TMEM, but never when the 100-μm window does not contain a TMEM in PyMT (Fig. 2F and G). Similar results were observed in the human TN1 model (Supplementary Fig. S3C and S3D), highlighting the importance of the TMEM in transient vascular permeability and tumor cell intravasation.

Tumor vasculature has been previously described as abnormal with increased permeability, which has been attributed to larger vascular intercellular openings (1, 12, 16). However, vascular permeability is not spatially or temporally uniform, with hotspots at vascular branch points (4, 12). Here, we demonstrate that vascular permeability is transient, occurs exclusively at TMEM sites, and is temporally heterogeneous, explaining the previously unresolved heterogeneity in vascular permeability (Fig. 2F and Supplementary Fig. S2E). Events of spontaneous, local vascular permeability and tumor cell intravasation at TMEM occur predominantly at vascular branch points, consistent with previous reports of vascular permeability (Supplementary Fig. S2G). If tumor blood vessels were uniformly leaky, high–molecular weight vascular probes would extravasate immediately and continuously after injection. Although the high–molecular weight probe, 155-kDa dextran–TMR, remains in the vasculature in the absence of transient TMEM-associated permeability events for the duration of the time-lapse imaging, a low-molecular weight dextran, 10-kDa dextran–FITC, which is below the molecular-weight cutoff size of the endothelium (1, 14), leaks from blood vessels and clears from the vascular space (Supplementary Fig. S5).

Further, transient permeability events are distinct from mechanical damage to the endothelium. After creating a 2-μm hole in the endothelium with a laser, 155-kDa dextran–TMR extravasates continuously, filling the field of view (Fig. 3A). In contrast, VEGFA-mediated permeability is transient (12). Intravenous injection of VEGFA₁₆₅, the soluble isoform of VEGFA with properties of native VEGF (17), results in vascular permeability with peak intensity of extravascular dextran at 20 minutes (Fig. 3B). Spontaneous vascular permeability at the TMEM follows similar kinetics to VEGFA₁₆₅-mediated permeability with peak intensity of extravascular dextran at 20 minutes, but is restricted to individual TMEM sites (Fig. 3C). The curves obtained for average intensity of extravascular 155-kDa dextran–TMR after laser damage, VEGFA₁₆₅, and spontaneous permeability were fit to an exponentially modified Gaussian function (Fig. 3D and E). Although the curve for laser damage does not have a clearance term as dextran continues to extravasate for the entire time lapse, both the VEGFA₁₆₅ and spontaneous curves have similar extravasation and clearance rates. A significant difference between VEGFA₁₆₅ and spontaneous TMEM-mediated permeability is that permeability at the TMEM is highly local, whereas VEGFA₁₆₅ results in dextran extravasation from all blood vessels within a field of view. Thus, the area of extravascular 155-kDa dextran–TMR from local TMEM-mediated permeability is markedly less than permeability from VEGFA₁₆₅ or laser-induced damage (Fig. 3F), further emphasizing the local nature of TMEM-mediated vascular permeability.

To determine if TMEM macrophages regulate vascular permeability and tumor cell intravasation, macrophages were depleted in the mammary tumor using the previously characterized mouse model, macrophage fas-induced apoptosis (MAFIA; refs. 18, 19), with orthotopic MMTV-PyMT tumor implants. Depletion of macrophages is systemic, including within the mammary tumor, thus resulting in a depletion of TAM and TMEM by 67% and 72%, respectively (Fig. 4A–C). When macrophages are depleted, extravascular dextran decreases, as does the number of circulating tumor cells (Fig. 4D–F). These data demonstrate that macrophages are essential for vascular permeability and tumor cell intravasation at the TMEM.

Because blood vessel permeability observed by IVM is restricted to TMEM, we examined whether vascular junction protein localization was altered in the absence of macrophages, reflecting a requirement for macrophage-dependent signaling events to induce vascular permeability. Staining for vascular junction proteins ZO-1 and VE-Cadherin increased in the tumor vasculature after depletion of macrophages in MAFIA mouse tumors (Fig. 4D, G, and H), indicating that macrophages are involved in vascular junction disassembly during vascular permeability events at the TMEM.

