Spatially organized, functional 3D networks at single cell resolutions

Abstract

An approach for forming organized 3D single-cell networks where their real-time signaling responses to a range of stimuli can be accurately captured using simple cell seeding and easy-to-handle microfluidic chips. Crosslinked collagen within multi-chambered microfluidic chips is formed, followed by femtosecond laser ablation of 3D microchannel networks and cell seeding. Cells migrate within ablated networks within hours, self-organize and form viable, interconnected, 3D networks in custom architectures such as square grid, concentric circle, parallel lines, and spiral patterns. The functionality of cell networks can be studied by monitoring the real-time calcium signaling response of individual cells and signal propagation when subjected to flow stimulus alone or a sequential combination of flow and biochemical stimuli. Furthermore, user-defined disrupted networks can be generated by lethally injuring target cells within the 3D network and analyzing the changes in their signaling dynamics.

Description

TITLESPATIALLY ORGANIZED, FUNCTIONAL 3D NETWORKS AT SINGLE CELL RESOLUTIONSSTATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0001] The invention was made with government support under Grant Nos. GM141573, AR083466 and DK136083 awarded by the National Institutes of Health (NIH). The government has certain rights in the invention.BACKGROUND OF THE INVENTION1. FIELD OF THE INVENTION

[0002] The present disclosure relates to microfluidic chips and, more particularly, to an approach for forming three-dimensional single-cell networks within an Extra Cellular Matrix (ECM).2. DESCRIPTION OF THE RELATED ART

[0003] Recreating the 3D spatial organization of single-cell networks will help elucidate underlying mechanisms related to their emergent functional properties exhibited at the tissue level. Although technological advances have led to an array of methods to arrange cells in 2D and 3D to mimic tissue specific architectures, single-cell resolution with control over cell-to-cell connectivity remains challenging. A common strategy is to pattern 2D adhesive micropattems using conventional lithography methods followed by cell seeding to generate interconnected single cell networks. Other methods, such as laser-assisted bioprinting, have also been used to directly deposit single cells and generate 2D cellular arrays, however, arranging single cells in 3D has proved difficult. Methods based on 3D photo-patterning of adhesive peptides within specialized semi-synthetic hydrogels have emerged, but they have failed to realize 3D single cell networks. At present, only methods based on multi-photon absorption (MPA) can pattern features at single-cell resolutions although limitations related to specialized photosensitive materials, water soluble low- toxicity photoinitiators, cell viability during laser scanning in the presence of cells, and low scalability, have limited its use in the field. Thus, the field continues to rely on self-assembly based methods, that involve mixing relevant cells in natural ECM like collagen or fibrin. This however results in randomly organized 3D cell networks without precise control over network density, connectivity, and architecture, making systematic mechanistic study on networks’ emergent properties challenging. Bioprinting methods can provide spatial control over cell placements within 3D ECMs, however single cell resolution is not possible.122104442.v2-8 / 13 / 25Micro fluidic chips integrated with acoustic, dielectric, and magnetic field stimulation have also been used to directly manipulate cells in a contactless manner. However, these methods cannot achieve single-cell resolution or user-defined multi-layer 3D patterns, and its control over intercellular connectivity remains poor. In summary, current methods are unable to generate tissue-specific, 3D, single-cell networks within natural, unmodified extracellular matrix (ECM), like collagen.BRIEF SUMMARY OF THE INVENTION

[0004] The present invention is a new approach, referred to herein as Cellnet, for the generation of 3D, single cell, functional networks in custom architectures within multichambered chips using laser ablation of 3D microchannel templates within collagen matrix. First, Digital Light Projection (DLP) is used to rapidly design and print master molds which are used to generate custom three-chambered PDMS chips. Second, ECM of interest (type I collagen) is perfused into central chambers and thermally crosslinked to generate a barrier between chambers 1 and 3. Third, Two Photon Ablation (TPA) is used to create 3D microchannel network within collagen. Model cells, seeded within the chip, self-assemble within this network, and generate an interconnected, 3D, functional circuit coined as Cellnet. Cellnets are compatible with standard imaging methods (brightfield, immunostaining, timelapse microscopy), co-culturing cells and in situ manipulation such as application of fluid flow and / or biochemical stimuli, or injury to target cells to generate user-defined disrupted networks. Cellnet, generated using tissue-specific cells, ECM, and single-cellular spatial organization, can be potentially used as a deterministic model to study how short-term signaling results in durable functional or yet unknown emergent properties.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING(S)

[0005] The present invention will be more fully understood and appreciated by reading the following Detailed Description in conjunction with the accompanying drawings, in which:

[0006] FIG. 1 is: (A) a schematic showing 3D printing of Master molds using Digital Light Projection (DLP) followed by replica casting to generate three-chambered micro fluidic devices using PDMS (picture). Type I collagen solution is perfused and crosslinked within chamber #2; (B) a process flow with top and cross-sectional views to generate a two-layer interconnected cellular networks or Cellnets within Type I Collagen in the central chamber of the micro fluidic chip.

[0007] FIG. 2 is a series of images of femtosecond laser ablation was performed in variety of bioinks crosslinked within chamber #2 of PDMS multi-chambered micro fluidic222104442.v2-8 / 13 / 25chips. (A) Composite (i) and fluorescence (ii) images showing perfusion of green microbeads though ablated channels in bovine collagen; (iii-iv) top and reconstructed isometric views obtained using confocal microscopy showing lumen sizes from 6pm to 35 pm. (B) Brightfield, composite and fluorescence images showing perfusable channels within photocrosslinked hydrogels; (i) Gelatin Methacrylate (GelMA), (ii) Polyethylene glycol diacryate (PEGDA, 6k MW), (iii) a mixture of GelMA-PEGDA, (iv-v) perfusion of red microbead solution through ablated channels in hybrid GelMA-PEGDA hydrogel in Ch#2 of the device.

[0008] FIG. 3 is (A) a schematic of Digital Light Projection (DLP) setup to print Master Molds for making multi-chambered PDMS chips; (B) design files and a representative picture of the Master Molds made using polyethyelene glycol diacrylate (PEGDA 250MW); (C) pictures of defective molding process in the absence of post-processing steps of ethanol treatment and UV exposure for 24 hours; and (D) a composite brightfield- fluorescence image showing parallel line pattern (top row) and square grid pattern (bottom row) with and without surface modification of PDMS-collagen devices. Without surface modifications, crosslinked collagen detaches from the PDMS roof and glass bottom surfaces. Scale bar: 100pm.

[0009] FIG. 4 is a characterization of microchannels ablated within collagen. (A) Schematic showing lateral scanning (XY) with varying dosage, and reflectance confocal images of ablated channels. Scale bar: 25pm (B) Plots showing the relationships between ablated line width and laser dosage at a depth of 50pm, while (C) shows the effect of varying depth on ablated line width. (D) Schematic and plot showing vertical ablation of microchannels and relationship between ablated line width (lumen size) and laser dosage. (E) Schematic of a 3D ablated grid with lumen size of ~8pm in XY (lateral) and Z (vertical) directions within crosslinked collagen in chamber #2 of a three-chambered micro fluidic chip. (F-G). Top view and isometric confocal images after perfusion of a solution of fluorescent microbead (c|)=0.5pm) into the ablated channels. Scale bar: 100pm.

[0010] FIG. 5 is (A) a schematic of two-photon ablation setup; (B) confocal reflectance microscopy images of ablated channels in lateral (XY) directions at varying powers (0.8W, 1W, 1.4W) and at varying depths (50pm, 100pm, 200pm inside collagen from the bottom glass coverslip); (C) confocal reflectance microscopy images of ablated channels in vertical (Z) directions under varying conditions. Note that the laser dose is calculated from power of laser right before objective lens and the scanning speed of the stage. (Scale bar: 20pm).

