2.1.

Computational Model

Most damage rate process models are based on the chemical kinetic rate constant most famously associated with the Swedish chemist, Svante Arrhenius. The Arrhenius equation [Eq. (1)] linearly relates the rate ($k$) of an isothermal reaction (production of product $P$) directly to the frequency factor (A; ${\mathrm{s}}^{-1}$) and exponentially to the energy of activation ($E_{\mathrm{a}}$; $\mathrm{J}{\mathrm{mol}}^{-1}$), temperature ($T$; K), and the universal gas constant ($R$; $8.31\mathrm{J}{\mathrm{mol}}^{-1}{\mathrm{K}}^{-1}$).

When “[P]” is generalized to damage [$D(t)$], then $\frac{dD(t)}{dt}=A{\mathrm{e}}^{-E_{\mathrm{a}}/RT(t)}$ describes the instantaneous damage rate (${\mathrm{s}}^{-1}$), and integrating produces $\mathrm{\Omega }$, the damage integral [Eq. (2)], which is a measure of accumulated damage. The value of $\mathrm{\Omega }$ can be determined at each time step of a thermal profile (temperature versus time). Detailed derivations of $\mathrm{\Omega }$ have been provided elsewhere^29–33 and are not presented here. There are other photothermal damage rate process models^27,28 that use a kinetic approach to illustrate how brief heating, via melanin absorption, may lead to a cascade of oxidative chemical reactions that eventually lead to death several hours postlaser exposure. In medicine, where discrete hyperthermic treatments are used in anticancer therapies,^34,35 converting time-at-temperature data to a normalized cumulative number of equivalent minutes at 43°C (CEM 43°C) has proven useful.^13,36,37 In general, this metric is most valuable for long duration hyperthermic treatments and is not assessed here.

The Arrhenius rate parameters $A$ and $E_{\mathrm{a}}$ ($A/E_{\mathrm{a}}$) must be determined empirically. Taking the natural logarithm of the Arrhenius equation [Eq. (1)], linearizes it for the condition of constant (absolute) temperature [Eq. (3)]. Measuring chemical rates at discrete temperatures permits the Arrhenius plot of $\ln k$ versus $T^{-1}$, which provides values for $A$ and $E_{\mathrm{a}}$ in the $y$ intercept and slope, respectively:

However, in photothermal damage studies, where the rate of generating a chemical product through thermal conversion cannot be measured in cells, and the reaction is seldom isothermal, a similar linearization is used on the damage integral [Eq. (2)]. When $\mathrm{\Omega }=1$, indicating a level of damage accumulation critical to cells, and integration is carried out over $\tau$, assuming constant peak temperature [$T(t)=T_{\mathrm{p}}$], the linearized form [Eq. (4)] can be rearranged to provide Eq. (5), and a modified Arrhenius plot of $\ln \tau$ versus $T_{\mathrm{p}}$ provides values for the Arrhenius rate parameters ($A/E_{\mathrm{a}}$):

Tabulated values of experimentally determined Arrhenius rate coefficients are available³¹ for a variety of thermally induced biological transformations, ranging from induction of heat shock proteins³⁸ to the immediate creation of intracellular bubbles.³⁹ The tabulated data demonstrate a wide range of values for the coefficients that can be explained, to a limited degree, by the various tissue types and the observational sensitivity for which an effect is determined. For example, retinal lesions are scored differently for damage when an ophthalmoscope is used, as opposed to evaluating histological sections at high magnification. The empirical $E_{\mathrm{a}}$ values listed³¹ vary by about one order of magnitude ($200-\mathrm{2,000}\mathrm{kJ}{\mathrm{mol}}^{-1}$), and the frequency factor displays a variance of 254 orders of magnitude. Even with the wide discrepancies in $A$ coefficients, within each empirical dataset ($A/E_{\mathrm{a}}$), there does exist a logarithmic correlation between the $A/E_{\mathrm{a}}$ values.^30–32 These relationships, shown in Eq. (6)³⁰ (using $\mathrm{kcal}{\mathrm{mol}}^{-1}$) and Eq. (7)⁴⁰ (using $\mathrm{J}{\mathrm{mol}}^{-1}$), are fit from data derived at the molecular (protein denaturation), cellular, and tissue levels, and include the tabulated data described above.³¹ Comparing empirically derived rate coefficients with a line demarked by Eq. (6) or Eq. (7) can provide a useful measure of validity for experimental approaches:

Alternatively, from $\frac{dD(t)}{dt}=A{\mathrm{e}}^{-E_{\mathrm{a}}/RT(t)}$, one can calculate the critical temperature ($T_{\mathrm{crit}}$) for an experimental system, which mathematically represents the temperature at which $dD(t)/dt=1$ for the first time during an exposure:

As can be shown with Eq. (5), the $T_{\mathrm{crit}}$ equates mathematically to the threshold temperature for a 1-s exposure.³¹ The critical temperature value should reflect the type of biological transformation measured and the level of sensitivity provided by the methods chosen. Equation (8) defines $T_{\mathrm{crit}}$ and is calculated by taking the ratio of the slope to the $y$-intercept for data plotted per Eq. (5). Alternatively, the critical temperature can be calculated from any set of $A/E_{\mathrm{a}}$ pairs reported in the literature. For example, the $T_{\mathrm{crit}}$ for the photothermal induction of heat shock proteins in cultured cells³⁸ is low (48°C) relative to photocoagulation in the retina^41,42 (57.6°C–59.7°C) and whitening of egg albumin using water bath heating⁴³ (76°C).

