Bimetallic Nanocatalysts Immobilized in Nanoporous Hydrogels for Long-Term Robust Continuous Glucose Monitoring of Smart Contact Lens

Abstract

Smart contact lenses for continuous glucose monitoring (CGM) have great potential for huge clinical impact. To date, their development has been limited by challenges in accurate detection of glucose without hysteresis for tear glucose monitoring to track the blood glucose levels. Here, long-term robust CGM in diabetic rabbits is demonstrated by using bimetallic nanocatalysts immobilized in nanoporous hydrogels in smart contact lenses. After redox reaction of glucose oxidase, the nanocatalysts facilitate rapid decomposition of hydrogen peroxide and nanoparticle-mediated charge transfer with drastically improved diffusion via rapid swelling of nanoporous hydrogels. The ocular glucose sensors result in high sensitivity, fast response time, low detection limit, low hysteresis, and rapid sensor warming-up time. In diabetic rabbits, smart contact lens can detect tear glucose levels consistent with blood glucose levels measured by a glucometer and a CGM device, reflecting rapid concentration changes without hysteresis. The CGM in a human demonstrates the feasibility of smart contact lenses for further clinical applications.

1. Introduction

Continuous glucose monitoring (CGM) can be used for real-time detection of glucose levels in the body fluid such as tears, sweat, and interstitial fluid (ISF) with greatly improved patient compliance.^[1–4] CGM devices can detect hypoglycemia, as well as, hyperglycemia, enabling the rapid treatment of serious side effects such as tremor, trouble talking, confusion, loss of consciousness, epileptic seizure, and death.^[5,6] However, minimally invasive CGM cutaneous patch devices suffer from limited accuracy even with sophisticated data filters and algorithms.^[7] In addition, prior studies on glucose-sensing contact lenses revealed the inconsistent correlation between tear and blood glucose concentrations without supporting the full requirements of a medical device.^[1–4] The contact lens-based approaches have several critical issues including low stability of glucose oxidase, low and inconsistent sensitivity of glucose sensors, small volume of tears (≈5–7 μL) on the ocular surface,^[8] and slow diffusion rate of tears through hydrogel substrates. The accurate detection of glucose without hysteresis is one of the most important clinically unmet needs for glucose monitoring in the body fluid to track the blood glucose levels.

The detection of glucose using minimally invasive CGM devices are typically performed using electrochemical sensors consisting of hydrogels with immobilized glucose oxidases (GOx).^[9] To improve glucose sensor sensitivity, various nanomaterials have been harnessed such as carbon nanotubes,^[10] gold doped graphene and gold porous structures.^[11,12] Commercial CGM devices estimate blood glucose contents from current changes resulting from the electrochemical reaction between oxygen (O₂), glucose in the ISF and GOx in the polymer matrix.^[9] In this process, the electrical current change with increasing glucose concentration correlates with the diffusion rate of glucose and O₂ into the hydrogel and the decomposition rate of hydrogen peroxide (H₂O₂). The highly dense polymer matrices in the commercial CGM devices result in long warming-up times (≈2 h),^[13] and very slow diffusion and swelling rates. In addition, when being implanted subcutaneously, CGMs are subject to biofouling on their electrochemical sensor surfaces.^[14] GOx and nanocatalysts can also be released from the polymer matrix over time,^[9] which cause changes in sensitivity, diffusion rate, and warming-up times.

Tears are generated from lacrimal glands at a constant flow rate of 0.15 μL min^–1 without external stimulation^[8] and drained to tear duct without additional circulation system unlike other body fluids such as sweat and saliva. Accordingly, tears have been investigated as an important medium for non-invasive diabetic diagnosis in the field of wearable devices.^[1–4] The most important critical issue of contact lens devices for glucose monitoring is insufficient c orrelation between blood and tear glucose concentrations. Google Lens, which was developed by the collaboration of Google and Novartis, suffered from the insufficient consistency in the correlation between tear glucose and blood glucose concentrations.^[1] Among various possible reasons for the failure of Google Lens, the low stability of glucose oxidase, the low and inconsistent sensitivity of glucose sensor, the slow diffusion rate of tears, and irritation in the eyes might cause the inaccuracy between tear and blood glucose concentrations.

