A microwell-based impedance sensor on an insertable microneedle for real-time in vivo cytokine detection

Impedance-based protein detection sensors for point-of-care diagnostics require quantitative specificity, as well as rapid or real-time operation. Furthermore, microfabrication of these sensors can lead to the formation of factors suitable for in vivo operation. Herein, we present microfabricated needle-shaped microwell impedance sensors for rapid-sample-to-answer, label-free detection of cytokines, and other biomarkers. The microneedle form factor allows sensors to be utilized in transcutaneous or transvascular sensing applications. In vitro, experimental characterization confirmed sensor specificity and sensitivity to multiple proteins of interest. Mechanical characterization demonstrated sufficient microneedle robustness for transcutaneous insertion, as well as preserved sensor function postinsertion. We further utilized these sensors to carry out real-time in vivo quantification of human interleukin 8 (hIL8) concentration levels in the blood of transgenic mice that endogenously express hIL8. To assess sensor functionality, hIL8 concentration levels in serum samples from the same mice were quantified by ELISA. Excellent agreement between real-time in vivo sensor readings in blood and subsequent ELISA serum assays was observed over multiple transgenic mice expressing hIL8 concentrations from 62 pg/mL to 539 ng/mL.

It is noted that the Rsolution resistors changed between the antibody immobilization and protein detection steps as listed in Supplementary Table 1. This change could be due to multiple factors, including small differences in the ionic strength of various solutions applied to the sensing region during different steps. However, if this occurs, one might expect Rsolution01 and Rsolution02 would change by the same percentage when the sensor was interacting with different solutions. In general, this was not observed. We attribute this to the observation that Rsolution01 in this sensor geometry contributes negligibly to sensor impedance in comparison with Cox02 and Cox03 within the utilized EIS frequency range. As a result, the circuit model fitting is not strongly dependent on Rsolution01. A perturbation analysis was performed to verify this, wherein we fixed the Rsolution01 to increase by the same percentage as Rsolution02 (e.g., for one data set, an increase of 8%, from 29.7 kOhm to 32 kOhm). Fitting results without and with fixing the percentage change in Rsolution01 during the target protein detection step are represented by solid lines and dashed lines, respectively in Supplementary Fig. S1 and parametric values corresponding to the circuit components are summarized in Target

Operating frequency determination
To illustrate the impedance behavior of sensors as biological binding events occur, and to determine a suitable operating frequency range for the sensor for real-time impedance monitoring, the equivalent circuit model was modified based on the parametric values from the above fitting results. The circuit branch comprising the outside sensing region is simplified to a single capacitor Cox01 as the impedance represented by Rsolution01 was found to be negligible in comparison; Rct01, Rct02 are removed as they are of many orders of magnitude larger compared to the impedance from CPE01 and CPE02 within the range of EIS frequency as from 1kHz to 1MHz. Also, both constant phase elements have a n value that are close to 1, thus they are substituted with two double layer capacitors Cdl01 and Cdl02. The simplified circuit model is shown in Supplementary Fig. S2 The parametric values of circuit elements for this simplified model resulting from the lumped element combinations described above, as well as from a fit of the simplified model directly to the data, are summarized in Supplementary Table 2. Reasonable agreement between these two methods of estimation is observed, further validating the assumptions underlying the simplified model.

Supplementary Fig. S2
A physical representation of the simplified equivalent circuit model after the immersion of the label-free sensor into electrolytes for impedance behavior analysis.
The basic principle of this impedance sensor is that the biological binding events occurring inside microwells will influence ion transport between the two electrodes, as well as alter the dielectric properties of the electrolyte on the gold surface. As a result, the impedance behavior of the sensor will change, and real-time monitoring of the target protein can be achieved by continuously capturing this change. For example, referring to the simplified circuit model of Supplementary Fig. S2, it is expected that protein absorption would result in an increase in Rwell and a decrease in Cdl02 (as εprotein is ~ 20 while εPBS is ~ 80). It is further expected that these phenomena may be more or less dominant at different operating frequencies. To obtain an operating frequency region that leads to a maximum change in impedance along with the change in Rwell and Cdl02, the impedance behavior of the simplified model is analyzed utilizing the parametric values of circuit components from the fitting results as summarized in

