Genipin-crosslinked double PLL membranes overcome the strength-diffusion trade-off in cell encapsulation without compromising biocompatibility

Abstract

Cell microencapsulation technologies allow non-autologous implantation of therapeutic cells for sustained drug delivery purposes. The perm-selective membrane of these systems provides resistance to rupture, stablishes the upper molecular weight limit in bidirectional diffusion of molecules, and affects biocompatibility. Thus, despite being a decisive factor to succeed in terms of biosafety and therapeutic efficacy, little progress has been made in its optimization so far. Here we show that, compared to other usually used coating designs, genipin-crosslinked double poly-L-lysine (GDP) membranes are able to simultaneously improve mechanical and mass-transport properties of the microcapsules, without causing any significant increase in the foreign body response when implanted in vivo. In particular, we show that GDP membranes confer capsular integrity under high pressures, both internal and external. Furthermore, this membrane design allows for more efficient bidirectional diffusion of molecules in the 20–40 kDa range while preserving the molecular weight cut-off required for exerting an effective immunobarrier. These findings may also be useful for optimizing the membrane characteristics of multiple drug delivery systems.

1. Introduction

Cell microencapsulation is a technology with great potential for the treatment of various chronic and degenerative disorders (Gonzalez-Pujana et al., 2017; Gonzalez-Pujana et al., 2018). The cells of therapeutic interest are included in biocompatible spherical hydrogel matrices, normally surrounded by polymeric perm-selective membranes. Once implanted in the body, this cell-shielding strategy allows the access of nutrients and oxygen to the interior, as well as the diffusion to the exterior of different therapeutic products produced by the encapsulated cells. The ultimate goal is always to protect allo- or xenogeneic cells from the host immune response. Thus, encapsulated cells are used to secrete hormones, neurotransmitters and growth factors — among others — for long time periods, reducing or even avoiding immunosuppression protocols. In recent years, the advances made in the field of cell microencapsulation have turned this technology into more than just a promise, leading to several clinical trials supported by significant investments from both the public and the private sectors (Lopez-Mendez et al., 2021; Lee et al., 2021).

However, one of the aspects that critically affects the biosafety and durability of the microcapsule-based therapeutic strategies is the low mechanical resistance of the semipermeable membranes that cover them (Lopez-Mendez et al., 2021). Ideally, the microcapsule coating should be able to resist shear and compression forces coming from the exterior, such as those occurring during manipulation or those exerted by surrounding tissues once implanted. Additionally, microcapsule coating should withstand stretching forces originated from the inner core, such as those generated by enclosed cells. The classic membranes based on a coating of poly-L-lysine (PLL) or poly-L-ornithine (PLO) do not withstand abrupt changes in osmotic pressure and are hardly capable of containing a possible erratic and uncontrolled growth of the cellular content of the capsule (Gonzalez-Pujana et al., 2017; Hajifathaliha et al., 2020). Therefore, it is fundamental to establish new coating designs that ensure capsular integrity, avoid the release of the encapsulated cells with the subsequent tissue invasion and increase implant lifespan.

To date, many researchers have attempted to reinforce capsule coatings following different strategies, including layer-by-layer designs, covalent crosslinking, conformal coating or core–shell spherification (Hajifathaliha et al., 2020; Johnson et al., 2019; Gattás-Asfura et al., 2014; Ahn et al., 2021; Gattás-Asfura et al., 2020; Harrington et al., 2021). Unfortunately, it is widely assumed that increasing membrane strength — e.g. by membrane thickening — adversely affects the rest of the functional properties of the capsule, including mass transport capacity and biocompatibility. In fact, the main strategy traditionally adopted to resolve such incompatibility has been to sacrifice the ideal desirable quality ranges for these opposing competing parameters in order to achieve an optimal balance between them. In this sense, some studies have been able to decouple mechanical properties and permeability at hydrogel matrix level, but — to the best of our knowledge — no one has focused so far on the semipermeable membrane (Cha et al., 2010; Morton et al., 2023; Cha et al., 2011; Kim et al., 2017).

As a possible solution, in the present study we investigated the performance of our genipin-crosslinked double poly-L-lysine (GDP) membranes, which have previously shown zero cytotoxicity with encapsulated cells and great utility for in vivo imaging (Santos-Vizcaino et al., 2018). To this end, we performed a complete evaluation of the newly developed membrane resistance in two steps. To measure the resistance to internal pressures, we carried out osmotic explosion tests; and to evaluate the resistance to external pressures, we used a novel microfluidic test and atomic force microscopy (AFM). Afterwards, we tested the bidirectional permeability of the membranes: We studied the diffusion of dextrans with different molecular weights from the outside into the capsules; and we examined the secretion of erythropoietin (EPO) — a model therapeutic factor — towards the exterior of the capsule. Finally, we evaluated the foreign body response (FBR) elicited by the implanted microcapsules in Balb / c mice by means of flow cytometry, “western blot” and a complete set of histological studies. Our results demonstrate that these membranes simultaneously increase capsule resistance and enhance mass transport properties without causing any negative effect on the FBR when compared to the classic PLL membranes.

