1.1.

Cancer Models for In-Vivo Imaging Studies

In-vitro study of cancer cell lines has contributed tremendously to our understanding of the genetics and biochemistry of the malignant phenotype. In-vitro studies offer the advantage of a controlled environment, where one can design experiments to study one variable at a time. However, malignancies occur naturally in the complex environment of a living organism where many stimuli interact with cancer cells simultaneously. Growth of cancer cells in culture does not necessarily translate into tumor growth in vivo. The hospitable in-vivo environment for tumor cells includes appropriate cell signaling through external stimuli, access to nutrients and blood supply, and avoidance of the immune system. To study the myriad of interactions that a developing tumor undergoes requires the use of appropriate animal models that recapitulate key aspects of human tumors. Small rodents, in particular, are useful, as they have been genetically characterized, and strains with desirable genetic backgrounds have been developed to study tumor progression.

Traditionally, in-vivo experiments have been limited to looking at whether or not a tumor grows in a particular host,¹ without being able to characterize specific interactions in the process. Typically, superficial tumor growth has been monitored by caliper measurement, while identifiable internal tumors have been assessed by a single end-point volume measurement. These experiments required sacrifice of the animal to detect, characterize, and quantify the tumor.

Bioluminescence imaging (BLI) is a noninvasive, quantitative method that enables longitudinal studies of the changes in tumor volume and response to treatment in an individual animal over time. BLI measures visible light that is emitted by luciferase-catalyzed reactions on the luciferin substrate in the presence of oxygen.² It has been used to image the development of implanted tumors in mice^{3, 4, 5, 6, 7} and spontaneous tumors in transgenic mice,⁸ to assess the tumorigenicity of cell lines,⁹ and to monitor metastasis and response to chemotherapy.^{6, 9}

With BLI, the pattern of tumor spread can be followed in the same animal over time. However, BLI lacks the sensitivity and spatial resolution to examine events at the single cell level. Intravital microscopy (IVM), on the other hand, permits direct visualization of individual living cells and tissues with submicrometer resolution within an intact organism. Its capability is further enhanced by 3-D optical sectioning techniques such as confocal and multiphoton microscopy. Imaging of structures deeper than the surface of the skin requires surgical exposure to allow optical access due to the limited penetration depth of these imaging modalities, although advances in endoscopic microscopy allow minimally invasive imaging of internal organs through natural orifices or through small openings in the skin.^{10, 11} Similarly, imaging of bone marrow has been difficult due to the thickness of the cortical bones, but can be done through the more translucent calvarial bone of the mouse skull.¹²

IVM has been used to study processes in cancer metastasis that were previously inaccessible by traditional in-vivo and in-vitro assays. Movement of cells within the tumor, the interaction of tumor cells with vascular endothelium, intra- and extravasation of tumor cells, their organ preference through local cytokine attraction, and tumor induction of angiogenesis have all been documented using IVM.^{13, 14, 15, 16} Similar techniques have also been used to obtain insight into angiogenesis, blood flow, cell adhesion, and interstitial diffusion.¹⁷ Most of these studies require labeling of specific structures or cell populations with fluorescent probes or stably expressed fluorescent proteins to provide contrast. Fluorescent reporter proteins, such as green fluorescent protein (GFP), are passed on to daughter cells so that growth of the tumor can be documented,^{16, 18} whereas fluorescent cytoplasmic and membrane dyes are diluted by cell proliferation and are best used in short-term studies.

1.2.

Multiple Myeloma

Multiple myeloma (MM) manifests itself as a malignant proliferation of plasma cells, the terminally differentiated cells in the B cell lineage. It generally affects older individuals¹⁹ and occurs at a frequency of one in 25,000 individuals.²⁰ Symptoms of the disease are a reflection of the biology of the affected cell type. Plasma cells are the antibody producing cells of the immune system. In MM, the malignant plasma cells are clonal and produce one immunoglobulin product, rather than an array of immunoglobulins with a range of specificities. As the disease progresses untreated, the increased number of malignant plasma cells results in the overproduction of immunoglobulin product and subsequent blood hyperviscosity, potential visual problems, and kidney failure.¹⁹ Furthermore, as the tumor outcompetes normal plasma cells, overproduction of the single MM immunoglobulin over the normal antibody array results in compromised immunity in the patient.

