Research ArticleImaging

Intravital microscopy of osteolytic progression and therapy response of cancer lesions in the bone

See allHide authors and affiliations

Science Translational Medicine  01 Aug 2018:
Vol. 10, Issue 452, eaao5726
DOI: 10.1126/scitranslmed.aao5726

Bone tumors revealed

Noninvasive imaging can help monitor cancer metastasis and tumor-stroma interactions but is challenging for thick, dense tissues such as bone. Dondossola et al. studied prostate cancer metastasis to bone using tissue engineering and intravital multiphoton microscopy in mice. A skin window overlaying implanted engineered bone constructs injected with cancer cells allowed for observation of osteolysis in the bone constructs, with osteoclasts localized at the tumor-bone interface. Treatment with zoledronic acid slowed osteoclast activity (bone resorption) without affecting cancer growth. This engineered bone and imaging method gives a glimpse into tumor-bone interactions that could be useful to test therapies for bone remodeling and cancer metastasis.


Intravital multiphoton microscopy (iMPM) in mice provides access to cellular and molecular mechanisms of metastatic progression of cancers and the underlying interactions with the tumor stroma. Whereas iMPM of malignant disease has been performed for soft tissues, noninvasive iMPM of solid tumor in the bone is lacking. We combined miniaturized tissue-engineered bone constructs in nude mice with a skin window to noninvasively and repetitively monitor prostate cancer lesions by three-dimensional iMPM. In vivo ossicles developed large central cavities containing mature bone marrow surrounded by a thin cortex and enabled tumor implantation and longitudinal iMPM over weeks. Tumors grew inside the bone cavity and along the cortical bone interface and induced niches of osteoclast activation (focal osteolysis). Interventional bisphosphonate therapy reduced osteoclast kinetics and osteolysis without perturbing tumor growth, indicating dissociation of the tumor-stroma axis. The ossicle window, with its high cavity-to-cortex ratio and long-term functionality, thus allows for the mechanistic dissection of reciprocal epithelial tumor-bone interactions and therapy response.


Bone metastases account for many complications experienced by patients suffering from advanced stages of different cancers, such as prostate, breast, and other epithelial tissues, with generally incurable outcomes (1, 2). After seeding in the bone, cancer cells interact with the microenvironment and undergo molecular reprogramming, which supports both tumor growth and bone remodeling in a vicious cycle (3). To date, analysis of bone metastasis of solid tumors is mostly obtained by molecular approaches combined with histology and immunohistochemistry ex vivo on fixed tissue material and macroscopic imaging studies (1). Despite moving the field forward, these strategies lack the sensitivity to characterize dynamic and reciprocal interactions between malignant cells and the bone stroma during progression and therapy response and the three-dimensional (3D) adaptive niches, which provide this cross-talk. Thus, strategies for the longitudinal monitoring of the tumor-stroma interface in the bone are much needed to dissect the pathophysiology of bone metastasis and mechanisms of therapy outcome in both preclinical research and mechanistic studies accompanying clinical interventions (1).

3D and time-resolved analysis of the kinetics of tumor-stroma interactions at the single-cell level has been performed by multiparametric intravital multiphoton microscopy (iMPM) in small animals using higher harmonic light scattering and near-infrared and infrared fluorescence excitation (4, 5). In soft tissues, optical access to primary or secondary tumors is achieved by innovative window systems (4, 610); however, iMPM in mature bone is limited by poor optical accessibility due to its cortical thickness, which causes light to scatter, and is further complicated by the complex topology of bone and multifocal organization of bone metastases (1116). In the mouse calvaria, iMPM enables effective monitoring of osteoclasts and osteoblasts, hematopoietic stem cell functions (1114), and their interaction with hematopoietic tumors (12, 17). However, calvarial bone lacks the cavity space for direct implantation and growth of solid tumor lesions. To visualize cells in the tibia, iMPM requires the surgical removal of both muscle and cortical bone for retrieving the kinetics of hematopoietic progenitor and myeloma cells, mostly as a one-time procedure (15, 16, 18). Because of its invasive nature and reduced imaging and mechanical stability (hence, fracture risk), the tibia window has not been exploited for 3D longitudinal monitoring of the complex topology and slow progression kinetics of bone metastasis.

