Supplementary Material for: In situ collagen assembly for integrating microfabricated 3D cell-seeded matrices

Transcription

1 Supplementary Material for: In situ collagen assembly for integrating microfabricated 3D cell-seeded matrices Brian M. Gillette, Jacob A. Jensen, Beixian Tang, Genevieve J. Yang, Ardalan Bazargan- Lari, Ming Zhong, and Samuel K. Sia* Department of Biomedical Engineering Columbia University 351 Engineering Terrace 1210 Amsterdam Avenue New York, NY, USA *

2 Supplementary Methods Cell Culture Cryopreserved primary human umbilical vein endothelial cells (HUVEC), endothelial cell culture medium, and endothelial cell detach kit (containing Trypsin/EDTA, HEPES balanced salt solution, and soybean trypsin inhibitor) were purchased from Promocell (Heidelberg, Germany). HUVEC were passaged or used at approximately 80% confluence. HUVEC at passage 3-5 were used for experiments. Cryopreserved mouse 3T3 fibroblasts were purchased from ATCC (Manassas, VA) and cultured in DMEM (ATCC) supplemented with 10% calf serum (fibroblasts) and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA). Cells were passaged or removed for use when flasks reached approximately 80% confluence using a 1x trypsin-edta solution (Invitrogen). For the indicated experiments, HUVEC growth medium was supplemented with 50 ng/ml each VEGF, bfgf, and PMA (Invitrogen) 1. PDMS mold fabrication The following were purchased from the manufacturer: Poly(dimethylsiloxane) and curing agent (Dow Corning Sylgard 184 Silicone Elastomer Kit), SU photoresist and SU-8 developer (Microchem, Newton, MA), 2 and 3 silicon wafers (Silicon Sense, Nashua, NH), photomasks (24,000 dpi, CAD/Art Services, Bandon, Oregon). Master molds in SU-8 were developed using multilayer photolithography. SU photoresist (1 ml/sq. in.) was applied to clean silicon wafers and coated in a CEE

3 Brewer 100 Resist Spinner using the following parameters: spread cycle rpm at 100 rpm/s, total time 10 s; spin cycle 1000 rpm (for 165 μm thickness) or 2500 rpm (for 60 μm thickness) at 300 rpm/s, for 30 s. The photoresist was exposed to UV (195 W, 360 nm) using a Karl Suss MJB3 Contact Mask Aligner for 45 s for 165 μm thickness and 30 s for 60 μm thickness. Baking of the resists was performed on a hot plate according to the manufacturer specifications. The base layer to support the microchannels was created in the first step using a 165 μm thick resist layer, and was exposed under a mask defining the outer edges of the construct and the microfluidic inlets. The post exposure bake step was skipped, and a second 60 μm resist layer was coated over the base layer. The soft bake step for the second layer cured the exposed areas in the base layer, allowing visualization of the first mask pattern and alignment with a second mask defining the microchannel pattern. The masters were immersed in SU-8 developer to remove the uncrosslinked photoresist. PDMS and curing agent were mixed 10:1 by mass and cast on the SU-8 masters to create the molds used to cast the hydrogel constructs (Supplementary Fig. 1a). Prior to casting hydrogels, the PDMS molds were rinsed and sterilized with 70% ethanol. Hydrogel construct fabrication Sodium alginate (Pronova UltraPure MVG, 69% guluronate) was purchased from NovaMatrix (Drammen, Norway). Calcium chloride (CaCl 2 ), fibrinogen, and thrombin (from human plasma) were purchased from Sigma (St. Louis, MO). Dulbecco s phosphate buffered saline (DPBS - Ca++, Mg++ free, GIBCO) was purchased from

4 Invitrogen (Carlsbad, CA). Type I collagen from rat tail (High Concentration, 10 mg/ml) and Matrigel were purchased from BD (Franklin Lakes, NJ). Dialysis membranes (6000 Da MWCO) were purchased from Fisher Scientific (Fairlawn, NJ). Purified water was filtered using a Milli-Q system (Millipore, Billerica, MA). Untreated 8-well polystyrene plates and chambered coverglasses were from Nalge NUNC International (Rochester NY). Hydrogel construct microfabrication methods are described in the main Methods section. Imaging of live constructs Cells were stained with Celltracker Blue, Orange or Green (Invitrogen) to label the cytoplasm prior to construct fabrication. Brightfield, differential interference contrast (DIC), and fluorescence images were acquired using a Leica DMI 6000b inverted microscope fitted with a motorized stage (Leica Microsystems, Bannockburn, IL). High resolution images (1600 * 1200 pixels) were acquired using QImaging Retiga 2000R monochrome camera (QImaging, Surrey BC, Canada) and In Vitro software (Media Cybernetics, Bethesda, MD). Confocal images were taken using a Zeiss LSM 510 META microscope using 25x 0.8 NA and 40x 1.3 NA oil objectives (Zeiss, Thornwood NY). Celltracker Orange was excited at 543 nm and emission was collected using a 560 nm LP filter. Collagen reflectance was excited at 488 nm and collected using a 470 nm nm bandpass filter 2. 3D reconstructions of confocal Z-stacks were generated using

