Development of receptor binding domain-based double-antigen sandwich lateral flow immunoassay for the detection and evaluation of SARS-CoV-2 neutralizing antibody in clinical sera samples compared with the conventional virus neutralization test

Abstract

Vaccination is an effective strategy to fight COVID-19. However, the effectiveness of the vaccine varies among different populations in varying immune effects. Neutralizing antibody (NAb) level is an important indicator to evaluate the protective effect of immune response after vaccination. Lateral flow immunoassay (LFIA) is a rapid, safe and sensitivity detection method, which has great potential in the detection of SARS-CoV-2 NAb. In this study, a fluorescent beads-based lateral flow immunoassay (FBs-LFIA) and a latex beads-based LFIA (LBs-LFIA) using double antigen sandwich (DAS) strategy were established to detect NAbs in the serum of vaccinated people. The limit of detection (LoD) of the FBs-LFIA was 1.13 ng mL^{− 1} and the LBs-LFIA was 7.11 ng mL^{− 1}. The two LFIAs were no cross-reactive with sera infected by other pathogenic bacteria. Furthermore, the two LFIAs showed a good performance in testing clinical samples. The sensitivity of FBs-LFIA and LBs-LFIA were 97.44% (95%CI: 93.15%–99.18%) and 98.29% (95%CI: 95.84%–99.37%), and the specificity were 98.28% (95%CI: 95.37%–99.45%) and 97.70% (95%CI: 94.82%–99.06%) compared with the conventional virus neutralization test (cVNT), respectively. Notably, the LBs-LFIA was also suitable for whole blood sample, requiring only 3 μL of whole blood, which provided the possibility to detect NAbs at home. To sum up, the two LFIAs based on double antigen sandwich established by us can rapidly, safely, sensitively and accurately detect SARS-CoV-2 NAb in human serum.

Graphical abstract

1. Introduction

SARS-CoV-2 is extremely contagious and cause respiratory failure and multiple organ dysfunction and even death [1,2], which has caused over 610 million infections and 6.5 million deaths as of September 23, 2022. Vaccination remains the crucial prevention method for SARS-CoV-2 infection because it can induce the human immune system to produce neutralizing antibodies (NAbs) against the S protein. NAbs can block the binding of the receptor-binding domain (RBD) to angiotensin-converting enzyme 2 (ACE₂) [[3], [4], [5]], so NAbs act as an effective predictor of immune protection [6,7]. A key issue to be addressed is the determination of the correlation between protection against SARS-CoV-2 infection and NAb levels. Furthermore, similar to natural infection, the NAbs titre decreases over time after vaccination [[8], [9], [10], [11]]. Subsequently, a sensitive assay for detecting NAb levels is urgently needed to determine whether the body has developed sufficient protection against the SARS-CoV-2 to adjust epidemic prevention and control measures and vaccination policies.

Presently, NAbs are primarily detected using virus-based neutralization tests and protein-based surrogate assays. The former mainly include the conventional virus neutralization test (cVNT) and pseudo-virus neutralization test (pVNT). CVNT is the gold standard for detecting NAbs but requires the culture of live SARS-CoV-2, causing an infection risk to operators. Hence, it must be conducted in a biosafety level 3 (BSL-3) laboratory and need 3–5 days to complete it [6]. As for pVNT, the highly infectious SARS-CoV-2 is replaced by an engineered virus expressing the S protein, allowing the test to be performed in a BSL-2 laboratory [[12], [13], [14]], while it is still time-consuming. The second type is protein-based surrogate assay, primarily includes enzyme-linked immunosorbent assay (ELISA) and lateral flow immunoassay (LFIA), both of which generally use the ACE₂–RBD interaction. Tan et al. coated ACE₂ on the well and labeled RBD with horseradish peroxidase (HRP) to generate a detection signal and using the inhibition rate to determine the NAb level, achieved a specificity of 99.93% and a sensitivity of 95%–100% [15]. However, the ELISA is still necessary to perform in the laboratory because requires a variety of instruments, so it cannot meet the needs of point-of-care testing (POCT) [16,17].

