Engineering Modular, Oxygen-Generating Microbeads for the in situ Mitigation of Cellular Hypoxia

Abstract

Insufficient oxygenation is a key obstacle in the design of clinically scalable tissue-engineered grafts. In this work, an oxygen-generating composite material, termed OxySite, was created through the encapsulation of calcium peroxide (CaO₂) within polydimethylsiloxane and formulated into microbeads for ease in tissue integration. Key material parameters of reactant loading, porogen addition, microbead size, and an outer rate-limiting layer were modulated to characterize oxygen generation kinetics and their suitability for cellular applications. In silico models were developed to predict the local impact of different OxySite microbead formulations on oxygen availability within an idealized cellular implant. Promising OxySite microbead variants were subsequently co-encapsulated with murine β-cells within macroencapsulation devices, resulting in improved cellular metabolic activity and function under hypoxic conditions when compared to controls. Additionally, the co-injection of optimized OxySite microbeads with murine pancreatic islets within a confined transplant site demonstrated ease of integration and improved primary cell function. These works highlight the broad translatability delivered by this new oxygen-generating biomaterial format, whereby the modularity of the material provides customization of the oxygen source to the specific needs of the cellular implant.

Graphical Abstract

An oxygen generating biomaterial with customizable oxygen kinetics, termed OxySite, is fabricated into a microbead format. Via discreet manipulation of material parameters, including reactant (CaO₂) dosage, porogen inclusion, modulation of microbead size, and the addition of an outer, rate-limiting PDMS layer, resulting OxySite microbeads are capable of delivering oxygen to support the nutritional cellular needs of different devices and tissues.

1. Introduction

Oxygen is a crucial parameter in the design of biologically-active and clinically-scalable engineered tissues^[1]. Endogenously, oxygen is provided to three-dimensional tissues and organs through a dense network of vessels and capillaries that deliver oxygenated blood^[2]. This efficient and comprehensive perfusion prevents cellular hypoxia. When implanting avascular tissue engineered devices, however, intra-device vascularization can take weeks^[3]. As a result, most cell-containing implants larger than only a few hundred microns exhibit substantial hypoxia-induced cell loss during the early engraftment period, resulting in ineffective treatments and elevated immunological responses to the dying tissue^[4–6]. Multiple strategies have been employed in an attempt to accelerate intra-device vascularization, from the local release of pro-angiogenic therapeutics that accelerate the engraftment process to pre-vascularization approaches that “prime” the site for transplantation; however, shortening angiogenesis fails to eliminate the graft loss that occurs immediately (12–72 hours) post-implantation^[7–9]. In addition, surviving cells exposed to hypoxic stress may experience prolonged functional impairment, reducing their therapeutic benefit^[10].

One promising solution to “bridge the gap” between transplantation and engraftment is the use of oxygen-supplementing systems that deliver in situ oxygen to the transplanted cells^[4,11–14]. Oxygen-generating biomaterials, created through material encapsulation of hydrolytically reactive chemicals, are a promising approach. Commonly used reactive components include hydrogen peroxide (H₂O₂)^[15], sodium percarbonate (Na₂CO₃·1.5H₂O₂)^[16,17], and solid peroxides, such as calcium peroxide (CaO₂) and magnesium peroxide (MgO₂)^{[11,12,18,19]}. CaO₂, a favorable option due to its high oxygen release potential and thermal stability, decomposes to release oxygen in contact with water via the following reactions^[4,20]:

$$2{CaO}_2+{2H}_{2}O\to {2Ca\left(OH\right)}_2+O_2$$ Equation 1

$$2{CaO}_2+{4H}_{2}O\to {2Ca\left(OH\right)}_2+{2H}_2{O}_2\to {2Ca\left(OH\right)}_2+{2H}_{2}O+O_2$$ Equation 2

The encapsulation of CaO₂ within a polymer serves to control oxygen generation and temper reaction kinetics to avoid hyperoxide conditions and elevated levels of toxic reaction byproducts, such as reactive oxygen species (ROS) and hydroxides. In our laboratory, we have used polydimethylsiloxane (PDMS) to encapsulate CaO₂, creating a controlled oxygen-generating biomaterial termed OxySite (Figure 1A)^[4,11,12]. PDMS was selected due to its hydrophobicity, which serves to control water infiltration, biostability, which inhibits undesired solid peroxide release, and high oxygen permeability, which ensures the efficient release of oxygen from the material. The addition of OxySite cylindrical slabs into cell cultures and implants resulted in significant protection against hypoxia-mediated cell death and dysfunction, demonstrating the benefits of this in situ oxygenation material^[11,12].

Figure 1.

Schematic concept of oxygen generating microbead reactivity, modulation, and predicted kinetic release. a) Schematic depicting the reactivity of the OxySite microbeads, whereby encapsulated CaO₂ within the composite OxySite material hydrolytically reacts to generate in situ oxygen. b) Summary of modifications to the OxySite microbeads material and c) the hypothesized impact of these changes on global oxygen release rates (OGR) over time.

To elevate the modularity of the OxySite platform, a microbead geometry was explored. This geometric format has the benefits of ease of integration into implantable devices and direct tissue injection; however, altering the material geometry and length scales can dramatically affect controlled oxygen generation. Specifically, the elevated surface-to-volume ratio and smaller scale of a microbead can lead to burst kinetics and accelerated expiration of the oxygen source. Customization of peroxide reactivity and oxygen release is feasible by leveraging material features, as demonstrated for classic PDMS-based drug-delivery systems^[21–25]. Thus, to provide the discreet oxygen kinetic control needed to ensure optimal cellular support, the impact of four encapsulating material features on CaO₂ reactivity within PDMS was interrogated: 1) CaO₂ loading; 2) porogen inclusion; 3) microbead size; and 4) the addition of an outer pure PDMS layer (Figure 1B). It was anticipated that modulation of the OxySite microbead features would alter the overall oxygen generating rate (OGR) (Figure 1C), as well as the generation of reactant byproducts.

To characterize the role of these material features on OxySite microbead performance, global material characterization, oxygen generation kinetics, and byproduct assessment studies were performed. Oxygen kinetic results were integrated into finite element modeling (FEM) scenarios to illustrate the modularity of the different OxySite variants on local oxygen gradients and its potential benefits for avascular, bulk tissues. Promising prototypes were translated to two tissue engineered platforms to determine their therapeutic benefit. In one application, specific OxySite microbead variants were dispersed alongside pancreatic β-cells within a thick, avascular hydrogel. In another approach, smaller OxySite microbead variants were co-injected with primary pancreatic islets into a confined transplant site. Results illustrate the benefits of these distinct OxySite microbead designs for supporting cellular grafts. This work provides clear material parameters for the customization of oxygen-generating materials for a unique microenvironment, as well as demonstrates the newfound potency of the platform to support highly metabolically active cells under hypoxia.

2. Results

2.1. Fabrication of OxySite Microbeads Variants

OxySite microbeads were fabricated by emulsifying uncured PDMS silicone containing the reactant CaO₂, with or without NaCl porogen, within a stirred bath. To evaluate alterations in the composite PDMS formulations, nanoCT imaging of the resulting OxySite microbead variants was performed (Figure 2A). Control PDMS-only microbeads showed a low intensity, homogenous CT signal (Figure 2A ii), while images of baseline OxySite microbeads (25% CaO₂, 0% Porosity) exhibited high-density particles present throughout the microbead structure (Figure 2A iii). The addition of porogens within the microbeads (22% CaO₂, 13% Porosity) resulted in a visual increase in total encapsulated particulates (Figure 2A iv), while increased CaO₂ loading (50% CaO₂, 0% Porosity) created a qualitative increase in larger particulates within the microbead structure. (Figure 2A v).

Figure 2.

Fabrication of base OxySite microbead modifications and material validation. a) Schematic of microbead emulsion method (i) used to fabricate four key microbead material prototypes. Schematic (top row) and nano-CT images (bottom row; scale bar = 100 μm) of ii) control pure PDMS; iii) standard OxySite (20% CaO₂); iv) porogen OxySite (22% CaO₂; 13% NaCl); and v) high CaO₂ loading (50% CaO₂) microbeads. b-d) Coulter counter particle analysis of microbead diameter for b) “Large”, 300–600 μm microbeads and c) “Small”, 106–212 μm microbeads. d) Summary of average microbeads size for both groups. Predicted average size is shown as a dotted line.

