Solid lift-off method
To optimize the proposed solid lift-off protocol, two aspects including the surface treatment of substrates and time length of protein (taking fibronectin (FN) as an example in this work) incubation, were carefully considered. FN patterns with good uniformity were achieved on the substrate with hydrophilic modification before FN incubation for 30 min, as displayed in Fig. 2a2. However, some defects arose (discontinuity and nonuniformity) in the obtained FN patterns after FN incubation for 30 min, as shown in Fig. 2a3, if the surface of the substrate was not treated with oxygen plasma for hydrophilic modification, which would interfere with the cell adhesion and spreading in cellular studies. Moreover, when the FN incubation time increased from 30 to 60 min, the FN expanded more out of the pore area (on the shadow mask), resulting in a larger deviation in size between the practically obtained and designed patterns, as shown in Fig. 2a4. Thus, the optimized performing conditions for high-quality FN patterns comprised hydrophilic modification of substrates prior to adhesion of the Parylene C micropore arrays and incubation of the FN solution for 30 min. With the optimized protocol, FN patterns of various shapes and sizes could be well prepared (Fig. 2c). All the subsequently presented FN patterns were prepared using the optimized condition unless otherwise stated. Additionally, the differences between the designed and practically obtained sizes of protein micropatterns (d) prepared using the present solid lift-off and previously reported methods were compared, as shown in Fig. 2b. The data in Fig. 2b indicate that the proposed solid lift-off method is more reliable in achieving a small size of protein patterning with a high size precision, with d < 0.5 μm (red dots in Fig. 2b), i.e., only a small percentage (6.9 ± 6.1%) change was observed in the experimental values compared to the designed ones.
However, the microcontact printing technique prepares small protein micropatterns with a larger value of d (varying from 0.35 to 3.8 μm, blue triangles in Fig. 2b), i.e., a larger percentage (19.4 ± 11.8%) change relative to the designed value. It is conceivable that the size of the prepared patterns in the microcontact printing process is easily influenced by the applied pressure during the transfer of FN from stamps to substrates, as shown in Supplementary Fig. S1, which often varies among operators. Therefore, the differences between the experimental and designed values were large. In contrast, the proposed solid lift-off method is free of any extra pressure loading and thus more robust and stable, supporting its wide applications, particularly in fields with high requirements for size precision.
Protein patterning on flat and curved substrates
As shown in Fig. 2c, FN patterns of various sizes and shapes on flat substrates were successfully prepared. The typical sizes/arrangements generated in this study included small diameter/small space (2.47 ± 0.07 μm/1.74 ± 0.07 μm), small diameter/large space (4.02 ± 0.38 μm/15.23 ± 0.22 μm), large diameter/small space (11.02 ± 0.11 μm/1.32 ± 0.31 μm), and large diameter/large space (11.48 ± 0.36 μm/14.71 ± 0.46 μm). The shapes generated in this study included uniformly distributed square or hexagon arrays, a combination of squares and hexagons, and arbitrary shapes (e.g., the pattern of a smile, ). However, some of the above distribution types (e.g., large diameter/small space and small diameter/large or small space) of protein micropatterns were unachievable by the previously reported methods. The failures resulted from collapses and deformations of PDMS microstructures with a high width/height ratio and a high height/width ratio, respectively, schematically shown in Supplementary Fig. S1 for the microcontact printing method, and poor precision in the fabrication capacity for inkjet printing and UV-induced crosslinking.
In our present solid lift-off technique, the success in achieving various arrangements of protein micropatterns with a high size precision, even for a small feature size (<2 μm), is attributed to the spatially confining protein transfer process (i.e., protein molecules contacted substrates only after traveling through micropores). Further studies of the detailed mechanism to illustrate how the protein molecules reach the substrate through micropores without extra driving force are ongoing. Moreover, protein micropatterns on curved substrates, e.g., the inner wall surface of PDMS tubes with various radiuses of curvature (0.5–3 mm), were also successfully obtained, benefiting from the flexibility of Parylene C micropore arrays (compatibility of the tube shadow mask). The typical fluorescence images of the obtained FN patterns on curved substrates with various radius of curvatures (r) are displayed in Fig. 2f and Supplementary Fig. S2a-b, with views of different angles. The patterning capacity on curved substrates will facilitate the promising applications of this method in organ-on-chips and tissue engineering, which involve complicated 3D structure constructions, particularly for ex vivo and in vivo studies.