The LFIA is by far one of the most successful analytical platforms to detect the target substances owing to its rapidity, simplicity, relative cost-effectiveness, the possibility to be used by nonskilled personnel [[18], [19], [20], [21]]. The vast majority of LFIAs for detecting SARS-CoV-2 NAbs are based on ACE₂-RBD interaction, thus the detection signal correlates with NAbs concentration negatively. Lake et al. established a LFA based on the RBD-ACE₂ with colloidal gold as a signal reporter and achieved 90% sensitivity and 100% specificity in detecting the blood samples compared with pVNT [22]. The colorimetric LFIA (PDA-LFIA) developed by Tong et al. showed higher sensitivity with a LoD of 160 ng mL⁻¹ compared with the commercial ELISA [23]. Despite rapid detection of the NAbs, the sensitivity remains low because when the sample contains a few NAbs, the T-line signal is difficult to distinguish from negative. We also learned that some researchers used a combination of blocking and indirection strategy to detect NAb. For example, Duan et al. developed a DFIA combining blocking and indirection strategy for detecting NAbs, but it only achieved a sensitivity of 88.14% and a specificity of 93.24%, compared with the commercial ELISA kit [21].

In addition, double antigen sandwich (DAS) strategy is also an effective method to improve the sensitivity. Several studies have showed that the major epitopes for potently NAbs reside in the spike RBD [24,25], and qualitatively, positive agreement between NAbs titres and binding assays was highest for RBD assays [26]. Therefore, it's feasible to assess the NAbs titre by measuring the level of RBD binding antibodies. Fulford et al. established a LFIA using blocking and double antigen sandwich strategy, with a sensitivity of 96% and a specificity of 100% (n = 126), which significantly improved the performance [27]. However, biotinylated RBD protein and anti-biotin antibody increases the cost and complicates the production process. Cost and stability are important considerations for the clinical application of LFIA. We learned about some LFIAs for detecting NAb of SARS-CoV-2, and compared these methods in terms of principle, signal output, tracer and performance (Table 6).

Table 6.

Comparison of different LFIAs for detection of NAbs.

Method	Principle	Signal output	Tracer	Sensitivity	Specificity	Reference method	Ref.
LFA	Blocking	-a	AuNPs	90%	100%	pVNT	[22]
PDA-LFIA	Blocking	–	PDA	/	/	ELISA	[23]
LFIA	Blocking	–	AuNPs	100%	87.5%	ELISA	[30]
DFIA	Blocking and Indirection	- and +b	FBs	88.14%	93.24%	ELISA	[21]
QD-LFIA	Blocking and Indirection	- and +	QDs	86.1%	100%	NATc	[31]
Nab-testTM	Blocking and DAS	- and +	AuNPs	96%	100%	cVNT	[27]
Finecare™	DAS	+	FBs	92%	100%	PCR	[32]
FBs-LFIA	DAS	+	FBs	97.44%	98.28%	cVNT	This work
LBs-LFIA	DAS	+	LBs	98.29%	97.70%	cVNT	This work

^a“-” = Reverse signal.

^b“+” = Forward signal.

^c“NAT” = Nucleic acid test.

In order to quickly and simply detect neutralizing antibodies in blood samples with low-cost and sensitive, in this study, we developed two LFIAs based on double antigen sandwich system. In these two devices, one receptor binding domain (RBD) recombinant protein was coated on T-line as capture antigen, and another was labeled with fluorescent beads or latex beads as label antigen (Fig. 1 ). Different from other double antigen sandwich systems, the RBD proteins of SRAS-CoV-2 spike protein used in this work are all commercially produced on a large scale, with the advantages of stable performance and low-cost. We purchased a variety of commercial RBD proteins and obtained the best choice through pairing screening and process optimization. First, we developed a fluorescent beads-based LFIA (FBs-LFIA). Fluorescent beads containing lanthanide europium can significantly improve the sensitivity because it has the advantages of narrow emission spectrum, high quantum yield, large Stokes shift, long fluorescence lifetime and low environmental interference [28,29]. FBs-LFIA showed good analytical performance and the detection results of clinical samples showed that the FBs-LFIA reached high sensitivity and specificity compared with the cVNT (Table 3). In addition, the FBs-LFIA was no cross-reactive with sera infected by other pathogenic bacteria. These results indicate that FBs-LFIA achieves the goal of rapid and accurate detection of neutralizing antibodies in human sera samples. Although the FBs-LFIA has a good advantage of sensitivity and specificity, it still needs an immunochromatographic card scanner (ICS) to record the test result. Moreover，in order to further simplify the testing process, reduce the dependence on instruments, expand the applicable scenarios, and achieve home self-test, a latex beads-based LFIA (LBs-LFIA) was developed. The carboxyl modified latex beads can form stable chemical bonds with the RBD recombinant protein, which improves the stability of the product compared with the colloidal gold. The detection of clinical samples shows that LBs-LFIA has excellent performance compared with cVNT (Table 5). At the same time, LBs-LFIA was no cross-reactive with sera infected by other pathogenic bacteria. Furthermore, LBs-LFIA was suitable for whole blood samples so provided the possibility of detecting NAbs at home by using a micro needle to collect fingertip blood. These results show that LBs-LFIA does not rely on instruments and has the advantages of simple operation and broad application scenarios. In general, we have developed two novel diagnostic tools with good advantages of rapid, sensitive, specific, cheap, stable, and user-friendly for detecting SARS-CoV-2 NAb and have good clinical application prospects in assessing vaccine efficacy and recommending vaccination.