Here, we present smart contact lenses containing bimetallic nanocatalysts immobilized in nanoporous hydrogels for longterm and robust CGM (Figure 1). After redox reaction of GOx, hydrogen peroxide is rapidly decomposed on the surface of goldplatinum bimetallic nanocatalysts modified with hyaluronate (HA-Au@Pt BiNCs), which generates two electrons for amperometric responses.^[15] In addition, HA-Au@Pt BiNCs facilitate nanoparticle-mediated charge transfer in nanoporous hydrogels, resulting in high sensitivity,^[16] fast response time, and low detection limit. The nanopores in the hydrogels enable the reversible and continuous monitoring of glucose concentration with a fast diffusion rate of reaction species and a rapid swelling rate.^[17,18] Herein, we present our assessments of the sensitivity, reaction reversibility without hysteresis, and correlation between tear and blood glucose concentrations of our glucose-sensing contact lenses. We demonstrate that the smart contact lenses can accurately monitor increasing and decreasing blood glucose levels with 94.9% acceptable data in diabetic (n = 3) and normal (n = 3) rabbits. Finally, we evaluate the clinical feasibility and safety of the smart contact lenses on the eyes of a human patient.

Figure 1.

Schematic illustration of smart contact lens for diabetes monitoring. The structure and glucose sensing mechanism of bimetallic nanocatalysts (BiNCs) in nanoporous hydrogels. Flavin adenine dinucleotide (FAD) inside glucose oxidase in the hydrogels undergoes a redox reaction with diffused glucose and O₂, which is then reduced to FADH₂. H₂O₂ are rapidly decomposed on the surface of HA-Au@Pt BiNCs and generate two electrons which are rapidly transported to the electrode surface by nanoparticle-mediated charge transfer. The nanopores in the hydrogels play important roles for fast diffusion of reaction species and rapid swelling with superabsorbent properties.

2. Results and Discussion

2.1. Synthesis and Characterization of HA-Au@Pt BiNCs

The successful synthesis of end-group thiolated HA (HA-SH) was confirmed by proton nuclear magnetic resonance (¹H-NMR) (Figure S1, Supporting Information). Transmission electron microscopic (TEM) images in Figure 2a,b show the effect of HA-SH on the dispersion of Au@Pt BiNCs. While Au@Pt BiNCs were significantly aggregated (Figure 2a, Figure S2A, Supporting Information), HA-Au@Pt BiNCs were found to be uniformly dispersed on the photographs and TEM images (Figure 2b, Figure S2B, Supporting Information). The stable dispersion of HA-Au@Pt BiNCs might be due to the strong negative surface charge and the steric hindrance of HA. We analyzed the hydrodynamic size, polydispersity index, and zeta potential of HA-Au@Pt BiNCs to confirm the correlation between the surface charge and the dispersion (Figure 2c,d). As the surface charge of nanocatalysts decreased, the dispersion became more stable in accordance with the DLVO theory.^[19]

Figure 2.

Characteristics of HA-Au@Pt BiNCs. Photographs and TEM images of a) Au@Pt and b) HA-Au@Pt BiNCs in DI water (4.5 mM, scale bar: 200 nm). DLS analysis for c) the hydrodynamic particle size and d) the zeta potential of AuNPs, Au@Pt, and HA-Au@Pt BiNCs in DI water (n = 3). e) XPS spectra of the Au 4 f in AuNPs, Au@Pt, and HA-Au@Pt BiNCs. f) XPS spectra of the Pt 4 f in PtNPs and HA-Au@Pt BiNCs. g) Valence band XPS spectra of AuNPs, PtNPs, and HA-Au@Pt BiNCs. h) Structural analysis of HA-Au@Pt BiNCs showing the lattice structure of fcc Au[111], Au[220], and Pt[111] plane by HR-TEM (scale bar: 2 nm). i) HR-TEM image of HA-Au@Pt BiNCs and the corresponding EDS mapping images: Au in red, Pt in green, and Au, Pt merged (scale bar: 20 nm).