MM grows and spreads in the bone marrow (BM), the spongy hematopoietic (i.e., blood-forming) tissue inside bone cavities. Normally, fewer than 0.5% of BM cells are plasma cells.²¹ However, myeloma patients have greatly increased numbers of plasma cells, and usually present with multiple diffuse plasma cell tumors in the marrow of several bones.²² Since MM cells within a patient are clonal, the presence of multiple affected sites suggests that the MM cells metastasize to new distant marrow homes.^{23, 24} Experimental xenograft models of human multiple myeloma exist that measure localized tumor progression (i.e., the SCID-hu and subcutaneous plasmacytoma models)^{25, 26, 27} as well as systemic diffuse bone marrow disease. Whole body imaging (GFP or bioluminescence)^{28, 29} has been successfully used in both of these model systems.

Like some other malignant cell types^{30, 31, 32} and nonmalignant hematopoietic stem cells,^{33, 34} it has been shown that MM cells rely on the surface protein CXCR4 to direct them to the source of the cytokine SDF-1 in the bone marrow.^{35, 36} Binding of SDF-1 to CXCR4 on the surface of a circulating cell induces a cascade of events, including upregulation of integrin binding proteins.³⁷ These events allow MM cells to roll and stick to the bone marrow endothelial surface, enabling the MM cells to exit through the endothelial layer into surrounding tissue. Once in the marrow, MM growth results in activation of the digestive processes normally involved in bone remodeling^{38, 39, 40} but without significant new bone synthesis, leading to lytic bone lesions around the tumor that can be detected by x-ray and that often result in fractures as the disease progresses.¹⁹ While the tumor is mostly confined to BM spaces, late in the disease process MM cells can be detected in the peripheral circulation,⁴¹ and soft tissue organs can be affected.

Recent advances in therapies, both directed toward the tumor cell and its interaction with the microenvironment, have extended patient life span.^{42, 43} Instead of affecting the replication machinery of the cell, these new therapies work by different mechanisms, including inhibiting tumor-directed angiogenesis, inducing MM cell death through apoptosis, interfering with MM binding to bone marrow stromal cells (BMSC) and extracellular matrix proteins, inhibiting cytokine secretion by MM cells, stimulating an immune response, and/or altering transcription of the genome in the tumor cells.^{44, 45, 46} In addition, combining new drugs with older standard MM treatments has proven more effective than either single treatment alone.⁴⁷

We use a xenograft mouse model together with BLI, IVM, and in-vivo flow cytometry (IVFC) to study tumor progression and regression following treatment with one of the new drugs, bortezomib. Bortezomib is a clinically approved agent that acts by interfering with the degradation of proteins by the 26S proteasome complex within the cell. Protein degradation is a component of a complex regulatory network in cells that controls the cell cycle, DNA repair, growth arrest, and apoptosis.^{48, 49} Because cancer cells produce defective proteins at a higher rate than normal cells, they are more dependent on the proteosome to remove these toxic products, and subsequently are more sensitive to blockage of the proteosome complex than normal cells.^{48, 49, 50, 51, 52} Blockage of protein degradation ultimately results in cell death through apoptosis.^{51, 52}

2.1.

Animals

Animal studies were conducted under the approval and guidelines of the Dana-Farber Cancer Institute (DFCI) and Massachusetts General Hospital (MGH) Institutional Animal Care and Use Committees. Young nonirradiated male SCID beige (SCID/Bg) and female BALB/c mice were obtained from Charles River Laboratories (Wilmington, Massachusetts). Col2.3GFP mice⁵³ were used to demonstrate in-vivo homing and engraftment of MM cells with respect to localization of BM osteoblast lineage cells, and were kindly provided by Rowe at the University of Connecticut and maintained by LoCelso at MGH, Boston, Massachusetts. To better approximate physiological conditions under which multiple myeloma normally grows, we used mice that had not been irradiated throughout these experiments.

Anesthesia was performed by intraperitoneal injections of a ketamine (Bedford, Laboratory, Bedford, Ohio)/xylazine (Lloyd Laboratories, Shenandoah, Iowa) cocktail at 80-mg/kg/12-mg/kg body weight. Mice were sacrificed by cervical dislocation while under anesthesia after the appropriate imaging session.

