December 15, 2006 / Vol. 31, No. 24 / OPTICS LETTERS 3623

In vivo photoacoustic flow cytometry formonitoring of circulating single cancer cells

and contrast agents

Vladimir P. Zharov, Ekaterina I. Galanzha, and Evgeny V. ShashkovUniversity of Arkansas for Medical Sciences, Little Rock, Arkansas 72205

Nicolai G. Khlebtsov and Valery V. TuchinInstitute of Optics and Biophotonics, Saratov State University, Saratov 410012, Russia

Received July 17, 2006; revised September 7, 2006; accepted October 2, 2006;posted October 11, 2006 (Doc. ID 73127); published November 22, 2006

A new photoacoustic flow cytometry was developed for real-time detection of circulating cells, nanoparticles,and contrast agents in vivo. Its capability, integrated with photothermal and optical clearing methods, wasdemonstrated using a near-infrared tunable laser to characterize the in vivo kinetics of Indocyanine Greenalone and single cancer cells labeled with gold nanorods and Indocyanine Green in the vasculature of themouse ear. In vivo applications are discussed, including selective nanophotothermolysis of metastatic squa-mous cells, label-free detection of melanoma cells, study of pharmokinetics, and immune response to apop-totic and necrotic cells, with potential translation to humans. The threshold sensitivity is estimated as onecancer cell in the background of 107 normal blood cells. © 2006 Optical Society of America

OCIS codes: 170.0180, 170.1530, 170.6920, 170.1790.

Flow cytometry (FC) is a well-established diagnosticmethod that revolutionized cell diagnostics in vitro.Nevertheless, the invasive extraction of cells from aliving organism may introduce artifacts and make itimpossible to conduct long-term monitoring of thecells in the complex natural environment. Recently,in vivo FC with fluorescent and photothermal (PT)techniques in the visible spectral range was success-fully used for monitoring blood and cancer cells in thevasculature of the mouse ear and rat mesentery.1–3

However, fluorescent labeling is still subject to cyto-toxicity, while the label-free PT technique is cur-rently limited to the transillumination (forward)mode. We introduce a new FC with near-IR photoa-coustic (PA) detection of circulating cells in live ani-mals in a backward model, either without labeling orwith nontoxic gold labels.

In photoacoustic flow cytometry (PAFC), the indi-vidual cells in blood or lymph flow are irradiated withone or a few focused (or fiber-delivered) laser beamsof different wavelengths; PA pulses from cells are de-tected with an ultrasonic transducer attached to theskin (Fig. 1). The PAFC system was built on the plat-form of an Olympus BX51 microscope (OlympusAmerica, Inc.) and a tunable optical parametric oscil-lator (OPO) pumped by a Nd:YAG laser (both fromLotis, Ltd., Minsk, Belarus). Laser pulses had an8 ns pulse width, a repetition rate of 10 Hz, and awavelength in the range of 420–2300 nm. PA signalsfrom an ultrasonic transducer (Model XMS-310,Panametrics) and amplifier (Model 5662, Panamet-rics) were recorded with a Boxcar (Stanford ResearchSystems, Inc.) and a Tektronix TDS 3032B oscillo-scope. The Boxcar technique provided averaging ofPA signals from cells and time-resolved discrimina-tion from background signals in skin and vesselwalls. The signals from the oscilloscope were re-corded with a digital video camera (JVC, Inc.). The

PAFC was integrated with a photothermal flow

0146-9592/06/243623-3/$15.00 ©

cytometer2,3 (PTFC). In the thermolens mode of thePTFC, laser-induced, temperature-dependent varia-tions of the refractive index around cellular absorb-ing structures or contrast agents caused defocusingof a collinear He–Ne laser probe beam (Model 117A,Spectra-Physics, Inc.), which was detected with aphotodiode (Model C5658, Hamamatsu Corp.) (Fig.1). Navigation of the pump beam on vessels was con-trolled with a Cascade 650 CCD camera (Roper Sci-entific, Inc.).

The capability of this new system was evaluatedusing a nude mouse ear model, which has a thin��270 �m� and relatively transparent structure witha well-developed vasculature [Fig. 2(a)]. After stan-dard anesthesia (ketamine/xylasine, 50/10 mg/kg),the animal was placed on the microscopic stage aftertopical application of glycerol. Glycerol providedslight adherence of the ear to the slide, acousticmatching between the transducer and the ear, andreduction of scattered light, mainly from the epider-mis. The last “optical clearing effect” significantly im-proved images of blood microvessels [Figs. 2(b) and2(c)], allowing us to distinguish single white blood

Fig. 1. Integrated PA/PT FC with near-IR laser.

2006 Optical Society of America

3624 OPTICS LETTERS / Vol. 31, No. 24 / December 15, 2006

cell (WBC) [Fig. 2(c), arrows] and red blood cell(RBC) shapes in capillaries [Fig. 2(d)]. The examinedblood vessels had diameters in the range of10–40 �m, blood velocities of 0.5–4 mm/s, anddepth locations of 30–100 �m. Laser beams had acircular geometry with diameters comparable withvessel diameters and could be adjusted in a linearconfiguration �6 �m in width. To detect rare cells,beam diameter was expanded to preclude missingcells due to low pulse rates.

