ES cell differentiation is directed by a variety of environmental stimuli mediated through the ECM, cell-cell interactions, soluble factors, and physical stimuli [30
]. In EBs, such stimuli may be particularly sensitive to aggregate size and shape. For example, cells on the EB surface are exposed to growth factors in the surrounding medium, whereas cells within an EB perceive soluble factor signals that are determined by internal diffusion gradients. EB size influences the number and degree of cell-cell contacts and the extent of ECM deposition. Some evidence suggests that mechanical forces and shear stresses may also affect differentiation outcome [35
]. The integration of these stimuli can dictate differentiation fate. For example, cells in the periphery of EBs frequently differentiate into primitive endoderm while cells at the center tend towards primitive ectoderm [5
To uniformly specify EB cell fate, the cellular microenvironment must be precisely controlled. We had previously developed a platform of PEG microwells for culturing EB populations [27
]. However, the homogeneity of EBs grown on this platform was suboptimal due to nonspecific cell adhesion, which led to EB overgrowth, fusion of neighboring EBs, and even formation of monolayers (). In this study, we adapted our previous system to optimize EB homogeneity. It should be noted that here we have not done extensive lineage analysis to confirm that the aggregates grown on our platform are indeed EBs, though our previous studies indicates that this is the case.
Nonspecific cell adhesion was a significant hurdle to achieving homogenous aggregate populations. Aiming to enhance the cell-repellence of our substrate, we tested PEGs of various average MWs for ES cell adhesion, finding that PEG surfaces were less adhesive than non-tissue culture treated polystyrene, the standard in vitro
substrate for EB suspension cultures. We noted that the cell-repellence of PEG surfaces increases with increasing average MW (). PEG 1000 surfaces manifested virtually no adhered cells. When cultured for extended periods of time, however, cells can secrete proteins and adhesion molecules that may change the properties of the surface on which they grow. This process is termed “surface remodeling” [38
]. Surface remodeling may lead to enhanced nonspecific adhesion during prolonged culture. To address this possibility, we analyzed PEGs of various MWs for protein adsorption, finding that protein-repellence also increases with increasing average MW (). Thus, we would expect high MW PEG surfaces to resist surface remodeling. For these reasons, PEG 1000 was selected as a desirable substrate for microwell fabrication. Successful application of our microwell platform requires that the PEG arrays remain stably fixed to an underlying substrate. Silane and acrylate chemistries have been widely used to graft PEG onto surfaces such as glass and to ensure the integrity of the glass-polymer interface [39
]. We treated glass supports with TMSPMA to introduce terminal acryl functional groups onto the glass surface. During free radical driven polymerization, these acryl groups established bonds with acrylate groups of the polymer, thus covalently anchoring the microwell array to its support. Acrylation increased the stability of PEG microwell arrays of low MW (), but could not alone ensure the stability of PEG 575 and PEG 1000 microwell arrays ().
We reasoned that detachment of the high MW PEG arrays might be caused by swelling upon exposure to an aqueous environment. For certain hydrogels, the degree of swelling can be predicted based on temperature and the average MW between crosslinks [41
]. High MW PEG polymers are relatively diffusive, permitting rapid absorption of water. Rapid water uptake and swelling might create forces that stress the glass-polymer interface and lead to detachment of the microwell array. Accordingly, we used prepolymer solutions made from PEG diluted in PBS at different concentrations to generate hydrated polymers that would swell less when incubated. When a prepolymer solution containing 20% PEG 1000 was used, the resulting microwell arrays were stable during incubation and could be used for extended culture periods ().
Cell aggregates were successfully grown inside the optimized microwell arrays and remained viable even for prolonged culture periods. After 6 days of culture on PEG 1000, cells had not attached to the polymer surface, and the troublesome formation of monolayers associated with PEG 258 arrays was not observed (). Along with improved cell-repellence, the increased resistance of PEG 1000 microwell arrays to protein adsorption might explain the reduction in cell outgrowth. Aggregate populations obtained from the microwells were more homogenous in size than suspension culture EBs (). An analysis of aggregates grown in the PEG 1000 platform revealed narrow diameter distributions with mean aggregate diameters falling close to the microwell diameters. Statistical analysis confirmed a significantly higher degree of diameter homogeneity among aggregates in PEG 1000 microwell arrays than in PEG 258 ().
For an EB culture system to be widely applicable in differentiation studies, retrieval methods should preserve cell viability. We used a gravity- and agitation-driven retrieval method, and compared retrieval rates from PEG 258 and PEG 1000 microwell arrays of identical geometries. Retrieval efficiency was found to increase with increasing PEG MW (). Aggregates retrieved at day 6 remained viable after removal (). Easier harvesting from PEG 1000 arrays allowed for gentler retrieval methods, thus decreasing the probability of damaging cells.
EB formation hinges on the existence of a critical cell density within microwells. Below this density, aggregates form infrequently. To successfully grow large numbers of aggregates within a microwell array, it is thus necessary to achieve a critical cell density inside a large number of microwells. The need to achieve requisite cell densities should be balanced against the drawbacks of excessive cell seeding (namely expense and potential for overgrowth). To this end, we sought to establish a seeding protocol to optimize EB formation. Seeding at low densities yielded few EBs, and cells without cell-cell contact died (). At higher densities, more aggregates formed, but cells also settled on the PEG surface between microwells and had to be washed away (). If not washed away, these excess cells formed EBs in suspension that sometimes aggregated with EBs grown inside the microwells, reducing homogeneity of the EB population. Furthermore, cells that settled between the microwells were prone to nonspecific adhesion. Interestingly, the seeding density for maximal aggregate formation and the saturation density (the seeding density at which the microwells were saturated with cells) were identical (~4 × 102 cells/mm2) and independent of diameter (). Subsequent experiments (data not shown) have demonstrated that formation efficiencies can be further increased by modifying microwell geometry.
ES cells preferentially docked in the shear-protected microwells. Cells within the microwells were shielded from stresses generated during routine washings and media changes needed for culture maintenance (). Computer simulations predicted that shear stresses in 50 μm deep micowells would be lower than in microwells of 20 μm depth. However, for a shallow microwell depth of 20 μm, simulated shear stresses in 50 and 150 μm diameter microwells remained comparable for a wide range of velocities (). As such, we could apply identical washing procedures to microwells of different diameters without introducing a potential bias for a given microwell geometry. We therefore used 20 μm deep microwells for this study. The shallow wells also attenuated vertical diffusion limits and eased aggregate retrieval. However, dependent on purpose, geometries of the microwells can be changed readily [27