Microfluidic screening is gaining attention as an efficient method for evaluating nanomaterial toxicity. Here, we consider a multiparameter treatment where nanomaterials interact with cells in the presence of a secondary exposure (UV radiation). The microfluidic device contains channels that permit immobilization of HaCaT cells (human skin cell line), delivery of titanium dioxide nanoparticles (TNPs), and exposure to a known dose of UV radiation. The effect of single-parameter exposures (UV or TNP) was first studied as a benchmark, and then multiparameter toxicity (UV and TNP) at different concentrations was explored. The results demonstrate a concentration-dependent protective effect of TNP when exposed to UV irradiation.
Microfluidics offers a powerful approach to toxicity screening, since multiple tests can be carried out using tiny sample volumes in a compact microchip format. These chips may exploit various features of flow (laminar profile) and geometry (cell capture) to mimic biological environments in a way that is not possible using conventional well plates. Microfluidic toxicity screening has been demonstrated for chemicals such as potassium cyanide,1 cycloheximide,2 and chemotherapy drugs.3,4 Microfluidic devices have also been utilized for drug testing in skin-on-a-chip5 and organ-on-a-chip6 models with examples of specialized architecture such as micromixers and multilayer 3D environments. However, nanoparticles present a much greater challenge due to the diverse functionality depending on the chemistry, size, and shape.7 This is in the broader context of explosive growth in consumer applications (pharmaceuticals, cosmetics, and antibacterial additives) and the need for reliable regulation of product safety.
Single-parameter nanotoxicity screening in microchannels has been demonstrated based on cell immobilization and laminar delivery of particles.8–11 Microfluidic devices have been shown to have many advantages for cellular and nanoparticle analysis—small sample volumes, reduced cost, controllability, and reproducibility being chief among them. Microfluidics also offers the ability to introduce multiple biological conditions in a single device and replicate in vivo conditions.12–15 The microfluidic system approach offers an exciting opportunity because current nanotoxicity screening methods (e.g., static conditions) are laborious, time-consuming, and cannot keep pace with the rate of discovery of new nanomaterials due to the advances in nanotechnology.
Multiparameter toxicity screening has received less attention but offers greater insight and relevance to real-world exposures. Biological exposure to nanomaterials can occur in the presence of many different external stimuli. For instance, a nanomaterial included in a personal care product16–19 may remain on the body for many hours and be exposed to air, moisture, light, and heat, along with sweat and sebum from the skin. The nanomaterials will exhibit different physical, chemical, or biological properties resulting in either enhanced or diminished toxicity.
Studies of novel cancer therapies have shown that gold nanoparticles are sensitive to gamma waves on a size-dependent basis,20 and their uptakes into breast cancer cells can be increased based on their surface coating property.21 Similarly, iron oxide nanoparticles have been investigated for their ability to induce hyperthermia in a magnetic field and show a size dependency of their heating rates.22 To perform screening experiments for these kinds of parameters in a microfluidic-based nanotoxicity screening device, it requires the ability to introduce external stimuli as well as observe the internal stimuli interacting with the cells. Thus, multiparameter screening is essential and will be necessary for the envisaged microfluidic technology.
Practical screening methods for nanotoxicity will rely on multiparameter testing, due to the vast matrix of parameter space that must be explored. Microfluidic techniques have already demonstrated great potential in creating parallel, low-cost testing methods for both chemicals and nanomaterials.23–27 Here, the potential for multiparameter nanotoxicity screening in a microfluidic chip was investigated by exposure of HaCaT cells (a human epidermal keratinocyte cell line) to both UV radiation and commercially available titanium dioxide nanoparticles (TNPs). By utilizing a low-cost microfluidic device with a quartz substrate for its UV transparency, we show the potential for combining different materials in order to facilitate exposing multiple external parameters into lab-on-a-chip technologies for toxicity measurements.
