Ultrasensitive lateral-flow assays via plasmonically active antibody-conjugated fluorescent nanoparticles

Abstract

Lateral flow assays (LFAs) are rapid and inexpensive, yet they are nearly 1,000-fold less sensitive than laboratory-based tests. Here, we show that plasmonically active antibody-conjugated fluorescent gold nanorods can make conventional LFAs ultrasensitive. With sample-to-answer times within 20 min, plasmonically enhanced LFAs read out via a standard benchtop fluorescence scanner attained about 30-fold improvements in dynamic range and in detection limits over four-hour-long gold-standard enzyme-linked immunosorbent assays, and achieved 95% clinical sensitivity and 100% specificity for antibodies in plasma and for antigens in nasopharyngeal swabs from individuals with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Comparable improvements in the assay’s performance can also be achieved via an inexpensive portable scanner, as we show for the detection of interleukin-6 in human serum samples and of the nucleocapsid protein of SARS-CoV-2 in nasopharyngeal samples. Plasmonically enhanced LFAs outperforming standard laboratory tests in sensitivity, speed, dynamic range, ease of use and cost may provide advantages in point-of-care diagnostics.

Lateral flow (immuno)assays (LFAs) are amongst the simplest, fastest, and cheapest point-of-care (POC) diagnostic methods, and offer broad potential for population-level screening for disease.^1–2 However, this potential has not yet been fully achieved. Although numerous LFAs for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibodies^3–5 and antigens^6–7 have been introduced, none have sensitivity and quantitation comparable to laboratory-based diagnostics such as real-time reverse transcription polymerase chain reaction (RT-PCR) and enzyme-linked immunosorbent assay (ELISA), which constrains their widespread use.^8–10 In general, conventional colorimetric-LFAs are ~1000-fold less sensitive than these standard laboratory tests,^11–12 and diagnosis using LFAs requires an additional confirmatory laboratory-based test to correctly establish negative results. Colorimetric-LFAs offer limited quantification ability owing to the limited color change with respect to variation of the target analyte’s concentration.¹³

The COVID-19 pandemic highlights the need for improved LFAs for precise and rapid clinical diagnoses, mass screenings and epidemiological studies.^14–15 RT-PCR^16–17 and direct antigen tests^18–19 have been the mainstay for diagnosis of COVID-19, and serological assays are important for determination of infection stage and vaccine efficacy, and epidemiological studies.^{3, 20–21} These diagnostic assays are available only in qualified microbiology laboratories and remain expert-dependent, labor-intensive, and time-intensive, limitations that have precluded conduction of the millions of tests per day needed during epidemiological surges.^22–23 Therefore, a critical need exists for diagnostic and screening tools that are not only as accurate as laboratory-based assays, but also rapid, easy-to-use, inexpensive, readily available (e.g., home-based and POC), and scalable for rapid population-level screening.

Efforts to improve bioanalytical performance of LFAs have included using fluorescent molecules or quantum dots as reporter elements.^24–25 Although fluorescent reporters improve quantification, their relatively weak signal intensity limits their sensitivity and point-of-care diagnostic utility, and their low light absorption compared to conventional colloidal gold nanoparticles (AuNPs)²⁶ precludes the direct visual detection available in conventional LFAs, and moreover requires use of LFA readers with highly sensitive detectors or powerful excitation light sources. These limit the utility of fluorescent LFAs in mass screening and resource-limited settings.¹⁰

We envision a “bimodal” LFA in which an initial screening can be performed with a visual test, and subsequent quantitative testing can be performed when needed on the same LFA strip using a fluorescence reader. To achieve this, we employed an ultrabright fluorescent nanoconstruct that we have recently introduced,²⁷ called plasmonic-fluor, as a bimodal colorimetric and fluorescent reporter in LFAs (Fig. 1A). These nanoconstructs harness plasmon-enhanced fluorescence^28–32 to achieve nearly 7000-fold brighter fluorescence signal compared to conventional molecular fluorophores. We conjugated plasmonic-fluors with detection antibodies and used them to enable rapid and ultrasensitive colorimetric and fluorescent detection of analytes, using human IL-6 (LOD: 93 fg ml⁻¹), SARS-CoV-2 S1 (subunit of the spike protein) antibodies (LOD: 185 pg ml⁻¹), and SARS-CoV-2 antigen (nucleocapsid) protein (LOD: 212 pg ml⁻¹). We validated the clinical efficacy of the plasmonic-fluor-based LFAs (p-LFAs) by testing plasma, serum and nasopharyngeal (NP) swab samples for detection of SARS-CoV-2 S1 antibodies, IL-6, and SARS-CoV-2 antigen, respectively, and achieved high clinical specificity and sensitivity. We also demonstrate the quantitative ability of p-LFA employing a portable self-designed scanner compatible with plasmonic-fluor to demonstrate its versatile POC application. Substantially, this technology can be immediately deployed as an alternative to laboratory-based test for the diagnosis of clinically-relevant pathogenic infections and possible future pandemics.

Fig. 1 |. Gold nanoparticles and plasmonic-fluors as nanolabels for LFA.

(a) Schematic illustration of plasmonic-fluor, employed as a bimodal nanolabel (colorimetric+fluorescent) in LFAs, comprising of gold nanorod as plasmonic core, polymer layer as spacer, molecular fluorophores (800CW), and biotin as recognition element. Transmission electron microscopy image of (b) AuNPs and (c) plasmonic-fluors. (d) Mean gray values obtained from nitrocellulose membrane drop-casted with different concentrations of AuNPs. Inset shows the 8-bit ImageJ-processed image of the nitrocellulose membrane. Fluorescence intensities obtained from nitrocellulose membrane drop-casted with different concentrations of (e) plasmonic-fluors and (f) molecular fluorophores. Inset shows the corresponding fluorescence image of the nitrocellulose membrane. (g) Mean gray values obtained from nitrocellulose membranes, with biotinylated-BSA used as capture-ligand at test sites, after exposure to different concentrations of streptavidin-conjugated AuNPs. Inset shows the schematic illustration of streptavidin-conjugated AuNPs. (h) Fluorescence intensities obtained from nitrocellulose membranes, with biotinylated-BSA as recognition elements at test sites, after exposure to different concentrations of streptavidin-conjugated plasmonic-fluors. Inset shows the schematic illustration of streptavidin-conjugated plasmonic-fluors. Purple arrows indicate the direction of flow of the nanoconjugates. Data are mean ± standard deviation, n = 4 repeated tests.

Results and discussion

Plasmonic-fluors increases sensitivity over AuNPs by 10000-fold in LFAs

Plasmonic-fluors were first applied to overcome three fundamental limitations of the 30-40 nm AuNPs used as conventional colorimetric labels in LFAs. AuNPs have low capture rate (<5%), low signal-to-background ratio, and thus relatively low sensitivity.^33–34 Even with use of 100 nm AuNPs, shown recently to improve LFA sensitivity,³⁵ these problems persist. Because of these three limitations, color changes in AuNP-based LFAs are limited to qualitative analysis or simply a binary output, indicating the presence or absence of the target analyte.