High-throughput and high-efficiency cell patterning
The results of cell patterning are shown in Figs. 3, 4. Previously, we confirmed the high efficiency of the solid lift-off method in achieving high-throughput single-cell capture with a simple shadow mask43. In this work, improvements have been made to realize multicell alignments along with simultaneous control of cell adhesion and spreading via an ingeniously designed shadow mask. The design philosophy of the composite shadow mask in the present solid lift-off method comprises the large micropores (capture pores) in the central areas controlling cell capture/alignments and the small micropores (spreading pores) in the surrounding areas controlling cell adhesion/spreading. The size of the capture pores strongly influences the efficiency of cell alignments, as shown in Fig. 3a. The reasons why the present solid lift-off method can achieve such high efficiencies in high-throughput cell patterning are discussed and schematically illustrated in Fig. 3b. In the process of cell loading and incubation (2 h), only a single cell could be captured in a capture pore, although some cells nonspecifically adhered on the membrane (shadow mask) in the areas beyond the capture pores. Subsequently, the nonspecifically adhered cells were removed together with the lift-off of the shadow mask (Parylene C micropore arrays), leaving only cells in the capture pores to adhere and spread (12 h) in the patterning areas confined by the spreading pores (Fig. 3b1). The removal of nonspecifically adhered cells is a critical factor to obtain high-efficiency cell alignments at a high throughput. It is easy to be thought that if the cells in the micropores adhered on the wall of micropores, the cell alignment efficiency would be impacted when the mask was lifted off from the substrate. Nevertheless, this impact has been minimized via careful experimental optimizations with separate investigations of incubation duration lengths after the FN protein solution and cell solution loading for the high-efficiency protein and cell patterning.
In contrast, in the other protein micropattern-based methods, to meet the requirement of cell spreading, the size of the pattern was always larger than the diameter of an individual cell in suspension status (a result of the aforementioned trade-off). Therefore, two or more cells were often captured on protein micropatterns (Fig. 3b2, b3), without means to remove the undesired cells, resulting in poor controllability and a low efficiency of cell alignment. To the best of our knowledge, this is the first report to simultaneously establish the precise control of cell alignment and adhesion/spreading with a high efficiency at a high throughput through only a single-step operation, i.e., the developed solid lift-off method.
For optimization, different sizes of capture pores (20, 25, and 30 μm) were thus first investigated. As shown in Fig. 4a, a 25 μm capture pore presents the highest efficiencies for single-cell, double-cell, and triple-cell alignments of 86.2 ± 3.2%, 56.7 ± 9.4% and 51.1 ± 4.0%, respectively, compared to those of 20 μm (68.5 ± 3.7%, 41.0 ± 1.7%, and 33.7 ± 1.6%, respectively), and 30 μm (46.7 ± 9.1%, 21.8 ± 4.2%, and 22.8 ± 5.8%, respectively). All subsequent cell patterning results shown in this study were obtained with a capture pore size of 25 μm, unless otherwise stated. The confinement area for cell spreading (the total area spanned by capture pores and spreading pores) was 39 μm × 39 μm. The pitches were 69 μm, 44 μm, and 44 μm (i.e., spaces were 30, 5, and 5 μm) for single-cell, double-cell, and triple-cell alignments, respectively. Cell patterning on curved substrates (the inner wall surface of PDMS tubes with various radius of curvature) was also investigated, and the preliminary results are shown in Supplementary Fig. S2. In addition to the patterning of the same typed cells, the proposed solid lift-off method in this paper is also applicable to the patterning of different typed cells. The patterning of different typed cells is feasible referring to the previously reported multistep solid lift-off process with multilayer Parylene C micropore arrays as shadow masks. Taking the patterning of two typed cells as an example, the schematic illustration is shown in Supplementary Fig. S3. The previously reported results demonstrated that the cells were of high viability after multiple solid lift-off operations44,45. Therefore, the integration of our novel design for simultaneous control of cell alignment and adhesion/spreading and the multiple solid lift-off processes shows promise for fulfilling the patterning of multiple typed cells, and be of great potential and broad interests for applications in various fields including regenerative medicine and tissue engineering.
