2.1.

LED Module Design and Assembly

The LED module is designed around an aluminum-substrate printed circuit board (PCB) to maximize heat flow away from the LEDs. This design choice constrains the selection of electronics to only surface-mounted components. Circuits were designed to connect four rows of four 690-nm LEDs (1 W Infrared LED, Shenzen Fedy Technology Co.) or six 635-nm LEDs (PLR3535AA000, Plessy Semiconductors) in parallel with 45 W, $10\mathrm{\Omega }$ current limiting resistors (TKH45P10R0FE-TR, Ohmite). A thermistor [B57452V5103J062, Epcos (TDK)] and a $499\mathrm{\Omega }$ resistor (RNCP0805FTD499R, Stackpole Electronics Inc.) circuit were also included, with the thermistor placed in the center of the LED array to enable real-time measurements of the board temperature. The full circuit required an 8-pin connector that was made from a Dupont connector kit (WYTP07-KIT, WayinTop). The complete circuit board was designed using electronic AutoCAD software (Eagle, Autodesk) and manufactured by a PCB fabrication service (PCBCart).

To assemble the board, the components were applied to the PCB via solder paste (EP256, Kester) and, after all the components were added, a hot air gun (898D Soldering Station, Vivohome) set to 400°C was used to solder each component in place. To do this, the PCB was supported with a third hand (01902, Neiko) and the hot air gun was directed at the PCB from 3 to 4 cm below. A digital multimeter (DMM, WH5000A, AstroAI) was used to confirm that successful connections were achieved. The PCB was then mated to a heat sink (M-B012, Cincon) with a thin layer of thermal paste (Protronix-PST-D Series 7, Protronix), and six drops of super glue (PR40, 3M), three on each side, were added along the sides of the PCB to secure it in place (Fig. 1).

Fig. 1

Custom-designed, high-power, modular LED array for PDT. (a) A 690-nm LED module aside a 3D-printed nozzle for air cooling. (b) A 635-nm LED module mounted on a 3D-printed brace with focusing optic above. A plexiglass surface with an aperture (not shown) is secured directly above the lens on which a well plate is placed for PDT experimentation.

2.2.

Supporting Electronics and Optomechanics

The LED circuit was connected to a 30 V/5 A power supply (HY3005F-3, Dr.meter) with an intermediate single channel 5 V relay (MK1LU5V6P, Ficbox) to enable programmable LED modulation via a USB DAQ card (USB-6001, National Instruments). The relay was wired to the 5 V power supply, digital ground, and digital I/O channel on the DAQ. The thermistor circuit was also wired to the 5 V DAQ power, with differential signal measured between analog input and ground channels. A subminiature version A (SMA) coaxial connector (8541674577, Maxmoral) was augmented with signal and ground wires in order to read the analog power meter signal through the DAQ from the SMA output on the power meter. The DAQ was connected to a laptop placed next to the experimental setup via USB to USB-c cable (National Instruments). A complete wiring diagram is provided in Fig. S1 in the Supplementary Material.

The LED module was placed on the surface of an optics table via a custom 3D-printed brace (Ender 5, Comgrow) [Fig. 1(b), Fig. S2(a) in the Supplementary Material] that secures the module to a vertical cage mounting system (CPVM, Thorlabs). Alternatively, an aluminum breadboard (MB12, Thorlabs) may be used for a mobile-cart design or if an optics table is not available. A diffuser-lens pair was placed above the LED module using a 2-in. cage mount (LCP01, Thorlabs) to focus and smooth the light field. The choice of the focusing lens was determined empirically between a Fresnel lens (FRP232, Thorlabs) and an aspheric condenser lens (ACL5040U-DG6-A, Thorlabs), with a 600-grit diffuser (DG20-600, Thorlabs) mounted just below each optic. A power meter sensor (S130C, Thorlabs), connected to a compatible power meter (PM100D, Thorlabs), was placed above, but off-center to, the mounted optic. The lens was then adjusted along the $z$ axis to maximize the power transmission and determine the height of maximum collimation (i.e., the focal point of the lens). Both lenses allowed similar power transmission, but the Fresnel lens was found to produce a more uniform light field with a larger spot size than the condenser lens and was therefore chosen for the ensuing experiments.

A platform on which to place a well plate for typical in vitro PDT experimentation was fashioned from a transparent plexiglass board (SL-AS6, $12\times 12\times 1/4\text{-}\mathrm{in.}$, SimbaLux). The board was covered in light-blocking tape (T205-1.0, Thorlabs), except for a $3.85\times 3.85\text{-}\mathrm{cm}$ opening in the center. Two $3/8\text{-}\mathrm{in.}$ holes were drilled on the side of the board and secured to two $4\text{-}\mathrm{in.}$ optical posts (TR4, Thorlabs) mounted (UPH1, Thorlabs) to the optics table using $5/8\text{-}\mathrm{in.}$ screws (SH25S063, Thorlabs). This fixed the plate height at 4 in. above the table and $x-y$ adjustments were made to align the aperture above the lens.

