Lightsheet-based flow cytometer for whole blood with the ability for the magnetic retrieval of objects from the blood flow

: Detection and extraction of circulating tumor cells and other rare objects in the bloodstream are of great interest for modern diagnostics, but devices that can solve this problem for the whole blood volume of laboratory animals are still rare. Here we have developed SPIM-based lightsheet flow cytometer for the detection of fluorescently-labeled objects in whole blood. The bypass channel between two blood vessels connected with the external flow cell was used to visualize, detect, and magnetically separate fluorescently-labeled objects without hydrodynamic focusing. Carriers for targeted drug delivery were used as model objects to test the device performance. They were injected into the bloodstream of the rat, detected fluorescently, and then captured from the bloodstream by a magnetic separator prior to filtration in organs. Carriers extracted from the whole blood were studied by a number of in vitro methods. indistinguishable in the camera images. The second one works well for sparse separated objects, or when the number of objects in the aggregate can be evaluated by its size. Effective magnetic separation of objects from the PBS buffer stream has been shown with 0.3 T magnetic field near the tip of the field concentrator at a flow rate of 20–50 µL/min. The developed system allowed us to monitor fluorescently-labeled microcapsules inside the flow cell shunting two large blood vessels of a rat limb after their injection into the carotid artery. The speed of blood flow measured by the DOCT was 1.4 ± 0.8 cm/s inside the shunting polymer tube; 2.1 ± 1.2 and 1.1 ± 0.6 cm/s inside the artery and the vein, respectively. We have shown the separation of about 4 × 10 5 microcapsules in vivo from blood by a 0.3 T permanent magnet with a concentrator. The extracted objects were measured afterward by imaging flow cytometry, CLSM, and AFM methods. An unexpected effect of reducing the microcapsule shell thickness after passing through the bloodstream was found along with the adsorption of a large amount of debris on it.


