This article describes the detailed methodology to prepare a Multiplexed Artificial Cellular MicroEnvironment (MACME) array for high-throughput manipulation of physical and chemical cues mimicking in vivo cellular microenvironments and to identify the optimal cellular environment for human pluripotent stem cells (hPSCs) with single-cell profiling.
Cellular microenvironments consist of a variety of cues, such as growth factors, extracellular matrices, and intercellular interactions. These cues are well orchestrated and are crucial in regulating cell functions in a living system. Although a number of researchers have attempted to investigate the correlation between environmental factors and desired cellular functions, much remains unknown. This is largely due to the lack of a proper methodology to mimic such environmental cues in vitro, and simultaneously test different environmental cues on cells. Here, we report an integrated platform of microfluidic channels and a nanofiber array, followed by high-content single-cell analysis, to examine stem cell phenotypes altered by distinct environmental factors. To demonstrate the application of this platform, this study focuses on the phenotypes of self-renewing human pluripotent stem cells (hPSCs). Here, we present the preparation procedures for a nanofiber array and the microfluidic structure in the fabrication of a Multiplexed Artificial Cellular MicroEnvironment (MACME) array. Moreover, overall steps of the single-cell profiling, cell staining with multiple fluorescent markers, multiple fluorescence imaging, and statistical analyses, are described.
Human pluripotent stem cells (hPSCs)1,2 self-renew unlimitedly and differentiate into various tissue lineages, which could revolutionize drug development, cell-based therapies, tissue engineering, and regenerative medicine3,4,5,6. General culture dishes and microtiter plates, however, are not designed to enable precise physical and chemical cell manipulation at the cellular level with the range of nano- to micro-meters, which is a critical factor for cellular expansion, self-renewal, and differentiation. To address this drawback, studies have investigated the roles of cellular microenvironments in regulating cell-fate decisions and cell functions4. In recent years, an increasing number of studies have been conducted to reconstruct cellular microenvironments in vitro7,8. Nano- and micro-fabrication processes have established these microenvironments through the manipulation of chemical9,10,11,12,13,14,15,16,17 and physical18,19,20 environmental cues. Until now, there were no reports to systematically investigate the underlying mechanisms of chemical and physical environmental cues on cell-fate decisions and functions within a single platform.
Here, we introduce a strategy based on simple design principles to establish a robust screening platform (Figure 1). First, we describe the development procedure of an integrated platform for creating versatile, artificial cellular microenvironments by using a nanofiber array and a microfluidic structure: The Multiplexed Artificial Cellular MicroEnvironment (MACME) array (Figure 1A and 2A). The nanofiber array has 12 different microenvironments in varying combinations of nanofiber materials and densities. Electrospinning was used to fabricate nanofibers. The nanofiber materials, such as polystyrene (PS)21, polymethylglutarimide (PMGI)22, and gelatin (GT)23, were designed to test their chemical properties, which might affect cell adhesion and maintenance of pluripotency (Figure 2B). Nanofiber densities were varied by changing electrospinning time and the generated nanofibers were defined according to their densities (DNF, with D = XLow/Low/Mid/High). The microfluidic structure is composed of polydimethylsiloxane (PDMS) harboring 48 cell-culture chambers, which can be positioned along the standard dimensions of the 96-well microplate. PDMS is a biocompatible and gas-exchangeable polymer generally used to fabricate microfluidic devices24. Each microfluidic channel was designed to be 700-µm wide and 8.4-mm long and had two inlets at its edges (Table 1). The chambers had different heights (250, 500, and 1000 µm) to manipulate the initial cell-seeding densities (0.3, 0.6, and 1.2 × 105 cells/cm2), which might correlate with survival, proliferation, and differentiation of hPSCs25 (Figure 2C). The number of cells seeded into a chamber is proportional to the column density above the chamber floor, and thus initial cell seeding density was controlled by introducing the same cell suspension into culture chambers with different heights. All channels were designed to be ≥ 250-µm-high26 to minimize the effects of low-oxygen tension27 and shear stress28 on the cells. Channel heights of 250, 500, and 1000 µm are abbreviated here as XCD with X = Low, Mid, and High, respectively. The environments with distinct nanofiber densities and initial cell-seeding densities were shortened as "Material_NF density_Cell density" (e.g., GT_HighNF_HighCD: an environment characterized by high-density GT nanofibers and high initial cell-seeding density).
