Gold Nanourchins Improve Virus Targeting and Plasmonic Coupling for Virus Diagnosis on a Smartphone Platform

Abstract

Point-of-care detection of pathogens is critical to monitor and combat viral infections. The plasmonic coupling assay (PCA) is a homogenous assay and allows rapid, one-step, and colorimetric detection of intact viruses. However, PCA lacks sufficient sensitivity, necessitating further mechanistic studies to improve the detection performance of PCA. Here, we demonstrate that gold nanourchins (AuNUs) provide significantly improved colorimetric detection of viruses in PCA. Using respiratory syncytial virus (RSV) as a target, we demonstrate that the AuNU-based PCA achieves a detection limit of 1,400 PFU/mL, or 17 genome equivalent copies/μL. Mechanistic studies suggest that the improved detection sensitivity arises from the higher virus binding capability and stronger plasmonic coupling at long distances (~10 nm) by AuNU probes. Furthermore, we demonstrate the virus detection with a portable smartphone-based spectrometer using RSV-spiked nasal swab clinical samples. Our study uncovers important mechanisms for sensitive detection of intact viruses in PCA and provides a potential toolkit at the point of care.

Graphical Abstract

Respiratory infectious diseases have significantly burdened millions of people each year, especially in the current COVID-19 pandemic.¹ Among them, respiratory syncytial virus (RSV) is one of the leading causes of lung diseases among infants and young children under 5 years of age.² Currently available RSV detection methods include laboratory tests³ (e.g., polymerase chain reaction and viral culture assays) and rapid antigen tests (e.g., lateral flow assay⁴ and direct fluorescent antibody testing⁵). While sensitive and specific, laboratory tests have a long turnaround time and are inaccessible in resource-limited settings.⁶ In contrast, rapid antigen tests suffer from low sensitivity⁷ (Fig. S1, 10⁴ PFU/mL compared with the viral load range 10³-10⁶ PFU/mL among the hospitalized infants⁸) and are not recommended for adults and elder children.⁹ Therefore, there is an unmet and urgent need for simple and sensitive detection of RSV.

Homogeneous assays are simple and rapid detection methods without the need for immobilization, separation, and washing, and thus are promising for point-of-care (POC) diagnosis.¹⁰ Plasmonic nanomaterials, such as gold and silver nanoparticles (Au and AgNPs), have been employed for homogeneous plasmonic coupling assays (PCAs), due to their unique localized surface plasmon resonance (LSPR) properties.¹¹ Upon binding to the target and in close proximity to each other, they display color changes, where the peak shift, amplitude of a given peak, or the ratio of two peaks are frequently used for the quantification. However, PCAs have so far suffered from low detection sensitivity for intact pathogen detection in clinical applications, making them inadequate as a standalone diagnostic tool. Previous efforts to tackle this sensitivity limitation mainly focused on sample enrichment or signal amplification strategies by making use of spectroscopy analysis such as surface-enhanced Raman spectroscopy (SERS)¹² and single-particle inductively coupled plasma–mass spectrometry (ICP-MS)¹³, dynamic light scattering (DLS)¹⁴, and dark-field imaging¹⁵(see the comparison in Table S1), with many of these techniques towards POC applications.

The second important strategy to improve PCA relies on the optical properties of plasmonic NPs, which are highly dependent on their geometry (e.g., size and shape).¹⁶ For example, anisotropic AuNPs such as nanorods and nanocubes have additional oriented assembly (tip-to-tip and side-by-side) and produce controllable aggregates with a lower detection sensitivity.¹⁷ The nanostars or nanourchins have an even larger surface area due to the abundance of sharp spikes and dips, giving them a higher payload of probe agents (e.g., antibodies and oligonucleotides) and valuable hot spots for signal enhancement.^{18, 19} However, it remains unclear how these nanostructures can fundamentally alter the PCA kinetics and performance, and elucidating this would lead to new avenues to further engineer sensitive POC tests.

Here, we systematically investigated whether the gold nanostructure shape can substantially improve the detection sensitivity and the underlying mechanisms and developed a simple smartphone spectrometer for POC testing. We evaluated the PCA performance using AuNPs with spherical, rod, and urchin-like shapes for the colorimetric detection of RSV. We demonstrated that Au nanourchins achieve a limit of detection (LOD) of 1,400 PFU/mL, or 17 genomes equivalent copies/μL, which is significantly lower than that of nanorod and nanospheres. By electron microscopy imaging of the nanoparticle-virus binding, we revealed that rod and urchin nanoparticles bind to the virus particle more effectively than spherical nanoparticles. Furthermore, modeling of the plasmonic interactions shows that coupled Au nanourchins show the highest plasmonic coupling (i.e., the largest change in the extinction spectrum) at long distances (~10 nm) that represent the distances for immunorecognition. We further demonstrated that the detection can be performed on a smartphone-based spectrometer and the single-step detection works on clinical specimens spiked with RSV. With improved sensitivity and portability and a simple one-step assay, our point-of-care diagnostic platform has great potential to facilitate virus detection at its early presentation and combat viral infections (Scheme 1).

