A wearable aptamer nanobiosensor for non-invasive female hormone monitoring

Abstract

Personalized monitoring of female hormones (e.g., estradiol) is of great interest in fertility and women’s health. However, existing approaches usually require invasive blood draws and/or bulky analytical laboratory equipment, making them hard to implement at home. Here we report a skin-interfaced wearable aptamer nanobiosensor based on target-induced strand displacement for automatic and non-invasive monitoring of estradiol via in situ sweat analysis. The reagentless, amplification-free, and “signal-on” detection approach coupled with gold nanoparticles-MXene-based detection electrode offers extraordinary sensitivity with an ultra-low limit of detection of 0.14 pM. This fully-integrated system is capable of autonomous sweat induction at rest via iontophoresis, precise microfluidic sweat sampling controlled via capillary bursting valves, real-time estradiol analysis and calibration with simultaneously collected multivariate information (i.e., temperature, pH, and ionic strength), as well as signal processing and wireless communication with a user interface (e.g., smartphone). We validated the technology in human subjects. Our data suggests a cyclical fluctuation in sweat estradiol during menstrual cycles, and a high correlation between sweat and blood estradiol was identified. Our study opens up a potential for wearable sensors for non-invasive, personalised reproductive hormone monitoring.

Female hormones impact much of women’s health, from menstruation to pregnancy to menopause and more (e.g., depression)^1,2. They are regulated by the feedback mechanism of the hypothalamic-pituitary-ovarian (HPO) axis, which governs all the female hormonal events related to reproductive activity. Estradiol is the primary form of the female hormone estrogen, and the most potent and prevalent female hormone during reproductive years. During the menstrual cycle, estradiol provides feedback to the HPO axis to regulate the production of luteinizing hormone (LH) and ovulation (Fig. 1a)³. In addition to its centrality in sexual development, it significantly affects major organs, including the blood vessels, bones, muscles, and brain for both males and females^4–7. Therefore, estradiol monitoring is important in human biology from the cradle to the grave and is an essential component of infertility management as well as general physiological surveillance^8,9. Compared to other alternative approaches for fertility monitoring based on the basal thermometer or urine LH, serum estradiol provides the most timely and accurate information¹⁰. Moreover, serum estradiol analysis is a gold standard for monitoring the induction of ovulation with gonadotrophins¹¹; routine estradiol monitoring plays a crucial role for women in menopause and individuals undergoing hormone replacement therapy^12,13. Despite the widespread importance of female hormones, our ability to measure them properly has not kept pace with their growing importance in clinical medicine^8,14.

Figure 1. A wearable nanobiosensor based on strand-displacement aptamer switch for non-invasive reagentless female reproductive hormone analysis.

a, Female hormones play an important role in women’s health; non-invasive monitoring of estradiol can be realized through sweat analysis using a skin-interfaced wearable sensor. GnRH, gonadotropin-releasing hormone; LH, luteinizing hormone; +, stimulatory effect; –, inhibitory effect. b, The reagentless in situ quantification of estradiol using a AuNPs-MXene sensor coupled with a target-induced strand displacement aptamer switch. WE, working electrode; CE, counter electrode; RE, reference electrode; aptamer-ssDNA, aptamer single-stranded deoxyribonucleic acid; SH-ssDNA, thiolated ssDNA; MB-ssDNA, methylene blue-tagged ssDNA. c,d, Schematic (c) and layered design (d) of the flexible and wireless microfluidic wearable patch for automatic sweat induction via iontophoresis, precise sampling via capillary bursting valves (CBVs), and reagentless aptamer-based estradiol analysis. I to VII represent plastic substrate interfacing with the flexible printed circuit board (FPCB) (I), inkjet-printed sensor array (II), carbachol-loaded hydrogels (carbagels) (III), microfluidic channels (IV), biorecognition interface (V), sweat accumulation layer (VI), and the skin (VII), respectively. e,f, Photos of a disposable sensor patch (e) and a fully-integrated wireless wearable patch worn on a finger (f) for female hormone monitoring. Scale bars, 1 cm.

Quantitative measurements of female hormones have been obtained by mass-spectrometry and immunoassays which require bulky equipment and have low detection throughput for routine clinical use¹⁵. Quantifying pM-level estradiol in the menstruation cycle appears to be a crucial and challenging issue for infertility identification (Fig. 1a)^16–18. Moreover, these approaches are typically based on invasive blood draws and complex sample preparation, making them impractical for at-home remote use¹⁹. Although there are commercially available “at-home” estrogen test kits based on finger-prick blood drop, saliva, or 24-hour urine, they all require the collected samples to be sent out for lengthy laboratory analysis. Highly sensitive and accurate on-site quantification of estradiol levels in non-invasively accessible biological fluids is highly desired to ensure optimal medical decision-making at the individual level²⁰.

Human sweat contains rich molecular information on an individual’s health state^21–25. The transition from traditional laboratory analysis to wearable sweat analysis could enable non-invasive and remote female hormone monitoring^26–30. In situ monitoring of sweat female hormones is challenging due to the extremely low concentration (pM level) and variations in sweat accessibility and sample matrix³⁰, and thus has yet to be realized.

