Microfluidic photoreactor to treat neonatal jaundice

Abstract

While in most cases, jaundice can be effectively treated using phototherapy, severe cases require exchange transfusion, a relatively risky procedure in which the neonate's bilirubin-rich blood is replaced with donor blood. Here, we examine extracorporeal blood treatment in a microfluidic photoreactor as an alternative to exchange transfusion. This new treatment approach relies on the same principle as phototherapy but leverages microfluidics to speed up bilirubin removal. Our results demonstrate that high-intensity light at 470 nm can be used to rapidly reduce bilirubin levels without causing appreciable damage to DNA in blood cells. Light at 470 nm was more effective than light at 505 nm. Studies in Gunn rats show that photoreactor treatment for 4 h significantly reduces bilirubin levels, similar to the bilirubin reduction observed for exchange transfusion and on a similar time scale. Predictions for human neonates demonstrate that this new treatment approach is expected to exceed the performance of exchange transfusion using a low blood flow rate and priming volume, which will facilitate vascular access and improve safety.

INTRODUCTION

Virtually all newborns develop hyperbilirubinemia, defined as a serum total bilirubin (TB; comprised of unconjugated and conjugated bilirubin) exceeding 1 mg/dl. Neonatal jaundice is a subset of newborn hyperbilirubinemia where the TB level manifests as jaundice and is very common, affecting approximately 60% of newborns.¹ Approximately, 5% of newborns experience significant hyperbilirubinemia (TB > 95th percentile).²

Because of its lipophilicity, unconjugated bilirubin can cross the blood–brain barrier and accumulate in the tissue of the central nervous system, where it can cause neurologic damage, which manifests as bilirubin-induced neurologic dysfunction (BIND). BIND takes two forms: acute bilirubin encephalopathy with a possibility of reversible damage and chronic bilirubin encephalopathy (formerly referred to as kernicterus), with symptoms of motor function disorders, auditory dysfunction, seizures, and, in some cases, death.³ Risk of neurologic damage is related to the TB level, rapidity of rise, and time of exposure. In the USA, extreme (>25 mg/dl) and hazardous (>30 mg/dl) hyperbilirubinemia occur in approximately 0.14% and 0.01% of infants, respectively,^4,5 while globally the incidence of extreme and hazardous hyperbilirubinemia is much higher.⁶ This places neonates at an extremely high risk of BIND. As a consequence, at least one determination of the TB level is very common in the first few days of newborn life.

The most common treatment for significant neonatal hyperbilirubinemia is whole body phototherapy in which unconjugated bilirubin undergoes photoreaction to form isomers, predominantly lumirubin, that are more readily excreted from the body.^1,7 This treatment is sufficient to control or decrease TB in most cases of significant neonatal hyperbilirubinemia and is currently the best practice in terms of safety and simplicity. However, phototherapy decreases bilirubin levels relatively slowly and prolonged exposure to high bilirubin levels increases risk of BIND. Therefore, for extreme cases of hyperbilirubinemia that do not respond to phototherapy and for hazardous hyperbilirubinemia, exchange transfusion is indicated to prevent BIND.⁸ An exchange transfusion is a relatively risky procedure in which the infant's bilirubin-rich total blood volume is exchanged twofold (double volume exchange) with donor blood in small increments.⁹ Exchange transfusion is a complex and labor-intensive process. In the USA, it is typically only available in higher level neonatal intensive care units, sometimes necessitating transport to receive this treatment. Moreover, because of intensive screening for neonatal hyperbilirubinemia to avoid BIND, the need for exchange transfusion is rare, even at these centers. Globally, exchange transfusions are much more routinely performed, especially in lower and middle income countries where the rate of exchange transfusion can exceed 1 per 100 live births.⁶ Thus, there is a need for a safer and more convenient alternative to exchange transfusion that can be quickly administered in response to severe cases of neonatal jaundice.

To address this need, several research groups have explored the use of extracorporeal circuits for the removal of bilirubin from blood.^10–15 These extracorporeal designs have a wide range of functionality, including enzymatic conversion of bilirubin to less toxic products and adsorption of bilirubin to a resin. While none of these treatments have been put into clinical practice with neonates, several treatments have been tested on animals, demonstrating their ability to reduce bilirubin levels in vivo.^13–15 However, there are drawbacks with these previous approaches. Previous studies of bilirubin removal by adsorption to a resin have used columns with internal volumes exceeding 25% of the animal's blood volume, and the resin not only adsorbed bilirubin but also blood components (e.g., platelets) and hormones (e.g., cortisol).^14,15 Another device used the enzyme bilirubin oxidase to convert bilirubin to less toxic products, but one of the products—biliverdin—can be converted back to bilirubin in the body, and the other products are not well characterized and their toxicity is not known.¹³

In this study, we explore the potential for using a microfluidic photoreactor for the treatment of extreme or hazardous neonatal hyperbilirubinemia. The photoreaction we propose is the same as that carried out by phototherapy lights but instead targets the bilirubin in the blood directly for greater efficiency. This novel approach may offer improved safety compared to previous approaches^10–15 by enabling the use of an extracorporeal circuit with a reduced blood flow rate and priming volume. To examine the potential of this approach, we built a laboratory scale microfluidic photoreactor and carried out in vitro experiments with human blood and in vivo experiments in Gunn rats to examine the effects of the microchannel thickness, light wavelength, and light intensity on photochemical conversion of bilirubin, as well as damage to blood cells. We also developed a mathematical model of bilirubin conversion and used it to predict the performance of a clinical scale photoreactor for treating neonatal hyperbilirubinemia.

