Microscale 3-D collagen cell culture assays in conventional flat-bottom 384-well plates

Abstract

Three-dimensional culture systems such as cell-laden hydrogels are superior to standard 2-D monolayer cultures for many drug-screening applications. However, their adoption in high throughput screening (HTS) have been lagging, in part due to the difficulty of incorporating these culture formats into existing robotic liquid handling and imaging infrastructures. Dispensing cell-laden pre-polymer solutions into 2-D well-plates is a potential solution, but typically requires large volumes of reagents to avoid evaporation during polymerization, which increases cost, makes drug penetration variable and imaging complex. Here we describe a technique to efficiently produce 3-D ‘microgels’ using automated liquid handling systems and standard, non-patterned, flat-bottomed, 384-well plates. Sub-millimeter-diameter, cell-laden collagen gels are deposited on the bottom of ~2.5 mm-diameter microwell with no concerns over evaporation and meniscus effects at the edges of wells, using aqueous two-phase system patterning. The microscale cell-laden collagen-gel constructs are readily imaged and readily penetrated by drugs. Cytotoxicity of chemotherapeutics were monitored by bioluminescence and demonstrates that 3-D cultures confer chemoresistance, as compared to similar 2-D culture. This data hence, demonstrates the importance of culturing cells in 3-D to obtain realistic cellular responses. Overall, this system provided a simple and inexpensive method for integrating 3-D culture capability into existing HTS infrastructure.

Introduction

Conventional in vitro high throughput screening (HTS) and high content screening (HCS) platforms have mostly been based on 2-D cell culture platforms due to their compatibility with robotics, liquid handling systems, and imaging platforms. In parallel, 3-D culture platforms such as cell-laden hydrogel have gained much attention as alternative, and in many ways, more physiologically-accurate culture models. Cells maintained in 3-D culture display altered gene expression profiles^{1, 2}, metabolic functions^3–5 as well as sensitivities towards drugs^{6, 7} and physical stimuli^{8, 9}. Despite these advantages, the adoption of 3-D culture methods into industrial HTS platforms has been slow, partly due to the cumbersome hydrogel handling techniques and challenges in maintenance, automated data collection and analysis. The most straightforward approach to introducing 3-D matrix in cell-based assay is to embed cells in a hydrogel matrix. Common matrices used in 3-D platforms include naturally derived extracellular matrix (ECM) proteins such as collagen, fibrin and Matrigel. Collagen type I is the most abundant of these ECMs found in the body¹⁰ and would be valuable to incorporate into a 3-D HTS format. Cell-laden collagen gels can be formed directly on non-patterned culture dishes¹¹, but requires large gel volumes (usually tens of µL) and the throughput is low. Other methods such as tube casting of collagen modules¹² and microfluidic-based generation of collagen microbeads¹³ can achieve much higher throughput but requires specialized equipment and expertise, and are not robust or mature enough technologies to support the demanding nature of HTS assays. Hence, adoption of these novel technologies into the HTS industry is limited. Ideally, techniques to generate low-volume collagen microgels in conventional HTS multiwell plates using existing robotic liquid handling infrastructure would greatly aid adoption of 3D cultures in HTS industry.

Fabricating collagen gel at the microscale within conventional multiwell plate can be challenging¹⁴, primarily due to evaporation during the thermal gelation process. Once extracted from its source, usually from rat tail or bovine skin, collagen is kept in solution and stored at low temperature and low pH to prevent gelation. Even for small volumes of material, neutralized collagen solution takes 30–40 minutes at 37°C to completely gel. Microscale constructs in 384-well plates would need to be prepared with just a few microliters of collagen solution exacerbating this evaporation problem and significantly reducing viability of any embedded cells¹⁵. Although evaporation may be minimized by tightly monitoring and controlling the atmospheric humidity during gelation, it would require specialized equipment and complicates its integration into existing HTS infrastructure. The effects of evaporation can be partially alleviated by increasing the gel volume. However, the large wall surface area to internal volume ratio of a 384-well would lead to the formation of significantly concaved meniscus, even with just tens of microliters of gel. Such curvature may lead to complications and optical interference during microscopy and other analysis modalities.

