Real-Time Monitoring of Cancer Cells in Live Mouse Bone Marrow

Disseminated tumor cells in the bone marrow environment are the main cause of systemic metastasis after curative treatment for major solid tumors. However, the detailed biological processes of tumor biology in bone marrow have not been well defined in a real-time manner, because of a lack of a proper in vivo experimental model thereof. In this study, we established intravital imaging models of the bone marrow environment to enable real-time observation of cancer cells in the bone marrow. Using these novel imaging models of intact bone marrow and transplanted bone marrow of mice, respectively, via two-photon microscopy, we could first successfully track and analyze both the distribution and the phenotype of cancer cells in bone marrow of live mouse. Therefore, these novel in vivo imaging models for the bone marrow would provide a valuable tool to identify the biologic processes of cancer cells in a real-time manner in a live animal model.

Disseminated tumor cells in the bone marrow environment are the main cause of systemic metastasis after curative treatment for major solid tumors. However, the detailed biological processes of tumor biology in bone marrow have not been well defined in a real-time manner, because of a lack of a proper in vivo experimental model thereof.
In this study, we established intravital imaging models of the bone marrow environment to enable real-time observation of cancer cells in the bone marrow. Using these novel imaging models of intact bone marrow and transplanted bone marrow of mice, respectively, via two-photon microscopy, we could first successfully track and analyze both the distribution and the phenotype of cancer cells in bone marrow of live mouse. Therefore, these novel in vivo imaging models for the bone marrow would provide a valuable tool to identify the biologic processes of cancer cells in a real-time manner in a live animal model.
Keywords: two-photon microscopy, intravital imaging, bone marrow microenvironment, tumor cell dormancy, cancer cell inTrODUcTiOn Early systemic metastasis is a major feature of major cancers even after margin-negative resection of a primary cancer lesion (1,2). Circulating tumor cells derived from primary cancer lesion can be disseminated to secondary organs including bone marrow blood, lymph node, and distant organ via blood vessels or lymphatic channels and can cause early systemic recurrence regardless of definite treatment (3,4). The majority of cancer-related deaths are due to the involvement of metastatic tumors originating from disseminated cancer cells (5,6). Contrary to other malignancies showing a dormant tumor phenotype in secondary organs, most notably in breast cancer, it has been known that aggressive cancer types such as pancreatic cancer can overcome easily or bypass suppressive microenvironment compared to other malignancies having a relatively long dormant period (3,(7)(8)(9). Therefore, the mechanism of an awakening or activation of cancer cells in secondary organs should be investigated in order to prevent early systemic metastasis even in resectable cancer. The tumor biology of cancer cells in the bone marrow environment, the main site of minimal residual disease after curative cancer treatment, should be investigated in order to identify the specific process of systemic metastasis in most solid cancers (3,10,11). However, the lack of an in vivo experimental model for cancer cells in the bone marrow environment has been a major barrier to exploring the initial process of systemic metastasis (4,12,13). The aim of this study is to establish intravital imaging model of cancer cells in the bone marrow environment.

Mouse imaging Models for Two-Photon intravital imaging
To observe the bone marrow environment effectively, we established two imaging window models, a calvarial window in the skull and a dorsally transplanted femur bone marrow window. A natural imaging window for calvarial bone marrow in skull bone had been previously established. Meanwhile, however, the dorsally transplanted femur bone marrow model was newly created to investigate the exact biology of cancer cells in bone marrow. Whole surgical procedures were performed carefully by a single hepatobiliary surgeon who has abundant experience in microsurgery. The operative procedures for each intravital imaging model were performed as described in the following sections.

establishment of the imaging Window for calvarial Bone Marrow
After anesthesia of fluorescence expressing mice (CXCR1-GFP or LysM-GFP mouse) using Zoletil ® (Virbac Korea, Seoul, Korea) via intraperitoneal injection, mice was placed onto a stereotactic heating plate (Live Cell Instrument, Seoul, Korea) to maintain body temperature. The scalp was removed with scissors with a 1.5-cm radius ( Figure S1A in Supplementary Material), and then, acrylic resin was applied around the exposed skull area ( Figure  S1B in Supplementary Material). The fixation ring was attached to the pre-applied acrylic resin and then assembled with a stereotactic head fixation device (Live Cell Instrument, Seoul, Korea) attached to the heating plate ( Figure S1C in Supplementary Material).

