Complementary, High-throughput Methods for Creating Multi-dimensional, PEG-based Biomaterials

Tunable biomaterials that mimic selected features of the extracellular matrix (ECM), such as its stiffness, protein composition, and dimensionality, are increasingly popular for studying how cells sense and respond to ECM cues. In the field, there exists a significant trade-off for how complex and how well these biomaterials represent the in vivo microenvironment, versus how easy they are to make and how adaptable they are to high-throughput fabrication techniques. To address this need to integrate more complex biomaterials design with high-throughput screening approaches, we present several methods to fabricate synthetic biomaterials in 96-well plates and demonstrate that they can be adapted to existing liquid handling robotics. These platforms include 1) glass bottom plates with covalently attached ECM proteins, and 2) hydrogels with tunable stiffness and protein composition with either cells seeded on the surface, or 3) laden within the three-dimensional hydrogel matrix. This study includes proof-of-concept results demonstrating control over breast cancer cell line phenotypes via these ECM cues in a high-throughput fashion. We foresee the use of these methods as a mechanism to bridge the gap between high-throughput cell-matrix screening and engineered ECM-mimicking biomaterials.

affected by both stiffness and ECM protein composition on 2D hydrogels. [6][7]20 We have also shown that tumor spheroids encapsulated into 3D hydrogels are drug resistant compared to TCPS. 10 Additionally, biomaterials can be used to predict stage-specific phenotypes in disease. For example, early stage melanoma cell drug response was shown to be insensitive to the dimensionality of the platform, but metastatic melanoma cells show resistance in 3D environments. 21 These studies demonstrate controlled in vitro ECMs are useful tools for preclinical studies.
The need for engineered biomaterials in the drug development pipeline is clear, since only 8% of filed Investigational New Drug applications will be approved by the Food and Drug Administration, 22 due in part to false positives missed early in the drug development pipeline. Therefore, there is a critical need to scale up these types of biomaterials for them to be feasible for high-throughput screening. To achieve a paradigm shift in high-throughput screening with biomaterials, they must be formatted to use existing equipment and technology, such as liquid handling robotics and plate readers. 23 To address this need, we integrated three biomaterials our lab has developed into 96-well plates and liquid handling robotics ( Figure 1). In this study, we give a detailed overview of their development, use, and how they can be made amenable to high-throughput technologies. They range from the simplest, chemically modified glass surfaces, to a significantly more complex tunable 3D hydrogel cell culture system. We propose that each of these platforms are incredibly useful for in vitro studies of cell-ECM interactions, and the level of complexity chosen by an individual lab should be guided by the biological question at hand. Custom 96-well plate fabrication. We used bottomless polystyrene 96-well plates (Greiner Bio-One, Kremsmünster, Austria) as the base of the custom well plate. We manually cut a rectangular piece of ARcare 90106 (Adhesives Research, Glen Rock, PA) 24 medical-grade, double-sided adhesive to fit the well plate bottom and it was attached to the plate by hand ensuring that there were no air bubbles present. Holes were cut in the adhesive from a custom template designed in Adobe Illustrator CC 2014 (Adobe Systems, San Jose, CA) with a laser cutter (Epilog Mini 18 Laser, Golden, CO). We purchased glass pieces cut to size (110 mm x 75 mm, S.I. Howard Glass, Worcester, MA) with the same thickness as a number 1.5 coverslip (0.17 mm) to optimize microscopy applications. We adhered the glass to the plate by removing the remaining backing to the adhesive and pressing the glass firmly to ensure a tight seal around each well. The glass was attached to the plate before or after chemical modification depending on the biomaterial platform ( Figure S1).