MATERIALS AND METHODS

Experimental setup for in vitro studies

The experimental test loop shown in Fig. 1 was used for in vitro experiments. Blood was circulated from a reservoir through the photoreactor and then back to the reservoir using a peristaltic pump. The blood in the reservoir was gently mixed by periodically moving a magnetic stir bar up and down by hand. Before each experiment, the reactor was first rinsed with a 5% pluronic solution for 1 h at a flow rate of 0.5 ml/min, followed by a 20 min rinse with isotonic saline.

FIG. 1.

Experimental setup for in vitro assessment of microfluidic photoreactor. (a) Expanded view of the photoreactor components. (b) Picture of the assembled photoreactor. (c) Schematic of the in vitro setup showing recirculation of blood from the reservoir, through the photoreactor and back to the reservoir.

A series of experiments were performed to examine photoreactor performance under different conditions. To examine the effects of light intensity and wavelength, blood from a 5 ml reservoir was circulated at 0.3 ml/min through a 250 μm thick reactor for 2 h using various light intensities at two different wavelengths: 470 and 505 nm. To examine the effect of channel height, the bilirubin concentration in the blood reservoir was measured after circulating blood for 2 h through photoreactors with channel heights of 125, 250, and 500 μm using a wavelength of 505 nm and a light intensity of 125 mW/cm². The blood flow rate for each channel height was adjusted to keep a consistent reactor residence time of 1 min, and the volume of the blood reservoir was scaled by the blood flow rate to maintain a constant ratio of flow rate to reservoir volume. Specifically, the blood flow rates were 0.15, 0.3, and 0.6 ml/min and the reservoir volumes were 2.5, 5, and 10 ml for channel heights of 125, 250, and 500 μm, respectively.

The blood flow rates and reservoir volumes were adjusted in this way for the following reasons. For clinical treatment, blood flow rate and device volume are both important parameters that affect safety and ease of vascular access. Therefore, the best way to compare channel heights is to use the same blood flow rate to evaluate devices with the same volume but different channel heights. This results in the same residence time (defined as the ratio of device volume to flow rate) for all channel heights. However, in our experiments, it was necessary to hold the channel width and length constant so that the size of the reactor channel matched the size of the LED array. Therefore, it was not possible to change channel height without also changing device volume. To address this, the blood flow rate was adjusted for devices with different channel heights to keep residence time constant. To ensure fair comparison between devices with different channel heights, it was also necessary to adjust the reservoir volume to keep the ratio of blood flow rate to reservoir volume constant; as shown in Eq. (4) below, the rate at which the bilirubin concentration changes in the reservoir is proportional to this ratio. After adjusting the blood flow rate and the reservoir volume, the resulting bilirubin concentration changes are expected to be equivalent to those for the corresponding experiments with constant device volume, blood flow rate, and reservoir volume. This can be illustrated by comparing the experimental conditions for devices with channel heights of 250 and 125 μm. Reducing the channel height to 125 μm reduces the device volume by twofold, and we also reduced both blood flow rate and reservoir volume by twofold. This is equivalent to the corresponding experiment with constant device volume, flow rate, and reservoir volume, but it is scaled down by a factor of two.

The results of these experiments were used to develop a mathematical model of bilirubin conversion (see below), which was validated with an additional experiment using a wavelength of 505 nm, a light intensity of 125 mW/cm², and a channel height of 250 μm, with double the reservoir volume compared to previous experiments and a longer treatment period of 8 h instead of 2 h.

Experiments were also performed to examine the potential for light-induced damage to blood cells. These experiments involved the circulation of blood at 0.3 ml/min through a 250 μm thick photoreactor channel for a prescribed duration, followed by assessment of DNA damage (as described below). Two wavelengths of light, 470 and 505 nm, at two different levels of intensity, 25 and 125 mW/cm², were examined. The experiments were designed to deliver a total light dose that is similar to conventional phototherapy (∼1 mW/cm² over ∼5 days),¹ but using a higher light intensity over a shorter time period. For experiments using the lower light intensity setting of 25 mW/cm², the total treatment time was 5 h. For the higher light intensity setting (125 mW/cm²), the treatment time was 1 h, which corresponds to the same total light dose.

Microfluidic photoreactor for in vitro studies

The device contained two channels, one for perfusion of blood and the other for perfusion of a cooling fluid from a refrigerated water bath (Fig. 1). These channels were created by compressing laser-cut Kapton (polyimide) sheets between two parallel plates of 0.25 in. acrylic (PMMA). The geometries of the reactor and cooling channels were both constant at 1.2 cm wide × 10 cm long, with 1.75 cm headers at either end. Devices with reactor heights of 125, 250, and 500 μm were created by using Kapton sheets with thicknesses of 125, 250, and 500 μm. The cooling channel thickness was held constant at 250 μm. The reactor channel and cooling channel were separated by a 0.5 mm piece of polycarbonate. The entire reactor assembly was compressed tightly using 22 bolts spaced 14.4 mm on center, tightened to 80 cN m with a torque wrench. Fluid inlets and outlets were provided via quick disconnect threaded Luer lock connectors fed directly from the top and bottom of the reactor.