In PyMT mammary carcinoma, a subpopulation of TAMs has been identified as TIE2^hi perivascular macrophages (7, 20, 21). TEMs have been shown to upregulate the TIE2 tyrosine kinase receptor by 100-fold after recruitment to the tumor (22). TEMs have features of protumorigenic macrophages and promote tumor angiogenesis (7). TEMs are further characterized as MRC1⁺/CD11b⁺/F4/80⁺/CD11c^– and are associated with CD31⁺ tumor blood vessels (20). Thus, we sought to determine if TEMs are located in the TMEM. Immunofluorescence of the TMEM markers MENA (tumor cells), CD31 (endothelial cells), and CD68 (macrophages; Fig. 5A) compared with TIE2, VEGFA, and CD31 in sequential tissue sections demonstrates that TIE2^hi/VEGFA^hi macrophages are enriched in TMEM structures (Fig. 5B and Supplementary Fig. S6A–S6E). VEGFA is elevated in TIE2^hi macrophages, as compared with the adjacent endothelial cells and surrounding tumor tissue (Fig. 5C and D). Further, 100% of TIE2^hi/VEGFA^hi TMEM-associated macrophages express the TEM markers MRC1, CD11b, and F4/80 while lacking CD11c (Fig. 5E and F and Supplementary Fig. S6).

To investigate the importance of VEGFA in TMEM function, we blocked VEGFA binding to VEGFRs using a neutralizing antibody (B20-4.1.1), and found a decrease in extravascular dextran and circulating tumor cells (Fig. 6A–C). Binding of VEGFA to VEGR2 leads to junction disassembly (23). Vascular ZO-1 and VE-Cadherin immunostaining increased during VEGFA inhibition, suggesting an increase in integrity of endothelial adherens and tight junctions from reduced bioavailability of VEGFA, including VEGFA from the TMEM (Fig. 6A, D, and E).

To determine if the subpopulation of TIE2^hi/VEGFA^hi macrophages in the TMEM is an essential source of VEGFA in the tumor microenvironment required for transient vascular permeability at the TMEM and tumor cell intravasation, VEGFA was selectively ablated in monocytes and macrophages using the Vegfa^flox/flox;Csf1r-Mer-iCre-Mer transgenic mouse depletion model of Vegfa that targets myeloid cells expressing CSF1R, including both Ly6C^hi and Ly6C^lo populations, and the TEM population (24). Macrophage-specific depletion of VEGFA reduced transient vascular permeability and circulating tumor cells and restored vascular junctions (Fig. 6F–J). Immunofluorescence of sequential sections demonstrates that blood vessels adjacent to CD68⁺/TIE2^hi/VEGFA^hi TMEM macrophage have significantly reduced vascular VE-Cadherin/CD31 relative intensity compared with regions of vasculature away from TMEM sites in Vegfa^flox tumors (Fig. 7A and B). Further, when VEGFA has been ablated in Vegfa^flox;Csf1r-Cre tumors, VE-Cadherin/CD31 relative staining intensity is the same along the tumor vasculature as in regions away from the TMEM (Fig. 7C and D). Therefore, vascular junction integrity, as measured by VE-Cadherin/CD31 relative staining intensity, is significantly reduced only in regions of vasculature adjacent to VEGFA^hi TMEM macrophages in the TMEM (Fig. 7E). Further, pericyte coverage of the vasculature is reduced in regions of VEGFA^hi TEMs in the TMEM as compared with regions away from VEGFA^hi TMEM structures (Supplementary Fig. S6F–S6H). A decrease in pericyte coverage of vasculature has been correlated with increased metastasis and vascular permeability (25).

To establish the relevance of TIE2^hi/VEGFA^hi macrophages in TMEM structures in mediating vascular permeability and tumor cell dissemination in metastatic breast cancer, vascular junction staining was measured in human breast cancer patient samples. Staining of sequential sections demonstrates that blood vessels adjacent to TIE2^hi/VEGFA^hi macrophages in the TMEM have significantly reduced vascular VE-Cadherin fluorescence intensity compared with regions of vasculature away from the TMEM (Fig. 7F and G and Supplementary Fig. S7).

Together, these data establish that the TIE2^hi/VEGFA^hi TMEM macrophages interact with endothelial cells through VEGFA signaling to mediate local, transient blood vessel permeability, demonstrating the mechanism underlying the clinically demonstrated association of TMEM density with metastatic recurrence of breast cancer.

