1.

Introduction

To provide guidelines for eye-safe exposure to lasers, the laser safety community has relied mainly on damage assessment in nonhuman primate studies. Results of in vivo studies have shown that laser damage in the retina depends upon wavelength, power level, and duration of the exposure.^{1, 2} Histopathologic analysis of retinal damage indicates that damage to the retinal pigment epithelial (RPE) cells, due to absorption by intracellular melanosomes, is universal in all three known mechanisms leading to laser-induced lesion formation.^{1, 3} Thus, pigmentation plays an important role in laser absorption and damage.

Analyzing trends in damage threshold data can provide information regarding mechanisms, as described in the review by Stuck.⁴ The familiar action spectrum, relating threshold to incident laser wavelength, illustrates a dramatic increased efficacy in the blue and ultraviolet region of the spectrum, indicating a transition to a photochemical (actinic) mechanism. Distinguishing trends can also be found when relating threshold (laser radiant exposure) to exposure duration on log-log axes, an analysis we describe as the “temporal action profile.” Thermal damage mechanisms are thought to predominate for the shorter exposures, where there is a nonlinear increase (power function) in threshold radiant exposure with increasing exposure time. At longer exposure durations (especially for actinic wavelengths), the trend is for the threshold radiant exposure to remain constant with increasing exposure duration. This identifies the principle of reciprocity, whereby a reciprocal relationship exists between threshold irradiance and exposure duration.^{5, 6} In the transitional phase between these opposing trends (around $20\mathrm{s}$ ), there is most likely a complex interplay of both thermal and chemical outcomes.

Over the years, there have been theoretical models developed to describe photothermal damage,^{7, 8} and the in vivo threshold data has played a vital role in validating these models. Specifically, computer-simulated retinal temperature rises, coupled with the generally accepted temperature dependence for the denaturation of proteins, has provided support to the basic data trends in temporal action profiles as to when damage mechanisms transition from thermal to photochemical.^{2, 4} Currently, there are no computer modeling and simulation programs that accurately predict photochemical damage. Biochemical analysis of ocular lesions is needed to identify key components, such as reactive oxygen species (ROS), responsible for photochemical injury. Once time-dependent and wavelength-specific components are identified, a rate-process approach to modeling and simulation can proceed. Even with the utility of in vivo–based laser-capture microdissection,⁹ the prospect of biochemically analyzing sufficient biomaterial of a retinal-visible lesion generated in the intact eye is an arduous task. A simple in vitro RPE model system, if found to respond to lasers in similar fashion to RPE cells in vivo, could lead to validation of modeling refinements and provide ample tissue for proteomic and transcriptomic analysis.

In the present work, we have used our previously described⁹ in vitro cell model to measure estimated dose for 50% lethality ( ${\mathrm{ED}}_{50}$ ) thresholds for four exposure durations at two different wavelengths, twice varying the melanosome density within the cells. We show that although the absolute values of the thresholds differ from minimal visible lesion (MVL) data taken from the literature, we see trends in the temporal action profile that are consistent with that from ocular exposures in rhesus monkeys. We also note that our analysis of 532-nm exposures precedes safety studies in animals. Overall, this is a second validation of the in vitro model and a starting point for both refining existing computer models and identifying interesting laser exposures to study at the molecular level.

2.

Materials and Methods

2.2.

Lasers, Beam Delivery, and Cell Exposures

A large-frame argon laser (Innova 200, Coherent) was used for its 457.9-nm line (referred to hereafter as $458\mathrm{nm}$ ). A diode-pumped ND:YVO4 laser (Millennia Xs, Spectra Physics), with intracavity doubling, was used for the 532-nm exposures.

Figure 1 provides a schematic representation of the laser delivery to cells in the 96-well plates. Attenuation of laser power was achieved by the combination of a half-wave plate $(\lambda ∕2)$ and polarizing beamsplitter (Pol). All beams were then coaligned to a common optical path using apertures and a flip-up mirror (F). The optical path included a telescope (T), a beam shaper (BSh, model GBS-AR14, Newport Corporation), a computer-controlled shutter (S), and a single lens (L) imaging system (88-mm focal length) generating a beam diameter of about $250\mathrm{\mu }\mathrm{m}$ at the cells. The telescope allowed for collimated beam expansion to $4.7\mathrm{mm}$ prior to entry into the beam shaper, which converted the beam to a flat-top profile. The imaging system was designed to image the beam at the near-field output of the beam shaper (8-mm diam) via $0.05\times$ magnification. The effect of the column of HBSS above the cells during exposure was taken into account when identifying laser beam diameter (knife-edge method).

