Protein Micropatterning in 2.5D: An Approach to Investigate Cellular Responses in Multi-Cue Environments.

The extracellular microenvironment is an important regulator of cell functions. Numerous structural cues present in the cellular microenvironment, such as ligand distribution and substrate topography, have been shown to influence cell behavior. However, the roles of these cues are often studied individually using simplified, single-cue platforms that lack the complexity of the three-dimensional, multi-cue environment cells encounter in vivo. Developing ways to bridge this gap, while still allowing mechanistic investigation into the cellular response, represents a critical step to advance the field. Here, we present a new approach to address this need by combining optics-based protein patterning and lithography-based substrate microfabrication, which enables high-throughput investigation of complex cellular environments. Using a contactless and maskless UV-projection system, we created patterns of extracellular proteins (resembling contact-guidance cues) on a two-and-a-half-dimensional (2.5D) cell culture chip containing a library of well-defined microstructures (resembling topographical cues). As a first step, we optimized experimental parameters of the patterning protocol for the patterning of protein matrixes on planar and non-planar (2.5D cell culture chip) substrates and tested the technique with adherent cells (human bone marrow stromal cells). Next, we fine-tuned protein incubation conditions for two different vascular-derived human cell types (myofibroblasts and umbilical vein endothelial cells) and quantified the orientation response of these cells on the 2.5D, physiologically relevant multi-cue environments. On concave, patterned structures (curvatures between κ = 1/2500 and κ = 1/125 μm-1), both cell types predominantly oriented in the direction of the contact-guidance pattern. In contrast, for human myofibroblasts on micropatterned convex substrates with higher curvatures (κ ≥ 1/1000 μm-1), the majority of cells aligned along the longitudinal direction of the 2.5D features, indicating that these cells followed the structural cues from the substrate curvature instead. These findings exemplify the potential of this approach for systematic investigation of cellular responses to multiple microenvironmental cues.

Introduction

In vivo, cells are subjected to an orchestra of physical, (bio)chemical, and mechanical cues present in the surrounding extracellular environment, which have been separately shown to influence cellular behavior.[1−4] For example, one of the most studied cellular responses is contact guidance, which is the phenomenon of cell and stress-fiber alignment in the direction of anisotropic geometrical cues.[5−13] Contact-guidance cues can be emulated in vitro and presented to cells by patterning extracellular matrix (ECM) proteins on an otherwise non-cell-adhesive two-dimensional (2D) substrate. 2D protein patterns are often produced using microcontact printing, where a protein-inked polydimethylsiloxane (PDMS) stamp containing a desired micrometer-sized feature is brought into contact with a substrate in order to transfer the protein ink.[14] This technique enables patterning of relatively large areas in one go and can be used for a wide range of substrate materials.[14,15] 2D protein patterns can also be created using deep UV (ultraviolet) patterning, where UV light is applied through a physical mask to degrade passivation polymers at specified locations on the substrate, which can then be coated with the desired protein.[16−19] Using these approaches, it has been shown that cell differentiation can be directed by modulating the cell shape and that focal adhesion, stress fiber organization, and cell migration are all strongly influenced by the adhesive environment.[20−23]

The role of the three-dimensional (3D) environment on cell behavior has also gathered increasing attention.[24] In particular, the in vivo ECM often presents cells with topographical cues at length scales of μm to hundreds of μm; examples include collagen fibers in the tendon (Ø = 1–20 μm) and also overall tissue curvatures such as capillaries (Ø = 6 μm), arterioles (Ø = 30 μm), small arteries (Ø = 1000 μm), and alveoli (75–300 μm).[25,26] Experiments using engineered cell environments, such as 3D substrate curvatures of electrospun fibers, have been shown to elicit a cell-type-dependent orientation response. On fibers with diameters of at least 5 μm, endothelial cells were able to wrap around fibers, whereas endothelial colony-forming cells aligned in the fiber direction.[27] Similarly, the degree of substrate curvature can influence cellular orientation as shown by Hwang et al.(28) Common bottlenecks in gaining mechanistic insights into the role of the 3D structure and geometry are the limited degree of control and precision over the environmental cues presented to the cells and the difficulty in obtaining high-resolution microscopic readout due to the limited penetration depth and Z-resolution. To address these issues, we propose using microfabricated platforms containing two-and-a-half-dimensional (2.5D) cues. 2.5D substrate cues contain structural geometries and topographies that represent relevant 3D environments cells typically encounter in vivo. Since the topographies are partially embedded in the substrate surface, it is neither 2D or 3D and hence 2.5D. Microfabricated 2.5D platforms, made with techniques such as hot embossing, thermoforming, photolithography, soft lithography, or electron beam lithography, provide cells with a structurally more representative biomimetic environment compared to 2D substrates, while still retaining a high degree of control over parameters such as material, coating, roughness, stiffness, and dimensions of features.[29−33] 2.5D platforms are rapidly gaining popularity and show great promise in unraveling the influence of substrate geometry on cell behavior.[24,34−39]

