A Sensitive DNA Bowtie Sensor for Calibration-Free Sex Hormone Detection Using Analyte-DNA Conjugates and Multiple Redox Labels

Abstract

Using a modular and reconfigurable DNA-based bowtie sensor structure, several advancements in electrochemical (EC) biosensing are reported. Raw current and sensitivity were enhanced up to 13-fold using a novel redox reporter design, where a DNA strand was decorated with five methylene blue molecules (MB₅) instead of the typical single label (MB), and flexible spacers further improved signal. Leveraging the bowtie sensor’s modularity, custom synthesized analyte-DNA conjugates (testosterone-DNA, estradiol-DNA, and progesterone-DNA) and a control strand (digoxigenin-DNA) were incorporated into the complete DNA structure on electrode surfaces by enzymatic ligation. Upon binding antibodies, each of these sensors showed enhanced current with the MB₅ label, and responses consistently followed a Langmuir curve. A previously inaccessible signal-on responsesustained current at longer timeswas observed at low square wave voltammetry frequencies using MB₅, allowing new differential signal corrections. The limit of detection (LOD) for direct antibody sensing was as low as 0.08 nM (80 pM) for the antidigoxigenin antibody (control system). A mathematical model was devised based on uncoupled equilibria among the analytes, antibodies, and the electrode surface. After training the model on binding curves, calibration-free measurements of small molecule hormones were possible, with observed LODs for testosterone, estradiol, and progesterone at 0.2, 0.8, and 1 nM, respectively, extending within human clinical ranges. For example, the testosterone sensor’s range was 200 pM to 70 nM, covering the clinical range in both females (0.52 to 2.43 nM) and males (9.0 to 34.7 nM), and the sensor functions with 90% recovery in 50% human serum. This work expands the usage of the EC bowtie sensor, employing custom synthetic organic methods to introduce new analyte-DNA conjugates and further validate the sensor’s generalizability, alongside developing a unique model of the binding mechanism. Beyond this, the newly introduced MB₅ label should be functional with a range of other DNA-based EC sensors or aptamer-based sensors.

Introduction

Sex hormones with steroid-based structures interact with steroid hormone receptors and play important roles in controlling various functions and behaviors in the body, in addition to their roles in regulating the reproductive system. − As major sex hormones, progesterone, estradiol, and testosterone control and report on conditions such as pregnancy, puberty, menstruation, menopause, and sex drive, and they show phenotypic effects ranging from body weight to mood. , The hormones are mainly secreted from testes or ovaries, with small contributions from the adrenal gland.

Testosterone is the most abundant sex hormone in males (9.0 – 34.7 nM, total serum level including bound and free), and it controls, maintains, and develops secondary sex characteristics, libido, growth, and prevention of osteoporosis. In women, testosterone is produced in lower amounts (0.52 – 2.43 nM), yet it remains a clinically important hormone to measure. Abnormalities in testosterone levels can cause androgen deficiency and hypogonadism in males or androgen excess with polycystic ovary syndrome in women. Other than its important effects in the reproductive system in women, progesterone plays an important role in several tissues, such as the mammary gland in preparation for breastfeeding, the cardiovascular system, the central nervous system, and in bones. Estradiol controls the growth of the uterus, fallopian tubes, and breast development. It also plays a role in genital change, distribution of body fat, and menopause. Anomalies of these hormone levels can cause significant adverse effects in women. As an environmental effect, commercial or local sources of milk and water can contain high amounts of progesterone, and consumption can lead to side effects such as breast tenderness, headache, upset stomach, constipation, and diarrhea.

Clearly, clinical measurements of testosterone, estradiol, and progesterone are important for a variety of human conditions. Unfortunately, due to the inconsistencies often observed with immunoassays in clinical practice, the accepted gold standard methods for measuring these hormones are typically based on isotope dilution liquid chromatography tandem mass spectrometry (ID-LC-MS/MS), which require significant sample preparation and analysis workflows. − For example, the gold standard method for testosterone quantification from a blood sample involves many steps in the workflow: addition of a deuterium labeled testosterone standard, extraction with ether, evaporation, reconstitution, derivatization, mixing, heating, solid-phase extraction and elution, evaporation, and reconstitution into a mobile phase for injection into the LC-MS/MS system. A much simpler electrochemical (EC) binding assay would greatly reduce the costs and workflow needs and would be highly beneficial from an analytical methods point of view. Below, we discuss the steps involved in adapting of our analyte-flexible, DNA bowtie-based binding assay − to the quantitative measurement of testosterone, estradiol, and progesterone.

DNA-based EC sensors have been extensively leveraged for measurement of biologically relevant analytes. In particular, structure-switching aptamer-based sensors, − or “aptasensors,” have opened a new sensing paradigm, enabling continuous monitoring of biomolecules, even in living animals. Feedback-controlled personalized pharmacokinetics could feasibly be achieved in humans with these sensors, which most assay platforms cannot accommodate. Although this approach could be generalizable using the SELEX processwhich can theoretically find an aptamer for any target of interestthe selection process can be very expensive and can sometimes end without a suitable structure-switching aptasensor even after significant reengineering efforts. As they are much less chemically diverse compared to antibodies, aptamers are not yet available for many clinically relevant analytes. Real-time sensing needs also place a lower limit on aptamer/analyte dissociation constants (usually in the μM range), since fast equilibrium off-rates are needed. These structure-switching EC aptasensors are not yet available for many important analytes, and although they continue to find great success in continuous measurement, they are not usually applied to analytes with clinical importance in the pM or nM range.

As an alternative, EC, nonstructure switching assays such as E-DNA scaffolds, our modular DNA nanostructures, − pendulum sensors, − and steric hindrance assays have been incorporated to quantify clinically significant bioanalytes in complex fluids within minutes, with low limits of detection (LOD) ranging from nM to pM. These sensors typically come with a trade-off between affinity and measurement time, since the highest affinity binders (e.g., antibodies, DNA hybridization, and low K_d aptamers) will have slow kinetic off-rates. While less suitable for real-time measurements, these sensors are more adaptable to many analytes and have been proven well-suited for measurement of clinical samples with analytes of low concentration. Interestingly, this caveat may have been overcome in recent work by the “active reset” approach of the Kelley group, in which clever programming of electrode potential was used to essentially shake off bound analyte from high-affinity sensors. It remains to be seen if the active reset method can be applied to other sensor architectures, although these results bode well for the further development of higher-affinity, surface-confined, reusable EC biosensors as presented herein.