The catalytic activity of HA-Au@Pt BiNCs was analyzed on the basis of the d-band center (E_d) model of Hammer and Nørskov,^[20] which explains the difference in the molecular orbital hybridization of adsorbed molecules according to the d-band center of transition metals. The down-shifting of the d-band center of Pt from the Fermi level (E_F) is important to improve the catalytic efficiency.^[21–24] We verified that the d-band center of Pt was downshifted in the X-ray photoelectron spectroscopy spectra of HA-Au@Pt BiNCs. The peaks of HA-Au@Pt BiNCs in Au⁰4f_7/2 and Au⁰4f_5/2 were blue-shifted compared to AuNPs and the peaks in Pt⁰4f_7/2 and Pt⁰4f_5/2 were red-shifted compared with PtNPs (Figure 2e,f). We analyzed the shift of binding energy in Au and Pt with the charge compensation model, considering the compensation of 6s orbital conduction electron gain and 5d orbital charge depletion.^[25,26] The d-band center of HA-Au@Pt BiNCs was calculated from the valence band. In Figure 2g, we confirmed that the d-band center of HA-Au@Pt BiNCs (4.25 eV) located between the d-band centers of PtNPs (3.66 eV) and AuNPs (5.23 eV).

The morphology of HA-Au@Pt BiNCs was analyzed by transmission electron microscope (TEM) and energy dispersive spectroscopy (EDS) mapping. HA-SH was well-coated on the HA-Au@Pt BiNCs with a thickness of ≈2.5 nm and PtNPs with a diameter of ≈3.2 nm were well dispersed on the surface of AuNPs (Figure 2h). PtNPs preferentially grew at the location with a high surface energy such as the side or grain boundary of AuNPs^[22] and didn’t grow outside AuNPs as shown in the EDS mapping (Figure 2i). We analyzed the synthesis efficiency of HA-Au@Pt BiNCs according to the concentration of H₂PtCl₆ (Table S1 and Figure S3, Supporting Information). From the results, we confirmed that HA-Au@Pt BiNCs at 4.5 mm were stably dispersed with the highest loading efficiency of Pt on AuNPs.

2.2. Preparation and Characterization of Nanoporous Hydrogels Containing BiNCs

To develop a fully reversible CGM system, we synthesized nanoporous hydrogels with fast swelling and superabsorbing characteristics by forming nanopores on the hydrogels containing BiNCs. The time required for the hydrogels to reach equilibrium is determined by the porosity of the hydrogel. The higher the porosity, the higher the diffusion rate between the reactant and the electrode. The rate of water absorption and swelling rate of the polymer matrix are highly depending on the porosity and the hydrophilic functional groups. In this context, we prepared a polymer matrix composed of hydrophilic polymers such as hyaluronic acid, chitosan, and poly(vinyl alcohol) (PVA). In addition, to introduce the well-distributed pore structures, we used and optimized the amount of polystyrene (PS) beads in the hydrogels. The minimum volume of PS beads for uniformly forming nanopores in the hydrogels was analyzed to be 6 μL (Figure S4, Supporting Information). HA-Au@Pt BiNCs were chemically crosslinked by the reaction between aldehyde group (−CHO) of glutaraldehyde and hydroxyl group (−OH) of HA.^[27] Remarkably, the enzyme layer with HA-Au@ Pt BiNCs showed the acetal bond peak in the 1085–1150 cm^–1 region than the enzyme layer without the nanocatalysts (Figure S5, Supporting Information).^[28] Most of the pores in the hydrogels with HA-Au@Pt BiNCs were blocked due to chemical bonding and physical interaction between the hydrogels and HA-Au@Pt BiNCs (Figure 3a upper images). When uniform nanopores were formed in the hydrogels containing HA-Au@Pt BiNCs, the swelling and diffusion rate were accelerated due to the enhancement of the capillary force with a greatly improved porosity (Figure 3a lower images).^[17,18] To understand the morphology of four different hydrogels with and without nanocatalysts and nanopores, we analyzed the porosity, pore density, average pore area, and pore to pore distance using the ImageJ (Table S2, Supporting Information).

