pO₂ Reporter Composite Hydrogel Macroencapsulation Devices for Magnetic Resonance Imaging Oxygen Quantification

Abstract

Hydrogel cell encapsulation devices are a common approach to reduce the need for chronic systemic immunosuppression in allogeneic cell product transplantation. Macroencapsulation approaches are an appealing strategy, as they maximize graft retrievability and cell dosage within a single device; however, macroencapsulation devices face oxygen transport challenges as geometries increase from preclinical to clinical scales. Device design guided by computational approaches can facilitate graft oxygen availability to encapsulated cells in vivo but is limited without accurate measurement of oxygen levels within the transplant site and graft. In this study, we engineer pO₂ Reporter Composite Hydrogels (PORCH) to enable spatiotemporal measurement of oxygen tension within macroencapsulation devices using the Proton Imaging of Siloxanes to map Tissue Oxygenation Levels (PISTOL) magnetic resonance imaging approach. We engineer two methods of incorporating siloxane oximetry reporters within hydrogel devices, an emulsion and microbead-based approach, and evaluate PORCH cytotoxicity on co-encapsulated cells and accuracy in quantifying oxygen tension in vitro. We find that both emulsion and microbead PORCH approaches enable accurate in situ oxygen quantification using PISTOL magnetic resonance oximetry, and that the emulsion-based PORCH approach results in higher spatial resolution.

1. Introduction

Therapeutic-delivering cell therapies are emerging as a promising and effective method to treat chronic diseases such as type 1 diabetes ⁽¹⁾ and hemophilia ⁽²⁾. Allogeneic cell products enable off-the-shelf availability of cell therapies ^(3),(4), but require chronic systemic immunosuppression to prevent rejection. For many applications and patients, the acute risks of immunosuppression such as cancer and infection outweigh the benefits ^(5),(6). As such, it is critical to reduce or eliminate the need for chronic systemic immunosuppression in cell replacement therapies to widen their applicability and the eligible patient population.

Cell encapsulation within immune-isolating hydrogels has been explored since the 1970s as a means to reduce or eliminate the need for immunosuppression, with extensive preclinical and clinical investigation with insulin secreting cells for the treatment of type 1 diabetes ^(7),(8). Hydrogel encapsulation of cells in microcapsules, typically in materials such as agarose, alginate, or poly(ethylene glycol) (PEG) ^(9)–(11), enables the exchange of glucose, insulin, and oxygen, but prevents the direct recognition of encapsulated cells by host immune cells ^(9),(12). Hydrogel microcapsules are typically spheres of 300 to 1000 μm diameter containing one to three islets per capsule ⁽¹³⁾; the resulting volume of hydrogel and cells necessitates delivery to large cavities such as the intraperitoneal space. Delivery of thousands of individual capsules results in poor retrievability in the case of adverse events or graft failure ⁽¹⁴⁾. Additionally, hydrogel microcapsule oxygen availability, and therefore cell survival and function, depends on exchange with the intraperitoneal fluid, which can be impeded by capsule surface fibrosis and capsule adherence to abdominal organs and tissues ^(15),(16).

Macroencapsulation devices are an alternative approach that maximizes encapsulated graft retrievability by containing many cells within a single device ⁽¹⁰⁾. The containment of large numbers of cells at higher densities enable transplantation into alternative transplant sites ⁽¹⁷⁾ such as the subcutaneous or omentum sites ⁽¹⁰⁾. Several macroencapsulation approaches have advanced to clinical trials, such as Viacyte’s subcutaneously implanted VC-02 device ⁽¹⁾. Macroencapsulation devices face challenges in oxygen transport to encapsulated cells due to (1) large diffusion distances between cells and the surface of the device, as macroencapsulation devices typically have lower surface area-to-volume (SA:V) ratios relative to microcapsules ^{(10),(13),(18)}; and (2) variable oxygen availability within alternative transplant sites, estimated as ranging from 40 Torr in the subcutaneous space ⁽¹⁹⁾ to 136 Torr in the intraperitoneal cavity ⁽²⁰⁾. Computational approaches can be used to predict in vivo oxygen transport and to design macroencapsulation devices that reduce encapsulated cell hypoxia, as well as approximate the oxygen transport achieved with microcapsule designs ^{(10),(13),(21)}. However, computational approaches are limited without accurate measurements of oxygen in vivo in device transplant sites. As such, there is a critical need to spatiotemporally monitor oxygen levels within macroencapsulation devices noninvasively in vitro and in vivo to improve device design and clinical outcomes.

Methods to monitor real-time oxygen concentration within cell encapsulation devices in vitro and in vivo have been limited. Several groups have attempted to develop methods of noninvasive imaging of islet graft oxygen concentration via ¹⁹F magnetic resonance (MR) imaging, using perfluorocarbon labeling of microcapsules dispersed within the recipient’s abdominal cavity ^(22)–(25). None of these investigations explored macroencapsulation devices in vivo, nor correlated in vivo measurement to islet viability or function. While the absence of background ¹⁹F signal is a major advantage of this technique, a major limitation is the requirement of additional hardware to register the ¹⁹F on ¹H MR scanners, potentially limiting widespread adoption in the clinical setting. A promising alternative is the use of EPR-based oximetry, which uses an analogous linear relationship between the spin– lattice relaxation rate (R1) of oxygen-sensitive spin probes such as trityl OX071 and pO₂, to generate oxygen maps ⁽²⁶⁾. While a promising research tool, the EPR-based method is limited in spin probe clinical translatability and the absence of an ability to provide anatomical imaging.

