2.1.

In Vivo Flow Cytometry with Photoacoustic Schematics

In PAFC, laser-induced PA waves (referred to as PA signals) from the cells in the blood (or lymph) flow are detected with an ultrasound transducer attached to the skin (Fig. 1 ). The PAFC system was built on the platform of an Olympus BX51 microscope (Olympus America, Inc. Center Valley, Pennsylvania) and a tunable optical parametric oscillator (OPO) pumped by a $\mathrm{Nd}:\mathrm{YAG}$ laser (both from Lotis Limited, Minsk, Belarus). Laser pulses had an 8-ns pulse width, a regular repetition rate of $10\mathrm{Hz}$ with ability to provide short-term $50\mathrm{Hz}$ , and a wavelength in the range of 420 to $2300\mathrm{nm}$ (Fig. 2 ). Laser energy was delivered to vessels with conventional lenses, or fibers.²⁸ PA signals from the ultrasound transducers (unfocused XMS-310, Panametrics, $10\mathrm{MHz}$ , (Olympus NTD Inc., Waltham, Massachusetts); focused cylindrical V316-SM, Panametrics, $20\mathrm{MHz}$ , focused lengths of 6.5 and $25\mathrm{mm}$ ; and customized resonance transducers) and amplifier (models 5662 and 5678, Panametrics, bandwidth $50\mathrm{kHz}$ to $5$ and $40\mathrm{MHz}$ customized amplifiers with adjustable high and low frequency boundary in the range of 50 to $200\mathrm{KHz}$ and 1 to $20\mathrm{MHz}$ , respectively; resonance bandwidth of 0.3 to $1\mathrm{MHz}$ ) were recorded with a boxcar (Stanford Research Systems, Inc., Sunnyvale, California) and a Tektronix TDS 3032B oscilloscope, and were analyzed with standard and customized software. The boxcar technique provided averaging PA signals from cells in the blood vessels and time-resolved discrimination from background signals from surrounding tissue because of the different time delays. The signals from the oscilloscope screen were also recorded with a digital camera (Sony Corporation of America, New York, NY) and video camera (JVC Company of America, Wayne, New Jersey).

Fig. 1

Principle of PA detection of single absorbing targets in blood flow in vivo.

Fig. 2

Schematic of the integrated PA/PT flow cytometry.

Because PAFC is a relatively new tool in biological cell study, it was integrated with well-established PT flow cytometry (PTFC) (Fig. 2) ^{14, 15, 16, 17, 18} to validate PAFC parameters. In particular, in the thermolens mode of PTFC, laser-induced temperature-dependent refractive heterogeneity around cellular endogenous absorbing structures (or contrast agents) cause defocusing of a colinear $\mathrm{He}–\mathrm{Ne}$ laser probe beam (model 117A $633\mathrm{nm}$ , $1.4\mathrm{mW}$ ; Spectra-Physics, Inc., Mountain View, California), and a subsequent reduction in the beam’s intensity at its center, which is detected with a photodiode (C5658; Hamamatsu Corp., Hamamatsu City, Japan) with a 0.5-mm-diameter pinhole as an integrated PT response from a whole cell.

The laser beams in most of the experiments had a circular geometry with diameters comparable to vessel diameters (see Sec. 4.1.3 for details). In specific cases to distinguish closely located cells in flow, the laser beam was adjusted in a linear (elliptical) configuration approximately $6\mathrm{\mu }\mathrm{m}$ wide with the use of an additional cylindrical lens and a slit. For the detection of rare cells, the beam diameter was expanded to preclude any missing cells because of the relatively low laser-pulse rates.

Navigation of the laser beams was controlled with a high-resolution $(\sim 300\mathrm{nm})$ transmission digital microscopy (TDM) with a Cascade 650 charge-coupled device (CCD) camera (Photometrics, Roper Scientific, Inc., Trenton, New Jersey). A fluorescence module with colored CCD cameras (Nikon DXM1200) was added to the integrated system (Fig. 2) to verify PA/PT data using selected cells tagged with specific fluorescent labels.

A feasibility study was first performed in vitro with cells in suspension in conventional microscope slides. To model flow conditions, a flow module fitted with a syringe pump-driven system (KD Scientific, Inc., Holliston, Massachusetts) was used with glass microtubes of different diameters in the range of 30 to $150\mathrm{\mu }\mathrm{m}$ to 1 to $4\mathrm{mm}$ .²⁶ This phantom provided flow velocities of 1 to $10\mathrm{mm}∕\sec$ , which are typical for blood flow in animal microvessels.²⁹

2.2.