2.2.

Cells

MM.1S cells⁵⁴ were the kind gift of Rosen at Northwestern University. This cell line was genetically modified to carry GFP, firefly luciferase, and neomycin genes (MM.1S-GFP-Luc-neo cells).³⁶ Primary MM cells were obtained from bone marrow samples from patients with the use of CD138 microbead selection (Miltenyl Biotec, Auburn, California). Informed consent was obtained from all patients in accordance with the Declaration of Helsinki. Approval of these studies was obtained by the Dana-Farber Cancer Institute Institutional Review Board.

2.3.

Anti-VLA 4 treatment

MM.1S were reacted with mouse antihuman CD49d (clone HP2/1, AbD Serotec, Raleigh, North Carolina) according to the manufacturer's recommendation before the cells were injected into mice to quantify circulating cells. This antibody binds to the α subunit of the very late antigen (VLA) 4 integrin.

2.4.

Bortezomib treatment

Mice carrying MM tumors were treated with 1-mg/kg bortezomib (PS-341, Millennium Pharmceuticals, Incorporated, Cambridge, Massachusetts) twice weekly beginning at the indicated time. Bortezomib was suspended in phosphate buffered saline at 1-mM concentration, and the mice were injected intraperitoneally.

2.5.

DiD/DiO Cell Labeling

MM.1S cells were labeled for 30 min at 37°C with 1 μM of the lipophilic cyanine dyes, DiD or DiO (Vibrant DiD or DiO, Molecular Probes, InVitrogen, Eugene, Oregon) in RPMI with 0.1% bovine serum albumen (BSA) according to the manufacturer's recommendations and as previously described.³⁵ The cells were injected into mice through the tail veins. The mice were then either subjected immediately to in-vivo flow cytometry, or they were imaged immediately or within a few days to capture the membrane dyes’ fluorescence before dilution through cell division.

2.6.

Imaging Reagents

For delineating the vasculature in imaging experiments, mice were injected intravenously with either 2-ng/0.1ml Angiosense-750 (VisEn Medical, Woburn, Massachusetts) or 0.1 ml of 0.67-μM Qtracker nontargeted Quantum Dots 800 (Molecular Probes, InVitrogen, Carlsbad, California), depending on the specific experiment, as indicated by the manufacturer's directions. The endosteal surface was labeled in some experiments with the fluorescent bisphosphonate Osteosense 750 (VisEn Medical, Woburn, Massachusetts) according to the manufacturer's recommendation.

2.7.

In-Vivo Flow Cytometry

In-vivo flow cytometry is a new technology that allows real-time, continuous monitoring of fluorescent cells in the circulation of live animals without the need to draw blood samples.⁵⁵ In these experiments, mice were anesthetized, 2 to 3 × 10⁶ cells were injected intravenously, and then the mouse was positioned on a temperature-controlled stage (32°C). An appropriate arteriole in the ear was then chosen for obtaining measurements. The GFP or DiO fluorescence of circulating cells was excited by a 473-nm diode-pumped solid-state laser (85 BCA 015, Melles Griot, Carlsbad, California) focused as a slit across the ear vessel. Signal was detected by a photomultiplier tube (R3896 with socket C6271, Hamamatsu Corporation, Bridgewater, New Jersey), through a 525+/−25-nm bandpass filter (BP525/50, Chroma Technologies, Rockingham, Vermont), and then digitized for analysis on a personal computer equipped with Matlab software (The MathWorks, Natick, Massachusetts). Likewise, fluorescence from DiD-labeled cells was excited with a 633-nm HeNe laser (1144/P, JDS Uniphase, Milpitas, California) and detected through a 670+/−20-nm filter (XF3030 670DF40, Omega Optical, Brattleboro, Vermont).

2.8.

Bioluminescence Imaging

Mice were intraperitoneally injected with 75-mg/kg luciferin (Xenogen, Hopkinton, Massachusetts) and imaged for bioluminescence 10 min after the injection. The noncommercial bioluminescence system used an electron multiplying CCD camera (E2V TECH CCD87, Andor Technology, Belfast, United Kingdom) with an exposure time of 15 s, an electron multiplication gain of 600, 5 × 5 binning, and background subtraction. Images were analyzed with ImageJ software, version 1.38 (National Institutes of Health, Bethesda, Maryland, open source at http://rsb.info.nih.gov/ij/index.html).