[0011] FIG. 6 is a series of (A) schematics for laser ablation inside collagen via cavitation. A focus femtosecond laser pulse generates a cavitation bubble through rapid322104442.v2-8 / 13 / 25energy deposition. The bubble expands and creates shockwave to disrupt collagen matrix, followed by bubble collapse resulting in a hollow void in collagen. Repeated cavitation events along the laser scanning direction creates an open microchannel. (Bi) Cross-sectional schematic of micro fluidic chips showing microchannel connecting chambers #1 and 3. (Bii) Representative brightfield image of microchannel in collagen matrix is shown (Red color is due to the fdter used during imaging). (C) Image of crosslinked collagen in Ch#2 between a pair of micropost arrays is shown. (D) Brightfield images of microchannels generated by laser scanning (i) immediately after crosslinking (~1 hour) show widely varying microchannel lumen sizes as compared to (ii) those generated with fully-crosslinked collagen (after 24 hours). Scanning conditions were identical for both groups.

[0012] FIG. 7 is a series of (A) schematic and representative brightfield images from two z-planes (located at 50pm and 100pm from the bottom glass surface inside collagen matrix) showing seeding of model cells (10Tl / 2s) in chamber #1 or #3, and their migration within microchannel networks to form viable, interconnected, 3D single-cell networks or Cellnets; (Bi) viability plot and (Bii, iii) live / dead assay of Cellnets on Day 7 taken from both z-planes of the 3D network (Green (calcein) stains for live cells, red (ethidium homodimer- 1) for dead cells); and (Ci) connectivity plot (Day 7), (Cii, iii) morphology of cells (actin=red; nucleus=blue). (Scale bar: A 20pm, B and C 100 pm)

[0013] FIG. 8 is a series of (A) brightfield images in 5 different samples showing fibroblast migration within ablation microchannel networks generated within crosslinked collagen in Ch#2 of PDMS chip. Parallel line and square grid networks are shown; and (B) representative images from 2 different samples showing clearance of cells from the side wall through trypsin treatment (0.25%, 5 mins) and prevent unwanted aggregation of cells in Ch#l and Ch#3. Scale bar: 100 pm

[0014] FIG. 9 is a series of: (A) composite, fluorescence, reconstructed 3D images showing the morphology of 3D (2-layered) CELLNETs using 10T1 / 2 fibroblast-like cells. (Lumen size ~8pm); (B-C) 3D reconstructed confocal image and top-view of z-stacked image of Saos2 osteoblast Cellnet. (Lumen sizes ~16pm)

[0015] FIG. 10 is series of: (A) representative images showing single cells organized within user-defined square grid and parallel line architectures in two different z-planes within collagen matrix; and representative image of cell morphology organized in various patterns: (B) concentric rings, (C) out-of-plane spiral and (D) double-helix. Scale bar: 100pm. (Actin: Red; Nucleus: Blue).422104442.v2-8 / 13 / 25

[0016] FIG. 11 is series of images of: (A-C) morphology of Cellnets using three model cells. (A-i) Orthogonal views showing 10T 1 / 2 Cellnet’s 3D morphology. (A-ii) Cross- sectional image showing actin-labelled cellular processes extending between two z-planes ~50 pm apart. (B) Orthogonal views showing 3D SaoS-2 Cellnet. (C-i) Orthogonal views showing 3D MLO-Y4 Cellnets, (C-ii) 3D reconstructed image and (C-iii) high-resolution image of MLO-Y4 Cellnets on one plane. Scalebar:25pm. (Actin: Red; Nucleus: Blue).

[0017] FIG. 12 is a series of images and graphs of: (A) MLO-Y4 CELLNET with lumen size of 8pm. (i) Live / Dead fluroscence image of cell network located ~50pm inside Type I collagen in Ch#2 of micro fluidic chips, (ii) Quantification of live (91.39%) and dead cells (8.61%) from 3 independent chips with error bar ±1.08 (iii) Cell connectivity. Multi cells MLO-Y4 cell circuit Histogram values. Values in x axis are the number of cells connected to each other (5 cells: 64.06% error bar: ±1.92%, 4 cells: 30.73% error bar: ±2.39%, 3 Cells: 2.08% error bar: ±0.90%, 2 cells: 3.13 error bar: ±1.56%) (iv) Stitched images showing MLO-Y4 network spanning the entire width of Ch#2; and (B) MLO-Y4 CELLNET with lumen size of ~12pm. (i) Actin Nuclei-stained MLO-Y4 cells at different z planes (ii) Cell Connectivity: Single cell MLO-Y4 cell circuit Histogram values (5 cells: 22.45% error bar: ±2.04%, 4 cells: 20.07% error bar: ±0.59%, 3 cells: 25.17% error bar: ±2.36%, 2Cells: 11.56% error bar: ±1.56%, 1 cell: 5.44 error bar: ±3.28%, none: 15.31 error bar:±1.77). Scale bar A, B: 50pm.

[0018] FIG. 13 is a schematic and series of images of heterotypic Cellnets. Design and generation of heterotypic Cellnets. (A) Schematics for top and crosssectional views of cell seeding for (i) same cell types distinctly tagged with green and red colors and (ii) different cell types tagged with unique colors (green and red). (Bi) Fluorescence image of green and red MLO-Y4 cell seeded in chamber #1 and chamber #3. (Bii) Composite image combining fluorescence and brightfield to display cells inside ablated microchannels. (Ci)Fluorescence image of green MLO-Y4 and red MC3T3 cell seeded in chamber #1 and Chamber #2. (Cii) Composite image combining fluorescence and brightfield to show ablated microchannels. Scale bar: 100pm.

[0019] FIG. 14 is a series of images and graphs of real-time signaling within osteocyte (MLO-Y4) Cellnets. (Ai) Schematic showing application of a biophysical fluid flow stimuli in Chamber #1. (Aii) Time-lapse fluorescence images showing calcium signaling at specific time-points. ( Aiii) Schematic of model Cellnet. Here, 1.2 represents a single cell in row 1 and column 2, located at the interface of Ch#l-2. Upon stimulation, 1.2 initiates a Ca signal that propagates to adjacent connected cells embedded deeper within collagen (2.2 and522104442.v2-8 / 13 / 253.2) (Bi-iii) Normalized single cell signals from three representative local networks (timestamps are shown in seconds). (C) Signaling response of individual osteocytes (Cell #1 and #2) and the entire network can be accurately analyzed in the presence of sequential biophysical and biochemical stimuli. Scale bar: (Aii), C: 50pm.

[0020] FIG. 15 is a series of images of (A) Cumulative heat-map intensity of MLO- Y4 Cellnet during flow-stimuli evoked Calcium signaling experiment. (B) Representative time-lapse fluorescence image showing calcium signaling at specific time-point. (C) Multiple local cell-circuit signaling results to calculate the average velocity of calcium signal propagation within Cellnets. Individual cell signals are marked using simplified nomenclature ‘x.y’ where x point to row number and y point to column number for a particular cell within Cellnet. Average velocity for each column is calculated by taking difference of time in onset of signal in first and last cells in the column. (The asterisksand “**” are used to annotate the cells at the interface of Ch#l and Ch#3).

[0021] FIG. 16 is a series representative fluorescence images showing calcium spikes in MLO-Y4 Cellnet when subjected to sequential (A) biophysical and (B) biochemical stimuli. (C) About 15 individual single cells within Cellnets were randomly selected and corresponding fluorescence intensities were plotted. Results show the calcium signaling response in selected single-cells when subjected to biophysical fluid flow stimuli (black) followed by a biochemical (ionomycin) stimulation (grey).

[0022] FIG. 17 is a series of schematic and graphs of in-plane (~70pm inside collagen) real-time signaling within user-disrupted osteocyte Cellents. (A,B) (i) Schematic showing application of a biophysical fluid flow stimuli in Chamber #1 after femtosecond laser ablation was used to lethally injury three cells (shown by cross mark). (C) Simplified representation of how signal propagation is diverted due to targeted disruptions. (D) Intensity plots showing delayed propagation of signal to cell 2.2 as signal travels from cell 1.4 to 2.4 to 3.4 to 2.2. Scale bar: B: 50pm.