In addition to the material modifications, two distinct microbead size ranges were created: 300–600 μm and 106–212 μm, with the latter size being more favorable to injectability. Bead ranges were created using selected sieves. Coulter counter particle analysis was used to validate size ranges. Beads sorted to within the 300–600 μm size range exhibited a mean size of 475.8 ± 176.8 μm (Figure 2B&D); the mode of the microbeads population was 429.2 μm, skewed to the right, and was leptokurtic. Beads fabricated within the 106–212 μm size range showed a mean size slightly larger than predicted at 202.2 ± 65.70 μm (Figure 2C&D); the mode of the microbeads population was 203.5 μm, skewed to the right, and was leptokurtic.

2.2. Quantification of OxySite Microbead Short-term Oxygen Generation and Byproducts

To determine the impact of material modifications on microbead reactivity, oxygen generation kinetics and byproducts of CaO₂ hydration were measured. As both H₂O₂ and alkaline pH can result in dysfunction and cytotoxicity in sensitive cell populations, quantified byproducts were classified based on their expected transient effect^[26–28]. H₂O₂ concentrations below 50 μM (green) were considered safe for transient exposure under buffered physiological conditions (e.g., cell culture, in vivo transplant)^[28,29]. The elevation of H₂O₂ levels to the range of 50 – 100 μM (yellow) may be a significant cytotoxic risk for sensitive populations (e.g., highly ROS-sensitive cells), but should be suitable for bulk tissue applications^[30]. Finally, concentrations > 100 μM (red) present a significant cytotoxic risk to bulk tissue, requiring intervention to mediate impact^[30]. Additionally, pH levels greater than 7.6 were considered not ideal based on the homeostatic pH range, although cultured cells support a broader range (i.e., 7–7.87) with select cells, such as fibroblasts, preferring basic conditions (7.4–7.7)^[31–33].

As expected, increasing the overall CaO₂ loading within the microbead resulted in an increase in daily oxygen generation rates (OGRs) (Figure 3A). For example, day 2 OGR values for the baseline 25% CaO₂ OxySite microbeads were 1.85 × 10⁻⁴ mmol of O₂/day, while increasing CaO₂ loading to 35% (“Mid”) or 50% (“High”) resulted in 4.9- and 14.3-fold higher OGR, respectively (P < 0.0001 and 0.006 compared to base OB). Increasing CaO₂ density within the microbeads also significantly altered the kinetic curves (P = 0.008; based on the statistical interaction between CaO₂ loading and time). Overall, increasing peroxide loading positively correlated with cumulative oxygen release over the one-week study period (Figure 3B; P < 0.001 one-way ANOVA). Quantification of H₂O₂ byproduct generation showed non-significant changes in concentration on Day 1, with all groups averaging in the 50 – 64 μM range (Figure 3C). By Day 3, the H₂O₂ concentration for both the “Low” and “Mid” CaO₂ groups were significantly lower compared to the “High” CaO₂ loading (P = 0.013 and 0.04, respectively); this trend continued to Day 7 (P = 0.029 and 0.046, respectively). Assessing changes in pH under buffered conditions (Figure 3D) found the highest CaO₂ loading resulted in a significant elevation in pH over the entire study period (P < 0.0001), with values ranging between 7.83 and 7.97. Comparing the baseline OxySite and “Mid” CaO₂ loaded microbeads, significant changes in pH at all time points were measured, with the most significant being at Day 7 (P < 0.0001).

Figure 3.

Summary of oxygen generation kinetics and byproduct quantification for OxySite microbeads formulations with varying CaO₂ loading and microporosity. The impact of CaO₂ loading on a) daily oxygen generation rates (OGRs), b) overall oxygen generated (mmol) during the study period, and c-d) temporal release of byproducts H₂O₂ and pH for OxySite microbeads loaded with 25% (OB); 35% (Mid), or 50% (High) CaO₂. The impact of porogen loading on e) daily oxygen generation rates (OGRs), f) overall oxygen generated (mmol) during the study period, and g-h) temporal release of byproducts H₂O₂ and pH for OxySite microbeads for OxySite microbeads loaded with 0% (OB); 7% (Mid), or 13% (High) NaCl. N = 3–9 samples for each group; data represented as mean ± SD; solid lines represent trends in data over time; two-way ANOVA with Tukey’s HSD multiple comparison test for daily mean comparisons. Statistics for a & d: * represents comparisons between OB and High groups, ρ represents comparisons between OB and Mid groups, and Φ represents comparisons between Mid and High groups; for b & g: * represents comparisons between OB and High groups, ρ represents comparisons between OB and low groups, and Φ represents comparisons between Low and High groups. Color bands in c & g represent H₂O₂ concentration ranges considered safe (green), potentially cytotoxicity (orange), or likely cytotoxic (red) support sensitive cell populations (green), bulk tissue (orange), and potential cytotoxicity (red).

The inclusion of sacrificial porogens is an established method for enhancing both release rates and duration for monolithic drug delivery systems. To explore the impacts of a porogen within the OxySite material, finely milled salt particles (~5 μm) were integrated into the CaO₂ and PDMS mixture prior to microbead fabrication. OxySite microbeads containing 0% salt porogen (“Non”) were compared to beads containing 7 (“Low”) or 13 (“High”) wt% NaCl. For both dosages, porogen inclusion resulted in significant changes in the OGR kinetic profiles, with increasing OGR over the study period (Figure 3E; P < 0.0001 two-way ANOVA). Porogen levels also positively correlated with overall oxygen generation (Figure 3F; P < 0.0001 one-way ANOVA). Contextualizing porogen inclusion with H₂O₂ release (Figure 3G), increasing porogen levels from 0 to 13% resulted in significantly higher H₂O₂ concentrations on day 1, with concentrations of 50.6, 102, and 159.9 μM H₂O₂ for NaCl loadings of 0, 7, and 13%, respectively (P = 0.016 & < 0.0001 compared to 0% controls). This burst release of H₂O₂, however, was resolved by day 3 with no significant difference between porogen and non-porogen beads and all values < 50 μM. While the pH significantly increased with porogen inclusion (Figure 3H), values remained under 7.45 throughout the measurement period.

With the vision of translational applications that require modulation in OxySite microbead size to accommodate inclusion within large, engineered implants or injected into tissues, the impacts of microbead size on oxygen kinetics was examined. Two size ranges were examined, a larger microbead, 300 – 600 μm diameter, that was more relevant to inclusion within macroscale devices, as well as a smaller microbead, 106 – 212 μm diameter, suitable for injectable applications. Examination of oxygen generation curves indicated a subtle impact from altering size, with only days 1 and 2 exhibiting a significant change between groups (P = 0.002 and P < 0.0001, respectively) (Figure 4A). Performing a two-way ANOVA showed that the interaction between size and time was not significant (P = 0.062). Comparing cumulative oxygen release, the smaller microbead variant resulted in a ~1.5-fold increase over the larger microbeads (P < 0.0001) (Figure 4B). Surprisingly, despite modest impacts on oxygen generation rates, size strongly altered byproduct production levels. Release of H₂O₂ from the smaller OxySite microbeads was within the toxicity range (163 μM) and significantly higher than that measured from the larger microbeads (P < 0.0001). While H₂O₂ release (Figure 4C) from the smaller microbeads precipitously declined by Day 3, levels were still significantly higher than their larger counterparts (P = 0.001). By day 7, both groups reached comparable concentrations. Measuring buffered pH (Figure 4D), smaller microbeads resulted in significantly higher pH at day 1 (P = 0.003), although both sizes stayed within the cell compatible range during the measurement period.

Figure 4.

