Sample-Invariant Salivary Urea Quantification across the Clinically Relevant Range Using a Silicon Device-Based pH Sensor

Abstract

Urea is a clinically important biomarker for monitoring chronic kidney disease (CKD). Saliva provides a noninvasive alternative to blood for urea measurements, making it attractive for routine disease monitoring. Urease-based pH sensing has been extensively investigated for urea diagnostics due to its transduction simplicity and clinical relevance. However, this approach has two key challenges that are critical for quantitative accuracy but remain unexplored in existing literature. First, significant variations in salivary pH and buffering capacity across populations and within the same person lead to sample-dependent pH-urea calibration curves, thereby compromising quantitative accuracy. Second, pH sensor responses saturate at elevated urea concentrations, thus limiting quantitative detection across the full clinically relevant range (2–45 mM). To address these challenges, a saliva dilution strategy was investigated using bipolar junction transistor (BJT)-based sensors with enhanced pH sensitivity. The optimal conditions, comprising a 60-fold dilution of saliva in 4 mM MES buffer (pH 5.88), produced a saliva-independent linear pH–urea calibration curve spanning the entire clinically relevant urea concentration range. The BJT sensor-based urea measurements were validated against a commercial gold-standard colorimetric assay, with Bland–Altman analysis showing a mean difference of −0.858 mM. In summary, this work provides a pathway toward robust, scalable, and miniaturized urea sensors for portable CKD diagnostics.

1. Introduction

Urea is one of the biomarkers for kidney function monitoring in chronic kidney disease (CKD). In clinical practice, urea is assessed in blood as blood urea nitrogen (BUN). Recently, salivary urea has emerged as a promising noninvasive alternative due to its ease of collection and demonstrated correlation with blood urea levels. − Importantly, independent cohort studies have observed that salivary urea levels show a progressive association with CKD severity: mean salivary urea increases from ∼2–3.4 mM for healthy individuals ,, to 15.85–26.65 mM across CKD stages 3–5. , In dialysis-dependent end-stage renal disease, predialysis salivary urea can reach up to ∼40–45 mM in individual patients. , Hence, across health and disease, the clinically relevant salivary urea range spans from ∼2 to 45 mM.

There has been increasing research effort toward developing salivary urea sensors for home care diagnostics. − A popular approach for urea sensing is based on the urease-catalyzed hydrolysis of urea, which produces an alkaline shift in the solution as given by eq .

$$\begin{array}{l}\mathrm{N}{\mathrm{H}}_3+{\mathrm{H}}_2\mathrm{O}\leftrightharpoons \mathrm{N}{{\mathrm{H}}_4}^{+}+\mathrm{O}{\mathrm{H}}^{-}\\ \mathrm{C}{\mathrm{O}}_2+{\mathrm{H}}_2\mathrm{O}\leftrightharpoons \mathrm{H}\mathrm{C}{{\mathrm{O}}_3}^{-}+{\mathrm{H}}^{+}\end{array}$$ 1

where urea hydrolysis is catalyzed by urease, producing two ammonia molecules and one CO₂ molecule, which are further hydrolyzed to produce a net increase in pH, thus providing a measure of urea concentration. Notably, the concentrations of the hydrolysis products and the resulting pH changes are pH dependent.

The pH-sensing approach (eq ) offers several advantages such as high enzyme specificity, simple readout based on pH change, no redox reactions needed, and compatibility with a wide range of biofluids such as saliva, sweat, urine, and blood. , As a result, this approach has been extensively investigated using field-effect transistor (FET)-based pH sensors, which combine high sensitivity, cost-effective scalability, low power consumption, and system-level integrability, offering significant advantages over alternative sensing approaches.

In recent years, several studies have demonstrated urea detection using FET-based pH sensors across various concentration ranges. For example, Yang et al. used a silicon FET pH sensor to measure urea concentrations from 0.2 to 7 mM in a phosphate-buffered saline solution, with signal saturation observed at higher concentrations. Another recent study demonstrates FET-based pH sensing of urea at lower concentration range (0.0167–1.7 mM) with saturation at higher levels (>2 mM). Choi et al. reported silicon FET pH sensor measurements in phosphate buffer, achieving a detection limit of 0.6 mM but exhibiting signal saturation at urea concentrations ≥10 mM. Graphene oxide FET-based pH sensors have also been used for urea detection in the 0.001–1 mM range in HEPES buffer. Additional studies have adopted FET-based pH sensing for urea detection using various test conditions. − Majority of these studies were performed in buffer solution with a few exceptions that were performed in artificial saliva or saliva. Studies involving saliva have highlighted inherent challenges arising from nonspecific protein adsorption and enzyme biofouling caused by the complex mixture of biomolecules present in raw saliva. To mitigate these effects, microsieve membrane filtration has been integrated into sensing platforms to remove larger proteins from the sample prior to analysis. Despite these advances, two basic limitations continue to hinder the clinical translation of pH sensor-based urea diagnostics. First, most studies evaluate sensor performance in a single test solution with a fixed pH and buffering capacity. In contrast, human saliva exhibits substantial inter- and intraindividual variability in both pH and buffering capacity, − which can strongly influence the pH versus urea calibration curve. Consequently, the impact of test sample pH and buffering capacity on pH-urea calibration remains poorly understood. Second, as discussed above, many studies have observed that pH sensor responses saturate at higher urea concentrations, which prevents the coverage of the full clinically relevant range spanning from 2 to 45 mM. ,, Hence, addressing these limitations is essential to enable saliva-invariant calibration in the clinically relevant urea range.