The sensitivity of EC sensors can be improved by modifying or optimizing the molecular assay, the instrument, or the interrogating technique. Nanostructured surfaces , have been evaluated for improved signal-to-noise ratio (S/N) by increasing the sensor’s surface area. Introducing roughness and material shrinking ,, are other techniques that can be utilized to improve S/N. Alternatively, changing the redox reporter, incorporating dual redox reporters for ratiometric readout, , and polymerized redox reporters , have all provided sensitivity improvement and stability. Interrogation techniques like square wave voltammetry (SWV) and differential pulse voltammetry (DPV) are commonly used due to high sensitivity; , and specialized approaches such as Hot-SWV, pendulum time constant analysis, cSWV, differential potentiostat circuits, and kinetic differential measurement (KDM) have been adapted more recently. Another possible way to improve sensitivity would be to include multiple of the same redox molecules at the labeling site. Indeed, a DNA origami-based collection of many reporters was recently reported, although this approach is inherently complex to implement. To date, the more direct approach of chemically synthesizing a multilabeled redox reporter strand has not yet been explored in the realm of DNA-based EC sensors. In this work, we show clear evidence that this approach is practical and enables improved detection modes.

To improve the potential for sensor generalizability toward a wider variety of analytes, we have shown that a modular construction of a DNA nanostructurecontaining modules for electrode attachment, target recognizing, and signalingonto electrode surfaces can provide a valuable assay platform. Herein, we refer to this sensor structure as the “bowtie sensor.” With the same core structure and signaling mechanism, this DNA bowtie platform has been used thus far to quantify small molecules, peptides, antibodies, − and proteins , in complex biofluids. The approach is flexible and can be utilized as either a simple one-step, reagent free method for antibody or protein sensing or in a two-step competitive assay workflow for small molecule and peptide sensing, sometimes in protein sensing. Stable signal generation has been demonstrated from as low as 3 min up to 30–60 min, making this sensor viable as a potential point-of-care (POC) or do-it-yourself (DIY) assay. ,

Through several important advancements, including custom organic chemistry methods for analyte-DNA conjugate synthesis, this work reports the first adaptation of the DNA bowtie sensors toward clinically relevant hormone quantification in the pM or low nM range in human serum. Because the sensor’s modular design enables adaptation to new analytes, we demonstrate several approaches in chemical synthesis that can be used to make analyte-DNA conjugates, ultimately allowing sensing of the chosen analyte. Three important sex hormonestestosterone, progesterone, and estradiolwere covalently linked to oligonucleotides and incorporated into the DNA structure via enzymatic ligation.

We additionally report a novel method to improve sensitivity by replacing the typical single methylene blue label (MB) with a multilabeled strand (MB₅). This method moved the sensor’s response into the microampere range, gave significant improvements in reproducibility, afforded from 6- to 13-fold improvements in the analyte sensitivity, and enabled both signal-on and signal-off mechanisms with a nonstructure switching sensor for the first time (inaccessible with standard MB labels). With these benefits realized, it was possible to develop a mathematical model of the bowtie sensor’s operation which allowed calibration-free sensing of small molecule hormones.

By introducing improved chemical methods for analyte-DNA conjugate synthesis and for introducing more redox reporters onto surface-bound sensor structures, this work represents an important bioanalytical advancement toward small-molecule hormone sensing and for more generalizable, high-sensitivity, and calibration-free sensing in the future.

Materials and Experimental Methods

Reagents and Materials

All solutions were prepared using ultrafiltered DI water (Fisher Scientific). The reagents used in this study were as follows: magnesium chloride hexahydrate, 6-mercaptohexanol (MCH), tris­(2-carboxyethyl) phosphine hydrochloride (TCEP), gold etchant, and chromium etchant from Sigma-Aldrich; sodium chloride (NaCl) from BDH; dimethyl sulfoxide (DMSO) from VWR; and chromium­(VI) oxide from Sigma-Aldrich. Gold-sputtered gold on glass (GoG) (5 nm Cr adhesion layer, 100 nm Au layer) with dimensions of 25.4 mm × 76.2 mm × 1.1 mm was purchased from Deposition Research Lab, Inc. (St. Charles, MO). AZ 40XT (positive photoresist) and AZ 300 MIF developer were obtained from Microchemicals. Polydimethylsiloxane (PDMS) was purchased from Ellsworth Adhesives. Testosterone, progesterone, estradiol, testosterone 3-(O-carboxymethyl) oxime, and progesterone 3-(O-carboxymethyl) oxime, sodium phosphate, anhydrous DMF, deuterated methanol, and O-(carboxymethyl)­hydroxylamine hemihydrochloride were purchased from Sigma-Aldrich. Customized MB and MB₅ conjugated DNA were purchased from Biosearch Technologies (Novato, CA), and thiolated DNA was obtained from Integrated DNA Technologies (IDT; Coralville, Iowa) with purity confirmed by mass spectroscopy. T4 DNA ligase (400,000 units) and adenosine triphosphate (ATP, 10 mM) were from New England Biolabs. Prescreened human serum samples were acquired from BioIVT. The sequences of DNAs used in this study are listed in Table S-1 and Table S-2.

Antitestosterone antibody (anti-Tes) (ab116597), antiprogesterone antibody (anti-Prg) (ab20236), and antietradiol antibodies (anti-Est) (ab131413 and ab116648) were purchased from Abcam. Anti-Prg antibodies (P1922) and (MA-1–83403) were obtained from Sigma and Thermo Fisher Scientific, respectively. Antibodies were used as received.

Patterning of Gold-on-Glass (GoG) Electrodes

To create electrodes of 2 mm diameter, a photomask with 18 identical electrodes was designed in Adobe Illustrator and sent to Fineline Imaging (Colorado Springs, CO) for printing a positive photomask (see Figure S-8). An optimized photolithographic process was followed to pattern the gold electrodes on GoG slides, as reported previously. To remove excess gold and chromium, the slides were treated with gold and chromium etchant for 30 and 15 s, respectively. The remaining positive photoresist on the GoG slides was removed by immersing them in DMSO at 150 °C for 30 min. Finally, the electrodes were rinsed with deionized water for 30 s, followed by ethanol, and dried with nitrogen.

DNA Monolayer Assembly

First, electrodes underwent a cleaning process to remove impurities. The gold surface was treated with freshly prepared piranha solution (H₂SO₄:H₂O₂, 3:1) for 1 min and then rinsed with deionized water. The PDMS wells and gold electrodes were sonicated in methanol for 2 min and heated at 100 °C for 5 min. An air plasma was used to clean and oxidize both the electrodes and PDMS wells for 45 s (Harrick Plasma Cleaner), and the wells were aligned and affixed to the GoG slide within 30 s.