Preparation of Donor Mouse and Processing of Femur Bone graft
After harvesting cultured cancer cells in vitro using a 100-mm dish, 1-ml syringes with a 31-G needle were prepared with 0.5 ml of injectable normal saline mixed with a predefined number of cancer cells. The cancer cells were injected into the tail vein of a donor mouse (fluorescence expressing mouse: CX3CR1-GFP or LysM-GFP mouse) for femur bone transplantation with a prefilled 1-ml syringe without anesthesia. After a period of mouse breeding in an animal facility for mice over 1-7 days after the injection, the femur bone of the donor mouse was harvested carefully after euthanasia in a CO2 chamber ( Figure 1A). The extracted femur bone of the donor mouse was immediately processed for long bone transplantation. First, both epiphyses of the femur bone were excised using a surgical scalpel blade (No.10) and one side of the femur bone cortex

establishment of the imaging Window for Transplanted Femur Bone Marrow in recipient Mice
During the femur bone preparation in the donor mouse, formation of the imaging window for the dorsal chamber in recipient mice (C57BL/6 or BALB/c nude mice) was simultaneously carried out in keeping with the progress of femur bone harvesting and processing in donor mouse. Recipient mice were anesthetized by Zoletil ® injection into the intraperitoneal cavity and placed on a heating plate to maintain body temperature. For stable installation of a dorsal chamber kit (SM100, 27 mm titanium, APJ trading, USA), the back skin was sutured and retracted in an upward direction. The unilateral back skin with round shape was removed by fine micro-dissecting spring scissors along the position for the microscope cover glass (GL100, 12 mm, 0.13 mm thick, APJ trading, USA) for intravital imaging ( Figure S2C in Supplementary Material). The dorsal chamber kit was assembled into the retracted back skin of recipient mice, except for cover glass equipment. Immediately after completing femur bone processing of the donor mouse, the exposed bone marrow side of the femur graft was placed onto the back skin of a recipient mouse after being excised in a round shape for cover glass positioning. After delicate dropping of PBS around the femur bone graft, the cover glass was placed and fixed to the dorsal chamber (Figures S2B,C in Supplementary Material).

Two-Photon intravital Microscopy
Mice were anesthetized using via intraperitoneal injection of Zoletil at a dose of 30 mg/kg during imaging procedures. Long-term anesthesia was performed using the inhalation agent isoflurane.
A staging system (manual XY stage and microscope mounting plate) was set up using Live Cell Instrument Korea (Seoul, Korea), and two-photon microscopy (LSM7MP, Carl-Zeiss, Germany) was used for imaging data generation. Zen software (Carl-Zeiss) was used for image acquisition and basic image analysis. For two-photon excitation, light of 880-900 nm wavelength was used for imaging green, red, and second harmonic generation. Images were acquired at a resolution of 512 × 512 pixels using step sizes of 1 µm to a depth of 30-50 µm every 30-60 s.

imaging Data analysis
Fiji/ImageJ software was used for image analysis and basic image processing. IMARIS version 7 (Bitplane, USA) and Volocity software (PerkinElmer, USA) were used for 3D and 4D imaging data analysis.

Flow cytometry analysis
Chronological flow cytometry analysis at baseline (control) and 1 and 7 days after cancer cell injection via the tail vein. Injectable saline without cancer cells was injected to control mice. The acquisition of bone marrow for flow cytometry was performed by aspiration from bone marrow of the femur bone at the day of injection (control) and 1 and 7 days after injection via the tail vein. Then, mouse bone marrow was obtained 1 and 7 days after cancer cell injection. Harvested bone marrow cells were washed with PBS and fixed with 4% paraformaldehyde for 10 min. The fixed cells were chilled on ice and permeabilized with 90% methanol for 30 min. Then, 1 × 10 6 cells per experimental condition were aliquoted, washed, and resuspended in 100-µl solutions of fluorochrome-conjugated primary antibodies against DAPI (564907, BD Biosciences, USA), MHC class II (550750, BD Biosciences, USA), and CD11b (553311, BD Biosciences, USA) at the manufacturer's recommended concentrations, and incubated for 1 h. For isotype control, fluorochrome-conjugated rabbit IgG was used at the same concentration. Cells were washed, resuspended in PBS, sorted by fluorescence, and analyzed using a FACSAria cell sorter.