Subsequently, we describe how to perform single-cell analyses to systematically investigate cell behavior in response to environmental factors (Figure 1B). As a proof-of-concept, we identified the optimal cellular environment for hPSC self-renewal, which is a key function for hPSC maintenance (Figure 1B)29. Image-based cytometry, followed by statistical analyses, allows for quantitative interpretation of individual cellular phenotypic responses to cellular environments. Among a variety of cellular functions, this paper provides a detailed procedure to identify the optimal conditions for maintaining hPSC self-renewal.
1. Fabrication of MACME Array
Note: All materials and equipment are listed in the Materials Table.
2. Loading hESCs into MACME Array
3. Quantitative Single-cell Profiling
MACME arrays: Design and fabrication: In combination with nanofiber technology, we used microfluidic cell culture and screening techniques employed previously to identify optimal conditions for hPSC self-renewal or differentiation35,36 (Figure 1). This is well suited for establishing robust high-throughput cell-based assays because the cell culture chambers and conditions are precisely controllable and expandable42,43,44. Here, the nanofiber array was prepared by a simple nanofiber deposition method, electrospinning with 3D-printed masks. The microfluidic part was fabricated with a 3D-printed mold to easily design the chamber structure through many trials and errors. The MACME array was constructed by combining the two parts and contained 48 environments, providing different biophysical and biochemical cues by varying combinations of both nanofiber ECMs (3 types of nanofiber materials and 4 different densities or 4 control matrix proteins) and cell-cell interactions (3 ranges of initial cell-seeding density) as shown in Figure 2.
Evaluation of overall effects of nanofiber properties and cellular density on hESC phenotypes: Following cell culture on a MACME array and on-plate fluorescent staining (Figure 6A), single-cell measurements were performed with the acquired images. For comparing homogeneity and heterogeneity of expression levels for the four phenotypic markers in each dataset, this protocol used SOM analysis38,39, which converts high-dimensional, multiparametric datasets into low-dimensional 2D maps. Notably, the results of PS nanofibers and 2DGT matrices were omitted from this analysis given that most cells did not adhere. Moreover, excluded from this analysis were those experiments with low CD where cells did not have enough direct (e.g., juxtacrine) or indirect (paracrine) interactions to survive. Following SOM analysis, unsupervised hierarchical clustering was performed for the SOM nodes of all analyzed samples (Figure 6B). By considering the height of the cluster dendrogram, the phenotypic differences allowed us to categorize the samples into three groups: Group i, most of the sample niches comprising of GT nanofibers, MG_HighCD, and VN_HighCD; Group ii, samples containing GT nanofibers (GT_XLow and MidNF_MidCD), MG_MidCD, or VN_MidCD; and Group iii, all PMGI_NF samples.
To study the differences among the groups, we examined one representative sample from each group (Figure 6C; Group i-GT_MidNF_HighCD, Group ii-GT_MidNF_MidCD, and Group iii-PMGI_MidNF_HighCD). In terms of OCT4 signals, all groups showed a higher expression than that of MG (MG_MidCD). Group i was characterized by high EdU signals, indicating that most cells were actively proliferating. Thus, microenvironments in Group I were characterized as conditions suitable for hPSC maintenance because undifferentiated hPSCs typically proliferate rapidly when cell-cycle gap phases were shortened45,46.
GT_MidNF_HighCD and GT_MidNF_MidCD are distinguished by their distinct initial cell-loading densities. High initial cell densities increased cells' opportunities to interact with neighboring cells, which enhanced their survival and proliferation47. Cells seeded at an insufficient initial density (3.0 × 104 cells/cm2) neither survived nor grew during the experimental period (4 days). Therefore, we did not incorporate these samples in the SOM analysis; cell numbers were below the cut-off of 1000 living cells/chamber. The cells on 2DGT scaffolds were also categorized as Group ii cells; these conditions do not support hPSC self-renewal48. The GT_MidNF_MidCD cells' EdU signal was lost and Annexin V levels were slightly increased, indicating that the cells had lost their stemness and had gradually become apoptotic. PMGI_MidNF_HighCD, which represents the microenvironments comprising PMGI nanofiber matrices, showed the larger variations in OCT4 and EdU signals compared with the other conditions.
