Waffle‐inspired hydrogel‐based macrodevice for spatially controlled distribution of encapsulated therapeutic microtissues and pro‐angiogenic endothelial cells

Abstract Macro‐encapsulation systems for delivery of cellular therapeutics in diabetes treatment offer major advantages such as device retrievability and high cell packing density. However, microtissue aggregation and absence of vasculature have been implicated in the inadequate transfer of nutrients and oxygen to the transplanted cellular grafts. Herein, we develop a hydrogel‐based macrodevice to encapsulate therapeutic microtissues positioned in homogeneous spatial distribution to mitigate their aggregation while concurrently supporting an organized intra‐device network of vascular‐inductive cells. Termed Waffle‐inspired Interlocking Macro‐encapsulation (WIM) device, this platform comprises two modules with complementary topography features that fit together in a lock‐and‐key configuration. The waffle‐inspired grid‐like micropattern of the “lock” component effectively entraps insulin‐secreting microtissues in controlled locations while the interlocking design places them in a co‐planar spatial arrangement with close proximity to vascular‐inductive cells. The WIM device co‐laden with INS‐1E microtissues and human umbilical vascular endothelial cells (HUVECs) maintains desirable cellular viability in vitro with the encapsulated microtissues retaining their glucose‐responsive insulin secretion while embedded HUVECs express pro‐angiogenic markers. Furthermore, a subcutaneously implanted alginate‐coated WIM device encapsulating primary rat islets achieves blood glucose control for 2 weeks in chemically induced diabetic mice. Overall, this macrodevice design lays foundation for a cell delivery platform, which has the potential to facilitate nutrients and oxygen transport to therapeutic grafts and thereby might lead to improved disease management outcome.


| INTRODUCTION
Encapsulation of exogenous therapeutic cells in hydrogel-based devices is a promising strategy enabling long-term, immunosuppressantfree cellular therapy for the treatment of hormone and proteindeficiency diseases, such as type 1 diabetes, hemophilia A, or neurodegenerative diseases. 1,2 Transplanted cells need to be protected from the host immunological attack while maintaining sufficient metabolic exchange and secretion of their respective therapeutic agents. 3 The semi-permeability and high water content of hydrogel facilitate the diffusion-dependent exchange of nutrients and metabolic wastes between encapsulated cells and their surrounding environment while protecting these therapeutic cells from attack by host antibodies and immune cells. 4,5 Therapeutic cells or microtissues could be embedded in microencapsulation systems consisting of spherical microcapsules or in macroencapsulation systems comprising a variety of shapes and structural designs. 6 A macro-device, which consists of a single compartment containing multiple therapeutic microtissues, is removable from the recipient either for cell replenishment or in the event of unexpected complications after transplantation. 7,8 Thus, macro-encapsulation devices offer remarkable ease of retrieval over microcapsules, which are often scattered in uncontrolled distribution at the implantation site, rendering it more difficult to retrieve them completely. Moreover, in contrast to the limited design of only spherical geometry for hydrogel microcapsules, macro-encapsulation systems exploit a wider range of structural designs for the microtissue-containing compartment. Particularly for islet transplantation, a wide range of designs for diffusion chambers or cell-laden hydrogel sheets have been evaluated for their immuno-isolation performance in multiple preclinical studies yielding variable outcomes. 4,[8][9][10] For example, the bi-layered polytetrafluoroethylene (PTFE) TheraCyte™ macrodevice comprised an inner immuno-isolating membrane and an outer vascularizing membrane, which were designed to facilitate the ingrowth of new blood vessels to enhance survival of encapsulated islet grafts. 11 Even though diabetic mice transplanted with islet-loaded TheraCyte devices remained euglycemic for 6 months, 12 the hydrophobic PTFE component of these devices elicited macrophage activation and fibrotic reaction in rats. 13,14 To deliver sufficient tissue dosage for clinically relevant therapeutic efficacy, a macro-encapsulation device is typically burdened with a high cell packing density, increasing the risk of cell or microtissue aggregation that hampers their viability. 7 Yang et al. reported decreased viability of microtissues encapsulated in a macrodevice at higher cell loading densities, albeit with the incorporation of convective transport for oxygen and nutrients. 15 Furthermore, the uncontrolled spatial distribution of therapeutic microtissues in macrodevices also exacerbates the tendency for microtissues to cluster, thus, potentially limiting the mass transfer of oxygen to the encapsulated cells and subsequently resulting in hypoxic cell death as well as an ultimate loss of therapeutic function of the transplanted cells. 16,17 A multitude of oxygenation strategies such as exogeneous oxygen supply or delivery of angiogenesis agents for vascularization have been explored to tackle the challenge of diffusion-dependent hypoxia. 18 However, only a limited number of studies focused on solving the issue of nonhomogenous distribution. 10,19,20 For example, organized distribution of loaded cells or microtissues in macro-encapsulation devices might facilitate efficient allocation of nutrients and oxygen, mitigating the need for additional supplies. Specifically, Lee et al. reported an alginate-collagen composite hydrogel sheet with arrays of homogeneously distributed spheroids for the regulation of blood glucose levels in diabetic mice. 20 Yet, large spacing between distributed spheroids in this device lowered tissue loading capacity as 11 microtissueloaded hydrogel sheets with an area of 1 Â 1 cm each were required to restore the glycemic control of one mouse. Transplantation of such large devices not only burdens diabetic recipients but also potentially increases device-induced host immune reaction due to a higher volumetric ratio of encapsulating materials to the dosage of therapeutic cells. 6 In addition to the limitation in oxygen diffusion induced by their aggregation due to poor spatial distribution, therapeutic microtissue, or cells also suffer from the inherent lack of vasculature at the implant site surrounding the tissue-containing device. 21 In their native biological niche, insulin-secreting primary islets are supported by rich intraislet capillaries in the pancreas, allowing the islets to receive about 10% of the pancreatic blood supply despite these islets constituting only 1% of the mass of this organ. 21 Therefore, to achieve optimal cellular survival, an encapsulation system should also ideally facilitate the formation of new vascularization in close proximity to the encapsulated cells throughout the device while protecting them from host immune attack. To this end, the incorporation of pro-angiogenic agents has been explored to promote neovascularization of transplanted therapeutic cells, [21][22][23] albeit without sufficient spatial guidance to direct the growth pattern of these new blood vessels. As such, the resultant neo-vasculature could be disorganized, potentially leading to unorderly blood flow and inadequate supply to the therapeutic cells. 24 Alternatively, a device embedded with a vascular network for the delivery of hepatocyte aggregates was fabricated by photopolymerization of gelatin methacryloyl (GelMA) using a customized stereolithography apparatus for tissue engineering (SLATE). 25 Although the transplanted cellular graft survived and remained functional 14 days post-transplantation in mice, the vascular layer in the device was positioned underneath the tissue-encapsulating layer, potentially impairing diffusion of oxygen and nutrients to the cellular aggregates located at the uppermost plane away from the vascular network. 25 To address the challenge of microtissue aggregation associated with the requirement for high loading of therapeutic cells, we developed a hydrogel-based macro-encapsulation device with micropatterned features directing spatially homogeneous distribution of therapeutic microtissues. We also investigated the feasibility of concurrently incorporating vascular inductive cells for potential establishment of a co-planar, intra-device vascular network in close proximity with therapeutic microtissues. The viability and functional characteristics of both therapeutic and vascular-inductive cell types incorporated in this device were evaluated in vitro with immortalized cell lines. This device, which contained primary rat islets and was reinforced with an additional alginate coating, was also evaluated for its ability to restore glycemic control in chemically induced immunocompetent diabetic mice.