Fig. 1

Laser delivery for in vitro damage threshold experiments. M, mirror; F, flip-up mirror; Pol, polarizing cube; T, optical telescope; BSh, beam shaper; S, shutter; L, lens; PM, power meter; ND, neutral density filter; CCD, charge-coupled device camera; Obj, microscope objective.

Cells were systematically exposed to laser irradiation for time intervals between $0.1\mathrm{s}$ and $100\mathrm{s}$ at irradiance ranges useful for determining viability thresholds. The 96-well plates were suspended (without lids) in the beam path using a specialized holder attached to $x\text{-}y$ translational stages equipped with computer-controlled motors. Ambient temperatures during the experiments ranged from 18 °C to 25 °C.

3.

Results

3.1.

Laser Damage Thresholds

Figures 2a to 2d show fluorescence [panels (a) and (c)] and bright field [panels (b) and (d)] images for each of the two levels of pigmentation in our hTERT-RPE1 cells. As expected, there was substantially more pigmentation distributed throughout the monolayer when adding more melanosomes. Because we saw no residual MP after the phagocytosis period, we expect the two levels of pigmentation to be about tenfold different. Also note that there was sufficient calcein-AM dye in the cells containing the high density of MP to assay for damage. Figures 2e to 2l provide representative fluorescence images (overlay of red and green channels) for the laser exposures used in our threshold determinations. Note the altered fluorescent staining pattern (red and devoid of green) in the center of the image, indicating cell injury in the region of the monolayer exposed to the damaging effects of a laser.

Fig. 2

Microscopic analysis of RPE pigmentation and laser damage. Panels (a) to (d) represent unexposed hTERT-RPE1 monolayers at $160\mathrm{MP}∕\mathrm{cell}$ [(a) and (b)] and $1600\mathrm{MP}∕\mathrm{cell}$ [(c) and (d)]. Images were either bright field [(b) and (d)] or stained for viability [(a) and (c)]. Panels (e) to (l) represent cells stained for viability after laser exposure at $458\mathrm{nm}$ [(e), (f), (i), (j)] or $532\mathrm{nm}$ [(g), (h), (k), (l)]. Pigmentation of exposed cells was either $160\mathrm{MP}∕\mathrm{cell}$ [(e) to (h)] or $1600\mathrm{MP}∕\mathrm{cell}$ [(i) to (l)]. Laser irradiances $(\mathrm{W}∕{\mathrm{cm}}^2)$ were as follows: (e), 40; (f), 60, (g), 200; (h), 500; (i), 30; (j), 80; (k), 50; (l), 120. All images were acquired using a $20\times$ objective and images (a) to (d) were enlarged to show detail. All spatial bars represent 250 microns.

Damage threshold $\left({\mathrm{ED}}_{50}\right)$ values for various laser exposure conditions are listed for hTERT-RPE1 cells pigmented with doses of MP equivalent to $160\mathrm{MP}∕\mathrm{cell}$ (Table 1) and $1600\mathrm{MP}∕\mathrm{cell}$ (Table 2). Irradiance and radiant exposure thresholds are listed in the tables, but the Probit output data are provided only for the analysis of radiant exposure. Overall, the Probit results indicated statistical confidence in the data sets for each exposure condition. Fiducial limits (representing 95% confidence intervals) within each data set were 5% to 33% from their respective ${\mathrm{ED}}_{50}$ value.

For exposures to cells containing $160\mathrm{MP}∕\mathrm{cell}$ , the damage thresholds over the 1-s to 100-s range of exposure durations for $458\mathrm{nm}$ were consistently lower than those for $532\mathrm{nm}$ . This trend indicates enhanced killing efficiency by exposure to the shorter wavelength, except at the 0.1-s duration. Irradiance thresholds within each wavelength decreased as exposure durations were extended, while the radiant exposure thresholds increased.