Scheme 1.

Schematic illustration of advances in Point-of-care testing (POCT) for a patient using the smartphone-based homogeneous assay.

MATERIALS AND METHODS

Materials.

Tetrachloroauric (III) trihydrate (HAuCl₄·3H₂O, 16961-25-4, 99.9%), sodium citrate tribasic dihydrate (Na₃CA·2H₂O, 6132-04-3, ≥99%), and hydroquinone (123-31-9, ≥99%) were purchased from Sigma-aldrich. 3,3’-dithiobis (sulfosuccinimidyl propionate) (DTSSP, 21578) and borate buffer (1M, 28341) were purchased from Thermo Scientific. Gold nanorod (Citrate, GRCH800-1M) were purchased from nanoComposix, Inc. Amicon^™ ultra centrifugal filter units (UFC510024) were purchased from Fisher Scientific. Palivizumab/Synagis (MedImmune, Gaithersburg, MD) was purchased from Children’s Health Retail Pharmacy. Bovine Serum Albumin (BSA) was purchased from BioPharm Laboratories, LLC. WarmStart^® LAMP 2×Master Mix (M1700S) and LAMP Fluorescent dye (50×concentrate, B1700A) were purchased from New England Biolabs Inc. LAMP primer sequences were customized from Sigma-Aldrich. RSV A2 quantitative genomic RNA (VR-1540DQ) was purchased from American Type Culture Collection (ATCC). Viral RNA Extraction Buffer (VRE100) was purchased from Sigma-Aldrich. All aqueous solutions were prepared via ultra-pure (UP) water with a resistivity of 18.0 MΩ·cm.

Propagation and purification of RSV A2 strain.

Human respiratory syncytial virus strain A2 (ATCC, Cat# VR-1540) was propagated by HEp-2 cells in 5% FBS/EMEM media (Fig. S2). Briefly, cells were inoculated with RSV at an MOI of 0.01 tissue culture infectious dose (TCID50)/cell. After 5-6 days of replication in cell culture, RSV A2 strain was purified from culture supernatant using 30%-60% (w/v) non-continuous sucrose density centrifugation (Beckman Coulter SW-28 rotor, 28000 rpm, 4°C for 90 minutes). Plaque titration by immunohistochemical staining technique (IHC) was used to determine the purified viral titers.²⁰ Briefly, plaques formed on infected A549 cells were detected with a primary anti-RSV Fusion protein antibody (Palivizumab/Synagis, MedImmune, Gaithersburg, MD) and a secondary HRP conjugated anti-human antibody (Jackson Immuno Research Labs, Inc., West Grove, PA).

Size-controlled synthesis of Au nanourchins and Au nanospheres.

Au nanourchins and Au nanospheres of different sizes were prepared based on a seed-mediated growth using 15 nm spherical AuNPs as seeds (see supplemental information for details ).^21–23 Typically, the UP water, HAuCl₄·3H₂O, Na₃CA·2H₂O, and AuNPs seeds with a certain amount (Table S2) were added into a clean 250 mL Erlenmeyer flask orderly upon vigorous stirring at the room temperature. Then the hydroquinone was added subsequently. The reaction takes place immediately as indicated by the rapid color change. After stirring for over 2 hours, the final solution was kept in the dark for future use. The concentration of spherical AuNPs has a linear correlation with the O.D. values using empirical formula.²⁴ For urchin-like AuNPs, we estimated their concentrations by the amount of spherical Au seeds added to the reaction.

Antibody functionalization on AuNP probes for respiratory syncytial virus (RSV).

Synagis (Palivizumab) is a monoclonal antibody to RSV surface F protein²⁵ and has been administrated for use in patients due to its significantly higher affinity and potency in neutralizing both A and B subtypes of RSV. We covalently conjugated Synagis onto plasmonic AuNPs with different shapes by DTSSP. Detailed protocol was provided in Supplemental Information.

RSV detection by AuNP probes.

The as-prepared AuNP probes were incubated with purified RSV in a volume ratio of 2:1 (60 μL: 30 μL). For the specificity study, each type of AuNP probe was tested with RSV at 1×10⁵ PFU/mL. Other close relative respiratory viruses including influenza type A virus (IAV, 3.0×10⁷ PFU/mL), human metapneumovirus viruses (hMPV, 2.8×10⁶ PFU/mL), and parainfluenza viruses (PIV, 3.5×10⁵ PFU/mL) were used as negative controls. The spectra were monitored by a microplate reader after 30 minutes of incubation.