Introduction
Direct analysis of rare pathogens and embolus in the blood flow should provide extremely important capabilities for the next generation of diagnostic systems. This task is essential for both diagnostic purposes and pre-clinical testing of modern therapy methods. Circulating tumor cells (CTCs) in the case of metastatic cancers [1,2]; microorganisms [3] and protozoa [4] during blood infections; atherosclerotic plaques [5] and red blood cell clots during cardiovascular diseases [6] can be detected by direct measurement of patients' blood. But in the early disease stages, when the detection of such objects is extremely important, the possibility of doing so is utterly low. Thus, the major challenge of modern diagnostics is the detection of such rare events in the patient's whole blood. In therapy, it can be applied for thorough comprehensive pre-clinical testing of novel methods, especially based on targeted drug delivery platforms.
Currently, several approaches for in vivo detection and separation of such objects have been developed. At first glimpse, a search for rare cells in the blood flow and investigation of processes between numerous drug delivery carriers have completely different engineering requirements. However, the sensitivity of the system designed to detect single objects in a large amount of blood cells provides necessary reliability for the complex testing of drug delivery systems. On the other hand, a large number of circulating carriers for targeted drug delivery with well-defined parameters can be detected much more easily than single unique objects with unknown features. Moreover, testing of the system using well-characterized objects greatly simplifies its development.
Large amounts of blood should be analyzed in order to capture CTCs at the early stages of cancer, or specific pathogen germs at the beginning of acute blood infection. According to the literature, the amount of CTCs contained in 1 mL of whole blood can range from a few to several hundred objects depending on the cancer stage and correlates with the disease progression [7,8]. Meanwhile, the typical pathogen germs concentration during sepsis is approximately 10 3 colony-forming units (CFU)/mL in blood samples [9]. Moreover, no objects of interest may be captured into the blood sample, which conventional size is 7.5 mL [10,11], as there are fluctuations in the rate of objects appearance in circulation [12,13].
In vivo cytometry systems overpass this restriction by the analysis of the total volume of whole blood. In combination with photoacoustic microscopy, such systems can be applied for early malaria diagnostics [4], detection of circulating melanoma cells [14,15], and other kinds of emboli [16]. However, their functioning in most cases is provided by the characteristic features inherent in the objects of interest. For instance, such features can be a high optical absorbance and a specific spectrum for circulating melanoma cells, providing their in vivo photoacoustic flow cytometry detection. In other cases, such features should be created by cells labeling directly in the blood flow in vivo with following photoacoustic [17] or fluorescent [10,18,19] detection.
Another possible way to search for rare objects in a large volume of blood is based on the imaging flow cytometry. Various approaches have used 3D detection capabilities for imaging flow cytometry and have reduced the need for hydrodynamic focusing, which requires diluting whole blood with the sheath fluid. They include lightsheet scheme of SPIM-Fluid device [20], line-field quantitative phase microscopy with digital refocusing [21], orthogonal light-sheet scanning illumination with spatiotemporal transformation detection producing 3D cell image reconstruction from a cameraless single-pixel photodetector readout [22], virtual-freezing fluorescence imaging flow cytometry [23], and other schematics.
Despite all the advantages of non-invasive in vivo flow cytometry, it has one significant drawback -the inability to retrieve the analyzed objects from the bloodstream for further additional research. The most mature approach to solve this problem is the extracorporeal separation of such objects from the flow of whole blood. Magnetically-labeled objects [24][25][26] can be trapped from the blood flow using the magnetic separator [27][28][29][30] integrated with the flow cell. Systems for the magnetic separation of objects from the patient's blood pumped by an additional peristaltic pump are being actively developed now [31]. The main purpose of such devices is the cleaning of blood from pathological cells labeled by functionalized magnetic beads.
Our approach is based on a combination of two methods. The first is the visualization of fluorescent objects in undiluted blood performed in the external flow cell by the lightsheet system based on SPIM-Fluid design. The second is the separation of magnetically-labeled objects using a permanent rare-earth magnet with concentrator, similar to the one used in our previous studies [25,26]. This approach combines the advantages of an in vivo flow cytometer, which analyzes the entire blood volume with detection of rare objects, and a magnetic cell sorter, which captures magnetically-labeled objects from the bloodstream. In addition, this technical solution does not require focusing the object flow in the caustic region of the laser and focusing the camera, which allows working directly with undiluted blood and performing flow cytometry measurements in a flow cell that connects two blood vessels of a laboratory animal. The external flow cell eliminates the scattering and attenuation of laser radiation by biological tissues and their autofluorescence compared to in vivo cytometry. Technology approbation was performed both in vitro and in vivo using the biodegradable fluorescently-labeled magnetic capsules as model objects.

Device design
The optical design of the device is based on a simplified variant of the SPIM-Fluid imaging flow cytometer [20]. The major features are shown in Fig. 1. The device is composed of three main units: • a light-sheet illumination subsystem that consists of CW diode lasers: 488 nm 60 mW

Synthesis of citrate-stabilized magnetite nanoparticles
Magnetite nanoparticles (MNPs) were synthesized by chemical precipitation method from diand trivalent iron salts into an alkaline media as described before [32,33]

Synthesis of FITC-labeled BSA
First, 160 mg of BSA were dissolved in 40 mL of 0.1 M PBS buffer (pH 8). Afterward, the solution of FITC in ethanol (5 mg/mL) was prepared. 40 mL of BSA solution was added to 5 mL of FITC alcoholic solution under gentle stirring and the mixture was left with constant steering under 4°C in the dark for 12 h. Finally, freshly prepared FITC-conjugated BSA was dialyzed for 3 days in deionized water.