2.2.

Biological Model

Successful in vitro models retain phenotypic characteristics and respond to external stimuli in a manner common to cells in the source organ. Cells of the retinal pigment epithelial (RPE) layer contain melanosome particles (MPs), which are highly absorptive to visible light and are the reason the RPE layer is the principal target for photothermal damage in the retina. Unfortunately, RPE cells taken from adult animals have lost their ability to produce endogenous MPs, and in extended culture, the MPs are diluted with each cell division. Our studies with the hTERT-RPE1 cell line show that they retain their phagocytic functions and can be artificially pigmented.^8–10 Additionally, this artificially pigmented RPE model⁸ exhibits damage trends similar to those in nonhuman primates.^44,45

Well-crafted comparative experiments provide trends in damage sensitivity. Unfortunately, differences in optical and thermal properties between the cell culture model and intact retina make comparing damage thresholds based on laser dose difficult. After photon absorption, and the subsequent temperature rise, the processes of cellular damage likely ensue the same as heat transfer induced without photons, unless there is a subsequent chemical reaction(s) based on the concentration of melanin pigment.^27,28 Therefore, a method that follows the damage rate processes relative to temperature rise, rather than laser dose, avoids complications from varying absorption and diffusivity in samples, regardless of wavelength. Without the inconsistencies of absorptivity and diffusivity of samples, temperature-based damage thresholds will be more comparable across inhomogeneous samples. It is interesting to consider if, based on a fundamental photophysical response rather than laser dosimetry, a thermal damage threshold could be similar in all cell types.

Regardless of the subtleties of cell type, there is the issue of how to assign the thermal threshold. Following typical convention, identifying the temperature rise at the center of a laser exposure site leading to a minimum lesion defines the threshold temperature, regardless of laser irradiance. However, if laser irradiance is too great, the lesion size becomes larger than the minimum, and the central temperature is greater than the threshold. To determine temperature thresholds, one must use large sampling numbers to attempt to get the perfect minimum lesion scenario. This method leads to the discarding of many data points when exposures lead to either no damage or lesions that are too large.

One solution is to use thermal methods that provide data with spatial resolution, which correlate to damage, which is likewise spatially resolved. There are few reports of empirical laser damage studies that combine the two methods. Welch and Polhamus⁴¹ used carefully placed microthermocouples in animal eyes, and Simanovskii et al.⁴⁶ spatially modeled the thermal distribution about a centrally measured temperature rise for comparison with fluorescence damage images. Our group has directly measured spatially resolved thermal maps using infrared cameras at high magnification (c.a. $8\times 8\mu \mathrm{m}$ effective pixels) and high speed (800 fps) for correlation with fluorescence damage images,^47–49 which has been termed microthermography. In essence, these methods provide an estimate of temperature history at the boundary between cells surviving the laser exposure and those that go on to die.

The prime advantage of determining threshold temperature at the boundary of cell death is that the method does not depend upon laser dose. Central temperatures do rise with increasing laser irradiance, whereas the temperature at the boundary of cell death should remain relatively stable because spatially, it represents the biological threshold regardless of the size of damaged tissue (Fig. 1).⁴⁷ Also, one can achieve a much greater experimental efficiency using damage boundary data because all exposures resulting in damage, smaller than the laser footprint, are part of the dataset, not just those that produce a minimal lesion.

Fig. 1

Definition of threshold temperature at the boundary of cell death in response to photothermal damage mechanisms. Artificial data are shown to illustrate the relationship between temperature at the center of a laser irradiated region (demarked as an X) and the temperature at the boundary of cell death (arrow). (a) Near-threshold irradiance yields a small area of damage 1-h postexposure and the temperature at the boundary ($T_{\mathrm{B}}$) is only slightly lower than that produced at the center of the exposure site ($T_{\mathrm{X}}$). (b), (c) As laser irradiance is increased above threshold ($E_1<E_2<E_3$), $T_{\mathrm{X}}$ rises due to increased absorption. However, $T_{\mathrm{B}}$ remains constant in all exposures because it corresponds to the true biological threshold temperature. Reproduced with permission from Ref. 47.

Temperature history collected at the boundary of cell death is well suited to study the kinetics of cell death. Our hypothesis is that the boundary regions represent the same damage rate processes, not only within the same damage event (perimeter of a damaged region) but also for all boundaries of death generated for a given sample and environment. Being identical in damage rate processes, we can, therefore, assign the same damage accumulation, such as $\mathrm{\Omega }=1$. This simple point provides the basis for comparing how changing laser parameters, ambient temperature, and the optical and thermal properties of the tissue effect damage rate processes, if at all.

3.1.

Cell Culture and Laser Exposure

3.1.1.

