Laser-fabricated cell patterning stencil for single cell analysis

Precise spatial positioning and isolation of mammalian cells is a critical component of many single cell experimental methods and biological engineering applications. Although a variety of cell patterning methods have been demonstrated, many of these methods subject cells to high stress environments, discriminate against certain phenotypes, or are a challenge to implement. Here, we demonstrate a rapid, simple, indiscriminate, and minimally perturbing cell patterning method using a laser fabricated polymer stencil. The stencil fabrication process requires no stencil-substrate alignment, and is readily adaptable to various substrate geometries and experiments.


Background
The ability to manipulate and selectively localize cells into patterns or distinct microenvironments is critical for single cell analysis [1][2][3][4], tissue engineering [5,6], cell signaling studies [7][8][9], drug screening [10][11][12], and cell migration assays [13,14]. Exploring the population dynamics and communal contributions within heterogeneous cell populations is fundamental to furthering our understanding of disease pathology [15][16][17][18]. In recent years, much effort has been focused on developing innovative, active and passive cell patterning methods and applications thereof. Many active cell patterning and isolation methods utilize microfluidic systems, in which cells are manipulated and transported using fluidic forces. Inkjet-based cell 'printing' and deposition methods have proven effective at sorting and patterning cells at the bulk and single cell level, but are typically low throughput and raise concerns about cell stress responses [19][20][21][22]. A variety of microfluidic geometries have been used to pattern cells into hydrodynamic traps at single cell capture efficiencies nearing 100% for capture rates on the order of thousands of cells per minute [23][24][25][26][27][28][29]. While trap-based approaches are very high throughput, they may discriminate against particular cell morphologies or sizes with relevance for human disease [30]. Microfluidic trap environments also impose difficulties in delivering single cells to isolated microenvironments for further experimentation.
Droplet based microfluidics, which encapsulate single cells within medium-oil emulsion droplets, are highly effective at isolating cells at hundreds of droplets per second [31][32][33] and are cost-effective for biomolecular analysis of single cells. However, these approaches are poorly suited for studying temporal processes in live cells due to the limited supply of gas and nutrients in the droplet environment. It is also unclear how droplet technology can be integrated with on-chip analysis that require multistep processes such as single cell PCR [34]. An additional shortcoming of all microfluidic patterning and isolation approaches is that they subject cells to shear stress that can effect cell health, function, and gene expression [35].
Several non-hydrodynamic methods have also proven effective at actively patterning cells. Magnetic spot microarrays can localize magnetically labelled cells onto complementary features of cell patterning substrates [36,37]. Non-uniform electric fields have been demonstrated to polarize single cells thus creating a mechanism by which they can be patterned or even rotated in the absence of a label [38][39][40]. Laser and optical fiber based systems have been used to assemble, sort, and pattern live cells [41][42][43]. A prominent concern with these optical approaches is the large power output required to trap cells and the physiological damage that cells may incur due to heating [44]. Recently, fluidic devices utilizing acoustic fields have proven effective at spatially patterning [45,46], and rotating [47] cells with 5 × 10 5 times lower power exposure than optical systems [48]. However, all of these approaches require specialized equipment and expertise at the implementation step.
Many passive cell patterning methods achieve localization through chemical [49][50][51][52] or topographical [53,54] surface modifications, deterring adhesion to undesired regions and/or promoting adhesion to desired regions. This preferential adhesion patterning strategy has also been demonstrated with dynamic substrates, where surface properties can be modulated in real-time to alter adhesion susceptibility [55]. However, substrate surface modification is prone to select for cells with a particular adhesive behavior, and may discriminate against certain phenotypes [56]. A large body of evidence suggests that the distribution of adhesive phenotypes within cell populations has profound implications in biological development and disease pathology [57,58]. The biased nature of surface modification cell patterning suggests it may be ill suited for high throughput single cell analysis methods where isolation of representative populations is desirable. Further, it is well understood that extracellular matrix components that promote cell adhesion also profoundly influence cell physiology [59,60]. Other passive patterning strategies utilize traditional random seeding approaches but with physical barriers (stencils) to pattern cells onto accessible regions of substrates [61][62][63][64][65]. Because stencil patterning relies upon physical barriers, there is little to no phenotypic discrimination imposed upon the seeded population, so long as the stencil through-holes are large enough to be cell-size indiscriminate. However, the use of a cell patterning stencil to seed cells into predefined features of a substrate typically requires microscale stencil-substrate alignment, presenting a challenge in the utilization of typical stencils. Therefore, despite significant recent advances, major challenges remain in obtaining large, representative quantities of isolated single cells. Here, we demonstrate the use of a rapidly produced laser-fabricated polymer stencil to pattern cells into wells of microarrays without the need for stencil-substrate alignment. Laser ablation is used to create holes in the polymer film, while it is immobilized and stretched across the microwell array. The stencil remains fixed relative to the underlying substrate during cell seeding. Seeded cells can only access the microwell array by settling through the holes, and are blocked by the film from accessing other areas of the substrate. In our geometry, this results in cells being restricted to the inside of the microwells. After cell adhesion, the stencil is removed and the patterned cells are ready for experimental treatments and analysis. This approach does not require specialized expertise or equipment for the implementation steps. Further, it does not introduce any additional stress to the cells relative to normal cell culture and does not decrease cell viability. We demonstrate that this method greatly increases efficiency of seeding cells into defined locations on a substrate.

