Multiplexed Biosensing of Proteins and Virions with Disposable Plasmonic Assays

Our growing ability to tailor healthcare to the needs of individuals has the potential to transform clinical treatment. However, the measurement of multiple biomarkers to inform clinical decisions requires rapid, effective, and affordable diagnostics. Chronic diseases and rapidly evolving pathogens in a larger population have also escalated the need for improved diagnostic capabilities. Current chemical diagnostics are often performed in centralized facilities and are still dependent on multiple steps, molecular labeling, and detailed analysis, causing the result turnaround time to be over hours and days. Rapid diagnostic kits based on lateral flow devices can return results quickly but are only capable of detecting a handful of pathogens or markers. Herein, we present the use of disposable plasmonics with chiroptical nanostructures as a platform for low-cost, label-free optical biosensing with multiplexing and without the need for flow systems often required in current optical biosensors. We showcase the detection of SARS-CoV-2 in complex media as well as an assay for the Norovirus and Zika virus as an early developmental milestone toward high-throughput, single-step diagnostic kits for differential diagnosis of multiple respiratory viruses and any other emerging diagnostic needs. Diagnostics based on this platform, which we term “disposable plasmonics assays,” would be suitable for low-cost screening of multiple pathogens or biomarkers in a near-point-of-care setting.

using a tungsten halogen light source attached to a monochromator.The monochromatic light passes through a polariser to generate a linearly polarised monochromatic light to excite the nanostructures.Images are then captured at each wavelength point.The images are then analysed in the Labview software to separate out the 18 regions and calculates the average intensity of the pixels for each array region, for each polarisation state.Using Stokes equation (below), the optical rotation can then be calculated and plotted for each wavelength producing the ORD plot.

𝑂𝑅𝐷 =
1 2  −1 ( 45 −  135 ) ( 0 −  90 ) Where  represents the intensity and the subscript represents the polarisation angle.The monochromator steps by 1nm for each wavelength point and the entire spectrum is measured in ~5mins.The software performs peak detection using the ORD spectrum and stores the resonance wavelength position for each nanoarray.
The microfluidic chamber (Supplementary Figure S1 (A)) is customised to adhere directly to the gold surface of the sample.A handheld pipette is used to inject 100 uL of solution through the ports into the well over the array of nanostructures.
For multiplexing, 0.2uL of solution was spotted onto the 4 outer corners of the array (A, C, G and I) as shown in Supplementary Figure S1 (B).The surface chemistry of the solutions adhering to the gold surface allow such droplets to be formed -omitting any cross contamination between solutions.
To estimate the performance of the instrument and sensing substrate for a multiplexing application, we need to evaluate the noise performance and establish error limitations that will impact sensitivity and performance limits of the platform.Multiple measurements of water, followed by 2M MgSO4, were performed hourly and used to estimate the measurement error and potential drifts when no changes are made in the fluidic chamber.A 2 M MgSO4 solution was prepared in water.An initial reference measurement is taken using water pipetted into the fluidic well.An initial measurement was taken at 0 minutes, and every 30 minutes thereafter.After 3 hrs, the water is removed and the salt solution is added for measurements at 0 minutes, and every 30 minutes thereafter.
Supplementary Figure S2 (A and B) and Table S1, show that the largest maximum-minimum (max-min) range is of 0.3 nm or less for both water and salt.Hence, we estimate measurement error to ±0.15 nm for Δλ of both peaks.Errors can be larger when solutions are exchanged and this may occur due to variations in incident light wave fronts, temperature and concentration gradients, or changes on the metal-liquid interface (electrical double layer) over the 3mm x 3mm measurement area that have been shown previously to affect plasmonic sensing.Another useful parameter to consider is the value S that represents spacing in the wavelength values between the two peaks.ΔS is the change in the spacing in comparison to the initial measurement and this has previously been used as an additional parameter to measure protein interactions at the surface.The large ~20 nm resonance shift due to MgSO4 is accompanied by ΔS of 0.2-0.3nm indicating a slight difference between Peak 1 and Peak 2 in terms of the sensitivity to refractive index changes.However, for proteins and buffer solutions, we expect resonance shifts to be an order smaller as the proteins form only a monolayer and the buffer solutions are far lower in salt concentration than the solutions used here.The differences in sensitivity, and hence ΔS, seen in the salt solutions would be near negligible when resonance shifts are in the order of ~2-5nm.