Here, we design a 3-D culture solution that has been adapted for an automated 384-well plate format, using a previously described method to fabricate collagen microgels in an aqueous two phase system (ATPS)¹⁵. Our ATPS system consists of two immiscible, phase-forming aqueous polymer solutions, poly(ethylene) glycol (PEG), and dextran (DEX). A 3-D microgel is formed by adding collagen into the DEX phase and dispensing the mixture into a solution of PEG, where the collagen will partition to the interface and undergo thermal gelation to form a microgel. We characterized the size and morphology of the collagen microgel as a function of dispensing volume. The aqueous environment reduces evaporation-driven cell death and enables the formation of collagen microgels with volumes as small as 0.5 µL. We also compared sensitivity to cytotoxic compounds between cells embedded in collagen microgels and cells seeded on a 2-D surface using a bioluminescent assay.

Materials and Methods

Cell culture

Breast ductal carcinoma cell line MDA-MB-231 expressing Click Beetle green luciferase CBG99 (Promega) expressed from lentiviral vector FUW (gift of D. Baltimore), denoted here on as 231-LUC, was described in previous study¹⁶. Cells were cultured at 37 °C in 5% CO₂ using Dulbecco’s Modified Eagle’s Medium (DMEM) (Invitrogen) supplemented with 10% Fetal Bovine Serum (FBS) and 1% antibiotic-antimycotic (Invitrogen). Cell stock were maintained in 100mm plastic culture dishes until they reach ~80% confluence. For seeding and passaging, cells were washed with PBS (Gibco) then detached using 0.25% Trypsin/EDTA (Gibco) and placed into the appropriate cell culture vessels.

ATPS collagen microgel printing in 384-well plate

Collagen microgels were printed onto the floor of 384-well plates using the aqueous two phase printing method. Briefly, stock solutions of dextran T500 (20% w/w DEX in PBS; DEX from Pharmacosmos) and polyethylene glycol (20% w/w in DMEM; PEG from Sigma) were prepared as previously described¹⁵. Collagen-DEX solutions were prepared by diluting Type I bovine Collagen (BD Biosciences) to 2 mg mL⁻¹ in a sterile solution of 10% v/v 10× DMEM, 1% v/v 3M NaOH and 3% DEX in deionized water. This solution was kept on ice to prevent gelation. To prepare cell-laden collagen microgels in 384-well plates, 231-LUC cells were detached and mixed with collagen-DEX solution at a concentration of 10⁶ cells/mL and loaded into a chilled 96-well round bottom dispensing plate. A small volume (1–5µL) of cell collagen-DEX mixture was dispensed into a clear, flat-bottom 384-well plate (Greiner Bio-one, #781098) containing 50µL of PEG (6% w/w) using a Cybi®-Well 96-channel simultaneous pipettor (CyBio, Jena, Germany) fitted with sterile Cybi®-TipTray 96 tips (25µL, #OL 3800-25-633-S). Liquid handling routines were compiled and executed by CyBio Composer software (v2.13 CyBio, Jena, Germany). Dispensing height refers to the distance between the tip of the pipette to the bottom of the well. Two different heights, 0.5mm and 2.5mm, were tested for the delivery of cell-laden collagen-DEX solution into PEG. We also tested two dispensing speeds at high (10) and low (1) setting within the CyBio Composer. After dispensing the cell-laden collagen-DEX solution, the 384-well plates were then placed in an incubator for 30 minutes to allow the collagen to polymerize. Once gelation is complete, the wells were rinsed 7 times with medium, each time replacing 25µL, which corresponds to 50% of the total liquid volume in each well. This gives a theoretical dilution such that less than 1% of the original PEG remains. For 2-D culture system, cells were also seeded directly into the wells of the 384-well plate at the same cell seeding density and using the same Cybi-Well setting as 3-D collagen microgel.

Chemotheraputic dose response studies

The cytotoxic effects of three chemotherapeutics toward 231-LUC were compared between cells seeded in 2-D culture and cells embedded in 3-D collagen microgels. At one day post-seeding, cells were treated with clinical formulations of Cisplatin (NDC-0703-5748-11), Paclitaxel (NDC-55390-304-50), and Doxorubicin (NDC-0069-3030-20), from the University of Michigan Hospital Pharmacy. Drugs were added by replacing 25µL of medium with working solutions at 2× final concentrations. Appropriate series of concentrations spanning 4-orders of magnitude were chosen for each compound. Cell viability was assessed using a bioluminescence assay. After incubation for two days the 384-well plates were rinsed 3 times with medium. For bioluminescence imaging 25µL of a 2× working solution of luciferin (Promega) was added to each well resulting in a final concentration of 1.5µg/mL luciferin. The plates were kept at room temperature and imaged after 30 minutes. Capture parameters including aperture, binning and exposure time were adjusted so that the luminescence values falls between 9000 and 600 counts sec⁻¹. Bioluminescence signals were collected using an IVIS Spectrum Imaging System (Perkin Elmer) (Figure S4). Data was acquired and analyzed using Living Image 4.0 (Perkin Elmer). Signals from replicates (N=8) were averaged, normalized against no-drug control and plotted against the log of drug concentrations (+/− SEM).