| Synthesis of GelMA
An amount of 7 g of gelatin (Sigma-Aldrich) was dissolved by heating in 70 mL of Dulbecco's Phosphate-Buffered Saline (DPBS) at 50 C. Afterwards, 7 mL of methacrylic anhydride (Sigma-Aldrich) was added dropwise. After 3 h of reaction, 200 mL of DPBS (Gibco Laboratories) was added to dilute the mixture, and the whole solution was stirred for another 15 min at 50 C. Subsequently, the solution was dialyzed against deionized water in a dialysis tube (Shanghai Rebus Network Technology Co., Ltd.) with a cut-off molecular weight of 12-14 kDa for 1 week at 37 C. The deionized water was replaced every 2 days.
Lastly, solid GelMA was obtained by lyophilization of the dialyzed solution and subsequent storage at room temperature. were cultured in RPMI-1640 medium (Hyclone) supplemented with 10% heat-inactivated FBS (Gibco Laboratories), 10 mM HEPES, 1 mM sodium pyruvate, 2 mM L-glutamine, 100 U mL À1 penicillin, 100 μg mL À1 streptomycin and 50 μM 2-mercaptoethanol. Cells were kept in a standard incubator with a 37 C humidified atmosphere of 95% air and 5% CO 2 . Cells were detached at a confluency of $80% after being treated with 0.25% trypsin-EDTA (Gibco Laboratories) at 37 C for 1 min and replated with a density of $2 Â 10 4 cells cm À2 for subculture. Culture medium was replaced every 3 days.