In this work, we introduce a wireless wearable sensor based on a target-induced strand-displacement aptamer switch for in situ automatic electrochemical monitoring of female hormones with sub-picomolar sensitivity (Fig. 1b and Supplementary Fig. 1). The flexible sensor contains a biorecognition interface modified with an estradiol-selective deoxyribonucleic acid (DNA) aptamer facing a gold nanoparticles-MXene (AuNPs-MXene)-based detection working electrode (WE) modified with capture single-stranded DNA (ssDNA) for indirect target detection. These parallel faces function as walls of the microfluidic reservoir and are bridged upon sweat filling. When the sensor is exposed to body fluids such as sweat, estradiol molecules competitively bind with the estradiol aptamer, resulting in the release of the redox probe methylene blue (MB)-tagged ssDNA (MB-ssDNA) molecules from the biorecognition interface, which are then captured by the opposite detection WE through complementary DNA hybridization. Sensitive and selective estradiol quantification is realized through the detection of the MB redox signal on the AuNPs-MXene WE (Fig. 1b and Supplementary Fig. 2). The use of AuNPs-MXene improves the electron transport efficiency for the MB reduction, enabling the assay analysis with ultra-high sensitivity. Importantly, this nanoengineered sensor patch is mass-producible at low cost through inkjet printing with custom-developed nanomaterial inks (Supplementary Fig. 3, Supplementary Tables 1 and 2). Ti₃C₂T_x MXene, a two-dimensional (2D) transition metal carbide, was chosen owing to its highest conductivity among solution-processed 2D materials and the surface hydrophilicity for ink formulation and antifouling^31,32.

A fully-integrated wearable system was developed to realize automatic sweat estradiol sensing. This system contains a microfluidic iontophoresis module for autonomous sweat induction and sampling at rest (without the need for vigorous exercise), a potentiometric pH sensor, a resistive skin temperature sensor, and an impedimetric ionic strength sensor for real-time sensor calibration (to mitigate the influence of interpersonal/intrapersonal variations), as well as electronic circuitry for signal processing and wireless communication (Fig. 1c,d and Supplementary Fig. 4). The miniaturized wearable system was designed to be mechanically flexible and conform to the human skin (Fig. 1e,f and Supplementary Fig. 5). We validated the sensor against gold-standard measurements and in human subjects for personalized female hormone analysis during menstrual cycles.

Results and discussion

Design and characterization of the estradiol nanobiosensor

The reagentless estradiol sensor was designed based on a strand-displacement aptamer switch. At the biorecognition interface, the estradiol aptamer-ssDNA is immobilized on a AuNPs-decorated surface, forming a partially hybridized duplex with an MB-ssDNA probe molecule, which acts as a competitive redox probe (Fig. 2a and Supplementary Fig. 6). Considering that a short MB-ssDNA with weak but selective affinity reaction to the estradiol aptamer is desired for efficient target-induced strand displacement reaction, a 25-base MB-ssDNA was selected based on secondary structure analysis of the estradiol aptamer^33,34: substantial hybridization reaction was obtained from the 25-base MB-ssDNA but not from the molecules with 24 or fewer bases (Supplementary Fig. 7). On the detection WE, a capture thiolated ssDNA (SH-ssDNA) molecule, the antisense strand of the MB-ssDNA probe molecule, is immobilized on a AuNPs-MXene electrode (Supplementary Fig. 8). In the presence of estradiol, the MB-ssDNA is released by the aptamer-ssDNA owing to the higher affinity of the aptamer to estradiol than the partially hybridized sequence³⁵, and then anneals to the complementary sequence at the WE (Fig. 2a). The recaptured MB-ssDNA probe molecules can be quantified by the redox signal measured electrochemically via square wave voltammogram (SWV). The reagentless “signal-on” detection approach, coupled with highly sensitive low-background SWV measurements, offers extraordinary sensitivity and applicability for ultra-low-level sweat estradiol analysis in situ³⁶.

Figure 2. Design and characterization of the reagentless amplification-free aptamer estradiol sensor.

a, Schematic of the reagentless electrochemical aptamer sensor based on competitive redox probe displacement and recapture. b, Scanning electron microscopy (SEM) image of the inkjet-printed AuNPs-MXene electrode. The images are representative of three independent experiments. Scale bar, 200 nm. c, Transmission electron microscopy (TEM) characterization of the AuNP-MXene. The images are representative of three independent experiments. Scale bar, 5 nm. d,e, Differential pulse voltammogram (DPV) (d) and open circuit potential-electrochemical impedance spectroscopy (OCP-EIS) (e) characterizations of the WE in 0.1X PBS (pH 7.4) containing 2.0 mM K₄Fe(CN)₆/K₃Fe(CN)₆ (1:1) after each surface-modification step: AuNPs, AuNPs-MXene, SH-ssDNA, and 6-mercapto-1-hexanol (MCH). f, Square wave voltammogram (SWV) response of a modified AuNPs-MXene-based sensor before and after incubation with 50 pM estradiol. g, Comparison of the detection performance of WE substrates including Au, AuNPs, AuNPs-MXene, and AuNPs-MXene-mercaptopropionic acid (MPA). Error bars represent the s.d. of the mean from three sensors. h,i, SWV response of the aptamer estradiol sensors in artificial sweat (0.2X PBS, pH 7.4) with 0, 0.1, 0.5, 1, 5, 10, 50, 100 pM estradiol (h) and the corresponding calibration plot based on the peak current height of the SWV voltammograms (i). Each SWV voltammogram was obtained from an independent estradiol sensor (total 8 sensors) in h. E2, estradiol. Error bars represent the s.d. of the mean from three sensors. j, Selectivity of the aptamer sensor to potential interferences (50 pM) in human sweat. Error bars represent the s.d. of the mean from three sensors. k, Validation of the aptamer sensor toward estradiol quantification in iontophoresis-induced human sweat samples with enzyme-linked immunosorbent assay (ELISA) (n=25). Dashed line represents the linear-fit trendline. r, linear correlation coefficient. l, Calibration plots of the aptamer sensors with 10-minute incubation measured with and without the assistance of an external electric field. Error bars represent the s.d. of the mean from three sensors. m, Experimental and simulation results of the sensor response with and without the assistance of an external electric field after 10-minute incubation. Inset, simulated molecular diffusion and recapture of MB-ssDNA around the WE. Exp., experimental result; Sim., simulation result; E, electric field; BI, biorecognition interface. Error bars represent the s.d. of the mean from three sensors. n,o, The SWV voltammograms (n) and the corresponding peak current height (o) showing the regeneration and repetitive use of the aptamer estradiol sensors with 50 pM estradiol. Error bars represent the s.d. of the mean from three sensors.

