A microfluidic systems biology approach for live single-cell mitochondrial ROS imaging

Abstract

Most current studies of ROS production report globally averaged measurements within the cell; however, ROS can be produced in distinct subcellular locations and have local effects in their immediate vicinity. A microfluidic platform for high-throughput single cell imaging allows mitochondrial ROS production to be monitored as varying in both space and time. Using this systems biology approach, single cell variability can be viewed within a population. We discuss single cell monitoring of contributors to mitochondrial redox state - mitochondrial hydrogen peroxide or superoxide - through the use of a small molecule probe or targeted fluorescent reporter protein. Jurkat T lymphoma cells were stimulated with antimycin A and imaged in an arrayed microfluidic device over time. Differences in single cell responses were observed as a function of both inhibitor concentration and type of ROS measurement used.

1. Introduction

Reactive oxygen species (ROS), such as hydrogen peroxide and superoxide, have critical roles in numerous cellular processes including signal transduction and have been found to be abnormally high in many diseases such as cancer¹ and autoimmune disorders². A major source of ROS is the electron transport chain in the mitochondria, which produces superoxide and hydrogen peroxide. It has been estimated that mitochondrial respiration accounts for 50-500 μmol·kg⁻¹·min⁻¹ cellular ROS, depending upon the metabolic rate of the cell³.

Two ROS produced in the mitochondria are superoxide and hydrogen peroxide. Superoxide is generated in the mitochondria as a by-product of complex III in the electron transport chain (Figure 1) and can be disproportionated to hydrogen peroxide via manganese superoxide dismutase (MnSOD or SOD2). Hydrogen peroxide can both diffuse through the mitochondrial membrane and be transported via aquaporins⁴. Although superoxide cannot passively diffuse through the membrane, transport has not been definitively excluded from anion transporters. Both ROS are implicated in different signaling processes, such as proliferation, apoptosis, and the cell cycle⁵. Fluorescent indicators exist for both species. MitoSOX is an irreversible dye capable of localizing to the mitochondria and fluorescing upon oxidation by superoxide. This dye is commonly used as it is well studied and shows exclusive sensitivity to superoxide with 4 × 10⁶ M⁻¹s⁻¹ as the rate-limiting step of oxidation by superoxide^6,7. Hydrogen peroxide has historically been more difficult to image with controversy surrounding H₂DCF-DA measurements^8,9. The recombinant protein, HyPer, has been developed from cpYFP and Oxy-R and is capable of changing conformation upon oxidation by hydrogen peroxide¹⁰. Once transfected into cells, the reporter protein is capable of providing ratiometric measurements with the correct microscope set-up with two excitation filters and one emission filter.

Figure 1. Schematic of mitochondrial ROS production.

The electron transport chain is composed of four main complexes that allow electrons to be transferred, driving ATP production. Superoxide is a byproduct of respiration and can be disproportionated to hydrogen peroxide via SOD2. Antimycin A is a complex III inhibitor shown to increase the rate of production of mitochondrial ROS by inhibiting the flow of electrons to cytochrome c.

Stochastic fluctuations in transcriptional and translational regulators within a cell are now considered influential to differences in cell behavior¹¹. Such differences can be masked by techniques that analyze populations of cells instead of single cells. For instance, flow cytometry analyzes individual cells at single time points but cannot measure a single cell through multiple time points. With single time point measurements, kinetic differences between individual cells in the response to a stimulus are not observed. High magnification imaging through time can address this problem for adherent cells, but for T cells and other suspension cells, this can be difficult as the cells may drift out of the focal plane. Advances in microfluidic design offer an alternative approach to studying these differences by i) passively trapping and analyzing the fluorescence of cells through time¹² and ii) allowing quantification of components of signaling networks within a single cell and then applying these findings to an overall system⁹.