Animal Model

The feasibility studies of PAFC in vivo involved mouse (a nude) and rat (White Fisher, F344) models. Most experiments for PA detection of circulating cells were performed in thin $(\sim 270\mathrm{\mu }\mathrm{m})$ , relatively transparent (i.e., also suitable for PTFC in transillumination mode) mice ear with well-distinguished blood microvessels [Figs. 3a and 3b ]. The ear blood vessels examined were located 30 to $100\mathrm{\mu }\mathrm{m}$ deep and had diameters in the range of 30 to $70\mathrm{\mu }\mathrm{m}$ and blood velocities of 1 to $5\mathrm{mm}∕\sec$ . After standard anesthesia (ketamine/xylazine, $50∕10\mathrm{mg}∕\mathrm{kg}$ ), an animal was placed on the customized heated microscope stage, together with a topical application of warm water, which provided acoustic matching between the transducer and the ear. Some experiments were performed also with rat ear [Fig. 3a], which still had a distinguished blood microvessel in spite of its thicker structure compared to mice ear [Fig. 3c]. Selected experiments were performed with the rat mesentery [Fig. 3d], which has an almost ideal biostructure for PAFC in vivo because it consists of very thin (7 to $15\mathrm{\mu }\mathrm{m}$ ) transparent connective tissue with a single layer of blood and lymph microvessels [Fig. 3d]. For this procedure, the rat was anesthetized (ketamine/xylazine, $50∕10\mathrm{mg}∕\mathrm{kg}$ ), and the mesentery after exposure was placed on a heated $(37.7°\mathrm{C})$ stage and bathed in warm Ringer’s solution ( $37°\mathrm{C}$ , pH 7.4).

Fig. 3

(a) Mouse ear photograph with the linear probe laser beam (red) and circular pump laser beam (green). (b) Mouse ear blood microvessels at $10\times$ magnification. (c) Rat ear blood microvessels. (d) Rat mesenteric structures.

2.3.

Samples

The PAFC’s capacity was estimated by intravenous injection of nanoparticles, contrast dyes, and bacteria into the animal tail vein. The gold nanoparticles such as GNRs and nanoshells (GNSs) were provided by the laboratory of Nanoscale Biosensors at the Institute of Biochemistry and Physiology of Plants and Microorganisms, Saratov, Russia. The GNRs were prepared using the seed-mediated growth method, as originally developed by Jana, Gearheart, and Murphy³⁰ and Nikoobakht El-Sayed³¹ with minor modifications.³² Both “seed” and growth solutions included a cetyltrimethylammonium bromide (CTAB) surfactant. The silica/gold nanoparticles were prepared by the well-established two-step protocol.^{33, 34} On the basis of TEM and the dynamic light scattering analyses, the geometrical parameters of nanoparticles were estimated to be of average diameter $15\mathrm{nm}$ and length $45\mathrm{nm}$ for GNRs, and average silica core diameter $60\mathrm{nm}$ and 20-nm gold coating (total diameter around $100\mathrm{nm}$ ) for GNSs. For both sorts of particles, the concentration was about ${10}^{10}\mathrm{particles}∕\mathrm{ml}$ . Two types of GNRs and GNSs were used in our experiments: 1. nanoparticles without surface modification, and 2. nanoparticles functionalized by thiol-modified polyethylene glycol (PEG) according to the established protocol.³⁵

Single-walled and multiwalled carbon nanotubes (CNT) were purchased from Carbon Nanotechnologies Incorporated (Houston, Texas) and Nano-Lab Incorporated (Newton, Massachusetts), respectively. The CNT samples used in this study were processed as described previously.^{36, 37} The average length and diameter were 186 and $1.7\mathrm{nm}$ for the single-walled CNT, and 376 and $19.0\mathrm{nm}$ for the multiwalled CNT.

The bacterium S. aureus strain designated UAMS-1 was isolated from a patient with osteomyelitis at the McClellan Veterans Hospital in Little Rock, Arkansas.³⁸ The strain was deposited with the American Type Culture Collection and is available as strain ATCC 49230. UAMS-1 was cultured in tryptic soy broth and grown aerobically for $16\mathrm{h}$ at $37°\mathrm{C}$ . Cells were harvested by centrifugation, resuspended in sterile phosphate buffered saline (PBS) and incubated with CNTs or ICG as follows. Before incubation, ICG was filtered through a $0.22\text{-}\mathrm{\mu }\mathrm{m}$ pore-size filter. Aliquot $\left(150\mathrm{\mu }\mathrm{l}\right)$ of bacteria in suspension was incubated with $150\mathrm{\mu }\mathrm{l}$ $\left(375\mathrm{\mu }\mathrm{g}\right)$ of ICG solution for $30\min$ and $2\mathrm{h}$ at $37°\mathrm{C}$ . Labeled bacteria were centrifuged ( $5000\mathrm{rpm}$ ; $3\min$ ) and the pellet was resuspended in PBS. Before incubation, the CNT solutions were further treated with interrupted ultrasound (US) for $10\min$ : $1.5\min$ of US (power $3\mathrm{W}$ ) and $0.5\min$ of interruption. Aliquot $\left(150\mathrm{\mu }\mathrm{l}\right)$ of bacteria in suspension was incubated with $150\mathrm{\mu }\mathrm{l}$ of CNT solution for $30\min$ and $2\mathrm{h}$ at room temperature. Labeled bacteria were centrifuged ( $\mathrm{10,000}\mathrm{rpm}$ ; $5\min$ ) and the pellet was resuspended in PBS.

E. coli K12 strain was obtained from the American Type Culture Collection (Rockville, Maryland). E. coli was maintained in Luria-Bertani (LB) medium (1% tryptone; 0.5% yeast extract; 0.5% $\mathrm{NaCl}$ ; pH 7). Aliquot $\left(100\mathrm{\mu }\mathrm{l}\right)$ of E. coli in PBS was incubated with $100\mathrm{\mu }\mathrm{l}$ of the aformentioned CNT solution for $60\min$ at room temperature.

