Micropatterning the organization of multicellular structures in 3D biological hydrogels; insights into collective cellular mechanical interactions

Control over the organization of cells at the microscale level within supporting biomaterials can push forward the construction of complex tissue architectures for tissue engineering applications and enable fundamental studies of how tissue structure relates to its function. While cells patterning on 2D substrates is a relatively established and available procedure, micropatterning cells in biomimetic 3D hydrogels has been more challenging, especially with micro-scale resolution, and currently relies on sophisticated tools and protocols. We present a robust and accessible ‘peel-off’ method to micropattern large arrays of individual cells or cell-clusters of precise sizes in biological 3D hydrogels, such as fibrin and collagen gels, with control over cell–cell separation distance and neighboring cells position. We further demonstrate partial control over cell position in the z-dimension by stacking two layers in varying distances between the layers. To demonstrate the potential of the micropatterning gel platform, we study the matrix-mediated mechanical interaction between array of cells that are accurately separated in defined distances. A collective process of intense cell-generated densified bands emerging in the gel between near neighbors was identified, along which cells preferentially migrate, a process relevant to tissue morphogenesis. The presented 3D gel micropatterning method can be used to reveal fundamental morphogenetic processes, and to reconstruct any tissue geometry with micrometer resolution in 3D biomimetic gel environments, leveraging the engineering of tissues in complex architectures.


Introduction
Precisely locating and patterning the organization of cells in-vitro have enabled scientific and technological advancement in diverse areas in cell biology, tissue engineering, and drug testing.For example, culturing cells on defined micropattern adhesive substrates, led cells to adapt their cytoskeleton architecture and contractility, which impacted cell migration, growth, and differentiation [1][2][3][4].In the field of tissue engineering, patterning cells at precise organizations allows to reconstruct specified tissue architectures, whether they be blood vessels or skin layers, as examples.
As the patterned cells develop, they maintain cellcell and cell-extracellular matrix (ECM) contacts that facilitate the formation of a functional tissueconstruct [5,6].In addition, micropatterning of individual cells or cell clusters in defined array configurations can be used to create controlled niches or microwells to analyze drug efficacy and toxicity of drug components [7,8].
Most available micropatterning techniques are performed on 2D substrates, such as glass or plastic, to generate a defined cell adhesion pattern.A common 2D patterning technique is microcontact printing that uses a polydimethylsiloxane (PDMS) stamp with desired micro-features to print ECM proteins onto a culture substrate to which cells adhere [9,10].Alternatively, photo-patterning methods involve projection of a UV light with or without a photomask, onto a light-sensitive polymer coating to define regions for cell adhesion [11].However, in a plastic dish or on top of a glass coverslip, cells encounter a homogeneous, flat, and highly rigid adhesion substrate which does not reflect the physiological microenvironment.A multitude of studies have demonstrated drastic differences in cell morphology, migration, and gene expression when cells are cultured on a 2D rigid as opposed to embedded in a 3D soft hydrogel [12,13].
The differences in cell behavior between 2D and 3D systems have opened a whole field that aims at constructing 3D cell culture model systems, where hydrogels serve as central materials that more closely mimic tissue properties [14].When culturing cells in 3D hydrogels, cells are typically embedded randomly throughout the 3D gel, resulting in an environment that is highly artifactual, uncontrolled, and nonreproducible [15].The random positioning of cells in the 3D hydrogel results in poor control over the cell microenvironment, due to the variability in cell density and cell-cell distances across the gel regions.The number of neighboring cells and their mutual distances may have a profound effect on cell-cell interactions via biochemical [16,17] or mechanical signals [18,19].Mediated by the elasticity of the soft gel, cells can mechanically interact over long distances by applying pulling forces on their environment and deforming it [15,[20][21][22][23][24][25][26].In this cell-cell mechanical interaction process, the distance between cells directly influences the ability of neighboring cells to sense and respond to each other [15,20].Hence, micropatterning of cells in precise distances and spatial arrangements in hydrogel environments, is an important step toward the development of in vitro model systems that better reflect the complexity of the native microenvironment and for studies of individual cell responses to microenvironment conditions and biomechanical stimuli.
Early attempts to pattern cells in hydrogels was introduced in polyacrylamide (PAA) gels using microfabricated stencils or stemps [27,28].However, synthetic hydrogels such as PAA fail to undergo enzymatic degradation, lack cell adhesion moieties and are limited to surface-based patterns (2.5D).Cell micropatterning in natural biological hydrogels, such as collagen or fibrin, is more physiologically appealing as they better recapitulate the fibrous structure and mechanical ECM properties, and support full embedment of cells within the 3D gel.The use of natural hydrogels is also more applicable for generating grafts for clinical studies.Cells can be patterned in 3D collagen gels by creating defined cavities in the gel using a PDMS stamp and seeding cells in these cavities [29].However, this technique is rarely used to pattern individual cells, likely because of the challenges associated with creating micro-sized cavities and maintaining their stability after stamp removal.Several advanced technological attempts have managed to pattern cells in 3D hydrogels by first generating a defined cell pattern on a 2D substrate and then lifting or inverting the cell pattern into a supportive 3D hydrogel [30][31][32].However, in these studies, the 2D cell pattern was established using sophisticated procedures such as DNA linkers [31] or dielectric forces [32].More recently, the force of gravity was used to transfer cells trapped in microwells into a collagen hydrogel [33].Other advanced techniques have used bioprinting technologies, such as inkjet or extrusion based bioprinting which can reach fabrication resolution on the order of 50 µm [34,35].Lightbased bioprinting methods such as laser-based direct writing [36], or two photon laser scanning [37] can reach micrometer scale resolution [35].Surface acoustic waves [38] and magnetic forces have also been applied to locate cells at desired positions in 3D collagen hydrogels [39].Altogether, most methods for cell patterning in 3D hydrogels demand special fabrication techniques, expensive devices, or challenging preparations and because of that their use is more limited and less common than the use of 2D patterning techniques.
Here, we present an accessible and efficient method for micropatterning of large-scale arrays of individual cells and cluster of cells in biological fibrin and collagen hydrogels, with precise control over cell cluster size, cell-cell separation distances, and number of neighboring cells.With this method, a dedicated silicone strip that mechanically supports the 3D gel, is placed over a 2D glass carrying a predefined cell pattern, created by standard photolithography tools, and then peeled-off, thereby transferring the organized cell pattern from the 2D glass to a 3D gel.The final pattern features a reproducible, large-scale, and physiologically relevant environment that is suitable for microscopy imaging and biomechanical stretching.It can be used to explore fundamental mechanisms of cell-cell interactions as well as for various biomedical applications in tissue engineering and drug screening.Here, we describe the design and fabrication of the micropatterned gel and use it to study the contribution of cell-induced forces and matrix remodeling in collective cellular repsonse relevant to tissue morphogenesis.

