Progresses in cell-based assays and related areas of science are often guided by advancements in analytical protocols and instrumentation applicable to the analysis of individual cells or their organelles. For instance, developments in flow cytometric techniques (1, 2) have allowed rapid measurements and multiparametric analysis of various cell attributes from a large number of cells, which have led to widespread application of these techniques in many recent studies on cell proliferation and death.
Among various cell-based assays, the cell cycle analysis has an important role in both biological research and clinical applications (3, 4). The most common approach to determine the cell cycle phase of individual cell is by staining and quantifying their total DNA contents using flow cytometry. Currently, the majority of cell cycle assays are relying upon the flow cytometry method, in which cells are classified into G1, S, and G2/M phases based on their DNA content.
Although flow cytometry is considered as a well-established tool for cell cycle analysis in a single cell mode, this technique has several drawbacks. For example, high assay cost, low assay throughput, large sample volume, and lack of in situ monitoring capability have limited more widespread applications of this technique as a HT-HCS method (5, 6). Thus, for further enhancement of assay throughput and reduction of assay costs, miniaturization, integration, and automation of cell-based assays are urgently needed. Among the numerous efforts to improve efficiency of the current cell-based assays, the microfluidic device (μFD) is considered as one of the most promising approaches capable of improving current in vitro cell-based assays (2). Several recent studies have reported promising application examples of μFD, demonstrating their advantages in manipulating small volume samples and integrating multiple experimental steps (6, 7). For instance, morphology- (8), and MTT- (9) based microfluidic image cytometry (μFIC) techniques have demonstrated the capability of this technology as a potential HTS platform for simple, efficient, and in situ evaluation of the cell death process.
Here, we present our recent effort to develop, evaluate, and apply a high throughput cell cycle analysis method using fluorescence-based μFIC. A μFD for measuring cellular DNA content was designed, fabricated, and applied for the assessment of paclitaxel-induced cell cycle changes of HeLa cells. Throughout this study, the advantages of μFIC-based cell cycle analysis, as a simpler, cheaper, and faster method, were demonstrated, which should have significant implications as a future high throughput cell cycle analysis platform.
MATERIALS AND METHODS
Design and Fabrication of the Microfluidic Device (μFD)
The μFD used in this study was fabricated by using soft lithography technique. It was composed of two polydimethylsiloxane (PDMS) layers, which were chemically bonded on the top of a glass coverslip (24 mm × 60 mm) substrate. SU-8 2150 photoresist (Microchem, Newton, MA), spin-coated and photolithographically-patterned on a silicon substrate, was used as a cast for the top layer (for cell culture chambers). After patterning, masters were silanized by exposure to tridecafluiri-1,1,2,2-tetrahydrooctyl (trichlorosilane) (Sigma Aldrich, St. Louis, MO) and were positioned into Petri dishes (90 mm × 15 mm). Sylgard 184 PDMS prepolymer and curing agent (Dow Corning, Midland, MI) were thoroughly mixed in a 10:1 weight ratio. These mixtures were poured into casts positioned in a Petri dish and were used as the top layer. Next, the casts were degassed in a vacuum chamber for 10 min before being baked in an oven (60°C, 4 h). To prepare the bottom layer, a mixture of PDMS prepolymer and curing agent was poured on a Petri dish to produce an approximately 0.3 cm thick PDMS layer. After curing (60°C, 4 h), both PDMS layers were combined via oxygen plasma treatment (100 W, 0.2–1 mbar, 4 min; CUTE, Femto Science, South Korea). The combined PDMS layers were then cut with a razor blade, peeled from the master and trimmed to size. Holes were punched out of the PDMS using flat-tip needles to form fluidic connection ports. The combined PDMS layers as well as the glass coverslip (24 mm × 60 mm) were treated with oxygen plasma and finally bonded to each other. Before use, the μFD was filled with deionized water and sterilized by UV light for 10 min.