The abovementioned studies demonstrate the usefulness of in vitro minimal models to dissect cells’ ability to sense and respond to individual physical and geometrical cues through contact guidance and curvature sensing.[40]In vivo, however, multiple cues are present simultaneously. To improve mimicking multi-cue cellular environments and understand cell behavior in such complex settings, it is imperative to create substrates that contain multiple cues in a controlled manner. Here, we focused on the combination of contact-guidance and curvature-guidance cues as a prime example of how multiple structural features can concurrently modulate cell response. Previously, contact guidance has been proven to influence cellular behavior in multiple ways. For example, cells’ orientation,[21,23] migration,[41] polarization,[42] cell shape,[18] and cell fate[43] are affected by contact-guidance cues.[44] Identically, curvature guidance is an established physical cue that among others influences cell alignment,[45] migration,[38,46] differentiation,[46] and polarization.[24,35,36] Combining these cues on one cell-culture substrate has proven to be particularly challenging, primarily due to technical limitations. Contact-mediated protein patterning is not possible on structured substrates with features in the micrometer range due to alignment difficulties. Besides, as microcontact printing and deep-UV patterning make use of soft lithography methods and physical masks during substrate preparation, experiments are typically time- and labor-intensive and limited in terms of pattern flexibility. Waterkotte et al. introduced an approach involving maskless lithography, chemical vapor deposition, and thermoforming to combine surface topographies with protein patterns.[47] This method, however, is limited to materials that are suitable for thermoforming and has a relatively low pattern resolution (7.5 μm), while contact-guidance cues as small as 0.1 μm2 have been shown to affect cell response.[43,48] Another approach was developed by Sevcik et al.,[49] who made use of microcontact printing on sacrificial poly(N-isopropylacrylamide) (PIPAAm) layers that were subsequently used for transfer of the ink from the thermally transformed PIPAAm onto a PDMS substrate containing topographies. Unfortunately, this method is also limited in terms of pattern flexibility and resolution of the final pattern due to its need for photolithographic methods and microcontact printing. Overall, we note that existing methods (e.g., microcontact printing, deep-UV patterning, and photolithography) have been crucial in uncovering the influence of individual substrate cues on cellular behavior but lack the ability to fabricate and systematically change substrate cues to investigate the effect of multiple cues at the same time. In order to advance the current understanding of cellular behavior both in vitro and in vivo, new methods are needed that can be applied in a flexible, high-throughput, and robust manner. Eventually, the obtained knowledge can contribute to steer cell behavior in, for example, engineered scaffolds that can be applied for tissue engineering. Especially in this field, the formation of functional tissues (e.g., for cardiac tissues, muscles, and the vascular wall) relies on the specific organization of cells and the ECM.[10,50,51]

In the present study, we introduce a new approach that circumvents the limitations associated with the existing approaches through a rapid, contactless, and maskless patterning of ECM proteins on a culture chip containing a library of micro- to mesoscale 2.5D structures. This approach enables an improved resolution of protein patterns on structured substrates and allows use of a wide range of substrate materials. Moreover, the use of a 2.5D chip containing curved substrates allows for easy microscopic evaluation of cells and their intracellular components. We first explored and validated this novel patterning approach by looking at the response of human bone-marrow-derived mesenchymal stromal cells (hBMSCs) on convex cylindrical substrates, for which the cellular response to contact-guidance and curvature-guidance cues has been previously studied using another method.[37,38] In this case, the hBMSCs are used as a cell model to test the developed protocol for adherent cells. We then show a proof-of-principle demonstration for this method in a specific application area (i.e., blood vessel), for which the two cues are particularly relevant. We found that human myofibroblasts (hmFBs), the cells responsible for tissue growth and remodeling in blood vessels, distinctly respond to both contact- and curvature-guidance cues in a substrate-type-dependent manner. Subsequently, the orientation behavior of human umbilical vein endothelial cells (HUVECs) on the combination of cues is measured and compared to the response of hBMSCs and hmFBs.

Materials and Methods

Experimental Approach

In our approach, we employed a cell culture chip containing a library of 2.5D structures (Figure S1) and added patterning of the ECM protein on top of these 2.5D structures to provide contact-guidance cues. Patterning on 2.5D structures was enabled using light-induced molecular adsorption of proteins (LIMAP).[52] This technique enables spatial control of a user-defined UV-light pattern that is projected onto a substrate using a digital micromirror device (DMD). Figure summarizes the main steps involved in the patterning process. After fabrication of the structured PDMS cell culture chip using soft lithography, the surface of the substrates was passivated to prevent unspecific attachment of cells and proteins. By means of photoinitiated cleavage of the passivation layer using the DMD for user-defined patterning, regions of a predefined shape and size were created in this passivation layer.[52] This approach enables the fabrication of cell culture materials with patterned areas of any desired size and design. Incubation of the substrate with a desired protein solution [e.g., fibronectin (FN), collagen, or fibrinogen] subsequently leads to protein adsorption at the patterned regions. Cells were then seeded and could attach to the adhesive regions where the protein adsorbed to the surface. Confocal microscopy was used to image the 2.5D chip with protein patterns and cells, after which the response of single cells to the multi-cue environment was analyzed.

