A glucose-responsive cannula for automated and electronics-free insulin delivery

Abstract

Automated delivery of insulin based on continuous glucose monitoring is revolutionizing the way insulin-dependent diabetes is treated. However, challenges remain for the widespread adoption of these systems, including the requirement of a separate glucose sensor, sophisticated electronics and algorithms, and the need for significant user input to operate these costly therapies. Herein, we report a user-centric glucose-responsive cannula for electronics-free insulin delivery. The cannula —made from a tough, elastomer-hydrogel hybrid membrane formed through a one-pot solvent exchange method— changes permeability to release insulin rapidly upon physiologically relevant varying glucose levels, providing simple and automated insulin delivery with no additional hardware or software. Two prototypes of the cannula were evaluated in insulin-deficient diabetic mice. The first cannula —an ends-sealed, subcutaneously inserted prototype— normalized blood glucose levels for three days and controlled postprandial glucose levels. The second, more translational version —a cannula with the distal end sealed and the proximal end connected to a transcutaneous injection port— likewise demonstrated tight, three-day regulation of blood glucose levels when refilled twice daily. This proof-of-concept study may aid in the development of “smart” cannulas and next-generation insulin therapies at a reduced burden-of-care toll and cost to end-users.

Graphical Abstract

In this study, we present a glucose-responsive cannula for automated insulin delivery without the use of electronics. This cannula, composed of a resilient elastomer-hydrogel hybrid, changes its permeability in response to glucose levels. Tested in diabetic mice, it effectively maintains blood glucose levels within target ranges for approximately three days, offering a potential solution for user-centric insulin delivery.

INTRODUCTION:

In good health, the well-functioning interactions of the pancreas play a key role in ensuring glucose homeostasis by producing and releasing hormones in response to physiological fluctuations in blood glucose levels (BGLs) (1–3). In the case of type 1 diabetes (T1D), the autoimmune destruction of insulin-producing β-cells in the pancreas results in a loss of glycemic control (4–6). In the absence of a cure, reversing insulin deficiency is primarily achieved through the lifelong administration of exogenous insulin via subcutaneous insulin infusions or multiple daily insulin injections (7, 8). Unfortunately, insulin replacement therapies have their own limitations; the risk of hypoglycemia and hyperglycemia from patient-directed excess or insufficient dosing, respectively, continues to pose a significant barrier in achieving tight BGL control with insulin monotherapy (9–11). Not to mention that ideal diabetes management is contingent upon strict and frequent compliance to psychologically straining caretaking protocols, imposing an incredible burden of self-care on T1D patients (12, 13). As such, there is a need for (new) technologies and therapies that can mimic dynamic β-cell function by affording tight glycemic control through continual, autonomous insulin delivery with minimal burden to the patient.

To this end, significant efforts have been devoted to the development of closed-loop insulin delivery systems, or artificial pancreas (AP) systems, namely bioartificial AP technologies based on polymer-encapsulated islets/stem cells (14–16), fully synthetic AP therapies based on insulin-releasing polymers (17–23), and electromechanical AP systems based on a continuous glucose monitor (CGM) and computer-controlled algorithms connected to an insulin delivery device (24–27). Since T1D occurs from a loss of function of ß cells, restoration of endogenous insulin secretions though bioartificial AP therapies is a promising approach to achieve optimal glycemic control, particularly since even small doses of endogenous insulin secretions in T1D patients lead to less frequent severe hypoglycemic episodes and fewer overall diabetes related complications (28). Isolated islets transplanted to the liver via the portal vein have achieved years-long insulin independence in many patients (29). However, this cell therapy is limited by the need for potent immunosuppression and donor pancreata, which is a scarce source of tissue (30). Several groups, including our own, have investigated encapsulating (enclosing) islets within protective membranes to remove the need for immunosuppressive medication, as has been extensively reviewed (16, 17, 31). Despite several promising animal studies and pre-clinical advances, it has been challenging to develop a clinically viable, long-term functional encapsulation device.

In recent years, rationally focused development efforts have yielded longer wear times for CGMs (32, 33) and an assortment of commercially available electromechanical AP systems (34–36). Despite clinical trials not consistently concluding that pump therapy yields better glycemic control than multiple daily injections (37), it is generally accepted that the subcutaneous route for glucose sensing and insulin delivery via electromechanical APs is currently the most promising for widespread clinical use (38–41). Surveyed diabetic patients and their caretakers ideally desire an all-in-one AP that is ‘small [in] size,’ ‘discreet [in] appearance’ and ‘effective’ (42); however commercially available electromechanical AP systems fall short of these favored attributes, with several factors contributing to the hinderance of their widespread adoption including their large size (36), their susceptibility to inaccuracy during unavoidable behaviors (e.g., sleeping) (43, 44), and the higher monetary costs compared to that of multiple daily insulin injections (45). In addition, one disadvantage of current electromechanical AP systems is that glucose-sensing and insulin delivery are performed by separate, stand-alone hardware components that have different lifetime spans and require separate insertion at two different subcutaneous tissue sites (46). With clinical studies demonstrating that glucose concentrations measured at the site of subcutaneous insulin infusion closely reflect glucose levels in the blood (47–49), further integration of AP components by combining insulin delivery and glucose sensing into a single platform, for example (50–59), might broaden the appeal of AP technologies and encourage their adoption. Thus, the objective of our study was to develop a user-friendly, easy-to-implement, and minimalistic system for automated insulin delivery.

We specifically sought to engineer an electronics-free, single-port glucose-responsive cannula for insulin delivery. Cannulas —a critical component of electromechanical AP systems through which insulin is infused— are typically inserted subcutaneously through an introducing needle and held in place by an adhesive patch placed on the skin (60). We hypothesized that a cannula that itself changes permeability to insulin in response to glucose level changes may provide a minimally sophisticated approach for automated insulin delivery. The bottleneck remains that most insulin-permeable, glucose-responsive materials do not exhibit sufficient strength and toughness to withstand handling, insertion, and retrieval as a thin-walled cannula (17). To overcome this challenge, we developed a tough, elastomer-hydrogel hybrid material capable of being formed and shaped into cannulas through a one-pot solvent exchange method. Combining elastomers and hydrogels at a molecular level through an interpenetrating network hybrid material offers a balance between mechanical strength and selective permeability towards insulin, with further fluoro-phenylboronic acid (FPBA) functionalization endowing the cannula with glucose-responsive permeability. We first demonstrated the glucose binding capacity of the material and glucose-responsive insulin release from the cannula under physiologically relevant glucose environments. Next, we examined the in vivo therapeutic efficacy of the cannula in a T1D mouse model. When evaluated as either a subcutaneous implant with both ends sealed or as a transcutaneous, externally refillable insert, both cannulas maintained glycemic levels in mice within normal ranges for up to three days (the current lifetime of conventional insulin delivery cannulas recommended by the American Diabetic Association) (60, 61). Additionally, the ends-sealed cannula maintained normoglycemia when facing glucose challenges, providing a proof-of-concept for the potential application of this glucose-responsive cannula for electronics-free, user friendly T1D glycemic management.

