A Versatile Biomimic Nanotemplating Fluidic Assay for Multiplex Quantitative Monitoring of Viral Respiratory Infections and Immune Responses in Saliva and Blood

Abstract The last pandemic exposed critical gaps in monitoring and mitigating the spread of viral respiratory infections at the point‐of‐need. A cost‐effective multiplexed fluidic device (NFluidEX), as a home‐test kit analogous to a glucometer, that uses saliva and blood for parallel quantitative detection of viral infection and body's immune response in an automated manner within 11 min is proposed. The technology integrates a versatile biomimetic receptor based on molecularly imprinted polymers in a core–shell structure with nano gold electrodes, a multiplexed fluidic‐impedimetric readout, built‐in saliva collection/preparation, and smartphone‐enabled data acquisition and interpretation. NFluidEX is validated with Influenza A H1N1 and SARS‐CoV‐2 (original strain and variants of concern), and achieves low detection limit in saliva and blood for the viral proteins and the anti‐receptor binding domain (RBD) Immunoglobulin G (IgG) and Immunoglobulin M (IgM), respectively. It is demonstrated that nanoprotrusions of gold electrodes are essential for the fine templating of antibodies and spike proteins during molecular imprinting, and differentiation of IgG and IgM in whole blood. In the clinical setting, NFluidEX achieves 100% sensitivity and 100% specificity by testing 44 COVID‐positive and 25 COVID‐negative saliva and blood samples on par with the real‐time quantitative polymerase chain reaction (p < 0.001, 95% confidence) and the enzyme‐linked immunosorbent assay.

Interior view of unit fully assembled with all device components. For each assay, the current response of the system is recorded, passed through an analog-to-digital converter (ADC), and transmitted back to the Bluetooth module for interfacing to a custom Android smartphone application. The output current measurements are used to calculate the frequency-dependent impedance magnitude on a Bode plot. The entire device can be powered via a standard lithium-polymer battery or via micro-USB, which also recharges the battery when power is 7 depleted. All the potentiostat components are contained within a 3D-printed housing unit (127 mm x 106 mm x 39.5 mm). Figure S4. Digitization of the EIS readout signal. To perform the electrochemical measurements, we designed an EIS smartphone application (based on the open-source Android application provided by Jenkins et al. [1] ) that communicated with a Bluetooth Low Energy (BLE) module; set parameters include the sample per frequency decade as 1, DC bias voltage as 0, signal amplitude as 0.01 V, and electrode configuration as "3-electrode".

Sample collection cartridge
There is no established protocol at the point-of-care for saliva collection on a microfluidic chip. The primary challenge remains the high viscosity of saliva caused by the presence of mucin glycoproteins, which result in a matrix-like and stringy consistency in saliva that can lead to inaccuracies when pipetting and aliquoting a sample. [2,3] A high viscosity results in high internal friction of the fluid, which impedes its ability to flow in microchannels. We proposed the use of a filter-based technique that can remove large glycoproteins from the saliva while effectively reducing its viscosity with results comparable to that of from the centrifugation [4,5] . This is being done via an integrated self-collection funnel that connects to the microfluidic device using custom 3D-printed attachments. kit that combines sample collection, pre-treatment, and microfluidic flow on a single 9 apparatus, (b) Perspective view of the sample collection cartridge with labelled components; a saliva capture funnel for direct self-collection of saliva, a blood collection window that exposes the inlet of the blood microchannel to the finger prick blood from the user, a singlerelease trigger that is used to press down on the PDMS soft lithography buttons (only when the trigger is removed, the buttons will be lifted to enable the suction-based flow), (c) real image of 3D-printed cartridge with inserted electrochemical microfluidic device and corresponding dimensions, (d) Sectional view of the sample collection workflow and pointof-care automated biofluid flow.