The ability to track individual cells through time will lead to a more complete understanding of redox signaling and ultimately more insight into diseased states. In this chapter, we discuss methods for utilizing microfluidics to analyze mitochondrial superoxide and hydrogen peroxide responses to an oxidative stimulant, antimycin A, among single cells in a high-throughput manner. While our analysis is limited to one reporter molecule at a time, ultimately other fluorescent measurements can simultaneously be performed in multicolor live imaging microscopy (e.g. calcium, pH, mitochondrial membrane potential, etc.) to provide insight in how variations in mitochondrial function influences behavior across an array of cells.

2. Microfluidic Platform

Densely arrayed single-cell trapping device

A previously developed microfluidic device was utilized for imaging mitochondrial ROS production in the Jurkat T cell line through time¹². This high-throughput single-cell trapping device has the capability to hold approximately 4000 total cells in eight different trap arrays¹². Each array contains 25 traps per row and 20 rows. The device is compatible with any mode of optical microscopy, so imaging can be done at different magnifications and, with a motorized stage, multiple chambers can be imaged within seconds (Figure 2). The traps can also be placed downstream of different microfluidic platforms, allowing for different stimulus conditions. In this study, a linear serial dilution generator¹³ was used upstream to create a range of stimulus concentrations while simultaneously maintaining consistency in flow rate, dye loading, etc. between chambers containing cells from the same cultured population.

Figure 2. Single-cell analysis of mitochondrial superoxide production using MitoSOX Red Mitochondrial Superoxide Indicator (Invitrogen) dye in microfluidic cell traps.

(a) 10x view of approximately 160 single cell traps with MitoSOX labeled Jurkat cells. (b) 20x view of traps.

Device preparation

Devices were molded in a polydimethylsiloxane (PDMS) A and B mixture of 10:1 using a SU-8 master mold developed in previously described methods¹². Devices were cut and holes were punched using a 19-gauge needle. The prepared devices were bonded via oxygen plasma to a 1 mm glass slide for imaging and polyethylene (PE3) tubing (Scientific Commodities) was used for all connections with solutions. Initially, the device was primed with a 2% bovine serum albumin (BSA) in PBS solution. This removes any air bubbles from the channels and prevents unintentional cell adhesion to the walls or glass slide. Once primed, the cells were loaded at a concentration of 4 × 10⁶ cells/mL using gravity flow and the optimized flow rate of ~ 2 μL/hour found previously was used for all treatments¹².

3. Microscope System and Image Analysis

Microscope Set-Up

Once bonded to a glass slide, the cell trap device is placed on a 37°C heated stage and imaged using a Nikon Eclipse Ti inverted epifluorescent microscope (Figure 3). Inlet tubing for flow of buffer and/or stimulus was set approximately 40 cm above the outlet tubing to allow for a gravity-based pressure-driven flow of liquid through the chip. Time-lapse microscopy was performed on an automated stage with a 0.7 s delay between imaging the separate chambers with an exposure time of 900 ms. Images were collected every 30 s for 70 minutes. All images within a given video were set to the same look-up tables (LUTs) to avoid digital differences between images during image analysis.

Figure 3. Device position with respect to microscope.

Device is plasma bonded to a glass slide with tubing connections to treatment solutions and outlet. Slide is placed on a motorized stage set to 37°C. Filter cubes are utilized for appropriate excitation and emission wavelengths.

Image Analysis

Matlab (MathWorks) scripts were written and utilized for image analysis of the time-lapse videos. The fluorescent image was converted to a binary image to identify the fluorescent cells in the trap array. To ensure cells present during the entire experiment were selected, both the first and last images were converted to a binary image using a Matlab built-in global threshold function, “graythresh”. Next, manual selection was performed to select the cell areas to analyze, based on presence in both the first and last binary images. The mean fluorescence was then calculated for each region of interest. To remove differences in background fluorescence between frames, the average fluorescence of a non-occupied portion of the trap was calculated and subtracted from each region of interest at each time point. This ensures differences in intensities were less reliant on background noise. To normalize each cell individually, measurements are divided by the first mean intensity for that region of interest.