Digital Loop mediated-isothermal amplification (dLAMP) to quantify RSV A2 genomic RNA.

LAMP primer sequences were designed using conserved regions of the matrix gene for RSV A.²⁶ Detailed primer sequences were provided in Table S3. Intact RSV A2 viruses were lysed to release the RNA (see supplemental information for details). LAMP solution was prepared with a final volume of 20 μL containing 15 μL of isothermal master mix (including LAMP 2× Master Mix buffer, 10× primer mix, LAMP Fluorescent dye, and UP water with a volume ratio of 10:2:1:2) and 5 μL of target RNA. After gently mixing with a pipette, 15 μL of LAMP solution was then loaded onto QuantStudio^™ digital chip and sealed by oil. The reaction was carried out at 65°C for 35 minutes, then followed by 2 minutes of heating at 95° to re-anneal the amplified DNA. Then the chips were observed under the fluorescent microscope.

Boundary element method to simulate plasmonic properties of AuNPs with different geometries.

The plasmonic properties of AuNPs were calculated using the boundary element method (BEM) approach.²⁷ Three models of AuNPs (i.e., spherical, rod, and urchin-like) were embedded in water as surrounding dielectric environment. The dielectric functions of gold used in the simulations were obtained from Johnson and Christy.²⁸ All simulations were conducted via the MNPBEM toolbox in MATLAB (version 2021b).²⁹

Smartphone spectrometer device.

The smartphone spectrometer was assembled using a low-cost HTC U11 smartphone (1/2.55’’ sensor size). The fabrication process was optimized and referred to a previous work.³⁰ Briefly, the rear-face camera was used as an array detector. A custom-designed cradle with a diffraction grating holder, cuvette holder, and two plastic capillaries have been fabricated by a 3D printer using Acrylonitrile Butadiene Styrene (ABS) polymer. One of the plastic fibers (1.5 mm) carries the phone back flashlight to the cuvette holder, and another fiber collects transmitted light (0.25 mm) to the linear diffraction grating slide (Amazon, 300-700 nm, 1000 lines/mm). Incoming light hits the diffraction grating with ~ 42° to the normal, and first-order diffracted light falls on top of the phone camera. Three lasers with different wavelengths (405, 532, and 641 nm) were used to calibrate the pixel index and the corresponding wavelength. The conversion factor adapted from the linear relationship between wavelength and pixel index was 0.293 nm/pixel. The spectrophotograms are collected as RAW image format and processed by customized MATLAB script. The RGB information from RAW images were converted to the Hue–Saturation–Value (HSV), where the value (V) represents the light transmission through the solution. The absorbance intensity (A) was calculated based on Beer–Lambert law:

$$A={log}_{10}\frac{I_0\left(\lambda \right)}{I_s\left(\lambda \right)}$$

where I₀ is the transmitted light intensity of the blank control solution (i.e., 2 mM borate buffer in this study), and I_s is the transmitted light intensity of the testing solution.

Biosafety statements.

The research project was approved and performed strictly in adherence to CDC/NIH guidelines. The experimental protocols and the use of human specimens were approved by Institutional Biosafety & Chemical Safety Committee (IBCC, #190517) and the Institutional Review Board (IRB, #MR15-189) from the University of Texas at Dallas.

RESULTS AND DISCUSSIONS

Au nanourchin with long protrusions provides the most sensitive detection among three different nanoparticle shapes.

