Is this the Simplest Possible Biosensor? A Surface Imprinted Polymer EIS Sensor for Detecting Alpha-synuclein, a Parkinson’s Disease Biomarker

2.1. Device Fabrication

Stamp: 400 nm of polyvinyl alcohol (PVA) is static deposited spin coated from a 10 wt.% solution of PVA in chloroform onto a glass slide. The stamp was prepared by dropping 50µL of 1 mg/mL αSyn in deionized water (DI) onto a 0.5cm² area and dried at room temperature for 2 hours (Figure 1a).

PCL SIP: Kapton substrates (500 EN, Dow Chemicals, USA) are patterned with 100 nm of Aluminum (Al) and 100nm of chromium (Cr) using standard lift-off photolithography techniques. The interdigitated electrodes (IDE) with an area of 25mm², with 20 fingers and a measured spacing of 102 µm and width of 153 µm (Figure 1b). A 600 nm thick layer of PCL is deposited on the IDEs by dynamic deposition spin coating (at 6000 rpm). The stamp is placed in contact with the PCL surface, heated to 60⁰C and pressed using a 200g weight for 2 minutes. The structure is then removed from the heat and allowed to cool. The PCL surface and stamp are submerged in water allowing the PVA stamp polymer to dissolve, releasing the EIS device without damage. The SIP EIS biosensors are washed with 0.5 mM ascorbic acid and deionized water (DI) to ensure removal of the stamp biomolecules prior to testing.

PCL microfluidic channel: A microfluidic channel was made by melting PCL into a mold (Figure 1c). The well depth was 2 mm deep, and luer lock tubing was melted into the microfluidic channel for simple sample loading and waste removal. The microfluidic channel was then adhered to the PCL microfluidic channel using chloroform as a solvent. The PCL of the microfluidic channel and the surface forms an excellent seal after the chloroform off gases, leaving an integrated sample handling method.

2.2. Sample Preparation

αSyn in DI test samples were created by serial dilutions of from 10 mg/L of dried αSyn material in DI. 10-fold serial dilutions were produced from 10 mg/L to 100 ag/L. Constant ionic concentration solutions were created using αSyn and ßSyn in varying ratios to produce solutions of serially decreasing αSyn solutions, but with a constant total ionic concentration. The concentrations of αSyn in these solutions was 10-fold dilutions from 50 µg/L down to 100 ag/L.

αSyn monomer and ßSyn were supplied by the LADDER group in Chemistry Department, Carleton University. To avoid unwanted aggregation of the material, all materials were stored at -20⁰C when not in use, and vortexed prior to use.

2.3. Testing Processes

Impedance analysis is performed using an Agilent 4294A impedance analyzer (Fig 1d). The impedance magnitude and phase angle are collected during a logarithmic frequency sweep from 40 Hz to 100 MHz with an amplitude of 500mV. During testing, 10µL of sample is incubated on the surface for 1 minute prior to testing. Each data collection was repeated three times. The surface is then rinsed with DI, followed by 0.5 mM ascorbic acid, and a final DI rinse and N₂ drying to ensure all material is removed from the surface between tests. The device is then ready for the next test.

The impedance magnitudes were converted to the real (Z’) and imaginary (Z”) components and plotted as a Nyquist plot for parameter extraction during sensor optimization. Experimental data was analyzed using MATLAB.

2.4. Atomic Force Microscopy of Soft-Printed SIP Surface

SEM was performed with a Tescan Vega-II XMU VPSEM. Figure 2a shows the SEM topography of a 238.1 µm by 238.1 µm scan of a SIP on EIS electrode post testing and regeneration. The scale-like appearance of the PCL is a factor of the heat-melt process involved in the stamping process. The important factor here is the size of the crystals formed. Without the presence of a stamp, we observe crystals on the scale of 100 µm – mm. In the presence of the stamp, we observe significantly smaller crystals (scale of 2-5 µm) formed by the stamp protein acting as nucleation points. Clear cavities are observed well distributed across the surface. These are the binding sites of the PCL SIP. Based on the partial specific volume, calculating the approximate volume occupied by a protein of mass M (kDa) is volume (nm3) =1.212*M, giving a volume of 17.5nm3, so assuming a globular protein the diameter should be 2.78nm for αSyn monomers [26]. We expect the surface cavities to be in this range for single αSyn monomers, which we further examined using atomic force microscopy.

Atomic Force measurements took place in air using a Veeco Dimension 3100 AFM in tapping mode with a silicon probe tip. Nanoscale AFM lateral resolution is dependent on tip sharpness and profile, lateral feature size is inflated for adjacent particles or rough surfaces. z dimension deflection is a reliable indicator of feature size. Figure 2b shows the AFM topography of a 1 µm by 1 µm scan of a PCL SIP post testing and regeneration. The largest surface cavities have depths of 9.2 ± 5.5 nm, with the smallest cavities down to a few nm. This indicates that there is some agglomeration of αSyn. Figure 2c shows the AFM topography of a 1 µm by 1 µm scan of an αSyn Stamp with lyophilized αSyn on the surface. The z dimension sizes of molecules were between 3 - 9.2 nm. Material size variation is observed on the slides. This confirms that there is some anticipated agglomeration. Thus, the imprints on the PCL SIP shown in figure 2b were consistent with the size of lyophilized material on the surface of the stamp. We observed surface cavities consistent with effective stamping of the PCL surface.