2. Material and methods

2.1. Cell culture

D1-MSCs (ATCC CRL12424) genetically engineered to secrete human EPO (D1-MSC-hEPO) were cultured in T-flasks with Dulbecco’s modified Eagle medium (DMEM) (ATCC 30–2002) supplemented with 10 % fetal bovine serum (FBS) and 1 % penicillin/streptomycin (P/S). Cells were incubated at 37°C in a 5 % CO₂ / 95 % air atmosphere and passaged every 2–3 days when reaching 70–90 % confluence. All reagents were acquired from Gibco (Life Technologies).

2.2. Creation of D1-MSC-hEPO clone

The D1-MSC-hEPO clone was created at the University Clinic of Navarra (CUN) under the generous instruction of Dr. Prosper’s group, as previously described (Gurruchaga et al., 2015). Briefly, the Lenti-X^™ 293 T Cell Line (Clontech, Cat. #632180) was transfected with 9 μg pSIN-EF2-Epo-Pur, 3 μg of pMD2.G (Addgene plasmid 12259), 6 μg of psPAX2 (Addgene plasmid 12260) and 45 μL of lipofectamine^® 2000 (Invitrogen, Carlsbad,CA,USA) following manufactureŕs protocol. After 72 h, virus-containing supernatant was collected and added to D1-MSC cells with 4 μg/mL of Polybrene (hexadimethrine bromide; Sigma, Cat. #H9268). For the following month, cells were maintained in medium containing puromycin (12.5 μg/mL) to select the successfully produced clones.

2.3. Encapsulation

Microcapsule matrices were obtained by means of an electrostatic droplet generator (Nisco, Var V1), extruding a solution of 1.5 % alginate (ProNova UP LVG, NovaMatrix, Cat. #4200006) through a 0.35 mm needle (~23G) at a constant flow rate of 5.9 mL/h. When encapsulating cells, the alginate solution was loaded with 5 × 10⁶ cells/mL. This cell load was selected based on the duration and conditions of the in vitro experiments, in accordance with the findings of previous studies conducted by our group (Lopez-Mendez et al., 2017). The droplets were collected in a 100 mM CaCl₂ solution and kept under stirring until reaching ionic gelation equilibrium.

To obtain different membranes, alginate matrices were coated as follows: (1) Alginate-poly-l-Lysine (AP) microcapsules: alginate matrices were suspended in a 0.05 % PLL (Sigma-Aldrich, Cat. #P7890) solution for 5 min and then washed with DPBS containing Ca⁺⁺ and Mg⁺⁺ (Lonza, Cat. #BE17–513F). (2) AP microcapsules crosslinked with genipin (APG): AP microcapsules were suspended in 0.1 % genipin (Wako, Cat. #078–03021) solution for 5 min and then washed. (3) AP microcapsules coated with a second layer of PLL (APP): AP microcapsules were suspended in a 0.05 % PLL solution for 5 min and subsequently washed. (4) GDP: APG microcapsules were suspended in a 0.05 % PLL solution for 5 min, washed, and suspended again, this time in a 0.1 % genipin solution for 5 min to crosslink the second PLL layer. Finally, the resulting microcapsules were washed.

All solutions were adjusted to physiological pH and osmolarity ranges. Aseptic conditions were maintained during the whole process and all reactions were performed at room temperature. The microcapsules, with or without cells, were kept in T-flasks with complete medium as described in “Cell Culture” section.

The distributions of microcapsule diameters and initial cell loads (million cells/mL capsule) are presented in Supp. Figs. 1 and 2, respectively.

2.4. Explosion assay and swelling

To evaluate the osmotic resistance of the coatings — i.e., the resistance to inner pressure forces —100 μL of cell-laden microcapsules (see 2.3. Encapsulation section) were sedimented in a 1.5 ml eppendorf tube. The supernatant culture medium was discarded, 1 mL of ddH₂O was added and the tubes were maintained in a rocker for 5 min. Subsequently, the microcapsules were pelleted again and the supernatant was replaced with fresh ddH₂O water. This process was repeated up to 5 times. After every washing step, supernatants were collected in 15 mL falcon tubes and the resulting cell suspension was spun down at 300 × g. Centrifuged cells were resuspended in PBS and subsequently counted in a TC20^™ Automated Cell Counter (Bio-Rad) to determine the total number of released cells. Results are shown as mean ± SD of 3 independent experiments for every group.