Preparation of fluorescent magnetic microcapsules
Spherical porous CaCO 3 particles (average size of ∼0.8-1.2 µm) with the co-precipitated MNPs were synthesized similarly to that described earlier [34]. In brief, 0.4 mL of 0.5 M CaCl 2 and 0.5 M Na 2 CO 3 solutions were injected into 4 mL of glycerol and 2 mL of MNP suspension under vigorous stirring. One hour later stirring was stopped, the particle suspension was centrifuged and washed two times by deionized water. Layer-by-Layer assembly was used to obtain microcarriers. Microcapsules were formed by the sequential adsorption of 1 mL of pArg (1 mg/mL), DS (1 mg/mL), and BSA-FITC (1 mg/mL) from 0.15 M NaCl solutions onto the surfaces of the spherical CaCO 3 cores. The cores were then dissolved with EDTA (pH 7.3). The obtained shell structure was [MNPs/pArg/DS/{pArg/BSA-FITC} 2 /pArg/DS].

In vitro experiments
During in vitro study, the fluorescent magnetic capsules suspended in PBS or in the rat blood (the ratio of blood to capsule PBS suspension was 3:1) were used as test samples. The flow cell was connected to two tubes, one connected to a constant flow syringe pump and the other to a waste container.

Laboratory animal experiment in vivo
Animal care and all experimental procedures were carried out in accordance with the "Guide for the Care and Use of Laboratory Animals" [35]. For general animal anesthesia, the mix of Zoletil solution in the concentration of 40 mg/kg (Virbac SA, Carros, France) and 2% Rometar in the concentration of 10 mg/kg (Spofa, Czech Republic) was injected intraperitoneal. The experimental protocol was approved by the Committee for the Care and Use of Laboratory Animals at Saratov State Medical University (wide approval no. 5 on 29.12.2018). The study was carried out on adult male white mongrel rats with an average weight of 200-250 g.
The microcapsule suspension was injected into the branch of the carotid artery (about 0.5 mm in diameter) through a thin polyethylene tube (inner diameter 0.28 mm, outer diameter 0.61 mm, Portex, Smiths Medical International Ldt., UK). Prior to the implantation, the tube was filled with isotonic NaCl (0.15 M) aqueous solution. Before injection, the capsules were also transferred into 0.15 M NaCl aqueous solution and diluted 3 times to prevent blockage of circulation due to their aggregation. The total injection volume was 20 µL.
Evaluation of blood flow velocity The blood flow velocity in large vessels of laboratory rats was measured by the optical coherence tomography (OCT) method using Thorlabs PSOCT1300SS system (Thorlabs, USA) in the Doppler measurement mode. The transverse and axial resolution of the system were 20 and 5 µm, respectively. The A-scan rate of the system was set to 16 kHz. To ensure sufficient spatial sampling for both veins and arteries, 2048 A-scans were performed over a 5 mm section. The angle between the velocity vector and the probing beam was set to 12 degrees. The phase difference between two adjacent A-scans was used to calculate blood flow values. Due to the nature of arctan function, reconstructed phase values were limited by the [−π, π] interval. A built-in NumPy function unwrap was used to unwrap the phase values thus making them usable for blood flow velocity calculation. The blood flow velocity measurement was performed on the femoral artery and vein of the right hind paw of the laboratory animal. In the area of femoral vessels, a small skin dissection was performed to provide better visualization quality (see Fig. S1 in Supplement 1).
Magnetic separation from the blood flow For the in vivo experiment, the input tube was connected to the femoral artery and output to the femoral vein of the rat's hind paws by catheters with an internal diameter of 0.28 mm (Fine Bore Polythene Tubing, Portex, USA). All inner surfaces of the flow channel were treated with heparin to prevent blood coagulation. Before the connection of catheters with blood vessels, all volume of the fluidic system was filled with the physiological salt solution to prevent air embolism.
In the first step, the flow cell was connected to the artery to ensure the blood pressure is enough for the blood to flow through the channel. Next, the output tube was connected to another paw's vein to provide the blood flow through the flow cell. Before and during the experiment, small portions of the heparin solution were injected through an implanted catheter into the left carotid artery to prevent thrombosis. Subsequently, the same catheter was used to inject capsules' suspension.
The capture of magnetic FITC-labeled microcapsules from the bloodstream was performed by 0.3 T magnet during 10 min. Then, the flow cell was disconnected from the rat's vascular system, the magnet was removed, and trapped capsules were collected into the tube by washing the flow channel with PBS. Finally, the 50-times diluted sample of blood containing captured capsules was obtained and studied using the imaging flow cytometer Amnis ImageStream X MkII (Luminex, USA).