This study addresses the key limitations hindering accurate urea quantification arising from inter- and intraindividual variations in salivary pH and buffering capacity. We use a previously demonstrated bipolar junction transistor (BJT)-based pH sensor that not only retains the advantages of silicon FET sensors , but also has additional attributes that are particularly well-suited for scalable and portable sensing applications. Using these BJT pH sensors, artificial saliva formulations were benchmarked against human saliva and shown to exhibit comparable chemical characteristics, thus ensuring the clinical relevance of reported results. Comprehensive studies of artificial saliva samples with varied initial pH and buffering capacities were performed to identify optimal conditions that yield a saliva-invariant, linear pH-urea calibration curve with a single slope across the full clinically relevant urea range (2–45 mM). This study also provides insights into sensing signal saturation at higher urea levels. The BJT pH sensor-based urea concentration measurements were validated against a commercial gold-standard colorimetric assay. In summary, this study addresses key technological gaps in the development of reliable, scalable urea sensors for CKD monitoring.

2. Materials and Methods

2.1. Reagents and Chemicals

The following reagents and materials were used in the experiments: artificial saliva with mucin, custom pH (pH 5.63, 6.36, 7.25) (Pickering Laboratories), urease from Canavalia ensiformis (Jack Bean) type IX (50,000–1,00,000 units/g, Sigma-Aldrich), Hydrion Buffer Chemvelope (pH 5, 6, 7, 7.4, 8) (Micro Essential Laboratory), PBS, pH 7.4 (GibcoTM), MES (1.0 M buffer solution, pH 6.0, Thermo Scientific Chemicals), Urea Nitrogen (BUN) Colorimetric Detection Kit (Invitrogen). All experiments were performed using Protein LoBind Tubes (Eppendorf); pH levels of the reagents were measured upon receiving using pH meter.

2.2. BJT pH Sensor Components and Electrical Sensing Measurements

The BJT pH sensor was constructed using a commercially available NPN bipolar junction transistor (2N2222A), with its base connected to a titanium nitride (TiN)-coated tungsten tip (radius ∼0.5 mm). The TiN coating was sputter deposited and was 100 nm thick with a resistivity of 0.0484 Ω cm. A commercial Ag/AgCl reference electrode (Innovative Instruments, Inc.; 1 mm diameter) was used for all measurements. Electrical measurements were performed using a Keysight B1500 source measurement unit (SMU). The sensing current was measured by applying voltages V_E, V_C, and V_B to the emitter, collector, and reference electrode, respectively. For all measurements, V_C = V_B = 0 V, while V_E (<0 V) was either varied or held at a fixed value. All sensing measurements were made at room temperature.

2.3. Artificial Saliva and Buffer Titrations

Titrations of artificial saliva and MES buffer were performed using the BJT pH sensor to monitor the pH change during the addition of titrant (0.1 M NaOH), and the solution was mixed by pipetting 20 times after each addition.

2.4. Urea Sensing Process for Undiluted Samples

1.2 mL of artificial saliva spiked with various urea concentrations (0–45.8 mM) was loaded into a low-bind microcentrifuge tube. The TiN sensing electrode and reference electrode (RE) were inserted into the sample, and the time-series sensing current I _C(t) was recorded continuously with a time interval of 1 s. Once the baseline I _C signal was established, 2 μL of urease solution (containing 15 units (U) of urease prepared in 4 mM MES) was added directly into the sample. The solution was mixed by pipetting 20 times over ∼20 s. The I _C(t) signal was recorded continuously throughout urease addition, mixing, and enzymatic reaction, until a postreaction steady-state value was reached.

2.5. Urea Sensing Process Using Optimized Protocol

For urea sensing under optimized conditions, 1.2 mL of the diluent containing 5 U urease in 4 mM MES was first loaded into a low-bind microcentrifuge tube. The TiN and RE were then inserted into the sample, and the time-series sensing current I _C(t) was recorded continuously with a time interval of 1 s. Once the baseline I _C signal was established, 20 μL of the urease-containing MES solution was removed and replaced with 20 μL of the AS sample (resulting in a 60× dilution of the added sample). The solution was mixed by pipetting 20 times over ∼20 s. The I _C(t) signal was recorded continuously throughout urea sample addition, mixing and enzymatic reaction, until a postreaction steady-state value was reached.

The 5 units (U) urease was selected to provide sufficient enzymatic activity while minimizing reagent consumption. According to the supplier specification (Sigma-Aldrich): one unit of urease liberates 1.0 μmol of ammonia from urea per minute at pH 7.0 and 25 °C under the supplier-defined assay conditions. This is equivalent to hydrolyzing 0.5 μmol of urea per minute; thus, 5 U has a theoretical hydrolysis capacity of 2.5 μmol urea/min. For the highest tested salivary urea concentration of 45.8 mM, with 60× dilution in the 1.2 mL measurement solution, approximately 0.92 μmol of urea is present. Thus, 5 U urease provides excess enzymatic activity relative to the maximum tested urea amount, while also accounting for potential enzyme activity reductions under the actual measurement conditions compared with the supplier-defined assay conditions.

The repeatability of the sensing protocol was evaluated across 80 independent samples measured over 10 days, yielding a mean ⟨ΔpH⟩ of 0.000 with a standard deviation of 0.009 pH units (Figure G,H), confirming that the protocol introduces no measurable systematic drift.

1.