Pigmented cells (532-nm exposures)

The in vitro retinal model employed in this study was developed previously.⁸ RPE cells immortalized with hTERT (hTERT RPE-1, ATCC, CRL-4000) were seeded into 24-well glass-bottom (No. 0) tissue culture plates (MatTek, P24G-0-13-F) at $7.0\times {10}^4\text{cells}/\text{well}$ in DMEM/F-12 media (Cellgro, 15-090-CM), excluding well A1 for monitoring temperature during exposures. After about a day, each well was observed for optimal cell health and confluency (50%–60%). For cultures meeting optimal conditions, MPs were added based on expected cell density at the time of laser exposure (two population doublings postseed would yield $2.8\times {10}^5\text{cells}/\text{well}$). Stock solutions of extracellular MPs from a bovine source were obtained by the isolation method of Dontsov et al.⁵⁰ To achieve target pigmentations greater than $200\mathrm{MPs}/\text{cell}$, multiple additions of $200\text{–}250\mathrm{MPs}/\text{cell}$ aliquots (in 0.5 mL complete medium) were added to wells 1.5–2 h apart. Upon completion of MP additions, fresh medium was added to wells to a final volume of 2.0 mL. Cells were then incubated at 37°C, 5% ${\mathrm{CO}}_2$ atmosphere overnight to allow for phagocytosis of MPs. Adhering to this schedule provided monolayers with consistent cell density and good overall viability. In preparation for laser exposures on the second day postseed, wells with optimal MP uptake and overall health (manual inspection of each well) were carefully washed twice with 0.5 mL Hank’s balanced salt solution (HBSS) with 10 mM HEPES at pH 7.4 and without sodium bicarbonate (Corning 20-023-CV) (collectively called the exposure buffer) at room temperature. A final rinse with prewarmed exposure buffer in a glove box incubator (37°C) was performed near the exposure box. Wells then received 0.5-mL prewarmed exposure buffer and equilibrated in the exposure chamber ($35°\mathrm{C}\pm 1°\mathrm{C}$) for at least 10 min before laser exposures.

Laser delivery for exposures of cells with varying pigmentation (100, 200, 400, $800\mathrm{MPs}/\text{cell}$) is depicted in Fig. 2. As described previously,^49,51 two lenses (L5 and L6) and a fiber coupler were used to launch the 532-nm beam into a $100\text{-}\mu \mathrm{m}$ core multimode fiber (NA 0.22) and the output of the fiber tip was relay-imaged ($1.25\times$ magnification) to the sample plane as a nominal flat top $480\pm 5\text{-}\mu \mathrm{m}$ diameter image using L3 and L1. Beam diameter was verified using a CCD camera at the sample plane and Spiricon software. In addition to the delivery of the laser, the microscope served to image (L4 and L1) and record video (60 fps) of cell samples, and image (L2 and L1) and record microthermography. A Plexiglas enclosure chamber^49,51–53 provided consistent ambient temperature and 60%–75% relative humidity. As measured by microthermography, the average ambient temperature for all 91 individual 1-s and 3-s exposures was $34.3°\mathrm{C}\pm 0.3°\mathrm{C}$. Immediately prior to the automated process at each well, cells at the sample plane were brought into focus with a z-micrometer using the video camera. Imaging cells through glass bottom cell culture dishes eliminated steps for complex liquid handling prior to microthermography that were needed previously.⁴⁷ Once exposures were completed, wells in the plates were assessed for cytotoxicity as described below.

Fig. 2

Laser delivery to pigmented RPE cells. Output of a 10-W Millennia Pro laser (Spectra Physics, Mountain View, California) was launched into a $100\text{-}\mu \mathrm{m}$ core multimode fiber. BS3 provided real-time monitoring of laser power. Output from fiber was imaged through the microscope to the wells inside the warmed enclosure. Thermal (FLIR) and video (CCD) imaging utilized beam splitters (BS) 1 and 2. M, mirror; L, lens; W, optical window, F, laser rejection filter (OG570).

3.1.2.

Cells without pigment ($2\text{-}\mu \mathrm{m}$ exposures)

We used the same hTERT-RPE1 cell line for laser exposures at $2\mu \mathrm{m}$ without the addition of MPs. Cells (100,000) were seeded (1.0 mL total) into sterile AttoFluor chambers (Molecular Probes) containing 25-mm diameter No. 0 borosilicate coverslip bottoms (Deutsch Deckglaser, Cat. GC-25-0-oz). Up to six AttoFluor chambers were prepared together for laser exposures. On the following day, after cells were carefully washed twice with 1.0-mL exposure buffer, 1.0-mL exposure buffer was added and the chambers were transferred to a 37°C culture incubator (no ${\mathrm{CO}}_2$) adjacent to the exposure bench until each was exposed separately. A custom LabVIEW program automated all imaging and laser delivery processes. Immediately after receiving three replicate exposures (1 mm apart) of a given laser energy at a given ambient temperature, the chambers were placed back into the incubator. Once all six (or fewer) chambers received replicate exposures, the chambers were assessed for cytotoxicity 1 h later, as described below. Using this method, the cells were only at the given ambient temperature for the short time needed to achieve the desired temperature and carry out the laser exposures.