Summary of oxygen generation kinetics and byproduct quantification for OxySite microbeads formulations with varying size without or with microporosity. The impact of microbead size on a) daily oxygen generation rates (OGRs), b) overall oxygen generated (mmol) during the study period, and c-d) temporal release of byproducts H₂O₂ and pH for small (106–212 μm diameter) or large (300–600 μm diameter) OxySite microbeads loaded with 25% CaO₂. The impact of size and porogen on e) daily oxygen generation rates (OGRs), f) overall oxygen generated (mmol) during the study period, and g-h) temporal release of byproducts H₂O₂ and pH for small (106–212 μm diameter) or large (300–600 μm diameter) OxySite microbeads for OxySite microbeads loaded with loaded with 22% CaO₂ and 7% NaCl. N = 3 samples for each group; data represented as mean ± SD; solid lines represent trends in data over time; two-way ANOVA with Tukey’s HSD multiple comparison test for daily mean comparisons. * represents comparisons between the two graphed groups. Color bands in c & g represent H₂O₂ concentration ranges considered safe (green), potentially cytotoxicity (orange), or likely cytotoxic (red) support sensitive cell populations (green), bulk tissue (orange), and potential cytotoxicity (red).

To interrogate the modularity of material features, the parameters of size and porogen inclusion were combined. The same “Small” and “Large” size ranges were examined, but 7% NaCl was included in the OxySite formulation. The inclusion of the porogen resulted in a significant alteration of the oxygen kinetic curves between the two sized microbeads (Figure 4E; P = 0.018). While the OGR for the two sizes were similar on day 1, the smaller, porogen-loaded microbeads released more oxygen per day than their larger counterpart on days 2 – 5 (P = 0.001 – 0.013). Globally, this resulted in almost a 2-fold increase in cumulative release rates for the smaller porogen-loaded microbeads, when compared to the larger porogen-loaded microbeads (P = 0.0003) (Figure 4F). Similar to the trends observed for non-porogen beads, the smaller, porogen-loaded microbeads exhibited higher H₂O₂ levels than their larger counterpart on day 1 (P = 0.026) (Figure 4G). After this burst release, H₂O₂ concentration was comparable between groups for the rest of the study period. The smaller porogen microbeads also significantly elevated pH when compared to the larger sized porogen beads (P < 0.0001), with pH values approaching levels of concern (7.5) on day 7 (Figure 4H).

2.3. Interactions between Microbead Material Modifications

To contextualize the material parameters that contributed to key global changes in OGR, as well as the interactions between the different material modifications, principal component analysis, or PCA, was used to analyze all short-term kinetic data, with the continuous variables of time (days), CaO₂ loading (%), porogen inclusion (%), average microbead diameter (μm), and OGR (mmol/day) considered pertinent. Generating the model (Figure S1) led to the creation of 5 principal components (PCs) representing 100% of the total variance of the data set, with PCs 1–3 accounting for over 88% of all variances with values of 46.4, 22.2. and 19.7% respectively. PC1 was composed of positive loadings from CaO₂ loading (0.976) and OGR (0.942), with a negative contribution from porogen inclusion (−0.604). PC2 was primarily composed of the variables porogen inclusion (0.573) and microbead diameter (0.777), with a negative loading from CaO₂ loading (−0.417). PC3 was primarily composed of measurement time (0.904) and microbead diameter (0.378).

2.4. Analysis of model tissue mathematical modeling

Following the detailed characterization of the oxygen generating properties of the OxySite bead formulations, finite element modeling (FEM) was implemented to evaluate their potential therapeutic impact on improving oxygen availability within an implanted graft. The computational model consisted of a thick avascular graft implanted within a vascularized site (Figure 5A). Oxygen availability within the graft, with or without selected OxySite microbead prototypes, was modeled. In addition, the impact of cellular demand/loading within the avascular implant was evaluated via modulation of the implant’s global oxygen consumption rate (OCR). Figure 5B–E illustrates the predicted oxygenation of a tissue slice located at the center of the 3D implant. For implants lacking the OxySite material (Figure 5B), the lowest cellular OCR condition ($R_{Max,O_2}=1mmol/(m^3\bullet s)$) resulted in a hypoxic core (defined as oxygen < 0.01 mM and designated in white) within the simulated tissue. The total hypoxic region increased as the global cellular OCR demand increased, with up to 52.3% hypoxia within the tissue slice. Supplementing with baseline OxySite microbeads (25% CaO₂; Figure 5C) resulted in the complete resolution of hypoxia for the two lower OCR simulations and a 1.5-fold reduction in the overall hypoxic area over OxySite-free models for the highest OCR (Figure 5C iii). Swapping the OxySite variant to a porogen formulation (24% CaO₂ and 7% NaCl₂; Figure 5D) further enhanced local oxygen availability. At the highest OCR, the inclusion of the porogen OxySite microbead reduced the overall hypoxia by 2.94-fold over OxySite-free models (Figure 5D iii). Finally, the use of “Mid” loaded OxySite beads (35% CaO₂; Figure 5E) further enhanced local oxygenation of the model tissue, with complete resolution of hypoxic regions for the highest OCR. To provide additional insight into the impact of OxySite microbead integration on oxygen gradients, models exploring bead placement within the implanted tissue were created (Figure S2). When OxySite microbeads were asymmetrically placed at the base of the model tissue, a large hypoxic core emerged within the construct; this highlights the importance of bead placement and/or distribution within the implant to ensure optimized in situ oxygenation. Overall, utilization of these in silico models can provide clear guidance on the selection of an OxySite microbead variant to support the oxygen needs of a specific implant design, with ease in the manipulation of key parameters that drive oxygen availability including device length scales, overall graft OCR depending on cell type and loading, and the anticipated oxygen tension at the implant site.

Figure 5.

Computational modeling predicts the impact of OxySite microbead formulations on the in situ oxygenation of a representative tissue graft model. a) Schematic of representative cell-based implant placed between two vascularized native tissues. b-e) Oxygen concentration heatmaps at steady-state conditions for b) control grafts or grafts containing c) OxySite microbead, d) low porosity microbeads; e) or mid CaO₂ loading microbeads. The cell loading density was further modulated to example the impact of low (i), medium (ii), and high (iii) cell oxygen consumption rates (OCR) on oxygen availability. Scale represents oxygen values at 10⁻² mol/m³. White regions represent oxygen concentrations below the hypoxia threshold (O₂ < 0.01 mM).

2.5. Addition of Outer Layering onto OxySite Microbeads for Enhanced Kinetic Control

For PDMS-based drug delivery platforms, the inclusion of an outer coating of pure polymer is a common method used to provide additional control over therapeutic release kinetics. In response, a methodology for applying a thin layer of PDMS to the outside of OxySite microbeads was developed, with the goal of further controlling OGRs and reducing byproduct production. Utilizing emulsion-based methods, discreet and complete PDMS coatings were generated around the OxySite microbeads, as visualized via SEM, NanoCT, and fluorescence imaging (Figure 6A). Per SEM, PDMS coated OxySite microbeads exhibited visually smoother outer bead topography (Figure 6A ii), when compared to unlayered MOBs (Figure S3). Discreet layers were also visualized using confocal microscopy of rhodamine B – doped PDMS coatings (Figure 6A iii) and stereomicroscopy imaging of water-soluble dyes emulsified into the outer PDMS coating (Figure S4). Image analysis of NanoCT images (Figure 6A iv) quantified an average layer thickness of 17.5 ± 1.9 μm.

Figure 6.

Coating of OxySite microbeads with an outer, reaction-limiting layer. a) Layer formation was achieved using i) microfluidic method, resulting in layered OxySite microbeads as validated via ii) scanning electron microscopy (SEM) images of layered microbeads, iii) confocal microscopy of fluorescent layer, iv) and nano-CT imaging of layer (pseudo-colored added to highlight outer layer). The impact of the outer PDMS layer on a) daily oxygen generation rates (OGRs), b) overall oxygen generated (mmol) during the study period, and c-d) temporal release of byproducts H₂O₂ and pH for OxySite microbeads (22% CaO₂ with 13% NaCl). N = 3 samples for each group; data represented as mean ± SD; solid lines represent trends in data over time; two-way ANOVA with Tukey’s HSD multiple comparison test for daily mean comparisons. * represents comparisons between the two graphed groups. Color bands in c & g represent H₂O₂ concentration ranges considered safe (green), potentially cytotoxicity (orange), or likely cytotoxic (red) support sensitive cell populations (green), bulk tissue (orange), and potential cytotoxicity (red).