Once slides and wells were permanently attached, thiol DNA was incubated on the electrodes to create a self-assembled monolayer of thiol-DNA. Prior to monolayer assembly, the thiol-DNA was reduced by mixing a 10 μM solution with TCEP (10 mM) for 1 h, followed by diluting the reduced solution to a final concentration of 30 nM in buffer (10 mM HEPES with 10 mM MgCl₂ at pH 7.0). To prevent sodium interference with the ligation reaction, the buffer did not contain NaCl. The diluted solution was added to the electrodes and incubated for 1 h. Afterward, 3 mM of MCH in buffer was added and incubated for 1 h. To prevent protein and antibody adsorption onto PDMS or glass, the entire electrochemical cell was filled with 0.1% BSA in buffer and left for 30 min. The electrodes were then ready for the construction of the DNA bowtie sensor on the gold surface.

For the construction of the DNA bowtie sensor, 100 nM of antibody-binding DNA strand (various analyte-DNA conjugates), 100 nM MB-DNA or MB₅-DNA, 1 mM ATP, and T4-DNA ligase enzyme were mixed. Twenty μL of the mixture was incubated on each electrode (having immobilized thiol-DNA), and ligation was allowed to proceed for 6 h at room temperature. To prevent evaporation, the wells were covered with parafilm and placed inside a Petri dish containing DI water. The electrodes were then washed with DI water for 45 s to remove nonligated nanostructures and other impurities. The measuring buffer (0.5 M NaCl, 0.1% BSA, and 10 mM HEPES) was introduced, and the electrodes were stored until used.

Electrochemical Measurements

Sensor responses were recorded using a Gamry Reference 600 potentiostat operating in square-wave voltammetry (SWV) mode using a step size of 1 mV and a pulse height of 25 mV, and an in-house-built Peltier control system was used to maintain temperature stability. A range of frequencies from 4 to 900 Hz were scanned, and potential scanning was carried out from −0.450 to 0.000 V versus an Ag/AgCl reference (3 M KCl) (BASi), using a Pt counter electrode (CH Instruments). Empirical observations show that sensors can be stored at 4 °C and interrogated for several days and up to a few weeks without loss of performance in buffer solution. In this study, typical experiments on a given electrode were carried out at varying times after monolayer assembly, as short as 30 min but not extending beyond 12–14 h (for calibrations).

Following data acquisition, SWV data was imported into Microsoft Excel, and each curve was processed by fitting the nonfaradaic baseline to a third-order polynomial, as described in our prior work. , After baseline subtraction, the faradaic peak current was determined for each measurement by averaging the 5 largest points in the processed SWV peak. Additional details on our custom signal processing approach and the Lovric analysis are provided in the Supporting Information document.

Results and Discussion

Generalizable Sensing with a Modular DNA Bowtie

The modular nature of the bowtie sensor is a key advantage of the platform, − enabling quantitative sensing of small molecules, peptides, and proteins. A unique aspect of this structure is that the position and movement of the redox label is essentially independent of the identity of the targeted analyte. The position of the MB label with respect to the electrode is mostly fixed into a hemisphere of tethered diffusion , which can be slowed by antibody binding, and indirect sensing of smaller analytes can be achieved by controlling the extent of antibody binding compared to a blank. These sensors fit into the category of higher affinity binders that are useful for lower limits of detection compared to other DNA-based sensors, while they are less applicable for real-time sensing in their current formthat is, unless active reset can be applied in the future.

As shown in Figure A, the sensor’s DNA structure is immobilized on gold electrodes with a common thiol-DNA strand, and EC readout is provided through tethered diffusion of the common signaling DNA strand. The central strand, an analyte-DNA conjugate, can be substituted depending on the analyte of interest. Once all three strands are enzymatically ligated together (either on the surface or in bulk solution), the bowtie sensor surface consists of an array of highly stable, single DNA molecules, each with three targeted chemical modifications. Compared to aptamer-based sensors, which require laborious selection and screening of new aptamers for each intended target, the bowtie sensor can simply be modularly updated with a different central strand to bind antibodies, which are much more well-developed and commercially accessible as bioaffinity reagents.

1.

Schematic and initial testing of the enhanced EC bowtie sensor. (A) Modular construction of the DNA-based bowtie allows generalizable sensing by simply substituting the central analyte-DNA conjugate strand (magenta). This strand is enzymatically ligated to both a thiol-DNA (dark blue) for surface immobilization and a multiply labeled signaling DNA (brown) for reporting binding events. (B) Testosterone (Tes) sensing workflow and initial test data using SWV; blanks (purple), 10 nM Ab (magenta), and 10 nM Ab + 100 nM Tes (blue). Current and sensitivity enhancements with the MB₅ label (left) are obvious when compared to the standard MB label (right). Example data shows single experiments, while numerical sensitivities report triplicate runs during later calibrations. Sample matrix was BSA-containing, neutral buffer solution.

In this work, we present chemical strategies to synthesize the three relevant analyte-DNA conjugates shown in Figure A and demonstrate that the same sensor architecture can be used to quantify three different sex hormones: testosterone, progesterone, and estradiol.

Signaling with Multiple Internal Redox Labels

By modifying available redox reporters or utilizing new redox reporters, several groups have improved the signal-to-noise ratios and the stability of E-DNA based sensor readout in complex media. − While other labels and derivatives have been evaluated, , methylene blue (MB) remains the default label for most DNA-based EC sensors, presumably due to its repeated validation and superior stability against a large number of voltammetric scans in complex media. In this study, we maintain the advantages of the MB label while increasing the signal significantly by modifying the signaling site of the DNA-oligo with multiple, strategically spaced redox reporter molecules. We hypothesized that decorating the signaling oligo with multiple MB molecules (MB₅) would result in proportionally higher signal than the conventionally used single MB molecule (MB).

The experimental data support our hypothesis. Examples are shown by the square-wave voltammograms in Figure B. When the bowtie sensor was operated as a direct antibody sensor, the sensitivity toward testosterone antibody concentration was 130 ± 10 nA nM^–1 with the MB₅ label and 10 ± 1 nA nM^–1 with the typical MB label, an approximately 13-fold improvement in raw current with high statistical significance. Similarly, when operated as an indirect electrochemical immunosensor responsive to testosterone, the sensitivity was significantly improved by 8-fold from 5 ± 1 to 42 ± 6 nA nM^–1 using the MB₅ label, compared to MB. Raw currents with the MB₅ labeled bowtie reached nearly 2.5 μA compared to only about 400 nA with an otherwise equally fabricated sensor.