Tissue engineering allows for the generation of a shape-defined, morphologically, and functionally intact mature bone organ within the mouse skin to dissect mechanisms of cancer metastasis to the bone, including breast and prostate cancer (PCa) (1923). Here, we combined a miniaturized tissue-engineered bone construct (mTEBC) with an optical skin window (10, 24) to create an ossicle with high optical accessibility for noninvasive time- and space-resolved long-term iMPM of tumor-stroma interactions, osteolytic bone remodeling, and the reprogramming of the bone microenvironment during experimental therapy in mice.


mTEBC generation

To enable microscopic, nondestructive monitoring of both solid tumors and the bone microenvironment, a sufficiently large volume of the bone cavity must coincide with optically favorable bone anatomy, supporting tumor cell implantation and expansion and high optical accessibility. To achieve these requirements, we miniaturized medical-grade polycaprolactone (mPCL) scaffolds with an interconnected honeycomb pore network, functionalized with calcium phosphate (CaP) (Fig. 1, A and B), as square microscaffolds (5 mm × 5 mm × 0.2 mm) (1921), which were implanted into the dorsolateral flank of nude mice, underneath the future position of the skin imaging window (Fig. 1C). To induce ectopic bone formation, mPCL-CaP scaffolds were coimplanted with fibrin glue and recombinant human bone morphogenetic protein 7 (BMP7), which created an ectopic ellipsoidal bone within 30 days (Fig. 1D). The volume of the bone increased with the amount of coadministered BMP7, reaching 30 mm3 with a relatively flat outer cortical bone surface parallel to the body surface (fig. S1), which represented an ideal geometry for juxtaposing an imaging window.

Fig. 1 Generation and characterization of a tissue-engineered ossicle in vivo.

(A) Macroscopic overview of melt electrospun mPCL-CaP scaffold by bright-field microscopy. Box: Magnification shown in (B). Scale bar, 1 mm. (B) Maximum intensity projection of mPCL-CaP scaffold by multiphoton-excited SHG detection (z depth of 100 μm). Box: Magnification. Scale bar, 100 μm. (C) Schematic representation of the model. Nude mice were implanted subcutaneously with an mPCL-CaP scaffold embedded in fibrin glue and BMP7 (10 μg per mouse) that generated a mature, vascularized ossicle over time. (D) Localization of the ossicle (mTEBC) by micro–computed tomography (μCT) at day 30. Ve, vertebra. Box: Representative 3D reconstruction and bright-field image shown on the right. Scale bars, 5 and 1 mm, respectively. (E to H) Histological overview, cortical detail, and tartrate-resistant acid phosphatase (TRAP) staining showing mTEBC maturation over time [d10 (E), d20 (F), d30 (G), and d40 (H); d, day after implantation]. Scale bars, 1 mm (overviews) and 100 μm (zooms). Insets show higher magnifications of boxed regions in (H). Scale bar, 25 μm. (I) Quantification by bone histomorphometry. Parameters analyzed: osteoblast surface (percentage of total bone surface; Ob. S/BS, %), osteoclast surface (percentage of total bone surface; Oc. S/BS, %), erosion surface (percentage of total bone surface; ES/BS, %), bone volume (percentage of tissue volume; BV/TV, %), and trabecular separation (Tb. Sp.). Mean ± SD is shown (n = 4 to 6 per group). Bone cortex measurements: five cortical areas per mTEBC, 350 × 150 μm each; medullary region measurements, four areas per mTEBC, 250 × 250 μm each. *P < 0.05; ***P < 0.001 by one-way ANOVA, followed by Tukey’s post hoc test.

The generation of the mTEBC occurred as a triphasic process. Initial bone deposition was associated with high osteoclast density and multiple sites of bone erosion, forming an incipient interosseous cavity with primordial cellular content (Fig. 1, E and I, and fig. S2, A to C). Until day 20, bone remodeling and maturation of the bone marrow led to a continuous cortical bone surrounding a well-defined cavity interspersed with trabecular bone, hematopoietic bone marrow, and bone stroma, including adipocytes, osteoclasts, osteoblasts, and osteocytes (Fig. 1, F and I, and fig. S2, A to C). By day 30, an ossicle with cortical bone of even surface geometry and consistent thickness, trabecular bone, and niches of bone marrow was formed, which remained morphologically stable until up to 300 days (Fig. 1, G to I, and fig. S2, A to D). In summary, a miniaturized bone organ that stably persists over months in the mouse skin was developed.