[0023] FIG. 18 is a representative set of real-time fluorescence images show an absence of Ca signal (Fluo-4) in ablated cells forcing signaling through non-ablated connected cells within the network.

[0024] FIG. 19 is a series of schematics and graphs of out-of-plane (3D) real-time signaling within user-disrupted osteocyte Cellnets. (A) Schematic of 2-layered network with cells organized using a grid patterns (X). Top views (i) and cross-sectional view (ii) showing out-of-plane signal propagation, (iii) Actin (red) Nuclei (blue) showing cell morphology. Cell network at 130pm sense flow stimuli and transmit signals to cell network at 70pm via out-of-622104442.v2-8 / 13 / 25plane cell connections, (iv) Snapshot taken during live fluorescence imaging of Ca signaling at z plane location at 70pm. (B) (i) Schematic of 2-layered network with two different cell patterns (X=grid, Y=parallel line) located near each other at (i) z=130pm and (ii) at 70pm in the same sample. Cross-sectional views for (iii) Y-pattem and (iv) X-pattern showing out-ofplane signal propagation at different planes, (v) Snapshot taken at a z-plane of ~70pm during calcium signaling experiment. Cells a, b in pattern Y and c, d in pattern X have been marked by a red circle, while dashed white arrows indicate signal propagation directions, (vi) Signal intensity peaks of cells ‘c’ and ‘d’ in pattern X are delayed as the signal travels from the cellnetwork @z=130pm to cell-network @z=70pm as compared to cells ‘a’ and ‘b’ in pattern Y where flow stimuli has a direct path via cell network @ z=70pm. Note: In pattern Y, cells are directly exposed to flow stimuli while in pattern X, only the top layer is directly exposed to stimuli.

[0025] FIG. 20 is a set of images and graphs of representative snapshots of real-time fluorescence images showing Ca signaling within the cell network located at z=70pm in a 2- layer osteocyte network with both parallel-line (Y) and square-grid (X) pattens.DETAILED DESCRIPTION OF THE INVENTION

[0026] Referring to the figures, wherein like numerals refer to like parts throughout, there is seen in FIG. 1, an approach for generating 3D, single cell, functional networks in custom architectures within multi-chambered chips using laser ablation of 3D microchannel templates within collagen matrix. First, digital light projection (DLP) was used to rapidly design and print master molds which were used to generate custom three-chambered PDMS chips. Second, ECM of interest (type I collagen) was perfused into central chambers and thermally crosslinked to generate a barrier between chambers 1 and 3. Third, two photon ablation (TPA) was used to ablate 3D microchannel network within collagen. Model cells, seeded within the chip, self-assemble within ablated network, and generate interconnected, 3D, functional circuits coined as Cellnets.

[0027] Cellnets are compatible with standard imaging methods (brightfield, immunostaining, time-lapse microscopy), co-culturing cells and in situ manipulation such as application of fluid flow and / or biochemical stimuli, or injury to target cells to generate user- defined disrupted networks. Cellnet, generated using tissue-specific cells, ECM, and single- cellular spatial organization, can be potentially used as a deterministic model to study how short-term signaling results in durable functional or yet unknown emergent properties.

[0028] All tissues are composed of single cells, yet how tissues sense, respond, and adapt to stimuli cannot be predicted by studying individual cells as higher-level emergent722104442.v2-8 / 13 / 25property of the tissue likely rely on complex dynamic interactions between cells. To do this, the field needs new ways to generate tissue-specific 3D single cell networks, apply defined stimuli, and analyze networks’ adaptations. As a result, the link between the individual cells and cell-networks’ (tissue) function remains unclear and difficult to predict. The present invention, coined as Cellnet, generates normal and disrupted 3D single-cell networks within type I collagen matrix in user-defined configurations and study the real-time signaling of single cells and signal propagation across the entire network. This template -based strategy works with many cell types, is highly reproducible and provides user control over cell-cell connectivity and cell-network layout. Use of DLP allows rapid and inexpensive iteration of master molds enabling rapid production of custom internal microchannel designs within standard micro fluidic chips. Real-time Ca signaling of individual cells and signal propagation within Cellnets can be monitored when subjected to biophysical and biochemical stimuli. Moreover, femtosecond laser irradiation can ablate target cells within Cellnets at defined locations and time -points to design custom disrupted networks and study real-time changes in their signaling dynamics. This allows entire Cellnets or individual cells within the network to be manipulated in a non-invasively, contactless, and sterile manner during active culture.

[0029] To test the capability of studying real-time signaling within Cellnets, osteocytes were chosen as our model cells, due to prior experience with bone tissue engineering. Like many tissues, in bone, stimuli evoke Ca2+signals within 3D, organized, single-cell osteocyte networks while signal disruptions have been linked to many pathologies. Similar to other cell types, single-cell osteocyte networks have also been generated using micropatterned cell-adhesive self-assembled monolayers (SAMs) or micro-chambers, or two- photon laser based modifications, however generating single-cell network in 3D remain difficult. As a result, culture cells within collagen or fibrin matrices remain the ‘gold standard’ to generate 3D cell networks, however the randomly organized cell-cell connections with poor reproducibility makes a systematic study about single cell and its relationship to network signaling difficult. In contrast, structurally defined 3D MLO-Y4 (osteocyte) Cellnets simplify image processing and enable accurate mapping of how single cells respond to various types of stimuli and their contributions to changes in signaling wavefronts within the networks.

[0030] Previously, femtosecond laser ablation has been used to modify local properties within a 3D cell-laden hydrogel matrix to guide the alignment of single cells. For instance, laser based densification of partially crosslinked GelMA was used as guidance cues to align encapsulated cells in 3D. Another study used modified PEGDA hydrogels to generate822104442.v2-8 / 13 / 25cell network. In both these cases, a native extracellular matrix like collagen cannot be used due to the requirement of the photo-sensitive hydrogels to enable laser-based biochemical or biophysical modifications. Both laser processing conditions and hydrogel properties must be optimized to maximize viability of encapsulated cells; this enforces strict constraints on hydrogel type and processing conditions and limits its utility in the field. Laser scanning also generates reactive oxygen species (ROS) which decrease cell viability by interfering with its metabolism. Since ROS generation is strongly dependent on applied laser dosage and the material photosensitivity, new semi-synthetic hydrogels with better photosensitivity are needed to satisfy the contrasting requirements of rapid hydrogel modifications (either ablation or degradation) while maintaining high cell viability. With Cellnet, a user-defined template in collagen is generated before cells are seeded / pipetted in target microfluidic chambers. This decoupling provides the flexibility to generate 3D cell networks in any bioink, including native and unmodified ECM like collagen, which in turn results in close to 100% cell viability as laser scanning is not performed in the presence of cells. In the present invention, type I collagen (4mg / ml) was used as the model ECM as this natural and thermally cross-linkable biomaterial has been widely used in the field. Other ECM analogs have been tested such as bovine collagen and semi-synthetic PEGDA-GelMA hydrogels, as seen from FIG. 4. Cellnet is thus an ECM- and cell-agnostic technology that can be broadly used to design tissue-specific Cellnets for a range of applications.

[0031] Cellnet could also emerge as an ideal tool to study signaling heterogeneity, a potentially important yet unstudied phenomenon. Previously, it has been shown that selfassembled and randomly connected osteocyte networks exhibit spatial variation and temporal heterogeneity of Ca2+waveforms when subjected to fluid flow stimuli. Due to difficulty in identifying individual cells in self-assembled and randomly organized networks, the relationship between cell connectivity, stimulus conditions, and signal heterogeneity cannot be studied. In the present invention, calcium signal is used as a proxy to monitor real-time signaling of individual osteocytes within spatially organized 3D single cell networks. Cellnets within micro fluidic chips provides systematic application of biophysical, biochemical or injury stimuli that could be used to assess single cell signaling heterogeneity within other cell types.

