In the fields of bioengineering and cell biology, three-dimensional (3D) cell culture provides a means to more accurately resemble the physiological in vivo environment for preclinical studies (e.g. drug screening, cellular assays).^[1-3] Specifically, multicellular spheroids have been extensively used for studying embryogenesis in the form of embryoid bodies (EBs),^[4-6] adult tissue growth and organogenesis,^[7,8] cancer progression and liver toxicity.^[9,10] To date, technologies that generate multicellular spheroids are limited in culture duration (requiring spheroid transfer), optical clarity issues for imaging, or broad size distributions.

The hanging-drop method is a commercially available technique that has been extensively utilized in spheroid culture, yet this process is labor intensive due to the need for spheroid transfer and sometimes lacks reproducibility.^[11] Micromolding and photolithography have been used to create microwells made of PDMS,^[12,13] or hydrogels such as poly(ethylene glycol) (PEG)^[14,15] and agarose.^[16] But these technologies sometimes require multiple labor steps and produce microwells with limited optical transparency for imaging, protein adsorption issues, size restrictions or sample loss with media exhcange, thus resorting to spheroid transfer to another plate.

Here, we created hydrogel microstructures made of photocrosslinkable PEGDA with gradual concave topohraphies that are optically clear and can be utilized for long-term (e.g. with media-exchange, for durations beyond 2-3 days) cell spheroid culture. PEG is an FDA approved biomaterial and often utilized in cell culture for its low immunogenicity, minimal protein adsorption, lack of adhesive peptides (which in turn limits cell-material interaction and promotes cell aggregation), as well as optical clarity.^[17] The structures are fabricated with a 3D projection printer that uses nonlinear UV light exposure. We demonstrate their feasibility for spheroid culture in two distinct models – breast cancer spheroids and induced pluripotent stem cells (iPSC) EBs. In the breast cancer model, we grow the spheroids to 10 d, noting size changes and staining of hypoxia and necrosis, important markers in tumor progression.^[9] Next, we use the platform to generate EBs of iPSCs. iPSCs have become a desirable cell type as they are autologous (patient-derived) by nature and thus have the potential to be used in a multitude of patient-specific in vitro models and therapies. We show tight uniformity in EB size after 3 d, with important undifferentiated markers expressed. Expanding the culture to 10 d, we witness the EBs' spontaneous differentiation into the three germ layers, as evidenced by immunofluorescent staining. Importantly, EBs remained within the concave hydrogels during the entire process. This platform opens the door for more biological models to be developed of many cell types, including, but not limited to, cancer, embryogenesis, and patient-derived disease models using iPSCs.

Concave hydrogel microstructures for spheroid culture were fabricated using a light-based, continuous 3D projection printing technology adapted with nonlinear UV light epxosure. (Figure 1a).^[18,21] A 2D image of a gradient circle pattern is converted to a series of layer slices (53 layers in total) based on its grayscale intensity at each pixel (Figure 1a). Each layer represents a cross-sectional image in the series in proportion to the height of the structure (500 μm). The series is then fed to the digital micromirror device (DMD) for UV projection onto the photocurable prepolymer solution – in this case, 20% (w/v) poly(ethylene glycol) (PEG) diacrylate – in a continuous fashion. Importantly, this 3D printing technology permits the creation of any complex and precisely defined concave structure simply by changing the design or gradient of the inputted pattern (Figure S1). This feature represents a major advancement to previous 3D printing platforms, which rely on printing one dot or one layer at a time, while overcoming limitations associated with micromolding of soft biomaterials with complex designs.

A schemata of the 3D printing process at the molecular level is displayed in Figure 1b. For the first 15 layers, or masks, UV light is projected onto the entire prepolymer solution, photocrosslinking the base of the microwell structure. Subsequent optical masks with increasing areas of non-exposure (black, as indicated in Figure 1a) are displayed on the DMD. The concave hydrogel is therefore built in a continuous layer-by-layer fashion, alongside a continuously moving z-stage that coordinates its movements in the z direction with changes in the optical masks. Because we set the z-height to be 500 μm and there are 53 layers, the stage moves 9.4 μm for each layer, maintaining the same projection plane within the prepolymer solution as it moves through the layer series.

