Net-Shaped DNA Nanostructure-Based Lateral Flow Assays for Rapid and Sensitive SARS-CoV-2 Detection

Abstract

Lateral flow assay (LFA)-based rapid antigen tests are experiencing extensive global uptake as an expeditious and highly effective modality for the screening of viral infections during the COVID-19 pandemic. While these devices have played a significant role in alleviating the burden on the public healthcare system, their specificity and sensitivity fall short compared with molecular tests. In this study, we endeavor to address both limitations through the utilization of DNA nanotechnology in LFA format, wherein we substitute the target-specific antibody with designer DNA nanostructure-based molecular probes for recognizing the SARS-CoV-2 virus via multivalent, pattern-matching interactions. We meticulously designed a Net-shaped DNA nanostructure and strategically arranged trimeric clusters of aptamers that specifically recognize the spike proteins of SARS-CoV-2. This approach has proven instrumental in bolstering virus-binding affinity on the LFAs. Our findings indicate high LFA sensitivity, enabling the detection of viral loads ranging from 10³ to 10⁸ viral copies/mL. This notable sensitivity is maintained across various SARS-CoV-2 viral strains, obviating the need for intricate sample preparation protocols. The significance of this heightened sensitivity lies in the crucial role played by the designer DNA nanostructure, which facilitates the detection of extremely low levels of viral loads. This not only enhances the overall reliability of self-testing but also reduces the likelihood of false-negative results, especially in cases of low viral load within patient samples.

Graphical Abstract

The COVID-19 pandemic has brought into sharp focus on the urgent and critical need for accurate and accessible diagnostic platforms that enable individuals to test themselves and obtain rapid but still sensitive results in point-of-care (POC) and home settings.^1,2 As the world grapples with the multifaceted challenges posed by the pandemic, the demand for reliable and convenient diagnostic methods has exponentially surged.^3,4 Among the array of available diagnostic methods, lateral flow assays (LFAs) have emerged as the preferred and widely accepted gold standard platform for self-testing owing to their remarkable attributes such as portability, cost-effectiveness, user-friendliness, and expeditious results. Consequently, LFAs have become the most extensively utilized diagnostic platform for COVID-19 self-testing due to their versatility in accommodating different bodily fluids and suitability for both home and POC settings.⁴ The majority of commercially available LFAs utilize antibody or antigen-based testing methodologies. While antibody-based LFAs do not directly detect the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus itself, they instead assess the host immune response to SARS-CoV-2 infections, aiding in the identification of individuals with past exposure to the virus.⁵ By detecting immunoglobulin antibodies produced by B cells in bodily fluids, these LFAs provide valuable insights into the body’s immune response during viral infection.^6,7 However, it is noteworthy that these tests exhibit low specificity during the initial week of viral exposure, with specificity improving during the subsequent second and third weeks, owing to the time required for immunoglobulin antibodies to develop following disease onset.^7-10 To surmount this limitation, antigen-based LFAs have been developed, targeting the nucleocapsid protein in the case of SARS-CoV-2 detection.^11,12 Antigen-based LFAs not only offer convenience and scalability but also harbor the ability for viral detection in self-testing, thereby playing a crucial role in monitoring and curbing the dissemination of highly contagious mutant viral strains.¹³ Nevertheless, they do possess certain drawbacks when compared to nucleic acid tests, such as real-time reverse transcription-PCR (RT-PCR). These limitations encompass reduced test specificity using N protein targeting antibodies and the inability to detect low viral loads from noninvasive samples.^14-16 The lack of specificity can lead to both false-positive and false-negative test results. For instance, Gremmels et al. demonstrated that the “PanbioTM COVID-19 Ag rapid test” could detect high viral loads in nasopharyngeal samples of SARS-CoV-2-infected individuals and also produced several false-negative rapid test results.¹⁶ This becomes particularly problematic during the presymptomatic phase when viral loads in patient samples are exceedingly low, often resulting in false-negative antigen test outcomes.^10,17 In addition to false-negatives, rapid antigen tests have also shown instances of false-positives for SARS-CoV-2.^18-20 Yamaniha et al. reported an intriguing case where a rapid antigen test demonstrated recurrent positive results for SARS-CoV-2 in an individual with acute human immunodeficiency virus (HIV) infection.¹⁴ Interestingly, the authors observed consistent negative outcomes in RT-PCR tests performed on nasopharyngeal swabs. These results may be attributed to the presence of shared epitope motifs and a certain degree of homology between the surface glycoproteins of HIV and SARS-CoV-2.²¹ False-positive results are concerning as they can lead to inappropriate patient care and infection control measures. Conversely, false-negative test results for asymptomatic close contacts can engender a false sense of reassurance, potentially leading to the relaxation of isolation measures and a subsequent heightened transmission rate. Thus, the test specificity of antigen-based LFAs using antibodies becomes a significant concern as they allow the virus to go undetected during community-wide screening, thereby perpetuating the burden on the public healthcare system.¹⁷ This underscores the imperative need to develop innovative solutions capable of enhancing the performance of antigen-based LFAs, augmenting specificity, and enabling the detection of low viral loads from noninvasive samples such as saliva.^22,23 By addressing these challenges, we can ensure effective screening and early detection of viral infections, ultimately reducing the burden on healthcare systems and facilitating timely interventions to control the spread of the virus.^24,25

