A Nanobody-Based Lateral Flow Assay for Point-of-Care Diagnostics

Abstract

Lateral flow assays (LFAs) are among the most successful technologies for point-of-care and at-home testing, but further advances are needed to reduce costs and accelerate development. Alpaca-derived nanobodies (Nbs), single-domain antibody fragments, are promising immunoassay reagents across diverse applications. Their small size and ease of recombinant production make them particularly well suited for diagnostics. Here, we present a paper-based LFA targeting the SARS-CoV-2 nucleocapsid (N) protein that exclusively uses Nbs for direct antigen detection. We also demonstrate in-house synthesis of Nb-coated gold nanoparticles, enabling instrument-free visual readout and detection of N protein down to 40 ng/mL. This design avoids components that require mammalian cell culture and can be produced entirely from in-house reagents, simplifying manufacturing and lowering component costs. Because the assay is read visually without an external reader, it is well suited for deployment in resource-limited settings. Together, these results highlight the speed and practicality of developing Nb-based LFAs and suggest a broadly applicable strategy for detecting other clinically important disease biomarkers.

1. Introduction

The development of low-cost diagnostics is crucial for case finding and disease surveillance in resource limited areas. The accessibility of diagnostic tools can be greatly improved by utilizing materials that are able to be produced with straightforward methods and minimal equipment. The ability to produce all of the necessary reagents on site, particularly those with limited shelf lives, would allow communities to more quickly address local concerns rather than relying on external supply chains. The World Health Organization (WHO) and other healthcare institutions have identified local manufacturing of diagnostics tests as a key factor for reducing health inequality in low- and middle-income countries (LMICs) [1,2].

Multiple technologies exist for rapid diagnostics, though material limitations often force reliance on symptom-based diagnosis instead of microbiological confirmation, which is critical for effectively targeted treatment and antibiotic stewardship [3]. The REASSURED criteria, which evolved from the WHO’s ASSURED criteria, provide guidance on ideal product characteristics for point-of-care testing (POCT) in order to achieve the best outcomes [4,5,6]. Rapid antigen tests, also known as lateral flow assays (LFA), are widely recognized as ideal for POCT due to their low equipment and training requirements, allowing them to be used outside of a medical setting. However, LFAs typically require monoclonal antibodies (mAbs) as their affinity reagent, which require complex manufacturing processes involving mammalian cell culture expression systems.

Nanobodies (Nbs), also known as single-domain antibodies, have emerged as an exciting alternative to traditional mAbs in situations where size and/or complexity are limiting factors [7,8]. Derived from camelids and cartilaginous fish (e.g., sharks), Nbs are the binding domain fragments of heavy-chain-only antibodies, making them a fraction of the size while fully retaining their binding properties (Figure 1A) [9]. They consist of a single protein chain with no need for glycosylation. This allows Nbs to be produced using highly cost-effective bacterial expression systems. The lack of an Fc region or light-chain also contribute to their greater average thermal and chemical stability compared to conventional mAbs [10]. Further, their small size means that the same amount of protein contains a proportionally greater density of binding surfaces for antigen capture and detection [8,11]. Together, these properties make Nbs ideal candidates for the development of LFAs, and there has been increasing interest in incorporating them into novel LFA designs [12,13,14,15,16,17,18].

In this study, we demonstrate the development of an LFA using Nbs that we previously generated against the SARS-CoV-2 nucleocapsid protein [19]. We first identify an optimal capture and detection Nb pair using a sandwich ELISA screen, then confirm its performance across multiple sample matrices. Next, we synthesize gold nanoparticles (AuNPs) in house, conjugate them to the detection Nb, and validate binding by biolayer interferometry (BLI). We then incorporate these reagents into a paper-based, visual-read LFA and evaluate assay performance. Importantly, our intention was not to create an additional COVID-19 diagnostic test, but to use this well-established antigen as a practical test case to demonstrate how straightforward it can be to translate Nbs into point-of-care LFA reagents, highlighting that the core components can be rapidly produced, validated, and deployed without reliance on mammalian cell culture.