This significant increase in sensitivity should make measurements easier to interpret and should result directly into lower LODs, which can be particularly useful for the indirect immunosensor workflow. As discussed below, when decorating the bowtie with MB₅ labels, additional unexpected benefits emerge, such as the ability to operate as either a “signal-on” or “signal-off” sensor at different frequencies.

Antibody Binding Response of the Bowtie Sensor

To rigorously test the hypothesis on MB₅ improvements, a digoxigenin-modified bowtie sensor was initially prepared to serve as a control system (see experimental section in Supporting Information). For the antibody-binding strand, both the MB- and MB₅-labeled DNA (Table S-1) were used for comparison in otherwise identical sensing systems. To gain a deeper understanding of the systems, a comprehensive SWV frequency study was carried out from 4 to 900 Hz at up to 18 different frequencies.

As shown in Figure , unexpectedly, the MB₅ label appeared to introduce a nonresponsive frequency (f_NR) into the bowtie sensor’s response to antibody binding, a feature that was not observed in any of our prior work with the typical MB labels. − Using the 4A spacer, the f_NR was seen at ∼50 Hz, as observed by the crossing point in Figure A, the positive signal change percentages in Figure C (<50 Hz), and the crossing point in Figure E. Below the f_NR, or at times longer than the crossover in Figure E, the current remained higher with the slower moving, antibody-bound bowties. Effectively, we can say that the SWV voltage pulses begin at zero time in Figure E. Thus, a higher signal at longer times represents a “signal-on” sensing mechanism, where antibody binding slows the movement of the bowtie structure enough that electron transfer continues at longer times beyond each voltage pulse. The MB₅ label thus enhanced both the current and the mechanistic operation of our bowtie sensors, which should permit improved analyses and may even be translatable to other sensor types.

2.

Significant improvements in antibody-binding response of the DNA bowtie sensor using MB₅ labels. A digoxigenin-labeled bowtie sensor’s response before (blue colors) and after (red colors) addition of 10 nM antidigoxigenin antibodies. The raw current responses are shown for (A) MB₅ labels and (B) MB labels at a wide range of SWV frequencies. (C) The signal change percentage versus frequency shows that MB₅ uniquely allows the bowtie to operate as a “signal-on” sensor below about 50 Hz. (D) Lovric analysis showed that MB₅ labels clearly increased the effective electron transfer rates and shows that antibody binding slows the movement of the system. Comparing the temporal response, the (E) MB₅ sensor showed a clear crossover time, while the (F) MB sensor did not. Sample matrix was BSA-containing, neutral buffer solution.

By calibrating the sensor against antibody (Ab) concentration and reporting the relative change in signal, it was possible to further evaluate the mechanistic response. The bowtie sensor with the MB₅ label responds typically to digoxigenin antibody (anti-Dig) binding in the low nanomolar range with a decreasing faradaic current at high f_SWV, thereby operating as a signal-off sensor at high frequency (Figure A).

3.

Concentration-dependent binding of antibodies to the bowtie sensor. (A) The high-frequency (high f_SWV) response of the Dig-modified bowtie sensor showed signal-off behavior upon addition of more anti-Dig antibodies. (B) Signal-on behavior, novel to the bowtie sensor, was observed with MB₅ labels as a function of antibody concentration. (C) Nonlinear fitting showed good agreement of antibody binding to a Langmuir binding curve. (D) The nonresponsive frequency, f_NR, showed no appreciable change in response to antibody. (E) With the standard MB label, high and low f_SWV ranges showed signal-off behavior, although (F) this response also followed a Langmuir binding curve. Sample matrix was BSA-containing, neutral buffer solution. All x-axes represent antidigoxigenin (antidig) antibody concentration; y-axes for part A and B are normalized signal ratio for signal-off (R_off,norm) or signal-on (R_on,norm) configurations, respectively.

At lower frequency, such as 20 Hz, the MB₅ label gave a previously unattainable signal-on response to the bowtie sensor which was [Ab] dependent (Figure B). A nonresponsive frequency (f_NR = 50 Hz) was also observed as a function of [Ab], as shown in Figure D, which is similar to that observed by others using aptamer-based EC sensors. ,, By analogy to an isosbestic point in spectroscopy, the clear and reproducible observation of a f_NR with the bowtie sensor (Figures D, D&H, and S-6C&F) gives strong mathematical evidence that there are only two bowtie sensor states, unbound or bound, and that these two states do not interfere with one another. More specifically, the data suggests that the signal decay curves over time (as in Figure E) represent a linear sum of only two states, a faster decay state (unbound bowties) and a slower decay state (antibody-bound bowties). Considering the Lovric analysis (details in SI) in Figure D and the temporal data in Figure E-F, it appears that the MB₅ label increased the effective electron transfer rate significantly (compared to MB), such that the current decay is even more rapid when free and slows in a measurable way when antibody-bound. Thus, with the MB₅ label, a time point exists where the temporal decay curves cross each other (Figure E), giving us access to signal-on (low f_SWV), nonresponsive (f_NR), and signal-off (high f_SWV) regions and thereby expanding the informational content available from these sensors.

5.

Thorough characterization of the EC bowtie sensor for testosterone (Tes). Calibration of anti-Tes antibody binding showed (A) a signal-off response at high f_SWV, (B) a signal-on response at low f_SWV, and (C) a good fit to the Langmuir curve using differential analysis to determine antibody coverage (θ). (D) Nonresponsive behavior was also observed at an intermediate frequency (f_NR). Switching the workflow to an indirect Tes sensor showed recovery of the signals at (E) high f_SWV and (F) low f_SWV, and (G) the differential coverage data showed a good fit to our custom mathematical model and gave an LOD of only 200 pM Tes. Measurements in 50% human serum showed 90% recovery compared to the buffer calibration. (H) The nonresponsive behavior was observed here as well. (I) Test set experiments confirmed the model to allow calibration-free measurements. (J) SWV frequency responses with and without antibody-binding showed much higher currents with MB₅ compared to MB, along with (K) a differential response with a nonresponsive point with MB₅ that (L) is not present with the MB label. Note that parts K and L share a y-axis label with part J, and that the time axes were generated using 1/f_SWV for each point. Sample matrix was BSA-containing, neutral buffer solution in all but part G, where 50% serum is labeled as the matrix.

Accordingly, using both signal-off and signal-on responsesa novel feature of the MB₅-labeled bowtiewe could calculate a differential response similar to the kinetic differential monitoring (KDM) method developed by Plaxco et al. ,, Here the differential response, R_on,norm – R_off,norm, was calculated and normalized to the maximum antibody-induced change to provide a plot of surface coverage (θ) of antibody, as shown in Figure C, which nicely followed a Langmuir curve. As described further in SI, these data were fit by nonlinear least-squares fitting to a Langmuir curve, allowing extraction of the parameters for initial amount of surface bowtie structures (Γ_NS,0) and the Langmuir binding constant (K_L).