Minimal cortical thickness and maximal bone marrow cavity with mature marrow and vasculature

Cortical bone thickness critically confounds the optical resolution of cells located in the bone cavity (15, 25). We thus quantified the evolution of the cortical bone layer by second-harmonic generation (SHG) microscopy, which detects collagenous bone matrix (Fig. 2A) (26). During development, the average thickness varied between 50 and 80 μm and uniformly stabilized at 53.3 ± 3.3 μm after 30 days (Fig. 2, A and B). This thickness was validated by stereomicroscopy of native cross-sectioned bone (Fig. 2, B and C). Despite attenuation of laser light by the bone (75% of power loss at the inner interface), this thickness enabled full-penetrance iMPM microscopy (Fig. 2D). In practice, we compensated light attenuation by z-power adaptation and reached ~2-fold improved excitation underneath the bone (Fig. 2E). Besides thickness, the quality of cortical bone improved over time as quantified by measuring the area of the microcavities decorating the bone surface (Fig. 2, A and F). The geometry of micro-ossicles further offered a superior cavity (cross section, 17 to 27 mm2) and ratio to cortical thickness when compared to natural bones, including the mouse tibia, vertebra, and rib, which are typical locations of PCa metastases, or the calvaria, which is the most accessible location for iMPM of hematological neoplasias (Fig. 2, G and H) (15, 25).

Fig. 2 Analysis of the mTEBC morphology.

(A) Maximum intensity projection of the SHG signal to detect the cortical bone of the ossicle ex vivo at days 10, 20, 30, and 40. Horizontal (xy; left) and orthogonal view (xz; right) are shown (z depth of 150 μm). Scale bar, 100 μm. (B and C) Quantification of cortical bone thickness by MPM or stereomicroscope [image in (C)]. Mean ± SD; six areas per mTEBC were averaged; three ossicles from independent mice per time point. (D) Measurement of the signal intensity in dependence of the laser power at the outer and inner cortical bone surface. Scanning wavelength, 1090 nm. One representative of two independent experiments is shown. (E) z-power adaptation (z-PA). Representative orthogonal view (xz) images of increased intensity and signal distribution are shown. Scanning wavelength, 1090 nm. (F) Quantification of surface porosity, represented as mean ± SD; individual data points are shown (six areas per mTEBC were averaged; three ossicles from independent mice per time point). (G) Histology (hematoxylin and eosin) of the ossicle compared to the mouse tibia, vertebra, rib, and calvaria. Scale bar, 50 μm. (H) 3D SHG reconstruction of the indicated bones (z depth of specimens: TEBC, 1 mm; tibia, 800 μm; vertebra, 1 mm; rib, 400 μm; calvaria, 800 μm). Yellow dashed line, cavity. Box: Magnification of cortical bone. Red dashed line, cortical bone. Data are means ± SD (n = 4 to 6 independent samples). (I) Left: Intravital detection of blood vessels in the mTEBC [70-kDa Alexa Fluor 750–conjugated dextran (Dx)]. Center and right: MPM detection of blood vessels (laminin) in cleared mTEBC, tibia, and calvaria ex vivo. Right: Quantification (mean ± SD, n = 3). Scale bar, 100 μm. *P < 0.05; ***P < 0.001 by one-way ANOVA, followed by Tukey’s post hoc test.

We next verified the physiological composition and functionality of the bone marrow in the engineered bone. The proportions of platelets, red cells, and white cells and the population distributions of granulocytes (CD11b+GR1+ cells), monocytes (CD11b+GR1 cells), and B cells (B220+ cells) present in the engineered bone were similar to the bone marrow isolated from tibia (fig. S3, A to D). Furthermore, bone marrow transplantation using donor cells isolated from mTEBCs 40 days after implantation successfully reconstituted the bone marrow in lethally irradiated recipients (fig. S3, E and G). Thus, similar to existing strategies to supplement bone-free hydrogels with type I collagen and bone marrow stromal cells (24) and large TEBCs functionalized with mesenchymal stem cells and BMP7 and BMP2 (20, 23), this mTEBC recreates a functional environment to recruit precursor cells and establish mature bone marrow (20, 23, 24).

The vasculature within the mTEBC consisted of perfused blood vessels detected by intravital microscopy after intravenous injection of Alexa Fluor 750 dextran (70 kDa) (Fig. 2I and fig. S4A) and expression of the surface marker endomucin (fig. S4B), confirming vascular network organization and sinusoidal differentiation typical of natural bone (27). The vessel density present in the mTEBC showed a space occupancy of ~30% (Fig. 2I and fig. S4C), which matched the vascular density of natural bones used for intravital imaging in other contexts, including tibia and calvaria, as well as the mouse femur (28). In aggregate, the mTEBC develops a large bone cavity that harbors mature bone marrow and sinusoidal blood vessels and, because of the thin, optically transparent cortex, presents a suitable geometry for monitoring solid bone tumor biology by iMPM.