Figure 3.

Characteristics of nanoporous hydrogels containing BiNCs. a) scanning electron microscope (SEM) images of the hydrogels containing HA-Au@Pt BiNCs without and with nanopores: Surface, surface zoom, and cross-section images (scale bar, black: 2 μm, white: 10 μm). b) In vitro release of GOx from 4 different hydrogels in PBS (25 °C, pH 7.4) for 7 days (n = 3). Gray line: w/o nanopores and w/o HA-Au@Pt BiNCs, green line: w/nanopores and w/o HA-Au@Pt BiNCs, blue line: w/o nanopores and w/ HA-Au@Pt BiNCs and pink line: w/ nanopores and w/ HA-Au@Pt BiNCs. c) In vitro release of Au@Pt and HA-Au@Pt BiNCs from the hydrogels in PBS (25 °C, pH 7.4) for 7 days (n = 3): Full color bars before removing unreacted nanocatalysts and check pattern bars after removing unreacted nanocatalysts. d) HR-TEM images of Au@Pt and HA-Au@Pt BiNCs in the nanoporous hydrogels and the corresponding EDS mapping images: Au in red and Pt in green (scale bar: 50 nm).

We quantitatively analyzed the amount of GOx and nanocatalysts released from the nanoporous hydrogels containing BiNCs. In vitro release of samples was monitored for 7 days because commercial CGM devices are generally used for 7 days.^[13] The released amount of GOx was maximal in 24 h without immobilization during the hydrogel crosslinking reaction. In the nanoporous hydrogels containing HA-Au@Pt BiNCs, the released amount of GOx was only 8.9% compared to nanoporous hydrogels without HA-Au@Pt BiNCs (18.0%) due to the interaction between HA-Au@Pt BiNCs and GOx in the nanoporous hydrogels (Figure 3b). The released amount of BiNCs increased in the presence of nanopores. As shown in Figure 3c, the HA-Au@Pt BiNCs (17.2%) were released significantly less than Au@Pt BiNCs (57.8%) from the nanoporous hydrogels because of chemical bonding and physical interaction between HA-Au@Pt BiNCs and components in the nanoporous hydrogels.^[29] After removing uncrosslinked BiNCs in the hydrogels in phosphate-buffered saline (PBS, pH 7.4) for 24 h, in vitro release of BiNCs from the hydrogels was monitored in PBS for 7 days. We verified that HA-Au@Pt BiNCs in nanoporous hydrogels were rarely released (2.03%) and stably immobilized, compared to Au@Pt BiNCs in the nanoporous hydrogels with higher release of 8.28%. According to TEM, Au@Pt BiNCs were heavily aggregated in the hydrogels, but HA-Au@Pt BiNCs were uniformly dispersed throughout the hydrogels (Figure 3d). We also confirmed that the nano particles on the TEM image were HA-Au@Pt BiNCs by the EDS mapping for Au and Pt elements.

2.3. Electrochemical Properties of Nanoporous Hydrogels with BiNCs

We analyzed electrochemical properties of nanoporous hydrogels containing BiNCs as a function of bimetallic nanocatalysts, nanopores, and solvents to dissolve PS beads (Table S3, Supporting Information). The volcano plot in Figure S6A,B, Supporting Information, showed the correlation between the sensitivity and the d-band center of HA-Au@Pt BiNCs. We found that the adsorption strength should not be so strong or so weak to achieve the highest activity of BiNCs.^[21–24] We measured the electrochemical properties according to the volume of PS beads solution in the hydrogels immobilized with 4.5 mm HA-Au@Pt BiNCs. A significant increase of sensitivity was confirmed in 6 μL PS beads which could make uniform nanopores (Figure S6C, Supporting Information). Acetone appeared to be better than ethanol which dissolved the PVA and chitosan as well as the PS beads, resulting in the decreased sensitivity (Figure S6D, Supporting Information).