Indocyanine green (ICG) was purchased from Akorn Incorporated. ${\mathrm{Lymphazurin}}^{™}$ 1 % (Isosulfan Blue) was manufactured by Ben Venue Labs Incorporated.

3.1.

Evaluation of Photoacoustic Flow Cytometry’s Noninvasiveness

For noninvasive cell studies, the principal task was to evaluate the laser-induced damage threshold at the single cell level. We performed these important measurements using recently developed methodology, which assumes the specific changes in PT images and PT responses from individual cells as a function of the pump-laser energy.^{21, 39} The method was verified with a conventional assay such as Trypan Blue exclusion. In agreement with the previous study,³⁹ for relatively strong absorbing RBCs in the visible spectral range, the photodamage threshold ${\mathrm{ED}}_{50}$ (50% of cells damaged) in vitro varied in the range of 1.5 to $5\mathrm{J}∕{\mathrm{cm}}^2$ with a laser wavelength in the range of 417 to $555\mathrm{nm}$ (for a $20\text{-}\mathrm{\mu }\mathrm{m}$ laser spot). However, for other cells with lower absorption, the thresholds were higher: 12 to $42\mathrm{J}∕{\mathrm{cm}}^2$ for lymphocytes and 36 to $90\mathrm{J}∕{\mathrm{cm}}^2$ for K562 blast cells. In the NIR spectral range, where most cells have minimal absorption, the ${\mathrm{ED}}_{50}$ damage thresholds were lower than those for wavelengths in the visible-spectral range. Specifically, the damage thresholds for RBCs and WBCs in the spectral range of 860 to $920\mathrm{nm}$ (Table 1 ) were more than 1 order magnitude less compared to those in the visible spectral range. The measurement of laser-induced cell damage in vivo using change in PT responses¹⁶ revealed that within the accuracy of experiments $(\sim 15\%)$ , the photodamage cell thresholds in vitro and in vivo were similar. A PAFC’s noninvasiveness was achieved in most studies by choosing laser fluence much below the estimated cell photodamage threshold according to the laser safety standard of $34\mathrm{mJ}∕{\mathrm{cm}}^2$ in the NIR spectral range for the condition of our experiments.⁴⁰

3.2.

Photoacoustic Flow Cytometry’s Parameter Verification with Conventional Contrast Agents

The capability of PAFC was estimated first with standard contrast vital dyes, which are widely used to study physiological function of blood and lymph system in normal or for disease diagnostics, such as cancer or cardiovascular and lymphatic disorders, among others.^{7, 27} Specifically, in the current study we extended PAFC on the kinetic study of Lymphazurin, which was used before, mostly for the delineation of lymphatic vessels.^{41, 42} Lymphazurin’s exact pharmacological action and its kinetics in the blood circulation and surrounding tissue are still unclear.

An aqueous solution of Lymphazurin ( $200\mathrm{\mu }\mathrm{l}$ ; 1% concentration) was injected into mouse and rat blood circulatory systems through the tail vein. The observed dynamic effects were very similar in both mice and rats. In particular, a few minutes after the injection, mouse tail, nose, and lips turned blue quickly with complete coloration of the whole skin body within $15\min$ [Fig. 4b ]. For approximately $1\mathrm{h}$ , the animal’s skin remained blue. Then, the animal’s skin became lighter and returned to normal after $\sim 3\mathrm{h}$ . After anesthesia ended, visual qualitative monitoring revealed no adverse effects, including no changes in vital signs and no differences in the animal behavior compared to the control group with no dye injection. These findings are in line with the published data of low toxicity and safety regarding this dye at the doses used.⁴¹ Selected animals were sacrificed $3\mathrm{h}$ after injections and were anatomically examined. No organ accumulation of the dye was seen except the urine tract and billary excretion, which is in agreement with other previously reported data.⁴¹ This study suggested extremely high permeability of dye through the blood vessel walls, followed by fast distribution into tissue.

Fig. 4

Blue mouse: mouse under anesthesia (a) before and (b) after 15-min intravenous administration of Lympazurin.

Lymphazurin has a relatively flat absorption band in the spectral range of 570 to $660\mathrm{nm}$ with a sharp boundary (i.e., outside this range, the absorption quickly drops down). For PA detection, we have chosen the wavelength of $650\mathrm{nm}$ , where absorption of Lymphazurin is close to maximum, while the blood absorption drops down to the level which is almost 1 order magnitude less compared to maximum blood absorption near $580\mathrm{nm}$ .⁴³ The PA signals from blood vessels [e.g., $240\mathrm{mV}$ in Fig. 5a ] were slightly higher than the PA background signals from surrounding tissue [e.g., $160\mathrm{mV}$ in Fig. 5b], i.e., approximately 1.5 times. The continuous monitoring of PA signals from the ear blood microvessels with average diameter $50\mathrm{\mu }\mathrm{m}$ after intravenous administration of Lymphazurin with the standard (1%) concentration in $200\mathrm{\mu }\mathrm{l}$ solution revealed fast (within a few minutes) Lymphzurin appearance in blood flow, followed by its clearance half-life of $30\min$ and total disappearance from the blood within approximately $50\min$ (Fig. 6 ). Maximum PA signals from the blood vessel after dye administration [Fig. 5c] were approximately three times higher than the PA signals from blood vessels before dye administration [Fig. 5a]. The PA signals from tissue around the vessels after dye injections [Fig. 5d] gradually increased 2 to 2.5 times during 15 to $20\min$ and then remained relatively constant for approximately 1 to $1.5\mathrm{h}$ . This level was comparable to or even a little higher than the background PA signals from blood vessels when Lymphazurin was cleared from the blood pool. The measurement of PA signals during the scanning laser beam across ear tissue revealed the readable increase in the PA signals in some local areas that can probably be associated with dye uptaken by lymphatics. We did not observe readable PA signal changes in amplitude and shape as precursors of photodamage phenomena³⁹ that suggest noninvasive condition for one pulse irradiation at used laser fluence in the range $30\mathrm{mJ}∕{\mathrm{cm}}^2$ , although this issue requires further verification.