UV photopolymerization and gelation of PEGDA is a nonlinear process, where free radical initiation, polymer chain propagation, and termination take place on multi-order kinetics.^[22] Thus we sought to create a 3D printing process that allows for nonlinear UV exposure (see Experimental). Figure S2 depicts the changes to the nonlinear fabrication parameters as well as the outputted structure, maintaining the same gradient circle deisgn throughout.

As T₀ decreases, the well shape becomes wider and less polymerized, and largely unpolymerized in the middle of the concave hydrogel (Figure S2a, panels i to ii). This lack of polymerization is presumably due to a lower exposure time for the 15 base layers, where the entire solution is exposed to UV light. We hypothesize that a longer exposure time to the base layers is required to generate free radicals for the rest of the structure. While increasing exposure to the base layers can be achieved by increasing T₀ in a linear fashion, this method overpolymerizes the remaining layers of the hydrogel structure allowing for an undefined shape that is not optically clear (Figure S2a, panel i). We thus modulated the nonlinear factor, A₂, to vary the exposure time for each layer. When A₂ is negative, every successive layer is exposed for a shorter duration than the previous layer, in turn speeding up the entire fabrication process as it proceeds through the entire 53 layers (Figure S2a, panels iii to v). By increasing T₀ and making A₂ more negative, the bulk of the UV irridation shifts to the earlier layers, allowing a longer duration for free radical generation in the base layers (where the entire prepolymer solution is exposed to UV light).

We empirically determined the optimal T₀ and A₂ values to be 0.95 s and -0.023, respectively, fitting our aforementioned design criteria – that is, a an opticially clear, concave hydrogel that permits single spheroid formation in its center (Figure S2a, panel v). Figure S2b provides a graphical understanding of the cumulative exposure time in accordance to the layers for each of the five cases shown in Figure S2a. It is interesting to note the cumulative exposure time for the first 15 base layers increases from 6.0 s for linear exposure to 10.2 s for nonlinear exposure in panels ii and v of Figure 2a, respectively (Figure 2b inset). Thus, we believe that a longer duration of UV exposure to the base layers is required to initiate the free radical polymization process throughout the prepolymer solution. Below this time, we observed unpolymerized sections in the microwell center.

We used scanning electron microscopy to assess the topography of the hydrogels. (Figure 2a). The hydrogel displays a gradually increasing slope from the center to the edge and steep walls, indicating a concave shape. Atomic force microscopy was used to characterize the stiffness profile on the concave hydrogel surface (Figure 2b). The structure displayed a soft, low modulus center (10 Pa) that stiffened to the edge of the well (∼200 Pa). The tallest part of the structure – the wall of the hydrogel – represented the stiffest region (1-2 kPa). We hypothesized that the soft center correlates to earlier layers of UV exposure during the fabrication process, and as it proceeds through the layers, increasing UV exposure drives additional crosslinking to stiffen the hydrogel. We confirmed this by taking stiffness measurements of flat hydrogel structures with different UV exposure to the base (Figure S3). It appears that the flat wells with 15 base layers has an average stiffness of 20 Pa, while 24 base layers and 34 base layers have higher moduli profiles of 151 Pa and 203 Pa, respectively. Thus, it is likely that the gradient UV exposure in our concave hydrogels is due to the variable light exposure in the continuous layer-by-layer 3D printing process.

For preliminary cell studies, we fabricated flat or concave hydrogels and seeded BT474 breast cancer cells to examine the effect of concavity on spheroid generation (Figure 2c and d). When flat hydrogels were used in cell culture, several spheroids of varying sizes formed within each well, while the desired single spheroid formation was achieved in the concave hydrogel microstructures. Expanding on our first cell experiments, BT474 breast cancer cells were seeded at various densities and used to assess tumor spheroid generation and growth within the concave hydrogels (Figure 3a). At day 2, LOW (250 k mL^-1) and HIGH (750 k mL^-1) cell seeding densities produced spheroids with diameters 146 ± 11 μm and 213 ± 16 μm, respectively (Figure 3b). However, over the course of the next several days, spheroids from the HIGH group began to plateau at a size of around 250-275 μm, while the smaller spheroids from the LOW group continued to grow in size, albeit smaller than the 250 μm threshold. Growth rates for each group confirmed this trend (Figure 3b, inset). At day 10, spheroid diameters for both groups were within standard deviations of each other – 269 ± 17 μm and 273 ± 12 μm for LOW and HIGH groups, respectively.