2.9.

In-Vivo Video Rate Confocal and Multiphoton Microscopy

MM cell growth in the bone marrow of calvarial bones was analyzed using fluorescence confocal microscopy, as previously described.³⁰ Briefly, skin flaps are made in the scalps of the mice to expose the underlying dorsal skull surfaces. Images of the tumors or individual tumor cells were captured in approximately 2 to 3-h long sessions. High-resolution images with cellular detail were obtained through intact mouse skulls at depths of up to 200 μm from the skull surfaces using a 30× 0.9-NA water immersion objective lens (Lomo, Saint Petersburg, Russia), which corresponded to a field of view of 660 × 660 μm. Blood vessels were imaged using blood pool agents, Angiosense-750 (VisEn Medical, Woburn, Massachusetts) or Qtracker nontargeted Quantum Dots 800 (Molecular Probes, Invitrogen, Carlsbad, California). Angiosense-750 was excited with a 735-nm diode laser (Melles Griot, Carlsbad, California), and the fluorescence intensity was captured by a photomultiplier tube (PMT) (R3896 with socket C7950, Hamamatsu Corporation, Bridgewater, New Jersey) detector after passing through a 770-nm long-pass filter (HQ770LP, Chroma Technologies, Rockingham, Vermont). GFP was excited with a 491-nm solid-state laser (Dual Calypso 20, Cobolt AB, Stockholm, Sweden) and detected with a PMT through a 528+/−19-nm bandpass filter (FF01-528/38-25, Semrock, Rochester, New York). Both DiD labeled cells and Qtracker nontargeted Quantum Dots 800 (Invitrogen, Carlsbad, California) were excited with a 635 nm helium-neon laser (Radius, Coherent Incorporated, Santa Clara, California) and detected through a 695+/−27.5-nm bandpass filter (XF3076 695AF55, Omega Optical, Brattleboro, Vermont) or 770-nm longpass filter (HQ770LP, Chroma Technologies, Rockingham, Vermont), respectively. Images of the bone were acquired through a 370+/−18-nm bandpass filter (FF01-370/36-25, Semrock, Rochester, New York) using second harmonic generation (SHG) imaging at 775-nm excitation (Mai Tai HP, Spectra-Physics, Irvine, California). The optical section thickness of the confocal microscope, as measured by translating a mirror through the focus of the microscope and calculating the corresponding full-width at half-maximum signal intensity, is 12 μm at 491 nm and 24 μm at 635 nm. The nominal section thickness of the SHG images is calculated to be 1.4 μm.

Similarly, imaging of Col2.3 mice, as shown in Fig. 1, was done with two-photon and second harmonic generation microscopy, as previously described.⁵⁶ High-resolution images with cellular detail were obtained through the intact mouse skull using a 30× 0.9-NA water immersion objective lens (Lomo, Saint Petersburg, Russia), which corresponded to a field of view of 330 × 330 μm on this instrument. All images were acquired using 840-nm excitation (Mai Tai HP, Spectra-Physics, Irvine, California). Images of bone were acquired using second harmonic generation imaging of the scattering signature from the bone through a 457+/−25-nm bandpass filter (FF01-457/50-25, Semrock, Rochester, New York). GFP-expressing osteoblasts were imaged by two-photon microscopy, detecting the fluorescence through a 545+/−37.5-nm bandpass filter (XF3105 545AF75, Omega Optical, Brattleboro, Vermont). Circulating quantum dots (Qtracker 655) were also imaged by two-photon microscopy through a 710+/−50-nm bandpass filter (HQ710/100M, Chroma Technologies, Rockingham, Vermont).

Fig. 1

Architecture of the normal bone marrow cavity. Selected images from a 3-D stack taken with second harmonic generation microscopy (bone, collagen: blue) and two-photon excited fluorescence (osteoblasts, GFP: green and vasculature, quantum dots: red). Shown here are every 15th image from an original stack acquired with 1-μm z step. The appearance of vasculature within a small bone marrow cavity and osteoblasts lining the cavity wall can be seen beginning at approximately 60 μm below the bone surface. Scale bar equals 100 μm. (Color online only.)