A light emitting diode (LED) array was situated on top of the photoreactor housing. Identical arrays were created for the two different wavelengths to be tested, 470 and 505 nm. The 505 nm LEDs and 470 nm LEDs (OSRAM Opto Semiconductors, models LV CK7P-JYKZ-25-Z and LB CPDP-GZHX-35, respectively) had spectral bandwidths of 30 and 25 nm, respectively. A customized PCB carrier board was designed to generate uniform light output over the target region, and provide heat dissipation for the LEDs. The LEDs were arranged in 14 banks of 8 LEDs, resulting in a total of 112 LEDs per LED array. Each bank of LEDs is controlled by a high voltage constant current driver, and the 14 LED drivers are connected in parallel to an external 48 V power supply as well as a pulse-width modulation dimming controller.

Preparation of bilirubin-spiked blood

To create a concentrated bilirubin stock solution, bilirubin powder (Sigma-Aldrich, St. Louis, MO, USA) was dissolved at a concentration of 2 g/l in 0.2M sodium hydroxide solution. This stock solution was added to whole human blood to create a working bilirubin concentration of about 0.4 g/l plasma (40 mg/dl). This concentration is consistent with the bilirubin levels in neonates with hazardous hyperbilirubinemia.⁵ The appropriate volume of 2M hydrochloric acid was also added to the blood in order to neutralize the sodium hydroxide from the bilirubin stock solution. Blood was obtained via venipuncture from healthy volunteer donors into 4.5 ml 3.2% citrate vacutainer collection tubes using a protocol that had been approved by the Oregon State University Institutional Review Board. Blood draws were carried out in accordance with relevant guidelines and regulations, and informed consent was obtained from all blood donors.

Measurement of bilirubin concentration

For measurement of bilirubin concentration, the blood was centrifuged to isolate the plasma, and the plasma was diluted 50-fold in isotonic phosphate buffered saline (pH 7.2) containing 4% w/v bovine serum albumin (MP Biomedical, Santa Ana, CA, USA). The bilirubin concentration in this solution was determined using spectrophotometry, as described below.

Bilirubin has an absorbance maximum at 468 nm in the presence of bovine serum albumin,¹⁶ and conversion of bilirubin to photoproducts is expected to result in a decrease in absorbance at this wavelength. However, the bilirubin concentration could not be determined directly by measuring the absorbance at 468 nm because of interference caused by the presence of hemoglobin. Hemoglobin has an absorbance maximum at 415 nm.¹⁷ Thus, to estimate bilirubin concentration, we used absorbance measurements at both 415 and 468 nm. Bilirubin's photoproducts may also absorb light at these wavelengths. When exposed to light bilirubin undergoes configurational and structural isomerization, as well as oxidation reactions. Bilirubin's configurational isomers and normal bilirubin have similar absorption spectra, so the formation of configurational isomers is not expected to cause appreciable changes in absorbance. However, formation of the structural isomer lumirubin and oxidation products is expected to decrease the absorbance at 468 nm. For the purposes of estimating bilirubin concentration, we treated the mixed photoproducts (consisting of lumirubin and oxidation products) as a single species with average absorption properties. The implications of this assumption are discussed in the supplementary material.

Therefore, to determine the bilirubin concentration, we needed to account for the contributions of bilirubin, hemoglobin, and bilirubin's photoproducts to the absorbance values at 415 and 468 nm. Separate equations can be defined for the absorbance at 415 nm (A₄₁₅) and 468 nm (A₄₆₈) in terms of the concentrations of bilirubin (C_B), hemoglobin (C_H), and photoproducts (C_P),

$$A_{415}={\epsilon }_{415}^B{C}_B+{\epsilon }_{415}^H{C}_H+{\epsilon }_{415}^P{C}_P,$$ (1)

$$A_{468}={\epsilon }_{468}^B{C}_B+{\epsilon }_{468}^H{C}_H+{\epsilon }_{468}^P{C}_P,$$ (2)

where $(\epsilon )$ is the extinction coefficient, the subscripts 415 and 468 refer to the light wavelengths (in nm) and the superscripts B, H, and P refer to bilirubin, hemoglobin, and photoproducts, respectively. The initial concentration of photoproducts was assumed to be zero at the start of the experiment. This initially allows for Eqs. (1) and (2) to be solved simultaneously for the concentrations of bilirubin and hemoglobin. After exposure to light in the photoreactor, we expect photoproducts to be produced, which introduces an additional unknown. To address this, we can use the following equation:

$$C_{B,0}=C_B+C_P,$$ (3)

where $C_{B,0}$ is the bilirubin concentration before photoreactor treatment. After photoreactor treatment, the concentrations of bilirubin, photoproducts, and hemoglobin can be determined by simultaneously solving Eqs. (1)–(3).

The extinction coefficients for hemoglobin (assumed to be fully oxygenated) and bilirubin at these wavelengths were obtained from the literature.^17,18 To estimate the extinction coefficients for bilirubin's photoproducts, we performed experiments in which a known amount of bilirubin was added to plasma, and this plasma was exposed to light in the photoreactor until the absorbance at 468 nm no longer changed. At this point, it was assumed that all bilirubin had been converted to photoproducts. The extinction coefficients were then estimated from linear fits to plots of absorbance vs photoproduct concentration. Table I provides extinction coefficient values.

TABLE I.

Extinction coefficient of bilirubin, hemoglobin, and bilirubin's photoproducts (cm²/mg).