Interestingly, live/dead staining with calcein AM/ethidium homodimer at day 10 showed that the HIGH group exhibited a 10-fold increase in its dead core area, compared to the LOW group: 25,394 ± 5514 cm² and 3,385 ± 1,565 cm² for HIGH and LOW groups, respectively (Figure 3c). This observation suggests a necrotic core forming in the HIGH group, correlating with regression in spheroid growth. It has been well documented that tumor spheroids greater than ∼200 μm in diameter demonstrate a hypoxic core due to a nutrient and gas transport gradient, which in turn can lead to necrosis.^{[9, 23]} The presence of a hypoxic core in the tumor spheroid provides a more physiologically relevant tumor model for cancer screening applications, as tumor hypoxia in vivo drives a pro-angiogenic cascade for continued growth and invasion.^[24] Hypoxia was confirmed with immunostaining of the spheroid cross-sections for HIF-1α, a biomarker for hypoxia (Figure 3d), and necrosis was observed in hematoxylin and eosin staining (Figure 3e). The spheroids showed considerable hypoxia and necrosis more prevalently in spheroids from the HIGH group. These data are in good agreement with previous literature regarding tumor spheroid progression (e.g. hypoxia and necrosis).

Human iPSCs were utilized in subsequent experiments for generating and culturing EBs. iPSCs, derived by retroviral transduction of a combination of four transcription factors, Oct4, Sox2, c-Myc and Klf4, are stem cells with an equivalent self-renewal and differentiation capacity as embryonic stem cells.^[14] In addition to their pluripotency, iPSCs provide a superior platform for clinical translation because they are autologous by nature (patient-specific). This facilitates their use in personalized disease modeling, drug testing, and regenerative medicine development, as well as minimizing any ethical concerns.

iPSCs were seeded on top of the concave hydrogels at a density of 100 k mL^-1. Single EBs formed after three days of culture, with an average diameter of 155 ± 17 μm (Figure 4a). Flat microstructures, conversely, generated a broader distribution of EB sizes, such that an initial seeding density of 200 k mL^-1 produced EBs of 129 ± 48 μm. We reported similar observations for flat hydrogels with our breast cancer spheroids (Figure 2d). This is also consistent with previous literature on flat microwells that EBs only form at a critical cell density proportional to the microwell size, below which they form infrequently or at varied sizes.^[25] At day 3, EBs showed pluripotency by immunostaining for Nanog and Oct4, transcription factors highly expressed in embryonic stem cells (Figure 4b). Grown to day 10, EBs displayed morphological changes in their size, shape, and appearance in the form of intra-organoid cavities (Figure 4c and Figure S4). We hypothesized that this was due to spontaneous differentiation that can occur in these pluripotent cells, based on similar observations in the literature.^[5] Immunostaining confirmed EB differentiation to all three germ layers for both cell seeding densities, as evidenced by their co-expression of SOX-17 (endoderm), SOX-1 (ectoderm) and brachyury (mesoderm) (Figure 4d). These differentiation results serve to only show the possibility of visualizing differentiation of a single EB housed in the concave hydrogel. Further studies will be needed to address and quantify the different stages of embryogenesis and differentiation, as well as a more focused review on the necessary components in cell culture (e.g. media, growth factors) contributing to tissue-specific differentiation.^[4,26]

The concave hydrogel platform described here can be a valuable tool in the development of a multitude of spheroid-based cell culture models, especially for longer timepoints beyond the first media exchange. These may include tumor progression (e.g. proliferation, hypoxia, necrosis), migration and angiogenesis as well as various EB, and in particular iPSC, studies such as embryogenesis, organogenesis, toxicity, and patient-specific disease models. Due to its high reproducibility, low cost (material and time), ease of fabrication, and retention of the spheroids for long-term culture, this technology could also be adapted for high-throughput screening if individual hydrogel microstructures were to be printed into a high-throughput plate.