An alternative approach to spatiotemporal in vivo oxygen MR measurement is a siloxane probe-based ¹H MRI technology, Proton Imaging of Siloxanes to map Tissue Oxygenation Levels (PISTOL) magnetic resonance imaging ⁽²⁷⁾. Siloxanes have high oxygen solubility, a low toxicity ⁽²⁸⁾ that decreases as siloxane size increases, and are commercially available in a variety of forms in terms of chain length, orientation (linear or cyclic), and with or without functional groups ⁽²⁸⁾. Symmetric siloxane methyl protons exhibit a single peak at around 0.1 ppm, regardless of the chain length and orientation, and this chemical shift makes the siloxane signal distinguishable from that of water and fat ^(29),(30). The highly hydrophobic nature is also advantageous as it restricts the diffusion of aqueous ions, making this oximetry method insensitive to the presence of ions and metals in the tissue or cell culture ⁽²⁷⁾. We have evaluated a range of siloxanes that display a pO₂ sensitivity that is higher than perfluorocarbon labels ^{(27),(30)–(33)}, and we previously demonstrated the capability of PISTOL to be applied to biological systems by measuring cellular oximetry using hexamethyldisiloxane (L2, MW= 162.4 g/mol) emulsions ^(34),(35). Given the short elimination time in vivo for L2 based emulsions ⁽³⁴⁾ (T_1/2 = 15 ± 1 hr in rat muscle) and the correlation of stability of nanoscale “oil-in-water” type emulsions with the molecular weight of the “oil” phase ⁽³⁶⁾, it is essential to consider the emulsions longer chain siloxanes for biomedical applications where oximetry is needed.

In this work, we engineer pO₂ Reporter Composite Hydrogels (PORCH) (Figure 1) to enable spatiotemporal measurement of local oxygen tension within macroencapsulation devices using linear siloxanes octamethyltrisiloxane (L3, MW= 236.5 g/mol) and tetradecamethylhexasiloxane (L6, MW = 459 g/mol). We evaluate the retention of oximetry probe emulsions of varied composition within agarose, alginate, and PEG-based macroencapsulation devices, and we assess the cytotoxicity of L3 and L6 emulsions to cells encapsulated within the devices. To further isolate oximetry probes from cells and reduce cytotoxicity, we engineered a poly(dimethyl siloxane) (PDMS) microbead approach and evaluate cytotoxicity of co-encapsulated cells. Finally, we demonstrate the ability of the various PORCH devices to generate quantifiable oxygen measurements in vitro and compare the efficacy and accuracy of emulsion and microbead-based approaches. Here we demonstrate the feasibility of spatial resolution of oxygen within cell-laden macroencapsulation devices suitable for future application in in vivo spatiotemporal oxygen measurement.

Figure 1. Engineering pO₂ Reporter Composite Hydrogels (PORCH) to detect oxygen within macroencapsulation devices in vitro and in vivo.

Schematic illustrating two approaches to incorporating siloxane oximetry probe within macroencapsulation devices: (1) entrapment of siloxane emulsion within the hydrogel matrix, or (2) absorption of siloxane within solid PDMS microbeads that are entrapped within the hydrogel matrix. After fabrication, PORCH are characterized for compatibility with encapsulated cells and capacity to generate quantifiable oxygen measurement.

2. Materials and Methods

2.1. Materials

All chemicals were obtained from Sigma-Aldrich and all cell culture reagents were obtained from ThermoFisher unless otherwise noted.

2.2. Siloxane nanoemulsion fabrication and characterization via dynamic light scattering

Siloxane emulsions (L3 or L6-based) were prepared following the previously reported protocol ^(34),(35) using polyethylene glycol (15)-hydroxystearate (HS-15) as surfactant. L3 and L6 were obtained from Sigma-Aldrich and Combi-Blocks, respectively. Two surfactant concentrations were tested (2% and 5%) to assess cytotoxicity. Briefly, Nile Red/siloxane was prepared by first dissolving Nile Red into ethanol at the concentration of 3.9 mg/mL. 41 μL of the HS 15/siloxane mixture was then dissolved in 2 mL of siloxane. HS 15 was heated to liquid state before mixing with deionized water (DI-H₂O) in 1:27.5 (for 2% surfactant) or 1:11 (for 5%) ratio at 70 °C for 2 min. Nile Red/siloxane solution was then added dropwise to the DI-H₂O/HS 15 solution. The mixture was heated at 70 °C for 15 min with stirring, followed by sonication for 45 min (15 min intervals at 100% duty cycle) using Omniruptor 4000 Ultrasonic Homogenizer (Omni International, Tulsa, OK). The resultant nanoemulsion was filtered using a syringe filter (pore size: 0.22 μm) 11 times. Nanoemulsions were characterized by dynamic light scattering (DLS) (Delsa Nano Particle Size Analyzer, Beckman Coulter, Pasadena, CA). The average modal diameter was determined based on three repeated measurements per sample.