[0032] Overall, the modular nature of Cellnet allows ‘swapping in’ relevant cell types, ECM composition, network designs, optical tracers for other signals (NO, ROS, ATP) and / or fluorescent reporters or inhibitor drugs and acquire new knowledge using standard imaging and culture methods. In the future, transfected variants of cell lines with genetically922104442.v2-8 / 13 / 25encoded calcium indicators can facilitate fluorescence imaging without exogenous labels and allow us to study the dynamic evolution of Cellnets over time. The simple and easy to use cell seeding strategy to generate Cellnets will enable broad utility and adoption in the field to test new hypothesis across cell, tissue, or stimulus types and develop personalized tissuespecific models for drug testing and diagnostic applications. It is envisioned that Cellnet will transform the study of tissue-scale system biology by linking individual cells and tissue networks’ function thereby elucidating higher-level emergent property

[0033] EXAMPLE

[0034] Design and development of microfluidic chips using DLP-printed Master Molds.

[0035] DLP was used to print master molds which were used to replica cast final PDMS micro fluidic chips FIGS. 1 and 3. Details related to the DLP printing setup, prepolymer formulation, printing conditions, glass modification protocols, are reported in the Methods and SI sections. Briefly, a CAD model of a reverse master mold of the intended microfluidic chip was designed in SolidWorks; this consist of positive features that would be replicated into three microfluidic chambers with negative spaces, each with inlet and outlet ports. (FIG. 3B). CAD model was then sliced to generate a series of virtual masks which were converted to corresponding light patterns by the Digital Micromirror Device (DMD). Upon irradiation onto liquid PEGDA (250MW) photo-polymer solution in a layer-by-layer manner, final master mold was realized. Glass surface was modified to prevent delamination of PEGDA mold during the printing process. Molds were printed at the step size of 20pm with 0.8s exposure time per layer with constant light intensity of 6mWcm”2. Post-printing, molds were cleaned with 100% ethanol to remove any uncured resin. Molds were cured for an additional 20s under UV light to ensure strong crosslinking between the mold and the bottom glass slide to prevent any delamination during the PDMS casting steps. PDMS prepolymer solution was casted onto the molds followed by thermally curing at low and high temperature cycles; this ensured high repeatability, since without the low temperature curing step, molds show tendency to crack and warp. Prior to casting, PEGDA molds were submersed in 100% ethanol for an hour followed by exposure to ambient light for 24 hours to generate defect-free PDMS chips. Without these post-processing steps, PDMS casting led to many defects possibly due to leaching of free radicals from the interior of the mold causing inhibition of PDMS crosslinking at the mold-liquid PDMS interface. (3C) Polymerized PDMS was carefully peeled off, trimmed to suitable size, and 6 inlet-outlet holes were punched (2 holes per chamber) before irreversibly bonding to a glass coverslip (0.15 mm1022104442.v2-8 / 13 / 25thick) to generate final PDMS chips. (FIG. 1 A) The final chips consist of three-chambers; a central chamber (Ch#2; 0.85mm wide, to house crosslinked collagen) flanked on either side by two chambers (Ch#l and Ch#2; 1mm wide, for media exchanges), separated by an array of microposts (width = 200pm) with a spacing of 200pm. The height for all the chambers were 240pm.

[0036] Characterization of laser ablated 3D microchannel networks within crosslinked collagen.

[0037] Type I collagen was chosen as model ECM due to its abundance in in vivo tissues, and wide use in the field to generate 3D cell culture models. Before pipetting collagen solution within PDMS chips, they were surface modified to prevent delamination of crosslinked collagen from the channel surfaces during active cultures. First (3- aminopropyl)triethoxy silane (APTES) was used to silanize the glass / PDMS surfaces to generate self-assembled monolayer (SAM) with reactive functional groups (-NH2) followed by glutaraldehyde (GA) to form reactive functional groups (-COOH).29 Without these modification steps, the crosslinked collagen showed a tendency to detach from the PDMS roof and glass bottom surfaces. (3D) Then, type I collagen solution (4mg / ml) was crosslinked within the central chambers of UV sterilized chips. (Fig. IB) After 24 hours, a focused femtosecond laser was used to ablate 3D microchannels architectures (800nm, 1.2W; ablation setup is shown in 5 A). To reliably generate 3D channels of defined lumen sizes, laser ablation was performed at varying power and depths inside collagen (50 - 200pm) in both lateral (XY) (FIG. 4A, 5B) and vertical (Z) (FIG. 4D, 5C) directions. Reflectance microscopy images show top and cross-sectional views of ablated channels embedded in collagen matrix. Laser dosage below 2 x 105 Jem-2 results in ablated microchannels with lumen size of 0.5-1 pm, while above this threshold, laser irradiation results in cavitation and formation of larger sized lumens (8-10pm). (FIG. 4B) During laser scanning above this threshold, the radial expansion of the bubble generates a shockwave which locally breaks down the collagen network and leaves behind a hollow lumen in its wake. (FIG. 6A,B) Gas bubbles were elongated in the scanning direction and often remain in the channel from 10s of seconds to 1-2 hours. If ablation was carried out immediately after collagen crosslinking, scanning induced bubble collapses and the channels close, possibly due to partially crosslinked collagen. Thus, 24 hrs were given for completion of collagen crosslinking before ablation was performed; this ensured repeatable and stable channel formation without collapse. (6D) As expected, for constant laser dosage, lumen size decreased with increased depth in collagen. (FIG. 4C) Laser processing plots (shown in FIG. 4B,D) were used as a design guide to repeatably generate 3D1122104442.v2-8 / 13 / 25microchannel networks with a lumen size of ~8pm within collagen; this lumen size was found to be ideal for single cell migration. Lumen size >8 gm led to migration of multiple cells within the channels while lumen size <8 pm decreased single cell migration and resulted in network occupancy below 50%, as discussed in later sections. For this work, 4mg / ml collagen concentration was used with a laser power of 1.4 W and 40x water objective (0.8NA) to generate various configurations of Cellnets. Fluorescent beads (~0.5pm) were used to verify perfusion within 3D ablated channel network (lumen size~8pm) within the central chamber (#2) of a microfluidic chip. (FIG. 4E-G).

[0038] Generation of viable and interconnected single-cell 3D networks within micro fluidic chips.

[0039] Fibroblast-like model 10T 1 / 2 cells were used to demonstrate the formation of Cellnets. First, cell solution (IM / ml) was pipetted in one of the side chambers (#1 or #3) of the microfluidic chip, and brightfield microscopy was used to monitor cell migration within the ablated channels. Migration as early as 3 hours post-seeding was observed. Between Day 1-2, cells migrated inside the collagen (500-1000pm from the interface between Ch#2 and Ch#l / 3), and assemble into a 3D, interconnected single cell network within chamber #2 of the microfluidic chip. (FIG. 7A, 8A) On Day 2, extra cells were flushed out using 0.25% trypsin treatment for 5 minutes followed by pipetting fresh media in the side chambers; this step was performed to prevent aggregation of cells in Ch#l or Ch#3 that could block diffusion of reporter dyes or antibodies and generate unwanted fluorescence during the characterization of cell networks. Post-trypsin wash, cell processes within the microchannel network retract for a few hours before recovering back to their normal spread morphology after 12 hours. (FIG. 8B).

[0040] To evaluate viability and morphology of cells within Cellnets, a 3D microchannel network (lumcn=8pm) with a two-layer connected grid architecture was used. The upper- and lower-layer, located 50pm and 100 pm respectively from the glass bottom substrate, connect with each other. Viability of cells, calculated from cells from both layers, was found to be 93.23 ± 2.87% (FIG. 7B, i-iii). Fluorescence image of cells fixed on DAY 7 shows morphology (actin / nucleus) of cells assembled into pre-templated square-grid microchannel networks that provide defined control over cell-to-cell connectivity within collagen. (FIG. 7Ci-ii) For instance, in the template used here, for a target cell within the Cellnet, a maximum of 5 cell-cell connections can be generated: 4 in-plane connections (in either the upper or lower layers) and one-out of plane (between the two layers). (FIG. 7C) Results show that 70.56 ± 1.92 % cells were connected to 5 cells, while 16.67 ± 3.33 % were1222104442.v2-8 / 13 / 25connected to 4 cells, 6.67 ± 1.67 % were connected to 3 cells, while 5 ± 0.96 % were not connected to any cells. This shows that around 95% cells within Cellnets are connected to at least three neighboring cells. 9 shows composite brightfield and fluorescence images, and a reconstructed 3D image.