Device fabrication
Fused silica wafers (4 in. diameter, 500 μm thickness) were etched using standard photolithography to create 37 arrays of 2980 microwells (20 μm deep, 80 μm inner diameter, 120 μm outer diameter) hexagonally packed within 9 mm by 9 mm footprints. The wafer was partitioned into the 37 individual microwell array cell-seeding substrates using a dicing saw. The substrates were sonicated for 30 min in 1X alkaline detergent to remove particles and fibers, followed by rinsing and sonicating in deionized water for an additional 30 min. Substrates were then dried in a 105°C oven and stored for later use. X-Ray Fluorescence (XRF) film (3090, Chemplex, Palm City, FL) was stretched using an XRF sample cup, rinsed with ethanol, dried with nitrogen, and secured to the surface of the microwell array substrate. A purpose-built fixture was used to clamp the polymer film onto the microwell array substrate and also provide a reservoir for holding cell culture medium during equilibration and cell seeding (Fig. 1a, b). The area of the reservoir footprint that cells are seeded onto was 236 mm 2 . The fixture ensures that the position of the film and underlying substrate are fixed relative to one another through the fabrication and cell seeding process. The fixture was secured to the mechanical stage (ATS250, Aerotech, Pittsburgh, PA) of a laser fabrication environment where a 355 nm UV laser (AVIA 355-3000, Coherent, Santa Clara, California) was focused through the substrate and onto the surface of the polyester XRF film (Fig. 1c). The laser was tuned to output radiation below the ablation threshold of fused silica (48 μJ/pulse, 40 ns pulse width, 6 kHz repetition rate, defocused to 30 μm) to avoid substrate ablation, while still having sufficient energy to form a pore in the XRF film. Poration is achieved through the redistribution of film away from the site of exposure (through ablation or heat induced polymer restructuring) leaving a pore slightly larger than the irradiated area (~35 μm) (Fig. 1d).
The fabrication environment was programmed to traverse the geometry of the cell seeding substrate, aligning to the center of each well and perforating the film through 500 ms UV exposures. The porated substrate assembly was then placed into a 35 mm petri dish and plasma treated to hydroxylate the surface of the substrate and XRF film, sterilize the cell seeding surface, and promote cell adhesion [66].