Fig. 1.

The workflow and schematic diagram of the LFIAs for detection of NAbs. a) The serum is diluted by the sample diluent, and then the mixture solution is dropped into the sample well of LFIA. After 15 or 20 min, observe and record results. b) Schematic diagram of the FBs-LFIA and the guide for interpreting the results. c) Schematic diagram of the FBs-LFIA and the guide for interpreting the results.

Table 3.

Predictive agreement values of the FBs-LFIA for the detection of NAbs in sera samples.

	FBs-LFIA
Sensitivity	97.44% (93.15%–99.18%)
Specificity	98.28% (95.37%–99.45%)
PPVb [95% CIa]	97.44% (93.15%–99.18%)
NPVc [95% CI]	98.28% (95.37%–99.45%)

^aCI, confidence interval.

^bPPV, positive predictive value.

^cNPV, negative predictive value.

Table 5.

Predictive agreement values of the LBs-LFIA for the detection of NAbs in sera samples.

	LBs-LFIA
Sensitivity	98.29% (95.84%–99.37%)
Specificity	97.70% (94.82%–99.06%)
PPVb [95% CIa]	97.96% (95.39%–99.17%)
NPVc [95% CI]	98.08% (95.32%–99.29%)

^aCI, confidence interval.

^bPPV, positive predictive value.

^cNPV, negative predictive value.

2. Materials and methods

2.1. Materials

Nitrocellulose (NC) membrane was purchased from Millipore (Shanghai, China) and Sartorius (Germany). Sample pads, conjugate pads, absorbent pads and PVC plates were purchased from Jiening Biotechnology (Shanghai, China). Fluorescent beads (1% (w/v)), latex beads (4% (w/v)),1-(3-Dimethyaminopropy)-3-ethylcarbodiimide hydrochloride (EDC) and N-hydroxy succinimide (sulfo-NHS) were purchased from Thermo Fisher Scientific (USA). 4-Morpholineethanesulfonic acid (MES), bovine serum albumin (BSA) and Tween-20 were purchased from Sigma-Aldrich (USA). The RBD recombinant protein and anti-SARS-CoV-2 IgG purchased from Guangzhou FeiPeng Technology Co., Ltd. And Huizhou BaiXi Technology Co., Ltd. The chicken IgY and goat anti-chicken immunoglobulin Y (IgY) were obtained from Hangzhou Qitai Technology Co., Ltd. (Hangzhou, China). All reagents were of analytical grade.

2.2. Instruments

An ultrasonic cell crusher (Ningbo, China), XW-80 Avortex Mixers (Shanghai, China), and centrifuge (Eppendorf, Germany) were used to prepare signal probes. The XYZ 3000 series dispensing system (Biodot, USA) was used to prepare conjugate pads and reaction membrane, and the blast drying oven (Shanghai Boyuan Industrial Co., Ltd, China) was used for drying all components of LFIA and for stability study. The HGS201 programmable strip cutter (Hangzhou Fenghang Technology Co., Ltd, China) was used to assemble the LFIA. A field-emission transmission electron microscope (TEM, Philips, Holland) and a Zetasizer Nano instrument (Nano Series, UK) were used to characterize the signal probes. An ICS (AFS330 M, Guangzhou Lab Biotech, China) was used to record the fluorescence signal of FBs-LFIA. A smartphone was used to take all photos.