[0041] To demonstrate that Cellnets of complex topology can be generated, microchannel templates such as parallel lines, spider webs, out-of-plane spirals, and double helixes were designed. (FIG. 10) Results show that cells migrate and self-organized into single cell 3D networks in custom orientations.

[0042] To demonstrate that Cellnets can work with many cell types, 3D square-grid microchannel templates within collagen were generated, and seeded fibroblasts-like 10Tl / 2s, MLO-Y4 osteocytes, and Saos-2 osteoblast-like cells. Top, side and 3D reconstructed views of Cellnets show formation of Cellnets with in-plane and out-of-plane connections. (FIG.11 A-C). FIG. 11 Aii highlights a cross-sectional section ~75pm from the bottom glass substrate - a plane between the upper and lower microchannel grids showing -80% of out-of- plane actin-labelled cellular connections that span -50pm distance in the vertical direction. High resolution images show that some regions within the Cellnets house more than one cells; nodes where 4 microchannels meet tend to slightly larger than the channels, thus have a higher likelihood of housing more than one cells. Cell connectivity is a function of lumen size and cell type and need to be optimized to get a single cell connectivity close to 100%. (FIG. 14Aiii and Bii). For instance, for MLO-Y4 Cellnets, an ablated channel size of -8pm ensuring single cell occupancy with -85% cell-to-cell connectivity while a lumen size of ~12pm increases connectivity to -97% at the cost of having more than one cells in the larger ablated channels. Furthermore, the capability of generating Cellnets using two different cell types was tested by seeding cells tagged with different fluorescent dyes in either side of the central chamber (Ch#2) (FIG. 13). A 3D square-grid microchannel network was ablated within collagen. One study involved fluorescently tagging of the same cell type (MLO-Y4) with green and red dye, while the other study tagged MLO-Y4 with green dye and preosteoblasts (MC3T3) with red dye. In both cases, cells migrate towards each other within the 3D ablated microchannel network and form heterotypic Cellnets.

[0043] Testing functionality of Cellnets using stimuli evoked real-time Ca signaling studies.

[0044] MLO-Y4 Cellnets were chosen to mimic an organized 3D osteocyte interconnected networks found within bone tissue and study the real-time signaling responses of individual cells and the entire network when subjected to biophysical and biochemical1322104442.v2-8 / 13 / 25stimuli. Before monitoring of real-time signaling within Cellnets, the viability and morphology of osteocytes were analyzed. Results show an organized square grid network of cells with high viability of 91.39 ± 1.08%. (FIG. 12A) Osteocytes form 3D in-plane and out- of-plane cell-to-cell connections as evidence by the 3D reconstruction and orthogonal views. As explained earlier, a number of cell-to-cell connections were characterized. Specifically, 22.45±2.04% cells connect with 5 neighboring cells (4 in-plane and 1 out-of-plane connections), 20.07±0.59% connect with 4 cells, 25.17±2.36% connect with 3 cells, 11.56±1.56% with 2 cells, 5.44±3.28% with one cell, while 15.3H1.77 were not connected to any neighboring cells. Thus, for this 2-layer design, it was found that -85% of cells were connected to at least one neighboring cell within the network while maintaining single cell occupancy within ablated channels. Since a lumen size of -12pm resulted in many occurrences of multiple cells in the same location within the ablated networks, ( FIG. 12B), a lumen size of -8 pm and a square grid network design was chosen to test signaling within osteocyte Cellnets. First, MLO-Y4 Cellnets were generated in the central chambers (#2) of microfluidic chips and treated with Fluo-4 Ca indicator via media chambers (#1, #3). (FIG.14 A) Timelapse fluorescence images show an increase in calcium signals in cells proximal to the stimuli, and subsequent transmission of the signals to cells embedded deeper within collagen. (FIG. 14Aii) To track signaling response of individual osteocytes within Cellnets, a simplified nomenclature ‘x.y’ was used, where x point to row number and y point to column number. (FIG. 14Aiii) For instance, row 1 represents osteocytes at the interface of central and side chamber that could directly experience the stimuli, while rows 2 and 3 represent osteocytes embedded at increasing distance within collagen (away from the stimuli). Control over laser scanning path allows the generation of a network with deterministic cell-to-cell connectivity. For instance, a black line represents a direct microchannel connection between cell 1.1 and embedded cell 1.2 while an absence of a black line between cell 1.1 and cell 2.1 indicates that these cells are not connected. For individual cells within the network, calcium signaling, expressed as fold change in fluorescence over baseline, was plotted and time-lags between locally connected osteocyte circuits were analyzed. (FIG. 14Bi-iii) For instance, signals travel from cell 1.2 to 2.2 to 3.2 (a distance of -100pm) with peaks recorded at 50.22s, 59.21s and 72.23s respectively (total time of 22.01s for a -100pm resulting in speed of - 4.54pm / s). (FIG. 14Bi) Each cell within a 3D connected network function as a node, which when connected to other cells in the network, relay stimuli evoked Ca signals. Consider another locally connected network (marked in blue), here a clear time delay is seen as signal travels from cell 1.4 (41.54s) to 2.4 (57.35s) to three connected cells 2.3(70.37s),1422104442.v2-8 / 13 / 253.4 (72.54s) and 2.5 (61.07s). (FIG. 14Bii) Since all cells are not always present at the nodes with equal spacing between them, the location of each cell was analyzed using FIJI image processor to get accurate measurements about propagation speeds. (FIG. 15) Lastly, consider all cells in third row; here cells 3.1 and 3.6 are not connected to cells from previous rows, while cells 3.2, 3.3, 3.4 and 3.5 are connected to cells in row 2. As expected, the signal travels to the connected cells first before propagating to 3.1 and 3.6 via 3.2 and 3.5 cell-cell connections respectively. (FIG. 14Biii).

[0045] A sequential flow stimulus was then provided, followed by rest, followed by a chemical stimulant (ionomycin, 5 pl, 45pmol.L-1, commonly used ionophore known to enhance Ca signals). (FIG. 14C). The sequence of fluid flow, rest, and ionomycin stimulus was performed on the same sample, and resulting changes in calcium flux were captured. Here, two osteocytes (marked as #1, #2) within the Cellnet were picked and their response monitored over time when the network is subjected to a defined set of sequential stimuli. Results show signaling heterogeneity in magnitude and signal profiles between the two cells and between the stimuli type for the same cell. FIG. 16 shows calcium intensity plots for 15 individual osteocytes within the network. For all cells within the Cellnet, ionomycin generates a stronger Ca signaling response often lasting for more than 10 minutes as compared to the response evoked by the flow stimulation. Like earlier results, cells closest to the stimuli are triggered first, followed by signal propagation inside the entire network. Time delays within the network are directly related to the user-templated microchannel design and associated cell-to-cell connectivity.

[0046] Study of Ca signaling and signal propagation within custom designed disrupted Cellnets.