2. Materials and Methods

2.1. Protein Purification

Nbs and the N protein were expressed and purified as previously described [19,20]. Briefly, Nb genes were expressed from a pHEN vector with a N-terminal pelB signal sequence, and C-terminal LPETG sortase tag followed by a 6× His tag. Nbs were produced in WK6 E. coli and grown in Terrific broth (TB) until reaching an OD₆₀₀ of 0.7, before being induced with 1 mM IPTG overnight at 30 °C. Nbs were purified from the periplasm by osmotic shock by resuspending in HES buffer (200 mM HEPES, 0.65 mM EDTA, 500 mM Sucrose, pH 8) for 2 h at 4 °C, then adding an equal volume of ultrapure water for another 2 h with nutation. The solution was twice clarified by centrifugation at 8000× g for 15 min, and the released protein was bound to Ni-NTA beads for 1 h, washed with three column volumes of wash buffer (50 mM HEPES, 150 mM NaCl, pH 8.0), and eluted with 500 mM imidazole in the wash buffer. Purified Nbs were then dialyzed into PBS and stored at −80 °C.

N was expressed from a pET28a vector (BEI resources, Manassas, VA, USA, NR-53507) and produced in BL21(DE3) E. coli. Cultures were grown to an OD600 of 0.7 and induced with 0.5 mM IPTG for 4 h at 37 °C. Cultures were pelleted at 5000× g for 10 min and frozen at −80 °C. Pellets were then thawed in the buffer W0 (50mM Na₂HPO₄, 300mM NaCl, pH 7.8) supplemented with a protease inhibitor (Sigma Aldrich, Burlington, MA, USA, S8820) and 25 µg/mL DNase I. Cells were lysed by two passes through a French press and clarified by centrifugation at 20,000× g for 10 min, then treated with 0.3% polyethyleneimine (PEI) to remove excess nucleic acids, and centrifuged again. The supernatant was then treated with NH₄SO₄ (0.352 g per mL lysate) for 1 h and centrifuged again. The protein pellet was then dissolved in the buffer W0 with a protease inhibitor and purified by Ni-NTA chromatography as above, washed with the W0 buffer, eluted in W0 with 500 mM imidazole, and immediately dialyzed into 20 mM Tris, 500 mM NaCl, 5 mM beta-mercaptoethanol, and 10% glycerol, with pH 8.0.

2.2. Nb Biotinylation

Nbs were biotinylated using the sortase enzyme to perform transpeptidation of their included LPETG sortag [21,22]. Nb stocks were diluted to make reaction mixtures containing 1 mg/mL Nb, 1 mM GGG-biotin, and 1 mg/mL sortase A enzyme. The reaction was incubated at 37 °C for 2 h. Ni-NTA beads were then added and incubated for an additional 1 h at room temperature to remove the His-tagged sortase enzyme and any unreacted Nbs. The supernatant was then desalted using a 7 kDa Zeba spin column (Thermo Fisher Scientific, Waltham, MA, USA, 89877) to remove unreacted GGG-biotin.

2.3. ELISA

For direct ELISA, 96-well high-binding plates (Thermo Fisher Scientific) were coated with 2 µg/mL purified recombinant N protein overnight at 4 °C. Plates were blocked in the wash buffer (1% BSA, 1% PVP, 0.1% Tween-20 in PBS) for 30 min at RT. Half-log dilutions starting from 100 µg/mL of Nb were incubated for 1 h at RT. Plates were washed with PBST five times between each antibody addition. A biotinylated anti-VHH antibody (Jackson ImmunoResearch, West Grove, PA, USA) and streptavidin–HRP (Jackson ImmunoResearch) secondary antibodies were used at a dilution of 1:10,000 in the blocking buffer for 1 h each at RT. Plates were developed with an OPD substrate (Sigma Aldrich), stopped with 1M HCl, and read at 492 nm using a CLARIOstar Plus plate reader (BMG Labtech, Ortenberg, Germany).

Sandwich ELISAs were performed similarly with the following adaptations. Plates were coated with 2 µg/mL of coating Nbs overnight at 4 °C. After blocking, five-fold dilutions of N protein were incubated for 1 h at RT. Plates were washed five times with PBST, then 2 µg/mL of biotinylated detection Nb was added for 1 h at RT. Plates were washed another five times with PBST, then streptavidin–HRP was added at a dilution of 1:10,000 in bocking buffer. Plates were developed similarly to direct ELISAs.