MATERIALS AND METHODS
ECM protein well plates. ECM proteins were attached to the glass surface in a 96-well plate format by modifying a procedure described previously. 19 24 (Thermo) for two hours to prevent non-specific protein adsorption, the wells were rinsed three times, and UV sterilized for one hour before cell seeding. MDA-MB-231 or Hs578T cells were seeded in serum free cell culture medium at 34,000 cells/cm 2 and were imaged with an Axio Observer Z1 (Carl Zeiss AG, Oberkochen, Germany) one day post-seeding, N ≥ 85 cells were analyzed per condition. either collagen I alone (Thermo) or a "collagen-rich" solution containing 65% collagen I (Thermo), 33% collagen III (recombinant human, FibroGen, San Francisco, CA), and 2% fibronectin (Millipore). 6 The plates were stored at 4°C overnight and rinsed three times with pH 3.8 PBS, followed by three rinses with pH 7.4 PBS, UV sterilized for one hour, and then rinsed with cell culture medium prior to cell seeding.

RESULTS
Three distinct biomaterial platforms adapted to 96-well plates and liquid handling robotics.
We previously developed a biomaterial system with ECM proteins covalently attached to glass, providing control over the integrin-binding of the cells by changing the protein composition on the surface. 19 This chemistry was initially developed for single coverslips, and this method was incompatible with highthroughput applications. To optimize this chemistry for multi-well plates, we applied this same silane chemistry on a well plate-sized piece of glass ( Figure 1a). We used oxygen plasma treatment to promote surface hydroxylation followed by overnight chemical vapor deposition of APTES. We functionalized the surface with DSC to provide the amine-reactive succinimidyl ester and then attached the glass to a bottomless 96-well plate with a medical grade adhesive ( Figure S1a-b). Then, the desired single protein or protein combination was covalently attached within each well using liquid handling robotics. The remaining surface area was blocked with non-specific PEG to ensure that the cells would only bind to the intended proteins on the surface. Then, cells were seeded into the wells for experiments. To make protein-modified glass-bottom plates, we chemically modified a well plate-sized piece of glass. This glass was attached to a bottomless 96-well plate and ECM proteins (purple and blue) were attached in solution with three hours of surface contact followed by a PEG blocking step for two hours before cell (green) seeding. b. 2D PEG-PC hydrogels were covalently attached to fully assembled glass bottom 96-well plates functionalized with a methacrylate silane. The hydrogels were functionalized with full length proteins via sulfo-SANPAH before cell seeding. c. 3D PEG-MAL hydrogels were covalently attached to fully assembled glass bottom 96-well plates functionalized with MPT. The PEG-MAL (4-arm, purple) pre-polymer solution containing cell-adhesion (black/blue) and cell-degradable (black/orange) peptides was mixed with single cells and the PEG-dithiol crosslinker (gray), and allowed to polymerize on the silane-functionalized glass surface.
2D PEG-PC hydrogels were synthesized in 96-well plates with an efficient photoinitiator ( Figure   1b). We improved this platform from our initial adaptation in a 96-well plate 6  We also adapted 3D PEG-MAL hydrogels 10 to a 96-well plate format (Figure 1c). Briefly, fully assembled glass bottom 96-well plates were oxygen plasma treated and functionalized with a solution of MPT. The MPT presents a thiol group, which reacts with the maleimide to covalently attach the 3D hydrogel to the plate. This 3D platform consisting solely of PEG and peptides is the quickest platform to fabricate: the plate functionalization, pre-polymer solution, and cell seeding can all be completed within a few hours.

Custom-built 96-well plates with covalently bound, full-length ECM proteins and protein
combinations. To create a high-throughput system with precisely tailored integrin-binding ECM proteins, we modified commercially available glass with aminated silane (Figure 1a). 19 To ensure that the silane chemistry was working as advertised, we performed XPS at each chemical step (Figure 2a). We covalently attached the cell-binding peptide RGD (GRGDSPCG) to the surface of the glass in lieu of a full-length protein due to the limitations of XPS resolution and demonstrate a predictable change in the surface chemistry across all the reaction steps in this process (Figure 2b).
To demonstrate the ability to tune ECM protein concentration and combinations, we coupled 0-2 μg/cm 2 of full-length collagen I to the amine-reactive glass surfaces. We quantified the protein coupling with a modified ELISA and observed a statistically significant increase in collagen I over the range of concentrations that we tested (Figure 2c). Real tissues are comprised of multiple proteins, 29-30 therefore, we performed the same collagen I range quantification against a background of a constant concentration of fibronectin in the mixture. Again, we detected a statistically significant change in collagen I concentration across this range while in this ECM protein mixture (Figure 2d).
We then used this tunable glass bottom plate system to demonstrate changes in cell phenotype