Although the abnormality and permeability of tumor vasculature have been well characterized, the mechanism leading to spatial and temporal heterogeneity in permeability has not been resolved. The use of high-resolution multiphoton microscopy has allowed for the study of vascular permeability and tumor cell dissemination in mammary carcinoma at unprecedented spatial and temporal resolution. Our data show that in the PyMT authochthonous mouse mammary carcinoma and human patient–derived xenograft TN1 models, vascular permeability is dynamic, localized, and restricted to the TMEM. These data are consistent with previous findings that hyperpermeability of tumor vasculature is heterogeneous and often in the presence of perivascular macrophages (4), but further explain the observed heterogeneity and that tumor cell intravasation occurs at sites of vascular permeability.

The sites of dynamic tumor vascular permeability have been identified at sites of VEGFA^hi perivascular macrophages at the TMEM. The clinical significance of TMEM density in predicting metastatic risk has been recently expanded to a large cohort of patients, further emphasizing the importance of the TMEM in breast cancer metastasis (11). These data demonstrate that TIE2^hi/VEGFA^hi perivascular macrophages in TMEM share the characteristics of the proangiogenic and prometastatic TEMs (7). Thus, we have been able to expand our understanding of the function of this subset of TAMs in the tumor microenvironment in promoting metastasis.

Mechanistically, macrophage/tumor cell streams migrate to TMEM sites through the EGFR/CSF1R paracrine loop (26). Elevated expression of VEGFA in the TIE2^hi TMEM macrophage results in transient permeability of tumor blood vessels proximal to the TMEM that occurs by disassembling endothelial cell junctions. The simultaneous attraction of migratory tumor cells and transient blood vessel permeability results in a concurrent spike in tumor cell intravasation with vascular permeability at TMEM sites (Fig. 7H). These data, together with the clinical association of the TMEM with distant metastatic tumor recurrence in human breast cancer patients, explain why TMEM density can predict metastasis and argue for the development of therapeutic approaches targeted against both TMEM formation and function.

All studies involving mice were carried out in accordance with the NIH regulation concerning the care and use of experimental animals and approved by the Albert Einstein College of Medicine Animal Care and Use Committee. PyMT (MMTV-PyMT) transgenic mice were bred in house. MAFIA mice [C57BL/6-Tg(CSF1R-EGFP-NGFR/FKBP1A/TNFRSF6)2Bck/J] were obtained from The Jackson Laboratory and were implanted with tumor pieces (2 mm × 2 mm) into the fourth mammary fat pad on the left side. For multiphoton microscopy, transgenic mice were generated to label the myeloid lineage and mammary tumor cells by crossing MacBlue mice [Csf1r–GAL4-VP16/UAS-enhanced cyan fluorescent protein (ECFP); ref. 27] in a C57BL/6 background with Tg(MMTV-iCre)-Tg(loxP-stop-loxP-PDendra2)jwp (28) mice of FVB background. The Friend Leukemia Virus, strain B (FVB) macrophage-specific (Csf1r promoter), tamoxifen-inducible Cre-expressing Tg(Csf1r-Mer-iCre)1jwp transgenic mice were crossed with Vegfa^flox/flox mice. Depletion of Vegfa in myeloid cells was induced by daily subcutaneous injection of 3 mg tamoxifen in corn oil per mouse for 2 days before injection of 155-kDa TMR–dextran (24). TN1 cells were isolated from a pleural effusion sample from a patient with triple-negative breast cancer (ER⁻/PR⁻/HER2⁻) and transduced to express GFP (29). TN1-GFP patient-derived cells were only passaged in vivo in NOD.SCID mice (The Jackson Laboratory) by orthotopic injection of 1 × 10⁶ dissociated tumor cells mixed with 50% Matrigel (BD Biosciences) into the fourth mammary fat pad on the left side of the mouse. No cell lines were used in this study.

Multiphoton IVM was performed using a skin flap procedure as previously described (30) with a custom-built 2-laser multiphoton microscope (28). The animal was placed in a heated chamber maintained at physiologic temperature during the course of imaging and monitored using MouseOx (Starr Life Sciences). Three milligrams of 155-kDa TMR–dextran or 100 μL of 8 μmol/L Qdots 705 [obtained from A. Smith, University of Illinois Urbana-Champaign (UIUC) or Life Technologies Qdots ITK 705; refs. 31–33] was administered via a tail vein catheter.

