1. Introduction

Diabetes (hyperglycemia) endangers the health of millions of people all over the world. In 2017, the International Diabetes Federation (IDF) reported that one of two adults with diabetes is undiagnosed, and currently, there are a total of 425 million adults with diabetes. This number is expected to increase to 649 million by 2045. Diabetes has become a worldwide chronic disease problem, and there is no cure for diabetes at this time. Therefore, people with diabetes need to take their medication regularly to maintain blood glucose levels in the acceptable range and prevent diabetes complications. Extreme fluctuation of glucose concentrations in the blood can cause serious health problems such as blindness, kidney failure, stroke, and nerve issues [1,2]. A person unable to produce sufficient insulin to regulate the level of glucose in the blood may need insulin injection daily to determine the correct dosage of insulin, consequently continuous monitoring of blood glucose levels is essential. Therefore, a variety of glucose biosensors has been developed [3, 4]. In the past, most commercially available glucose sensors required needle-sticking to collect blood samples [5–7].

Recently, there have been many minimally invasive sensors developed. Of these, the commercially available ones include: FreeStyle Libre Flash (Abbott Diabetes Care Inc., USA) Gluco Track (GlucoTrack Integrity Applications Ltd., Israel), and Dexcom G5 (Dexcom, Inc., USA) [8]. These devices do not require a skin prick. FreeStyle Libre Flash determines glucose levels in diabetic adults and is advertised as giving a 40% false positive for hypoglycemia detection in a clinical study. Gluco Track uses a combination of ultrasonic, electromagnetic and thermal technology to measure glucose from the earlobe [9]. The multi-modal detection device is expected to be less susceptible to interferents, but still requires an individual calibration, and it is only intended for people with pre-diabetes and type 2 diabetes. Dexcom G5 is the first FDA (U. S. Food and Drug Administration) approved continuous glucose monitoring (CGM) system for non-adjunctive use by patients (two years of age and older) with type 1 and type 2 diabetes [8–10].

Completely non-invasive devices include OMELON B₂ (Omelon, Russia), which is based on how blood sugar affects the state of the blood vessels. It operates by measuring blood pressure, pulse wave, and vascular tone on two hands. However, this device is not suitable for people with type 1 diabetes. Also, it gives errors for people with acute fluctuations in blood pressure, with advanced atherosclerosis, or with sharp fluctuations in blood sugar.

Additional prick-free glucose monitors that are still under development include the smart contact lens from Google [11], which measures the glucose level from tear fluid. GlucoWise (MediWise Ltd., United Kingdom), which uses low power radio waves transmitted through a section of the human body such as the earlobe. SugarBEAT (Nemaura Medical Inc., UK) that measures the glucose level by using a daily disposable adhesive skin-patch connected to a rechargeable transmitter and could be used by diabetic and non-diabetic people. Combo Glucometer (Conga Medical, Ltd., Israel) that is based on four light emitting diodes (LEDs) and is intended to be used by people over 18 years old with type 2 diabetes; however, this device cannot detect hypoglycemia because it cannot read below 70 mg/dL of glucose. And, Dexcom G6 CGM (Dexcom, Inc., USA), which is a new generation of Dexcom products that extended the sensor lifetime to 10 days compared to 7 days of their previous generations.

Nowadays, there is increasing interest toward long-term caring and monitoring for diabetes that enables real-time monitoring capabilities by using fully implantable continuous glucose monitoring (ICGM) systems, such as the Eversense (Senseonics company, USA) and the Eclipse ICGM System (GlySens, USA). Eversense is a small glucose sensor about the size of a pill that stays implanted under the skin for 90 days. It will be the first FDA approved ICGM available in the U. S. (expected in the middle of 2018), and it has been in use in Europe since 2016. GlySens is also developing an ICGM that can be used for one year.

Optical sensors are of special interest for both external and implantable glucose monitoring due to their potential longer standoff range. One class of optical sensors is based on the change in emission intensity and/or fluorescence-lifetime of fluorescent particles as a function of glucose concentrations [12]. These include organic dyes, quantum dots (QDs), and noble metal nanoparticles (NPs) functionalized with boronic acid, which provide sensitive and reversible glucose level monitoring systems [13–16,17]. However, most of these optical sensors are limited by photostability, especially for dyes, and toxicity, especially for QDs. Furthermore, noble metal nanoparticles are sensitive to other factors in the local environment, besides glucose levels, for example their emission can be affected by pH levels [18–20,31].