2.3. pO₂ Reporter Composite Hydrogel (PORCH) fabrication

2.3.1. PEG-based composite hydrogel

20 kDa four-arm PEG-maleimide (Laysan Bio) powder was combined with sterile Dulbecco’s phosphate buffer solution without magnesium chloride and calcium chloride (DPBS −/−) in a 10% solution. RGD (GRGDSPC, Genscript) powder was combined with DPBS −/− to form a 0.125 mg/μL solution. The 5% PEG solution was functionalized with RGD (1mM) to enhance cell attachment. PEG was mixed with RGD at a 1:1 ratio and vortexed until homogenous. MR oximetry probe was then added to the PEG/RGD solution at various concentrations for various experiments (5%, 10%, 20%, and 50% v/v). PEG and RGD concentrations were adjusted to account for an extra dilution caused by the addition of MR oximetry probe to the solution. PEG-DT powder was combined with DPBS −/− at a concentration of 0.0225 mg/μL. PEG-DT was vortexed until homogenous. The PEG/RGD/MR oximetry probe solution and PEG-DT solutions were combined in a cylindrical mold in a 1:1 ratio to form MRI oximetry probe composite composed of PEG hydrogel at a final concentration of 5% w/v.

2.3.2. Alginate-based composite hydrogel

Slow-gelling alginate was fabricated using 3% (w/v) alginate, 30 mM calcium carbonate (CaCO₃), and 100 mM gluconic-delta-lactone acid (GDL). Alginate and GDL solutions were made in DPBS −/− and stored at 4°C for 24 hours to allow for full solubilization. After 24 hours, CaCO₃ was added to the alginate solution and mixed well. MR oximetry probe was added to the GDL solution at various concentrations (5%, 10%, 20%, and 50% v/v). The concentration of the gel components was adjusted to account for the MR oximetry probe dilution at 50% concentrations. GDL/MR oximetry probe solution was added in a 1:1 ratio to form the final alginate gel (1.5% w/v alginate, 15 mM CaCO₃, 50 mM GDL) in a cylindrical mold at all MR oximetry probe concentrations. The alginate hydrogel was washed with a 1.5% BaCl₂ solution fabricated using PBS −/− to further crosslink the gel.

2.3.3. Agarose-based composite hydrogel

Low-gelling temperature agarose powder was combined with sterile DPBS with magnesium chloride and calcium chloride (DPBS +/+) (2% w/v) in a 50 mL beaker in a double boiler system. The solution was stirred with a stir bar at 600 rpm and 100°C until powder was completely incorporated into DBPS +/+. For experiments that required a 50% concentration of nanoprobe a 4% w/v agarose solution was made to account for the dilution caused by the addition of MR oximetry probe. For 5%, 10%, 20%, and 50% MR oximetry probe concentrations in the agarose hydrogel, 2% w/v agarose solution was used. The MR oximetry probe was mixed into the hydrogel solution by combining both solutions into an Eppendorf tube. The resulting mixture was vortexed to ensure homogeneity and allowed to gel in a cylindrical mold.

2.4. MR oximetry probe retention measurement

Oximetry probe composite hydrogels (composed of 1.5% alginate, 5% PEG, and 2% agarose, respectively) incorporated with L6 emulsion (fabricated with 2% surfactant and 5% surfactant) were evaluated for probe retention. The oximetry probe was added at concentrations (v/v) 5%, 10%, 20%, and 50% as described in section 2.4 within the various hydrogels. Hydrogels were incubated in Dulbecco’s phosphate buffer solution without magnesium chloride and calcium chloride (DPBS −/−) for 5 days. The released oximetry probe (Nile Red-tagged) from the hydrogel was determined by measuring the fluorescence of the Nile Red in DPBS −/− solution each day at the same time. To further evaluate oximetry probe retention, fluorescent images of the various hydrogels were taken each day at the same time. Average fluorescence of each hydrogel was calculated by drawing a line across the diameter of the hydrogel in ImageJ and averaging the fluorescence values along that line.

2.5. PDMS microsphere fabrication and MR oximetry probe loading and retention

The microsphere mixture was prepared as previously described ⁽³⁷⁾ by mixing PDMS Part A and Part B (Cat. No. 04019862, Sylgard) at a 4:1 ratio in both a centrifuge and manually. Once the components were thoroughly mixed, it was injected into 120°C PEG 8000 liquid at 1 mL/min using a syringe pump (Cat. No. NE-300, New Era Pump Systems) while the PEG 8000 was stirred at a rate of 60 RPM by a mixer (Cat. No. 04555–00, Cole-Parmer). The microbeads were cured at 100°C in liquid PEG 8000 for 1 hour before being washed with Pluronic F-127 solution and collected using sieves to ensure the bead size ranged from 67 to 253 micrometers. The microbeads were sterilized and then loaded with OMTSO nanoprobe in a sterile environment. The size of the microbeads was confirmed using image analysis and manual size annotation on an ECHO Revolve microscope. Oximetry probe retention in the microspheres was measured on day 0, day 1, day 2, day 3, day 4, day 5, day 7, day 15, and day 19. The microbeads were incorporated into hydrogel materials by adding 0.75, 1.5, and 3 mg of microbeads corresponding to concentrations of 5%, 10%, and 20% w/v in a cylindrical mold.