The biorecognition interface and the detection WE were prepared by scalable inkjet printing of AuNPs (~22 nm) which offered a high electrochemical active surface area (ECSA) for subsequent nucleic acid modifications (Fig. 2b, Supplementary Figs. 9 and 10). On the WE, a thin layer of MXene, was inkjet-printed on the surface of AuNPs to further improve the sensor performance (Fig. 2b,c and Supplementary Fig. 11). The MXene nanosheets used in the ink were largely single- or few-layered (Supplementary Fig. 12), and the sheet size was smaller than 0.45 µm (Supplementary Fig. 13). SEM characterizations and experimentally obtained double layer capacitance (linearly correlated with ECSA) confirmed higher ECSA of the AuNPs-MXene electrode than those of evaporated Au and printed AuNPs electrodes (Supplementary Fig. 10). Compared to traditional screen printing, inkjet printing has multiple advantages here including lower cost owing to less ink consumption, lower risk for sample contamination, more precise spatial and thickness control of the printed nanomaterials, and higher electroactive surface area owing to the porous structures to realize optimal estradiol sensing (Supplementary Figs. 14 and 15)³⁷. Differential pulse voltammogram (DPV) and open circuit potential-electrochemical impedance spectroscopy (OCP-EIS) were used to further characterize the AuNPs-MXene electrode surface after each modification step (Fig. 2d,e). The increased peak current height in DPV voltammograms and the decreased resistance in Nyquist plots after MXene printing indicate the increased electrode conductivity and electrochemical catalytic activities. The following increased resistance in Nyquist plots and decreased DPV peak current height and after SH-ssDNA and 6-mercapto-1-hexanol (MCH) modification indicate that the self-assembled monolayers and the SH-ssDNA molecules impede the charge transfer at the electrode/solution interface due to the increased surface coverage by non-conductive molecules. It should be noted that despite MXene’s high conductivity, the bare inkjet-printed MXene electrode (without AuNPs support) showed large impedance due to the lack of percolating pathway in the ultrathin film of MXene flakes. After incubation with 50 pM estradiol, a clear MB redox peak appeared at ~-0.28 V, demonstrating the successful capture of MB-ssDNA released from the biorecognition interface due to estradiol-aptamer interaction (Fig. 2f). The incubation time, reagent concentration, and measurement frequency were further optimized to reach the optimal sensor performance with high sensitivity and repeatability (Supplementary Figs. 16 and 17, Supplementary Note 1).

Our study shows that both AuNPs and MXene play a crucial role in sensor performance (Fig. 2g and Supplementary Fig. 18). Compared to the plain Au electrode, the AuNPs coating increased the effective electrode area, thus increasing the SH-ssDNA loading capacity for improved sensitivity. A further 40% signal enhancement was obtained with the additional MXene layer. This can be explained by MXene’s ultrahigh electrical conductivity³⁸, which facilitates the electrode conductivity and the efficiency of charge transfer and transport for the reduction of MB molecules. The importance of MXene was further validated by introducing 2-mercaptopropionic acid (MPA) to block the surface of MXene via the esterification interactions between the −COOH groups in the MPA and the −OH groups in the MXene. The resultant AuNPs-MXene-MPA electrode showed a substantially decreased signal, indicating the crucial role of the MXene layer for highly sensitive estradiol analysis (Fig. 2g). It should be noted that the possible MB reduction caused by MXene doesn’t affect the final SWV estradiol quantification as the initial phase of the negative SWV scan will transform MB to the oxidized state; highly stable SWV measurements were observed on MB-ssDNA/SH-ssDNA modified AuNPs-MXene electrodes over 5000 repetitive scans and after 0–120-minute solution conditioning (Supplementary Figs. 19 and 20). The estradiol sensors also displayed high operation stability in solution and high dry-state long-term storage stability under cold and room temperature conditions (Supplementary Figs. 21 and 22). The sensors from five different batches displayed high repeatability with a low coefficient of variation of 0.013 (Supplementary Fig. 22a).

The sensor performance was evaluated by SWV in artificial sweat (0.2X PBS, pH 7.4) containing physiologically relevant estradiol levels (0.1–100 pM) (Fig. 2h). The sensor exhibited a log-linear relationship between peak current density height of the SWV voltammograms and target concentrations with an ultralow limit of detection of 0.14 pM (Fig. 2i). The estradiol sensor demonstrated high selectivity to estradiol over a variety of potential interferences present at much higher concentrations (Fig. 2j). The accuracy of the aptamer sensors for sweat estradiol analysis was validated by the enzyme-linked immunosorbent assay (ELISA) using iontophoresis-induced sweat samples (Fig. 2k), and a high linear correlation coefficient of 0.921 between the ELISA and biosensor results was obtained.

Considering that the optimal incubation time for the competitive interaction and recapture is over 60 min, we explored the use of an external electric field (a positive potential bias between the working and counter electrodes prior to incubation) to reduce the incubation periods for rapid estradiol analysis. Since ssDNA is negatively charged, applying a positive potential at the WE results in electrophoresis-based enhanced transport of the released MB-ssDNA redox probe across the sensing gap and substantially reduces the necessary incubation time (Supplementary Fig. 23a,b). As a result, electrochemical measurements reveal a substantially enhanced sensor signal after a 10-minute incubation period with the application of a bias-potential (1 min duration at +0.5 V prior to incubation) (Fig. 2l and Supplementary Fig. 24), in agreement with the numerical simulation results (Fig. 2m, Supplementary Fig. 23c,d, and Supplementary Video 1). The applied bias potential leads to ~300 nA of current flowing through the electrodes (Supplementary Fig. 25). Thus, the power dissipated due to the low current flow is at the nW level. It should be noted that the hybridized MB-ssDNA dominates the signal of the final measurement while electrostatic adsorption or chelation of MB-ssDNA onto the positively charged detection WE only contributes to a small portion of the final measurement result (Supplementary Fig. 26). Considering that all free MB-ssDNA molecules are resulted from the selective estradiol-induced strand displacement reaction, such electrostatic adsorption only contributes positively to sensitivity without compromising sensor selectivity. The estradiol sensor can be readily regenerated in deionized water (Fig. 2n) or in acidic conditions (Supplementary Fig. 27) after 1-minute incubation to perform repetitive estradiol quantification (Fig. 2o). Successful estradiol sensing and sensor regeneration showed only 6.2% signal drift after 5 cycles of repetitive measurement/regeneration (Fig. 2o and Supplementary Fig. 28).

