Content uploaded by Michael L. Denton
Author content
All content in this area was uploaded by Michael L. Denton on Jun 23, 2014
Content may be subject to copyright.
Determination of threshold average temperature for cell death
in an in vitro retinal model using thermography
‡
Michael L. Denton
a
, Michael S. Foltz
a
, Gary D. Noojin
a
,
Larry E. Estlack
b
, Robert J. Thomas
c
, and Benjamin A. Rockwell
c
a
Northrop Grumman, Warfighter Concepts and Applications Department,
San Antonio, TX, USA, 78228-1330
b
Conceptual MindWorks, San Antonio, TX, USA, 78228
c
Air Force Research Laboratory, 711 HPW/RHDO, Brooks City-Base, TX,
USA, 78235-5214
ABSTRACT
Even though laser exposures of 1 s or less are non-isothermal events, researchers have had to rely upon the
isothermal treatise of Arrhenius to describe the laser damage rate processes. To fully understand and model
thermal damage from short exposure to laser irradiation we need to experimentally obtain the temperature
history of exposed cells and correlate it with the cellular damage outcomes. We have recorded the thermal
response of cultured retinal pigment epithelial cells in real-time with laser exposure using infrared imaging
(thermography). These images were then overlaid with fluorescence images indicating cell death taken 1 hr
post laser exposure. The image overlays allowed us to define the thermal history of cells at the boundary
(threshold) of laser-induced death. We have found a correlation between the onset of cell death and the
average temperature over the course of the laser exposure.
Keywords: laser, damage, Arrhenius, thermography, cultured cells, RPE, fluorescence, threshold
1. INTRODUCTION
Many computational methods for modeling and predicting thermal laser-induced damage in various biological
tissues presently implement the damage integral,
1,2
which is based on the Arrhenius formulation. Because
the Arrhenius equation was originally adopted from the van’t Hoff equation describing the isothermal and
time dependence of chemical equilibrium rates (thermodynamic analysis), its utility in correlating thermal
rates with measured physical processes such as protein denaturation upon chronic heating, has been
established.
3,4
However, its use in calculating rates of non-isothermal laser damage has been largely
empirical and difficult,
5,6
and there is the need for new concepts regarding rate processes in the field of laser-
tissue interaction.
6,7
In an effort to more fully understand the thermal requirements for cellular damage we have devised a
method to identify a thermal metric that can replace the damage integral as a threshold indicator for death.
The key components of the method include thermal imaging at high magnification, damage assessment
using fluorescence imaging, and the careful registration of pixels from the two types of images. It is
assumed that cells at the periphery of laser-damaged regions of cultured cells have all seen equivalent
threshold temperature-time histories leading to cell death. Our findings suggest that the average
temperature over the course of the laser exposure duration qualifies as a very simple metric for
distinguishing whether a cell is ultimately destined to die due to thermal processes.
‡
Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed
by the United States Air Force.
Invited Paper
Optical Interactions with Tissue and Cells XX, edited by Steven L. Jacques, E. Duco Jansen, William P. Roach,
Proc. of SPIE Vol. 7175, 71750G · © 2009 SPIE · CCC code: 1605-7422/09/$18 · doi: 10.1117/12.807861
Proc. of SPIE Vol. 7175 71750G-1
2. METHODS
2.1 Cell culture
The in vitro retinal model was used as described previously,
8
except that 48-well microtiter dishes were used
rather than 96-well plates. Cells (human derived hTERT-RPE1 cell line purchased from BD Biosciences
ClonTech Labs, Palo Alto, CA) were seeded in 48-well plates at 70,000 cells per well, pigmented the
following day with isolated bovine melanosome particles (MPs) such that there were approximately 200 MPs
per cell, and exposed to the laser on the second day post-seed. Adhering to this schedule provided
monolayers with consistent cell density with good overall viability. Prior to exposure to the laser, cells were
transferred to a heated glove box (37 ºC) and twice rinsed with preheated Hank’s Balanced Salt Solution
(HBSS). Each well then received 100 µL HBSS and taken to the bench for laser exposure.