One promising solution lies in harnessing the great potential of DNA nanotechnology and integrating aptamer-based DNA nanostructures as a replacement for antibodies on the test line of LFAs.^26-28 Designer DNA nanostructures, products of the rapidly evolving field of DNA nanotechnology, possess unparalleled programmability and self-assembly properties.^29-31 These ingeniously engineered designer nanostructures, composed of synthetic DNA molecules, can be precisely designed to form desired shapes and patterns with highly controlled dimensions, shapes, and surface functionality.³² By leveraging the unique characteristics of DNA nanostructures, our previous research has demonstrated the multivalent, pattern-matching strategy built on custom-designed star-shaped DNA nanostructures.^33,34 This innovative approach mimics the complex spatial distribution of virus epitopes with nanoscale precision and efficiently binds with dengue viral particles with high avidity, present in noninvasive specimens including blood samples. However, the flexible and mobile nature of envelope glycoproteins in membraned viruses, including coronaviruses like SARS-CoV-2, necessitates a different approach to enable effective pattern matching and detection using DNA nanotechnology.³⁵

In this article, our innovative strategy centers around the implementation of a “multilayer” Designer DNA Nanostructure (DDN) design principle. This principle empowers us to selectively identify and capture intact SARS-CoV-2 virions with remarkable precision and strong affinity, resulting in a multifold increase in the sensitivity of the LFAs. We achieve this by employing a Net-shaped DDN, henceforth referred to as the “DNA Net” for simplicity. This innovative approach combines pattern-matching and multivalent interactions between aptamers specific to the receptor binding domain (RBD) of SARS-CoV-2 spike proteins present on the DNA Net and the trimeric spikes displayed on the viral surface.^35,36 By strategically arranging these DNA aptamers into trimeric clusters on the DNA Net, we bolster the virus binding affinity through dynamic clustering of spike proteins and precise matching of the interspacing of the trimeric clusters. To enable colorimetric or naked-eye detection, we employ DNA Nets in two distinctive ways. First, we incorporate them onto the test line of LFAs, serving as capture agents for the target virus. Second, the DNA Nets are embellished with ultrabright gold nanoshells, endowing them with remarkable signal amplification capabilities, thereby serving as exquisite reporter assays.³⁷ Upon introduction of a sample containing SARS-CoV-2 spike proteins or viruses, a sandwich complex is formed among the DNA Nets, the target analyte, and the reporter assays, resulting in a significantly amplified colorimetric signal. This enhanced signal can be interpreted either qualitatively or quantitatively, providing valuable insights into the presence and concentration of the target analyte.^38,39

The findings elucidated herein hold great potential in transforming self-testing-based viral diagnostics, ensuring timely and accurate diagnoses of infectious diseases. By capitalizing on these advancements, the presented technology possesses the capacity to wield a significant impact on healthcare systems, enabling early detection, efficacious disease management, and targeted interventions, consequently augmenting the overall health and well-being of individuals and communities across the globe.

EXPERIMENTAL SECTION

Printing the DNA Net Conjugates and Spike Protein on the Paper Membrane.