Primary human umbilical vascular endothelial cells (HUVECs)
were purchased from Lonza Bioscience, Singapore. HUVECs at passages of 8-11 were used in this work. Cells were cultured in EGM™-2 Endothelial Cell Growth Medium-2 BulletKit™ (Lonza Bioscience Singapore) in 5% CO 2 at 37 C. At a confluency of $80%, cells were treated with 0.25% trypsin-EDTA at 37 C for 1.5 min and replated with a density of $1 Â 10 4 cells cm À2 for subculture. Culture medium was replaced every 2 days.

| Ex situ fabrication of islet-like microtissues on agarose micromolds
INS-1 E microtissues were fabricated using agarose molds as previously described. 26 First, solid Ultrapure© Agarose (Invitrogen) was autoclaved and dissolved in autoclaved deionized water at 150 C to form a solution with the concentration of 2.3% (w/v). The dissolved agarose solution was pipetted onto silicone templates purchased from 3D Petri dish ® (Sigma-Aldrich) and allowed to solidify at room temperature for 20-25 min. Subsequently, the solidified, micromolded agarose hydrogels were separated from the silicone template using a spatula before being transferred to a 6-well plate and equilibrated with 3 mL of culture medium overnight prior to cell seeding. Every micromolded agarose hydrogel contained a 16 Â 16 array of spheroid-fabricating recesses with each having a round bottom of 300-μm diameter and a depth of 800 μm. Before cell seeding, the culture medium in each agarose mold was removed by aspiration. Subsequently, 175 μL of INS-1E cells suspension at a density of 2.19 Â 10 6 cells mL À1 was seeded into each agarose mold and cells were allowed to sink into the recesses for 5 min. Afterwards, 3 mL of fresh culture medium was supplemented to each agarose mold. The seeded cells were allowed to assemble on the agarose mold for 48 hours at 37 C and 5% CO 2 . Afterwards, microtissues from 3 agarose molds were collected and subsequently rinsed 3 times with culture medium and kept in a 6-well plate prior to fabrication of each WIM device. The images of the collected microtissues were taken under an inverted microscope (Olympus CKX53) at 4Â magnification and analyzed using ImageJ.

| Fabrication of waffle-inspired GelMA "lock" component with or without HUVECs
GelMA was patterned to form a waffle-inspired hydrogel network interspersing square or circular microwells of defined dimension by photolithography ( Figure S1). Briefly, 50 μL of 10% (w/v) GelMA prepolymer solution with 0.5% (w/v) I2959 photoinitiator (Sigma-Aldrich) was deposited onto the lid of a petri dish. Afterwards, this GelMA prepolymer solution was sandwiched in the gap between the lid and a glass slide (Biomedia, Singapore) placed on top of two stacks of spacers each comprising two pieces of coverslips with a total thickness of 150 μm. Subsequently, a photomask with desired micropatterns was placed on the top of the glass slide. Upon exposure to UV light (360-480 nm; 7.9 mW cm À2 ) for 19-25 s, the UV-exposed portion of GelMA was crosslinked to form the micropatterned network of sidewall with a total area of 1 cm Â 1 cm. Afterwards, the glass slide bearing the UVcrosslinked GelMA pattern was rinsed with DPBS to remove the uncrosslinked GelMA residue that was not exposed to UV light, leaving behind microwells.
In this study, 6 groups of GelMA "lock" patterns were designed with circular and square microwells labeled as C-200, C-300, C-400, S-200, S-300, and S-400 with microwell arrays of 32 Â 32, 23 Â 23, 18 Â 18, 32 Â 32, 23 Â 23, and 18 Â 18, respectively. The label "C" and "S" denoted circular and square microwells, respectively, and the number denoted either the diameter of the circular microwell or the side length of the square microwell. For example, C-200 is the GelMA "lock" component with circular microwells of diameter 200 μm. The width of the GelMA sidewall separating adjacent microwells for each pattern was designed to be 50 μm.
The bright field images of patterns were taken under an inverted microscope (Olympus CKX53), and the fluorescent images of patterns containing FITC-Dextran (Sigma-Aldrich) were taken under an inverted fluorescence microscope (Olympus IX71). The FITC-Dextran was mixed into GelMA prepolymer solution before the mixture was pipetted onto the petri dish.
For fabrication of each HUVEC-encapsulating GelMA "lock" component, HUVECs were mixed with GelMA prepolymer solution at a density of 2.5 Â 10 7 cells mL À1 before this mixture was dispensed on the lid of the petri dish prior to UV exposure. The HUVECs-laden GelMA "lock" component was cultured for 7 days before incorporation with of the alginate key encapsulating INS-1E microtissues. imaged under an inverted microscope (Olympus CKX53). In a parallel experiment, a control device was also fabricated using a similar approach as that used for WIM device. First, a square GelMA frame without the central microwell array was formed on a glass slide using photolithography. Afterward, a mixture of approximately 720-750 microtissues and F I G U R E 1 Schematic illustration of the procedure to fabricate Waffle-inspired Interlocking Macro-encapsulation device (WIM device). Photolithography was employed to pattern the GelMA "lock" component, followed by addition of a mixture of microtissues and alginate solution, which was subsequently crosslinked to form the alginate "key" component. 35 μL of 1% (w/v) alginate solution was pipetted onto the glass slide within the interior of the GelMA frame followed by addition of BaCl 2 solution for alginate gelation.
To fabricate WIM device co-laden with HUVECs and microtissues, HUVECs were mixed with GelMA prepolymer solution at a density of 2.5 Â 10 7 cells mL À1 before this mixture was dispensed on the lid of the petri dish prior to UV exposure. The HUVECs-laden GelMA "lock" component was cultured for 7 days before incorporation with the alginate "key" containing INS-1E microtissues. Post-encapsulation live/dead staining, glucose-stimulated glucose secretion (GSIS) and CD31 immunofluorescent staining were performed after an overnight incubation of 18 h. Islets were purified by gradient centrifugation using 1100 g/mL Histopaque, prepared with the mixing of 1077 and 1119 g/mL Histopaque (Sigma), followed by six times of sedimentation. Islets were handpicked and cultured in RPMI-1640 (Hyclone) supplemented with 10% heat-inactivated FBS (Gibco Laboratories) and 1% penicillin/ streptomycin (Gibco Laboratories). Area of islets was measured by ImageJ and effective islet diameter was calculated as the diameter of a circle with the same area. 27 Based on calculated diameter, islets were converted into islet equivalent (IEQ) using conversion factors as previously described. 28 Islet circularity was calculated using an ImageJ function based on the formula Circularity = 4π Â Area Perimeter 2 , 29 where area and perimeters were automatically determined by the same function.