Longitudinal intravital microscopy of solid tumors in the bone

Preclinical analyses of therapy response require sufficiently consistent and metronomic cohort behavior. Spontaneous bone colonization from orthotopically implanted tumors or by cancer cells administered in the circulation recapitulates naturally occurring bone metastases; however, these routes display low penetrance of macrometastases and high interindividual variability (29). Thus, we applied PCa cells by direct injection into the ossicle. This route maximizes the comparability of tumor growth and therapy response between mice. It further minimizes metastatic dissemination to other organs, which increases the risk of premature euthanasia due to systemic tumor burden and compromises reliable therapy response monitoring.

To generate solid tumors in the bone, we implanted 150,000 human osteolytic PC3 or osteoblastic C4-2B PCa cells as a dense cell suspension in ~3 μl of phosphate-buffered saline into the ossicle using a microneedle (Fig. 3A). PCa lesions within the mTEBC recapitulated the histotopology of patient biopsies (Fig. 3B and figs. S5 and S6A). To allow iMPM, we surgically applied a skin window to the mouse 2 weeks after tumor implantation without disrupting the outer cortex and the periosteum of the mTEBC (Fig. 3A). Using combined near-infrared and infrared iMPM through the intact cortical bone, PC3 lesions expressing histone H2B (H2B)/enhanced green fluorescent protein (eGFP) and DsRed2 (30) signals were detected with subcellular resolution, revealing cell shape and nuclear states (including mitosis, apoptosis, and interphase), as well as the bone structure using SHG microscopy (Fig. 3, C and D). The penetration depth reached typically up to 200 μm (four to six cell layers) into compact-shaped, bulky tumor lesions (Fig. 3C and movie S1). Similar microscopic penetration depth and resolution were reached for C4-2B osteoblastic PCa tumors expressing H2B/mCherry and LifeAct/GFP, which detects F-actin (fig. S6, D and E). Through the intact cortical bone, differential actin organization along cell-cell junctions and in subcellular protrusions was detected with subcellular detail (fig. S6E), indicating efficient monitoring of subcortical tumors in osteolytic and nonosteolytic tumor types. Tumor growth and suitability for iMPM were further confirmed using breast cancer cells (MDA-MB-231), which are known to metastasize to the bone and induce osteolysis (fig. S6, F to I) (31).

Fig. 3 PC3 tumor cell administration and monitoring by histology and iMPM.

(A) Schematic representation of the model. PC3 cells were injected into the mature ossicle 30 days after implantation into nude mice. A dorsal skinfold chamber (DSFC) was applied 10 to 14 days after tumor cell implantation, adjacent to the ossicle to enable iMPM. (B) Histology of PC3 cells in mTEBC and patient biopsy. T, tumor. Scale bar, 100 μm. (C) iMPM detection of PC3 cells at different scanning depths (z-stack at z = 0, −50, and −96 μm). Overviews shown as xy and xz, yz orthogonal sections, and detail with subcellular resolution (dotted box). Dashed line in top left denoted the edge of the ossicle with fluorescence signal originating from the tumor. (D) XZ intensity profile of individual channels with increasing scanning depth into the ossicle. (E) Quantification of interphase (IF), mitotic, or apoptotic cells in 3D stacks captured by iMPM. Mean ± SD is shown (n = 5 independent lesions). Nucleus, H2B/eGFP (green); cytoplasm, DsRed2 (red); collagen fibers: bone, SHG (cyan). Scale bar, 100 μm.

As in natural bone and large TEBCs, PC3 tumors caused macroscopic and microscopic osteolysis (21, 32), with focal osteoclast recruitment and erosion of the mineralized matrix (Fig. 4A), followed by progressing cortical bone defects that fused and ultimately led to near-complete dissolution of the mTEBC by day 34 (Fig. 4B and fig. S7). As a consequence of cortical bone erosion, improved imaging penetration depth in osteolytic lesions was achieved (Fig. 4C versus Fig. 3C).

Fig. 4 Monitoring of osteolysis and dynamics of Cat K+ cells.