Figure 1

Outline of the experimental procedure of patterning on structured cell culture substrates. Substrates containing 2.5D features are fabricated by PDMS replica molding of a glass-etched negative mold and treated with oxygen plasma to clean the PDMS surface (step 1). The surface is then passivated to prevent unspecific attachment of proteins and cells (blue, step 2). After washing, the chip is transferred to the protein-patterning setup and turned upside down in a droplet of the photoinitiator (PLPP, green). UV light (purple) is projected in a pattern on the substrate (circumferential lines as a proof of principle) which initiates the cleavage of the passivation layer at the illuminated locations (step 3). As the pattern is user-defined, other patterning possibilities can be applied as well. Example images of crosshatches, longitudinal lines, or text patterns are shown (top: x–y views, bottom: x–z views). The fluorescently labeled protein (red) is incubated on the chip and can only adsorb at the locations where the passivation layer was cleaved (step 4). After protein incubation, the cells are seeded and only attach to the protein-coated areas on the structured substrate (step 5). The zoomed-in image shows a maximum intensity projection of hmFBs that are stained for F-actin (green) and nuclei (blue) on a cylindrical substrate (concave, κ = 1/250 μm–1) patterned with 10 × 10 μm (width × spacing) rhodamine-labeled FN lines (red). Scale bar: 50 μm.

Fabrication of Cell Culture Chips

The cell culture chips that provide 2.5D structured substrates for the cells were produced as described previously.[37] A negative mold containing semi-cylinders (convex and concave) with curvatures of κ = 1/125, 1/175, 1/250, 1/375, 1/500, 1/1000, and 1/2500 μm–1 and a length of 1000 μm, surrounded by flat areas, was designed using computer-aided design software (Rhinoceros 3D, McNeel Europe). This mold was produced from glass by FEMTOPrint (Muzzano, Switzerland) using a femtosecond-laser direct-write technique. The negative glass mold was used for the fabrication of a positive PDMS (Dow Corning, a Sylgard 184 Silicone Elastomer Kit, a prepolymer and curing agent, a 10:1 weight ratio) cell culture chip by pouring the PDMS prepolymer onto the glass mold, followed by degassing and curing for 3 h at 65 °C in a laboratory oven. After unmolding, a thin additional layer of PDMS was applied to the chip in order to create a smooth surface.[37] The positive PDMS chip was subsequently used for the fabrication of a negative PDMS mold, again by degassing and curing at 65 °C in a laboratory oven. In order to prevent bonding between the mold and the chip, tridecafluoro(1,1,2,2-tetrahydrooctyl)trichlorosilane (AB111444, ABCR) was used as silanization agent. Next, the negative PDMS mold was utilized for the production of multiple cell culture chips suited for the patterning and cell culture further downstream the experimental procedure described in Figure . In this study, cell culture chips made from PDMS, a commonly used cell substrate, were used as a proof of principle; other soft polymers can in principle also be used.

Scanning Electron Microscopy of the 2.5 Substrate

The topography of the cell culture chip was characterized using scanning electron microscopy (SEM) (Quanta 600, FEI). Under low vacuum, a large field detector and backscatter electron detector were combined to obtain images.

Fabrication of Flat PDMS Substrates

Flat PDMS surfaces were fabricated to optimize the passivation protocol and served as a control for the cell experiments. For this purpose, glass slides (Menzel-Gläser, 32 mm Ø, #1) were spin-coated with a PDMS solution (a 10:1 weight ratio) for 10 s at 500 rpm, followed by 50 s on 5000 rpm to obtain a final thickness of 10 μm. Samples were cured overnight at 65 °C.

Optimization of Substrate Passivation

Several passivation protocols were evaluated to optimize the quality of the passivation layer on PDMS substrates as well as the quality of the patterned protein after protein incubation (Table ). 3-Aminopropyltriethoxysilane (APTES; A3648, Sigma-Aldrich) and poly-l-lysine (PLL, P4707, Sigma-Aldrich) were used in combination with mPEG-succinimidyl valerate (mPEG-SVA; MW 5000 Da, Laysan Bio) as well as Pluronic F-127 (P2443, Sigma-Aldrich) to passivate the PDMS substrates in various ways.