Since cooling of the module is paramount, we used a compressed air line to provide active cooling of the LED module. A fan could be used in place of the compressed air. Approximately 1 m of $3/8\text{-}\mathrm{in.}$ tubing was connected to the air outlet valve on one end and to a 3D-printed nozzle on the other [Fig. 1, Fig. S2(b) in the Supplementary Material]. The nozzle was designed to match the cross-section of the heat sink and provide uniform airflow through the fins. It was secured to the table via two 2-in. optics posts (TR2, Thorlabs) combined with a right-angle clamp (RA90, Thorlabs) and a fixed position lens mount (NRC MH-2P, Newport) such that the air was directed through the heat sink fins.

2.3.

Software

We employed MATLAB’s (R2019b, MathWorks) DAQ toolbox to interface with the DAQ via custom software application, ExpressPDT, written with MATLAB’s application programming software App Designer (Fig. S3 in the Supplementary Material). Note that compatibility of the DAQ toolbox restricts full operation of the software (i.e., connection to the DAQ and relay control) to Windows operating systems. However, ExpressPDT may still be used as a PDT planning tool on any operating system.

2.4.

Thermistor Model and Calibration

A thermistor is a resistor whose resistance changes predictably with temperature. We employed this device in a simple voltage divider circuit to measure the on-board temperature of the LEDs in real time. For negative temperature coefficient thermistors, the resistance $R$ varies exponentially with temperature $T$ according to

The parameters $\beta$ and $R_{\infty }$ were determined experimentally for each LED module by measuring the thermistor voltage across a range of temperatures. First, approximately 0.2 mL of thermal paste was placed on the thermistor, and the module was refrigerated at 4°C for 2 h. The module was then reconnected to the power supply and a thermocouple connected to the DMM was inserted into the thermal paste to determine the temperature of the thermistor. Using ExpressPDT’s calibration mode, the temperature and thermistor signal were recorded simultaneously as the module returned to room temperature. The LEDs were then turned on at low power to facilitate further warming of the module at approximately 1°C every 5 to 10 s. The LEDs were turned off and recording ceased once the module reached 65°C; 40 to 50 data points were collected in total for each module. Once the calibration procedure was complete, the thermal paste was cleaned from the LED module using cotton swabs and optic wipes dampened with isopropyl alcohol. Data were linearized and fit to Eq. (3) using Prism 8 (GraphPad); best-fit values $\beta$ and $R_{\infty }$ are reported with their standard errors (SE). These values were programmed into ExpressPDT for automated temperature monitoring using Eq. (5).

2.5.

Module Characterization

The emission spectrum of each LED module was measured with a spectrometer (Amadeus AMA01338, Ocean Optics) at room temperature (21°C), and then again at 38°C to characterize the effect of temperature on the spectral emission. The LED supply voltage was then adjusted so that the power meter read $\sim 100{\mathrm{mW}/\mathrm{cm}}^2$ peak power at the plate surface. Once the temperature stabilized, the temperature and power were recorded for 30 min at 1-s intervals using ExpressPDT. To assess the power delivered to each well of a cell culture 24-well plate, the module was turned on and allowed to stabilize at 39°C. The power at the plate surface in each quadrant of the aperture was measured thrice with a power meter to determine the intensity given to each well in $2\times 2$-well experimental group. Average temperature and intensity are reported as $\mathrm{mean}\pm \mathrm{standard}\mathrm{deviation}$ (SD) with the coefficient of variation (CV, defined as SD divided by mean) provided where useful.

2.6.

Cell Culture and PDT

Human primary high-grade serous ovarian cancer line (Powder, Cellaria Biosciences) was cultured in T75 Flasks (1256685, Thermo Scientific) according to a protocol recommended by Cellaria Biosciences in a humidified incubator at 5% ${\mathrm{CO}}_2$ and 37°C. Powder cells were cultured in Renaissance essential tumor medium (RETM) and RETM supplement (CM-0001, Cellaria Biosciences) completed with 6.3% heat-inactivated fetal bovine serum (FBS, SH30071.03HI, Hyclone™ GE Healthcare Life Sciences) and 1% penicillin/streptomycin (BP295950, Fisher BioReagents).