To assess whether plasmonic-fluors (length 98 ± 8.7 nm; diameter 29.2 ± 3.1 nm) could overcome these limitations, we compared their performance to AuNPs (diameter 104 ± 13.4 nm) on a nitrocellulose membrane. The localized surface plasmon resonance (LSPR) wavelength of plasmonic-fluors (and the gold nanorod (AuNR) core) was tuned to match the excitation and emission wavelengths of the molecular fluorophores²⁷ by modifying their aspect ratios^36–37, and the optimal dimensions of the nanostructures were chosen to maximize fluorescence enhancement, based on our previous study³⁸. We set out to determine the minimum number of AuNPs and plasmonic-fluors required to produce a detectable visible or fluorescence signal. When serially diluted AuNPs (Fig. 1B and Supplementary Fig. 1) and plasmonic-fluors (Fig. 1C and Supplementary Fig. 1) of known concentration were drop-casted onto nitrocellulose membrane, accumulations of ~10⁶ AuNPs and plasmonic-fluors were needed to produce a discernable visible signal (Fig. 1D and Supplementary Fig. 2). However, only ~10² plasmonic-fluors were required to produce a detectable fluorescence signal (Fig. 1E and Supplementary Fig. 3). Further, accumulations of ~0.6x10⁶ molecular fluorophores (800CW, the fluorescent unit of plasmonic-fluors) were required to produce detectable fluorescence signal (Fig. 1F), indicating ~6000-fold lower concentration threshold for a detectable fluorescence signal with plasmonic-fluors compared to molecular fluorophores.

Plasmonic-fluors exhibited colorimetric signal nearly identical to that of AuNPs (Supplementary Fig. 4). The colorimetric signal enabled qualitative visual detection (by naked eye), obviating the need for specialized read-out equipment at a relatively high concentration of the target analyte, while the fluorescence signal enabled ultrasensitive detection and quantification of low abundance analytes. Thus, plasmonic-fluor function as a bimodal nanolabels (colorimetric+fluorescent) and offers ultrasensitive detection in a biological assay representative of LFAs.

Next, to compare the performance of plasmonic-fluors and AuNPs in LFA format, we employed the well-characterized biotin-streptavidin conjugate pairing, known to exhibit extremely high binding affinity.³⁹ Both AuNPs and plasmonic-fluors were functionalized with streptavidin and biotinylated bovine serum albumin (BSA) was used as a capture-ligand. LFA strips were then subjected to different known concentrations of streptavidin-conjugated AuNPs and plasmonic-fluors for 20 min (Supplementary Fig. 5). Nanolabels flows along the nitrocellulose membrane by capillary force and gets captured by the capture-ligand, leading to the accumulation of nanoparticles at the test spot. Accumulation of sufficient number of nanolabels converts the color at the test site to red, indicating a positive result and the presence of the target analyte. The average grayscale intensity of the colorimetric signal at the test site with AuNPs and the fluorescence signal with plasmonic-fluors monotonically increased with the concentration of the nanolabels (Fig. 1G and 1H). Notably, for both AuNPs and plasmonic-fluors, approximately ~10⁷ nanoparticles are needed to produce a discernable visible signal, however, only ~10³ plasmonic-fluors are enough to produce a detectable fluorescence signal. The four-order magnitude lower concentration threshold for a detectable signal with plasmonic-fluors compared to AuNPs in the LFA format is consistent with the drop-casting approach discussed above. These results manifest the fundamental basis that plasmonic-fluors can serve as ultrabright nanolabels for ultrasensitive detection of target analytes in an LFA.

Bioanalytical parameters of p-LFA compared to LFA

We optimized the bioanalytical performance of LFA by tuning concentration of capture ligand and nanolabels. We employed biotin-streptavidin as a model system. Both AuNPs and plasmonic-fluors were biotin functionalized, streptavidin and biotinylated BSA were utilized as target analyte and capture ligand, respectively (Supplementary Fig. 6). It was observed that as the concentration of capture-ligand (i.e., biotinylated BSA) increased, both mean grayscale intensity and fluorescence intensity of the test spot corresponding to AuNPs (Supplementary Fig. 7) and plasmonic-fluors (Supplementary Fig. 8), respectively, increased. These results suggest that higher concentrations of capture-ligand results in better signal intensity. Further, as the number of nanolabels increased, both mean grayscale intensity and fluorescence intensity of the test spot corresponding to AuNPs (Supplementary Fig. 9) and plasmonic-fluors (Supplementary Fig. 10), respectively, increased, implying better signal intensity with higher number of nanolabels. However, in both cases, the background signal (signal from the LFA strip outside the capture spot) also increased with the number of nanolabels. Therefore, the optimum number of nanolabels for both AuNPs-based LFA and p-LFA was determined by subtracting the background signal from the test spot signal. As expected, the optimum number of plasmonic-fluors (1.2 x 10⁶) was four-orders magnitude lower than the AuNPs (1.78 x 10¹⁰).

Next, we compared the bioanalytical parameters (limit-of-detection (LOD), limit-of-quantitation (LOQ) and dynamic range) of biotin-streptavidin AuNPs-based LFA and p-LFA. It is worth noting that colorimetric signal, obtained from the 8-bit ImageJ processed images of LFA strips, from both AuNPs and plasmonic-fluors exhibit similar LOD, suggesting no loss in visual detection capabilities in p-LFAs (Supplementary Fig. 11). The LOD (defined as mean + 3σ of the blank) of colorimetric LFA was calculated to be 4.8 ng ml⁻¹ (Supplementary Fig. 12, five-parameter logistic). In contrast, the fluorometric p-LFA enabled the detection down to 2.3 pg ml⁻¹ (Supplementary Fig. 13, five-parameter logistic fit), representing ~2000-fold improvement in the LOD. The LOQ (defined as mean + 10σ of the blank) of fluorometric p-LFA is ~2500-fold better than the LOQ of colorimetric LFA. Further, the fluorescent component of plasmonic-fluor augmented the dynamic range of the assay by three orders of magnitude. Therefore, owing to the ultrabright fluorescence signal of the plasmonic-fluors, the p-LFAs enable ultrasensitive detection of target analyte over a much broader range of analyte concentration.

p-LFA for quantitative detection of human IL-6

Cytokines are small (5-26 kDa) proteins, involved in cell signaling and immuno-modulation and are critical indicators of health and disease.⁴⁰ Several diseases including cancer, sepsis, HIV, chronic inflammation and auto-immune diseases are known to be associated with dysregulation of immune system, leading to disruption of the subtle balance between pro-inflammatory and anti-inflammatory cytokines.^41–42 The pro-inflammatory cytokines include IL-1 (interleukin-1), IL-6, IL-12, TNFα (tumor necrosis factor α) and IFNγ (interferon γ), while the anti-inflammatory cytokines include TGFβ (transforming growth factor β), IL-10 and IL-4. Rapid monitoring of the immune status by analyzing serum cytokines and early diagnosis of these diseases is essential for prompt clinical intervention and for inhibiting disease progression. Though few LFAs for IL-6 detection have been introduced recently,^43–44 none provide sensitivity and quantitation comparable to gold-standard ELISA. Therefore, we employed IL-6 as a model target analyte to investigate the applicability of our p-LFA.

Human IL-6 capture antibodies and sheep anti-immunoglobulin G (IgG) antibodies were immobilized on a nitrocellulose membrane to form test and control spots, respectively (Fig. 2A and Supplementary Fig. 14). The LOD of AuNP-based colorimetric LFA (Fig. 2B) and of molecular fluorophore-based LFA was calculated to be 166 pg ml⁻¹ (Fig. 2C, five-parameter logistic fit) and 362 pg ml⁻¹ (Supplementary Fig. 15), respectively. In contrast, the fluorometric p-LFA (Fig. 2D) enabled the detection down to 93 fg ml⁻¹ (Fig. 2E, five-parameter logistic fit), which represents a 1785-fold improvement in the LOD compared with conventional AuNP-based LFAs and at least an order magnitude higher than the previously reported LFAs ^43–46. The LOQ of fluorometric p-LFA (298 fg ml⁻¹) is 2288-fold better than the LOQ of colorimetric LFA (682 pg ml⁻¹). Further, the plasmonic-fluor improved the dynamic range of the LFA by nearly three-order magnitude. The colorimetric signal from both AuNPs and p-LFA exhibited similar LODs, suggesting no loss in visual detection capabilities in p-LFAs (Fig. 2F, 2G and Supplementary Fig. 16). Additionally, the fluorescence signal from the plasmonic-fluors enabled ultrasensitive detection and quantitative analysis over a much broader range of analyte concentration (Fig. 2E and 2H).

