Chen Cao, Dong Zhou, Tao Chen, Aaron M. Streets *, and Yanyi Huang * Ⅱ

Transcription

1 Chen Cao, Dong Zhou, Tao Chen, Aaron M. Streets *, and Yanyi Huang * Ⅱ Biodynamic Optical Imaging Center (BIOPIC), School of Life Sciences, Peking University, Beijing , China College of Engineering, Peking University, Beijing , China Ⅱ Peking-Tsinghua Center for Life Sciences, Peking University, Beijing , China * astreets@berkeley.edu (A. M. S.); yanyi@pku.edu.cn (Y. H.) This Supporting Information file includes experimental details of microfluidic chip fabrication, cell culture in dishes, mixing efficiency evaluation, stimulated Raman scattering microscopy, statistical analysis, and Figures S1-S13. S-1

2 EXPERIMENTAL SECTION Microfluidic Chip Fabrication. All microfluidic devices were fabricated with multilayer soft lithography 1. The manufacture process for the cell culture microfluidic array includes the fabrication of molds and PDMS devices. The flow and control molds were patterned on separate silicon wafers by photolithography. The silicon wafers were cleaned with acetone (140123, MOS Grade, BICR, China) and isopropanol (135261, HPLC Grade, Fisher Scientific, U.S.) thoroughly, and baked at 200 C for 5 min to dehydrate the wafers. After cooling down the wafers to room temperature, negative photoresist (SU8-2005, MicroChem, U.S.) was deposited onto the wafers by a disposable pipette and spun at 500 rpm for 5 s and 3000 rpm for 30 s to generate a dummy layer. Then the wafers were baked at 65 C for 1 min and 95 C for 2 min, and exposed to UV radiation for 30 s with a photolithography aligner (URE-2000B; Institute of Optics and Electronics, Chinese Academy of Sciences, China). After exposure, the wafers were baked at 65 C for 1 min and 95 C for 5 min. Then the wafers with dummy layer were ready for the following flow and control mold fabrication. The flow mold required two layers of resist in order to obtain flow channels and cell culture chambers with different heights. Both positive and negative photoresist were used. The wafer was first coated with positive photoresist (AZ50XT, AZ Electronics Materials, Luxembourg) by spinning at 500 rpm for 5 s and 2000 rpm for 60 s, and baked at 95 C for 1 min and 105 C for 10 min. Then the wafer was exposed to UV radiation through a positive mask defining the flow channel features for 30 s twice, and developed in AZ400K developer (DI water: developer=2: 1). The wafer was rinsed by blow drying and baked by ramping from 65 C to 190 C at 10 C /h for 17 h to reflow the positive photoresist. The height of the flow channels was about 15 μm. Negative photoresist (SU8-2025, MicroChem, U.S.) was then used for making cell culture chambers. SU was poured on the wafer with flow channels and spun at 500 rpm for 5 s and 2000 rpm for 30 s. Then the wafer rested on a flat surface for 10 min to ensure the uniformity of the photoresist. The wafer was baked at 65 C for 2 min and 95 C for 6 min, and exposed to UV radiation for 25 s twice after aligning the chamber mask and the channel mask. The wafer was baked again at 65 C for 1 min and 95 C for 6 min, and developed by SU-8 developer and gently washed by isopropanol followed by blow drying. The flow mold was then baked by ramping from 65 C to 160 C at 50 C /h for 5 h. The height of cell culture chamber was about 30 μm. To fabricate the control mold, negative photoresist (SU8-2025, MicroChem, U.S.) was poured onto the wafer with dummy layer and spun at 500 rpm for 5 s and 3500 rpm for 30 s. After baking at 65 C for 2 min and 95 C for 6 min, the wafer was exposed to the UV light for 30 s, and baked again at 65 C for 1 min and 95 C for 5 min. The control mold was developed by SU-8 developer and baked on a hotplate by ramping from 65 C to 160 C at 50 C /h for 5 h. Both the flow and control molds were incubated with trichloro-(1h, 1H, 2H, 2H-perfluorooctyl) silane (448931, Sigma-Aldrich, U.S.) vapor for 10 min to adhesion of the PDMS (polydimethylsiloxane, RTV615, MOMENTIVE, U.S.). The chip was designed in a push-down manner. PDMS mixture (potting agent: crosslinking agent=20: 1) was poured onto the clean flow wafer, and the wafer was spun at 500 rpm for 5 s and 1800 rpm for 60 s and rested for 10 min in room temperature. Before alignment, the flow wafer was incubated at 80 C for 15 min and UV light sterilized for 5 s. About 30g of PDMS mixture (potting agent: crosslinking agent=5: 1) was S-2

