In this study, we explored the possibility of using a portable intravital fluorescence microscopy strategy to study the dynamics of circulating tumor cells in living subjects. Using non-invasive bioluminescence and fluorescence imaging, we established an experimental mouse model of metastatic breast cancer and showed that it leads to multiple metastases and the presence of CTCs in blood samples. We utilized a novel miniature intravital microscopy (mIVM) system and demonstrated that it is capable of continuously imaging and computing the dynamics of CTCs in awake, freely behaving mice bearing the experimental model of metastasis.
Besides other advantages described previously, 
the mIVM system presented here offers three major advantages over conventional benchtop intravital microscopes: (1) it presents a low cost alternative to IVM that is easy to manufacture in high number for high throughput studies (multiple microscopes monitoring multiple animals in parallel), (2) its light weight and portability allow for in vivo
imaging of blood vessels in freely behaving animals, (3) overcoming the requirement for anesthesia is a novel feature that allows us to perform imaging over extended periods of time, making it ideally suited for real-time monitoring of rare events such as circulating tumor cells. For many applications, mIVM might still be a complementary technique to IVM. However, for CTC imaging, mIVM presents clear advantages when compared to conventional IVM: mIVM is ideally suited for imaging CTCs as it fulfills the needs for (1) cellular resolution, (2) a large field-of-view, (3) a high frame rate and (4) continuous imaging without anesthesia requirements.
The current approach developed here to image CTCs presents several limitations. First of all, due to the current single-channel imaging capabilities of the mIVM, a green fluorescent dye (FITC-dextran) was needed in low concentrations in order to focus the microscope onto blood vessels, but hampered the visualization of eGFP expressing CTCs. Indeed, even though the eGFP expression in the cancer cells was very strong and sustained (), the signal-to-background ratio by mIVM imaging in vitro
was relatively low (< 2; ). Since the mIVM excitation source is based on a LED, this was expected. However, since a higher signal-to-background ratio was required in order to detect CTCs in the background of FITC-dextran circulating in plasma, we decided to label the cancer cells with a bright green fluorescent dye in addition to reporter gene expression which provided sufficient signal to background to image single 4T1-GL cancer cells both in vitro
() and in vivo
in the background of FITC-dextran (Fig. S2A
). However, even though we were able to image CTCs circulating in vivo
using the mIVM, there might be a possible signal-to-background issue limiting our capability to image all
the CTCs circulating in a vessel.
Labeling the cells exogenously with a fluorescent dye would not be amenable to the study of CTCs in an orthotopic mouse model of metastasis, where CTCs would spontaneously arise from the primary tumor. In order to avoid this issue, we envision two solutions. The first one, based on our current imaging setup requires waiting for 1–2 hours post - FITC-dextran injection to start imaging CTCs. Indeed we have observed that the FITC-dextran is almost completely cleared of blood vessels 2h-post injection (Fig. S2B
). The second approach rely on the next-generation design of mIVM setups capable of multicolor imaging, similarly to benchtop IVM systems. Using a dual-channel mIVM currently under development, the blood plasma could be labeled using a dye with different excitation/emission spectrums and circumvent the need for double labeling of the CTCs.
Another limitation of the mIVM is its penetration depth/working distance of max. 200 µm, 
allowing imaging through a 55–80 µm thick coverslip of superficial blood vessels of diameter up to 145 µm (the skin layer was removed as part of the window chamber surgery). For the 150 µm and smaller vessels – which are typical vessel sizes for IVM setups – our miniature microscope is capable of imaging the entire blood vessel’s depth. However in the case of the largest vessel of 300 µm diameter imaged here (), the penetration depth might have limited our capabilities to image all the CTCs circulating in this vessel. Therefore, the mIVM system is not intended to measure deep vessels, and should focus on smaller superficial blood vessels.
In this manuscript, we do not intend to image all the CTCs circulating in a mouse’s bloodstream, nor do we intend to image all the CTCs circulating in a particular vessel, as there might be depth penetration, fluorescence variability and signal-to background issues preventing us from recording all the CTCs events. Instead, we demonstrate here that we can image a fraction of the CTCs circulating in a particular superficial blood vessel. Assuming that the blood of the animal is well-mixed, the circulation dynamics of this fraction are representative of the circulation dynamics of CTCs in the entire blood pool. This assumption is common to all existing CTC detection methods that detect CTCs in a fraction of the entire blood pool (a blood sample, or an imaging time-window for in vivo flow cytometers) and/or detect a fraction of all the bona fide CTCs that are expressing a specific marker (e.g. EpCAM, CK, melanin, a fluorescent label). Since we are focusing on one small superficial blood vessel, we are not able to detect all the CTCs injected but only a small fraction of them, whose circulation dynamics we believe to be reflective of the dynamics of all the CTCs in this mouse model. In order to estimate this fraction and therebye estimate the sensitivity of our method, we estimated the total number of CTCs events detected over 2 hours: over 2 hours, we were able to detect an average of 2930 CTC events in a vessel, out of 1×106 cells injected, that is 0.29% of the CTCs injected. However, we believe that this number is not able to really reflect the true sensitivity of our method since the number of CTC events detected is dependent on (1) the size of the blood vessel imaged, (2) the relative location of the blood vessel in the circulation system, (3) the unknown fraction of CTCs circulating multiple times, that are therefore counted multiple times, (4) the unknown fraction of CTCs dying, (5) the unknown fraction of CTCs arresting/extravasating in organs. All these parameters require a complex mathematical model to relate the number of CTCs detected over a period of time to the actual sensitivity of our method at detecting CTCs.