Multiple sequential imaging depths were acquired at 10-μm intervals, except for distance determination experiments. In those cases, 2-μm sequential z-axis stacks were made. The frame rate of the confocal microscope was 30 frames per second. Images were captured, after averaging 30 frames, using a Macintosh computer equipped with an Active Silicon snapper card (CBL-25D-SNP, Active Silicon, Chelmsford, Massachusetts). Time points were taken at days 0, 1, and 3 after MM cell injection, then twice weekly beginning at 4 days after cell injection for the first two weeks, and finally followed by weekly imaging sessions.

2.J.

Image Processing and Three-Dimensional Distance Determinations

Each channel was acquired individually, but simultaneously, in 8-bit grayscale and merged to form an RGB image using custom-developed software (iPhoton). To measure cavity distances from the bone surface (Fig. 1), multiple image stacks from the same bone cavity were concatenated based on coordinate information. We assumed a refractive index of bone close to that of water (1.33) for the purposes of determining the axial location of the optical section. The refractive index of osteones isolated from the femoral diaphyses of oxen was estimated by Ascenzi and Fabry⁵⁷ at various stages of mineralization to range from 1.555 to 1.564. Thus, the refractive index mismatch at the interface between the immersion fluid and the bone will cause a slight axial displacement of the focus away from the bone surface. However, this effect should be somewhat mitigated by the curvature of the calvarial bone. Therefore, we assumed that the curved surface of the bone compensated for the increased index of refraction of the bone, and used an index of refraction of 1.33 for bone for the purpose of determining the axial location of each image in the image stack.

The image stack was imported into ImageJ (NIH, Bethesda, Maryland), resliced into the vertical planes (y–z, x–z), and if necessary, digitally rotated to remove tilt resulting from positioning of the animal relative to the imaging plane. The RGB channels were separated. The grayscale intensity profile along the z-direction of the bone was measured and an exponential decay curve was fit to the data as a function of depth. The inverse of the calculated exponential decay was then multiplied to the bone (blue) channel to correct for the decrease in signal due to increased tissue absorption and scattering with depth. A similar function was applied to the osteoblast (green) and the vascular (red) channels using an empirically derived exponent. A multiplicative constant was incorporated to normalize the intensity between the channels. 3-D Euclidean distance measurements (Fig. 2) between MM cells and the endosteum, vasculature, and osteoblasts were performed using ImageJ after reslicing the z stacks in the vertical planes.

2.K.

In-Vivo Homing Imaging Experiments

Col2.3 mice were injected with 1×10⁵ DiD labeled MM.1S cells. The mice were imaged between 2 to 6, 24, and 72-h post-MM cell injection. Each mouse was used twice for imaging so that specific BM regions could be followed temporally. After the first imaging session, the scalp was sutured, treated with CVS (Consumer Value Stores) Pharmacy brand triple antibiotic salve, and allowed to recover.

2.L.

Long-Term Tumor Growth Experiments

SCID/Bg mice were injected with 5×10⁶ MM.1S-GFP-Luc-neo cells intravenously through the tail vein. Subsets of the mice were imaged twice weekly for the first two weeks, and weekly thereafter. Each mouse was sacrificed after the imaging session.

2.M.

Immunohistochemistry

Mice used for long-term MM growth were sacrificed immediately after imaging. Skulls and femurs were extracted and submitted to the Specialized Histopathology Core Laboratory of the Dana-Farber/Harvard Cancer Center for confirmation of the in-vivo results. The calvarial bones and femurs were dissected and preserved in 10% formalin. The bones were decalcified using Kristensen's solution and then processed according to standard pathologic procedures. The following antibodies were used for immunohistochemical staining (antibody dilution, antibody clone, manufacturer): mouse antihuman CD38 (1:300, SPC32, Abcam, Cambridge, Massachusetts), mouse antihuman CD138 (1:50, MI-13, DAKO, Carpinteria, California).