For the preparation of the test line conjugates, we mixed 3 × 3 DNA Net and nitrocellulose-streptavidin (NC-SA) with a 1:10 ratio in 1× TAE-Mg-K buffer, resulting in a total volume of 200 μL. Subsequently, the test line conjugates were incubated at room temperature for 45 min on a rotator with intermittent shaking to ensure optimal binding. As for the control line, we utilized 200 μL of trimeric spike proteins (TSPs) at a concentration of 0.15 mg mL⁻¹. To ensure an even distribution of the test and control solutions onto the nitrocellulose membrane (NCMs), we loaded both reagents into the glass syringes connected to an ALFRD dispenser. The LFA device was securely fixed to the dispenser table for stability. We employed a controlled flow rate of 0.7 μL cm⁻¹ during the dispensing process. Following the printing, the NCMs were dried at 37 °C for 30 min under a relative humidity of 17%. The amount of input material allowed for the printing of approximately 6 full-length LFA devices, each measuring 254 mm in length and 25 mm in width. Finally, the printed LFA chips were carefully stored in a laminated aluminum bag supplemented with desiccants to maintain optimal conditions. Prior to each experiment, the sealed aluminum bag was opened, and the LFA chips/devices were precisely cut to dimensions of 60 mm in length and 4 mm in width, ensuring consistency and uniformity.

RESULTS AND DISCUSSION

Optimization of DNA Net on Gold Nanoshells.

Rapid antigen-based LFAs are commonly employed for POC or self-testing purposes, aiming to detect viral antigens or antibodies for confirming infection. However, the limited sensitivity and selectivity of these assays often lead to the generation of false-negative or false-positive results, both of which have significant implications. False-negative outcomes create a false sense of security, increasing the risk of unknowingly transmitting the virus, while false-positive cases require appropriate patient care. To tackle these challenges and enhance the specificity of LFAs for SARS-CoV-2 virus detection, which can be readily expanded to other viral detections, we propose a distinctive approach that harnesses the power of designer DNA nanostructure (DDN)-based viral capture probes. Despite the emergence of DNA nanotechnology nearly four decades ago, accompanied by the pioneering development of the first home pregnancy test kit (“Clearblue One Step”) based on LFA technology during a similar time frame, the convergence of these two scientific disciplines has been surprisingly scarce.^40,41 The untapped potential of incorporating DNA nanostructures into LFAs represents a vast reservoir of scientific innovation and progress in this realm. In this study, we present an application of DNA nanotechnology that significantly enhances the sensitivity of LFAs, thereby revolutionizing the field of early viral detection.

To achieve this goal, we applied DDN design principles to create intricate net-shaped DNA nanostructures (called “DNA Net”), composed of rhombus units with a ~15 nm long edge. The DNA Net was designed to arrange RBD-targeting aptamers into an intricate array of triaptamer clusters with a ~6 nm intra- and a ~15 nm inter-triaptamer spacing, matching those of the spike timers on viral particles. The resulting DNA Net architecture offers excellent flexibility and broader coverage, enabling the high-affinity binding of multiple trimeric spike proteins on viral surfaces for sensitive and selective detection of the SARS-CoV-2 virus via multivalent, pattern-matching interactions.³⁵ DNA Nets with various sizes, including '2 × 2' (30 nm × 30 nm), '3 × 3' (45 nm × 45 nm), and '4 × 4' (60 nm × 60 nm) Nets, were constructed and characterized herein to determine the best Net size we should use on the LFA devices (Scheme 1A).³⁵ To confirm the successful formation of DNA Nets, we utilized atomic force microscopy (AFM) imaging and agarose gel electrophoresis, as illustrated in Figures 1A, S1, and S2. The yields of forming the monomeric 2 × 2 Net, 3 × 3 Net, and 4 × 4 Net are estimated as 96.61, 88.01, and 77.51%, respectively, as shown in Table S2. It is worth noting that due to the delicate nature of the thin tile-shaped DNA Net structures, they can be susceptible to deformation or damage during the process of AFM characterization.⁴² However, despite this inherent fragility, all the desired DNA Net structures were effectively formed with a high degree of yield and purity as shown by our gel electrophoresis assay, enabling their direct utilization in our assays without the need for further purification.

Scheme 1. Illustration of DNA Net-Based LFA Device Design and Operation^a.

^a(A) Rhombus-shaped DNA Nets with different sizes, including “2 × 2 DNA Net (30 nm × 30 nm)”, “3 × 3 DNA Net (45 nm × 45 nm)”, and “4 × 4 DNA Net (60 nm × 60 nm)” were created. ★ Shows DNA aptamer cluster positions across the DNA Net. Triaptamer clusters are positioned with a ~6 nm intra- and a ~15 nm inter-triaptamer spacing, mirroring the arrangement of spike trimers on the viral particle. (B) A 3 × 3 DNA Net (10 nM) was implemented on the test line using biotin–streptavidin chemistry, while 0.15 mg/mL of SARS-CoV-2 trimeric spike protein (SARS-TSP) was placed on the control line. Streptavidin-coated gold nanoshells (SA-AuNS) were coated with a 3 × 3 DNA Net and used as a reporter assay. (C) Upon introducing virus particles into the LFA device, they bind in a sandwich format, utilizing the reporter assay and DNA Net present on the test line. the reporter assay also binds with the TSP on the control line through aptamer–TSP interaction. The aggregation of AuNS results in an enhancement in colorimetric signals in both the test and control lines.