As far as the specificity of our method is concerned, we are assuming here that only the cancer cells labeled with CFSE will generate a strong green fluorescence signal. We acknowledge that there could be some autofluorescence issues that would make tissue appear fluorescent as well. Therefore, we programmed our CTC detection algorithm to only count as a cell an object of the right fluorescence level harboring a circular shape of the right diameter (10–20 µm). Furthermore, any fluorescent object that is not moving at all over the imaging window (10 min – 2h) is going to be considered as background. We tested and optimized the algorithm on small imaging datasets before applying it to a larger dataset as presented on .
This study provides a proof-of-principle for mIVM imaging of CTCs in awake animals. However, we only explored the experimental model of metastasis, where 4T1 metastatic cancer cells are injected into the tail vein and allowed to circulate and seed metastasis sites. In this model, we imaged CTCs as they circulate during the first 2 hours post-injection. We were able to identify key features of the dynamics of CTCs: variations in speed and trajectory, rolling phenomenon when CTCs are in contact with the vessel edges (), half-life of CTCs in circulation in awake animals, representative fraction of CTCs still circulating 2 hours post-injection in awake animals (). Our measurements of the half-life of 4T1-GL cells (7-9 min) is in the same range than previous half-life measurements done on other metastatic cancer cell lines as measured with IVM methods. 
Similarly the rolling phenomenon we observed with the 4T1-GL cells has been demonstrated and studied in-depth in previous litterature. 
We were not able to image CTCs in the same mice around day 12, where the re-circulation of CTCs seems to happen because at that time, animals were showing signs of distress and needed to be sacrificed. It would be interesting to apply the mIVM method to a breast cancer model where the primary tumor is naturally shedding CTCs into the circulation. We envision that the mIVM will be particularly useful to explore the dynamics of CTCs in orthotopic metastasis models, since it has the ability to continuously monitor a blood vessel for sporadic and relatively rare CTC shedding events.
Our current mIVM setup is weighing less than 3 g, and is mounted on a titanium DSWC weighing less than 3 g, amounting the total weight to less than 20% of the mouse’s body weight (for a 30 g mouse). This setup would certainly be considered heavy for long term imaging of the superficial skin and smooth muscle on the back of the mouse. For longer imaging sessions, we envision that the setup would be placed on a cranial window chamber instead of the DSWC. Our collaborators, Ghosh et al., have previously demonstrated the feasibility of brain imaging using the mIVM in a cranial preparation. 
This preparation could be used similarly to image CTCs in the brain and alleviate the weight of the setup on the skin. Another strategy to offset the weight of the system is to use a counterweight system in the cage, similarly to the one used for the RatCAP head-mounted PET imaging system. 
We describe here how mIVM imaging allows enumerating CTCs as they circulate in a mouse’s bloodstream. This in vivo CTC enumeration method offers several advantages over in vitro interrogation of CTCs in blood samples. First of all, as the imaging is relying on the endogenous expression of the eGFP protein by the CTCs, there need not be reliance on a given CTC marker for CTC imaging or capture. Furthermore, the blood volume that can be analyzed by continuously imaging a blood vessel can potentially be much larger than that of a blood sample, enabling the potential capture of even more rare events. Assuming a blood flow speed of 1 mm/sec in a blood vessel of 100 µm diameter (typical parameters measured in our mIVM experiments), we estimated that we are able to analyze 28 µL of blood per hour. If we perform continuous imaging over 24 hours, we will be able to sample 672 µL of blood. Over 1 week, we will be able to sample over twice of the total mouse blood volume (~2 mL), versus 5% as allowed per veterinary guidelines for blood sampling. If we image larger vessels with higher frame rates, we will be able to achieve even higher blood volume analysis. The current mIVM system will also be particularly useful to image tumor cells as they are leaving a primary tumor and entering the bloodstream – this can be achieved by implanting a primary tumor at the site of the dorsal skinfold window chamber. This strategy will also increase the probability of observing naturally occurring CTCs.
Previously, other in vivo
CTC imaging methods have been used to interrogate CTCs in living animals, namely in vivo
and multiphoton intravital flow cytometry 
. Both techniques are benchtop systems and have been able to detect single CTCs as they are flowing in a mouse’s ear blood vessels. Multiphoton microscopy harbors much higher signal-to-background ratios (~22) than mIVM for detecting dye-labeled cells (~2). 
However, since both methods are based on time-consuming laser-scanning, they had to rely on a one-dimensional line scanning through a slit in a blood vessels in order to detect fast flowing CTCs. Our mIVM method has the advantage of combining high speed detection (up to 100 Hz) and two-dimensional imaging. In our mIVM setup, an image of the detected CTCs can be formed, to confirm that the signal detected is indeed coming from CTCs. Moreover, thanks to its miniaturization, our mIVM system is the first setup we know of allowing to image CTCs in awake, freely-behaving animals. Eventual use of these and related devices to monitor CTCs in humans (e.g., for monitoring for tumor recurrence) may also be possible by combining these devices with implantable patches that periodically inject fluorophores that target CTCs for continuous monitoring strategies.
To shed light on the potential clinical relevance of CTCs, complex questions about tumor metastasis need to be answered: (1) how and when a breast tumor infiltrates the bloodstream, (2) how inefficient the process of metastasis is for a particular carcinoma and (3) which properties of CTCs enable them to successfully colonize distant organs. Here we have demonstrated that our new mIVM system is capable of continuously imaging blood vessels for CTCs in awake animals. Our system has the potential to shed light on some of the fundamental questions raised above. We are currently exploring the possibility of using an optoelectronic commutator for long term use of the mIVM system in awake freely moving subjects as well as developing a real-time analysis algorithm that will only keep and store the data corresponding to CTCs events. This method will enable the in vivo long term study of CTCs dynamics in orthotopic mouse models of metastasis.