Animals were administered 5 mg/kg B20-4.1.1 VEGFA-neutralizing antibody (Genentech) or antibody isotype control by intravenous injection 6 hours before termination of the experiment.

Animals were administered 10 mg/kg B/B homodimerizer (AP20187; Clontech) diluted in 4% ethanol, 10% PEG-400, and 1.7% Tween-20 or vehicle control by intraperitoneal injection daily for 5 days. Twenty-four hours after the last injection of B/B homodimerizer, animals were administered 3 mg of 155-kDa dextran–TMR for 1 hour and 100 μL of anti-CD31 antibody for 10 minutes.

One hour before the termination of the experiments with inhibitors in MAFIA mice, 3 mg of 155-kDa TMR–dextran was administered by i.v. tail vein to label sites of vascular permeability. In tumor tissue, several transient permeability events may occur at any given time due to the spatial and temporal heterogeneity of vascular permeability, thus quantitation of extravascular dextran over the course of 1 hour will capture these dynamic events. Anti-mouse CD31-biotin was administered by i.v. tail vein for 10 minutes to label flowing blood vessels. Tumors were fixed in 4% paraformaldehyde and cryoprotected in 30% sucrose in PBS before freezing in optimal cutting temperature. Sections (5 μm) of tumors were cut, and immunofluorescence was performed.

CTCs were isolated from anesthetized mice from blood drawn from the right ventricle of the heart. Blood burden experiments obtained by cardiac puncture were used for endpoint experiments to capture all of the CTCs in the animal blood at the experiment endpoint. Red blood cells were lysed using RBC lysis buffer (multispecies; eBioscience) before the cells were placed in culture with Dulbecco's Modified Eagle Medium:Nutrient Mixture F-12 (DMEM/F-12) with 20% FBS. Adherent tumor cells were counted at a time of no tumor cell growth, which is a count of CTCs. Tumor cells were identified by fluorescence microscopy as Dendra2- or CFP-expressing tumor cells as previously described (34, 35). CTCs were also scored by quantifying tumor cells in blood vessels adjacent to the TMEM as seen in Fig. 2 and Supplementary Fig. S4 in live animals. This method is described under IVM image analysis in Supplementary Methods.

Tumor sections were fixed and permeabilized with cold acetone, washed with PBS, and blocked with block solution (1% BSA, 10% FBS, and 0.0025% fish skin gelatin in PBS/Tween). The following primary antibodies were used: rat anti-mouse CD68 (1:200, clone FA-11; Serotec), AlexaFluor647-conjugated CD68 (1:100, eBioscience), mouse anti-MENA (1:400, from F. Gertler), rat anti-ZO-1 (1:100, clone R40.76; Millipore), goat anti–VE-Cadherin (1:200, clone C-19), rabbit anti-VEGFA (1:200, clone A-20; Santa Cruz Biotechnology), hamster anti-CD11c (1:100, clone HL-3; BD Bioscience), rat anti-CD11b (1:200, clone ICRF4; BD Biosystems), rat anti-F4/80 (1:50, clone BM8; eBioscience), goat anti-MRC1/CD206 (1:200, R&D Systems), rat anti-TIE2 (1:50, clone TEK4; eBioscience), and rabbit anti-NG2 (Millipore). Sections were washed with PBST, and the primary antibodies were detected with AlexaFluor488, 555, or 647 secondary antibody conjugates (Molecular Probes/Invitrogen) and nuclei stained with 4,6-diamidino-2-phenylindole (DAPI). All fluorescently labeled samples were mounted with Prolong Gold antifade reagent (Molecular Probes/Invitrogen) and analyzed with a compound fluorescent microscope (Zeiss Axio Observer; 40× objective with numerical aperture 1.3). Images of tumor sections were acquired using mosaic tiling of 40× images in AxioRel version 4.8. All images were acquired as 16-bit TIFF images, and all quantitative analyses were performed on the raw 16-bit TIFF images in ImageJ.

Mouse tumor sections were fixed overnight in 10% neutral buffered saline prior to embedding in paraffin. Tumor sections (5 μm) were deparaffinized and stained for hematoxylin and eosin (H&E) or TMEM using anti–IBA-1 (macrophages), anti-endomucin (blood vessels), and anti-MENA (tumor cells; ref. 36), and the TMEM was quantified as previously described (8).