Automatic microfluidic estradiol analysis

For automatic sweat estradiol analysis in situ, we designed a disposable sensor patch consisting of a pair of carbachol gel-loaded inkjet-printed carbon electrodes for iontophoresis-based autonomous sweat induction³⁹, a microfluidic module coupled with one collection reservoir and two capillary bursting valves (CBVs) to precisely control the sweat sampling, and a multiplexed inkjet-printed sensor array in the microfluidic sensing reservoir for estradiol quantification and calibration (Fig. 3a). The induced sweat will enter the microfluidic channel via the inlet, travel into and fill the sensing reservoir (with a sample volume of ~6.66 µl); when the sensing reservoir is filled, estradiol analysis will initiate while newly secreted sweat will be rerouted to the outlet due to capillary effects of the CBVs (Fig. 3b,c and Supplementary Fig. 29). Such microfluidic valving design allows highly stable biosensing in the quiescent sweat matrix free from the influence of sweat flow. Compared to the bulk solution analysis, no apparent decrease in the sensitivity and selectivity of the estradiol sensors was observed when operating in the microfluidics (Fig. 3d,e and Supplementary Figs. 30 and 31).

Figure 3. Microfluidic wearable system integration for automatic in situ hormone analysis.

a, Schematic illustration of multiplexed microfluidic sensor patch for automatic reproductive hormone analysis. IMP, electrode for impedimetric ionic strength sensor. b,c, Schematic illustrations (b) and optical images (c) showing phases of CBV-modulated in situ sweat sampling process towards automatic estradiol analysis: sweat entry via inlet (I), sweat collection in the reservoir (II), estradiol recognition and analysis (III), and sweat outflow via outlet (IV). Scale bar, 2.5 mm. d,e, The SWV voltammograms of the aptamer estradiol sensors (d), and the corresponding calibration plot based on the peak current height (e) in a laboratory microfluidic flow test using artificial sweat (0.2X PBS, pH 7.4) containing 0, 1, 5, 10, 20, 40, 60, 80, 100 pM estradiol. Error bars represent the s.d. of the mean from three sensors. f, Potentiometric response and corresponding calibration plot (inset) of the pH sensors in artificial sweat (0.2X PBS). g, Impedimetric response and the corresponding calibration plot (inset) of ionic strength sensors in artificial sweat (0.2X PBS, pH 7.4). h, Block diagram of wearable electronic system. IP, iontophoresis; GPIO, general-purpose input/output; SPI, serial peripheral interface; BLE, Bluetooth Low Energy; AFE, analog front-end; Soc, system-on-chip. i, Optical images of a fully-integrated wearable system with high mechanical flexibility. Scale bar, 1 cm. j, performance of the aptamer sensor patch during the bending tests for both x- and y-axis. I and I₀ represent the peak current height of SWV voltammograms under and in the absence of mechanical deformation, respectively. R, radius of bending curvature. The error bars represent the s.d. of the mean from 10 measurements in 50 pM estradiol. k, Time-lapse images of the microfluidic sampling of the iontophoresis-induced sweat using a finger-worn wearable sensor for automatic hormone analysis. Scale bar, 5 mm.

Considering that the large interindividual variability in sweat compositions (i.e., pH and ionic strength) could have a major influence on the target recognition and electrochemical measurement, a polyaniline-based potentiometric pH sensor and an impedimetric ionic strength sensor were developed and integrated into the sensor patch. In vitro sensor evaluation revealed linear relationships between the measured potential and pH for the pH sensor, and between measured admittance and electrolyte levels for the ionic strength sensor (Fig. 3f,g). The accuracy of the pH sensor for sweat sample analysis was validated with a pH meter (Supplementary Fig. 32).

Wireless wearable system integration and characterization

The fully-integrated wearable estradiol sensing system was developed based on the assembly of the disposable sensor patch and a reusable flexible printed circuit board (FPCB) (Fig. 3h and Supplementary Fig. 33). The FPCB contains a high compliance-voltage current source for iontophoresis to extract sweat on-demand and a programmable Bluetooth Low Energy (BLE) module that integrates a microcontroller for signal processing and wireless communication. To perform multiplexed sensing, this integrated system incorporates an electrochemical analog front-end (AFE) that was able to perform multimodal electrochemical measurements (including voltammetry, impedimetry, and potentiometry for estradiol, pH, and ionic strength sensing, respectively). In addition, it could also provide skin temperature information continuously via a built-in temperature sensor in the AFE. During in situ sweat analysis, the obtained pH, ionic strength, and skin temperature levels were used to calibrate the estradiol sensor readings in real-time (Supplementary Fig. 34). A full measurement cycle consumes ~1 J in total (Supplementary Fig. 35); a rechargeable 3.8 V lithium coin cell battery with a capacity of 8 mAh could perform approximately 30 cycles. By exploring efficient wearable energy harvesting system such as biofuel cells⁴⁰ and solar cells⁴¹, it is possible to realize a fully self-powered wearable system.