Table 1

Stepwise Description of Five Passivation Strategies for PDMS Surfaces

passivation strategy	description	duration and temperature
gas-phase passivation with APTES	(1) place an open vial with APTES and the PDMS substrate in a hermetically sealed Petri dish	1 h, room temperature
	(2) gently wash 3× with HEPES (0.1 M) (8 < pH < 8.5)
	(3) incubate the PDMS substrate with 50 mg/mL mPEG-SVA in HEPES 0.1 M (8 < pH < 8.5)	1 h, room temperature
	(4) wash 3× with PBS
liquid-phase passivation with APTES	(1) immerse the PDMS substrate in a 1% APTES solution in Milli-Q	1 h, room temperature
	(2) wash 3× with PBS	5 min/wash
	(3) wash 3× with HEPES (0.1 M) (8 < pH < 8.5)
	(4) incubate the PDMS substrate with 50 mg/mL mPEG-SVA in HEPES 0.1 M (8 < pH < 8.5)	1 h, room temperature
	(5) wash 3× with PBS
vacuum passivation with APTES	(1) place the PDMS substrate and an open vial of APTES in a vacuum desiccator	overnight, room temperature
	(2) wash 3× with HEPES (0.1 M) (8 < pH < 8.5)
	(3) incubate the PDMS substrate with 50 mg/mL mPEG-SVA in HEPES 0.1 M (8 < pH < 8.5)	1 h, room temperature
	(4) wash 3× with PBS
passivation with poly-l-lysine and mPEG-SVA	(1) pretreat the PDMS substrate with oxygen plasma at 20–100 W	30–60 s
	(2) incubate the substrate with 500 mg/mL poly-l-lysine	30 min, room temperature
	(3) wash 3× with HEPES (0.1 M) (8 < pH < 8.5)
	(4) incubate the PDMS substrate with 50 mg/mL mPEG-SVA in HEPES 0.1 M (8 < pH < 8.5)	1 h, room temperature
	(5) wash 3× with PBS
passivation with Pluronic F-127	(1) immerse the PDMS substrate in a 1% Pluronic F-127 solution in PBS	5 min
	(2) wash 3× with PBS	5 min/wash

X-ray Photoelectron Spectroscopy Measurements

Flat PDMS samples {experimental conditions: PDMS, passivated PDMS, passivated PDMS patterned using LIMAP, and passivated PDMS patterned using LIMAP and incubated with FN [10 μg/mL in phosphate-buffered saline (PBS)]} were prepared as described above, washed with Milli-Q, and dried using N2. X-ray photoelectron spectroscopy (XPS) spectra of samples were recorded using a Thermo Scientific K-Alpha spectrometer equipped with a monochromatic, small-spot X-ray source and a 180° double-focusing hemispherical analyzer with a 128-channel detector. Spectra were recorded using an aluminum anode (Al Kα, 1486.7 eV, 72 W). Pass energies of 200 eV for the survey scans and 50 eV for the region scans were used. CasaXPS software (version 2.3.23) was used for the analysis and quantification of the spectra.

Patterning of Flat and Structured PDMS Substrates

Patterning of flat PDMS substrates was performed following an established protocol.[52] UV photopatterning was realized by means of LIMAP using Leonardo software (Alvéole, version 4.13, Paris). To pattern cell culture chips containing 2.5D structures, the chips were flipped upside down into a droplet of the photoinitiator [4-benzoylbenzyl-trimethylammonium chloride (PLPP), Alvéole, Paris] on a glass slide (Menzel-Gläser, 32 mm Ø, #1). During patterning, the UV light projection at a dose of 1000 mJ/mm2 was focused on the top and bottom of the convex and concave features, respectively. After patterning, the substrates were washed 1× with PBS before being incubated with fluorescently labeled FN (FNR01-A, Cytoskeleton, Inc.). Incubation times (5–30 min) and protein concentrations (10–50 μg/mL in PBS) were varied under sterile conditions. Finally, the substrates were washed again with PBS by pipetting up and down extensively.

Fabrication, Passivation, and Patterning of Polystyrene Cell Culture Chips

Polystyrene (PS) chips were fabricated by clamping a sheet of PS (190 μm thick, GoodFellow) to a negative PDMS mold of the cell culture chip. The PS imprint of the cell culture chips was formed after 30 min in the laboratory oven at 140 °C. PS cell culture chips were passivated using PLL and mPEG-SVA and patterned (lines, a 10 μm width and a 10 μm gap size, UV dose: 1000 mJ/mm2) and coated using 0.3 mg/mL collagen type I (bovine PureCol, 3 mg/mL, Advanced BioMatrix).

Cell Seeding and Culture on the Cell Culture Chip

Human bone marrow stromal cells (hBMSCs), as a cell model to test our system, were cultured at 37 °C and 5% CO2 in an expansion medium [Dulbecco’s modified Eagle’s medium (DMEM; Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS; Biochrom AG), 1% penicillin/streptomycin (Biochrom AG), and 1% l-glutamine (glutaMAX; Invitrogen)]. 1 mL of the cell suspension containing 20,000 cells was added to the photopatterned chip and incubated at room temperature for 10 min and at 37 °C and 5% CO2 for 1 h. After this incubation step, the culture medium was replaced and the cells were cultured on the chips for 24 h at 37 °C and 5% CO2 before fixation.

