Article pubs.acs.org/ac
Study of Phospholipids in Single Cells Using an Integrated Microfluidic Device Combined with Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Weiyi Xie,†,‡ Dan Gao,*,‡,§ Feng Jin,¶ Yuyang Jiang,‡,∇ and Hongxia Liu*,‡,§ †
Department of Chemistry, Tsinghua University, Beijing 100084, China State Key Laboratory Breeding Base-Shenzhen Key Laboratory of Chemical Biology, Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, China § Key Laboratory of Metabolomics at Shenzhen, Shenzhen 518055, China ¶ Neptunus Pharmaceutical Technology Center, Shenzhen 518057, China ∇ School of Medicine, Tsinghua University, Beijing 100084, China ‡
S Supporting Information *
ABSTRACT: Single-cell trapping and high-throughput mass spectrometry analysis remain challenging now. Current technologies for single-cell analysis have several limitations, such as throughput, space resolution, and multicomponent analysis. In this study, we demonstrate, for the first time, the combination of microfluidic chip and matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) for high-throughput and automatic single-cell phospholipids analysis. A microwell-array-based microfluidic chip was designed and fabricated for cell array formation on an indium tin oxide (ITO)-coated glass slide. Mass spectrometry imaging measurement with 25 μm pixel size was performed with a MALDI ion source. Eight phospholipids in a single A549 cell were detected, and their structures were further identified by MS/MS spectra. Selected ion images were generated with a bin width of Δm/z ± 0.005. The selected ion images and optical images of the cell array showed excellent correlation, and mass spectrometry information on phospholipids from 1−3 cells was extracted automatically by selecting pixels with the same fixed interval between microwells on the chip. The measurement and data extraction could be processed in several minutes to achieve a high-throughput analysis. Through the optimization of different microwell sizes and different matrices, this method showed potential for the analysis of other metabolites or metabolic changes at the single-cell level.
I
life sciences,11,12 and it has been regarded as a promising technique for single-cell analysis.13 Until now, different strategies have been established for single-cell capture, such as the droplet method,14,15 microwells,16−18 microvalves,19,20 and microtraps.21,22 Microfluidics allow an efficient, highthroughput and well-organized isolation and docking of single cells, which provide convenience for subsequent treatments and measurements. Until now, the most widely used online detection techniques for single-cell analysis are fluorometry23,24 and spectroscopy.25,26 The fluorescence microscope technique relies on the molecular probes or reporters that need to be preselected, and the components that can be detected by fluorescence analysis are also limited, because of the restricted range of fluorescent dyes. Meanwhile, spectroscopy analysis
t is well-established that individual cells, even from the same origin, differ from each other in many aspects, because of stochastic biological processes and differences in environmental perturbations.1 Since population-level measurement techniques inherently average out the property differences between individual cells, the underlying biochemical mechanisms cannot be revealed by these measurements.2 Moreover, the ignored differences between each cell may be of great importance in several research fields, such as early diagnosis,3 treatment of diseases,4 and screening of effective drugs.5 Single-cell analysis has attracted increasing interest during the past few decades.6 However, new analytical strategies with an ultrahigh detection sensitivity and high throughput at the single-cell level are still urgently needed to be developed to explore the heterogeneity between cells.7,8 The traditional methods for single-cell capture, such as microdissection9 and micromanipulators,10 often require complex manual manipulation. In recent years, microfluidics has become a widely used technology for various applications in © XXXX American Chemical Society
Received: January 3, 2015 Accepted: June 25, 2015
A
DOI: 10.1021/acs.analchem.5b00010 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 1. (A) Partial scheme of three-layer microfluidic chip; inset shows the optical image of the porous membrane. (B) Schematic diagram of the microfluidic chip fabrication via an imbalanced cross-linking ratio bonding method. (C) Schematic illustration of single-cell capture and cell array detection by MALDI-MS.