4. Imaging Mitochondrial Superoxide Production

Treating cells with MitoSOX

The Jurkat E6-1 human acute T cell lymphoma cell line (American Type Culture Collection) was cultured at 37°C in a humidified 5% CO₂ incubator in RPMI 1640 Medium without Phenol Red and with L-glutamine (Sigma-Aldrich), with 10 mM HEPES buffer, 1 mM sodium pyruvate, 100 units/mL penicillin-streptomycin (Cellgro), 1x MEM Non-Essential Amino Acids, and 10% fetal bovine serum (Sigma-Aldrich). To visualize the presence of mitochondrial superoxide, Jurkat cells were incubated with 5 μM MitoSOX Red Mitochondrial Superoxide Indicator (Invitrogen) for 10 minutes at 37°C, according to previous protocols⁶. Following incubation, cells were washed three times with 4°C sterile phosphate-buffered saline (PBS) and resuspended in 250 μL RPMI Phenol Red-free media to be loaded in the device at a final concentration of 4 × 10⁶ cells/mL.

Live cell imaging of MitoSOX Red

Once loaded in the device, cells were imaged using the TRITC (540/605 nm) filter cube (Nikon) and stimulated with antimycin A (Sigma-Aldrich) at various concentrations. Individual cells were tracked through time and a heat map was created to visualize the change in intensity, related to change in mitochondrial superoxide production, through time (Figure 5). Within a chamber with uniform antimycin A concentration, we observed individual cellular differences in mitochondrial superoxide production (Figure 4). With an irreversible dye, a monotonic increase is expected over the 70 min experiment. In several instances, lower fluorescence was observed with time; this may be due to rotation of the trapped cell with respect to the focal plane of imaging, photo bleaching of the dye, or inaccuracies in creation of the binary images. On average, we observed an increase in MitoSOX Red oxidation with increasing antimycin A concentration (Figure 4 a,c) as well as the number of cells that respond to the stimulatory condition as defined by an increase greater than 40% the original fluorescence. In the example shown, there were 7 responsive cells in the 1.6 μM inhibitor treatment compared to 16 responsive cells in the 50 μM inhibitor concentration.

Figure 5. Mitochondrial hydrogen peroxide production with antimycin A stimulation (blue arrow).

(a, c) 1.6 μM treated cells (c, d) 50 μM treated cells. (a, b) Heat map of normalized mean fluorescence of 25 analyzed cells through time. (c, d) Average and single cell traces of mean fluorescence.

Figure 4. Mitochondrial superoxide production with antimycin A stimulation (blue arrow).

(a, c) 1.6 μM treated cells (c, d) 50 μM treated cells. (a, b) Heat map of normalized mean fluorescence of 25 analyzed cells through time. (c, d) Average and single cell traces of mean fluorescence.

5. Imaging Mitochondrial Hydrogen Peroxide Production

Stable transfection of HyPer-Mito into Jurkat cell line

To visualize the dynamics of mitochondrial hydrogen peroxide through time, the pHyPer-dMito plasmid (Evrogen) was transfected into Jurkat cells using the Neon Transfection System (Life Technologies). Jurkat cells were cultured until logarithmic growth was observed and then washed in PBS without Ca²⁺ and Mg²⁺. Cells are resuspended in Resuspension Buffer R (Invitrogen Neon Kit) at a final concentration of 1 × 10⁷ cells/mL with 10 μg DNA per 100 μL transfection. Four 100 μL transfections were completed using the Neon protocol of 3 pulses of 1325 V with a 10 ms pulse width.

Once transfected, cells were cultured for 3 days without antibiotics. Selection was completed using the neomycin resistant gene on the pHyPer-dMito plasmid. On day 4, 1.4 mg/mL neomycin (G418) (KSE Scientific) was added to the media with a cell concentration of approximately 0.2 × 10⁶ cells/mL. The selection was continued for 14 days with washing and addition of fresh media and antibiotics every 3 days, maintaining the cell concentration between 0.2-0.6 × 10⁶ cells/mL. The concentration of G418 was calibrated via a cytotoxicity curve with the same lot of G418 (KSE Scientific). Following selection, a maintenance concentration of 0.6 mg/mL G418 was continued in cell culture.

