Effects of sustained GABA releasing implants on pancreatic islets in mice

Abstract

Gamma-aminobutyric acid (GABA) is an inhibitory neurotransmitter that is strongly and selectively synthesized in and secreted from pancreatic beta cells. Exogenously delivered GABA has been proposed to induce beta cell regeneration in type 1 diabetes, but these results have been difficult to replicate and may depend on the specifics of the animal model and drug delivery method used. Here, we developed a GABA-releasing ethylene-vinyl acetate polymer implant for sustained GABA delivery to the intraperitoneal space as an alternative to injected or oral GABA. We explored the effect of the GABA-releasing implants on islet size in non-diabetic, outbred mice. We also attempted to monitor in vivo GABA release using HPLC on blood samples, but these measurements were confounded by high variability within treatment groups and unexpectedly high serum GABA levels in mice receiving GABA-negative implants. The ethylene-vinyl acetate polymer implants became heavily fibrosed with abdominal adhesion tissue, while the osmotic pumps had no macroscopic fibrosis. Histological analysis showed no significant effect of the sustained GABA delivery polymer or osmotic pumps on islet size, alpha cell to beta cell ratio, or the number of Ki67-positive islet cells. The GABA treatment time course was limited to two weeks due to the drug-release window of the polymer, while others reported islet-trophic effects of GABA after ten to twelve weeks of treatment. In summary, our study is consistent with the concept that exogenous GABA administration does not significantly alter islet cell mass in non-diabetic CD-1 mice in the short-term. However, more data are needed, including higher GABA doses and more prolonged treatment regimens for a better comparison with contrasting reports.

Graphical Abstract

Introduction

Type 1 diabetes (T1D) is an autoimmune disease characterized by autoreactive T cells that target and destroy insulin-producing beta-cells in pancreatic islets. Pancreatic islets are clusters of alpha, beta, and delta endocrine cells which serve to release the major metabolic hormones produced in the pancreas that control blood glucose homeostasis [1]-[3]. Beta-cells are responsible for the production of insulin, which is necessary for the uptake of glucose [1]. Common methods to restore glycemic control in patients with T1D include insulin replacement therapy and islet transplantation [3]-[6].

GABA is a neurotransmitter secreted by pancreatic beta cells to regulate islet function and homeostasis. Acute physiological signaling effects of GABA on pancreatic islet behavior are mediated mainly by two receptor families: GABA_A and GABA_B receptors [7], [8]. GABA has been demonstrated to regenerate lost beta-cell mass, synchronize insulin secretion, and regulate autoimmune responses when injected daily in the intraperitoneal space for several weeks [9]-[18]. However, there is little consensus as to the exact effects of long-term GABA administration on the regulation of pancreatic islet mass. Ben-Othman et al. demonstrated that extended daily GABA administration for one to three months might induce the conversion of glucagon-expressing or formerly glucagon-expressing alpha cells to insulin-positive beta cells in Glucagon-Cre-ROSA mice [17]. The Glucagon-Cre-ROSA modification permanently tags any cell that has ever expressed glucagon [19]. It was also demonstrated that daily injection of GABA into the intraperitoneal space over periods of 1-3 months restores normoglycemia in streptozotocin-induced (STZ-induced) diabetic Glu-Cre-ROSA mice [17]. Other studies have found that three-month oral administration of artemisinin and its derivatives, a group of anti-malarial drugs, may also induce similar effects as GABA due to suppression of the Arx gene via GABA_A receptor signaling, which theoretically leads to the alpha-to-beta cell conversion phenomenon and the reduction of blood glucose in diabetic mice [20]-[23]. In another study, Feng et al. demonstrated that a relatively short 12-day time-course of GABA injection (10 mg/kg daily) in streptozotocin (STZ)-treated C57BL/6 mice prevented the typical STZ-induced changes to alpha and beta cell mass [10].

