MULTIMODAL IMAGING OF THE TUMOR MICROENVIRONMENT AND BIOLOGICAL RESPONSES TO IMMUNE THERAPY

Abstract

Beyond heterogeneous cancer cells, the tumor microenvironment includes stromal and immune cells, blood vessels, extracellular matrix and biologically active molecules. Abnormal signaling, uncontrolled proliferation and high interstitial pressure all contribute to a chaotic, non-hierarchical vascular organization. Using an immune competent 4T1 breast adenocarcinoma murine model, this study fully characterizes the architecture and immunocyte milieu of the tumor microenvironment. Heterogeneous vessel distribution, chaotic connectivity, limited perfusion, cancer cell density, immune phenotype, and biological responses to immune therapy are presented. Cancer cell density mirrored the distribution of large, perfusable vessels, both predominately in the tumor periphery. Intratumoral administration of the proinflammatory cytokine IL-12 led to an increase in CD45⁺ leukocytes, with a specific increase in CD4⁺ and CD8⁺ T cells, and a decrease in the percentage of Gr-l^lo myeloid-derived suppressor cells. Concomitantly, serum G-CSF and VEGF decreased, while CXCR9 and interferon gamma increased. The distribution pattern of infiltrating monocytes/macrophages, visualized using a fluorescent perfluorocarbon emulsion, indicated that macrophages predominately localize in the vicinity of large blood vessels. Electron microscopy supports the presence of dense tumor cell masses throughout the tumor, with the largest vessels present in the surrounding mammary fat pad. Overall, the large vessels in the 4T1 tumor periphery support high vascular perfusion and myeloid accumulation. The pro-inflammatory cytokine IL-12 stimulated a transition towards T helper 1 cytokines within the tumor and in serum, supporting suppression of tumor growth and angiostatic conditions.

INTRODUCTION

The goal of cancer immunotherapy is to boost or restore immune function for effective recognition of tumor-associated antigens by antigen presenting cells (APC), activation of effector cells, and blockade of negative regulators that inhibit active effector cell function. Recent clinical trials support paradigm-changing successes in treatment outcomes using checkpoint inhibitors (CPI) for alleviating immune suppression. CPI therapy has resulted in remarkable antitumor responses in some patients, unfortunately, others fail to benefit and often suffer from side effects that range from mild to life threatening.[1, 2] Adverse events include inflammation and autoimmunity, with CPI antibodies capable of inducing high levels of inflammatory cytokines (i.e. cytokine release syndrome (CRS).[3] CRS occurs in response to cytotoxic damage of immune cells, such as monocytes, macrophages and lymphocytes, with IL-6 believed to play a key role in the pathophysiology [4]. Methods to monitor responses to immunotherapy and identify cancer phenotypes that respond favorably to immune therapy or with immune-related adverse events (IRAEs) are desperately needed.

Recently, molecular and pathologic features of the tumor microenvironment have been shown to contribute to tumor heterogeneity and variability in therapeutic responses.[5] Tumor stromal cells, which include myofibroblasts and endothelial cells, influence tumor growth by providing structural support and cellular contiguity, and via secretion of cytokines and growth factors.[6] Fibrogenic factors, such as vascular endothelial growth factor (VEGF), regulate angiogenesis and alter the tumor microenvironment impacting therapeutic responses. Other contributions to cancer progression and dissemination include cancer desmoplasia. Desmoplasia, characterized by increased proliferation of alpha-smooth muscle actin-positive fibroblasts and deposition of distinct extracellular matrix (ECM) components, creates unique tissue heterogeneity.[7] The unique tumor architecture also includes heterogeneous vessel distribution, chaotic connectivity, limited perfusion and regions of hypovascularity; all influencing drug distribution and therapeutic responses.

To evaluate the ability to characterize the tumor microenvironment and monitor biological responses to therapy, we have employed computed tomography, bioluminescent imaging (IVIS), magnetic resonance imaging (MRI), tissue scanning electron microscopy, immunohistochemistry, flow cytometry and Luminex technology using an immune competent 4T1 model of murine breast adenocarcinoma. Biological responses to the proinflammatory cytokine IL-12 were monitored with respect to vascular traits and immunocyte and cytokine profiles.