A Nikon TE300 microscope stage plate (Linkam part L-PE100/NI, distributed by McCrone Microscopes, Westmont, Illinois) with thermoelectric heating/cooling ($10°\mathrm{C}/\min$) was used to obtain various ambient temperatures (20°C, 25°C, 30°C, and 40°C) of cells in AttoFluor chambers prior to laser exposure. Preliminarily, cells held at each of the ambient temperatures for an excessive time (15 min) relative to that required to perform exposures (2–5 min) showed no loss in viability as determined by the method used after laser exposure (data not shown). To provide feedback for thermal camera offset (Sec. 3.3), temperatures of samples were estimated with a microthermistor (EPCOS, Cat. B57540G1103F000) held off center and near the surface of the cells via a custom Teflon lid. A heated Plexiglas enclosure similar to that used for the pigmented cell exposures was used to reduce the time needed to achieve the ambient temperatures of 30°C and 40°C.

Both laser delivery and microthermography were carried out from beneath the cells in a similar manner to that described,⁵³ where the final imaging optic prior to cell samples was a 75-mm ${\mathrm{CaF}}_2$ lens (CF-PX-38-75, ISP Optics, Corp., Irvington, New York). Figure 3 provides a schematic of the experimental set up for the variable ambient temperature experiment. The output of a thulium fiber laser (IPG Photonics, TLR-20-2000-LP) provided $2\text{-}\mu \mathrm{m}$ irradiation focused to an approximately Gaussian beam with a diameter of $620\pm 7\mu \mathrm{m}$. The internal laser shutter was computer controlled to deliver exposure durations of 0.05, 0.25, 1.0, and 20 s.

Fig. 3

Laser delivery to RPE cells without pigment. Output of an IPG Photonics TLR-20-2000-LP fiber laser was delivered to RPE cells (sample) from below. Cells were imaged from below using bright field (PULNiX TM-6701 AN Camera) and microthermography (FLIR thermal camera). Mirror 1 (M1) and lens 1 (L1) were common to both legs of the microscope. Exposures were carried out in a Plexiglas enclosure, which was only closed with heat control for exposures at $T_{\mathrm{amb}}$ of 30°C and 40°C.

3.2.

Damage Assessment

Cell death was determined by postlaser exposure in the same manner for all sample types. After exposures, the exposure buffer was replaced with complete medium and cells were incubated at 37°C and 5% ${\mathrm{CO}}_2$ for at least 1 h to allow recovery or damage progression. After recovery, cells were stained for 15 min at 37°C for viability using $1.7\text{-}\mu \mathrm{M}$ calcein-AM (C3100MP, Life Technologies, Grand Island, New York) and $1.4\text{-}\mu \mathrm{M}$ ethidium homodimer 1 (EthD1, E3599, Life Technologies) in complete HBSS. Laser-damaged sites were identified by $10\times$ imaging on an Olympus CK-40 using EthD1 channel (bandpass exciter of 475–545 nm and a barrier filter at 590 nm) and Calcein-AM channel (bandpass exciter of 460–490 nm and a bandpass emitter of 490–530 nm). Monochrome images were taken using a Hamamatsu ORCA 100 camera.

Damage threshold irradiance values (${\mathrm{ED}}_{50}$) were determined using the Probit^54,55 method, as previously described.^8–10 The Probit output included 95% confidence intervals [fiducial limits (FLs)] related to the ${\mathrm{ED}}_{50}$ value. Probit slope values were determined by taking the first derivative of the Probit probability curve at the 50% value.

3.3.

Microthermography

A ThermoVision SC4000 (532-nm exposures) or SC6000 ($2\text{-}\mu \mathrm{m}$ exposures) mid-wave infrared camera (FLIR Systems, Inc. Boston, Massachusetts) imaged cells at 800 fps during laser exposure. Effective pixel pitch values were 8.61 and $7.41\mu \mathrm{m}/\text{pixel}$ for the SC4000 and SC6000, respectively. Knowing that our thermal imaging system involves transmission through thin borosilicate coverslips (no. 0, 0.08–0.12 mm), we characterized the transmission of a clean no. 1 coverslip (average thickness of 0.15 mm) by placing it in front of a blackbody reference source (M316, Mikron Instruments). We found 86% transmittance at the wavelength range specific to the InSb camera detector ($3-5\mu \mathrm{m}$). During each calibration of the thermal camera, we placed a piece of MatTek well-bottom No. 0 glass (a window) in front of a portable blackbody source (M316, Mikron Instruments). This then accounted for the true transmission of heat by the cells through the bottom of the wells during experiments. After the final optical configuration was established, the nonuniformity correction and the calibration of both the camera and overall thermal imaging system were performed, as suggested by FLIR systems. Specifically, we used reference plates provided by FLIR at two temperatures, different by at least 10°C, and recorded short movies for comparisons of defined RoIs. The calibration was verified by applying the calibration in the ExaminIR software and imaging the black body with the window in place. At several points across the calibration range, the temperature of the black body was compared to the temperature measured by the thermal camera. Our calibration was correct to within 1°C.