[0047] Cellnets can be disrupted in custom configuration by ablating individual cells within the network by targeted femtosecond laser irradiation. First, ablation threshold to lethally injure target osteocytes was identified using via standard live / dead assay. Osteocyte monolayer, labeled with green cell tracker, were identified, and irradiated with femtosecond laser at varying power (10-200mW) and exposure times (0.5- 1 s). Ablation threshold to cause permanent injury was identified to be 0.22J / cm2. To compensate for a decrease in laser dosage while irradiating cells deep in collagen (~ 150pm), ablation was performed with 3x of the threshold used under monolayer conditions. Thus, a lethal dosage of 0.64J / cm2(Objective: 40X, NA:0.8, Power: 150mW, exposure time:0.5s) was used for all disruption experiments in this work. Two network designs were tested. First design involves single-layer MLO-Y4 Cellnets in square grid orientation located ~70pm inside collagen in the central1522104442.v2-8 / 13 / 25chamber (#2) of the microfluidic chip. (FIG. 17A) On Day 5, focused femtosecond laser was used to lethally injure individual osteocytes. (FIG. 17B) Then, Fluo-4 Ca indicator treatment was applied and timelapse fluorescence imaging was used in the presence of fluid flow stimuli to monitor real-time Ca signaling within the disrupted Cellnet. Real-time fluorescence images show an absence of Ca signal and propagation for injured cells within the network. Based on the simplified nomenclature described earlier (FIG. 17C, 18), cells 1.2, 1.3 and 2.3 were lethally injured forcing the signals to propagate from cell 1.4 (proximal to stimuli) to cell 2.4 (row 2, column 4) to other cells within the network. Plots shows the time-delays of signaling peaks as the signal travels from cell 1.4 (50s) to cell 2.4 (56s) to cell 3.4 (72s) to cell 2.2 (78s). (FIG. 17D) In an undisrupted Cellnet, signal would have chosen the shortest path to propagate to cell 2.2 (i.e, from cell 1.2 to cell 2.2), however due to user-defined disruptions, the signal is forced to circumnavigate using the only possible connected path available, as indicated by shifts in the signal intensity peaks and associated time-delays.

[0048] Design 2 involves a 2-layer Cellnet in a square grid orientation indicated by ‘X’ in FIG. 19, where the top and bottom layers are located at a depth of 130pm and 70pm inside collagen respectively. For pattern ‘X’, only cell network located at z=130pm has direct physical connection to chamber #1 while the cell network located at z=70pm can only indirectly sense the stimuli through out-of-plane cell-to-cell connections. Fluorescence images show osteocyte morphology (FIG. 19Aiii) and a snapshot of Ca signaling (FIG.19Aiv) from the imaging plane located at z=70pm. These results clearly show that there is no direct contact between the flow stimuli applied in Ch#l and cell network located at z=70pm.

[0049] To compare signal propagation speeds between directly and indirectly connected cell networks, two network 3D patterns ‘X’ and ‘Y’ were tested. (FIG. 19Bi-ii) While both patterns are located at z=130pm and z=70pm inside collagen, for parallel-line pattern Y, both top and bottom cell networks can directly sense flow-stimuli evoked signaling response while for pattern ‘X’, cells located at z=70pm can indirectly receive signals from stimuli-evoked cells located at z=130pm. Fluorescence image showing a snapshot of Ca signaling within the cell network located at z=70pm. (FIG. 19Bv, 20) (Arrows indicate the signaling direction for both patterns). Cells ‘a’ and ‘b’ from pattern Y and cells ‘c’ and ‘d’ from pattern X are chosen as representative cells, and their signaling responses are compared when subjected to a flow stimulus. (FIG. 19Bvi) Peak delays observed in ‘c’ and ‘d’ as compared to ‘a’ and ‘b’ indicate that pattern X enforces Ca signal propagation from the top- layer (z=130pm that directly senses the stimuli applied in Ch#l) to cells in the bottom layer1622104442.v2-8 / 13 / 25(z=70pm). These results show the capability of studying real-time signaling within user defined disrupted cellular networks.

[0050] Materials and Methods

[0051] Fabrication of PDMS micro fluidic chips using DLP-printed PEGDA Master Molds.

[0052] Prepolymer solution was prepared using polyethylene glycol) diacrylate (PEGDA, MW= 250Da; Sigma-Aldrich) as the base material, with Phenylbis(2,4,6- trimethylbenzoyl)phosphine oxide (commonly known as Irgacure 819; Sigma- Aldrich) and 2- Isopropylthioxanthone (ITX; Tokyo Chemical Industry) as the pho to initiator and photosensitizer respectively, while 2,2,6,6-tetramethyl-l-piperidinyloxy (TEMPO; Sigma- Aldrich) served as the free radical quencher. A stock solution was prepared by mixing 100 mg of Irgacure 819, 200 mg of ITX, and 4 mg of TEMPO with 40 ml of PEGDA in a centrifuge tube, which was then wrapped in aluminum foil and vortexed to ensure thorough mixing of the chemicals in the solution. This stock solution was then stored at room temperature until use.

[0053] Prior to printing, the glass slides were subjected to cleaning in piranha solution (H2SO4 and H2O2; 7:3) with constant stirring at 125 rev / min for 30 mins, followed by rinsing with ethanol and water and subsequent drying at 65 °C in a vacuum oven for an hour. The surfaces of the glass slides were then modified with methacrylate groups by immersing them in a solution containing 3-(trimethoxysilyl)propyl methacrylate (TMSPMA; Sigma- Aldrich) and Toluene (Sigma-Aldrich) (9:1) at 50°C, while maintaining a constant stirring at 125 rev / min. The acrylated glass slides were dried at 65°C in a vacuum oven overnight, and the modified cover slips were affixed to an aluminum print head using a double-sided tape for printing.

[0054] Custom Digital Light Projection (DLP) platform was designed and built as seen in FIG. 3A. Briefly, the optical setup consists of a 405 nm CW laser light source (iBEAM SMART 405, Toptica Photonics), DLP development kit (DLP 1080p 9500 UV, Texas Instruments, USA), and a Z-stage (25 mm Compact Motorized Translation Stage, ThorLabs). A rotating diffuser was added to the setup to eliminate laser speckles generated by the Gaussian beam profile, and to convert Gaussian intensity profile into a uniform hatshaped distribution beam profile before expanding, collimating, and projecting the beam onto the DMD which consists of an array of 1920 x 1080 micromirrors with a single pixel resolution of around 10 pm. A 3D CAD model of positive master mold of a three-chambered microfluidic chip was designed in SolidWorks (FIG. 3B). Custom MATLAB code was used1722104442.v2-8 / 13 / 25to slice the model and generate a stack of binary portable network graphics (PNG) image files. These files were uploaded onto DMD and converted to virtual masks to spatially modulate the laser beam onto a vat of liquid prepolymer solution through an oxygen permeable PDMS window. Lab VIEW code was used to control DMD and synchronize it with the upward movement of the print head to photo-polymerize the master mold onto a surface- modified glass coverslip.

[0055] PEGDA master molds were used to make the final PDMS micro fluidic chips. First, polydimethylsiloxane (PDMS, Sylgard 184: Dow Corning Corporation) base and curing agent were thoroughly mixed in a 10:1 mass ratio, degassed under vacuum, and poured onto the master molds. The molds were initially placed in a 52 mm petri dish, then PDMS was poured directly over the micro holes (reverse feature for micro-pillars) and degassed to ensure uniform filling of PDMS in all negative spaces of the mold. PDMS-mold setup was then cured for 10-12 hours at a low temperature (35-40°C) and for an additional 2 hours at a higher temperature (65 °C). After cooling, the PDMS layer was gently peeled from the molds and trimmed to size. Inlet and outlet holes were punched out of the PDMS to form 3 inlets and 3 outlet ports. Finally, the PDMS was irreversibly bonded to a glass coverslip (0.15mm thick) using oxygen plasma treatment (60-seconds exposure).

[0056] Generation of user-defined microchannels within crosslinked collagen.