To overcome these drawbacks, lanthanide ion (rare earth elements) doped nanocrystals have recently been investigated, especially those that function as upconversion nanoparticles (UCNPs). UCNPs are preferred because they are almost immune to the ubiquitous bio-fluorescent background [21–24]. Many UCNPs operate via energy transfer upconversion (ETU) in which a sensitizer ion with a large near-infrared (NIR) absorption cross section (typically ytterbium, Yb³⁺) absorbs usually at 980 nm and sequentially transfers the absorbed energy to an upconverting ion such as (Er³⁺, Ho³⁺, or Tm³⁺) [25–27]. The upconverting ion stores this energy in a metastable state until an NIR photon is absorbed and transferred, at which point the excited-state absorption (ESA) brings the upconverting ions to a higher energy level that can emit visible to NIR emission, preferably in the tissue transparency region. Since almost all autofluorescence from biological tissues is red-shifted relative to the laser excitation, it is easily filtered out.

Nevertheless, there are two main issues that might limit the use of the UCNPs as an optical glucose sensor. First, to make UCNPs sensitive to glucose, they must be functionalized, for example with manganese dioxide (MnO₂)-nanosheets [28] or molecules like boronic acid. However, these can be affected by bloodstream conditions, such as pH level, resulting in reduced measurement sensitivity and erroneous readings [4,12,29]. Second, the absorption maximum of the typical sensitizer of UCNPs (Yb³⁺) overlaps with the strong absorption of water at 980 nm and can cause heating [30, 31, 50]. This can also interfere with glucose sensing because UCNPs respond similarly to temperature changes [32–37]. Therefore, core/shell UCNPs have been developed that can be efficiently excited at a wavelength (808 nm) where heating due to water absorption is minimized [38].

Here, we report the use of water-tolerant YVO₄ and GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ upconversion nanoparticles in a core/shell structure for accurate, selective, and reversible optical glucose sensing. The key advance is that these particles sense glucose with no surface functionalization. We find similar glucose sensitivity for both particles (after taking into account size differences), which suggests that the ${\text{VO}}_4^{-}$ ion is likely responsible for the glucose sensitivity. Interestingly, other studies show that the vanadate ion can interact with glucose in a way that can be used to control blood sugar [39–42]. In addition to being naturally glucose-sensitive, vanadate crystals like YVO₄ and GdVO₄ have shown high upconversion efficiency and good photostability in water, as reported in [35,43–45].

2. Methods

The YVO₄ core/shell particles are discussed in detail in previous work [36]. Briefly, the core is doped with Yb and Er, and the shell with Yb and Nd. They have an average size of 20 nm, The GdVO₄ particles are also core/shell with similar Yb, Er, and Nd dopants.

2.1. Synthesis of (Gd/Y)VO₄ : Yb, Er@Nd core/shell nanoparticles

Briefly, for both systems, core nanoparticles were prepared following a hydrothermal synthesis procedure described in [42–44]. Their synthesis is described below.

2.1.1. Core synthesis

A 10 mL solution of Gd(NO₃)₃.4H₂O (78% mol/L), plus Er(NO₃)₃.5H₂O (2% mol/L), and Yb(NO₃)₃.5H₂O (20% mol/L) was prepared and slowly added dropwise using a peristaltic pump to a 7.5 ml solution of sodium citrate (0.1 mol/L). This rare earth (Re³⁺ sodium citrate mixture was kept under vigorous stirring until a white precipitate was formed. Then, a second solution consisting of 10 mL of sodium orthovanadate (Na₃VO₄, 0.1 mol/L) dissolved in distilled water was added slowly to the (Re3+) sodium citrate mixture under constant stirring for 1 hour at room temperature with a few drops of 1 M NaOH to maintain the pH above 11. This process results in core nanoparticles. For the synthesis of core-only particles, the mixture was then transferred into a polytetrafluoroethylene (PTFE) vessel. The vessel was placed into a stainless steel, high-pressure autoclave chamber and kept under 7 MPa pressure, and 230 °C for 24 h. The core nanoparticles were collected after washing three times with pure water and centrifugation (10000 rpm (5600 g’s) for 10 min).