Cell Culture and Toxin Exposure Within the μFD
The μFD was initialized with media to remove dead volume and bubbles, then placed inside a humidified incubator (37°C with 5% CO2, Forma Scientific, Waltham, MA) at least 6 h prior to use. The HeLa cell line was purchased from KBRC (Korea Biological Resource Center) and cultured in fresh DMEM medium (Gibco, Grand Island, NY) containing 10% fetal bovine serum (Gibco, Grand Island, NY) and 1% penicillin-streptomcyin (Gibco, Grand Island, NY). A cell suspension prepared at a concentration of 104–105 cells/mL was introduced into the loading port of the μFD, and the seeded μFD was placed into a humidified incubator (37°C with 5% CO2) for 12 h, after which the media was changed two to three times per day. This cell culture cycle was repeated several times until the appropriate numbers of cells were achieved within the μFD. Because of the gas permeability of the PDMS, oxygen diffusion was sufficient for cell culture, even though most of the cell culture were performed under static-flow conditions.
Generation and Exposure to Toxin Gradient
Cytotoxicity experiments were performed at least in triplicates by establishing a concentration gradient of paclitaxel within the μFD for 12 h. One inlet was supplied with culture media, while another was fed with a 5 or 10 nM paclitaxel solution in cell culture media. Concentration gradients of paclitaxel were formed inside the multiple compartment chambers using a Concentration Gradient Generator (CGG). The paclitaxel concentrations formed in each compartment was estimated from the previously measured absorption profile of trypan blue (4%, Gibco, Grand Island, NY), which has a comparable molecular weight (i.e., diffusion coefficient) and high absorption coefficient. During the toxin exposure period, the μFD was placed inside a CO2 incubator with controlled humidity and temperature.
Cell Cycle Analysis
For the analysis of cell cycle, the following fixation protocol was used. The medium was carefully aspirated from the μFD and was replaced with 70% ethanol in PBS (−20°C). Cells were stored in 70% ethanol at −20°C for 30 min, then the device was washed twice with PBS. After the second wash, the PBS was removed by aspiration, and 200 μL of RNAse solution (0.1 mg/mL, DNase free) in PBS was added to the device. The device was then incubated for 1 h at 37°C. The RNAse solution was aspirated off the device and replaced with a 1.5 μM propidium iodide (PI) in PBS. The device was then sealed with a black cover-seal and incubated in the dark for 10 min at room temperature.
Image Acquisition and DNA Content Analysis
Fluorescence images (100×) were acquired for each compartment using an inverted type fluorescence microscope (IX51, Olympus, Japan) equipped with a cooled CCD camera (Retiga-2000RV 12-bit Mono Fast 1394 Cooled, QImaging, Canada). The acquired fluorescence images were further analyzed to determine the number of cells as well as individual cellular information (e.g., integrated density parameters of individual cells) using Java-based image processing and analysis software (ImageJ 1.41n, http://rsb.info.nih.gov/ij, Wayne Rasband, NIH, USA). For automated processing of the multiple images obtained from each compartment, an ImageJ macro, involving cell segmentation, counting, and integrated density (IntDen) measurement functions, was made and applied to extract IntDen from each fluorescence image. Among the measured parameters, the IntDen of each cell was used for further cell cycle analysis. Here, the IntDen of each cell reflected total DNA content. Using this parameter, all of the single cells in the device were classified into the different phases of the cell cycle.
The design of the μFD used in this study is shown in Figure 1A. The device consisted of two inlets, a CGG, eight straight channels for cell culture, and an outlet used for cell loading and waste removal. As shown in Figure 1B, concentration gradient of paclitaxel was generated via laminar flow and diffusive mixing of media and toxicant streams in a CGG, which was composed of six successive steps of mixing to ensure complete mixing of toxicant (10). Paclitaxel concentrations at each channel were estimated from the previously measured optical density profile of trypan blue, which had a comparable molecular weight and diffusion coefficient with paclitaxel. The concentration gradient at the end of the CGG was found as linear, as shown in Figure 1B. The media, each containing eight different levels of paclitaxel, were fed into each cell culture channel. Because of these capabilities to form linear concentration gradients via tree-shaped microfluidic network design, the needs for labor-intensive manual operations or high cost robotic-based HTS systems were significantly reduced. The cell cycle assay protocol used in this study (see the schematic diagram in Fig. 1C) was based on the typical μFIC assay protocol (4) combined with a PI staining method for cell cycle analysis in flow cytometry approach. Cell culture time for the μFIC assay was optimized as 48 h, during which monolayers of adherent cells were formed with a minimum population of floating cells and intercellular overlaps. Fixation, RNAse treatment, and PI staining were followed by fluorescence image acquisition and quantification of cellular DNA content.