First, we compared the analytical performance for three nanoparticle shapes for colorimetric detection of intact viruses. Nanoparticles clustering around the virus leads to a color change due to plasmonic interactions between the nanoparticles (Fig. 1A). We prepared Au nanourchins (AuNUs), Au nanospheres (AuNSs), and Au nanorods (AuNRs, nanoComposix) and characterized the nanoparticles with transmission electron microscopy (TEM, Fig. 1B–D). Fig. S3A–C show their corresponding sizes to be 75 nm ± 6 nm (defined as tip-to-tip length of an urchin), 50 nm ±5 nm (diameter of spheres), and 50 nm±7 nm and 15 nm ± 2 nm (length and width of rods). All the AuNPs have citrate ions as capping ligands on the surface that facilitates the subsequent antibody (i.e., Synagis/Palivizumab) conjugation via a covalent crosslinker DTSSP (Fig. S3D). The Synagis as a passive immunoprophylactic agent and specifically targets the fusion (F) proteins on the RSV surface.³¹ Synagis coating on AuNPs was verified by the increase of hydrodynamic sizes (Fig. S3E–G). Then the Synagis-modified AuNPs were ready for the RSV detection. Here, all AuNP probes with the same optical density (O.D. = 5) were incubated with serial dilutions of RSV stocks at room temperature for 30 minutes and measured by UV-Vis spectrometer (Fig. 1E–G). The spectra show an obvious drop in the peak intensity for all cases with higher RSV concentration. We then applied a ratiometric method to quantify the results. Taking AuNUs as an example, we generated the calibration curve by plotting the absorbance intensity (I_Abs) ratio of 800 nm to 660 nm against the viral titers (Fig. 1H). A linear relationship (R² = 0.99) was observed in the range of 2,500-50,000 PFU/mL (Fig. 1H inset). The limit of detection (LOD) was determined to be 1400 PFU/mL by calculating the concentration corresponding to a signal that is 3 times the standard deviation above the blank control calibrator.³² The same strategy was applied to other AuNP probes for the LOD calculation (Fig. 1F for AuNSs and Fig. 1G for AuNRs). Their corresponding linear detection ranges and LOD were determined to be 10,000-200,000 PFU/mL and 8,000 PFU/mL, and 10,000-100,000 PFU/mL and 4,300 PFU/mL, respectively. AuNUs provided the most sensitive detection for RSV, which was about 5.7- and 3.1-times lower than that of AuNSs and AuNRs.

Fig. 1. Respiratory syncytial virus (RSV) detection using plasmonic AuNPs with different shapes.

(A) Schematic illustration of AuNPs aggregation induced colorimetric detection of viruses. (B-D) Transmission electron microscopy (TEM) image of (B) Au nanourchins (AuNUs), (C) Au nanospheres (AuNSs) and (D) Au nanorods (AuNRs). Insets show the measured size for each NP. (E-G) Colorimetric detection using (E) AuNUs, (F) AuNSs, and (G) AuNRs as probes. Each inset shows the model of the probe for detection of RSV with serial dilutions (0-100,000 PFU/mL for AuNUs, and 0-500,000 PFU/mL for AuNSs and AuNRs). (H-J) Corresponding calibration curves by plotting the absorbance intensity (I_Abs) ratio at two wavelengths against RSV titers. The error bars indicate the standard deviations (n = 3). Insets show the linear detection range and calculated limit of detection (LOD).

To verify that the assay is specific for RSV detection, we incubated the AuNP probes with other closely related respiratory viruses. These include influenza type A virus (IAV), human metapneumovirus (hMPV), and parainfluenza viruses (PIV) with viral titers above 10⁵ PFU/mL (Fig. 2A–C). Spectral analysis suggests that only the RSV of 10⁵ PFU/mL leads to the peak intensity change (Fig. 2D), while the other control groups give signal below the LOD thresholds. This result confirms that the as-prepared AuNP probes have good detection specificity against RSV.

Fig. 2. Selective detection of RSV among closely relative respiratory viruses.

(A-C) Spectral measurements of RSV detection using AuNSs, AuNRs, and AuNUs. The concentration of RSV is at 1×10⁵ PFU/mL. Other closely relative respiratory viruses are at stock concentration (see details in methods). (D) The absorbance ratio obtained from corresponding spectrum results (A-C) demonstrate that all the AuNP probes allow selective detection of RSV from other respiratory viruses including influenza type A virus (IAV), human metapneumovirus (hMPV), parainfluenza viruses (PIV). The error bars indicate the standard deviations (n = 3). The dashed lines mark the thresholds determined by the LOD values of the different probes.

To fully understand the impacts of AuNPs’ physical properties on the assay performance, we further evaluated the effect of AuNPs size and concentration on RSV detection. Since the size of the urchin-like and spherical NPs can be readily tuned, we chose them as a model to investigate the effect of particle size. By adjusting the amount of the reagents in the reaction, we obtained AuNUs with different tip lengths (65 nm and 75 nm, Fig. S4). The same method was adapted to the size-controllable synthesis of AuNSs. Then we incubated serial dilutions of RSV with AuNU (Fig. S5A–B) and AuNS probes (Fig. S5C–F) of different sizes and kept them at the same concentration (5 O.D.). In addition to 75 nm AuNU and 50 nm AuNSs in previous section (Fig. 1E–F and H–I, LOD = 1,400 PFU/mL and 8,000 PFU/mL), Fig. S6 shows that the LODs for 65 nm AuNUs, 15 nm AuNSs, and 100 nm AuNSs were 8,000, 34,000, and 8,700 PFU/mL, respectively. Smaller size AuNUs result in less sensitive detection (8,000 PFU/mL versus 1,400 PFU/mL). For spherical NPs, increasing the AuNSs size from 15 nm to 50 nm improves the detection sensitivity (4.25-fold enhancement), while further increasing the size to 100 nm leads to less sensitive detection (8,000 PFU/mL versus 8,700 PFU/mL).