Fig. 5

PA signals from blood flow in (a) mouse ear vessel with diameter $\sim 50\mathrm{\mu }\mathrm{m}$ and (b) from surrounding skin before dye injection. (c) PA signal from blood flow with Lymphazurin $5\min$ after injection (d) PA signal from skin $20\text{-}\min$ after dye injection. Laser parameters: wavelength $650\mathrm{nm}$ , fluence $30\mathrm{mJ}∕{\mathrm{cm}}^2$ . Amplitude/time scale: (a), (b), and (c) $200\mathrm{mV}∕\mathrm{div}∕1\mathrm{\mu }\mathrm{s}∕\mathrm{div}$ , (d) $500\mathrm{mV}∕\mathrm{div}∕1\mathrm{\mu }\mathrm{s}∕\mathrm{div}$ .

Fig. 6

Quantitative PA monitoring of Lymphazurin in blood flow of rat ear microvessel with diameter $50\mathrm{\mu }\mathrm{m}$ after intravenous injection of dye in the tail vein. Laser parameters: wavelength $650\mathrm{nm}$ , fluence $30\mathrm{mJ}∕{\mathrm{cm}}^2$ , beam diameter $40\mathrm{\mu }\mathrm{m}$ .

3.3.

Detection of Circulating Gold Nanoparticles

The $250\text{-}\mathrm{\mu }\mathrm{l}$ suspension of GNRs in concentration of ${10}^{10}∕\mathrm{ml}$ was injected to the rat blood circulation through the vein tail followed by continuous monitoring of PA signals from blood vessels with $50\mathrm{\mu }\mathrm{m}\mathrm{diameter}$ in the rat mesentery. The chosen laser wavelength of $830\mathrm{nm}$ was close to the maximum absorption of GNRs. Uncoated GNRs were rapidly cleared from the blood circulation within 1 to $3\min$ preferentially by the reticuloendothelial system (RES). After PEGylated GNR injection, the fluctuated strong PA signals appeared with the amplitudes significantly exceeding the PA background signals from blood vessels approximately within the first minute and lasted from $14$ to $25\min$ , depending on the individual animal (Fig. 7 ). Simultaneously, the slightly fluctuated PA background increased approximately 1.5 to 2 times above the background from blood at $3\min$ and lasted around $4\min$ . These measurements were performed in three animals, and the average data of the number of circulating GNRs per minute are shown in Fig. 8 . In average, the maximum number of individual PA signals was achieved 3 to $5\min$ after injection, with gradual decrease in the rate of their appearance as time elapsed during $15\min$ . In one animal, during the clearance process, a relatively strong random fluctuation of the PA signals was observed. This can be associated with a nonoptimal tail injection when some GNRs could accumulate near the injected site and then enter into the vein vessels with some delay compared to the initial injection. The verification of these data is currently under investigation.

Fig. 7

PA signals from circulating GNRs in blood flow in the rat mesentery microvessel with diameter of $50\mathrm{\mu }\mathrm{m}$ as a function of time postinjection. Laser parameters: wavelength $830\mathrm{nm}$ , laser fluence $100\mathrm{mJ}∕{\mathrm{cm}}^2$ . The amplitude/time scale on oscilloscope signals is $100\mathrm{mV}∕\mathrm{div}∕4\mathrm{\mu }\mathrm{s}∕\mathrm{div}$ .

Fig. 8

Normalized number of circulating GNRs in blood microvessels of the rat mesentery as a function of time postinjection. Dashed curve shows an average data with three rats.

3.4.

Detection of Single Circulating Bacteria

3.4.1.

Detection of Staphylococcus aureus

The capability of PAFC to monitor circulating single S. aureus in blood flow was estimated with the mouse ear model. Because the endogenous absorption of bacteria was relatively weak compared to blood absorption even in the NIR spectral range, bacteria were labeled with NIR absorbing contrast agents such as ICG, GNRs, GNSs, and CNTs. The incubation of S. aureus with these agents was performed for $30\min$ and $2\mathrm{h}$ at $37°\mathrm{C}$ without antibodies to avoid the potential influence of immunogenicity on bacteria circulation. Except CNTs, the other labeling agents showed slight toxicity at 2-h incubation. Hence, 30-min incubation was used in most experiments to avoid their potential cytotoxic effects. The labeling efficiency was estimated using methodology of the PT aggregation/clusterization assay.^{26, 44} In particular, Fig. 9 shows the PT thermolens signals in the time-resolved mode (Sec. 2.2) from bacteria alone [Fig. 9a] and from bacteria labeled with ICG [Fig. 9b] and CNTs [Fig. 9c]. At relatively high laser fluence $(\sim 24\mathrm{J}∕{\mathrm{cm}}^2)$ , which nevertheless was below the bacteria photodamage in the NIR spectral range used $\left(795\mathrm{nm}\right)$ , the classic linear PT response²¹ was observed for the control bacteria alone, as indicated by an initial peak due to fast laser-induced heating of the bacteria and a slower exponential tail corresponding to bacteria cooling through heat diffusion into the surrounding solution [Fig. 9a]. The high laser fluence required, and nevertheless, still realitively low PT and PA signal amplitudes [data not shown; similar to those in Fig. 5b] prevented us using bacteria without labeling for their detection in vivo in the background of RBCs, because of the possible RBC damage and the influence of background absorption from blood. Staining S. aureaus with ICG led to significant PA signal amplitude increase [Fig. 9b], on average almost 2 orders of magnitude. This allowed us to reduce the laser fluence to safety level $34\mathrm{mJ}∕{\mathrm{cm}}^2$ .