METHODS

Animal models

Athymic nude and Balb/c mice (6-8 weeks old) were obtained from Charles River Laboratories, Inc. (Wilmington, MA). Breast tumors were established by injection of 1x10⁵ 4T1-luc2-td Tomato Bioware® Ultra Red (Caliper Life Sciences, Hopkinton, MA, USA) mouse mammary cancer cells into either the dorsolumbar or inguinal mammary fat pad. For intratumoral injections, 5 μg of recombinant mouse IL-12 (R&D Systems Inc., Minneapolis, M, USA) or 50 μl PBS vehicle control (n=3) were administered to female BALB/c mice when tumors were approximately 750 mm³. Injections were performed 24 h after intravenous injection of V-sense and 24 h prior to MR imaging as described in the succeeding section on MR imaging. Bioluminesence data was acquired using the Xenogen IVIS System following i.p. injection of 150 mg/kg RediJect D-Luciferin. All procedures were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee at Houston Methodist Research Institute, Baylor College of Medicine, and the University of New Mexico according to the NIH Guide for the Care and Use of Laboratory Animals.

Immunohistochemistry

Excised tumors were fixed in 10% buffered formalin and histology sections, 4 μm, were deparaffinized with xylene and ethanol and prepared for staining using heat-mediated antigen retrieval and blocking with 10% goat serum. Sections were stained with anti-CD31 rabbit polyclonal antibody (ab28364) purchased from Abcam (Cambridge, MA) at a 1/50 dilution (1 μg/ml). Secondary enzyme-linked antibody and chromogenic staining were used to visualize antibody-epitope interactions. Images were acquired using a 10x objective and stitching performed to create whole tissue section montages.

Vital microscopy

3D tumor vasculature imaging was performed in nude mice with 4T1 breast tumors using a Nikon confocal vital microscope. Mice were anesthetized using 2-3% isoflurane and 4T1 tumors, approximately 500-750 mm³ in size, were exposed by a small midline incision, the fascia between the skin and muscle was disrupted using a cotton swab, and an inverted skin flap was elevated using rolled cotton guaze. The tumor was moistened with saline and scanned at a video-rate of 30 fps using a Nikon A1R Multiphoton-ready laser scanning confocal microscope equipped with a resonance scanner, CCD color camera (Nikon DS-F1), motorized stage (Prior Scinetific ZDeck), Nikon long-working distance air plan-apochromat ojective (4x, NA 0.2, WD 2 mm), an isoflurane (Aerrane; Baxter Healthcare) vaporizer system (E-Z Systems), and image acquisition software (Nikon NIS Elements 4.0). Field of views were imaged using a frame rate of 30 fps at 250 x 250 μm every 1-5 minutes FITC Dextran (Life Technologies, Grand Island, NY) was administered by retro-orbital injection of 10 mg/kg in 50 μl PBS and imaged for 60 minutes using an excitation wavelengths of 640 nm. The tumor was imaged at 561 nm and data was collected using band-pass filter widths of 30-50 nm centered at 488 nm for FITC and 579 for mCherry. 512 x 256 bit two-channel images were acquired with a pinhole of 1.0 Airy unit.

Fluoroescent and brightfield microscopy

RAW cells were seeded onto glass cover slips in 6-well plates at a density of 5 x 10⁵ cells per well in media containing 10 ng/mL IFN-γ and 200 ng/mL LPS. After 24 hours incubation, V-Sense (360 mg/ml stock diluted to 45 μg/m1) was added and cells were incubated for an additional 24 h. During the final 30 min of incubation, CellTracker Green (Invitrogen) was adding to the cell culture media according to the manufacturer’s instructions. After incubation, cells were washed with PBS, fixed with 4% paraformaldehyde and mounted on glass slides using Vectashield with DAPI (Vector Laboratories, Burlingame, CA, USA). Confocal images were acquired with a 63X/1.4NA oil objective in sequential scanning mode using a Leica TCS SP8 confocal microscope. Three dimensional cell images were isosurface rendered using the Leica Application Suite Advanced Fluorescence 3D analysis software.