(A) TRAP staining of PC3 lesion in mTEBC. Box: Inset (magnification). Arrowheads indicate osteoclasts. Scale bar, 100 μm. (B) Evaluation of osteolysis by ex vivo μCT and SHG detection by MPM. Magnifications of numbered boxed regions are shown. Scale bar, 1 mm. (C) iMPM detection of an osteolytic lesion. Merged xy and xz, yz orthogonal views, single channels, and details are shown. Box: Magnification. Nucleus, H2B/eGFP (green); cytoplasm, DsRed2 (red); collagen fibers: bone, SHG (cyan); bone remodeling, OsteoSense (OS; white); osteoclast/phagocyte, cathepsin K (Cat K; yellow). Solid arrowhead, mitotic event; void arrowhead, apoptotic nucleus; asterisk, Cat K+ cells. Scale bar, 100 μm. (D) Imaging analysis, xy intensity profile for single channels and quantification of the distance between tumor border (first layer of cells, DsRed2) and osteoclast/phagocyte (Cat K) or bone (OS). Scatter diagrams represent measurements from different regions in three independent lesions. (E) Dynamics of Cat K+ events monitored by time-lapse iMPM and analyzed by tracking. “+” and diamonds indicate the start, red dots indicate the final position of each cell, and numbers indicate individual cells. Right: Heat map of the speed of 15 representative cells. Scale bar, 20 μm. (F and G) Cat K signal detected over time by time-lapse iMPM (F) and particle imaging velocimetry (PIV) obtained by whole-field analysis (G). Two sequential frames obtained at different time points (0 and 8 min) from movie S2 are shown. Scale bar, 10 μm. Magnitude, vector map, and speed frequency distribution (three to four areas per tumor, from three mice).

As a direct proxy for co-registering the osteolytic activity along the tumor-bone interface by iMPM, remodeled bone was identified by fluorescent pamidronate (OS) and osteoclasts were visualized by Cat K (Fig. 4C) (33, 34). Cat K is a cysteine protease predominantly localized in the cytoplasm and endosomal/lysosomal compartments of osteoclasts (35). The 3D topology analysis of OS, Cat K, SHG, and dual-color PC3 cells was used to map zones of the tumor-bone interface, where nodular PC3 lesions were bordered by complex-shaped Cat K–positive (Cat K+) phagocytes and osteoclasts (Fig. 4, C and D; fig. S8; and movie S2). The density of osteoclasts, detected by TRAP staining in tissue sections or fluorescent Cat K probe in vivo, decreased with distance from the tumors (fig. S9), suggesting that solid PC3 tumors recruited Cat K+ cells locally.

To map the dynamics of Cat K+ cells along the zone of bone resorption, we performed time-lapse recordings to track the cell positions, followed by particle image velocimetry for subcellular dynamics (movie S3). Osteoclasts in lacuna-like bone resorption niches, detected by SHG (fig. S7), remained positionally stable (Fig. 4E) but retained vigorous cytoplasmic and vesicle dynamics with a variable speed of up to 1.2 μm/min (Fig. 4, F and G). Similar positional stability of bone-resorptive osteoclasts was reported for cancer-free calvarial bone (14). The general lack of protrusive activity of bulky PC3 lesions growing adjacent to osteolytic phagocytes/osteoclasts suggests that bone destruction occurs without signs of active tumor invasion.

Disconnecting the tumor-stroma interface by bisphosphonate treatment

To demonstrate the suitability of iMPM for longitudinal monitoring of therapy responses in the micro-ossicle, established PC3 lesions with incipient osteolysis were monitored sequentially during treatment with zoledronic acid (ZA) (Fig. 5A). ZA inhibits osteoclast-mediated bone resorption by modifying prenylation of proteins, including regulators of the cytoskeleton and vesicle trafficking involved in bone remodeling (36, 37). When given therapeutically to patients with metastatic cancer, ZA reduces bone pain and delays skeletal degeneration (38). In control mice receiving vehicle, osteolysis progressed over time, with local resorption speeds of hundreds of micrometers per day when monitored by serial iMPM (Fig. 5, B and C). As expected, ZA near-instantaneously and completely halted bone resorption at the macroscopic level (Fig. 5, B and C). However, when monitored microscopically, residual dynamics of the bone interface in ZA-treated mice revealed ongoing focal microremodeling (Fig. 5D), with subregions of the same lesion progressing at 3 to 10 μm/day (Fig. 5, E and F). ZA administration did not alter the density of Cat K+ cells within 6 days of treatment (Fig. 5G), consistent with stable osteoclast density in clinical bone specimens (39, 40). However, particle image velocimetry of time-lapse recordings revealed that the subcellular kinetics of Cat K+ cells was significantly decreased, resulting in both cell stasis and ablation of fast subcellular movements (Fig. 5, H to J; fig. S10; and movie S4 versus movie S5). ZA caused no change to PC3 cell growth, detected as the frequency of apoptotic and mitotic cells at the tumor-bone interface in the same lesions over time (Fig. 5K). This reveals that the primary effect of this bone-stabilizing therapy acts on the stromal but not on tumor compartments.