Antigen retrieval was performed on 4-μm-thick formalin-fixed, paraffin-embedded tissue sections in a pressure cooker using DAKO retrieval solution (for CD138), or 10-mM citrate, pH 6.0 (for CD38) (Zymed, South San Francisco, California). Following incubation in a hydrated chamber for one hour, the primary antibodies were detected with the mouse Envision⁺ kit (for CD38; CD138, DAKO). Immunostaining was visualized with DAB chromogen (DAKO). All stained sections were also counterstained with hematoxylin prior to evaluation by a hematopathologist (Rodig) in a blinded fashion.

3.1.

Normal Vascular/Endosteal Compartment in Bone Marrow

Initially, we imaged normal mouse skull BM in Col2.3 mice (Fig. 1). These reporter mice were genetically engineered to express GFP under the osteoblast-lineage-restricted collagen 1α promoter.⁵³ Osteoblasts are known to reside intermittently along the inner surface of the BM cavity (the endosteal surface) where bone remodeling occurs.⁵⁶ We used second harmonic imaging to detect the collagen signal inherent in structural bone and two photon-excited fluorescence to detect the osteoblast-GFP and blood pool-quantum dot signals. As Fig. 1 demonstrates, we were able to image through the bone to the BM sinusoidal vasculature and endosteal cavity, the roof of which appeared around 60 μm below the surface of the skull. The last frame in this figure demonstrates that the compartment extended to 105 μm below the skull surface for this small cavity. All subsequent BM images in this work were taken between 50 and 200 μm below the bone surface.

3.2.

Trafficking of Circulating Multiple Myeloma Cells to the Bone Marrow

Within minutes after injection, MM cells could be seen interacting with the endothelium of the calvarial bone marrow vasculature by intravital confocal microscopy [Fig. 2a]. In the real-time video (available online) taken 90 min after injection of 3 × 10⁶ cells, some cells have firmly attached to the endothelial wall while others briefly attached before returning to the circulation, and still others remained in the circulation without interacting with the vascular wall. The arrows in Fig. 2a follow the course of a single rolling MM cell.

In-vivo flow cytometry was used to quantify the number of circulating tumor cells as a function of time after cell injection. We have previously shown in this mouse xenograft model that after the first hour, the MM.1S cell number was reduced to approximately 20% of the original injected cell number.³⁵ To confirm that the behavior of the MM.1S cell line in circulation was representative of malignant MM cells, we isolated cells based on the MM CD138⁺ phenotype from MM patient bone marrow biopsies, and subjected them to in-vivo flow cytometry. Figure 2b shows that the MM.1S cell line homing behavior is similar to that of MM patient samples. Previously, we have shown that clearance of the MM cells from the circulation was dependent on the CXCR4/SDF-1 interaction³⁵ and was not the result of nonspecific interactions or MM cell trapping.

Others have shown that the VLA 4 integrin is important in homing and binding of stem cells to bone marrow endothelium, and that the process is completed through VLA 4 activation by SDF-1. Antibodies to VLA 4 inhibit stem cell migration through bone marrow endothelial cells.³⁷ Therefore, to test the role of VLA 4 in the bone marrow trafficking of MM cells that express high levels of this molecule on the cell surface,⁵⁸ we treated MM.1S cells with neutralizing anti-VLA 4 antibody before injecting cells into mice and performing in-vivo flow cytometry. Treatment with the antibody greatly increased the number of MM cells remaining in circulation after one hour compared to the number in control mice [Fig. 2c], further demonstrating that the numbers of circulating MM cells detected by in-vivo flow cytometry are inversely correlated with the homing potential of the MM cells.

To better identify the sites of extravasation of the MM cells and spatial distribution in the bone marrow, we imaged Col2.3 mice that had been injected with DiD-labeled MM.1S cells. In these intravital images, as in Fig. 1, the vasculature is surrounded by bone, the surface of which is intermittently coated with osteoblasts. To identify the preferred homing sites for MM cells, we injected a much lower number of MM cells (10⁵) and followed these for three days. Figure 3 shows images of the injected cells in Col2.3 calvarial BM at 2 [Fig. 3a], 6 [Fig. 3c], and 72 [Fig. 3d] h post-injection. The MM.1S cells were found closely associated with the BM vasculature. 3-D analysis of z stacks allowed us to determine the distance between myeloma cells and BM microstructures. Distance measurements taken at six hours post cell injection showed that most MM.1S cells could be found within 15 μm of a blood vessel, compared to the distance between MM cells and the nearest osteoblast, which were as far as 120 μm away [Fig. 3b]. We found that the myeloma cells continued to remain closely associated with the vasculature over the first three days [Figs. 3c and 3d, and data not shown]. Similarly, injected cells from an isolated patient sample were found in the regions corresponding to the perisagittal areas in mouse calvarial bone 72 h after injection [Fig. 3e].