Cell seeding
Prior to use, plasma treated fixtures were equilibrated with cell culture medium at 37°C for 4-18 h. Medium was then removed by pipetting and 600 μL of cell suspension at the indicated concentration was added to the reservoir ( Fig. 1e). Cells were allowed to adhere to the substrate for 18 h under normal culture conditions. Then the fixtures were disassembled and the stencil film was peeled off with forceps, effectively removing cells not localized to wellinteriors (Fig. 1f). The disassembled fixtures were disinfected with 70% ethanol, rinsed 3× with dH 2 O, then air dried for reuse. The stencil films were discarded after a single use. Substrates were visually inspected prior to cell seeding to verify that the stencil and microwell array were well aligned (Fig. 1g). A second visual inspection was performed prior to removing the stencil film to qualitatively evaluate cell health and morphology on the stencil film, and within the wells (Fig. 1h). A similar procedure was used to assemble the no-stencil control assays wherein the microwell array substrate was placed atop the polymer film inside the fixture. Cells were then seeded into the reservoir onto the uncovered substrate.  labeled after the 18 h adhesion period and prior to removal of the seeding stencil. In order to evaluate the distribution of cells on the microwell array substrate, cell nuclei were labeled with the NucBlue® Live ReadyProbes® Reagent (R37605Thermo Scientific). The fixtures were disassembled and the stencil film removed as described, then the microwell array substrates were transferred to a six well plate containing fresh cell culture medium and the nuclear label. Cells were incubated for 10 min and then imaged via wide-field microscopy (Fig. 1i). The microwell array was imaged by phase contrast and the nuclear and viability stains were imaged via wide-field epifluorescence. The excitation and emission wavelengths in nm were: 360/460 to detect nuclei (blue), 485/540 to detect live cells (green), and 540/600 to detect dead and dying cells (red). All images were collected using an inverted, Nikon TE2000-U fluorescence microscope with a 4× plan apo lens, NA = 0.2, and a Hamamatsu Orca Flash 4.0 digital CMOS camera. The entire microarray was imaged in each channel by stitching individual image fields. Image acquisition was automated using NIS-Elements software.

Image analysis
Cell viability and localization were evaluated by counting cells from 400 wells from each microwell array. The 400 wells are a randomly sampled subset from the total of~2980 well detections provided by a normalized 2D cross correlation algorithm implemented in National Instruments LabVIEW software. For cell distribution, labeled nuclei were identified as localized to the interior or the exterior of each well. The exterior of the well was defined as the outside of the well or the lip of the well. Cells located outside of wells were automatically assigned to the nearest well by comparing the Euclidean distances between the cell and the centroids of the 400 sampled wells. The efficacy of the cell patterning stencil was evaluated by calculating the localization efficiency, which was defined as the percentage of cells localized to the interior of wells relative to the total cell count. The biocompatibility of the polymer stencil and the stencil removal process was evaluated by examining the viability of cells in wells, seeded with or without a stencil. Dead (or dying cells) were determined as the percent of red-labeled cells relative to total (blue) nuclei. Live cells were visually confirmed by the presence of green-labeled cytoplasm, however, this label was not used for quantification.

Statistical analysis
Data were pooled from a minimum of three independent experiments. Data were analyzed by Student's t-test and Mann-Whitney U test using the R statistical computing environment. P-values of <0.05 were considered significant.

Biocompatibility of polymer film stencil
We used a commercially available cell viability assay to investigate the biocompatibility of the laser fabricated polymer stencil. We compared viability of cells seeded through the stencil to cells seeded directly onto an uncovered microwell array. Biocompatibility was measured in terms of cell viability, defined as the percentage of cells within wells that were neither dead nor dying, as indicated by absence of red labeling of the nuclei. Cells were manually counted in the blue (all nuclei) and red fluorescent images. The red, dead cell, channel was used for counting since this signal is confined to the nucleus, and thus there is negligible overlap or contact between individual cells, allowing greater accuracy in counting. Figs. 2a and b show representative images of nuclei and LIVE/DEAD labeled HME1 and MDA-MB-231 cells respectively. The stencil reduced the total number cells because it blocks the majority of cells from adhering to the substrate. Use of the stencil did not reduce cell viability in either cell line; in fact, the stencil seeded cells had a very slight increase in survival. For both the stencil and no-stencil seeding environments, cell viabilities were found to be in excess of 95% on average across all experimental seeding densities (Fig. 2c).