2.4. AuNP Synthesis

AuNPs were prepared by a citrate reduction adapted from the Trukevich–Frens method [23,24]. An aqueous HAuCl₄ solution (0.5 mM, 150 mL) was heated to reflux, followed by the rapid addition of prewarmed trisodium citrate (40 mM, 7.5 mL). The mixture was refluxed for 1 h, and cooled to RT with continued stirring overnight. AuNPs were stored at RT until they were ready for conjugation.

2.5. AuNP Conjugation

AuNPs were diluted to an OD of 1 (523 nm) with water and adjusted to pH 8.5 with 0.1 M borate buffer. They were then centrifuged at 12,000 rcf for 6 min, resuspended in the stabilization buffer containing 1 mg/mL BSA in 0.1× PBS for 1 h at RT with gentle agitation, centrifuged again and resuspended in 0.1× PBS with 0.05% Tween-20 (PBST). For conjugation, the stabilized AuNPs were centrifuged and resuspended in 10 µg/mL purified recombinant Nb N42 in 0.1× PBS for 2 h at RT with gentle agitation. The AuNP conjugates were then washed twice by centrifugation and resuspension in PBST, and finally stored in PBST plus 1 mg/mL BSA at 4 °C until use.

2.6. Electron Microscopy (EM)

AuNP samples were diluted 1:1 to 1:10 in water, bath sonicated for 5 min at room temperature on a Branson 1210 bath sonicator, and applied for 5 min onto nickel formvar grids (Ted Pella, Redding, CA, USA, 01800N-F). Grids were dried and samples were imaged on a FEI (Thermo Fisher Scientific) Tecnai F12 transmission electron microscope (TEM) at direct magnifications of 9300–4900×. Images were collected as 4640 × 3400 pixel 8 bit tagged image-format file (tiff) images on an AMT charge-coupled device (CCD) camera equipped with AMT image acquisition software (version 602.591).

2.7. UV–Vis Spectroscopy

Spectroscopy was performed using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, ND-ONE-W). A total of 2 µL of plain AuNPs or AuNPs coated with Nb N42 at approximately 1 OD (peak absorbance) were measured in UV–vis mode and the full spectrum was exported to a USB drive. The wavelength with the highest absorbance (below 400 nm) was noted for each AuNP, and the spectrum curves were normalized to 1 OD by dividing by the peak absorbance.

2.8. Biolayer Interferometry (BLI)

BLI experiments were performed using previously described methods [21]. Briefly, streptavidin (SA) biosensors were soaked in deionized water for 30 min prior to use. Analyte solutions were prepared in running buffer (RB: 10 mM HEPES, 150 mM NaCl, 3mM EDTA, 0.005% Tween-20, 0.1% BSA, pH 7.5). Pre-soaked biosensors were blocked with RB; loaded with biotin-Nb44; washed in 10 mM glycine, with pH 1.7, baseline in RB; loaded with 4 µg/mL N, baseline in RB; then reacted with AuNP-N42 or AuNP-BSA. Data were analyzed using Data Analysis HT (v10.0, Sartorius, Göttingen, Germany).

2.9. Lateral Flow Assay

Paper-based LFAs were developed using the principles and protocols outlined by Parolo et al. [25]. The nitrocellulose (NC) membrane (Millipore, Burlington, MA, USA, HF135 membrane) was striped with N44 at 1.5 mg/mL and N protein at 1.4 mg/mL, both in PBS. Stripes were made using an IsoFlow flatbed reagent dispenser (Imagene, Lebanon, NH, USA) at 0.07 μL/mm. The membrane was then dried on a hot plate at 37 °C until stripes were no longer visible. Glass fiber conjugate pads (Millipore, GFCP000800) were used as conjugate and sample pads. N42-conjugated AuNPs supplemented with sucrose to a final concentration of 1% were then added to the conjugate pads and dried, similarly to the NC membrane. We cut and assembled individual strips on the adhesive back of white label tape, each with a glass fiber sample pad, a AuNP-impregnated conjugate pad, a striped NC membrane, and a blotter paper absorbent pad. The sample buffer consisted of PBS with 0.1% Triton-X100 and 7 mM EDTA.