| Fabrication of alginate-coated WIM device encapsulating primary rat islets
To fabricate each islet-encapsulating WIM device following the procedure illustrated in Figure 1, 250 IEQs of primary rat islets were suspended in 30 μL of 1.5% (w/v) SLG20 alginate and dispensed onto S-300 GelMA "lock" component and followed by alginate crosslinking with a 20 mM BaCl 2 solution. The whole device was further coated with another layer of alginate ( Figure S3). Specifically, the WIM device was detached from the glass slide in Figure S1 and placed on top of a droplet of 200 μL of 2% (w/v) SLG20 alginate which was previously dispensed on another glass slide. The WIM device was then flipped over to expose the other side of the device to the remaining portion of the dispensed  Live/dead fluorescent staining was performed at 24 h post-seeding.

| Visualizing the interlocking of device components
The interlocking of the GelMA "lock" component and alginate "key" component was examined by fluorescent labelling of different components of the WIM device. A suspension of red or green fluorescent polystyrene microbeads (Thermo Fisher) with the diameters of 1 μm were mixed at a concentration of 0.005% (v/v) in the alginate "key" and GelMA "lock" component respectively, while the microtissues were stained with Hoechst dye. Fluorescent images were taken with confocal microscopy (ZEISS LSM 800, Carl Zeiss) and analyzed using ImageJ.

| Static GSIS of microtissues encapsulated in WIM device
After an overnight incubation of 18 h following device fabrication, the microtissue-encapsulating WIM device in the presence or absence of HUVECs was transferred into a 12-well plate, rinsed three times with 2 mL of 2.8 mM glucose-contained RPMI-1640 complete medium, and then pre-incubated with 1 mL of 2.8 mM glucose-contained RPMI-1640 complete medium at 37 C for 1.5 h. Subsequently, the sample was rinsed three times with 2 mL of glucose-free Krebs Ringer Buffer Hepes (KRBH; 0.5 mM NaH 2 PO 4 , 0.5 mM MgCl 2 , 135 mM NaCl, 5 mM NaHCO 3 , 3.6 mM KCl, 1.5 mM CaCl 2 , 10 mM HEPES and 0.1% (w/v) BSA, pH 7.4). Afterwards, the sample was consecutively subjected to 1 mL of 2.8 mM glucose-contained KRBH for 1.5 h at 37 C, 1 mL of 16.8 mM glucose-contained KRBH for 1.5 h at 37 C and 1 mL of 2.8 mM glucose-contained KRBH for 1.5 h at 37 C. In between two consecutive incubations, the sample was rinsed three times with glucose-free KRBH buffer. A volume of 100 μL of the supernatant was aspirated at the end of each incubation and stored at À20 C. Lastly, the insulin concentration in each sample was quantified using Ultrasensitive Insulin ELISA kit (ALPCO Diagnostics).

| Statistical analysis
All values were averaged and expressed as the mean ± SD. Comparison of insulin secretion following exposure to different glucose concentrations was performed using the one-way repeated measures ANOVA analysis with Tukey's post hoc test. Comparison of microtissue distribution in C-200, C-300, C-400, S-200, S-300, S-400 WIM devices was performed using one-way ANOVA with Tukey's post hoc test. A p-value of less than 0.05 was considered statistically significant.