Tissue scanning electron microscopy

Fixed tumors were sectioned using a razor blade, washed twice with 0.1% sodium cacodylate and incubated in tannic acid overnight. Tissue was then washed, post-fixed in osmium tetroxide for 30 min, washed twice, and then dehydrated in ascending concentrations of ethanol in water (30, 50, 70, 90, 95, and 100%; 15 min each), followed by 50% ethanol:50% t-butanol for 15 min, and then a final incubation in 100% t-butanol for 30 min. Tissues were dried overnight in the fume hood and sputter-coated with gold-palladium (5 nm) using a Leica ACE600 sputter-coater. Secondary electron images were acquired using an FEI Qunata 3D FEG Dual Beam (FEI, Hillsboro, OR, USA).

In Vivo magnetic resonance imaging (MRI)

For in vivo ¹⁹F imaging, V-Sense, a commercially available perfluoropolyether (PFPE) emulsion (VS-1000H, Celsense, Pittsburg, PA, USA), was injected 48 h prior to imaging. Each mouse was administered a single, 150 μL dose of V-Sense via tail vein injection. According to the manufacturer, V-Sense is a specially formulated, non-toxic emulsion containing 30% v/v PFPE with a mean droplet diameter between 135 and 145 nm.

Magnetic resonance imaging was performed on a 9.4 T, 20-cm horizontal-bore Bruker BioSpec /AVANCE III micro-imaging system driven by ParaVision 5.1 (Bruker Biospin, Billerica, MA, USA). A 40-mm, dual-tunable birdcage volume resonator was used to acquire both ¹H and ¹⁹F images at approximately 400 and 376 MHz, respectively. Mice were anesthetized using 1.5% isoflurane and 1.5 L/min O₂ inhalation gas. Animal temperatures were kept at 37°C, and respiratory gating was employed to minimize motion artifacts using an MR-compatible small rodent heater and gating system (SA Instruments, Inc., Stony Brook, NY, USA). To obtain high-resolution ¹H images, a 2D rapid acquisition with relaxation enhancement (RARE) sequence was used with a repetition time (TR) of 2500 ms, an echo time (TE) of 11 ms, and a RARE factor of 2. The number of excitations (NEX) was set to 1, field of view (FOV) to 6.00 cm x 6.00 cm, slice thickness to 2.00 mm, and acquisition matrix sized at 256 x 256, giving an in-plane spatial resolution of 0.234 mm x 0.234 mm. ¹⁹F images were obtained similarly with a 2D RARE sequence but with a TR/TE of 3500 ms/11 ms and RARE factor of 20, resulting in an effective echo time of 77.0 ms. NEX was also increased to 64. FOV and slice thickness were maintained as they were for ¹H imaging at 6.00 cm x 6.00 cm and 2.00 mm, respectively, while the acquisition matrix was resized to 128 x 128, giving an in-plane spatial resolution of 0.469 mm x 0.469 mm. ¹⁹F images were rendered in pseudo-color (“hotiron”) and then co-registered and fused with ¹H images using OsiriX Imaging Software (Pixmeo, Bernex, Switzerland). The windowing and degree of fusion between the ¹H and ¹⁹F images were adjusted to accentuate the locations of ¹⁹F signal.

For dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), we used a 3D GRE sequence with a TR = 6.0 ms; TE = 1.931 ms; Eff spectral bandwidth = 98684.2 Hz; FOV = 3.0 cm x 3.0 cm x 2.5 cm; and matrix size = 128 x 128 x 16. The resulting voxel resolution = 234 microns x 234 microns x 1.562 mm. The scan time per measurement was 9 seconds and with a 175 measurements acquired for a total imaging time of 30 minutes.

Micro Computed Tomography (CT)

Microinjection Technique.