Due to the relatively high magnification and speed of our microthermography imaging, coupled with low spectral irradiance for the range of the thermal camera ($3-5\mu \mathrm{m}$) at around room temperature, there was a minor drift in the camera’s offset over time. We devised methods to adjust the thermal camera’s digital offset to a known temperature immediately prior to each laser exposure at the two experimental systems. For the pigmented cell exposures, a small cylindrical piece of aluminum was fit snugly into well A1, which effectively placed a coverslip between the camera (SC4000) and the aluminum. The painted (flat black) aluminum was fitted with a resistance temperature detector (RTD) directly into the top. When the $x$, $y$ translational stage was in the “home” position (well A1 centered over the microscope) between laser exposures, we could adjust the thermal camera offset to the temperature verified by the RTD. For the $2\text{-}\mu \mathrm{m}$ exposures, the real-time temperature feedback from a microthermistor placed just above the cells was used to adjust the thermal camera (SC6000) offset, as also described in Sec. 3.1.2.

3.4.

Thermal Data Analysis

Thermal data were processed as previously described,⁴⁷ with the primary goal of identifying temperatures correlated spatially to the boundary of cell death, which identifies those cells receiving the lowest dose to produce cell death ($\mathrm{\Omega }=1$). Thermal images were used to extract and calculate thermal data from full-frames ($192\times 192\mathrm{pixels}$) and the boundary of cell death regions of interest (RoIs). After the overlay of thermal and damage images, with proper orientation and stretching protocols, boundary RoI mask files were output. Boundary RoI masks were then used to extract temperatures for pixels corresponding to the boundary of cell death from original FLIR files. Mean thermal profiles for individual exposures were averaged with others within a data ($\tau$) set, and the final mean thermal profiles were reported with both standard deviation (SD) and standard error of the mean (SEM).

4.1.

Effects of Ambient Temperature on Damage

4.1.1.

Thermal profiles at the boundary of cell death

Figure 4 depicts steps during the method of extracting thermal data from the boundary of cell death for two representative $2\text{-}\mu \mathrm{m}$ laser exposures of hTERT-RPE1 cells without pigment. The fluorescence damage images (panels 4b and 4h) show clear demarcation for laser damage and high cell viability outside the damaged regions. The orientation of the damage maps (panels 4c and 4i) relative to the damage images (panels 4b and 4h) reflect multiple flip/rotate functions required for correct overlay (shown) with thermal images (panels 4d and 4j). The resulting overlay of thermal and damage images provided thermal single-pixel RoI maps (panels 4e and 4k) corresponding to the perimeter of the damage map. Mean thermal profiles were derived from averaging the RoI pixels from the original thermal movie of the laser exposure (panels 4fand 4l). Thermal profiles are shown at the boundary (lower-temperature rise profiles with SDs), which can be compared to temperatures at the center of the exposed region (upper-temperature rise profiles).

Fig. 4

Determining threshold thermal profiles at the boundary of cell death. Shown are images representing various stages of extracting thermal data at the boundary of cell death for exposures to $2\text{-}\mu \mathrm{m}$ laser irradiation for (a)–(f) 50 ms and (g)–(l) 250 ms. (a), (g) Phase contrast images postlaser exposure. (b), (h) Overlaid fluorescence images (calcein and ethidium homodimer) identifying damage. (c), (i) Damage masks generated from calcein images. (d), (j) Thermal maps of corresponding thermograms at the end of laser exposure. (e), (k) Thermal image mask of pixels at the boundary of cell death. (f), (l) Thermal profiles for the 50- and 250-ms laser exposures. The upper profile in each represents the central temperature of (d) and (j) the example exposure; the lower profiles used the $\text{mean}\pm \mathrm{SEM}$ threshold temperature values for all similar exposure durations.

As an example of constant sample thermal and optical properties, the nonpigmented RPE cells were exposed at $2\text{-}\mu \mathrm{m}$ for 0.05, 0.25, 1.0, and 20 s. Each set of the four exposure durations was performed at four different ambient temperatures (20°C, 25°C, 30°C, and 40°C). The large number of equivalent pixels (1100-3–207) associated with the 12–20 replicates within each of the 16 types of exposures validated the use of SEM values (Table 1). However, even though the small SEM values indicated the threshold $T_{\mathrm{p}}$ values were statistically significant, they were smaller than the accuracy of the FLIR camera (1°C), and we will refer to SD values because they realistically convey experimental uncertainties.

Table 1

Exposure parameters and thermal data for nonpigmented and pigmented cells exposed to laser irradiation.