BJT-based pH sensor schematics and pH measurement principles. (A) Schematic diagram of the bipolar junction transistor (BJT) based pH sensor, where BJT device is the transducer, TiN is the pH-sensitive sensing surface and RE is the miniaturized AgCl reference electrode; V_E (<0 V), V_C (= 0 V), V_B (= 0 V) are voltages applied at the emitter, collector and base, respectively; collector current (I _C) is the sensing signal. (B) Dependence of I _C on applied voltage V_BE (= V_B–V_E) for different pH buffer solutions. (C) Dependence of I _C at a fixed V_BE = 0.730 V on the buffer solution pH; symbols are the extracted data from (B), and the solid line is the fit with α = 0.949; inset shows extracted parameter mean α dependence across a range of applied voltage V_BE. (D) Representative measured I _C dependence on time before and after the start of urea–urease reaction. (E) Conversion of sensing current I _C from (D) to corresponding pH change (ΔpH­(t)) using eq with α = 0.949, where ΔpH­(t) is the pH difference between before and after the reaction. (F) Conversion of ΔpH­(t) to pH­(t): pH­(t) = pH_initial + ΔpH­(t), where pH_initial is the test solution pH before the start of the reaction. (G) Method for evaluating pH measurement variabilities, including that induced by solution mixing (via pipetting) in the absence of reaction. Representative time series shows continuous pH measurement over ∼1000 s, interrupted by three mixing events; steady-state pH values after 3 mixing events are labeled as pH₁, pH₂, and pH₃; ΔpH_i is the difference between two consecutive steady-state pH values before and after ith mixing event. (H) Dependence of <ΔpH> for 80 different samples of artificial saliva and buffer solutions measured over 10 days. Each sample was measured using the method shown in (G), where <ΔpH> is the mean of three ΔpH values. Combining the data for 80 samples measured over 10 days, mean <ΔpH> ± <S.D.> = 0.000 ± 0.009, where the dashed line is the mean and the shaded region indicates the standard deviation.

2.6. Gold Standard Assay

Urea reference measurements were obtained using the Invitrogen Urea Nitrogen (BUN) Colorimetric Detection Kit (Thermo Fisher Scientific), following the manufacturer’s protocol. Briefly, standards and diluted samples were added to a 96-well plate, mixed with color reagents, and incubated for 30 min at room temperature. Absorbance was measured at 450 nm using a benchtop spectrophotometer. The manufacturer-reported analytical sensitivity is 0.030 mg/dL BUN (0.011 mM urea); standard curve range of 0.156–10 mg/dL BUN (0.056–3.57 mM urea), with a linear absorbance response across all standard concentrations (0–10 mg/dL); intra-assay CV of 1.9–2.8%; and interassay CV of 3.1– 4.3%. The kit reports urea nitrogen (UN) concentration, which was converted to urea concentration using the molecular weight ratio between urea (60.06 g/mol) and urea nitrogen (28 g/mol).

3. Results and Discussions

3.1. BJT-Based pH Sensor Evaluation

This study uses a previously proposed bipolar junction transistor (BJT)-based pH sensor to measure the changes in the pH of the test solution due to enzymatic urea–urease reactions, thereby measuring urea concentrations in test solutions. In this section, the BJT pH sensor is evaluated, including sensing principle, pH calibration, signal repeatability, and drift with time.

The BJT pH sensor has been previously demonstrated and validated. , In comparison to widely used FET-based sensors, BJT sensors have higher sensitivity and signal-to-noise ratio that are independent of use voltages, thus resulting in simpler calibration. Additionally, the BJT pH sensor fabrication is compatible with silicon processing technology and can be cost-effectively mass-produced in a semiconductor foundry. Furthermore, similar to FET sensors, the BJT pH sensor can be readily integrated with automated data acquisition and wireless communication circuitry powered by a 3.2 V battery to form a portable pH-sensing platform. Hence, BJT pH sensors were selected over FET sensors.

Figure A shows the schematics of the BJT pH sensor with its three components: (i) a BJT device as the transducer, (ii) the pH-sensitive TiN sensing surface connected to the BJT base, and (iii) a miniaturized reference electrode (RE). The TiN sensing surface and RE are immersed in the test solution. The collector current (I _C) was the sensing signal. I _C is measured by applying voltages to the emitter, collector, and RE, respectively; V_C = V_B = 0 V, and V_E < 0 V.

Figure B shows the sensing current I _C as a function of the applied voltage V_BE (= V_B–V_E) measured in different pH buffer solutions. I _C is observed to increase exponentially with V_BE with a subthreshold slope of SS = 59 mV/decade; this measured SS value is close to the theoretical limit at room temperature. Since both sensor sensitivity and signal-to-noise ratio improve with decreasing SS, measured SS = 59 mV/decade results in enhanced pH sensing characteristics. Figure B also shows that the I _C curve shifts by 57 mV per unit change in pH (Supporting Information Figure S1), indicating the Nernst limit pH sensitivity of TiN sensing surface. It may be noted that the applied voltage magnitudes are less than 1 V, thus making BJT sensors well suited for portable applications. Using data at a fixed V_BE = 0.730 V from Figure B, the sensor calibration equation is derived as shown in Figure C, where the symbols denote data and the solid line is the fit. From this fit, the calibration equation can be written as

$${\mathrm{I}}_{\mathrm{C}}=k·{10}^{-\alpha ·\mathrm{pH}}$$ 2

where mean α = 0.944 ± 0.006 and is independent of applied voltage V_BE as shown in the inset in Figure C. This value is voltage independent because measured I _C curves in different pH buffer solutions are parallel to each other. k is a constant that depends on V_BE.