Light wavelength	Bilirubin	Hemoglobin	Photoproducts
415 nm	48	8.1	44
468 nm	110	0.54	30

Measurement of DNA damage

A Trevigen CometAssay Kit (Gaithersburg, MD) was used to determine the effects of light exposure on DNA damage in peripheral blood mononuclear cells (PBMCs). DNA damage was quantified as the percent increase in the DNA “comet” tail compared to a control sample of freshly drawn blood that was set aside for the duration of the experiment. PBMCs were isolated by placing 3 ml of whole blood onto 3 ml of Histopaque 1077 in a 15 ml centrifuge tube, followed by centrifugation at 800 g for 30 min. The resulting PBMC layer was then removed and diluted 1:50 with phosphate buffered saline (PBS). A portion of this cell suspension was used for a positive control by treating the cells with 100 μM H₂O₂ for 10 min, followed immediately by centrifugation and resuspension in PBS to prevent further damage to the cells.

The comet assay was carried out according to the instructions provided in the Trevigen kit with minor modifications. Agarose at 37 °C was mixed at a 2:1 ratio with PBMC solution, and 53 μl of the resulting solution was pipeted onto the Trevigen Comet Slide sample area. Slides were stored in the dark at 4 °C for 10 min, and then immersed in lysis solution and DMSO in a 10:1 ratio and incubated for an additional 24 h at 4 °C. The slides were then immersed in 200 mM NaOH unwinding solution for 30 min. The resulting slides were then subjected to electrophoresis. The gel rig was filled with 4 °C unwinding solution, and electrophoresis was run for 40 min at 15 V (1 V/cm). Upon completion the slides were dipped twice in de-ionized (DI) water and then once in 70% ethanol, and left to dry for 15 min. The slides were then stained with SYBR green for 30 min, dipped once more in DI water and left to dry completely. The slides were then imaged using a GFP filter on a microscope at 100×. Images were taken at five random locations per treatment. Each image was analyzed with the CometScore software to determine the percentage of DNA in the tail for each cell in the image.

Microfluidic photoreactor for in vivo studies

The photoreactor used for in vivo experiments is shown in Fig. 2. Two arrays of 512 LEDs were arranged to project light from opposite sides of a removable cartridge containing the cooling and blood channels. The cooling and blood channels were both 250 μm thick, with a width of 2 cm and a length of 16 cm. We used the same 470 nm LEDs that were used for the smaller scale in vitro device. The cartridge was cooled using water that was constantly pumped through the device and circulated through a radiator to remove heat. The LED arrays had directly mounted heatsinks where the thermal load was dissipated. A digital controller supported by a touch screen interface allowed for the adjustment of the LED intensity as well as monitoring the temperature of the arrays via thermocouples.

FIG. 2.

Experimental setup for in vivo assessment of microfluidic photoreactor. (a) Schematic of the experimental setup. (b) Picture of the experimental setup. See Fig. S2 in the supplementary material for a close up picture of the photoreactor cartridge and the LED arrays.

Gunn rat study

The performance of the photoreactor was tested in vivo using Gunn rats as a model for hyperbilirubinemia.^19,20 The study was carried out at Absorption Systems (San Diego, CA, USA) in accordance with relevant guidelines and regulations, including the ARRIVE guidelines. The study protocol was approved by their Institutional Animal Care and Use Committee (IACUC). A schematic and picture of the experimental setup can be seen in Fig. 2. Male Gunn rats were used at an age of 14–17 weeks and weight of approximately 240 g. The animals were purchased from the Rat Resource and Research Center (Columbia, MO).

The Gunn rats were subjected to three different treatments, each of which was carried out for 4 h: (1) sham, in which blood was pumped through the photoreactor with the LED lights off, (2) photoreactor, in which blood was pumped through the photoreactor with the LED lights on, and (3) double volume exchange transfusion. For the sham and photoreactor treatments, blood was circulated through the photoreactor at 0.4 ml/min using a peristaltic pump. For the photoreactor treatment, the LED panels delivered an incident light intensity of 125 mW/cm². The double volume exchange transfusion was carried out by slowly infusing ∼30 ml of heparinized donor blood into the Gunn rat over 4 h using a syringe pump while withdrawing blood at the same rate. The volume of donor blood was adjusted for each Gunn rat based on its weight to deliver twice the rat's expected blood volume.²¹ Donor blood was collected from 12-week-old male Wistar rats on the day of the experiment. The Wistar rats were anesthetized using ketamine/xylazine or isoflurane prior to exsanguination using either cardiac puncture or carotid artery cannulation, after which the rats were euthanized while under anesthesia by cervical dislocation.

To prepare the Gunn rats for experiments, animals were first anesthetized using an intraperitoneal cocktail of ketamine (80 mg/kg) and xylazine (12 mg/kg). Cannulation of the animal was performed through the neck, using either both jugular veins, or the right jugular vein and left carotid artery. To prevent coagulation during the experiment, heparin (1000 U/ml) was administered via the tail vein. Approximately, 1 ml of blood was collected from the cannulated vein before initiating treatment and at the final time point of the experiment and analyzed for total bilirubin concentration. At the end of the experiment, the animals were euthanized by cervical dislocation while under anesthesia.

Six rats were allocated for each of the treatment groups. However, three rats from the sham group died while establishing anticoagulation and extracorporeal perfusion procedures, and an additional two rats in the exchange transfusion group died due to procedural complications. The final number of rats in each treatment group was three for the sham treatment, six for the photoreactor treatment, and four for the exchange transfusion treatment.

Mathematical model of bilirubin conversion

A mathematical model was developed to evaluate the conversion of bilirubin to photoproducts within the photoreactor, as well as the resulting changes in bilirubin concentration within the blood reservoir over time. A material balance on the blood reservoir results in

$${\displaystyle \frac{d{C}_b}{dt}={\displaystyle \frac{\dot{V}}{V}(C_{b2}-C_b),}}$$ (4)

where $C_b$ is the concentration of bilirubin in the blood reservoir, $C_{b2}$ is the bilirubin concentration exiting the photoreactor and returning to the reservoir, $\dot{V}$ is the blood flow rate through the photoreactor, and V is the reservoir volume. This equation shows that to determine the bilirubin concentration in the reservoir, we need a model for bilirubin conversion in the photoreactor that allows the prediction of the bilirubin concentration at the reactor outlet ( $C_{b2}$).