Cell culture
NIH 3T3 fibroblasts expressing green fluorescent protein (GFP)-actin and GFP-labeled cell line derived from murine epithelial cells that were infected with an HRAS mutation [40], a key oncogene associated with cancerous tumors (obtained from Prof. Moshe Elkabets, Faculty of Health Sciences, Ben-Gurion University) were cultured in DMEM supplemented with 10% fetal bovine serum, nonessential amino acids, sodium pyruvate, L-glutamine, 100 U ml −1 penicillin, 100 µg ml −1 streptomycin, and 100 µg ml −1 neomycin, in a 37 • C humid incubator.In all experiments, the 3T3 fibroblasts cells were used in passage 27-35.

Micropatterning process (all indicated steps refer to figure 1(A))
Step#1: patterned coverslips (CSs) were fabricated using a photolithography technique implemented with a commercialized micropatterning kit (4DCell © , Montreuil, France).A 24 mm diameter CS was passivated with poly-lysine Poly(ethylene glycol) (PEG) solution (PLL-g-PEG, 0.1 mg ml −1 , from 4DCell © ) and incubated for 30 min.The CS was then rinsed with distilled water and placed on a customized photomask made of a quartz substrate with a thin chrome layer.The photomask and the CSs were placed on top of each other inside a dedicated aluminum frame, which was tightly closed using screws, and then placed inside a UV ozone cleaner, ProCleaner™ Plus system (220 nm wavelength) for 10 min.After UV exposure, the patterned CSs were gently removed and rinsed with distilled water.
Step#2: silicone rubber (0.5 mm thick polysiloxane, from McMaster-Carr) was cut using a laser cutting machine (Universal) into a 24 mm diameter circle containing four cutouts of 6 mm diameter each.The cut-out silicone was then washed with 10% sodium hypochlorite and sterilized by soaking it in 70% ethanol.The silicones were dried and attached to the 24 mm patterned CS to serve as molds for cell suspension and gel formation.
Step#3-4: cell suspensions (3T3 or HRAS cancer cells) were added to each silicone mold (∼9 × 10 3 cells per silicone cutout).Then, the cells were incubated for several hours, and the unattached cells were washed away.The samples were then incubated for up to 48 h in a 37 • C humid incubator until acquiring the desired patterns.
Step#5-7: after acquiring the cell pattern on the CS, 14 µl of fluorescently-labeled fibrinogen (10 mg ml −1 ; Omrix Biopharmaceuticals) was added to each silicone mold, and mixed gently with 14 µl of thrombin (10 U ml −1 ; Omrix Biopharmaceuticals).Samples were placed in the incubator for 30 min to polymerize.After polymerization, warm medium was added to cover the gels and samples were incubated for an additional 10 min.The silicone was then gently lifted with the micropatterned gel using tweezers.
The following table 1 summarizes the cell pattern specifications that were used throughout this study:

Confocal microscopy
The gel-patterned cells were imaged with Zeiss LSM 880 confocal microscope with an Airyscan fast detector, using 40× water immersion and 10× objective lenses (Zeiss) with a 30 mW argon laser (wavelengths 488 and 514 nm).Throughout imaging, cells were maintained in a 37 • C, 5% CO 2 incubation chamber.