3 poured onto the clean control wafer. Then the control wafer was degassed and incubated at 80 C for 30 min. The PDMS layer on the control wafer was carefully peeled off and cut along the edge of the wafer. A 20-gauge rounded punch was used to punch holes for the control channels inlets. After sterilization by UV light for 5 s, the control layer was aligned to the PDMS coated flow wafer under a stereo microscope (SMZ645, NIKON, Japan), and the aligned double-layer device was incubated at 80 C for 2 h. The bonded layers were then carefully peeled off from the flow wafer, and inlets of the flow channels were punched. Finally, a 75% alcohol-sterilized glass coverslip (0.3 mm thickness) and the two-layer chip were exposed to oxygen plasma (PDC-MG, MingHeng, China) for 30 s, and the chip was placed onto the glass coverslip quickly to complete the bonding. Cell Culture in Dishes. HeLa cells were purchased from the Cell Bank of the Chinese Academy of Sciences ( MCF7 and Chang Liver cells were obtained from Dr. X. Zheng and Dr. C. Yi labs (School of Life Sciences, Peking University), respectively. MDA-MB-231 cells were contributed by Dr. N. Zhang lab (Tianjing Medical University Cancer Institute & Hospital). The four types of cells were cultured in DMEM (high glucose, , Life Technologies, U.S.) supplemented with 10% FBS (vol/vol, , Gibco, U.S.) and 1% Penicillin-Streptomycin ( , Gibco, U.S.). Cells were cultured at 37 C in a humidified incubator (CO-150, New Brunswick Scientific, Germany) with 5% (vol/vol) CO 2, and passaged every 2-3 days. For comparison experiments between cells cultured in dish and chip, cells were cultured in 35 mm glass bottom dishes (D N, In Vitro Scientific, China), and imaged when ~50% confluence was reached. For cell suspensions prepared for the chip, cells were cultured in 60 mm dishes (430166, Corning, U.S.), and digested with 0.05% trypsin ( , Gibco, U.S.) when ~70% confluence was reached. The typical concentration of cells seeded on chip was ~ 10 7 cells/ml. Oleic acid (O3008, Sigma Aldrich) was thoroughly mixed with culture medium before incubating with cells. Mixing efficiency evaluation. To confirm efficient mixing, cell culture chambers were injected into clean water and oleic acid chambers were filled with colored dye (see Video S1 in the Supporting Information). Bright-field images and videos were acquired with a 10 objective of an inverted microscope (TE2000-S, Nikon, Japan) and a charge-coupled device (CCD) camera (QHYCCD, China) connected to a computer. As indicated in Figure S7, we calculated the relative concentration of both chambers with the Beer-Lambert law: I I I0 I f log 10 I 0 = 10 εlc, C C f = log 10 I0 (1) The mean pixel intensity (I ) in the labelled regions of both chambers was calculated for every 112 s. I0 was measured as the mean pixel intensity of a control area outside of the chambers to be mixed. If represents the mean pixel intensity of a region with pure dye. In the equation S1, ε denotes the dye s extinction coefficient, and l is the depth of the chamber. According to our calculation, the liquid in two chambers was fully mixed when the mixing valves were actuated for 672 s. In order to ensure thorough mixing, the mixing time was extended to 30 min (1800 s). After mixing, interface valves (pressurized with 30 psi) closed and the mixing valves were pressurized with 5 psi to function as hydration channels. Then the microfluidic device was placed on an onstage incubator (Chamlide TC-A, LCI, Korea; with a home-made adaptor to fit the shape of our device) for imaging. The containment valves between chambers remained closed during imaging. S-3

4 Stimulated Raman scattering (SRS) microscope. The home-built SRS system used a dualwavelength laser source (picoemerald, APE, Germany) that contains a pump laser integrated optical parametric oscillator (OPO). It provided two spatially and temporally overlapped pulse trains, with the synchronized repetition rate of 80 MHz. One beam, fixed at 1064 nm, was used as the Stokes light. The other beam, tunable from 780 to 990 nm, served as the pump light. The intensity of the Stokes beam was modulated at 20.2 MHz by a resonant electro-optical modulator (EOM). The overlapped lights were directed into an inverted multi-photon scanning microscope (IX81, Olympus, Japan). The collinear laser beams were focused into the sample by a water immersion 60X objective (UPlanSAPO 60XW, Olympus, Japan). Transmitted light was collected by a condenser (WI-DICD, NA 0.8, Olympus, Japan). After filtering out the Stokes beam, the pump beam was directed onto a large area photo diode (FDS 1010, Thorlabs, U.S.). The photo diode voltage output was sent into a lock-in amplifier (HF2LI, Zurich Instruments, Switzerland) to demodulate the SRS signal. Images were reconstructed using software provided by the microscope manufacturer (FV10 ASW, Olympus, Japan). Statistical Analysis. Some single LDs were comparable in size with droplet aggregates, and the size of LD aggregates greatly deviated from the sample mean. In order to avoid statistical bias, we introduced the concept of circularity to distinguish single LDs from LD aggregates. The circularity of a three dimensional object was calculated as follows: C = P2 4πA (2) Here C denotes the circularity, P is the perimeter and A is the area. The circularity was calculated for the projection of each 3D connected LD. For single LD analysis, LDs with (A<100 & C<1.5) and (A 100 & C<1.2) were treated as single LDs. LD number was estimated by dividing the total intensity of LDs by the median value of single LD in the same field of view (or in one single cell). To independently compare the intensity and number distribution, estimated LD number was calculated by dividing the total size of LDs by the median size of single LD in each cell. In the LD distribution comparison between single cells and averaged cells, the intensity of averaged cells was calculated by dividing the total intensity of LDs by the cell number in the same field of view. The cell number in images generated only with the lipid band was manually counted. Error bars were computed using the standard deviation of the mean. The heatmap was plotted by the R package 'pheatmap'. The RPCL clustering algorithm was written in MATLAB. For single LD measurements, P values were calculated by the median value of LD morphology parameters in each field of view between 0 mm and 1.2 mm. For single cell measurement, P value was calculated by the median value of LD morphology parameters among single cells between 0 mm and 1.2 mm. S-4