Nude mice were anesthetized and placed in the supine position. A 1 cm skin incision was made in the inguinal region on the side opposite to the tumor and the femoral vascular bundle was exposed under a dissecting microscope. A 30G 1 inch needle was connected by an extension tube to a 1 ml syringe and a hub of the needle was fixed on a holder to a micromanipulator (MN-153, Narishige International, NY, USA). The needle was positioned parallel to the femoral vein. The needle was inserted into the lumen by screw handle of the micromanipulator. The needle and distal portion the vein were secured by 10-0 nylon to prevent leakage from the punch hole. A silicone rubber-lead chromate compound (Microfil, Flow Tech Inc., Carver, MA) was injected with manual pulsating of a 1 ml syringe. The mouse was dead soon after the injection however the compound was perfused to the entire tissue during heartbeat. Resistance in the syringe signaled the end of the injection and volume of the injected compound was 0.1 ml/g per mouse on average. The compound was red or yellow and therefore status of the perfusion to the vessels could be evaluated through the skin.

Micro-CT imaging.

High-resolution micro-computed tomography was performed using a Siemens Inveon Preclinical Multimodel SPET/PET/CT System, and the Inveon acquisition workplace (IAW) software (Siemens, Erlangen, Germany). Tumor tissues were imaged with resolution of 38um effective pixel size. The voltage was 80 kV and the current was 500 μA. A filtered back projection algorithm was employed for reconstruction. Imaging was viewed and analyzed using Inveon Research workplace (IRW) software (Siemens, Erlangen, Germany).

Alternatively, mice and tumor were imaged using the Explore Locus SP pre-clinical Specimen Scanner (GE Medical Systems, London Ontario), a specimen-dedicated cone-beam volume CT system which uses a tungsten source X-ray tube operating at 80 kV and 80 microamperes. In this scanner, objects are rotated in 1.0-degree increments on a holder between the X-Ray source and CCD-based detector. Typical scans generate eight to 28 mm isotropic voxels and require two to six hours for data acquisition. The scanner is designed to non-destructively acquire high-resolution 3D images of ex vivo specimens, ranging in size from 1 to 30 mm in diameter.

Amira 3D image analysis

Amira 5.6.0 by FEI software was used to analyze the tumor vasculature network. Using DICOM formatted data, image stacks were imported into the program with a voxel size of 0.15mm x 0.15mm x 0.15mm. The data was then analyzed by using the function ‘OrthoSlice’. This function is a tool for visualizing scalar data fields on a uniform Cartesian grid. This allows us to view the 2D CT images from the XY, XZ, and YZ planes as well as create a backdrop for our eventual three-dimensional reconstruction model. To be able to visualize the tumor data, a contrast must be created, and to do so the color map grayscale range was set from −146 to 760.

The data was then manipulated using the proprietary algorithms Isosurface, Volren, and Voltex. Isosurface is a module that creates a three-dimensional (3D) reconstruction of a scalar field with regular Cartesian coordinates using polygons. The draw style used was shaded, the color map was set from 628 to 695, and the threshold was set to 726.259. The Volren module is another method for creating a 3D reconstruction of a scalar field by determining the amount of light emitted and absorbed at each point in a data volume. To accomplish this reconstruction, the color map was set at 596 to 1065, alpha scaled at 1, the VRT, DDR, and MIP mode were used in three separate reconstructions. VRT visualizes based on classic texture based methods, DDR simulates a radiograph display from arbitrary views, and MIP visualizes based on the highest or lowest intensity in the data volume. The VRT mode shading was set to Specular, and light angled from top left. DDR mode had gamma set to 7.94118. The Voltex module is similar to the Volren module based upon how the 3D reconstruction images are generated. However, the Voltex module also allows for texture mapping techniques to be applied to the images. For Voltex we used the MIP option, alpha table lookup, and a color map from 596 to 1065. The alpha was scaled to 1, texture mode set at 3D, and the number of slices created maxed out at 512. We also created Voltex reconstructions using the color table option, using both luminance and rgba look up tables. The color map was set from 628 to 695, alpha scaled to 1, texture mode set at 3D, and slices maxed out to 512. If the original CT data contained more noise than desired, the data could be filtered using an Edge-Preserved Smoothing filter or Gaussian filter before performing any manipulations or reconstructions to the CT data.