The impact of layering on OxySite microbead oxygen kinetics and byproduct production was characterized using the “High” porosity OxySite microbead variant (22% CaO₂ with 13% Porosity). The addition of the PDMS layer to this microbead formulation imparted a significant effect on the magnitude of daily oxygen release but not its kinetic profile (Figure 6B; P < 0.0001 and 0.2097, respectively; two-way ANOVA). The layer also significantly reduced overall oxygen generation when compared to its unlayered counterpart (P = 0.0007; Figure 6C). This control over oxygen generation at early time points due to the outer PDMS layer resulted in a substantial reduction in generated H₂O₂ (Figure 6D) on day 1 (13-fold decrease; P < 0.0001). Layering also significantly altered pH levels (Figure 6E), although both groups maintained a pH below 7.6 throughout the study.

2.6. Microbead Characterization and Application in a Macroencapsulation Platform

The benefits of OxySite microbeads within cell-containing devices was explored using a macro-scale encapsulation system. For these studies, the larger microbeads, 300–600 μm diameter, were selected, as their size should translate to a more durable oxygen release. Long-term OGR kinetics were captured for two relevant OxySite formulations: standard OxySite microbeads (OxySite MB; 25% CaO₂, 0% Porosity); and Layered Microporous OxySite microbeads (LMOBs; 22% CaO₂, 13% Porosity, Layered) (Figure 7A–B). OxySite MBs exhibited a burst release of oxygen in the first three days, followed by a gradual reduction to a plateau ranging from 4.53 × 10⁻⁴ to 1.24 × 10⁻⁴ mmol O₂/day over the 25-day study period. In comparison, the LMOB prototype, which included both a porogen and outer layer, exhibited an attenuated oxygen burst release, resulting in significantly different OGRs on Days 1, 3 – 8 compared to OxySite MBs (Figure 7A; P < 0.0005, see Table S1); the two variants also exhibited significantly different kinetic curves (P < 0.0001). Of note, the introduction of the outer PDMS layer had significant impacts on byproduct production when compared to uncoated OxySite microbeads (Figure S5), with a 4-fold reduction in H₂O₂ production on Day 1 (P = 0.0005) and significantly decreased pH at Days 1, 3 and 7 (P < 0.04; see Supplementary Text 3).

Figure 7.

In vitro application of modular OxySite microbeads within a 3D hydrogel macroencapsulation device. Characterization of unmodified OxySite microbeads (MBs) and Layered, Microporous OxySite MBs (LMOBs) on a) daily oxygen generation rates (OGRs), and b) overall oxygen generated (mmol) during the study period. c) Schematic of macroencapsulation device containing MIN6 β-cells and microbeads encapsulated within agarose. Quantification of impact of OxySite microbead prototypes on d) cellular metabolic activity and e) total insulin content. Results expressed as fold-change to control devices containing PDMS-only microbeads. f) Schematic illustrating the general location of live/dead images collected from f) control or g) OxySite loaded macroencapsulation device, with representative images i) outer, ii) intermediate, iii) and center locations within the device. N = 3–6 samples for oxygen kinetic studies; N = 6–14 samples for metabolic activity measurements and total insulin quantification; data represented as mean ± SD; solid lines represent trends in data over time; two-way ANOVA with Tukey’s HSD multiple comparison test for daily mean comparisons.

OxySite MB and LMOB prototypes were then incorporated into agarose macroencapsulation hydrogels containing murine β-cells and cultured under critical hypoxia (1% O₂). Cells were assessed after three days in culture, as the two OxySite variants exhibited a high disparity in oxygen delivery and byproduct production at that time range. The inclusion of either OxySite MBs or LMOBs resulted in significant increases in β-cell metabolic activity (2.6-fold or 1.7-fold increase, respectively; P < 0.0001 or 0.0002) compared to constructs containing control CaO₂-free PDMS beads, albeit devices containing LMOBs exhibited significantly lower metabolic activity than those containing OxySite MBs (P < 0.0001) (Figure 7D). To capture functional activity of the encapsulated β-cells, total insulin content was measured. Both OxySite microbeads formulations significantly elevated β-cell function by over 1.5-fold, when compared to control (P = 0.0004 or 0.002 for baseline and LMOBs, respectively) (Figure 7E). Of interest, no significant difference in total insulin content was measured between treated groups, despite their significant difference in metabolic activity. Spatial analysis of cell viability was captured via live/dead confocal imaging of discreet positions within the agarose construct (Figure 7F–H). As expected, hypoxic gradients within control OxySite-free devices drove spatial patterns in cell viability, with most viable cells present at the outer edges of the encapsulating material (Figure 7G). The inclusion of OxySite microbeads resulted in viable cells throughout the construct, indicating significant shifts in oxygen availability throughout the macroscale device (Figure 7H).

2.7. OxySite Microbead Characterization and Application in an Injectable Platform

Engineering OxySite microbeads for co-delivery with cells within confined sites required unique design constraints. Specifically, prototypes needed to provide: 1) a size range suitable for injection and site retention; 2) robust oxygen generation with minimal material volume; and 3) reduced byproduct production to avoid deleterious impacts at the graft site. OxySite microbead material formulations of smaller sized microbeads (106–212 μm diameter) were screened to identify a variant that provided an optimal combination of these parameters. Four smaller microbead prototypes were interrogated: baseline OxySite microbeads (OBs; 25% CaO₂, 0% porosity); High Loading OxySite microbeads (HOBs; 50% CaO₂, 0% porosity); Microporous OxySite Microbeads (MOBs; 24% CaO₂, 7% Porosity); and High Loading, Layered, Microporous OxySite microbeads (HLMOBs; 46% CaO₂, 9% Porosity, outer PDMS layer) (Figure 8A). The impact of the material formulations on both short-term potency and the durability of oxygen generation, as well as byproduct generation, were explored. The addition of sacrificial porogens to the baseline OxySite microbeads altered the oxygen generation kinetic curve (Figure 8B), with elevated OGR for MOBs between days 4 and 13 when compared to OB (P < 0.026). Alternatively, increasing the reactant dosage within the smaller microbeads from 25 to 50% CaO₂ further elevated oxygen generation, with a highly significant increase in HOB OGR for the first 15 days of the study when compared to OB (P < 0.0001). Quantification of cumulative oxygen release over the 20-day time period found the HOB formulation to generate the greatest amount of oxygen (4.3-fold higher than OB), followed by the MOB (2.3-fold higher than OB). Contextualizing these oxygen release curves with reaction byproducts revealed a more complete picture, as byproduct production by these smaller, more potent microbeads was of particular concern. Quantification of H₂O₂ and pH (Figure S6A) found that OBs and MOBs levels were below 55 μM and 7.45 respectively, indicating cytocompatibility of these microbead variants. The local changes in H₂O₂ and pH induced by HOBs, however, indicated potential toxicity, with H₂O₂ levels well over 100 μM and pH changes ranging from 7.97 to 8.11. Thus, while the HOBs generated a high amount of oxygen, this came with the cost of greatly elevated pH and H₂O₂.

Figure 8.

In vivo application of injectable OxySite microbeads co-delivered with pancreatic islets into the murine kidney capsule. a) Schematic of the four OxySite microbead formulations used for in vivo transplants: unmodified OxySite microbeads (MBs); Microporous OxySite microbeads (MOBs); High Loading OxySite microbeads (HOBs); and High Loading, Layered, Microporous OxySite microbeads (HLMOBs). Characterization of b) daily oxygen generation rates (OGRs), and c) overall oxygen generated (mmol) during the study period of tested OxySite formulations. d) Schematic of in vivo transplantation of co-infused OxySite microbeads and primary pancreatic islets within the confined kidney capsule implantation site. e) Quantification of total insulin content (ng/IEQ) from explanted grafts, 24 hrs post-transplantation. N = 3 samples for oxygen kinetic studies; N = 4–9 samples for total insulin content; data represented as mean ± SD; solid lines represent trends in data over time; two-way ANOVA with Tukey’s HSD multiple comparison test for daily mean comparisons; one-way ANOVA with Dunnet’s multiple comparison test for explant total insulin content comparisons.