hmFBs were harvested from the vena saphena magna obtained from patients, as previously described and in line with Dutch guidelines for secondary use of materials, and previously characterized as myofibroblasts.[53] hmFBs were cultured in advanced DMEM (Gibco) supplemented with 10% FBS (Merck Millipore, BCBV7611), 1% penicillin/streptomycin, and 1% l-glutamine at 37 °C and 5% CO2. For seeding on the photopatterned PDMS chip, we used hmFBs at passage 4. A total of 20,000 cells per chip were seeded and directly placed in the incubator for 30–45 min. After initial attachment of the cells, the chips were gently washed to remove cells that did not attach during the first 30–45 min. After medium replacement, the samples were cultured at 37 °C and 5% CO2 for 24 h. Pooled HUVECs (Lonza, C2519A) were cultured in the endothelial cell growth medium with growth supplements (ECGM2 kit; PromoCell, C-22111) and 1% penicillin/streptomycin at 37 °C and 5% CO2. For seeding on the photopatterned PDMS chip, we used 20,000 HUVECs at passage 4. After seeding, chips were directly transported to the incubator for 60 min. After medium replacement, the samples were cultured at 37 °C and 5% CO2 for 24 h.

F-Actin and Nuclear Staining and Confocal Imaging

Samples were fixed with 3.7% paraformaldehyde (formalin 37%; 104033.1000, Merck) for 15 min at room temperature. After washing with PBS, F-actin was stained using Phalloidin-Atto (65906, Sigma-Aldrich) and nuclei were stained using 4′,6-diamidino-2-phenylindole dihydrochloride (D9542, Sigma-Aldrich). Stained chips were imaged using a Leica TCS SP5 or an SP8 confocal microscope with 20×, 0.7 NA or 10×, 0.4 NA objectives. Z-stacks were recorded at a 4 μm Z-spacing.

Image Analysis

To correctly map the cell images made on the cylindrical areas of the 2.5D features to a 2D plane for further morphometric analysis, we used a custom-made script in MATLAB (The Mathworks, Natick, USA). First, all cylinders were aligned vertically in the images, after which the edges of the cylinder were identified and used as boundaries for the mapping algorithm. Subsequently, all intensity values were mapped to the correct horizontal positions, depending on the known curvature and height of the cylinder and on the relative pixel location to the centerline of the cylinders. Morphological parameters of single cells, such as orientation, were analyzed using an automated algorithm in Wolfram Mathematica 11.1 (Oxfordshire, UK).[54] Cells located at the edges of the protein pattern or curved substrate or cells touching neighboring cells were excluded.

Statistical Analysis

To test for significant differences in the orientation data, the Kruskal–Wallis test with Dunn’s post hoc analysis (Holm’s adjustment method) was performed using R (version 3.4.3). Differences were considered significant for p-values < 0.05.

Results and Discussion

Surface Passivation of PDMS Substrates

To obtain contact-guidance cues for cells on PDMS substrates, protein adsorption and cell adhesion outside of the patterned areas should be minimized and ECM patterns must provide a suitable substrate for the cells to adhere. Therefore, several passivation methods based on Pluronics (Pluronic F-127) and PEG-based agents (APTES + mPEG-SVA, PLL + mPEG-SVA) were tested on flat PDMS substrates. Three different patterns—lines (Figure A), a homogeneous rectangle, and circles/squares (Figure S2)—were subsequently used to visually assess the pattern on each passivated substrate using rhodamine-labeled FN (50 μg/mL, 5 min) and with identical fluorescence imaging settings. The approaches that yielded substrates with homogeneous pattern intensity, well-defined edges, and no unspecific protein attachment were considered successful.

Figure 2

Digital pattern design (A) and patterning outcomes of PDMS passivated with (B) PLL + mPEG-SVA, (C) liquid-phase APTES + mPEG-SVA, (D) gas-phase APTES + mPEG-SVA, (E) vacuum APTES + mPEG-SVA, and (F) Pluronic F-127. After patterning, rhodamine-labeled FN is incubated (50 μg/mL, 5 min) and visualized using fluorescence microscopy. Scale bar: 50 μm.

The surfaces treated with gas-phase APTES and vacuum-phase APTES in combination with mPEG-SVA and Pluronic F-127 showed both a low pattern intensity and unspecific protein adsorption outside of the patterned area (Figures D–F and S2D–F). Passivation using liquid-phase APTES + mPEG-SVA led to a reduction in unspecific protein adsorption but resulted in inconsistent pattern homogeneity. For instance, the line and circle/square patterns were considered homogeneous (Figures C and S2C) but not the rectangular pattern (Figure S2C). In addition, blurred pattern edges were observed for all pattern designs (Figures C and S2). Substrate passivation and subsequent patterning on PDMS treated with PLL + mPEG-SVA resulted in patterns with consistent intensity and well-defined edges, without the presence of unspecific protein adsorption (Figures B and S2). This suggests that the oxygen plasma treatment, which was only applied on the surface incubated with PLL + mPEG-SVA, enabled stabilization of the PLL coating on the PDMS surface.[55] Thus, the coupling with mPEG-SVA can be expected to be more efficient, resulting in less unspecific protein adsorption.