II. MATERIALS AND METHODS
A. Cell culture
HaCaT keratinocytes (CLS Cell Lines Service, Eppelheim, Germany) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) containing 4.5 g/l D-glucose and Glutamax (Gibco), supplemented with 10% fetal bovine serum (FBS) (Gibco), 100 units/ml penicillin, 100 μg/ml streptomycin, and 1mM amphotericin-B (Gibco). Cells were maintained in an incubator at 37 °C, 5% CO2, and 95% relative humidity for 6–7 days until they reached ∼80% confluence before passaging. HaCaT cells were used between passages 25 and 38. Trypsin-EDTA (Gibco) and a wash with 1× phosphate buffered saline (PBS) (Gibco) was used to detach the cells from culture flasks.
B. Microfluidic device preparation
A 4 in. 〈100〉 P-type silicon wafer was cleaned with acetone and isopropyl alcohol (IPA) while revolving on a DeltaSUSS spin coater, and then SU8-50 photoresist (MicroChem, Westborough, MA, USA) was spin-coated at 500 rpm for 10 s to spread, then spun at 1100 rpm for 30 s to achieve a thickness of 155 μm. A soft bake procedure was performed at 65 °C for 8 min and 95 °C for 25 min before using a direct laser writer (Kloe Dilase 650) to expose a master pattern onto the photoresist with a 10 μm beam size. Postexposure bake was performed at 65 °C for 2 min and 95 °C for 8 min, before development in propylene glycol methyl ether acetate (PGMEA) (MicroChem, Westborough, MA, USA) for 8 min, and then 1 min in a fresh PGMEA solution and rinsing with IPA. Finally, the wafer was hard baked at 65 °C for 1 min, 95 °C for 3 min, 130 °C for 10 min, 95 °C for 3 min, and 65 °C for 1 min. The silicon master was then hydrophobized by chemical vapor deposition of trimethylchlorosilane (Sigma-Aldrich) inside a desiccator.
Poly(dimethylsiloxane) (PDMS) microfluidic channels were prepared using a Sylgard 184 PDMS kit, mixing in a 10:1 ratio by weight of the monomer to the curing agent. This mixture was stirred vigorously in a plastic container until bubbles formed and then poured gently over the silicon master in a plastic Petri dish. The Petri dish was transferred to a desiccator and placed under vacuum. Once the majority of air bubbles had escaped the polymer, the Petri dish was transferred to an oven at 70 °C for 1 h to thermally cross-link the polymer and increase its rigidity. The cured PDMS chip was cut using a scalpel and peeled off from the silicon master, and then inlet/outlet ports were cored using a 4 mm Miltex biopsy punch (Integra Life Sciences, Plainsboro, NJ, USA).
Quartz wafers of 6 in. diameter (Shin-Etsu, Chiyoda, Tokyo) were diced on an Automatic Dicing Saw DAD321 (Disco, Ota-ku, Tokyo) to sizes of 60 × 30 mm2. These pieces were cleaned with absolute ethanol (Fisher Scientific) and then 18.2 MΩ/cm milli-Q water, and then dried with N2 gas and placed in a plastic Petri dish (Nunc). In order to bond the PDMS channels to the quartz substrate, the quartz pieces were exposed to an oxygen plasma for 3 min in a plasma cleaner (Harrick Plasma, Ithaca, NY, USA). The PDMS channels were then placed channel-face-up and exposed along with the quartz plate for another 15 s. The PDMS device was then flipped and pressed down onto the coverslip to form a watertight bond with the quartz.
C. Cell seeding
For well plate nanotoxicity and irradiation studies, 96 well plates (Corning) were seeded at 4 × 104 cells per well and incubated for 48 h until the HaCaT cells reached ∼80% confluence (assessed visually). Before adding any nanoparticles or exposing to UV irradiation, the cell culture media was removed and replaced with a fresh cell culture medium to prevent any dead cells in suspension from interfering with the test.