Supplementary Table 2.
The total impedance between the gold electrodes as predicted by the simplified model of Supplementary Fig. S2 can be expressed as: To assess the ability of this simplified circuit in modeling the electrochemical properties of the sensor, a representative set of EIS data were fitted and analyzed. The EIS data were acquired when anti-hIL8 functionalization and specific binding of hIL8 to pre-immobilized anti-hIL8 were occurring within microwells.
Each step included ten EIS data points within a time period of ten minutes. The changes of Zreal at 100 kHz and Zimag at 1 MHz during antibody immobilization and target protein detection were subsequently obtained from these EIS data as shown in Supplementary Fig. S3g-j. The fitting routine used in this study followed a two-step procedure. First, the measured impedance data when the sensor was initially immersed in PBS were fitted to the circuit and the values of corresponding circuit components were determined (Supplementary Table 2). Of these, Cox01, Cox02, Cdl01, and Rsolution were assumed to be largely unaffected by biological events inside microwells. Next, the impedance data obtained during biological binding events occurring inside microwells were fitted to the model  Fig. S3e), Rwell showed an increment of 14.1 k and Cdl02 decreased from 125 pF to 34.5 pF. As discussed above, this is consistent with the proposed basic principle of the microwell sensor; i.e., the affinity of protein on the gold surface inside the microwells would block ion transport across two electrodes as well as alter the dielectric properties of the gold-electrolyte surface. When target protein solution was introduced to sensing wells, specific antibodyantigen binding occurred and an increase of 7.5 k in Rwell was observed, as shown in Supplementary Fig. S3f.
During this step, Cdl02 did not change as much as when anti-hIL8 was first introduced on to the gold electrode; we hypothesize that this could be due to the smaller size of the hIL8 protein molecule compared to its antibody. One potential explanation for the increase in Cdl02 (Supplementary Fig. 3f) post addition of hIL8 solution might be the detachment of the anti-hIL8 from the electrode surface.
In addition to mouse sera tests, in vitro experiments with laboratory-prepared samples, which contained a second type of cytokine protein (i.e., nontarget protein), as a negative control instead of blank PBS buffer were also conducted to assess sensor selectivity. As show in Supplementary Fig. S5b, the sensor was first pre-immobilized with anti-hIL8. With the introduction of nontarget protein (IL6 solution), the sensor had a negative response just as if blank PBS solution had been added. When adding the target protein solution (hIL8), the sensor responded with an increase in the Zreal, as expected. (We note that this particular sensor had a different electrical interconnect geometry (Supplementary Fig. S5a), but shared the same microwell working principle.) These results were repeatable over different protein resources, further supporting sensor specificity in vitro.
Supplementary Fig. S5 a Microscopic image of sensor sensing region with an array of microwells. b A summary of representative sensor responses from an experiment using IL6 solution as a negative control. Data were collected using a potentiostat at 100 kHz. Similar to blank PBS buffer, when sensor was interacting with IL6 solution, a drop in Zreal was observed. The sensor responded with an increment in Zreal following the addition of target antigen (hIL8) solution.

Effect of temperature on sensor sensitivity
Temperature plays an important role in molecular transport and reaction. Regarding temperature effects on detection results, some insight can be gained by examining the difference between the sensor sensitivity (as assessed by the slope of the regression line against ELISA assessments) in in vitro (room temperature, 23 ℃) and in vivo (mouse body temperature, 37 ℃) experiments as shown in Supplementary Fig. S6. It is noted that at the higher operating temperature (i.e., in vivo), a slightly higher sensitivity (i.e., slope of normalized percentage change in Zreal vs hIL8 concentration assessed by ELISA) was observed. There are several potential mechanisms to explain this sensitivity effect, such as temperature dependent biomolecule delivery rate and antigen binding reaction rate, both of which would lead to higher sensitivity at higher temperature. For in vivo application of the sensor, this temperature effect is somewhat ameliorated by the constant temperature environment of the body; but if the sensor is to be used in environments where the temperature is not constant, temperature compensation schemes may be considered.