2.1.2. Shell mixture preparation

A 10 mL solution of Gd(NO₃)₃.4H₂O (78% mol/L), plus Nd(NO₃)₃.5H₂O (2% mol/L), and Yb(NO₃)₃.5H₂O (20% mol/L) was prepared and slowly added dropwise to a 7.5 ml solution of sodium citrate (0.1 mol/L) using a peristaltic pump. This rare earth(Re³⁺) sodium citrate mixture was kept under vigorous stirring until a white precipitate was formed. Then, a second solution consisting of 10 mL of sodium orthovanadate (Na₃VO₄, 0.1 mol/L) dissolved in distilled water was added slowly to (Re³⁺ sodiumcitrate) mixture under constant stirring with a few drops of 1 M NaOH to maintain the pH above 11 for 1 hour at room temperature to form the shell solution.

2.1.3. Core/shell synthesis

For core/shell structure synthesis, the Nd³⁺ containing shell solution was added to the GdVO₄ : Yb³⁺, Er³⁺ core nanoparticles to overgrow a shell layer (3–4 nm) of Y or GdVO₄ : Yb³⁺, Nd³⁺ at 10% Yb³⁺ and 10% Nd³⁺ doping ratio. This is done by mixing the core mixture and shell mixture (as described above in the core and shell syntheses, respectively) as follows: the shell mixture was added slowly dropwise to the core nanoparticles mixture and kept under constant stirring for one hour. Next, the final mixture was transferred into an autoclave for hydrothermal treatment at 230 °C and pressure 7 MPa for 24 hours. After that, it was left to cool down until reaching room temperature. The concentrations of Yb³⁺ and Nd³⁺ in the shell layer were carefully optimized, as reported in [31]. Finally, the YVO₄ and GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell nanoparticles were obtained and stored at room temperature.

3. Results and discussion

Prior to the optical experiments, we performed transmission electron microscopy (TEM) imaging of the YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ and GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ nanoparticles to determine their sizes. To verify upconversion fluorescence, a thin layer of GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs was uniformly distributed on a quartz slide. Fig. 1(a) shows the resulting upconversion spectrum of GdVO₄ : Er³⁺Yb³⁺@Nd³⁺ core/shell UCNPs at a relatively low laser intensity of 200 W/cm². The upconversion optical spectrum shows peaks located at 525 nm, 552 nm, and 650 nm corresponding to ²H_11/2 → ⁴I_15/2, ⁴S_3/2 → ⁴I_15/2, and ⁴F_9/2 → ⁴I_15/2 transitions of Er³⁺ ion [49].

 

Fig. 1 (a) Upconversion luminescence spectrum of GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs. (b) The TEM image of GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs. (c) Upconversion luminescence spectrum of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs. (d) The TEM image of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs. (e) An illustration of a home-made confocal microscope equipped with visible-near-infrared objective, NIR laser, and spectrometer. (f) Energy transfer mechanism and electronic structure of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs.

Download Full Size | PDF

For GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ UCNPs, a droplet was placed on a carbon grid for TEM imaging. Fig. 1(b) shows the TEM image that reveals dispersed and well-crystallized nanoparticles with average size of 60–80 nm. The larger size than expected was likely due to the high temperature treatment of 1000 °C for 10 min. To avoid this problem in the future, we note that silica-gel encapsulation has been reported to prevent the nanoparticle size from increasing during high temperature annealing [42–44]. The TEM and optical characterization of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ were discussed in detail in previous work [36]. Fig. 1(c) shows the resulting upconversion spectrum of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs, and the related TEM image in Fig. 1 (d).

For optical characterization of YVO₄ and GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs, a custom-made confocal laser scanning microscope setup was built, as shown in Fig. 1(e). The optical confocal microscope is equipped with a 90×, NIR microscope objective with NA=0.8, and an 808 nm laser diode for optical excitation. After filtering, the excitation laser with a short-pass (750 nm SP) filter, the upconversion fluorescence of the UCNPs was collected through the same microscope objective and analyzed with a custom-made spectrometer equipped with a starlight camera (Trius camera model SX-674), and a photon counter (Hamamatsu photon counter model H7155-21).