Cell-cluster quantification
The volume and diameter of the clusters were measured from the 3D images using the Imaris software (version 8.4.1; bitplane).The surface analysis tool was used to calculate the cluster volume.Cluster diameter was calculated using the spot analysis tool.

Quantification of micropatterning efficiency
Micropatterning efficiency was determined by calculating the number of clusters in a specific patterned area on a CS and comparing it to the number counted in the same patterned area in the gel after 'peeloff ' .The clusters in each image were identified using the open CV blob detection algorithm.

Collagen gel preparation
Rat tail collagen type I (Corning ® ; 8-11 mg ml −1 ) was dissolved in 1 N NaOH and diluted with a mixture of ddH2O, HEPES, NaHCO 3 and cell medium to a final concentration of 3 mg ml −1 .Then, 30 µl of the collagen mixture was placed into each silicon hole and incubated for 30 min to allow for polymerization.Following polymerization, warm medium was added to cover the gels.All collagen gel preparation procedures were performed on ice.

Quantifying cell migration dynamics in a cell-patterned gel
To quantify the dynamics and preferential direction of cell migration, the entire cell pattern was divided into separate domains, each surrounding a single cluster.First, cluster positions in each image were manually identified and only those with exactly six nearest neighbors were selected for analysis.A rectangular domain around each chosen cluster, which extended to its neighbors, was defined.The brightness of the cell-GFP fluorescence channel was normalized to the fluorescence signal originating from the cluster center, with a distribution between 0 and 1.Then, all clusters were superimposed together to obtain an average cell distribution C (r, θ) with r being the radial distance from the cluster and θ the angular direction, around a typical cluster.
To examine the directional pattern of cell migration, an annular domain extending from r1 = 5/12 to r2 = 1/2 (with r = r/R being the normalized radial distance from the cluster, and R being the distance to the nearest neighbors) was chosen.The average cell distribution as a function of direction < C (r, θ) > r1< r<r2 was obtained.r2 was chosen as 1/2 because for r > 1/2, C mostly contained cells originating from neighbors, and r1 = 5/12 was chosen in order to include in the analysis the cells that had migrated the farthest.The radial cell distribution was obtained by averaging C over all angles: < C (r, θ) > θ .

Cell micropatterns on 2D glass for migration analysis
To create a cell array pattern on a 2D CS to serve as a control experiment for the gel micropatterns, we followed steps #1-4 mentioned above in section 2.2 'Micropatterning process' .Cell migration and spreading on the glass initiated after about 48 h from the micropatterning process despite having the PEGpassivation layer.This was probably because of the natural degradation of the passivation layer and/or deposition of ECM proteins by the cells themselves.We note that cell adhesion to the CS in this case was not optimal but still permited some degree of spreading and migration, sufficient for analysis of directional migration.To analyze migration directionality, cells were imaged after 48 h from the time they started to migrate out of the clusters.

Analysis of integrins, stress fibers and gel fibers orientation
For analysis of integrins, stress fibers, and gel fiber orientation, images were taken with Zeiss LSM 880 confocal microscope using the Airyscan fast detector and 40× water immersion objective.Image analysis was done by ImageJ software.Each image was divided into different region of interests (inside a cell cluster or in a matrix band), and each region was rotated so gel fibers would be located horizontally, then the images were split into separated channels of gel fibers, and integrins/stress fibers.For anlaysis of orientation, we used the directionality plugin in ImageJ sofware.

Results
We describe a robust and accessible method to precisely position individual cells or cell-clusters in predefined configurations in biological hydrogels of fibrin and collagen.These hydrogels are routinely used as 3D matrices to embed cells for tissue engineering applications and cell-matrix interaction studies [15,41].We created cell arrays in varying cell-cell spacing to study the mechanical interaction between contractile cells through deformation of the fibrous gel.This biomechanical process is relevant to biological processes such as morphogenesis, cancer metastasis and wound healing where cells can mechanically sense and respond to neighboring cells through elastic deformation of the matrix [42].