RESULTS AND DISCUSSION

Vital microscopy imaging of macro tumor vasculature

Confocal microscopy was used to confirm homogeneous labeling of 4T1-luc2-td Tomato Red tumor cells in culture and in established tumors in female BALB/c mice (Figure 1a). Temporal intravital confocal imaging of tumor tissue (red) pre and post intravenous administration of FITC dextran (yellow) revealed a large perfusable network of vessels at the tumor periphery (Figure 1b). Spatial 3D imaging of a tumor section, achieved by sequential z-plane imaging post intravenous injection with Alexa Fluor 647 albumin (green), revealed dense peripheral vasculature in the 3D reconstructed image at the bottom image in Figure 1b.

Figure 1. Analysis of 4T1 tumor vasculature by microscopy.

a) Phase contrast and fluorescent micrographs of 4T1-luc2-td Tomato Red tumor cells in culture (left) and a micrograph showing merged td Tomato Red and DAPI signals in a BALB/c tumor (right). b) Intravital micrograph series showing first pass filling of blood vessels with FITC dextran and a lower magnification image to the right that highlights the peripheral location of large vessels. The bottom image is a 3D confocal micrograph of a unique 4T1 tumor showing similar peripheral localization of large vessels (green; AF647 albumin) in a td Tomato red tumor. c) Stitched micrographs of CD31 antibody-visualized vasculature in a tumor section. d) Green and pink masks highlight the location of vasculature and necrotic regions, respectively. Concentric rings were created to map vessel density across a tumor section, with the boxed region in the upper right image enlarged below to illustrate numbering in relation to location. Vessel density and percent vessels per region are shown graphically to the right.

Wide-field microscopy imaging of micro tumor vasculature

Histological imaging of vessels is the gold standard for ultra-high resolution imaging. The density and location of microvasculature in tumor sections was studied using immunohistochemistry (IHC). Excised tumors were sectioned and stained with anti-CD31 antibody and chromogenic staining. Images were acquired from labeled sections using the ImageXpress® Micro XL (Molecular Devices, Sunnyvale, CA, USA) optical imaging system equipped with RGB filters and a 10x objective. Micrographs were stitched together (Figure 1c). and color masks (Figure 1d) created to enable quantification of vessels (green) located in concentric rings across the tumor. Analysis of vessel density and fractional vascular area supported an increase in small vessels with increases in tumor depth. Necrotic regions (pink) were seen across the tumor including regions near the tumor periphery.

Computed tomographic imaging of macro tumor vasulature.

While vital microscopy is limited to imaging vessels at the surface of the tumor because of limited penetration depth, computed tomography (CT) imaging enables imaging the vasculature across the entire tumor. Nude mice with established 4T1 tumors were perfused with the contrast agent microfilm (yellow) via injection into the femoral vein, located contralateral to the tumor. Post administration, the yellow-colored vasculature could be visualized in the intact mouse, with the tumor region magnified in the inset (Figure 2a). Saggital, axial and coronal CT images of the perfused mouse are shown in Figure 2b, with the tumor area circled in each image. Post whole animal imaging, tumors were excised for secondary CT imaging. A reconstructed 3D image of the excised tumor is shown in the bottom right image. Clear omission of large vessels in the central region of the tumor supports that the larger vasculature is restricted to the tumor periphery. Two additional mice with tumors in the left inguinal fat pad were perfused and imaged. A 3D micro CT rendered image on one of the mice is shown in Figure 2c. 3D lateral and ventral views of excised tumors are shown in Figure 2d. All tumors showed similar peripheral localization of large vasculature.

Figure 2. Computed tomography (CT) imaging of tumor vasculature.

a) Nude mouse perfused with yellow microfil via the inguinal vein The tumor region is magnified in the inset. b) Saggital, axial, and coronal CT images of a perfused mouse, with tumors highlighted by circled regions. A processed CT image of the excised tumor is shown at the bottom right. c) Rendered CT image of a mouse. d) Processed CT images of excised tumors from two mice.