Further, to evaluate surface distance dependences, we utilized ssDNA spacers as in our related work with other sensors, ,,, where the 3′ end of the thiolated DNA was modified with either four (4A) or six adenines (6A), respectively (Table S-2). Further optimization with polyethylene glycol (PEG) linkers is discussed later in this study in relation to sex hormone bowtie sensors. One way to compare the MB₅- and MB-labeled sensors is to take the ratio of the changes in current induced by antibody binding, as shown in Figure S-1 as a function of frequency. According to these results, our hypothesis relating to a 5-fold improvement is valid at higher frequencies (>400 Hz) with either spacer, where the ratio leveled off near a value of about 5. But at lower frequencies (4 to 400 Hz), interesting behavior was observed. With the 4A spacer, the signal-on response gave negative ratios from 4 Hz to the f_NR of 50 Hz, after which the ratio increased to a positive maximum of around 12, then decreased to the asymptote of about 5. With the 6A spacer, similar behavior was shown, just shifted toward lower frequencies as expected for a sensor spaced further from the surface. The f_NR was 20 Hz for the 6A system, and the antibody-binding current change ratio reached up to 16 at 100 Hz before leveling off just below a value of 5. This behavior suggests that the spacing of the bowtie relative to the electrode surface is a critical feature that defines the frequency dependence of the system. We attribute these effects to the electrochemical diffusion layer thickness decreasing as frequency is increased, passing through a spacer length-dependent region before settling on a spacer-independent region at high frequency. These distance-dependent results should be explored further in future work.

While digoxigenin antibody sensing is not the main purpose of this report, it is noteworthy that the 4A version of this bowtie sensor provided a sensitivity toward antibody binding that was improved by up to 12-fold using MB₅ compared to MB. As a result, the 3σ LOD for antidigoxigenin sensing was 0.08 nM (80 pM) for MB₅ versus 1.3 nM for MB. While the MB₅ label gave new capabilities to the bowtie sensor, the MB label only exhibited signal-off behavior, at either low or high f_SWV (Figure E). The coverage data in Figure F was calculated from the average signal-off response rather than a differential response.

These data on antibody binding to the Dig-modified bowtie sensor provided clear evidence in favor of MB₅ and suggested that the enhanced label should provide significant improvements to other bowtie sensors, whether they are operated as antibody sensors or as indirect analyte sensors. In this study, due to the comparably larger raw currents, the 4A spacer or the equivalent length PEG spacer (PEG₁) was used throughout. Since the bowtie sensor’s signaling mechanism should be mostly analyte-independent, − we hypothesized that novel analytes should be measurable at higher sensitivities by synthesizing custom analyte-DNA conjugates (magenta strand in Figure A) and by using MB₅-labeled DNA as the signaling moiety (brown strand in Figure A). To test this hypothesis, we chose to target clinically important sex hormones as analytes for the EC bowtie sensor, since these molecules are difficult to detect without using enzyme-based, amplified assay formats.

Using Flexible Spacers to Further Improve Sensor Performance

In our previous reports we showed that when the MB label was placed too close to the electrode, i.e. within the EC diffusion layer, lower initial currents were observed. Presumably, this occurs because the diffusion layer can interrupt either the DNA hybridization of bowtie components during structure assembly or directly disrupt the signaling mechanism. Maximized starting current was observed using thiol DNA strands with a spacer of four adenine bases (4A) between the thiol and the rest of the bowtie structure (Table S-2), and further spacing showed reduced currents.

Several of our recent studies , have suggested that adding more flexible spacers made of polyethylene glycol (PEG) into DNA-based structures serves to increase their diffusional mobility and thereby significantly enhance the flow of current in EC assays. , To evaluate this approach in the current work, we thus replaced the 4A spacer (∼2.9 nm long) with a “1-PEG” spacer (internal spacer 18 from IDT; formally 6 PEG units and ∼2.8 nm long) and hypothesized that this would provide an even further enhancement in signal while keeping roughly the same distance from the surface (Table S-2). To explore the spacer’s properties, we also evaluated a 2-PEG spacer (∼5.6 nm).

The 1-PEG spacer gave a significantly higher current for the baseline bowtie signal (not bound to Ab) with both MB and MB₅ labels (Figure S-4A). The MB-labeled bowtie’s current at high f_swv increased by 29%, while the MB₅-labeled bowtie increased by 59%. The 2-PEG spacers resulted in a lower current, near the original levels with 4A, presumably due to delays in completing electron transfer with the longer tether. Upon antibody addition, the percentage decrease was essentially the same for all these systems (MB, MB₅, 1-PEG, and 2-PEG), around 60–70%, suggesting that the PEG spacers were not interfering with antibody binding. With the large increases in current observed, we chose to employ the 1-PEG spacers with all bowtie sensors in the experiments that follow, focusing mostly on the enhanced MB₅-labeled sensors.

Analyte-DNA Conjugate Synthesis and Initial Testing

To construct different bowtie sensors for each of the three analytes, testosterone (Tes), progesterone (Prg), and estradiol (Est) were each conjugated to different batches of the antibody-binding DNA strand (Figure A)as described in detail in the supplementary methods section (Figures S-9 to S-13). With antibody-binding assay systems, careful selection of the bioconjugation site is often crucial, as it can have a significant impact on antibody affinity and recognition, especially in the case of small molecules with fewer functional groups to participate in binding. In Figure A, we highlight four analyte-DNA bioconjugation sites that were evaluated in this study. The selection of the most suitable bioconjugation sites was achieved through custom methodology in synthetic chemistry, which is further detailed in the SI.

4.

Custom synthesis of analyte-DNA conjugates. (A) The four analyte-DNA conjugates shown here were synthesized and tested in this study for development of hormone-responsive bowtie sensors. (B) A custom approach for synthesis of the 17β-estradiol-NHS ester (S4) was developed for site-directed conjugation of DNA, mimicking antibody selection conditions and improving the sensor response.

In initial sensor tests using the position 3 bioconjugation site on testosterone (Figure ), the resultant bowtie sensor exhibited a significant decrease in current upon anti-Tes antibody binding (Figure S-2A), as shown by the examples above in Figure B. This Tes-DNA conjugate was therefore ready to be used in bowtie sensors (more below).