Photographs were taken through a microscope (Nikon TMS), with 4X magnification, before and after the assay to measure diameters and calculate the swelling percentage of non-broken microcapsules. At least 15 randomly selected microcapsules were analyzed in 3 different experiments and results are expressed as mean ± SD per group.

2.5. Capsule compression assays

The capsule compression system consists of a main tubing of larger diameter than the microcapsules, at the end of which a smaller piece of capillary tubing (i.e., of a diameter smaller than the microcapsules’) is manually fitted (Virumbrales-Munoz et al., 2021). More specifically, this microdevice was composed of a Ø800 μm tubular channel, drastically reduced to Ø 200 μm. The tubing of the system was perfused with serum-free medium. The fitted piece of smaller diameter tubing was then uncoupled for microcapsule injection. A single microcapsule was pipetted through the bigger tube outlet, while avoiding bubble injection. Successful injection was verified using optical microscopy (AmScope 20X & 40X Trinocular Stereo Microscope on Single Arm Boom Stand). Thereafter, the tubing was recoupled. Air pressure was coupled to the system and constantly monitored by means of a precision manometer. Low air pressure was firstly applied, allowing the microcapsule to position by the capillary tube outlet. From that point on, the pressure was raised in a stepwise manner, while the microcapsule was inspected for signs of structural damage. For each group, results show the average value of 5 to 9 different microcapsules ± SD.

2.6. Young’s moduli of the microcapsule surface by AFM

Microcapsule force-spectroscopy-based imaging was performed using a JPK Instruments (Brucker), using the Nanowizard 3 AFM module and the QI Mode ^™. The AFM system was mounted over an optical inverted microscope (Nikon-Eclipse). PNP – TR cantilevers (Nano-AndMore, GMBH) with tip A were chosen for their nominal spring constant (0.32 N/m), which has been previously described as appropriate for the Young’s moduli expected for microcapsules (Virumbrales-Muñoz et al., 2019). Tips were calibrated on glass by thermal noise method before each measurement. For stable force spectroscopy measurements in liquid, we used a nylon mesh to secure the microcapsules (330 μm openings (Labopolis S.L.)). To this end, the mesh was glued inside of a Petri dish (TPP Techno Plastic Products AG, 93040) using a standard two-component cyanoacrylate adhesive only on the edges of the plate, to avoid interfering with AFM measurements. Once the cyanoacrylate adhesive had dried, the Petri dish was filled with serum-free Dulbecco’s modified Eagle medium (Cultek, 12–707F), and bubbles were dislodged from the mesh through iterative gentle pipetting. The dish was then placed and secured on a Petri dish heater (JPK instruments) assembled onto the microscope stage and incubated for 15 min at 37 °C. 50 μL of capsule suspension was added to the Petri dish and allowed to settle in the mesh openings for 5 min. Subsequently, the cantilever was positioned on top of a microcapsule for the arrayed force spectroscopy measurements. Quantitative imaging (QI) maps were obtained using high Z-lengths (10 μm) and slow speeds (10 μm⋅s–1), to achieve stable measurements with a high signal/noise ratio. Force curves were fitted using the Hertz model for a quadratic pyramidal indentor and averaged per microcapsule. For every group, a number of microcapsules (3–5 per group) was assessed and resulting data are given as mean ± SD.

2.7. FITC-dextran diffusion

Microcapsule membrane molecular weight cut-off (MWCO) and permeability were determined by measuring the entrance capacity of FITC-labeled dextrans of increasing molecular weights (MW) (MW 20, 40, 70 and 150 kDa, Sigma, Cat. #FD20S, FD40S, FD70S and FD150S respectively). For each group, 10 μL of microcapsules were washed with DPBS w/Ca⁺⁺ and Mg⁺⁺ and subsequently resuspended in a FITC-dextran solution (0.5 mg/mL in DPBS). After a 24 h incubation at room temperature, samples were observed by confocal microscopy. Resulting micrographs were further analyzed with ImageJ. Dextran diffusion percentage was obtained by calculating the relative fluorescence intensity of equal area squares defined in the center of the microcapsules and in the background. Results are shown as mean of 5 randomly acquired samples ± SD for each group and FITC-dextran size.

2.8. Measurement of hEPO secretion

Aproximately 135,000 encapsulated cells/well were cultured for 24 h in 12 well plates with 1 mL of culture medium. To measure the hEPO content of the supernatants a Quantikine IVD Human Erythropoietin ELISA Kit (R&D Systems, Cat. #DEP00) was used following manufacturer instructions. As specified in the kit, both standards and samples were assayed in duplicate.