5 Figure S1 Setup of the microfluidic chip. (a) A photograph of the microfluidic chip which shows the tubes connecting to the control valve inlets (scale bar: 1 cm). The cells suspension and reagents are input through pipette tips. (b) A schematic of the titration array unit. The hydration channel is designed to encircle the cell culture chambers. The interface valve controls the connection between oleic acid chambers and cell culture chambers. The containment valve separates the cell culture chambers. (c) The chamber size and the volume ratio between each oleic acid chamber and cell culture chamber. The chamber height is ~30 μm. Each cell culture chamber harbored a volume of 5.9 nl, which is six orders of magnitude smaller than traditional 30 mm cell culture dish. S-5

6 Figure S2 HeLa cells cultured on the microfluidic chip were imaged for three consecutive days in the same field of view (bright field). In both low and high density seeding conditions, cells were attached and grew healthily. S-6

7 Figure S3 The comparison between cells cultured in dish and chip. (a) The SRS images of cells cultured in different platforms. (b) Average Lipid droplet intensity per cell was calculated by averaging the total lipid droplet intensity to the cell number in the same field of view. S-7

8 Figure S4 Comparison between and images. (a) The same field of view of cells was imaged with two different resolutions. Blue circles indicate a single cell for demonstration. (b) Magnified regions from (a) and corresponding lipid droplet mask. Blue circles indicate the lipid droplets for demonstration. (c) A magnified comparison of LD detection in the image and image. The LDs move during image acquisition. This results in an inflated LD mask in the images which have a slower acquisition time. Cells shown here were cultured in dishes. S-8

9 Figure S5 Background removal of lipid specific images by morphology top-hat filtering. (a) Images before and after background smoothing. (b) Pixel intensity is plotted along the horizontal and vertical lines labelled in (a) before and after background processing. S-9

10 Figure S6 Background Removal of protein specific images by morphology top-hat filtering. (a) Images before and after background smoothing. (b) Pixel intensity is plotted along the horizontal and vertical lines labelled in (a) before and after background processing. S-10

11 Figure S7 Mixing efficiency evaluation. (a) Photographs of mixing scheme. Oleic acid chambers were filled with dye and the average transmitted intensity of labelled areas (red square) was calculated during mixing (SI Materials and Methods). (b) The relative concentration of the dye in both oleic acid chamber and cell culture chamber was plotted during 672s of mixing. S-11

12 Figure S8 The discrimination between single big lipid droplets and lipid droplet aggregates. (a) A lipid specific image and its lipid droplet mask. (b, c) Three big lipid droplets from two cells were selected for demonstration. The size of the three lipid droplets are comparable, but the first and the second lipid droplets are both single while the third one is an aggregate of many small lipid droplets, which is reflected by the circularity value. S-12

13 Figure S9 The single LD size distribution of four types of cells. The P value of these cells was (Hela), (Chang Liver), (MCF7), (MDA-MB-231). S-13

14 Figure S10 Cell-cell heterogeneity. The background heatmap represents the total intensity per cell, and the lipid droplet intensity distribution within each cell is indicated by a violin plot. Under the same concentration of oleic acid, individual cells exhibit a variety of lipid droplet distributions. S-14

15 Figure S11 RPCL clustering and the cell count of different culture conditions in each cluster. (a) RPCL clustering of Chang Liver cells based on two parameters we measured: the lipid droplet number and the total lipid droplet intensity of single cells. Each cell is indicated by a circle, and the clustering centers are plotted as asterisks. Clustering tags are plotted below the clustering centers. (b) In each cluster, the count of cells cultured under different concentrations of oleic acid was calculated. The ratio of cells from higher concentration of oleic acid increased from cluster 1 to cluster 11 (clustering tags are indicated on the right). S-15

16 Figure S12 RPCL clustering (a) and the MCF7 cell count of different culture conditions in each cluster (b). S-16

17 Figure S13 RPCL clustering (a) and the MDA-MB-231 cell count of different culture conditions in each cluster (b). S-17

18 (1) Unger, M. A.; Chou, H. P.; Thorsen, T.; Scherer, A.; Quake, S. R. Science 2000, 288, S-18