Fig. 2 |. Quantitative p-LFA of human IL-6.

(a) Schematic illustration of IL-6 LFA strips comprising an IL-6 capture antibody test spot and a sheep IgG control spot. (b) Schematic illustration of AuNP-based IL-6 LFA, and (c) dose-dependent mean gray values, corresponding to different IL-6 concentrations, acquired from these AuNP-based LFAs. (d) Schematic illustration of IL-6 p-LFA and (e) dose-dependent fluorescence intensities of IL-6 p-LFA. 8-bit, ImageJ processed images of (f) AuNP-based IL-6 LFAs and (g) IL-6 p-LFAs, depicting the visual readout mode. (h) Fluorescence images of the IL-6 p-LFA strips depicting the fluorescence readout mode. (i) Resolution of molecular concentration (RMC) curves for ELISA, p-FLISA and p-LFA (see Supplementary Information for calculations). The dashed lines indicate RMC cutoffs at μ=2 and μ=5; intersections of dashed lines and RMC curves indicate the range of concentrations over which a specific quantitative performance of the assay is achieved. For IL-6 p-LFA, μ < 2 over a concentration range over of 0.13–86.0 pg/mL, suggesting that IL-6 p-LFA can distinguish signals corresponding to any two concentrations within that range that differ by at least 100% with at least 99% confidence. The relevant RMC parameters are listed in Supplementary Table 1. (j) Stability of IL-6 p-LFA over 7 months, as evidenced by the error in concentration estimates of IL-6 concentration deduced using four different standard curves obtained over a span of seven months.

We also compared the sensitivity and LOD of fluorometric p-LFA with gold-standard ELISA and plasmonic-fluor linked immunosorbent assay (p-FLISA) implemented on a microtiter plate (Supplementary Fig. 17). The LOD of p-LFA is nearly 30-fold lower compared to conventional sandwich ELISA (2.9 pg ml⁻¹) and only 5-fold inferior to that of p-FLISA (16.8 fg ml⁻¹) (Supplementary Fig. 18). However, the sample-to-answer time for p-LFAs was 20 min whereas ELISA and p-FLISA require 4 h.

To evaluate the ability of fluorometric p-LFA to accurately resolve changes in concentration of human IL-6, we quantified the resolution of molecular concentration (RMC), a recently introduced metric that indicates whether changes in analyte concentration can be discriminated with statistical significance.⁴⁷ This metric is complementary to LOD: whereas the low LOD represents the smallest analyte concentration that can be distinguished from the background, RMC represents the smallest fold change in concentration that can be discriminated with 99% certainty.⁴⁷ We compared the RMC of ELISA, p-FLISA and p-LFA for resolution of two-fold changes in concentration of human IL-6 (RMC parameter μ=2, meaning a two-fold change in concentration could be resolved). The RMC curves for p-LFA exhibited μ ≤ 2 over a concentration range of 0.13–86.1 pg/mL, two orders of magnitude lower than that of ELISA (Fig. 2I), and nearly identical to that of p-FLISA. This suggests that IL-6 p-LFA can distinguish signals corresponding to two concentrations that differ by at least 100% within that range with at least 99% confidence. The RMC function and other bioanalytical parameters of p-FLISA and p-LFA, listed in Supplementary Table 1, indicate that the performance of the 20 min POC-compatible p-LFA is nearly identical to 4 h lab-based p-FLISA.

Next, to establish the stability of fluorometric p-LFA for quantitative detection without the use of standards, multiple IL-6 standard curves were acquired over a span of seven months (Supplementary Fig. 19). All standard curves attained similar RMC (Supplementary Fig. 20) and bioanalytical parameters, suggesting excellent repeatability and reproducibility. Using these standard curves, IL-6 concentrations ranging from 1 pg ml⁻¹ to 50 pg ml⁻¹ were quantified with less than 20% deviation (Fig. 2J and Supplementary Fig. 21).

Overall, the POC assay showed performance comparable to that of the lab-based assay, and showed the ability to accurately quantify the analyte concentration in a standard-free manner. This has not been reported previously with LFA technology, ascertains that p-LFAs overcome the long-standing limitations of LFAs – limited sensitivity, low accuracy and smaller analytical range compared to laboratory tests, and limited quantitation ability.

Ultrasensitive p-LFA for SARS-CoV-2 serology

To assess the potential for clinical translation of our p-LFA, we next optimized it for detection of SARS-CoV-2 antibodies. A pressing need persists for sensitive, rapid and POC serological assays for SARS-CoV-2, both for epidemiological studies and for vaccine efficacy against SARS-CoV-2 studies.^{3, 20} Several LFAs^{3–4, 48} and other assay methods⁴⁹ exist that employ SARS-CoV-2 spike protein as recognition element for detection of SARS-CoV-2 antibodies. Using p-LFA, our goal was to extend the sensitivity and limit of detection beyond the range possible with current assays, and into the range of ELISA.

Recombinant SARS-CoV-2 S1 subunit of spike protein was immobilized at the test spot and sheep IgG was used for control spot (Fig. 3A and Supplementary Fig. 22). We first determined the bioanalytical parameters of AuNP-based LFA (Fig. 3B) and p-LFA (Fig. 3D) for detection of SARS-CoV-2 S1 antibody. Using the colorimetric signal obtained from LFA strips, the LOD of AuNP-based LFA was determined to be ~ 1.05 μg ml⁻¹ (Fig. 3C). In contrast, fluorometric p-LFA exhibited an LOD of 185 pg ml⁻¹ (Fig. 3E, five-parameter logistic fit), which represents a nearly 5675-fold improvement. Further, as expected, the mean grayscale intensities obtained from both AuNP and p-LFA exhibited similar sensitivity, suggesting no compromise in the visual detection capabilities (Fig. 3F, 3G and Supplementary Fig. 23). However, the fluorescence signal from plasmonic-fluors enabled ultrasensitive detection and quantitative analysis over a much broader (four-orders of magnitude higher) range of analyte concentration (Fig. 3E and 3H). Fluorometric p-LFA displayed 165-fold improvement in LOD as compared to conventional sandwich ELISA (Fig. 3I) and comparable LOD to p-FLISA (Fig. 3J).

Fig. 3 |. SARS-CoV-2 serological p-LFA.

(a) Schematic illustration of the SARS-CoV-2 S1 antibody LFA strips comprising recombinant SARS-CoV-2 S1 protein as capture element at the test spot and sheep IgG at the control spot. Schematic illustrations of (b) AuNP-based SARS-CoV-2 S1 antibody LFA and (d) p-LFA. (c) Dose-dependent mean gray values, corresponding to different concentrations of SARS-CoV-2 S1 antibody, acquired from AuNP-based LFA. (e) Dose-dependent signal-to-noise ratio of SARS-CoV-2 S1 antibody p-LFA performed in 20 min. 8-bit ImageJ processed images of (f) AuNP-based SARS-CoV-2 S1 antibody LFA and (g) SARS-CoV-2 S1 antibody p-LFA, depicting the visual readout mode. (h) Fluorescence images of SARS-CoV-2 S1 antibody p-LFA strips, depicting the fluorescence readout mode. Dose-dependent optical densities and fluorescence intensities, corresponding to different SARS-CoV-2 S1 antibody concentrations, obtained by standard (i) ELISA and (j) p-FLISA implemented on a microtiter plate, performed in 4 h.