2.9. Cell metabolic activity

Nearly 15,000 encapsulated cells/well were cultured in a 96 well plate with 10 μL/well of CCK-8 reagent (Sigma, Cat. #96992). Cells were incubated for 3 h at 37 °C and the color development was read at 450 nm with correction at 690 nm. The mean background value (only medium) was subtracted to all wells to obtain the normalized metabolic activities.

2.10. LIVE/DEAD assay

Cells within the microcapsules were stained using the LIVE/DEAD kit (Thermo Fisher Scientific, Cat. #L3224), which includes the calcein-AM and ethidium homodimer-1 dyes, following the guidelines provided by the manufacturer. Briefly, after a 30 min incubation period at room temperature while protected from light, fluorescence micrographs were captured using a confocal fluorescence microscope (Leica TCS SP2 AOBS Spectral Confocal Scanner mounted on a Leica DM IRE2 inverted fluorescence microscope). The acquired micrographs were z-stacked using ImageJ software.

2.11. Animal experiments

For in vivo biocompatibility and foreign body response studies 48 female, 8-week-old, Balb / c mice (Harlan Laboratories) were used as immunocompetent murine models (4 × 3 mice per group, see 2.16. Data analysis and statistics section). Animal studies were performed by qualified personnel attending to the ethical guidelines established by our European and national institutions and under an approved animal protocol (CEEA/368/2014/SÁNCHEZ ARIZMENDIARRIETA). General anesthesia was induced in mice by inhalation of 4 % isoflurane in oxygen, and administration procedures were carried out under a maintenance dose of 1.5–3 % isoflurane/oxygen. Prior to administration, the dorsal area of mice (between the shoulder blades) was disinfected with a Betadine wipe. Then, the animals were implanted subcutaneously with a total volume of 300 μL of microcapsules (suspended in additional 300 μL of DPBS) using a sterile 20-gauge catheter (Nipro; Nissho Corp, Belgium). Implantations were performed with the catheter tip displaced 2–3 cm away from the puncture point following the spinal line to the tail.

14 days post-implantation, mice were sacrificed by cervical dislocation. A longitudinal dorsal incision was made using forceps and scissors until implanted microspheres were localized. In all cases, microcapsules formed flatted and vascularised aggregates in the subcutaneous space. The whole implant was retrieved with a scalpel and transferred to different buffers for processing, following the different protocols described below

2.12. Confocal immunofluorescence microscopy

Microspheres and the surrounding tissue harvested from mice were fixed with 4 % formaldehyde (PanReac AppliChem, Cat. #252931.1211) for 10 min at room temperature, washed twice with DPBS and permeabilized with 0.1 % Triton X-100 (Sigma Aldrich, Cat. #T8787) solution (in DPBS with 1 % BSA) for 5 min. The explant was washed 3 times with DPBS and stained with Alexa Fluor 488-conjugated Phalloidin (ThermoScientific, Cat. #A12379) at a final concentration of 165 nM in DPBS with 1 % BSA. The reaction was allowed to take place for 40 min protected from light. Afterwards, samples were washed twice and counterstained with 300 nM DAPI (ThermoScientific, Cat. #D1306) for 5 min in the dark. After staining, samples were washed twice and kept in 50 % glycerol (Sigma Aldrich, Cat. #G5516) solution protected from light. Explants were imaged using a confocal fluorescence microscope with a 10X objective and obtained micrograph sections (≈50 μm thickness) were processed and rendered as z-stacks using ImageJ. For each group, implants from 4 mice were analyzed.

2.13. Flow cytometry assays

Single-cell suspensions for flow cytometry studies were obtained following the protocol kindly provided by the group of Prof. Daniel G. Anderson from MIT (Veiseh et al., 2015) < sup > RW.ERROR – Unable to find reference:798</sup><sup > RW.ERROR – Unable to find reference:798</sup >. This method has been reported to extract more than 90 % of cells adherent to the surface of the microcapsules. Briefly, freshly removed implants were immersed in C-Tubes (Miltenyi Biotec, Cat. #130–093–237) with passive PEB dissociation buffer (1X PBS, 0.5 % BSA and 2 mM EDTA) and processed in a gentleMACS^™ Dissociator (Miltenyi Biotec, Cat. #130-093-235) following the manufacturer’s instructions. Resulting suspensions were passed through a MACS Smart-Strainer (70 μm) (Miltenyi Biotec, Cat. #130–098–462) to obtain single-cell suspensions. One volume of cell suspensions was diluted with 10 volumes of 1X Red Blood Cell Lysis Solution (Miltenyi Biotec, Cat. #130–094–183), vortexed for 5 sec and incubated for 2 min at room temperature. Remaining cells were centrifuged at 300 × g at room temperature and supernatants were aspirated completely.