Trends for thresholds in cells containing $1600\mathrm{MP}∕\mathrm{cell}$ were different from those described for cells with $160\mathrm{MP}∕\mathrm{cell}$ . The thresholds for exposure to $532\mathrm{nm}$ were consistently lower over the range of exposure durations than those for $458\mathrm{nm}$ , indicating that the longer wavelength was more efficient at killing cells. This pigment-dependent difference in efficiency was especially apparent at the 100-s data point, where the fold-difference between $458\mathrm{nm}$ and $532\mathrm{nm}$ was 3.7 $(160\mathrm{MP}∕\mathrm{cell})$ and 0.77 $(1600\mathrm{MP}∕\mathrm{cell})$ .

Table 3 summarizes the pigment-dependent changes in the laser damage threshold. In all conditions compared, a reduction in threshold was found when the pigment density was increased tenfold. The pigment dependence appeared to be less pronounced at the shorter wavelength. The longer exposure duration had the most dramatic effect between the two wavelengths (4.9-fold). Within each wavelength, the general trend was for the pigmentation dependence to be more pronounced at the shorter exposure durations.

Figures 3, 4, 5 summarize our data as temporal action profiles, where radiant exposure was plotted versus exposure duration on log-log axes, and which allowed for comparisons with in vivo laser damage studies reported in the literature. Figure 3 shows our 458-nm temporal action profile in comparison with that from Ham ² To extend our in vitro data set to $3600\mathrm{s}$ , we have included a data point from an earlier publication.¹⁰ Our in vitro thresholds in cells containing $160\mathrm{MP}∕\mathrm{cell}$ showed a trend with a striking resemblance to the in vivo data. Both curves increased linearly (log-log plot) over the shorter exposure durations (up to and including $10\mathrm{s}$ ), and both leveled off with respect to radiant exposure for the longer exposures ( $10\mathrm{s}$ and longer). This occurrence has been termed “reciprocity”⁵ and indicates a change in damage mechanisms from one predominately thermal to one that is more photochemical in nature. The trend for in vitro exposure of cells having greater pigmentation did not show reciprocity at the 100-s duration point.

Fig. 3

Temporal action profiles for in vitro thresholds and in vivo (Ref. 2) thresholds at $458\mathrm{nm}$ . The equations (power function with $R^2$ correlation coefficient) provided in the figure represent only those data corresponding to the linear trends (log-log plot) in each data set. The data point for $160\mathrm{MP}∕\mathrm{cell}$ $3600\mathrm{s}$ is taken from Denton ¹⁰

Fig. 4

Temporal action profiles for in vitro $\left(532\mathrm{nm}\right)$ thresholds and in vivo ( $580\mathrm{nm}$ and $514.5\mathrm{nm}$ from Ref. 2) thresholds. The equations (power function with $R^2$ correlation coefficient) provided in the figure represent only those data corresponding to the linear trends (log-log plot) in each data set.

Fig. 5

Temporal action profiles for in vitro $\left(532\mathrm{nm}\right)$ thresholds and in vivo ( $514.5\mathrm{nm}$ from Ref. 15) thresholds. The equations (power function with $R^2$ correlation coefficient) provided in the figure represent only those data corresponding to the linear trends (log-log plot) in each data set.

Table 3

Pigment dependence of laser-induced cytotoxicity in vitro.

Exposure condition	Fold reduction in ED50 (160MP∕cell1600MP∕cell)
$532\mathrm{nm}$
$100\mathrm{s}$	5.9
$10\mathrm{s}$	6.9
$1\mathrm{s}$	8.6
$458\mathrm{nm}$
$100\mathrm{s}$	1.2
$10\mathrm{s}$	4.7
$1\mathrm{s}$	4.2

When best-fit equations for a power function were determined for the linear portions (log-log plot) of the three curves in Fig. 3, it was found that the slopes (exponents) were similar in magnitude (0.86). The data for the in vitro exposures in Fig. 3 showed the pigment-dependent reduction in threshold values for cells containing more MP. The comparison also explained the apparent lack of a pigment dependence at $100\mathrm{s}$ for the 458-nm exposures as being due to no reciprocity at the higher level of pigmentation.