2.2 Laser exposures of cells
Figure 1 illustrates the laser delivery scheme used to expose RPE cells to the 514-nm line of a large-frame
argon laser (Model Innova 200, Coherent). Verification of laser wavelength was performed with a
spectrometer (Ocean Optics). Attenuation of laser power was achieved by the combination of a half-wave
plate and polarizing beam splitter. The beam was launched into a multimode 300 µm fiber, the output of
which was imaged to the cell monolayer using a 3:1 imaging system. A mirror located beneath the 100-mm
IR imaging lens was used to redirect the beam to the cells. Using a spatially calibrated imaging system, the
diameter of the laser beam was measured to be 0.93 mm at the cell monolayer.
Each 48-well plate was suspended (without lids) in the laser beam path using a specialized holder attached
to x-y translational stages equipped with computer-controlled motors. Ambient temperature was held
constant (35 – 36 ºC) throughout each experiment using a plexiglass enclosure, which also provided an
environment of consistent relative humidity (60 – 70%). Cells (one exposure per well) were systematically
exposed to the laser for 0.1, 0.25, or 1.0 s durations at irradiance ranges useful for determining viability
thresholds using the Probit method (see Section 2.4). We used a computer software program to
simultaneously trigger the real-time imaging (Section 2.3) and the mechanical laser shutter. We insured that
the cells were without HBSS for no longer than 30 – 45 s throughout the procedure for each well, including a
quick step of focusing the cells on the CCD camera using a z-directional micrometer.
2.3 Imaging during laser exposure
We used a FLIR SC6000 IR camera to obtain IR images (3 - 5 µm) at 800 fps. With the 100-mm IR lens in
place (Figure 1), the effective pixel pitch and image magnification are calculated to be 8.12 um/pixel and
3.08x, respectively. Radiometric Calibration of the thermal camera was achieved using a calibrated portable
blackbody source (M316, Mikron) placed at the focus of the imaging system. At the exposure setup, the
calibration was verified by placing a painted aluminum plate with known emissivity and temperature at the
image plane of the thermal camera. The temperature of the painted plate was directly measured using a
thermocouple located inside the plate. A thermal camera offset was then adjusted until agreement between
the thermocouple and thermal camera measurements was achieved. At the end of each exposure set, the
painted plate was inserted at the image plane and the initial calibration verified. This procedure was
repeated for each exposure set.
A long-working distance 5x Mitutoyo microscope objective placed beneath the cell culture plate was used to
image (9 fps) the cell monolayer with a Hamamatsu ORCA 100 CCD camera. Once a positive test target
was in focus for both the CCD and IR cameras, we could use the CCD camera to focus on cells prior to laser
exposures and have confidence that the cells were also in focus of the IR camera. It was at this location in
space within the chamber that we measured the spot size of the laser.
2.4 Damage assessment
After laser exposures to cells in the microtiter plate, the HBSS was replaced with complete growth medium
(pre-warmed) and the cells were placed at standard growth conditions for 1 hr. At this time, cells were
assayed for viability using 1.7 µM calcein-AM and 1.4 µM Ethidium homodimer 1 (EthD1) in 0.1 mL HBSS
Proc. of SPIE Vol. 7175 71750G-2
A
Acryl
En
high capacity
heater
/
Computer-driven
x-y translational
stage
IR
Camera
Optical Window
(MgF
Cells in 48-well plate
Large-frame
Argon Laser
w/ fiber delivery
Video
Microscope
WA.
edge of
encIosure
Laser
Fiber
Figure 1. Laser delivery and imaging systems within the environmentally controlled enclosure. A. Schematic diagram
showing how the cells were suspended at the focal point of both the IR and video cameras. B. Photograph of camera
placement and laser fiber delivery.