Other studies were unable to reproduce the effects of GABA shown in the works above. In a study by van der Meulen et al., C57BL/6 mice were treated with artemether, but it was instead found that the expression of alpha cell, beta cell, and delta cell genes were suppressed as a result of artemether treatment [21], [24]. Ackermann et al. used glucagon-CreERT2;Rosa-LSL-eYFP mice in a study to demonstrate that neither GABA nor artesunate administration for three months stimulate alpha to beta reprogramming or improve islet function [25].

An additional recent study by Untereiner et al. demonstrated that GABA does promote beta cell proliferation, but the effect depends on the mouse strain. For ten weeks, GABA was supplemented in the drinking water of both non-diabetic C57BL/6 mice and non-diabetic CD-1 mice. It was found that beta cell mass, glucose tolerance, and insulin secretion were unaffected by GABA treatment in the C57BL/6 mice. However, in non-diabetic CD-1 mice, GABA treatment increased glucose tolerance, increased beta cell number, and increased insulin secretion [26]. Additionally, Untereiner et al. found that CD-1 mice with diet-induced diabetes, rather than STZ-induced diabetes, did not experience improved glucose tolerance or increased beta cell mass from GABA treatment. Thus, it is possible that the major disparities in the findings of these investigations arise from several factors: strain-dependent responses to GABA, where CD-1 mice may be more responsive to GABA than other commonly used strains such as C57BL/6, whether or not the mouse is initially diabetic, the method of induction of diabetes, and the route, dose, and duration of GABA administration [26].

The mode of drug delivery is an important consideration when implementing GABA treatments to increase beta cell mass. Previous strategies used either daily intraperitoneal bolus injections or oral ingestion in the drinking water [17], [20], [21], [26]. Long-term delivery of GABA to the pancreas via oral or intraperitoneal injection administration could be challenging due to rapid clearance of GABA due to its small size and polar nature [11], [27]. A sustained method of GABA delivery may be useful for observing consistent effects of GABA on islet size/beta cell mass. The properties of GABA, including its low molar mass and high water solubility, make it a challenging substance for constant-rate drug delivery from biomaterials. Elvax^™ 40W (E40W) is an ethylene-vinyl acetate copolymer that is unreactive in polar solutions. Ethyl-vinyl acetate copolymers are non-toxic, non-biodegradable, and pharmacologically inert substances that have been previously explored as materials for long-term drug delivery [28]-[36]. When homogenized with polar drugs such as GABA and allowed to dehydrate, ethylene-vinyl acetate implants can release low molecular weight polar drugs over long stretches of time due to the osmotic leaching [29]-[32], [36], [37].

Here, we hypothesized that the administration of GABA-releasing implants placed in the intraperitoneal space of non-diabetic CD-1 mice would induce increases in islet size due to the combination of previously reported strain-specific responses of non-diabetic CD-1 mice to GABA and the sustained IP drug administration (Fig. 1) [17], [20], [26]. The objective of this project was to engineer and optimize an E40W ethyl-vinyl acetate polymer drug-releasing biomaterial implant to achieve an extended period of sustained and controlled GABA administration. Osmotic pumps were explored as an additional positive control GABA delivery vehicle due to their ability to release a soluble drug at a steady rate from a reservoir compressed by a high osmotic pressure difference [38]-[41].

Figure 1:

Schematic summarizing the experimental methods. GABA and Elvax 40W are first homogenized with an organic solvent, which is then allowed to evaporate. This leads to the formation of a GABA-infused E40W film. Through OPA derivation, GABA release from the implant over time was quantified using a plate reader. GABA-infused E40W implants are partitioned according to the individual weights of each mouse. After successful implantation, tail-tip blood samples were extracted every two days from mice, and mouse weight was measured every three days. Post-euthanasia, blood samples were analyzed for GABA content via high-performance liquid chromatography. Mouse pancreata were extracted, preserved, and sectioned for analysis via confocal microscopy.