Because there have been no threshold studies reported in animals for ocular laser exposures at $532\mathrm{nm}$ , we have made comparisons with studies using $514.5\mathrm{nm}$ and $580\mathrm{nm}$ . Figure 4 makes comparisons with $514.5\mathrm{nm}$ and $580\mathrm{nm}$ taken from Ham ² Both in vitro temporal action profile curves (power function) were linear (log-log plot) over the 100-s exposure duration, with average slopes of approximately 0.85. The in vivo data representing 580-nm exposures gave a linear temporal action profile (log-log plot) over the 1000-s range (slope of 0.73), whereas the data for 514.5-nm exposures showed reciprocity ( $10\mathrm{s}$ to $1000\mathrm{s}$ ). Between the 1-s and 10-s data points for the 514.5-nm exposures, the slope was 0.88. From Ham’s data in Fig. 4, the wavelength transition for achieving reciprocity in vivo fell somewhere between $514.5\mathrm{nm}$ and $580\mathrm{nm}$ . Our in vitro data would suggest that this transition occurs at a wavelength shorter than $532\mathrm{nm}$ . Although Figs. 3 and 4 suggest that conditions in the intact eye lead to greater sensitivity for damage from laser exposure, we provide other explanations in Sec. 4.

Figure 5 shows 514.5-nm data from a different animal MVL study¹⁶ for comparison with our 532-nm exposures. Again, the slope of the power function equation for the animal data (0.80) approximated our in vitro data at $532\mathrm{nm}$ . Unlike the 48-h post-exposure data points given by Ham, ² Gibbons and Allen¹⁶ assessed for damage at both $1\mathrm{h}$ and $24\mathrm{h}$ post-exposure. The extended time of recovery after 514.5-nm exposure was sufficient to allow cells to respond and follow the principle of reciprocity, presumably increasing the degree of photochemical damage relative to thermal damage.

4.

Discussion

Our goal was to identify a cellular model that metabolically reacts to laser exposure similarly to RPE cells in nonhuman primate models. Our approach was to determine threshold $\left({\mathrm{ED}}_{50}\right)$ radiant exposures for damage over a broad range of durations of exposure and compare with animal (in vivo) thresholds reported in the literature. We found that our threshold values differed from two animal studies, most likely because of differences in several exposure parameters that are discussed later. We contend that validation of our model system should primarily focus on how it responds to varying pigmentation, wavelength, and exposure duration (i.e., trends in the threshold data). As described in Sec. 1, threshold trends for in vivo exposures indicative of transitions from photothermal to photochemical damage mechanisms have been correlated with calculated temperature-rise data, thus making such comparisons between in vivo and in vitro threshold trends meaningful.

Our data showed an increase in sensitivity to laser exposure at both $458\mathrm{nm}$ and $532\mathrm{nm}$ when intracellular melanosome density in our hTERT-RPE1 cells is increased. The trend (8.6-fold) was most prominent at exposure conditions most consistent with thermal damage (longer wavelength and shortest exposure duration). This type of trend was not evident (1.2-fold) at conditions consistent with photochemical damage (shorter wavelength and longest exposure duration), although Fig. 3 shows that this is because reciprocity had not yet occurred $\left(100\mathrm{s}\right)$ for cells having the higher pigmentation.

Thresholds for cells having both levels of pigmentation followed the general trend, where less irradiance was required as exposure duration was extended. When expressed in a temporal action profile, threshold radiant exposure increased as described by a power function. Within each wavelength, the slope (exponent of equation) for the $1600\mathrm{MP}∕\mathrm{cell}$ thresholds was slightly greater than the slope for the $160\mathrm{MP}∕\mathrm{cell}$ thresholds, which is expected considering the results shown in Table 3. One surprising result was how the ${\mathrm{ED}}_{50}$ values for exposures to cells with $1600\mathrm{MP}$ showed an opposite trend with respect to wavelength. Notice in Table 2 that the 532-nm thresholds were lower than the 458-nm thresholds for each given exposure duration.

Among our in vitro damage threshold analyses, only results from exposures to RPE cells containing $160\mathrm{MP}∕\mathrm{cell}$ revealed significant features in common with data taken from animal studies in the literature (Figs. 3, 4, 5). At this level of artificial pigmentation, the transition to reciprocity for exposure to $458\mathrm{nm}$ (Fig. 3), but not $532\mathrm{nm}$ (Fig. 5), for our 1-h post-exposure endpoints was considered supportive of our in vitro model. Because the $1600\mathrm{MP}∕\mathrm{cell}$ system failed to follow reciprocity $\left(458\mathrm{nm}\right)$ or increased sensitivity at the shorter wavelength, we do not consider it a good universal model for studying laser bioeffects. Another positive feature found in Fig. 3 was how the transition from thermal to photochemical damage occurred at around the 20-s exposure duration for both the in vitro and in vivo exposures at $458\mathrm{nm}$ .

