1.

Introduction

The detection and quantitation of single moving blood cells in vivo in their native state is potentially important for the early diagnosis of many diseases (e.g., cancer, diabetes, and cardiac diseases) and for the study of the influence of various factors (e.g., drugs, smoking, and radiation) on individual cells. This approach, however, has not yet been well established in vivo,¹ although in vitro, on the technical platform of flow cytometry (FC), it has revolutionized many biological and medical studies.^{2, 3} With in vitro FC, cells move in flow at high speed (up to $20\mathrm{m}∕\mathrm{s}$ ) through an area irradiated with one or more focused laser beams. The forward and sideways scattering of fluorescent light in different spectral regions is recorded with several photodetectors. This technique is now in wide use in basic and clinical research, providing, among other applications, rapid analysis of large populations of cells, information about the relative sizes and shapes of cells, detection of rare cells, distinction of subpopulations of cells according to their properties, and evaluation of cell-drug interactions.^{2, 3}

The potential adaptation of the principle of FC from in vitro to in vivo requires many precautions. On the one hand, cell velocity in blood and, especially, in lymph flow is much slower (from $0.3\text{to}5\mathrm{mm}∕\mathrm{s}$ in blood microvessels to $0.2\text{to}0.5\mathrm{m}∕\mathrm{s}$ in large blood vessels of animals and humans⁴) than the maximum speed of analysis achieved in FC (see previous paragraph). Therefore, from this perspective, FC’s technical platform is easy to adapt to an in vivo setting. On the other hand, in vivo study may be limited by (1) poor optical conditions, such as scattering and absorption of laser radiation by the vessel wall and surrounding tissues (e.g., skin layers, connective tissue, muscles, fat, and other tissue components), compared with the ideal optical conditions in in vitro FC; (2) difficulties of noninvasively accessing deep vessels; (3) problems with the use of a transillumination optical scheme, which is more convenient for achieving high resolution; and (4) instability of blood and, especially, lymph flow parameters (e.g., fluctuation of cell velocity due to bifurcations, nonuniform vessel cross sections, etc.), compared with the well-controlled cell-flow parameters of in vitro FC. Additionally, (1) the shapes of the velocity profiles in venules and arterioles differ from the parabolic⁵ (i.e., are asymmetric); (2) the resistance to blood flow in living microvessels is higher than that in uniform tubes of the same diameters;⁶ and (3) the distribution of cells in flow depends on the flow velocity and the cell concentration.^{4, 7, 8} Nevertheless, as we show in the following, in selected blood vessels with a relatively fast flow velocity, flow is close to laminar, with most red blood cells (RBCs) concentrated near vessel axes with a mean centerline velocity 1.6 times higher than that of the mean blood flow⁹ (i.e., natural hydrodynamic focusing occurs in vivo as it does in in vitro FC).

The first few attempts to develop in vivo FC in an animal model were relatively successful.^{10, 11, 12, 13} In particular, use of a confocal scheme with fluorescence detection enabled continuous monitoring of circulating labeled RBCs, white blood cells (WBCs), and cancer cells in the upper layer of blood microvessels in mouse ear skin.^{10, 11} Currently, the very powerful fluorescence labeling used in this work (as in many other in vitro FC platforms) is potentially subject to photobleaching effects and the escape of fluorescent indicators through diffusion. Moreover, growing evidence shows that fluorescence labeling may seriously distort genuine cell properties through unwanted indicator-cell or indicator-drug interactions.^{14, 15} This issue gains importance in in vivo studies in humans.

In our alternative approach for in vivo FC, we used a photothermal (PT) technique without fluorescence labeling.^{12, 13} Indeed, PT imaging (PTI) has already demonstrated in vitro the capability of visualizing, at the living single-cell level, nonfluorescent absorbing cellular structures that are not visible by other techniques.^{16, 17} In addition, the PT technique has been able to characterize a liquid-flow profile (PT velocimetry) without the use of probes or scattering objects¹² and to provide selective multispectral detection of moving absorbing compounds in capillary chromatography.¹⁸ In in vivo experiments in an animal model (rat mesentery), we demonstrated PTI’s capability, without laser scanning (with just one laser pulse) or conventional fluorescence labeling, to image (1) specific intracellular absorbing structures of circulating RBCs and WBCs in microlymphatic flow; (2) laser-induced thermal effects in blood flow at the single-RBC level; and (3) gold nanoparticles in lymphatic vessels as prospective models of both PT probes and sensitizers for the purpose of selective nanodiagnostics integrated with nanotherapeutics (e.g., selective nanophotothermolysis¹⁹) of metastatic cancer cells and bacteria in vivo.^{13, 20} In general, PTI of individual cells in vivo could be crucial to morphologic analysis and identification of different cells (most conventional in vitro FC platforms do not provide image analysis), although this technique is relatively sophisticated (e.g., requiring synchronization of the laser pulse with flow and a high-speed CCD camera), and significant memory storage is required to image even just a selected number of cells. For these reasons, we present here an adaptation of the simpler and potentially more rapid thermolens method (developed previously for studying single nonmoving cells in vitro^{19, 21}) for detecting moving cells in vivo with guidance by transmission digital microscopy (TDM) and fluorescence microscopy.