statistical analysis
Graphical data generation and basic statistical analyses were performed using GraphPad Prism 6 software. Comparisons between the separate groups were conducted using the Mann-Whitney U test for non-parametric test of continuous variables to calculate statistical probability of non-parametric test. p-Values less than 0.05 were considered statistically significant.

resUlTs anD DiscUssiOn
To monitor the entire biological process of the cancer cells in the bone marrow in real-time manner, we established two different intravital imaging models to track cancer cells in the bone marrow environment of live mice by employing specific windows for the calvarium and transplanted femur, respectively. An imaging window for indigenous calvarial bone marrow in the mouse skull ( Figure S1 in Supplementary Material) was adopted with a stable fixed-staging system (16). In addition, a novel imaging window to observe bone marrow in a transplanted femur was contrived using a dorsal chamber (17) (Figure 1A). To investigate whether our newly designed femur bone transplantation model works for engraftment, we transplanted MCF7-RFP cells injected bone to healthy LysM-GFP mice. Using two-photon intravital microscopy of the MCF7-RFP-carrying bone marrow transplanted LysM-GFP mice, we confirmed that the transplanted femur bone marrow was successfully engrafted as a viable live tissue after 7 days from the implantation, and intact vascular connections with a vascular network system were formed at the fascia layer of the recipient mouse ( Figure 1B; Videos S1 and S2 in Supplementary Material). The transplanted bone marrow tissue was viable at all-time points at days 1-30 after the surgery. Therefore, it was feasible to perform intravital imaging of the interaction of cancer cell with leukocytes from the implanted bone marrow in the recipient mice. Interestingly, we found that cancer cells were mainly located in the perivascular area in the bone marrow environment after intravenous injection of cancer cell lines. A total of 5 × 10 5 of cancer cells from human (MCF7 and AsPC-1) and mouse (4T1, MMT060562, SL4, Lewis Lung Carcinoma, and Pan02) cancer cell lines were used. From 4D image analyses (3-dimensional image analyses in time lapse manner), the migration trajectory of cancer cells was significantly shorter than those of primary cells in the bone marrow environment. The relative location of cancer cells from the blood vessel was closer to blood vessel. Migration length of cancer cells in an hour was statistically shorter than the patrolling macrophage crawling around the vessel wall and resident macrophage in the bone marrow parenchyma of CX3CR1-GFP mice (Figures 1C,D; Video S3 in Supplementary Material). We also confirmed similar localization and migration pattern of other cancer cells closer to blood vessels, when other cancer cells from human (AsPC-1) and mouse (4T1, MMT060562, SL4, Lewis Lung Carcinoma, and Pan02) cancer cell lines were used for the same experiment (data not shown). When we used red-labeled Pan02 cells (Pan02-CMPTX) to 0.94 ± 0.28 µm in MCF7-RFP, **p < 0.01). The Mann-Whitney U test was used to compare two groups. A representative data were shown from independently repeated experiments three times. (e) Chronological flow cytometry analysis showing control and 1 and 7 days after cancer cell injection via the tail vein in a mouse that was not used for intravital imaging. Injectable saline without cancer cells was injected to control mice. The acquisition of bone marrow for flow cytometry was performed by the aspiration from bone marrow of the femur at the day of injection (control) and 1 and 7 days after the injection via tail vein. MHC II high cells and MHC II low CD11b + Ly6C hi Ly6G − cells were significantly increased in day 7. The Mann-Whitney U test was used to compare two groups (cell count of MHC II high ; 39.24 ± 8.84% in control vs. 67.26 ± 12.45% in day 7, MHC II low CD11b + Ly6C hi Ly6G − ; 35.45 ± 8.74% in day 1 vs. 53.48 ± 17.31% in day 7, *p < 0.05). Data were averaged from independently repeated experiments three times (see Figure S5 in Supplementary Material). (F) Qualitative analysis of relative fluorescence of MCF7-RFP with or without GM-CSF injection before injection and 1, 4 days after injection in dorsally transplanted bone marrow model.  (Figures 1E,F; Videos S7 and S8 in Supplementary Material). Two-photon intravital imaging of cancer cells in calvarial bone marrow also verified that cancer cells were in active contact with resident macrophages in the bone marrow environment of CX3CR1-GFP mice (Figure 2A; Video S9 in Supplementary Material). Even after serially sustained contact with CX3CR1-GFP or LysM-GFP expressing cells, the cancer cells were consistently observed to be viable without undergoing immune clearance. Interaction between cancer cells and CX3CR1-GFP or LysM-GFP expressing cells were significantly decreased 24 h after i.v. injection of cancer cells in both LysM-GFP and CX3CR1-GFP mouse, compared to that within 1 h after i.v. injection of cancer cells (Figure 2B). This difference in interactions between cancer cells and stromal cells in the bone marrow may result from the fact that temporal change of local innate immunity toward cancer cells might be the reason of cancer dormancy in the bone marrow environment. Further evaluations of this finding should be done to confirm these observations. Flow cytometry analysis also revealed that the subpopulation of myeloid lineage changed over time after cancer cells entry into the bone marrow environment (Figure 2E; Figures S5A,B in Supplementary Material).