The quantum yield, Y, is a parameter that defines the number of bilirubin-to-photoproduct reactions that occur per photon of light absorbed by bilirubin. Thus, to determine the bilirubin conversion rate, we first need to determine the amount of light absorbed by bilirubin. To estimate the amount of light absorbed by bilirubin, we assumed the blood contained only three absorbing species, hemoglobin, bilirubin, and bilirubin's photoproducts, and that concentrations of these species were constant across the reactor height. This results in the following equation for the fraction of light absorbed by bilirubin (f):

$$f={\epsilon }_b{C}_b/({\epsilon }_b{C}_b+{\epsilon }_h{C}_h+{\epsilon }_p{C}_p),$$ (5)

where $\epsilon$ is the extinction coefficient specific to the light wavelength of the photoreactor LEDs, C is the concentration, and the subscripts b, h, and p refer to bilirubin, hemoglobin, and photoproducts, respectively. The extinction coefficients for bilirubin in human serum albumin at 470 and 505 nm are 73 and 12 cm²/mg, respectively.²² The extinction coefficients for hemoglobin at 470 and 505 nm are 0.515 and 0.313 cm²/mg, respectively.¹⁷ The extinction coefficients of bilirubin's photoproducts in human serum albumin were estimated from published spectra of bilirubin solutions after irradiation with green light, resulting in values of 7 and 5 cm²/mg at 470 and 505 nm, respectively.²³ A material balance can be used to express the concentration of photoproducts C_p in terms of the bilirubin concentration C_b [see Eq. (3)], which leaves the hemoglobin concentration as the only remaining parameter to define. To make predictions corresponding to the in vitro experiments, we used a hemoglobin concentration of 150 mg/ml, which is between the standard levels for male and female adults. To make predictions for photoreactor treatment of a newborn, the concentration of hemoglobin was assumed to be 193 mg/ml.²⁴

The total photons of light absorbed per area (B) can be calculated in terms of the absorbance A as follows:

$$B=E(1-{10}^{-A}),$$ (6)

where E is the incident light intensity. The absorbance is defined as

$$A={\epsilon }_b{C}_{b}h+{\epsilon }_h{C}_{h}h+{\epsilon }_p{C}_{p}h,$$ (7)

where h is the reactor height.

Combining Eqs. (5) and (6), we can estimate the bilirubin conversion rate per area as

$${\dot{R}}_{b\to l}=YfE(1-{10}^{-A}).$$ (8)

This reaction rate can be used to develop a differential equation for how the bilirubin concentration changes along the length of the reactor, resulting in

$${\displaystyle \frac{d{C}_b}{dx}=-{\displaystyle \frac{w{\dot{R}}_{b\to l}}{\dot{V}},}}$$ (9)

where w is the width of the photoreactor. This equation was integrated between the reactor inlet and outlet to determine the bilirubin concentration at the reactor outlet $C_{b2}$.

To solve for the bilirubin concentration in the blood reservoir, Eqs. (4) and (9) were solved simultaneously using the built-in differential equation solver ode45 in MATLAB, which implements the Runge–Kutta (4,5) method.²⁵ For each time step, the current value of bilirubin concentration in the reservoir was fed to the solver for Eq. (9) as the inlet concentration at x = 0, and Eq. (9) was solved for the concentration at the reactor outlet $C_{b2}$.

To determine the best-fit quantum yield, the value of Y was varied to minimize the sum of the error squared between the predictions and the experimentally measured values of bilirubin concentration in the blood reservoir. The built-in MATLAB function fminsearch was used to minimize the sum of the error squared; this function uses a Nelder–Mead simplex method.²⁶

Statistical analysis

Results are reported as averages and standard errors of the mean. Treatment effects were examined using analysis of variance (ANOVA) followed by Tukey's tests for pairwise comparisons. For comparison of initial and final bilirubin levels in the animal study, a paired two tailed t-test test was used. Statistical comparisons yielding a p-value less than 0.05 were considered to be significant.

RESULTS AND DISCUSSION

The potential for treating hyperbilirubinemia using a microfluidic photoreactor was first evaluated with in vitro experiments using the small-scale device illustrated in Fig. 1. The purpose of these experiments was to assess the effects of light wavelength, light intensity, and channel thickness on bilirubin isomerization, as well as the potential for damage to blood cells.

Pilot in vitro experiments were initially performed without a heat exchanger to control the temperature of the blood. This resulted in excessive heating for high LED power settings, causing significant destruction of red blood cells. Therefore, all subsequent experiments were carried out using a heat exchanger comprising a second microfluidic channel beneath the blood channel (see Fig. 1), which was perfused with 20 °C water from a refrigerated water bath. This heat exchanger maintained the blood temperature at the reactor outlet below 25 °C, and the temperature on the top surface of the acrylic housing below 40 °C.