Analysis of the tumor vasculature

A reconstructed micro CT image of the tumor is shown in Figure 3a. The algorithm ‘Filament Editor’ in Amira was used to isolate the vasculature network from the background tumor information. The tracing algorithm was used to isolate structures with a bright on dark filament type and the filaments were used to reconstruct the vasculature network (shown overlaying the tumor in Figure 3a). Next, the display model ‘SpatialGraphView’ was used to visualize the filament segments as tubes that displayed the length and approximated thickness of each isolated filaments. Vasculature nodes are shown in Figure 3b and c, with node type differentiated by color in Figure 3c. 3D images of the reconstructed tumor vasculature are shown at increasing magnification in Figure 3d. For reference, a dimensionless scale bar was attached to the images.

Figure 3. 3D reconstruction of tumor vasculature.

a) A processed 3D image of 4T1 tumor vasculature derived from CT data slices is shown in the upper left merged with 3D projection image created using Amira software. The Amira OrthoSlice function was used to visual scalar data on a uniform Cartesian grid. Isosurface, Volren and Voltex algorithms were used to generate 3D images that highlight the location of vascular vessel length/diameter (b) and nodes (c). d) Scaled 3D vasculature images are shown at 4 magnification levels. Data derived from the vascular data were graphed to show vessel dimensions across the tumor (e), node types (f), and vessel directionality (g).

The final 3D isolated vasculature network was then evaluated for length, radius and volume using the ‘3D Length’ and ‘3D Angle’ functions (in voxels). A graph containing each element across the tumor is shown in Figure 3e. Vessels with the longest diameter are clustered near the central region of the tumor. Radius ranged from 0.075 to 0.52, with 198 of the 512 vessels (39%) having diameters of 0.075. Starosolski et al. reported clear CT delineation of cerebral blood vessels as small 40 μm.[8] A pie chart of node types is displayed in Figure 3f, with most nodes being branching nodes (257) and 38 terminal nodes. Vessel orientation is shown in Figure 3g. The large diversity in vessel orientation supports reports of chaotic vasculature in tumor.[9–11] Grabowska-Derlatka et al.[12] used multi-row detector computed tomography to demonstrate a correlation between tumor vasculature parameters, including chaotic configuration of vessels, and differentiation between borderline ovarian tumors and ovarian cancer.

A representative H&E stained 4T1 tumor section was imaged using an Olympus microscope equipped with a 10x objective. Images were stitched to show cellular features across the entire tumor sections (Figure 4, top left). High amounts of leukastasis and neutrophilic accumulation were present across tumors, with large areas of necrosis. High-resolution scanning electron microscopy images of tumor sections showed regions supported the existence of large blood vessels in the mammary fat pad and regions of high cellularity and immune cell infiltration.

Figure 4. Tumor landscape.

a) Representative H&E stained tumor from a mouse bearing 4T1 tdTomato Red Luc tumor was imaged using an Olympus microscope equipped with a 10x objective. Images were stitched to show cellular features across entire tumor section. b) High resolution scanning electron micrographs of 4T1 tumor sections showing features including surrounding adipose tissue and areas of high cellularity and neutrophillic infiltration.

Beyond the chaotic vasculature, structural and functional abnormalities exist in tumor blood vessels. Regions exist with intermittent stagnation and high-flow,[10] and variations exist in vascular permeability, extracellular space and perfusion. Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) was used to track contrast agent (Gd) from the arteries to the microvasculature and into the extracellular space in 4T1 tumors. The contrast agent shortens the relaxation time of the tissue in which it is found, thus producing a hyper-enhanced signal in T₁-weighted images. Figure 5a shows a tumor pre and post intravenous injection of Gd. The majority of the enhanced positive contrast (white, indicated by the arrow) was seen at the tumor periphery, indicating that the vessels with the highest perfusion are those at the tumor periphery where we previously showed the largest vessels exist. Springer et al.[13] showed promising DCE-MRI results in the discrimination of regional differences within the tumor that may be related to therapeutic responses and/or changes in metabolic alterations. 3D Bioluminescence imaging revealed of a mouse with a 4T1 tdTomato Red Luc tumor showed the highest localization of tumor cells at the tumor periphery (Figure 5b)