However, the initial sensor constructed with a similar position 3 modified Prg-DNA conjugate showed very little response to 100 nM of antibody (anti-Prg 1) and only about a 40% decrease in current upon adding 500 nM anti-Prg 1 (Figure S-3A). This antibody also showed no evidence of competitive binding with free Prg, thus it was abandoned. After surveying the literature and consulting with the commercial supplier, we found that bioconjugations at the 11th position (Figure A) should be more fruitful, since many other antibodies were produced using haptens from this position. Indeed, with the 11α-Prg-DNA conjugate strand incorporated, the bowtie sensor responded nicely to the new antibody (anti-Prg 2) (Figure S-2B and S-3B), and the sensor was also responsive to free Prg analyte in the indirect immunoassay workflow (Figure S-3B). Interestingly, a third antibody that was tested showed excellent binding to the 11α-Prg-DNA bowtie, with 31% suppression to only 10 nM antibody (anti-Prg 3); however, this sensor was not responsive to free Prg in the indirect workflow (Figure S-3C). Thus, the 11α-Prg-DNA conjugate was chosen and paired with anti-Prg 2 (herein referred to simply as “anti-Prg”) for use as a bowtie sensor for Prg.

Likewise, the bioconjugation position of estradiol (Est) to DNA was selected according to information from the commercial sources (details in SI), where the 17β-Est-DNA conjugate was chosen. A custom approach was developed for synthesis of the NHS ester, S4 (Figure B and S-12), which was used to make the conjugate by coupling with amine-DNA (Figures S-12 and S-13). The first selected antibody (anti-Est 1) at induced a modest suppression of current, however, it showed no observable binding affinity toward free Est even at 10-fold higher levels (Figure S-3D). A second antibody was sourced (anti-Est 2), and adding it to the 17β-Est-DNA bowtie sensor gave a significant signal suppression in response to antibody (Figures S-2C and S-3E). This antibody was also able to function nicely in a competitive immunoassay workflow in the presence of free Est competing for antibody binding sites (Figure S-3E). Therefore, the 17β-Est-DNA conjugate was chosen and paired with anti-Est 2 (herein referred to simply as “anti-Est”) for this version of the bowtie sensor.

Information gained from the synthesis and initial testing of three novel analyte-DNA conjugates highlighted the critical importance of choosing 1) the site of chemical modification in analyte-DNA bioconjugates and 2) an appropriate antibody that binds to the resultant bowtie nanostructure while also providing competitive binding to the free analyte. Below, we expand the testing and validation of these sensors.

Testosterone Sensing with the Bowtie Sensor

The antibody-binding response curve of the MB₅ version of the testosterone bowtie sensor (Figure A-D) was similar to the digoxigenin control system (Figure ), supporting our assertion that the sensors’ responses are largely independent of analyte identity. The key difference was that the anti-Tes antibody exhibited higher sensitivity at low concentrations (Figure A-B) and steeper Langmuir binding curve (Figure C), suggesting either higher affinity or higher initial loading of bowtie structures onto the surface. At high f_swv (464 Hz), the raw current sensitivity was 130 ± 10 nA nM^–1 (absolute value) for anti-Tes. Relative changes, R_off,norm, correcting for electrode-to-electrode variation, showed an absolute sensitivity of 8.9 ± 0.4% nM^–1 alongside a σ_blank = 0.7%.

As with the test system, anti-Tes antibody binding to the bowtie induced a signal-on response at low f_swv (Figure B), as well as a nonresponsive curve at f_NR = 46 Hz (Figure D). Because raw SWV currents are smaller at low frequencies, the raw sensitivity in the linear range was lower at 4.5 ± 0.6 nA nM^–1, although currents remained in the measurable range with the MB₅ label. Using the signal-on response, the relative change, R_on,norm, showed an increase of 10 ± 1% nM^–1 with a σ_blank = 1%.

The signal-off and signal-on responses were then used to calculate a differential response (described above and in SI) and similar to others’ work, ,, and the surface coverage (θ) data is shown in Figure C. The Langmuir fit again allowed extraction of the parameters for initial amount of surface bowtie structures (Γ_NS,0) and the Langmuir binding constant (K_L). Using the coverage data, the 3σ LOD for anti-Tes antibody sensing was determined to be 0.2 nM (200 pM).

Configuring the same bowtie structure as a small-molecule sensor (Figure E-H), we simply modified the workflow into an indirect EC immunoassay, as in our prior work, , by premixing a fixed amount (10 nM) of anti-Tes antibodies with varying Tes analyte samples before adding to the electrode. The response at high f_swv (Figure E) showed a [Tes]-dependent recovery of the signal that was “lost” by antibody binding, where signal change percentage moved from around −80% back toward zero as a nonlinear function of [Tes]. The dynamic range of the Tes bowtie sensor extended up to about 70 nM, with a high f_swv absolute sensitivity of 42 ± 6 nA nM^–1 in the pseudolinear range at low concentrations. As anticipated, a fully nonresponsive curve was measured at an intermediate frequency, f_NR = 50 Hz (Figure H), and the signal-on response at low f_swv returned from about 80% and approached zero with higher [Tes] (Figure F). Surface coverage (θ) was determined from the differential data as before (Figure G). The selection of [Ab]_Total = 10 nM (marked in the figures) was strategic, since the Langmuir based surface coverage during antibody loading began to saturate above this level (Figure C), and the indirect immunoassay could be operated in the nearly linear region below θ = 0.8. Indeed, Figure G shows that addition of increasing levels of [Tes] caused the coverage to drop from about 0.8 and approach zero.

These coverage data were fit to a novel, bowtie sensor binding model curve shown in Figure G. The two-step, uncoupled equilibrium model is based on a free-solution binding model involving antibody and Tes, followed by a Langmuir based surface binding model using the remaining antibody ([Ab]_Rem), and it is described in more detail in SI (Equations S1 to S4). From the experimental data, we extracted a bowtie surface loading (Γ_BS,0) of 0.46 pmol and an antibody-Tes dissociation constant (K_D) of 4.9 nM. The LOD for Tes sensing was determined to be 0.2 nM (200 pM). From this model, we could derive an expanded expression relating [Tes]₀ (initial concentration) to the measured coverage (θ) in a calibration-free manner; the expression is given as Equation S4. By fitting the data to a more comprehensive training set (shown in Figure G), we were able to apply Equation S4 from the model to a group of test measurements from three different spiked Tes concentrations, and the calculated [Tes]_model values were well-correlated with [Tes]_spiked (Figure I). These data confirm that our sensors can operate in a calibration-free manner, akin to aptamer-based sensors reported by others. ,,