The particle concentration related to the hook effect is another important factor for an optimal assay.³³ We thus performed parallel experiments using AuNPs (i.e., AuNUs, AuNSs, and AuNRs) of different optical densities (1-10 O.D.) to assess the effect of NP’s concentration on assay sensitivity (Fig. S7). Fig. S8 shows that increasing the particle concentration from 1 O.D. to 5 O.D. leads to lower LOD values (better detection sensitivity) for all cases. Among all tested cases, 5 O.D. AuNUs with longer protrusions show the lowest LOD at 1,400 PFU/mL. When further increasing the NP concentration to 10 O.D., a U-shaped dependence of the LOD on particle concentration was observed for the 15 nm AuNSs and AuNRs. This is probably due to the “prozone hook effect”, where a high concentration of unbound probes will contribute a strong background signal, resulting in reduced sensitivity.³³ We also noticed the “postzone hook effect” (virus in excess) when using low concentration probes, where the high-concentration viruses give small signal changes (Fig. S9).³⁴ This is because the strength of plasmonic coupling among NP probes was weakened in the presence of excess virions, e.g., one probe bound to a few virions without coupling. Notably, we observed the “hook” shifting to lower RSV concentrations upon increasing the probe concentrations rather than the right shift. The latter is more likely to occur since adding more probes requires more viruses to generate a detectable signal or saturation. To understand this phenomenon, we analyzed the signal responses for different probe concentrations based on two simplified simulations (Table S4 and Fig. S10). Briefly, in these two cases, the binding capacities of probes and analytes are set differently (i.e., probe-independent and dependent binding), resulting in the shift of “hook” with opposite directions. The right shift appears only with a constant binding capacity. We envision that in the RSV detection, the hook effect can be varied with multiple parameters, such as the speed of the formation of maximal aggregates, the size/shape-dependent NP transporting kinetics, the uniformity of the virus (RSV has different morphologies in our case), etc. Therefore, the left shift of the hooks is also reasonable. Collectively, the binding stoichiometry between NPs and viruses plays a critical role in assay performance. Altering the probe concentration alleviates the hook effect and enables a tunable dynamic range for homogeneous assay. Only an optimal ratio of the NP probes to analytes results in the most sensitive detection.

Considering the viral particle concentration (i.e., plaque-forming unit (PFU) ) is highly related to the culture conditions, we performed the digital loop-mediated amplification (dLAMP) and quantified the RNA copy numbers for cross-checking. This allows us to compare the assay performance without relying on reference values and eliminates the batch-to-batch variations. Fig. S11A–H show the dLAMP images using quantitative genomic RSV A2 RNA as standards. A calibration curve was plotted by counting the fraction of positively fluorescent wells (f_on) over the total wells against the RNA concentrations (Fig. S11I). The LOD is calculated at 2 copies/μL. We then lysed serial dilutions of RSV samples and purified the resulted viral RNA before taking them into the dLAMP reactions. Fig. S12 show the corresponding detection results. The LOD is calculated to be 141 PFU/mL. Based on the f_on values, we can correlate RSV titers (PFU/mL) and RNA concentrations (copies/μL, Fig. S13), suggesting that 100 PFU/mL is approximately 1.2 copies/μL. Then we calculated the detection limit of RSV by 75 nm AuNUs as 17 copies/μL. We summarized our findings in Table 1 for the three nanoparticle shapes with various sizes and concentrations. Overall, the AuNU provides a highly sensitive and specific detection for intact virus with a one-step homogeneous assay.

Table 1.

Comparison of RSV detection performance for three different nanoparticle geometries (sphere, rod, urchin) and concentrations (1, 5, 10 OD).

AuNP shape	Dimension (nm)	Concentration (O.D.)	LOD (PFU/mL)	LOD (copies/μL)	LOD (PFU)	LOD trend
						Low	High
Sphere	15	1	280,000	3360	8400
		5	34,000	408	1020
		10	200,000	2400	6000
	50	1	29,000	348	870
		5	8000	96	240
		10	7900	95	237
	100	1	14,000	168	420
		5	8700	104	261
		10	9600	115	288
Rod	50 (length) × 15 (width)	1	8,300	100	249
		5	4400	53	132
		10	5100	61	153
Urchin	65	1	16,000	192	480
		5	8000	96	240
		10	7300	88	219
	75	1	2100	25	63
		5	1400	17	42
		10	1500	18	45

Electron microscopy imaging reveals high virus binding capability for AuNU with longer protrusions and AuNR.