For a microfluidic nanotoxicity study, HaCaT cells were resuspended in culture medium at a concentration of ∼1 × 106 cells/ml and pipetted directly into the microchannel. Straight microchannels of 2 cm length and 2 mm width (surface area = 0.4 cm2) were used in this experiment to provide enough room for the cells to bind and proliferate. The microchannel volume (excluding the ports) was calculated to be 6.2 μl, and so volumes of >10 μl were applied to each inlet to completely fill the channel. The cells flowed along the length of the channel due to the positive pressure from the inlet volume and created a relatively homogenous coverage. The devices were incubated for 1 h at 37 °C and 5% CO2, until the cells were settled on the surface and began to attach. At this point, a 150 μl culture medium was gently pipetted into the inlet and outlet ports of the microchannel. This method allowed the majority of cells in the microchannel to remain static due to equalized pressures. The device was then incubated for a further 48–72 h until the cell reached ∼80% confluence.
For a microfluidic irradiation study, a quartz substrate was used due to its UV-transparent nature. HaCaT cells were found to bind slower to quartz than to glass, and so a layer of Collagen I (Gibco) was deposited onto the quartz substrate to improve cell binding. In brief, the collagen solution (0.03 mg/ml in PBS) was pipetted into the inlet of the assembled microfluidic device. The solution flowed to fill the microchannel due to positive pressure and was left at 4 °C overnight in order to fully coat the surface. This treatment was found to greatly accelerate the attachment of HaCaT cells to quartz substrate and improved the cell seeding efficiency (Fig. 1). Additional media could be added at 30 min without causing bound cells to detach.
D. Cell viability assay
A LIVE/DEAD™ Viability/Cytotoxicity Kit for mammalian cells (Thermo Fisher) was used to fluorescently stain the living and dead cells. The Calcein-AM component (Ex/Em = 495/515 nm) stained viable cells green by permeating the cell membrane and enzymatically degrading to a brightly fluorescent green calcein product, while the ethidium homodimer-1 component (Ex/Em = 528/617 nm) permeated only damaged cell membranes and fluoresced red when bound to nucleic DNA.
For 96 well plate and microfluidic cell viability measurements, the media were aspirated by gentle suction and replaced with 100 μl of PBS. In the 96 well plates, 5 μl of viability dyes were added to each well and the well plate was mixed on a rotary shaker at room temperature for 20 min, protected from light. For microfluidics, 5 μl of the viability dyes were added to each channel inlet and flowed through by aspirating a small volume at the outlet, then incubated at room temperature protected from light for 20 min. 10× magnification microscopy images were taken on a Nikon Eclipse TI inverted microscope. The images were analyzed by cell counting with Fiji (ImageJ)28 to determine the number of green cells (for live) and red cells (for dead).
E. Nanoparticle exposure
Uncoated TNP dry powder of 15–30 nm spherical diameter size range and of anatase (tetragonal) crystallinity was purchased from Nanostructured and Amorphous Materials, Inc., Houston, TX, USA. A stock solution of 500 μg/ml was constituted in PBS. The purchased TNP were not sterile when received. To ensure that the particles were suitable for use in cell culture treatment, gamma irradiation at 50 kGy was performed by Steritech Inc., VIC, Australia. This treatment sterilized the TNP and showed no difference in the resuspension and so was used for these experiments.
TNPs were introduced to cells on well plates by first aspirating the media via gentle suction. Cell culture medium was added to each well, and then the required volume of TNP stock solution was added to make up a 100 μl suspension. For the microfluidic study, TNPs were introduced to the microchannels by premixing the stock solution with the DMEM medium in a well plate and then aspirating the media from the channel inlets and outlets via gentle suction, which did not dry out the channel itself. Pipetting 100 μl of known concentrations of TNP suspensions into each inlet then allowed the samples to flow through the channel due to positive pressure, until equilibrium was reached.