The wearable patch was also designed to have high mechanical flexibility that enables conformal contact with the skin (Fig. 3i). It was able to maintain stable sensor responses under different bending radii, making it promising for practical at-home on-body monitoring during daily activities (Fig. 3j). When attached to the finger as a ‘smart ring’, the wearable system allowed efficient sweat extraction and sampling toward real-time estradiol analysis (Fig. 3k, Supplementary Fig. 36, and Supplementary Video 2). Considering that the spot-checking of estradiol every day is sufficient to monitor menstrual status, a disposable estradiol sensor patch design coupled with a reusable wearable electronic system could fulfil the practical application needs of automatic, immediate, and non-invasive estradiol detection at any given time. When necessary, continuous wearable estradiol sensing could be realized through integrating a multi-reservoir CBVs design⁴² and estradiol sensor arrays into a single sensor patch.

In vivo evaluation of the wearable estradiol sensor

During a menstrual cycle, estradiol level in the blood rises and falls twice as illustrated in Fig. 4a: it gradually increases in the mid-follicular phase and reaches the highest point right before ovulation; then the estradiol level drops quickly after ovulation, followed by a secondary rise during the mid-luteal phase and a secondary decrease at the end of the menstrual cycle⁴³. Compared to invasive and lengthy blood assay, analyzing sweat estradiol using wearable technology offers a highly attractive approach for remote at-home female hormone monitoring (Fig. 4a). To validate the clinical values of the sweat estradiol measured by the sensor, human studies were conducted over two consecutive menstrual cycles on two healthy female subjects by simultaneously monitoring urine LH, body temperature, and blood estradiol (Fig. 4b,c). The results revealed that estradiol levels in both sweat, and serum reached the peak right before the ovulation while the urine LH and temperature levels peaked during the ovulation period, confirming estradiol’s potential for early ovulation prediction. The main and secondary rise of the estradiol was also observed in both sweat and serum in all menstrual cycles. This indicates that sweat estradiol follows cyclical fluctuation during the menstrual cycles. In addition, a strong correlation coefficient of was identified between sweat and blood estradiol levels from the pilot study (Fig. 4d), suggesting the high potential of sweat estradiol as a non-invasive biomarker for fertility and ovulation monitoring.

Figure 4. Evaluation of the wearable sensor for non-invasive female hormone monitoring in human subjects.

a, Hormonal fluctuations over the menstrual cycle. b,c, Continuous female hormone monitoring for two menstrual cycles in two female subjects. Urine LH levels were tested by the commercial LH strips kit while estradiol levels in serum and sweat were tested by the enzyme-linked immunosorbent assay (ELISA) and the aptamer sensors, respectively. Error bars represent the s.d. of the mean from three measurements. Subject 1: female, 32 years old; Subject 2: female, 28 years old. d, Correlation of estradiol levels between human sweat and serum (n=51). Dashed line represents the linear-fit trendline. e, Optical images of a female human subject wearing the wireless wearable sensor during an on-body test. The collected data was wirelessly sent to a smartphone and displayed in the custom-developed mobile app. Scale bar, 1 cm. f–h, On-body multiplexed physicochemical sensing and estradiol quantification with real-time sensor calibrations using the wearable aptamer sensor in the follicular, ovulation, and luteal phases of menstrual cycles for three female human subjects. Subject 1: female, 31 years old; Subject 2: female, 28 years old; Subject 3: female, 32 years old.

On-body evaluation of the wearable technology for real-time in situ estradiol analysis was performed on three female subjects on Day 5, Day 13, and Day 20 during a menstrual cycle with the sensor patch conformally attached onto the skin (Fig. 4e and Supplementary Video 3). During the study, sweat was induced via the built-in iontophoresis module, and sampled by the microfluidics. Multiplexed sensor data was collected wirelessly using a user interface (Fig. 4f–h). The calibrated estradiol levels were converted in real-time and displayed in a custom-developed mobile app based on the obtained multimodal data (Fig. 4e). As expected, the lowest estradiol levels were observed on Day 5 while the highest values appeared on Day 13 in all subjects; moderate estradiol levels were observed on Day 20 due to the secondary rise of the estradiol during the menstrual cycle. Additional control studies were performed on three male subjects, very low estradiol levels were obtained on all subjects with no apparent fluctuations (Supplementary Fig. 37).

Conclusion

This work demonstrates an aptamer nanobiosensor with sub-picomolar sensitivity for ultrasensitive and reagentless wearable analysis of the female hormone estradiol in sweat. A AuNPs-MXene conductive scaffold enhanced signal transduction by enhancing the WE’s electroactive surface area and charge transfer efficiency. For automatic on-body monitoring, a fully-integrated wireless wearable system was designed that incorporates iontophoretic hydrogels for localized sweat stimulation, microfluidics for sweat collection, and functionalized sensors for estradiol sensing and calibration. We identified a direct correlation between sweat and serum estradiol levels as well as clear cyclical fluctuation trends in sweat estradiol levels during the menstrual cycle. This device thus enables convenient at-home reproductive hormone monitoring and can be reconfigured to monitor other trace-level biomarkers toward a wide variety of personalized medicine applications.

Methods

Materials and Reagents

β-Estradiol, cortisone, progesterone, folic acid, serotonin, cortisol, tobramycin, kanamycin, glucose, lactic acid, 6-mercapto-1-hexanol (MCH), 3-mercaptopropionic acid (MPA), potassium ferrocyanide (II), tris(hydroxymethyl)aminomethane hydrochloride (Tris–HCl), tris (2-carboxyethyl) phosphine hydrochloride (TCEP), and ethylenediaminetetraacetic acid (EDTA) were purchased from Sigma-Aldrich. Magnesium chloride, potassium chloride, isopropyl alcohol (IPA), ethylene glycol, sulfuric acid, and sodium chloride, were purchased from Fisher Scientific. The enzyme-linked immunosorbent assay kit for estradiol was purchased from Invitrogen. Iron (III) chloride, potassium ferricyanide (III), silver nitrate, and 0.1 M phosphate-buffered saline (PBS) (pH 7.4) were purchased from Alfa Aesar. Medical adhesives were purchased from Adhesives Research and 3M. Polyethylene terephthalate (PET) films were purchased from McMaster-Carr. The estradiol aptamer and other DNA oligonucleotides with an initial concentration of 100 μM were ordered from Integrated DNA Technologies (detailed sequences shown in Supplementary Fig. 6a). Carbon ink (5 wt %) and silver ink (25 wt %) were purchased from NovaCentrix. Gold ink (10 wt %) was purchased from C-INK Co. Ltd. All chemicals are used as received. Screen-printed Au electrodes were purchased from Shenzhen Haoyang Technology Co., Ltd (HY-PA100). Plain Au electrodes (as the control) were fabricated on a PET substrate by photolithography and electron-beam evaporation (30 nm Cr/100 nm Au).