Second, we examined the extent of NPs accumulation on the virus surface due to its critical role in the colorimetric assay. We took TEM images from completed assay solutions using AuNP probes with different shapes, sizes, and concentrations (Fig. 3A) under the same virus titer (10⁶ PFU/mL). The virions show different morphologies³⁵ in the negative stained images (highlighted by red arrows). An important observation is that the AuNUs with longer protrusions and AuNRs show much more dense particle coverage on virions than AuNSs. For a quantitative comparison, we measured the particle coverage on the virion-probes complexes and determine the binding effectiveness (BE, defined as the area of the viruses covered by the NPs in the 2D image). Fig. 3B shows the BE results for all AuNP probes and confirms that 75 nm AuNU and AuNR overall have higher binding to the viruses at different nanoparticle concentrations (1, 5, and 10 O.D.). The particle coverage on virions increases with higher nanoparticle concentrations (1 vs 5 O.D), while the coverage saturates for AuNRs and AuNUs when further increasing the concentration from 5 to 10 O.D. We also noticed excess free NPs in the cases of high AuNP concentrations (e.g., 10 O.D. AuNSs and AuNRs), consistent with the increased background color in the assay. The TEM analysis suggests that non-spherical nanoparticles with higher surface area, either by rod shape or long protrusions on the nanourchin surface, increase the virus binding capability. The high virus-binding capability for AuNRs and AuNUs reveals a key mechanism for the high detection sensitivity with these nanoparticles.

Fig. 3. AuNUs with longer protrusions and AuNRs show higher virus binding effectiveness (BE) than AuNSs.

(A) TEM images of AuNUs, AuNRs, and AuNSs probes with different optical density (O.D.) after incubating with RSV at 10⁶ PFU/mL. Scale bar = 200 nm. Red arrows indicate negative stained RSV A2 strains via 2% uranyl acetate. Note that RSV is variable in shape and size (e.g., spherical, or filamentous). (B) Binding effectiveness (BE) analysis for different AuNP probes. BE is defined as the area of the viruses covered by the NPs in the 2D image. NPs counting and area measurements were performed by Image J. The total number of TEM images used to analyze the AuNP probes at each concentration is n=6 from 3 samples. ns, P > 0.05; **, P < 0.01; ****, P < 0.0001.

Modeling the plasmonic coupling shows that AuNU with longer protrusions has strongest interaction at long distances (10 nm).

Third, we investigated the plasmonic coupling strength for the nanoparticles, another key factor in determining the colorimetric assay performance, with computational simulations. Among different computational techniques (e.g., Mie theory,³⁶ discrete dipole approximation,^{37, 38} finite element method,³⁹ finite difference time domain⁴⁰), boundary element method (BEM) provides simple and fast simulation for AuNPs with complex geometries. Particularly, we calculated the extinction cross-section area as a metric for the AuNP plasmonic coupling, as colorimetric assays essentially measure the extinction or absorbance change of the AuNPs probes. For simplicity, we calculated the plasmonic coupling strength between two coupled AuNPs. We first built a model of two identical AuNSs with interparticle distance d of 0.1-10 nm (Fig. 4A). Fig. S14 show the simulated extinction cross-section for coupled AuNSs with different sizes. Decreasing the interparticle distance leads to a red shift in the plasmonic peak (to longer wavelength) and an increase in the extinction cross-section. Different particle sizes show the same trend. Furthermore, the larger AuNSs have a greater peak shift than smaller ones (e.g., 15 nm). Fig. 4B–C show the change of extinction cross-section (ΔExt_@SPR) and the peak shift of coupled AuNSs against that of a single particle. Clearly, both values of ΔExt_@SPR and peak shift decrease as the interparticle distance (d) increases for a given particle size. While fixing d (< 5 nm), larger particle size has higher ΔExt_@SPR and peak shift. Notably, for d > 5 nm, the ΔExt_@SPR becomes close for 50 nm and 100 nm AuNSs. This suggests that further increasing AuNP size does not necessarily improve the plasmonic coupling.

Fig. 4. BEM simulation of plasmonic coupling properties for AuNP probes with different shapes.

Models for two coupled (A) AuNSs, (D) AuNRs, and (G) AuNUs. The calculated change of extinction cross-section (ΔExt_@SPR) and peak shift for coupled AuNSs (B, C), AuNRs with side-to-side (STS) and end-to-end (ETE) orientations (E, F), and AuNUs (H, I) as a function of interparticle distance d. The extinction cross-section and the peak position of a single AuNP were used as a reference. (J-K) Comparison of peak shift and extinction cross-section change (ΔExt_@SPR) for AuNP probes at d = 10 nm.