Data processing
2.5.1. Detecting average fluorescence in the lightsheet plane The first and one of the simplest detection methods is based on the calculation of the average fluorescence within the region of interest (ROI) representing a lightsheet plane crossing the square capillary. If a large number of objects are simultaneously present in the image frame as a non-resolvable cluster, the average fluorescence value fairly well represents the amount of the fluorescent material in this cluster. For instance, this method performs well for analyzing fluorescent signals in the region directly alongside the tip of the magnetic concentrator in the case of numerous fluorescent magnetic objects in the sample flow. In this case, if there is a large accumulation of indistinguishable objects, the intensity of the fluorescent signal within the ROI can give an estimate of their number even if the fluorescent signal is nonlinearly dependent on the number of objects.

Analysis by computer vision methods
The second detection approach is based on the identification of object contours against the background. The applicability of this method depends on several characteristics of the video stream of the object flow. First, the presence of a uniform background against which the objects of interest are clearly distinguishable. Second, all counted objects belong to the same class, other types of objects are not visible in the frame (thus, there is no need for an additional classification task). Third, these objects have the same simple shape (elongated ellipsoidal contour). Fourth, the frame rate during shooting (33 frames per second) and the rapid unidirectional movement of objects in the fluid flow ensures that in the absence of a magnet, the same object will occur in one or two frames and then disappear.
There are two cases in the video stream analysis: each object is clearly separated from the others, and clusters of aggregated objects forming large nonseparable regions are present. Therefore, if the objects do not stick together in the video, we can find all the contours in a separate frame and calculate their areas using the corresponding functions of the OpenCV library. Based on the values of these areas, we evaluate the average area of the contour corresponding to a single object. Then, we subtract adjacent frames to reveal moving objects and count their number by the number of contours (one contour per object). If object adhesion is present, we determine the area of bright spot representing several objects sticking together and divide it by the calculated average contour area of a single object. Since the areas of the spots change in proportion to the number of adhering (or washed off) objects, it allows us to approximately calculate the number of objects in the frame. Despite simplicity, these algorithms show high accuracy in object counting. This was validated by the video obtained from the sample with a known number of objects. The algorithms have found 89-93% of objects depending on the presence of object agglomeration. A lower detection rate is observed in the presence of one or more large bright spots since the calculation through the average object area is inevitably less accurate. However, the results obtained here using computer vision methods correlate with the results obtained by measuring the average intensity in the lightsheet plane. Atomic force microscopy (AFM) topography images of microcapsules before injection and captured from the rat blood were obtained using NTEGRA Spectra microscope (NTMDT-SI, Russia) and NSG10 tips (TipsNano, Russia) in tapping mode. All samples were dried on a glass substrate and measured in a completely dried state. All image processing, measurements of microcapsule parameters, and correction of AFM artifacts were done using Gwyddion software [36].

Confocal laser scanning microscopy
Fluorescence images of microcapsules captured from the whole blood were obtained using a confocal laser scanning microscope (CLSM) Leica TCS SP8 X (Leica Microsystems, Wetzlar, Germany).

Flow cytometry
Captured magnetic capsules were counted and their optical properties were evaluated using the imaging flow cytometer Amnis ImageStream X Mk II. As shown in previous works [37,38], fluorescence intensity of FITC-labeled capsules does not overlap with the autofluorescence intensity of blood cells, which allows us to separate capsules from cells by fluorescence intensity gating in the range of 480-560 nm. Fluorescent dye was excited by a 488 nm laser at 100 mW power. Flow cytometry data were processed using IDEAS software (Luminex, USA).