To balance high oxygen generation with mitigation of deleterious byproducts, a new formulation was created that combined high CaO₂ loading and porogen inclusion with an outer PDMS layer; this resulted in High Loading, Layered, Microporous OxySite microbeads (HLMOBs; 46% CaO₂, 9% Porosity, PDMS layer). As shown in Figure 8B, despite containing similar amounts of reactant CaO₂, the outer PDMS coating significantly attenuated burst OGRs seen in HOBs, resulting in significantly decreased OGRs during the first week of hydration (P < 0.0150; see Table Supplementary Text 3). The combination of high reactant loading and porogen mitigated the robust OGR suppression observed when layering beads, resulting in an OGR higher than the baseline OBs between day 3 – 11. Overall, the release curves for the HLMOBs prototypes were similar to the MOBs, with the exception of higher daily OGRs on days 4 and 6 (P = 0.045 and 0.021, respectively); cumulative oxygen release for HLMOBs and MOBs were not significantly different (Figure 8C; P = 0.219). The inclusion of the outer PDMS coating, however, significantly enhanced potential cytocompatibility, with an over 20-fold decrease in day 2 H₂O₂ levels when compared to HOBs, despite their similar CaO₂ loadings. In fact, average H₂O₂ levels were the lowest for the HLMOBs than any of the other formulations.

The potential benefits of the injectable OxySite microbead variants were explored via their co-delivery with primary pancreatic islets. Islets were used in this study due to their high sensitivity to hypoxia, with rapid loss of insulin secretory function (< 8 hours)^[11]. As such, the therapeutic impact of in situ oxygen supplementation was characterized via quantification of changes in total insulin content within the cellular graft. The murine kidney capsule was selected as the transplant site, as recent work implicates that poor oxygenation at this site during early engraftment (i.e. 24–72 hours post-transplant) may result in functional islet loss^[34]. Islets and the four different OxySite microbead variants were co-injected within the kidney capsule site and grafts were explanted 24 hours post-transplant for quantification of total insulin content. For all prototypes, OxySite microbeads were easily consolidated with the islet spheroids and co-injected through the syringe and tubing used for localization within the kidney subcapsular site; the volume of OxySite microbeads was targeted at ~10% of the total transplant volume for all conditions. The co-transplantation of islets with OBs or HOBs resulted in no significant change in insulin content (5.248 or 8.437 ng/IEQ, respectively; P = 0.99 or 0.86) compared to control (7.762 ng/IEQ). In addition, the co-transplantation of islets with the porous oxygen generating formulation MOBs resulted in a non-significant change in total insulin content compared to control (11.68 ng/IEQ; P = 0.7700). The co-infusion of islets with High Loading, Layered, Microporous OxySite (HLMOBs), however, significantly increased total insulin content within the graft (20.45 ng/IEQ; P = 0.003), resulting in over a 2.6-fold increase in insulin levels when compared to control transplants.

3. Discussion

Distinct OxySite microbead formulations within targeted size ranges were consistently formulated using emulsification techniques, via modulation of stirring speed and post-fabrication sieving. Resulting microbeads provided dexterity in application, with larger scale microbeads (300–600 μm) suited for larger macro-scale devices and smaller microbeads (106–212 μm) supportive of injectable applications. Based on the sieving method employed and these results, it is postulated that various size ranges could be generated by selecting different sieve sizes. The fabrication of OxySite microbeads containing increased CaO₂ loadings was straightforward via elevated inclusion of the CaO₂ powder within the polymer. For porogen integration, NaCl was selected, given its common use for this purpose^[35–37]. As smaller porogen sizes can increase the effective surface area within the microbead and their subsequent overall reactivity^[38], the salt porogen was milled into fine microparticles (~5 μm). While potential cytotoxicity and immunogenicity of NaCl should be taken into consideration^[39], calculations of maximum potential NaCl release (assuming no loss during fabrication) indicates concentrations would be well below cytotoxic levels (~55 mM)^[38]. Overall, the modulation of both CaO₂ and NaCl loadings within the tested ranges resulted in no gross changes in microbead morphology and computed tomography imaging indicated successful encapsulation of the agents within the cured PDMS polymer matrix.

Detailed characterization of the impacts of reactant and porogen loading, as well as surface area, on oxygen kinetics provided key engineering parameters for implementation into living tissues. Based on previous studies using PDMS-based drug delivery systems, OxySite microbeads were expected to behave as a non-erodible, osmotic pressure-controlled release system, due to the physical trapping of CaO₂ and/or NaCl as distinct aggregates within the polymer matrix^[21,40]. As a result, one can infer that oxygen generation would be primarily mediated by osmotic pressure creating channels within the microbead structure as water infiltrates and reacts with the CaO₂^[41,42]. Thus, modifications of increased CaO₂ loading or porogen inclusion were expected to not only increase daily OGRs, but also increase total “potential” oxygen generation by either increasing the overall payload of oxygen to be generated or by removing the resistance to water migration into the matrix.

For the lower CaO₂ (“Low” and “Mid”) loadings and porogen inclusion (“Low”) ranges, daily oxygen generation values and overall kinetics were consistent with an osmotic pressure-controlled release model, with kinetic profiles exhibiting an early burst release behavior and an increased reactant or porogen presence resulting in increased overall oxygen generation. For the highest CaO₂ (50%) or porogen (13%) loadings, alternative kinetic curves with a delayed or ramping up of OGR over time were measured. This lag in oxygen generation, in conjunction with the non-linear increase in cumulative oxygen generation seen with these higher loadings, indicates swelling as the primary driver of water infiltration^[41], as observed for other silicone-based release systems containing high levels of porogens or other impurities^[43]. This phenomenon, often explained in terms of percolation theory, results in nonlinear shifts of reactivity as the reactant loading density increases^[42]. The dominant impact of CaO₂ loading on oxygen kinetics was further contextualized via PCA analysis. Examination of size effects on oxygen generation, particularly when combined with porogen inclusion, also revealed interesting trends. Standard 25% CaO₂ “small” and “large” microbeads exhibited typical diffusion-based kinetic curves, with the higher surface area of “Small” microbeads yielding higher OGRs when compared to their “Large” counterpart only at early time points. These trends were supported by PCA analysis, which did not strongly correlate size with CaO₂ loading or OGR, indicating that it was not a dominant driver of kinetic changes. The combination of porogen and the smaller microbead size, however, yielded a surprising interaction, with ablation of the early burst release profile and a steady OCR over the study period, albeit with a greater magnitude. This observation was supported by PCA analysis, which suggested a relationship between porogen inclusion and size. Thus, a multiplicative effect from changes in the external (reduced diameter) and internal (porosity) surface area is indicated.

For use within living tissues, there is a need to contextualize oxygen kinetics with byproduct production, as this is critical for determining what modifications would be suitable, or even reasonable, for different biological applications. For CaO₂ degradation, particular attention was paid to H₂O₂ concentrations and pH changes, which can cause dysfunction or cell death unless properly managed^[44–48]. Collectively, OxySite variants that exhibited a swelling-driven reaction model showed a dramatic increase in pH and H₂O₂, likely driven by accelerated exposure of CaO₂ to water. Specifically, increasing CaO₂ past 35% resulted in a non-cytocompatible prototype. Furthermore, 13% porogen inclusion with 22% CaO₂ lead to toxic H₂O₂ release at early time points, likely due to increased reactant surface area. Both modifications, in conjunction with the overall increase in byproducts seen with reducing microbead size/increasing surface area, present a potential risk to local cells. For the other prototypes, only a transient elevation in H₂O₂ was observed, which may be compatible with most cell types, as the duration of exposure plays a key role in cytotoxicity particularly given the buffering capacity of culture media and physiological tissues^[49–51]. It should be noted, however, that cells exhibit a range of sensitivities to ambient ROS and pH. Thus, care should be taken to contextualize these ranges to the targeted cells and/or tissues of interest. If a concern, deleterious byproducts can be mitigated through several approaches. Overnight pre-incubation of OxySite microbeads prior to cell culture or implantation should avoid transient exposure to elevated H₂O₂ created during initial material hydration. Alternatively, antioxidants could be added to the OxySite microbeads to scavenge locally generated ROS, as used for other CaO₂-based material prototypes^[52–55].