Figure 1.

Reporter assay optimization. (A) AFM was employed to capture high-resolution images of DNA Net structures with different dimensions, including 2 × 2, 3 × 3, and 4 × 4 DNA Nets. The obtained AFM images reveal the intricate arrangement and organization of the DNA Net structures at a nanoscale level and scale bars indicating a length of 50 nm. (B) Different sizes of DNA Nets were coated on gold nanoshells (AuNS), resulting in an increase in the size of the AuNS-conjugated DNA Net reporter assay as the size of DNA Net increased. (C) Zeta potential decreases as the size of DNA Nets increases, indicating improved assay stability. Data are presented as the mean ± sd, n = 5 biologically independent samples.

Our approach aimed to effectively incorporate DNA Nets into LFAs, serving two primary objectives: first, employing them as reporter assay and, second, integrating them onto the test line. In pursuit of the former objective, we evaluated DNA Nets of various sizes on AuNS to assess the overall stability of the DNA Net-AuNS conjugates. In this work, we explored the use of commercially available 150 nm gold nanoshells as an alternative to 40 nm gold particles commonly used in LFA. Gold nanoshells are composed of a 20 nm thick gold shell deposited around 110 nm silica spheres that exhibit unique optical properties that yield highly contrasting signals. The presence of the silica core reduces particle density, enabling easy resuspension in water and smooth flow through the LFA without precipitation.⁴³ To synthesize the reporter assay, we attached DNA Nets with 150 nm streptavidin-coated gold nanoshells (SA-AuNS) via a biotin strand. We performed zeta potential and hydrodynamic diameter measurements to evaluate the characteristics of the resulting conjugates. The size of the AuNS increased from approximately 159.94 ± 2.96 to 178.04 ± 3.16 nm upon the introduction of the '2 × 2' DNA Net, followed by further growth to 264.84 ± 5.55 and 270.66 ± 6.18 nm with the '3 × 3' and '4 × 4' DNA Nets, respectively (Figure 1B). Of note, we noticed that our DLS measurements for AuNS coated with 2 × 2 and 4 × 4 Nets did not simply agree with the theoretical sizes of the corresponding Nets. We speculate that geometric orientation (2 × 2 Net) and steric hindrance (4 × 4 Net) may influence these size measurements.

These findings substantiated the effective coverage of the AuNS surface area by the DNA Nets, as anticipated based on their theoretical dimensions. Such coverage is pivotal for stabilizing individual AuNS, which initially exhibits a very low zeta potential (−18.16 ± 2.67 mV). Consequently, the overall zeta potential of the AuNS gradually decreased with the increasing sizes of the DNA Nets, indicating the stability of the AuNS-DNA Net complex (Figure 1C). More specifically, the zeta potential of the AuNS decreased from −18.16 ± 2.67 mV (without DNA Net) to −27.56 ± 1.67 mV in the presence of the '2 × 2' DNA Net, further dropping to −35.6 ± 2.95 mV with the '3 × 3' DNA Net and ultimately attaining a value of −40.0 ± 1.36 mV when the '4 × 4' DNA Net was employed. Notably, no significant alterations in zeta potential were observed between the '3 × 3' and '4 × 4' DNA Nets. Consequently, for the subsequent experiments in this study, we opted to utilize the more cost-effective '3 × 3' DNA Net to develop both the reporter assay (DNA Net – AuNS) and integrating DNA Nets onto the test line of the LFA.

LFA Device Testing Using SARS-CoV-2 Trimeric Spike Protein as a Target.

In previous investigations, we have demonstrated the binding affinity of DNA Nets to TSPs immobilized on an surface plasmon resonance (SPR) chip within the femtomolar to nanomolar range.³⁵ Encouraged by these findings, we embarked on a comprehensive exploration to ascertain whether DNA Nets, when immobilized on the test line of a NCM, would exhibit comparable binding behavior toward free TSP in solution. We strategically affixed test line conjugates, consisting of a 3 × 3 DNA Net, as well as control line conjugates of TSP onto the NCM, utilizing working concentrations of 10 nM and 0.15 mg mL⁻¹, respectively (as expounded upon in the Experimental Section of the Supporting Information as well as in Figure 2A).