XPS measurements show that in the passivated areas on PDMS, the carbon signals were higher compared to non-passivated areas, while the oxygen and silicon signals decreased (Figure S3), indicating the presence of PEG chains on the passivated samples. Furthermore, a nitrogen signal emerged in the spectra of the passivated samples at 400 eV, which is contributed to the N-hydroxysuccinimide group of mPEG-SVA. After photopatterning of the passivated areas, a decrease in carbon content was observed, while the oxygen and silicon content increased. This confirms the successful removal of the passivation layer on PDMS. After incubation with FN, the carbon signals increased and the oxygen and silicon signals decreased again in the XPS spectra of the passivated patterned samples (Figure S3 and Table S1).

Application of Protein Patterns in 2.5D and the Response of Human Bone Marrow Stromal Cells

Our previous work has shown that when human bone-marrow stromal cells (hBMSCs) were seeded on convex structures covered by collagen networks, the directionality of cell migration was tuned by directionalities of both the mesoscale structures and the nanoscale collagen fibers.[37] This demonstrates that hBMSCs can sense curvature- and contact-guidance cues and reorient their cell bodies according to these cues. Here, this mechanosensitivity of hBMSCs was used to explore the proposed 2.5D patterning approach.

First, we show that microscale contact-guidance cues alone could induce hBMSC alignment by patterning linear FN patterns (lines, a 10 μm width and a 10 μm gap size) on flat substrates (glass slides) using conventional microcontact printing (50 μg/mL, 5 min) (Figure A,B). The cell bodies and their F-actin fibers aligned in the direction of the protein lines, showing a clear contact-guidance response (Figure B). To see whether the quality of protein patterns on microcontact printed glass is comparable to that on PDMS substrates, we patterned both flat PDMS and glass slides using either LIMAP or microcontact printing. The LIMAP method showed better consistency in fluorescence intensity across multiple substrates compared to microcontact printing (Figure A). The LIMAP method was used for the patterning of linear FN lines in the circumferential direction on convex cylindrical PDMS substrates. By using rhodamine-labeled FN (50 μg/mL, 5 min), we showed that the line pattern covered the 2.5D features completely (Figure S4). We observed that hBMSCs attached to and formed adhesion points on the FN lines on the patterned convex cylinders (Figure C). Moreover, unlike the orientation behavior observed on flat substrates that were patterned with the conventional microcontact printing method, hBMSCs on curvatures of κ = 1/500 μm–1 and κ = 1/375 μm–1 did not align in the direction of the contact-guidance cues (Figure S4). Instead, cells were primarily oriented along the longitudinal axis of the cylindrical substrate.

Figure 3

(A) Intensity profiles of four lines (10 μm width, 10 μm gap size) of rhodamine-labeled FN (50 μg/mL, 5 min) using microcontact printing and the LIMAP approach on PDMS and glass substrates. (B) hBMSCs on microcontact-printed rhodamine-labeled FN lines (a 10 μm width, a 10 μm gap size) on glass surfaces. The merged image shows the F-actin cytoskeleton (green), nuclei (blue), and FN (red). hBMSCs show a contact-guidance-dominated response by aligning in the direction of the patterned lines. Scale bar: 50 μm. (C) hBMSCs on rhodamine-labeled FN lines (a 10 μm width, a 10 μm gap size) patterned on a convex cylindrical structure with κ = 1/500 μm–1. The merged image shows the F-actin cytoskeleton (green), nuclei (blue), and FN (red). Scale bar: 50 μm.

We note that this result of hBMSC orientation is in line with our earlier observation of cell migration directionality on curvatures with nanoscale contact-guidance cues (fibrillar collagen).[37] Specifically, for curvatures in the range of 1/125 ≥ κ ≥ 1/250 μm–1, cells aligned in the longitudinal direction of the cylinders instead of in the direction of the microscale contact printed lines (as shown here) and nanoscale cues from aligned collagen fibrils in the study by Werner et al.(37)

Distinct Responses of hmFBs on Patterned Convex and Concave Substrates

We next sought to apply the principle to study the orientation response of hmFBs to a combination of curvature- and contact-guidance cues. Our earlier experiments with adherent cells on curved substrates showed that the cells clearly formed adhesions and attached to the patterned FN lines (Figure C). However, we also qualitatively observed that the contact-guidance response of hBMSCs on LIMAP-patterned flat parts of the cell-culture chip was weaker than expected. In addition, we noticed viable cells adhering outside the patterned areas, suggesting that there may still be unspecific adsorption of FN on the unpatterned areas that was not significant enough to be detected using fluorescence microscopy. We also note that each cell type has individual properties (such as size, morphology, and contractility)[56] that may require prior optimization of the protein adsorption protocol in order to evoke a representative contact-guidance response by the LIMAP approach. For these reasons, we first performed optimization of the protein adsorption step with hmFBs.