We now discuss the procedure for converting the measured sensing current I _C into corresponding pH values. Figure D shows a representative time dependent I _C curve measured at a constant applied voltage V_BE in an aqueous test solution before and after the start of urea–urease reaction. As shown in the figure, the enzyme reaction caused the sensing current I _C to sharply decrease over ∼50 s and start to reach a steady-state value ∼250 s after the start of the reaction. These observed I _C changes are attributed to pH changes, occurring due to the urea–urease reaction and an equation for time dependent ΔpH­(t) can be derived from eq as shown below.

$$\mathrm{\Delta }\mathrm{pH}(t)=-\frac{1}{\alpha }\log \left(\frac{I_{{\mathrm{C}}_{,\mathrm{a}\mathrm{fter}}}(t)}{I_{{\mathrm{C},}_{\mathrm{b}\mathrm{efore}}}}\right)$$ 3

where I _C,After(t) is the time-dependent sensing current after the initiation of urea–urease reaction, and I _C,Before(t) is the constant current before the start of the reaction. Using eq , the I _C time series data were converted to corresponding ΔpH­(t) as shown in Figure E. As shown, the enzyme reaction causes ΔpH­(t) value to sharply increase and reach the steady-state values at longer times. Since pH­(t) is the sum of ΔpH­(t) and initial pH before the start of the reaction, the corresponding pH­(t) is shown in Figure F. In subsequent sections, this method is used for converting measured I _C into corresponding ΔpH­(t) and pH­(t).

To evaluate BJT pH sensor performance, experiments were conducted to evaluate pH measurement repeatability and drift with time in the absence of enzymatic reactions. pH was continuously measured for over 1000 s, interrupted with three mixing steps that would mimic the addition and mixing of reagents to initiate urea–urease reactions. Figure G shows a representative pH time series with the mixing process. Prior to any mixing event, the initial steady-state pH was recorded as pH₀. After each mixing event, the pH signal recovered to a steady-state value, denoted as pH₁, pH₂, and pH₃. The difference between two consecutive steady-state pH values before and after the ith mixing event is defined as ΔpH _i = pH _i –pH _i–1. The resulting ΔpH _i is a measure of the nonbiochemical signal fluctuations after each event of reagent addition and solution mixing via pipetting.

Using the protocol described in Figure G, the systematic signal drift over time was evaluated by measuring 80 independent samples of artificial saliva and buffer solutions measured over a 10-day period. Figure H shows <ΔpH> and standard deviation values for all the samples; <ΔpH> is the average of three ΔpH _i values obtained using the protocol discussed above. Combining the data for 80 samples, the mean <ΔpH> value is 0.000 and the standard deviation is 0.009 pH units. These repeated pH measurements made over 10 days show that the sensor and measurement protocol do not introduce measurable long-term cumulative drift, and the pH can be measured with ±0.009 pH unit precision. Additionally, over the course of the entire study, the pH-sensitive TiN was immersed in diluted saliva for ∼100 h, with no measurable changes in pH sensitivity.

3.2. Artificial Saliva Comparison with Human Saliva

Commercially available artificial saliva (AS) formulations (Supporting Information, Table S1) with three different initial pH values (pH 5.62, 6.36, and 7.26) were used in the study. These AS formulations are designed to mimic the composition and pH of human saliva. Since the buffering capacity of the saliva sample can influence enzymatic reaction-driven pH changes, it is important to experimentally confirm that the AS samples used in this study also reproduce the buffering properties of human saliva.

To assess this, titration experiments were performed by incrementally adding NaOH and recording the steady-state pH after each addition. Figure A shows measured titration curves for three AS samples with different initial pH values, where symbols represent the measured steady-state pH values. All AS samples show a continuous increase in pH with increasing hydroxide addition.

2.

Artificial saliva (AS) buffering capacity (β) measurement. (A) Titration curves of three different AS samples with initial pH values of 5.62, 6.36, and 7.26; symbols represent steady-state pH, measured after each addition of NaOH. (B) Comparison of measured buffering capacities of artificial saliva samples with literature values for human saliva. Buffering capacities of AS were estimated from (A).

Using the data in Figure A, the buffering capacity (β) of each AS sample was estimated from eq shown below.

$$\beta =\frac{\mathrm{\Delta }{n}_{{\mathrm{OH}}^{-}}}{V·\mathrm{\Delta }\mathrm{pH}}$$ 4

where Δn _OH^– is the amount of hydroxide added (mmol), V is the saliva volume (L), and ΔpH is the corresponding change in pH. The buffering capacity is expressed in units of mmol·L^–1·pH^–1. Figure B shows the estimated buffering capacities of AS (filled symbols) and compares them to those of human samples (open symbols). As shown, buffering capacities of AS are comparable to those reported in the literature for both unstimulated and stimulated human saliva across the physiologically relevant pH range. Thus, the artificial saliva used in this study not only contains composition (Supporting Information, Table S1) and pH similar to those of human saliva, but also closely reproduces its buffering behavior. Hence, artificial saliva experiments can be translated to expected behavior in human saliva.

3.3. Urea Measurements in Undiluted Saliva

We first investigated pH-urea calibration curves using three undiluted artificial saliva samples with different initial pH values of 5.62, 6.36, and 7.26 and buffering capacities (Figure B). Each sample was spiked with urea concentrations ranging from 0 to 45.8 mM, and time-dependent pH measurements were performed following enzyme addition. Figure A–C shows the time-dependent ΔpH­(t) curves before and after enzyme additions, where ΔpH­(t) is extracted from eq and the sensing current I _C. For all samples, ΔpH­(t) increases rapidly at shorter times (t < 100 s) and reaches a steady-state value at longer times (t > 250 s). Using the steady-state values at t = 300 s, the dependence of ΔpH on urea concentration is plotted in Figure D. As shown, ΔpH-urea calibration curves exhibit a strong dependence on the initial pH values of saliva samples. For each sample, ΔpH increases linearly at lower urea concentrations (<10 mM) and then saturates to constant values at higher concentrations, corresponding to the chronic kidney disease range (yellow shaded region). The slope shows a weak dependence on initial pH and buffering capacities of saliva. In contrast, the saturated ΔpH values show a strong dependence on initial saliva pH, increasing by approximately a factor of 3 as the initial pH decreases from 7.26 to 5.62. The observed ΔpH saturation at higher urea levels limits disease monitoring and therefore needs to be examined in greater detail.