Figure 2.

LFA device testing with free trimeric spike protein (TSP) as a target. (A) Schematic representation of the operation of the LFA device and the detection of free TSP using DNA Net. (B) A range of TSP concentrations from 1.7 to 52 nM were tested, leading to a gradual increase in the intensity of the test line on the LFA device. (C) Absorbance values from the test line were calculated and plotted against the corresponding TSP concentrations. The binding isotherm of TSP was analyzed using Hill fit, yielding a dissociation constant (K_D) of 7.25 ± 0.59 nM. Additionally, a linear detection range for TSP was observed, spanning from 1.7 to 13 nM (inset, Figure 2C). The data are presented as the mean ± SD (standard deviation), with n = 3 biologically independent samples.

Subsequently, we assessed a spectrum of TSP concentrations ranging from 1.7 to 52 nM as input samples in the test strips. Remarkably, we observed distinct gray lines with high contrast forming on both the test line and control line as we added TSP at increasing concentrations, ultimately reaching the saturation state (Figure 2B).^43,44 Representative images depicting the colorimetric changes in the test line (T) and control line (C) signals in the presence of TSP are presented in Figure 2C. Our findings revealed that the LFA exhibited a wide linear range of detection spanning from 1.7 to 13 nM with a dissociation constant (K_D) of 7.25 ± 0.59 nM, indicating its ability to accurately detect free TSP in solution within the tested concentration range, thereby establishing its potential for virus particle detection. Notably, our method does not necessitate the use of blocking reagents to passivate or pretreat the NCM and conjugate pad, as reported in previous reports.^23,27,45-47 This approach not only enhances the overall assay sensitivity but also reduces the preparation time required for the assay.

LFA Performance in Saliva, Serum, and Urine Environment.

An ideal LFA device must possess the capacity to navigate the intricate biological environments encountered during patient sample testing. To assess the compatibility of our LFA device with such environments, we introduced 4 nM TSP targets spiked with various levels (ranging from 1 to 75%) of serum, saliva, and urine samples. Remarkably, when utilizing saliva and urine samples at concentrations ranging from 1 to 25%, we observed successful binding between the TSP target and the 3 × 3 DNA Net anchored on the test line, as evidenced by the discernible colorimetric signals (Figure 3A, B).

Figure 3.

LFA device performance in different biological environments. (A) LFA device was tested in a saliva environment with varying concentrations ranging from 1 to 75%. A negative symbol (−) indicates the absence of the TSP target, while a positive symbol (+) represents the presence of a 4 nM TSP target. (B) Similarly, the LFA device was tested in a urine environment with varying concentrations ranging from 1 to 75% using the same TSP target indication symbols. (C) Furthermore, the LFA device was also tested in a serum environment with varying concentrations ranging from 1 to 75% using the same TSP target indication symbols. (D) Absorbance values from the test line on the LFA device are plotted against different percentages of the target analyte (saliva/serum/urine), providing a visual representation of the device’s performance in different biological environments. The data are presented as the mean ± SD (standard deviation), with n = 3 biologically independent samples.

However, upon increasing the concentrations of saliva and urine samples beyond 25%, we noticed a discernible gray color on the test line in the absence of TSP targets, indicating the presence of false-positive signals. An analogous pattern of false-positive signals was observed when the concentration of human serum samples exceeded 10% (Figure 3C). We hypothesize that the presence of excess proteins in concentrated serum/saliva/urine may interfere with the test line, contributing to the enhanced color bands observed in the absence of TSP. Furthermore, when the concentrations were further elevated to 75% in both saliva and urine samples, the signals on the control line were absent in the presence of the target protein. This phenomenon can be ascribed to the heightened sample viscosity, impeding the flow of the reporter assay. Based on these discernments, we established the optimal analyte quantity (serum/saliva/urine) that can be employed on the LFA device, effectively minimizing the occurrence of false-positive signals (Figure 3D). These findings provide invaluable insights into the design of assays that proficiently navigate complex biological environments encountered during the testing of patient samples.

Comparing the Target Binding Isotherm in the LFA Device.