cell analysis approaches with direct MALDI-MS detection need to be developed. For MALDI-MS analysis, plenty of compounds were used as the MALDI matrix, such as 2,5dihydroxybenzoic acid (DHB), α-cyano-4-hydroxy cinnamic acid (CHCA), and 9-aminoacridine (9-AA), and various species (such as peptides,33 lipids,27,34 and other metabolites35) have been successfully detected, upon which lipids were mainly focused. Lipids are regarded as one of the most popular targets in single-cell research, because of their high concentrations in cell membrane,36 as well as their significant roles in cell functions, including growth and transformation,37 apoptosis,38 and phagocytosis.39 Moreover, variations in lipid metabolism indicated several disease states,40 which suggested that lipids could be used as biomarkers of diseases. Much research about the detection of phospholipids in single cells using MALDI-MS has been reported. Nevertheless, high-throughput MALDI-MS
cannot achieve a direct analysis of single cells, which was incapable of detecting specific components. Recently, mass spectrometry (MS)-based single-cell analysis methods have been attracting increased interest, because of their label-free detection and their ability to detect unknown molecules. Among these analytical methods, matrix-assisted laser desorption ionization mass spectrometry27−29 (MALDI-MS) has been demonstrated suitable for providing both the chemical and spatial information inside single cells. Direct analysis of cellular different species at the single-cell level using MALDI-MS have been reported, such as single yeast cell30 and mammalian cell.31 MALDI mass spectrometry imaging (MALDI-MSI) was also used to determine in situ both the distribution and relative abundance of specific components in a single cell.8,32 However, the cell should be manually selected, which can lead to extremely low throughput. Therefore, high-throughput singleB
DOI: 10.1021/acs.analchem.5b00010 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
medium was then added on the captured cells and incubated in the CO2 incubator for another 30 min. After incubation, the medium was removed and the cell array was washed twice with pure water within 2 s. The schematic illustration of cell capture and cell array formation on the chip is shown in Figure 1C. The Live/Dead assay reagent containing final concentrations of 2 μM Calcein AM and 4 μM ethidium homodimer-1 was used to evaluate the viabilities of cells on the ITO-coated glass slide. The cell array was incubated with the Live/Dead assay kit at 37 °C for 30 min and then washed with fresh PBS for three times. The images of the array were observed under an inverted fluorescence microscope (Model Olympus IX5, Olympus America, Inc., Center Valley, PA, USA) at an excitation wavelength of 450−490 nm. Lipid Extraction and MALDI-MS Detection. The chloroform−methanol system for lipid extraction reported by Folch et al.43 was used in our experiment, and the detailed procedures are presented in the Supporting Information. The lipid extraction sample for MALDI-MS analysis was prepared via a dried droplet method, using 9-AA or DHB as the MALDI matrix. Two matrix solutions were prepared in advance: 5 mg/ mL 9-AA in 70:30 ethanol/water, and 20 mg/mL DHB in TA30 (water:acetonitrile = 70:30, v/v) with 0.1% trifluoroacetic acid (TFA). 0.5 μL of lipid extraction sample and 0.5 μL of matrix solution were premixed, pipetted on the MALDI Anchorchip standard target plate, and allowed to dry for further MS analysis. MALDI mass spectra were acquired on an UltrafleXtreme MALDI-TOF/TOF instrument (Bruker Daltonik, Inc., Billerica, MA, USA) with a 355 nm and 200 Hz solidstate Smartbeam Nd:YAG UV laser (Azura Laser AG, Berlin, Germany). Calibration was performed with CHCA and Peptide Calibration Standard II with the m/z range of 100−1500. Mass spectra were acquired over the mass range from 100 to 1500 in the reflective positive-ion mode. Acceleration voltage was set at 25 kV, and deflection up to m/z 100 was used to suppress matrix and other low-molecular-weight signals. MS/MS experiments were also carried out to help identify the potential structures of phospholipids. The MS/MS spectra were acquired in positive LIFT mode and in the mass range of 100−1500 with a “small” focus setting Smartbeam II laser. No matrix suppression was conducted in MS/MS measurement. MALDI-MSI Analysis of Phospholipids in Cell Array. For the MALDI-MSI analysis of phospholipids in the cell array, the cells on the ITO coated glass slide were first washed with PBS and deionized water separately. The cells then were covered with 2 μL of 9-AA solution (10 mg/mL in 70:30 ethanol/water) for MALDI-MSI analysis. Before analysis, mass spectra were also calibrated using CHCA and Peptide Calibration Standard II. For imaging analysis, a “small” focus setting Smartbeam II laser was used, and the step size of the sample stage was set to 25 μm. A total of 1000 laser pulses were summed for each mass spectrum. Mass spectrometric measurements were performed in an automatic fashion in reflective positive-ion mode and in the mass range of 100−1500. The acquired mass spectra were processed using the software FlexImaging to generate MSI images of specific ions. Selected ion images were created with a bin width of Δm/z ±0.005. Intensity values in ion images were normalized to the highest intensity measured for each ion species separately. No other post-processing steps such as interpolation or normalization to matrix signals were applied to the images, to demonstrate the original data quality.
