2.1.

Calibration

To study the relationship between intracellular temperature and PA image amplitude, a coverslip with adherent HeLa cells was uniformly heated by a custom-made heating pad, as shown in Fig. 1(a). The heating pad was made by connecting a thin metallic platform to a resistive heater (PN: HT10K, Thorlabs). A thermocouple was placed in contact with the coverslip and close to ($\sim 1\mathrm{mm}$) the cells. The environment temperature was gradually raised by increasing the voltage applied to the heating pad. At each temperature, after enough time had elapsed to ensure that thermal equilibrium had been reached among the cells, the thermocouple, and the environment, a PA image was acquired and the temperature was measured by the thermocouple. Since the thermal and stress relaxation times for a 2.5-μm particle are 270 ns and 1.7 ns, respectively, which are longer than the laser pulse duration ($\sim 1\mathrm{ns}$), both thermal and stress confinement conditions were satisfied. After fluence correction, the acquired PA image amplitude (peak-to-peak value of each A-line signal) was given by

Fig. 1

System layout of PA-based cellular temperature-sensing during (a) calibration and (b) photo-thermal heating. In (b), the cell was heated by a 100-mW CW laser via a multimode fiber. The intracellular temperature was monitored by a voice-coil PAM system in real time.

The measured PA signal versus temperature for a single cell is shown in Fig. 2. The PA signal of the cell was calculated by averaging all pixels within the cellular boundary. Since the internalized micro-particles were distributed in cytoplasm, the cellular boundary was estimated by applying a threshold to the acquired PA image. Four independent measurements [Fig. 2(b) to 2(e)] were carried out on the same cell pointed by an arrow in Fig. 2(a). The results indicate that the cellular temperature rise is accompanied by a linear increase in its PA signal. The slope $k{\mu }_{a}B$ and intercept $k{\mu }_{a}A$ thus can be solved by linear regression. The regressions from the four independent measurements are approximately the same, a result that proves the reproducibility of the experiment. Note that the heterogeneous distribution of particles causes a variation in PA amplitude within a cell [Fig. 2(a)]; however, since the cell was treated as a whole, the PA amplitude was averaged inside a cell. Since the cellular temperature has a one-to-one correspondence to its PA signal, the PA-recovered cellular temperature is calculated as

Fig. 2

PA amplitude versus temperature for a single cell. (a) Photoacoustic image of the cell at 33 °C. (b)-(e) Four independent measurements for the same cell pointed by an arrow in (a). The slope $c_1$ and intercept $c_2$ were solved by linear regression. The coefficients $c_1$ and $c_2$ were labeled with 95% confidence bounds. These measurements were made approximately 10 min apart in time.

In Eq. (5), the coefficients $k{\mu }_{a}A$ and $k{\mu }_{a}B$ are cell-specific, so it is critical that the calibration is done at the single-cell level. Here the four measurements were performed approximately 10 min apart in time; however, we expect that the calibration of these coefficients remains valid as long as the loaded particles are still within the cell. Additionally, since $k{\mu }_{a}A$ and $k{\mu }_{a}B$ are calculated by linear regression, more measurements would result in a higher coefficient of determination, $R^2$ (presuming that random, instead of systematic, error remains a major noise source), yielding a more accurate PA-recovered temperature. However, this effort would adversely increase the calibration time, complicating operational procedures.

2.2.

Single-Cell Temperature Sensing During Photo-Thermal Heating

Localized photo-thermal heating is widely used in cancer therapy.²³ Measurement of cellular temperature during photo-thermal heating is important to understanding the thermodynamics of cancer cells. The findings on cellular pathology and physiology during this process, in turn, can contribute to the development of new cancer diagnostics and therapeutics. To simulate the heating process, the output from a 100-mW CW laser (wavelength: 532 nm, PN: MLL-III-532, General Optoelectronic) was guided to the sample by a multimode fiber (PN: BFL37-600, Thorlabs) to heat up the HeLa cancer cells [see Fig. 1(b)].

The experiment was divided into three steps. First, we calibrated the relation between PA amplitude and temperature for a chosen single cell [Fig. 3(a)]. Then a control experiment without photo-thermal heating was performed to provide a baseline. The cellular temperature was monitored by the PAM at constant room temperature [see Fig. 3(b)]. The image acquisition speed was 3 s per frame. As shown in Fig. 3(b), with this imaging speed, the temperature resolution – defined as the standard deviation of the measured control temperature – was 0.2°C. Note that by using a slower imaging speed or a finer imaging step size, the temperature resolution can be further improved. Next, the heating CW laser was turned on and the cellular temperature was continuously measured by the PAM at the same acquisition speed. As seen in Fig. 3(c), the PA-recovered cellular temperature rose from 23°C to 25°C during the 300-s heating period. After the heating laser was turned off, the cellular temperature gradually decreased. To test the cells’s viability after these procedures, the cells were stained with green-fluorescence calcein-AM (cat #: L-3224, Life Technologies, Inc.) to indicate intracellular esterase activity and red-fluorescent ethidium homodimer-1 to indicate loss of plasma membrane integrity. The cells were re-imaged on a fluorescence microscope, and the results show that 84% of cells were still viable after photoacoustic imaging.