To assess the translational potential of fluorometric p-LFAs, we tested 79 plasma samples obtained from COVID-19 positive individuals and 48 archived de-identified serum/plasma samples which were collected pre-COVID-19 (March-October 2019) under HRPO 201102546⁵⁰ for the presence of SARS-CoV-2 S1 antibodies. All 127 plasma samples were diluted 500-fold and tested using fluorometric p-LFA. Out of 79 IgG positive samples (tested positive by ELISA), 76 were tested positive (sample SNR ≥ blank SNR + 3σ of blank) with p-LFA, indicating 96.2% sensitivity. All pre-COVID-19 samples tested negative with LFA for SARS-CoV-2 S1 IgGs, indicating 100% specificity (Supplementary Table 2). Thus, the p-LFAs for SARS-CoV-2 antibodies detection offers POC applicability with accuracy comparable to gold standard ELISA and with potential applicability to vaccine efficacy and epidemiological studies.

p-LFA for SARS-CoV-2 antigen detection

Next, we evaluated the potential of p-LFAs to fill the critical need for a highly sensitive and specific POC SARS-CoV-2 antigen test. In serological testing of virus-specific immunoglobulins, the antibody responses to viral antigens are usually detected in the late stage of infection (7–14 days after virus exposure), therefore serological antibody tests cannot achieve accurate screening of asymptomatic populations or early stages of infection.⁵¹ Further, RT-PCR, the current gold standard in diagnosing COVID-19, has proven highly successful in identifying individuals who have contracted the SARS-CoV-2 virus, however, they may fail to distinguish between infectious patients and noninfectious individual, and may yield false positive results for months even after a patient has recovered from the disease.^52–53

Since antigens are expressed only when the virus is actively replicating, the antigen-based tests may have better correlation with infectiousness than RNA detection by RT-PCR. Current antigen detection tests for diagnosing COIVD-19 are scalable and convenient but are limited by their low and wide-ranging accuracy.^54–57 LFAs for detection of SARS-CoV-2 antigens can be the most important tool in addressing the infection outbreaks owing to their ease of use, lower-cost and better correlation with infectivity. Currently, several LFA-based antigen^{6–7, 58} assays have been reported and are widely used but none offers the optimal sensitivity,⁵⁹ thus, a negative result with such assays in a symptomatic patient requires a confirmatory RT-PCR test or frequent retesting. Therefore, there is an urgent need for a more sensitive POC antigen assay that would be just as reliable and accurate as the RT-PCR method.

p-LFA provided the accuracy and sensitivity needed for this in samples from patients who simultaneously had PCR tests performed. Our test focused on the detection of SARS-CoV-2 nucleocapsid protein (N protein). Test and the control spots on the LFA strips were prepared by immobilizing N protein capture antibodies and sheep IgG, respectively (Fig. 4A and Supplementary Fig. 24). Both colorimetric and fluorescence signals obtained from p-LFAs increased monotonically with an increase in the concentration of N protein standard (Fig. 4B, 4C). However, the LOD and LOQ of fluorometric p-LFA were calculated to be nearly 400-fold better than colorimetric counterpart, ascertaining the importance of plasmonic-fluors as ultrabright fluorescent nanolabels (Fig. 4D). Further, fluorometric p-LFA displayed 37-fold improvement in LOD as compared to conventional sandwich ELISA and comparable LOD to p-FLISA (Fig. 4E and Supplementary Fig. 25).

Fig. 4 |. p-LFA for SARS-CoV-2 N protein and variants-of-concern.

(a) Schematic illustration of the nucleocapsid (N) protein p-LFA strips comprising of N protein capture antibody as test spot and sheep IgG as control spot. (b) Colorimetric and (c) fluorometric readout modes of p-LFA for N protein detection. (d) Dose-dependent mean gray values, corresponding to different concentrations of N protein, acquired from colorimetric p-LFA (black) and dose-dependent signal-to-noise ratio of N protein fluorometric p-LFA performed in 20 min (red). (e) Dose-dependent optical densities and fluorescence intensities, corresponding to different N protein concentrations, obtained by standard ELISA (black) and p-FLISA (red) implemented on a microtiter plate, performed in 4 h. (f) Comparison of fluorometric p-LFA (red) and commercial point-of-care rapid antigen kit (BD Veritor^™) (black). (g) N-protein signal-to-noise ratio in PCR-positive NP swab samples (wild type SARS-CoV-2) determined by colorimetric p-LFA (gray), fluorometric p-LFA (black) and BD Veritor^™ (marked with #). (h) Comparison of colorimetric (gray) and fluorometric (black) p-LFA in terms of their ability to quantify N protein concentrations present in NP swab samples of 35 PCR-positive samples (19 wild type SARS-CoV-2 and 16 Delta variant). (i) N protein signal-to-noise ratio in NP swab samples tested negative for COVID-19 and positive for different seasonal coronaviruses and other respiratory viruses.

Next, to demonstrate the advantage of p-LFAs over an existing commercial FDA EUA approved rapid, point-of-care antigen testing method, we compared the analytical sensitivity of p-LFAs with BD Veritor^™ assay, which indicated samples with concentrations below 50 ng/ml as “Presumptive Negative” (Fig. 4F). This implies that the fluorometric p-LFA offers nearly 235-fold better analytical sensitivity as compared to the commercial antigen test. p-LFA outperformed the FDA-approved BD Veritor^™ antigen kit when analyzing PCR-positive COVID-19 patient samples (wild type SARS-CoV-2). BD Veritor^™ antigen kit and colorimetric p-LFA correctly identified 8 out of 19 PCR-positive NP swab samples (analytical sensitivity: 42.1%), whereas fluorometric p-LFA correctly identified 18/19 samples (analytical sensitivity: 94.7%) (Fig. 4G). 13/14 patient samples in the early stage of illness (<10 days since symptoms onset) were tested positive by fluorometric p-LFA (93% sensitivity), while only 7 tested positive by BD Veritor^™ (50% sensitivity) (Supplementary Table 3). Notably, FDA-approved BD Veritor^™ antigen kit can only be used in negative/positive format, however, fluorometric p-LFA enabled quantitative detection of target analyte in patient samples (Supplementary Table 4). We also compared the quantitative performance of colorimetric and fluorometric p-LFA. While only 3/19 samples were quantifiable (above LOQ) via colorimetric p-LFA, 18/19 samples were quantifiable via fluorometric p-LFA (Fig. 4H).

To further substantiate the clinical translational potential of fluorometric p-LFAs for the detection of N protein, we tested 16 PCR-positive Delta B.1.617.2 variant (confirmed by gene sequencing) NP swab patient samples. Colorimetric p-LFA detected N protein in 7/16 Delta variant positive samples, of which only 3 were quantifiable. However, fluorometric p-LFA detected N protein in all 16 samples, of which 15 were quantifiable (above LOQ) (Fig. 4H, Supplementary Fig. 26 and Supplementary Table 5). We also tested 17 PCR-positive Omicron BA.1 (confirmed by gene sequencing) samples and observed that fluorometric p-LFA returned positive results for 16/17 Omicron (Supplementary Fig. 27 and Supplementary Table 6) variant samples. All Omicron-positive patient samples were collected within a short duration from the onset of symptoms (1-2 days) and all, but one patient had very mild illness (Supplementary Table 7). These findings establish the efficacy of p-LFA in early detection of N protein.

A total of 52 PCR-positive samples was tested. While only 15/52 returned positive result with colorimetric p-LFA, indicating 28.8% clinical sensitivity, 50/52 tested positive with fluorometric p-LFA (SNR > mean + 3σ), indicating 96.2% analytical sensitivity. The diagnostic sensitivity of p-LFA for samples with low viral load (cycle threshold (CT) values ≥ 25) was 91.7% (11 out of 12) and for samples with high viral load (CT values < 25) was 97.5% (39 out of 40). This diagnostic sensitivity was substantially higher than those previously reported for rapid antigen/POC SARS-CoV-2 tests (~80% for samples with CT values < 25 and 20-40% for samples with CT values ≥ 25).^{7, 59–61}

Finally, to evaluate the specificity of p-LFA to SARS-CoV-2 N protein, we tested 19 PCR-negative NP swab samples. The negative NP swab samples comprised a mix of healthy samples, and samples tested positive for seasonal coronaviruses and other respiratory viruses. All the 19 PCR-negative samples tested negative (SNR < mean of blank + 3σ) using p-LFA, suggesting 100% analytical specificity to COVID-19 N protein and no cross-reactivity with different seasonal coronaviruses and other viruses (Fig. 4I). These results substantiate that p-LFAs enable ultrasensitive, accurate, rapid, inexpensive, and point-of-care diagnosis of COVID-19 antigen and antibodies and thus can be a potential tool for rapidly and quantitative diagnosis of symptomatic and asymptomatic infections.