5 3D Constructor software in Image Pro (Media Cybernetics). DIC images shown in Figures 1e-j were processed using a DIC restore algorithm (Autodeblur, Media Cybernetics) followed by highpass and Gaussian filtering to sharpen the fibers, contrast enhancement, rotation, and cropping in Image Pro (see Supplementary Fig. 2 for unprocessed DIC images). Time lapse DIC imaging of collagen polymerization in the microchannels was performed by pipetting collagen suspension into the construct while positioned on the microscope at room temperature, and several images in a Z-stack were acquired centered approximately 20 μm from the coverslip surface. DIC movies were contrast enhanced and sharpened. Quantification of cell spreading and collagen fiber morphology Cell outlines in phase contrast images (10x magnification) were manually traced. Traces were filled to create binary images for segmentation and measurement in Image Pro software (Supplementary Fig. 5). Cell aspect ratio was calculated as the ratio of the major to minor axis of a best fit ellipse, and cell area was calculated from the total pixel area. We measured all segmented objects in each image (122 cells in the control group at both time points, and 107 and 102 cells in the Na-citrate group for 0.25 and 3 hour images respectively). The measured values in the groups were compared using ANOVA followed by a Tukey-Cramer post-hoc test in MATLAB (The MathWorks, Natick, MA). Trajectories of collagen fibers were manually traced in DIC images (8 images from separate experiments) using the software Image Pro and the angles calculated relative to the channel axis (0º defined as in the direction of the channel axis). In DIC

6 images, some fibers were likely less visible than others due to lack of contrast perpendicular to the DIC shear angle 3 ; all construct channels were aligned in the same direction for DIC imaging (within ~2 degrees). Patterned collagen fibers were traced inside the channel boundary, and the bulk fibers outside the boundary (see Supplementary Fig. 3 for representative images of the tracing process). Fiber angles were grouped in 10 degree intervals and patterned phase fiber distributions were compared to the bulk phase distribution using a one-way χ 2 -test in MATLAB. Fiber widths and the number of fibers per channel length were measured using the software Image Pro.

7 References 1. Kamei, M. et al. Endothelial tubes assemble from intracellular vacuoles in vivo. Nature 442, (2006). 2. Brightman, A.O. et al. Time-lapse confocal reflection microscopy of collagen fibrillogenesis and extracellular matrix assembly in vitro. Biopolymers 54, (2000). 3. Zvi, K. Microscopic differential interference contrast image processing by line integration (LID) and deconvolution. Bioimaging 6, (1998).

8 Supplementary Figure and Video Legends Supplementary Video 1. Time-lapse microscopy of HUVEC-seeded collagen flowing through a branched microchannel pattern in a fibroblast-seeded alginate bulk phase. Images (4x objective, brightfield) were taken at 1 frame per second for 34s. Displayed area is 2.9 mm * 1.5 mm. Supplementary Video 2. Time lapse DIC imaging of patterned collagen flowing and polymerizing in a collagen-doped alginate bulk phase. Fibers in patterned collagen (left) polymerize off collagen fibers present at the collagen-doped alginate bulk phase (right) interface. Flow direction is upward. Images (40x objective) were captured every 10 seconds; real time duration was 5 minutes 40 seconds. Displayed area is 40.4 µm * 40.4 µm. Supplementary Video 3. Time lapse DIC imaging of patterned collagen flowing and polymerizing in a pure alginate bulk phase. Fibers in patterned collagen (left) polymerize in the flowing ECM solution and do no nucleate at the pure alginate boundary. Flow direction is upward. Images (40x objective) were captured every 10 seconds; real time duration was 5 minutes 40 seconds. Displayed area is 40.4 µm * 40.4 µm. Supplementary Video 4. Time-lapse DIC imaging of contraction of a collagen matrix from a bare alginate bulk phase interface. Cellular contractile activity of HUVEC (stimulated with 50 ng/ml each VEGF, bfgf, and PMA) separates collagen from