Figure 1. Screening strategy to identify optimal cellular microenvironments for regulating cellular functions. (A) Overview of the components of the multiplexed artificial cellular microenvironment (MACME) array. The array is comprised of two main parts: nanofiber beds patterned on a basal substrate (nanofiber array) and a polydimethylsiloxane (PDMS)-based microfluidic structure. The MACME array is able to prepare nanofiber ECMs with a variety of features (i.e., nanofiber materials and densities) and various cell seeding densities. The PDMS microfluidic structure of the MACME array features three microfluidic-channel heights (250, 500, and 1000 µm) to regulate the initial cell-seeding density. (B) The impact of cellular microenvironments on cell phenotypes is quantitatively evaluated using an image-based single-cell assay and the statistical data analysis by self-organizing map (SOM) followed by hierarchical clustering. This figure was adapted from reference49 with permission from Wiley. Please click here to view a larger version of this figure.
Figure 2. Design of the MACME array. (A) Overhead and high-angle shots of the entire MACME array and the conceptual diagram of a microfluidic channel (light blue) on a nanofiber matrix (pink). Each channel is composed of a 700 µm wide and 8.4 mm long culture chamber and two inlets for introduction of medium and cells. (B) Comparison of sizes of cells and nanofiber matrices. The heights of cells and nanofiber beds range from 1-3 µm and 0.05-5.0 µm, respectively. (C) A cross section of each microfluidic channel with a 250/500/1000 µm height. Cells and nanofiber beds are represented as blue circles and orange lines, respectively. Each chamber has the same width and length but different heights (250, 500, and 1000 µm), and each chamber volume was 1.6, 3.2, and 6.4 µL, respectively. The black dotted rectangle in Figure 2C denotes Figure 2B. Figure 2 was adapted from reference49 with permission from Wiley. Please click here to view a larger version of this figure.
Figure 3. Fabrication of MACME array. (A) Fabrication of a nanofiber array. Patterning multiple nanofiber matrices on a baseplate was performed by modified electrospinning using a set of masks bearing patterned holes. Platinum-coating was performed to facilitate nanofiber deposition on a plastic baseplate. The masks with distinct hole-patterns were prepared by a 3D printer. Following masking of the platinum-coated area of the plate, the plate-mask set was put on the collection screen, and normal electrospinning was performed. The initial nanofibers were formed at the Pt-coated positions on the plate through the holes in the mask. Additional nanofiber beds were fabricated on the plate by replacing the existing mask and repeating the procedure. (B) Fabrication of a microfluidic structure. The mold printed by a 3D printer with UV-curable resin shaped the PDMS layer of the microfluidic device. After casting, the MACME array was assembled by attaching a PDMS-based microfluidic layer with a surface-activated nanofiber array. Figure 3 was adapted from reference49 with permission from Wiley. Please click here to view a larger version of this figure.
Figure 4. Images of graphical user interfaces (GUI) for microscopic image acquisition. Please click here to view a larger version of this figure.
Figure 5. Images of graphical user interfaces (GUI) of computer-guided image processing for single-cell phenotyping. The image analysis process contains 5 steps; (A) identification of primary objects in DAPI images, (B-D) identification of secondary objects in OCT4, Annexin V and EdU images, respectively, and measurement of object intensity. Please click here to view a larger version of this figure.
Figure 6. Phenotyping and classification of the phenotypes of H9 hESCs altered by microenvironmental factors. (A) Immunofluorescent images of H9 hESCs cultured at high initial seeding density on gelatin (GT) nanofibers (GT_HighNF_HighCD), stained with three cellular phenotypic markers (OCT4, pluripotency; EdU, cell proliferation; Annexin V, apoptosis) and DAPI for cell nuclei. (B) A heatmap and dendrograms of unsupervised hierarchical clustering. The tested microenvironments were categorized into three groups with distinctive features based on similarities of the phenotypes. Group i included GT nanofibers, MG_HighCD, and VN_HighCD. Group ii consisted of samples of GT nanofibers (GT_XLow and MidNF_MidCD), MG_MidCD, or VN_MidCD. All PMGI_NF samples were clustered into Group iii. In the heatmap, pink and blue colors indicate high and low cellular frequencies in SOM count plots, respectively. (C) Distribution of quantified marker expression levels of 1000 cells in one representative sample from each group (Group i-GT_MidNF_HighCD, Group ii-GT_MidNF_MidCD, and Group iii-PMGI_MidNF_HighCD). The 25th and 75th percentiles were indicated as the box limits. The whiskers extend to 1.5 times the interquartile range from the 25th and 75th percentiles. Experiments were repeated three times for each group. Figure 6A-C were adapted from reference49 with permission from Wiley. Please click here to view a larger version of this figure.