Preparation of the MXene nanosheets ink

MXene nanosheets were synthesized using the minimally intensive layer delamination (MILD) method⁴⁴. First, 2 g of LiF was dissolved in 40 ml of 9 M HCl. 1 g of Ti₃AlC₂ MAX phase powder was then gradually added into the etchant under continuous stirring at room temperature. The mixture proceeded to react for 48 h at 40 °C under continuous stirring. Once the reaction was completed, the acidic mixture was washed multiple times with deionized (DI) water by centrifugation at 2044 relative centrifugal force (RCF) for 5 min each cycle until the pH turned neutral. The dark-green supernatant was discarded, and 20 ml DI water was added to the sediment. This was followed by vigorous shaking using a vortex machine for 1 h. The dispersion was then centrifuged at 327 RCF for 5 min and the supernatant was collected. The suspension containing exfoliated MXene nanosheets was centrifuged again at 2044 RCF to remove away the residual impurities. 10 ml of DI water was added to the sediment and an MXene suspension with a concentration of ~15 mg ml⁻¹ was obtained. To prepare the MXene ink, an aliquot of the MXene suspension was diluted to a concentration of 2 mg ml⁻¹. The suspension was then probe sonicated for 1 h under the protection of N₂ to reduce the size of the nanosheets. The suspension was then filtered using a 0.45 µm filter and ethylene glycol was added in a volume ratio of 1:1.

Fabrication and characterization of disposable sensor patch

The fabrication procedures of the inkjet-printed sensor patch are illustrated in Supplementary Fig. 3. Firstly, the PET substrate was washed with IPA and dried with compressed air flow. The PET substrate was treated with a 10-minute O₂ plasma using Plasma Etch PE-25 (20 to 30 cm³ min⁻¹ O₂, 100 W, 150 to 200 mTorr) to enhance the surface hydrophilicity. The multimodal sensor patch was fabricated via serial printing of gold (reference/working/counter electrodes, pH sensor, and ionic strength sensor), silver (interconnects and reference electrode), carbon (iontophoresis electrodes), and MXene (WE) using an inkjet printer (DMP-2850, Fujifilm). The thickness of MXene layer is crucial to the sensor performance (Supplementary Fig. 15) and 4-layer MXene printing coupled with 2-layer Au printing was determined to be optimal. The plate temperature was set to 40 °C to ensure the rapid ink solvent vaporization during printing.

The morphology and material properties were characterized by transmission electron microscopy (TEM, Talos S-FEG FEI), and scanning electron microscopy (SEM, Nova Nano SEM 450). The Raman spectrum of the electrodes was recorded using a 532.8 nm laser with an inVia Reflex (Renishaw). X-ray diffraction (XRD) patterns were recorded with an X-ray diffractometer (D/max-Ultima IV, Japan).

Construction and modification of sensing platform

The biorecognition interface and WE were assembled face-to-face on the sensing patch. We first diluted both of thiol-modified aptamer-ssDNA and SH-ssDNA to 5 μM with 20 mM Tris-HCl buffer (containing 50 mM NaCl and 25 mM MgCl₂, pH 8.0), and then reduce them for 3 h in a solution of 500 μM TCEP at room temperature in the dark. To form the aptamer’s functional secondary structure, the estradiol aptamer-ssDNA was then heated at 95 °C for 5 min, and then cooled slowly down to room temperature. As shown in Supplementary Fig. 8, SH-ssDNA (5 μM, 2 μl) and aptamer-ssDNA (5 μM, 2 μl) were immobilized onto the inkjet-printed AuNPs-MXene (WE) and AuNPs (biorecognition interface), respectively, via the Au-SH binding overnight, respectively. MCH (100 mM, 2 μl) was employed to block the unbound active sites of AuNPs for 30 min, subsequently, followed by rinsing with DI water to wash away the non-electrostatically adsorbed ssDNA and aptamer-ssDNA. Then, the redox probe MB-ssDNA (1 μM, 2 μl) was attached to the biorecognition interface for 30 min to form a partially hybridized duplex with an aptamer-ssDNA, which acts as a competitive redox probe.

The pH sensing electrode was prepared by electrodeposition of polyaniline pH sensing membrane on the inkjet-printed Au electrode via cyclic voltammetry between −0.2 and 1.2 V at scan rate of 0.1 V s⁻¹ for 25 cycles.

Electrochemical characterizations of the biosensors

Square wave voltammetry (SWV), cyclic voltammetry (CV), differential pulse voltammetry (DPV), and open circuit potential-electrochemical impedance spectroscopy (OCP-EIS) were carried out on a CHI820 electrochemical station. All the electrochemical measurements were performed at room temperature in artificial sweat (0.2X PBS, pH 7.4 unless otherwise noted) and SWV was performed over a voltage range of 0 to −0.6 V at a frequency of 90 Hz (unless otherwise noted). To characterize the sensor performance, estradiol was introduced in between the biorecognition interface and WE for a 90-minute incubation (unless otherwise noted). The limit of detection was calculated based on the mean of the blank, the standard deviation of the blank, the sensitivity, and a defined confidence factor of 3.2 (for a confidence of 95%).