Materials and Methods:

Materials

The following products, chemicals, and reagents were used in this study: Acrylamide (AAm), N, N’-methylenebis(acrylamide) (MBAA) and thionyl chloride were purchased from Sigma-Aldrich (Saint Louis, U.S); Irgacure 2959 was purchased from BASF (Florham Park, U.S.); HydroMed^™ polyurethane (D3) was purchased from AdvanSource Biomaterials (USA); 4-carboxy-3-fluorphenylboronic acid was purchased from Combi-Blocks (San Diego, USA); Pierce^™ Coomassie Plus (Bradford) Assay Reagent and human recombinant insulin (catalog no. A11382II) were purchased from Thermofisher Scientific (Eugene, US); Ethylene diamine was purchased from VWR International (Corning, US); and N-octyl-glucopyranoside was purchased from Krackeler Scientific, Inc (US). Cyanine5 NHS ester was purchased from Abcam (US). Glucose was purchased from Mallinckrodt Pharmaceuticals (Dublin, Ireland). SuperWeld^™ Light Activated Instant Glue (UV glue), 22-gauge mouse VAB^™ buttons, and silicone tubes were purchased from J-B Weld (Texas, US), Instech Laboratories, Inc. (US), and SMi Specialty manufacturing, Inc. (Saginaw, US), respectively.

Synthesis of 4–2-acrylamidoethylcarbamoyl-3-fluorophenyl boronic acid (FPBA)

FPBA was synthesized according to a previously described method (62–64). Briefly, 4-carboxy-3-fluorphenylboronic acid (8 g) was added to a dried, three-neck round bottom flask and refluxed under a nitrogen atmosphere at 88 °C. Thionyl chloride (150 mL) was then added, and the suspension stirred and refluxed under a nitrogen atmosphere at 88°C for one hour. After an hour, nitrogen was removed, and the reaction proceeded for 24 h at 88°C to produce 4-carbonyl-3-fluorocyclohexyl boronic acid. The remaining thionyl chloride was evaporated, and the contents of the flask was resuspended in 60 mL of distilled tetrahydrofuran. The suspension was then added (dropwise) to cooled ethylene diamine (200 mL) in the presence of triethylamine (10 mL). The suspension was stirred in ice for 30 minutes and then at room temperature overnight. After overnight stirring, the ethylene diamine was evaporated, and the solution was acidified (pH 4) using 1M HCL. A white precipitate byproduct was filtered out. The remaining solution was then concentrated and stored at 4 °C overnight. The collected crystals were dissolved in water and recrystallized twice to produce 4-aminoethylcarbamoyl-3-fluorophenylboronic acid (AECPBA). AECPBA crystals were collected and dissolved in 150 mL of sodium bicarbonate (100 mM, pH 10) and 62.5 mL of 1 M NaOH in an ice cooled bath. Acryloyl chloride (5 mL) was then added dropwise. The mixture was stirred for 30 minutes in the ice bath, and then at room temperature for 5–6 hours. After stirring, the solution was concentrated and acidified (pH 4) using 1M HCL. The solution was once again concentrated and stored at −20°C overnight to produce a white crystalline product (FPBA).

Preparation of elastomer materials

A commercially available polyurethane (PU, HydroMed D3) was dissolved in a 95:5 mixture of ethanol (EtOH) and Mili-Q water to obtain an elastomer solution (w/v%). For mechanical characterization tests (Figure 2), the concentration of PU was kept at 12.5% (w/v%). For all glucose-responsive materials, the concentration of PU was kept at 18% (w/v%). Films and tubes were prepared using a solution-casting method at room temperature After evaporation of the solvent mixture, molds were immersed in water, allowing the material to crosslink, swell, and release from the molds. The prepared tubes or films were then stored in water for a subsequent three days, with water changed daily to allow for unreacted monomers to leach out.

Figure 2. Mechanical properties of the hydrogel-elastomer material.

(A) Images of a cannula after removal from the fabrication mold (top left) and imaged to demonstrate uniformity along the cannula wall (top right; scale bar= 2000 μm). The cannula is mechanically robust and can be manually stretched without breakage (bottom). (B) Graph of tensile stress vs tensile strain of the hybrid material as a function of hydrogel concentration (w/v%). (C) Loading-unloading curve of a hydrogel-elastomer sheet made from 12.5% (w/v%) elastomer and 12% (w/v%) hydrogel. (D) Water content of the hybrid material as a function of hydrogel concentration. (E) Ultimate tensile strength, and (G) Young’s modulus of the hybrid material as a function of hydrogel concentration. For all experiments, the elastomer (PU-D3) concentration is fixed at 12.5% (w/v%). Data points are means ± SD (n = 3).

Preparation of hydrogel-elastomer materials

To introduce glucose sensitivity to the system, FPBA monomers (6% (w/v%)), MBAA (0.01% (w/v%)), and IR2959 (0.4% (w/v%)) were added to PU-D3 solutions as the glucose binding moiety, crosslinker, and initiator, respectively; AAm (2% (w/v%)) was added to facilitate mixing and incorporation of the hydrogel material. Contents were stirred until the pre-gel solution became clear. Solutions were then sonicated for 30 minutes at room temperature using a VWR Ultrasonic Cleaner for degassing purposes. Films and tubes were prepared using a solution-casting method at room temperature. After evaporation of the solvent mixture under ambient room temperature conditions, UV-polymerization of the hydrogel material was triggered using an Omnicare UV for 300 seconds. After UV treatment, the elastomer-hydrogel thin films/tubes were immersed in DI water for three days to ensure full swelling and crosslinking of the elastomer material. Unreacted monomers were removed through daily water changes.

Characterization of mechanical properties

Mechanical property tests of the hybrid materials were carried out in air at room temperature. Tensile strength tests were carried out using an Instron 4680 mechanical testing instrument with a grip-to-grip speed of 15 mm/min and a 10 N load cell. The formation of an interpenetrating hybrid network was evaluated using FT-IR (Fourier transform infrared spectroscopy, Bruker Vertex V80V vacuum FT-IR system). Young’s Modulus was measured from stress-strain curve fitting using the Bluehill^® Universal Materials testing software. Water content was evaluated as the ratio of the weight of water to the weight of the sample. Specifically, after fully swollen in water, the weight of swollen samples was measured and recorded as M_s. The samples were then stored and dried in an 80°C oven for 6 hours. The weight of the dried samples was then recorded as M_d. The water content of the samples was then calculated as the amount of water (M_s - M_d) divided by the total weight of the sample (M_s), expressed as a percentage. To evaluate the microstructure of the membranes, samples were incubated in glass vials with 10 mL phosphate buffer saline (PBS, pH 7.4) containing different glucose concentrations (100 or 400 mg/dL) for 24 hours. Membranes were then lyophilized, followed by imaging using a field emission scanning electron microscope (LEO 1550 FESEM) to observe membrane morphology.