Micropatterning of cells in gels using a 'peel-off ' process
To micropattern cells in 3D biological hydrogels in precise arrangements, we first pattern the cells on a 2D glass surface (CS) and then transfer the cell pattern to the hydrogel, as schematically shown in figure 1(A).Initially, cells are micropatterned in the desired geometries on a glass CS, using an established photolithographic-based approach.A PEGpassivated CS (see methods) is exposed to UV light through a custom photomask designed with the desired pattern geometry.This results in a patterned CS, in which cells only attach to the exposed adhesive areas (for example the 5 × 4 spot array shown in step#1, figure 1(A)).The size of the adhesive spot will determine the number of cells that will attach at each spot and eventually the size of the resulting cell cluster.Of note, other methods for 2D micropatterning can be used, such as the PDMS stamping method [27,43].Afterward, a punctuated silicone strip is attached to the patterned CS to serve as a mold (figure 1(A), step#2).Cell suspension is then added to the silicone mold and incubated for several hours, allowing cells to attach to the designated adhesive spots (figure 1(A), step#3).The unattached cells are washed away, and the pattern is stabilized for several hours to a few days (figure 1(A), step#4).Then, a hydrogel precursor solution (i.e.fibrin or collagen components) is poured into the silicone mold and left to polymerize while in contact with both the underlying CS-patterned cells and the silicone strip (figure 1(A), step#5).The natural tendency of fibrin and collagen gels to stick to the silicone strip and to the cells themselves allows to 'peel off ' the gel with its embedded cell pattern, thereby facilitating the 2D-to-3D transfer (figure 1(A), step#6).The patterned gel can then flipped upside down, such that the patterned cells are positioned in the top layer of the gel, slightly First, a specific geometric pattern is generated on a CS (step 1).As an example, a 5 × 4 spot array is described.A silicone strip with a cutout (hole) is attached to the patterned CS to serve as a mold to sustain liquids (step 2).Cell solution is poured into the mold (step 3) and cells are allowed to settle on the adhesive spots (step 4).Hydrogel precursor solution is added to the mold and allowed to polymerize (step 5).The cell pattern is then transferred to the hydrogel by peeling-off the silicone (step 6).Finally, the micropatterned gel can be placed such that the cell pattern faces up (step 7).(B) Images of the actual system: incubation of cells in the silicone mold positioned on top of the micropatterned CS and after peeling-off the silicone mold embedded with a fibrin gel and patterned cells, using tweezers.Here, the silicone contains 4 gels.embedded within it (figure 1(A), step#7, SI movies 1 and 2).Of note, the gel is 3D but cells stay in a plane, making it convenient for fast confocal imaging with reduced z-stacks, and enabling imaging with a regular epifluorescence microscope (not confocal).Images of the actual system are shown in figure 1(B) where we use a silicone strip with four punctuated holes, to run several gels in parallel.The number of holes in the silicone strip and their sizes can be easily modified using a laser cutting machine.

Generating arrays of individual cells and cell-clusters in 3D biological hydrogels
We used the 'peel-off ' process described in figure 1 to pattern actin-GFP 3T3 fibroblasts in large array formats within fibrin and collagen hydrogels.The cell arrays were designed in hexagonal configuration, with each cell-cluster surrounded by six close clusters at various separation distances of 200 µm, 300 µm and 400 µm (figure 2(A)).To pattern individual cells, we used a 10 µm-diameter spot size that led to the positioning of about 1-3 cells in each spot (figure 2

(C)).
To pattern larger cell-clusters, we used spot sizes of 20 µm, 60 µm and 100 µm in diameter.Figure 2(A) shows an example of the 60 µm-spot pattern on a glass CS and the same region after being transferred to a 3D fibrin gel.It can be seen that the spatial arrangement of cells was accurately maintained during the 'peel-off ' transfer process.After the transfer, the cell pattern was slightly embedded within the 3D gel (figure 2(B), movies 1 and 2).Further, we also created similar patterns in collagen gels (figure 2(D)) and demonstrated partial control over the position of cells in the z-dimension by stacking two micropatterned layers with varying distances between the layers (figure 2(E) and supplementary figure S1).Finally, we demonstrate the possibility to pattern other cellular geometries, such as branched cell networks in square and triangular structures while reaching single-cell resolution (figure 3).

Characterization of the gel micropatterns
We evaluated the efficiency of the 2D-to-3D transfer process by quantifying the number of cell clusters successfully transferred from the CS to the gel, which was determined by comparing the same cellular pattern before (on glass, figure 2(A)) versus  after the transfer (in gel, figure 2(A)).For different cell cluster sizes (figure 4(A)), about 90% of the cell clusters on average, were successfully transferred from the CS to the gel (figure 4(B)).To characterize the size and homogeneity of the resulting clusters, the radius and volume of the clusters in the gel were estimated using image analysis (figures 4(C) and (D)).The calculated cluster diameter was 26 ± 4 µm, 56 ± 6 µm and 75 ± 6 µm corresponding to the original 20 µm, 60 µm and 100 µm-diameter designed adhesive spots, respectively.The small deviation indicates that the clusters were relatively homogenous in their sizes.Cluster volumes were also found to be relatively uniform  for the 20 µm-and 60 µm-diameter clusters, while the 100 µm-diameter clusters showed more volume variability.

Cell dynamics in 2.5D versus 3D gel micropatterns
Cells can behave differently when they are cultured on the surface of a thick gel (2.5D) or fully embedded in the gel (3D), in terms of their morphology, spreading, migration or gene expression [14,44,45].
Our patterning system delivers cells to the top layer of the gel, a situation that resembles a '2.5D' system.We note, however, that this system terminology usually refers to the situation where cells are situated on top of the gel's surface, however, in our case cells are slightly embedded inside the gel (figure 5(A), see the xz-section).To create a system where cells are fully embedded in the gel environment, we also explored the possibility of covering the gel-patterned cells with an additional gel layer.Immediately after the 'peel-off ' process and flipping the gel such that the cell pattern faces upwards (figure 1, step#7), we added additional fibrin gel and let it polymerizes while in contact with the underline patterned gel (figure 5(B), SI movie 3).The cell array can be observed at the interface of the two gel layers, and no voids are present between the layers (see xz-section, figure 5(B)).In both the 2.5D and 3D systems, cell viability was high indicated by extensive cell spreading and migration over time (figures 5(C) and (D)).To compare cell behavior in the two dimensionality systems, we monitored cell morphology over time in the gel within the 200 µm-spaced array.Cell spreading and migration were more pronounced and initiated earlier in the 2.5D system (figure 5(C)) as compared to the fully 3D system (figure 5(D)).In addition, in the 2.5D system, it was evident that cells preferentially migrated out of the clusters in a direction toward neighboring clusters.This behavior was in striking contrast to the 3D system, in which cells minimally invaded the surrounding gel and by 13 h, only minor cell extensions were observed emerging from the clusters.This difference likely stems from the fact that in the 2.5D system, the gel material interferes less with cell movement, while, in the 3D system, cells are completely entrapped in the material, and must generate sufficient forces and degrade the matrix in order to invade it.