3.

Urea concentration measurements in artificial saliva (AS). (A–C) Time series ΔpH­(t) curves for urea–urease reactions for varying urea concentrations (0–45.8 mM) in three different artificial saliva samples with initial pH of 5.62, 6.36, and 7.26. (D) Measured steady-state ΔpH (at t ∼ 300 s) dependence on urea concentration for different AS samples; ΔpH values are estimated from A to C. Measured ΔpH saturates at clinically significant CKD urea levels (yellow highlighted region) for AS samples. (E) Measured ΔpH dependence on urea concentration for phosphate-buffered saline (PBS) at two different concentrations of 1× PBS and 0.25× PBS. (F) pH (at t ∼ 300 s) dependence on urea concentration for both AS and PBS.

To investigate whether the steady-state ΔpH saturation behavior is specific to saliva or a basic property of the urea–urease reaction, similar experiments were repeated in phosphate-buffered saline (PBS) solutions at two different ionic strengths: 1× PBS and 0.25× PBS. When diluting 1× PBS to 0.25× PBS, the buffering capacity was reduced by ∼3.5-fold (Supporting Information, Figure S2), and pH values changed from 7.4 to 7.2. Figure E shows ΔpH dependence on urea levels for the two PBS solutions. As shown, measured ΔpH increases at lower urea levels and saturates to a constant value of 1.25 at high levels. This indicates that saturation behavior is independent of ionic strength and sample composition. Furthermore, this saturation behavior has been reported in other literature studies, though its cause is not discussed in detail. ,−

To gain insights into its cause, we plot post-reaction steady-state pH values for data shown in Figure D,E; pH = ΔpH (t = 300 s) + initial pH. Figure F shows the steady-state pH dependence on urea concentration for both artificial saliva and PBS samples. The pH values from different samples converge into a single curve (within experimental error) with the pH saturation occurring at pH ∼8.5 for urea levels >5.8 mM. Hence, this observed convergence is independent of sample composition, initial buffering capacity and pH. The saturation behavior observed in Figure F is consistent with prior studies by Ray et al. on urea hydrolysis in urine, where pH saturates at ∼8.5–9. This behavior can be understood as follows. Urea hydrolysis produces NH₃ and CO₂ (eq ), causing OH^– concentration ([OH^–]) to increase with concomitant pH increase. eq and enzyme activity are both pH dependent, thus causing the net OH^– increase (Δ­[OH^–]) to start decreasing for pH > 8. Since ΔpH = log­(1 + Δ­[OH^–]/[OH^–]), decreasing Δ­[OH^–] and increasing [OH^–] would cause the ΔpH value to become increasingly small as the test solution pH increases. As the solution pH ∼ 8.5, ΔpH becomes negligible, thus causing the pH value to saturate as observed in Figure F.

3.4. Diluent Selection and Optimization

To enable quantitative measurement across the full physiological and CKD-relevant concentration range, samples were required to be diluted into a medium with sufficiently low buffering capacity to produce a measurable postreaction ΔpH. The diluent must satisfy two key requirements: (i) minimal inhibition of urease activity, and (ii) its initial pH has to be well below the enzymatic reaction saturation pH value of 8.5, as shown in Figure F. MES was selected as the diluent since it has not been shown to inhibit urease activity, whereas other commonly used buffers such as PBS and borate buffer have been reported to inhibit urease activity. , Additionally, the initial pH of 4 mM MES is 5.88. The provides environment that is sufficiently above the pH range where urease activity has been reported to be compromised (∼pH 4.5–5) , while remaining well below the reaction saturation ceiling of ∼pH 8.5.

To ensure that the diluent does not suppress the pH increase due to urea hydrolysis, MES buffer capacity must be reduced by adjusting the buffer concentration. Figure A shows titrations of MES solutions ranging from 2 to 100 mM with incremental additions of OH^–. Lower concentrations (2 and 4 mM) exhibit steeper pH changes upon titration, whereas higher concentrations (10–100 mM) show strong buffering and limited pH changes. Although 2 mM MES shows a steeper pH rise upon titration, it reaches the saturation pH of 8.5 with less added OH^–. In contrast, 4 mM MES provides weak buffering while maintaining below saturation pH over a broader added OH^– range. Hence, 4 mM MES was chosen as the diluent.

4.

Diluent selection by optimizing MES buffer concentration and dilution factor. (A) Diluent concentration selection: titration curves for MES buffer solutions of different concentrations (2, 4, 10, 20, and 100 mM); 4 mM MES was selected as the diluent. (B) Dependence of pH change (ΔpH) on the dilution factor for three different artificial saliva (AS) samples in the absence of enzymatic reaction; AS diluted in 4 mM MES. (C) Comparison of urea calibration curves for AS samples with two different dilution factors (60× and 100×).

After selecting 4 mM MES as the diluent, the dilution factor must also be optimized to meet two criteria: (i) saliva addition to 4 mM MES (with 0 mM urea and 5 U urease) should minimally change the pH of the mixture, and (ii) the resulting diluted saliva solution should generate measurable ΔpH responses across both the healthy and CKD salivary urea ranges without reaching the saturation regime. To evaluate these criteria, artificial saliva samples with different initial pH were diluted by 5×, 10×, 60×, and 100× into 4 mM MES, and the resulting pH changes were measured. Figure B shows measured pH changes as a function of different dilution factors. At low dilution factors (5× and 10×), substantial ΔpH was observed upon mixing artificial saliva (AS) and 4 mM MES. This suggests that ΔpH remains strongly dependent on the original saliva pH, indicating that the intrinsic pH and buffering capacity of the saliva were not sufficiently suppressed. At higher dilutions (60× and 100×), this dependence is greatly reduced, demonstrating that dilution effectively minimizes the impact of saliva. This ensures that subsequent ΔpH changes are dominated by urea–urease reactions rather than the intrinsic differences in undiluted saliva characteristics. Additionally, the minimum required dilution factor was also constrained by the saturation behavior observed in undiluted saliva. As shown in Figure D, the ΔpH response begins to saturate between 1.9 and 5.8 mM urea. Therefore, to quantify salivary urea concentrations up to 45.8 mM while keeping the effective urea concentration below approximately 1.9 mM, a dilution factor greater than approximately 24× is required. This indicates that low dilution factors such as 5× and 10× are insufficient for high-urea quantification, while higher dilution factors are needed to avoid saturation.