Preparation of insulin stock solution

Insulin was prepared according to a previously described protocol (65). Briefly, insulin (25 mg) and n-octyl-glucopyranoside (3.65 mg) were dissolved in 0.1 M NaOH aqueous solution (600 μL). 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) (HEPES) (12.6 mg) was then added, and the volume was brought to 1 mL by a slow addition of 0.1 M HCL. A transparent solution was obtained at ~ pH 7. Sterile saline was used to further dilute the sample as needed.

Formation of cannulas

Cannulas were formed using a solvent casting method as describe above. Ends-sealed cannulas were thermo-sealed on one end, filled with insulin, and thermo-sealed on the filling end to create a closed device. Transcutaneous, externally refillable cannulas were formed by adding a silicone tube (SMI Silicone Tubing Tube 0.025” X 0.047” 50D CL) to an open end of the solvent-casted cannula. The silicone tube was secured in place with the addition of UV glue. The other end of the cannula (not attached to the silicone tube) was then sealed via thermo-sealing. Then, the silicone-end of the cannula was slipped on to the 22-gauge adapter of the mouse VAB^™ button (INSTECH), creating the final, ready-to-use product.

Blood glucose reduction potential studies for insulin stability

The stability of insulin in the devices was evaluated using a blood glucose reduction protocol. Devices were first prepared as described above and stored under static conditions at either 4°C or 37°C to evaluate long term storage and in vivo temperature effects of the enclosed insulin, respectively. At pre-determined time points, insulin from the devices was extracted and subsequently subcutaneously injected into diabetic mice at a set dose of 0.05 mg. The blood glucose of the mice (as measured with a Clarity GL2Plus glucose meter) was monitored prior-to and one hour after the insulin injection. The reduction in BGL was assumed to occur from the injected insulin. ‘As prepared’ indicates an ends-sealed closed cannula that was loaded with insulin, thermo-sealed, and immediately had the insulin extracted after thermo-sealing. The BGL reduction of a freshly prepared insulin solution (or native insulin solution) was used as control.

Glucose absorption study with glucometer

Membranes (1cm x 1cm x 0.5 mm) were placed in glass vials with 10 mL phosphate buffer saline (PBS, pH 7.4) containing different glucose concentrations (100, 200, or 400 mg/dL). The vials were incubated at room temperature, and the glucose concentration of the supernatant was monitored at timed intervals using a glucose meter (Clarity BG 1000, CD-BG1). The concentration of the solution was calculated using an established standard curve.

Glucose absorption study with glucose oxidase kit

Glucose absorption was determined following the outlined protocol for the Glucose Oxidase (GO) Kit (Sigma-Aldrich, catalog no. GAGO20). Briefly, membranes (1cm x 1cm x 0.5 mm) were placed in glass vials containing 10 mL PBS (pH 7.4) and different glucose concentrations (100, 200, or 400 mg/dL). The vials were incubated at room temperature. At a timed interval, a 1 mL supernatant solution from each sample was collected and pipetted into a test tube. Assay Reagent (2mL, GO kit) was added to the collected supernatant, an incubated at 37°C for 30 minutes. The reaction was stopped after 30 minutes by the addition of 2 mL of 6 M H₂SO₄. The absorbance of each sample was measured at 540 nm using UV/Vis spectrophotometry (Beckman Coulter DU 730). The concentration of the solution was calculated using an established standard curve.

In vitro insulin release study

Insulin release was determined as a function of glucose concentration over time. Devices loaded with insulin were placed in centrifuge tubes containing the release medium (2 mL of PBS, pH 7.4) at different concentrations of glucose (0, 100, or 400 mg/dL). Samples were thermo-sealed, incubated at 37°C, and rotated at 70 revolutions per minute (RPM) for the duration of the experiment. At timed intervals, a clear supernatant (50 μL) was collected and added to Coomassie blue (300 μL); 50 μL of fresh release medium was added to each centrifuge tube following collection of the supernatant. Absorbance of the solution was measured at 595 nm, with the concentration calibrated using an established standard curve. The glucose-responsiveness of the devices (R) was calculated as the ratio of insulin release under hyperglycemic glucose (400 mg/dL) to normal glucose (100 mg/dL).

Animal Studies

All animal experiments were approved by the Cornell University Institutional Animal Care and Use Committee. The in vivo efficacy of the glucose-responsive insulin delivery devices was evaluated on streptozotocin (STZ)-induced adult diabetic mice (male C57BL/6J mice; eight weeks old; The Jackson Laboratory). Typically, after overnight fasting, mice were intraperitoneally injected with STZ (140 mg/kg). After one week, mice with a fasting blood glucose level higher than 300 mg/dL were confirmed as diabetic mice and used for further experiments.

Cannula implantation

Cannulas were implanted in the subcutaneous space of mice. Briefly, a sharp sterile surgical blade was used to make a small incision along the back of the mice approximately one centimeter from the spine. A subcutaneous pocket was then made by blunt dissection. The insulin or saline containing cannulas were then placed into the pocket using blunt-end tweezers. The subcutaneous pocket was closed with Nylon 6 sutures.

Blood glucose control in diabetic mice

Diabetic mice were allocated to different groups and were treated with subcutaneously injected native insulin or with implanted cannulas containing 1.5 mg of human recombinant insulin (100 μL of a 15 mg/mL solution) for ends-closed cannulas or 0.375mg (50μL of a 7.5 mg/mL solution) for transcutaneous cannulas. FPBA cannulas were used as glucose-responsive test groups, and PU cannulas were used as controls. The BGLs were monitored using a Clarity GL2Plus glucose meter.

Cy5-Insulin and in vivo imaging

30 mg of human recombinant insulin (catalog no. A11382II) were dissolved in 10 mM sodium carbonate buffer (30 mL, pH ~8). 150 μL of Cyanine5 NHS ester (Cy5) (5 mg/mL dimethyl sulfoxide) was added to the mixture and stirred overnight; the reactive dye contains an NHS ester group, enabling labeling of primary amino-bearing groups present at the N terminus and lysine side chains on insulin (66). The contents were then dialyzed against DI H₂O (3× 2L). The resultant solution was lyophilized to obtain Cy5-labeled insulin. The fluorescently labeled insulin was imaged using an IVIS-Spectrum optical imager.