Figure 3A demonstrates the effective binding of the DNA Net, located on the test line of the LFA, with the target TSP. To compare the binding isotherms in biological (25% saliva) and aqueous (0% saliva) environments, we introduced TSP into the LFA device under both environments (Figure 4A, B). We found a K_D value of 5.72 ± 0.28 nM in the biological environment, which closely matched the value obtained in the aqueous environment (K_D of 7.25 ± 0.59 nM, Figure 2C). The minimal disparity in K_D values may be ascribed to the presence of diluted salivary proteins within the samples, exerting an influence on the binding of TSP to the DNA Net. These findings unequivocally substantiate the excellent binding capability of the LFA device to its intended target, in this case, the spike proteins of SARS-CoV-2, even within biologically relevant conditions.

Figure 4.

LFA device performance and selectivity. (A) Performance of the LFA device was evaluated in a 25% saliva environment, with a range of TSP concentrations tested from 4.3 to 52 nM. (B) The absorbance values from the test line were calculated and plotted. The binding isotherm of TSP was analyzed using Hill fit, resulting in a K_D value of 5.72 ± 0.28 nM. (C) The selectivity of the LFA device was assessed by using different trimeric spike proteins, including Influenza-HA and HIV GP-120, each at a concentration of 1 μM, along with 25 nM TSP and 1% serum. (D) Corresponding absorbance values from the test line were plotted against the different targets, demonstrating the selectivity of the LFA device. The data are presented as the mean ± SD (standard deviation), with n = 3 biologically independent samples.

Target Selectivity of the LFA Device.

Specificity stands as a pivotal determinant in effectively mitigating false-positive occurrences resulting from cross-reactivity. The LFA device we developed is meticulously tailored for the precise detection of the TSP target. It is imperative to consider the potential for cross-reactivity with other trimeric-shaped proteins. Previous studies have reported instances of cross-reactivity using viral antigen-targeting antibody-based LFA with influenza hemagglutinin (HA) trimers and HIV-1 GP120 trimers.^14,48-50 Cross-reactivity can occur when the target molecule shares structural similarities or exhibits some degree of homology with other molecules present in the sample. To comprehensively evaluate the cross-reactivity and selectivity of our LFA built on DNA Net probe, we conducted tests using influenza HA trimer (H1N1) and HIV-1 GP-120 trimer proteins at a much higher concentration of 1 mg mL⁻¹. The results were satisfactory as the LFA device exclusively manifested a positive colorimetric signal in the presence of TSP while displaying minimal test line signals in response to other proteins and serum samples (Figure 4C, D). These findings provided conclusive evidence ruling out any potential for protein cross-reactivity within our LFA, thereby establishing its great specificity. This is a crucial feature considering the specificity concerns associated with commercially available lateral flow assay kits.^51-54

For sensitivity assessments, we evaluated the LFA device performance in a 25 nM TSP with a 25% saliva environment that comprised 102 samples. The device exhibits a sensitivity of 96.08%, with 98 true positives (TP) and 4 false negatives (FN). Regarding specificity, we examined the LFA device under the same environmental conditions, excluding TSP from the solution. The device demonstrates a specificity of 91.18%, with 93 true negatives (TN) and 9 false positives (FP), as outlined in Supporting Information, Table S3.

Device Performance with SARS-CoV-2 Viruses with Different Strains.

The principal aim of this study was to develop a rapid, reliable, and sensitive LFA device capable of detecting minute viral loads of SARS-CoV-2 viruses while adeptly managing an intricate biological assay. We conducted a comprehensive assessment of SARS-CoV-2 viruses, encompassing four strains of substantial concern, namely, WA1/2020/Washington, B.1.1.7/2020/UK, B.1.351/2020/South Africa, and Omicron variant XBB.1.5/2022/USA, all rendered noninfectious through UV inactivation but with the binding function of viral surface spikes maintained.⁵⁵ The selection of SARS-CoV-2 variants in our study was driven by two principal considerations: (1) the alpha and beta variants had previously been classified as highly contagious variants and (2) the XBB.1.5 Omicron variant had maintained its status as a circulating variant of interest according to the World Health Organization as of August 17, 2023.⁵⁶ Furthermore, quick accessibility of these variants was facilitated through the RADx program. The DNA aptamer employed in our investigation was initially developed against the RBD of the WA1/2020/Washington SARS-CoV-2 strain.³⁶ Subsequently, we conducted SPR assays to assess the binding affinity between this aptamer and the spike proteins of different SARS-CoV-2 variants. Our findings show the K_D values of 1.55 × 10⁻⁹ M for XBB.1.5, 8.81 × 10⁻⁸ M for B.1.351, and 2.15 × 10⁻⁷ M for B.1.1.7 SARS-CoV-2 variants (Figure S5). These results demonstrate the versatility of the aptamer as a ligand binding different spike variants, which can be integrated into a 3 × 3 DNA Net for the effective detection of various variants of interest. As these viral strains were prediluted in cell media, we performed 10-fold serial dilutions of SARS-CoV-2 viruses by adding 10 μL of the virus stock into 90 μL of a reaction buffer containing 1× phosphate-buffered saline (PBS) and 25% saliva (Table S4). A normal buffer, consisting of 1× PBS with 25% saliva, served as the negative control. Subsequently, we subjected the diluted viral strains, in triplicate, to a 20 μL reporter assay solution in 96-well plates, exposing each well to our LFA device (Scheme 1B, C).