We then examined a model for coupled AuNRs with dimension of 50 nm × 15 nm. Because of their anisotropic shape, different orientations of the AuNRs lead to various modes of coupling. For simplicity, we built two representative models (Fig. 4D) to calculate their plasmonic coupling strength, i.e., side-to-side (STS) and edge-to-edge (ETE). Fig. S15 shows the simulation results of extinction cross-section with STS and ETE coupling. Three linear polarization directions (i.e., x-, y-, and z-directions) were considered in the simulations since the optical properties of AuNRs are polarization-dependent (Fig. S16 and S17). The two coupling modes lead to opposite peak shift, where a blue shift of longitudinal peak occurs for the STS and a red shift for ETE, consistent with previous work.⁴¹ Increasing the inter-particle distance (d) leads to an increase of ΔExt_@SPR and peak shift for STS, and a decrease of ΔExt_@SPR and peak shift for ETE (Fig. 4E–F). Interestingly, ETE mode has higher ΔExt_@SPR values than the STS mode, regardless of the coupling gap.

Next, we examined for AuNUs which have irregular tips on the surface. For simplicity, we constructed the model with a core-tips structure and evaluated the influence of a set of parameters on the optical response, including tip number (n), tip length (L), and core size (D) based on the TEM (Fig. S18A). We first analyzed the model with D = 50 nm and L = 20 nm for different tip numbers (n = 1, 6, and 14). Fig. S18B shows the simulation results of the extinction cross-section for individual AuNUs. It is worth mentioning that the simulated spectra of AuNUs can be assumed as the main plasmon mode confined on the tips (LSPR at around 700 nm) and a secondary mode confined within the core (LSPR at around 520 nm).^{42, 43} We found that additional tips on the surface increase the extinction cross-section at the longitudinal resonance (~700 nm) and minor blue shift of the peak. Moreover, the appearance of multiple peaks can be observed as the shape of the AuNUs becomes more complicated. On the other hand, increasing the tip length L (Fig. S18C) or core size D (Fig. S18D) of an AuNU with a fixed tip number (n = 14) leads to the increase of extinction cross-section as well. Detailed polarization-dependent plots of each case are shown in Fig. S19–S21. Fig. S22 shows the simulated extinction cross-section at different interparticle distances for coupled AuNUs (tip-to-tip, D = 50 nm, L = 15 nm or 20 nm, and n = 14). Interestingly, we found that the magnitude of the extinction cross-section (ΔExt_@SPR) changes with the interparticle distance while there is minimal peak shift regardless of the interparticle distance between two coupled AuNUs (Fig. 4G–I). Longer tips lead to a larger change in the extinction cross-section (ΔExt_@SPR), and therefore plasmonic coupling strength.

Other factors such as the binding orientation among anisotropic nanoprobes can influence plasmonic coupling. For simplicity, we again calculated the plasmonic coupling strength for two anisotropic NPs (i.e., AuNRs and AuNUs) with different coupling angles (θ). The interparticle distance (d) was set at 10 nm since it represents the closest interparticle distance in the case of antibodies conjugation on the nanoparticle surface (e.g., the size of antibodies is 5~10 nm).⁴⁴ Fig. S23A, B show the models and corresponding simulation results for AuNRs with various θ. When altering the θ from 0° to 180°, the rod dimer shows anti-bonding and bonding resonance at about 700 nm and 750 nm, which agrees with previous studies.^45–48 Results suggest that θ = 180° as the ETE coupling case provides the highest plasmonic coupling strength. In the case of AuNUs (Fig. S23C, D), it can be seen that AuNUs show magnitude changes in extinction cross-section without obvious peak shifts as θ changes. While there is some orientation dependence, the overall trends of the extinction cross-section for both AuNRs and AuNUs are in the same range as the previous simple model (Fig. 4).

To better understand the effect of the plasmonic coupling mechanism on colorimetric sensing performance, we compared the simulation results at d = 10 nm for all AuNPs employed for RSV detection. Fig. 4J–K suggest that ΔExt_@SPR is a more important factor to evaluate the detection performance than that of the peak shift. Non-spherical AuNP probes with higher values of ΔExt_@SPR tend to provide better detection sensitivity, where AuNUs_L20 (longer tip length, L = 20 nm) show 4.8- and 3-times higher ΔExt_@SPR than that of AuNSs (50 nm) and AuNRs, respectively. Our results suggest that the extinction cross-section change (ΔExt_@SPR) is a good indicator for the plasmonic coupling strength and assay performance.