2.6. PORCH cytotoxicity assessment

PEG (5% w/v), agarose (2% w/v), and alginate (1.5% w/v) hydrogels were fabricated as described above with MR oximetry probe concentrations of 0%, 5%, 10%, and 20%. MR oximetry probes assessed were L3 emulsion, L6 emulsion with 5% surfactant, and L6 emulsion with 2% surfactant. The beta cell line INS-1E was cultured in RPMI media supplemented with 10% FBS, 1% penicillin-streptomycin, 1% HEPES buffer solution, 1% sodium pyruvate, and 0.09% β-mercaptoethanol. INS1E cells were passaged, counted, and aliquoted to ensure a uniform cellular density of 7500 cells/μL to mimic a density of 5 islet equivalents (IEQ)/μL for each hydrogel. The 15 μL hydrogel constructs were incubated in well plates under standard conditions (5% CO₂, 20% O₂). As a control, 112,500 cells were seeded in 48 well-plates to correspond to the number of cells in each hydrogel construct. These free cells were exposed to L3, L6 with 2% surfactant, and L6 with 5% surfactant at 0%, 5% (20 μL), 10% (40 μL), and 20% (80 μL) w/v in 400 μL of RPMI media. The cells were cultured under standard conditions (5% CO₂, 20% O₂). For metabolic activity assessment, at 48 hours post-construct fabrication, the alamarBlue assay was mixed in a 1:10 ratio with complete RPMI media per the manufacturer’s instructions and cultured with hydrogels for 4 hours prior to fluorescence measurement (exc/em 560/590 nm). A hydrogel with no cells was fabricated for each MR oximetry concentration and hydrogel type to account for any lingering fluorescence in the alamarBlue media potentially caused by probe being released into the media. Any background fluorescence was subsequently subtracted from the final alamarBlue fluorescent data. The unencapsulated cells were washed three times with PBS to ensure no MR oximetry probe was collected with the alamarBlue measurements. Live/dead confocal imaging was performed at 48 hours. Unencapsulated cells were stained with calcein AM and ethidium homodimer for 30 minutes. Encapsulated cells were stained with calcein AM for 30 minutes and TOTO-3 Iodide for 15 minutes. Images were collected with a Leica SP8 White Light Laser Confocal microscope.

2.7. MR oximetry measurements

MRI oximetry experiments were conducted on a preclinical 7 T scanner (Bruker BioSpec) with a rat volume coil. R₁ mapping was carried out using the PISTOL sequence as previously described ^(30),(31). Briefly, the PISTOL sequence consists of a combination of pulse-burst saturation recovery with frequency-selective spin-echo (SE) excitation of the siloxane protons followed by an echo planer imaging (EPI) sequence for readout. Repetition time (TR) was varied in the range of 0.1—55 seconds (for L3) or 0.1—20 seconds (for L6) to sample 16 data points within the T₁ relaxation curve with variable number of averages ⁽³¹⁾. The total scan time is 3 min 44 s for L3 and 2 min 25 s for L6. Other parameters used for PISTOL scanning were as follows: echo time (TE) = 34.4 ms, field of view (FOV) = 3 cm × 3 cm, matrix size = 64 × 64, slice thickness = 5 mm. Scout images were acquired using a multi-slice, multi-echo (MSME) pulse sequence with the following parameters: TR = 1000 ms, TE = 10 ms, FOV = 3 cm × 3 cm, matrix size 128 × 128, slice thickness 1 mm.

PORCH composite samples (disk shape, diameter and height were approximately 6 mm and 2 mm, respectively) were bubbled in PBS for at least 40 min using nitrogen or air prior to being placed into a custom sample holder that allows maintaining the temperature at 37 °C and controlling the oxygen concentration. After bubbling, the PORCH samples and the bubbled PBS were transferred into the sample holder inside a glove bag filled with gas (air or nitrogen) and placed into the MRI scanner. Samples were maintained under nitrogen or air (corresponding to pO₂ = 0 and 159.6 torr, respectively) before and during the PISTOL scans.

Image reconstruction and analysis were carried out using a custom-made code on MATLAB R2021a (MathWorks, Natick, MA).

Inverse Fourier transform was applied to the filtered k-space data to generate the images used for subsequent analysis. The images were generated by first loading the k-space data to MATLAB and denoising by applying a Fermi filter (matrix size 64 × 64)

$$F\left(v\right)={(1+\mathrm{e}\mathrm{x}\mathrm{p}|\frac{v-r_f}{w_f}\left|\right)}^{-1}$$

The radius $r_f$ and edge-width $w_f$ of the filter were set to 40 pixels and 1 pixel, respectively and v is co-ordinate in pixels with respect to the center of k-space. The spin-lattice relaxation rate R₁ (= 1/T₁) at each pixel was computed by fitting the signal intensities (16 data points) to a single exponential, 3-parameter magnetization recovery equation.

$$M\left(t\right)=M\left(0\right)+\left(M_0-M\left(0\right)\right)*(1-{exp}^{-t*R_1})$$

Where $M\left(t\right)$ is magnetization at time $t,M\left(0\right)$ is the initial magnetization after the RF pulse, and $M_0$ is the magnetization at equilibrium state.