Materials and Methods

Implant Formulation and In-Vitro Testing

Elvax^™ 40W polymer beads were purchased from Dow, Inc. We used ≥ 99% purity GABA from Millipore Sigma (cat# A2129) for all experiments and HPLC standards/assays. E40W beads were sterilized by soaking in 70% v/v ethanol for one week. For the GABA treatment group (G+), 1000 mg of E40W beads were dissolved in 10 ml methylene chloride then rapidly vortexed with 62 mg of GABA dissolved in 1 ml H₂O. Based on in-vitro implant testing, implants produced using this formulation to deliver approximately 10 mg GABA / kg mouse delivered per day when portioned according to the weight of each mouse. For the E40W implant negative control group (G−), 1000 mg of E40W beads were dissolved in 10 ml methylene chloride then rapidly vortexed with 1 mL H₂O without any GABA. The E40W / GABA emulsions were poured into thin films in glass petri dishes. Methylene chloride, ethanol, and water were evaporated for 48 hours in a fume hood. After all liquids were fully evaporated from the mixture, a solid polymeric film with embedded GABA molecules was formed. Portions of this film were cut and weighed based on the desired GABA dosage for each mouse.

The GABA release profiles from the Elvax 40W film were characterized in-vitro by suspending E40W films in ultra-pure water and collecting aliquots from the supernatant at timed intervals. The concentration of GABA in each aliquot was measured by derivatization with o-Phthaldialdehyde (OPA), which renders amino acids fluorescent in a plate reader [42]. GABA-release rates from polymer implants were quantified in-vitro using fluorescence-based spectroscopy (340 nm excitation, 455 nm emission) and comparison to a standard curve of serial 1:2 GABA dilutions.

Osmotic pumps (Alzet Micro-Osmotic Pump Model 1002) were each filled with 100 μl of 650 mM GABA dissolved in saline under sterile conditions to achieve an approximate delivery rate of 10 mg GABA/kg mouse delivered per day. As recommended by the manufacturer, the osmotic pumps were incubated in sterile saline at 37°C overnight prior to surgery [38].

In-Vivo Studies

For in-vivo studies, outbred CD-1 IGS mice were purchased from Charles River Laboratories. The CD-1 IGS Mouse is commonly used as a multi-purpose model for research in toxicology, oncology, and aging [43]-[45]. For this study, mice were non-diabetic. Four groups were designated to determine the effectiveness of GABA-releasing E40W implants. Mice were anesthetized with isoflurane, and the IP space was opened with a 1 cm midline laparotomy under sterile surgical conditions. Sham mice were 67 days old at the time of surgery. G+. G−, and OP mice were 81 days old at the time of surgery. Sham mice underwent surgery, but no implants were placed in the IP space. The G+ group received GABA-releasing 50 mg/kg implants placed in the intraperitoneal space to achieve the desired GABA dosage of 10 mg GABA/kg body weight each day based on the in vitro measured release rates. This dosing of 10 mg GABA/kg body weight each day was based on previous reports of GABA injection at 10 mg GABA/kg mouse delivered per day being effective for short-duration GABA treatment [10]. The G− group received 50 mg/kg E40W implants without GABA. The OP group received osmotic pumps loaded with 66.7 mg/ml GABA in sterile saline implanted into the peritoneal space in order to administer 10 mg of GABA per kg of mouse mass each day.

Tail-tip blood samples were collected every two days starting on the day of surgery and stored at −80°C. Mice were weighed on the day of surgery and every three days thereafter. Because the E40W implants were not shown to be effective at releasing GABA for more than 14 days in-vitro, the mice were euthanized, and pancreata were extracted for analysis 14 days after surgery.

Whole mouse pancreata were fixed with 4% paraformaldehyde then embedded in a single paraffin block (the small size of the mouse pancreas allows for complete organ sections). Eight to ten paraffin sections were collected from each mouse at well-spaced intervals. A single section from the center portion of the organ was stained using immunofluorescence for insulin, glucagon, and Ki67 and imaged using a Leica SP8 confocal microscope with a resonant scanner. Merged tile scans of entire pancreatic sections were generated using the Leica LASX Navigator software feature [46]. All remaining sections were stained with hematoxylin and eosin (H&E). All sections were imaged and quantified for islet size and diameter. Islet areas were measured using ImageJ [46], [47]. Endocrine to exocrine ratios were calculated according to the following formula, where $I$ denotes insulin-positive regions and D denotes DAPI positive regions:

$$\frac{Endocrinearea}{Exocrinearea}=\frac{\sum I}{\sum D-\sum I}$$

Tail-tip blood samples were analyzed for GABA using high-performance liquid chromatography (HPLC) using an EICOM HTEC-500 HPLC-ECD with autosampler, online automated o-phthalaldehyde-derivatization, and Eicompak FA-3ODS separation column [42]. By comparison with GABA standards, HPLC chromatograms were used to determine the GABA concentration in serum samples. Approximately 2 μl of mouse serum was diluted in 8 μl of mobile phase (8.275 g anhydrous NaH₂PO₄, 1.56 g anhydrous Na₂HPO₄, 5 mg EDTA, 70 ml methanol, 130 ml acetonitrile, 800 ml H₂O). Samples were centrifuged at 14,000 x g to remove precipitated proteins. GABA bioavailability in mouse serum was quantified by integration of the area under GABA peaks, which were identified by comparison to sequentially diluted GABA standards.

Results

GABA-loaded E40W Implants follow Higuchi drug release kinetics

GABA-releasing E40W implants (Fig. 2A) were found to have a loading efficiency of approximately 53%; for every 1 mg of GABA infused into the implants during film formulation, 0.53 mg of accumulated GABA was found in aqueous solution after two weeks. E40W films exhibited a pattern of gradual GABA release for at least a ten-day time period (Fig. 2B). After ten days, GABA release rates approached zero due to payload depletion. Analysis of E40W films with differing GABA to polymer ratios and batch sizes indicated that a 3:50 (mg/mg) GABA to E40W ratio was the formulation most suitable for our experimental needs (Fig. S2A-B).

Figure 2:

(A) GABA-infused E40W resin film. Scale bar 1 mm. (B) GABA release from GABA-infused E40W film. Loading efficiency ≈ 53% (mg/mg). n=3. (C) Higuchi model of drug release: cumulative % GABA release versus square root of time. n = 3. CGR: Higuchi model approximation for cumulative GABA released as a function of time. Sy.x: standard error of estimate (%). R²: coefficient of determination. n = 3. Error bars: s.e.m.

The pharmacokinetic behavior of the E40W implant correlated strongly with the Higuchi drug release model, which assumes that the drug release material (in this case, E40W) does not degrade as the embedded drug is released (Fig. 2C). The Higuchi model was selected due to its prior use to model the drug release behavior of other ethylene vinyl acetate-based drug release biomaterials. The Higuchi model assumes that cumulative release is proportional to the square root of time [37], [48]-[52].

Serum GABA is slightly but not significantly elevated in mice that received osmotic pumps and GABA-negative E40W implants

Mouse serum GABA throughout the 14 day period was measured via HPLC and comparison to GABA standards (Fig. 3A-D) [9], [42]. Unexpectedly, HPLC analysis of GABA content in mouse serum revealed that G− mice that received implants with no GABA had a similar concentration of GABA in the bloodstream as the mice in the osmotic pump group (OP). Both G− and OP groups had slightly higher GABA concentrations in serum than those that received GABA infused implants (G+) and the sham group (Fig. 3E,F). However, a mixed-effect model analysis (two-way ANOVA) with Tukey correction indicated that any differences were statistically insignificant. The sham group experienced gradual weight gain over the two-week period. Implant receiving treatment groups G+, G−, and OP all showed a loss in weight three days after surgery, followed by stabilization of weight for the rest of the study. Repeated measures (two-way ANOVA) with Tukey correction show significant weight loss of mice in the G+ group compared to sham for days 3 and 6.

Figure 3:

(A, B) Chromatograms of purified GABA obtained via high-performance liquid chromatography. The initial peak at 1.5 min corresponds to the mobile phase. Voltage increases as GABA concentration is increased. GABA peaks occur within the 12-14 minute timeframe. (C, D) Mouse tail tip blood plasma chromatograms showing GABA peaks resolved between 12 and 14 minutes. Left: G+ vs. G−. Right: OP vs. Sham. (E, F) Mouse serum GABA concentration. Left: OP vs. Sham. Right: G+ vs. G−. Mixed model analysis with Tukey correction (two-way ANOVA) did not indicate any significant differences among serum GABA means at any time points. (G) Mouse weight gain as a fraction of initial weight. Repeated measures ANOVA indicates significantly lower weight in the G+ group compared to the sham group on days 6 and 9, but no significant (p ≤ 0.05) differences were found at all other time points. n = 5.