Intraperitoneal glucose tolerance test

The intraperitoneal glucose tolerance test (IPGTT) was performed to confirm the in vivo glucose-responsive nature of the implanted devices. Prior to the experiment, mice were fasted overnight and then allocated to different groups and treated with subcutaneously implanted cannulas. Once normoglycemia (BGL: 100 mg/dL < x 200 mg/dL) was achieved approximately one hour after device implantation, a glucose solution was injected intraperitoneally into all mice at a dose of 1.5 g/kg. Glucose levels were then monitored at specific time points to assess the return to normoglycemia following the glucose injection. The IPGTT was performed on healthy, non-diabetic mice as control.

Serum Insulin Level Measurements

Serum insulin levels were measured when performing intraperitoneal glucose tolerance tests to evaluate glucose homeostasis and insulin secretions in vivo using ends-sealed cannulas. In the experiment, five STZ-induced diabetic mice were given glucose-responsive (e.g., FPBA containing) ends-sealed cannulas for the experimental group and five healthy mice were used as controls. To measure the serum insulin concentration of the mice, 50 μL of whole blood were drawn from the submandibular vein of the mice at indicated time points. The samples were stored at −20°C until assayed. After collection of the whole blood, samples were allowed to clot by leaving the samples undisturbed at room temperature for 30 minutes. Samples were then centrifuged at 3000 RPM for 15 minutes, and then the serum supernatant was collected. The serum insulin concentration was then measured using a Human Insulin ELISA kit (ALPCO).

Histological Analysis

For histological analysis, samples were embedded in HistoGel^™ (Thermo Scientific) before being fixed with neutral buffered formalin and embedded in paraffin. Sections were subsequently stained with Hematoxylin and Eosin (H&E) and Masson’s Trichrome and sectioned by Cornell’s Histology Core Facility. H&E and Masson’s Trichrome staining slides were imaged by digital inverted microscope (AmScope IN300TC).

Statistical analysis

All data shown are means ± standard deviations. Sample size (n) is 5 unless otherwise indicated. Student’s t-tests or ANOVA with Tukey post hoc tests were used to analyze the difference between two or more groups, respectively. Statistical significance was evaluated as *, **, ***, ****, and ns, which represented p-values < 0.05, < 0.01, < 0.001, < 0.0001, and > 0.05, respectively. The GraphPad Prism software used for statistical tests.

RESULTS:

Design of a glucose-responsive, elastomer-hydrogel hybrid cannula

To develop a cannula suitable for glucose-responsive insulin delivery, two design features were essential: (1) high mechanical strength to support subcutaneous insertion and replacement, and (2) ready permeability to insulin that is sensitive to environmental glucose conditions. These two properties generally do not coexist within the same material, and oftentimes compromise each other: conventional steel or Teflon cannulas have robust mechanical properties but have low permeability; hydrogels are permeable but are challenging to form and shape into cannulas (67). Furthermore, while several different hydrogel-based materials have been explored for insulin delivery (68–71), their path to clinical translation remains unclear, as insulin therapies require precise and repeated dosing, an attribute that is difficult to achieve with hydrogels given that they often exhibit slow degradation rates that make them unsuitable for repeated administration (72–74). Moreover, non-degradable hydrogels tend to suffer from poor mechanical properties that are further exacerbated by the harsh in vivo milieu (75–77), making their insertion and replacement every few days unplausible (78, 79).

To address the first design feature, we developed a facile, one-pot strategy for constructing a robust elastomer-hydrogel hybrid cannula through a solvent exchange process (Figure 1). We hypothesized that combining elastomers and hydrogels at a molecular level through an interpenetrating network would offer tunable balance between mechanical strength and selective permeability towards insulin (Figure 1a). Additionally, to endow the glucose responsiveness of this material, we functionalized the hydrogel with fluoro-phenylboronic acid (FPBA), satisfying design feature number two. Under hyperglycemic conditions, glucose can reversibly and dynamically bind to FPBA, rendering the phenylboronic acid more hydrophilic, while increasing the negative charge density of the system (80–82). This increase in hydrophilic character increases the permeability of the carrier material (e.g., the cannula) to release enclosed insulin (Figure 1b, S1). Meanwhile, swelling is reversed under hypoglycemic conditions from a loss of negative-negative charge repulsion, mitigating a potential risk of over-dosing on insulin on the return to normoglycemia. Although most phenylboronic acids have a pKa ranging from 4.5 to 10 (83), the chemical structure of FPBA was selected and synthesized (Figure S2) such that the monomer could easily be incorporated into our final material via UV irradiation-triggered crosslinking, while undergoing glucose-responsive swelling-behavior under physiologically relevant conditions (e.g., pKa ~7) (63, 84, 85).

Figure 1. Schematic of glucose-responsive mechanism and one-pot fabrication procedure of elastomer-hydrogel cannula.

(A) To form the hydrogel-elastomer hybrid material, hydrogel monomers are dissolved within a precursor elastomer network. Crosslinking of the hydrogel network via UV irradiation traps the hydrogel within the elastomer material. After crosslinking of the elastomer network, a final interpenetrating “hybrid” material is formed. This hybrid material offers a tunable balance between mechanical strength and permeability towards insulin. (B) Binding of glucose to the hydrogel-elastomer material occurs via the fluoro-phenylboronic acid (FPBA) monomer incorporated in the hydrogel network. Binding of glucose and FPBA renders the matrix more hydrophilic while increasing the negative charge density of the system. This increase in negative charge density increases the osmotic pressure in the system, leading to volumetric swelling of the material. The subsequent increase in permeability allows for enclosed insulin to be released. (C) Schematic illustration of the one-pot, dip-coating cannula fabrication procedure. (D) Images of different sized ends-closed cannulas filled with water to demonstrate scale-up potential of the one-pot dip coating fabrication procedure.

This diffusion-based insulin release mechanism provides a simple method to replicate pancreatic function, as compared to more complicated fabrication, storage, and algorithm-controlled sensing mechanisms of enzyme-based “smart-insulin” delivery platforms (17). Additionally, the entire surface of the device is composed of the glucose-sensing material, consequently providing a continuous and large surface area for glucose-binding without the need for harsh, multi-step surface modification procedures or additional mechanically reinforcing components within the device. This design should improve the absorption of insulin by distributing the liquid into the subcutaneous tissue along the entire length of the cannula; the permeable structure allows insulin to percolate to any area where the tissue will expand, and this flow should continue despite kinking of the cannula due to continual permeability along the wall of the cannula. This should alleviate the burden of cannula-non-function experienced by pump users because of occlusions or kinks (61, 86, 87). Ideally, such a cannula should facilitate more rapid and robust glucose-detection and subsequent insulin release.