Results and discussion
We have developed a SPIM-based flow cytometry system capable of visualizing objects in the flow of whole blood without dilution or hydrodynamic focusing. The flow cell of our system can be connected directly to the blood flow by shunting two large vessels in the animal's paw. It can also provide on-demand magnetic separation of objects from the flow using a rare earth permanent magnet with their following extraction.
To test the capabilities of the developed system, we used the polyelectrolyte microcapsules, functionalized with magnetite nanoparticles and labeled with FITC-BSA, as test objects. Polyelectrolyte microcapsules, in general, are well-suitable from the different points of view: they are biocompatible and do not significantly affect normal blood flow [38]; can be easily functionalized by different agents providing fluorescent [39,40], Raman [41], or photoacoustic [42,43] modalities of contrast. The microcapsules that were used in our experiments have well-defined fluorescent properties, size, and good magnetic separation ability. These hollow microcapsules were formed using CaCO 3 microparticles as templates followed by dissolution of the cores. According to AFM data, the microcapsule shell thickness was about 33 nm, and the average diameter of the capsules was 1.34 ± 0.23 µm, which was also confirmed by CLSM measurements.
Two main methods of image processing were employed to analyze data from the developed SPIM-based flow cytometry system. The raw video stream of the lightsheet and surrounding area was captured by the sCMOS camera. The first analysis method is somewhat similar to traditional fluorescence flow cytometry, which measures average fluorescence intensity inside the ROI covering all the lightsheet plane cross-section of the flow channel. The average fluorescence intensity was calculated and used as flow cytometry channel values. This method works well if objects cannot be separated due to their high concentration or aggregation in the area near the tip of the magnetic field concentrator. But signals, in this case, can be affected by the fluorescence quenching, fluorescent background of the flow, and extinction of light by the media in the flow cell. However, in general, it is still proportional to the amount of fluorescent material in the lightsheet plane. The second analysis method is based on object separation against the background followed by counting various shape parameters. This method uses the difference between two adjacent camera frames to find the background and objects visible on it, even in difficult cases of blurry images and high background fluorescence typical of blood. While working with objects of known size and shape (for example, polyelectrolyte microcapsules), small aggregates can still be separated by measuring the object's area on the image plane. It is generally similar to imaging flow cytometry data processing. However, the same objects can be detected in several planes if the camera speed is high enough and the flow rate is low, but in our case this problem does not arise. Optical flow methods should probably be applied if a higher-speed camera is used in the experiment.