F. UV exposure
For the well plate study, UV irradiation was performed on 96 well plates (Corning) by placing each plate beneath the lamp of a EuroClone (Pero, Milan, Italy) TopSafe 1.2 biological safety cabinet [Fig. 2(a)]. The lamp was a Sylvania G3018 30 W UVC germicidal lamp, quoted with a peak wavelength of 253.7 nm. The lamp intensity was measured with a power meter (Jasco), with the detector placed on the cabinet surface and directed upward toward the lamp. The well plates were placed in the center of the cabinet directly under the lamp at a distance of 70 cm, where the power of the UV light was measured at 0.8 mW/cm2.
For experiments with different exposure doses of UV, a custom aluminum photomask was applied over the course of the exposure. Negative control of UV dose of 0 mJ/cm2 was achieved by covering a triplicate set of wells from the beginning of the experiment, preventing any UV radiation from reaching them. UV was then applied for the whole well plate for the required period of time until the first desired level of energy was reached, at which point the lamp was switched off and the photomask was applied to block the first triplicate set of wells. This was repeated until all desired energies had been applied to at least one triplicate set of wells.
The values of UV exposure dosage (mJ/cm2) were calculated by multiplying the lamp irradiance (mW/cm2) by the exposure time (s). The maximum time required to expose cells to 200 mJ/cm2 under the cabinet lamp was 2.66 min. After irradiation, the well plates were returned to the incubator for 24 h prior to running cell viability assays to allow the UV-irradiated cells time to recover or undergo apoptosis and express markers of cell stress.
For the microfluidic study, irradiation was performed on microfluidic devices by placing them on an ETX-20.C UV table (Extech Equipment Pty. Ltd., Boronia, VIC, Australia) [Fig. 2(b)]. The lamps inside were 6 × 15 W UV tube lamps, quoted with a peak wavelength of 254 nm. The transmission spectrum of the UV table and microfluidic substrates was measured with a fiber-optic cable attached to a UV-vis spectrophotometer (OceanOptics) and recorded in OceanView software. The UVC peak at 255 nm was observed to be lower in intensity than the two other peaks discovered at 311 nm (UVB) and 364 nm (UVA), as shown in Fig. 3. The UV intensity was measured with a power meter (Jasco) with the detector facing down onto the surface, and the area directly above each lamp was found to be where the maximum intensity was located. The power of the UV light directly over one of the tube lamps was measured at 3.0 mW/cm2, and this location was chosen for the experiments.
For the control experiments, a custom-made aluminum photomask was used to block all UV lights coming from the surface, with the power meter reading of 0.0 mW/cm2. Any microchannel that required irradiation was placed on the exposed area of the UV table and was left there for the required exposure time (<3 min). After irradiation, the microfluidic devices were returned to the incubator for 24 h prior to viability assays to allow the UV-irradiated cells time to recover or undergo apoptosis or necrosis.
III. RESULTS AND DISCUSSION
A. Single-parameter exposure
1. Ultraviolet (UV) light irradiation
UV radiation was investigated as a physical cell stressor in both well plate and microfluidic systems. UV was of interest not only for a single-parameter test but also in its potential for its inclusion in a multiparameter study between direct physicochemical stressors (i.e., nanoparticles) and external stressors (i.e., irradiation). UV radiation is known for its detrimental effect on biological matter and is used in laboratory biosafety cabinets and professional and personal sterilizers. UVC (100–280 nm) is the most common form of UV used in germicidal applications, as it induces mutations into the genetic material of micro-organisms and causes them to undergo apoptosis.29,30 UVB acts to disrupt the nucleic acids of mammalian cells,31,32 as its short wavelength (280–315 nm) is absorbed by the deoxyribonucleic acid (DNA) within cell nuclei. UVA (315–400 nm) is also known to cause drying and longer-term damage to human skin cells.33–35
The most damaging form of UV radiation is UVC, and although it is commonly blocked from interacting with living skin cells by the outer epidermal layer called the “stratum corneum,” some skin samples show this layer’s thickness to be less than the penetration depth of UVC (∼50 μm depth at 254 nm wavelength), and as such there can be damage to epidermal cells.36 For this reason, UVC-containing light sources were used in this experiment to provide a damaging source of external cell stress. As an initial test with the well plate method, top-down irradiation was performed using a biosafety cabinet’s germicidal UV lamp by placing a well plate with its cover removed beneath the lamp for predetermined exposure durations. Analysis of the Live/Dead staining of the cells (Fig. 4) shows a calculated half maximal inhibitory concentration (IC50, 50% viability) value of 35.6 mJ/cm2 for HaCaT cells in well plates.