As shown in Figure 2, we have performed systematic comparisons of image analysis results for channel “a” to find optimum conditions for image-based cell cycle analysis, such as the minimum number of cells with reasonable sampling statistics and appropriate sampling positions in cell culture channels. Further fitting analyses were also performed and the goodness of fit parameters (e.g., R2) and the peak positions for 2n and 4n, were plotted as a function of analyzed cell numbers (Supp. Info. Fig. S1A) and positions along the channel (Supp. Info. Fig. S1B). While ∼ 5000 cells were observed per channel (i.e., > 300 cells per image and 12 images per channel), fitting analysis with ∼ 1000 cells seems to provide reasonable results and those with ∼ 2500 cells provided fitting results with reasonable statistics (see Fig. 2A and Supp. Info. Fig. S1A). In addition, no significant changes in the 2n and 4n peak positions were observed along the channel, indicating that most of the regions in cell culture channels can be used for cell cycle analysis (see Fig. 2B and Supp. Info. Fig. S1B). Therefore, eight images from the central regions of the channels (image no. 3–10, correspond to ∼ 2500 cells) were chosen for the cell cycle analysis. Moreover, the volume of the channel used for these analyses corresponds to only 450 nL, whereas typical FACS analysis requires at least 500 μL. This reduction in sample volume should also contribute to μFIC-based HTS assays by facilitating integration and parallelization of μFIC assays.
Figure 3 displays histograms of cellular DNA distributions and PI fluorescence images of representative cells at different stages of the cell cycle. The first image and histogram shown in Figure 3 corresponds to the control channel (channel “a”, [Paclitaxel] = 0.0 M) of the microfluidic device. Under this condition of control channel, cells repeatedly cycled between G1 and G2/M phases, with a dominant contribution from the G1 phase. Thus, the histogram is comprised of two major peaks located at 2n and 4n levels of cellular DNA contents that correspond to the cells in the G1 and G2/M phase, respectively. The nuclei of cells in the G1 phase were mostly oval in shape with lower PI fluorescence intensities (see the inset image of Fig. 3A), whereas cells in the G2/M phase display much higher PI fluorescence intensities (see inset image of Fig. 3C). However, upon exposure to a low level of paclitaxel (channel “c”, [Paclitaxel] = 1.53 nM), the appearance of new peaks in the sub-G1 phase and a decrease of the G1 peak were observed (see the histogram of Fig. 3B). The corresponding PI fluorescence image (see the inset image of Fig. 3B) also displayed overlapped nuclei within a single cell in sub-G1 phase, with significant changes in nuclear sizes and PI fluorescence intensities. Fragmented nuclei were observed to have relatively low cellular DNA contents according to the associated histograms. The histogram shown in Figure 3C indicated that most of the cells were arrested at the G2/M phase when the cells were exposed to higher levels of paclitaxel (channel “g”, [Paclitaxel] = 4.48 nM). Cellular DNA contents under this phase were approximately twice that of the G1 phase, indicating that the progress in cell cycle was arrested by paclitaxel during mitosis period.