We observed a gradual increase in the intensity of the test line signals as the viral load in the samples increased (Figure 5A-D). Our LFA device exhibited a broad detection range, spanning from 10² to 10⁷ viral copies per 100 μL reaction volume (or 10³ – 10⁸ viral copies/mL), thus satisfying the acceptable analytical sensitivity delineated within the World Health Organization’s target product profile.⁵⁷ To determine the LOD, we utilized a previously reported statistical method employing a four-parameter logistic curve fit.⁵⁸ By using this method, we calculated theoretical LOD values to be 42,830 viral copies; 26,720 viral copies; 5,590 viral copies; and 2,280 viral copies/mL for the B.1.1.7, B.1.351, XBB1.5, and WA1 viral strains, respectively (Figure 5F). To further benchmark our LFA device, we assessed a commercially available gold-nanoparticle-based rapid antigen test for SARS-CoV-2 (iHealth COVID-19 Antigen Rapid Test), which has a reported LOD of 350 TCID50/swab. We selected this product because it had received authorization from the FDA under an Emergency Use Authorization and was widely adopted in the market and available for purchase on various e-commerce platforms. To facilitate a comparison between this kit and our LFA, we employed the SARS-CoV-2 virus derived from the WA1/2020/Washington strain in an antigen rapid test. We generated a serial dilution of the viral stock by combining 10 μL of the virus stock with 90 μL of the manufacturer-supplied extraction solution. Relative to our developed LFA device, the commercial antigen rapid test demonstrated a suboptimal linear detection range, ranging from 10⁵ to 10⁷ total viral copies per 100 μL reaction volume (or 10⁶ – 10⁸ viral copies/mL). Notably, the test line signals were markedly weak for the viral stock containing 10³ – 10⁵ viral copies/mL (Figure 5E, F). These results highlight the critical role of DNA nanostructures in improving the sensitivity of LFAs to detect extremely low levels of viral loads. Importantly, we achieved this level of LFA sensitivity without using extensive sample preparation steps such as sample lysis, purification, and nucleic acid amplification steps.^59,60 We wish to emphasize that in the specificity test involving H1N1 (Figure 4C), the test line signals exhibited intensities comparable to those observed with a low concentration of SARS-CoV-2 (Figure 5). This observation raises the concern of potential false-positive cases attributable to other viral infections. To address this issue, several experimental strategies can be employed to minimize the occurrence of false-positive results. For instance, pretreating nitrocellulose membranes (NCM) with various blocking reagents, such as polymers, surfactants, and proteins (e.g., gelatin, casein, and bovine serum albumin), has shown promise in preventing the occurrence of false positive results.²³ However, it is worth noting that NCM pretreatment may impact assay sensitivity. In commercial settings, addressing these challenges can involve pretreating the conjugate pad with these blocking reagents before the application and drying of reporter assays.

Figure 5.

LFA device’s performance was evaluated using different strains of SARS-CoV-2 viruses, including (A) WA1/2020/Washington, (B) B.1.1.7/2020/UK, (C) B.1.351/2020/South Africa, and (D) Omicron variant XBB.1.5/2022/U. The viruses were serially diluted (10⁷–10²/ 100 μL reaction volume) with 1× PBS and 25% saliva. A buffer composed of 1× PBS and 25% saliva was used as the negative control (0 virus). (E) To facilitate comparison, a commercially available COVID-19 antigen rapid test kit (iHealth) was utilized with the SARS-CoV-2 WA1/2020/Washington strain. (F) Absorbance values from each test line (A–E) were plotted against the total number of viral copies in the solution, allowing for an assessment of the LFA device performance. The data are presented as the mean ± SD (standard deviation), with n = 3 biologically independent samples.