Point-of-care p-LFA using an inexpensive, portable fluorescence scanner.

Finally, to determine the applicability of this biodiagnostic technology in POC settings, we validated the performance of p-LFA using a portable, inexpensive fluorescence scanner. Note that the Stokes shift corresponding to plasmonic-fluors is much smaller (~15 nm) than those corresponding to commonly employed fluorescent nanoparticles such as quantum dots and europium nanoparticles (100s of nm)^62–63. To the best of our knowledge, no inexpensive, portable fluorescence scanner compatible with plasmonic-fluor is available commercially. Therefore, we developed an inexpensive portable fluorescence scanner for reading p-LFA employing plasmonic-fluors as nanolabels. The scanner prototype, with dimensions 25 x 25 x 19 cm (LxBxH), was built using routinely available, off-the-shelf optical components (see Methods section for detailed description Fig. 5A and Supplementary Fig. 28). The total cost of the scanner is $1429, and the most expensive component is the laser, which costs $919. It is worth noting that for all the measurements described in this work, the laser (excitation source) power was set to 1% of the maximum power to avoid fluorescence signal saturation. Thus, these components can be miniaturized and replaced with less expensive components for commercialization. Also, the portable scanner can run from a battery and thus can be immediately deployed in resource-limited settings.

Fig. 5 |. Validation of p-LFA using an inexpensive, portable fluorescence scanner.

(a) Photograph of the portable fluorescence scanner. (b) Fluorescence intensities (black spheres) and area under the curve values (red spheres) obtained from LFA strips, drop-casted with different concentrations of plasmonic-fluors, scanned using benchtop and portable scanners. (c) Schematic illustration of the LFA cassette employed in the study and the workflow of p-LFA. S, C and T correspond to the sample pad, test line and control line, respectively. The blue arrow represents the direction of the fluorescence measurements made on the LFA cassette using the portable scanner. (d) Representative positive (black) and negative (red) signals obtained using the portable scanner. (e) 8-bit ImageJ processed image of the full strip IL-6 colorimetric p-LFA depicting the visual readout mode. (f) Fluorescence image of the full strip IL-6 fluorometric p-LFA depicting the fluorescence readout mode. (g) Dose-dependent signal of 15 min IL-6 fluorometric p-LFA measured by benchtop (black) and portable scanners (red). (h) Linear regression plot of IL-6 concentration in serum samples determined by fluorometric p-LFA, and measured benchtop and portable scanners. (i) Linear regression plot of IL-6 concentration in serum samples determined by 4 h lab-based p-FLISA and a benchtop fluorescence scanner, compared to measurements made using 15 min fluorometric p-LFA and the portable scanner. (j) Linear regression plot of N protein concentration in NP swab samples determined by fluorometric p-LFA and measured using the benchtop and portable scanners. Data are mean ± standard deviation., n = 2x2 repeated tests.

We fabricated full-strip LFAs with separate sample and conjugate pads along with test membranes and absorbent pads. The assembled strip was embedded into a standard LFA cassette (Fig. 5B and Supplementary Fig. 29). Fluorescence measurements were performed by translating the cassette using travel actuator along the optical system of the portable scanner in the direction of the blue arrow in Fig. 5B. This produced trace of pixel value (signal intensity), averaged from 10 images, versus the travel length of the test membrane, taken in 100 μm increments (Fig. 5C). Consequently, a valid positive result has peaks at test and control lines, and a negative result has a peak only at the control line. A test without a peak at the control line is considered to be an invalid result (Supplementary Fig. 30).

First, to compare the performance of the portable scanner with the benchtop scanner, we determined the minimum number of nanolabels that could be detected by each scanner. When serially diluted plasmonic-fluors (Fig. 5D and Supplementary Fig. 31) and 800CW molecular fluorophores (Supplementary Fig. 32 and 33) of known concentration were drop-casted on the test membranes, accumulations of ~100 plasmonic-fluors were needed to produce detectable fluorescence intensity (mean of blank + 3σ) when measured using the benchtop scanner, and ~200 plasmonic-fluors were needed for the portable scanner (Fig. 5D). Accumulations of ~0.6×10⁶ molecular fluorophores (not plasmonically enhanced) were needed to produce detectable fluorescence intensity when measured using either the benchtop or portable scanner. Data acquired by the portable scanner and subsequent data processing methodology is discussed in detail in the supplementary information (Supplementary Fig. 34 and 35). These observations indicate nearly identical performance of the benchtop and portable scanners in detecting the fluorescence signal from plasmonic-fluors.

Next, to demonstrate the POC-compatible workflow of p-LFA and compare the performance of the portable and benchtop scanners, we employed human IL-6 as a model analyte. Human IL-6 capture antibodies and sheep anti-immunoglobulin G (IgG) antibodies were printed on a nitrocellulose membrane to form test and control lines, respectively (Supplementary Fig. 36). The LOD of the colorimetric IL-6 p-LFA in full strip format (Fig. 5E) was calculated to be ~ 526 pg ml⁻¹ (Supplementary Fig. 37). In contrast, the fluorometric p-LFA (Fig. 5F and Supplementary Fig. 38) enabled the detection down to 813 fg ml⁻¹ (Fig. 5G in black, five-parameter logistic fit), measured using the benchtop scanner. Notably, with the portable scanner, the IL-6 p-LFA exhibited similar LOD, 916 fg ml⁻¹ (Fig. 5G in red, five-parameter logistic fit). The near identical performance of benchtop and portable scanners was further confirmed by comparing the N protein dose-response curve (Extended Data Fig. 1, Supplementary Fig. 39 and 40). Note that the LOD of the full-strip LFAs is higher compared to the half-strip format discussed above due to the shorter time (15 min vs 20 min) and smaller analyte volume (70 μl vs 100 μl) available for binding of the analytes to the capture antibody conjugated-nanolabels in the full-strip format.

Finally, to demonstrate the clinical translational potential and the possible POC application of p-LFA with the portable scanner, we tested 28 serum and 14 NP swab samples from COVID-19 PCR-positive individuals for detection of IL-6 (Extended Data Fig. 2 and 3) and N protein (Supplementary Fig. 41 and 42), respectively. These samples were tested by 15 min p-LFAs and measured using benchtop and portable scanners. Quantitative results from the benchtop and portable scanners exhibited excellent correlation with a Pearson’s r value of 0.97 for IL-6 (Fig. 5H) and 0.94 for N protein concentrations (Fig. 5J, Supplementary Table 8). Equally important, the IL-6 concentrations determined by 15 min p-LFA and measured by the portable scanner also exhibited excellent correlation with those determined by 4 h long lab-based p-FLISA (Pearson’s r value of 0.91) (Fig. 5I and Supplementary Table S9).

This observation, along with the nearly identical bioanalytical parameters of the p-LFA standard curve generated using the benchtop and portable scanners for IL-6 and N protein, suggests that the sensitivity and quantitative detection ability of p-LFA is not compromised by the use of an inexpensive, portable fluorescence scanner. Results using this portable scanner were comparable to those obtained using the 4 h long, lab-based tests performed using the expensive, non-portable benchtop fluorescence scanner. These results highlight the simple workflow of p-LFA and its potential for biodiagnostics in POC settings.