| Structural design of micropatterned macroencapsulation device
We developed a dedicated platform to encapsulate therapeutic microtissues with controlled spatial distribution while concurrently supporting an organized intra-device network of vascular-inductive cells.
Termed Waffle-inspired Interlocking Macro-encapsulation (WIM) device, this platform comprised two modules with complementary topography features that fitted together in a lock-and-key configuration ( Figure 1). Based on a design inspired by the grid-like pattern on the surface of a waffle, the "lock" component consisted of an interconnected vascular-supporting hydrogel network, which served as dividing sidewalls separating evenly spaced microwells. The "key" component consisted of therapeutic microtissues encapsulated in a semi-permeable alginate hydrogel that fitted into the microwells of the "lock" component. We postulated that the waffle-inspired micropattern of the "lock" component could be designed to direct spatially homogenous distribution of therapeutic microtissues such that each microtissue can be ideally entrapped in one microwell to prevent undesirable aggregation. Furthermore, the interlocking design of our WIM device facilitates the proximity of the vascular-inductive cells in the "lock" component and the encapsulated therapeutic microtissues in the "key" component. Figure 1 illustrated a typical procedure to fabricate the WIM device. To fabricate the "lock" component, photolithography was leveraged to design a network of photo-crosslinked GelMA with desired geometric pattern (Figures 1 and S1) and containing HUVECs, which serve as precursors for potential formation of new blood vessels. An array of uniform empty microwells was left behind after uncrosslinked region of GelMA was removed. This microfabrication technique has been commonly employed to produce hydrogel microstructures with well-defined geometry and controlled resolution within the sub-micron to millimeter range. 30,31 GelMA was selected as the vascular-supporting hydrogel due to its ease of photopolymerization and ability to support HUVEC proliferation and organization into tubular structures. 32,33 Subsequently, the "key" component was fabri- 3.2 | Optimizing design parameters of GelMA "lock" component for efficient entrapment and homogenous spatial distribution of microtissues To the optimize WIM device for efficient entrapment and homogenous distribution of islet-like microtissues, we fabricated several designs of its GelMA "lock" component by varying its micropattern to yield microwells with different shapes and sizes ( Figure S4). Square and circular microwell shapes were chosen due to their simple geometries, which facilitated their microfabrication with high fidelity from photomask designs. Microwell dimensions, specifically the side-length of square microwells and the diameter of circular microwells, at 200, 300, and 400 μm were evaluated. The lower limit of microwell dimension was specified as 200 μm to sufficiently accommodate potential therapeutic microtissues, such as rodent and human islets, the mean diameters of which range from 100 to 150 μm. 36 The upper limit of microwell size was set at 400 μm to ensure a distance of at most 200 μm between cells in the microwell and the vascular inductive cells in the surrounding GelMA sidewalls. This maximum distance of 200 μm has been postulated as the critical diffusion limit between cells and surrounding blood vessels to ensure adequate supply of oxygen, nutrients, and metabolites. 37  For each "lock" design, we quantitatively evaluated its microtissue entrapment efficiency, which is defined as the ratio of the total number of microtissues each effectively entrapped within the interior of a microwell to the total number of microtissues encapsulated in the entire device. A higher microtissue entrapment efficiency was more desirable as fewer microtissues settled on top of the GelMA sidewall or outer edges of the device. Figure 2c shows that for the same device size (1 Â 1 cm), a larger total interior area of microwells per device type generally correlated with a higher microtissue entrapment efficiency. For the same microwell shape of either circular geometry (C-200, C-300, and C-400) or square geometry (S-200, S-300, and S-400), the larger microwell size correlated with a higher entrapment efficiency. In addition, square microwell design resulted in a higher entrapment efficiency of the "lock" component compared to that of  where n is the number of microtissue(s) for a particular microwell and ranges from 1 to 6, to the total number of microtissues encapsulated in the entire WIM device. Since a spatial distribution with one microtissue entrapped in each microwell is considered ideal, a high microtissue distribution ratio for microwells of type M 1 is considered optimal.
About 18%-25% of the microwells in each device remained unoccupied ( Figure S6) while the remaining microwells in the same device receive a range of 1 to 6 microtissues in each microwell. Among the three WIM devices evaluated (Figure 2d), S-300 WIM device presented a higher microtissue distribution ratio for microwells of type M 1 compared to the other two devices so this design was chosen for further experiments.

| Characterization of structural components of WIM device
To examine the structural interlocking between the GelMA "lock"   (Figure 4a(iii)), further confirming negligible cell death observed for these encapsulated microtissues.