Fig. 3

Single-cell temperature-sensing during (b) control (c) photo-thermal heating. (a) shows the calibration relation between PA amplitude and temperature. The coefficient of determination, $R^2$, was 0.995. (c) through (e) show the cell images during photo-thermal heating at $\text{time}=0\mathrm{s}$, 320s, and 400s. The cell is pseudo-colored based on its PA-recovered temperatures.

During raster scanning, repetitive probe laser pulses may cause a local temperature rise. For spheres, the thermal diffusion length $d$ during laser-pulse interval time $\tau$ (0.2 ms) was estimated by $d=\sqrt{27\tau \alpha }$,²⁴ where $\alpha$ is the thermal diffusivity ($1.4\times {10}^{-3}{\mathrm{cm}}^2{\mathrm{s}}^{-1}$). If all laser-pulse energy is absorbed and converted to heat, the temperature rise by the previous laser pulse is approximated as

With $n$ previous laser pulses, the total temperature rise became

We also evaluated the dependence of the system’s pressure sensitivity on the coupling water’s temperature. First, since the ultrasound transducer was in the air and the room temperature was considered stable during the experiments, the transducer’s pressure sensitivity can be considered constant. Next, the coupling water’s temperature change may alter the speed of sound, acoustic focal distance and focal spot size, so the system’s pressure sensitivity can be affected. Here we estimated the acoustic focal distance and spot size at 25°C and 33°C with an assumption of a uniform temperature distribution in the coupling water. The focal distance was calculated from

The sensitivity to local temperature in PA imaging rises from the intrinsic photoacoustic generation mechanism. Consequently, the PAM-based method features the unique advantage of not requiring additional temperature-sensitive contrast agents to sense cellular temperature – the loaded iron oxide particles simultaneously play dual roles as a photo-thermal heating source and a local temperature sensor. This simplifies the experimental procedures and setup, as compared to the fluorescence-based approach.⁸

The single-cell photoacoustic thermometry requires that the cell stays in the same solution and cellular environment during calibration and temperature measurement experiments. For in-vivo studies, since the tissue surrounding the cell may also contribute to the PA signal,²⁵ the correspondence between PA amplitude and cellular temperature acquired by in-vitro calibration experiment may not be applicable. To reduce the effect of surrounding tissue on the in-vivo cellular temperature measurement, one can minimize the ratio of the tissue’s absorption to the sensing particle’s absorption by choosing an exciting wavelength.

4.1.

Cellular Sample Preparation

The HeLa cells were maintained in Dulbecco’s Modified Eagle Medium with 10% fetal bovine serum, 2 mM glutamine and 1% penicillin/streptomycin supplement. The cells were incubated at 37°C, in 5% ${\mathrm{CO}}_2$, and were divided every 72 h. After being dispersed in 0.25% EDTA-trypsin, they were seeded at $2\text{-}4\times {10}^4\text{cells}/{\mathrm{cm}}^2$. As a PA-imaging contrast agent, iron oxide micro-particles (PN: PI21353, Fisher Scientific) were added to the cell medium in a concentration of 0.5 pM and incubated for 24 h. The diameters of these particles range from 1 to 4 μm with a mean of 2.5 μm. Since the amount and rate of internalization were sufficient, the particles were not coated to increase uptake. Coverslips with adherent cells were washed with PBS before imaging. The optical microscopic imaging showed that the micro-particles were distributed in cytoplasm.

4.2.

Photoacoustic Microscopy

A recently developed voice-coil PAM system¹³ was employed to acquire the temperature-dependent cellular images. The voice-coil PAM system consists of a PA probe and a raster scanner. In the PA probe, short laser pulses (PN: 10-200-532, Elforlight Ltd.) were focused onto the sample surface through a set of optics. The generated photoacoustic signals were collected by an ultrasound lens (6 mm aperture, 0.5 NA in water) and received by a high-frequency ultrasound transducer (PN: V2022 BC, Olympus NDT). To achieve high sensitivity, the optical and acoustic foci were confocally and coaxially aligned. The photoacoustic probe was mounted onto a voice-coil-based scanner (PN: VCS-1010, Equipment Solutions) to implement fast imaging. In experiments, the sample was scanned at a 10-Hz cross-sectional imaging speed and a $1/3\text{-}\mathrm{Hz}$ volumetric imaging speed. The lateral resolution was measured as 3.4 μm. A higher spatial resolution is possible by increasing the optical numerical aperture.