The effects of chemotherapeutic agents on dormant cancer cells in the bone marrow environment are not yet known because of the absence of a suitable experimental model. Therefore, an interventional experiment was performed using our novel model to investigate the impact of a chemotherapeutic agent on cancer cells, as well as on the bone marrow environment. After injection of the chemotherapeutic agent gemcitabine, the main drug used for first or second-line chemotherapy of several solid cancers, such as pancreatic cancer, ovarian cancer, lung cancer, and breast cancer, we observed that the cell number and movement of various bone marrow components were significantly decreased. However, the number of cancer cells remained consistent; cancer cells appeared to be less affected by the chemotherapeutic agent than other normal cells in bone marrow (Figures 2C,D; Videos S10-S12 in Supplementary Material). These initial findings from the intravital imaging model indirectly suggest that current clinical policy regarding adjuvant chemotherapy might be creating a paradoxical effect, in which chemotherapeutic agents induce the relative activation of cancer cells in suppressive microenvironments, such as cancer cells in the bone marrow environment. A confirmative study with a robust experiment design is mandatory to prove these initial findings. Another example of an interventional approach using our model of the bone marrow environment showed that the morphology and movement of immune cells and stromal cells in the bone marrow environment were significantly increased immediately after i.v. injection of granulocyte-macrophage colony factor (GM-CSF). In coculture with murine cancer cell lines, active movement and sudden disappearance of the cancer cells from bone marrow environment was observed in a small portion of the engulfed tumor cells in the perivascular area of the mouse bone marrow. Bone marrow stimulation by GM-CSF was a significant factor for reactivation of cancer cells in the mouse bone marrow environment (Figure 2F). These findings indicate that GM-CSF administration due to neutropenia after chemotherapy could unintentionally cause the activation of cancer cells in the bone marrow environment. Further investigations into the detailed mechanisms of this paradoxical phenomenon should be performed to address these concerns.
Bone marrow is a representative depot for the distribution of cancer cells clinically known as minimal residual disease (18)(19)(20)(21). Nonetheless, only a few studies on cancer cells in the bone marrow environment have shown direct experimental evidence of their biology. In this study, we have successfully established two imaging models for intravital observation of the bone marrow in both the skull calvarium and femur. These intravital imaging models can elucidate the detailed biology of cancer cells, including their distribution and biological behavior in response to specific stimulation as well as reactivation phenomenon. Femur bone marrow transplantation from a cancer cell-injected mouse to the dorsal area of a healthy mouse was first developed to identify the biology of the cancer cells in the bone marrow using two-photon intravital microscopy. The combination of the calvarial window for observing indigenous healthy bone marrow and the dorsally transplanted cancer cell-bearing bone marrow model could capture the entire process, from the bone marrow dissemination of cancer cells to systemic metastasis, as novel experimental animal models using intravital imaging. These findings are consistent with current theories of cancer cells behavior in the bone marrow environment. Additionally, our results verified that even after constant exposure to neutrophils and macrophages, the cancer cells were robustly viable in the bone marrow environment. This phenomenon provides new insights for cancer immunology in the bone marrow environment favoring immune tolerance for cancer cells. Chemotherapeutic agents caused significant damage to in situ bone marrow components but did not influence the viability of cancer cells in the bone marrow environment. The presence of cancer cells in bone marrow in patients has been strongly suggested to be a risk factor for metastasis (22)(23)(24). However, clinical implications of cancer cells in bone marrow-targeting strategies for preventing metastasis could only be addressed when detailed mechanistic analysis of cancer cell biology in bone marrow was undertaken in preclinical models (4,25,26). It is, therefore, inevitable that an in vivo bone marrow cancer cell model should be developed for real-time monitoring and tracking (27).
There are several limitations in this study regarding insufficient experimental validations for specific biologic process and limited scope to perivascular niche for the cancer cells in the bone marrow environment. Although the experiments of intravital imaging were conducted as far as possible from the interface of implantation, the method of bone marrow transplantation, which can create artificial wound repair-like process, can also affect the result of experiments using the femur bone transplantation model. The experimental set-up and initial validations, however, were primary milestone in this study. Therefore, vigorous experimental validation and subsequent functional study should be followed. Taken together in this study, we showed that the specific biologic process of cancer cells in bone marrow can be elucidated in high resolution using two-photon intravital imaging. This novel method offers a very useful tool for gaining new insights into cancer cell biology in bone marrow, especially, the identification of both dormancy and reactivation of cancer cells in bone marrow in vivo.     showed restricted proliferation compared to active proliferation in monoculture and coculture with NIH/3T3 cells in the Mann-Whitney U test. *p < 0.05. See Videos S10-S12 in Supplementary Material. Representative data were shown from independently repeated experiments three times.