We next assessed the effect of nanopores and HA-Au@Pt BiNCs on the sensitivity, response time, and detection limit (Figure 4a). In the presence of nanopores, the sensitivity increased from 27.8 to 65.3 μA∙cm^–2∙mmol^–1, the response time decreased from 6.8 to 3.5 s and the detection limit was reduced from 1 to 0.1 mg∙dl^–1. These results might be attributed to the improved diffusion rate of reaction species in the presence of nanopores. The continuous phase of the hydrogels might locally inhibit the diffusion of glucose, O₂, and H₂O₂.^[14] The high porosity and homogeneously dispersed nanopores facilitated the full diffusion of reaction species into the hydrogels. In addition, the formed nanopores seemed to augment the role of individual nanoelectrodes, resulting in the improved sensitivity.^[16,30] In the presence of HA-Au@Pt BiNCs, the sensitivity increased from 27.8 to 82.08 μA∙cm^–2∙mmol^–1, the response time increased from 6.8 to 9.5 s and the detection limit was reduced from 1 to 0.1 mg dl^–1. The improved sensitivity and detection limit might be caused by the high conversion efficiency of H₂O₂ into O₂ and 2 electrons in the presence of HA-Au@Pt BiNCs. HA-Au@Pt BiNCs improved the function of GOx by actively removing H₂O₂, resulting in the improved sensitivity and detection limit. However, HA-Au@Pt BiNCs blocked the pores of the hydrogels and showed a slow response time due to the decreased diffusion rate of reaction species. In the presence of both nanopores and HA-Au@Pt BiNCs, the sensitivity was 180.18 μA∙cm^–2∙mmol^–1, the response time was 3.6 s and the detection limit was 0.01 mg∙dl^–1. The best sensitivity, fast response time and low detection limit might be the result of a synergistic effect between the fast rate of diffusion by nanopores and the improved decomposition efficiency of H₂O₂ in the presence of HA-Au@Pt BiNCs.

Figure 4.

In vitro continuous glucose monitoring of smart contact lens. Grey line for w/o nanopores and w/o HA-Au@Pt BiNCs, green line for w/ nanopores and w/o HA-Au@Pt BiNCs, blue line for w/o nanopores and w/o HA-Au@Pt BiNCs, and pink line for w/o nanopores and w/o HA-Au@ Pt BiNCs. a) Real-time CGM in PBS (25 °C, pH 7.4). b) Hysteresis curve with increasing and decreasing glucose concentrations in the range from 5 to 50 mg dl^–1 (n = 50). c) Statistical correlation analysis between the glucose concentration and the measured current (n = 50). d) Cyclic voltamograms between −0.25 and 0.2 V at a scan rate of 25 mV s^–1 in 10 mm K₃Fe(CN)₆. e) Impedance spectra of the hydrogels with the nyquist plot and analysis for the charge transfer resistance (R_ct) between 100 and 100 000 Hz at 10 mV in 10 mm K₃Fe(CN)₆. f) Effect of the immersion time (0, 2, 4 h) of the hydrogels with nanopores and HA-Au@Pt BiNCs into PBS on the current change with increasing glucose concentration for the warming-up time analysis (n = 50).