With high customization of the OxySite material to create distinct oxygen generation kinetics, in silico models can be used to predict the potential therapeutic benefit of a given OxySite microbead variant under different in vivo conditions and implant features. A robust model should support ease in the customization of the key parameters of implant geometry and length scales, cell loading and metabolic demand, and the local tissue microenvironment^[13,56]. To interrogate the impact of OxySite microbead variants on implant cell survival, a 3D avascular tissue graft model was generated, whereby the parameters of oxygen demand within the tissue, OxySite microbead integration and placement, and OxySite kinetics could be modified. As shown, in silico models allowed for the identification of suitable OxySite microbead variants that meet the specific oxygen demands of the idealized implant. For lower cell loadings, the inclusion of baseline (25% CaO₂) OxySite microbeads may be the most appropriate, while implants containing high cell densities or cells with elevated oxygen demands may require the use of more potent OxySite microbead variants. Computational models were also able to highlight the effect of microbead placement within the implant. The asymmetry generated when OxySite beads “settled” at the bottom of the implant site illustrates the importance of microbead distribution within thicker tissue implants to provide optimal in situ oxygen delivery. Overall, the model provides ease in screening OxySite variants for different implant characteristics and the dexterity of the OxySite formulations to match the distinct requirements of a graft. Future studies can include a time-dependent OGR, which can be used to study the temporal oxygen supplementation needs for closed implants that do not support intra-device vascularization. Additional parameters of interest that can be added to these models include the examination of byproduct release within the tissue space and interrogation of local oxygenation effects on cellular function such insulin production.

Based on results from the material formulation and microbead size study, it was clear that increasing the percentage of solid particulates within the PDMS microbead shifted kinetics to a swelling-based reaction model. While this increased the potency of the system, it also enhanced exposure to potentially deleterious reaction byproducts. To effectively limit the rate at which oxygen could be generated, an outer hydrophobic PDMS coating/layer was proposed. By providing a physical barrier to water migration, this pure PDMS layer would serve as a rate-limiting factor^[57,58]. The resulting prototype should then behave more like a reservoir-based release system, whereby a highly potent inner oxygen generating material is encased within an outer layer that mediates global release^[41].

To generate the outer layer on the microbeads, several approaches were explored. Initially, channel-based microfluidics appeared as an attractive option, whereby individual microbeads could be coated in a manner similar to published reports^[59–61]. In practice, however, the optimization of several microbead variants, each with its own size distribution and densities, was exponentially challenging and would likely not yield a scalable process. Alternatively, bulk emulsification supported ease in scale and in modulation of parameters such as emulsifier molecular weight, stirring speed, and the ratio of microbeads to pure PDMS. Iterative optimization found a combination of elevated stirring speed and a 1:2 ratio of microbeads to PDMS led to the most effective coating of microbeads without excess waste products or clumping. This “spin coating” method relied on the same polar/non-polar interactions that drive the base microbead emulsion, with the non-aqueous PEG emulsifier driving the formation of a thin, PDMS conformal coating^[62]. Multiple imaging modalities validated both the micro scale and uniformity of the resulting layer.

The impact of layering was striking, more notably for microbead variants previously identified as highly reactive and at risk of matrix swelling. For example, the addition of the PDMS layer to “high” porogen inclusion microbeads reduced early OGRs and decreased burst generation of byproducts, reducing H₂O₂ concentrations well below the proposed threshold for cytocompatibility (<50 μM). This dampening effect on oxygen generation kinetics was consistent with previous implementations of layers and coatings, albeit on a smaller scale as pure silicone barriers have historically been utilized on a macro-scale, e.g. bulk silicone membranes for controlled hormone release^42,64. Compared to OxySite microbead variants exhibiting first-order kinetics (e.g., baseline OxySite or “low” porogen inclusion formulations), layering created a temporary (1–3 day) “lag” period in oxygen generation kinetics during in vitro studies. This hindrance to osmotic pressure was previously identified in monolithic PDMS coatings, with resolution of this dampening effect following the inclusion of soluble channeling agents within the layer^[63]. However, in the context of decreasing reactivity and gaining greater kinetic control, this “rate-limiting” effect was proposed to be beneficial for maintaining zero order-like kinetics under varying culture and in vivo conditions.

The therapeutic benefit of the modular in situ oxygenation platform was explored using two models: a macroscale, avascular encapsulation model and an injectable application. For the macroencapsulation study, two OxySite prototypes with distinct oxygen kinetics at early time points were studied. Baseline OxySite microbeads (OB) exhibited a burst release over the first ~5 days while OxySite microbeads possessing both porogens and an outer layer (LMOB) displayed a steady release rate over the 25-day study period. Thus, interrogating cellular viability and function at early time points should highlight the impact of the OxySite variants within this macroscale encapsulation platform. Both prototypes significantly improved global β-cell metabolic activity and insulin production, demonstrating the advantages of in situ oxygenation via this microbead format. In addition, the inclusion of the oxygen generating microbeads mitigated central necrosis, which is commonly observed for constructs reliant on only external oxygen diffusion. When compared to OxySite microbead prototype results, the impact of the OB formulation on metabolic activity appeared more favorable; however, the observation of increased metabolic activity without improved theoretical function indicates potential metabolic stress, which may be caused by elevated exposure to pH and H₂O₂, as supported by previous work^[43]. Overall, the strong cellular benefits of OxySite microbead integration within large macroencapsulation devices were clearly demonstrated, with the feasibility of customization of microbead prototypes to the oxygen demands, as well as potential ROS sensitivities, of the co-encapsulated cells.

For the injection of OxySite microbeads with cells within a confined transplant site, several smaller microbead variants were screened. While baseline OxySite microbeads (25% CaO₂) exhibited the most consistent oxygen generation rates in vitro, total oxygen supplementation was insufficient to impart a measurable benefit in vivo. To elevate in situ oxygen availability without increasing microbead dosage, variants with elevated CaO₂ (HOBs) or porogen inclusion (MOBs) were tested. Elevating reactant dosage using the HOBs formulation did not alter islet function in vivo. This was likely due to the elevated presence of H₂O₂ and hydroxide ions in the local microenvironment that were not mitigated by physiological buffering mechanisms. In vitro, porogen microbeads (MOBs) increased cumulative oxygen levels 2-fold over the OB formulation without elevating byproduct levels; however, the MOBs formulation did not significantly raise insulin content within the graft. This discrepancy between in vitro oxygen kinetics and in vivo impact may be due to differences in their environment. In vitro, oxygen kinetics were measured under closed conditions, which may dampen CaO₂ reactivity over time due to oxygen saturation, despite daily refreshments of the system. The transplant environment likely mimics a sink condition, as reactants are cleared and/or consumed by the system. In addition, the efficiency of water migration into the PDMS-based microbead may be altered by the composition of the interstitial fluid, leading to changes in the hydration rate of encapsulated CaO₂ particulates. Future studies will explore the optimization of in vitro models of cell therapy to more accurately model in vivo conditions.

To provide an elevated controlled release system, a final prototype was created that combined porogens and high CaO₂ loading with an outer PDMS layer. It was hypothesized that this material prototype would deliver a robust oxygen reservoir with controlled reactivity. In vitro oxygen kinetics and byproduct generation for this formulation were similar to the MOB variant, with the exception of a gradual ramping OGR within the first three days. The inclusion of the HLMOB formulation within the islet graft resulted in a substantial improvement in total insulin content post-transplant. This result further indicated that kinetics may be altered under in vivo conditions, with accelerated oxygen availability from what was predicted in vitro. Furthermore, the addition of an outer layer imparts benefits likely correlated to decreased burst reactivity of the encapsulated reactant.

After establishing the early cellular benefits of in situ oxygenation using OxySite microbeads, future work is focused on examining the therapeutic benefits of this elevated engraftment survival using diabetic models. We are also exploring the translation of this technology to other tissues and implants. Furthermore, the newfound flexibility provided by the microbead format allows for its use within previously infeasible sites. Of particular interest is the subcutaneous space, which is appealing as an easily accessible site but is oxygen deficient^[7,64]. The small size and customizability of the new OxySite microbeads support their subcutaneous injection or co-transplantation into cell-containing pouches that are recently receiving considerable clinical attention^[65,66]. In addition, transient oxygen supplementation for ischemic events initiated by trauma or disease could be another application, particularly for tissues experiencing high metabolic demand or conditions that delay angiogenesis (e.g., diabetic ulcers)^[67–70]. Finally, detailed studies investigating the in vivo oxygen generation and byproduct production from these OxySite microbead variants would further contribute to our understanding of these materials and allow for additional optimization and/or customization of this in situ oxygen generating material platform.