2.

Methods and Materials

2.1.

General Concept of Integrated in vivo FC

The concept of integrated multispectral in vivo FC developed in our laboratory combines the basic principles^{2, 3} of in vitro FC with recent attempts^{10, 11, 12, 13} at in vivo FC (Fig. 1 ). Cell detection occurs in superficial microvessels in the skin or in thin, relatively transparent living biostructures that enable the measurement of light transmitted through the vessels. To illuminate selected vessels, different optical sources are used, ranging from a conventional lamp with filters to lasers in pulsed or continuous-wave (cw) mode. One beam or several beams simultaneously from such a source are focused on a small area of microvessels (the analysis point). To detect all cells in a cross section of vessels, the laser beam diameter must cover whole vessels or at least illuminate a localized vessel area where most cells move in the flow stream (e.g., near the axis in high-speed blood flow, or in an area of lymphatic valves, which act as natural nozzle for the cells; see Fig. 1, bottom left). In some cases, to minimize simultaneous irradiation of several cells along the vessel axis, the beam has an elongated ellipsoidal shape produced with a cylindrical lens or slit as in conventional FC. The high-speed (up to $170\mathrm{fps}$ ), highly sensitive CCD camera is used to visualize moving cells, together with the vessel’s structure, and to control the position of the laser beam in the vessel with the use of a reflection, transmission, or fluorescence (if appropriate) mode.

Fig. 1

Schematic of integrated multispectral FC in vivo combining PT, photoacoustic, transmission and fluorescent techniques. Bottom left, position of the laser beam in a lymph microvessel with a valve, which acts as a natural nozzle focusing flowing cells near the vessel’s center.

Optical absorption by nonfluorescent cellular structures is measured with PT and photoacoustic (PA) methods by the detection of absorbed energy transformed into heat and acoustic waves, respectively, through nonradiative relaxation.¹⁸ Specifically, we incorporated a PT deflection module with a position-sensitive photodetector to measure bioflow velocity (PT velocimetry¹²), a PT thermolens module to measure the integrated, time-resolved PT response^{19, 21} from moving cells, and a PTI module to image moving cells alone or with PT probes (e.g., gold nanoparticles conjugated with specific antibodies) in combination with thermolens²² and phase-contrast imaging.^{16, 17} Additionally, a PA transducer (PZT ceramic, sensitivity $50\mathrm{mV}∕\text{bar}$ , $8\text{-}\mathrm{mm}$ diameter, bandwidth $0.5\text{to}30\mathrm{MHz}$ , produced by Dr. Karabutov), or transducer array, enables measurement of PA signals from single cells alone or with attached nanoparticles in flow, as well as measurement of PA signals from cells accumulated at specific sites (vessel wall, surrounding tissue, or specific organ) with a relatively broad laser beam and according to the principles of PA tomography.^{23, 24} Here, we focus rather on laser-based PT methods that provide imaging of single moving cells with the highest absorption sensitivity (threshold for the absorption coefficients of the order of ${10}^{-2}{\mathrm{cm}}^{-1}$ ) and resolution $(\sim 0.7\mu \mathrm{m})$ . This is crucial for detecting WBCs, which have low absorption in the visible-spectral range compared with strongly absorbing RBCs.^{12, 16} This sensitivity threshold corresponds to temperature variations of $\sim {10}^{-1}\text{to}10°\mathrm{C}$ in a cell at the laser energy range typically used for noninvasive mode.²⁵ Thus, the temperature fluctuation is small enough not to damage cell structures,²⁵ especially during the short time of irradiation while cells cross the illuminated area. Because PT methods are relatively new in biological studies, fluorescence techniques (widely used in conventional in vitro FC), together with TDM, are employed to verify PT data in the initial experimental stage by detecting labeled cells. In particular, we found that all these techniques may beneficially supplement each other and, in combination, may increase the capability of an integrated FC system.

In most experiments, PT data were obtained for 100 to 500 cells over $10\text{to}100\mathrm{s}$ , depending on the flow rate (see the preceding), at a pump-pulse repetition rate of $1\text{to}10\mathrm{Hz}$ . The experiments were performed at least three times under similar conditions, and statistical data were calculated according to established procedures, including estimation of the ${\mathrm{ED}}_{50}$ photodamage threshold.²⁵

2.3.

Experimental Setup

The scanning of a focused laser beam across a cell, as in confocal microscopy, is not quite suitable for detecting and imaging moving cells. This is because this scanning technique would require sophisticated, time-consuming processing algorithms¹ (taking at least seconds, if not minutes) to accommodate even the relatively slow flow in capillaries as cells cross the area of detection in ${10}^{-1}\text{to}{10}^{-3}\mathrm{s}$ . To realize an integrated PTFC, we employed the principle of time-resolved, nonscanning PTI^{16, 17} using a recently described advanced experimental setup⁵¹ that we further modified and upgraded with the incorporation of a PT deflection module, a fluorescence module with a sensitive CCD camera (PixelFly, Cooke Corp.), high-speed photodetectors, a stabilized pilot laser, a high-resolution TDM module with a powerful objective [ $100\times$ ; numerical aperture (NA), 1.25], and accessories for in vivo studies (Fig. 2 ). Here, we focus more intently on the thermolens module, which in our opinion is crucial for realizing PTFC in vivo.