Cell pellets were resuspended in 45 μL ice-cold PEB buffer and co-stained with 5 μL of each of the following antibodies: Anti-Ly-6G-FITC (Miltenyi Biotec, Cat. #130-107-974), Anti-CD68-PE (Miltenyi Biotec, Cat. #130-102-923) and Anti-CD11b-APC (Miltenyi Biotec, Cat. #130-098-088). All samples were incubated for 10 min at 4 °C protected from light. Cells were then washed 3 times by adding 2 mL of PEB buffer and centrifuging at 300 × g for 5 min. Afterwards, supernatants were completely aspirated and cell pellets were resuspended in 500 μL of MACSQuant Running Buffer (Miltenyi Biotec, Cat. #130-092-747) for acquisition in a MACSQuant Analyzer 10 (Miltenyi Biotec, Cat. #130-096-343). Background intensities and compensation settings were adjusted using blank and single antibody samples, respectively. As isotype controls, samples stained with REA Control (S)-FITC (Miltenyi Biotec, Cat. #130-104-610), Rat IgG2a-PE (Miltenyi Biotec, Cat. #130-103-098) and Rat IgG2b-APC (Miltenyi Biotec, Cat. #130-103-085) were also run. Single-cell suspensions from 4 different mice per group were analyzed and results are showed as mean ± SD.

2.14. Histological processing for H&E and Sirius Red staining

Retrieved microcapsules and surrounding tissues were fixed in 4 % Formaldehyde (PANREAC, Cat. #A3684.5000) overnight at 4° C. Afterwards, slides were dehydrated overnight in 70 % ethanol at 4 °C and embedded in paraffin. Samples were sectioned at a 3 μm thickness for subsequent histological analysis. Hematoxylin-Eosin sections were stained in Harriś Hematoxylin (PANREAC, Cat. #254766.1610) for 15 min, gently washed and immersed in Eosin (PANREAC Cat. #256879.1210) for 45 sec, dehydrated and mounted in DPX (PANREAC, Cat. #255254.1608). Sirius Red (Kit LABOLAN, Cat. #210923) staining was performed following the supplier’s instructions. The degree of tissue inflammation and fibrosis were blindly evaluated by an expert pathologist.

2.15. Western blotting

Microcapsule aggregates retrieved from mice were directly immersed in ice-cold Pierce RIPA Lysis and Extraction Buffer (Thermo Scientific, Cat. #89901) with Halt Protease inhibitors (Thermo Scientific, Cat. #78430). Then, samples were lysed by 2 sonication cycles (30 sec on/30 sec off) at 70 % amplitude on ice, and maintained in agitation for 2 h at 4°C. Afterwards, lysates were spun down at 12,000 rpm for 20 min in a pre-chilled (4°C) centrifuge. Supernatants were collected in fresh tubes on ice and stored at − 80 °C for posterior analysis.

30 μg of protein – previously quantified by Pierce BCA protein assay kit (Thermo Scientific, Cat. #23225) – were first mixed with 1/10 parts of DTT (Sigma-Aldrich, Cat. #10197777001) and 1/4 parts of 4x Laemmli Sample Buffer (Bio-Rad, Cat. #161–0747). The mixture was boiled at 85 °C for 10 min and subsequently quenched on ice for 3 min. Samples were run on 10 % Criterion^™ TGX Stain-Free^™ Precast Gel (Bio-Rad, Cat. #5678033) and transferred onto Polyvinylidene fluoride (PVDF) membranes using the Trans-Blot^® Turbo^™ Midi PVDF Transfer Packs (Bio-Rad, Cat. #170–4157). Blots were then assayed with anti-α-Smoth Muscle actin (α-SMA) antibody (rabbit polyclonal, 1:400 dilution) (AbCam, Cat. #ab5694), and anti-β-actin antibody (mouse monoclonal, 1:4000 dilution) (Sigma Aldrich, Cat. #A1978) as a loading control, followed by horseradish peroxidase (HRP)-conjugated Goat Anti-Rabbit (Bio-Rad, Cat. #170–6515) and Goat Anti-Mouse (Bio-Rad, Cat. #170–6516) secondary antibodies (1:10,000 dilution for both). Finally, Clarity^™ Western ECL Substrate (1:1) (Bio-Rad, Cat. #170–5061) was added and blots were visualized using a ChemiDoc^™ MP Imaging System (Bio-Rad, Cat. #170–8280). Obtained band signals were processed for semi-quantitative analysis using Image Lab^™ Software 4.0.1 (Bio-Rad). Protein lysates from 4 different mice per group were assayed and results are presented as mean ± SD.