Magnetic microcapsule separation from the PBS solution
We carried out several control experiments to test the capabilities of the developed system. The first one was performed to test the magnetic microcapsule separation from PBS at different flow rates. The suspension of microcapsules in PBS buffer media was pumped through the flow cell and visualized in the lightsheet plane. Then the magnet was placed near the channel and the carriers' capture in the strong magnetic field area was registered. Due to high capsules' aggregation in the magnet tip area, we used the average level of fluorescence to detect objects. At the same time, the video of the flow was also recorded for further analysis by computer vision methods. The estimated amount of fluorescent microcapsules can be converted to a dose of a biologically active substance accumulated in the region of a strong magnetic field and extracted from the liquid flow. After registration, the magnet was removed and the desorption of carriers from the accumulation area was observed.
The velocity of the flow affects the processes of magnetic accumulation and separation of objects (Fig. 2). The magnetic field distribution from the top of the concentrator tip (in the direction perpendicular to the flow cell channel) is shown in Fig. 3(a). The field was 0.3 T measured at a distance corresponding to the channel position. At the lowest flow rate of 10 µL/min, the microcapsules did not accumulate significantly under the magnetic field influence. The average fluorescence intensity values upon the application of a magnet were close to the background baseline and did not exceed 50 a.u. (Fig. 2, top row, 10 µL/min). The number of capsules also did not show significant changes under the magnetic field influence (Fig. 2, bottom row, 10 µL/min). It was probably caused by a low number of capsules passing through the flow cell during the accumulation period. With the increase in the flow rate to 20 µL/min, the accumulation increased, but this process was relatively slow. It can be seen from the increase in the average fluorescence intensity (Fig. 2, top row, 20 µL/min) and the number of capsules detected in the lightsheet area per second (Fig. 2, bottom row, 20 µL/min). Both curves had a characteristic slope before reaching the saturation region. The accumulation at 30 and 50 µL/min rate showed the same behavior: when the magnet was applied to the flow cell, we could see a sharp increase of signal intensity (Fig. 2, top row, 30 and 50 µL/min) and a number of capsules detected per sec (Fig. 2, bottom row, 30 and 50 µL/min) followed by a saturation region. The magnet removal led to the immediate decrease in the fluorescence intensity and the number of capsules, respectively. However, the microcapsule accumulation happened 1.5 times faster at 30 µL/min rate if compared with 50 µL/min. This can be caused by washing of magnetically captured objects by multiple collisions with microcapsules flowing through the channel. After removing the magnet, most of the adsorbed microcapsules were washed out. At 100 µL/min rate, nothing was captured by the magnet, since the impulse of the flowing microcapsules colliding with the captured ones presumably exceeded the force holding the microcapsules in the trap (Fig. 2, 100 µL/min).
Whether an object will be held in a region of a strong magnetic field is determined by the balance of forces acting on it. The object is held in place by the magnetic field and adhesion from adjacent objects. It should also be noted that the residual magnetization of objects and adhesion forces continue to operate even some time after the magnet removal from the flow cell. However, the object can be knocked out of the magnetic trap due to the impact of the fluid flow itself and collisions with other objects carried by the flow. Moreover, the accumulated volume of capsules makes the stream alongside the magnetic concentrator more narrow, therefore increasing flow speed in this area and the washing effect of the stream.

Magnetic microcapsule separation from the whole blood in vitro
The efficient capture of magnetic objects in vivo will require to separate them from the flowing blood, which contains about 8-9×10 9 erythrocytes per mL as well as leukocytes, thrombocytes, proteins, ions of Ca + 2 , Mg + 2 , Na + , etc. The high concentration of these objects in the blood can affect the ability and efficiency of the separation of individual magnetic capsules, as they must overcome mechanical tension and impulses from multiple collisions with other objects when crossing the region of the strongest magnetic field.
Furthermore, in vivo procedures place some restrictions on the size of used objects to ensure safety of their circulation in the bloodstream. As it was shown in our previous paper [38], micron-sized polymeric capsules do not disrupt the blood flow, and the organism can adapt to the appearance of the carriers in the blood as well as their accumulation in blood vessels and organs. This makes the magnetic capsules of micron size a relevant, safe, and preferable choice for further in vivo experiments.
The testing of the ability to efficiently separate magnetic microcapsules from the blood was required before we started in vivo experiments. We evaluated the dependence of microcapsule capture efficiency by the magnetic field on the flow rate in the cytometer flow cell in vitro. The magnetic objects were represented by the microcapsules used above. The whole rat blood, diluted 3:1 with PBS, represented the control sample without any microcapsules. For other samples, the whole blood was diluted with microcapsule suspension in PBS in the same ratio. The final concentration was about 14.65 × 10 6 carriers per mL. The separation ability was tested at the suspension flow rates: 5, 10, and 30 µL/min. The amount of captured carriers was assessed using the imaging flow cytometer Amnis ImageStream X Mk II (Fig. 3(b)).
Therefore, the efficiency of microcapsule capture from the blood slightly decreases with the increase in the sample flow rate (Fig. 3(b)). For example, at the 30 µL/min 35.6% of carriers was trapped from the initial amount compared to the 44.6% at 5 µL/min rate.