To evaluate the accuracy of CGM, we analyzed the hysteresis curve with increasing and decreasing glucose concentrations between 5 and 50 mg∙dl^–1 (Figure 4b). In the presence of nanopores, hysteresis decreased from 7.2–17.6% (gray line) to 1.4–3.2% (green line). In contrast, in the presence of HA-Au@Pt BiNCs, the hysteresis notably increased from 7.2–17.6% to 9.7–50.2% (blue line). In the presence of both nanopores and HA-Au@Pt BiNCs, the hysteresis was 1.2–4.5% (pink line). Remarkably, the higher the porosity, the lower the hysteresis. The sample with the highest porosity showed the lowest degree of hysteresis (green line). As shown in Figure 4c, the hydrogels containing both nanopores and HA-Au@Pt BiNCs showed the highest linearity (r² = 0.9982) by synergistic effect. We investigated the effect of nanopores and HA-Au@Pt BiNCs on the cyclic voltammetry (Figure 4d). At 25 mV∙s^–1, hydrogels containing both nanopores and BiNCs showed a quasi-steady state by hemispherical diffusion via nanopores^[30] and the nano particle mediated charge transfer via HA-Au@Pt BiNCs.^[16] At 25–200 mV∙s^–1, the sigmoidal voltammogram response was observed by the shielding effect (Figure S6E, Supporting Information).^[16,30] The decrease of charge transfer resistance (R_ct) by nanopores and HA-Au@Pt BiNCs was confirmed by electrochemical impedance spectroscopy. The R_ct decreased significantly due to the synergistic effect between nanoparticle-mediated charge transfer and nanoelectrode (Figure 4e). The decrease of R_ct by nanopores was remarkable in the case of 6 μL PS beads (18.6 Ω) (Figure S6F, Supporting Information). We also evaluated the warming-up time of the hydrogels in the presence of nanopores and HA-Au@Pt BiNCs. The hydrogels without nanopores showed a long warming-up time with a deviation of 19.6% (Figure S7A, Supporting Information) and 28.8% (Figure S7B, Supporting Information) between 2 and 4 h in PBS (pH 7.4). In contrast, the hydrogels with nanopores showed a short warming-up time under 2 h with a small deviation of 1.98% (Figure S7C, Supporting Information) and 4.28% (Figure 4f) between 0 and 2 h.

2.4. In Vivo Continuous Glucose Monitoring and Safety Evaluation of Smart Contact Lens

The smart contact lens was fabricated by using a silicone elastomer with high oxygen permeability and surface-modified with hygroscopic grafted HA-NH₂ (Figure S8, Supporting Information). In vitro, we confirmed that the smart contact lens reacted with few tear components other than glucose and remained stable for up to 21 days (Figure S9, Video S1, Supporting Information). In vivo CGM was conducted with diabetic (n = 3) and normal (n = 3) rabbits at the intervals of 5 days for 3 weeks. We performed CGM by wearing a smart contact lens on rabbit’s eyes and a commercial CGM on the abdomen (Figure 5a). Figure 5 and Video S2, Supporting Information, show the representative data. To analyze the correlation between blood glucose levels and tear glucose levels in response to the blood glucose concentration changes, 10 mL of 20% glucose solution or 1 unit of Humulin R was subcutaneously injected into normal and diabetic rabbits. The smart contact lens was used to measure tear glucose levels after a lag time of 10 min including a warming-up time. For the warmingup time, the glucose in tears slowly spread to the surface of the glucose sensor and the glucose concentration increased in response. In Figure 5b,c, the tear glucose level showed a lag time of ≈10 min from the blood glucose level in accordance with previous reports.^[31] For all conditions of the constant, increasing, or decreasing blood glucose levels, the trends of tear glucose levels with the lag time of 10 min were similar to the trends of blood glucose levels in diabetic and normal rabbits (Figure 5b,c, pink dashed line). The correlation between the glucometer and the smart contact lens was analyzed with a Pearson correlation coefficient in consideration of the lag time of 10 min.^[31,32] In both diabetic and normal rabbits, there was a strong correlation (ρ > 0.70) in all cases (Figure S10A,B, Supporting Information). Especially, it showed a very strong correlation (ρ > 0.90) when blood glucose levels were increased or decreased. The high correlation between blood and tear glucose levels might be attributed to the enhanced diffusion rate, reaction rate, and sensitivity by HA-Au@Pt BiNCs immobilized in nanoporous hydrogels for the electrochemical glucose sensor of the smart contact lens.

Figure 5.