| Evaluation of cellular viability and function of encapsulated microtissues
A successful encapsulation strategy for therapeutic cells allows not only the preservation of cellular viability but also their desired function. Therapeutic cells after encapsulation must still secrete insulin in an appropriate amount according to physiological glucose levels, because both excessive or inadequate insulin secretion can result in abnormal glycemia in diabetic recipients. 39 Microtissue-encapsulating S-300 WIM device was subjected to the static GSIS assay to evaluate the insulin-secreting function of encapsulated INS-1E microtissues. A sequential incubation of the device at 2.8, 16.8, and 2.8 mM glucose levels simulated the physiological condition in diabetic patients, alternating from basal condition to hyperglycemia and back to basal condition respectively. As shown in Figure 4b and Figure S8, the amount of insulin secreted by the encapsulated microtissues increased 6.6-fold when they were exposed to the higher level of glucose in the second incubation compared to the first incubation at a lower glucose level.
This increase was comparable to the outcome from published studies investigating the responsive insulin secretion of INS-1E spheroids. 40 During third incubation, the microtissues exhibited a return to basal insulin secretion, indicating their preserved function and responsiveness.  Figure S2). Nonetheless, primary rat islets were more heterogeneous in shape as measured by their circularity ( Figure S9). This parameter, defined as the ratio of area to squared perimeter of an islet, was used to quantify the resemblance of the two-dimensional (2D) projection of islet geometry to an ideal circle. 36 In our experiment, primary rat islets were of significantly lower level of circularity (0.76) in contrast to that of uniform INS-1E microtissues (0.85) ( Figure S9), hence probably rendering it more difficult to achieve the same level of homogeneous spatial distribution for the islets in the S-300 WIM device.

| Preliminary in vivo evaluation of S-300 WIM device function containing primary rat islets in chemically induced diabetic mice
After gelation of the alginate "key," the device was further coated ( Figure S3) in another layer of alginate that covered the whole device to increase its mechanical strength ( Figure S10) for subcutaneous transplantation. The alginate-coated S-300 WIM device preserved the interlocking feature ( Figure 5D

| Assembly of monodispersed cells for in situ formation of microtissues with homogenous spatial distribution
Islet-like microtissues or pseudo-islets, which are engineered microtissues assembled in situ or "on device" from mono-dispersed primary islet-derived cells, might be used in place of irregularly shaped primary rat islets. 41,42 In situ fabrication of microtissues comprises the initial step of microtissue assembly on one platform and the subsequent step of encapsulating microtissues on the same platform with semipermeable hydrogel to form the final macroencapsulation device. 20 In contrast, ex situ or "off device" fabrication of microtissues involve assembly of microtissues on another platform prior to transferring them onto a second platform to be encapsulated in semi-permeable hydrogel. 26 Hence, in situ microtissue fabrication eliminates the risk of cell mass loss associated with microtissue transfer and replating between the two platforms used in ex situ fabrication approach. Therefore, we evaluated the feasibility of fabricating the S-300 WIM device using an in situ approach for fabrication of islet-like microtissues on GelMA "lock" component ( Figure S11). GelMA "lock" did not aggregate (Figure 6a(i)). Confocal microscopy after fluorescent live/dead staining of the microtissues confirmed that they were highly viable and located completely within the interior of the microwells in a uniform array, as shown by their projected confocal images (Figure 6a(iii-v)). However, the microtissues were irregularly shaped and each exhibited a pointed protrusion towards the corners of the microwells. We postulate that this cell assembly behavior might be due to the increased attachment of cells at the peripheral of microtissues to the stiffer substrate 43 associated with exposure to a higher UV intensity 44  to prevent microtissue aggregation and potentially enhance nutrient access for the embedded microtissues.