9 alginate after a short time in culture. Constructs were maintained in a humidified, 37 C, 5% CO 2 stage incubator. Images (40x objective) were acquired at the edge of the microfluidic inlet every 5 minutes for 12 hours 30 minutes. Displayed area is 164 µm * 280 µm. Supplementary Video 5. Time-lapse DIC imaging of cellular contraction of patterned collagen attached to a collagen-alginate bulk phase interface. Cellular contractile activity of HUVEC within a branched channel shows distortion of fibers attached to the collagenalginate bulk phase interface (note fiber network movement in lower and upper left channels by cells at the branch point). Images (40x objective) were captured every 30 seconds; real time duration was 38 minutes 30 seconds. Displayed area is 144 µm * 144 µm. Supplementary Video 6. Time lapse microscopy of HUVECs cultured in 3D collagen patterned in a collagen-doped alginate bulk phase (corresponding to image in Fig. 3b). HUVEC (stimulated with 50 ng/ml each VEGF, bfgf, and PMA in culture medium) migrate and interact within the patterned phase. A multicellular structure appears to form at the branch point. Constructs were maintained in a humidified, 37 C, 5% CO 2 stage incubator and images acquired every 15 minutes for 15 hours (10x objective, brightfield). The focus was readjusted toward the end of the movie. Displayed area is 210 µm * 210 µm.

10 Supplementary Video 7. Time lapse microscopy of fibroblasts encapsulated in a collagen-alginate bulk phase exposed to sodium citrate. After exposure to sodium citrate, which chelates crosslinking calcium ions in alginate, the fibroblasts are able to rapidly spread and migrate in the collagen ECM. Time lapse imaging was initiated just after 5 minutes exposure to 5% w/v sodium citrate. Images (10x objective) were captured every 10 minutes; real time duration was 3 hours. Displayed area is 710 µm * 710 µm. Supplementary Video 8. Time lapse microscopy of fibroblasts encapsulated in a collagen-alginate bulk phase that is not exposed to sodium citrate. Fibroblasts in the collagen-alginate composite appear to slightly distort the gel, but are unable to extend processes, spread, or migrate, likely due to the nanometer-scale pore size of crosslinked alginate. Images (10x objective) were captured every 10 minutes; real time duration was 3 hours. Displayed area is 710 µm * 710 µm. Supplementary Figure 1. Fabrication of constructs for patterning multiple ECM phases. a, Schematic diagram of the microfabrication process. i) A two-tier mold is fabricated in PDMS using two-step photolithography of the master followed by replica molding (see Methods). ii) Alginate or collagen-alginate is pipetted onto the mold and a semipermeable membrane is flattened over the top surface. The alginate is crosslinked by pipetting a 60 mm CaCl 2 solution onto the membrane. iii) The gel is removed from the mold and placed on a substrate to seal the channels, then cell-seeded ECM is pipetted into the inlets and flow proceeds by hydrostatic pressure. iv) The construct is incubated at 37ºC to gel the ECM in the microchannels (a thrombin solution is added on top of

11 constructs with patterned fibrinogen to gel fibrin), and then culture medium is added to complete fabrication. Total fabrication time for completed constructs was approximately hours. b-e, Images of fabrication method. b, A multilayer PDMS mold is fabricated that independently defines microchannel height and overall construct thickness, and that pre-forms microfluidic inlets. The microchannel layer and base layer thicknesses are set independently using different spin coating speeds; in this case the base layer is 165 μm thick and the microchannel layer 60 μm thick, yielding an overall construct thickness of 225 μm. c, Alginate or collagen-doped alginate solutions are pipetted into the mold and a dialysis membrane is used to define the top surface of the construct. d, A 60 mm calcium chloride solution is added on top of the membrane to crosslink the alginate. If collagen was mixed with alginate, the collagen was polymerized at room temperature for 40 minutes prior to addition of calcium chloride solution. e, The microfluidic hydrogel is removed from the PDMS mold and placed on a flat substrate to seal the channels (here a glass coverslip). Multiple cell and ECM combinations can be rapidly pipetted into multiple independent microchannels. Such a design was used in Figure 1 to pattern multiple ECM within a single collagen-doped alginate bulk phase. Supplementary Figure 2. Unprocessed DIC images from Figure 1e-j. a, A pure alginate bulk phase (15 mg/ml) with 3D collagen (4 mg/ml) patterned in microchannels. b, Collagen-alginate (6 mg/ml collagen:6 mg/ml alginate) with 3 mg/ml collagen patterned in microchannels. c-f, Collagen-alginate (3 mg/ml collagen:30 mg/ml alginate) bulk phase with patterned 2.5 mg/ml collagen:4.25 mg/ml Matrigel (c), 4.25 mg/ml Matrigel (d), 2.5 mg/ml collagen:5 mg/ml Fibrin (e), and 5 mg/ml Fibrin (f).