Fig. 3

Very early after injection, the MM cells position themselves in proximity to the vasculature. (a) DiD stained MM.1S cells were injected intravenously into Col2.3-GFP mice at a dose of 100,000 cells per mouse. Immediately before imaging, the mice were injected with the vascular marker Quantum Dots 800. The mice were imaged within 2-h (a), 6-h (c), and 72-h (d) post-MM cell injection. Images (a), (c), and (d) demonstrate the relationship of the MM cells (white) to the vasculature (red), osteoblasts (green) and bone (blue) during the first 72 h after cell injection. Arrowheads in (a) and (c) point to representative MM cells. (b) Z stacks were acquired from multiple regions in the calvaria of the mice. Distances were measured and tabulated between MM cells and osteoblasts or endosteal surface for the first 6 h after MM cell injection. (e) CD138⁺ patient tumor cells were purified, stained with DiD, and injected into a SCID/Bg mouse. 72-h post-CD138⁺ cell injection, fluorescently labeled MM cells (arrowheads) could be found in the perisagittal regions along the central vessels (arrow). Less bright cells in the same regions are endogenous autofluorescent cells. Scale bars equal 100 μm. (Color online only.)

3.3.

Engraftment and Disease Progression

For long-term tumor cell growth and development, we injected SCID/Bg mice with MM.1S cells that carried retroviral constructs for GFP and luciferase genes (MM.1S-GFP-Luc-neo cell line), allowing us to follow the tumors locally using fluorescence confocal microscopy, in whole animals using bioluminescence, and in the circulation using in-vivo flow cytometry. To investigate MM engraftment under normal BM conditions, we used mice that had not been irradiated prior to MM cell injection. This allowed imaging of unaltered bone marrow infrastructure, avoiding the irradiation-induced destruction of the resident stem and progenitor cells. To delineate the location of MM cells in relation to blood vessels, fluorescent labeling with the blood pool marker Angiosense was performed at serial time points during tumor progression. At four days post-injection, we observed single cells as well as small clusters of GFP⁺ MM cells within vascular beds [Fig. 4a]. Although the entire skull was imaged for the presence of GFP⁺ cells in the x and y planes as well as serially (sectioning through 200 μm) at each x-y position, we could only find a few regions of GFP⁺ MM cells at early times post-injection. Eight and 14 days after the cells were injected, limited numbers of MM clusters were still closely associated with the vasculature [Figs. 4b and 4c]. By the third week, however, the tumor cells had expanded to take up significant areas of the localized sinusoidal bone marrow [Figs. 5a and 5b]. Large area montage images of median skulls are shown in Fig. 5, demonstrating the relationship between the central vessels, sinusoidal vasculature, and expanding myeloma population. Individual myeloma cells (short arrows) were detected in association with the vasculature hundreds of micrometers away from the tumor-dense areas. In the fifth week, however, plasmacytomas filled some of the sinusoids in the calvarial bone [Fig. 6a], and expansion occurred along the sinusoidal vasculature (Fig. 6b]. In addition, MM cells associated closely with the central vein and artery [Fig. 6c]. Plasma cell infiltrate of the marrow was confirmed by histologic preparation of the same mouse calvarium [Fig. 6d], which stained positive for the plasma cell markers CD38 and CD138 [Figs. 6e and 6f, respectively].

Fig. 5

Plasmacytomas are evident in the calvaria by the third week. Mosaics of individual fields of view imaged by confocal microscopy show that at days (a) 20 and (b) 22 post-MM.1S cell injection, confocal imaging detects the formation of plasmacytomas in the BM sinusoids. Individual MM cells exist away from the bulk of the tumors (short arrows). Long arrows indicate the positions of the central arteries and veins under the metopic sutures of these mice. Scale bars equal 100 μm. (Color online only.)