Name | Channel length (mm) | Channel width (μm) | Channel height (μm) | Inlet/Outlet diameter (μm) | Inlet/Outlet height (μm) | Mould length (mm) | Mould width (mm) | Mould height (mm) |
Mold | 8.4 | 700 | 250/500/1000 | 800 | 3000 | 127.76 | 85.48 | 14.35 |
Name | Thickness (μm) | Hole length (mm) | Hole width (mm) | Mask length (mm) | Mask width (mm) | Mould height (mm) | Row Offset (mm) | Column Offset (mm) |
Mask | 1000 | 6 | 5 | 123.5 | 80.2 | 14.35 | 11.24 | 14.38 |
Table 1. Dimensions of mold for a microfluidic structure and mask for a nanofiber patterning.
This protocol demonstrates the first screening method to establish a robust culture system for the maintenance of qualified hPSCs. First, we described how to prepare a platform featuring diverse artificial ECMs and cell seeding densities by using a microfluidic device integrated with a nanofiber array, the MACME array. Second, quantitative image-based single-cell phenotyping was performed50 to evaluate individual cellular outcomes and behaviors altered by distinct biochemical and biophysical features. In this protocol, the environment consisting of GT nanofiber and positive control ECM proteins in the condition of high initial cell seeding density was characterized by features of an undifferentiated state; rapid proliferation45 and stable OCT4 expression51. This observation indicated that molecular mechanisms relating to "cell-extracellular matrix" and "cell-cell" interactions could affect pluripotency of hPSCs.
This strategy can be customized to suit the user's purpose. For example, cell manipulation and throughput can be adjusted for a variety of experimental settings by re-designing the shapes and patterns of the microfluidic structure. As another way for throughput improvement, the inlet and outlet positions of each chamber are designed as per the microplate standard of 96-well plates, standardized by the Society of Biomolecular Sciences; thus, an automated robotic liquid dispenser can be used for high-throughput screening. To further examine the molecular mechanism concerning "cell-ECM interactions", the artificial cellular microenvironments can be made to more precisely mimic in vivo conditions by chemically modifying the nanofibers with signaling molecules52.
To date, most platforms developed for screening cellular microenvironments have been aimed at screening soluble factors (e.g., growth factors and chemical compounds)42,43,44, but not nanofiber ECMs owing to difficulties in combining the distinct fabrication techniques. For example, a nanofiber sheet produces a small gap between a microfluidic structure and a baseplate and causes their unstable bonding, which results in cross-contamination between different samples because a culture medium mix with another chamber's one through the small gap. Our new method facilitates simple and precise fabrication of nanofiber matrices at specific sites and allows their direct bonding with the baseplate, which prevents both unstable bonding and cross-contamination. Moreover, few platforms53 integrated with microfluidics and nanofiber matrices have been accessible for the systematic manipulation of both chemical and physical cues to investigate their effects on cell functions, because technologies needed for the integration were highly sophisticated. Our single-cell profiling with the MACME array can be performed with an easily attainable apparatus and simple techniques and holds great potential for use in modeling cellular environments for developmental and cell biological studies and drug discovery/screening.
The authors have nothing to disclose.
We thank Prof. N. Nakatsuji at iCeMS, Kyoto University, for providing human ES cells. We also thank Prof. A. Maruyama at Tokyo Institute of Technology for his support in the use of the atomic force microscope. Funding was generously provided by the Japan Society for the Promotion of Science (JSPS; 22350104, 23681028, 25886006, and 24656502); funding was also provided by the New Energy and Industrial Technology Development Organization (NEDO) and the Terumo Life Science Foundation. The WPI-iCeMS is supported by the World Premier International Research Centre Initiative (WPI), the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan. A part of this work was supported by Kyoto University Nanotechnology Hub and the AIST Nano-Processing Facility in “Nanotechnology Platform Project” sponsored by MEXT, Japan.
