Interindividual- and blood-correlated sweat phenylalanine multimodal analytical bio-chips for tracking exercise metabolism

Amino acids (AAs) in sweat originate from endogenous loss of plasma AAs and from natural moisturizing factors (NMFs) on the skin surface. NMF contamination has challenged wearable sweat AA sensing, but prolonged exercise diminishes skin-leached AA contributions, particularly for non-NMF AAs, enabling sweat AA concentrations to reflect blood levels. Wearable AA biosensors remain limited in the number of AAs detected and often neglect sweat rate, impeding quantitative metabolic assessment and robust blood–sweat correlations. Phenylalanine (Phe), an essential AA and non-NMF species with low skin prevalence and favorable physicochemical properties for diffusion, is a promising sweat biomarker that may correlate with blood Phe. Clinically, Phe monitoring informs nutritional status and diseases such as phenylketonuria (PKU), exercise-related muscle protein metabolism, liver dysfunction in obesity, and infection severity. Simultaneous measurement of Phe concentration and sweat rate can help determine Phe sources during different exercise stages, elucidate partitioning mechanisms, and improve interindividual sweat–blood correlations by accounting for sweat dilution effects.

Prior portable Phe detection approaches often rely on affinity elements (antibodies, aptamers), which face challenges including cost, stability, washing, and potential bioreactivity. Non-affinity electrochemical sensors are attractive for wearables but Phe is non-electroactive on common electrodes (Au, carbon, graphene). Reported wearable methods include (i) derivatization to render Phe electroactive (one-time use, unsuitable for continuous sensing) and (ii) indirect detection with redox probes and molecularly imprinted polymers (MIPs), which suffer lower sensitivity and inverse nonlinear responses versus analyte concentration. Simultaneous measurement and combined analysis of sweat AA levels with sweat rates have been largely neglected due to lack of convenient sweat rate quantification. The present work addresses these gaps by enabling direct electrocatalytic Phe detection and concurrent sweat rate monitoring within a wearable multimodal platform.

Device architecture: A skin-mounted wearable multimodal biochip integrates (i) an electrochemical Phe sensor based on an electrocatalytically active molecularly imprinted polyaniline (PANI) layer (E-MIP) for direct Phe oxidation, (ii) a vertically assembled multipurpose microfluidic module for rapid sweat sampling, pH buffering, and visualized flow for sweat loss (volume/rate) quantification, and (iii) a wireless flexible circuit with smartphone app for DPV/OCP acquisition and data display.

E-MIP Phe sensor fabrication: Cr/Au (20/50 nm) electrodes were thermally evaporated on PET and patterned by laser-engraved masks. The working electrode was cleaned (piranha) and PANI-MIP was formed by electropolymerization in PBS (pH 7.0) containing 10 mM L-Phe and 10 mM aniline using cyclic voltammetry from −0.4 to 0.9 V vs SCE (4 cycles, 50 mV s⁻1). Template extraction and over-oxidation/electro-degradation were performed via multi-potential steps between 0.9 and −0.9 V (40 alternations, 100 s each), yielding an electro-degraded MIP (E-MIP) with selective, catalytically active sites. Non-imprinted controls (NIP/E-NIP) were prepared identically without Phe template. The chloride sensor used Ag/AgCl; a PVB/NaCl/F127/MWCNTs cocktail formed a stable reference electrode.

Electrochemical measurements: Differential pulse voltammetry (DPV) parameters: 0.4–0.7 V range, 1 mV step, 30 mV amplitude, 50 ms pulse width, 20 ms sampling width, 100 ms period, sensitivity 1×10⁻⁵ A V⁻¹. A 0.4 V pre-bias for 2 s enhanced Phe binding and mitigated interferents. CV, EIS, and selectivity studies were conducted in relevant media. Baseline correction was applied to DPV before peak extraction.

Microfluidics: A five-layer, laser-engraved vertical microfluidic stack (transparent PET with μ-dots for optical contrast; double-sided tapes forming inflow/outflow serpentine channels; dark PI isolation layer; skin adhesive) incorporated a chamber with embedded, laser-cut filter paper preloaded with phosphate buffer (pH 7.5–7.8). Design features: multiple elliptic inlets to increase sampling area; reduced chamber volume via filter paper for faster filling; pH/ionic strength buffering to stabilize electrochemistry; and a serpentine outflow channel with visual readout (1 μL per meander; ~0.5 μL resolution) leveraging reflectivity/transmittance differences at μ-dots for sweat loss quantification. COMSOL simulations (two-phase level set; porous media; mass transport) guided filling and concentration refreshing behavior.

Wireless electronics and app: A flexible PCB centered on an STM32L15C8T6 MCU generated DPV waveforms via 12-bit DAC, acquired currents via TIA to 12-bit ADC, and performed differential OCP with an AD8227 amplifier. Data were filtered/baseline-corrected on the smartphone app and converted to concentrations. Bluetooth module (E104-BT5005A) enabled telemetry. Power was supplied by a 3.7 V Li-ion polymer battery with LDO regulation (3.3 V analog/digital rails).

On-body studies: Human studies were IRB-approved with informed consent. Healthy males (23–27 years) were recruited and stratified by BMI (lean: 18.5–24.9; overweight: 25–30 kg m⁻²). For dynamic tests, subjects performed moderate-intensity cycling (and jogging in supplementary tests). Forehead skin was pre-cleaned; sensors were worn on forehead and forearm in some measurements. Every 10 min during exercise, DPV for Phe and OCP for chloride were acquired; sweat loss was imaged for algorithmic sweat rate estimation. Parallel sweat samples were collected from adjacent sites for ELISA and colorimetric AA assays. A pilot diet intervention (protein intake during exercise, followed by rest) was conducted on two representative subjects (one lean, one overweight) to track sweat and serum Phe (LC-MS) and evaluate sweat–serum correlations with and without sweat rate normalization.