Effect of channel height on bilirubin conversion

To test the effect of channel height on bilirubin conversion, experiments were carried out with microchannel heights of 125, 250, and 500 μm for a period of 2 h, using a wavelength of 505 nm and a constant incident light intensity of 125 mW/cm². The results of these experiments are shown in Fig. 3(a). To inform the design of a device for clinical use, we wish to compare the performance of devices with the same internal volume and blood flow rate, but with different channel heights. However, in our experiments, it was necessary to hold channel width and length constant so that the size of the reactor would match the LED array, making it impossible to change channel height without also changing device volume. To ensure a fair comparison between devices with different channel heights, the flow rate was adjusted to maintain a consistent residence time of 1 min, and the reservoir volume was adjusted to maintain a constant ratio of flow rate to reservoir volume (see Materials and Methods for more details). Channel height had a statistically significant effect on the final bilirubin concentration in the blood reservoir (p = 0.015). The final bilirubin concentration for the 125 μm thick photoreactor was 24 ± 5% of the original concentration, indicating that 76% of the bilirubin had been converted to photoproducts after a 2 h treatment period. In comparison, the final bilirubin concentration for the 500 μm thick channel was significantly higher at 47 ± 3% of the original concentration. This is consistent with the expected decrease in bilirubin conversion in the thickest channel as a result of light attenuation.

FIG. 3.

Factors affecting conversion of bilirubin in the photoreactor. (a) Effect of reactor height on bilirubin conversion after 2 h treatment. (b) Effect of light intensity and wavelength on bilirubin conversion after 2 h treatment. (c) Validation of photoreactor model for the following conditions: h = 250 μm, V = 10 ml, V˚ = 0.3 ml/min, E = 166 mW/cm². For all figure panels, symbols represent experimental data, lines show model predictions, blue triangles correspond to 470 nm, and green circles correspond to 505 nm. Sample size n = 3 for all data points in panels (a) and (b); n = 1 in panel (c). Each replicate corresponds to a different run using a separate blood sample.

Effect of light intensity and wavelength on bilirubin conversion

To examine the effects of light intensity and wavelength on the conversion of bilirubin to photoproducts, blood was circulated through the photoreactor for 2 h under various LED power settings at two different wavelengths, 470 and 505 nm. For these experiments, the microchannel height was held constant at 250 μm. As shown in Fig. 3(b), the final bilirubin concentration after 2 h of treatment decreased as the LED power increased for both wavelengths. This is an indication of a faster conversion of bilirubin to photoproducts at higher light intensities. Statistical analysis revealed that this effect was significant for both wavelengths (p < 0.0001).

Figure 3(b) shows that blue light at 470 nm outperformed green light at 505 nm in the conversion of bilirubin to photoproducts, resulting in lower bilirubin concentrations at the end of the treatment period. The best results were obtained for a wavelength of 470 nm at the maximum light intensity tested (125 mW/cm²), which resulted in a final bilirubin concentration of 18 ± 6% of the original concentration. Previous studies have also shown that blue light is more effective in the treatment of neonatal jaundice.²⁷ Currently, standard phototherapy primarily delivers light within the 430–490 nm band because the maximum absorption of bilirubin occurs near 470 nm, but historically green light at 505 nm had been used because it is theoretically more effective at penetrating the skin.^28,29 Our results are consistent with more recent studies showing that blue light is more effective than green light.²⁷

Mathematical model of bilirubin conversion

A mathematical model was developed to enable prediction of the factors affecting bilirubin conversion and inform the design of a device that could be used at a neonatal scale. When bilirubin is exposed to light, a complex set of reactions occur, including isomerization to structural and configurational isomers (e.g., lumirubin) and formation of oxidation products. For simplicity, our model treats this as a single reaction in which bilirubin absorbs light and then forms “photoproducts” (see Materials and Methods section for details). This simplified model was fit to the experimental data shown in Figs. 3(a) and 3(b) to estimate best-fit quantum yield values for conversion of bilirubin to photoproducts. A separate fit was performed at each light wavelength, resulting in best-fit quantum yields of 5.9 × 10⁻⁴ and 8.2 × 10⁻⁴ at 470 and 505 nm, respectively. These values are similar to, but about 30% lower than, previously published quantum yields for isomerization of bilirubin to lumirubin.³⁰ As shown in Figs. 3(a) and 3(b), the best-fit model predictions are consistent with the overall trends in the data, although bilirubin conversion is slightly underestimated at low light intensities and overestimated at high light intensities.

In order to demonstrate the predictive ability of the model, another experiment was performed with a device configuration that deviated from the conditions used to determine the best-fit quantum yield values. In particular, we used double the reservoir volume compared to previous experiments and a longer treatment period of 8 h instead of 2 h. The resulting experimental data are plotted against model predictions in Fig. 3(c). The model predictions are in good agreement with the experimental data.

Effects of light exposure on DNA damage

Our LED arrays can deliver light at an intensity that is more than 100 times higher than that used in conventional phototherapy. In order to examine the potentially damaging effects of high intensity LED light, a comet assay was used to quantify DNA damage. Blood was exposed to light in the photoreactor under various conditions designed to deliver a total light dose comparable to conventional phototherapy, but using a higher light intensity over a shorter duration. In particular, the blood was exposed to either 470 or 505 nm light at an intensity of 25 mW/cm² for 5 h, or at an intensity of 125 mW/cm² for 1 h. The results of these experiments are shown in Fig. 4. All of the conditions tested, including the sham treatment with the LED lights off, resulted in a slight increase in DNA damage compared to a control blood sample that was set aside for the duration of the experiment. The slight increase in DNA damage for the sham treatment could possibly be explained by cell damage caused by the peristaltic pump. There was not a statistically significant effect of light exposure on DNA damage (ANOVA, p = 0.51). Overall, these results suggest that short duration exposure to high intensity LED light at these wavelengths does not cause appreciable DNA damage.