analysis and data acquirements from single cells could not be realized. Here, we developed a high-throughput and label-free singlecell phospholipids analysis method, using a microwell-based microfluidic device and MALDI-MS combination system to conquer the above disadvantages. Several functions including single cells capture, cell array formation, cell culture and sample pretreatment prior to MALDI-MS analysis could be integrated on the device, which makes it possible to efficiently capture cells and process MALDI-MS measurement. The dimensions of the microwells and MALDI-MS detection conditions were optimized to acquire an optimal cell capture efficiency and optimal mass spectrometric signals. Under the optimized conditions, 8 phospholipids in 1−3 cells were automatically and quickly measured by MALDI-MSI in the positive-ion mode. Selected ion images were generated with a bin size of Δm/z = 0.005. Mass spectrometry information on phospholipids were acquired and extracted from MALDI-MSI results with high throughput, and the heterogeneity between cells was revealed by phospholipids signals in 1−3 cells. Compared to the existing approaches for single-cell analysis with MALDIMS, our established system not only provides automatic and high-throughput analysis ability, but also has the potential to expand to involve a broader range of analytes.
■
EXPERIMENTAL SECTION Microfluidic Device Fabrication. The PDMS microfluidic chip was fabricated by standard soft lithographic and replica molding techniques.41 The three-layer structure of microfluidic chip and the schematic diagram of the microfluidic chip fabrication are shown in Figure 1A and Figure 1B, respectively. Briefly, two silicon wafers were used to fabricate the microfluidic device. In addition, an imbalanced cross-linking ratio bonding method was used to bond the two PDMS layers.42 For the top PDMS layer, a flat silicon wafer without any mold was used. PDMS prepolymer with an excess of crosslinker was poured on the wafer, cured at 65 °C for 30 min, and carved into frames with fixed sizes. For the middle layer, a 60μm-height post array was fabricated on a silicon wafer. PDMS prepolymer with a deficiency of cross-linker was spin-coated on the wafer and cured at 65 °C for 10 min to form a PDMS porous membrane. The top layer with fixed sized frames then was aligned to the porous membrane. After curing overnight, the porous membrane was peeled from the top layer, and the two PDMS layers were attached onto an ITO-coated glass slide to fabricate a three-layered microfluidic device for cell capture. Cell Culture and Device Operation. A549 cells were cultured in Ham’s F-12K (Kaighn’s) medium containing 10% fetal bovine serum (FBS), 100 μg/mL penicillin, and 100 μg/ mL streptomycin. Cells were maintained at 37 °C in a 5% CO2humidified air atmosphere and passaged every 2 or 3 days. For the formation of a cell array on a microfluidic device, the chip was processed in oxygen plasma for 2 min to improve the hydrophilicity and then exposed under UV light for 30 min to sterilize. Approximately 10 μL of A549 cell suspension (1 × 108 cells/mL) in F-12K culture medium was injected into the microfluidic chip and mixed using a pipet. After confirming that a majority of the cells were captured inside the microwell arrays, the redundant cells were replaced by F-12K medium with 20% FBS. The microfluidic chip with captured cells was then placed in a 5% CO2 incubator at 37 °C for 6 h. After cell adhesion, the PDMS portion was peeled off from the ITOcoated glass slide. Two hundred microliters (200 μL) of F-12K C
DOI: 10.1021/acs.analchem.5b00010 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 2. (A) Workflow of cell capture and cell array formation. (B) Fluorescence images of cell array formed by 40 and 50 μm microwell-based microfluidic devices after cell adhesion on ITO-coated glass slides. Cells were stained by the Live/Dead assay kit for viability assay. (C) The percentage of 0−4 cells occupancies in the 40 μm microwell array. There were no more than four cells in each position of the array.