Cell patterning localization efficiency
To evaluate the effectiveness of the cell patterning stencil for localizing cells, localization efficiencies were compared between stencil-seeded and directlyseeded substrates. Cells seeded at 100 K cells/mL through a stencil were highly localized to wellinteriors ( Fig. 3a) with localization efficiencies of 97% for K562 cells, 82% for HME1 cells, and 92% for MDA-MB-231 cells. The stencils increased the percentage of cell in microwells by approximately 3-fold and 4-fold for HME1 and MDA-MB-231 cells, respectively. This represents a significant increase in localization efficiency (p < 0.01) relative to controls seeded without stencils (Fig. 3b). Seeding through the stencil had an even more profound effect on the cell distribution of K562 cells. This cell line is generally considered non-adherent and is grown in suspension culture. As expected, when these cells were seeded onto bare microwell array substrates, few cells were observed after the substrate was removed from the fixture and placed in fresh medium in a petri dish. Generally, 0-5 cells were observed across the entire array. Interestingly, when these cells were seeded through the stencil, many cells were retained in the microwells after removal of the stencil and transfer to fresh medium. The mechanism for this is not completely clear, but one possibility is that the stencil reduces turbulence of the medium close to the substrate surface. With this reduction in fluid motion, some very low level of adhesion to the substrate is sufficient to immobilize these cells in the microwells. It should be noted that though they are retained through the disassembly of the fixture and transfer to fresh medium, rinsing these cells with moderate force, or multiple medium changes will dislodge them.

Effects of seeding density
We found that the enhanced localization efficiency provided by the seeding stencil was relatively independent of seeding density with a 4-fold average increase in localization across all seeding densities for MDA-MB-231 cells and 3-fold increase for HME1 cells (t-test; p < 0.0001) (Fig. 4). These findings confirm that the stencil is highly effective at controlling the localization of cells using various cell lines and across a wide range of seeding densities.
The ability to efficiently pattern cells to defined locations is of significant value in single cell analysis. With this in mind, we next sought to investigate the frequency of single cell occupancy in microwells seeded through our stencil. The relationship between the seeding density and the percentage of sampled wells containing single cells (single cell occupancy) was found to exhibit different trends for each cell line. The K562 cell line, which requires the use of a stencil to achieve successful adhesion, exhibited very little correlation between seeding density and single cell occupancy, which remained at approximately 17% through all experimental densities  (Fig. 5). Single cell occupancies of HME1 cells trended toward a loosely parabolic dependence on seeding density, with densities near 50 K cells/mL yielding the largest fraction of single-cell wells for both stencil seeded and control cells. Stencil-seeded and directly-seeded HME1 cells did not differ significantly in single cell occupancy (Mann-Whitney U test: p > 0.1) (Fig. 5). In contrast, single cell occupancies of MDA-MB-231 were positively impacted by stencil-seeding. MDA-MB-231 stencilseeded substrates with seeding densities above 50 K cells/mL were found to be significantly higher (Mann-Whitney U test: p < 0.05) than directly-seeded controls, averaging at about 31% single cell occupancy. We observed that at high concentrations, these cells tend to clump together into loosely adherent aggregates. We therefore speculate that the stencil acts as a cellular sieve, favoring single cells or doublets, while inhibiting the passage of larger aggregates. The exhaustive distributions of empty, single, double, triple, and quadruple occupied wells under various seeding conditions was also investigated (Fig. 6). These data support the idea that the stencil shifts the balance of cell distribution in wells toward single cells at the expense of wells containing four cells in the MDA -MB-231 cell line.