Live cell imaging of HyPer-Mito

Cells were loaded into the device at a concentration of 4 × 10⁶ cells/mL. Once sufficiently loaded, antimycin A (Sigma-Aldrich) was introduced at various concentrations. Cells were imaged using the QMAX GR TE-10 filter set (Omega Optical), as recommended by Evrogen for non-ratiometric imaging of the Hyper reporter protein. The filter set for excitation between 450-490 nm and emission filter at 535 nm primarily represents the oxidized form of the mitochondrial hydrogen peroxide sensor, HyPer-Mito. Images were collected over the course of 70 min and reveal differences in hydrogen peroxide signaling between different cells under the same antimycin A concentration (Figure 5). As with the MitoSOX Red, we observed an increase in fluorescence intensity associated with oxidized HyPer-Mito as the concentration of antimycin A increased (Figure 5 a,c). With a reversible ROS reporter, however, the observed kinetics were more diverse at the single cell level. At the 1.6 μM inhibitor concentration, the cells that respond tend to do so similarly in a steadily increasing manner (Figure 5 a,c), whereas cells at the 50 μM concentration vary in both the time to respond, and whether the H₂O₂ is sustained or changes with time. Differences can also be seen in the number of cells to respond, as defined by a 40% change from baseline. In the example shown, there were 5 responsive cells in the 1.6 μM inhibitor treatment compared to 12 responsive cells in the 50 μM inhibitor concentration. The quantity of cells responding was lower than that observed with MitoSOX.

6. Conclusion

We have demonstrated the potential to use microfluidics to study time-dependent changes of two different mitochondrial ROS in single cells, thus supporting redox systems biology at the level of single cells. Further, we provide the ability to study heterogeneity of these processes in different cells within a population. We report that cells exposed to the same concentration of stimulus exhibited variation in mitochondrial ROS production, which could ultimately result in different cellular responses. This variation could be the result of stochastic processes within the cell, such as transcriptional or translational regulation of key components of the mitochondrial ROS pathway. Additionally, the nature of the responses differed between O ⁻₂ and H₂O₂ kinetics. Because two different types of fluorescent reporters (irreversible small molecule vs. YFP-fusion protein) were used, it is difficult to ascertain without further analysis whether variation at the single cell level is attributed to properties of the reporter or due to subcellular concentrations of the respective ROS. To our knowledge, this is the first description of the use of microfluidics to image and quantify single-cell redox states in mammalian T cells, a method that may help elucidate underlying dysfunctions in different cell types and diseases. Accounting for distributions of mitochondrial redox state within populations of cells will ultimately allow for a better understanding of signaling processes associated with diseases that implicate mitochondrial dysfunction, such as Alzheimer’s disease¹⁴, autoimmune disorders¹⁵, and cancer¹.

7. Potential Pitfalls

Device Operation and Image Analysis

As the device is operated, cells can shift slightly in position and location. A small number of cells have been observed to squeeze through the trap and either get stuck or leave the trap completely. With these discrepancies in cell position, the image analysis methods must address changing location. Currently, the first and last images are used to ensure all cell locations are recorded through time. However, a more optimized program may track the cell at every time point for mean fluorescence calculations.

Imaging with MitoSOX

MitoSOX provides an irreversible signal of oxidation and hence it is difficult to analyze the dynamic changes in superoxide production through time. One workaround with sufficiently resolved temporal imaging is to analyze the derivative of the signal intensities in order to visualize changes production rates. This approach is limited by the potential saturation (i.e. full oxidation) of the probe.

Imaging with HyPer-Mito

Transfection efficiency becomes an issue in difficult to transfect cell lines as the utility of the microfluidic device decreases when only few cells can be analyzed. Stable transfection with antibiotic selection was performed to avoid low transfection efficiency. Basal differences in the oxidized HyPer-Mito fluorescence intensity may be due to either transfection levels or degree of oxidation. Although HyPer-Mito is designed for ratiometric measurements with a dual excitation system (which would discriminate between these possibilities), we used a single excitation filter in the findings reported here. Future directions include expanding the microscope setting to contain an excitation wheel capable of dual excitation.