| Incorporation of spatially guided vascularinductive cells into GelMA "lock" component
The WIM device could be further enhanced by incorporating neovascularization which might facilitate passive transportation of nutrients and oxygen to the encapsulated microtissues. Specifically, we postulate that the micropatterned waffle-inspired "lock" of GelMA, which is a substrate suitable for cellular encapsulation and proliferation, 45 could be loaded with vascular-inductive cells such as HUVECs to support their development into new vasculature. Therefore, we examined the feasibility of embedding primary HUVECs in the GelMA sidewalls of the S-300 "lock" component prior to encapsulating therapeutic microtissues and evaluated the viability of these loaded HUVECs using fluorescent live/dead staining assay followed by confocal microscopy ( Figure S12a).
HUVECs were mixed with the GelMA prepolymer solution prior to the fabrication of the S-300 "lock" component ( Figure S1) following in vitro culture for 7 days, embedded HUVECs exhibited good cellular viability and pro-angiogenic function as indicated by CD31 expression within the GelMA "lock" component ( Figure S12b). This observation suggested that the waffle-inspired interconnected micropattern of the GelMA "lock" component has the potential to provide controlled and uniform spatial guidance for the subsequent growth of an organized vascularized system throughout the entire device to ultimately reach individual therapeutic microtissues entrapped in the microwells. Specifically, HUVECs embedded in S-300 "lock" components ( Figure S13) fabricated using 5% and 7.5% (w/v) GelMA formed vessel-like structures after 2 days of in vitro culture while endothelial sprouting was also observed for HUVECs embedded in S-300 "lock" component fabricated from 10% (w/v) GelMA. However, this strategy of co-loading an additional vascular inductive cell type adjacent to the therapeutic cell type within the same encapsulation device also raises arguable concern for potential competition for oxygen or adverse cellular interaction between these two cell types. Therefore, we subsequently investigated the feasibility of co-loading both vascular-inductive HUVECs and insulin-secreting INS-1E microtissues in the same S-300 WIM device by evaluating cellular viability and preservation of intended function for each cell type to address such concern (Figure 7).   We further demonstrated that transplanting the alginate-coated S300 WIM devices containing primary rat islets at a dose of 500 IEQs per mouse resulted in glycemic correction in chemically induced, immuno-competent diabetic mice for up to 2 weeks. This data ( Figure 5) was sufficient to support our conservative claim that the dimension of the alginate-coated S-300 WIM device allowed sufficient release of insulin in vivo to restore blood glucose control with a relatively low dose of rat islets. Arguably, though the duration of glycemic correction in for the in vivo study might be considered relatively short, this data were sufficient to demonstrate that the alginate-coated WIM device has adequately shielded the transplanted rat islets from destruction by the immune system of diabetic mice during this period of time. Specifically, if the rat islets were not sufficiently protected from the immune system of immunocompetent diabetic mice, islet death would have occured within the first few days resulting in immediate loss of blood glucose control as observed in published studies. 48,49 Furthermore, it is important to highlight that the duration of glycemic correction depends on multiple factors. Specifically, comparing duration of glycemic control for different device designs requires standardization of therapeutic dosage, the partial oxygen pressure at the implant site and the presence of any enhancing factor such as oxygen-generating materials or supporting cells.