Matrix-mediated mechanical interaction in micropatterned fibroblast arrays
The forces that cells generate deform and remodel the fibrous microenvironment, facilitating the propagation of mechanical signals between distant cells.This forms a unique mechanism for intercellular mechanical communication [42].This long-range interaction is relevant to numerous biological processes where cells interact through the ECM and not by direct cell-cell contacts.Particularly, cells were observed to generate densified and aligned bands of deformed gel connecting neighboring cells, and subsequently respond by migrating, invading or reorientating toward each other [15,[46][47][48][49][50][51][52].Despite the accumulated knowledge on mechanical interaction between cells through matrix deformation, this process was mainly investigated in individual cell pairs.How tissue-level multicellular interactions may give rise to collective organizational and migratory response, is not well understood [53].The ability to precisely control and manipulate cell array geometry, including array spacing, number of neighbors and size of the cell cluster, can serve as an invaluable tool for investigating matrix-mediated mechanical signaling in morphogenesis of multicellular systems.
A system of fibrin gel contracted by fibroblasts can serve as a relevant model system of wound healing where fibroblasts locally stiffen the gel by generating contractile forces [54] and the transmitted forces induce remodeling of matrix fibers that extend between neighboring cells [15].To study multicellular interaction of fibroblasts in fibrin gels and its possible impact on collective and morphogenetic cellular response, we micropatterned 3T3 actin-GFP fibroblasts in hexagonal arrays composed of 60 µmdiameter clusters with increasing inter-cell spacing of 200 µm, 300 µm and 400 µm (figure 6(A)).We used fluorescently-labeled fibrin gels to follow the cell-induced remodeling of the fibrous matrix and correlate it with cell dynamics.At 200 µm spacing, when cells were more closely packed, visible matrix bands of aligned and dense fibers formed between almost all neighboring clusters.Along these bands, cells preferentially migrated toward their close neighbors, as indicated by clear fluorescent peaks in the distribution of cells in the direction of the six close neighbors (figure 6(B) blue curve, and supplementary figure S2).Closer inspection and image analysis revealed that cells oriented along the remodeled matrix bands had distinct actin stress fibers directed along the aligned fibers of the band whereas cells that remained in the cluster center had more diffusive and randomly oriented actin (figures 6(C) and (D) and supplementary figure S3).When the intercell array spacing increased to 300 µm, a smaller number of bands formed, and the directed migration was overall reduced but still notable (figures 6(A) and (B)).Almost no bands were evident when the cell-cell spacing increased to 400 µm, which correlated with a lower extent of cell migration out of the clusters and with loss of the preferential orientation in cell migration (figures 6(A) and (B)).Thus, formation of intercellular bands by cells, generated when spacing was within 200 µm, served as 'highways' for directed cell migration, guiding collective cellular organization.
We next focused on the 200 µm inter-cell spacing array, a distance in which most neighboring cells formed connecting bands, and monitored the timeevaluation of band formation and cell dynamics.Band formation between close neighbors was highly dynamic with bands forming at different times and having different intensities (figure 7(B)).A large-scale phenomenon was observed in which cell-clusters that were initially separated transitioned into a connected group via matrix bands, where cell migration was more pronounced and oriented over time toward near neighbors (figure 7(C)).
The preffered cell migration along the connecting bands can be due to mechanosensitivity of cells to the deformed and aligned fibers in the band region, as well as due to biochemical signals presented by neighboring clusters.To shed light on the underlying mechanism, we first verified the involvement of cell contractile forces by treating the micropatterned gels with blebbistatin (150 µM), a myosin II inhibitor, added to the cell media immediately after gel formation.Blebbistatin-treated gels showed that cells maintained in their original cluster without leaving the clusters after 9 h (supplementary figure S4), indicating that cell contractile forces are essential in cell invasion into the gel and migration along the bands.Next, we applied immunofluorescence staining for integrin β1, a key transmembrane receptor in focal adhesions known to mediate cellular mechanosensing [55].In fixed samples, we evaluated integrin expression in cells located along the band during 0-12 h (figure 7(D)).As cells started to invade the fibrous gel and form bands (t > 3 h), we found clusters of integrin at the invading cell edge and over time these integrin clusters oriented along the aligned fibers of the band (supplementary figure S5).In contrast, the expression of integrin in cells inside the original cluster was more diffuse without clear directionality.Furthermore, to examine the possibility that biochemical signals (e.g.diffusing molecules secreted from neighboring clusters) contributed to the observed directed migration, we created the same cellular pattern (hexagonal array, 200 µm spacing) but this time on a 2D glass, a substrate that does not mediate elastic deformation as it is too rigid (supplementary figure S6).Cell migration on the glass was more random without a clear preference to neighboring clusters, unlike the situation in the gel, suggesting that the involvement of biochemical signals in directing cell movement is less prominent.These results collectively suggest that cell migration directionality is driven by cues presented by the deformed band regions, such as the increased local stiffness, anisotropy or topology of the aligned fibers [56][57][58].