12 Images were acquired using a 40x 0.6 NA air objective with a 2x optical zoom. Scalebars are 10 μm. Supplementary Figure 3. Analysis of collagen fiber angles. a, Distribution of angles of collagen fibers residing within a bulk collagen-alginate phase (red circles), collagen fibers patterned within a collagen-alginate bulk phase (blue diamonds), and collagen fibers patterned within a pure alginate bulk phase (black squares). Processed DIC images were manually traced (8 images from separate experiments), and the collagen fiber orientation normalized between 0º and 90º relative to the long axis of the microchannel. The decrease in orientations around 45º for all three sets of data is likely due to the lack of contrast perpendicular to the DIC shear angle 3, which renders some fiber angles less visible (and possibly accounting for some deviation from an expected uniform distribution of bulk fibers). All construct channels were aligned in the same direction for DIC imaging (within ~2 degrees). The angle distributions of patterned fibers were compared to the distribution of fibers residing entirely within a bulk collagen-alginate phase using a one-way chi-squared test; the number of fibers, chi-squared and p-values are shown for each comparison. b-d, Representative images of fiber tracing performed in Image Pro. b, Patterned phase collagen fibers in a collagen-alginate bulk phase (traces in blue). c, Bulk phase collagen fibers in a collagen-alginate bulk phase (traces in red). d, Patterned phase collagen fibers in a collagen-alginate bulk phase (traces in black). e-f, Raw images (e) and sharpened and contrast enhanced images (f) of collagen patterned in a collagen-alginate bulk phase. Scalebars are 10µm.

13 Supplementary Figure 4. Resolution and patterning fidelity of multiple ECM phases. a- b, Patterning of multiple cell-seeded collagen matrices in an acellular alginate gel (1.5%). a, Fibroblasts (blue) were flowed into the straight channel (left) and HUVEC (red) were flowed into the curved channel (right). Image is a mosaic of 20x fluorescence images overlaying DIC images. b, High magnification DIC image of boxed region in (a) (40x 0.6 NA air objective). The two cell types in their respective 3D collagen phases were patterned within 40 μm, with the alginate barrier between collagen phases being approximately 20 μm wide at the narrowest point (which approaches the resolution limit of the photolithographic setup). DIC imaging reveals the phase boundary and shows the alginate bulk phase having a smooth texture while the collagen has a course appearance (due to polymerization at 37 C which produces small fibers) with some larger fibers visible. The alginate maintained very high fidelity features as indicated by reproduction of microscale coarseness present in the master mold. Scalebars are 500 μm in (a) and 30 μm in (b). Supplementary Figure 5. Contraction of collagen matrices in bulk phase inlets. a-b, After 24 hours in culture, cell seeded collagen had separated from a pure alginate interface (a)(arrow), but maintained a stable interface with collagen-alginate, even while contraction distorted the circular inlet (b). In this design, to simplify the alignment during photolithography and prevent overhangs in photoresist, the top layer of the inlet (black arrowhead) was made smaller than the bottom layer (channel layer white arrowhead), resulting in the two visible lines. c, The interface (arrow) between collagen (seeded with HUVEC) and collagen-doped alginate (acellular) at three weeks in culture

14 remains stable. Multicellular HUVEC structures are observed in the microfluidic inlet. HUVECs on the construct surface can be seen behind the focal plane. d, After 36 hours culture in a pure alginate bulk phase, cell seeded collagen contracted extensively throughout the inlets and channels, and in some areas fully separated and floated away from the bulk phase (arrow). e, Cell free collagen did not visibly contract from the pure collagen interface in the inlets (left) or channels (right) over the same culture period. Scalebars are 500 µm (a-b, d), 200 µm (c), and 100 µm (e). Supplementary Figure 6. Representative images of cell tracing and segmentation. a, Phase contrast image (10x objective) of fibroblasts in a collagen-alginate matrix 3 hours following exposure to sodium citrate. b-d, Cell outlines were manually traced (b), the traces were filled to create a binary image (c), and the objects segmented and measured in Image Pro software (d). Scalebar 100 µm.