In order to characterize the electrode performances of screen-printed Au, Au, AuNPs, and AuNPs-MXene, and to confirm the surface modification after each step, DPV and OCP-EIS readings were executed in 0.1X PBS (pH 7.4) containing 2.0 mM K₄Fe(CN)₆/K₃Fe(CN)₆ (1:1) under the following conditions: potential range, −0.2–0.6 V; pulse width, 0.2 s; incremental potential, 4 mV; amplitude, 50 mV; frequency range, 0.1–106 Hz; amplitude, 5 mV. [Fe(CN)₆]³⁻/[Fe(CN)₆]⁴⁻ was chosen as the redox couple for electrochemically monitoring the stepwise modification of electrode surfaces owing to its low redox potential, low surface contamination, and negative charge for optimal electrode performance characterization^45–47. Besides, the electrodes surface of Au, AuNPs, and AuNPs-MXene were compared through CV characterization in the range from 0.5 to 0.7 V at different scan rate (5, 10, 50, 100, 150, 200 mV s⁻¹) in artificial sweat (0.2X PBS, pH 7.4).

The electrodes used in these studies were prepared on the same day and stored at 4 °C (unless otherwise noted). To shorten the incubation time by enhancing the molecular transport, a bias-potential (1 min duration at +0.5 V) could be applied prior to the incubation between WE and counter electrode.

Numerical simulation of electric field enhanced bioanalysis

Numerical simulation of the chemical reaction and mass transport of estradiol and MB-ssDNA were conducted through Monte-Carlo method using MATLAB R2020b. Initially, estradiol molecules were randomly distributed in the 0.335 mm × 4 mm 2D rectangular area and doing random walk. Once estradiol reached the biorecognition interface, it could be captured by the aptamer-ssDNA, resulting in the release of MB-ssDNA; the probability here, P₁, was set to be 6 × 10⁻³ (value was chosen according to the experimental data). The released MB-ssDNA will also do random walk in the system. The mean squared displacement (MSD) of species in the system is described by:

$$MSD= {\left(\mathrm{\Delta }x\right)}^2=4D\times \mathrm{\Delta }t$$

Where $\mathrm{\Delta }x$ is the displacement within one time step $\mathrm{\Delta }t$, and $D$ is the diffusion coefficient. Here, the time step was set to be 1 s, and the diffusion coefficient of estradiol and MB-ssDNA are 6.2 × 10⁻¹⁰ m² s⁻¹ and 5.0 × 10⁻¹¹ m² s⁻¹ respectively^48,49. Once an MB-ssDNA reached the WE, there is a probability P₂ for the MB-ssDNA molecule to be captured by a complementary DNA on the WE within one time step. Since MB-ssDNA is a large molecule, the effective binding sites are limited. Each bound MB-ssDNA will occupy one effective reaction site, affecting the surrounding complementary DNAs to capture MB due to steric hindrance. So, the capture probability of the WE is:

$$P_2=k_{capture}\times {\left(1-\frac{n_{bounded}}{n_{initial}}\right)}^2\times \mathrm{\Delta }t$$

Where $k_{capture}$ is a constant and was set to be 2 × 10⁻³ s⁻¹, $n_{initial}$ is the initial number of effective bounding sites and was set to be 3.2 × 10⁶, and $n_{bounded}$ is the number of effective bounding sites already bounded to MB-ssDNA ($k_{capture}$ and $n_{initial}$ values were chosen to match the experimental data without electric field). Since the binding process is reversible, for each bound MB-ssDNA, there is a probability P₃ that it will be released from the WE, and here we set P₃ to be:

$$P_3=k_{release}\times \mathrm{\Delta }t$$

Where $k_{release}$ is a constant and was set to be 1.3 × 10⁻³ s⁻¹.

The electrophoresis velocity of MB-ssDNA is described by:

$$v=\mu \times E$$

Where $\mu$ is the electrophoretic mobility of MB-ssDNA, which was set to be 3.4 × 10⁻⁸ m² (V·s)^{−1 50} and $E$ is the electric field strength in the system. Since discharging is going to occur in the system, the electric field strength is described by:

$$E=E_0\times e^{-\frac{t}{t_c}}$$

Where $E_0$ electric field strength at $t=0$ and was set to be 4.5 × 10⁻² V mm⁻¹, and $t_c$ is a constant and was set to be 20 min.

Fabrication and characterization of microfluidics

A 50 W CO₂ laser cutter (Universal Laser System) was used to fabricate the microfluidic module. For all microfluidic layers, the iontophoresis gel outlines were patterned to enable the contact between the hydrogel and the skin. The bottom accumulation layer (the double-sided adhesive layer in contact with the skin) was patterned with a sweat accumulation chamber (130 µm 3M 468MP, laser parameters: power 60%, speed 90%, PPI 1,000) and microchannels including the capillary bursting valves (CBVs). The chamber has an ellipse shape (6.4 mm in minor axis and 7.5 mm in major axis) to fully enclose the sweat collection reservoir. The dimension of the CBVs is demonstrated in Supplementary Fig. 29. The second layer as the biorecognition interface with the inkjet-printed AuNPs was patterned with an inlet (100 µm-thick PET, laser parameters: power 30%, speed 90%, PPI 1,000). For the third layer (collection layer), in contact with the biorecognition interface layer, an assembly of PET film with the thickness of 75 µm sandwiched between double-sided medical adhesives (130 µm 3M 468MP) was cut through to make the inlet and outlet, channels with capillary bursting valve, and sweat collection reservoir (laser parameters: power 60%, speed 75%, PPI 1,000). The thickness of the collection layer is ~335 µm, the reservoir area is ~19.87 mm², and the reservoir volume totals ~6.66 µl, which is calculated as the area multiplied by the thickness of the reservoir layer.

Fabrication of iontophoresis hydrogels

The iontophoresis hydrogels on the anode and cathode are made of agarose (3% w/w). The agarose was fully dissolved in DI water and then heated to 250 °C. 1% carbachol as the muscarinic agents for the anode hydrogel and 1% KCl for the cathode hydrogel were added when the solution was cooled down to 165 °C. Subsequently, the mixture was slowly cast into the microfluidic patch and solidified for 10 min at room temperature. The prepared hydrogels were stored at 4 ℃ for long-term use.