■
diameters of 30, 40, and 50 μmwas also tested. In a 50 μm microwell, ∼10 cells were captured; in a 40 μm microwell, ∼6 cells were captured; and in a 30 μm microwell, 1−2 cells were captured. After capture, the microfluidic chip was incubated in an incubator at 37 °C in a 5% CO2-humidified air atmosphere. Because of the limited space in the microwell, only a few cells could be attached on the glass slide, while other cells in the microwell were attached on the side face of PDMS. The microwell-based PDMS membrane was then peeled from the glass slide with caution. All cells that attached on the PDMS side face were also removed in this procedure, whereas cells that attached onto the ITO-coated glass slide were left on the ITO surface. In the 30 μm microwells, cells were more likely to attach on the PDMS part during cell culture, because of the similar size of microwell and cell, which led to a failure in forming the cell array after PDMS removal (data not shown). However, the overall cell occupancy increased significantly in the 40 and 50 μm microwell array (Figure 2A). The 40 μm microwell showed a better result, in that >90% of all of the available positions were occupied by 1−4 cells (Figure 2B). The final single-cell capture efficiency (i.e., the percent of microwells filled with cells) was ∼25% (Figure 2C). After cell array formation, the cells were stained by a Live/Dead assay kit to test their viabilities. The fluorescence images indicated that cell viability was not affected after incubation in the microwells for 6 h (see Figure 2B). After PDMS removal, a small amount of culture medium was dropped on the array and incubated under the same conditions for another 30 min. Operations must be very careful to avoid drift of cells on the glass slide. This procedure could strengthen the adhesion between cells and glass slide, as well as improve the hydrophilicity of glass slide surface for matrix deposition. After incubation, the culture medium was removed and the cells were washed using pure water within 2 s, instead of buffer solution, to avoid salt crystallization on the sample. Matrix Screening for Phospholipids Detection. Matrices differ in the amount of energy they impart to the molecules during desorption and ionization and hence the degree of fragmentation that they cause.44 Therefore, the
RESULTS AND DISCUSSION Cell Capture and Array Formation on the Microfluidic Chip. We fabricated a well-established microfluidic chip with three layers which integrated functions of cell capture, culture, and matrix deposition for MALDI-MS analysis. The top layer was a PDMS frame in which cell suspension could be filled and flowed to achieve a sufficient contact with the microwells. The middle layer was a PDMS membrane with microwell arrays. The microwell was ∼40 μm in depth and 30, 40, or 50 μm in diameter, respectively. The interval between the adjacent microwells was 100 μm, yielding a feature density of 100 microwells per square millimeter. The bottom layer was an ITO-coated glass slide, which was attachable for cells and could be used as a MALDI target for MALDI-MS analysis. The fluid could be easily injected into or removed from the microfluidic device with pipet from the top layer of the device. A549 lung cancer cell line was used to conduct the experiment due to its moderate size and strong attachment to the substrate. For cell capture on the microfluidic device, 10 μL of the A549 cell suspension was introduced into the chip to completely cover the surface of the microwell array. Because of the hydrophobicity of PDMS, cell suspension could hardly fill into the microwells. Therefore, an oxygen plasma treatment was performed prior to improve the hydrophilicity of the microwells to help the cell load. After this treatment, the cell suspension could be quickly filled into the microwells within 10 min. Cell suspension concentration ranging from 1 × 105 cells/ mL to 1 × 108 cells/mL were tested on a microwell