Islet size is not significantly affected by IP GABA administration

Mouse pancreata were sectioned and immunostained for insulin, glucagon, and Ki-67. Confoical microscopy images of whole slides were taken at low magnification (Fig. 4A-C). Tile scan images of the entire pancreas sections acquired with this method are shown in the supplementary data (Fig. S1). Islets were identified in low magnification whole slide scans by insulin positivity [53] (Fig. 4A). Higher magnification confocal images were acquired for insulin, glucagon, and Ki-67 as a marker of cell proliferation [54] and quantified (Fig. 5A-E). An additional six to eight pancreas sections from each mouse were stained by hematoxylin and eosin, imaged, and quantified for islet sizes (Fig. S3).

Figure 4:

(A) Low magnification confocal microscopy tile scans of mouse pancreas sections stained for insulin and Ki-67 (see also Fig. S1) Ki-67 is not shown in upper low magnification images due to insufficient resolution. Ki-67 positive cells are circled in lower high magnification images of individual islets. All inset panels of single islets are shown at the same magnification. Some sections contained large islets, but the large islets did not contain higher numbers of insulin⁺ / Ki-67⁺ cells, and their presence was not dependent on GABA treatment (see also Fig. 5). Scale bars upper panels = 500 μm, lower panels = 50 μm. (B) CD-1 mouse islet areas. Each dot is one islet. Data are pooled from 5 mice per group. (C) CD-1 mouse endocrine to exocrine area ratios (μm²/μm²). n = 5 mice. Each dot represents the ratio of the insulin-positive area to the insulin-negative area in each mouse. One-way ANOVA statistical test with Tukey posthoc analysis indicates no significant (p < 0.05) differences in the endocrine to the exocrine area between treatment groups [55]. Bartlett’s test to compare variance indicates no significant differences in variance among treatment groups.

Figure 5:

(A) Confocal microscopy images of mouse pancreas sections stained for insulin, glucagon, and Ki-67. Arrowheads indicate Ki-67 positive nuclei. (B) Islet areas from individual confocal scans of single islets (C) Islet areas from images of H&E-stained slides (see also Fig. S3). (D) Ki-67+ cells per islet. (E) Ratio of glucagon positive area to insulin-positive area per islet. One-way ANOVA statistical test with Tukey posthoc analysis indicated no significant (p < 0.05) differences for all plots. Each dot is one islet. Data are pooled from 5 mice per group.

Islet areas were compared to the total pancreatic area for each mouse in order to account for pancreas sections of varying size. One-way ANOVA statistical test with Tukey posthoc analysis indicated no significant (p < 0.05) differences in islet size, alpha cell to beta cell ratio, Ki-67⁺ cells per islet, or endocrine to exocrine ratio between treatment groups [55]. Bartlett’s test to compare variance indicated no significant differences in variance among treatment groups.

While Ki-67 positive cells could be identified in many islets, the presence of Ki-67 positive cells was not substantially different between all treatment groups. Ki-67 and glucagon were not resolvable in the low-magnification view of whole pancreas sections (Fig. 4A, Fig. S1), so these markers are only shown in the higher magnification images captured of individual islets (Fig. 4A, 5A,D,E).

Discussion and Conclusions

Elvax 40W-based GABA-releasing implants produced GABA release patterns consistent with the Higuchi drug release model for approximately ten days. Previous studies using similar polymeric ethylene-vinyl acetate substances for long-term drug release have demonstrated the relevance of the Higuchi drug release model due to the lack of degradation of the polymer [36], [37], [48]-[52].