Fig. 9

PT signals from (a) S. aureus cluster, (b) S. aureus stained with ICG, and (c) S. aureus labeled with carbon nanotubes (CNTs). Amplitude/time scale (laser pump wavelength)/(laser fluence): (a) 10 mV/div/1 $\mathrm{\mu }\mathrm{s}$ /div/795 nm/24 $J∕{\mathrm{cm}}^2$ ; (b) 100 mV/div/1 $\mathrm{\mu }\mathrm{s}$ /div/795 nm/80 $\mathrm{mJ}∕{\mathrm{cm}}^2$ ; and (c) 50 mV/div/1 $\mathrm{\mu }\mathrm{s}$ /850 nm/24 $\mathrm{mJ}∕{\mathrm{cm}}^2$ .

Labeling S. aureus with CNTs and their clusters yielded more dramatic increase of PT signal amplitude, which was several times higher than that from bacteria with ICG at the same laser energy. The amplitude of the PT response rapidly increased with increases in the pulse energy, until the overheated local effects around CNTs clusters led to the bubble formation, which resulted in a negative peak [Fig. 9c] due to bubble-dependent changes in local temperature, refractive index, and light scattering phenomena.^{21, 38} The degree of increase in PT signal amplitudes and bubble-associated effects were used to optimize the labeling procedure almost in real time with different nanoparticles. The best results were obtained with GNRs, but its toxicity related to the preparation procedure required further studies.²⁷ Minimal bacteria binding was obtained both with nonconjugated and PEGyted GNSs. Thus, for next experiments in vivo, we have chosen ICG and CNTs because of their relatively high labeling efficiency and low toxicity.

Labeled bacteria in suspension, with volumes of $100\mathrm{\mu }\mathrm{l}$ and concentration of $5\times {10}^5∕\mathrm{ml}$ , were injected into the mouse’s circulatory system through the tail vein, followed by clearance monitoring with the PAFC in mouse ear blood microvessels with an average diameter of $50\mathrm{\mu }\mathrm{m}$ (Fig. 10 ). The obtained results were similar to both ICG and CNT contrast agents: the rapid appearance of bacteria in the ear blood microvessels after injection within the first minute, followed by their elimination from the blood circulation after 3 to $5\min$ . The periodical PA monitoring of mice blood vessels within a few days revealed that bacteria with CNTs kept appearing for more than 2 days with the average rate of one PA signal per $3\min$ . The complete clearance with no sign of circulating bacteria was observed $60\mathrm{h}$ after injections.

Fig. 10

Normalized number of circulating S. aureus in blood microvessels of mouse ear as a function of time postinjection, and label type (ICG and CNTs). Laser parameters: wavelength 795 and $850\mathrm{nm}$ , respectively. Laser fluence: 50 and $20\mathrm{mJ}∕{\mathrm{cm}}^2$ , respectively.

Laser fluence range in these particular experiments was in the range of 20 to $50\mathrm{mJ}∕{\mathrm{cm}}^2$ , which is much less than the damage threshold for RBCs at wavelengths of 790 to $850\mathrm{nm}$ (Table 1). However, multiple laser exposure to bacteria with CNTs at $50\mathrm{mJ}∕{\mathrm{cm}}^2$ led to a gradual decrease (several percents) in PA signal amplitude from pulse to pulse that can be associated with strong absorption by CNTs, leading to thermal and bubble-related disintegration of CNT clusters.³⁷

3.4.2.

Photoacoustic detection of Escherichia coli labeled with carbon nanotubes

The methodology similar to the aforementioned was also applied for real-time monitoring of E. coli labeled with CNTs. Figure 11a exemplifies TEM images of an E.coli before and after 0.5-h incubation with CNTs, which demonstrates the excellent binding efficiency of CNTs at the bacterial surface with formation of local CNT clusters. Furthermore, using the standard viability kits revealed that bacteria viability was not affected by CNTs.³⁷ The CNT clusters provided more effective heating, and thus generation of PT signals from individual bacteria which exceeded the background PA signals from the blood pool [Fig. 11b, top and bottom oscilloscope signals, respectively]. Bacteria labeled with CNTs have been intravenously administrated to mice through the tail vein and monitored through the CNT’s intrinsic NIR absorption at the wavelength of $850\mathrm{nm}$ with PAFC. After fast appearance of the bacteria, their concentrations in the blood decreased almost exponentially within a half-life of $6\min$ [Fig. 11b]. The absence of bi- or multiexponential kinetics indicates that there was no significant temporary accumulation of bacteria in a specific area (e.g., vessel wall or liver) that could act as reversible reservoirs. After a few days, the experimental animals displayed normal behavior and no evidence of adverse effects from bacteria injections. Therefore, we deduce an absence of acute toxicity for the dosage used here.