[0057] TyPeI collagen solutions at varying concentrations (Rat tail tendon, ibidi, Germany) were prepared using established protocols. Briefly, stock solution of collagen (10 mg / ml) was diluted to 4 mg / ml with IX medium (DMEM, ThermoFisher) and 10X medium (HBSS, ThermoFisher) such that the final salt concentration is IX and it is neutralized to pH 7.2-7.4 by adding calculated amount of IM NaOH. All the reagents, media and collagen were placed in ice-bath throughout the neutralization process. Final pH was measured using pH strips. Before pipetting the collagen solution into the PDMS microfluidic chips, the channel surfaces were modified with (3-aminopropyl)triethoxysilane (APTES) and glutaraldehyde (GA) to immobilize the collagen and prevent detachment from channel surfaces. First, chips were sequentially immersed in a solution of (i) 10% (APTES, Sigma- Aldrich) in ethanol for 1 hour, followed by three ethanol rinses, and (ii) 2.5% glutaraldehyde (GA, Sigma- Aldrich) solution in deionized water for 1 hour, followed by three rinses using deionized water, and (iii) sterilization under UV light overnight. Collagen solution was pipetted into chamber #2 of the chips, followed by incubation at 37°C for 20 minutes to allow gelation of the matrix. Then, IX PBS supplemented with 1% Penicillin-Streptomycin (10,0001822104442.v2-8 / 13 / 25U / mL, Thermo Fisher Scientific) was pipetted into chambers #1 and #2 to prevent crosslinked collagen from dehydration and bacterial contamination. The chips were stored in incubator for 24 hours to ensure robust crosslinking of collagen and consistent channel sizes during laser ablation experiments. All femtosecond laser ablation experiments were performed after 24hours at user-defined time-points.

[0058] Custom-built fs-laser ablation setup was designed and built by combining a Ti:Sapphire fs laser (Coherent, Chameleon, USA) with a Zeiss Microscope (Observer Zl, Germany). (FIG. 8A). In this setup, an 800 nm femtosecond laser beam with a repetition rate of 80 MHz was focused inside crosslinked collagen within the microfluidic chips using a water immersion objective (40x, NA=0.8, Leica). User-defined patterns, designed using visual basic code, were used to scan the laser focus within the collagen to ablate 3D microchannels. To control the microscope stage, a Visual Basic script was developed that was integrated with Axio Vision software (Zeiss, Germany) to precisely manipulate the XYZ stage to trace complex trajectories. For example, to create a helical channel, it was determined the locus of points (x, y, z) from online resources and used it to write a visual basic code to automate microscope stage movement in 3D. A custom written algorithm was used to synchronize the stage and laser shutter to write 3D patterns at controlled speeds and laser powers. The dosage was changed by modulating the average power of the laser using a polarization-based power-tuning system or by modulating the scanning speed of the stage. Ablated patterns in both lateral and vertical directions were visualized in real time under bright-field microscopy. (FIG. 6A-B).

[0059] Laser ablation of other types of hydrogel matrix was also performed. Bovine collagen (Advanced Biomatrix, 7mg / ml) was used similarly to the rat tail collagen described earlier. Additionally, semi-synthetic gelatin methacrylate (GelMA, Advanced Biomatrix), synthetic polyethylene glycol diacrylate (PEGDA, 6KDa, Advanced Biomatrix), and their combinations were tested. Here hydrogel solutions were pipetted in Ch#2 and crosslinked for 15 seconds using LED light Cure Box (B9A-LCB-010: B9Creations) that emits light at 390- 410. Microchannels were then created using femtosecond laser ablation. Here, Lithium Phenyl (2,4,6-Trimethylbenzoyl) Phosphinate (LAP; Signa- Aldrich) was used as photoinitiator.

[0060] Characterization of 3D cell networks within target chambers of microfluidic chips.

[0061] Murine osteocyte-like cell line MLO-Y4 (Kerafast, Inc. Boston, MA) was cultured according to the suppliers recommended protocol. Briefly, T75 vented cell culture1922104442.v2-8 / 13 / 25flasks were coated with 4 pg cm2rat tail type I collagen (Sigma- Aldrich, Inc. St. Louis, MO) for 30 minutes at 37°C before cell seeding. MLO-Y4 cells were cultured in these coated flasks using minimum essential media (a-MEM, GIBCO#12571-063) containing L- glutamine, Ribonucleoside and deoxyribonucleosides; supplemented with 2.5 % heat inactivated Fetal Bovine Serum (FBS, R&D Systems, Minneapolis, MN), 2.5% calf serum (Cytiva Life Sciences, Marlborough, MA) and 1% penicillin / streptomycin (Thermo-Fisher Scientific). The fibroblast-like 10T 1 / 2 cell line, derived from a C3H mouse embryo cell and exhibiting fibroblast morphology, was bought from American Type Culture Collection (ATCC, Manassas, VA) and cultured in T75 vented culture flasks using a cell growth media consisting of a Basal Medium Eagle (BME, GIBCO#21010-046) supplemented with 10% fetal bovine serum (FBS), 1% penicillin / streptomycin and 1% GlutaMAX (GIBCO#35050- 061). MC3T3-E1, pre-osteoblast cells, provided by Dr. Horton (SUNY Upstate Medical University), were grown in a-MEM (GIBCG#A1049001) supplemented with 1% GlutaMAX, 1% sodium pyruvate, 1% penicillin / streptomycin and 10% heat inactivated Fetal Bovine Serum. Saos-2 (ATCC, Manassas, VA), human osteosarcoma cell line, was cultured in Dulbecco's modification of Eagle's medium (DMEM, Gibco#l 1965-092) as the foundational medium, supplemented with 10% heat-inactivated fetal bovine serum (FBS), 1% GlutaMAX and 1% penicillin-streptomycin. All the cells were maintained in a controlled environment at 37°C with 5% CO2 and 90% humidity.

[0062] For micro fluidic chips, all the cell types were harvested by trypsin treatment (0.05 x 10"3M in IX PBS, 5 min), centrifugation, and media washing. Cell seeding was performed by pipetting of harvested cells (1 x 106cells / ml) into the inlet ports of one of the side chambers (#1 or #3). Capillary action and differential pressure facilitate the perfusion of cell solution within the microchambers as well as the 3D ablated microchannel networks. Cells were allowed to settle in the network for 30 mins before flushing non-adherent cells from the chip using corresponding growth media, and chips were cultured in 35mm cell culture petri dish under static culture conditions (37°C, 5% CO2 and 90% humidity). Daily monitoring of cell growth using brightfield microscopy was conducted, with media channels being flushed each time cells were observed, and periodic brightfield images were captured. (FIG. 8A,B) If necessary, media channels were flushed with trypsin to prevent unwanted cell accumulation in the side chambers. For heterotypic cell seeding, MC3T3 cells were tagged with red CM-Dil dye (Invitrogen#C7001) and MLO-Y4 cells were tagged with either red CM-Dil dye or green CMFDA dye (Invitrogen#C2925) as required. (FIG. 13) Briefly, cells2022104442.v2-8 / 13 / 25cultured in T75 vented flasks were harvested and incubated in the corresponding culture media containing 2pg / ml of cell tracker dye for 30 minutes in 1.5 ml centrifuge tubes. Cells were then diluted by adding 5 ml PBS and collected by centrifuging the cell solution. Tagged cells were then seeded in channel #1 and channel #3.

