To image orthotopic (in the natural, mammary gland environment) breast tumors intravitally at high resolution for prolonged times, we developed a MIW that can be placed on top of the mammary gland of a mouse (Fig. 1a). The protocol for these animal studies was approved by the Institutional Animal Care and Use Committee for the Albert Einstein College of Medicine. The MIW consists of two plastic rings which form a mount for a glass coverslip. The mount has holes which facilitate suturing into the skin, whereas the glass coverslip assures the optimal working distance and refraction index for high resolution imaging (for more details on the equipment used for imaging, see Supplementary Fig. 1 online). While surgical dissection of the skin overlaying the imaging site allows for several hours of data to be collected and is usually a terminal procedure, imaging through the MIW extended the imaging time to multiple days (up to 21 days). Tumors with MIW implants did not show inflammation, or a change in growth and microenvironments scored at 1–9 days after the implantation procedure (Supplementary Fig. 2 online). In order to locate the same subpopulation of cells in each of the imaging sessions, reference points were required⁶. In the fast changing tissue topology of the tumor, the use of fixed reference points is limited, and therefore we used photoswitchable fluorescent proteins^{7, 8} as photomarkers of the cells of interest. These proteins represent a new group of GFP-like fluorophores which allow labeling and tracking of a single cell or a group of cells^{9, 10, 11, 12}. We stably expressed the photoswitchable protein Dendra2 in the metastatic breast cancer line MTLn3. Dendra2 resembles GFP in its spectrum prior to photoswitching, but exposure to blue light (e.g. 405 nm) can induce an irreversible red shift >150 nm in the excitation and emission spectra of the chromophore¹³. Following the photoswitch, the red fluorescence stably increases up to 250 fold both in vitro and (Fig. 1b) and in vivo (Fig. 1c), resulting in red/green contrast of up to 850 and allowing us to track cells marked in this way. Five days after photoswitching, the red fluorescence of the photoswitched cells is still 31 fold higher than the red fluorescence of non-switched cells, which enables us to recognize the highlighted cells in vivo for extended times after the photoswitch (Fig. 1d).

Regions containing one to hundreds of cells can be photoswitched and imaged through the MIW (Fig. 1e, Supplementary Movies 1 and 2 online, Supplementary Fig. 3 online). As cells in the tumor migrate and invade, the distribution of these cells relative to blood vessels and other tumor cells changed over time. By selectively photoswitching the fluorophore in a group of cells as demonstrated in Fig. 1e, the changes in distribution of cells in the tumor microenvironment was visualized. Twenty four hours after photo-switching, we recorded images of the non-photoswitched tumor cells (green), photoswitched cells (red), extracellular matrix (collagen was visualized with reflectance in Fig. 2a, or second-harmonic-generation in Supplementary Movie 3 online) and blood vessels (fluorescent dextrans were used for vessel labeling, see Supplementary Movie 4 online). Interestingly, some photoswitched regions showed dramatic migration and invasion of the surrounding microenvironment (Fig. 2a).

The tumor perivascular microenvironment (tumor tissue surrounding blood vessels) is enriched in tumor-associated-macrophages and extracellular matrix, which supports metastatic behavior, including inhibition of proliferation and stimulation of migration, invasion and intravasation^5,14. This suggests the existence of distinct mammary tumor microenvironments within the same tumor with different rates of invasion and intravasation. However, the long term implications of these observations required the ability to revisit distinct microenvironments inside the same mammary gland over a 24- hour period. In order to quantify invasion and intravasation within distinct mammary gland microenvironments, we photoswitched square regions (~300 cells) in different tumor microenvironments of the same orthotopically grown tumor (Supplementary Fig. 3 online), focusing on regions lacking (Fig. 2b) and containing (Fig. 2c) detectable blood vessels. The location of the photoswitched red cells was determined by acquiring Z-stacks of red and green images of the same regions at 0, 6 and 24 hours after photoswitching. In regions not containing detectable vessels (Fig. 2b) there was limited migration (Fig. 2d) and the number of cells increased (Fig. 2e), suggesting that this microenvironment does not support metastatic behavior. In contrast, cells in the vascular microenvironment (Fig. 2c) have infiltrated larger areas (Fig. 2d) and even migrated to sites outside of the field of view. Moreover, cells in this vascular microenvironment lined up along the blood vessel (Fig. 2c), with a concomitant decrease in the number of red tumor cells (Fig. 2e), and the appearance of red tumor cells in the lung (Fig. 2f). From these experiments, we conclude that cell behavior is determined by the surrounding microenvironment, and that the vascular microenvironment promotes invasion and intravasation of tumor cells.

While the existence of different tumor microenvironments has been reported previously ⁵, the quantitative analysis of such microenvironments in the same tumor with spatial and temporal resolution, is not possible with previous techniques. The combination of photoswitchable proteins with the MIW allowed for such analysis, since a distinct group of cells can be photomarked in any location in the primary tumor, and tracked over time without long term anesthesia. Furthermore, the high stability of Dendra2 enabled us to freeze fix the tissues and analyze them by microscopy without additional labeling. A limitation of Dendra2 is that a limited number of excitation wavelengths can be used. For example, since violet light causes a switch, DAPI stains should only be used after imaging other wavelengths as DAPI imaging would cause all green cells to then become red. Nevertheless, through the MIW, the formation of tumors from injection of cells in the mammary gland can be followed for days, which is not possible in surgically dissected areas. This also opens the possibility to study fluorescently-tagged proteins that have lethal effects if stably expressed, but can be studied in transiently transfected cells. For example, transiently transfected cells expressing membrane targeted GFP and injected into the mammary gland can be imaged with high resolution through the MIW (Supplementary Fig. 4a online). Although this would also be possible with the dorsal skinfold chamber⁴, imaging through the MIW allows studies of cell behavior in their physiological breast microenvironments and moreover, the MIW technology can be extended to tumors of transgenic origin, such as the MMTV-PyMT tumor model (Supplementary Fig. 4b online). Such tumor models allow the investigation of different stages of tumor progression¹⁵ in contrast to cell-line derived xenografts. The combined use of MIW and photomarking cells to revisit chosen subpopulations of cells is an important capability not only in tumor studies but any studies related to cell motility and morphogenesis. Visualization of infectious agents and immune responses, or the progression of chronic inflammation, are other examples of the potential applications of the technique described in this manuscript. This technique will also be helpful in monitoring of identification and proliferation of mammary stem cells, mammary gland growth and morphogenesis or testing artificial tissue heterotransplants.

Supplemental Figures

Movie 1

Movie 2

Movie 3

Movie 4