Fabrication of multiplex fluidic chip embedded with test assay
A two-step aligned standard lithography was used to pattern the electrochemical electrodes on an indium tin oxide (ITO) glass-coated wafer while a single-step lithography was utilized to pattern the fluidic channels. Initially the ITO-coated glass was deposited with a 5-10 µm silicon dioxide insulating layer using plasma-enhanced chemical vapor deposition (PECVD) at a deposition rate of 10 nm. s -1 . Then the electrochemical reference electrode (RE) and counter electrode (CE) were patterned in an AZ9245 photoresist followed by etching the patterned electrodes in the SiO 2 via BOE etching (Figure S6a). A thin-film consisting of oxide, an attachment layer, and gold at a ratio of 6:1:10 was deposited via electron-beam deposition (BJD 1600) followed by the second lithography step to pattern the electrode in a Shipley photoresist. A wet etching step using HF was utilized to remove the un-patterned gold thin-film and develop well isolated conductive gold electrodes upon photoresist lift-off ( Figure S6b) with a final RE and CE dimensions corresponding to those of a standard screenprinted electrode to confer compatibility with the adapters assembled in the PCB. Next, a tertiary lithography step was used to fabricate the multiplex fluidic channels in an SU-8 layer with a thickness of ~50 µm aligning the sensing chamber over the electrochemical electrodes ( Figure S6c). A bottom-up fabrication method based on chronoamperometric growth of gold nanostructures from HAuCl 4 solution was used to fabricate the working electrode with gold nano/micro island (NMI) structures in the sensing chamber of the microfluidic device. [6][7][8] To complete the assay fabrication on the working electrode, a thin-film of o-PD polymer (5-10 nm) was electrodeposited over the NMI surface via cyclic voltammetry approach ( Figure   S6d) followed by the electropolymerization with SARS-CoV-2 SP and antibody binding sites (fabrication and optimization described in Section S1.4. the NMI/MIP assay). Finally, a PDMS layer with the same size as the wafer (57 mm x 24 mm) was bonded to the wafer to encapsulate the channels via plasma treatment ( Figure S6e).  Cross infection caused by mixing of blood and saliva biofluids on the microfluidic device was avoided by constructing microchannels with unique reaction chambers for each impedimetric assay. Separated fluidic chambers allow for the multiplexed treatment of unique biofluids with controlled fluidic manipulation in individual chambers ( Figure S7, Movie S1), resulting in spatially distinct readouts that have provided 100% sensitivity and 100% selectivity for our system.

COMSOL simulations of fluid flow
The flow profiles of the proposed fluidic channels are evaluated using COMSOL Multiphysics with geometries imported from AutoCAD ( Figure S8). The implementation of suction-based flow via flexible PDMS buttons allows for pressure to become a function of the compressed volume (Equation S1).

( ) Equation S1
Where is the negative pressure caused by the suction-button deformation, is the volume of the undeformed pressure chamber, is the volume of the deformed pressure chamber and is the atmospheric pressure. [10] Notably, ( ) expresses the compression that causes negative pressure-driven flow with the volumes calculated from the suction button (diameter: 2.5 mm, height: 1 mm). We assumed that the suction button is compressed almost completely (± 5% uncompressed); this yielded a pressure drop of about 5000 Pa, which was input as the negative pressure at the device outlets. Blood was simulated with a density of 994 kg. m -3 and dynamic viscosity of 0.004 Pa. s, while saliva was simulated with a density of 1012 kg. m -3 and a dynamic viscosity of 0.00157 Pa. s. [11][12][13] The 3D simulation based on these conditions yielded a velocity distribution shown in Figure S8a; surface integration of the velocity magnitude yielded a volumetric flow rate of 0.801 mL. s -1 , which is reasonable given that the channels have volumes of ~1 mL. The velocity in the chamber containing the assay remained low, which can improve the detection of proteins in the targeted biofluids. [14] The analysis assumed creeping flow of incompressible fluids with unchanging material properties in a stationary simulation. To determine the effect of creeping flow on the microfluidic-based electrochemical assay, we assumed based on the scanning electron microscopy (SEM) analysis (Figure 2b) that the assay height was 2 µm and the RE/CE connection for on-chip detection height was 5 µm; only a 50 µm by 25 µm region of interest was studied. in the proximity of the RE/CE connection, the incident pressure on the assay was dampened; zoom-in view of the assay. Only a 50 µm by 25 µm region of interest was studied.

Cost of NFluidEX device
The cost of the device was based on the fabrication of the printed circuit board, electrochemical microfluidic device, 3D-printed housing unit and the 3D-printed sample collection kit. The cost for one handheld signal transduction unit with a fully assembled PCB was $533, which is cheaper than current potentiostats used for EIS analysis. 29 Table   S1; all prices are given in Canadian dollars and assumes the use of research research-grade materials and fabrication protocols.