The data in Figs. 4 and 5 support the notion that our in vitro system $(160\mathrm{MP}∕\mathrm{cell})$ responds in a fundamentally similar fashion to 532-nm laser irradiation, as does the nonhuman primate retina. Combined, the in vivo data in Figs. 4 and 5 predict that a break point for photochemical damage mechanisms occurs at or around $514\mathrm{nm}$ and is dependent upon whether one assesses for acute ( $1\text{-}\mathrm{h}$ post-exposure) or chronic (minimum of 24-h post-exposure) damage. From the data trends shown, the presumed thermal response of the hTERT-RPE1 cells exposed to $532\mathrm{nm}$ complied with expected results in the animal model.

Additionally, the data found in Figs. 3, 4, 5 serve to highlight the importance of our analyses of data trends, rather than the absolute values for thresholds at each wavelength and exposure duration. Our thresholds fell consistently above the values from Ham ² and below those of Gibbons and Allen.¹⁶ We attribute the discrepancies to differences in experimental parameters. For example, Ham did not use the Probit analysis for identifying their thresholds the way it was done in both this study and that of Gibbons and Allen. Rather, Ham determined the lowest irradiance that led to an ophthalmoscopically observable lesion $48\mathrm{h}$ post-exposure. This would obviously lead to a lower threshold value than the estimated dose that is predicted to produce a lesion 50% of the time $\left({\mathrm{ED}}_{50}\right)$ . We have already made note of the effect of extending the time of examination for damage, which allows for added physiological responses, typically decreases thresholds in the retina, and is expected to be more pronounced for photochemical damage mechanisms. In regard to the data of Gibbons and Allen that we used in Fig. 5, uncertainty has arisen regarding their ability to image a small laser beam to the same spot at the retina for such extended lengths of time. A “wobble” of the beam would lead to an increase in the power required to achieve damage;¹⁷ however, it is difficult to make direct comparisons when only one group determines ${\mathrm{ED}}_{50}$ values.

Some basic physical properties that differed between our in vitro experiments and the animal studies reported in the literature include laser beam profile, ambient temperatures during exposure, and the depth and composition of materials preceding the RPE cells. The difference between average and maximum threshold irradiances for the flat-top beam profiles used in our study is minimal compared to the difference for the Gaussian beam profiles used by Ham This alone could account for about a twofold difference in threshold value. Initial temperature of exposed tissue would influence dose requirements to achieve both thermal and photochemical damage (threshold) to cells. A lower initial temperature (such as the 18 °C to 25 °C used in our cell culture experiments), relative to retinal (“body”) temperature of the nonhuman primate is expected to increase the required laser dose to achieve the temperature rise associated with thermal damage. The known temperature dependence of all chemical reactions (rate doubles with each increase of 10 °C) would also predict an increase in threshold laser dose for exposures at lower ambient temperatures. Ongoing experiments in our laboratory are designed to examine these temperature-dependent effects on the laser damage threshold in cultured cells.

Last, the depth and composition of liquid media and tissues that the laser beam must pass through before reaching the RPE cells differed greatly between the cell culture method and the intact globe. The cell culture method eliminated much of the scattering that occurs when light passes through the various layers of the eye. Indeed, we have begun an assessment of the effects of both buffer depth and presence of macular carotenoids in our model system in an effort to aid computational modeling of exposures of blue laser light to the paramacula and macula.

We conclude that the artificially pigmented system described (hTERT-RPE1 cells at $160\mathrm{MP}∕\mathrm{cell}$ ) is a good model for studying laser bioeffects, although subtle modification in pigmentation may optimize results relative to MVL data. Wavelength and exposure duration trends in our data indicate that our artificially pigmented RPE cells respond to laser dosimetry via fundamentally similar mechanisms when compared to native RPE cells. The cell culture model has an easily manipulated environment, provides laser delivery with relative ease, and can generate large data sets quickly.