2.3. Clinical samples and ethics statement

The clinical samples used in this study were obtained from vaccinated people and non-vaccinated people of the SARS-CoV-2 vaccine, and the NAbs titre was determined using cVNT performed by Guangdong Provincial Center for Disease Control and Prevention. The collection of all clinical samples was approved by the Scientific Research Ethics Review Committee of the Shunde Hospital of Guangzhou University of Chinese Medicine (Ethical Approval No.KY-2020128) and was performed according to the WHO standard protocol. Informed consent was obtained from the patients or their legal representatives.

2.4. The assembly of the LFIAs

The complete LFIA consists of five parts: (1) Sample pad: used for adding, filtering, and adjusting samples; (2) Conjugate pad: carrier of signal probes; (3) Reaction membrane: site of the immune reaction occurs and where the results are observed with the test line (T-line) and control line (C-line); (4) Absorbent pad: maintains capillary effect to ensure smooth liquid flow through the strip; (5) PVC plate: support material to ensure all components are bonded to form a complete strip. Each component is assembled on the PVC plate with a 2 mm overlap, then cut to a width of 4 mm using a programmable HGS201 strip cutter and placed in a plastic case for the assay. The preparation process of each component of two LFIAs are detailed in Supplementary Material.

2.5. The performance of the LFIAs

The optimization process of the two LFIAs has been detailed in Supplementary Material. To evaluate the linear range of LFIAs, we added the anti-SARS-CoV-2 IgG to negative serum to prepare a series of known concentration samples. Subsequently, 100 μL of solution of each concentration was added to the FBs-LFIA or LBs-LFIA, respectively. The F_T/F_C (the ratio of T-line fluorescence intensity to C-line) of FBs-LFIA was recorded by ICS after 15 min. Similarly, the results of LBs-LFIA were observed by naked eye and recorded using a smartphone after 20 min, and then, the T-line gray value was analyzed using ImageJ. The LoD was calculated as Mean + 3SD, in which SD is the standard deviation of the 20 cases of negative sample signal, and Mean is the average value of the 20 cases of negative sample signal. To verify the specificity of two LFIAs, sera samples infected pathogenic bacteria were tested. Briefly, 10-fold dilutions of each serum were added to FBs-LFIA and 40-fold to LBs-LFIA, respectively, and results were recorded in the same way as LoD.

2.6. Clinical application of the FBs-LFIA

According to the results of cVNT that NAbs can't effectively block the binding of RBD to ACE₂ when the NAbs titre is below 1:4, so these samples was defined as negative (titre <1:4), otherwise as positive (titre ≥1:4). A total of 389 clinical samples were tested to validate the feasibility of FBs-LFIA for detecting SARS-CoV-2 NAb in human serum. All samples were diluted 10 times and then took 100 μL solution into the FBs-LFIA, respectively. After 15 min, the F_T/F_C of each sample was recorded by ICS and then obtained the receiver operating characteristic (ROC) curve using IBM SPSS Statistics. To balance the sensitivity and specificity of FBs-LFIA, the Youden index was calculated for getting the cut-off value. Finally, the test results of FBs-LFIA were compared with cVNT to calculate the negative, positive, and total concordance rate.

2.7. Clinical application of LBs-LFIA

A total of 181 samples were continuously diluted (10 times, 20 times, 40 times, etc.). Determine the titre level of NAb in the samples by the red strip of the T-line. If the T-line has red bands visible to the naked eye, it is positive, otherwise it is negative. Dilute all clinical samples and add 100 μL solution to LBs-LFIA. After 20 min, observe the results with naked eyes and take photos. The samples with cVNT titre ≥1:4 are positive samples, and the samples with cVNT titre <1:4 are negative samples. Through the test results of 181 samples, the appropriate dilution ratio was determined, and more clinical samples were tested (n = 554).

3. Results and discussion

3.1. Detection principle of the LFIAs

Based on the double antigen sandwich strategy, the RBD recombinant protein labeled with fluorescent beads (FBs) and latex beads (LBs) was used as the label antigen, and another RBD recombinant protein was coated on T-line as the capture antigen. In the device of FBs-LFIA (Fig. 1b), RBD-Ag2 and goat anti-chicken IgY labeled with FBs were sprayed on the conjugate pad, meanwhile the RBD-Ag4 and chicken IgY were coated on the T-line and C-line, respectively. Similarly, in the LBs-LFIA (Fig. 1c), the conjugate pad contained LBs labeled RBD-Ag6 and goat anti-chicken IgY, as well as the RBD-Ag8 and chicken IgY were coated on T-line and C-line, respectively. When the sample contains NAbs, it combines with FBs-RBD or LBs-RBD to form an immune complex. The immune complex continues to move to the T-line and is intercepted by the capture antigen, forming a fluorescent signal (FBs-LFIA) or visualized red band (LBs-LFIA) on the T-line. The control probe (FBs or LBs-goat anti-chicken IgY) will be captured by chicken IgY, so there is always a fluorescent signal or visualized red bands on the C-line.