Fabrication of the cannula

The glucose-responsive cannula formed from the hydrogel-elastomer hybrid material was fabricated through a one-pot solvent exchange method (Figure 1c). The fabrication process began by incorporating and dissolving the hydrogel network components —in our case a glucose binding moiety (FPBA), crosslinker N, N’-methylenebis(acrylamide) (MBAA), and photo-initiator Irgacure 2959 (IR2959)— to a pre-prepared elastomer (Hydromed, D3 polyurethane) solution. Given the hydrophobic nature of the elastomer and FPBA, acrylamide (AAm) (2 w/v%) was added to the solution to facilitate mixing and incorporation of the hydrogel material. A solvent-casting method was then employed in which a mold with the desired geometry and shape was selected and dip-coated with the dissolved hybrid solution (Figure 1c). Then, the solvent was evaporated at ambient temperature before UV treatment using an Omicare UV. Under UV irradiation, the hydrogel matrix was photo-crosslinked to entrap the hydrogel within the elastomer network, thereby forming an interpenetrating network material (Figure 1a). The presence of a second network in the hybrid samples was evaluated using Fourier-Transform Infrared Spectroscopy (FTIR), with the FTIR spectrum of the hybrid elastomer hydrogel exhibiting three distinct peaks at 1663 cm-1, 1407 cm-1, and 1375 cm-1 corresponding to the C=O, B-O bonds in the FPBA polymer chains (Figure S3). After immersing the cross-linked material in water, the final cannula swelled and could be peeled off from the fabrication mold, producing a uniform and ready-to-use cannula (Figure 1c). The simple fabrication process is conducive to scale-up or scale-down requirements, as the final form of the device is contingent upon the mold selected for fabrication (Figure 1d). While we explored the application of this material for glucose-responsive insulin delivery, functionality of the material is contingent on the monomers used in the hydrogel component. As such, one could functionalize the material with other ‘responsive’ properties (i.e., pH-responsiveness, temperature sensitivity, etc.) so long as the selected hydrogel monomer(s) are water soluble and form a network through UV irradiation. Such materials might have other interesting stimuli-responsive therapeutic applications.

Characterization of mechanical properties

To quantify the mechanical robustness of the hydrogel-elastomer material, we measured the tensile strength of the hybrid material as a function of hydrogel concentration (Figure 2). The concentration of the elastomer (Hydromed, D3) was fixed at 12.5% (w/v%) for these mechanical characterization tests. This was because preliminary tensile tests of the pure elastomer material (polyurethane or PU) showed that maintaining the concentration of material between 10–30% (w/v%) provided an ideal balance between mechanical strength (Figure S4) and viscosity conducive to forming cannulas through the dip-coating fabrication procedure.

Compared to a pure hydrogel sample, the mechanical strength of the hybrid material was greatly enhanced (Figure 2b) and demonstrated remarkable recoverability from deformation under cyclic loading (Figure 2c). While the energy dissipated in the second cycle was observed to be lower than that of the first cycle, it remained approximately constant for subsequent cycles (Figure 2c). It is possible that rearrangements in the interpenetrating network structure during the first cycle contributed to this phenomenon, and this reconfiguration contributed to the material’s ability to maintain elasticity and strength under subsequent cyclic loading (88). Advantageously, these features contributed to the material’s ease of handling compared to its pure hydrogel counterpart when formed into a cannula (Figure 2a). Moreover, while the elastomer concentration was fixed, the mechanical properties of the final hybrid material were broadly subject to the amount of hydrogel in the system. We postulate that incorporating a higher concentration of the hydrogel component dilutes the density of the original elastomer network, thereby softening the material (Figure 2f) and compromising the overall tensile strength of the final product (Figure 2e). Additionally, incorporating a higher concentration of hydrogel into the final construct increased the overall water content and hence permeability of the material (Figure 2d). Therefore, a balance between hydrogel content and the mechanical properties of the material should be considered for case-specific scenarios. The final concentrations of hydrogel (8% w/v%) and elastomer (18% w/v%) were selected for an optimal balance of mechanical strength and FPBA content for glucose responsiveness for our application as an insulin infusion cannula.

Insulin stability and glucose-responsiveness in vitro

After confirming the mechanical robustness of the elastomer-hydrogel material and its ability to be formed into cannulas, we explored the suitability of the material for insulin delivery. This first entailed evaluating the stability of insulin within the cannula (Figure 3). It has been shown that highly concentrated insulin solutions can quickly undergo aggregation (e.g., fibrillation) when contacting hydrophobic surfaces (89, 90). This was of concern given both the high insulin concentration loaded in the cannulas and the hydrophobic nature of the hybrid-material, as insulin aggregation could potentially block the pores of the glucose-responsive membrane, leading to poor insulin release or the release of non-active insulin. To prevent insulin aggregation, a previously established strategy was implemented to stabilize insulin through the addition of the non-ionic surfactant, n-octyl-β-glucopyranoside (65, 91). Incorporating the surfactant permitted insulin to be stored stably within ends-sealed cannulas at 4°C for one month and at 37°C for one week (Figure 3b, 3c). Specifically, the blood glucose reducing potential of a freshly prepared insulin solution was comparable to the reducing potential of insulin stored in the device, demonstrating that neither the fabrication procedure nor the long-term storage of insulin within the ends-sealed cannulas affected the insulin’s therapeutic potential (Figure 3a).

Figure 3. In vitro evaluation of glucose-responsive properties and insulin stability tests.

Blood glucose level (BGL)-reducing activity in diabetic mice of (A) freshly prepared insulin and insulin extracted from freshly prepared ends-sealed cannulas, (B) from insulin stored at 37 °C for one week and (C) at 4°C for 4 weeks. Initial BGLs were compared with BGLs at 60 min post injection of insulin (0.05mg). Glucose concentration-dependent glucose binding capacity of (D) control and (E) glucose-responsive sheets as measured via a glucometer. (F) Comparison of glucose absorbed by control (PU) and glucose responsive (FPBA) sheets following 1h exposure to 200 mg/dL of glucose measured via a glucometer (blue) and a GOx assay (red). In vitro accumulated release of insulin from ends-closed cannulas made of control (G-H) and glucose-responsive (J-K) materials loaded with 0.750 mg (50 μL of 15 mg/mL) of insulin under different glucose concentrations. Pulsatile insulin release from control (I) and glucose-responsive (L) ends-closed cannulas loaded with 1.5 mg of insulin by alternating the glucose concentration for three consecutive cycles; cannulas were incubated in each solution for 15 min. Data points are means ± SD (n = 3). ns p > 0.05, * p < 0.05, * p < 0.01, *** p < 0.001, **** p < 0.0001.