3.1. Impedance Spectroscopy Data Analysis

The SIP EIS biosensor impedance response for five separate devices was repeat tested n=3 times with each test solution. Given that the EIS biosensor tests electrolytes, it is suitable to analyze the impedance response data with a Randles-Ershler equivalent circuit model (Figure 3a) [27]. The expected graph shape from the Randles-Ershler Nyquist plot is shown in Figure 3b. This is a basic model that is applied to both faradaic and non-faradaic EIS biosensors. Faradaic biosensors are defined as having a redox species that generates charge. Non-faradaic biosensors do not rely on charge generation and are generally label free. It is important to note though that there is not necessarily a direct correspondence between circuit elements and underlying physical processes; for example, the simplified Randles-Ershler circuit model lumps the entirety of the sensing mechanism processes into a single element C_G.

There are 4 main parameters, R_S or solution resistance, C_G or geometric capacitance, Z_W the Warburg element and R_CT, the charge transfer resistance. The prevalence of the elements is dictated by the device architecture and materials. Solution resistance (R_S) is dependent on the finite conductance of ions in bulk solution. Therefore, it is affected by concentration, but not by binding processes. The Warburg impedance, Z_W is usually physically insignificant in non-faradaic biosensors, as it is a delay arising from the diffusion of electroactive species to the electrode. Thereby, it only has an appreciable effect at low frequencies, and is affected by convection. The ideal Warburg element has a phase shift of -45⁰. R_CT captures two effects, the energy barrier to redox species (caused by electrostatic repulsion or steric hinderance) and the overpotential. In non-faradaic EIS biosensors, it also models the leakage current from imperfect insulator dielectrics.

C_G is the capacitance between the electrodes and the electrolyte solution. It can be modelled as a series of capacitances including surface insulators, double layer capacitances and surface modifications. The electric double layer is created by the alignment of charged materials in solution to electrodes of opposite charge. Thus, electric fields in ionic solutions decay exponentially because the alignment of ions negates the effective field. The length of this decay is called the Debye length and is proportional to the square root of ion concentrations. Another contributor to the C_G is the adsorbed molecules on the surface. In the absence of charge production, C_G is the dominant capacitance term. The C_G also contains a constant phase element that dominates at low frequency that can account for the complex double layer capacitance of the remaining fluid on surface, adsorbed molecules, and porous surface structures.

$$Z\left(\omega \right)=\frac{V\left(t\right)}{I\left(t\right)}=\frac{V_0\sin \left(\omega t\right)}{I_0\sin \left(\omega t+\phi \right)} Z^{\prime }=\left|Z\right|cos\phi Z^{″}=\left|Z\right|sin\phi$$

(1)

$${\mathrm{C}}_G=\frac{1}{{\mathsf{\omega }}_{\mathrm{peak}}{\mathrm{R}}_{\mathrm{CT}}} Z_W=\frac{1}{\omega \left|Z\right|}$$

(2)

In the ideal situation of non-faradaic biosensors, R_CT would be theoretically infinite as no charge would be crossing the perfect insulator. However, due to the polarizability of polymers and confirmational changes of materials R_CT is finite. Under these conditions, the imaginary portion of the impedance is inversely proportional to the electrical double layer capacitance [28]. This creates the incomplete semicircular shape with the slow transition to the linear behavior even in non-faradaic biosensors. This deviation from the ideal can be attributed to surface non-uniformity, roughness, and potentially porosity. These kinds of surface effects can create sub-microscopic areas each with a unique resistance-capacitance contribution to the overall behavior. Parasitic impedances and frequency dispersion—transformation of dielectric response from one mode of polarization to another—are usually described in the Z_W [29]. As the sensing mechanism is from the change in near surface effects and particularly the double layer capacitance in the geometric capacitance, repeatable C_G extraction is imperative. Using the Randles equivalent circuit model and parameter extraction methods has been established in literature for approximating the non-faradaic biosensors.

The x intercepts of the semicircle represent the contact resistance (R_s) and surface resistance (R_ct). Eqns. (2) are used to extract the C_G and Z_w, with the tail slope used to determine α, the phase change of the constant phase element. The peak frequency value is used to determine C_G, the geometric capacitance. MATLAB is used to identify the high corner frequency of the Nyquist plot and calculate the R_S and R_CT. Real data is shown in Figure 3c for device 5 tested with a 100 pg/L αSyn constant ionic concentration solution.