Fig. 2

Schematic of the integrated PT flow cytometer.

Briefly, cells in vessels were irradiated with a short, focused pump laser pulse [tunable optical parametric oscillator (OPO): spectral tuning range, $420\text{to}570\mathrm{nm}$ ; pulse width, $8\mathrm{ns}$ ; pulse energy, $0.1\text{to}300\mu \mathrm{J}$ ; Lotis Ltd., Minsk, Belarus]. Laser-induced temperature-dependent variations of the refractive index in the cells were detected by the thermolens technique with the use of an Olympus BX51 microscope. In this technique, a pump-laser-induced refractive heterogeneity (called a virtual thermolens⁵²) caused a focusing or defocusing of a collinear cw stabilized He–Ne laser probe beam (model 117A; Spectra-Physics Inc.; wavelength, $633\mathrm{nm}$ ; power, $1.4\mathrm{mW}$ ) and hence a reduction in the beam’s intensity at its center as detected by a photodiode (C5658; Hamamatsu Corp.) supplied with a $0.5\text{-}\mathrm{mm}$ -diam pinhole, and recorded with a Tektronix TDS 3032B oscilloscope. This mode enabled recording the integral PT response simultaneously from several cells or from single cells and even from localized zones inside single cells, depending on the laser-beam diameter. The diameters of pump- and probe-beam spots with stable, Gaussian intensity profiles⁵¹ (controlled by a CCD camera: AE-260E; Apogee Inc.) in the sample plane were in the ranges of 15 to 50 and $5\text{to}50\mu \mathrm{m}$ , respectively. The beam diameter could be adjusted by an axially moving condenser or a change of the micro-objective itself (with magnification values of $4\times$ , $10\times$ , $20\times$ , $60\times$ , and $100\times$ ). Because of the laser beams’ relatively broad sizes, resolution was determined by conventional diffraction-limited resolution of the microscope objective itself (e.g., $\sim 0.7\mu \mathrm{m}$ at $20\times$ , NA 0.4; and $\sim 0.3\mu \mathrm{m}$ at $100\times$ , NA 1.25). In selected experiments, we used additional optics to create a linear beam shape. Additionally, for multispectral studies, we were able to use first $\left(1.06\mu \mathrm{m}\right)$ , second $\left(532\mathrm{nm}\right)$ , and third $\left(355\mathrm{nm}\right)$ harmonics of a pulsed Nd:YAG laser.

TDM images with a conventional microscope lamp were recorded with three CCD cameras: a black-and-white Cohu 2122 (Cohu Inc., San Diego, California) and a color Nikon DXM1200 with speeds of up to 22 frames per second (fps) for imaging lymph flow, and a high-sensitivity CCD camera (PixelFly, Cooke Corp.) with a speed up to $170\mathrm{fps}$ for imaging blood flow. Scion Image (Scion Corp., Frederick, Maryland), ACT-1 (Nikon), and CamWare (PCO Computer Optics, Kelheim, Germany) software packages were used to process, capture, measure, and edit images, as well as for analyzing the moving cells. Depending on the structures’ sizes, different magnifications were used to image relatively large whole lymphangions ( $4\times$ , $8\times$ , $10\times$ , and $20\times$ ), as well as single cells ( $40\times$ , $60\times$ , and $100\times$ ), in lymph and blood flow.

In the current experiments, we focus on studying the capability of the integrated PTFC system in three modes (alone and in combination): TDM, thermolens, and fluorescent. TDM mode was used for general observation of mesenteric structures and blood and lymph microvessels (diameter, indices of phasic contraction, valve function, etc.), including measurement of the absolute velocity of lymphocytes in lymph flow by the video recording of cell movement. Furthermore, TDM made it possible to maneuver pump and probe beams within mesenteric structures, particularly enabling focusing of the beams in the central part of a vessel (or other parts, if necessary) and estimating cells’ shapes, sizes, and aggregation types, as well as, possibly, their interaction in flow. Thermolens mode was used to measure the PT responses from different moving cells and PT probes (e.g., gold nanoparticles¹³), compared with PT responses from other mesenteric microstructures (e.g., vessel walls, interstitial tissue). Fluorescence imaging with different labels (fluorescein isothiocyanate-dextran, Alexa 488, or FP-3056-2 fluorescence dye; Sigma-Aldrich, St. Louis, Missouri) was used in selected experiments to increase the contrast of blood vessels and to monitor WBCs, labeled cancer cells (MDA-MB-231 human breast carcinoma cells), and polystyrene beads (nail-red fluorescent particles; diameter, $0.84\mu \mathrm{m}$ ; Spherotech Inc., Libertyville, Illinois).

3.