Moreover, the dilution factor must keep the reaction-induced pH change within the near-linear working region of the 4 mM MES diluent. As shown by the MES titration curve in Figure A, the pH response becomes increasingly nonlinear once the pH rises beyond approximately 6.5. Therefore, limiting the reaction-induced pH increase to approximately 0.5–0.6 pH units helps maintain a more linear relationship between urea concentration and measured ΔpH. Among the high-dilution conditions tested, both 60× and 100× show reduced saliva-induced pH change and avoid the saturation observed in undiluted saliva (Figure D). However, the curve for 60× dilution (filled symbols, Figure C) has a steeper slope than that for 100× dilution (open symbols), thus indicating a higher urea measurement sensitivity. Based on the above results, artificial saliva samples would be diluted by 60× in 4 mM MES (pH 5.88) for urea concentration measurements, as discussed in the next section.

3.5. Optimized Urea Concentration Measurements

In this section, urea concentrations were measured in artificial saliva (AS) using the optimized dilution conditions of Figure . In the following experiments, the starting solution is 4 mM MES with 5 U of urease. To initiate the enzyme reaction, 20 μL saliva (pH 5.62, 6.36, 7.26) spiked with different urea concentrations (0–45.8 mM) was added to the starting solution, resulting in 60× dilution; the sample addition and mixing step took ∼22 s.

Figure A shows the measured ΔpH­(t) dependence on time for 0–45.8 mM urea spiked into different diluted artificial saliva (AS) samples. As shown, sensor response curves depend on urea concentrations. Curves corresponding to AS with 0 mM urea (control) resulted in a negative ΔpH shift. This negative shift in the absence of urea indicates a net release of protons that lowers the pH due to the re-equilibrium of buffer species in the mixture. Curves for 1.9 mM of urea show similar behavior but a clear deviation from the 0 mM baseline. This indicates that the production of OH^– from urea hydrolysis partially counteracts the background negative ΔpH at 0 mM. This confirms the BJT sensor ability to distinguish clinically relevant low levels of urea from the blank. For higher urea levels (5.8 to 45.8 mM), ΔpH­(t) rapidly increases to peak values and then gradually decreases to steady values after urease additions. These observed peaks occur due to hydrolysis rates being dominant over MES buffering. In other words, urea is rapidly converted to generate net OH^– production. Since the rate of OH^– production exceeds the rate of MES buffering, a transient maximum ΔpH is observed. As the reactions progress with decreasing unreacted urea, MES buffer starts to neutralize the pH change with ΔpH­(t) decreasing toward steady-state values.

5.

Urea concentration measurement in artificial saliva (AS) using the optimized diluent (60× dilution in 4 mM MES, 5 U urease); urea concentration values refer to those in undiluted AS. (A) Measured ΔpH­(t) dependence on time for diluted artificial saliva samples with three different initial pH values of 5.62, 6.36, and 7.26 with varying urea concentrations. (B) Dependence of ΔpH on urea concentration in three different AS samples; symbols denote data extracted from (A), and the solid line is the linear fit to all the data from different AS samples (with R ² = 0.999), thus providing a common calibration equation; the yellow shaded region highlights the CKD range.

Using the data from Figure A, the calibration curve was constructed as shown in Figure B. For high urea concentrations (>5.8 mM), peak ΔpH values were used as they represent the maximum reaction-driven pH change prior to buffer re-equilibration. For low concentrations (0 and 1.93 mM), steady-state ΔpH values were used since no distinct peak was observed. Figure B shows the dependence of ΔpH on urea concentration in three different diluted AS samples; symbols are measurements, and the solid line is a linear fit to the data from all the samples. Hence, the common ΔpH-urea calibration curve enables the measurement of urea levels from healthy to CKD range in saliva samples with different initial properties. For a future point-of-care implementation, this workflow could be realized by using a disposable cartridge with MES buffer and urease preloaded. Since the protocol effectively suppresses intersample variability in salivary pH and buffering capacity, the sensor can be used without sample-specific recalibration. The user would only need to add a saliva sample to the cartridge inlet; the device then automatically meters the correct volume, mixes it with the preloaded reagents, and produces a sensor readout. This could minimize the need for sample preparation, calibration, and mixing.

3.6. Validation against Gold Standard Assay

To validate urea concentration measurements obtained with the BJT pH sensor, the results were compared with those from a commercial gold-standard blood urea nitrogen (BUN) assay. The BUN assay quantifies urea nitrogen using a colorimetric method. The measured BUN values were converted to urea concentrations (C _urea) using the equation below, enabling direct comparison with the BJT pH sensor results.

$$C_{\mathrm{urea}}(\mathrm{mM})=\frac{60.06}{28}\times \frac{10}{60.06}\times \mathrm{BUN}(\mathrm{mgd}{\mathrm{L}}^{-1})$$ 5

where molecular weight of urea is 60.06 g/mol and urea nitrogen is 28 g/mol.