Fig. 5 Response to zoledronic acid therapy monitored by iMPM.

(A) Schematic representation of the experimental schedule. Ten to 14 days after PC3 cells injection, mice were implanted with a DSFC and treated with ZA (20 μg per mouse in saline by a single intravenous injection) the subsequent day. (B and C) Longitudinal monitoring of osteolytic progression by SHG detection at iMPM in control (vehicle, saline) and ZA-treated mice and quantification of the resorbed area of lesions represented in (B). Scale bar, 100 μm. Note the different scale of the images representing control and ZA-treated tumors to accommodate the different size of the osteolytic region. (D) Imaging analysis of micro-osteolytic lesion progression in ZA-treated mice. (E to F) Quantification of the microspeed of bone resorption after administration of ZA in individual subregions (E) and corresponding average microspeed per subregion over time representing the changes from days 0 to 9 (F). Data are means ± SD (n = 10 microregions from three independent osteolytic lesions). μregion, microregion; bone μres, bone microresorption. (G) Density of Cat K signal over time. Mean ± SD (n = 3 independent lesions per time point). Scale bar, 100 μm. No significant difference was identified by one-way ANOVA, followed by Tukey’s post hoc test. (H) PIV analysis of Cat K signal in control and ZA-treated mice; two representative frames and magnitude map. Scale bar, 20 μm. Area, 65 × 65 μm. (I and J) Average speed in control and ZA-treated mice. Mean ± SD. For image analysis, the quantification of PIV was obtained from three to four areas per lesion, from four mice per group. Speed distribution obtained from representative cells highlighted by red outline in (I). ***P < 0.001, unpaired two-tailed Student’s t test. (K) PC3 cell replicative status. Representative images taken at the onset and 6 days after ZA application (left) and quantification of the ratio between mitotic and apoptotic cells at different time points (right diagram); mean ± SD (n = 4 independent lesions per time point). Scale bar, 10 μm. No significant difference was identified by one-way ANOVA, followed by Tukey’s post hoc test.


Combining a tissue-engineered mature micro-ossicle with skin window–based iMPM provides a new strategy for longitudinal analysis of solid tumor interactions with bone, including tumor-induced osteolysis and therapy response. The uniformly thin cortex of the ossicle surrounds a large bone marrow cavity and permits both tumor implantation and iMPM through the intact bone lamella without perturbing bone integrity. Implanting a cell-free, CaP-functionalized scaffold in combination with BMP7 was sufficient to induce ossicle maturation, followed by long-term stability for several hundred days. This indicates that BMP7, added during implantation, orchestrated neobone generation of the host tissue, in contrast to other mesenchymal stem cell–based strategies that depend on cell-based functionalization (41, 42).

Besides the reliably thin cortical lamella, key features of the miniaturized TEBC include (i) tailored, roundish shape optimal for juxtaposing the imaging window; (ii) recruitment and maintenance of a mature bone marrow and vascular networks; and (iii) a reactive bone response to tumor growth, including osteoclast activation and bone erosion. These characteristics differ from other systems aimed at monitoring bone biology in vivo, including implanted bone fragments (4345). Implanted bone fragments lack optical transparency and incompletely recreate vital bone because, over time, they fail to support active bone neo-apposition, microvascularization development, and recruitment of hematopoietic bone marrow precursor cells but instead frequently develop fibrosis (4345). Thus, the mTEBC window model combines mature bone topology and functionality with accessibility for longitudinal intravital microscopy of tumors growing inside the bone cavity.