Since diffusion-based insulin release from the cannulas was hypothesized to occur from the volumetric swelling of the matrix caused by the complexation of glucose to the FPBA monomers, we investigated whether the hybrid material was capable of binding to glucose after fabrication. To do so, sheets of the material were immersed in phosphate buffer saline (PBS, pH 7.4) containing different clinically relevant concentrations of glucose. Results were consistent for all membranes being able to absorb glucose (Figure 3d, 3e). However, only membranes containing FPBA, the glucose-binding element, absorbed glucose in a glucose responsive fashion, with more glucose binding occurring under hyperglycemic conditions than in hypoglycemic conditions (Figure 3e). Glucose absorption measured via both a glucometer and a glucose oxidase (GOx) assay supported the glucose concentration-dependent absorption of FPBA membranes (Figure 3f). Accompanying increases in permeability evaluated by changes in water content were also observed in the membranes, with only membranes containing the glucose-binding element showing increased swelling in response to rising glucose concentrations, although the response was only statistically significant under hyperglycemic conditions (Figure S5). These findings demonstrated that the incorporation of FPBA endowed the hydrogel-elastomer material with glucose-responsive changes in permeability.

Given our intention to use this tough, elastomer-hydrogel hybrid material as a glucose-responsive insulin cannula, we then assessed the insulin release potential using cannulas formed from the hybrid material. Thermo-sealed, ends-closed cannulas loaded with insulin were placed in centrifuge tubes containing a release medium (2 mL of PBS, pH 7.4) with different glucose concentrations (0, 100, or 400 mg/dL) and incubated at 37°C, RPM 70. While the rate of insulin release (R) was variable based on the concentration of insulin loaded in the cannulas (Figure S6, Table S1), generally, higher rates of insulin release were obtained at hyperglycemic (400 mg/dL) glucose levels than normoglycemic levels (100 mg/dL) for FPBA cannulas (Figure 3j, 3k). Moreover, the insulin release profile of the FPBA cannula exhibited a typical pulsatile pattern when glucose concentrations were alternated between normal (100 mg/dL) and hyperglycemic levels (400 mg/dL) for several cycles, supporting the on-off release function of the cannula (Figure 3l, S7). In contrast, non-glucose responsive cannulas (PU, without FPBA) used as controls demonstrated negligible glucose-responsive insulin release delivery, releasing similar amounts of insulin independently of the environmental glucose concentration (Figure 3g, 3h). Interestingly, there seemed to be an indication of ‘responsiveness’ in the first few cycles of alternating PU device exposure to high versus low glucose levels (Figure 3i). It is plausible that this pulsatile response is caused by osmotic pressure induced-insulin release from sudden changes in glucose concentrations in the environment. However, in contrast to FPBA devices, this on-off function is substantially more attenuated. Collectively, these results substantiated that the ends-sealed cannulas made of the hybrid material supported glucose-mediated insulin release in vitro under physiologically relevant parameters.

In vivo evaluation of glucose-responsive performance

After exploring the mechanical properties of the hydrogel-elastomer material and its corresponding in vitro glucose-responsive properties as a cannula, we explored its therapeutic potential (Figure 4). Using mice with diabetes induced by streptozotocin (STZ) as our T1D model, we evaluated the BGL correcting potential of the cannula as a subcutaneously implanted, ends-sealed cannula (Figure 4a, S8a). Diabetic mice were randomly assigned to be treated with either a subcutaneous injection of PBS or insulin (0.05 mg) or implanted (subcutaneously) with an ends-sealed cannula at an insulin dose of 1.5 mg (Figure 4b). The BGLs of all treated groups decreased below 200 mg/dL within one hour of receiving insulin, indicating the rapid blood sugar reducing capacities of insulin-based treatments (Figure 4c). However, mice treated with an insulin injection showed only transient glycemic control, with a return to hyperglycemic BGLs within two-to-three hours after receiving the insulin injection (Figure 4c). In contrast, mice with ends-sealed cannulas showed prolonged glycemic control, on average lasting for three days (Figure 4b). Importantly however, when comparing the glucose-responsive cannulas (FPBA) and the control cannulas (PU), the FPBA-based cannulas prevented hypoglycemic episodes, which remains one of the outstanding challenges of many “smart” insulin delivery technologies (17), and maintained glucose levels within the normal range (100 mg/dL < x < 200 mg/dL) for approximately 72 hours before the mice returned to hyperglycemic BGLs (Figure 4d, 4e). This BGL regulation was much longer and more robust than either insulin injections or control (PU) cannulas, as mice in the PU group experienced severe hypoglycemia from unregulated insulin delivery for the duration of the experiment, gravely diminishing their survival (Figure 4f). Live imaging using Cyanine5 NHS ester-labeled ‘active’ insulin (Figure S9a, S9b) further confirmed the prolonged delivery of insulin from reservoir devices compared to injected insulin (Figure S9c). “Failure,” or the return to hyperglycemic BGLs, likewise seemed to occur from decreasing insulin from within the cannulas, as evidenced by diminished average relative fluorescence intensity with increasing BGLs (Figure S9d).

Figure 4. In vivo evaluation of glucose-responsive properties of ends-closed cannula.

(A) Schematic of timeline for ends-closed cannula trial in diabetic mice. (B) Blood glucose levels (BGLs) in Streptozotocin (STZ)-induced diabetic mice after treatment with Phosphate Buffer Saline (PBS), subcutaneous insulin injections (0.05 mg), control ends-closed cannulas (PU), or glucose-responsive ends-closed cannulas (FPBA) loaded with 1.5 mg (100 μL of 15 mg/mL) of insulin. (C) Mice given subcutaneous insulin injections quickly return to hyperglycemic BGLs, while cannula-implanted mice continue controlling BGLs. (D) Glucose-responsive cannula treated mice (FPBA) controlled BGLs within the normoglycemic range throughout the three-day period, whereas control mice (PU) fell below the range into hypoglycemic levels. (E) Time spent in normoglycemia per treatment group; normoglycemia is defined as BGLs within 100 mg/dL < x < 200 mg/dL. (F) Survival curves for control and glucose-responsive ends-sealed cannulas. (G) In vivo intraperitoneal glucose tolerance test (IPGTT) in healthy mice and diabetic mice at 1 h post-administration of control ends-closed cannulas (PU, n=5) or glucose-responsive ends-closed cannulas (FPBA). Glucose dose: 1.5 g kg.⁻¹(H) Responsiveness in diabetic mice calculated as the area under the curve (AUC) from 0–120 min. (I) Serum insulin levels (red) and BGLs (black) following IPGTT test at a dose of 1.5 g kg⁻¹ of glucose for FPBA treated mice. Data points are means ± SD (n = 5 mice) * p < 0.05, *** p < 0.001, **** p < 0.0001. Grey area represents the normoglycemic range as defined by BGLs within 100 mg/dL < x < 200 mg/dL.