Figure 3d shows unfiltered data for dilutions of αSynuclein in DI, showing non-ideal Randles-Ershler impedance output curve for non-faradaic electrolytes under AC. The data trend shows a smaller semicircular curve and lower maximum real and imaginary capacitances for increasing concentration. There are two distinct behaviors that contribute to the shapes of the graph: increasing concentration and increasing binding. As R_S is a factor of solution concentration, it will decrease with increasing concentration of charged materials. With the increasing concentration, there is an increasing contribution to C_G from an increasing double layer capacitance, with decreasing capacitance from an increase in binding to the surface. There are multiple explanations for the decrease in capacitance with increasing binding. It could be that the presence of proteins changes the conductivity in the near surface region, the binding interrupts the formation of the EDL, or it could change the surface energy of the insulator [30]. The effect can be seen in the decreasing size of the semicircular portions of the Nyquist plots, and the increasing impedance with increasing concentration. In order to observe the effects of binding alone, the ionic concentration of solutions was kept constant.

3.2. Characterizing Sensor Performance in

Figure 4a shows real, unfiltered Nyquist plots showing the concentration dependent change in C_G for our PCL SIP EIS biosensors tested in a constant ionic concentration environment with varying concentrations of αSyn. The solutions all have a total synuclein protein concentration of 100 µg/L, but with a decreasing ratio of αSyn to ßSyn. The purpose of testing only in a constant ionic concentration environment is that these devices do have a non-specific response to electrolyte concentration. ßSyn is a homologous protein to αSyn that is structurally different, making it ideal as a control. As the ionic concentration remains constant, the capacitance change will be from increasing binding.

Plotting the Nyquist data clearly shows that not only is there a change in peak frequency (a dependent variable of C_G) but the R_S is clearly increasing with concentration. This is a good indicator that there are indeed changes occurring at the surface of the biosensor. As these devices are tested in an aqueous environment under an applied bias, the electrolyte forms an electric double layer. The freely moving materials in the electrolyte align themselves to the surface. The effective thicknesses of these layers are on the angstrom to nm level. As the non-Faradaic EIS biosensors do not have any charge transfer the biosensing mechanism is due to changes in the electric double layer capacitance. With αSyn binding into the stereo cavities, the development of the EDL is interrupted which decreases the capacitance. The impact of binding is great because the nm and angstrom scale thicknesses of the EDL layers make the EIS SIP sensitive to near surface interactions.

The concentration to percent change in geometric capacitance is shown in Figure 4b. The data was averaged across the 5 different sensor devices, with a 95% confidence interval. Parameters were extracted from the plots using the Randles Erschler equivalent circuit model and fitted to a four-parameter logistical curve. Data is normalized using the 500 fg/L limit of detection test completed on each device prior to experimental data collection to allow for comparison between tested SIP EIS devices. LoD was determined by linear fitting using the standard method using LoD = 3.3(Sy/S) where Sy is the standard deviation of the sensor response (Sy) extracted using linearly fitted data and S, is the slope of the sensor calibration curve. In a simplistic estimation, concentrations that deviate by more than three standard deviations are considered outside of the linear detectable range. The biosensor has a linear range of 5 pg/L to 5 µg/L, with a LoD of 5 pg/L. The clear concentration-dependent geometric capacitance shows that the simplistic soft-printed SIP fabrication process is sufficient for quantifying minute changes to αSyn monomer concentration. Re-usability of our devices was investigated further by repeatedly testing the baseline geometric capacitance where we found a low standard variation of 7.2 % (n=12), indicating that the dilute acid wash and DI wash are effective in removing bound targets from the template between tests.

3.3. Preliminary Data from Microfluidic Channel SIP EIS Biosensor

The final test was to create an EIS SIP biosensor device with the sensing area enclosed within a PCL microfluidic channel, with luer lock interconnects for efficient sample handling. The test volume of the PCL microfluidic channel device was kept at 100 µL. Repeating the same constant ionic concentration testing as with the previous device (shown in Figure 4) produced the data shown in Figure 5a,b.

The device was tested in the linear range established with the open face biosensor. The EIS SIPs showed the same linear behavior and range enclosed in a microfluidic channel as when tested on an open surface, a highly desirable outcome (Figure 5b). It also highlights a source of work for the future, as enclosing the sensor in the microfluidic channel affects the change in geometric capacitance. The percent change in C_G has decreased over the same linear range when the microfluidic channel was enclosed. Another significant change is the Nyquist plot shape. The absence of the low frequency linear range can be explained by the enclosed channel minimizing dielectrophoretic droplet spreading. Investigating the bulk effects from the microfluidic channel will be the next stage of device development.

The SIP EIS is therefore a promising biosensor device for detection of biomolecules and the device fabrication process is simple and facile, making it ideal for large-scale manufacturing and rapid prototyping. Not only are our biosensors capable of detecting αSyn in the dilute levels present in saliva, but the nature of non-invasive testing makes the devices desirable [31]. Saliva has fewer risk factors than serum or cerebral spinal fluid and can be repeatedly sampled [32]. Furthermore, the sensor can be easily interfaced with off-the shelf portable EIS readers making it point-of-care ready. We have previously reported a PCB-integrated EIS sensor for portable and quantitative analysis of 8-Isoprostane in exhaled breadth [14]. Integrating the proposed SIP EIS in a similar way will be adopted in the next phase of platform testing and validation.

Future work will focus on improving the repeatability of stamp production to reduce device variability. The binding between SIP and target biomolecule is dependent only on steric forces, which in a complex media such as whole blood, serum or interstitial fluid biological samples would be greatly influenced by high energy media components.