For comparison to the frequency- and time-dependent responses of the digoxigenin bowtie sensor (Figure ), the data in Figure J-L shows the differences between the MB and MB₅ versions of the testosterone bowtie sensor. Once again, the MB₅ label resulted in much higher raw currents (Figure J), along with introducing both signal-on and signal-off sensor responses and a nonresponsive frequency (Figure K) compared to the exclusively signal-off response shown by the single MB label (Figure L). Here again, the ability to measure on and off responses is the key element in permitting differential data processing, using a custom approach modified from the powerful KDM approach. ,,

Testosterone Sensing in Human Serum

Recalling that the normal human male has a total serum testosterone level in the range of 9.0 – 34.7 nM, while females have a range from 0.52 to 2.43 nM, the calibration against [Tes] in Figure G showed that the Tes-DNA-MB₅ bowtie sensor could already operate well within the human clinical ranges (shaded in the figure). The sensor exhibited an LOD of 0.2 nM (200 pM) with a dynamic range of over two logs up to 70 nM, and the operation in 50% human serum was only slightly attenuated compared to buffer (dark red squares in Figure G). One important implication of this result is that it clearly demonstrates high selectivity of the sensor toward Tes, because the blank was easily distinguished from spiked Tes in the highly complex background of human serum. This clinically relevant bowtie sensor detection range was achieved with a relatively fast, direct binding EC sensor without enzymatic amplification, offering a greatly simplified workflow compared to the standard ID-LC-MS/MS method.

In a real-world clinical setting, however, Tes sensing can be more complex than that of simple calibration standards in buffer. In human blood, much of the testosterone is bound to sex hormone-binding globulin (SHBG) and human albumin. In men, 44–65% of testosterone is bound to SHBG, 33–54% is bound to albumin, and only 1–2% of Tes is free. In women, 66–78% is bound to SHBG and 20–32% is bound to albumin. The free and albumin-bound fractions have been shown to be metabolically active, and their sum gives the “bioavailable testosterone,” which is correlated with the level of free Tes. Since clinical samples contain plasma proteins, it is important to evaluate Tes sensors’ response in relation to SHBG and albumin levels.

In essentially all measurements shown thus far, our sensors have been operating in the presence of a relatively high concentration of bovine serum albumin (BSA), thus the antibody-Tes interactions have already been shown to overcome the lower affinity albumin-Tes interactions. To study the effects of SHBG, we evaluated the bowtie sensor’s response to equal amounts (10 nM) of anti-Tes, SHBG, and Tes (Figure S-5). These data show that the SHBG affinity is high enough to prevent antibody binding to Tes, at least on the time scale used in our workflow. The sensor should therefore be responsive to bioavailable Tes (free + albumin-bound) rather than the total level of Tes (free + albumin-bound + SHBG-bound). Interestingly, we show that estradiol (Est) can be added as a competitor at a 10-fold higher level (100 nM) and effectively displace the Tes from SHBG, suggesting a means for measuring either bioavailable or total testosterone by a simple modification to the workflow, both of which are clinically important. Additional studies in this area are warranted in future work, where validation of the bowtie sensors with the gold standard ID-LC-MS/MS method would be ideal.

As shown by the additional data in Figure G, the Tes bowtie sensor was calibrated in 50% human serum, where steroid-free serum was spiked with known concentrations of Tes. The 50% serum response curve was similar to the buffer calibration response, with a linear range sensitivity of 2.0 ± 0.2% nM^–1 just a bit lower, representing a 90 ± 10% recovery of the 2.2 ± 0.2% nM^–1 response observed in buffer. While this recovery is within error of 100%, it is possible that the presence of some steroid-free SHBG in the serum samples bound to added Tes, resulting in artificially lower Tes concentrations compared to what was expected. Nonetheless, the 90 ± 10% signal recovery and the agreeable sensor response in human serum was encouraging for future applications of these sensors, matching what we have observed in previous reports. −

Estradiol Sensing with the Bowtie Sensor

Presumably, a generalizable sensor should be capable of switching to different analytes by following a similar strategy. After preliminary screening of conjugation sites and antibody binding responses of the sensors (see Figure and discussion), this approach was used to design and evaluate novel EC sensors for both estradiol (Est) and progesterone (Prg).

The selected anti-Est antibody was paired with the appropriate bowtie sensor made with Est-DNA conjugates and evaluated accordingly (Figure A-D). As before, the antibody calibration showed a Langmuir response curve (Figure A), and with a measured sensor surface coverage of just over 80%, 10 nM anti-Est was chosen for the indirect immunoassay workflow. The LOD for anti-Est was determined to be 0.6 nM (600 pM). In the indirect assay workflow (Figure B), the sensor’s response was sharper at low concentrations, indicating higher affinity of binding and/or lower initial surface loading of the bowtie structures, with a dynamic range extending up to about 15 nM Est. Additional data showing changes at high f_SWV, low f_SWV, and at the f_NR are shown in Figure S-6. Even with the varying shape of the response curve, our custom binding model again was able to nicely to the data (Figure B) when using the MB₅ label. Indeed, the model suggested that the average bowtie surface loading (Γ_BS,0) was lower at 0.22 pmol and that the antibody-Est binding K_D was 0.07 nM, explaining the sharper sensor response curve. These experiments showed an LOD of 0.8 nM (800 pM) for Est sensing. While this value does overlap some of the clinical range of serum Est in premenopausal women, the full range is lower, from 0.11 to 1.47 nM.

6.

Two additional hormone sensors were made using the same core MB₅-labeled DNA bowtie design by substituting a custom-synthesized analyte-DNA conjugate (left) into the DNA nanostructure (bottom left) on the electrode. (A) Anti-Est antibody binding followed a Langmuir curve, and (B) coverage data versus Est concentration showed a more sigmoidal response, still fitting well to the model. (C) Comparative data with the typical MB label showed high variance that was absent with MB₅. (D) Calibration-free Est sensing (open circles) was validated using our model and matched closely with the ideal results (line). (E) A simpler characterization of Prg antibody binding still gave (F) a good model fit for the indirect EC immunoassay for Prg, and (G) calibration-free sensing was validated with Prg as well. Sample matrix was BSA-containing, neutral buffer solution.

The lower surface loading could have resulted from lower yield of the Est-DNA conjugation, or perhaps more likely, lower yield of the surface-based enzymatic ligation step. Interestingly, a result was that the estradiol bowtie sensor gave an example in which the MB₅ label truly was enabling to improve the robustness and consistency of the sensor’s response. As shown in Figure C, since raw currents were low, the relative errors were high when using the standard MB label. While the sensor was responsive in a way that fit the model, and it showed a similar dynamic range, the high variance with the typical MB label rendered the calibration unreliable. Comparatively, the MB₅ label showed much more consistent sensor responses and low variance (Figure B).