Figure A compares urea concentrations measured by the BJT pH sensor with those obtained using the gold-standard (GS) colorimetric method for three diluted saliva samples with different initial pH (5.62, 6.36, 7.26) and buffering capacities (Figure B). The saliva samples were diluted using optimized conditions (60× dilution in 4 mM MES). As shown in Figure A, the data points fall close to the identity line (x = y); the root-mean-square error (RMSE) relative to the identity line is 3.61, 1.77, and 2.07 mM for AS with initial pH of 5.62, 6.36, and 7.26, respectively. Hence, this comparison indicates a strong quantitative agreement between the BJT sensor and the GS colorimetric methods across a salivary urea concentration range of 0–45.8 mM.

6.

Verification of BJT sensor urea concentration measurements versus the gold standard (GS) urea nitrogen reference colorimetric assay method. (A) Comparison of BJT sensor urea concentration measurements with those measured by the gold standard (GS) assay for artificial saliva samples with three different initial pH values. Each symbol denotes measured data, and the dashed line is x = y. The root-mean-square error (RMSE) relative to the identity line is 3.61, 1.77, and 2.07 mM for AS pH 5.62, 6.36, and 7.26, respectively. (B) Bland-Altman analysis: C _urea,GS and C _urea,BJT are the urea concentrations measured by the gold standard assay and BJT sensor, respectively, and ΔC is the urea concentration difference obtained from the two methods. The blue solid line represents the Mean Difference in urea concentration from the two methods, and the green dashed lines are the limits of agreement (Mean Difference ± 1.96 Standard Deviations); the green shaded region, bounded by the limits of agreement, is the range where 95% of the ΔC are expected to lie.

To further quantify the agreement between BJT sensor and gold standard (GS) colorimetric methods, Bland-Altman analysis was applied as shown in Figure B. In this figure, the difference between urea concentrations (ΔC) obtained from the two methods was plotted versus urea concentration measured by the GS method; symbols represent measured data. The Mean Difference was estimated by taking the average of all ΔC: Mean Difference = −0.858 mM (solid blue line) with standard deviation (SD) = 2.464 mM. The analysis defines the upper and lower limits of agreement as Mean Difference ± 1.96 × SD. In the figure, these two limits (dashed lines) define the limits of agreement (LoA) (green region) where 95% of data points are expected to fall. We observe all but one data point fall within the LoA, consistent with statistical expectation for a data set of 18 measurement pairs, where approximately 18 × 0.95 ≈ 17 points are expected to fall inside the 95% LoA by definition. The single outlier corresponds to an AS pH 5.62 sample with a positive difference of approximately +5.5 mM. Both measurement methods carry independent sources of uncertainty that may contribute to this deviation. The BJT sensor measurement precision of ±0.009 pH units (Figure H), propagated through the calibration sensitivity of 0.008 ΔpH/mM (Figure B), corresponds to ±1.1 mM urea. The gold-standard assay uses samples that were separately diluted 13×, placing the highest tested concentration (45.8 mM) at 3.52 mM after dilution, which is near the upper limit of the kit’s validated range (3.57 mM); thus, small pipetting errors can cause under-reading. The magnitude of the outlier deviation (+5.5 mM) substantially exceeds the BJT’s ±1.1 mM precision, indicating that the deviation is primarily attributable to variability in the sample dilutions or gold-standard measurement rather than sensor error. Importantly, the differences across all data points are randomly distributed about the mean difference (−0.858) with no proportional or concentration-dependent trend, confirming the absence of systematic bias across the measurement range. The single outlier therefore reflects random measurement variability inherent to both methods and does not affect the overall accuracy conclusion.

In summary, the data points are randomly distributed about the mean difference (−0.858 mM) with no observable trend as a function of urea concentration. Hence, there is no proportional or concentration-dependent bias across the measurement range, thus demonstrating the accuracy of the BJT sensor measurements for urea concentrations.

3.7. Comparison with Prior Work

Table compares the present study to prior urea sensing studies that employ enzymatic pH-based detection. As summarized in column 2, prior studies predominantly rely on FET-based pH sensors for urea concentration measurements. In contrast, this work utilizes a BJT-based pH sensor, which has been demonstrated to offer enhanced performance characteristics, making it particularly well-suited for portable diagnostic applications. Columns 3 and 4 compare sensor-related attributes that are critical for the successful commercialization of portable diagnostic products. Column 3 contrasts the operating voltage and power requirements of the urea sensors reported across various studies. The proposed BJT sensor operates at voltages below 1 V with power consumption under 10 nW. Although this power level is higher than the minimum value of 0.3 nW reported by Choi et al., the BJT sensor offers additional advantages in terms of scalability. Scalability, evaluated in column 4, requires that sensing device fabrication be compatible with low-cost, high-volume manufacturing. The BJT sensors with TiN as the sensing surface employed in this study are fully compatible with standard silicon processing, enabling cost-effective fabrication in commercial silicon foundries. In contrast, the ultralow-power FET sensors reported by Choi et al. employ silver as the extended-gate material; since silver is not permitted in silicon foundries, such sensors are not suitable for inexpensive, large-scale manufacturing. Lastly, the saliva dilution approach provides an additional advantage by reducing errors associated with nonspecific binding and enzyme biofouling from salivary biomolecules. Although dilution lowers the analyte concentration in the test sample, this is not a limitation for urea because of its relatively high salivary concentration. However, dilution may become a limiting factor for the detection of analytes present in significantly lower concentrations.