Fig. 11

(a) TEM images of an E. coli fragment before (left) and after (right) incubation with CNTs. Arrows indicate CNT clusters at the bacteria surface. Scale bars represent $500\mathrm{nm}$ . Normalized number of circulating E. coli in blood microvessels of mouse ear as a function of time postinjection. Oscilloscope signals: PA signals from labeled E.coli in blood (top) and from blood alone (bottom). Amplitude/time scale: $200\mathrm{mV}∕\mathrm{div}∕2\mathrm{\mu }\mathrm{s}∕\mathrm{div}$ . Scale bar represents amplitude/time scale: $200\mathrm{mV}∕\mathrm{div}∕2\mathrm{\mu }\mathrm{s}∕\mathrm{div}$ , respectively. Laser parameters: wavelength $850\mathrm{nm}$ , laser fluence $100\mathrm{mJ}∕{\mathrm{cm}}^2$ .

Table 1

Photodamage threshold for single blood cells in near-IR spectral range.

Wavelength, nm		740	760	780	800	820	840	860	880	900	920	960
Photodamage threshold ${\mathrm{ED}}_{50}$ , $\mathrm{J}∕{\mathrm{cm}}^2$ (for $8\mathrm{nsec}$ pulse width)	Rat RBCs	6.9	6.8	17.7	17.5	28.0	43.5	43.8	76.5	69.4	77.7	33.5
	Rat WBCs	21.7		152	219	251		730			357	48.8

4.1.

Features of Photoacoustic Flow Cytometry

The detection of fast moving absorbing micro- or nano-objects of different origins on the background of the absorption of blood cells and surrounding tissue presents a new challenge for the PA technique. For noninvasive real-time detection of living cells or bacteria, most laser parameters are important, including wavelength $\lambda$ , pulse duration $t_P$ , pulse repetition rate $f$ , and laser fluence $\Phi$ .

4.2.

Kinetics of Conventional Contrast Dyes

The conventional application of contrast dyes is based on their absorption or fluorescent properties, their good permeability, fast diffusion, and low toxicity. In particular, after administrating some dyes (e.g., Evans Blue or Lymphazurin) the vessels can be easily identified visually with the naked eye because they are becoming darker compared to the surrounding tissue. The ability of PAFC to detect conventional agents was related to their limited quantum yield, typically in the range of 1 to 20%,⁶ resulting in transformations of the most absorbed energy into heat, to which the PA technique is very sensitive. Recent studies demonstrated the capability of ICG and Evans Blue dyes to enhance the PA image contrast of the blood vessels.^{46, 47} However, most PA imaging algorithms currently in use for signal acquisition are not quite suitable for rapid real-time PA monitoring of contrast agent kinetics in fast blood circulation due to their time-consuming signal acquisition process (in most cases seconds, if not minutes). In addition, PA imaging still suffers from relatively poor spatial resolution (in the best case $\ge 50$ to $100\mathrm{\mu }\mathrm{m}$ ),²² preventing its cytometric application at the single cell level. As described, the time-resolved PAFC with its high spatial (6 to $20\mathrm{\mu }\mathrm{m}$ , see above, and 0.5 to $1\mathrm{\mu }\mathrm{m}$ in imaging PT mode¹⁷), temporal ( ${10}^{-4}$ to ${10}^{-1}\mathrm{s}$ , depending on the laser repetition pulse rate), and spectral (line width $\le 0.5\mathrm{nm}$ ) resolutions in the broad spectral range of 420 to $2300\mathrm{nm}$ has great potential for studying the kinetics of dyes with different absorption in the blood and lymph systems on different animals, including small mesentery models. We expect quick translation of this technology to humans, because some dyes (e.g., ICG) are already Federal Drug Administration (FDA) approved and lasers are used to meet the safety standard. PAFC should be applicable to monitor the kinetics of dyes simultaneously in blood flow, in tissue, and potentially in small lymphatics.¹⁸ The obtained data in this study strongly suggest the high possibility of extending the traditional fluorescent contrast agents as potential PA contrast agents. Furthermore, the combination of PAFC with such agents may extend PAFC’s applicability. Detection of both fluorescent light and heat can increase the diagnostic value of integrated in vivo FC with PA and fluorescence detection techniques, especially far-red and NIR spectral ranges coinciding with the window transparency of specific tissue. It is important to note that the high absorption sensitivity of the PA technique made it possible to use relatively low concentrations of most dyes, especially with low quantum yields and a low laser energy levels, without notable signs of dye toxicity or photobleaching.

4.3.