Before plating, RETM was prepared by diluting stock media to 3% FBS and was used throughout the experiment. During passaging, cells were lifted with 0.25% trypsin ethylenediaminetetraacetic acid (25053CI, Corning), washed in phosphate-buffered saline (PBS, 70011069, Gibco), and suspended at $\mathrm{20,000}\text{cells}/\mathrm{m}\mathrm{L}$ in RETM at 3% serum. One mL of cell solution was added to each well of a black-walled, 24-well plate (P241.5HN, Cellvis) and allowed to grow for 3 days. PS was administered at 0.5 mM ALA (A3785, Sigma Aldrich) or $0.1\mu \mathrm{M}$ verteporfin (Visudyne^®, QLT Phototherapeutics, Inc.) in media and incubated for 4.5 or 2 h, respectively. Just before illumination, all wells were aspirated and replaced with fresh media.

To increase protocol efficiency, experimental groups were organized into $2\times 2$-well groups (four biological replicates) to be illuminated simultaneously. Six treatment groups in each 24-well plate included three controls—no $\mathrm{PS}+0{\mathrm{J}/\mathrm{cm}}^2$, no $\mathrm{PS}+50{\mathrm{J}/\mathrm{cm}}^2$, and $\mathrm{PS}+0{\mathrm{J}/\mathrm{cm}}^2$—and three treatment groups—$\mathrm{PS}+10$, 20, or $50{\mathrm{J}/\mathrm{cm}}^2$. For ALA-PDT, the average irradiance of the 635-nm module in each quadrant was $86.6{\mathrm{mW}/\mathrm{cm}}^2$ at a stable module temperature of 39°C, with a total plate illumination time of 25 min. For BPD-PDT, the average irradiance of the 690-nm module in each quadrant was $79.7{\mathrm{mW}/\mathrm{cm}}^2$ at 39°C, with a total plate illumination time of 27 min and 10 s. Laser safety goggles (100-38-245, Laser Safety Industries) with optical density (OD) 2+ at $>630\mathrm{nm}$ were worn during illumination. All work with PSs was done in subdued light and plates were protected from light with aluminum foil except during PDT.

Cell culture viability was assessed with fluorescent live/dead staining 24 h after light treatment using flow cytometry (FC) and validated with confocal microscopy. Three out of the four wells from each group were collected and stained with live/dead fixable green (L-34970, Life Technologies) for 30 min at 4°C protected from light. Cells in suspension were included in this analysis by collecting the supernatant before lifting the cells. After staining, samples were washed a further 2 times in PBS, resuspended in $300\mu \mathrm{L}$ PBS, and immediately analyzed using a flow cytometer (Attune NxT, ThermoFisher). Both the concentration and viability of cells were measured by FC; relative viability is reported here as the $\text{average}\pm \mathrm{SE}$ concentration of viable cells normalized to the no treatment control.

Immediately after three replicates from each treatment group were collected from the plate for FC analysis, the fourth well was washed with PBS (discarding the supernatant) and stained with a 1:60 dilution of acridine orange (AO)/propidium iodide (PI) (F23001, Logos Biosystems) in media and immediately imaged using a laser-scanning confocal microscope (LSM 800, Zeiss). Images are considered as a strictly qualitative assessment of viability to validate quantitative FC results.

3.1.

Thermistor Calibration

The thermistors on each module demonstrated ideal behavior across the range of operating temperatures (20°C to 40°C, Fig. 2). Temperature and voltage were fit to a linearized model [Eq. (3)] with slope $\beta$ and intercept $R_{\infty }$ (Fig. S4 in the Supplementary Material). For the 635-nm array, we found $\beta =4442\pm 16\mathrm{K}$ and $R_{\infty }=3.34\pm 0.17\mathrm{m}\mathrm{\Omega }$. For the 690-nm array, we determined $\beta =4279\pm 7\mathrm{K}$ and $R_{\infty }=5.37\pm 0.13\mathrm{m}\mathrm{\Omega }$, in agreement with manufacturer specified values. These data were then programmed into ExpressPDT software using Eq. (5) to enable real-time module temperature monitoring during treatment.

Fig. 2

Thermistor calibration. Temperature and voltage measurements (symbols) were used to find the best-fit values for Eq. (5) (lines). For clarity, the 635-nm array data set is shifted horizontally by $+0.2\mathrm{V}$.

3.2.

LED Module Characterization

At 38°C, the LED emission underwent slight redshifting and loss of intensity compared to room temperature (Fig. 3). The 635-nm spectrum redshift was less than the spectrometer resolution ($<2\mathrm{nm}$), and the relative intensity was 94% of the room temperature spectrum. The 690-nm LED fidelity was slightly more impacted at 38°C with a redshift of 4 nm at 86% of room temperature power. These shifts were not enough to compromise the PS excitation efficiency.

Fig. 3

LED array spectral emission (left, $y$ axis) at room temperature (21°C) and operating temperature (38°C) align with their respective PS absorption spectra (right, $y$ axis). Relative intensities are preserved between low- and high-temperature spectra after normalization.