Smartphone spectrometer enables point-of-care diagnostics of clinical samples with similar sensitivity to laboratory spectrometer.

We further fabricated a 3D-printed device that can be integrated with a smartphone for virus detection at the point of care (Fig. 5A). Specifically, the device prototype includes a phone case to slide the low-cost smartphone (Android, HTC U11, sensor size 1/2.55’’), a compartment that holds a cuvette (Fig. S24) for testing and blocks out environmental light, and a bundle of plastic fibers for light propagation. The integrated device has a single diffraction grating to divide the light and contains no electronic components in a low-cost and simple design. For colorimetric measurements, we simply apply the flashlight from the smartphone and take the RGB images. This facilitates sample testing in a “plug and play” fashion. The pixel index from RAW images was converted to wavelength by three lasers of different wavelengths (405, 532, and 641 nm, Fig. 5B).

Fig. 5. Smartphone-based spectrometer enables point-of-care colorimetric detection.

(A) CAD drawing of the smartphone spectrometer. Insets show the backside of the design (i) and a photograph of the assembled smartphone spectrometer (ii). (B) Pixel index and wavelength calibration by three lasers. (i) The spectral images obtained by the plastic fibers from violet (405nm), green (532 nm), red (641 nm) lasers. (ii) The output spectrum was extracted from the image after calibration. HSV means the value in Hue-saturation-value (HSV) was plotted. (C) Serial dilutions of AuNU suspensions were imaged by smartphone spectrometer. (D-E) Absorbance spectrum obtained from smartphone (D) and commercial microplate reader (E) for serial dilutions of AuNU suspensions. (F) Comparison of the measurement performance of the smartphone spectrometer and the commercial microplate reader. The error bars indicate the standard deviations (n = 3). (G-I) RSV detection by the smartphone device. (G) The photographs and spectral images for RSV detection. (H) Corresponding absorbance spectrum were extracted from (G). (I) Calibration curves by plotting the peak absorbance at 660 nm against RSV titers. The error bars indicate the standard deviations (n = 3). Inset shows the linear detection range and calculated limit of detection (LOD) for AuNUs using smartphone spectrometer.

To validate the smartphone-based colorimetric reader, we measured serial dilutions of 75 nm AuNU suspensions (Fig. 5C) and compared the results to a commercial microplate reader. Fig. 5D and Fig. 5E show the absorbance spectra obtained from a smartphone and a commercial microplate reader, respectively. The smartphone measurements show diminishing values at the upper limit of the diffraction grating (~700 nm) but give similar peak values with the microplate reader. Herein we plotted the peak absorbance at 660 nm against the particle concentrations (Fig. 5F). Both devices showed similar detection results and confirmed the utility of smartphone-based reader for colorimetric measurements. Moving forward, we used this device to test completed assay solutions containing 75 nm AuNU probes and serial dilutions of RSV. We analyzed the smartphone data using the peak absorbance instead of the previous ratiometric method since the smartphone measurements are not accurate when approaching 700 nm. Also, AuNU probes mainly show intensity change and minimal peak shift in virus samples, as shown in previous experiments (Fig. 1E). Therefore, the peak absorbance values are preferred to quantify the limit of detection. Fig. 5G–I show that the detection limit was calculated to be ~ 1580 PFU/mL, similar to the microplate reader result (~1,400 PFU/mL).

Lastly, we explored the potential clinical application of infectious disease diagnostics using the smartphone-based spectrometer. We collected nasal swab samples from healthy adults and spiked RSV of different concentrations into the nasal swab sample solutions. We directly tested those samples with 75 nm AuNU probes and measured them by smartphone spectrometer. Our result suggests sensitive detection in the complex sample matrix with high accuracy (Table 2), demonstrating the feasibility of mobile detection at the point of care.

Table 2.

Analytical performance of the smartphone-based diagnosis of RSV-spiked nasal swab samples.

Spiked RSV (PFU/mL)	Estimated viral RNA (copies/μL)	Positive test #	Total test #	Accuracy
0	0	0	5	100%
2,000	24	4	5	80%
10,000	120	5	5	100%
100,000	1200	5	5	100%

CONCLUSION

In summary, we have developed a rapid and sensitive PCA for RSV detection. We demonstrated that AuNUs with long protrusions achieved the lowest detection limit among AuNRs and AuNSs. The improved analytical performance of AuNUs can be attributed to higher virus binding and stronger plasmonic coupling at long distances. We further assembled a 3D-printed smartphone spectrometer for portable and colorimetric measurements, allowing for sample-to-answer and quantitative tests. Therefore, the AuNU-based PCA has the great potential to be employed for large-scale tests and to combat infectious disease spread.