| Viability assessment of co-laden therapeutic and vascular-inductive cells
Published studies using a higher dose of rat islets or alternative implant sites with higher partial oxygen pressure P O2 such as the intraperitoneal space have resulted in treatment duration of months or longer. 35 However, most reported macrodevices that were implanted at the oxygen-poor subcutaneous site with a therapeutic dosage 500 IEQs of primary rat islets per immuno-competent mouse but without any enhancing factors either failed 50 or only achieved glycemic correction for up to 10 days compared to our reported 2-week duration. 51 Nonetheless, we acknowledge that the spatial distribution of primary rat islets within the transplanted device might not be ideally homogeneous due to the geometrical variation in shapes and sizes of these native microtissues (Figures 5b-d, S9). Specifically, the low circularity of primary rat islets (0.76) compared to INS-1E microtissues fabricated ex situ (0.85) ( Figure S9) which resulted in poorer spatial distribution observed in the WIM device ( Figure 5d). Furthermore, large islets are inherently susceptible to hypoxia and necrotic cores, 42 diminishing their survival in vivo and thus their long-term efficacy in diabetes correction.
To address the challenge of nonuniform geometry of therapeutic microtissues, multiple prior studies have previously demonstrated the feasibility of reaggregating monodispersed cells derived from large primary islets or stem cell-derived beta cell clusters 42,52,53 into smaller and more uniform microtissues with improved viability. 42,53 In this study, we demonstrated that "on device" assembly of monodispersed INS-1E cells on S-300 GelMA "lock" component ( Figure 6) resulted in in situ formation of microtissues with more uniform size and improved circularity ( Figure 6) compared to primary rat islets, thus facilitating more homogenous spatial distribution. Thus, the WIM device as an in situ platform, which enables both reaggregation of mono-dispersed insulin-secreting cells into uniform microtissues and their encapsulation in semi-permeable alginate (Figure 6), has the potential to achieve both superior cellular viability and optimal microtissue distribution homogeneity. This proof-of-concept demonstration with microtissues assembled from INS-1E cells ( Figure 6) lays a foundation for further studies of a next-generation macrodevice. We envision a future study to evaluate long-term in vivo efficacy for an improved device containing therapeutic microtissues, which are assembled in situ from a more clinically relevant cell source such as as beta-like clusters derived from human-induced pluripotent stem cells. We postulate that spatially homogeneous distribution of such therapeutic microtissues will result in differential cellular viability and functionality during assessment of long-term in vivo efficacy. Additional monitoring of in vivo insulin plasma level and ex vivo characterization of the retrieved macrodevice including histological analysis for immuno-compatibility and examination of microtissue morphology will also provide more comprehensive understanding of the device performance in preserving microtissue viability and performance.
In addition, prior publications have demonstrated that synthetic polymers such as poly (ethersulfone) (PES)/polyvinyl pyrrolidone (PVP) blend can be utilized to fabricate different microwell designs to facilitate spatial distribution of microtissues. 8,16 However, such strategy does not support the growth of an embedded vasculature network as pro-angiogenic endothelial cells cannot be incorporated into the dense walls of these hydrophobic polymers. Since GelMA was shown to support survival and function of endothelial cells, 32,33 we anticipated that the HUVEC-embedded GelMA matrix of the S-300 WIM device would mature into a penetrating vascular network coplanar with the encapsulated microtissues in future in vivo study.
Notably, other strategies for device vascularization have not achieved this co-planar spatial arrangement of both pro-angiogenic and therapeutic cell types, 54 Figure S13), even though a more mature capillary-like network could be formed from HUVECs embedded in S-300 GelMA "lock" components that were fabricated from lower concentrations of 5% and 7.5% (w/v) of hydrogel prepolymers, endothelial sprouting was also observed when a higher concentration of 10% (w/v) was used. This observation aligned with published studies demonstrating that sprouting of embedded HUVECs into 3D capillary network is facilitated by decreased stiffness of GelMA hydrogels formed from lower prepolymer concentrations. 55 In this study, the S-300 GelMA "lock" component was fabricated with the higher prepolymer concentration of 10% (w/v) to provide sufficient stiffness and thus mechanical strength for device handling. Nonetheless, we speculate that, upon device transplantation, in vivo degradation of GelMA might eventually decrease its stiffness to support maturation of endothelial sprouting into formation of capillary network. 54 Previous studies demonstrated that crosslinked GelMA hydrogel provided favorable environment for cellular proliferation and spreading due to its RGD motifs which were responsive to matrix metalloproteinases and supportive of cell attachment. 45 Nikkhah et al. also reported vascularization of linear GelMA microstructures with embedded HUVECs. 56 Furthermore, coculture of human vascular cells and human mesenchymal stem cells in GelMA hydrogel generated lumen-containing vascular network in vitro. 57 After its implantation into immunodeficient mice, this cell-laden GelMA hydrogel construct resulted in anastomosis between mouse vasculature and in vitro cultured human vascular network. 57 Together, these literature evidences corroborate the potential application of our WIM device design as a promising strategy to create an interconnected network of neo-vasculature interspersing therapeutic microtissues to enhance their performance.
The mechanical integrity of the WIM device is another aspect requiring further investigation. In this study, the S-300 WIM device without an additional alginate coating was sufficiently robust to be handled by a pair of tweezers for static in vitro evaluation with confocal microscopy or GSIS assay. Incorporation of the GelMA "lock" component with waffle-inspired micropattern improved the structural integrity of the uncoated WIM device (Figure 3bi) compared to a control device (Figure 3bii), which comprised microtissues encapsulated in a monolithic alginate block without the GelMA "lock." As shown in (Figure 3bii), such control device wrinkled and did not maintain its intended square shape due to the absence of a GelMA network, which could have otherwise acted as a backbone framework to prevent the hydrogel from wrinkling. Nonetheless, in a pilot in vitro experiment during which the uncoated S-300 WIM device was subjected to shaking motion to mimic potential mechanical challenge of the in vivo condition such as animal scratching behavior or body movement, we observed detachment of the two device components. Therefore, we designed a modified protocol to add an additional alginate coating to mechanically strengthen the device and preserve its structural integrity upon in vivo transplantation. For future development of next-generation devices, we postulate that the mechanical strength of the WIM device might be improved by incorporation of mechanically robust materials into the "lock" component of this design. For example, electrospun nanofibers might be incorporated to the GelMA "lock" component by mixing these fibers with GelMA prepolymers prior to UV crosslinking to enhance the strength of the micropatterned hydrogel. Alternatively, free-standing nanofibrous electrospun membrane formed by conformal deposition on an alloy template can also be employed as a robust structural framework to support the GelMA "lock" component. 58 Furthermore, additive manufacturing of more mechanically robust polymers by melt electrowriting or extrusion printing might also be utilized to print the waffleinspired design of the "lock" component. 59,60 Nonetheless, it would generally be more challenging to achieve the same resolution of the sidewall dimension by these 3D printing techniques instead of the lithography-based patterning of GelMA as demonstrated in this study.
Lastly, delivery of a clinically relevant dosage of therapeutic cells while maintaining the spatially homogeneous distribution of the encapsulated microtissues remain an unsolved challenge for our current WIM device. Nonetheless, this scalability issue is currently a universal challenge for most reported macro-encapsulation systems which also require large device size for translation of preclinical success into clinically relevant applications. 61 This study focused only on demonstrating the proof-of-concept that a waffle-inspired "lock" component could distribute microtissues evenly in a 2D arrangement. Similar approach might also be adopted to design a next-generation device to homogeneously distribute the microtissues in a 3D arrangement by stacking multiple layers each comprising co-planar therapeutic microtissues and endothelial cells. Such design might increase microtissue packing density thus reducing device area required while minimizing microtissue fusion for improved cellular viability.

| CONCLUSION
We fabricated a macrodevice to position encapsulated therapeutic microtissues in spatially homogeneous distribution while preserving their cellular viability and insulin-secretion function. The waffle-