As mentioned earlier, YVO₄ and GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs exhibit efficient upconversion energy-transfer, as reported in [36]. Briefly, as shown in Fig. 1(f), a near-infrared (NIR) photon (808 nm) excites Nd³⁺ to its ⁴F_5/2 excited state, followed by non-radiative relaxation to ⁴F_3/2 state. Energy-transfer populates the ²F_5/2 state of nearby Yb³⁺ in the shell, and then crosses the shell layer toward nearby Yb³⁺ in the core to populate their ²F_5/2 states, which then initiates the typical UC process in the Er ³⁺ ion. In the core, the energy transfer from the ²F_5/2 excited state of Yb³⁺ promotes to its metastable state where the energy is stored. Sequentially, a second NIR photon’s energy is transferred from the ²F_5/2 excited state of Yb³⁺ to further excite the Er³⁺ to highly excited states (²H_11/2,⁴S_3/2, and ⁴F_9/2) of Er³⁺, after subsequent multi-phonon relaxations. Consequently, two strong green emissions and one weak red emission occur corresponding to these transitions: ²H_11/2 → ²I_15/2, ⁴S_3/2 → ⁴I_15/2, and ⁴F_9/2 → ⁴I_15/2 respectively.

In our first study, we investigated optical glucose sensing using YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs with an average size of 20 nm (Fig. 1(d)), which has been previously reported by [37]. After synthesis, the particles were buffered by combining 150 μL of 0.1 M PBS buffer solution (pH = 7.4) with 20 μL of an aqueous solution of the YVO₄ particles. This suspension was then placed into a 2 mL colorimetric tube. Multiple identical tubes were then prepared in the same manner. Then, different quantities of glucose were added to each tube, as described in [45]. The mixtures in each tube were then diluted to a volume of 1 mL by adding water and mixing thoroughly. Instead of measuring fluorescence in solution, we waited one hour, then placed a 20μL droplet of each sample (with different glucose concentrations) onto a microscope coverslip, and allowed it to air dry for 10–15 minutes.

The fluorescence spectra of each sample was recorded by our custom-built confocal microscope. Optical spectra were collected from individual particles (not the clumped particles). As shown in Fig. 2(a), the UC luminescence of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs decreased gradually with increasing glucose concentrations from 0 to 30 mM. The influence of glucose concentration on the quenching efficiency at 552.8 nm is summarized by the plot in Fig. 2(b). Here, the quenching efficiency was defined by (I₀ − I)/I₀, where I₀ and I represent the fluorescence intensity without and with glucose; respectively, at the 552.8 nm line of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺. At first, the quenching increased linearly with the glucose concentration, throughout the range of normal glucose level in blood (from 4 to 6 mM). Then, it saturated above 10 mM concentration, which corresponded to 70% quenching. The experiment was repeated several times using dried droplets, and also using liquid droplets (data not shown) and the results were the same within the experimental error. Later, we also measured fixed particles in a flowing glucose solution (to be described next). Again, results were similar.

 

Fig. 2 (a) Upconversion luminescence spectra of YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs recorded as a function of different glucose concentrations (0–30 mM). (b) Influence of the glucose concentrations on the quenching efficiency at 552.8 nm for the glucose concentrations of 1 mM, 10 mM and 100 mM.

Download Full Size | PDF

One possible mechanism for glucose detection (suppression) by lanthanide ions doped vanadium oxide is the stimulation of glucose polymerization to form glycogen by VO₄ [46]. Glycogen exhibits a relatively good absorbance in the visible spectral region [47], which overlaps with UCNPs green emissions and thus can function as an effective energy quencher for UCNPs through reabsorption (inner filter effect) [48]. Fig. 2(a) shows a clear change in the ratio of erbium two emission bands centered at 525nm and 550nm. This effect happens because that the absorption spectrum of glycogen is not flat over erbium emission peaks, hence, the first peak at 525nm get quenched faster than the other peak at 550nm during glucose sensing process. However, further experimentation is needed to confirm this possibility.

The next phase of our study was to determine the reversibilty of the observed optical glucose sensing. In this study, a layer of dispersed UCNPs was fixed to a microscope slide such that the same particles could be investigated while the glucose concentration was changed. To make this fixed layer, 20 μL of GdVO₄ or (YVO₄) : Er³⁺, Yb³⁺@Nd³⁺ core/shell upconversion nanoparticles were mixed with polyvinyl alcohol (PVA) solution (1:1) and spin coated on a microscope coverslip. Then, the spin-coated nanoparticles were covered with another microscopic coverslip to create a flow channel to allow for incremental glucose concentration changes and afterward to return to the original concentration in order to verify UCNP’s reversibility.