(10 min at 37ºC). Exposure sites within wells were identified as stained positive for damage when nuclei
were fluorescent with EthD1 (red = dead) and as a region devoid of staining by calcein-AM.
Scoring of damage by three individuals was blind of dosimetry and a score (yes/no) for damage required a
consensus from two. These binary data were input into the Probit software package.
9,10
In addition to
probability-dose information (ED
50
), the Probit output includes uncertainty intervals (fiducial limits at 95%
confidence) related to the ED value, and the Probit slope (first derivative at a probability of 0.5 for ED
50
).
Furthermore, EthD1 or calcein fluorescence images from laser exposures generating cell death were used to
determine the extent of damage (area), and for overlaying with IR images.
3. RESULTS AND DISCUSSION
3.1 Laser damage thresholds
Figure 2 shows examples of damage results after laser exposure to the in vitro retinal model. Notice that
each well of the culture plate has only one laser exposure, and that some of the exposures do not lead to
damage (panel b). When damage was evident (red fluorescence), there were various sizes and shapes of
fluorescence regions (panels a, c, d, and f). Panel e was an unexposed control well.
Table 1. Threshold ED50 values in the in vitro retinal model when exposed to
0.93-mm beam at 514 nm.
Table 1 provides the Probit ED
50
threshold values for each of the
three laser exposure durations. Note
that the threshold irradiances
decrease as the exposure duration is
lengthened while the opposite
correlation holds true for the
threshold radiant exposures. The
variance in our Probit data
(comparing fiducial limit values with
ED
50
values) was low (8 - 13 %).
Exposure
Duration
(sec)
Threshold ED
Irradiance (W/cm )
LFL ED UFL
50
50
2
0.10
0.25
1.00
95 104 112
Threshold ED
Rad. Expos. (J/cm )
LFL ED UFL
50
50
2
9.5 10.4 11.2
80 91 103
33 38 42
20.0 22.8 25.8
33.0 38.0 42.0
#of
exposures
55
59
55
LFL; lower fiducial limit (95% confidence)
UFL; upper fiducial limit (95% confidence)
Proc. of SPIE Vol. 7175 71750G-3
e5 4 dual lii e6 4x dual.til
c5 4
dual.til
c
4x d'.jal.lil
d5 4x dual.tiI
d6 4x dual.til
Together, these results show that the in vitro retinal model was functioning properly, which validates our
measurement of cellular response to laser exposure using the model.
A comparison of the results in Table 1 with the 514-nm threshold data in
our previous publication,
8
where the spot diameter was 0.25 mm,
identifies a substantial difference in damage susceptibility. The
previous threshold irradiance values for 0.1-s (463 W/cm2) and 1.0-s
(292 W/cm2) exposures are about 4.5 and 7.7 fold greater than the data
presented here, respectively. This data supports the notion of a spot
size dependence in photothermal damage mechanisms.
3.2 Correlating thermal history with cell death
3.2.1 Full-frame “hottest” pixel method
In our first method for correlating thermal history with cell death, we
located the frame of each thermal movie corresponding to the end of
the laser exposure and identified the pixel with the greatest
temperature. The temperature (rise) value for this pixel was then
integrated over the duration of the laser exposure (T
int
). The T
int
(ºC * s)
and damage outcome (Probit input data) was plotted and correlated
with laser radiant exposure (Figure 3).
Figure 2. Fluorescence detection of laser
induced damage. (e) unexposed control well.
Figure 3. (a) Relationship between laser dose and thermal response in the in vitro retinal model. The T
int
was
calculated for the hottest pixel in each thermal movie and plotted against the laser radiant exposure corresponding to that
exposure. Damage results are shown, where open and closed symbols represent no laser damage and laser damage,
respectively. Diamond (teal) symbols, 0.1-s exposures; square (green) symbols, 0.25-s exposures; triangle (blue)
symbols, 1.0-s exposures. (b)
T
int
values associated with best-guess distinction between damage-no damage results in
(a) plotted versus the respective laser exposure duration.