By positioning the skin window near the intact cortical bone, repeated 3D microscopy could resolve the position and activity of individual malignant and nonmalignant cells inside the mTEBC, their subcellular kinetics (including mitosis, apoptosis, and shape oscillations), and the spatial evolution of bone remodeling. During bone resorption, Cat K+ cells accumulate at the tumor-bone interface, become largely stationary but retain vigorous subcellular dynamics, and establish local resorption niches, which coalesce until the bone is destroyed, similar to resorbing osteoclasts in vitro and in cancer-free calvaria in vivo (14, 46, 47). In vitro positional stability of osteoclasts is mediated by the podosome-rich sealing zone that focuses the lytic region to the bone surface and defines the bone resorption niche (46, 47). Concurrently, osteoclasts maintain highly dynamic substructures, including the actin-rich ruffle border engaged in Cat K release and resorption, lamellipodia exploring free bone surface (4850), and Cat K–containing secretory vesicles, which depend on the joint activity of the small guanosine triphosphatases (GTPases) Rac and Rab (37, 49). ZA inhibits the farnesyl diphosphate synthase and thereby prevents prenylation of small GTPases, including Rac and Rab, which inhibits their membrane localization and function (37). Consistently, ZA perturbs vesicular trafficking and essential actin structures in vitro, including the sealing zone, ruffling border, and lamellipodia dynamics, and this combined effect likely impairs bone resorption (36, 37, 51). The in vivo effects of ZA on osteoclast dynamics and function, including impaired subcellular dynamics and near-ablated bone resorption, thus validate actin and vesicular dynamics as central effectors of bone remodeling. ZA did not affect tumor cell viability and growth in vivo, which explains the clinically observed disconnect between the well-documented stabilization of bone despite unperturbed growth of the cancer lesions (52, 53) and clinical disease progression (54).

The conundrum between ZA-induced bone stabilization despite ongoing tumor growth further validates the use of the mTEBC model in support of bone remodeling as a secondary, largely osteoclast-dependent pathology, which can be uncoupled from tumor expansion. Beyond cancer, the ossicle window represents an anatomically stable and genetically versatile platform to enable a range of kinetic studies in basic bone biology, tissue engineering, and reparative medicine. Variation of the mTEBC will provide additional options for including human cells, such as hematopoietic stem cells or engineered stromal cells (mesenchymal stem cells/osteoblasts, osteoclasts, or vascular cells), and genetic modulation of the bone stroma (20). Noninvasive monitoring of the tumor-bone interface may deliver preclinical and coclinical insights into metastatic cancer progression, bone remodeling, and multitargeted therapy response.


Study design

The objective of this study was to establish a preclinical model for intravital microscopy of solid cancer in the bone. Tissue engineering was applied to generate an ossicle that combines (i) minimal cortical thickness, (ii) large bone marrow cavity, (iii) mature bone marrow and vasculature, (iv) positioning in the skin suitable for applying an imaging window, and that (v) provides sufficient optical transparency for detecting cells with a subcellular resolution by multiphoton microscopy through the intact cortical bone. Fluorescent microtumors were implanted directly, and intrabone tumor growth and osteolysis dynamics were 3D-reconstructed by multichannel intravital microscopy and further monitored by bone histomorphometry, flow cytometry of maturity of the bone marrow, and μCT for ossification. Using molecular reporters, the osteolytic tumor-bone interface was detected by intravital microscopy in the live mouse over time, including the composition and kinetics of collagen and bone matrix turnover (SHG), calcified bone (fluorescent bisphosphonates), osteoclasts (Cat K), and cancer cells (nuclear and cytoplasmic fluorescent reporters). As a proof of concept, the ossicle model was applied for iMPM of bisphosphonate therapy to halt bone remodeling and the effects on bone resorption, osteoclast, and tumor biology were recorded. Repeated microscopy of the same mouse allowed to obtain multiparametric longitudinal data from the same tumor, thereby maximizing data reliability and minimizing animal consumption. Accordingly, three to five mice per group were sufficient to reach statistical significance. Image acquisition in vivo and ex vivo from control and treated mice was performed in a blinded manner. Data analysis was performed in a nonblinded manner.

mTEBC generation, tumor cell injection, and DSFC creation

Animal studies were approved by the Institutional Animal Care and Use Committee of the University of Texas MD Anderson Cancer Center and performed according to the institutional guidelines for animal care and handling. To generate a micro-ossicle, male and female nu/nu athymic or male nonobese diabetic (NOD)–severe combined immunodeficient (SCID) mice (4 weeks old; provided by the Department of Experimental Radiation Oncology, MD Anderson Cancer Center) were anesthetized using ketamine/xylazine, and a 5 mm × 5 mm × 0.2 mm (length × width × thickness) mPCL-CaP scaffold was inserted into a small surgical subcutaneous pocket (4 to 5 mm) in the presence of fibrin glue (5 μl of fibrinogen/5 μl of thrombin; fibrin sealant TISSEEL, Baxter) and 10 μl of BMP7 (10 mg/ml; Gibco), followed by wound closure (one or two 4.0 polypropylene sutures; Covidien). The subdermal implantation site preserved the integrity of the muscular tissue and fasciae of the underlying ribs and spines. Ossicle formation was independent of implantation site in different locations in the back of the mouse. At least 30 days after surgery, tumor cells (1.5 × 105 cells in 3 to 5 μl of phosphate-buffered saline) injected into the mTEBC (30 gauge × ½ needle) were as follows: PC3 cells in male nu/nu athymic mice, C4-2B cells in male NOD-SCID mice, and MDA-MB-231 cells in female nu/nu athymic mice. When the lesion was identifiable by epifluorescence microscopy (10 to 14 days after implantation), a titanium DSFC with a glass window (diameter, 11.5 mm; Machine Shop, University Clinics Munich) was mounted adjacent to the tumor cell–bearing organ bone. The mTEBC was centered in the window system, and subcutaneous tissue deposited on the bone was carefully removed, avoiding any damage to afferent vessels and the periosteum, and the imaging procedure continued as described elsewhere (9, 10).