Results and Discussion

TDM mode using different magnifications (4 to $100\times$ ) enabled the imaging of moving blood cells in mesenteric microvessels: RBCs in blood vessels (Fig. 3 , top, first), single platelets traveling through capillaries (Fig. 3, top, second), rare single WBCs in blood flow (Fig. 3, top, third), and rare RBCs in lymph microvessels (Fig. 3, middle). In particular, TDM’s high spatial resolution (up to $300\mathrm{nm}$ at $100\times$ magnification with water immersion) made it possible to roughly estimate cell size and even shape, including continuous monitoring of RBC rotation during movement in lymph microvessels (Fig. 3, middle). We found, however, that the relatively low absorption sensitivity of TDM (due to the short optical path in thin structures) did not enable easy and rapid differentiation of cells with different absorption properties in the fast flow (e.g., rare moving WBCs in blood flow and RBCs in lymph flow). In contrast, using PT thermolens mode, we were able to realize many applications that were unachievable with TDM. Figure 3, bottom, shows a typical configuration and the positions of the laser beams in the rat mesentery. In most cases, when the distance between cells was larger than the laser beam diameter, we used a circular geometry for the laser beam (Fig. 3, bottom, first and second), although selected experiments were performed with an ellipsoidal beam configuration (Fig. 3, bottom, right), created by the use of an additional cylindrical lens and a slit in front of the microscope condenser. In this case, the size of the laser beam at the sample plane could be changed from $5.5\times 45\text{to}8\times 65\mu \mathrm{m}$ , which was sufficient to cover most of the selected microvessels.

Fig. 3

Typical TDM images of rat mesentery with blood and lymph microvessels. Top row: first, RBCs in blood microvessels (magnification, $100\times$ ); second, platelets with RBCs in a blood capillary (magnification, $100\times$ ); third, rare leukocytes in blood microvessels (blood velocity, $\sim 1\mathrm{mm}∕\mathrm{s}$ ; magnification, $100\times$ ). Middle row: rotation of RBC in lymph flow (magnification, $100\times$ ). Bottom row: typical position of a circular probe beam in a blood capillary (first; magnification, $100\times$ ), artery (second; magnification, $10\times$ ), and blood vessels with an ellipsoidal beam geometry (third; magnification, $10\times$ ).

We obtained integral PT responses from single moving RBCs and lymphocytes in lymph flow compared with PT responses from vessels and interstitial areas (no cells) at different pump laser energy levels (Figs. 4 and 5 ). At relatively low laser energy levels, the classic PT response was usually observed (Fig. 4, top; Fig. 5, top): a high initial peak, due to rapid heating of the cell, and a much lower exponential tail ( $6\text{to}10\mu \mathrm{s}$ for RBCs with an average size of $5\text{to}6\mu \mathrm{m}$ , $20\text{to}25\mu \mathrm{s}$ for $7\text{-}\text{to}8\text{-}\mu \mathrm{m}$ lymphocytes, and $0.5\text{to}2\mu \mathrm{s}$ for $1.5\text{to}2.5\mu \mathrm{m}$ platelets, not shown), corresponding to whole-cell cooling through heat diffusion into the surrounding plasma. The shapes of these responses were very similar to those obtained earlier in vitro,²⁵ although the cooling times were a little shorter (of the order of 20%). This discrepancy can be explained by the additional influence of flow on the cooling rate¹² (e.g., increased thermal diffusion). In the current experiments, the minimum rise time of the PT response was limited by the response time of the photodetector $\left(10\mathrm{ns}\right)$ and the pump laser pulse duration $(\sim 8\mathrm{ns})$ . The PT-response amplitude increased with increases in the pulse energy level, until local temperature reached the evaporation threshold $(>100°\mathrm{C})$ . At these energy levels, formation of a negative peak occurred, which indicated microbubble formation around localized absorbing cellular zones, as the refractive index inside the bubbles was much lower than that in the surrounding liquid and the temperature decreased during fast bubble expansion. After the initial expansion (rise time of the negative peak), the vapor rapidly cooled and condensed, and the bubbles collapsed (fall time of the negative peak). The lifetime of the bubbles, $1\text{to}4\mu \mathrm{s}$ , depended on the size of the endogenous absorbing cellular structures and on the laser energy level. At a high energy level, several bubbles might arise, overlap, and merge with each other. In accord with our previous finding, bubble formation detected with the PT technique led to cell damage.^{19, 25} The photodamage threshold, the fluence energy dose that induced damage in 50% of irradiated cells (100 cells) [median effective dose $\left({\mathrm{ED}}_{50}\right)$ ], was determined for blood cells in vivo with the use of analogous in vitro protocols.²⁵ We found that these parameters for single RBCs and lymphocytes were $3.5\pm 0.6$ and $41\pm 5.5\mathrm{J}∕{\mathrm{cm}}^2$ , respectively, at a laser wavelength of $532\mathrm{nm}$ , which was a little higher for cells in vitro (2.3 and $28\mathrm{J}∕{\mathrm{cm}}^2$ , respectively). This discrepancy can be explained by the different conditions of laser irradiation of cells in vitro and in vivo, as well as the slightly different cell sizes. Although additional, careful, independent verification is required because the transition from a linear to a nonlinear PT response occurred at a relatively broad range of pump pulse energy levels (depending on the cell type and the heterogeneity of the absorbing structures, among other factors), these experiments nevertheless demonstrate the ability of the PT technique for routine study of cell photodamage in vivo at the single-cell level.