LAP is an efficient photoinitiator to synthesize 2D PEG-PC hydrogels in a high-throughput
format. In addition to protein composition, 29-30 every tissue has a specific stiffness range 31-32 that can be mimicked using hydrogels. We had developed a 2D PEG-PC hydrogel that is polymerized with Irgacure 2959 (Irgacure), 33 a commonly used photoinitiator in hydrogel synthesis. [34][35][36] However, this initiator could not be used in a black-walled 96-well plate due to light intensity loss. To overcome this, we used LAP, 25 which forms free radicals more efficiently at 365 nm light than Irgacure. 25 We made 1, 3, 6, and 10 wt% PEGDMA hydrogels with a constant 17 wt% PC. In the multi-well plates, varying the crosslinker (PEGDMA) in the hydrogels allowed for control of the effective Young's modulus (Figure 3a). Of note, the modulus measured for hydrogels created in the 96-well plates was detectably lower than the modulus of the same condition made on glass coverslips ( Figure S2a). This is likely due to weaker light transmittance in black-walled plates. 34,36 We also tested the stability of the covalent attachment of these hydrogels to the glass bottom surfaces (Figure 3b). This attachment was similar comparing hydrogels attached to the glass-bottom multi-well plates and individual coverslips ( Figure S2b), suggesting that the efficacy of the silane treatment is similar between the two different glass formats.
To evaluate our ability to control protein attachment to these hydrogel surfaces, and to demonstrate a proof-of-concept study using this system to study cell behavior, MDA-MB-231 and Hs578T breast cancer cell lines were seeded on a 10 wt% hydrogel condition coupled with 5 μg/cm 2 collagen I alone or a "collagen-rich" mixture (65% collagen I, 33% collagen III, & 2% fibronectin). 6 After 24 hours, we fixed and stained the  (Figure 3c). Both cell lines spread out less on the soft hydrogels compared to TCPS, confirming the results of many other groups. 6,33,[37][38][39][40] The amount of fibrillar F-actin in cells cultured on the 2D hydrogels was much lower in both cell lines than on TCPS (Figure 3c). [41][42] Comparing the 2D hydrogels by qualitative visual inspection, we observed that more cells could adhere to the "collagen-rich" substrate over the collagen I only substrate for both MDA-MB-231 and Hs578T cell lines. This result is likely due to the variety of integrin-binding sites provided by the protein mixture. The most visible F-actin was present in the Hs578T cells on the "collagen-rich" substrate (Figure 3c), suggesting that this cell line exhibits more intracellular tension on the protein cocktail than the MDA-MB-231 cells. Largely in agreement with our previous work on PEG-PC gels, we saw very little evidence of punctate focal adhesions in any of the conditions tested, which appears to be a general feature of these breast cancer cell lines. 33

