Microscopy and image analysis
Co-cultures were imaged either shortly after generation (4–12 h) or maintained in their culture medium for 7 days, with medium exchange every 2 days, and then imaged. The fluorescent images were taken with an ImageXpress Micro Microscope (Molecular Devices, Sunnyvale, CA) at 4× magnification. For the purpose of visualizing the whole well, multiple images at 4× were acquired and stitched together using MetaExpress software (Molecular Devices).
To quantify the reproducibility of patterning, each well in a 24-well plate was patterned with ARPE-19-GFP cells as described above. Images of each well (the entire well) were acquired. We analyzed the location of the pattern interface by measuring the y-coordinate of the interface at different x-coordinates. The coordinates were then converted into distance from the bottom left corner of the screen (for clarity of presentation) and used to calculate mean and standard deviation. Results and discussion
Figure 1A shows a schematic of our approach for generating patterned co-cultures in multi-well plates based on differential deposition of BSA as well as the outcome of each step of the patterning process. BSA is a commonly used protein for blocking the adsorption of other cell-adhesive proteins, thus preventing both specific and non-specific binding of cells to a culture surface (20, 21). The first critical step in the patterning process is differential deposition of BSA on the tissue culture polystyrene (TCPS) surface. To achieve this, we placed a multi-well plate (24-well size in this case) on a slanted support angled at approximately 40° and added 200 μL of 0.5% BSA solution to each well using a multi-channel pipettor (Figure 1B and C). This volume was chosen to ensure that approximately half of the well was pre-patterned with BSA. The profile of the BSA solution in the angled well ultimately dictates the profile of the cellular patterns generated. We found that by controlling the angle under which the plate is placed (Supplementary Figure S1) or the amount of BSA solution placed into each well (Supplementary Figure S2), we could control the ratio of the well surface that was covered with BSA and, ultimately, the space available for cell type 2.
Plates were incubated for 2 hours in a slanted position at 37°C to allow BSA deposition onto the polystyrene surface. It was important to keep the angle of the plate constant at all times. After the incubation, we aspirated BSA from the wells and let them air dry. Successful BSA deposition and the subsequent establishment of a clear boundary between BSA-exposed and non-exposed regions are critical for the outcome of patterning. Conveniently, we also found that we were able to generate patterned co-cultures using BSA-coated plates prepared up to three weeks prior to patterning if stored in a dry place at room temperature, making this patterning method even more attractive for screening applications.
Following preparation of the plates, we seeded a suspension of cell type 1 in serum-free medium into normally oriented (non-slanted) plates. The use of serum-free medium was critical for our differential adhesion strategy to work robustly: serum or other proteins in the medium can adsorb on the substrate and mask the blocking effect of BSA, resulting in incorrect patterning. The number of cells added during this step was also crucial. The total amount of cells seeded should cover 100% of the well surface (Figure 1D). For most cell types tested, 2 × 106 cells suspended in 500 μL provided optimal results for seeding a 24-well plate. If the cell density is too low, cell type 1 does not generate a confluent sheet and cell type 2 will adhere to unwanted areas, resulting in disrupted cell adhesion patterns. Seeding densities may require adjustment depending on cell size.
Over time, cells seeded in the pre-patterned wells attached only to the BSA-free regions of the substrate. However, the duration of this incubation step is important for successful patterning. Sufficient time must be allowed for the cells to attach and spread on the BSA-free part of the plate, but extended incubations can result in undesirable cell attachment to BSA-covered regions. Once the cells were attached to the BSA-free region, we removed non-adhered cells by washing the plate 3–5 times with serum-free medium. This step produced a clear interface between the region where cell type 1 is attached and the cell-free region (Figure 1E). We also found that we could achieve cell patterning by tilting a solution of fibronectin in non-tissue culture treated polystyrene plates (Supplementary Figure S3). However, because the cost of BSA is significantly lower than that of fibronectin, we focused on optimizing our method using BSA deposition on tissue culture polystyrene plates.
Next, we rendered the BSA-covered portion of the well cell adherent again by incubating the plate in culture medium supplemented with 10 μg/mL fibronectin. A solution of cell type 2 (mCherry expressing ARPE-19 cells) in serum containing medium was then added to generate a patterned co-culture. Similar to the first seeding step, 1 × 106 cells per well were used, ensuring that 100% surface coverage immediately after seeding (Figure 1F). Cells were given time to adhere to the available surface before removing excess cells by gentle washing with culture medium. Once unadhered cells were removed, a co-culture with a clear boundary between the two cell types was visible (Figure 1G).
We next determined if our method is suitable for other cell types and if it was possible to generate heterotypic co-cultures. We were able to generate heterotypic co-cultures of HUVEC/ARPE-19 (Figure 2A), homotypic co-cultures of BJ cells (Figure 2B), as well as heterotypic cultures of ARPE-19/MDCK (Figure 2C), and ARPE-19/BJs (Figure 2D). Our results show that the method presented here is compatible for use with different cell types, including epithelial, endothelial, and fibroblast cells. Moreover, the patterning was equally successful in established cell lines and primary cells. In most cases, cell type 2 did not adhere to the areas occupied by cell type 1, resulting in clearly distinct populations (Figure 1, Figure 2A and C). For some cell types, however, the second cell type tended to attach to areas covered by the first cell type. This was the case when BJs were used as cell type 2. In homotypic co-culture of GFP-BJ and mCherry-BJ (Figure 2C), the adhesion of cell type 2 onto cell type 1 was more pronounced than in the heterotypic co-culture of GFP-ARPE-19 and mCherry-BJ (Figure 2D). This suggests the effect is cell type–dependent and can be stronger or weaker for different cell combinations. We speculate that cells which readily form aggregates will be more challenging to pattern as they will tend to adhere equally to both the exposed TCPS surface and the pre-existing confluent sheet of cell type 1. For this reason, when generating heterotypic co-cultures, if one cell type is particularly adhesive to either itself or the other cell type, this cell population should be patterned in the first seeding step to ensure optimal pattern quality.