Detection of Single Nanoparticles In Vivo

Different nanoparticles are intensively being developed for biomedical applications, including diagnostics and therapy. ^{33, 34, 35, 36, 44} Before any clinical application of nanoparticles can be feasible, it is imperative to determine critical in vivo parameters, namely pharmacological profiles, including the clearance half-time from the blood circulation.⁴⁸ In this study, we present previously unreported data on blood circulation of GNRs, which showed promise both as PA and PT labels and PT therapeutic agents for killing individual cancer cells and bacteria.^{27, 49} In particular, the PEGylated GNRs remained for a longer time in the blood circulation after intravenous administration, compared to the nonPEGylated nanoparticles. Fast clearing of noncoated GNRs can be associated with quick GNR aggregation in blood flow immediately after intravenous injection, and the influence of the incompletely purified and partly toxic particle precursors, especially CTAB, which lead to fast responses of RES and the immune system.²⁷ It is known that PEG coating leads to reduced nanoparticle aggregations, protein binding, and adhesion to blood and endothelial cells.^{33, 35} However, the observed clearance rate of around $15\min$ for the PEGylated GNRs was not as long as expected. It should be related to its nonoptimal preparation procedure, which leads to the presence of remaining toxic impurities or incomplete coating. Thus, the described technique should also be useful to the routine evaluation of the possible influence of the natural properties of nanoparticles, the protective materials, and the coating procedures on their clearing. It should be noted that in the current study, the nonspecific binding and uncontrolled aggregation, which should be avoided for biomedical applications, were the subjects of discussion. Therefore, specific binding and aggregation (clustering) should be the next logical reseach to be conducted using additional bioagents attached to the particles.⁴⁴

The detection threshold sensitivity of PAFC in this application can be estimated as follow. The 250- $\mathrm{\mu }\mathrm{l}$ GNR suspension $(\sim {10}^{10}∕\mathrm{ml})$ was injected in the blood pool of a rat with total blood volume of $25\mathrm{ml}$ . Assuming the ideal condition with no loss of GNRs during injection and circulation, the average GNR concentration in the rat blood pool should be ${10}^8∕\mathrm{ml}$ . The irradiated (detected) volume in blood vessels is limited by a laser beam diameter of $25\mathrm{\mu }\mathrm{m}$ and path length of around $45\mathrm{\mu }\mathrm{m}$ across the vessel (i.e., comparable with vessel diameter), which gives the volume around $2.2\times {10}^{-8}{\mathrm{cm}}^3$ . These parameters correspond to the estimated maximum number of GNRs in the detected volume around 2 to 3. The possible loss of GNRs during injections, such as adhesion in blood vessels and elimination by RES, may decrease this value several times, especially at the end of the elimination process. This strongly implies that the number of GNRs in the detection volume should be close to 1. Therefore, each pulse in Fig. 7 may be associated with the PA signals from individual GNRs. However, immediately after injection, local concentration of GNRs in the blood flow before redistribution in larger volume should be higher (i.e., many GNRs in the detected volume). This was confirmed by the increases in both individual PA signal rates and average PA background above the blood background in approximately 4 to $9\min$ (Fig. 7). In addition, in this time period, the PA signals with larger amplitudes were observed after each laser pulse. Essentially, these fluctuations were not accompanied by decrease of PA signal amplitude to zero, supporting the hypothesis of random appearance of several GNRs in the detected volume at the same time. This is similar to the aforementioned PA monitoring of Lymphazurin, when relatively large amounts of homogenously distributed absorbing molecules appear in the detected volume, thus providing relatively stable continuous (i.e., not fluctuated) PA signals. On the contrary, only rare PA signals appeared even after many pulses at $10\min$ , suggesting the observation of the GNR clearance.

Laser fluence used in this study was of $100\mathrm{mJ}∕{\mathrm{cm}}^2$ , which is at least 2 orders of magnitude less than the damage threshold for RBCs at the wavelength of $830\mathrm{nm}$ (Table 1). However, a recent study has suggested that high absorption of noble metal nanoparticles may lead to laser-induced nanoparticle modification, including their melting, evaporation, and explosion, accompanied by enhanced bubble and shock wave formation, even at relatively low laser fluence.⁵⁰ These nonlinear effects can lead to strong enhancement of PA signals from single gold nanoparticles due to the aforementioned highly localized shock wave and bubble formation, which may further increase PAFC sensitivity by the cost of the minimal invasiveness of the PA measurement at the current experimental conditions. The estimated PAFC sensitivity in the linear mode at lower laser energy is at least 3 to 5 gold nanoparticles in the detected volume. We also observed some additional PA signal enhancement for PEG-coated gold nanoparticles due to the change in their thermal properties and their average size increase (up to 50 to $100\mathrm{nm}$ ), leading to more effective nanoparticle heating and PA signal formation. This effect was more profound for GNSs due to probably more effective explosion of thin gold layers around a relatively large silica core (we called it the spherical piston explosion model). In addition, these effects might be important to increase the efficiency of PT nanotherapy with nanoparticles and their clusters used as a PT sensitizer for cancer cells or bacteria.⁵¹

The results obtained so far demonstrate unprecedented high sensitivity of the developed PAFC schematics, enabling single gold nanoparticle detection in blood flow in vivo. It also suggests the possibility of using a limited number of gold nanoparticles (e.g., 10 to 30) for cell labeling to achieve reliable PA detection of cells in the blood circulation.²⁷ In addition, the routine monitoring of the clearance rate of different nanoparticles may be combined with simultaneous real-time PA monitoring of nanoparticle accumulation in different organs in normal and pathological states (e.g., in tumors, lymph nodes, or infected areas).

4.4.