Values in nm
Δλ of Water over 3hrs

Δλ of MgSO4 relative to Water
Δλ

Simulations for maximum electric field intensities
Supplementary Figure S3: Electric field magnitude, |E| (V/m) along z-axis above shuriken nanostructures for two separate modes: 690 nm (A) and 713 nm (B).Inset shows the 2D surface plots for |E| values on gold surface (z=100 nm) at these two modes.
The shuriken nanostructures were simulated using COMSOL 5.6.At the two peak positions, the electric field intensities were studied by measuring the maximum |E| field in the xy plane along the lateral z direction moving from the bottom of the indentation (at z=20 nm) up to 150 nm from the surface (at z=250 nm).For Peak 1 we see two lateral points that have the highest E fields, one inside the indentation and one on the surface.Peak 2 has a peak only at the surface.Both show that the |E| fields decay to <15% of the maximum by ~25 nm above the surface indicating sensitivity to material up to 20-30nm and showing high surface sensitivity, hence, the structures are not strongly affected by the bulk media.

Additional biosensing data
To evaluate whether non-specific binding is completely inhibited using spacer molecules, and whether the Biotin PEG thiol/spacer SAM is binding to the streptavidin as expected.An experiment was carried out where the spacer molecule alone was used to generate a SAM for complete passivation of any nonspecific adsorption of proteins.The results showed that MT(PEG)4 concentrations below 50mM in PBS leads to potential non-specific binding of streptavidin, as seen in Supplementary Figure S4, some weak fluorescence is observed at 10mM.However, this is completely omitted upon increasing the concentration to 100mM.Therefore, based on these results, future experiments were carried out using MT(PEG)4 at a minimum concentration of 50mM.Bovine serum albumin (BSA) tagged with Alexa 647 was used as a control.
Supplementary Figure S4: Fluorescence Images of nanostructures with streptavidin Alexa 647 on surface, with varying spacer concentrations-0, 10 and 100mM.BSA with Alexa 647 as control.
To test streptavidin biotin interactions, streptavidin (with Alexa 647 conjugate) is immobilised to the surface using a biotin PEG thiol SAM.Results for resonance shift measurements relative to initial water measurements are shown in Supplementary Figure S5 (A and B) and Supplementary Table S2.
Supplementary Figure S5: Boxplot data for experiment with streptavidin conjugated with Alexa Fluor 647, binding to biotin PEG thiol SAM, for peak 1 (A) and peak 2 (B).Fluorescence from the streptavidin with Alexa Fluor 647 conjugate shown in (C).While protein is coating the entire surface, only the plasmonic enhanced fluorescence generates enough intensity to be visible hence the array pattern is visible.
The buffer solution shows an average resonance shift of 0.8 and 0.9 nm for P1 and P2 respectively, with ~0.1 nm standard deviation (σ) for both peaks.A SAM layer is then allowed to form onto the surface and rinsed with PBS.A mean Δλ of 1.8nm is observed for P1 but P2 shifts more, 2.7nm on average.The σ values are still low even though the range is above 0.5nm showing relatively good homogeneity in the coverage over the entire surface.The SAM layer shows an increase in the spacing by 0.8 nm for Δλ < 3 nm.This is significantly large in comparison to the spacing change shown by the salt solutions.We attribute this change in spacing to a change in coupling of the optical modes as shown in previous work.(Kelly et al., 2018) Streptavidin is then immobilised onto the surface, rinsed with buffer after 1 hr exposure and then measured.P1 and P2 respectively show Δλ of 4.4 nm and 6.2 nm, however, the larger σ values show a reduction in homogeneity of coverage, but such coverage is still relatively homogenous as also shown by fluorescence microscopy, Supplementary Figure S5 (C).ΔS values do not change indicating that the coupling may largely be affected by the SAM alone.The ΔΔλ values were also considered, however the σ values were matching or larger than the mean values in the case of SAM and streptavidin indicating a lack of statistical relevance.It is likely that this is a consequence of random orientations of the immobilised streptavidin and its own structure having overall higher symmetry in comparison to previously tested proteins such as enzymes and bovine serum albumin.In another experiment, unconjugated streptavidin is immobilised to the surface using the same protocol as in Exp1.An additional step is performed where biotin conjugated with Atto 655 is added to bind to the streptavidin.Results are shown in Supplementary Figure S6 and Supplementary Table S3.The results are discussed in the main text.Emission and absorption spectra for Alexa 647 and Atto-655 (taken from ThermoFisher SpectraViewer) are shown in Supplementary Figure S7.For Alexa Fluor 647, the excitation peak lies at 653 nm whilst the emission peak lies at 669 nm.In comparison, the maximum excitation wavelength (λex) for Atto 655 is 663nm, and the maximum emission wavelength (λex) is 684nm.