Fig. 4

Typical linear (top) and nonlinear (middle, bottom) PT responses from slow-moving single RBCs in lymph flow in rat mesentery with coaxial configuration of pump and probe beams. Laser parameters: wavelength, $525\mathrm{nm}$ ; pulse width, $8\mathrm{ns}$ ; pump and probe beam diameters, 23 and $15\mu \mathrm{m}$ , respectively. Laser energy/amplitude/time scale from top to bottom: $0.5\mu \mathrm{J}∕50\mathrm{mV}∕1\mu \mathrm{s}∕\mathrm{div}$ ; $1.5\mu \mathrm{J}∕100\mathrm{mV}∕2\mu \mathrm{s}∕\mathrm{div}$ ; and $3.5\mu \mathrm{J}∕100\mathrm{mV}∕1\mu \mathrm{s}∕\mathrm{div}$ , respectively.

Fig. 5

Typical PT responses from slow-moving single WBCs (lymphocyte) in lymph flow in rat mesentery with a coaxial geometry of pump and probe beams in linear (top) and nonlinear (second) modes. Linear PT response from lymph vessels alone without cells in the laser beam (third). Linear PT response from interstitial area (background noise) located close to lymph microvessel (fourth). Laser parameters: $525\mathrm{nm}$ , $8\mathrm{ns}$ ; pump and probe beam diameters, 23 and $15\mu \mathrm{m}$ , respectively. Laser energy/amplitude/time scale from top to bottom: $25\mu \mathrm{J}∕50\mathrm{mV}∕5\mu \mathrm{s}∕\mathrm{div}$ ; $75\mu \mathrm{J}∕100\mathrm{mV}∕5\mu \mathrm{s}∕\mathrm{div}$ ; $125\mu \mathrm{J}∕50\mathrm{mV}∕5\mu \mathrm{s}∕\mathrm{div}$ ; $130\mu \mathrm{J}∕20\mathrm{mV}∕10\mu \mathrm{s}∕\mathrm{div}$ , respectively.

The cooling time of the PT response from vessels alone (no cells in the laser beam) or from the interstitium was a little longer (compared with the PT responses from cells) than can be explained by the heating of a more extensive area within the laser beam (Fig. 5). In addition, the polarity of the PT response in the case of the interstitium was “negative,” compared to “positive” for PT responses from cells and vessels, because the shape of the interstitium (and hence its virtual thermolens spatial configuration⁵²) was close to platelike, compared to the cylinder- and spherelike shapes of small vessels and cells, respectively. The photodamage thresholds determined for interstitium and vessels wall were $95\pm 12$ and $56\pm 7.5\mathrm{J}∕{\mathrm{cm}}^2$ (for a $532\text{-}\mathrm{nm}$ wavelength), respectively, which are higher than those for WBCs and, especially, RBCs (see earlier). These significant differences in photodamage thresholds make it possible to study cells in vivo under noninvasive conditions with the PT technique. These differences also provide the basis for indirect identification of the nature of the dominant absorbing targets. Cells were also identified by monitoring the cooling times of laser-heated cells because of the significant differences for RBCs and WBCs (see earlier). Thus, we found that the PT amplitude in linear, noninvasive mode (without photodamage) was approximately 30 to 50 times lower during irradiation of interstitial structures than during irradiation of single RBCs for the same laser energy level (compare Fig. 4, top, with oscilloscope images in Fig. 5, third and fourth). This finding could be explained by the low level of absorption by connective tissue and thin vessel walls. Surprisingly, even for lymphocytes, which have a low level of absorption in the visible spectral range (in contrast to RBCs), the corresponding difference in amplitude was lower ( $\sim 4$ to 6 times) than in vessels alone but still noticeable (Fig. 5, first, third, and fourth). Thus, high absorption contrast in cells and relatively lower background absorption by interstitial tissues and vessel walls (background noise), together with greater photodamage threshold for biotissue made it possible for us to use the PT technique to study primary blood cells in vivo and, in particular, to monitor the presence of RBCs in microlymphatics during pathologic processes. This is in line with our previous finding of the minor influence of absorption from vessels, which can be further reduced by the subtraction of this noise from useful PT signal, by the focusing of the laser beams, or by their proper positioning.¹³ In this study, a pathologic state was induced by mechanical injury of blood microvessels or by a relatively high $(>150\text{-}\mu \mathrm{J})$ pump pulse energy level that caused local blood vessel damage (Fig. 6 ).

Fig. 6

Position of the probe beam in microlymphatic flow with RBCs whose concentration was increased by laser-induced localized blood vessel damage (see hemorrhage around vessel wall): left: magnification, $10\times$ ; right: magnification, $100\times$ , water; immersion. (Different vessel zones are shown on left and right.)