PT signals from single cells in a linear mode2,3 (i.e.,without notable cell photodamage) demonstrated astandard initial peak associated with rapid (ns scale)cell heating and a slower, �m-scale tail correspond-ing to cell cooling [Fig. 3(a), top]. The �m-scale PAsignal from the same cell had an initial bipolar shapetransformed to a pulse train due to reflection effects[Fig. 3(b), top]. These signals were compressed atslow oscilloscope rates (ms scale) and appeared asvertical lines [Figs. 3(a) and 3(b), bottom]. PAFC wasused to detect circulating cancer cells (human squa-mous carcinoma cell line SQ20B) labeled withstrongly near-IR absorbing gold nanorods (GNs),which were 15 nm�52 nm in size and had a maxi-mum absorption of nearly 840 nm. The GNs weresynthesized with cetyltrimethylammonium bromide(CTAB). Then the GNs were centrifuged to eliminatethe extra free CTAB in solution and were then incu-bated with cancer cells for 15 min at 37 °C. To avoidthe influence of immunogenicity on cell circulation,this was done without antibodies by using just elec-trical or endocytotic effects. The real-time accumula-tion of GN in the cells was monitored through an in-crease in PA or PT signals.

PT/PA spectroscopy at different pump laser wave-lengths in vitro [Fig. 4(a)] and in vivo revealed thatthe absorption contrast, estimated as a ratio of themaximum amplitudes of linear PA signals fromsingle cancer cells labeled with GN in blood flow [Fig.3(d)], from the vessel with blood flow [Fig. 3(c)], andfrom the background tissue near the vessel was ap-proximately 29:2.1:1. The change in PT/PA signal am-

Fig. 2. (a) Mouse ear. Transmission image of ear (b) beforeand (c) after topical administration of an optical clearingagent, glycerol (magnification 10�). (d) High-resolution im-ages of individual RBCs in a capillary (indicated by dashedcurves; magnification 100�).

plitude from the same cell with GN after one laser

pulse at 10–30 mJ/cm2 fluence was negligible [Fig.4(a), dashed line]. We injected 100 �l suspensionwith 104 cancer cells labeled with GN into themouse’s circulatory system through the tail vein. Us-ing PAFC, we found that the half-life of circulatingthese cells was �15–20 min [Fig. 4(b)]. The capabil-ity of GN was compared with that of conventionalcontrast agents such as Indocyanine Green4 (ICG).We first injected 3.5 mg/kg of this dye alone into the

Fig. 3. (a) PT and (b) PA signals from the same RBCs in ablood capillary of the mouse ear, (top) noncompressed and(bottom) compressed. (c) PA signal from blood flow in vesselwith diameter �30 �m. (d) PA signal from single cancercell labeled with GNs in blood flow. Amplitude/time scale/(laser pump wavelength)/(laser fluence): (a, top)20 mV/div/4 �s/div/530 nm/0.3 J/cm2, (a, bottom)20 mV/div/20 ms/div/530 nm/0.3 J/cm2; (b, top)20 mV/div/4 �s/div/530 nm/0.3 J/cm2; (b, bottom)20 mV/div/4 �s/div/530 nm/0.3 J/cm2; (c) 20 mV/div/4 �s/840 nm/25 mJ/cm2; (d) 200 mV/div/4 �s/div/840 nm/18 mJ/cm2.

Fig. 4. (a) PT spectra of single cancer cells labeled withICG or GN and single RBCs in solution in vitro at laser flu-ences 0.1, 0.028, and 0.09 J/cm2, respectively. The dashedline with open triangles shows PT spectra of GNs after la-ser exposure (1 pulse, 20 mJ/cm2). (b) Normalized numberof circulating cancer cells labeled with GNs and ICG inblood flow as a function of time postinjection monitored at840 and 805 nm, respectively. (c) PA monitoring of ICG inthe blood flow at 805 nm (maximum ICG absorption in

flow).

December 15, 2006 / Vol. 31, No. 24 / OPTICS LETTERS 3625

mouse. Then, a similar procedure was performedwith cancer cells stained with ICG(25 �g/mL for 1 h at 37 °C) [Fig. 4(b)]. The typicalclearance time from the blood pool of ICG alone wasapproximately 10–20 min, which is in accordancewith other data.4 However, we observed the appear-ance of rare, strong PA signals (not previously de-scribed) above the continuous background signalsfrom ICG in the blood [Fig. 4(c)]. These signals lastedmore than 1 h after the background signals from ICGhad already disappeared. These fluctuations can beassociated with the ability of individual reticulocytes(and probably neutrophils) to significantly uptakeICG directly in blood flow5 and to circulate longerthan pure ICG. Thus, PAFC with its high spatial(6–20 �m, see above, and 0.5–1 �m in imagingmode2,3), temporal (10−4–10−1 s, depending on thepulse rate), and spectral (linewidth �0.5 nm) resolu-tion has great potential for studying the kinetics ofdyes, drugs, and nanoparticles, in circulation.