Given that false positive and negative results are the major limitation of existing LFAs, the integration of machine learning, probabilistic algorithms, and automated image collection providing a measure of uncertainty can help reduce user confusion stemming from false positives or negative results, fostering wider LFA adoption. The existing infra-structure developed during the COVID pandemic could be leveraged to automate image collection of digital LFT data within these pipelines, facilitating the continuous refinement of image classification models. For instance, due to visual similarities in qualitative LFT, existing algorithms can be updated with a smaller data set once a substantial data set is accumulated for a specific test.

Navigating Scale-Up and Cost Challenges in LFAs.

Translating LFA-based diagnostics from the laboratory to the market and overcoming scale-up and cost challenges and materials availability entail a range of important considerations and obstacles. Our LFA device consists of three key components: DNA Net-based virus probes, AuNS, and paper strips. DNA oligos for the self-assembly of DNA Net can be rapidly synthesized via solid-phase oligo synthesis at a large scale using phosphoramidite chemistry. The cost of the DNA Net-based virus probe for building one LFA device is ~$0.20. The estimated cost for the components of the test strip, including the conjugation pad, absorption pad, nitrocellulose membrane, backing card, and cassette, amounts to approximately $1.13 per strip. For AuNSs, the overall cost can be minimized by optimizing different factors. For instance, in industrial production settings, cost efficiencies are achieved through substantial batch sizes, elevated nanoparticle concentrations, and high liquid volume capacity of centrifuges, all contributing to reduced labor costs and consequently overall AuNSs costs. Conversely, in research settings and at smaller laboratory scales, nanoparticle synthesis via chemical reduction proves more economically viable due to lower expenses.⁶¹

DNA oligonucleotides and AuNSs should ideally be obtainable at a reasonable cost, both during periods of routine circumstances and in the context of health crises, such as the COVID-19 pandemic. However, it is imperative to recognize that the cost and availability dynamics of the materials integral to LFA devices, particularly paper substrates, can be significantly affected during health emergencies such as the COVID-19 pandemic. Throughout the course of the pandemic, numerous research groups, ourselves included, grappled with formidable logistical challenges related to the procurement of essential paper-based components required for LFA devices. The unprecedented demand for LFAs led to substantial increases in the acquisition costs associated with these paper substrates, consequently contributing to an overall increase in the cost of LFA devices. This starkly underscored the intricate interplay between the cost and availability of paper materials and the broader considerations of affordability and accessibility within the realm of LFAs. To effectively curtail the mounting cost of LFAs, it becomes imperative to address the prevailing dependency on paper substrates, particularly during periods of heightened demand in pandemics. Exploring alternative substrate materials, streamlining production processes, and optimizing resource allocation represent promising avenues for mitigating these challenges. Furthermore, channeling efforts into research and development initiatives aimed at the creation of more cost-effective and sustainable materials for LFA devices holds the potential for yielding enduring benefits. Such endeavors have the capacity not only to ameliorate immediate cost concerns but also to ensure the continued accessibility and utility of LFAs, even in the face of protracted health crises.

CONCLUSIONS

The proposed technology addresses the limitations inherent in rapid antigen-based COVID-19 testing by introducing a DNA nanostructure-based assay that significantly improves sensitivity and allows for testing noninvasive sampling, such as saliva samples. Our LFA device exhibits a sensitivity of 96.08% and a specificity of 91.18% when subjected to testing with the SARS-CoV-2 TSP. Notably, the device demonstrates a broad detection range spanning from 10³ to 10⁸ viral copies/mL and the capability to detect as few as 2,280 viral copies/mL of the SARS-CoV-2 WA1 strain. In contrast, the commercial kit, when subjected to the same strains and reaction volume, failed to produce any detectable signals on the test line below 10⁵ viral copies/mL in our comparative analysis. These findings not only encourage the overall reliability of self-testing but also reduce the occurrence of false-negative results in the case of low viral load in patient samples. This innovative use of DNA nanostructures unlocks the potential for unparalleled improvements in the precision and accuracy of diagnostic testing, revolutionizing the management and control of infectious diseases, including the formidable challenge posed by COVID-19. Nevertheless, significant challenges persist in the areas of standardization and scalability, hindering the seamless integration of this technology into real-world settings. Through a comprehensive exploration of fundamental principles, innovative design strategies, and cutting-edge advancements, we aim to ignite a spark of inspiration that fuels further research and development endeavors in this field. By doing so, we aspire to accelerate the translation of DNA nanostructures or DNA origami-based LFAs into tangible and practical diagnostic solutions, effectively addressing the pressing and demanding healthcare needs of our time.