In summary, plasmonic-fluors were demonstrated as a bimodal (colorimetric+fluorescent) reporter element for overcoming long-standing limitations of LFAs. Specifically, p-LFA overcomes the limited sensitivity, low accuracy, small dynamic range, and limited quantitation ability of LFAs compared to laboratory tests. Plasmonic-fluors produced a discernable fluorescence signal at densities 10000-fold lower than those needed in conventional colorimetric AuNPs. p-LFAs for various analytes (IL-6, SARS-CoV-2 S1 antibodies, and SARS-CoV-2 antigen) exhibited ~1000-fold improvement in bioanalytical parameters (LOD, LOQ and dynamic range) over conventional LFAs. p-LFAs offered standard-free quantitative detection with over 10-fold better sensitivity than that of gold standard ELISA, with a much lower sample-to-answer time (20 min versus 4-6 hours) and similar ability to resolve molecular concentration as lab-based tests. p-LFAs for detection of COVID-19 antibodies and antigens present in plasma and nasopharyngeal swab samples achieved >95% sensitivity and 100% specificity, demonstrating clinical applicability. The inexpensive and portable fluorescence scanner we developed and optimized for reading p-LFA was as effective as the benchtop scanner we used. When applied to human specimens of COVID-19 positive individuals, concentrations of IL-6 and N protein measured for 15 min p-LFAs using the benchtop and portable scanners exhibited excellent correlation with each other, and also with concentrations determined by lab-based 4 h p-FLISA. We believe p-LFAs are highly attractive for realizing POC biodiagnostics that require accurate and quantitative detection of bioanalytes. The technology demonstrated here can be readily adapted for the detection of other infectious pathogens and disease biomarkers, and can complement or even replace laboratory-based tests for the diagnosis of pathogenic infections and other acute conditions.

Methods

Synthesis of plasmonic-fluors

Plasmonic-fluors consists of plasmonically active core, gold nanorod synthesized by seed-mediated method,⁶⁴ a polymer spacer layer, fluorophores and universal biorecognition element (biotin). Plasmonic-fluors were synthesized following the similar procedure described in our previous study.²⁷ Detailed stepwise procedure is discussed in the supplementary information.

Synthesis of gold nanoparticles (AuNPs)

Citrate-stabilized AuNPs were synthesized using seed-mediate synthesis method and using citrate as reducing agent. Au seeds (~15 nm) were synthesized as described previously by Frens et al.⁶⁵ Briefly, 20 ml 0.25 mM of HAuCl₄ (Sigma Aldrich, 520918) was brought to boil under vigorous stirring, 800 rpm. Immediately after the solution started boiling, 0.2 ml of 3% (w/v) sodium citrate (Sigma Aldrich, 1613859) aqueous solution was added and maintained under boiling condition until the solution color changed to wine red, indicating the formation of Au seeds. Next ~100 nm AuNPs were synthesized using hydroquinone (Sigma Aldrich, H9003) as reducing agent for reduction of ionic gold.

Materials characterization

TEM images were obtained using a JEOL JEM-2100F field emission instrument. The extinction spectra of plasmonic nanostructures were obtained using a Shimadzu UV-1800 spectrophotometer. Fluorescence mappings were recorded using LI-COR Odyssey CLx imaging system. Digital camera (Sony cybershot DSC HX300) and imaging software, ImageJ were employed to characterize mean gray intensities. SpectraMax iD3 (Molecular Devices) plate reader was used to measure the optical density in ELISA.

Functionalization of nanolabels

To functionalize nanolabels with streptavidin (Sigma Aldrich, SA101), 1 μl 10 mg ml⁻¹ of streptavidin (or BSA-Biotin or detection antibody) was added to 1 ml OD1 of nanolabels and incubated for 1 h on a shaker at room temperature. To stabilize the particles, 1 ul 10 mg ml⁻¹ of BSA (Sigma Aldrich, A7030) was added to the solution and further incubated for 20 min. Unbound protein was removed by washing the solution four times with pH 10 nanopure water (1 μl NaOH in 10 ml of water). Finally, nanolabels were redispersed in 1% BSA in 1X PBS solution for use in the LFAs. To functionalize nanolabels with antibodies (IL-6 and N protein detection antibody and anti-human IgG), similar process was employed.

Lateral flow immunoassay assembly and preparation procedures

Nitrocellulose test membrane and absorbent pads with adhesive backing material (GE healthcare, FF120HP) were employed for fabricating the LFA strips. The test membrane and absorbent pad was cut into 4 mm wide strips using a paper trimmer. For preparing the LFA strip, biorecognition element (e.g., capture antibody) solution was pipetted onto the test membrane and dried at room temperature for 30 min. Subsequently, the test membrane was blocked using 3% BSA in 1X PBS solution. Next, strips were washed with PBST (1X PBS and 0.5% Tween20 (Sigma Aldrich, P9416), followed by drying at room temperature in a vacuum desiccator for 1 h. After drying, absorbent pads (GE healthcare, CF5) were assembled onto the polystyrene adhesive backing next to nitrocellulose test membrane. To ensure efficient transfer of the solution from the test membrane to the absorbent pad, we ensured an overlap of 1-2 mm between both strips. Experiments were performed by dipping LFAs into 96-well plates filled with 100 μl of sample/standard solutions for 20 min. The visual signals of LFAs were obtained by a digital camera. The images were converted to 8-bit gray scale image using ImageJ. Mean gray values of the test spot were calculated by averaging the test spot grayscale intensities obtained from ImageJ. The fluorescence signals were obtained by averaging test dot fluorescence intensities obtained using LI-COR Odyssey CLx fluorescence scanner using the following scan parameters: laser power~L2; resolution 21 μm; channel 800 nm; height 0 mm.

Optimization of lateral flow immunoassay parameters

To determine the optimum concentration of biotinylated BSA on the test spot, different LFA strips with varying concentrations of biotinylated BSA (100 μg ml⁻¹ to 5 mg ml⁻¹) were prepared in duplicates. LFAs were then subjected to the same concentration of streptavidin (1000 ng ml⁻¹ for AuNP-LFA and 1 ng ml⁻¹ for p-LFA) and biotinylated nanolabels. To determine the optimal concentration of the nanolabels, LFA strips with the same concentration of biotinylated BSA (5 mg ml⁻¹) were prepared in duplicates. These LFA strips were then subjected to the same concentration of streptavidin (1000 ng ml⁻¹ for AuNPs and 1 ng ml⁻¹ for plasmonic-fluors) but different numbers of biotin-functionalized nanolabels (4.45x10⁶ to 3.56x10¹⁰ for AuNPs and 1.2x10⁴ to 6x10⁶ for plasmonic-fluors). The optimum number of nanolabels for colorimetric AuNPs-LFA and p-LFA, and fluorometric p-LFA was determined by subtracting the background signal from the test spot signal.

Biotin-streptavidin lateral flow immunoassay

Test spots were formed by pipetting 0.5 μl of 5 mg ml⁻¹ biotinylated-BSA onto the nitrocellulose membrane. The LFA strips were assembled as described above. For AuNP-based and plasmonic fluor-based Biotin-streptavidin LFA, 1 μl of biotinylated AuNPs and 1 μl of biotinylated plasmonic-fluors, respectively, were mixed with 99 μl of different concentrations of streptavidin standard solutions (0.1 pg ml⁻¹ to 1000 μg ml⁻¹) in 96-well plates to allow the binding of streptavidin with the biotinylated nanolabels. LFA strips in duplicates were then exposed to the sample/standard solution for 20 min.