PEG-Maleimide hydrogels with encapsulated cells in 3D created with liquid handling
robotics. Cells experience a 3D microenvironment in vivo. To recapitulate this feature in a highthroughput biomaterial format, we adapted our 3D PEG-MAL hydrogel 10 to 96-well plates via liquid handling robotics. As in our PEG-PC system, we could control the effective Young's modulus of the hydrogels within the well plates by tuning the polymer weight percent (Figure 4a). Also, there was no significant difference between the stiffnesses measured for hydrogels synthesized on coverslips or in multi-well plates ( Figure S3a). This result was expected since this reaction does not require UV light irraditation. 26 During gelation, the 3D PEG-MAL hydrogels covalently attached to the glass bottom plates by reacting to the thiol group we silane-functionalized to the surface. Most of the 3D PEG-MAL hydrogels remained adhered to the well surface for more than 30 days (Figure 4b), which was also observed for the 2D PEG-PC hydrogels (Figure 3b), and this stability was dependent on the polymer weight percent, likely due to the increased number of available maleimide groups (Figure 4b). It was not possible to synthesize these 3D PEG-MAL hydrogels in plates that were not silane treated, because the gels wicked to the sides of the wells. When we synthesized the hydrogels on silane treated coverslips, they adhered for more than As a proof-of-concept study, we encapsulated MDA-MB-231 or Hs578T cells in 3D PEG-MAL hydrogels with both manual pipetting ( Figure 4c) and liquid handling robotics (Figure 4d). We chose three peptide sequences, RGD, GFOGER, and DGEA, that represent cell-adhesion sites in collagen I (GFOGER & DGEA) [43][44][45] and fibronectin/vitronectin/osteopontin (RGD). 44,46 Additionally, an enzyme-degradable crosslinker (Pan-MMP, GPQG↓IWGQ) was incorporated into some of the hydrogels to allow the cells to degrade and move through the matrix. Cell viability was measured at one and four days postencapsulation via LIVE/DEAD staining (Figure 4c-d, S3d). Cells were over 80% viable within the hydrogels one day post-encapsulation regardless of whether the hydrogels were made manually or via liquid handling robotics (Figure 4c-d). The cells encapsulated in the Pan-MMP degradable hydrogel had higher viability (>60%) at day four than the PDT only crosslinking condition (<60%, Figure S3d), likely related to the increase in cell spreading. Overall, these results indicate that the PEG-MAL hydrogel system is adaptable to automated liquid handling robotics, and the incorporation of degradable sequences in the matrix improves cell viability.