Computational analysis: DFT (Gaussian 16, B3LYP-D3BJ) modeled Phe–polymer interactions and charge transfer under external fields to rationalize electrocatalytic oxidation on E-MIP vs controls.

• Sensor principle and performance: A PANI-based enzyme-mimicking MIP (E-MIP) enables direct, selective electrocatalytic oxidation of Phe with higher responses than E-NIP, PPY-MIP, and Au controls, consistent with DFT-predicted greater charge transfer (Δq=3.924e for E-MIP). DPV shows two linear regions: 10–300 μM with sensitivity 1.4 nA μM⁻¹ and LOD 4.7 μM; 300–1000 μM with lower slope (reported sensitivity 0.27 nA μM⁻²). The sensor is selective against abundant AAs (Gly, Ser, Ala, His, Tyr, Trp) and interferents (glucose, urea, lactate, ascorbic acid), exhibits L-Phe chiral selectivity, and maintains stable responses from pH 7.0–9.5 and across flow rates 0.5–2 μL min⁻¹.
• Microfluidic capabilities: Elliptic multi-inlets and embedded filter paper accelerate chamber filling (~8 min at ~2 μL min⁻¹) and homogenize concentration refreshing. Visual serpentine outflow channel resolves ~0.5–1 μL volume increments, enabling on-skin sweat rate quantification that agreed with syringe-pump benchmarks (0.5–2 μL min⁻¹). The embedded buffer maintained near-neutral pH for >40 min at 2 μL min⁻¹; Phe sensing remained stable (<10% variation) over the first ~80 μL of continuous sweat.
• Sweat rate–Phe concentration relationship: Among 16 healthy subjects at 20 min of exercise, sweat Phe concentration negatively correlated with sweat rate; correlation was weaker at low sweat rates due to persistent skin-derived Phe contributions.
• Metabolic risk assessment via Phe secretion rate: Defined sweat Phe secretion rate Sp= Cp × Rw (μmol min⁻¹ m⁻²) to reduce interindividual variability. Risk zones were proposed: low (<1), medium (1–2), high (>2 μmol min⁻¹ m⁻²). One subject exhibited notably high Sp, indicating elevated exercise-induced AA loss.
• BMI-related differences: In eight on-body tests (lean vs overweight), overweight (BMI≥25) subjects exhibited lower sweat Phe concentrations than lean (BMI<25), consistent with dilution at higher sweat rates; Wilcoxon rank-sum W=378, P<0.001.
• Relation of Phe to total AAs: Across subjects, sweat total AA levels correlated with Phe (Pearson r=0.862). In the intake–exercise pilot, sweat AAs vs Phe also showed strong correlation (r=0.883).
• Validation and serum correlation: Sensor sweat Phe matched ELISA (r=0.956). In two subjects, sweat–serum Phe showed strong correlations (r=0.878 and r=0.947). Normalizing sweat Phe by individual stabilized sweat rate (i.e., secretion-based normalization) yielded similar fitted slopes across subjects and strengthened interindividual sweat–serum correlations.
• Additional markers: Concurrent sweat chloride measurements trended upward during exercise, and sweat rate increased with workload, while Phe levels decreased over time as skin contributions waned and endogenous loss dominated.

The study addresses the challenge of translating sweat amino acid measurements into meaningful metabolic and blood-correlated insights by simultaneously quantifying sweat Phe concentration and sweat rate. The observed negative correlation between Phe concentration and sweat rate supports a diffusion-driven partitioning mechanism from interstitial fluid/blood, modulated by sweat dilution, especially once skin-derived Phe is depleted. Introducing the Phe secretion rate as a composite indicator reduces interindividual variability and enables quantitative estimation of exercise-induced AA loss and identification of individuals at higher metabolic risk who may benefit from targeted nutritional supplementation. The strong sweat–serum Phe correlations, further improved by sweat rate normalization, demonstrate that integrating sweat rate into analysis strengthens blood–sweat coupling and enhances the reliability of noninvasive monitoring. Together with robust on-body performance, selectivity, and microfluidic pH buffering, these findings validate the multimodal biochip as a practical tool for personalized exercise and dietary management, and as a platform for probing biomarker partitioning mechanisms.

This work presents a wearable multimodal biochip that concurrently measures sweat phenylalanine, chloride, and sweat rate, enabling direct electrocatalytic Phe sensing via an enzyme-mimicking PANI-based MIP and quantitative sweat loss readouts via vertically integrated microfluidics. Key contributions include: (i) demonstration of diffusion-dominated Phe partitioning with sweat dilution effects; (ii) introduction of a sweat Phe secretion rate to assess exercise-induced AA loss and metabolic risk with reduced interindividual variability; and (iii) establishment of strong sweat–serum Phe correlations that are further strengthened and unified across individuals by sweat rate normalization. Future work will expand subject cohorts with diverse metabolic phenotypes (e.g., PKU, obesity) to reinforce sweat–serum models, extend the approach to additional non-NMF amino acids, refine microfluidic buffering and sampling for low sweat rate conditions, and explore clinical translation for noninvasive, personalized diet (low-Phe) management.

The study acknowledges limited subject numbers for correlation analyses, with detailed sweat–serum comparisons conducted on two subjects, necessitating larger cohorts to generalize findings. Correlations between sweat Phe and sweat rate are weaker at low sweat rates due to residual skin-derived Phe contributions, which may confound early-stage measurements before depletion. The cohort comprised healthy young males, limiting generalizability across genders and ages. The stable pH-buffered sensing window was demonstrated over the initial ~80 μL of continuous sweat, and long-duration buffering under varying conditions may require further optimization.