Rapid in vivo glucose-responsive insulin release from the cannulas is essential for maintaining normoglycemia when facing prandial glucose challenges. As such, we then performed intraperitoneal glucose tolerance tests (IPGTTs) to assess the in vivo glucose-responsive performance of the cannulas (Figure 4g-i). Blood glucose peaks were observed for healthy mice and mice treated with FPBA cannulas, with both groups reestablishing normoglycemia in a brief period (Figure 4g). Notably, serum insulin levels increased in response to the acute glucose bolus for mice with an FPBA cannula (Figure 4i). On the other hand, a characteristic glucose peak following the glucose injection was visibly absent for the PU cannula groups. We postulate that this is likely because the glucose injection served as an immediate but brief reprieve from hypoglycemia caused by the unregulated insulin release from the PU cannulas. Without any insulin regulating abilities, this group returned to low BGLs soon thereafter (Figure 4g). Encouragingly, these data indicated the potential of FPBA cannulas at controlling ‘prandial’ glycemic levels through in vivo glucose-responsive insulin delivery behaviors.

Importantly, one-week post-implantation, the ends-closed cannulas could be completely retrieved without tissue adhesion or gross deformation (Figure S8c, Supplementary Video 1-3). Hematoxylin and Eosin (H&E) and Masson Trichrome staining showed a non-uniform, but mostly thin layer of cells deposited on the cannula surface (Figure S8b, S10, S11). It is possible that a combination of the local trauma caused by cannula insertion, active release of insulin, and the hydrophobic nature of the material used to fabricate the cannulas contributed to the areas of higher cell deposition seen in certain regions (Figure S11). It is important to note, however, that the incorporation of the HydroMed^™ polyurethane layer within our material makes the devices susceptible to ethanol-induced degradation. This makes the devices vulnerable to degradation during histology fixation steps, given that ethanol is a commonly used coagulant fixative. To minimize dissolution, prior to sample fixing in formalin, freshly explanted devices were embedded in HistoGel^™. Generally, incorporation of the HistoGel^™ preserved most of the samples for a preliminary analysis of the fibrotic response against the implanted devices (Figure S8b). Although the cannulas are intended to be used for short-term use, further investigation into fibrosis-reducing materials are warranted to minimize interference between the cannula surface and the environmental interstitial fluid to maximize the glucose-responsive insulin release performance of the cannula.

A transcutaneous, externally refillable cannula

While the glucose-responsive performance of the cannula in diabetic mice was encouraging, the translational potential of this ends-sealed cannula was quite limited; as currently designed, the fully sealed cannula was not refillable. As such, the cannula would have had to be inserted subcutaneously every three days to offer longer-term glycemic control, which given the volume of the device, is a cumbersome and invasive ordeal. Not to mention that the ends-sealed cannula holds three-day’s worth of insulin at a given time, posing severe safety concerns in the case of device failure; the risk of overdosing on insulin due to cannula damage leading to leaking, for example, becomes a significant concern given that severe hypoglycemia can cause brain damage or even death (92). To facilitate the translational potential of this technology, we designed a cannula that was more representative of how typical single-port cannulas infuse insulin in commercially available infusion sets.

To do so, we attached our glucose-responsive cannula to the mouse VAB^™ button (INSTECH) (Figure 5). The mouse VAB^™ is a transcutaneous button that —when attached to our glucose-responsive cannula— permits quick, painless, aseptic filling and refilling of insulin in the inserted cannula via a syringe (Figure 5a). By simply modifying the dip-coating fabrication method, we can slip-on a medical grade silicone tube to the non-sealed end of the cannula (Figure S11b), which is then used to tether our cannula to the 22ga connector under the disk of the VAB^™ transcutaneous button, creating a closed system (Figure 5a). Although not required in clinical settings, the polyester felt attached to the transcutaneous button additionally aids to hold the device in place under the skin for improved patency. The glucose-responsive cannula, when attached to the button, provides the composite device with glucose-dependent insulin delivery (Figure 5a). Collectively, the design of this new prototype is more akin to current insulin infusion sets, whereby a transcutaneous system infuses insulin through a subcutaneous cannula on a continual basis. Ideally, this design set up could be used to support glucose control in large animal models, as rather than creating larger cannula prototypes, one could simply more frequently refill the devices based on increased metabolic demands for insulin.

Figure 5. In vivo evaluation of externally refillable transcutaneous cannulas.

(A) Schematic illustration of the method of action of the externally refillable, transcutaneous cannula. (1) The mouse VAB^™ transcutaneous button has a 22ga connector through which the glucose-responsive cannula can be attached to via a silicone adapter. (2) The external adapter can then be used for quick, aseptic filling and refilling of insulin of the implanted cannula via a syringe. The cannula can then moderate insulin release in a glucose-responsive fashion. (B) Schematic of timeline for transcutaneous, externally refillable cannula trial in Streptozotocin (STZ)-induced diabetic mice injected with STZ (140 mg/kg) on day 0. (C) (Left) Blood glucose levels (BGLs) of diabetic mice implanted with glucose-responsive cannulas (FPBA) loaded with 0.375 mg (50 μL of a 7.5 mg/mL solution) of insulin (n=10) or injected subcutaneously (n=3) with insulin (0.05 mg) at times 0, 24, and 48 hours. (Middle) Individual BGL curves for mice with the transcutaneous cannula; black arrows indicate the time of a fresh infusion of insulin (0.375 mg); red arrow indicates when insulin was removed from the cannula. (Right) Individual BGL curves for subcutaneous injections of insulin; black arrows indicate when a fresh subcutaneous injection of insulin (0.05 mg) was administered at times 0, 24, and 48 hours. Grey area represents the normoglycemic range as defined by BGLs within 100 mg/dL < x < 200 mg/dL.

To evaluate this new version of the cannula, the transcutaneous button-connected cannulas were inserted under the skin of STZ-induced diabetic mice (Figure 5b, S12b-c). Approximately 10U (~50μL (7.5 mg/mL)) of fresh insulin were used to refill the cannula twice daily. The insulin solution remaining in the device was removed prior to administration of the fresh insulin dose. The BGLs of the mice started decreasing 30 minutes after the cannula insertion, with the stable establishment of normoglycemia (e.g., BGLs within the 100–200 mg/dL range) occurring after approximately one hour (Figure 5c, left). From a safety perspective, it is critical that membrane stably support any refilling action, as any failure could lead to large doses of insulin dumping that instead induce hypoglycemia. While there were noticeable drops in BGLs immediately after refilling of the devices with fresh insulin, the refilling action did not induce hypoglycemia during any of the five refilling actions (Figure 5c, middle). We speculate that had there been insulin leaking from device damage, failure to control BGL levels of the mice would have been observed, and the mice would have experienced severe hypoglycemia. On the other hand, several mice did experience temporary hyperglycemic peaks, but eventually returned to normoglycemic levels after the refilling action. This refillable insulin infusion set promoted normoglycemia in the mice for three days until hyperglycemia was re-established following the removal of insulin from the device at 72h. This functionality study was terminated after 72h, given the current recommended lifetime of most conventional insulin delivery cannulas is three days, as recommended by the American Diabetic Association (60, 61). Conversely, direct subcutaneous injections of insulin only supported transient glucose control, with BGLs of the mice returning to mildly hyperglycemic levels 4h post injection before returning to fully diabetic states at 12h (Figure 5c, right).