In addition, the cell patterning efficiencies obtained from the previous methods (e.g., microcontact printing), which controlled cell alignment totally depending on the protein adhesion confinement, were experimentally compared in parallel, as shown in Fig. 4b. Typical images of patterned cells are displayed in Fig. 4c. The achieved efficiencies with the present solid lift-off method are significantly higher than those of the protein micropattern-based methods. For single-cell alignment, the increase in proportions are 153.52% and 131.16% compared to the previous methods, simple and composite protein micropatterns26, respectively. The corresponding values for double-cell and triple-cell alignments are 141.27% and 76.6% and 250% and 111.15%, respectively.
Application in the functional study of cell skeleton distribution and cell–cell junctions
The cytoskeleton (F-actin) alignment and distributions of three typical proteins (vinculin, Cx43, and N-cadherin) in the cell–cell junction areas were analyzed with murine skeletal muscle myoblasts (C2C12), as shown in Fig. 5. The orientation distributions of F-actin filaments are shown in Fig. 5a–c, including three different types of cell alignments, in vertical (a1–c1), square (a2–c2), and lateral (a3–c3) arrangements. From a previous report, the actin cytoskeleton will be strongly reorganized before myoblasts fusing and differentiating into myotubes, where the actin filaments tend to organize into a dense actin wall structure that parallels and extends the length of the plasma membrane of the aligned cells46. Here the consistent phenomena were observed. For single-cell patterning, in the type of vertical arrangement (Fig. 5a1), most of the F-actin filaments were distributed along the direction of 90o, i.e., along the long axis of the micropatterns (cells). In contrast, in the type of lateral arrangements, i.e., 90o rotation from the vertical arrangement (Fig. 5a3), the F-actin filaments were mainly distributed along the directions of 0o and 180o, i.e., 90o rotation clockwise or counterclockwise rotation from that of the vertical arrangement (direction of 90o). In the type of square arrangement (Fig. 5a2), the F-actin filaments were randomly distributed along directions of 0o, 90o, and 180o (i.e., where the plasma membrane boundary of patterned cells is distributed). Here one point needs noting that the “random” in this work means “the random percentages of F-actin distributions among directions of 0o, 90o and 180o”, rather than “the F-actin distributions are in any angle along the circumferential direction”. In the cases of multicell patterning, in the vertical arrangement (Fig. 5a1–c1), the F-actin filaments were more significantly distributed along the direction of 90o, i.e., along the long axis of the entire alignment of micropatterns (cells), as expected. In contrast, in the other two groups (square and lateral arrangements in Fig. 5a2–c2 and a3–c3, respectively), the distributions of F-actin filaments became random. The distribution of the cytoskeleton (F-actin filament as a typical object) is considered important for cell–cell interaction and cell–cell junction formation. To verify this, distributions of three typical proteins in cell–cell junction areas were analyzed, as shown in Fig. 5d−f.
Vinculin, a membrane-cytoskeleton protein in focal adhesion plaques, participates in linkage of integrin adhesion molecules to the F-actin filament and is associated with cell–cell junctions by anchoring F-actin to the membrane. Therefore, the distribution of vinculin was predicted to follow that of F-actin, which was proved by the experimental results. In the vertical arrangement (Fig. 5d1), the distribution was focused in the direction of 90 o, i.e., along the long axis of the entire arrangement of patterns/cells. In contrast, random distributions were observed in the other two control groups (square and lateral arrangements in Fig. 5d2, d3, respectively).
Cx43 and N-cadherin, which are involved in the formation of junctions to bind cells with each other, were also analyzed. As shown in Fig. 5e1, h and Fig. 5f1, i, the distributions of Cx43 and N-cadherin molecules were both focused in the direction along the main axis of cell–cell junctions (the borders of adjacent cells) in the vertical arrangement. In contrast, in the square and lateral arrangements, the distributions of both Cx43 and N-cadherin molecules were rambling (Fig. 5e2–3, h, and Fig. 5f2–3, i, respectively). The focused distributions of Cx43 and N-cadherin along the cell–cell border and the distributions of vinculin and F-actin along the long axis of the entire arrangement of patterns revealed that C2C12 cells preferred to form cell–cell junctions in the end-to-end way, i.e., vertically, rather than laterally or at right angles, which is consistent with a previous report47.
The precise control of cell alignments achieved by only a single-step solid lift-off operation realizes the ability to regulate the distribution of the cytoskeleton and cell–cell junctions, which are critical factors affecting subsequent signaling pathways involved in cell–cell interactions, differentiation, and further development into an organ or organism. The promising potential of the developed simple but effective method will facilitate its extensive utilization in regenerative medicine and tissue engineering for basic mechanism studies as well as practical applications.