At the same laser energy, linear PA signals fromGN-labeled cancer cells were 5–7 times larger thanthe signals from cancer cells stained with ICG. Thisfinding indicates the advantage of using GN labels incases where the volume to be detected is small (i.e.,within one cell). PA signals from single cells labeledwith GN and from ICG alone in blood flow were com-parable. This means that, despite a significantlyhigher absorption coefficient of GN than that of ICG,this advantage in local absorption may be partly lostbecause of spatial integration of PA/PT signals fromlimited numbers of strongly absorbing GNs withinone cell. This is compared with the much larger num-ber of less-absorbing dye molecules homogeneouslydistributed in the same or larger (i.e. within the laserbeam) volumes. Cells stained with ICG were ob-served in circulation for a longer time �0.5–1.5 h�than cells labeled with GN [Fig. 4(b)]. This is prob-ably due to the influence of the incompletely purifiedtoxic component of CTAB, leading to faster clearanceof partially nonviable cells in apoptotic or even ne-crotic states. Despite the fact that GN is not toxic, thepreparation procedure requires further study. Also,one cannot exclude the possibility of partial GN de-tachment from labeled cells during their interactionwith other cells and molecules in flow. Nevertheless,these data demonstrate the potential to monitor theclearance rate of cells in different states (e.g., apop-totic or necrotic) at a different immune status.

The PT thermolens method demonstrated advan-tages over the PA method in the sensitivity of label-free detection of individual cells. The PA method pro-vided higher sensitivity in detecting nanoparticles ortheir nanoclusters alone or in cells [Fig. 3(d)]. PTthermolens signals from the single nanoparticleswere much weaker due to the influence of rapid heatdiffusion (responsible for thermolens effects). Accord-ing to our theoretical modeling, this effect is less im-portant to PA signal generation, which is associatedmore with the initial stage of thermal expansion (orevaporating) of a thin layer of liquid around GN (wecalled it the spherical piston model). To detect GN,

the pump–probe PT technique requires an extension

of the thermal field or bubble around the GN abovethe diffraction limit, while the PA method with a non-optical recording system is free of this optical limita-tion. Nevertheless, PT imaging provides higher reso-lution close to the diffraction limit,2,3 which isdifficult to achieve with PA imaging because of thetemporal limitation of transducers. On the otherhand, an advantage of the PA technique is its back-ward mode (i.e., its laser and transducer are on oneside), which is crucial for use on humans. PT and PAmethods may beneficially supplement each otherand, in combination, provide a powerful tool. For ex-ample, noninvasive PA diagnostics can be integratedwith PT killing of metastatic or residual cancer cells,either static or in flow, by use of more powerful laserpulses triggered by PA signals from these cells. Inparticular, at 840 nm, strong absorption of GN pro-vides complete killing of GN-labeled cancer cells��99% � with a very low photodamage threshold, cur-rently 58 mJ/cm2 with the potential to go to20 mJ/cm2, without harmful effects for surroundingRBCs, for which the photodamage threshold is muchhigher: around 22 J/cm2 at 840 nm (for leukocytes,219 J/cm2).

The described technique is capable of detecting 103

cancer cells in the blood pool of the mouse with an ap-pearance of single cells in peripheral circulation ev-ery few minutes, with the potential to improve thisparameter to 102 cells. The sensitivity provides detec-tion of one cancer cell in the background of 107−108

blood cells in small animals. Further optimizationmay provide the potential to detect one cancer cellamong 105−106 blood cells in peripheral human cir-culation.

Other applications of this tool in vivo may include(1) label-free detection of cells with differences innatural absorption (e.g., RBCs, WBCs, leukemia, andmelanoma), (2) study of therapeutic impacts (e.g., ra-diation, drug, or laser) on individual flowing cells,3

(3) aggregation assays because of the sensitivity tothe number of GNs and their orientation, and (4)time-resolved PT/PA angiography and lymphography.According to preliminary data, further improvementof PAFC may be realized using compact, inexpensivenear-IR pulsed laser diodes with a pulse energy (afew �J) and repetition rate �104 Hz� sufficient forlabel-free detection of every single pigmented squa-mous and melanoma cells in a peripheral flow.

This work was performed at the University of Ar-kansas for Medical Sciences with support from NIH/NIBIB grants EB000873/1858/5123 (to V. P. Zharov,zharovvladimirp@uams.edu).References

1. J. Novak, I. Georgakoudi, X. Wei, A. Prossin, and C. P.Lin, Opt. Lett. 29, 77 (2004).

2. V. Zharov, E. Galanzha, and V. Tuchin, in Proc. SPIE5320, 256 (2004).

3. V. P. Zharov, E. I. Galanzha, and V. V. Tuchin, J. Cell.Biochem. 97, 916 (2006).

4. X. Wang, G. Ku, M. A. Wegiel, D. J. Bornhop, G. Stoica,and L. V. Wang, Opt. Lett. 29, 730 (2004).

5. X. Wei, J. M. Runnels, and C. P. Lin, Invest.

Ophthalmol. Visual Sci. 44, 4489 (2003).