2.16. Data analysis and statistics

The data obtained were first tested by Shapiro-Wilk to verify whether they fit a normal distribution. If so, an unpaired two-tailed t-test was used for comparisons between two groups and one-way ANOVA for multiple comparisons. For the latter, depending on whether homoscedasticity was met or not, the Bonferroni or Tamhane post-hoc tests were applied respectively. If the data were not normally distributed, Mann-Whitney was applied as a non-parametric test. Results are presented as mean ± SD. The sample size (n) and the test used are specified in the figure captions. The level of significance was established at p < 0.05. All statistical analyzes were run in SPSS 22 (IBM SPSS, Chicago, IL). For in vitro studies, preliminary assays were conducted until ensuring results were reproducible. To test in vivo biocompatibility, experiments were repeated three times (n = 4 × 3 mice per group) using different techniques to confirm results were reproducible.

3. Results and discussion

In this study, we tested GDP microcapsules along with intermediate control groups at each step of the formulation — namely AP, APG and APP — to explore in detail the effect that each membrane component has on the final properties of the capsule (Fig. 1a). In particular, the mechanical properties, mass transport and biocompatibility of each of them were thoroughly studied (Fig. 1b).

Fig. 1. Design of the study.

a, Schematic representation illustrating GDP microcapsules and all the intermediate control groups examined in the study. b, Graphical scheme showing evaluated dimensions for the GDP microcapsules and the different control groups. Figure created with BioRender.com.

3.1. Mechanical properties

Semi-permeable membranes must resist internal — e.g. cell overgrowth — and external pressures — e.g. manipulation and implantation — alike to fulfill biosafety criteria (Nafea and A.M. Poole-Warren L.A., P.J. Martens,, 2011). In this regard, studies have reported inconsistent effects of genipin on membrane strength of immunoisolation devices (Hillberg et al., 2013; Paul et al., 2011; Chen et al., 2007). In light of these discrepancies, our first objective was to demonstrate unequivocally that GDP membranes substantially improved the mechanical properties of microcapsules. To do so, we studied the mechanical resistance of the capsules when subjected to internal and external pressures (Fig. 2a).

Fig. 2. Mechanical properties of GDP microcapsules.

a, Experimental scheme depicting the assays performed to evaluate the resistance of the membranes against outward and inward forces. AFM, atomic force microcopy. Created with BioRender.com. b, Representative bright-field images showing microcapsules before and after repeated osmotic shocks with ddH2O (5 cycles). Scale bar, 200 μm. c, Analysis plot from (b) expressing the number of cells released from microcapsules after explosion assay. Error bars, mean ± SD. ***, p < 0.001 in comparisons against the AP group; Unpaired, two-tailed t-test, n = 3 independent experiments per group. d, Analysis plot from (b) expressing the swelling degree of resistant microcapsules. ###, p < 0.001 for GDP-APP comparison; Mann–Whitney U test, n = 15 randomly taken microcapsules per group in 3 independent experiments. e, Data from compression assay showing the resistance of microcapsules to external pressures. Error bars, mean ± SD. *, p < 0.05; **, p < 0.01 in comparisons against the AP group; One-way ANOVA with Bonferroni multiple comparison correction, n ≥ 5 randomly taken microcapsules per group. f, Young’s Modulus of microcapsule surfaces measured by AFM. Error bars, mean ± SD. *, p < 0.05 for GDP-AP comparison; Unpaired, two-tailed t-test, n ≥ 3 randomly taken microcapsules per group.

To measure resistance to rupture under internal pressure, we exposed cell-laden microcapsules to drastic osmotic shocks in a repeated manner (up to 5 cycles) and quantified both the number of cells released into the medium and degree of swelling in those microcapsules that resisted pressure (Fig. 2b–d). Regarding microcapsule rupture and cell release, the groups with a single coating – namely, AP and APG — barely withstood the osmotic shock cycles, releasing approximately 2.5 × 10⁶ cells per group from the microcapsules. Groups with a double coating–APP and GDP— demonstrated remarkably superior behavior and maintained intact integrity in virtually all microcapsules. Despite both groups with double coating resisted the osmotic shock, we found significant differences in their swelling behavior. Interestingly, within the groups that experienced microcapsule rupture, we observed that APG microcapsules swelled more rapidly and broke before AP (Supp. Fig. 3).

To measure resistance to external pressures, we developed a novel low-cost microfluidics-based assay. We placed microcapsules one by one into a micro-capillary tube and subjected them to increasing air pressures under microscopic inspection until we observed signs of structural damage (Supp. Fig. 4). Once again, the double coated-groups APP and GDP showed significantly greater membrane integrity under stress when compared to AP control (p < 0.01 and 0.05, respectively) (Fig. 2 e). Strikingly, these results revealed that the double PLL coating rather than genipin accounted for increased resistance of microcapsules. Moreover, genipin cross-linking seems to weaken slightly the mechanical strength of membranes, which will be the subject of future studies.