FigUre s4
| Western blot analysis of ERK/p-ERK/p38/p-p38 in mouse cancer cell line monocultures vs. coculture with mouse bone marrow. (a) Western blot to confirm of cancer dormancy at the molecular level. Various cancer cell-lines originated from a C57BL6 mouse (MMT060562, LLC, Pan02, SL4) were cocultured with mouse bone marrow stromal cells that were aspirated from the femur bone of a C57BL6 mouse. Blotting images represent the relative protein expressions of p-ERK/ERK and p-p38/p38 in cancer cell-lines. (B) Quantitative analysis of western blot images with or without coculture with mouse bone marrow by the Mann-Whitney U test. Expression ratio of phospho-ERK with phospho-p38 between monoculture and coculture with BM were significantly different in MMT060562 and SL4 cell lines. *p < 0.05. The mean values were quantified from pooled experiments conducted using different lysates from independent samples three times. Injectable saline without cancer cells was injected to control mice. The acquisition of bone marrow was performed by the aspiration from bone marrow of the femur bone at the day of injection (control) and 1 and 7 days after injection via the tail vein. (B) Quantitative analysis of temporal changes for myeloid derived suppressive factors in myeloid lineage subpopulation MHC II lo CD11b + Ly6C hi Ly6G − . Relative expression of Arg-1 was significantly increased in days 1 and 7 compared to control (relative expression of Arg-1; 1.62 ± 0.73 in day 1 vs. 2.21 ± 0.48 in day 7, *p < 0.05, ***p < 0.001). The Mann-Whitney U test was used to calculate the statistical significance. The mean values were quantified from independently repeated experiments three times.
ViDeO s2 | 3D structural analysis for vascular connections between the donor bone marrow and recipient fascia layer.