Fig. 6

In the fifth week, tumors are detected by immunohistochemistry, confocal microscopy, and BLI. (a), (b), and (c) are confocal images, that reveal tumor that has filled sinusoidal regions (a), grown along the central artery and vein (c), and continued along more distal vessels (b). Scale bars equal 100 μm. Immunohistochemistry of the calvarium of the same mouse as in (a), (b), and (c) shows massive tumor infiltrate (d) that stained positively for (e) CD38 and (f) CD138. Original magnification in (d), (e), and (f) was 400×. Bioluminescence images of ventral and dorsal views of two representative mice showing (g) and (i) skull and vertebral involvement; and hip, femur, and rib involvement in (h) and (j). Color bar indicates the relative luminescence intensity. (Color online only.)

Bioluminescence images acquired at the 5-week time point indicated that in addition to the skull, tumor was growing in the vertebral column [Figs. 6g and 6i], as well as pelvis and femurs [Figs. 6h and 6j]. The pattern of tumor growth is similar to the human disease, which if left untreated, is marked by the spread of MM cells to the BM of the pelvis, femurs, vertebral column, and skull.^{19, 25}

In human patients, MM cells can also be found in the circulation as the tumors are established.^{23, 41} To enumerate MM cells in mouse circulation during disease progression, we used noninvasive in-vivo flow cytometry⁵⁵ to detect the number of GFP⁺ cells flowing through an ear arteriole over time. Figure 7 (solid line) shows the increase in circulating cell number during the first six weeks of tumor progression.

Fig. 7

The number of circulating MM cells is an accurate measure of tumor burden. Weekly in-vivo flow cytometry was repeatedly performed on mice that had been injected with MM.1S-GFP-Luc-neo cells. The number of GFP⁺ cells were counted and plotted against time. Beginning in the sixth week post cell injection, bortezomib was administered twice-weekly as is given clinically. Solid line indicates cell counts accumulated before the mice were treated. Dashed line indicates cell counts after twice-weekly bortezomib treatment. For each data point, n = 3.

3.4.

Response to the Therapeutic Agent Bortezomib

To monitor response to bortezomib treatment in vivo in the SCID/Bg MM model, we began treatment twice weekly when tumors became visible using BLI. After three 1-mg/kg bortezomib treatments, the tumor size decreased, as shown in the BLI images [Figs. 8g and 8h vs. Figs. 8e and 8f], and the number of circulating MM cells also decreased (Fig. 7, dashed line) as determined by in-vivo flow cytometry. When the tumor burden fell below the BLI detection level, a small number of remaining MM cells were still clearly evident by confocal imaging [Figs. 8d vs. 8c]. Even more dramatically, the number of areas in the calvarial BM where MM cells could be found decreased after bortezomib [Figs. 8b vs. 8a]. At this point, several areas of treated calvarial bone contained only one or few MM cells [Fig. 8d]. These results clearly indicate the potential for use of the increased sensitivity of in-vivo confocal microscopy in following tumor treatment and detection of residual tumor cells post-therapy in cancer models.

Fig. 8

In-vivo confocal imaging detects surviving cells after therapy. Throughout this experiment, diagrammatic maps were made of GFP⁺ MM sites in the calvaria of mice at selected times after the MM.1S-GFP-Luc-neo cells were injected. Detection of GFP⁺ MM cells was done in x, y, and z planes. (a) and (b) Representative map of GFP⁺MM sites from a mouse (b) after three bortezomib treatments and (a) a control untreated mouse at the same time point. Vertical line represents the central vein and artery; wavy horizontal line represents the coronal suture. For mapping purposes, each skull was divided into six regions (R numbers). Mosaics of images of the fields of view corresponding to maps (a) and (b) are shown in (c) and (d), respectively. For (c) and (d), the endosteum was delineated with osteosense, a fluorescent bisphosphonate that binds to hydroxyapatite at the bone surface (blue). Scale bars equals 400 μm for (c) and (d). Representative bioluminesence images of mice (g) and (h) treated with three bortezomib doses or (e) and (f) not treated. Color bar indicates the relative luminescence intensity. (Color online only.)