On the basis of our earlier experiments in vitro,¹² we also estimated the capability of a new, noncontact, dual-beam PT method for the measurement of flow velocity in vivo. Briefly, a focused pump laser pulse produced a small heated zone in the blood or lymph flow through absorption. This zone, moving with the flow, was detected with a second probe beam located parallel to and some distance from the pump beam (Fig. 7 , top). Deflection of the probe beam as a result of temperature-dependent changes in the blood’s refractive index was recorded with a PT-deflection scheme (position-sensitive photodetectors) or the thermolens method (photodetectors with a pinhole). Both methods provided similar data. Flow velocity could be measured with these schemes by measuring (1) the width of the PT signal and (2) the time of flight of the heated zone between spatially separated pump and probe beams. Because of the very fast cooling times of laser-heated zones and the relatively low flow rates, we used a modification of the second method that involved an increased beam diameter up to $50\mu \mathrm{m}$ (to increase the cooling time up to a few milliseconds) and a short distance between pump and probe beams when we were able to measure the time between pump pulse and maximum PT response (in flow PT velocimetry, it is derived from the original PT response; Fig. 7, bottom left and right). Local heating was produced in blood flow by the OPO laser (see earlier) at a wavelength of $532\mathrm{nm}$ , at which significant absorption by RBCs occurs, while in lymph flow, with a much lower level of absorption in the visible spectral range, we used the first harmonic of an Nd:YAG pulsed laser (wavelength, $1060\mathrm{nm}$ ; pulse width, $8\mathrm{ns}$ ). For example, Fig. 7, top, shows the positions of probe (red) and pump (green) beams on blood vessels. For illustration purposes, we increased the distance between the two beams to prevent their overlapping, although in many experiments they were closer to each other (in the range $5\text{to}20\mu \mathrm{m}$ ). In the current case, we used relatively broad beams $(\sim 40\mu \mathrm{m})$ to irradiate many RBCs (100 to 150) within the laser beam volume, although in capillaries a single RBC or fluid between rare RBCs might be irradiated and heated, respectively. Figure 7, bottom, shows two PT responses for a coaxial (left) and a parallel (right) configuration of pump and probe beams with a distance between them of just $5\mu \mathrm{m}$ . The blood velocity, determined by obtaining the time delay (around $1.5\mathrm{ms}$ ) between pump pulse and maximum PT response (Fig. 7, bottom left), was approximately $3\mathrm{mm}∕\mathrm{s}$ , which is consistent with known data.⁴ A similar experiment performed in lymph flow showed⁵⁰ an average velocity of $0.5\mathrm{mm}∕\mathrm{s}$ .

Fig. 7

Top: positions of probe and pump beams in a relatively large blood vessel (diameter, $\sim 55\text{to}60\mu \mathrm{m}$ ). Bottom left: linear PT response from moving blood cells in vessels with a coaxial geometry of pump and probe beams (diameter, $\sim 45\mu \mathrm{m}$ ). Right: linear PT response from moving blood (as derivative profile shown on left, velocity, $\sim 3\mathrm{mm}∕\mathrm{s}$ ) with a parallel geometry of pump and probe beams obtained with time-of-flight technique. Laser pulse parameters: wavelength, $532\mathrm{nm}$ ; pulse width, $8\mathrm{ns}$ . Laser energy/amplitude/time scale/division: $0.5\mu \mathrm{J}∕100\mathrm{mV}∕5\mathrm{ms}∕\mathrm{div}$ and $0.5\mu \mathrm{J}∕20\mathrm{mV}∕5\mathrm{ms}∕\mathrm{div}$ , respectively.

This approach—PT velocimetry—has the potential for noninvasively (maximum temperature elevation, $\sim 0.5$ to $5.0°\mathrm{C}$ ) characterizing flowing fluids at velocities ranging from $0.1\mathrm{mm}∕\mathrm{s}\text{to}10\mathrm{m}∕\mathrm{s}$ and higher, with a potentially high spatial resolution of $\sim 3\text{to}5\mu \mathrm{m}$ (depending on the focal size of the laser beam), without the probes or the scattering objects required by Doppler or speckle techniques.⁵³ To our knowledge, this was the first demonstration of this technique for measuring blood and lymph flow in vivo.

Finally, on the basis of the experiments that we have described, we realized continuous monitoring of moving blood cells with the PTFC system, analogous to in vitro FC.