3.2. Optimization of the LFIAs

To improve the sensitivity of LFIAs and reduce the interference of the serum matrix effect, we optimized the type and amount of RBD (Fig. S3), the ratio and spray volume of signal probes as well as reaction condition, etc. (Figs. S4 and S5). The optimal results were shown in Table 1 , and the detailed optimization process was provided in the Supplementary Material. In addition, in Fig. S1, we show the characterization results of FBs-LFIA, and in Fig. S2, we show the characterization results of LBs-LFIA.

Table 1.

The optimal parameters of two LFIAs.

Parameters	FBs-LFIA	LBs-LFIA
Label antigen	RBD-Ag2	RBD-Ag6
Capture antigen	RBD-Ag4	RBD-Ag8
Amount of label antigen	10 μg	30 μg
Amount of capture antigen	1 μg cm−1	1 μg cm−1
Probes ratio (RBD to goat anti-chicken IgY)	9:1	4:1
Probes spray volume	4 μL cm−1	6 μL cm−1
Sample pad treatment solution	0.1 M pH 9.0 B B	0.1 M pH 7.0 Tris
Sample diluent	1% PBST	1% PBST
Sample dilution folds	10-fold	40-fold
Reaction time	15 min	20 min

3.3. Performance of the LFIAs

The linear range of FBs-LFIA was measured by a series of known concentrations of anti-SARS-CoV-2 IgG. As shown in Fig. 2 a, the fluorescence intensity of the T-line increased with increasing anti-SARS-CoV-2 IgG concentration and showed a good linear range with 0 ng mL⁻¹ to 2000 ng mL⁻¹. Meanwhile, using the negative serum, positive serum and sera infected by other pathogenic bacteria (Fig. 2c), the FBs-LFIA showed a highly specific. The accuracy of the FBs-LFIA was evaluated by recovery rate, as shown in Table S1, the recovery rate of FBs-LFIA is between 95% and 107%. In order to determine the intra-batch and inter-batch repeatability of FBs-LFIA, three different concentration samples were detected by the same batch or 5 batch of the strip. As shown in Table S2, the coefficient of variation (CV) of the same concentration is less than 7%, indicating that the FBs-LFIA is relatively stable and have good repeatability. The stability of FBs-LFIA and LBs-LFIA were evaluated by real-time stability (25 °C) and accelerated aging study (50 °C). As shown in Fig. S7a, the detection performance of FBs-LFIA did not change significantly after the strip was stored at room temperature (25 °C) for 91days. After storing for 23 days at high temperature (50 °C), the FT/FC of positive samples began to decrease (Fig. S7b).

Fig. 2.

The performance of the LFIAs. a) Four-parameter fitting curve of FBs-LFIA, the x-axis is the concentration of anti-SARS-CoV-2 antibody. The y-axis is T-line fluorescence value/C-line fluorescence value. Y = 0.01743+(3.652–0.01743)/[1+(229.7/X) ^^1.155], R² = 0.9954, Linear range 0–2000 ng mL⁻¹ (n = 3). b) Four-parameter fitting curve of LBs-LFIA, the x-axis is the concentration of anti-SARS-CoV-2 antibody, the y-axis is T-line gray value. Y = 195.8+(5736–195.8)/[1+(173.8/X) ^^0.9012], R² = 0.9844, Linear range 0–2500 ng mL⁻¹ (n = 3). c) The specificity of FBs-LFIA (n = 3). d) The specificity of LBs-LFIA (n = 3).