Figure 4 shows the effects of paclitaxel concentrations on the progress of the HeLa cell cycle. As demonstrated in Figure 4A, each histogram was deconvoluted with three components (e.g., G1, G2/M, and sub-G1 phases), and paclitaxel dose-dependent variations of each component were plotted in Figure 4B. Similar to the previous observations (11, 12), the deconvolution results showed that, as the paclitaxel dose increased (<∼ 1.5 nM), the number of cells in the G1 phase rapidly decreased, whereas the number of cells in the G2/M phase slowly increased at doses of paclitaxel > 3 nM. Interestingly, the number of cells in the sub-G1 phase, which represents multinucleated cells with low cellular DNA content, were found mostly at the intermediate dose range of paclitaxel (1–3 nM). In addition, the effect of paclitaxel exposure time on HeLa cell cycle were also investigated and presented in Figure 5. Compared to the results shown in Figure 4B (i.e., 12 h exposure), Figure 5 showed paclitaxel dose-dependent variations of each component after a short (1 h) period of exposure. As shown in Figure 5A, when the cell cycle was analyzed immediately after paclitaxel treatment, no significant changes were observed in the relative proportions of the three cell cycle components. However, significant changes were observed when HeLa cells were further incubated with paclitaxel-free media for 10 h after the 1 h exposure to paclitaxel. As shown in Figure 5B, the contribution of the G1 phase continuously decreased, while the contribution from the sub-G1 phase showed a proportional increase with paclitaxel concentration.
In this study, we designed and fabricated a μFD combined with fluorescence-based image cytometry for measuring cellular DNA contents. To test feasibility of this fluorescence-based μFIC approach for cell cycle analysis, we evaluated and applied this novel technique for the assessment of paclitaxel-induced cytotoxicity of HeLa cells. Paclitaxel is a widely known anticancer drug with significant antitumor activity for the treatment of ovarian, lung, and breast cancers. In previous studies using flow cytometry (11–13), it is already well known that the paclitaxel binds and stabilizes microtubules and causes arrest of the cell cycle at the mitotic phase, which has been considered as the cause of paclitaxel-induced cytotoxicity. In addition, the mechanisms of paclitaxel-induced cell death are known as dose dependent. At high dose of paclitaxel, apoptotic cell death is induced by the blockage in the G2/M phase, while abnormal mitosis with the formation of multinucleated cells is observed for the cell death at high dose of paclitaxel (13). Since these dose dependent phenomena have been thoroughly investigated and previously confirmed using flow cytometry (11–13), we have selected paclitaxel as a reference compound of the feasibility test using fluorescence-based μFIC cell cycle analysis method.
As shown in Figures 2 and 3 and Supporting Information Figure S1, the fluorescence-based μFIC method was optimized for the cell cycle analysis. In addition, it was also demonstrated in Figures 4 and 5 that the μFIC was able to provide comparable cell cycle analysis results with flow cytometry. For instance, dominant appearance of sub-G1 phase at the intermediate dose (1.5–2.0 nM) as well as the dose dependent decrease in G1 (≤1 nM) phase and increase in G2/M (>3 nM) phase were observed in this study, indicating that the cell cycle arrest of HeLa cells is strongly dependent on the paclitaxel dose. Except that the changes in FCM results were observed at higher concentrations than current μFIC, the observed trends in μFIC were in good agreement with the previous results reported by Torres and Horwirtz (13), who observed similar paclitaxel dose-dependent responses of A549 cells with flow cytometry and suggested that paclitaxel-induced cell death can be ascribed to two different apoptosis pathways. The observed μFIC cell cycle changes at lower concentration of paclitaxel is mainly due to the differences in the exposure condition (e.g., continuous flow condition in μFIC and static exposure condition in FCM). Under the continuous flow conditions of μFIC device, fresh media solutions with constant concentrations of paclitaxel were continuously supplied, therefore, similar changes in cell cycle can be achieved at lower concentrations of paclitaxel under continuously flowing exposure condition. In contrast, cellular accumulation or consumption of paclitaxel under static exposure condition will result in reduced concentrations of paclitaxel around the cells, which will require higher concentrations of paclitaxel to induce similar cell cycle changes in FCM. Moreover, they also reported that, for those cells treated with low concentration of paclitaxel, cell death occur via an aberrant exit from abnormal mitosis and cause formation of multinucleated cell, while cells exposed to higher paclitaxel dose undergo terminal mitotic arrest via a Raf-1 dependent pathway. Therefore, these observations by Torres and Horwirtz (13) agree very well with the results from current μFIC study, which confirmed the compatibility of μFIC results with those from conventional flow cytometry.