Experiment with the animal model
The evaluation of magnetic capsules' capture efficiency in vivo was performed on the rat model. Blood vessels of a white rat were connected to the cytometer flow cell by the polyethylene catheters inserted into the femoral artery and vein of the rat's hind paw. 1.1 × 10 7 capsules suspended in 20 µL of physiological salt solution were injected into the carotid artery ( Fig. 4(a,  b)). The injection was performed through the catheter, accompanied by heparin injection to prevent blood coagulation.  Figure 4 represents the carrier circulation dynamics in the rat blood flow and a decrease in microcapsule concentration in the bloodstream due to both systemic filtration by internal organs (in particular, by the liver) and magnet-induced accumulation of capsules in the flow cell. Until the 120 th sec (Fig. 4(c) (left)) we could see the autofluorescence of erythrocytes in the flow channel. The 120 th sec point corresponds to the injection of microcapsules into the blood flow (the sharp decay of the signal could be caused both by the artifacts induced by the interaction between the rat and the optical system, and the optical clearing of the blood due to the decrease in erythrocytes number after its dilution by PBS). We could see the beginning of the carriers' circulation several seconds after microcapsule suspension injection into the bloodstream in the absence of a magnetic field (Fig. 4(c) (left)). Microcapsules were circulating in the blood circulatory system for a significant period of time (20-30 min and more). According to our previous data [37], the circulation time of 3-5 µm microcapsules (versus 1.34 µm in the current study) was much shorter, and the number of microcapsules in the blood rapidly decreased within the first 3 min. After the magnet was brought in (the 20 th sec in Fig. 4(c) (right)), the number of carriers in the blood flow downstream of the magnet decreased relative to the initial value, corresponding to the first minutes of recording. We estimated the amount of fluorescent material in the channel and converted it into a number of capsules (Fig. 4(d)).
Doppler optical coherence tomography (DOCT) cross-sections of the bypass polymer tube, artery, and vein are presented in Fig. 5. Phase unwrapping was done prior to the calculation of the blood flow velocity values to obtain correct results (see Supplementary Note 2 and Fig. S2 in Supplement 1). Due to the turbulent nature of arterial blood flow, it was impossible to apply the phase unwrapping algorithm; therefore the average phase difference over the entire vessel was used to calculate the values of the arterial blood flow velocity. According to the DOCT data, capsule injection in the blood-vascular system caused an insignificant slowdown of the arterial and vein blood flow with subsequent recovery to the initial values. Magnetic capture of microcapsules did not provide any impact on the blood flow velocity, which indirectly indicates the absence of vessel embolism and thrombosis. The speed of blood flow inside the capillary flow cell was 1.4 ± 0.8 cm/s. The speed of blood was 2.1 ± 1.2 cm/s inside the artery and 1.1 ± 0.6 cm/s inside the vein. Taking into account the vessels' diameters, that were ∼0.2 mm for artery, ∼0.35 mm for vein, and 0.28 mm inner diameter of the polymer tube, the flow rates were ∼0.04 mL/min in artery, ∼0.06 mL/min in vein, and 0.05 mL/min in the flow cytometry system. These flow rates are nearly identical as they are all connected and should have approximately the same volumetric flow rate. Magnetic trapping of microcapsules was carried out for 10 min by the 0.3 T magnet, followed by removal of the magnet and washing the trapped microcapsules into an outer tube by several rinses with PBS. The number of magnetically-trapped capsules was independently measured using the imaging flow cytometer Amnis ImageStream X Mk II. Fluorescence signals of FITC-labeled microcapsules and autofluorescence intensity of blood erythrocytes are well spectrally-separated [37,38].
The microcapsules' collection was carried out for a fixed sample volume, and then the total number of retrieved carriers was calculated ( Table 1). The initial capsule concentration in the approximate whole volume of rat blood (15 mL) was around 7.81 × 10 5 capsules per mL. During flow cytometry measurements, 20 µL of the whole blood was analyzed. We found that the number of capsules in the sample after magnetic capture was around 25 times higher (3.97 × 10 5 capsules per 20 µL) than the calculated number of capsules contained in a similar blood volume (0.15 × 10 5 capsules per 20 µL). Moreover, the actual number of circulating microcapsules was less than the calculated one, since we did not take into account the filtration and accumulation of microcapsules in internal organs (for instance, lungs, kidneys, liver) during the experiment. This indicates the possibility of using the developed magnetic separator for efficient capture of capsules. Fluorescence images of microcapsules before and after their injection to the blood flow showed the absence of significant differences between them, except for the higher aggregation degree of capsules captured from the whole blood ( Fig. 6(c, d)) in comparison with the water suspension ( Fig. 6(a, b)). This effect can be caused by the close interaction between the capsule shells during their accumulation in the region of a strong magnetic field and the interaction of the capsules with various blood proteins, which can play the role of binding molecules [37]. AFM images of capsules after the experiment showed slightly decreased thickness of the shell (50-56 nm of double-layer shell thickness against the 70-80 nm before the injection) probably due to some interaction of biodegradable microcapsule shell with the whole blood. We could also see the presence of some debris on the surface of microcapsule shells which could be adsorbed prior to extraction or as a part of the extracted suspension.
Thus, we have developed a new type of in vivo flow cytometry system. It allows objects to be measured and imaged in undiluted blood; it allows the long-term in vivo experiments with the total blood volume of an animal to be performed without taking blood samples; and it allows objects to be retrieved from the bloodstream. The developed technique is definitely more invasive compared to systems based on photoacoustics or diffuse fluorescence. But an external flow cell connected to large blood vessels gives us opportunities not available in other systems.
In this study, we demonstrated some basic set of features: counting objects for an experiment with an artificial CTC phantom; continuously extracting objects from animal whole blood; determining the characteristics of objects extracted from the bloodstream, which was not possible before. It can be extended to address more specific challenges in biomedical research, for instance, providing simultaneous measurements and imaging in multiple fluorescent channels, or separating and sorting objects on demand (based on images if needed). So, we hope that we have developed a useful new tool that expands the ability to work with rare objects in the bloodstream and, thus, allows us to better understand some of the basic biological processes in the body.