4. Conclusion

Oxygen supply is a crucial parameter in the design of functional engineered tissues. In situ oxygen generating materials may support cellular implants but requires high modularity to ensure optimal oxygen delivery. Herein, a customizable oxygen generation biomaterial in a microbead geometry was fabricated through the tailoring of the reactant (CaO₂) dosage, porogen inclusion, modulation of microbead size, and the addition of an outer, rate-limiting PDMS layer. Resultant microbeads achieved distinct and customizable oxygen generation kinetics in both short-term (7 days) and long-term (20–25 days) oxygen generation studies. Oxygen generation was contextualized with levels of potentially deleterious ROS and alkaline byproducts to identify material formulations suitable for tissue engineering applications. In silico, multiphysics models were developed and leveraged to predict the potential impact of different OxySite variants on local oxygen availability. Promising OxySite variants incorporated within two distinct translational cell therapy models imparted beneficial effects on metabolic and therapeutic activity. This work highlights the potential for a microbead format for potent, durable oxygenation with improved kinetic control and geometric flexibility compared to previous platforms. This modular material platform could be translated to numerous applications where in situ oxygen is desired, from supplementing implants or abating ischemic medical events to soil and water remediation.

5. Experimental Section

5.1. Fabrication and Characterization of Modular Microbeads

The base polymer matrix for the modular microbead platform was prepared by first mixing medical-grade Nusil PDMS (Cat. No. MED-6215; Nusil) parts A and B at a 4:1 (vol/vol) ratio in an automatic mixer (Model No. ARE-310; THINKY). Calcium peroxide (CaO₂; Cat No. 466271; Sigma Aldrich; Purity = 75%; 25–50% wt/wt) was mixed into the PDMS at the concentrations designated using a THINKY mixer, as previously described^[19]. For porogen studies, sodium chloride (NaCl; Cat. No. S271–500; Thermo Fisher Scientific) was milled to a fine powder (Retsch; Cyromill) and mixed into the PDMS-CaO₂ mixture. Next, the composite material was injected into a 100 °C PEG 8000 (Polyethylene glycol; Cat. No. BP2331; Thermo Fisher Scientific) bath while continuously stirring at 60 RPM via a laboratory mixer (Cat. No. 04555–00; Cole-Parmer). The suspended microbeads were cured at 100 °C for 2 hours before being immersed in 1% (wt/vol; aqueous) Pluronic F-127 (Cat. No. P2443; Sigma Aldrich) solution to dissolve the PEG 8000 and prevent microbead clumping. Beads were then sorted into desired size ranges (106–212, 300–600 μm) using macro-scale sieves (Cat. No. 106 μm: 0488110Z, 212 μm: 0488110U, 300 μm: 0488110T, 600 μm: 0488110P; Thermo Fisher Scientific) and washed overnight in deionized water. For storage, microbeads were dried using a 40 μm cell culture filter (Cat. No. 22363547; Thermo Fisher Scientific) and sealed with a desiccant (Drierite, Cat. No. 075783B; Thermo Fisher Scientific) until use. Microbeads were used within two weeks post-fabrication.

To deposit an outer PDMS layer onto the microbeads, dried microbeads were suspended in medical-grade Nusil PDMS (A:B; 4:1; vol/vol) at a ratio of 1:2 (wt/vol) and mixed in the THINKY mixer for 1 minute at 2000 RPM. A second emulsification was performed into 100 °C PEG 8000 stirring at 200 RPM via a laboratory mixer. Beads were sorted into desired size ranges as described above and washed overnight in deionized water before collection and subsequent storage under desiccant.

Microbead size was confirmed by coulter counter (N = 3; Coulter LS13320). Overall microbead structure, including microporous internal structure and CaO₂ distribution, were imaged using nanoscale computed tomography (nanoCT; v|tome|x m 240; General Electric). Additionally, the outer PDMS layer was visualized using nanoCT, confocal microscopy (Leica SP8) using a fluorescent probe (Rhodamine B; Cat. No. UB7203; J.T. Baker), scanning electron microscopy (SEM, Hitachi SU5000 Schottky Field-Emission), and stereoscopy using a stained outer PDMS solution. Microbeads used in kinetic studies and their properties can be found in supplementary documentation (Supplementary Text 2).

5.2. Characterization of Customizable Oxygen Generation Kinetics

Oxygen generation measurements from each microbead variant was accomplished using a custom-built, sealed system using a non-invasive oxygen-sensing probe (PreSens) connected to a fiber optic transmitter (OXY-10 mini; PreSens)^[12,19]. In brief, 10 – 25 mg of microbeads, dependent on variant and study, were immersed in 1 mL of phosphate-buffered solution (PBS; Cat. No. 10010023; Gibco) and incubated at 37°C in a sealed titanium chamber (Instech Laboratories). Oxygen partial pressure within the chamber was recorded in mmHg (torr) and study media was collected daily and stored at −80°C for later analysis. N = 3–9 independent replicates were measured for each variant. The daily average oxygen generation rate (OGR) was calculated as the slope of the oxygen generation curve between 1 and 2 h after initialization of recording. Recorded values were converted to mols of oxygen using Henrys Law [Equation 3] and the Bunsen coefficient of oxygen ($\alpha =1.27\cdot {10}^{-9}\frac{mol}{{cm}^3\cdot mmHg}$; oxygen at 37°C in the medium^[71]) to be used for subsequent calculations and mathematical modeling.

$$c=\alpha p$$ Equation 3

5.3. Quantification of Microbead Byproducts

H₂O₂, produced as an intermediate of the oxygen-generating reaction, was quantified using a colorimetric detection kit (Cat. No. ADI-907–015; Enzo; sensitivity = 1.51 μM) with n = 3 experimental replicates and n = 3 technical replicated per sample. Changes in pH due to oxygen generation byproducts were measured using a dedicated pH probe (Mettler Toledo) with n = 3 experimental replicates. For short-term generation studies, temporal byproduct generation was measured for both buffered (PBS + 2.4% HEPES) and unbuffered (PBS only, Figure S7) samples to capture byproduct measurements relevant to physiological and non-physiological systems. For microbeads used for in vitro, agarose macroencapsulation, byproducts were measured from unbuffered (PBS) from oxygen generation studies. For microbeads used for injectable transplant into the murine kidney capsule, the approximate dosage of microbeads used for transplantation (0.2 mg) was submerged in 100 uL of PBS and incubated for 72 hours under sealed culture conditions at 37 °C. The collected, unbuffered media was then used for byproduct quantification.

5.4. Principal Component Analysis (PCA) of Microbead Modifications

To further analyze and facilitate understanding of short-term kinetic data, a statistical model was built using PCA. Graphpad Prism 9.3.1 was used to generate the model. Temporal oxygen generation data was reorganized to format each data point in terms of measurement time (days), CaO₂ loading (%), porogen inclusion (%), average microbead diameter (μm), and OGR (mmol/day). All variables were considered continuous. PCs were chosen based on a total percentage of explained variance, with 75% of total variance chosen as the threshold. A total of 5 PCs were created, with 3 PCs accounting for over 75% of the total variance. PC loadings were used to determine primary sources of variance within the data set and potential relationships between measured variables.