Assay optimization using Finite Element Method (FEM)
A FEM simulation was performed using COMSOL Multiphysics to show the effect of the NMIs on electrochemical sensing. A single NMI was designed in SolidWorks 2021 based on SEM images (Figure 2b); we assumed a base with a 1 µm diameter and 3.14 µm 2 surface area, then designed eight protrusions separated by an equal angle (45 degrees) with 2 µm height. The electric current physics is used to solve the current conservation equation based on Ohm's law. A reference electrode was implemented by applying zero potential at infinity using an Infinite Element Domain boundary condition. Then, the molecularly imprinted polymer (MIP) layer made from nonconductive o-PD was modelled by first determining its impedance value. In general, an impedance value depends on the resistance and the capacitance of the electrode surface. [15] To determine the impedance of the o-PD layer, we used a control sample (Figure S9a, inset) as an experimental calibration to determine the unknown parameters. First, we considered the capacitance to have a negligible effect on the impedance value at high frequency (10 5 Hz) (i.e., the impedance magnitude is equal to the resistance parameters). Then, we applied known impedance values for a single gold NMI (2.96 µΩ for an active surface area of 0.49 cm 2 ). [16] The resultant obtained resistance and capacitance parameters for o-PD were 4.5e-4 Ω. µm 2 and 14 F. µm -2 , respectively.
With these known values, we assessed the role of the NMIs on electrochemical biosensing by evaluating the surface current density against that of a bare gold electrode. The simulation results showed that current density was increased more than five times by adding NMI structures ( Figure S9a), which was likely due to the higher geometric aspect ratio and isotropy of the protrusion surface. The sharp edges of the NMI electrodes provided a steep electric field gradient, which enabled a higher electrical current. Also, the NMI structures provided 10 times higher surface area (28.9 µm 2 versus 2.82 µm 2 ) for a high surface-tovolume ratio, resulting in predictably enhanced electrochemical biosensing.
Observing the corresponding impedance of the simulated electrode over the frequency sweep demonstrated the highest impedimetric response for low frequencies values, particularly at 0, 0.1 and 0.01 Hz ( Figure S9b). As such, the most sensitive response was expected to occur over these low probing frequencies.

Molecular docking simulation and characterization of o-PD
Due to the large sequence length of the 6VXX SP, we used 24 boxes to investigate docking events (Table S2); for the smaller 7BWJ antibody fragment, we used 2 boxes (Table S3)   pg. µl -1 . As such, 10 min is considered as the optimal incubation time as negligible differences were observed after 10 min incubation. For the whole blood, the incubation period was determined to be 1 min to prevent coagulation on the surface of the electrode to avoid an erroneously high impedimetric readout due to the presence of coagulated blood aggregates.       for whole blood are likely due to the presence of interferent molecules and blood cells, [39] but the electrochemical response remains within a comparable range of impedance magnitude.   . [40]

Selectivity and Cross Reactivity
where is the number of particles defined to be the total mass divided by the mass of individual atoms, is the coordinate of each atom and is the center of mass defined to be the average position of each atom. Interestingly, a higher impedance magnitude was recorded amongst proteins with a smaller R g , indicating that smaller proteins can likely bind with greater ease and generate a more robust electrochemical response while larger proteins cannot easily diffuse out of the polymer layer following template removal; [41] similar trends between protein size and electrochemical response were reported in other electrochemical assays. [23] The percent sequence similarity was measured from the PDB file using the Pairwise Structure Alignment tool using the Smith-Waterman 3D algorithm (gap opening penalty = 3; gap extension penalty = 5). [42] The percentage of locally identical residues in addition to residues with similar chemical properties between two structures was reported as the percent sequence similarity.
Physical properties are summarized in Table S9-S10. Structures with a higher similarity showed better binding to the assay (ex. SARS-CoV-1 SP); nonetheless, the sequence and physical similarity between SARS-CoV-2 and other similar viruses were sufficiently distinct to generate statistically significant differences in the impedance magnitude.

Surface plasmon resonance (SPR) to study affinity
The SPR results shown in Figure S25 were fit with a Hill function (Equation S3): [59]

Patient Sample Validation and a Field Study
Raw Impedance Response of Patient Samples

RT-qPCR Calibration Curve for Assessment of quantitative NFluidEX
To quantify the results obtained by the NFluidEX in comparison with the RT-qPCR gold standard method, we used the cycle threshold (Ct) values using the RNA-dependent RNA polymerase (RdRp) gene for amplification in RT-qPCR, to relate the concentration of SARS-CoV-2 viral particles with the RT-qPCR response. Figure S33 shows the calibration plot that relates the Ct values of RT-qPCR with viral particle concentration, which is consistent with the known trend of a decreasing Ct for increasing viral load. [41,[71][72][73][74]