The HPLC results suggest that our implant design did not effectively impact circulating GABA levels. Post-mortem collected G+ and G− E40W implants revealed a substantial fibrotic encapsulation for both conditions, likely due to the formation of intraperitoneal adhesions tissue. As ethylene-vinyl acetate implants are used successfully in other sites, including ophthalmic, subcutaneous, and contraceptive drug-releasing implants [28]-[36], this issue may be unique to our specific implant design and the tendency of peritoneal tissues to form adhesions. A macroscopic foreign body response was not observed in the osmotic pump group. It could be the case that fibrotic encapsulation adversely affected the GABA release rate in vivo due to insufficient contact with the IP fluid in fibrosed implants. We could likewise conclude that GABA simply has a short half-life and tight regulation in serum. We also used the outbred CD-1 mouse strain, which may have higher variability between individuals than inbred strains. Future studies should titrate the GABA dosage to optimize increases to serum GABA levels. We did not attempt to increase the dosage for this study because we were already delivering the maximum amount of GABA that could be loaded in the E40W films without causing burst release.

High-performance liquid chromatography revealed that the GABA concentration in the serum of mice receiving G− implants was higher than expected while G+ was lower than expected, although none of the groups were statistically different from sham. It is possible that the presence of a foreign body response or abdominal adhesions to E40W implants complicated the GABA release kinetics and serum GABA readings. Dietary intake of GABA and other amino acids related to mouse feeding behavior, as observed with mouse weight differences, may also complicate interpretation of circulating GABA levels. Further investigation is required to optimize GABA-eluting implants to alter serum GABA. We propose that the osmotic pump is a more reliable approach for sustained GABA release in future studies because it did not induce abdominal adhesions, behaved somewhat more predictably for the elevation of serum GABA levels, and has the ability to load much higher concentrations of GABA.

The use of a 3:5 GABA:E40W ratio was chosen based on preliminary testing of different methods of implant formulations due to its low burst release and relatively long-lasting release kinetics. Other implant methods were explored, including GABA-releasing microparticles and varying ratios of GABA to E40W, but these were limited to burst release within the first 24 hours in-vitro.

Unlike previous studies investigating the delivery of GABA to the IP space, we neither directly injected GABA into the IP space, nor did we administer oral GABA as a control group. While we acknowledge this as a limitation of our study, this project was not designed as a direct comparison to the works by Ben-Othman et al. or Li et al. [17], [21]. We also chose to study the effects after a two-week time period rather than the ten-week timeframe in the study by Untereiner et al. due to the pharmacokinetic performance of our implants [26]. While these relevant studies reported effects of GABA on islet cell mass after 1-3 months, at least one other group has reported positive results on beta cell mass following STZ treatment after only 12 days of GABA treatment [10]. Another limitation of our study is that the sham group received no implant. Future studies should include a negative control where osmotic pumps are loaded with saline to account for any adverse effects of the implant itself on mouse health or behavior. This control is justified by the observation that sham animals did not lose weight post-surgery while all implant-receiving animals lost weight for the first three days before stabilizing.

Interestingly, we observed many large >300 μm diameter islets in the pancreata of CD-1 mice, including some islets as large as 600 μm diameter, although the average islet diameter was still on the order of 100 μm. The typical islet diameter for mammals ranges between 50 μm and 150 μm [56]. Similar numbers of large islets were observed across all treatment groups in our study. This contrasts with the study by Untereiner et al., which reported a significant increase in the frequency of large islets (area >60,000 μm² ≈ diameter >300 μm) in non-diabetic CD-1 mice after ten weeks of GABA administration. However, Untereiner et al. also showed no change in the frequency of large islets after GABA treatment in high fat diet-induced diabetic CD-1 mice. This suggests that animal age, chow formulation, and activity level could affect GABA treatment results in this strain. We should also consider, as very large islets were observed in all groups in our study and for diabetic mice in the study by Untereiner et al., that large islets may be an intrinsic characteristic of the CD-1 mouse strain. Our results are limited by the fact that other studies reported an islet-trophic effect of GABA using much longer treatment regimens on the order of ten to twelve weeks and using different delivery routes. From here, additional studies are warranted to examine longer time-courses and higher doses of GABA administration, which are achievable with long-duration osmotic or refillable/programmable style infusion pumps.