Collective mechanical interaction in arrays of metastatic cancer cells
To further demonstrate the robustness of the presented gel micropatterning platform, a different cell type was micropatterned in fibrin gels to create large array formats.In this set of experiments, we followed GFPcancer metastatic cells (HRAS + cells) micropatterned at inter-cell spacing of 200 µm, at increasing time intervals and simultaneously observed matrix remodeling (figure 8).Within 9 h after cell transfer to the gel, the metastatic cancer cells generated intense bands between most neighboring pairs, seen as bright lines in the matrix, resulting in a large-scale network of bands (figure 8).However, bands did not form between every pair of cells, and some pairs remained unconnected at this time point (figure 8).Band network formation was a highly dynamic process, with some bands starting to form earlier than others, and which intensified over time, while new bands emerged between additional pairs.Moreover, in a strikingly different behavior than the 3T3 fibroblasts, the cancer cells did not migrate out of the clusters, but, rather, remained in a spherical cluster form.In addition, after 6 h, holes (cavities) in the gel were visible around the clusters, a process that was accelerated at 9 h, with large variability in the hole sizes.These holes could be caused by enzymatic degradation or mechanical rapture associated with the increased tension in the gel.In the example highlighted in the dashed box in figure 8, a group of six clusters was observed over time to form visible holes of various sizes.Interestingly, the specific pair that initiated band earlier (at 3 h) seemed to be pulled toward each other by the high tension in the band, more than with its other neighbors, resulting in closer proximity between this specific pair (see large-zoom images at 9 h in figure 8, the cell pair is indicated with an arrow).At the same time, the bands that formed with the other neighbors remained intact, although the central cell moved away toward the specific partner highlighted in an arrow (figure 8), implying that the deformation of the matrix in the bands is highly plastic and irreversible.