Exposure Duration (s)	Wavelength (nm)	MPs/cell	Exposures (n)	Ambient		Boundary of cell death RoI
				Temperature	Temperature	Pixels (n)	Peak temperature
				(°C)	SD		(°C)	SD	SEM
1.0	532	100	8	34.7	0.4	661	52.4	0.8	0.3
	532	200	13	34.2	0.4	794	51.9	0.7	0.2
	532	400	16	34.7	0.3	1223	52.2	0.8	0.2
	532	800	14	34.0	0.2	1369	52.5	0.9	0.2
3.0	532	100	12	34.2	0.3	894	50.4	0.8	0.2
	532	200	9	34.0	0.2	640	50.2	0.7	0.2
	532	400	11	34.0	0.2	1050	50.2	0.4	0.1
	532	800	8	34.0	0.1	989	50.5	0.7	0.3
0.05	2000	0	18	20.1	0.2	2017	62.2	5.2	1.2
	2000	0	16	24.7	0.2	3207	62.5	2.0	0.5
	2000	0	20	29.6	0.3	2049	60.1	1.3	0.3
	2000	0	16	39.9	0.2	2003	62.3	1.3	0.3
0.25	2000	0	15	20.0	0.1	1733	59.0	1.2	0.3
	2000	0	12	24.7	0.2	1438	59.6	1.5	0.4
	2000	0	13	29.6	0.3	1511	59.5	2.9	0.8
	2000	0	12	40.1	0.3	1486	61.2	1.9	0.6
1.0	2000	0	15	20.4	0.3	2105	55.5	1.4	0.4
	2000	0	15	25.0	0.5	1617	56.5	1.1	0.4
	2000	0	16	29.7	0.5	1408	54.9	1.8	0.6
	2000	0	12	39.9	0.2	1100	56.6	1.4	0.6
20.0	2000	0	13	20.1	0.2	2907	48.3	2.6	0.8
	2000	0	15	24.7	0.1	1944	49.2	1.5	0.5
	2000	0	15	29.4	0.2	1841	50.2	1.2	0.4
	2000	0	12	39.9	0.2	1390	56.1	1.5	0.5

The $T_{\mathrm{amb}}\pm \mathrm{SD}$ values in Table 1 provide the support that we achieved the target ambient temperature prior to laser exposure with precision. Again, it took only 2–5 min to reach these ambient temperatures and another 20–90 s to perform the exposures. Beyond this limited time at nonoptimal temperatures, cells either recovered or underwent damage progression at the optimal temperature (37°C). Laser energies were chosen to produce damage, with slight variations leading to damage regions of various sizes. Damaged regions that were nearly the area of the laser, and larger, were avoided in our analysis because they do not accurately represent energy deposition by the laser.

Mean temperature rise ($\mathrm{\Delta }T$) profiles, constructed by averaging temperature among replicates at each time point of the thermal movie, are given in Fig. 5. To provide separation between the shorter thermal profiles, Figs. 5(a)–5(d) truncate the 20-s data at 2 s. The diffusivity in our samples was sufficient for the thermal steady state for a significant portion of the 1-s exposure. As expected, the 20-s exposures appeared to reach thermal steady state within 1-2 s, but as shown in Fig. 5(e), temperatures drifted higher over the majority of the 20-s thermal profiles. Although the 20-s exposures were expected to represent isothermal reactions, a drift in laser power starting at about 2.5 s caused the moderate increase of about 10% of each peak $\mathrm{\Delta }T(\mathrm{\Delta }{T}_{\mathrm{p}})$ during the final 17.5 s [Fig. 5(e)]. In constant power mode, the laser output had a large power spike as the internal shutter opened, necessitating the use of the constant current mode. Unfortunately, being in the constant current mode caused the power drift seen most noticeably in the longer exposures.

Fig. 5

Threshold thermal profiles and $A/E_{\mathrm{a}}$ determination for nonpigmented RPE cells exposed at $2\mu \mathrm{m}$. Initial 2 s of thermal responses to $2\text{-}\mu \mathrm{m}$ laser exposure at (a) 20°C, (b) 25°C, (c) 30°C, and (d) 40°C. Panel (e) provides full thermal profiles for the duration of the 20-s exposures at each of the four ambient temperatures. All thermal profiles shown in panels (a)–(e) are the mean values at each time point taken from the boundaries of cell death across a wide range of laser irradiances and damage sizes. Panel (f) provides the determination of Arrhenius parameters $A/E_{\mathrm{a}}$ using average threshold $T_{\mathrm{p}}$ values (abscissa) at the four laser exposure durations (ordinate). The values of $A$ and $E_{\mathrm{a}}$ are taken from the $y$-intercept and slope of the resulting line, respectively.

Table 1 shows that, except for the 20-s exposure at 40°C, within each $\tau$ group, there were no statistically significant differences in threshold peak temperature values. This was less clear from Fig. 5 because one must add $T_{\mathrm{amb}}$ to the $\mathrm{\Delta }T$ values to calculate threshold peak temperature values. Combined, the results from Table 1 and Fig. 5 show that threshold peak temperature was a primary determining factor for damage within each exposure duration. The effect of varying ambient temperature was that it determined the $\mathrm{\Delta }T$ required for achieving the damage threshold peak temperature value. This result appeared to be a form of reciprocity between threshold $\mathrm{\Delta }T$ at $\tau$ ($\mathrm{\Delta }{T}_{\mathrm{p}}$) and the ambient temperature. When plotted as $\mathrm{\Delta }{T}_{\mathrm{p}}$ versus $T_{\mathrm{amb}}$, slopes of $-1.01$, $-0.89$, $-0.96$, and $-0.60$ were obtained for $\tau$ values of 0.05, 0.25, 1.0, and 20 s, respectively (data not shown). Except for the 20-s data, which were skewed by the inexplicably high threshold peak temperature at 40°C ambient, the current results follow a trend expected for reciprocity (i.e., a slope near 1).