2. TiO2 nanoparticles (TNPs)
Due to the contentious nature of TNP nanotoxicity in the current literature,16,18,37–43 we sought to investigate the effect of TNP on the HaCaT cell line. From the literature of other materials, e.g., silver nanoparticles, it has been shown that in the nanometer range, the smaller the particle, the higher the observed toxicity.44,45 Thus, we chose to investigate a sample of anatase TNP (Nanostructured and Amorphous Materials, Inc.) with diameters ranging from 15 to 30 nm. Those TNPs were shown to have negligible toxicity in both well plates and microchannels (Fig. 5). However, at the highest concentrations of 200–250 μg/ml, TNP aggregates formed and were settling as a white coating over the cells that was visible to the naked eye. Given that there was no dose-response curve within either test, these 15–30 nm anatase TNPs were determined to be nontoxic across the concentration range of 0–250 μg/ml with the exposure conditions studied in this experiment.
This result is of interest due to the public health concerns over nanomaterials in products like sunscreens. Macroscale (>100 nm) titanium dioxide is a widely used pigment in white paints and as a food colorant (E171) and has only come under recent scrutiny upon the inclusion of its nanoscale form in cosmetics and sunscreens, which stems from the desire for transparent skin treatments and sun protection for esthetic purposes. Well plate studies have been performed on many different sizes of TNPs for the purposes of discovering potential toxicity,16,18,39,43,46,47 and different parameters have been introduced to more closely match a real-world use case, i.e., utilizing a chemical environment that is an analog to human sweat.48 This is an example of a toxicity screening requirement that would benefit enormously from a multiparameter, parallel screening method as described in the following section.
B. Multiparameter exposure (UV and TNPs)
In order to prove the feasibility of using microfluidics for multiparameter nanotoxicity screening, moving to a microfluidic platform required the use of a bottom-up UV irradiation source as the microchannels were fabricated in PDMS, a material that is opaque to UVC radiation.49 Due to this requirement, a UV lamp table was used as the radiation source in the microfluidic experiment (refer to Sec. II F). The microfluidic devices were placed on the table surface and irradiated from beneath, through the quartz substrate material (UV transparent). Cells adhering to the substrate would then receive the radiation dosage that was transmitted through the material. The literature for UV experiments on HaCaT cells50–53 commonly uses dosages in the range of 0–200 mJ/cm2. Therefore, it was necessary to determine the experimental parameters that allowed for this range to be investigated.
After determining that the UV table was suitable for microchannel experiments, multiparameter experiments were conducted by exposing HaCaT cells simultaneously to nanoparticles and UV radiation in microchannels. A sublethal exposure dosage of UV was required for this experiment, as the combined effect of nanoparticles and UV was of interest; thus, a dosage of 25 mJ/cm2 was selected, as in the well plate testing it had been shown as sublethal against HaCaT cells and thus would show positive or negative influences from the addition of TNP.