2.10. Data Analysis

ELISA data were analyzed and plotted using a previously published python script [26]. UV–vis data were analyzed and plotted using Microsoft Excel 365. LFA images were analyzed using the ImageJ Gel Analysis Toolbox (version 1.54p), and the color intensity was quantified by integrating the area under each peak.

3. Results

3.1. Nb Selection and Testing

We recently generated a panel of high-affinity Nbs against the SARS-CoV-2 nucleocapsid (N) protein that recognize multiple distinct, non-overlapping epitopes [19]. N is a 44 kDa antigen with several structurally and functionally distinct regions (Figure 1B), including the N-terminal domain (NTD), an RNA-binding domain (RBD), a central linker region, a dimerization domain, and a C-terminal domain (CTD). To characterize our reagents under defined conditions, we expressed and purified recombinant N protein for use in binding and pairing assays (Figure 1C).

A sandwich-format LFA requires two Nbs that can bind the antigen simultaneously, meaning that they must recognize different epitopes and not compete with one another. To maximize the likelihood of identifying a compatible capture and detection pair, we prioritized Nbs with distinct domain specificities (Figure 1D). This set included CTD-targeting Nbs (N7 and N42), linker-targeting Nbs (N23 and N44), and one Nb with undetermined domain specificity (N32).

We expressed His-tagged Nbs and recombinant SARS-CoV-2 N in an E. coli system and purified each protein by Ni-NTA affinity chromatography. We then quantified Nb binding to recombinant N by direct ELISA and observed EC50 values spanning 0.09 to 0.78 µg/mL (Figure 1E), consistent with robust binding across multiple regions of N. With well-performing Nbs in hand and clear coverage across distinct N domains, we next focused on identifying non-competitive Nb pairs suitable for sandwich assay development.

3.2. Sandwich Assay Development

We developed a sandwich ELISA to identify Nb pairs that can capture antigens from complex mixtures and support sensitive detection (Figure 2A). In this format, an unmodified Nb is immobilized on the plate to capture the SARS-CoV-2 N protein, and a second, biotinylated Nb is used for detection. Signal is generated by adding streptavidin–HRP, which binds the biotinylated Nb and produces a colorimetric readout. A key advantage of this workflow is that our Nbs include a built-in sortase tag, enabling rapid, site-directed biotinylation and reducing variability that can arise from random chemical labeling.

To determine the optimal capture and detection orientation, we performed a matrix of sandwich ELISAs testing all pairwise combinations of our N-binding Nbs. This screen identified N44 as the best capture reagent and N42 as the best biotinylated detection reagent (Figure 2B), consistent with the Nbs binding distinct, non-competing epitopes on the antigen. We further verified the specificity of these Nb pairs by performing sandwich ELISAs with an unrelated antigen and found no reactivity (Appendix A Figure A1).

We next quantified the sensitivity of the N44 (capture) and N42 (detection) pair using a dilution series of recombinant N in PBST, yielding an EC50 of 24 ng/mL (Figure 2C). Because assay performance can vary substantially across sample types, we then evaluated this same pair in several matrices relevant to LFA development. Recombinant N was diluted into PBS, LFA running buffer, a BSA-based blocking buffer, and normal human serum (Figure 2D). Sensitivity was highest in LFA running buffer (EC50 40 ng/mL), followed by PBS (83 ng/mL), blocking buffer (130 ng/mL), and serum (830 ng/mL). Together, these data show that the N44/N42 pair performs robustly across multiple conditions while also illustrating the expected loss of sensitivity as matrix complexity increases, with serum having the strongest inhibitory effect.

3.3. Nanobody–AuNP Synthesis

For our full paper-strip LFA, we used Nb-coated AuNPs as the visible detection label. This is primarily because AuNPs are straightforward to produce, highly stable, and well suited for instrument-free visual readout (Figure 3A). Using the citrate reduction method, we synthesized uniform, well-dispersed AuNPs with an average diameter of 17.5 nm (SD 2.1 nm). Particle size and morphology were confirmed by electron microscopy (EM), 100 particles were measured to obtain the above size estimate. AuNP formation was independently verified by UV–vis spectroscopy by measuring the characteristic surface plasmon resonance peak (Figure 3B,C). The resulting colloid displayed a strong wine-red color and remained stable at room temperature for more than one year prior to conjugation.