One experimental artifact that may cause such discrepancy between 2-D and 3-D responses may be the presence of detached cells that contain active luciferase and are readily rinsed away during the pre-luciferin assay wash in 2-D rather than 3-D culture. To rule out this possibility, we compared 2-D culture with and without pre-luciferin assay rinse. The signals observed for the without rinse samples were in close agreement with the rinsed control (Figure S1), suggesting negligible contribution of residual luciferase activity to the observed signals.

Cytotoxicity of PEG and Z’-factor of ATPS collagen microgel bioluminescence assay

To examine the cytotoxic effect of PEG 35K solutions, we incubated ATPS collagen microgel (1µL) containing 1000 231-LUC cells in a series of half-diluted PEG 35K solutions. The final concentrations of PEG 35K used in the dose curve, in descending order, were 6.00%, 3.00%, 1.50%, 0.75%, 0.38%, 0.19%, 0.09%, 0.05%, 0.02% and 0% (w/w). Viability was assessed after 2 days using bioluminescence assay as described in previous sections. We also took the bioluminescence readings at 0% and 1.5% PEG as negative and positive controls, respectively, to calculate the Z’-factor of this assay. The Z’-factor is a commonly used statistical parameter characterizing an HTS assay which reflects both the signal dynamic range as well as data variations of signals. Because Z’ only uses data from positive and negative controls, it is an assessment of the quality of the assay itself without intervention of drug candidate. Using a previously described formula¹⁷, Z’-factor can be calculated as follows:

$$Z\prime -\mathit{\text{factor}}=1-\frac{3s_h-3s_l}{|{\overline{X}}_h-{\overline{X}}_l|}$$

where s is standard deviation, X̄ is average, and subscripts h and l represent positive and negative controls, respectively. The Z’-factor calculated from these value is 0.83, indicating the assay quality is very good^{17, 18}.

Mass transport simulations

Simulations of diffusion of molecules through collagen gels of various sizes were conducted in COMSOL v4.3 (Burlington, MA), using previously determined values for diffusion of large molecules in collagen-DEX (unpublished data). Diffusion coefficients for 0.1, 1, 10, and 100 kDa-sized molecules were approximated using this data and the Einstein-Stokes relationship between hydrodynamic radius and diffusion. Diffusivity coefficients ranged from 1.18 × 10⁻⁹ m²/s for 0.1 kDa molecules to 1.18 × 10⁻¹⁰ m²/s for 100 kDa molecules. Geometric parameters for cylindrical collagen gels (in well-plate), and hemispherical collagen gels (in ATPS) were estimated using previously published characterization data¹⁵. Simulations were set up such that the boundary of the hydrogel hemisphere was fixed at a finite concentration of the diffusing molecule, while the bulk of the hydrogel was initialized at a concentration of 0 ng/mL. Simulations report the equilibration fraction of the gel center, as a function of time. This setup approximates the case of adding soluble factors to the well-plate, but neglects any delays that might occur due to diffusion of the soluble factor through the media surrounding the hydrogel. Since dispersion rates of molecules dispensed into water is typically at least an order of magnitude faster than diffusion through hydrogels, this delay should be negligible.