Figure 5. Immunocyte milieu of the tumor microenvironment.

a) MR images of a mouse with a 4T1 td Tomato red tumor pre and post gadolinium injection. The arrow indicates the tumor location. b) IVIS bioluminescence imaging of a 4T1 tumor. c) 1H, 19F and fused coronal NMR slices showing fluorescent v-sense (48 h; intravenous) in control and IL-12 treated (24 h; intratumoral) mouse tumors. d) 3D isosurface rendered and original (bottom) fluorescent micrographs of CellTracker green treated RAW macrophages 24 h after incubation with v-sense (red). e) Representative flow cytometry gating and antibody panels used to quantitate myeloid and T cell populations (top). Myeloid and T cell populations in control and IL-12 (intratumoral) treated 4T1 tumors from mice 24 h after drug administration (lower). f) Comparison of serum cytokine levels in control and IL-12 treated (24 h, intratumoral) mice.

Imaging the tumor immune milieu

Macrophage accumulation has been documented to contribute to cancer progression through suppression of the immune response and promotion of tumor vascularization,[14], however, macrophages are also vitally important for immune surveillance, phagocytosis of foreign invaders, and antigen presentation. V-sense is a perfluorocarbon emulsion that is taken up by monocytes and macrophages, enabling in vivo detection using ¹⁹F MRI. Here we used V-sense to track migration of macrophages into 4T1 tumors following intratumoral delivery of IL-12. Proton and 19F fluorine signals in the tumor are shown independently and fused in Figure 5c. The tumor region is shown inclusive of other regions of the mouse body and magnified for a no treatment control mouse and at high magnification for an IL-12 treated mouse. The majority of the myeloid cells internalizing V-sense were located in the vicinity of the large tumor vessels.

To support of uptake of V-sense by myeloid cells, we treated activated murine RAW macrophage-like cells with V-sense and found high levels of internalization 24 h post introduction. Surface rendered 3D confocal images in Figure 5d show a RAW cell with internalized V-sense (red) and CellTracker Green (blue is DAPI). Phalloidin staining (white) is shown in the top image only to delineate the cell boundaries.

IL-12, predominately produced by macrophages and dendritic cells, stimulates proliferation and activation of cytotoxic CD8⁺ lymphocytes and natural killer (NK) cells, leading to the production of interferon gamma (IFN-γ), and stimulating antigen-specific and nonspecific immune responses. Here we used flow cytometry to phenotype cells in the tumor of control and IL-12 treated mice. IL-12 decreased Gr-1^lo myeloid derived suppressor cells (MDSC) and increased CD4⁺ and CD8⁺ T cells in the tumor 24 h post treatment (Figure 5e). As expected, IL-12 increased serum levels of IFN-γ and increased the chemokine CXCR9 (Figure 4f), a T cell chemoattractant induced by IFN-γ that recruits tumor-suppressive T and natural killer cells.[15] Conversely, serum levels of G-CSF, IL-10 and VEGF were reduced by IL-12 treatment. Each of these factors has been shown to support tumor growth, with tumor-derived G-CSF facilitating MDSC-dependent neoplastic growth[16]; IL-10 expression being correlated with locally advanced breast adenocarcinoma[17]; and VEGF supporting uncontrolled tumor neoangi ogenesis.[18]

CONCLUSIONS

Diverse imaging modalities ahowed that the majority of cancer cells, macrophages and metabolic activity occurs at the tumor periphery, near large blood vessels in an immune competent murine 4T1 breast adenocarcinoma model. While microvasculature existed throughout the tumor, necrosis was scattered but predominately centrally located. CT imaging supported the existence of chaotic blood vessel connectivity and MRI showed limited perfusion of vessels in the central region of the tumor. Intratumoral treatment with the proinflammatory cytokine IL-12 polarized the immune milieu toward a Th1 phenotype with the presence of serum cytokines and immunocytes supportive of tumor suppression. The most descriptive noninvasive indicator of biological changes in the tumor microenviroment, aside from tumor size, was serum cytokine analysis.