Additional data on multiplexing experiment for multi target diagnostics
The SARS-CoV-2 spike protein is a 1273 amino acid long (180-200 kDa) class I glycoprotein which can be cleaved into two subunits, spike 1 (S1) and spike 2 (S2), which play different roles in viral binding and entry into cells.For our experimental processes, the S1 subunit was used, which consists of an Nterminal domain (NTD; amino acid sequence 14-290), and a receptor binding domain (RBD; amino acid sequence 306-527).The 193-amino acid RBD constituent of the spike protein is responsible for recognising and binding to the angiotensin-converting enzyme-2 (ACE-2) on host cells.Prior to its use in a multiplexed DPA, we evaluated detection of anti-S1 antibodies in more realistic samples by mixing the antibodies in an artificially reconstituted mimic of human mucus, containing 0.2% mucin, plus 0.25 mg/ml haptoglobin and 0.50 mg/ml transferrin in phosphate buffer saline (PBS).The binders were tested against His-tagged recombinant S1-protein immobilised over the whole surface.Each measurement was taken relative to the SAM measurement after rinsing with PBS and Tween 20 solution to remove any non-specifically bound material on the surface, shown in Supplementary Figure S8.A mean Δλ value of 2.6 nm (Peak 2) was obtained for the successful immobilisation of the S1-protein onto the SAM.The artificial mucus was then applied for 15 minutes as a control to monitor any non-specific binding.Measurements with the artificial mucus itself shows non-specific binding with mean Δλ of 0.2 nm (Peak 2), values similar to resonance shift error (±0.3 nm) expected for changes in media performed previously.These are also much smaller than the expected shifts (>2nm) for anti-S1 antibodies (anti-S1 Ab) that are >100 kDa in mass.The artificial mucus was spiked with a 1 μM solution of anti-S1 Ab and applied to the biosensor for one hour in which positive shifts of 1.2nm (Peak 2) were observed for the change in the mean Δλ, portraying binding of the anti-S1 Ab to the S1 protein.From the data displayed in Supplementary Figure S8, there is minimal variance across the 9 structures, portrayed by the box plots (25-75% quartile shown by the box and max-min range by the whiskers) indicating a similar response from all arrays at each stage of the experiment.In particular, the anti-S1 Ab maximum and minimum data lies within a 0.3 nm range for Peak 2.

Array G & I Buffer
Array Supplementary Figure S2: (A) Resonance shifts (Δλ) with repeat water measurements and (B) repeat MgSO4 measurements, over 3-hour periods.(C) Δλ obtained for MgSO4 relative to water, with (D) change in ORD for both right-and left-handed structures.Boxplot whiskers show the max-min range of the data.
Supplementary Figure S6: Results from the second experiment, non-fluorescent streptavidin is used and an additional final step of binding biotin conjugated with Atto 655 to the streptavidin is performed, in which resonance shifts from peak 1 (A) and peak 2 (B) are also shown.(C) shows the fluorescence on the nanostructures from the biotin with Atto 655 conjugate.Initial measurements were taken in water and PBS, with all biosensing data taken relative to water.(D-F corresponds to Figure 3(B) in main text).