To generate the detection reagent, we conjugated the detection nanobody (N42) to AuNPs by passive adsorption, achieved by incubating purified recombinant N42 with the AuNP suspension. Following adsorption, we observed a consistent 4 nm red shift in the plasmon peak, consistent with Nb coating of the AuNP surface. Importantly, absorbance above 600 nm increased only minimally, indicating negligible aggregation and confirming the stability of the Nb-AuNP conjugates.

3.4. Sandwich Biolayer Interferometry (BLI) Validation

BLI enables real-time, label-free measurement of biomolecular interactions, and we used it as an orthogonal validation step before moving to strip-based testing. Our goals were to confirm that the selected Nb pair supports true sandwich formation on the SARS-CoV-2 N antigen and to verify that N42 remains functional after conjugation to AuNPs. To closely approximate the final LFA architecture, we designed a sandwich BLI workflow that mirrors the capture and detection sequence used on-strip (Figure 3D).

Streptavidin biosensors were first coated with biotinylated N44 (10 µg/mL), creating an immobilized capture layer with controlled orientation. After capture loading, sensors were briefly regenerated with pH 1.7 glycine to remove any loosely associated material and reduce background, then re-equilibrated in assay buffer. Sensors were then exposed to recombinant N protein (4 µg/mL) to form the captured antigen layer. Finally, we introduced either N42-coated AuNPs (OD 0.1) or BSA-coated AuNPs as a negative control (OD 0.1). We only observed rapid and strong association with the N42-AuNP condition, while the BSA-AuNP control produced minimal signal (Figure 3D). These data support two important conclusions: first, N44 and N42 bind distinct, non-competing epitopes on N and therefore function as a compatible sandwich pair; second, passive adsorption to AuNPs does not measurably compromise N42 binding activity, indicating that the Nb remains accessible and properly oriented for antigen recognition.

3.5. Full-Strip LFA Development and Testing

With the capture Nb, detection Nb, and AuNP conjugates validated, we next assembled a full paper-strip LFA using previously described approaches [25]. In brief, N44 was deposited as the test line to capture the N antigen, and an anti-species or tag-based reagent was deposited as the control line to confirm proper flow and reagent release (Figure 3A). N42-coated AuNPs were incorporated into the conjugate pad as the colorimetric detection reagent, allowing a visible line to develop upon accumulation at the test line in the presence of antigen.

Although LFA performance often requires extensive optimization across membrane types, line-dispensing concentrations, conjugate release chemistry, and running buffers, we found that only minimal tuning was necessary to obtain clear results with this Nb pair. For analytical testing, recombinant N was serially diluted into a sample buffer consisting of PBS supplemented with 0.1% Triton X-100 and 7 mM EDTA. This buffer was chosen to approximate the conditions commonly used in commercial SARS-CoV-2 antigen tests, where detergents promote antigen release and EDTA can help mitigate matrix effects. Each dilution was applied at 250 µL per strip, and strips were allowed to develop until bands were clearly visible.

Using this visual-read format, we observed a signal detectable by eye at 40 ng/mL of recombinant N protein (Figure 3E), and quantification of test lines imaged with a smartphone showed measurable signal at 4 ng/mL (Appendix A Figure A2). Importantly, this apparent sensitivity falls within the range reported for commercially available SARS-CoV-2 N antigen LFAs that use conventional monoclonal antibodies when benchmarked using recombinant N protein as the reference material (Appendix A 

Table A1. Comparison of our Nb-based LFA with commercial SARS-CoV-2 N antigen LFAs. Values indicate the lowest total input concentration of recombinant N protein (ng/mL) that produced a visually detectable test line. All sensitivities, except those measured in this study, were previously reported for commercially available tests using commercially available recombinant N protein diluted in kit-provided sample buffer [27].

Kit Manufacturer	ng/mL
Easy Diagnosis	9.8
GenSure	11
YHLO	16.7
EcoTest	16.7
Orient Gene	20
Panbio	34.4
This study	40
Deepblue	78.5

; [27]), supporting the conclusion that Nbs can achieve performance comparable to widely used mAb-based formats. Together, these results demonstrate that a Nb capture and AuNP-based detection system can be rapidly translated into a functional, instrument-free LFA with minimal optimization.