Formalin-fixed paraffin-embedded patient tissue from 5 invasive ductal carcinomas was collected under the Montefiore-Einstein Institutional Review Board approval. Paraffin-embedded human breast cancer tumors were cut into 5-μm sections, deparaffinized, and stained for H&E or TMEM. The sequence was anti-CD31 (clone JC70A; DAKO) and Vector Blue chromogen (for endothelial cells); anti CD-68 (clone PG-M1; DAKO) with and 3′,3′-diaminobenzidine (DAB) chromogen (for macrophages); and anti–pan-MENA with Fast Red chromogen (for carcinoma cells; refs. 8, 11). Sequential sections were cut for tyramide signal amplification (TSA) for quantitative immunofluorescence using the Opal 3-plex Kit (Perkin Elmer) according to the manufacturer's directions. The sequence was rabbit anti-VEGF (1:2,000; Rb 9031-P0-A; Thermo) with TSA Plus Cy3; rabbit anti-TIE2 (1:3,000; clone C-20; Santa Cruz Biotechnology) with TSA Plus Cy5; goat anti–VE-cadherin (1:200; clone C-19; Santa Cruz Biotechnology) with TSA Plus fluorescein and nuclei stained with DAPI. All quantitative analyses were performed on the raw 16-bit TIFF images, and images of TMEM were validated independently by a pathologist.

Vascular leakage is composed of two competing processes: an increase in extravascular signal due to leakage from the vasculature and a diffusive clearance. The Exponentially Modified Gaussian asymetic function (37) is composed of the product of a sigmoidal error function with an exponential decay. Least squares curve fitting was performed utilizing a nonlinear Generalized Reduced Gradient (GRG₂) solver and values were directly compared (38).

Individual animals are presented as individual points, a horizontal line indicates the mean, and error bars represent the SEM. One-way or two-way ANOVA analysis was performed for data sets with more than two groups to determine significance. Statistical significance was determined by the comparison of the means of two groups using an unpaired, two-sided t test using Prism (Graph Pad Inc.). Data sets were checked for normality (D'Agostino & Pearson omnibus normality test or Shapiro–Wilk normality test) and unequal variance using Prism (Graph Pad Inc.). Welch's correction was applied to t tests as needed. P values of less than 0.05 were deemed significant.

A.S. Harney reports receiving commercial research support from Deciphera Pharmaceuticals, LLC. J.G. Jones has ownership interest (including patents) in MetaStat and is a consultant/advisory board member for the same. J.S. Condeelis reports receiving a commercial research grant from Deciphera; has ownership interest (including patents) in MetaStat; and is a consultant/advisory board member for Deciphera and MetaStat. No potential conflicts of interest were disclosed by the other authors.

Views and opinions of, and endorsements by, the authors do not reflect those of the U.S. Army or the Department of Defense.

Conception and design: A.S. Harney, D. Entenberg, Y. Wang, J.W. Pollard, J.S. Condeelis

Development of methodology: A.S. Harney, D. Entenberg, Y. Wang, J.W. Pollard, J.S. Condeelis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.S. Harney, E.N. Arwert, D. Entenberg, Y. Wang, B.-Z. Qian, M.H. Oktay, J.W. Pollard, J.G. Jones, J.S. Condeelis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.S. Harney, D. Entenberg, P. Guo, M.H. Oktay, J.W. Pollard, J.G. Jones, J.S. Condeelis

Writing, review, and/or revision of the manuscript: A.S. Harney, E.N. Arwert, D. Entenberg, J.W. Pollard, J.G. Jones, J.S. Condeelis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.S. Harney, D. Entenberg, Y. Wang

Study supervision: J.S. Condeelis

Other (performed experiments and analyzed data): J.G. Jones

The authors thank H. Guziak at the Analytical Imaging Facility of the Albert Einstein College of Medicine for technical assistance. They also thank A. Smith (UIUC) for Quantum dots and F. Gerlter (MIT) for the gift of the pan-MENA antibody. Finally, they thank Jiufeng Li for assistance in generating the MMTV-Dendra mouse.

This research was supported by the Department of Defense Breast Cancer Research Program under award number W81XWH-13-1-0010 (to A.S. Harney), the Wellcome Trust (WT096084MA, to E.N. Arwert), NIH CA100324, NIH CA179507, the Integrated Imaging Program, and the NCI cancer center support grant (P30CA013330).

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.