In vivo continuous glucose monitoring of smart contact lens in diabetic rabbits. a) Photographs of the continuous glucose monitoring system in diabetic rabbits. i) A rabbit wearing the smart contact lens, close-up images of ii) our smart contact lens on rabbit’s eye (scale bar: 150 μm) and iii) a commercial CGM attached and inserted under the subcutaneous abdomen. Continuous glucose monitoring in blood and tears for 45 min in case of b) diabetic rabbits and c) normal rabbits for i) the stable change, ii) the increase, and iii) the decrease of blood glucose levels. d) Clark error grid analysis for clinical validation, pink circle: Blood-tears correlation. e) Continuous glucose monitoring with a smartphone application for the correlation equation between blood and tear glucose levels. i) Photograph of the measurement system and ii) continuous glucose monitoring with a glucomter, a commercial CGM and our smart contact lens for 30 min.

All the measured data were gathered to analyze the Pearson correlation coefficient and Clarke error grid.^[14,29] These methods are crucial tools to evaluate the clinical accuracy of CGM. We confirmed that the correlation value of the smart contact lens was very high (ρ = 0.82) (Figure 5d). In the Clarke error grid, 94.9% of data measured by the smart contact lens were located in A and B regions where data are clinically acceptable, and some data were located in D and C regions with underestimated values (Figure 5d).^[14,31] The decreased movement of muscles around the eyes due to anesthesia might result in the reduced tear circulation and inaccuracy.^[31] Finally, we evaluated the clinical feasibility of the smart contact lens compared with a glucometer and a commercial CGM (Figure 5e). The measured tear glucose level with the smart contact lens was converted to the blood glucose level by using the correlation equation between blood and tear glucose levels from Figure 5d. The correlation between blood glucose measured by a glucometer and tear glucose measured by the smart contact lens (ρ = 0.95) showed a high correlation similar to that between a glucometer and a commercial CGM (ρ = 0.96) (Figure S10C, Supporting Information).

Dry eye syndrome and physical irritation caused by wearing contact lenses are reported to increase the glucose concentration in the tear and cause the overestimation of the blood glucose levels.^[9] We fabricated smart contact lens with silicone elastomers with a high oxygen permeability and a low swelling ratio to prevent smart contact lens from deformation. Then, hygroscopic HA was coated on the contact lens surface to improve the surface water content. Regardless of the type of contact lens and the number of wearing times, the slit lamp test and histological analysis confirmed that there was no wound or inflammation in the cornea (Figure S11, Supporting Information). In addition, the volume of tears in both eyes showed no significant difference with no significant corneal thickness changes for all conditions (Figure S12, Supporting Information). The cornea of diabetic rabbits was also analyzed by TEM and EDS mapping to check whether HA-Au@Pt BiNCs were released from the smart contact lens or not. There was no HA-Au@Pt BiNCs on the TEM images of all samples, and Au and Pt were not measured in the EDS analysis (Figure S13, Supporting Information). During in vivo tests, the heat generated by the antenna for operating the smart contact lens was analyzed with an infrared camera. There was no temperature increase compared to the initial temperature for 30 min (Figure S14, Supporting Information).

2.5. Preliminary Feasibility Study of Smart Contact Lens in a Human

Figure 6a shows wireless power transmission and data communication via inductive coupling between the smart contact lens and the external reader in a 41-year-old human subject. Smart contact lens was evaluated in strict compliance with the protocol approved by the bioethics committee of Pohang University of Science and Technology (POSTECH, PIRB-2021-A001). The digitized glucose sensor outputs were then transmitted wirelessly through the same coil antenna via load modulation and monitored on a smartphone (Video S3, Supporting Information). The temperature change on the ocular surface was less than 0.7 °C after 10 min continuous operation of the system in human eyes (Figure 6b). The specific absorption rate (SAR) simulation using a high-frequency structure simulator was also conducted to estimate the effect of EM energy on the human eyes. The peak SAR value was obtained according to the distance between a homogeneous human head model and an antenna at the external reader under the wireless power of 14 dBm (Figure 6c). The peak SAR was 0.64 W∙kg^–1 at the distance of 1 mm (Figure 6d). It was much less than 1.6 W kg^–1 in compliance with IEEE standards.^[33] From the results, we could confirm that our smart contact lens system would be harnessed for further clinical applications without causing safety issues by heating and EM radiation.

Figure 6.