Again, using our custom model defined by Equations S1 to S4, an expanded expression was derived and used to calculate [Est]₀ directly from the measured coverage values (θ) in a calibration-free manner. This modified expression is given by Equation S4, where [Est]₀ was substituted in place of [Tes]₀, and other parameters were used as derived from the estradiol sensor characterization. Three test sets of data were shown to correlate nicely with the ideal response (Figure D), confirming that the mathematical model also enables calibration-free measurements of estradiol with the EC bowtie sensor. Higher errors observed with the 1 nM Est test value were reasonable, considering that the model fit showed a pseudosigmoidal shape and was moderately flat at very low concentrations.

Progesterone Sensing with the Bowtie Sensor

The same core sensing system was evaluated for measurement of a third clinically relevant, small molecule hormone, progesterone (Prg). After choosing the appropriate 11α-Prg-DNA conjugate and antibody, the bowtie sensor was constructed and characterized as before (Figure E-G). To evaluate a simpler sensor development workflow, we chose to test only three anti-Prg concentrations in the antibody binding study (Figure E). In this case, the average surface loading of sensor structures was apparently higher, with a Γ_BS,0 value of 1.2 pmol extracted from the model fitting. As a result, the use of 10 nM anti-Prg in the sensor workflow showed lower fractional coverage (θ) of around 44%. To maintain direct comparisons to the other two sensors, we chose to keep the 10 nM antibody concentration in the indirect sensor workflow, a value which could be optimized in future studies using the model as a guide.

Figure F shows that the generalizability of the bowtie sensor was verified once again, with the sensor’s response fitting the model curve nicely and exhibiting an LOD of 1 nM for Prg sensing. The dynamic range extended to about 40–50 nM, thus the sensor already nearly covers the full clinical range of females in the luteal stage of the menstrual cycle (6.4 to 79.5 nM). Pharmacological hormone therapies can increase the levels further, and in pregnant women, the serum Prg levels can exceed 900 nM. Because the sensor shows an obvious saturation level, dilution could be used to analyze such samples in future applications. Serum levels in females in the follicular stage of their cycle are lower (0.32 to 2.23 nM), thus the sensor would need about a 3-fold LOD reduction to encompass this range.

Additional data showing changes at high f_SWV, low f_SWV, and at the f_NR are shown in Figure S-6. The calibration-free method was also validated with this third sensor (Figure G), and the resultant model-calculated values were closely matching with the ideal, expected values. Note that the higher error observed at 15 nM in the calibration-free calculation is reasonable, since the calibration curve in Figure F begins to flatten and approach saturation near 15 nM.

Study on Cross-Reactivity

Finally, considering the structural similarities among the three steroid hormones that we have developed sensors for in this study (Tes, Prg, Est), it was important to evaluate whether each sensor would exhibit cross-reactivity by binding antibodies selected against the other hormones. Generally, even with commercial immunoassays, some cross-reactivity can be observed and may need to be mitigated in bioanalytical methodology. As shown in the cross-reactivity matrix image in Figure S-14, all matched antibodies exhibited the maximal binding and signal response to their targeted bowtie sensor, as expected. Minimal to zero cross reactivity was observed in all tests against mismatched antibodies, except for the significant 62% cross-reactivity observed with the anti-Prg antibody partially binding to the Tes-DNA bowtie sensor. Interestingly, this should not be problematic for male patients, who have higher Tes levels and lower Prg levels in their blood. However, this issue could create problems for the Tes-DNA bowtie sensor for use in female patients, who have lower Tes levels and fluctuating Prg levels which can reach much higher than Tes in their blood. For clinical sample preparation, in this case, it would be advisible to selectively remove Prg, i.e. with an immunoaffinity column, prior to measurement of Tes in female samples with this sensor. Otherwise, the three bowtie sensors should be usable for measurement in human serum samples with minimal sample preparation, as exhibited in Figure and corroborated in our other studies. −

Conclusions

From the bioanalytical perspective, this work further confirmed the modularity and generalizability of our EC bowtie sensor, while also introducing several new features which significantly improve the sensor’s performance. Custom synthesized analyte-DNA conjugates were used to create unique adaptations of the sensor toward small molecule hormone measurement. Flexible PEG surface spacers gave faster tethered diffusion and enhanced electron transfer rates, as shown in our earlier work with other systems. , Perhaps the most broadly impactful development was the introduction of the novel MB₅-DNA label. This label greatly increased raw currents and reduced sensor variance, enabling picomolar range sensing of several small molecule hormones with a simple, binding-based EC sensor, i.e. without any enzymatic or chemical amplifications. Considering the enhancements observed herein, one implication is that the MB₅ label should be useful to enhance the responses of various other EC sensors that leverage DNA structures, such as aptasensors, ,, E-DNA scaffold sensors, ,, steric hindrance assays, , or molecular pendulum sensors.

The bowtie sensor’s response was also enhanced mechanistically by MB₅, allowing signal-on, signal-off, and differential sensing for the first time with this platform. Leveraging this robust response, a new mathematical model was developed, and calibration-free sensing was achieved for the first time with the bowtie sensors. In the future, each sensor’s response could be further tuned with the mathematical model, and antibody concentrations could be optimized to shift the sensor’s optimal sensitivity to either low or high concentration ranges, depending on the application needs, after correcting for surface loading and binding affinities.

In the realm of clinical sensing, this work introduced a new and simpler means of quantifying testosterone, estradiol, and progesterone. The bowtie sensor’s workflow is greatly simplified compared to the gold standard ID-LC-MS/MS method, and it remains simpler than enzyme-enhanced immunoassay formats. Studies on SHBG’s interactions with testosterone showed that our sensors could be reconfigured to measure either bioavailable testosterone, total testosterone, or both. It is important, however, that future work focus on validation of the bowtie sensor with the gold standard ID-LC-MS/MS or with other immunoassay systems in analyzing real clinical samples, and studies on the effects of SHBG for other hormone sensors should be conducted. Fundamentally an EC measurement, the bowtie sensors could feasibly be miniaturized, multiplexed, and made portable in the future to adapt to point-of-care needs.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmeasuresciau.6c00042.

• DNA sequences, supporting experiments and methods, details of the custom equilibrium model of the bowtie sensor’s mechanism, synthetic method development details, cross-reactivity studies, and references (PDF)

The manuscript was written through contributions of all authors, and all authors have given approval to the final version of the manuscript. CJE and RRK contributed to the conceptualization, data curation, funding acquisition, formal analysis, project administration, supervision, and writing of the work. AG, MIM, KGO, and ASNK participated in data acquisition, investigation, methodology development, visualization, and writing. SS was involved in conceptualization, methodology development, and writing.

The authors declare no competing financial interest.