This result corroborates a previous assessment of the effect of pre-exposure fundus temperature ($T_{\mathrm{amb}}$) on threshold $\mathrm{\Delta }T$ in pigmented rabbit retinas exposed at 488-nm for 10 s.⁵⁶ Here, the authors used microthermocouples⁴¹ to determine threshold $\mathrm{\Delta }{T}_{\mathrm{p}}$ in the intact globe, plotted the thresholds (along with some extrapolated thresholds) versus their corresponding pre-exposure fundus temperatures (30°C–44°C), and found a slope of $-1.15$. This slope indicated reciprocity between threshold $\mathrm{\Delta }T$ and pre-exposure fundus temperature. The way these animal studies were set up, the ambient temperature was established for much longer than in our in vitro experiment. Also, of interest, the extrapolated line for the rabbit study crossed the abscissa at 52.4°C, to which the authors attribute the threshold temperature for damaging the fundus without laser exposure. Notice that this 52.4°C in vivo threshold value for 10 s falls between the in vitro peak temperature thresholds (boundary) for 1 and 20 s in Table 1.

An important implication of our results is that short intervals of nonoptimum pre-exposure temperature did little to affect the required threshold peak temperature to cause photothermal damage in the cultured cells. The 40°C (104°F) ambient temperature condition did not appear to contribute to the damage rate process, as dictated by peak temperature. Likewise, a short time at the ambient temperature of 20°C (68°F) did not increase the peak temperature requirement for damage at the boundary of cell death. Perhaps, for a more sensitive damage rate process (e.g., apoptosis), the 40°C or 20°C ambient temperature would influence the required threshold peak temperature.

The current study also points to scenarios in which laboratories, without the luxury of owning expensive thermal cameras, or the means of controlling environmental parameters, can perform photothermal damage experiments on cultured cells. Point detection of threshold peak temperature rise in the center of exposed sites can provide thermal data for minimum lesions or as a starting point for computationally building spatial temperature maps that can be overlaid later with damage images.⁴⁶ Decoupling laser dose from damage by measuring temperature and minimizing the time that cultures spend at room temperature are primary considerations for improving the quality of the data. A recent article⁵⁷ reports the feasibility of using real-time optoacoustic temperature measurements combined with a nonfluorescent spatially resolved assay for functional mitochondria after 10-s exposure to a thulium laser at $1.94\mu \mathrm{m}$. Although refinements to the method could aid in both temperature measurement and special resolution of damage assessment, the approach shows potential for assessing photothermal damage processes without the use of a thermal camera.

4.2.

Effects of Pigmentation on Threshold Temperature

Taking advantage of our ability to alter the amount of intracellular pigmentation in the hTERT-RPE1 cells, we investigated whether MPs undergo a chemical reaction during photothermal heating that effects the damage rate process. Although our method could not distinguish whether a proposed photochemical byproduct of photothermal exposure^27,28 was the result of photon absorption directly (photochemical) or the temperature rise due to photon absorption (photothermal), it was designed to determine if any alteration in threshold damage rate does occur.

Table 1 provides $T_{\mathrm{amb}}$ and threshold $T_{\mathrm{p}}$ values for 1- and 3-s exposures of RPE cells containing up to an eightfold difference in the number of melanosomes offered cells in culture the day prior to exposure. Note the well-controlled ambient temperature across all 91 exposures ($34.3°\mathrm{C}\pm 0.3°\mathrm{C}$) at 532 nm. Peak temperatures at the boundary of cell death were essentially identical within the two exposure duration groups. The thermal profiles within each exposure duration were strikingly similar, and Fig. 6 provides a plot with each average thermal profile for pigmentation of 100, 200, 400, and $800\mathrm{MPs}/\text{cell}$. Although the threshold peak temperature values were the same (Table 1), Fig. 6 clearly shows a change in the shape of the two thermal profiles resulting from exposure to cells having $800\mathrm{MPs}/\text{cell}$. At the lower pigmentation levels, there was no pigment-dependent effect. In fact, due to the overall similarities in thermal profiles, the data actually indicate a high level of precision for extracting thermal data from the many boundary RoIs.

Fig. 6

Threshold thermal profiles for pigmented RPE cells exposed at 532 nm. Lines represent mean temperature history for thermal camera pixels at the boundary of cell death for 1- and 3-s exposures with each level of pigmentation. Pigmentation levels correspond to normal cellular uptake of MPs when provided at 100, 200, 400, and $800\mathrm{MPs}/\mathrm{cell}$ on the previous day. Dashed lines correspond to data for $800\mathrm{MPs}/\mathrm{cell}$. Data incorporate laser irradiances of $60\text{–}260\mathrm{W}{\mathrm{cm}}^{-2}$ (1 s) and $45\text{–}300\mathrm{W}{\mathrm{cm}}^{-2}$ (3 s) to show irradiance independence at the boundaries of cell death.

Clearly, the various levels of pigmentation dictated different laser irradiances to generate damage. Additionally, for the threshold $T_{\mathrm{p}}$ characterizations, only those exposures resulting in damage were included. In general, laser irradiances of $60-260\mathrm{W}{\mathrm{cm}}^{-2}$ (1 s) and $45-300\mathrm{W}{\mathrm{cm}}^{-2}$ (3 s) generated damage for boundary analysis across the range of pigmentation. By expanding the range of laser irradiances lower, negative damage results were recorded and used in the Probit software to estimate threshold ${\mathrm{ED}}_{50}$ values. Table 5 provides Probit data for the exposure durations and pigmentation levels shown in Table 1 and Fig. 6, plus a 50 and $300\mathrm{MPs}/\text{cell}$ scenario for 3-s exposures.