Continuous glucose monitoring of smart contact lens in a human. a) Photographs for a man wearing the smart contact lens and the wireless CGM using a smartphone (scale bar: 1 cm). b) The heat profiles during in vivo operation for 0 and 10 min, respectively. c) Simulated peak specific absorption rate (SAR) as a function of distance between the human head and the external reader antenna. d) Simulation for the generated heat in the model human head.

In Table S4, Supporting Information, we compared the advantages and disadvantages of measuring glucose levels in the blood, ISF, and tears with a glucometer, a CGM patch, and the smart contact lens. The measurement of blood glucose levels with a glucometer is accurate, but it is very painful with blood sampling and cannot be used for CGM. Although a CGM patch can be used for the continuous monitoring of glucose in the ISFs, it has the disadvantages of infection risk and biofouling on surface of the sensor. When the commercial CGMs are inserted into the interstitial tissue space, proteins, cells, or other biomacromolecules are adsorbed to the glucose sensor surface by non-specific binding, which results in the inhibition of glucose diffusion to the surface of glucose sensors and the decreasing sensitivity over time with limited sensing applicability for only two weeks. This is the well-known wound healing process for the implanted materials. The main limitation for the measurement of tear glucose was the low glucose concentration in the small volume of tears, resulting in the low sensitivity and accuracy. To overcome this limitation, we developed our smart contact lens containing HA-Au@Pt BiNCs immobilized in the nanoporous hydrogels for the glucose sensor, which dramatically improved the sensitivity and accuracy. The smart contact lens showed a similar correlation value with that of a commercial CGM and reflected the changing blood glucose levels very sensitively in diabetic and normal rabbits. Finally, smart contact lens could measure the tear glucose levels in a human with data transmission to a smartphone, confirming the feasibility for further clinical applications. Our smart contact lens would have a great potential to be an excellent alternative to a glucometer and a commercial CGM.

3. Conclusion

We have successfully demonstrated that our smart contact lenses containing HA-Au@Pt BiNCs in nanoporous hydrogels could accurately measure rapidly changing blood glucose levels. HA-Au@Pt BiNCs were uniformly dispersed and stably immobilized in the nanoporous hydrogels with maximized efficiency of the BiNCs by controlling the d-band center of Pt. The nanoparticle-mediated charge transfers by the HA-Au@Pt BiNCs in the nanoporous hydrogels resulted in the high sensitivity, low detection limit, low hysteresis, and fast warming-up time. In the diabetic rabbits, the smart contact lens showed a similar correlation value with that of a commercial CGM device requiring calibration and data processing algorithms, and reflected rapidly changing blood glucose levels very sensitively. Our smart contact lens could measure the tear glucose levels in a human with data transmission to a smartphone. We confirmed the initial safety parameters of these smart contact lenses in terms of heat and electromagnetic radiation. Taken together, our smart contact lenses containing HA-Au@Pt BiNCs in the nanoporous hydrogels have successfully demonstrated the feasibility for further clinical applications to diabetes monitoring.

4. Experimental Section

All experimental methods are described in Supporting Information.

Study Approval:

All in vivo animal experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The animal experiment was approved by the Institutional Animal Care and Use Committee (IACUC, KPC-M2020042) at the Korea Preclinical Center (KPC). Smart contact lens was evaluated in human eyes in strict compliance with the protocol approved by the bioethics committee of Pohang University of Science and Technology (PIRB-2021-A001). Informed written consent of the participant was obtained.

Statistical analysis:

One-sided statistical analysis was carried out using Student’s t-tests and extracted p-values. The difference of *p < 0.05 was considered statistically significant. Statistical data analysis was performed by SigmaPlot (10.0 Version, Systat Software Inc).

Code Availability:

A full code availability statement is included in the manuscript. Custom algorithms are ancillary, but a full code is available in GitHub as a project name “SCL_2021.” Commercial software programs of Xilinx ISE Design Suite (ver.14.7) and Java (ver.1.8.0_131) were used. The LSK_v1.v is a Verilog code used in FPGA board to read data from RF receiver. The guimake4.java is a java code to receive and plot data from the FPGA board on the computer.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.