4. Discussion

In this study, we demonstrate the successful development of a Nb-based LFA strip using Nbs against SARS-CoV-2 N protein. N is a relatively conserved protein among coronaviruses, compared with spike, as well as being highly abundant during infection [28]. These features make it an excellent target for diagnostic assays, and indeed, many such N-targeted LFAs are commercially available and have shown successful deployment both in medical settings and for at home testing [29,30,31]. Accordingly, our intention was not to create yet another COVID-19 diagnostic, but to use this well-established target as a practical test case to demonstrate how readily Nbs can be translated into point-of-care LFA reagents and to highlight their broader potential as a platform for rapid diagnostics.

We were able to produce a functional LFA with minimal optimization, requiring only standard, platform-level adjustments rather than extensive assay re-engineering. We observed excellent translation from the sandwich ELISA into sandwich BLI, and finally into the LFA format.

In terms of analytical sensitivity, the final LFA limit of detection appears to fall within the range reported for commercially available antigen tests based on traditional mAbs (Appendix A Table A1), using the recombinant SARS-CoV-2 N protein as the reference analyte [27]. We intentionally restricted our benchmarking to studies that used recombinant N protein, because reported sensitivities can differ by orders of magnitude when alternative reference materials are used, such as inactivated virus, making cross-study comparisons difficult to interpret without a common calibrant. It is also possible that further optimization of our LFA would improve sensitivity; however, this is likely to be highly dependent on the specific antigen and Nb pair rather than providing general information about optimizing Nb-based LFAs.

As with traditional mAbs, the most important consideration is selection of matching capture and detection pairs [14,32], though the typically less sensitive competition LFA format can be used when only a single binding reagent is available [33]. We leveraged our recently published set of N-specific Nbs because of their established epitopes, covering distinct parts of the N protein; however, in practice, it is not necessary to know the precise epitopes in order to identify matching pairs using a sandwich ELISA or BLI experiment. Others have also successfully demonstrated Nb-based LFAs against Zika [15], T. congolense [16], and interferon alpha [17]. Our goal with this study was to produce all necessary reagents for successful test performance, including the AuNPs and control line protein. Some components, such as the nitrocellulose membrane and glass fiber pad, were not possible for us to produce in house; however, these do not require cold-storage and are not tied to any particular antigen or assay design. Further, while we found that the N protein itself made a stable and reliable test line, extremely high antigen concentrations may inhibit control line formation, leading to uninterpretable results. It can be replaced with an anti-Nb antibody when available.

A key advantage of Nbs for test manufacture is the ability to produce them in bacterial expression systems. High-yield fermentation can be performed with much less equipment and infrastructure than mammalian cell culture of mAbs. Our purification method using Ni-NTA chromatography can also be achieved with minimal additional equipment, and the final LFA does not require any chemical or enzymatic modification of the Nbs, just mixing with the raw AuNPs. Production of the AuNPs themselves is straightforward, and while EM is helpful for confirming nanoparticle size distribution, UV–vis spectrometry provides an easier measure of nanoparticle quality, and other methods such as aggregation testing allows for visual measurement of AuNP conjugate stability.

Biotinylated Nbs were useful for some aspects of assay development, and we therefore employed sortase-mediated, site-specific biotinylation. Although biotinylation is not required for LFA manufacturing or analytical performance, this approach provides a straightforward and reproducible means to generate uniformly labeled reagents, which can streamline Nb-based diagnostic development. More broadly, it highlights a practical advantage of Nbs—they can be engineered and modified in a controlled, modular manner that is often less variable than conventional mAbs, which frequently rely on non-site-specific chemical biotinylation that can yield heterogeneous products, and, in some cases, compromise antigen-binding activity by modifying residues near or within the paratope [34].

The primary bottleneck of LFA development has traditionally been mAb development, but high-throughput Nb discovery processes combined with advanced AI methods will greatly expand our ability to produce these reagents. Many advanced test designs are being developed using a wide variety of different nanoparticle chemistries and detection modes [14], and we hope to see Nbs incorporated into novel biosensor designs that may be difficult or impossible using traditional mAbs. However, it is important to use the simplest method which achieves a suitable analytical sensitivity for a particular antigen, and we believe that visually read Nb-AuNP LFAs represent a straightforward solution that can be efficiently produced locally in low-resource settings, maximizing accessibility and minimizing reliance on cold-chain supply networks.