[0063] For assessing cell viability, an Invitrogen Live / Dead assay kit, comprising Calcein AM and Ethidium homodimer, was utilized. Briefly, a solution containing 0.5 pl / ml of Calcein AM and 1 pl / ml of Ethidium homodimer in phenol red-free cell media (FluoroBrite, DMEM, GIBCO#A18967-Ol) was introduced into chambers #1 and #3. The chips were then incubated at 37°C for an hour, and images at different planes were captured using an inverted microscope (Nikon Eclipse). The captured images were processed using FIJI. Cells in two-layered connected square grid structures with 9 x 9 nodes for 10T1 / 2 cells and 7 x 8 nodes for MLO-Y4 cells were analyzed. Error bars were calculated as standard deviation of data collected from 3 different samples (n=3) for each cell types. To study cellular alignment, the cells were fluorescently stained for f-actin and nuclei and imaged under LSM 980 confocal microscope (Zeiss, Germany). First, 4% formaldehyde (Sigma- Aldrich) was pipetted in Ch#l and #3 for 30 minutes at room temperature and washed three times with IX PBS. Cells within Ch#2 of the chip were permeabilized with 0.5% TritonX- 100 in IX PBS for 30 minutes. Subsequently, the cells were stained with rhodamine- phalloidin (Invitrogen) at a dilution of 1 / 250 in 10% horse serum for 45 minutes at room temperature to visualize f-actin, and DAPI (ThermoFisher) at a dilution of 1 / 1000 for 5 minutes at room temperature to visualize cell nuclei. Samples were washed subsequently washed and stored in 35 mm petri dish covered with aluminum foil covered to protected from light and prevent bleaching of fluorophores. 3D images of Cellnets in connected two-layered square grid were acquired using confocal microscope (Zeiss LSM 980, Germany) and processed with FIJI image processing software. Grid of 8 x 8 nodes (area: 400pm2in each plane) in the Cellnet was analyzed. The total number of cells in the Cellnet were determined by manually counting the number of nuclei present in the nodes. In the Cellnet with ablated channel size of 8 pm, mostly had one cell was present per node, whereas for the channel sizes of 12- 15 pm, there were multiple cells per node. To access connectivity of cells, stack of f- actin-stained images were analyzed. The analysis was performed in two steps. First, cell connectivity in a single plane was analyzed by taking a single cell in a node and checking if its dendritic processes were connected to adjacent cells. In a two-layered network, single cells in are particular plane could potentially connect to a maximum of four cells (4 in-plane connections and 1 out-of-plane connection). Second, the z-stacked images were resliced to2122104442.v2-8 / 13 / 25obtain images in the X-Z plane to assess the connectivity of a particular cell to out-of-plane cell connections (z-planes). The newly acquired stacked images were analyzed frame-by- frame to check the connections of cells in different z-planes. Cell connectivity is presented as the percentage of cells connected to the number of adjacent cells in the Cellnet. Error bars represent the standard deviation of data acquired from 5 different samples (n=5, 10 grids and 640 nodes).

[0064] Real-time calcium signaling within ‘deterministic’ 3D single-cell network.

[0065] Calcium Signaling Fluo-4 Calcium Imaging Kit (Invitrogen#F 10489) dye was used to conduct calcium signaling experiments. Prior to mechanical loading or chemical stimulation, Fluo-4 calcium reporting dye was pipetted into side chambers of the chip to enable diffusion mediated loading of dye into MLO-Y4 Cellnets (3D cell networks in collagen in Ch#2). Briefly, the Fluo-4 AM loading solution was prepared by mixing 10 parts of 100X PowerLoad and 1 part of 1000X Fluo-4, AM, with 1000 parts of cell medium. Next, the cell medium was removed from all inlets and oulets of the chip, and side chambers (#1 and #3) were gently rinsed with IX PBS (37°C). lOOpL of Fluo-4 AM loading solution was added to both inlets and incubated at 37°C for 30 minutes, followed by 15 minutes at room temperature. Fluorescence data were collected at 1-3 frames per seconds under mechanical and / or chemical stimulation using confocal microscope (Zeiss LSM980) equipped with enclosed chamber maintaining cell culture conditions.

[0066] Mechanical stimulation. Fluo-4 AM loading solution was removed from the chips. Next, micro fluidic chips were mounted on the microscope stage for an additional 10 minutes allowing cells to settle down; this avoids recording of unwanted cell signals due to perturbations during the media removal steps. The data acquisition was started 30 seconds prior to applying mechanical stimulation to the cells. To apply mechanical stimulation to the cells, 150pl of fresh media was pipetted in the inlet of Ch#l. This causes the media to flow from the inlet to the outlet of Ch#l, generating shear stress at the interface of Ch#l and Ch#2 which is sensed by the cell network located at different z-planes in collagen matrix. Fluorescence signal (changes in calcium within each cell of the network) were recorded using time-lapse fluorescence microscopy, and relative changes in fluorescence intensity of each cell in the network was calculated using established methods. Briefly, calcium signal intensity was quantified with FIJI. Each cell in the 3D network was selected for intensity measurements. Intensity of first 30 seconds, prior to any stimulation, were averaged to generate an initial baseline intensity Io. Then the intensity of each frame was divided by the2222104442.v2-8 / 13 / 25baseline intensity to get fold increase of intensity I / Io . All the experiment were repeated at- least three times and the time lag between the signal transfer from one cell to another according to their location in the circuit was analyzed using Microsoft Excel and MATLAB.

[0067] Biochemical stimuli. For biochemical induced cell responses, 5 pl of ionomycin (45pmol.L-1) (Invitrogen#I24222) was introduced into the reservoirs containing 180pl of media, post baseline image acquisition, and confocal microscopy was used to record series of images for 10 minutes. Control experiments with 5 pl of media was pipetted into Ch#l of the chips to confirm that mechanical forces during pipetting does not generate any mechanical stimuli and associated calcium signaling responses; results showed no change in the signal intensity of loaded Fluo-4 AM. To model the effects of network discontinuity,2 femtosecond laser irradiation a dosage of 0.64J / cm was used; this is calculated based on a 40x objective, 0.8 NA, an aperture size of 9.3mm with power of 150mW and an exposure time of 0.5s. Femtosecond laser was used to lethally injure individual osteocytes within MLO-Y4 Cellnets and resulting changes to signal propagation and time-delays were analyzed.2322104442.v2-8 / 13 / 25

Claims

CLAIMSWhat is claimed is:

1. A process for generating 3D, single cell, functional networks, comprising: casting a polymer in a mold to provide a chip having a first chamber, a second chamber, and a third chamber, wherein the second chamber positioned between the first chamber and the third chamber; perfusing an amount of an extracellular matrix into the second chamber; thermally cross-linking the amount of the extracellular matrix in the second chamber to generate a barrier between the first chamber and the third chamber; forming a predetermined pattern of microchannels that extend within the amount of the extracellular matrix to connect the first chamber to the third chamber; seeding the first chamber with an amount of cells to be studied; and allowing the amount of cells seeded in the first chamber to migrate through the predetermined pattern of microchannels in the amount of the extracellular matrix in the second chamber.

2. The process of claim 1, wherein the extracellular matrix is collagen and can be any other ECM that is compatible with MPA3. The process of claim 1, further comprising forming the mold defining the chip having the first chamber, the second chamber, and the third chamber; second chamber positioned between the first chamber and the third chamber by using digital light projection to generate the mold from a photo-polymer solution.

4. The process of claim 3, wherein the photo-polymer solution is poly(ethylene glycol) diacrylate.

5. The process of claim 1, wherein the polymer comprises polydimethylsiloxane.

6. The process of claim 1, wherein the amount of cells is selected from the group consisting of osteocytes, osteoblasts and fibroblasts.

7. The process of claim 1, wherein the extracellular matrix is a hydrogel formed from poly(ethylene glycol) and gelatin methacrylate.

8. The process of claim 1, further comprising submerging the mold in ethanol for a first time period and then exposing the mold to ambient light for a second time period prior to casting the polymer in the mold.

9. The process of claim 1, wherein the step of forming the predetermined pattern of microchannels is performed a predetermined time period after the step of thermally crosslinking the amount of the extracellular matrix.2422104442.v2-8 / 13 / 2510. The process of claim 9, wherein the predetermined time period is twenty-four hours.

11. The process of claim 1 , wherein the step of ablating the amount of the extracellular matrix comprises ablating with 800 nm femtosecond laser beam with a repetition rate of 80 MHz.

12. A chip for generating 3D, single cell, functional networks, comprising: a chip having a first chamber, a second chamber, and a third chamber; second chamber positioned between the first chamber and the third chamber, wherein the chip is composed of a polymer; an amount of a cross-linked extracellular matrix positioned in the second chamber and providing a barrier between the first chamber and the third chamber; a predetermined pattern of microchannels extending within the amount of the crosslinked extracellular matrix and connecting the first chamber to the third chamber.

13. The chip of claim 12, wherein the polymer comprises polydimethylsiloxane.

14. The chip of claim 12, wherein the cross-linked extracellular matrix comprises collagen.2522104442.v2-8 / 13 / 25

Images