Conclusion
The developed SPIM-based imaging flow cytometry system is capable of detecting objects in whole blood both in vivo and in vitro, followed by magnetic retrieval. Two object counting methods were applied to detect objects on the video stream from the sCMOS camera. The first one is based on total fluorescence in the lightsheet plane and is similar to the data analysis in traditional flow cytometry based on point detectors. The second one is based on the object separation from the image background with subsequent object size evaluation. The first method works well if the concentration of objects is very high and they become indistinguishable in the camera images. The second one works well for sparse separated objects, or when the number of objects in the aggregate can be evaluated by its size. Effective magnetic separation of objects from the PBS buffer stream has been shown with 0.3 T magnetic field near the tip of the field concentrator at a flow rate of 20-50 µL/min. The developed system allowed us to monitor fluorescently-labeled microcapsules inside the flow cell shunting two large blood vessels of a rat limb after their injection into the carotid artery. The speed of blood flow measured by the DOCT was 1.4 ± 0.8 cm/s inside the shunting polymer tube; 2.1 ± 1.2 and 1.1 ± 0.6 cm/s inside the artery and the vein, respectively. We have shown the separation of about 4 × 10 5 microcapsules in vivo from blood by a 0.3 T permanent magnet with a concentrator. The extracted objects were measured afterward by imaging flow cytometry, CLSM, and AFM methods. An unexpected effect of reducing the microcapsule shell thickness after passing through the bloodstream was found along with the adsorption of a large amount of debris on it.