Polystyrene (PS) | Sigma | #182435 | Average Mw: 290,000, average Mn: 130,000 |
Polymethylglutarimide (PMGI) | MicroChem | G113113 | |
Gelatin (GT) | Sigma | G2625 | From porcine skin, type A |
Sylgard 184 silicone elastomer kit | Doe Corning Toray | #1064291 | PDMS curing agent and silicone elastomer base are components of this kit. |
OpenSCAD | This is a free 3D computer graphics software (http://www.openscad.org/) used for designing the mold of the microfluidic device. | ||
AutoCAD 2014 | Autodesk | This is a 3D computer graphics software (https://www.autodesk.com/products/autocad/overview) used for design of the mask used on nanofiber-array preparation. | |
3D printer, AGILISTA-3000 | Keyence | ||
UV-curable resin, AR-M2 | Keyence | This is used for 3D printing by Agilista. | |
Acetic acid | Sigma | #338826 | ≥99.99% |
Ethyl acetate | Sigma | #270989 | Anhydrous, 99.8% |
Tetrahydrofuran (THF) | Sigma | #401757 | |
MSP-30T | Vacuum Device | Magnetron sputtering machine | |
Nunc OmniTray | Thermo Fisher Scientific | #242811 | This is a polystyrene baseplate on which the nanofiber array is created. This plate size is typically 127.7 x 85.5 mm. |
Gun-type corona discharge machine | Shinko Electric & Instrumentation | CFG-500 | This handy device is used to generate corona for activation of the bottom surface of the PDMS layer at step 1.5 "Assembly of the MACME arrays" in the protocol. |
5 mL syringe | Terumo | SS-05SZ | |
Stainless-steel blunt needle (23-gauge) | Nipro | #2166 | Outside diameter and length are 0.6 and 32 mm, respectively. |
High-voltage power supply | TechDempaz | ||
1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide, hydrochloride | Dojindo | W001 | |
N-Hydroxysuccinimide | Sigma | #56480 | |
Matrigel hESC-Qualified Matrix | Corning | #354277 | This protein is refered as basement membrane gel matrix in the protocol. |
CellAdhere Vitronectin, Human, Solution | STEMCELL Technologies | #07004 | |
TeSR-E8 | STEMCELL Technologies | #05940 | Feeder-free, xeno-free culture medium for maintenance of human ES and iPS cells |
Y-27632 | Wako Pure Chemical Industries | #253-00513 | |
TrypLE Express Enzyme (1X), phenol red | Thermo Fisher Scientific | #12605028 | This ia a recombinant trypsin-like protease for dissociation of adherant mammalian cells. |
Click-iT EdU Imaging Kit with Alexa Fluor 647 Azides | Thermo Fisher Scientific | C10086 | The fluorescent labeling of proliferating cells in on-plate fluorescent staining was performed along the product manual of this kit. |
Annexin V, Alexa Fluor 594 conjugate | Thermo Fisher Scientific | A13203 | |
4',6-diamidino-2-phenylindole (DAPI) | Thermo Fisher Scientific | D1306 | |
Oct-3/4 Antibody (C-10) | Santa Cruz Biotechnology | sc-5279 | |
Donkey Anti-Mouse IgG H&L (DyLight 488) | abcam | ab96875 | This is a secondary antibody used in on-plate fluorescent cell staining. |
ECLIPSE Ti-E | Nikon | This is an inverted fluorescence microscope equipped with a CFI Plan Fluor 4×/0.13 N.A. objective lens (Nikon), CCD camera (ORCA-R2, Hamamatsu), mercury lamp (Intensilight, Nikon), XYZ automated stage (Ti-S-ER motorized stage with encoders, Nikon), and filter cubes for four fluorescence channels (DAPI, GFP HYQ, TRITC, Cy5; Nikon) | |
NIS-Elements Advanced Research | Nikon | This is a microscope imaging software used for automatic image acquisition. | |
CellProfiler, Version 2.1.0 | This is a free open software for cell image analysis (http://cellprofiler.org/). | ||
R | SOM analysis is performed by kohonen package of this software. This is freely available (https://www.r-project.org/). | ||
Cluster 3.0 | This is the open source clustering software (http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm). Unsupervised hierarchical clustering is performed with this software. | ||
Java TreeView | This open source software (http://jtreeview.sourceforge.net/) is used to visualize clustering data as a heatmap and a dendrogram. | ||
H9 human embryonic stem cell | WiCell Stem Cell Bank | WA09 |