Fig. 3 shows the glucose sensitivity in the flow-channel using GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ core/shell UCNPs (average size 60–80 nm). Similar to the dried UCNPs discussed in Fig. 2, Fig. 3(a) shows a strong quenching of the UC luminescence with increasing glucose concentrations. Similarly, the quenching efficiency vs. glucose concentration at 552.8 nm was plotted and is shown in Fig. 3(b), which reveals similar performance when comparing these results to Fig. 2(b) (YVO₄ particles) when taking into account the size differences of the particles (and therefore the surface to volume ratio).

 

Fig. 3 (a) Upconverion luminescence spectra of GdVO₄ : Yb³⁺, Er³⁺@Nd³⁺ core/shell UCNPs recorded as a function of high and low glucose concentration ranges. (b) Influence of the glucose concentrations on the quenching efficiency at 552.8 nm for the glucose concentrations of 1 mM, 10 mM, and 100 mM.

Download Full Size | PDF

Similar to YVO₄ particles, the quenching efficiency for GdVO₄ particles was defined by (I₀ − I)/I₀, where I₀ and I represent the fluorescence intensity without and with glucose, respectively; at the 552.8 nm line of GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺.

To verify that our glucose sensing is reversible, we flowed DI water over the YVO₄ particles after the aforementioned quenching measurements. Specifically, a 10 mM glucose solution was first added to produce quenching, then the particles were rinsed with DI water and the fluorescence intensity was measured again. As shown in Fig. 4(a), the UC intensity was, indeed, recovered after rinsing, demonstrating reversibility, as hypothesized. Fig. 4(a) shows that the UCNPs emission slightly increased after rinsing the glucose away with DI water compared to the original spectrum curve. This unexpected increase in the UCNPs emission can be attributed to unexpected movement of the sample in the optical spot or a small fluctuation of laser intensity.

 

Fig. 4 (a) Reversible upconversion luminescence spectra of GdVO₄ : Yb³⁺, Er³⁺@Nd³⁺ core/shell UCNPs after adding and then washing away 10 mM glucose concentrations (b) Unchanged upconversion spectra of GdVO₄ : Yb³⁺, Er³⁺@Nd³⁺ core/shell UCNPs after adding and then washing away various fructose concentrations. (c) Normalized upconversion spectra of YVO₄ : Yb³⁺, Er³⁺@Nd³⁺ core/shell UCNPs as a function of glucose concentrations.

Download Full Size | PDF

Next, we investigated the specificity of GdVO₄ : Yb³⁺, Er³⁺@Nd³⁺ core/shell UCNPs fluorescence quenching in order to estimate the performance in the presence of interferents. To do this, we repeated the above experiment, but this time with glucose being replaced by fructose. As shown in Fig. 4(b), we observed no quenching effect from fructose. This demonstrates glucose selectivity of UCNPs over the similar fructose molecule. Sensitivity to other chemicals will be investigated in the future.

Finally, to demonstrate the possibility of accurate glucose sensing in the presence of laser-intensity fluctuations and other environmental factors, we present a normalized (to the 520 nm peak) plot of the UC spectra quenching vs. glucose concentrations, in Fig. 4(c). As it can be seen, the ratio of emission peak heights is also sensitive to glucose, and hence, it is possible to develop a sensor that is immune to intensity fluctuations caused by the laser or some other interferents.

4. Conclusions

We synthesized small YVO₄ and GdVO₄ : Er³⁺, Yb³⁺@Nd³⁺ UCNPs in a core/shell structure where their excitation wavelength was in the biocompatible band. We showed that these UCNPs are sensitive, reversible, and selective glucose sensors, without the need for any functionalization or surface modifications that might cause toxicity, interference, or long-term instability. This feature makes YVO₄ : Er³⁺, Yb³⁺@Nd³⁺ upconversion nanoparticles due to its smaller size especially interesting for glucose sensing in biological systems. In particular, these nanoparticles have great potential to be used as optical sensors in both external and implantable continuous glucose monitoring systems.