Statistical analysis

Statistical analysis was performed using GraphPad Prism 7.0a, as described in the figure legends. To test differences between two populations, the unpaired two-tailed Student’s t test was applied. To test the differences among more than two populations, a one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was performed. Unless stated otherwise, data are means ± SD.


Materials and methods

Fig. S1. Optimization of tissue-engineered ossicle generation.

Fig. S2. Longitudinal characterization of mTEBC maturation.

Fig. S3. Functional analysis of bone marrow in TEBC and bone marrow transplantation experiments.

Fig. S4. 3D organization and functionality of blood vessels in mTEBC.

Fig. S5. Topology and position of solid PC3 tumors in mTEBC.

Fig. S6. C4-2B and MDA-MB-231 tumor cell administration and monitoring by μCT, x-ray, histology, and iMPM.

Fig. S7. Analysis of the bone morphologies ex vivo.

Fig. S8. Imaging analysis of PC3-mediated osteolytic lesions by iMPM.

Fig. S9. Distribution of osteoclasts and Cat K+ cells within mTEBC.

Fig. S10. PIV analysis of response to zoledronate therapy monitored by iMPM.

Movie S1. 3D reconstruction of PC3 lesion in mTEBC.

Movie S2. 3D reconstruction of a PC3 osteolytic lesion in mTEBC.

Movie S3. Kinetics of a Cat K+ cell.

Movie S4. Kinetics of Cat K+ cell overview.

Movie S5. Kinetics of a Cat K+ cell after ZA treatment.

References (5559)


Acknowledgments: We thank S. Alexander for the maintenance and support with the multiphoton microscope; C. Johnston for the histological processing of the samples; B. Weigelin for providing samples with endomucin stainings; and C. Kingsley, K. Maldonado, K. Michel, and J. Delacerda for the support with μCT application. Funding: E.D. is supported by the Cancer Prevention and Research Institute of Texas (RP140482), the Prostate Cancer Foundation (16YOUN24), and The Rolanette and Berdon Lawrence Bone Disease Program of Texas. P.F. is supported by Netherlands Science Organization (NWO-VICI 918.11.626), the European Research Council (ERC-CoG DEEPINSIGHT, project no. 617430), the NIH (U54 CA210184-01), and the Cancer Genomics Center. D.W.H. is supported by the Australian Research Council Training Centre in Additive Biomanufacturing, the National Health and Medical Research Council of Australia, National Breast Cancer Foundation (IN-15-047), and World Cancer Foundation. The Genitourinary Cancers Program of the Cancer Center Support Grant (CCSG)–shared resources at MD Anderson Cancer Center was supported by NIH/National Cancer Institute award number P30 CA016672. N.N. is supported by the University of Texas MD Anderson Moon Shots Program. Author contributions: E.D., S.A., B.M.H., N.N., C.J.L., D.W.H., and P.F. designed the research. R.M.H., E.M.D.-J.-P., and D.W.H. provided the materials. E.D., S.A., B.M.H., S.F., E.M.D.-J.-P., and D.W.H. performed the bone generation experiments. E.D., S.A., and S.F. performed the iMPM experiments. E.D. and S.F. performed the ex vivo MPM experiments and μCT. E.D. performed the bone marrow transplant experiments and functional analysis of mTEBC. M.W.S performed the TRAP staining and bone histomorphometry. E.D., S.A., S.F., M.W.S., D.W.H., and P.F. analyzed data. E.D. and P.F. wrote the paper. All authors read and corrected the manuscript. Competing interests: The authors declare that they have no competing interest. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.
View Abstract

Navigate This Article