DISCUSSION
Many groups have developed 3D cell culture platforms and suggested that future applications could include drug screening. [4][5]47 They could have significant impact, if adapted to current screening technologies. Towards this goal, we adapted our lab's existing biomaterial technologies to 96-well plates and automated liquid handling robotics.
Since the importance of cell culture dimensionality has been increasingly evident, 48 a tunable, reproducible, and easily fabricated high-throughput 3D in vitro cell culture platform is a valuable tool for drug screening. [48][49] In our previous 3D PEG-MAL hydrogels drug screening study, 10 we used 10 μL hydrogel volumes to avoid diffusion limitations, and in 96-well plates this volume wicks to the edge of the wells. Here, we overcame this challenge by functionalizing glass well surfaces with thiol-terminated silane to covalently attach the 3D hydrogels to the surface. The stiffness of the hydrogels in the plate is tunable (Figure 4a), and these hydrogels also stick reliably to the surface of the plate for over 30 days (Figure 4b), critical for long-term cell culture studies. The silane chemistry could be used to attach hydrogels up to 100 days ( Figure S3c), indicating that this surface preparation can be done well in advance of hydrogel synthesis.
The chemistry of PEG-MAL hydrogels allows for gelation at neutral pH, 26 and the encapsulated cells are not exposed to UV light, as is common in other systems. 9,11 The PEG-based hydrogel system allows the specific incorporation of cell-binding and cell-degradable functional groups, which is not possible with protein-based gels such as collagen 50 or Matrigel. 51 Moreover, these protein based gels have high amounts of batch-to-batch variability, 52 (Figure 4c). This is expected because these peptides are susceptible to cell-secreted enzymes used to degrade local matrix.
Cell viability was also maintained with the use of liquid handling robotics (Figure 4c-d). Looking specifically at Hs578T cells encapsulated with RGD, there was lower survival on day four than on day one. Perhaps due to a need for additional integrin-binding sites other than RGD or an increase in total binding sites for long-term survival in the hydrogel. This result is important for scaling up our 3D PEG-MAL hydrogel system for high-throughput applications.
For situations where full-length proteins are essential, we also demonstrate a silane-based approach to create multi-well plate 2D hydrogels. This also allows users to capture stiffness when 3D geometry is unnecessary to answer the biological question at hand. Others have developed 2D hydrogel platforms, 7,20 which produce perfectly flat hydrogels, but they are complicated to fabricate 7 or require a plate insert, 20 limiting the number of plates that can be made at once. Microarrays with multiple 2D hydrogel stiffnesses and protein binding surfaces have been developed to demonstrate that both factors impact mesenchymal stem cell behavior. 3 Hydrogels with incorporated nanofibers to control the spatial arrangement of biochemical signals have also been developed, 53 but this type of system is complicated to fabricate. Our original 2D PEG-PC hydrogel system design 6 was hindered for scale up because gelation required an oxygen-free environment. Irgacure is a photoinitiator that is commonly used for initiating hydrogel polymerization, [34][35][36] and it was used for 2D PEG-PC hydrogels when they were first developed for coverslips. 33 Irgacure is not a very efficient photoinitiator 25 because it does not absorb light readily at 365 nm, 34 but this wavelength is used in other platforms that include cell encapsulation to limit damage to the cells. 34,[54][55] Since the pre-polymer solution resides deep within the wells of black-walled plates, the efficacy of Irgacure was particularly problematic, and therefore it was not possible to form PEG-PC hydrogels in 96-well plates. Since LAP is more efficient than Irgacure, it was possible to make the PEG-PC hydrogels in 96-well plates using this photoinitiator.
Using LAP, we could control hydrogel stiffness within the plate (Figure 3a), and the silane treatment allowed for the long-term adherence of the hydrogels to the well surface ( Figure 3b). Both MDA-MB-231 and Hs578T breast cancer cell spreading was mechanosensitive, in agreement with observations by us and many other groups. 6,33,[37][38][39][40] Specific to our study, both cell lines adhered and spread more on the "collagen-rich" mixture than with collagen I alone (Figure 3c), perhaps due to the additional integrin binding sites provided. Both MDA-MB-231 and Hs578T breast cancer cell lines express high levels of β 1 integrin, 56 which pairs with multiple integrin α subunits to bind to collagen I and fibronectin, possibly explaining the result we observed.
To isolate cell-ECM binding, we developed a third form of silane-treated glass-bottom 96-well plate. Previous work with this platform had been limited to a 24-well plate, 19 where treated glass coverslips were prepared and inserted by hand. Unlike the 2D and 3D hydrogel platforms, the chemical functionalization to the glass cannot be performed in a fully assembled glass-bottom well plate due to the incompatibility of polystyrene and some of the solvents. Therefore, the custom 96-well plate format was created by first modifying one well plate-sized piece of glass and attaching it to a commercially available bottomless 96-well plate with a medical-grade adhesive. This approach presents a feasible method for scaling up to 384 or 1,536 well plates, also commonly used in industrial applications. Since the surface concentration of protein in each well can be carefully controlled (Figure 2c-d), the 2D biomaterial platforms described above could be used as a straightforward way to study tumor progression where collagen density increases and ECM composition changes. 57 We observed an overall trend of increasing cell area in MDA-MB-231 and Hs578T cells with increasing collagen I surface concentration (Figure 2e), alongside a decrease in the circularity of the Hs578T cells (Figure 2f). These types of studies might help identify a certain threshold of ligand density where efficacy of a compound is greatly enhanced or reduced.
The use of synthetic biomaterials allows for the capture of both biochemical and mechanical cues experienced by cells in vivo. These rationally designed in vitro models are beneficial because components of the microenvironment can be controlled individually, an impossible task in vivo. The likely application of interest for each platform will be dependent on the user's end goal. For example, the ECM proteins attached to glass and 2D PEG-PC hydrogels incorporate full-length proteins whereas the 3D PEG-MAL hydrogels only include peptides representative of integrin-binding sequences to specific proteins. Since the structure of the protein can impact integrin-binding, [58][59] it is an important consideration in biomaterial choice. Our goal was to present mechanisms to use each of these platforms in highthroughput applications. Specific features included in each, such as protein versus peptide, stiffness, and geometry will guide their exact applications in individual labs or companies.

CONCLUSIONS
We have adapted three unique, but complementary biomaterial platforms to 96-well plates for high-throughput screening applications. While more complex and expensive than traditional TCPS, these systems include specific microenvironmental features that could be key to better predict in vivo cell behavior, either for applications in preclinical drug screening or as better tools for hypothesis test-beds.

SUPPORTING INFORMATION
Supplemental Materials and Methods: DGEA synthesis Figure S1: Custom glass bottom 96-well plates can be made with commercially available materials.