10
100
100 ms Damaged
100 ms No Damage
250 ms Damaged
250 ms No Damage
1,000 ms Damaged
1,000 ms No Damage
0
1
1 10 100 1000
Radiant Exposure of Laser (J/cm )
2
T for Duration of Laser Pulse
Single Pixel ( C * sec)
int
o
(a)
y = 13.981x
R² = 0.9996
Exposure Duration
(b)
Crude Threshold T
int
(C*s)
0
12
10
8
6
4
2
0 0.2
0.4
0.6 0.8
1.0
0
14
0
Proc. of SPIE Vol. 7175 71750G-4
(a)
5 000 +00
4 SUE +00
400E+OU
4 40E +00
4200+00
3500+00
3.60 E.00
3 40E+OO
3 20E +00
3000+00
h
(b)
5.000*00
4.80E+00
4.60E*00
4.40E+0O
4.200*00
4.00E+00
3.80E*00
3.60E+0O
3.40E*00
3.20E*00
3005*00
The symbols in Fig. 3(a) are color-coded to allow the distinction between the 3 laser exposure durations.
Overall, there appears to be a good correlation between T
int
and laser radiant exposure, regardless of
exposure duration or damage result (y = 0.115x
1.17
with R
2
= 0.79 for all data points).
There was some overlap between data points for no damage with data points for damage of the next shorter
laser exposure. However, there was a more scatter in the data from the 1-s exposures that exceeded the
data for the shorter exposures. It is conceivable that the effect is the result of some photochemical damage
mechanism when the duration of the exposure is lengthened to 1.0 s.
A general approximation of the threshold T
int
for each exposure duration can be made by drawing a line
parallel to the x-axis of Figure 3(a) between the open (no damage reported) and closed (some damage
reported) symbols (dashed lines). When these values (“crude” threshold T
int
) are then plotted versus their
respective laser exposure duration (Figure 3(b)) we see a linear relationship. The slope of the line presented
in Figure 3(b) represents the average T
int
at each of the exposure durations, and is thus a measure of
threshold average temperature. Because we expected no cell death in the absence of laser exposure, we
set the line in Figure 3(b) to go through the origin, and the correlation was very good. This result was very
interesting because it implied that the threshold average temperature was the same for all three exposure
durations. Because the analysis shown in Figure 3 was an approximation based on a single pixel in the
thermography data, we describe the determination of the threshold average temperature values using two
rigorous methods in the following sections.
3.2.2 Determination of threshold average temperature using damage areas
In order to identify the values for T
int
that correspond to those cells having the minimum thermal dose to
cause death, we looked to the fluorescence images for each exposure. We identified pixels in the
fluorescence images that indicated regions of laser-induced death and calculated the corresponding damage
areas. A LabVIEW program (SAF analysis) was written to extract and analyze thermal movies created with
RTools (FLIR Systems). This program allowed the extraction of T
int
information for full-frames. Using our
LabVIEW program we calculated full-frame T
int
maps (Figure 4) for each exposure leading to cell death. We
moved an imaginary plane in the y-axis that intersected the T
int
map for each exposure until the area
delimited by the plane was equal to the damage area of the corresponding fluorescence image. Notice that
the thermal response of the cells to the flat-top laser beam was essentially Gaussian, except in the very
center where varied pigmentation caused heterogeneities in absorption (Figure 4(b)).
Figure 4. Full-frame T
int
map for a thermal movie recorded during laser exposure. T
int
values were calculated for each
pixel in a thermal movie using our LabVIEW program
Proc. of SPIE Vol. 7175 71750G-5
Dual fluorescence
The T
int
map with the matching area was then overlaid (appropriate registrations) with the fluorescence
image. Figure 5 provides an example of T
int
and fluorescence image overlays. Notice how well the T
int
map
correlates with the size and shape of the damaged region of the cell monolayer.
Figure 5. Comparison images for overlay between
fluorescence damage detection and
T
int
map with
damage area thresholding.