FIG. 4.

Effect of light exposure on DNA damage. (Top) Representative images of control and treated samples from the comet assay, showing evidence of DNA damage in the tail (see arrow). (Bottom) Experimental results (n = 3) from the comet assay after photoreactor treatment at wavelengths of 470 or 505 nm for either 5 h at 25 mW/cm² or 1 h at 125 mW/cm².

Gunn rat study

To test the in vivo capabilities of the photoreactor, a study was conducted using Gunn rats. These rats lack an enzyme called uridine 5′-diphospho-glucuronosyltransferase, which causes elevated levels of bilirubin, making them a good model for studying hyperbilirubinemia.³¹ To carry out the in vivo study, we designed a new photoreactor system that included a larger reactor channel, two 470 nm LED arrays (enabling illumination of the blood from both sides), a touch screen interface and a radiator cooling system. Each LED array was set to a light intensity of 125 mW/cm². Pilot in vitro experiments were performed to confirm that these LED settings did not cause hemolysis. The experimental setup for the in vivo experiments is illustrated in Fig. 2. Three treatments were examined, each of which was carried out for 4 h: sham, where blood was passed through the photoreactor with the LED lights off, photoreactor, where blood was passed through the device with the LED lights on, and double volume exchange transfusion, where donor blood from healthy Wistar rats was used to slowly replace the bilirubin-rich blood within the animal.

As shown in Fig. 5, photoreactor treatment decreased bilirubin levels by 41 ± 20%, slightly exceeding the bilirubin reduction observed after exchange transfusion (38 ± 28%). However, statistical comparison of the different treatment groups did not reveal a significant effect (ANOVA, p = 0.22), likely due to insufficient sample size for the sham and exchange transfusion groups. Nonetheless, bilirubin levels after photoreactor treatment were found to be significantly lower than initial bilirubin levels (two-tail t-test, p = 0.024), which demonstrates that photoreactor treatment is effective for reducing bilirubin levels.

FIG. 5.

Results of the Gunn rat study. Percent reduction in total bilirubin is shown after each 4 h treatment. Treatments include sham (n = 3), photoreactor (n = 6), and exchange transfusion (n = 4).

In one case, photoreactor treatment resulted in a negligible decrease in bilirubin concentration. There was also a single case where exchange transfusion did not reduce bilirubin levels. The reason for these observations is unclear but could be related to variability between animals, or variability in the bilirubin measurement method. In both of these cases, the data points showing negligible change in bilirubin levels had the lowest initial bilirubin concentration (before treatment) within that experimental group.

Predicted performance of a clinical device

To evaluate the therapeutic potential of the proposed microfluidic photoreactor for the treatment of hyperbilirubinemia, we used the mathematical model [Eqs. (4)–(9)] to predict the effects of photoreactor treatment for 4 h on a neonate's bilirubin levels [see Fig. 6(a)]. For comparison, we also made predictions for double volume exchange transfusion, which we assumed occurred continuously over a 4 h treatment period (see Fig. S3 in the supplementary material). A 3 kg newborn has a blood volume of approximately 240 ml. Bilirubin is known to distribute into multiple tissue compartments that interact with the plasma,¹³ but the rate of transfer between these compartments is slow and was assumed to be negligible over the 4 h treatment period. Therefore, to make predictions for both photoreactor treatment and exchange transfusion, we assumed a reservoir volume equal to the newborn's blood volume. This is expected to overestimate the reduction in plasma bilirubin levels but, nonetheless, enables direct comparison between photoreactor treatment and exchange transfusion.

FIG. 6.

Predicted performance of microfluidic photoreactor for treatment of neonatal hyperbilirubinemia. (a) Schematic illustrating model, where V = 240 ml is the neonate's blood volume, C_b is the bilirubin concentration in the neonate's blood, $\dot{V}$ is the blood flow rate through the photoreactor, and C_b2 is the bilirubin concentration at the photoreactor exit. (b) Effect of device volume and blood flow rate on reduction in bilirubin levels after 4 h treatment with a device with a channel height of 250 μm illuminated from both sides using LEDs at 470 nm and a light intensity of 125 mW/cm². (c) Combinations of device volume and blood flow rate predicted to meet or exceed the performance of double volume exchange transfusion.

Figure 6(b) shows the predicted reduction in bilirubin levels after 4 h photoreactor treatment for various combinations of photoreactor volume and blood flow rate. We assumed a photoreactor design similar to that used in the Gunn rat study, but scaled up for treating a neonate. In particular, the photoreactor channel height was 250 μm, and it was illuminated from both sides using 470 nm LEDs at 125 mW/cm². Devices with different volumes were modeled by varying the channel length and width. Predictions were the same for different combinations of length and width that produced the same device volume. As shown in Fig. 6(b), photoreactor treatment is predicted to reduce bilirubin levels to a greater extent for high blood flow rates and high device volumes. For a given flow rate, the extent of bilirubin removal plateaus as device volume goes up. This is because, for large device volumes, the residence time is long enough for complete photoconversion of bilirubin, resulting in a bilirubin concentration of zero at the device outlet. Once this occurs, further increasing the device volume does help reduce neonate bilirubin levels. When device volume is held constant, increasing blood flow rate causes the extent of bilirubin removal to go up. This is because increasing the flow rate increases the amount of the neonate's blood that is treated in the 4 h period.