Results and Discussions

Characterization of ATPS collagen microgel

Previous studies from our group have demonstrated that collagen microgels can be fabricated in aqueous medium using ATPS¹⁵. In the current study we adapt this technique to create 3-D cultures in 384-well format using robotic liquid handling infrastructure, similar to that already available in existing commercial HTS equipment. Using the described collagen-ATPS formulations, we were able to form collagen microgels in 384-well plates. Similar to our previous observations¹⁵, collagen microgels deposited on the bottom of the well took on a circular morphology and their sizes were proportional to printing volume (Figure 1A and B). Collagen microgels also maintained good circularity, which only began to diminish at higher printing volume of 5µL (Figure 1C and Figure S3), suggesting that maintaining uniform culture conditions between wells in HTS would be facilitated by using highly reproducible low-volume droplets. In addition to printing volume, dispensing height may also affect microgel morphology. We tested two printing heights, at 0.5mm and 2.5mm above the bottom of the well. With our formulation, lower dispensing height increased the likelihood of forming non-circular or satellite microgels, however, dispensing speed had no visible effect on microgel morphology (Figure S2). Maintaining similar morphologies between microgels in each well would help simplify automated imaging protocols. With optimized parameters (dispensing volume = 1µL, dispensing height = 2.5mm, speed setting = 1), the yield of ATPS microgel printing was 100% for all the plates used in our experiments, and all the microgels remained attached to the bottom of the well. Overall, the robotic pipettor was able to deliver droplets with a high degree of accuracy and reproducibility in terms of shape and volume.

Figure 1.

Physical characterization of collagen microgel. A) Representative brightfield micrograph of a 1µL collagen microgel in 384-well plates (scale bar = 250µm). B). Microgel areas at different dispense volumes. All subsequent experiments were performed using a 1µL microgel volume. C) Microgels remain uniform and circular at dispense volume up to 2µL.

One of the advantages of using microgels instead of collagen macrogels is the reduction of equilibration time needed for small molecules to penetrate through the bulk of the gel. The long diffusion times for large molecules in three-dimensional matrices may hinder transport of the test drug candidate to the cells encapsulated within the 3-D matrix, and cells located in the center of such gels may not even be stimulated by the drug of interest over the time-course of the experiment¹⁹. To determine the impact of our microgels on alleviating this diffusion problem, we constructed finite element models to simulate diffusion of different size molecules, from 0.1 kDa to 100 kDa, into bulk collagen gel of 100µL in a 96-well and two ATPS collagen microgels of different volume (10µL and 1µL) (Figure 2). As expected, the models predict that the time it takes to equilibrate in 3-D collagen gels is proportional to both the volume of the gel and the size of the molecule of interest. Small molecules including chemotherapeutics such as cisplatin (MW = 300 Da), doxorubicin (MW = 543 Da) and paclitaxel (MW = 854 Da), were able to penetrate and fully equilibrate in ATPS collagen microgels within 5 minutes, but in bulk collagen gel complete saturation takes approximately 6 hours. The small diffusion distance and high surface area-to-volume ratio of collagen microgels facilitates rapid saturation. This difference is magnified as the molecular weight increases. In our simulations, molecules weighing 1 to 100 kDa, (a range that encompasses several growth factors including insulin, TGF-b and VEGF), only reached 80% and 40% saturation concentration in bulk collagen macrogel after 4 hours, respectively. In contrast, the increase in molecular weight posed only a modest delay in equilibration, and both the 1µL and 10µL microgels become completely equilibrated in less than 1 hour. The simulations presented here demonstrate that the size of ATPS collagen microgel allows molecules with weights spanning a large range to diffuse rapidly and simultaneously to reach embedded cells. This can be a crucial feature in situations where one needs to observe the combined effects of compounds with drastically different molecular weights, or in a situation where the experiment requires the complete removal or substitution of compound during the time course. In these experiments, the cell-seeding densities were significantly lower than that in densely-packed tumor tissues or cell spheroids (<10⁶ cell/µL). This is expected to allow for greater oxygen penetration depth and minimize hypoxic regions within the sub-millimeter diameter microgel. Depending on application requirements, the method is flexibly in terms of the cell seeding density and it should also be possible in the future to create higher cell density constructs if more hypoxic conditions are desired.

Figure 2.

Finite element models showing percent saturation of molecules of different weights as a function of time. For all geometry and volumes, the time required for complete saturation increases as the molecular weight of compound increase. With the lowest surface area to volume ratio, conventional macrogel in 96-well plate (A) presents significant diffusion barrier to embedded cells. In contrast, ATPS collagen microgel both large (B) and small (C) were able to reach saturation much more quickly.