Because we assessed the T
int
map in this
manner, the edge of T
int
values after the
threshold was at the boundary of cell death
caused by the laser, and we calculated the
average T
int
values for each exposure duration
and plotted as shown in Figure 6. Again, the slope of the line produced represents the threshold average
temperature value at each of the three exposure durations. Also notice how closely the slope generated
from this rigorous method (Figure 6) matches the slope of our approximation method (Figure 3(b)).
Figure 6. Plot of average T
int
values
using the damage area thresholding
method versus their corresponding
laser exposure durations.
3.2.3 Determination of threshold average temperature using average pixel history
A second rigorous method was used to determine the average temperature values of cells at the threshold of
death. This method relied on our ability to overlay the thermal and fluorescence images (with proper
registrations). This allowed us to identify the pixels of any thermal image that correspond to the pixels (cells)
at the boundary of cell death in the fluorescence image. Once the pixels in the thermal image were mapped
to the region of interest (boundary), each of their T
int
values were calculated over the course of the laser
exposure and averaged. This process was independent of size and shape of the damage zones found by
fluorescence microscopy, and the number of pixels averaged for a given thermal image therefore varied from
100 – 650.
Figure 7 shows the results of the threshold average temperature values at each exposure duration for both of
the rigorous methods described. Both methods provided an identical result for each exposure duration.
Figure 7 indicates that, unlike the slope analysis found in the T
int
versus laser exposure duration plots, the
threshold average temperature values for the 0.1-s (11.7 ºC) and 0.25-s (14.5 ºC) exposures are significantly
different. The variance in the 1-s exposure data causes its threshold average temperature value to overlap
the data from with both shorter exposures.
y = 13.662x
R² = 0.9991
8
10
12
14
16
0
2
4
6
0 0.2 0.4 0.6 0.8 1
Laesr Exposure Duration (s)
o
Proc. of SPIE Vol. 7175 71750G-6
00
50 -
t.o -
eo -
80 -
Io.o -
-
.1t.o -
eo -
T
oc
b
gi
Figure 7. Comparison of the damage area thresholding (light shaded bars) and average pixel history (dark shaded bars)
methods for determining threshold average temperature values in the in vitro retinal model. Threshold average
temperature values were added to ambient temperature in the heated enclosure to obtain the actual average
temperatures.
When we add the ambient temperature of the exposure enclosure (35 ºC) to each of the threshold average
temperature values we find that the average temperature that killed cells from exposures of 0.1 s and 0.25 s
was 46.7 ºC and 49.5 ºC, respectively. These are neither maximum nor average temperatures of the central
exposure site, which are often modeled or simulated with computer programs. Our threshold average
temperature values are the average temperature achieved by cells just reaching the critical thermal history
required for cell death.
By keeping track of frame numbers during laser exposure, we also averaged all the temperature values at
each frame number for a given laser exposure duration. This means that the average temperature values in
frame n from all exposures (for an exposure duration) were averaged together, regardless of damage size,
shape, or laser irradiance. Figure 8 provides thermal profiles over the course of each exposure duration
using the average pixel history method. Figure 8 also shows the unexpected difference between the 0.1-s
and 0.25-s exposures. Additionally, the 1-s profile data shows that the temperature of the cells at the
boundary of cell death came to equilibrium. We emphasize that the data represented by the average pixel
history method comes from exposures to a wide range of laser irradiances, and from both large and small
damage areas.
Threshold ITTP From Boundary
of Cell Death (sec)
46.7 C
o
49.5 C
o
47.7 C
o
0.10 0.25 1.00
Laser Exposure Duration (s)
Average Temperature at
Boundary of Cell Death
(
o
C)
Proc. of SPIE Vol. 7175 71750G-7
Figure 8. Average temperature values of cells at the boundary of cell death over the course of each laser exposure.
This data does not represent “threshold” average temperatures.