It is commonly known that TNPs are used for sun protection as UV blockers in sunscreens and cosmetics, and so it is of particular interest to investigate whether the usage of TNP in these products will or will not lead to negative health effects. Titanium dioxide at the macroscale and the nanoscale has been established as an effective photocatalyst in the literature38,54–60 and interacts strongly with UV radiation. Given that this interaction may have led to the release of harmful reactive species in a biological system, it was hypothesized that higher concentrations of TNP would interact more strongly under irradiation and lead to lower cell viability if these reactive species were present. Figure 6 shows screening results for a range of 15–30 nm TNP concentrations with and without exposure to 25 mJ/cm2 UV.
It was observed that in the absence of TNP, a significant (p < 0.0001) decrease in viability was observed with cells exposed to the UV radiation alone. However, a trend of increasing viability with increasing nanoparticle concentration up to 250 μg/ml was seen. No significant toxicity was observed at TNP concentrations of 150 μg/ml and above between the blocked and exposed channels (Fig. 6), suggesting that the TNPs are protecting the HaCaT cells from the effects of UV radiation. These multiparameter exposure results may be of considerable interest in the context of public health concerns over TNP in personal care products. From these results, it appears that 15–30 nm anatase TNP was not only nontoxic up to 250 μg/ml but also acts to protect the skin cells from UV damage.
The action of titanium dioxide in UV protection has been investigated in the literature.16,18,42,43 It is widely accepted that titanium dioxide is an effective inorganic UV filter, as it absorbs and scatters a broad range of wavelengths in the UV spectrum, mostly in the UVA-UVB region. The UVA-UVB region is the range of wavelengths that is least blocked by the Earth’s ozone layer, and so it is the main target for sunscreens containing UV blocking compounds. We showed that the peaks for UVA and UVB contributed a significant portion of the total intensity emitted from the UV table. Thus, if the TNP were acting to block this wavelength range, the total dosage of UV that reached the HaCaT cells residing in the microchannels would have been significantly reduced. However, the irradiation was delivered bottom-up directly to the cells through the quartz glass, not top-down through the nanoparticle suspension, meaning that the UV would interact with the cells prior to interacting with the TNP.
Given that we observe a concentration-dependent increase of viability with the TNP dosage, the by-products of photocatalysis on the TNPs do not appear to play a role in our experiments. Penetration studies have been performed on HaCaT cells with 18 nm particles and show noncytotoxic uptake into the cell membrane but not the nucleus,61 and thus the nanoparticles are likely contributing to absorbance or scattering of UV radiation inside the cell before it can cause damage to the nuclear DNA. In both the literature and our device, the HaCaT cells were in a subconfluent monolayer and so the TNPs may have been able to migrate to the very lowest point of the cell (closest to the UV source) prior to interaction.
This result demonstrates how the proposed approach may help answer questions about mechanistic details regarding multiparameter toxicity. In this case, the UV-absorbing TNPs appear to reduce photoreactions in the aqueous solution that subsequently lead to the observed toxicity, e.g., reactive oxygen species, rather than direct UV damage. Further research using the method reported here may lead to a greater understanding of multiparameter toxicities.
Microfluidic screening of multiparameter nanotoxicity has been investigated for exposure of HaCaT cells to 15–30 nm anatase TNP and/or UV irradiation. The microfluidic device is a quartz-PDMS hybrid that allows UV light to enter the channel in specific locations via masking. Well plate UV exposure experiments reported an IC50 of exposure dosage of 35.6 mJ/cm2. TNP alone did not show any toxicity in well plate or microfluidic experiments. For the combination of UV exposure and TNP in microfluidic channels, a concentration-dependent protective effect of the TNP was observed, consistent with its widespread use in sun protection products. Thus, we have demonstrated this microfluidic platform approach as a promising tool for multiparameter screening of nanotoxicity and associated mechanisms.
S. McCormick would like to acknowledge the South Australian node of the Australian National Fabrication Facility (ANFF-SA) for their technical assistance and the use of their facility in the production of the devices in this project. This research was funded by the Australian Research Council (Grant No. DP150101774). The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.
The authors declare no conflict of interest.