In the case of hmFBs, a clear contact-guidance response is expected on line patterns (10 μm width and 10 μm gap size) as shown before by Buskermolen et al.(23) In order to optimize the FN patterning protocol for hmFBs, the incubation steps of FN on flat PDMS surfaces were varied for 5 min, 2 × 5 min, or 30 min in concentrations of 10 and 50 μg/mL. hmFBs on PDMS surfaces incubated with 50 μg/mL FN, as used earlier for hBMSCs, were observed to lack the alignment response typically seen for these cells (Figure S5A,C,E). These results suggest that hmFBs were able to adhere to the passivated layer or possibly unwanted adsorbed FN in between the line pattern. In contrast, for substrates incubated with 10 μg/mL FN, the cells aligned in the direction of the protein lines as expected (Figure S5B,D,F). Furthermore, for all surfaces incubated with 10 μg/mL FN, hmFBs were viable regardless of the duration of FN incubation, implying that 5 min of protein incubation is sufficient on PDMS surfaces. Based on these findings, we proceeded with the study of hmFBs on patterned 2.5D substrates using an FN incubation protocol consisting of 5 min incubation with a concentration of 10 μg/mL. Since hmFBs can be found throughout the thickness of the vessel wall, thereby experiencing both convex and concave substrates, we also developed patterned concave substrates alongside the convex substrates and examined hmFB adhesion on these structures.

We found that hmFBs predominantly align in the direction of the contact-guidance pattern on all concave curvatures (i.e., regardless of the magnitude of the curvature), similar to on flat PDMS substrates, as indicated by the cell bodies in Figure A (flat substrate), Figure 4B (curved substrates), and the red and black dots in Figure C. A closer inspection shows that the cells form long protrusions along the circumferentially oriented protein patterns, consistent with a contact-guided response on these concave structures. In contrast, on convex structures, a transition from a contact-guidance-dominated response (at a 90° cell orientation) on lower curvatures (κ = 1/2500 μm–1) to a curvature-guidance-dominated response (at a 0° cell orientation) on higher curvatures (κ > 1/2500 μm–1) was observed. A possible explanation for this behavior is given by Biton and Safran as the transition in orientation of cells on curved substrates can be caused by the competition of forces between the bending stiffness of stress fibers and shear stress resulting from cell adhesions and active contractility.[57]

Figure 4

(A) hmFB on a flat, patterned PDMS substrate with FN lines (red, 10 μm lines and a 10 μm gap size) stained for the actin cytoskeleton (green) and nucleus (blue). Scale bar: 50 μm. (B) Example outlines of single hmFBs on patterned curvatures used for quantification of the orientation response. The direction of the contact-guidance cues and curvature is indicated with red bars. Scale bar: 50 μm. (C) Orientation response of hmFBs on both flat and structured PDMS (concave and convex cylinders with curvatures ranging from κ = 1/2500 to 1/125 μm–1, with the degree of curvature schematically indicated on top, n = 11–42 cells). Individual cells are represented by dots and the median by a line. For concave features (red), a similar orientation response on the complete range of curvatures is observed compared to hmFBs on flat substrates (black). In the case of convex features (blue), a transition in the orientation behavior can be noted with increasing curvature. hmFBs tend to align in the direction of the contact-guidance pattern and curvature (90°) if κ = 1/2500 μm–1, whereas the curvature-guidance cue is dominant for curvatures of κ > 1/2500 μm–1 (0°). Asterisks indicate statistical differences compared to the orientation on flat substrates (*p < 0.05).

Endothelial Cells Show a Similar Orientation Response as hmFBs on Concave Substrates

Interestingly, hmFBs only show a transition in orientation behavior on convex, patterned substrates. A possible explanation for this behavior is that the cells lift the cell body upward to minimize the adhesion points and avoid a bent morphology on concave substrates.[36,46] On convex cylindrical substrates, however, the only way cells can circumvent a bent morphology is by means of reorienting in the longitudinal direction. This would imply that hmFBs are able to sense not only the degree of curvature but also its sign (concave vs convex). To further explore this implication, we also subjected endothelial cells to the combination of curvature- and contact-guidance cues on convex and concave cylinders. It has previously been shown that HUVECs have the ability to sense substrate curvature, but it is unknown how endothelial cells respond to the combination of contact-guidance and curvature-guidance cues.[27,58] The cytoskeletal organization of HUVECs is distinct from that of hmFBs, typified by peripheral stress fibers and less pronounced ventral stress fibers (Figures A and 5A), thereby potentially shifting the mechanical balance between stress fiber bending and contractility.[57] Unexpectedly, we observed that on convex, patterned substrates, HUVECs consistently exhibited aberrant morphologies and did not adhere and grow well. The reason for this behavior is as yet unclear and should be explored further. Here, we focused on the orientation behavior of HUVECs on the concave patterned substrates in order to further verify the lack of transition in the cell orientation behavior on concave substrates and whether this is a cell-type specific response. Figure B,C shows the orientation response of the HUVECs on concave cylinders patterned with 20 μm lines and a 20 μm gap size. It can be observed that HUVECs predominantly align in the direction of the contact-guidance cues (90°) across the complete range of curvatures, similar to the orientation response on flat, patterned substrates (Figure A). Statistical differences indicate a slightly less pronounced contact-guidance response of cells on patterned concave cylinders compared to flat substrates (except for κ = 1/175 μm–1). Despite differences in the cytoskeletal organization, the orientation response of HUVECs on concave, patterned substrates was comparable to the orientation response of hmFBs. We note that HUVECs are typically smaller than hmFBs and this difference in cell size may influence the ability to sense the substrate curvatures presented to cells in this study (κ ≤ 1/125 μm–1). It will therefore be interesting that future experiments examine HUVEC response on substrates containing even larger curvatures to unravel this intriguing observation.