Error range of the R₁ estimation was computed using the half width of the 95% confidence interval (CI) of the estimated R₁ value, using the equation below.

$$\mathit{\text{Error}}\mathit{\text{range}}=\frac{\left(\mathit{\text{upper}}\mathit{\text{limit}}of95\%CI\right)-\left(\mathit{\text{lower}}\mathit{\text{limit}}of95\%CI\right)}{\mathit{\text{estimated}}{R}_1\mathit{\text{value}}*2}$$

Pixels that reported error range larger than 0.2 were considered unreliable and were removed from the analysis. The resultant R₁ maps were converted to pO₂ maps using the predetermined calibration curves (i.e., the equations describing the relationship between R₁ and pO₂). Calibration curves for each PORCH type was determined by fitting the mean R₁ values obtained under nitrogen and air conditions (corresponds to pO₂ = 0 and 159.6 torr, respectively) at 37 °C to the following model ⁽³⁸⁾:

$$R_1=A+B*{pO}_2$$

2.8. Statistical analysis

Fluorescence intensity data for emulsion retention in hydrogels was analyzed via two-way ANOVA with a Tukey’s multiple comparison test to compare oximetry probe concentrations, or Sidak’s multiple comparisons test to compare between surfactant composition. All alamarBlue metabolic activity data was analyzed by non-parametric one-way ANOVA with Kruskal-Wallis multiple comparisons against 0% control. L3-loaded PDMS bead diameters were analyzed using ordinary one-way ANOVA with Dunnett’s multiple comparisons test. L6 and L3 pixel pO₂ quantification were analyzed by one-way ANOVA and Tukey’s multiple comparisons test. Emulsion droplet diameter and comparison of pO₂ quantification between hydrogels of same condition analyzed by unpaired t-test. All statistics were performed in GraphPad Prism 9.1.

3. Results

3.1. Engineering siloxane emulsion-based PORCH macroencapsulation devices

We have previously identified a range of siloxane probes (L3 and L6) capable of resolving oxygen concentrations with high sensitivity and resolution using the PISTOL MR technique ^{(30),(32),(33)}. We first sought to develop a PORCH system that would enable the measurement of spatiotemporal oxygen changes within hydrogel macroencapsulation devices in vitro and in vivo. To generate these composite hydrogels, we first investigated the retention of L6 siloxane probe nanoemulsions entrapped within common hydrogel encapsulation matrices alginate, agarose, and PEG (Figure 2).

Figure 2. Reduced surfactant concentration improves siloxane emulsion retention in common cell encapsulation hydrogels.

A) Schematic illustrating the experimental setup to measure siloxane oximetry probe retention in PORCH. B) Measurement of L6 nanoemulsion retention within PEG (5% w/v), agarose (2% w/v), and alginate (1.5% w/v) hydrogels (n = 6/group) when nanoemulsions were fabricated with 2% and 5% surfactant at varied siloxane oximetry probe concentrations (0 – 50% v/v). C) Fluorescence imaging (red = Nile Red, grey = brightfield (BF)) of nanoemulsion retention within hydrogels on day 4 post encapsulation (n = 3–6/group). D) Quantification of fluorescent images in (C). Data were pooled from three independent experiments. Fluorescence intensity (D) data was analyzed via two-way ANOVA with a Tukey’s multiple comparison test to compare oximetry probe concentrations, or Sidak’s multiple comparisons test to compare between surfactant composition. Error bars = SEM. Scale bars = 100 μm. * P < 0.05, ** P < 0.005, ^### or *** P < 0.0005, ^#### or **** P < 0.0001.

We generated siloxane emulsions using the L6 probe and HS 15 surfactant at surfactant concentrations of 2 and 5% (v/v). We have previously successfully used 5% surfactant siloxane emulsions for PISTOL MR oximetry ^(34),(35) but sought here to evaluate the efficacy of a lower surfactant concentration as we hypothesized that excess surfactant could reduce co-encapsulated cell viability. We first evaluated the impact of surfactant concentration on emulsion droplet size and found average droplet modal diameters of 245.8 ± 21.8 nm and 199.2 ± 35.1 nm for 2 and 5% surfactant concentrations, respectively (Supplemental Fig. 1A). Unsurprisingly, reducing surfactant concentration results in a statistically significant increase in emulsion droplet size (P = 0.004), but we found no change in emulsion droplet size stability up to 5 weeks (Supplemental Fig. 1B). Similarly, L3 emulsions fabricated with 2% surfactant demonstrated comparable droplet size to L6 immediately after fabrication (Supplemental Fig. 1A, bottom); however, after approximately 2 hr, we observed a second peak emerge in L3 DLS measurements (Supplemental Fig. 1B, bottom), indicating poorer stability over time than L6 nanoemulsions.

We next examined the retention of L6 nanoemulsions fabricated with 2% and 5% surfactant within common cell encapsulation hydrogels alginate (1.5% w/v), agarose (2% w/v), and PEG (5% w/v), at L6 nanoemulsion concentrations of 5, 10, 20, and 50% (v/v). The L6 nanoemulsion contained a NileRed reporter dye to enable quantitative detection of emulsion escape from the hydrogels (Figure 2A). Hydrogels containing L6 nanoemulsions were incubated in physiological conditions (37 °C) and fluorescence in the supernatant surrounding the hydrogels was measured over 4 days (Figure 2B). For all hydrogels and nanoemulsion concentrations, an initial burst release was observed on day 1 followed by limited nanoemulsion escape. We additionally used fluorescent imaging on day 4 post-release to confirm nanoemulsion retention within the hydrogels (Figure 2C). Quantification of fluorescence retained within the hydrogels on day 4 demonstrated statistically significant dose-dependent increases in fluorescence with increasing siloxane concentration in the both surfactant groups for all hydrogels up to 20% v/v nanoemulsion concentration, and a significant drop in fluorescence intensity at the highest concentration of 50% v/v in PEG and alginate groups (Figure 2D). We repeated retention studies within PEG, alginate, and agarose hydrogels with L3 nanoemulsions fabricated with 2% surfactant and observed the same trends of minimal burst release on day 1 (Supplemental Fig. 2A), and dose-dependent increases in fluorescence intensity within hydrogels (Supplemental Fig. 2B–C). Day 1 burst release fluorescence measurements correlate strongly with oximetry probe loading concentration (Supplemental Fig. 3A–B), with PEG exhibiting substantially greater slope for all nanoemulsion compositions (Supplemental Fig. 3C).