4. CONCLUSIONS
We have found that T
int
correlates with laser dose and exposure duration. The threshold average
temperature appears to serve as a good indicator of whether or not a lased cell will eventually die when
using our damage assessment scheme. The method is independent of absorption coefficient and laser
power density because the thermal response is measured, which is an outcome of both factors.
The data from the 0.1-s and 0-.25-s exposures would indicate that the damage rate processes do not follow
the Arrhenius model. It should be pointed out that the 0.1-s and 1.0-s data were all collected on the same
days (3 days back-to-back), whereas the 0.25-s data were collected over 2 days (1 week apart from each
other and 3 – 4 weeks prior to the 0.1-s and 1.0-s data). It has been noted in our prior publication
8
that the in
vitro retinal model is best suited for comparative analyses with minimal time between replicates. The best
comparisons are those collected together, like our 0.1-s and 1.0-s data, and thus, our confidence in
comparing these two data sets with the values in the 0.25-s data sets is not as high. Additionally, there were
signs in the viability images that the cells exposed for 0.25 s were not of the same quality as the other two
exposure durations.
Regardless of the issues associated with the 0.25-s data, the average peak temperatures at the boundary of
cell death for the 0.1-s and 1.0-s exposures (Fig.8) are not significantly different. Although this was
explained by the observation that the 1.0-s thermal profiles at the boundary came to apparent equilibrium,
the result was not expected. We have since determined that the average initial rate of heating for the 0.1-s
exposures was twice that of the 1.0-s exposures (data not shown). Likewise, a refined data analysis (data
not shown) has revealed that there is a statistically significant difference between the threshold T
ave
values
for the 0.1-s and 1.0-s data.
0
5
10
15
20
25
0 0.5 1 1.5
Time After Laser On (sec)
Average T
max
Average Temperature at the
Boundary of Cell Death
(
o
C)
Proc. of SPIE Vol. 7175 71750G-8
This new thermal metric for predicting cell death can be easily incorporated as an alternative end point
(rather than the damage integral) for computational modeling for laser-induced damage.
5. REFERENCES
[1] Henriques, F.C., "Studies of thermal injury. V. The predictability and the significance of thermally
induced rate processes leading to irreversible epidermal injury," Arch. Pathol. 43, 489-502 (1947).
[2] Moritz, A., F.C. Heriques, "Studies of thermal injury. II. The relative importance of time and surface
temperature in the causation of cutaneous burns," Am. J. Pathol. 23, 695-720 (1947).
[3] Jacques, S.L., "Ratio of entropy to enthalpy in thermal transitions in biological tissues," J. Biomed.
Opt. 11, 041108 (2006).
[4] Wright, N.T., S.S. Chen, J.D. Humphrey, "Time-temperature equivalence of heat-induced changes in
cells and proteins," ASME J. Biomech. Eng. 120, 22-26 (1998).
[5] Wright, N.T., "On a relationship between the Arrhenius parameters form thermal damage studies,"
ASME J. Biomech. Eng. 125, 300-304 (2003).
[6] Diller, K.R., J.A. Pearce, "Issues in modeling thermal alterations in tissues," Ann. N.Y. Acad. Sci.
888, 152-164 (1999).
[7] Gerstman, B.S., R.D. Glickman, "Activated rate processes and a specific biochemical mechanism for
explaining delayed laser induced thermal damage to the retina," J. Biomed. Opt. 4, 345-351 (1999).
[8] Denton, M.L., M.S. Foltz, K.J. Schuster, G.D. Noojin, L.E. Estlack, R.J. Thomas, "In vitro model that
approximates retinal damage threshold trends," J. Biomed. Opt. 13, 054014 (2008).
[9] Finney, D.J., [Probit Analysis], Cambridge University Press, New York (1971).
[10] Cain, C.P., G.D. Noojin, L. Manning, [A comparison of various Probit methods for analyzing yes/no
data on a log scale], USAF Technical Report AL/OE-TR-1996-0102 (1996).
Proc. of SPIE Vol. 7175 71750G-9