Real-Time Detection of Single Circulating Bacteria In Vivo

Despite major advances in medicine in the last decade, microbiologically based diseases continue to present enormous global health problems, especially due to the appearance of multidrug-resistant bacteria strains. The critical steps in the development of bacterial infections include their penetrations into the blood and lymph system, their interactions with blood cells in flow or with endothelial cells, and their toxin translocations in the host organisms.⁵² Increasing evidence has implicated the bacterial haematological and lymphatic translocation as the main sources for the induction of sepsis. Surprisingly, little is known about bacteria circulation in the blood pool, its clearance and adherence rates, and other kinetic parameters, which might be very important for understanding the transition from the bacteremic stage to the tissue-invasive stage of disease to develop effective therapies of infections.

In the current study, CNTs demonstrated great potential for labeling bacteria such as S. aureus and E. coli for FC diagnostics. High CNT binding efficiency may eliminate the bioconjugation steps with biological recognition components, such as antibodies. This alleviates the potential influence of the additional biocomponents on clearance rates. Besides binding efficiency, the advantage of CNTs as contrast agents is their natural strong absorption in the NIR spectra. Therefore, compared to gold nanoparticles such as GNRs or GNSs, which require time-consuming preparation procedures, CNT itself does not need to be modified in any way to make it responsive to the NIR laser irradiation. Thus, the intrinsic NIR label absorption allows unambiguous observation of unmodified nanoparticles in vivo. To the best of our knowledge, the results from this study demonstrate for the first time that CNTs serve as excellent NIR high-contrast agents for PT and PA diagnostics, in addition to the previous reports showing their potential as high contrast PT sensitizer for cancer⁵³ and antimicrobial therapy.³⁷

The clearance of bacteria with CNTs was a little slower than that with ICG (Fig. 10), which can be explained by a slight ICG toxicity and more effective RES response, on the one hand, or by more effective bacteria labeling with CNTs (i.e., the presence of more labeled bacteria in the blood) on the other hand. The observed advantage of CNTs as a PT label compared to ICG (Fig. 9) can be explained by the stronger absorption of individual CNTs, and especially by CNT clusters, compared to ICG absorption.^{21, 33, 49} However, the absorption band of ICG is more narrow and specific compared to the not quite selective absorption of CNTs. Thus, besides binding efficiency, the choice of nanoparticles and dyes may depend also on requirements to selectivity. We plan also to use nanoscale lyposomes filled up by ICG to increase the ICG local absorption as well as liposomes filled with CNTs and gold nanoparticles. However, such strong absorption may lead to the appearance of nonlinear phenomena, thus decreasing the labeled bacteria (and cell)²⁷ damage threshold. This implies that laser energy should be carefully selected to minimize such effects, although this should not be a problem during a single exposure of circulating bacteria with CNTs in blood flow.

The ultrasensitive, rapid PA detection of infection at a single bacteria level may supplement or replace conventional assays, for example, the time-consuming polymerase chain reaction (PCR) and others that are available currently in vitro only. The described PA technique is capable of detecting approximately ${10}^2$ bacteria in the peripheral circulation of the small animal, such as mice and rats every $\sim 15\min$ , with the potential to improve this parameter to ten bacteria. Theoretically, this technique enables us to detect just a single circulating cell or bacteria in the whole mice blood pool during continuous PA monitoring in the relatively large vessels approximately every hour. The obtained clearance rate for S. aureus (3 to $5\min$ ) is in line with data obtained with fluorescent-labeled bacteria monitored with the dorsal skin-fold chamber technique.⁵² However, the time-resolved PAFC revealed the fluctuation of clearance rates (Fig. 10). There are two possible explanations for the nature of these observations. First, this may be related to the bacterial adherence to endothelial cells, followed by bacteria extravasations in different organs, which are considered to play a central role in the invasive disease stage.⁵¹ Indeed, the scanning of the laser beam along vessels revealed the appearance of strong PA signals at different locations, indicating the adherence of bacteria with CNTs along the vessel wall. The adherence might be a dynamic process and can lead to periodical detachment of bacteria from blood vessel walls. We assume that the adherence can be influenced by bacteria labeling, especially with CNTs, which may prevent effective interaction of bacteria with other cells due to their high-density clustering on bacteria surfaces. Additionally, one cannot exclude the possibility of partial CNT detachment, especially in the relatively large CNT clusters, from bacteria surfaces during their interactions with other cells in fast blood flow. The detached CNTs could circulate longer than bacteria themselves, at least a few hours. In addition similar to the GNRs, the developed PAFC prototype demonstrated extremely high sensitivity in detecting individual CNTs and especially their small clusters. However, this topic is outside the scope of this work.

The scanning of the focused laser beam across ear tissue surrounding the examined blood vessels (PA mapping) revealed the strong local PA signals that can be associated with the presence of labeled bacteria in interstitium. This technique in vivo and ex vivo also revealed strong PA signals in liver (e.g., $24\mathrm{h}$ after administration), indicating the accumulation of bacteria in this organ. Additional studies are desired to elucidate the behavior of CNT-coated bacterial cells, as well as to optimize the processes of labeling bacteria with CNTs.

Compared to recent experiments with fluorescent or radioactive tracers with periodic examination of blood samples in vitro,^{48, 52} we are confident that our time-resolved PA method with continuous monitoring in vivo is better suited for studying clearance rates of bacteria, as well as conventional and novel contrast agents, especially when elimination time is very short, a few minutes if not seconds.