We further investigated the effects of genipin cross-linking on membrane mechanical properties using AFM force-spectroscopy measurements to quantify Young’s modulus of the microcapsule surface. Only GDP microcapsules exhibited significantly higher values than AP control (p < 0.05) (Fig. 2f). Therefore, we conclude that genipin cross-linking increases the stiffness of GDP microcapsules but is not primarily responsible for improving overall mechanical strength. In this sense, the resistance to both internal and external forces observed in the GDP design is due to its double PLL coating, as also reported previously for other microcapsule designs (Zheng et al., 2014; Thanos et al., 2007).

3.2. Mass-transport properties

The incorporation of multiple coating layers intended to improve the mechanical properties of capsules is generally related to negative effects on the perm-selective function of their membranes. Hence, we next tested the inward and outward mass transport properties of the GDP microspheres along with intermediate control groups (Fig. 3a). First, we used dextrans of different molecular weights to evaluate inward diffusion. GDP microcapsules presented the same 70 kDa molecular weight cut-off (MWCO) that the classic AP microcapsules show, but allowed an improved diffusion of smaller molecules (p < 0.05) (Fig. 3b–e and Supp. Fig. 5). Genipin cross-linking, rather than double PLL coating, was the main cause of such enhanced permeability, as we also noted significant differences between GDP and APP groups (p < 0.05).

Fig. 3. Mass-transport in GDP microcapsules.

a, GDP microcapsules were studied in detail to rule out any negative effects on the inward and outward perm-selective function of the new membranes. Created with BioRender.com. b-e, Inward permeability study with dextrans of different molecular weights (MW) (20, 40, 70 and 150 kDa) conjugated with FITC (n = 5 randomly taken microcapsules per group and MW). MWCO, molecular weight cut-off. Error bars, mean ± SD. *, p < 0.05 for GDP-AP comparisons; #, p < 0.05 for GDP-APP comparisons. Unpaired, two-tailed t-test was used for 40 kDa, whereas Mann–Whitney U test was used for 20 kDa (non-normal data distribution). f, Outward diffusion of model therapeutic factor (hEPO) secreted by D1-MSC-hEPO immobilized in AP and GDP microcapsules (n = 3 samples per group in duplicate) for 28 days after encapsulation. Values are expressed as the percentage relative to AP group for each time-point. Error bars, mean ± SD. *, p < 0.05; **, p < 0.01; Unpaired, two-tailed t-test. g, Metabolic activity of D1-MSC-hEPO encapsulated in AP and GDP microcapsules (n = 5 samples per group) assessed at same time-points of (f). Error bars, mean ± SD. N.S. non-significant differences; Unpaired, two-tailed t-test. h,i Representative confocal fluorescence images of cells encapsulated in AP (h) and GDP (i) microcapsules and probed with LIVE/DEAD viability kit (Green, living cells; Red, dead cells) 28 days after encapsulation. Scale bars, 500 μm.

To validate these results, we compared the outward diffusion of the GDP and AP microcapsules to investigate any possible restriction in the release of therapeutic factors secreted by cells immobilized in their matrix. Given the differences observed in the inward diffusion with molecules around 40 kDa, we quantified secretion of hEPO — a globular glycoprotein of 34 kDa — along with the viability of encapsulated cells over a period of 28 days. Again, we measured significantly higher levels of therapeutic product released from GDP microcapsules when compared to the AP control group (Fig. 3f). Such differences were the result of improved permeability, as the viability of cells in both groups did not differ during the whole experiment (Fig. 3g–i).

Enhanced permeability may be attributed to the spacing between PLL chains produced by dimerization of genipin. As reported by other authors, this may form short cross-linking bridges of up to 2.5 nm (Liang et al., 2004), the equivalent of ≈50 kDa for spherical proteins (Erickson, 2009), at physiological pH (Dimida et al., 2015). This property of genipin has the advantage of maintaining good mass transfer of growth factors from the surrounding environment and maximizing delivery of therapeutic molecules (Fig. 4).

Fig. 4. Proposed membrane nanostructure of GDP microcapsules.

The short cross-linking bridges mediated by genipin dimers create spaces between the PLL chains, thus enhancing the permeability of the membranes. In the proposed model, this structural modification gives rise to stable and more uniformly distributed pores of nearly 2.5 nm in diameter. This allows for maximized bi-directional diffusion of molecules below the well-established molecular weight cut-off (MWCO) of 70 kDa. Figure created with BioRender.com.

This study acknowledges the limitation of employing the D1-MSC cell line, which, while suitable for demonstrating the feasibility of the GDP microcapsules, does not encompass the full translational potential of the system. Future studies are warranted to validate these findings using primary MSCs or other clinically relevant cell types, thereby broadening the scope of applications and strengthening the clinical relevance of the proposed microcapsule design.