Discussion
This work presents a robust and accessible method to micropattern individual cells or cell-clusters in precise arrangements in biological hydrogels.We demonstrate the possibility of creating defined patterns of 3T3 fibroblasts and HRAS cancer metastatic cells in large array formats in fibrin and collagen gels and use them to study collective matrix-mediated interactions between contractile cells.In principle, however, any pattern geometry can be created, such as blood vessel geometries or specific tissue architectures of choice, for applications in tissue engineering or as substrates for drug screening.
The presented method is based on transferring a cell pattern from a 2D glass to a 3D gel by 'peeling-off ' cells with a gel carried by a silicone holder.The major advantage of the presented method is its simplicity.We pattern cells on 2D substrate using established photolithographic techniques, that are accessible in most institutions or can be commercially obtained.Other 'peel-off ' methods for micropatterning cells in 3D gels have been reported previously, but they rely on more sophisticated procedures, such as DNA linkers [31] or dielectric forces [32].Moreover, the transfer methodology that we use is relatively simple and relies on readily available materials, such as a commercial silicone rubber sheet.The majority of the patterned cells (90%) were successfully transferred to the 3D gel, for all examined cell cluster sizes.This level of efficiency is high compared to a previously described method that utilized the force of gravity to transfer cells from microwell arrays to 3D collagen or Matrigel [33].We note that in our system the calculated efficiency considers the percentage of cells transferred from the 2D glass to the 3D gel and does not take into account the inaccuracy associated with the 2D pattern itself (on glass), which depends on the accuracy of the chosen 2D micropatterning technique.In our case, micropatterning on the 2D substrate was achieved using a photolithographic-based approach, in which a UV light is directed through a custom mask, which resulted in high accuracy of the pattern and only rarely were cells missing from their designated positions in the 2D pattern.
At the end of the micropatterning process, the patterned gel is bounded by the silicone strip due to the natural stickiness of fibrin and collagen gels to the silicone.This coupled silicone-gel system imposes several advantages.Firstly, it allows for easy handling of the patterned gel, for example, transfer of the gel between different mounting dishes or placing it under the microscope.Without such support, the gels are too soft to act as free-standing and transferable material.The silicone carrier can also aid with in-vivo implantation of the gel in desired tissue region, for regenerative medicine applications.The silicone carrier can also be exploited in biomechanical stretching experiments, in which the gel can be strained externally through stretch of the silicone carrier, as we demonstrated in our previous studies [59][60][61].The silicone carrier also facilitates the stacking of several layers of patterned gels to create a 3D layered assembly of gel-patterned cells, as we have demonstrated.In such 3D assembly, cells can interact in the out of plane direction to form connected layered tissue in controlled microstructures.
The presented gel micropatterning platform was exploited to study how multicellular arrays of defined configurations and spacing mechanically interact through the ECM, and how it may guide collective cell migration and organization.It was demonstrated previously that by pulling on the matrix, contractile cells can mechanically sense and respond to other cells located hundreds of micrometers apart, thereby forming a unique mechanism of long-range mechanical interactions [42].The fibrous structure and nonlinear properties determine the capacity of the ECM to transmit forces and remodeling tracks over considerable distances [20,21,23,51].Indeed, it was shown that when embedded in fibrous biopolymer gels, such as collagen or fibrin gels, pairs of individual cells or cell clusters form long bands of densified and aligned fibers between them, extending over a distance much larger than the cell itself.These bands have often been referred to as bridges, tethers, tracks, bundles, fascicles or lines [15,[46][47][48][49][50][51].Groups of randomlyseeded cells can form 'networks' of bands, connecting many cells to one another [15].While previous studies have highlighted the importance of long-range mechanical signaling via the ECM in regulating cell morphology [52], migration [46] and invasion [50], most research in this field has focused on analysis of bands forming between individual pairs of cells or cell clusters and less is known about the mechanical interactions in multicellular systems, which better approximates tissue-scale and biological scenarios such as morphogenesis and wound healing.Few studies have highlighted the potential of local cell-cell mechanical interactions to coordinate global effects such as contraction of bulk gels [53] or tissue-level processes such as fibrosis expansion [62] or tissue folding [63].Thus, interactions between a multitude of cells may lead to exciting collective and global effects [53], as inspired by other natural systems, such as the coherent movement of birds or fish, where forces in the medium induce synchronization in the global movement of the group members without direct contact between its members [64,65].Here we took advantage of our system to precisely pattern multicell arrays in precise spacing.We demonstrate that a large cell population can be mechanically coupled via cell-induced deformed bands forming in the matrix between neighboring pairs that extend throughout the bulk gel.Such mechanical coupling of the cellular group via matrix bands took place when cell-cell spacing was 200 µm, and such mechanical coupling reduced significantly at larger separation distances.In the patterned arrays, fibroblasts migrated out of the clusters preferentially along the bands, giving rise to collective and preferred migration routes across the entire gel.In the supplementary note and figure S7, we provide theoretical prediction indicating that the observed polar cell distribution (figure 6(B), 200 µmspacing) indeed results from preferential migration of cells from the cluster, and not from cells traveling from neighboring clusters.
Cells respond to the deformed bands by aligning their stress fibers and integrins in the direction of the gel fibers, and this process is forcedependent as revealed by treatment with blebbistatin.The increased stiffness, anisotropy and alignment associated with remodeled band regions can impact cell morphology and directional migration [56][57][58].In parallel, the 2D glass control experiment provides further evidence that the directed cell migration along bands is not driven by biochemical signaling.We note, however, that this should be taken with caution, as the aligned and densified fibers of the bands in the gel may facilitate directed diffusion of molecules, as we computationally addressed previously [66,67], and thus the effect of molecular diffusion may be more propound in the gel than in the dish.Future studies should determine in more depth the mechanical versus biochemical signaling governing the observed collective migration of cells.
The cancer metastatic cells, in striking contrast to the fibroblasts, were less migratory, yet generated intense bands between cell pairs accompanied by visible local holes/cavities in the gel at later times.The difference between the cell types raises an interesting question why fibroblasts migrate along the bands whereas the cancer cells remain inside the clusters, although both cell types mechanically remolded the matrix and generated bands.The difference may relate to the secretion of ECM proteins by fibroblasts and different responses to ECM tension mediated by integrin expression.
Our micropatterning platform may set the stage for future in-depth studies on the emergence of collective and synchronized behavior of cell populations and the possibility of transition from stochastic behavior to population-level coherent behavior while being able to test critical parameters, such as the group spacing, and the number of interacting neighbors that may be associated with such transition.The observed collective behavior mediated by remodeling of the fibrous ECM is relevant to various biological processes, including tissue morphogenesis, tumor development and wound healing.

Figure 1 .
Figure1.'Peel-off ' gel micropatterning process.(A) Schematic illustration of the fabrication process, highlighting a single silicone well.First, a specific geometric pattern is generated on a CS (step 1).As an example, a 5 × 4 spot array is described.A silicone strip with a cutout (hole) is attached to the patterned CS to serve as a mold to sustain liquids (step 2).Cell solution is poured into the mold (step 3) and cells are allowed to settle on the adhesive spots (step 4).Hydrogel precursor solution is added to the mold and allowed to polymerize (step 5).The cell pattern is then transferred to the hydrogel by peeling-off the silicone (step 6).Finally, the micropatterned gel can be placed such that the cell pattern faces up (step 7).(B) Images of the actual system: incubation of cells in the silicone mold positioned on top of the micropatterned CS and after peeling-off the silicone mold embedded with a fibrin gel and patterned cells, using tweezers.Here, the silicone contains 4 gels.