The dashed black line in Fig. 6(b) shows the predicted reduction in bilirubin levels for double volume exchange transfusion, enabling comparison of photoreactor treatment to exchange transfusion. Note that the predicted 86% reduction in bilirubin levels for exchange transfusion is an overestimate due to the assumption that bilirubin is confined to the blood and does not distribute between the blood and tissues (see Fig. S3 in the supplementary material); in practice, exchange transfusion typically reduces bilirubin levels by less than 50%.³² Nonetheless, this bilirubin removal prediction can be used as the basis for comparison to photoreactor predictions, which use the same assumption. Several combinations of photoreactor flow rate and device volume are predicted to match the performance of exchange transfusion. For example, a flow rate of 4 ml/min and a device volume of 5 ml reduce bilirubin levels by 86%, as does a flow rate of 2 ml/min and a device volume of 15 ml.

The shaded area in Fig. 6(c) shows specific combinations of blood flow rate and photoreactor volume that are predicted to yield bilirubin levels after 4 h of photoreactor treatment that meet or exceed the performance of exchange transfusion. Overall, the predictions show that the microfluidic photoreactor is expected to enable effective blood treatment with a small priming volume and low blood flow rate. For example, the predictions show that the photoreactor is expected to match the performance of exchange transfusion when the blood flow rate is 4 ml/min and the device volume is 5 ml. This is such a small volume that the device could be pre-primed with sterile saline, avoiding the need for donor blood. This would not only dramatically improve safety but also greatly simplify the treatment process.

It may be desirable to reduce the photoreactor treatment time. Therefore, in Fig. S4 in the supplementary material, we show photoreactor configurations that are expected to yield bilirubin levels that match exchange transfusion after 1 h of treatment. Overall, higher blood flow rates and device volumes are needed to match the performance of exchange transfusion when the photoreactor treatment time is reduced to 1 h. For instance, it is possible to match the performance of exchange transfusion after 1 h photoreactor treatment for a blood flow rate of 18 ml/min and a device volume of 20 ml.

CONCLUSIONS

In this study, we highlight the potential for extracorporeal blood treatment in a microfluidic photoreactor as a promising approach for treating extreme neonatal jaundice, potentially offering a safer and more convenient alternative to exchange transfusion. This new treatment approach uses the same light-catalyzed reaction as phototherapy, but enables the use of ∼100× higher light intensity for much faster photoconversion of bilirubin. Our experimental results demonstrate that the rate of bilirubin conversion significantly increases as light intensity increases. Bilirubin conversion was more effective in thin microfluidic channels than in thick channels, and light at 470 nm was more effective than light at 505 nm. Our results also suggest that illumination of blood with high intensity LED light is safe, as we did not observe a significant increase in DNA damage, even for light intensities more than 100× higher than those used in conventional phototherapy. We also demonstrated that photoreactor treatment significantly reduces bilirubin levels in Gunn rats, at a level comparable to exchange transfusion. Theoretical predictions of neonate treatment highlight two advantages of the microfluidic photoreactor that will promote clinical acceptance of the device: (1) the small volume of the device will enable the establishment of an extracorporeal circuit without the need for priming with donor blood and (2) the low blood flow rate will enable safe and convenient vascular access, facilitating the initiation of treatment for medical practitioners and improving safety for the patient. The anticipated flow rates are compatible with umbilical catheters—both 3.5 and 5 French—that are often employed in double volume exchange transfusion.

Several factors will need to be considered in future studies before the microfluidic photoreactor is ready for clinical use. The device used here for treating Gunn rats will need to be scaled up for the treatment of human neonates, which are about 10 times larger. It will also be necessary to more rigorously demonstrate safety and efficacy—including in vivo measurement of DNA damage as a function of light intensity and exposure time—using a preclinical animal model that is more representative of hyperbilirubinemia in human neonates, such as the rhesus monkey.³³ The device used in the current study requires systemic anticoagulation, which is associated with bleeding risks in neonates. Future studies should examine the potential for reducing or eliminating the need for systemic anticoagulants by coating the device surfaces to locally inhibit coagulation.³⁴ The effects of shear rate on blood coagulation within the device should also be examined in future studies. It would also be useful to more rigorously examine temperature control in future studies to ensure that blood temperature is maintained at safe levels. Overall, further development of the photoreactor technology has the potential to offer a promising new approach for treating extreme neonatal hyperbilirubinemia.

SUPPLEMENTARY MATERIAL

See the supplementary material for additional information on the method we used to quantify bilirubin concentration, a close-up image of the photoreactor used in the Gunn rat study, and additional predictions of bilirubin concentration during neonate treatment.

AUTHOR DECLARATIONS

Conflict of Interest

J.M.L. and A.Z.H. are listed as inventors on a U.S. patent application filed by Oregon State University entitled “Microfluidic removal of excess bilirubin from blood” (application No: 16/431,928). R.A.F., H.Y.L., S.L.J., B.S., K.F.S., and J.E.B. have no competing interests.

Ethics Approval

Ethics approval was obtained for the studies involving human blood and for the animal study with Gunn rats. Human blood was obtained via venipuncture from healthy volunteer donors using a protocol that had been approved by the Oregon State University Institutional Review Board. Blood draws were carried out in accordance with relevant guidelines and regulations, and informed consent was obtained from all blood donors. The animal study was carried out at Absorption Systems (San Diego, CA) in accordance with relevant guidelines and regulations, including the ARRIVE guidelines. The study protocol was approved by their Institutional Animal Care and Use Committee.

Author Contributions

J. M. L. and R. A. F. contributed equally to this work.

DATA AVAILABILITY

The data that support the findings in this study are available from the corresponding author upon reasonable request.