Human IL-6 immunoassays

Human IL-6 DuoSet ELISA kit (R&D systems, DY206) was utilized in the study. For AuNP-based IL-6 LFA, AuNPs were conjugated with IL-6 detection antibody for the test spot and with anti-sheep IgG (R&D systems, BAF016) for the control spot. For p-LFA, plasmonic-fluors were conjugated with IL-6 detection antibody for the test spot and AuNPs were conjugated with anti-sheep IgG for the control spot, respectively. To prepare LFA strips for IL-6 immunoassay, 0.5 μl of 2 mg ml⁻¹ IL-6 capture antibody and 0.5 μl of 2 mg ml⁻¹ sheep IgG (R&D systems, 5-001-A) was pipetted onto the nitrocellulose membrane at different spots to create test and control spot, respectively. Subsequently similar steps, mentioned above, were followed for LFA preparation and assembly. For AuNP-based IL-6 LFA, 1 μl of IL-6 detection antibody-conjugated AuNPs and 1 μl of anti-sheep IgG conjugated-AuNPs for test and control spot, respectively, were mixed with 98 μl of different concentrations of human IL-6 standard solutions (64 fg ml⁻¹ to 5 ng ml⁻¹) in 96-well plates to allow the binding of the analyte with the detection antibody-conjugated nanolabels. LFA strips in duplicates were then exposed to the sample/standard solution for 20 min. For IL-6 p-LFA, 1 μl of IL-6 detection antibody-conjugated plasmonic-fluors and 1 μl of anti-sheep IgG conjugated AuNPs were mixed with 98 μl of human IL-6 standard solutions (1 fg ml⁻¹ to 1 ng ml⁻¹) in 96-well plates. The visual signals and the fluorescence signals were obtained according to the procedure described above.

Human IL-6 ELISA was carried out according to the procedure described in DuoSet ELISA kit manual and is discussed in detail in supplementary information. Plasmonic fluor-linked immunosorbent assay (p-FLISA) was performed by adopting a similar approach, expect that the HRP-labeled streptavidin was replaced by streptavidin-functionalized plasmonic-fluor. Instead of streptavidin-HRP, 100 μl of streptavidin-plasmonic-fluors (OD 1) was incubated for 30 min, and then the plate was washed three times with PBST. Both ELISA and p-FLISA were conducted in duplicates. The fluorescence signal was obtained by averaging the fluorescence intensities from the microtiter wells obtained using LI-COR Odyssey CLx with the following scan parameters: laser power~L2; resolution 169 μm; channel 800 nm; height 4 mm.

Lateral flow immunoassay quantitation study

Four p-LFA IL-6 standard curves (1 fg ml⁻¹ to 1 ng ml⁻¹) were generated over a span of 6 months and samples with varying IL-6 concentrations (0.5 pg ml⁻¹ to 62.5 pg ml⁻¹) were tested in duplicates in a standard-free manner. Their experimental concentrations were determined using each standard curve, and deviation from actual concentrations were calculated.

SARS-CoV-2 S1 antibody immunoassays

We pipetted 0.5 μl of 2 mg ml⁻¹ recombinant SARS-CoV-2 S1 protein (R&D systems, 10522-CV) and 0.5 μl of 2 mg ml⁻¹ sheep IgG onto the nitrocellulose membrane as test and control spot, respectively. Subsequently, we followed the same steps described above to prepare the LFA strips. For detecting SARS-CoV-2 S1 antibodies, AuNP-LFA and p-LFA, AuNPs and plasmonic-fluors were conjugated with biotinylated anti-human IgG (Rockland, 609-4617) for test spots, respectively. In both cases, AuNPs were conjugated with anti-sheep IgG for control spot. For AuNP-based SARS-CoV-2 S1 antibody LFA, 1 μl of anti-human IgG conjugated-AuNPs and 1 μl of anti-sheep IgG conjugated-AuNPs were mixed with different concentrations of standard solutions (16 pg ml⁻¹ to 25 μg ml⁻¹) in 96-well plates, prior to exposure to LFA strip for 20 min. For plasmonic-fluor-based SARS-CoV-2 S1 antibody LFA, 1 μl of anti-human IgG conjugated-plasmonic-fluors and 1 μl of anti-sheep IgG conjugated-AuNPs were mixed with different concentrations of standard solutions (16 pg ml⁻¹ to 1 μg ml⁻¹) in 96-well plates, prior to exposure to LFA strip for 20 min. Plasma samples were diluted 500-fold in reagent diluent (1X PBS containing 3% BSA, 0.2 μm filtered) before use. All experiments were done in duplicates. The visual signals and the fluorescence signals were obtained by employing the same procedure mentioned above.

SARS-CoV-2 S1 antibody ELISA was carried out according to the following procedure. Microtiter wells in duplicates were coated with 100 μl of 5 μg ml⁻¹ (in 1X PBS) recombinant SARS-CoV-2 S1 protein via overnight incubation at room temperature. For blocking, 300 μl of reagent diluent was added to the wells for a minimum of 1 h. Next, 100 μl of serially-diluted standard samples were incubated for 2 h, followed by incubation of 100 μl of 100 ng ml⁻¹ biotinylated anti-human IgG for 2 h. Next, 100 μl of 500 ng ml⁻¹ streptavidin-labelled HRP (Thermo Fisher scientific, N100) was incubated for 20 min, followed by the addition of 100 μl of substrate solution for 20 min. The reaction was stopped by addition of 50 μl of 2N H₂SO₄ (R&D Systems, DY994) and immediately the optical density at 450 nm was measured using a microplate reader. p-FLISA was carried out by adopting a similar procedure, expect that the HRP-labelled streptavidin was replaced by streptavidin functionalized-plasmonic-fluor. Instead of HRP, 100 μl of plasmonic-fluors (OD 1) were incubated for 30 min, and then the plate was washed three times with PBST. The fluorescence signal was obtained by averaging the fluorescence intensities from the microtiter wells obtained using LI-COR Odyssey CLx.

SARS-CoV-2 antigen (nucleocapsid protein) immunoassays

We pipetted 0.5 μl of 2 mg ml⁻¹ nucleocapsid protein capture antibodies (SinoBiologicals, 40143-MM08) and 0.5 μl of 2 mg ml⁻¹ sheep IgG onto the nitrocellulose membrane as test and control spots, respectively. For N protein p-LFA, plasmonic-fluors were conjugated with biotinylated N protein detection antibody (SinoBiologicals, 40143-R004) for the test spots. AuNPs conjugated with anti-sheep IgG were employed for control spot. Subsequently, similar steps mentioned above were followed to prepare and assemble the LFA strips. For plasmonic-fluor-based N protein LFA, 1 μl of detection antibodies conjugated-plasmonic-flours and 1 μl of anti-sheep IgG conjugated-AuNPs were incubated with different concentrations of standard solution (12 pg ml⁻¹ and 1 μg ml⁻¹; SinoBiologicals, 40588-V08B) spiked in universal transport media in 96-well plates prior to exposure to LFA strips for 20 min. p-LFAs were employed for the detection of N protein present in patient NP swab samples. The NP swab samples were in universal transport media and were used without any dilution or processing. All experiments were performed in duplicates. The visual signals and the fluorescence signals were obtained employing the similar process described above.

N protein ELISA was carried out by first coating the microtiter wells in duplicates with 100 μl of 100 ng ml⁻¹ N protein capture antibodies (in 1X PBS) via overnight incubation at room temperature. For blocking, 300 μl of reagent diluent was added to the wells for a minimum of 1 h. Next, 100 μl of serially-diluted standard samples were incubated for 2 h, followed by incubation of 100 μl of 200 ng ml⁻¹ biotinylated N protein detection antibody for 2 h. Next, 100 μl of 500 ng ml⁻¹ streptavidin-labelled HRP (Thermo Fisher scientific, N100) was incubated for 20 min, followed by the addition of 100 μl of substrate solution for 20 min. The reaction was stopped by addition of 50 μl of 2N H₂SO₄ (R&D Systems, DY994) and immediately the optical density at 450 nm was measured using a microplate reader. p-FLISA was carried out by adopting a similar procedure, expect that the HRP-labelled streptavidin was replaced by streptavidin-functionalized plasmonic-fluor. Instead of HRP, 100 μl of plasmonic-fluors (OD 1) were incubated for 30 min, and then the plate was washed three times with PBST. The fluorescence signal was obtained by averaging the fluorescence intensities from the microtiter wells obtained using LI-COR Odyssey CLx.