3.2. PORCH macroencapsulation devices reduce cytotoxicity of L6 oximetry probe

We next sought to evaluate the viability and metabolic activity of a model beta cell line (INS1E) exposed to oximetry probe alone or within emulsion PORCH macroencapsulation devices (Figure 3). We investigated the cytotoxicity of L6 (2% and 5% v/v surfactant) and L3 nanoemulsions (2% v/v surfactant) at various concentrations (0, 5, 10, and 20% v/v) entrapped within a 2% (w/v) agarose, 1.5% (w/v) alginate, and 5% (w/v) PEG hydrogels, as well as within the media of 2D cultured cell controls (“Free” group) (Figure 3A). We hypothesized (1) that oximetry probe emulsion entrapment within the hydrogel would limit cell exposure to the oximetry probe emulsion and mitigate potential cytotoxic effects and (2) that lower surfactant concentrations would yield greater cell viability. L6 nanoemulsions fabricated with 5% surfactant and introduced to free cells at concentrations from 5–20% v/v resulted in significant reductions in cell metabolic activity at 48 hours (Figure 3B, middle) and significant impacts on cell viability (Figure 3C, middle) in unencapsulated groups. L6 emulsion preparations with 2% surfactant exhibited significantly less cytotoxicity for 5% (P<0.0001) and 10% (P = 0.0136) emulsion concentration in unencapsulated groups compared to 5% surfactant preparations with 5% and 10% emulsion concentration respectively (Figure 3B and 3C, left). Conversely, L6 nanoemulsion entrapment within alginate, agarose, and PEG hydrogels mitigated cytotoxic effects at any surfactant concentration and any emulsion concentration within hydrogels. By contrast, L3 nanoemulsions (2% surfactant) resulted in significant impacts to metabolic activity and viability in both encapsulated and unencapsulated groups (Figure 3B–C, right).

Figure 3. L6 emulsion entrapment within PORCH reduces reporter probe cytotoxicity.

A) Schematic illustrating experimental setup to evaluate reporter probe emulsion cytotoxicity. B) Cell metabolic activity was evaluated at 48 hours (n = 9/group). C) Live/dead imaging of cell viability within hydrogel formulations (live=green, dead=magenta) (n = 3/group). Data was pooled from three independent experiments. Metabolic activity was analyzed by non-parametric one-way ANOVA with Kruskal-Wallis multiple comparisons against 0% probe control. Error bars = SEM. Scale Bar = 100 μm. ***P<0.001, **** P<0.0001.

3.3. PDMS microbead PORCH system reduces L3 oximetry probe cytotoxicity

Given the higher sensitivity and greater range of the T₁ relaxation times of L3 relative to L6 for the detection of oxygen via MR ⁽³⁰⁾, we sought to engineer a PORCH approach that minimized the L3 reporter probe cytotoxic effects (Figure 4). To achieve this, we hypothesized that entrapment of pure L3 within PDMS microbeads and distribution of microbeads throughout the hydrogel would enable L3 retention within the hydrogel while limiting cytotoxicity (Figure 4A). Microbeads were first fabricated from PDMS via an emulsion method (Figure 4B). Image analysis was used to evaluate the size range of PDMS microbeads both pre- and post-L3 loading (Figure 4B–C) as a measure to quantify L3 retention within PDMS microbeads. L3-loaded PDMS beads exhibit a significantly greater average diameter one day post-loading (164 ± 48 μm) relative to control PDMS beads (98 ± 32 μm). By day 19, the average diameter of L3-loaded beads had decreased to 94 ± 52 μm. We attribute the transient change in PDMS bead size after L3 loading to bead swelling as L3 is absorbed, and bead relaxation as the volatile L3 gradually escapes the beads over a period of approximately 2 weeks.

Figure 4. L3-loaded PDMS microbead PORCH approach abrogates L3 cytotoxicity.

B) Images of PDMS microbeads pre- and post-L3 loading. C) Quantification of L3-loaded PDMS bead diameter over time. Diameters pooled from two independent experiments. Data analyzed by ordinary one-way ANOVA with Dunnett’s multiple comparisons test. **** P < 0.0001. D) Metabolic activity of unencapsulated cells or L3-PDMS PORCH-encapsulated cells. Data analyzed by non-parametric one-way ANOVA with Kruskal-Wallis multiple comparisons against 0% control. * P<0.05, **** P<0.0001. E) Live/dead imaging of free or PORCH-encapsulated cell viability (live=green & dead=magenta). Scale Bar = 100 μm.

Finally, we evaluated the cytotoxicity of L3-entrapped microbeads at various concentrations of 0%, 5%, 10%, and 20% w/v on INS1E via metabolic activity and viability imaging at 48 hours post-exposure (Figure 4D–E). L3 entrapment in PDMS beads significantly reduced INS1E cytotoxicity relative to emulsion, as direct incubation of L3 beads with cells resulted in significant loss of metabolic activity only at 10% (57% metabolic activity) and 20% probe concentrations (14% metabolic activity) (Figure 4D). Conversely, entrapment of beads within hydrogels completely mitigated cell death, even at the highest L3 concentration. Cells exhibited significantly higher viability when co-cultured and encapsulated in hydrogels with L3-loaded PDMS beads (Figure 4E) relative to L3 in solution (Figure 3C). Cumulatively, this data demonstrates that the PDMS bead-PORCH approach is suitable for in vitro and in vivo oximetry studies with encapsulated cells.