While there is essentially no difference in cell-dispensing technique for seeding 384-well plates with conventional 2-D cultures compared to our 3-D collagen microgel assay, the presence of ATPS, specifically PEG, requires thorough washing to minimized cytotoxicity^{15, 20}. To avoid accidental removal or dislodgement of the collagen microgels, it is important to perform a series of half-volume sequential dilutions, with the pipettor set to withdraw at half the depth of the well (dispensing height = 5mm) and dispense at the slowest speed (setting = 1). One of the concerns is that such gentle rinse step would not be effective in removing the viscous PEG solution. To compensate for the poor rinsing efficiency we performed this rinse cycle 7 times over a total time of 20 minutes to ensure that the PEG would be sufficiently diluted to maintain cell viability. The number of rinses was chosen to yield a >99% theoretical PEG removal efficiency and the rinse time represents the minimum time required for the CyBi-Well to execute the serial dilution routine. Using the bioluminescence assay we confirmed that our post-rinsing protocol involving 7-times half volume serial rinse (theoretical final concentration of PEG 35K = 0.05% w/w) was sufficient to preserve the viability of 231-LUC cells embedded in collagen microgels (Figure 3). Though fewer rinse steps may also be sufficient, for this proof-of-concept demonstration, this rinsing protocol was adopted in all subsequent experiments.

Figure 3.

Cytotoxic effect of PEG on 231-LUC cells embedded in ATPS collagen microgel. Microgel were incubated in half-serial diluted PEG containing medium from 6% to 0.0234% w/w PEG 35K. After 2 days cell viability was determined by bioluminescence assay and plotted against log[conc. PEG] (+/− SEM, N=4). Dashed line indicates the concentration of PEG (0.0468% w/w) after 7× half-serial washes.

Differential responses to chemotherapeutic between 2-D and 3-D models

Beyond differences in mass transport kinetics, studies have also indicated that culturing cell in 3-D matrix instead of 2-D surfaces lead to alteration in cell phenotypes and function, including differentiation^{21, 22}, proliferation²³ and survival²⁴. For example, many tumor cells display reduced chemosensitivity in 3-D cultures²⁵. However, this increased resistance to chemotherapy may simply be due to increased diffusion transport barriers between the drug and cells embedded in a bulk 3D matrix, or protected by surrounding tumor cells, such as in a spheroid model system²⁶. Hence, the impact of cell culture ‘dimensionality’ alone remains unknown. Since our ATPS-microgel system enables culture of cells in microscale volumes of collagen, concerns over transport limitations can be eliminated. Moreover, culture of cells within 3D matrices that are attached to a surface enables the cells to generate tension within the matrix, without causing bulk matrix contraction, which in turn would alter the transport characteristics of the tissue, Cells that are allowed to generate tension can display differential response towards growth factors^{27, 28} as well as modulation of signaling networks²⁹. Since cells in the body typically exist in matrices under tension, these features must be considered in the design of in vitro drug-screening platform since the ultimate goal is to recapitulate the cellular responses in vivo and predict the outcome in the respective animal model.

Using bioluminescence assays, we compared the cytotoxic responses of mammary carcinoma cell expressing luciferase, 231-LUC, towards three commonly used chemotherapeutics, cisplatin, doxorubicin and paclitaxel. Both cisplatin and doxorubicin exert their cytotoxic effects by intercalating and/or crosslinking DNA molecules, thus inhibiting mitosis and triggering apoptosis³⁰. Paclitaxel operates by stabilizing microtubules during mitosis, thereby interfering with normal cell division³¹. Using the ATPS collagen microgel technique we seeded a 384-well plate where each well contained a single 1µL microgel embedded with 1000 231-LUC cells. The cells were treated with a serial dilution of the aforementioned chemotherapeutics and their viabilities were compared after 2 days against a 2-D control plate where each well contained 1000 231-LUC cells seeding in culture medium. At the end of treatment, luciferin was added to each well and the resulting bioluminescence was recorded (Figure 4A and B). In 2-D culture, cell viability decreased steadily as drug concentration increased, down to 20–30% at the highest concentrations compared to a no-drug control. In contrast, cell viability in 3-D culture dropped only to approximately 80% of control for all three drugs even at the highest concentrations. Our data agrees with previous findings^{25, 32, 33}, where cells cultured in 3-D showed a substantial decrease in chemosensitivity, and clearly demonstrates that these effects are related to the dimensionality of the cell culture environment, and not to transport limitations associated with conventional 3-D cultures. While uncovering the mechanism behind the differences in chemosensitivities between cells in 3-D versus 2-D is beyond the scope of this study, we believe that this difference may be caused by decreased cell proliferation in 3-D systems³⁴, thereby reducing drug uptake and activity

Figure 4.