array of 40 μm in diameter and 40 μm in depth to acquire a high-efficiency cell docking, especially the single-cell occupancy (data shown in Figure S2 in the Supporting Information). As a result, majority of the microwells could not be filled with cells when the cell loading concentration ranged from 1 × 105 cells/mL to 1 × 106 cells/mL, while the overall occupied microwells with one to four cells inside could reach up to 90% when the cell concentration increased to 1 × 107 cells/mL and 1 × 108 cells/ mL. Therefore, the cell suspension with a concentration of 1 × 107 cells/mL to 1 × 108 cells/mL was selected for the following experiments. In addition, the microwell sizewith different D
DOI: 10.1021/acs.analchem.5b00010 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
detect phospholipids in a single cell, and the result is shown in Figure S3 in the Supporting Information. Few signals were detected in negative-ion mode, which might be caused by the lower amounts of negative-ionized phospholipids in a single cell. The results indicated that positive-ion mode was more suitable for phospholipids detection. Therefore, 9-AA was finally chosen as a proper matrix for phospholipids identification and MALDI-MSI analysis in positive-ion mode. Phospholipids Identification. Because of the low absolute amounts of phospholipids at the single-cell level, acquiring the corresponding MS/MS spectra was very difficult. Therefore, the phospholipid structures were identified by additional MS/MS measurement of lipid extraction samples from a dense cell population. As shown in Figure 4A and Figure 4B, the mass spectrum obtained from lipid extractions showed a good alignment with that from a single cell and had complete coverage in the mass range of phospholipids, as well as a much higher mass spectrometry signal intensity. It is indicated that the identification of phospholipids using lipid extractions was credible. Besides, mass spectrum and MS/MS fragment measurement could provide reliable structural identification. A characteristic fragment ion of phospholipids (m/z 184) appeared in all measured MS/MS spectra (as shown in Figure S4 in the Supporting Information). By searching the LIPID MAPS database, it was obvious that eight different PC species were identified in the positive-ion mode, namely, PC(30:0) (m/z 706.5), PC(32:2) (m/z 730.4), PC(32:1) (m/z 732.4 and 754.4), PC(32:0) (m/z 734.5 and 756.4), PC(34:2) (m/z 758.5 and 780.5), PC(34:1) (m/z 760.5 and 782.5), PC(34:0) (m/z 784.5), and PC(36:2) (m/z 786.5 and 808.5). Details of all detected phospholipids are shown in Table 1, and all MS/MS spectra are shown in Figure S4 in the Supporting Information. The mass spectrum of a representative phospholipid PC(34:1), and its structural elucidation, are shown in Figure 4C. All eight identified phospholipids could be detected both in cell lipid extractions and in single cells, which indicated that MALDIMSI of these phospholipids in the cell array is feasible. High-Throughput MALDI-MS Analysis of the Cell Array. The cell array formed by the 40 μm microwell chip was used for MALDI imaging. Optical images of the array were acquired after matrix deposition. As shown in Figure 5A, matrix was distributed uniformly on the ITO-coated glass slide. Matrix
sensitivity difference is significant between matrices for disparate analytes. Several matrices had been used for phospholipids detection, in which DHB27 and 9-AA45 were reported as the most specific and efficient ones. In our experiment, both matrices were directly applied on the ITOcoated glass slide with or without cell arrays and measured using MALDI-MS. Representative positive ion mass spectra of matrix and matrix with a single cell are shown in Figure 3. As
Figure 3. Mass spectra of matrix and matrix with a single cell in the positive-ion mode. Mass spectrum of (A) 9-AA, (B) a single cell using 9-AA as the matrix, (C) DHB, and (D) a single cell using DHB as the matrix.