3.4. In vitro PISTOL MR characterization of PORCH macroencapsulation devices

Following the successful fabrication of L6 and L3 PORCH macroencapsulation devices that exhibit minimal cell toxicity, we next sought to evaluate the efficacy of our emulsion and PDMS bead PORCH strategies for the quantification of spatial oxygen distribution via MR imaging in vitro (Figure 5). L6 emulsion (2% surfactant, 20% concentration) exhibited the greatest retention and minimal cytotoxicity in alginate and agarose hydrogels, and these compositions were chosen to compare against the L3 microbead condition (Figure 5A). L3 emulsion-loaded hydrogels (2% surfactant, 20% siloxane probe concentration) were included as a control for the L3 microbead system. For consistent geometry, hydrogels were fabricated within cylindrical molds. PORCHs were equilibrated under ambient air (pO₂ ≈ 160 torr) and depleted oxygen conditions (N₂, pO₂ = 0 torr) at 37 °C and PISTOL scans were performed to obtain R₁ maps, which were subsequently converted to pO₂ maps using the calibration curves predetermined for each PORCH type (Figure 5B). Despite identical hydrogel geometries, the L3 microbead condition exhibited noticeably fewer measurable pixels via PISTOL (Figure 5B, Supplemental Figure 4); regardless, all hydrogel conditions demonstrated pO₂ measurements within the expected range for air or oxygen-depleted conditions. The mean pO₂ values quantified from all PORCH types under ambient air (152.9 – 162.5 Torr) and depleted oxygen conditions (−6.2 – 3.3 Torr) were within ± 10 Torr of the expected value (0 Torr under N₂ and 159.6 Torr under air), with high repeatability between replicates (Supplemental Figure 5), which demonstrates the accuracy of this approach (Figure 5C). Precision of the PORCH approach was assessed by comparing the interquartile range (IQR). Two-way ANOVA revealed that there were no significant effects from the type of hydrogels used. When quantifying pO₂ under N₂, the type of pO₂ probes exhibited a slight effect on the IQR (F value = 6.4), where the IQR is slightly lower for the L3 emulsion group (range 11.6 – 15.9) compared to L3 PDMS beads and L6 emulsion (range 20.2 – 30.5 and 17.7 – 40.3, respectively). Ranges of IQR values under air condition were 24.8 – 45.7, 30.9 – 35.1, and 25.6 – 37.8 for L3 beads, L3 emulsion, and L6 emulsion, respectively.

Figure 5: Evaluation of PISTOL MR oximetry accuracy in agarose and alginate PORCH constructs using the emulsion or microbead approach.

A) Schematic illustrating experimental procedure. (B) MR imaging of agarose and alginate L6 and L3 emulsion (2% surfactant, 20% oximetry probe concentration) and L3 microbead PORCHs under ambient (Air) or oxygen-depleted (N₂) conditions. C) Quantification of oxygen measurements within agarose (left) and alginate (right) PORCH devices under ambient (Air) or oxygen-depleted (N₂) conditions. Each data point represents a measured pixel, pixels were pooled from two independent hydrogel measurements. Data were analyzed by one-way ANOVA and Tukey’s multiple comparisons test. * P < 0.05, ** P < 0.01, *** P < 0.0005. $ P < 0.05 vs. agarose L3 bead air condition. Scale bar = 2 mm.

4. Discussion

In this work, we engineer two methods to incorporate siloxane MR oximetry probes within macroencapsulation devices for spatiotemporal oxygen quantification (Figure 1). In our first method, we entrap siloxane emulsions fabricated with two surfactant concentrations (2 and 5%) within common encapsulation hydrogels alginate, agarose, and PEG and quantified oximetry probe retention within the hydrogel matrix (Figure 2). We found a dose-dependent increase in burst release with increasing nanoemulsion concentration (Supplemental Fig. 3), and the most substantial burst release of siloxane emulsions was observed in the PEG hydrogel group, and particularly in 5% surfactant conditions relative to 2% surfactant conditions. Overall, limited nanoemulsion release was detected from 2% surfactant nanoemulsion groups, particularly from the agarose and alginate hydrogels, and 5% surfactant emulsions exhibited inconsistent fluorescence retention in all hydrogels at the highest concentration of nanoemulsion (50% v/v). These observations are likely due to the impact of surfactant concentration on nanoemulsion size (Supplemental Fig. 1), where larger droplet-size nanoemulsions (2% surfactant group) may be more efficiently retained within the hydrogel matrices. Overall, nanoemulsions with a larger droplet size (2% surfactant) exhibited greater overall retention within hydrogels, and this retention was not correlated with literature-reported hydrogel pore size (5% w/v PEG = 15–35 nm ⁽¹⁰⁾, 1.5% w/v alginate = 10–50 nm ⁽³⁹⁾, agarose = 200–500 nm ⁽⁴⁰⁾). Poor emulsion retention in hydrogels at higher emulsion concentration of 50% (v/v) and higher surfactant concentration of 5% (v/v) may be due to emulsion and surfactant interference with hydrogel crosslinking, potentially increasing hydrogel pore size and enabling emulsion escape.