In addition, the observations in Figure 5 suggest that a short exposure of paclitaxel can also induce mitotic arrest, although it requires a certain amount of lag time to induce changes in the cell cycle. Consequently, the response of HeLa cell appears to be strongly dependent on the exposure time as well as the concentration of paclitaxel. Similar to our observations, it was previously reported by Wang et al. (11) that a short period (1 h) of exposure to a low concentration (20 nM) of paclitaxel induces a reversible mitotic block without apoptosis, which indicate that paclitaxel at this exposure condition may only inhibit formation of mitotic spindles, resulting in G2/M phase arrest. Moreover, it was also noteworthy that the variations of each component observed after the 1 h exposure to a relatively high paclitaxel dose (less than 10 nM, see Fig. 5B) displayed some similarity to the results observed after a 12 h exposure to a very low paclitaxel dose (less than 1 nM, see Fig. 4B). Repeated observations of these phenomena in triplicated experiments confirmed that the response of the HeLa cell cycle was mainly dependent on the total exposure or accumulation of paclitaxel (i.e., dose × time).
In this study, we have demonstrated that the cell cycle analysis using fluorescence-based μFIC was able to provide comparable data with those of flow cytometry. Moreover, cell cycle analysis using μFIC can also provide further advantages over flow cytometry, such as higher throughput, lower assay cost, less generation of toxic waste, etc. First of all, automatic generation of concentration gradient allowed us to perform multiple assays in a single μFD and saved large amount of labor as well as time. In the current lab-based μFIC system designed for “proof of concept”, there were only eight channels with different drug concentrations. However, the number of channels following the CGG can be further increased in future commercialized μFIC system to conduct high-throughput assays without involving sophisticated manual operations. In addition, adoption of the tree-shaped microfluidic network design (10) provided more stable and linear concentration gradient, which is necessary to improve reliability and reproducibility of the μFIC assays. Second, the cell culture channels in this μFD were having much smaller volume (450 nL) compared to the typical sample volume (500 μL) used in flow cytometry. Thus, this more than 1000 times miniaturized chip design helped us to minimize the consumption of expensive chemicals as well as the production of toxic waste, which also contributed in reducing assay cost. Thirdly, the combination of fluorescence-based image cytometry with the μFD also provided valuable cellular information, such as those on the morphological changes and the localized staining of cells, which are not available in typical flow cytometry. Recently, we have demonstrated these advantages of μFIC as an image-based, multi-parametric cytometry approach (8, 9), where morphology- (i.e., circularity) and MTT absorbance- based analysis were applied for the multi-parametric assessment of cytotoxicity. In this study, following previous studies on μFIC (8, 9), we have expanded the capabilities of μFIC to the fluorescence-based μFIC for DNA content analysis. Similar to the previous multi-parametric μFIC analysis study by Lim et al. (9), combinatorial analyses of cellular responses, such as changes in cellular morphology and DNA content, are also possible using this μFIC system, which indicate the potentials of this μFIC system as a future high-content analysis (HCA) platform. Finally, since the artifacts generated during the cell detachment/transfer process of flow cytometry were recognized as one of the drawback of this method, the capability of the current μFIC-based assay with all of the treatments, measurements, and analyses of adherent cells with minimal disturbances also helped us to overcome such drawback of flow cytometry.
As previously mentioned, the μFIC used in this study is a laboratory-based platform with inverted microscope, cooled CCD camera, and PDMS-based microfluidic chips; however, due to the recent advancements in instrumentations, low-cost benchtop image-based cytometers are now commercially available at reasonable prices. Thus, we expect that the μFIC system can be easily incorporated into the above low-cost image-based cytometers with built-in light sources (e.g., LED), integrated optics, less expensive imaging detector, and mass-produced plastic chips. Therefore, in its future commercialized form, the μFIC can be a cost-effective platform with many other advantages over current flow cytometry, such as simpler, and faster assays with better resolved cellular information, while requiring smaller sample volume and generating less toxic waste. These advancements in μFIC system should have significant implications in pharmaceutical and biological studies.
This work was financially supported by the Nano R&D program through the Korea Science and Engineering Foundation (KOSEF), which was funded by the Ministry of Education, Science and Technology (MEST) (2008–02968). We appreciate Prof. Jeong-Hun Sohn of Chungnam National University for providing paclitaxel for this study.