shown in Figure 3A and Figure 3C, 9-AA presents a weaker background interference than DHB. As shown in Figure 3B and Figure 3D, several peaks in the mass range of 700−900 were detected when using the DHB or 9-AA as the MALDI matrix. However, more components in the mass range of 600−700 were detected in samples using 9-AA as the MALDI matrix (Figure 3B). Considering that most phospholipids in cells were in the mass range of 600−900, the result indicated that 9-AA was more appropriate for phospholipids detection. In order to choose proper ionization mode for MALDI-MSI, 9-AA was tested on both positive-ion mode and negative-ion mode to
Figure 4. MS and MS/MS spectra of a single cell or cell lipid extraction: (A) mass spectrum of a single cell, (B) mass spectra comparison between a single cell (red) and cell lipid extraction (blue), and (C) a typical spectrum of a phospholipid PC(34:1). E
DOI: 10.1021/acs.analchem.5b00010 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry Table 1. Identification of Phospholipids by MS/MS Fragment Analysis compound
measured m/z
calculated m/z
ion form
fragment ion m/z
PC(30:0)
706.4606
706.5381
[M+H]+
184.05, 523.29, 603.10
PC(32:2)
730.4339
730.5387
[M+H]+
184.04, 476.24, 547.39
PC(32:1) PC(32:1)
732.4327 754.4368
732.5543 754.5357
[M+H]+ [M+Na]+
184.04, 467.16, 494.28 184.00, 549.48
PC(32:0) PC(32:0)
734.5022 756.4293
734.5694 756.5514
[M+H]+ [M+Na]+
184.06, 467.17, 625.27 184.00, 467.09, 625.32
PC(34:2) PC(34:2)
758.4675 780.4504
758.5700 780.5514
[M+H]+ [M+Na]+
184.00, 504.33 184.09, 504.41
PC(34:1) PC(34:1)
760.4856 782.4670
760.5856 782.5670
[M+H]+ [M+Na]+
184.04, 478.25, 504.36 184.13, 504.38, 549.43
PC(34:0)
784.4852
784.5827
[M+Na]+
184.05, 478.28, 506.31
PC(36:2) PC(36:2)
786.5025 808.4851
786.6013 808.5827
[M+H]+ [M+Na]+
184.13, 504.31 184.05, 504.29
signal was detected in the average spectrum of MALDI-MSI (shown in the Supporting Information (Figure S9A)). Meanwhile, the selected ion image at m/z 782, which represents the distribution of PC(34:1), indicated that no interference was coming from the matrix or the ITO substrate (shown in the Supporting Information (Figure S9B)). To provide the reference of mass spectrometry signal intensity analysis of cells, an additional test was conducted to directly measure the phospholipid signal in one, two or three cells (as shown in Figure S10 in the Supporting Information). The result showed that there was a significant linear relationship (R2 = 0.997) between signal intensity and cell counts. In order to further investigate the content of cellular phospholipids in the array, mass spectrometry signals of specific phospholipid were extracted from the MALDI-MSI result in the ROI. Because of the inerratic arrangement of cells, signal extraction could be conducted by extracting signal intensities corresponding to pixels at a fixed interval in the selected ion images using FlexImaging software. In this experiment, the corresponding signal intensities of PC(34:1) obtained from the pixels at an interval of 100 μm were extracted, and the results are shown in Figure 5C. The extracted pixels were marked by yellow frames and numbered by yellow characters. The white numbers on the right of each extracted pixels represented the cell counts, which could be easily recognized in the optical image. The relative mass spectrometry signal intensities of PC(34:1) in each pixel were normalized and are shown in Figure 5D. For the single-cell measurement in the array, we found that the mass spectrometry signal intensity of PC(34:1) from a single cell in pixel No. 2 was significantly higher than other three single cells in pixel Nos. 4, 8, and 12, which indicated the heterogeneity between the four cells. The two-cell and three-cell clusters were also measured, in which the intensities of pixel Nos. 3 and 9 were significantly different from other two-cell clusters. Because of the linear relationship between signal intensity and cell numbers, the average intensities of two- and three-cell clusters were calculated and compared with the single-cell intensity. Besides, because the cell count was