Figure 8 shows the typical PT-signal tracings from cells as a function of time during flow through an irradiated area. The PT signals with a “pure” positive component from RBCs in flow (Fig. 8, first) indicate a standard linear PT response from the cells at a low laser energy level similar to that seen in Fig. 4, top. Because of the short duration of each PT response $(\sim 10\mu \mathrm{s})$ , the responses were compressed during the recording because of the relatively slow flow rate, and appeared as vertical lines. The amplitude differences indicate differences in average absorption and reflect natural RBC heterogeneities. Low-amplitude PT signals or the absence of PT signals on the PT tracings (Fig. 8, first) from rare WBCs in blood flow indicate that an absorption by these cells is much lower than that by RBCs. Because of differences in the optical properties of normal and abnormal cells, we can assume that PTFC also has the potential to distinguish these cells without conventional cell labeling by the monitoring of the PT responses in FC mode. The second graph in Fig. 8 shows the area of RBC detection (just one cell during a $9\text{-}\mathrm{s}$ observation) at a low laser energy level. Laser-induced injury of vessels led to a fast-growing number of RBCs in lymph flow (Fig. 8, third). The next graph shows that the PT signals from RBCs at a relatively high laser energy level leads to cell damage (Fig. 8, fourth). The presence of only a negative PT signal indicated the manifestation of strong, dominant nonlinear effects (bubble formation, see Fig. 4, middle) with a small positive amplitude. The presence of both positive and negative components in the PT signals (Fig. 8, fifth) indicated a noninvasive condition for WBCs (with a high photodamage threshold) and an invasive condition for RBCs (low photodamage threshold) at the same energy level. We used a relatively high laser energy level $\left(145\mu \mathrm{J}\right)$ and a relatively slow oscilloscope rate to demonstrate the capability of PTFC to distinguish cells with different photodamage thresholds. Similar results were obtained during selective detection of cancer cells labeled with gold nanoparticles conjugated with antibodies, as described before.²⁰

Fig. 8

Typical traces of PT signal from blood cells in blood and lymph flow: RBCs in blood flow (first); rear RBC in lymph flow (second), normal conditions; growing number of RBCs in lymph flow during laser-induced hemorrhage (third); laser-induced damage of RBCs in lymph flow (fourth); and lymphocytes in lymph flow in linear and nonlinear PT modes (fifth). Laser parameters: wavelength, $525\mathrm{nm}$ ; pulse width, $8\mathrm{ns}$ . From top to bottom: laser energy/amplitude/time scale/division: $0.3\mu \mathrm{J}∕50\mathrm{mV}∕100\mathrm{ms}$ ; $0.5\mu \mathrm{J}∕20\mathrm{mV}∕1\mathrm{s}∕\mathrm{div}$ ; $0.6\mu \mathrm{J}∕100\mathrm{mV}∕200\mathrm{ms}∕\mathrm{div}$ ; $5\mu \mathrm{J}∕500\mathrm{mV}∕4\mathrm{s}∕\mathrm{div}$ ; and $145\mu \mathrm{J}∕100\mathrm{mV}∕10\mathrm{s}$ , respectively.

It is important to compare the capability of the new tool that we have described with that of conventional PT cytometry, which was developed for in vitro study of nonmoving single cells. The technical platform for conventional PT cytometry is microscopy, which enables great flexibility in choosing the regime of signal acquisition and simplifies cell manipulation. This approach seems to be reasonable at the initial development stage of a new PT technique. Indeed, the PT image or the integral PT response from one cell is obtained in $5\text{to}10\mu \mathrm{s}$ after just one pump pulse with an $8\text{-}\mathrm{ns}$ width (Figs. 5 and 6) and a broad beam diameter $\left(15\text{to}25\mu \mathrm{m}\right)$ covering entire cells. Thus, with this technique, called PT scanning cytometry, there is no necessity for time-consuming conventional scanning with a strongly focused laser beam across each cell. To obtain PT images and responses from different cells, however, the position of the laser beam’s center must be moved from cell to cell; the laser then irradiates the cell, and so on. Discrete changes in the position of the laser beam in PT microscopy take a short time ( $0.5\text{to}5\mathrm{s}$ for one mode) compared with that for conventional continuous laser scanning¹ $\left(0.5\text{to}1\min \right)$ . PT analysis of many cells $(⩾1000)$ , however, may take from $10\text{to}15\min$ to hours, which is not convenient for routine study. Furthermore, the monitoring of both the PT image and the integral PT response from one cell, if necessary, takes at least $5\text{to}10\mathrm{s}$ (to switch from one mode to another).

In PTFC, as we demonstrated in the preceding and elsewhere,¹³ these limitations have already been partly overcome. Currently, the PTFC rate of $\sim 10\text{cells}∕\mathrm{s}$ is limited by the repetition rate of the pump laser $\left(10\mathrm{Hz}\right)$ . Nevertheless, compared with the existing nonflow PT mode (which requires at least a few seconds to obtain one PT response from nonmoving single cells), PTFC is approximately 50 times faster. This capacity is enough to detect every nucleate WBC circulating in $20\text{-}\mu \mathrm{m}$ blood vessels at a typical rate of $3\text{to}6\text{cells}∕\mathrm{s}$ (for a flow velocity of $2\mathrm{mm}∕\mathrm{s}$ ), as well as all cells in microlymphatic vessels, where flow velocity is slower. The detection efficiency (number of detected cells per total number of cells passing through the vessels) in small lymphatics and blood capillaries with an ellipsoidal beam shape and a beam diameter comparable with that of the vessels is expected to be close to 1, while for vessels with a larger diameter there is still a problem with obtaining a clear PT response from each individual cell. Simultaneous irradiation of several cells in the laser beam, however, is possible if rare cells with abnormally strong absorption (e.g., rare RBCs in microlymphatics or cancer cells labeled with PT probes¹⁹) are to be determined.