Thermal stability was assessed over a 30-min trial by measuring the temperature and peak power of the LED array at 1-s intervals. Both the 635- and 690-nm boards displayed $<1\%$ variation in both variables over that time frame (Fig. 4). Specifically, the 635-nm array was stable at $32.0\pm 0.1°\mathrm{C}$ ($\mathrm{CV}=0.42\%$) with an output of $108.2\pm 0.3{\mathrm{mW}/\mathrm{cm}}^2$ ($\mathrm{CV}=0.29\%$). Similarly, the 690-nm array was stable at $38.7\pm 0.3°\mathrm{C}$ ($\mathrm{CV}=0.66\%$) with an output of $102.8\pm 0.2{\mathrm{mW}/\mathrm{cm}}^2$ ($\mathrm{CV}=0.24\%$).

Fig. 4

LED stability. Both LED modules are stable at operating temperature and demonstrated $<1\%$ variation in on-board temperature (left, $y$ axis) and peak power (right, $y$ axis) over a 30-min illumination (measurements taken at 1-s interval). Peak power was measured along the optical axis at the well-plate surface.

The light field at the plate surface was analyzed for uniformity by acquiring measurements centered at the positions of each of the four wells being treated simultaneously (Fig. 5). It was determined that the variations in the power delivered to each well were 2.3% for both arrays. At operating temperature (39°C), the intensity was $86.6\pm 2.0{\mathrm{mW}/\mathrm{cm}}^2$ ($\mathrm{CV}=2.3\%$) for the 635-nm array and $79.7\pm 1.8{\mathrm{mW}/\mathrm{cm}}^2$ ($\mathrm{CV}=2.3\%$) for the 690-nm array. Individual well measurements were further analyzed to check for inhomogeneities in the light field. It was determined that for the 635-nm module, the power delivered to well A2 was $3.9{\mathrm{mW}/\mathrm{cm}}^2$ (4.3%) larger than to A1 ($p=0.0207$) and B2 ($p=0.0207$) but was no different from B1 ($p=0.2594$), indicating a minute nonuniformity in the photon flux through the aperture. No differences in light delivery from 690-nm module were detected ($p=0.3081$).

Fig. 5

Light field assessment of the (a) 635-nm and (b) 690-nm LED modules. Intensity was measured in each quadrant of the $3.85\times 3.85\text{-}\mathrm{cm}$ aperture to estimate the power delivered to each well of a $2\times 2$-well experimental group (A1 and B2). Circles represent approximate location and area of the power meter sensor within each quadrant of the aperture. Results are $\text{mean}\pm \mathrm{SD}$ of 3 measurements. One-way ANOVA of the four quadrant measurements was significant for the 635-nm array ($p=0.0142$) and not significant for the 690-nm array ($p=0.3081$). Follow-up analysis with Tukey’s multiple comparisons test on the 635-nm array group revealed the intensity in well A2 was significantly larger than A1 ($p=0.0207$) and B2 ($p=0.0207$), but not different from B1 ($p=0.2594$). Reported $p$-values are adjusted for multiple comparisons.

3.3.

PDT

A primary ovarian cancer cell line was successfully treated with ALA- and BPD-PDT using the 635- and 690-nm modular LED arrays, respectively. No statistically significant effects were observed across control groups in both cases, but a clear, light-dependent decrease in cell viability was observed via fluorescence microscopy and quantified with live/dead FC analysis (Fig. 6). Compared to the no-light, no-PS control, $10{\mathrm{J}/\mathrm{cm}}^2$ was sufficient to eliminate $>50\%$ of cells using each PS. At the maximum $50{\mathrm{J}/\mathrm{cm}}^2$ dose, cell viability was $0.52\pm 0.07\%$ ($p<0.0001$) after ALA-PDT and $0.07\pm 0.00\%$ ($p<0.0001$) after BPD-PDT. Confocal microscopy of treated cells with AO/PI staining is qualitatively consistent with the dose-response trend. Full images are provided in Fig. S5 in the Supplementary Material.

Fig. 6

Viability of powder cells 24-h post-LED-PDT. (a) AO/PI staining shows PDT-induced cell death. Full images are provided in Fig. S5 in the Supplementary Material. Scale bar: $200\mu \mathrm{m}$. (b) $\text{Relative viability}\pm \mathrm{SE}$ assesed by live/dead FC staining ($n=3$ biological replicates) confirms a light dose-dependant response to treatment. One-way ANOVA for each experiment was significant ($p<0.0001$) and Dunnett’s multiple comparisons test confirmed siginficant differences compared to the control group. ^†$p=0.0025$, ^‡$p=0.0002$, and ^§$p<0.0001$. $p$-values are adjusted for multiple comparisons.