DISCUSSION

Tremendous advancements have been made in tailoring the production of insulin analogues that mimic physiologic insulin secretions and engineering devices to administer these insulin analogues. Given that adherence to insulin therapy is a significant proponent to good glycemic control, the development of user-centric T1D therapies that improve quality of life and glycemic control has remained a central focus of next-generation diabetes therapies. This is because even though insulin continues to be the most powerful therapeutic option available for patients to manage hyperglycemia, several challenges are still faced by T1D patients, including complexity of treatment regiments, injection discomfort, prohibitive financial constraints, significant interference with daily life, and diabetes-related electronic fatigue (93–95). The growth of diabetes technology has accelerated the introduction of electromechanical AP systems that provide a new means of glycemic management for patients failing to achieve glycemic control with more classic therapies such as syringes for insulin injection or insulin pens. The potential advantages of electromechanical AP-based therapies include lower A1C levels, reduced total daily insulin dose, limited risks of hypoglycemia, lower BGL variability, elimination of daily insulin injections, and increased flexibility in meal timing and size (96–98). However, in addition to being more financially prohibitive, there are limitations still facing electromechanical AP systems that have limited their widespread adoption including issues associated with cannula blockages/kinking preventing insulin infusion, problems with infusion set handling, unreliability of Bluetooth-connectivity, user friendliness limitations, and the requirement of a high level of motivation and commitment to vigilant diabetes self-management (95, 99). These limitations can make the process of diabetes control quite challenging, stressful, and open to user errors that lead to non-optimal insulin dosing.

The glucose-responsive cannula described herein provides some reprieve from these burdens through two distinguishing features: (1) a singular glucose-sensing and insulin infusion cannula, and (2) an electronics-free method of glucose sensing and subsequent insulin dosing. To engineer the cannula, we first developed a novel one-pot solvent exchange protocol to combine elastomers and hydrogels at a molecular level. The final hybrid interpenetrating network material —capable of being formed into a cannula— showed robust mechanical strength and tunable permeability that can be functionalized with glucose-responsive properties. The material supported robust insulin delivery under physiologically relevant glucose environments in vitro. Furthermore, when tested in T1D murine models with STZ induced diabetes both as a fully implantable ends-sealed device or as an externally refillable, transcutaneous device, our technology consistently maintained glucose levels in mice within the normal range for ~72 hours, the current recommended lifetime of insulin cannulas.

We anticipate that an electronics-free method of glucose control can alleviate the hardware/software-related limitations of current insulin pumps on the market, although further design optimization is required to increase the translational potential of this technology. Firstly, the basal-release rate of insulin is impacted by the concentration of insulin enclosed within the cannula. This becomes critical given that the hydrogel-elastomer material is permeable to insulin even in the absence of glucose. More fine tuning of material feed ratios and insulin concentrations would be required to ensure that basal-release rates —using commercially available insulin(s)— are tailored to match basal insulin release rates required by patients. An alternative improvement route to explore would be to take advantage of electrostatic interactions within the material matrix by incorporating a positively charged monomer in the hydrogel network, as others have investigated (100, 101). Under low glucose conditions, a positively charged monomer could form a complex with negatively charged insulin, thereby minimizing insulin release under hypoglycemic conditions. In the presence of glucose, the binding of phenylboronic acid to glucose induces a negative charge to the network, which reduces the electrostatic attraction between the polymer and insulin from the decrease in positive charge density of the matrix. This weakened binding between the material and insulin can then trigger glucose-responsive insulin release. Preliminary results substantiated that the incorporation of a hydrophilic cationic monomer within a pure elastomer sample decreased insulin release with increasing monomer concentration (Figure S13), although more work would have to be undertaken to fine tune said response and confirm increased retention of insulin under low glucose conditions.

Another key consideration is the inflammatory response generated by the cannula. When a continuous insulin infusion cannula is subcutaneously inserted, the cannula causes an acute inflammatory response of the surrounding tissue, the response of which is influenced by catheter material, shape, and wear time (102, 103). All these factors may influence insulin absorption pharmacokinetics, as the layer of inflammatory tissue may act as a physical barrier to insulin flow to adjacent vascular tissue (103, 104). Given that the one-pot solvent exchange fabrication procedure of the cannulas is amenable to other materials, it might be interesting to explore improving the biocompatibility of the cannula directly through the incorporation of anti-inflammatory or immune-protective components, such as zwitterionic monomers for example (105). These monomers might reduce the inflammatory response against the cannula to minimize problems that would otherwise hinder a potential extended wear-time of the cannula past the currently recommended three-day use.

In considering the translational potential of this technology to large animal models and beyond, currently, the majority of T1D patients practice basal-bolus dosing, that is, administering a long-lasting insulin analogue at regular intervals (the basal dose), and a fast-acting insulin before meals (the bolus dose), adjusted according to meal composition (7, 8). Electromechanical AP systems can support this regimen, with most commercially available insulin pump technologies having separate infusion settings for basal and bolus insulin dosing (24, 26). Despite fully synthetic AP systems —like our cannula— demonstrating encouraging BGL regulation abilities, with several microneedle patch designs even showing promise in normalizing BGLs in large animal models (106, 107), bolus dosing is specifically not currently available with these therapies. Instead, fully synthetic systems rely on rapid changes in permeability to dose larger quantities of insulin, which may not support the physiological needs of multiple bolus doses required in a day by patients (17). Further investigation and design optimization to incorporate both basal and bolus insulin delivery within fully synthetic AP therapies might broaden the translational potential of these therapies.

In summary, we describe here a design of a novel elastomer-hydrogel hybrid cannula supporting electronics-free, glucose-responsive insulin delivery. The cannula provides a large surface area for continuous glucose-sensing and allowed for facile loading of large volumes of insulin. The mechanical strength of the cannula is such that it likewise does not need to be reinforced with other supporting structures (e.g., silicone tubing or reservoirs) for implantation and retrieval as used in other hydrogel-based, glucose-responsive membrane therapies (64, 108–110).When evaluated in vivo, the cannula normalized BGLs in diabetic mice effectively on average for three days without inducing hypoglycemia and controlled postprandial glucose levels. The minimalist style of the design is unique compared to other insulin infusion sets, glucose-responsive technologies, or electronic-based “closed-loop” therapies developed to date. This proof-of-concept study may aid in the development of other glucose-responsive cannula designs for simplistic and user-centric insulin delivery therapies.