Figure 2 .
Figure 2. Patterned cell arrays in fibrin and collagen gels.(A) Microscopy images of a 60 µm array of actin-GFP fibroblasts (green), spaced at 200 µm distances, on a glass coverslip and after its transfer to a 3D fibrin gel.Scale bar is 200 µm.(B) Confocal microscopy images of actin-GFP fibroblasts (green) in a 60 µm array in a fluorescently labeled-fibrin gel (gray).After transfer, the cell array is located at the top part of the gel.Top image shows top-view, and bottom image shows side view.Scale bar is 400 µm.See movie 2 for rotational view.(C) Enlarged view of an individual-cell pattern using 10 µm-diameter adherent spot sizes.Scale bar is 10 µm.(D) Confocal image of an array of actin-GFP fibroblast clusters (green) in a collagen gel (imaged in reflection, gray).(E) 3D stacking of two micropatterned fibrin layers of actin-GFP 3T3 fibroblast arrays (green).The distance between the layers (d) can be varied.See supplementary figure S1 for more details on the 3D stacking.Scale bar is 400 µm.

Figure 3 .
Figure 3. Micropatterns of branched cellular networks in fibrin gels.Confocal (left panels, actin-GFP fibroblasts in green) and bright field (right panels) images of cells patterned in square and triangular network geometries, reaching single-cell resolution (∼10 µm).Scale bar is 100 µm.

Figure 4 .
Figure 4. of the gel micropatterns.(A) GFP-actin fibroblast cells (green) micropatterned in a fluorescentlylabeled fibrin gel (gray), in a hexagonal array composed of 20 µm-, 60 µm-or 100 µm-diameter clusters.Scale bar is 100.(B) Percentage of cells transferred from the CS to the 3D gel for the three sizes of clusters.(C), (D) Average and standard deviation of cluster diameter (C) and volume (D) in fibrin gels for the three sizes of clusters.

Figure 5 . 2 .
Figure 5. 2.5D and 3D gel micropatterns.(A), (B) Confocal images of 3T3 GFP-actin fibroblasts (green) in the 60 µm -cluster array in the 2.5D (A) and 3D (B) systems.The fibrin gel is fluorescently-labeled (gray).See also SI movies 2 and 3. A XZ cross section cut along the thickness of the gel is shown to indicate the position of the clusters in respect to the gel top surface (A) or in respect to the interface between the two gel layers (B).(C), (D) Cell spreading and migration at 1 h, 7 h and 13 h after the transfer in 2.5D (C) and 3D (D) gel environments.Scale is 200 µm.

Figure 6 .
Figure 6.Cell dynamics and mechanical remodeling of the fibrous matrix for various cell array spacing.Fibroblast array patterns with increasing inter-cell spacings, at a fixed time point of 12 h.(A) Confocal images of 3T3 fibroblast cells (actin-GFP, green) micropatterned in fluorescently-labeled fibrin gels (gray) with inter-cell spacings of 200 µm, 300 µm, and 400 µm.Scale bar is 100 µm.(B) Quantification of cell migration orientation from the average actin-GFP cell maps (N > 50 clusters) with inter-cell spacings of 200 µm, 300 µm, 400 µm.See supplementary information figure S2 to see the overlaid GFP maps.(C, D) Zoom-in images of an array with an inter-cell spacing of 200 µm.Scale bar in C is 50 µm.(D) Confocal images of the marked orange box shown in C showing seperated and combined channels of stress fibers (actin-GFP, green) and gel fibers (gray).

Figure 7 .
Figure 7. Cell migration and matrix band formation in the 200 µm-spaced patterned cell array over time.(A) A large micropatterned array of 3T3 actin-GFP cells (green) in fibrin gel.About 70% area of a 6 mm diameter patterned-fibrin gel is shown.(B) Confocal imaging of actin-GFP fibrobalst cells (green) and fluorescently-labeled fibrin gel (gray) of the region highlighted in panel A at 0, 3, 7, and 12 h from the time cells were transferred to the gel.Scale bar: 200 µm.(C) Quantification of orientation-dependent cell migration at 3 h, 7 h, and 12 h (top), and the average cell GFP intensity as a function of distance from the cluster center (bottom).Number of analyzed clusters >50.(D) Immunofluorescence staining for integrin β1 (green) in cells located in matrix band regions (gray) at increasing time points.Cell nuclei are labeled in red.Scale bar is 20 µm.

Figure 8 .
Figure 8. Micropatterning of HRAS-cancer metastatic cells in large arrays in a fibrin gel.Cells were imaged at 1 h, 3 h, 6 h and 9 h intervals after being transferred to the gel.Cells are visible in actin-GFP (green) and the fluorescently-labeled fibrin gel in gray.Network of matrix bands formed over time, connecting most cell clusters.Holes in the gel emerged around the cell clusters at 6 h.An example of 6 specific cell clusters at increasing time points is highlighted by a dashed box.In the right-most column, an enlarged area at 9 h post-transfer showing two clusters getting closer by pulling on the intensified band (arrow).Scale bar is 200 µm.