Electronic system design and integration

A flexible printed circuit board (FPCB) with two layers was created using Eagle CAD and Fusion 360. The FPCB has a rectangular shape (13 mm × 46 mm) that can be wrapped around a finger like a ring. The electronic system of the FPCB includes a voltage regulator (ADP162, Analog Devices) for power management, a boost converter (TPS61096, Texas Instruments), BJT array (BCV62C, Nexperia), and analog switch (DG468, Vishay Intertechnology) for iontophoretic sweat induction, an analog front-end (AFE) (AD5941, Analog Devices) and an operational amplifier (LPV811, Texas Instruments) for sensor interface, and a Bluetooth Low Energy (BLE) module (CYBLE-222014-01, Cypress Semiconductor) programmed through PSoC Creator 4.3 for controlling the system and enabling wireless communication. Through its general-purpose input/output (GPIO) pins, the BLE module initiates current flow through the iontophoresis electrodes for sweat stimulation. The iontophoresis setup is designed to apply 55 µA across the skin with a compliance voltage of 18 V such that it can function across a wide range of skin impedances. Furthermore, the current goes through a switch and is interfaced with a transimpedance amplifier within the electrochemical AFE to monitor the current that is applied during iontophoresis, and to shut the current off if the current source malfunctions and applies a current above 55 µA. After stimulation, the BLE module controls the AFE to perform electrochemical techniques such as potentiometry, impedimetry, and SWV to acquire sensor data. After a certain amount of sensor data has been piled up in the AFE’s internal memory, the AFE interrupts to BLE module to fetch and transmit the sensor data to a customized mobile application or a BLE dongle (CY5677, Cypress Semiconductor) for data processing and display.

The power consumption profiles of the wearable system are demonstrated in Supplementary Fig. 34. During operation, the electronic system is connected to a user interface over Bluetooth low energy with a connection interval of 0.25 s. Iontophoresis consumes around 2.7 mW for 5 minutes. Prior to incubation, an electric field is applied for 1 minute, where 0.5 V is applied and the resulting current is measured with a sampling interval of 2 s. During the electric field application, the system which operates at 3 V consumes an average power of ~695 µW for 1 minute. During the incubation, the system goes to idle mode where a BLE connection interval of 4 s is maintained to consume a power of ~100 µW. Then, SWV is performed, where an average power of ~11.3 mW is applied for 2.3 s. Then, simultaneous ionic strength, pH, and temperature measurements, with a sampling interval of 2 s is applied for 1 minute, consuming an average power of ~1.8 mW.

Human subject recruitment

The evaluation of the wearable technology for non-invasive female hormone monitoring in sweat and serum samples for human subjects is following all the ethical regulations according to the protocol (ID 19-0892), which was approved by the Institutional Review Board (IRB) at California Institute of Technology (Caltech). The participating subjects (age range 18–65 years; premenopausal) were recruited from Caltech campus and the neighboring communities through advertisement by posted notices, word of mouth, and email distribution. All subjects gave written informed consent before participation in the study.

For sweat collection (used for in vitro sensor validation), an iontophoresis session was performed on the subjects using a Macroduct Sweat Collection System (Model 3700) to induce sweat. A Macroduct collector was then used to collect sweat sample over a period of 20 minutes. Fresh blood samples were collected at the same period of sweat collection using a finger-prick approach. After the clotting procedure, serum was separated by centrifuging (15 minutes at 1044 RCF), and instantly stored at −80 °C.

Human samples with the ELISAs

ELISA tests for estradiol analysis in serum and sweat samples were performed in an accuSkan FC Filter-Based Microplate Photometer at a detection wavelength of 450 nm according to the manufacturer’s instructions. In brief, 50 µl of standard, control or samples were added to microtiter plate wells, and then 50 µl of estradiol-HRP conjugate and 50 µl of anti-estradiol were added into all wells in sequence. The plate with above mixture was covered with aluminum foil or metalized mylar and incubated for 2 h at room temperature on a horizontal shaker set at 1.372±0.0168 RCF. After four times washing steps with wash buffer, 200 µl chromogen solution (3,3′,5,5′-tetramethylbenzidine, TMB) was added into each well, in which the liquid in the wells turned blue upon reaction. And incubated for 30 min at room temperature in the dark, the absorbance values were measured immediately after the addition of stop solution (50 µl, 1 M H₂SO₄) in each well.

Evaluate the wearable estradiol sensor in human subjects

To evaluate the wearable aptamer sensor for sweat estradiol analysis and investigate the clinical values of sweat estradiol, female subjects were recruited with periodic female hormone monitoring for two consecutive menstrual cycles. During each day’s study, we collected self-reported urine LH level analyzed by the subjects with a commercial strip kit, body temperature monitored with a medical thermometer, serum estradiol (serum samples were collected via invasive blood draw and analyzed with ELISA kits) and sweat estradiol (sweat samples were collected via iontophoresis and analyzed with the aptamer sensors).

For on-body system validation, the subjects were arranged with multiplexed physicochemical sensing and estradiol quantification with real-time sensor calibrations using the wearable aptamer sensor in the follicular, ovulation, luteal phases of menstrual cycles. The subjects’ arms/fingers were cleaned with alcohol swabs, and after 5-minute iontophoresis applied on the subjects, the aptamer sensor patch performed sweat sampling, incubation, and biosensing for 60 minutes. During the on-body trial, the collected SWV data and the physical data, including temperature, pH, and impedance, were wirelessly transmitted to a user device via BLE in real time, where the molecular data were extracted, calibrated, and converted to concentration levels. Microsoft Excel 2016 was used to analyze all data, plot the data and calculate the statistical parameters.

Data availability

The main data supporting the results in this study are available within the paper and its Supplementary Information. Source data for Figs. 2, 3 and 4 are provided with this paper.