Table 5

Threshold ED50 irradiance values for pigmented cells exposed at 532 nm. Probit54,55 analysis was performed on binary yes/no damage data. Upper FL (UFL) and lower FL (LFL) provide 95% confidence intervals, and slope of the Probit curve at 50% probability (Probit slope). Note that none of the LFL and UFL overlap with those of neighboring thresholds.

As expected, the higher pigmentation, which effectively increases the absorption coefficient (${\mu }_a$), required less laser irradiance to cause damage, as evidenced by the ${\mathrm{ED}}_{50}$ values in Table 5. The 95% FLs were within 10% of their respective ${\mathrm{ED}}_{50}$ values, except for the 400 and $800\mathrm{MPs}/\text{cell}$ data for 1-s exposures (11% and 24%, respectively). The large values of Probit slopes (Table 5), representing the first derivative at a 50% probability, indicate robust data.

The in vitro ${\mathrm{ED}}_{50}$ values did not track linearly with absorption coefficient (${\mu }_{\mathrm{a}}$), represented by the amount of intracellular pigmentation. As dictated by the heat-diffusion equation, a laser exposure much reach thermal steady-state in order for laser irradiance to correlate linearly with temperature rise. Referring to Fig. 6, the 3-s exposure was approaching steady state, so the data were plotted in Fig. 7. From the graph, both the 1- and 3-s data had power curve with negative exponents of less than 0.7, compared to the idealized case (${\mu }_{\mathrm{a}}^{-1}$). The figure indicates a complex dependence upon both exposure duration and pigment. These trends were derived from samples having the same sample boundary conditions (bottom to top; air, glass, cells, the same volume of buffer, air), so they are comparable. The in vitro sample boundaries differ from the tissue boundaries in the eye, and we would not expect the same ${\mathrm{ED}}_{50}$ values, but the trends shown here could translate to retinal or pigmented skin ${\mathrm{ED}}_{50}$ studies. Clearly, an advantage lies in characterizing damage rate mechanisms using threshold peak temperatures, where the key determinant is laser exposure duration.

Fig. 7

Threshold ${\mathrm{ED}}_{50}$ radiant exposure values ($\mathrm{J}{\mathrm{cm}}^{-2}$) versus pigmentation for cells exposed at 532 nm. Data for 1-s (dashed line) and 3-s (dotted line) exposures are compared to the idealized condition (solid line) in which threshold laser damage is reciprocally related to cell pigmentation ($x^{-1}$).

Although we had only two threshold temperature values (1 and 3 s) to plot as Eq. (5) (data not shown), we obtained the $A/E_{\mathrm{a}}$ values reported in Table 2. These values fit the ambient temperature trend described above for the $2\text{-}\mu \mathrm{m}$ data, falling between the $A/E_{\mathrm{a}}$ values for ambient temperatures of 30°C and 40°C without pigment. We applied the same scaling method to these two empirical threshold thermal profiles and found that they were farther from the Arrhenius ideal values than most of the nonpigmented profiles. The 1- and 3-s thermal data required a 16% (SF 1.160) and 10% (SF 1.100) increase at each integration time point to achieve $\mathrm{\Omega }=1$, respectively, which is also remarkable overall.

4.3.

Comparisons across Experimental Parameters

Comparing data within the nonpigmented and pigmented sets was straightforward because the laser parameters and the layers making up samples were the same. However, there were a few salient points for comparison between the nonpigmented and pigmented data. Both had a 1-s exposure. The threshold peak temperature values for these 1-s exposures differed by about 3.6°C, with the bulk water heating at $2\mu \mathrm{m}$ being the higher value. One explanation for this is that because the pixels of the thermal camera were $\sim 8\mu \mathrm{m}\times 8\mu \mathrm{m}$, it was unable to capture the expected higher temperature rise near the surface of the absorbing $1\mu \mathrm{m}\times 1\mu \mathrm{m}$ MP. Thus, the bulk water was being heated equally across the pixel area by the $2\text{-}\mu \mathrm{m}$ laser, but the heating around the tiny MP had diffused significantly to fill the area of the single pixel of the thermal camera. However, at the highest level of pigmentation ($800\mathrm{MPs}/\text{cell}$), where clumping of MPs was most common, the shape of the thermal profile was different (Fig. 6) even though the threshold peak temperature value was no greater than that generated by the lowest level of pigmentation.

Additionally, Table 2 includes the $A/E_{\mathrm{a}}$ values for the pigmented data at an ambient temperature of 35°C. Understandably, the data are unreliable due to having only two data points, but the values estimated for $A/E_{\mathrm{a}}$ fall into the expected trend (between 30°C and 40°C values for nonpigmented exposures). Also of note is how close the 1-s 35°C threshold peak temperature values (average of 52.3°C) is near the $T_{\mathrm{crit}}$ value (Table 2), as per expected from Eq. (5).