Figure 5

(A) HUVECs on flat, patterned PDMS substrates with FN lines (red, 20 μm lines and a 20 μm gap size) stained for the actin cytoskeleton (green) and nuclei (blue). Scale bar: 50 μm. (B) Example outlines of single HUVECs on patterned concave curvatures used for quantification of the orientation response. The direction of the contact-guidance cues and curvature is indicated with red bars. Scale bar: 50 μm. (C) Orientation response of HUVECs on both flat and concave PDMS (curvatures ranging from κ = 1/2500 to 1/125 μm–1, with the degree of curvature schematically indicated on top, n = 29–91 cells). Individual cells are represented by dots and the median by a line. Overall, a similar orientation response on the complete range of curvatures (red) is observed compared to HUVECs on flat substrates (black). Asterisks indicate statistical differences compared to the orientation on flat substrates (*p < 0.05).

Practical Considerations and Limitations

The developed 2.5D protein patterning approach is a useful tool for uncovering cell behavior in multi-cue environments. Despite its versatility, we emphasize that fine-tuning of several experimental parameters prior to use in combination with a new material or cell type of interest is necessary. The most important is optimization of substrate passivation, as shown in Figure , which may be required when switching between different types of cell-culture chip materials (i.e., films and polymers). To show an example that the presented approach is compatible with materials other than PDMS, we created PS imprints of the cell culture chip, followed by passivation, photopatterning, and protein incubation (Figure S6). Here, we used collagen type I to also show the versatility in terms of protein coating. When using different materials, one should consider the effect of material properties, such as stiffness, as these are known to influence the behavior of cells.[59,60] In this study, the stiffness of the patterned materials (glass, PDMS) is not expected to differentially influence cellular behavior as PDMS has an estimated stiffness exceeding 1 MPa.[61] hBMSCs, hmFBs, and HUVECs are all considered insensitive for stiffness changes in this order of magnitude. Next, cell-type-specific adaptations to the protocol may be necessary in order to establish representative cell behavior, for example, due to trace amounts of unspecific FN adsorption to unpatterned areas.

Conclusions

In this study, we report a novel approach to create micropatterned 2.5D substrates that allow for the investigation of multi-cue environments on cellular behavior. Until now, studies reporting cellular responses to contact guidance or curvature guidance investigate these aspects in isolation, thus neglecting the complexity of the cellular environment. Here, we present an approach that resolves a long-standing technical challenge in the field and offers a possibility for a systematic, high-throughput, and highly controlled study of cell behavior in physiologically relevant 3D geometries, while maintaining the practical convenience and benefits of 2D systems. Moreover, the developed method is highly versatile and can be applied to any cellular system of interest with diverse substrate materials, with simple tuning of a few experimental parameters, such as protein concentration, incubation time, UV-dose, and cell seeding density.

The proof-of-principle experiments highlight the utility of our approach to construct in vitro biomimetic environments to study cellular response to multiple environmental cues. When subjected to both ECM protein patterns and convex curvatures, hmFBs aligned along the longitudinal axis of the cylinders when κ > 1/2500 μm–1. On patterned concave substrates, however, hmFBs and HUVECs predominantly aligned in the direction of the contact-guidance pattern on all studied concave curvatures. Since the patterning approach we introduce here can be used with various ECM proteins, it offers an interesting possibility to further study the effect of contact guidance as induced by other ECM proteins for cells, such as collagen type I, V, VI, and XII and fibrillin or a combination of these, on curved substrates in the future.[62,63]

Eventually, we envision to use the proposed methodology to contribute to broader knowledge on the effect of geometrical cues on the cellular orientation response. The flexibility of the developed approach enables variation of the pattern design, for example, the direction of the lines, line width, interline spacing, or the use of completely different pattern designs. Studying cell responses on a variation of substrates and patterns leads to a better fundamental understanding of the cell’s geometry sensing machinery, which is critical for various downstream cellular functions, such as differentiation, directed migration, and matrix production.[11,64] More generally, the developed approach enables systematic exploration of different types of cellular responses to biomimetic multi-cue environments and can be further expanded to increase the level of complexity toward 3D physiological environments (i.e., different materials, substrate topographies, patterns, and cell types). The gained insights could ultimately be translated to smart designs of microfluidic organ-on-a-chip platforms and 3D tissue-engineered scaffolds.