5.5. Computational Modeling of Oxygen-generating Microbeads

A representative model of an avascular, tissue-engineered implant placed within a typical site was created in COMSOL Multiphysics^® 5.6. The purpose of the model was to predict the impact of local in situ oxygenation, provided by the microbead prototype, on the oxygenation of the engineered tissue. Parameters of interest included: the microbead prototype, overall tissue oxygen requirements, and spatial microbead distribution. A representative geometry was used. (900 W x 900 L x 1000 H; μm). Oxygen consumption was assumed to occur homogeneously throughout the structure. The oxygen consumption rate (OCR) of the tissue was modeled as a global Michaelis-Menten reaction [Equation 4] as previously described^[19,72]

$$R_{O_2}={-R}_{Max,O_2}\frac{c_{O_2}}{c_{O_2}+C_{{MM,O}_2}}·\delta (c_{O_2}>C_{cr})$$ Equation 4

Where $R_{Max,O_2}=0.001-0.0013mol/(m^3·s)$, $C_{{MM,O}_2}=0.001mol/m^3$, and $\delta \left(c_{o_2}>C_{cr}\right)$ represents a Heaviside function where $C_{cr}=0.0001mol/m_3$, based on previously published models^[72]. The OGRs for the three microbead variants tested were extracted by taking an average of measured values between days 1–3. To account for changes in volume from increasing density, the expected density for each microbead was calculated and accounted for in volumetric OGRs. (Equation S3, S4; Table S2). The diffusivity coefficient $\left(D_{eff}\right)$ of tissue and microbead were assumed to be $D_{eff,tissue}=2x{10}^{-9}{m}^2/s$ and $D_{eff,PDMS}=3.25x{10}^{-9}{m}^2/s$ respectively based on previously reported values^[72,73]. The top and bottom boundaries of the implant were assumed constant at typical tissue oxygen (0.05 mM). Simulations started under normoxic conditions (0.2 mM) and the symmetry condition was placed on the left, right, front, and back sides of the modeled geometry to simulate a larger bulk tissue. Models were run until equilibrium was reached (< t = 1 hour). Additionally, the placement of the microbeads was modeled in a “centered” and “one-sided” orientation to simulate the effect of spatial distribution on hypoxia with a set $R_{Max,O_2}=0.00115mol/(m^3·s)$. Additional details on calculations and model parameters can be found in the supplementary materials (Supplementary Text 1).

5.6. In vitro Cell Culture

MIN6 murine β-cells (Passage 20–50; Cat No. C0018008; AddexBio) were cultured as monolayers in tissue culture treated flasks with media changed every 3–4 days. Media was composed of Dulbecco’s Modified Eagle Medium (DMEM; Gibco) containing 5.5 mM glucose, 10% FBS, 1% P/S, 2.4% HEPES (Cat. No. 25–060-Cl; Corning), and 0.001% (v/v) beta-mercaptoethanol (Cat. No. 21985–023; Gibco). Upon reaching 80–90% confluence, cells were lifted using 0.25% trypsin (Cat. No. 15050–057; Gibco) and either passaged at low density for propagation (1:3 – 1:4) or co-cultured in agarose hydrogels alongside microbeads for experimental evaluation.

5.7. In vitro Agarose Macroencapsulation

For agarose macroencapsulation, MIN6 cells (11,300 cells/mm³) were co-encapsulated inside a 2% agarose hydrogel alongside 25 mg of microbeads (n = 6 – 14 per group), as previously described^[11,12]. OxySite microbeads (25% CaO₂, 0% Porosity, Unlayered, 300–600 μm) and Layered Microporous OxySite microbeads (LMOB; 22% CaO₂, 13% Porosity, Layered, 300–600 μm) were evaluated. In brief, microbeads were sterilized within a 40 μm cell culture filter using 70% ethanol for 30 minutes and 5x PBS 1x washes for 5 minutes each before being coated for 30 minutes in 100 μg/mL human plasma fibronectin (Cat. No. 33016–015; Gibco) at 37°C to improve hydrophilicity. After removing excess fibronectin with 3x PBS 1x washes, microbeads were transferred to 50 mL centrifuge tubes containing pre-prepared agarose and MIN6 cells at 40℃. Homogenously distributed microbeads and β-cells in agarose were then added to a 10 mm diameter, 4 mm thickness sterile mold and allowed to cure at 4℃ for 8 minutes. Finished constructs were carefully transferred to a 24-well plate and submerged in 1 mL DMEM media. These constructs were subsequently cultured under critical hypoxia (1% O₂) for 72 hours before portioning for quantification of metabolic activity via AlamarBlue assay and storage in total insulin release (TIR) buffer (7% dH₂O, 18% 1M HCl, 75% EtOH) at −80°C. Total insulin content was later quantified using an insulin ELISA (Cat. No. 10–1247-01; Mercodia) and a spectrophotometer (Spectromax M5; Molecular Devices). Representative cell viability was measured qualitatively using LIVE/DEAD viability/cytotoxicity assay (Cat. No. L-3224; Thermo Fisher Scientific) and confocal microscopy.

5.8. C57BL/6 Murine Islet Isolation and Culture

Islets were isolated from male C57BL/6 mice (Jackson Labs) as previously described^[12], under protocols approved by the University of Florida (UF) Institutional Animal Care and Use Committee (IACUC). Isolated murine islets were cultured overnight in CMRL 1066 media (Cat. No. 99–6630-CV; Corning) with 10% FBS (Cat. No. SH30396.03; HyClone), 1% P/S (Cat. No. 30–002-CI; Corning), 1% L-glutamine (Cat. No. 25–060-CI; Corning), and 25 mM HEPES buffer at 37°C and 5% CO₂. Islets were manually counted and quantified as IEQ. IEQ aliquots were prepared separately immediately prior to surgery.

5.9. Transplantation into the Murine Kidney Capsule and Insulin Quantification

Islet and microbead transplantation were conducted under protocols approved by the UF IACUC. OxySite microbeads (25% CaO₂, 0% Porosity, Unlayered, 106–212 μm), Microporous OxySite microbeads (MOB; 24% CaO₂, 7% Porosity, Unlayered, 106–212 μm), High CaO₂ loading microbeads (HOB; 50% CaO₂, 0% Porosity, Unlayered, 106–212 μm), and High Loading, Layered, Microporous OxySite microbeads (HLMOB; 46% CaO₂, 9% Porosity, Layered, 106–212 μm) were evaluated. Microbeads were sterilized using 70% ethanol for 30 minutes and 5x PBS 1x washes for 5 minutes each before being coated overnight in 100 ug/mL human plasma fibronectin (Cat. No. 33016–015; Gibco) at 37°C on a vortex to improve hydrophilicity. After removing excess fibronectin with 3x sterile PBS 1x washes, beads were transferred to CMRL 1066 and 40 beads were manually counted using a brightfield microscope and P200 pipette using wide-orifice tips. Counted microbeads were added to pre-prepared 400 IEQ murine islet aliquots and kept at 37°C until transplantation. For surgery, animals were anesthetized with isoflurane and a midline-dorsal incision was made to gently expose the kidney, which was kept hydrated for the duration of the surgery using sterile saline (Cat No. Z1376, Intermountain Life Sciences). Microbeads and islets were pelleted at 1000 rpm before being injected under the kidney capsule through a small incision on the kidney surface. After cauterization, the incision was closed using sutures and animals were observed until recovery. Evaluated microbeads (OxySite microbeads, n = 4; MOBs, n = 4; HOBs, n = 4; HLMOBs, n = 6) were compared to transplantation with islets and control, pure-PDMS microbeads (Control, n = 8). Animals were monitored for changes in body weight and behavior until the planned study endpoint (24 h), where then animals were sacrificed and the surface of the kidney was explanted and stored in total insulin release (TIR) buffer (7% dH₂O, 18% 1M HCl, 75% EtOH) at −80°C. Insulin was quantified using an insulin ELISA and results were measured using a spectrophotometer.

5.10. Statistical Analysis

In all studies, data were expressed as the mean ± SD. For comparison of two experimental groups, an unpaired Student’s t test was performed after confirmation of comparable variance. For comparison of more than two groups, a one-way ANOVA was performed followed by Tukey’s HSD multiple comparisons, Šídák’s corrected multiple comparisons, or Dunnett’s corrected multiple comparisons, based on the parameters of the study. For oxygen generation studies and byproduct quantification, a two-way ANOVA was used to determine the interaction between changing parameters. In all comparisons, a P value ≤ 0.05 was considered significantly different. Statistical analysis was performed using Graphpad Prism 9.3.1. Specific details on statistical tests performed for each comparison can be found in the supplementary materials (Supplementary Text 3).