1. Comparison of Transistor-Based Urea Biosensors Employing Urease-Mediated pH Transduction .

1. refs (year)	2. urea sensor	3. sensor use voltage& power	4. inexpensive scalability	5. test medium	6. sample-invariant calibration	7. detection range (clinical range2–45 mM)	8. gold-standard validation
Mu et al. (2014)	Si nanoribbon FET	<|0.5| V	yes	1× PBS (pH 7.4)	no	0–10 mM	no
Ahmad et al. (2015)	ZnO nanorod FET	<|6| V, ∼60 μW	no	0.05 M PBS; serum and blood	no	0.01–30 mM	no
Melzer et al. (2016)	pH-sensitive CNT FET	<|0.8| V, ∼100 μW	no	0.1 mM, 20 mM PBS	no	0.1–10 mM, signal saturates >10 mM	no
Piccinini et al. (2017)	rGO FET (LbL urease)	<|0.6| V, ∼600 μW	no	0.1 mM HEPES (pH 6)	no	0.001–1 mM	no
Choi et al. (2023)	Si EGT (nanowire)	<|1|V, 0.0003 μW	no	1× PBS (pH 7.4)	no	0.724–12.6 mM signal saturates >13 mM	no
Yang et al. (2024)	AlGaN/GaN HEMT (microsieve)	1.5 V, ∼15,000 μW	no	0.1× PBS	no	1 fM–100 mM	no
Neyra Recky et al. (2024)	PEDOT–PAH OECT	0.6 V, ∼600 μW	no	10 mM KCl (pH 7.4); diluted urine	no	0.1–5 mM, signal saturates >5 mM	no
Yang et al. (2025)	Si EGFET	<|1.4| V; current not specified	yes	not mentioned	no	0.3–3.7 mM, signal saturates >6 mM	no
Koike et al. (2025)	extended-gate FET	only reported ΔV, sensing current 100 mA	yes	1× PBS (pH 7.0)	no	0.017–1.7 mM, signal saturates >1.7 mM	no
this work	BJT-TiN sensing surface	< |1| V, 0.001–0.01 μW	yes	4 mM MES (pH 5.88); AS (3 pH variants)	yes (single calibration across AS pH5.62/6.36/7.26)	0–45.8 mM	yes (vs optical BUN assay)

^aAbbreviations in Table : AS, artificial saliva; BJT, bipolar junction transistor, FET, field effect transistor; BUN, blood urea nitrogen; CKD, chronic kidney disease; CNT, carbon nanotube; EGFET, extended-gate field-effect transistor; EGT, electrolyte-gated transistor; FET, field-effect transistor; HEMT, high-electron-mobility transistor; LbL, layer-by-layer; MES, 2-(N-morpholino)­ethanesulfonic acid; OECT, organic electrochemical transistor; PBS, phosphate-buffered saline; PEDOT, poly­(3,4-ethylenedioxythiophene); rGO, reduced graphene oxide.

Columns 5–8 compare attributes related to urea concentration measurement methodologies. Column 5 shows that most prior studies were evaluated in buffer solutions at a single pH value. Furthermore, none of these studies examined the impact of initial pH or buffering capacity of the test sample on urea calibration behavior (Column 6). In the present study, we systematically investigated the dependence of the calibration curve on both initial pH and buffering capacity and demonstrated sample-invariant calibration across samples with varying pH and buffering capacities (Figure ). Column 7 compares the measurable urea concentration ranges reported in previous studies. Most prior work exhibits signal saturation at higher urea concentrations, limiting measurements within clinically relevant disease ranges. While a similar saturation behavior was initially observed in this study, it was investigated in detail in Section . As shown in Figure , saturation occurs when the pH of test samples increases to approximately pH 8.5, regardless of the initial test solution property. The insights gained from this analysis enabled mitigation of saturation-related limitations through optimizing measurement procedures. Finally, Column 8 highlights that the present study is the only work to include validation against an independent gold-standard reference method. In summary, Table demonstrates that this work broadens sensor applications and advances measurement methodology, representing a significant step toward enabling scalable, low-power portable urea diagnostic products.

4. Conclusion

This study resolves key challenges that have impeded the clinical translation of pH-based salivary urea sensing. Using BJT-based pH sensors with high sensitivity and negligible drift, we demonstrate robust, time-resolved urea measurements under saliva-relevant conditions. Artificial saliva was benchmarked against human saliva and shown to exhibit comparable pH and buffering behavior, ensuring clinical relevance. Systematic analysis across broad ranges of initial pH and buffering capacity elucidates the mechanism of pH signal saturation at elevated urea concentrations and enables optimization of urea measurement conditions. Under these conditions, a sample-invariant pH-urea calibration with a single linear slope is achieved across the full clinically relevant urea range (0–45.8 mM). This optimized sensing is achieved using a simple protocol requiring a 60× dilution of 20 μL of saliva in 4 mM MES buffer (pH 5.88). Furthermore, this dilution also reduces errors due to salivary biomolecules associated with nonspecific binding and enzyme biofouling. The BJT sensor measurements show quantitative agreement with a commercial gold-standard colorimetric assay, thus confirming measurement accuracy. In summary, this study provides a pathway toward scalable, miniaturized, and portable salivary urea sensors for CKD monitoring. Lastly, this approach can be extended to other biofluids such as urine and sweat, and to other biomarkers whose detection relies on reaction-driven pH changes.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c05076.

• pH sensitivity of the TiN sensing electrode (Figure S1); buffering capacity comparison of PBS and artificial saliva samples (Figure S2); base composition of artificial saliva solutions (Table S1) (PDF)

†.

11F, no.37, Ln. 148, Sec. 3, Binhai Road, Tamsui, New Taipei 251, Taiwan

K.Y.H.: conceptualization, investigation, validation, methodology, formal analysis, visualization, writingoriginal draft, and writingreview and editing. S.Z.: conceptualization, supervision, investigation, validation, methodology, formal analysis, visualization, writingoriginal draft, and writingreview and editing.

During the preparation of this work, the author(s) used generative AI to edit the English language and improve the text clarity of the manuscript. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

The authors declare no competing financial interest.