The ultimate goal of detecting, without conventional labeling, circulating metastatic cancer cells at a desired threshold of $1:{10}^6$ cells requires further significant improvement in the PTFC speed of analysis and additional evidence that PTI can reliably discriminate cancer cells from normal cells. Under current conditions, which make it possible to study only slow flow, one cancer cell may appear in an irradiated area only once an hour. Nevertheless, a model for detecting circulating cancer cells labeled with PT probes (e.g., by injection into the blood stream or the localized primary tumor) at a relatively high threshold $(1:{10}^4)$ may be quite useful and realistic.

We also tested the combination of the PT method with the PA and fluorescence methods, as well as other animal models (mouse skin). Although these studies are outside the scope of this paper, we provide here a few brief comments to emphasize the capability of PTFC.

A preliminary estimation of the capability of the PA method (Fig. 1) with a customized acoustic transducer (Sec. 2(C)) to detect nonlabeled single RBC has shown its low sensitivity compared with that of PT methods ( $\sim 10$ times lower) and its inconvenience due to the necessity for mechanical contact of the transducer with the mesenteric tissue. Nevertheless, we believe that detection with the PA technique of metastatic cells labeled with gold nanoparticles may be a promising approach for monitoring their distribution in different tissues with a reasonable spatial resolution $\left(0.2\text{to}0.5\mathrm{mm}\right)$ . Comparison of the capability of TDM (Figs. 3, 6, 7) and the fluorescence technique (Figs. 9, 10, 11 ) with that of standard fluorescent dyes for high-resolution imaging of single cells and guidance of PTFC has shown that both TDM and fluorescence techniques may provide similar results in some specific circumstances [e.g., imaging of moving RBCs (Fig. 9), labeled breast cancer cells (Fig. 10), beads, and WBCs (Fig. 11)] and supplement each other. The advantages of TDM are simplicity, no need for labels, and high resolution (up to $300\mathrm{nm}$ ), although it is a transmission technique suitable only for selected animal models (e.g., rat mesentery). By contrast, the fluorescence technique is more universal and enables the use of reflectance mode, although fluorescent labeling may be harmful for some applications.^{14, 15, 54, 55} Using mouse skin with an optical chamber, we could not obtain high-resolution images of single cells in blood flow with either TDM or fluorescence techniques because of the relative thickness of the skin in this model. The use of glycerin as an immersion agent to reduce scattering effects improved the situation, but at the same time, significantly changed blood flow and cell parameters. According to our findings, the PT technique in combination with the rat mesentery animal model is the most promising system for obtaining high-resolution imaging of single cells in vivo as the basis for achieving our ultimate goal: high-resolution in vivo molecular imaging without labeling.

Fig. 9

Fluorescent image of blood flow (blood “traffic”) in mouse skin microvessels with visualization of single moving RBCs (as moving dark spots on a background of highly bright plasma). Image was obtained after injection of FITC-dextran into a mouse tail vein and by the use of an optical dorsal skinfold chamber and PixelFly CCD camera (magnification, $10\times$ ).

Fig. 10

Fluorescent image of breast cancer cells (MDA-MB-231) labeled with Alexa 488 fluorescent dye in model experiments after the dye was topically applied on the blood vessels and the intestinal background (mouse dorsal skinfold model) (magnification, $10\times$ ).

Fig. 11

Fluorescent monitoring of moving polystyrene particles labeled by FP-3056-2 fluorescence dye (mean diameter, $\sim 0.84\mu \mathrm{m}$ ) (left) and WBCs after intravenous injection of FITC-dextran (right) in blood flow of mouse skin microvessels (with the use of a dorsal skinfold chamber) (magnification: left, $10\times$ ; right, $40\times$ ).

Further development of integrated PTFC may include extending the spectral range from the visible spectral region to both the UV and the IR regions and the use of multispectral optical schemes. For example, this option may involve approaches that we have developed in the past, including fast spectral selection of different cells with a fast spectrum-scanning laser or with the simultaneous irradiation of cells with several laser beams at different wavelengths modulated with different frequencies (multiplex detection).¹⁸

4.

Conclusion

The experimental data that we presented demonstrate that TDM and PT microscopy in one integrated system, with rat mesentery as an animal model, beneficially supplement each other in in vivo FC applications without conventional labeling, including increased (at least four orders of magnitude) absorption sensitivity and improved spatial resolution (up to $300\text{to}350\mathrm{nm}$ ), as we demonstrated recently.¹³ These advantages clear the way for new potential applications such as the following:

With some modifications, this PTFC technique has the potential to be used in in vivo studies in humans. Noninvasive mode can be realized with thin structures transparent to light, such as eyelid or interdigital membrane, while an invasive approach might involve incorporation of a fiber-chip-based catheter in vessels or, by way of contrast, the use of a bypass whereby some amount of blood will pass through a capillary tube outside the body into the PTFC system. These approaches, however, must be very carefully verified first in an animal model.