Commercial antigen test

BD Veritor kit, Veritor System – For Rapid Detection of SARS-CoV-2, was used to analyze the presence of N protein in the patient samples. BD Veritor System was used in conjunction with the BD Veritor Plus Analyzer. NP swabs were eluted in Universal Transport Media (UTM) and Aimes (ESwab) transport medium. Internal validation and the assay precision was conducted and deemed acceptable for testing on clinical samples by the Barnes Jewish Clinical Microbiology Laboratory.

Patient sample acquisition

The clinical samples used in the study were acquired from the repository of saliva, serum, plasma and nasopharyngeal swab samples from individuals confirmed/suspected with COVID-19 disease, located at Washington University School of Medicine in St Louis, and from the Barnes Jewish Clinical Microbiology Laboratory, and supported by: the Barnes-Jewish Hospital Foundation; the Siteman Cancer Center grant P30 CA091842 from the National Cancer Institute of the National Institutes of Health; and the Washington University Institute of Clinical and Translational Sciences grant UL1TR002345 from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH). This repository was developed and is maintained by Jane O’Halloran, MD, PhD; Charles Goss, PhD, and Phillip Mudd, MD, PhD. Control NP swab samples from asymptomatic healthy volunteers were obtained with prior written consent. For evaluation of cross reactivity with seasonal coronaviruses, samples were obtained from adults at Barnes-Jewish Hospital who were tested positive with either of the four seasonal coronaviruses or respiratory diseases via clinically warranted NP samples tests. Washington University School of Medicine Human Research Protection Office (HRPO) approved the study. All clinical data pre-existed at the time of data collection. A prior waiver of consent was obtained for the clinical information and data on COVID-19 PCR results.

Preparation and assembly of full-strip LFAs

Full strip p-LFA components include - NC membrane: FF80HP on polystyrene backing (cat: 10547020, from Whatman, Cytiva) Sample pad: Fusion 5 (cat: 8151-9915, from Whatman, Cytiva) Conjugate pad: Whatman STANDARD 14 (8133-2250, Cytiva) Absorption pad: CF5 (cat: 8115-2250, Cytiva). Sample and conjugate pads were subjected to following pre-treatment process. Sample pad was soaked in 5% BSA, 0.5% Tween 20, 1X PBS and then dried in 37 °C oven for 2 h. Conjugate pad was soaked in 5% BSA, 10% sucrose, 0.5% Tween 20, 1X PBS and then dried in 37 °C oven for 2 h. After pre-treatment, sample and conjugate pad were cut into a strip of 15 mm * 25 mm and 13 mm * 25 mm dimensions, respectively. Absorption pads were used as received and were cut into 18 mm * 25 mm dimensions.

To prepare nanolabels for test line, 1-3 μl of biotinylated SARS-CoV-2 N protein or human IL-6 detection antibody of 1 mg ml⁻¹ concentration were added to 1 ml of streptavidin functionalized plasmonic-fluors of extinction 2. After 30 min incubation, 100 μl of 10% BSA in 1X PBS was added to this antibody-conjugated plasmonic fluor solution. After another 30 min incubation, the conjugated nanolabel solution was centrifuged three times to remove unbound detection antibodies and the subsequent solution was dispersed back to 2 mM sodium borate, pH 8.5 with 10% sucrose. For preparation of nanolabels for control line 1-5 μl of biotinylated anti-goat IgG of 2 mg ml⁻¹ concentration was added to 1 ml of streptavidin plasmonic-fluors of extinction 2. After 30min incubation, 100 μl of 10% BSA in 1X PBS was added to this antibody-plasmonic fluors conjugate solution. After another 30 min incubation, the conjugated nanolabel solution was centrifuged three times and dispersed back to 2 mM sodium borate of pH 8.5 consisting of 10% sucrose.

Next, the nanolabel for test and control line were mixed in 1:1 ratio. Thereafter, the resulting solution was sprayed on to the pre-treated conjugate pad. The nanolabel solution was air-jet sprayed with a dispense rate of 5 μl cm⁻¹ employing a reagent dispenser (XYZ Platform Dispenser HM3030, Kinbio, Shanghai). After spraying conjugate pads were dried in 37 °C oven for 2 h. Next, the test membrane were prepared by printing the capture antibodies specific to test and control lines. For test line SARS-CoV-2 Ag and human IL-6 capture antibody of 1mg ml⁻¹ concentration, and for control line goat IgG of 2 mg ml⁻¹ concentration were simultaneously printed on FF80HP nitrocellulose test membrane at a dispense rate of 0.5 μl cm⁻¹ and speed of 50 mm s⁻¹ by a reagent dispenser (XYZ Platform Dispenser HM3030, Kinbio, Shanghai). Thereafter, the membranes were dried in 37 °C oven for 2 h.

Finally, the pre-treated sample pad, the conjugate pad after spraying of nanolabels, and the membrane pad after printing of capture antibodies were assembled with a 2 mm overlap between each pad and cut to strips with a width of 3 mm using a strip cutter (Programmable Strip Cutter ZQ2002, Kinbio, Shanghai). For the schematic illustration of the design of p-LFA please refer to Supplementary Fig. 45.

Portable fluorescence scanner

An 80 mW 785 nm diode laser (Zlaser, Z80M18S3-F-785-pe) was used as an excitation source. The laser beam was attenuated with an ND 2.0 neutral density filter and shaped into a 4 mm wide line using the combination of the laser focus control and a 30 mm focal length cylinder PCX lens. Fluorescence was collected with a 30 mm focal length PCX lens (12.5 mm diameter) and passed through an 832/37 nm emission filter (Edmund Optics, 84-107). A 45 mm focal length achromatic doublet lens (Edmund Optics, 49-355) was used to form a 1.5x magnified image of the lateral flow strip on the sensor of the camera (ZWO ASI462MC). Fluorescence was measured at a 45° angle relative to excitation. Measurements from lateral flow cassettes were carried out by translating the sample (using Actuonix L16-R 50 mm travel actuator) through the optical system at 1 mm/s while streaming the camera video. Video was collected with 100 ms exposure (10 images per second). The average pixel value from each 10 images was used for analysis and corresponded to one point in the trace produced by this instrument. A Raspberry Pi 4 single board computer was used for controlling all hardware components of the instrument.

Reporting Summary.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Extended Data

Extended Data Fig 1:

(A) Dose-dependent fluorescence images of the nitrocellulose membrane corresponding to different concentrations of N protein solutions acquired from plasmonic-fluor-based N protein LFA and measured by (B) benchtop scanner and (C) portable scanner.

Extended Data Fig 2:

Schematic illustration of the full strip IL-6 p-LFA employed for the quantitative detection of IL-6 in the serum of COVID-19 positive (PCR confirmed) individuals.

Extended Data Fig 3:

Representative examples of data processing conducted on data acquired by portable scanner from IL-6 p-LFA strips exposed to the (A, B) serum sample of the PCR-positive individuals and (C, D) blank for quantitative analysis of IL-6 concentration.

Data availability

The main data supporting the results in this study are available within the paper and the supplementary Information. All data generated in this study are available from figshare with the identifier: https://figshare.com/s/d188ae9655f19238b0f3.

Code availability

The code for the RMC μ and RMC parameter calculation and the instructions on how to use the code for Langmuir and five-parameter logistic fitting is available at https://github.com/seanwangsalad/PythonRMC. The code for processing the data from the portable scanner and the instructions on how to use the code is available at https://github.com/seanwangsalad/AreaUnderCurveForLFAReader.