3.3. In vivo biocompatibility

We next analyzed the immune response against GDP and intermediate microcapsules implanted in the subcutaneous space of immunocompetent Balb / c mice. As established by other authors, we examined the foreign body reaction 14 days after implantation, given that this time is adequate to determine either immune or fibrotic responses to implanted biomedical devices (Veiseh et al., 2015; Doloff et al., 2017; Vegas et al., 2016). At this time point, Z-stacked confocal images using DAPI (nucleus marker) and Phalloidin-Alexa Fluor 488 (F-actin marker) showed similar cell deposition in retrieved microcapsules for all groups (Fig. 5a).

Fig. 5. GPD microcapsules in vivo biocompatibility.

The GDP microcapsules and the intermediate control groups were implanted in the subcutaneous space of Balb/c mice (300 μL in sediment volume) and retrieved 14 days after to examine the cellular and fibrotic response. a, Representative Z-stacked confocal images of pericapsular overgrowth stained for cell nuclei (DAPI, blue) and F-actin (Phalloidin-Alexa Fluor 488, green). Crosslinked genipin in the GDP group emits red fluorescence. Scale bar, 400 μm. b-d, Flow cytometry analysis for total cell deposition (b) and infiltrated innate response cells marked with specific markers for neutrophils (CD68− / Ly6G+) (c) and macrophages (CD11b+ / CD68+) (d). Error bars, mean ± SD. N.S. non-significant differences; One-way ANOVA with Bonferroni multiple comparison correction, n = 4 mice per group. e, Representative images of Sirius Red stained histological sections indicative of collagen deposition (light red) around the microcapsules (n = 4 mice per group). Scale bar, 200 μm. f, Western blot bands showing α-SMA (fibrotic marker) and β-Actin (load control) expression in lysates from microcapsule surrounding tissues. g, Semi-quantitative analysis of α-SMA band intensities shown in (f) normalized to β-Actin. Error bars, mean ± SD. N.S. non-significant differences; One-way ANOVA with Bonferroni multiple comparison correction, n = 4 mice per group.

We further evaluated the pericapsular overgrowth by means of flow cytometry to quantify the cellular response more accurately. In accordance with immunohistochemical assays (Fig. 5a), we found comparable levels of cell deposition in all assayed groups (Fig. 5b). When using specific markers for innate immune cells neutrophils — CD 11b and Ly-6G — and macrophages — CD 11b and CD 68 —, we found no significant differences with the AP control group (Fig. 5c,d and Supp. Fig. 6). Additional histological analysis also revealed similar cell infiltration and giant cells (fused macrophages or granulomas) around the microcapsules in all groups (Supp. Fig. 7).

Additionally, we evaluated the extent of fibrosis surrounding implanted microcapsules by histologic staining of collagen fibers with Sirius Red. All groups showed comparable levels of collagen deposition (Fig. 5e). Finally, we validated these observations by measuring the amount of smooth muscle actin (α-SMA) — a marker of activated fibroblasts — present in the lysates of the explants. Once again, we obtained non-significant differences when compared to AP control group (Fig. 5f,g). All these results suggest that neither the addition of double coatings, nor the presence of genipin have a significant negative impact on microcapsule biocompatibility in vivo.

4. Conclusions

Here we demonstrated for the first time that mechanical properties and mass-transport capacity of cell microencapsulation systems can be enhanced at the same time by means of GDP membrane design. In particular, the inclusion of genipin crosslinking in membranes reinforced with double polycation coatings enables a more efficient mass transport while preserving the immuno-barrier characteristics of the membrane. Furthermore, we also confirmed that the foreign body reaction elicited by GDP microcapsules in vivo is comparable to that of the widely used AP microparticles. Therefore, we conclude that the unique design of GDP membranes provide the immunoisolation devices with important improvements over the classic AP microcapsules without any negative effect. Similarly, these results may be relevant for the design of different nano-, micro- and macro-sized drug delivery systems that require the simultaneous optimization of their mechanical properties and perm-selectivity function.

Statement of significance.

This study presents a novel membrane design for cell encapsulation systems that simultaneously enhances both mechanical properties and permeability. The innovation lies in overcoming the typical trade-off between these two parameters, where improving one negatively affects the other. By combining the strength of a double layer of polycations with the optimized pore arrangement provided by genipin crosslinking, we achieve more resistant membranes with enhanced diffusion efficiency. Additionally, the modified membranes maintain the overall biocompatibility of cell encapsulation systems when implanted subcutaneously in mice.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijpharm.2025.125196.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. All materials are commercially available and their catalogue number are specified in the Methods section. Any step-by-step protocol required to reproduce the experiments are available upon request to the corresponding author.