Discussion
Cell patterning has broad applications, particularly in the field of single cell analysis. Though many approaches have been described, limitations remain such as expense, difficulty of use, and cell stress. The stencil method presented here resolves these challenges. We have implemented this approach in conjunction with single cell metabolic analysis. In related work, our group has developed a platform for high throughput analysis of single cell oxygen consumption rate and extra cellular acidification [67]. In this approach a microwell array, as described here, is seeded with cells, then sealed with a lid coated with an oxygen and pH sensitive, fluorescent sensor film [68][69][70][71]. Each well is temporarily, hermetically sealed with the sensor lid, allowing measurement of metabolic output from individual cells or small groups of interacting cells. For this particular application, exclusion of cells from the microwell lips is critical for effective sealing of the array as well as reducing cellular responses to components released from damaged cells caught between the sealing surfaces. After random seeding through the stencil, we employ automated image analysis to determine the number of cells in each well and correlate the number of cells with metabolic readouts. While higher single cell occupancy, approaching 100%, can be achieved by modern microfluidic methods [72], the use of a stencil has the benefit of not subjecting the cells to the high shear stress environments or impacts characteristic of high-throughput microfluidic devices. Biocompatibility of the stencil was demonstrated using a LIVE/DEAD assay, which indicated no increase in cell death. Cell viability was comparable to what has been previously reported for surface affinity patterning at 95% [50], micro-trap patterning at 94% [25], cell deposition patterning at 90% [19], and block-cell patterning at 95% [23] for various cell lines.
An additional advantage of our stencil approach is that it separates the cell patterning function from downstream analysis or additional experimental manipulations. For example, the design of our metabolic analysis platform [67], is optimized for sealing the microwells for efficient metabolite measurements. Engineering additional functionality for cell trapping such as a microfluidic component, would complicate both the device fabrication and experimental protocol, as well as likely compromise the analytic functionality. The stencil design also dramatically simplifies the cell loading procedure for the end user. Microfluidicbased cell patterning requires specialized equipment, expertise, and relatively demanding optimization. Use of the patterning stencil described here, requires only standard cell culture skills and no specialized or costly equipment.
Though we describe a specific application, this general approach, a removable stencil fabricated in situ on a functional substrate, is extensible to a wide range of applications and geometries. In recent years there have been remarkable advances in lab-on-chip and high throughput single cell analytic technologies [73]. Many of these platforms require precise localization of individual cells prior to cell lysis [74] or on-chip analyses such as qPCR [75,76], proteomics [77], genome sequencing [78] or cell signaling [79]. Integrating stencil-assisted cell seeding into these emerging technologies has the potential to simplify device fabrication and dramatically reduce the complexity of experimental protocols. This method retains the ease of random cell seeding, with equivalent or increased single cell deposition, while dramatically reducing the occurrence of "off-target" cells.

Conclusions
We demonstrate a laser-fabricated cell seeding stencil to be highly effective at patterning cells into features of microfabricated substrates. Stencils are readily adaptable to specific substrates or experimentally desirable geometries and multiple cell types. This approach does not expose cells to any physical stresses beyond those of Fig. 6 Well occupancy distributions. The occurrence (y-axis) represents the percentage of wells containing a specified number of cells (0-4, x-axis) at each seeding density. Note that K562 cells were non-adherent on the microwell array in the absence of a stencil. The data were pooled from three independent experiments. Error bars indicate SEM standard cell culture. Since the cell stencil approach relies on a physical barrier instead of differential cell adhesion to achieve isolation, it does not select subpopulations of cells based on adhesion properties or expose cells to reactive substrates that may alter cellular physiology. Preparation of the stencil is simple and high throughput; stencils for 2980-microwell arrays can be prepared in less than an hour. Because the stencil is fabricated while affixed to the cell seeding substrate, there is no need for microscale stencil-substrate alignment, dramatically simplifying its usage. Since the specificity of cell localization is based on the design of the stencil and not the cell seeding technique, this approach is readily accessible to researchers in biological fields using standard cell culture techniques.