Bioluminescence based cytotoxicity assay against three common chemotherapeutics (cisplatin, doxorubicin and paclitaxel). Bioluminescences from MDA-MB-231 cells expressing Click Beatle Green luciferase (231-LUC) seeded in 384-well plates were measured using IVIS Spectrum imager. Cells seeded in 3-D collagen microgels showed increased chemoresistant to all three compounds compared to their 2-D counterparts (*P<0.005, +/− SEM, N = 8).

Performance and advantages of the ATPS collagen microgel platform

The quality and robustness of an assay is crucial in HTS applications due to the large number of test condition and relatively small number of replicates. For this reason, statistical parameter known as the Z’-factor has been developed and widely adopted as a tool to assess the performance of HTS assays¹⁷. The Z’-factor is calculated based on the values obtained from the positive and negative controls of the assay. Here we calculated the Z’-factor for the ATPS collagen microgel cytotoxicity assay using bioluminescence values from 231-LUC cells at various concentrations of PEG (Figure 3). The calculated Z’-factor of 0.83 indicates that the assay is excellent with a large degree of confidence in identifying cytotoxic effects.

Aside from improved mass transport characteristics, the use of small volume collagen microgel also provides savings in both reagents costs and imaging time. When compared to conventional macrogel, a 1µL microgel platform provides a 100-fold reduction in collagen cost and cell number required per assay (Table 1). Simply pipetting collagen solution into non-patterened 384-well plates in an attempt to reduce collagen volume will cause most of the material to be lost in the corners of the well due to capillary action. Therefore the use of ATPS provides a unique solution to easily create collagen microgels on ordinary culture surfaces by localizing collagen gel and reducing interfacial forces that can cause collagen drop wicking near the walls of the well.

Table 1.

Cost and analysis time comparison between conventional macrogel and ATPS collagen microgel. Cost of collagen was calculated based on the collagen solution used in this study. Z-stack scan time for confocal imaging was calculated based on 5µm optical slices and a scan time of 10ms per slice.

	100µL macrogel in 96-well plate				1µL microgel in 384-well plate
Number of wells	Cost of collagen ($)	Number of cells (106/ml)	z-stack height (mm)	z-stack scan time	Cost of collagen ($)	Number of cells (106/ml)	z-stack height (mm)	z-stack scan time
1	1	105	3.13	6.3 sec	0.01	103	0.78	1.5 sec
103	1 thousand	108	3.13	1.73 hr	10	106	0.78	25 min
106	1 million	1011	3.13	72 days	10thousand	109	0.78	18 days

Small volumes of collagen in microgels also reduces the overall height of the gel compared to collagen macrogel and therefore would significantly reduce the time needed for z-stack based confocal imaging. For example, the estimated height of a hemi-spherical 1µL collagen microgel is 0.78mm. In contrast, a 100µL collagen macrogel in 96-well has a height of 3.13mm, which represents a 4-fold increase in z-stack height during confocal imaging. The smaller height of the microgel also means that the entire depth of the gel can be imaged by brightfield or epiflorescence microscopy using a low magnification objective lens with low numerical aperture. Hence, the small size of collagen microgel will also make this assay more easily compatible with a variety of microscopic techniques including automated high throughput imaging.

Conclusions

In this study we employed ATPS patterning techniques to deliver cell-laden collagen microgels into 384-well plates using common robotic liquid handling system used for HTS. We found that collagen microgels take on a predictable size and morphology as a function of dispensing volume. Cells embedded in collagen microgels remain viable, and when compared to cells seeded on 2-D surfaces, display significantly lower chemosensitivity. Our simulation also predicted that the small size of ATPS collagen microgel enables rapid and simultaneous delivery of compound over a wide range of molecular weights. Technologically, the method described here provides a simple and inexpensive way to incorporate microscale 3-D culture into existing HTS platforms. Compared to conventional macrogels, the small volume of ATPS collagen microgel can be more easily and rapidly analyzed with various imaging modalities and presents a significant saving in costs of materials. We believe the ATPS 3-D collagen microgel culture system will be widely applicable to many HTS application and represents a significant advancement in the field of microengineered culture platforms.