Next, we found that reducing surfactant concentration reduced cytotoxicity of L6 nanoemulsions incubated with free cells, and L6 nanoemulsion entrapment within hydrogels completely mitigated cytotoxicity; by contrast, the lowest surfactant concentration (2%) for L3 nanoemulsion resulted in significant cytotoxicity regardless of entrapment or free exposure conditions (Figure 3). This degree of cytotoxicity of L3 was expected, as previous studies have demonstrated significant cytotoxicity of L3 with a retinal cell line ⁽⁴¹⁾. This data demonstrates that L6 PORCH are suitable for in vitro and in vivo oximetry studies with encapsulated cells, and that L3 PORCH requires an alternative engineering approach to minimize cytotoxicity. As L3 has a greater sensitivity and dynamic range than L6 for the detection of oxygen with MR, we sought to engineer a method to use L3 for MR oximetry without compromising co-encapsulated cell viability. Consequently, we engineered the microbead PORCH approach (Figure 4), where L3 oximetry probe is directly loaded in PDMS microbeads, which ameliorated the severe cytotoxicity observed with the L3 emulsion PORCH approach. Measurement of L3-loaded PDMS microbead size was used as an indirect measure of L3 retention in the microbeads, and measurement of bead diameter over greater than two weeks indicates that L3 retention within the microbeads may be transient.

Finally, we evaluated the accuracy of emulsion and microbead PORCH methods for the detection of oxygen tension in vitro (Figure 5). We observed a lower number of pixels in the microbead PORCH method (Supplemental Fig. 4), though oxygen quantification of both approaches had comparable accuracy and repeatability (Supplemental Fig. 5). The lower number of pixels which were able to achieve a reliable pO₂ estimation in L3-loaded microbead PORCHs compared to emulsion-loaded PORCHs (Supplemental Fig. 4) could be due to the efficiency of dispersing the pO₂ probe-loaded microbeads within the PORCH. In the microbead method, 100–300 μm diameter PDMS beads were challenging to disperse evenly within the hydrogel due to their hydrophobicity, while 200–300 nm diameter emulsion droplets distributed evenly in the emulsion PORCH. When the L3-loaded microbeads are unevenly dispersed within the PORCH, pixels which are predominantly occupied by the beads can capture the siloxane signal with high SNR allowing a reliable curve fitting to estimate the pO₂, while pixels that are partially occupied by the beads result in lower SNR leading to decreased reliability in the pO₂ estimation. Additionally, the observed difference between L3 and L6 emulsions could be due to the differences in range of the spin lattice relaxation time T₁ (= 1/R₁) under different pO₂ conditions for L3 and L6. T₁ of the L3 can change from around 3.1 seconds (under air, pO₂ 160 torr) to 6.6 seconds (under N₂, pO₂ 0 torr), while the range of L6 is approximately 2.4 to 4.0 seconds under corresponding oxygen conditions. The larger range of L3 facilitates the detection of smaller pO₂ differences under different conditions. Under air condition, where T₁ values between probes are similar, IQR values were comparable among the three PORCH types. In general, while comparing the performance of multiple pO₂ reporter molecules, a useful indicator is the ratio η of the slope to the hypoxic intercept of the R₁ vs pO₂ calibration curve ⁽⁴²⁾. As reported previously ⁽³⁰⁾, η for linear siloxanes generally decreases with chain length with small differences in pO₂ sensitivity. It is possible that if additional signal averaging had been employed for L6 PORCH devices such that the total data acquisition time was similar between L3 and L6 devices, smaller differences in curve-fits and IQR would have been observed, but this was not tested in the current study. While we have applied k-space filtering to reduce image noise prior to parametric fitting and exclude pixels with error range greater than 0.2, further enhancements in data quality are possible by adopting an approach (such as used by Yang et al. ⁽⁴³⁾) where T₁ recovery time points that deviate by 3 standard deviations from monotonicity in the fit curve are eliminated, followed by refitting. Additionally, faster siloxane T₁ mapping approaches using a Look-Locker approach ⁽⁴⁴⁾ could be used in conjunction with more signal averaging to improve data quality.

5. Conclusion

We successfully engineered PORCH cell encapsulation hydrogels incorporating and retaining ¹H MR oximetry probes to enable the spatial quantification of oxygen within hydrogel macroencapsulation devices. We demonstrated that L6 emulsions can be entrapped and retained within alginate and agarose PORCH devices at up to 20% (v/v) concentration with minimal cytotoxicity, and can effectively be used to quantify oxygen within macroencapsulation devices. We also demonstrate that the highly cytotoxic lower molecular weight L3 can be incorporated and retained within hydrogels using a PDMS microbead approach, which eliminated cytotoxicity concerns; however, the microbead PORCH method reduces oximetry spatial resolution relative to the emulsion PORCH approach. Future studies will use the emulsion PORCH approach to quantify spatiotemporal oxygen changes within cell macroencapsulation devices in vitro and in vivo. The incorporation of ¹H MR oximetry probes may enable prognostic imaging of transplant devices and early detection of implant engraftment or failure. Separately, the assessment of pO₂ may enable the design of improved devices via optimal control of geometry, cell density, and implantation site.

Data Availability.

The raw or processed data required to reproduce these findings are available upon request.