To achieve instrument-free detection, a visual LBs-LFIA based on latex beads was developed. As shown in Fig. 2b, the red band could be observed on the T-line and became more obvious with increasing concentrations of anti-SARS-CoV-2 IgG. The T-line gray value analyzed by ImageJ, showing a good linear relationship with 0 ng mL⁻¹ to 2500 ng mL⁻¹. Unsurprisingly, the LBs-LFIA was no cross-reactive with sera infected by other pathogenic bacteria. (Fig. 2d). Three different concentrations of samples were prepared to determine the intra-batch and inter-batch repeatability of LBs-LFIA, as shown in Fig. S6. The results of Image J indicate that the coefficient of variation (CV) of the same concentration is less than 10% (Table S3), so LBs-LFIA is stable in sample testing and has a good repeatability. The stability of LBs-LFIA were evaluated by real-time stability (25 °C) and accelerated aging study (50 °C). The red band on the T-line of each concentration sample did not significantly change after the LBs-LFIA stored at 25 °C (Fig. S7c) and 50 °C (Fig. S7d) for 4 weeks, and the experiment needs to be continued to determine the shelf life of the LBs-LFIA.

3.4. Clinical application of the FBs-LFIA

Overall, 389 clinical samples were determined by cVNT and grouped according to vaccination status and time. The results (Fig. 3 b) showed that the NAbs titre in all participants is below 1:4 within 14 days of receiving the first dose of SARS-CoV-2 vaccine, while after 28 days, more than half of the subjects have a titre greater than 1:4. For the second SARS-CoV-2 vaccine dose, the NAbs titre increased significantly in the 14 days, exceeding 1:8 in most people. However, 28 days after the second dose, the NAbs titre gradually decreased. Moreover, 6 months after the second dose most samples turned negative indicating a need to further strengthen immunization.

Fig. 3.

Clinical application of the FBs-LFIA. a) The grouping scatter diagram of NAbs titre detected by FBs-LFIA. b) NAbs titres were grouped according to different inoculation status: Unvaccinated, 14 days after the first dose, 28 days after the first dose, 14 days after the second dose, 28 days after the second dose, 180 days after the second dose. c) The ROC curve of consistency analysis of FBs-LFIA and cVNT.

According to the results of cVNT that NAbs can't effectively block the binding of RBD to ACE₂ when the NAbs titre is below 1:4, so 233 out of 389 samples were negative (<1:4) and 156 were positive (≥1:4). Each sample was tested using the FBs-LFIA and recorded the F_T/F_C. The area under the curve (AUC) was calculated as 0.997 (95% CI: 0.995–1.000; p < 0.0001) using IBM SPSS Statistics with receiver operating characteristic (ROC) curve analysis (Fig. 3c). Furthermore, to balance the sensitivity and specificity of the FBs-LFIA, the Youden index was calculated, and the maximum (7.4 ng mL⁻¹) was chosen as the cut-off value (Table S4) to normalize all the test results based on this cut-off value (negative if F_T/F_C < 7.4 ng mL⁻¹, otherwise positive.

The test results of FBs-LFIA were shown in Fig. 4 a: 4 negative sample exceed 7.4 ng mL⁻¹, and 4 positive sample below 7.4 ng mL⁻¹. Overall, compared with the cVNT, among 233 negative samples 4 cases yielded a false positive, and among the 156 positive samples 4 cases yielded a false-negative (Table 2 ). The FBs-LFIA achieved 97.44% (95%CI: 93.15%–99.18%) sensitivity and 98.28% (95%CI: 95.37%–99.45%) specificity (Table 3 ). Additionally, the Cohen's kappa coefficient was 0.957 (Table S5), indicating a significant correlation between the FBs-LFIA and the cVNT. These results demonstrate that the FBs-LFIA is a rapid, simple, and accurate diagnostic tool for detection SARS-CoV-2 NAb.

Fig. 4.

Sera samples with different titres were continuously diluted and detected by LBs-LFIA.

Table 2.

Concordance analysis between FBs-LFIA and cVNT.

Samples		cVNT
		Negative	Positive	Total
FBs-LFIA	Negative	229	4	233
	Positive	4	152	156
	Total	233	156	389
	Concordance	98.28%	97.44%	97.94%

3.5. Clinical application of LBs-LFIA

181 clinical samples were continuously diluted (10 times, 20 times, 40 times, etc.) and tested by LBs-LFIA. If there is red band visible to the naked eye on the T-line, the result is positive, otherwise the result is negative. The test results of sera samples in the non-vaccinated group were all negative, but some of the sera samples with titre <1:4 in the vaccinated group had false positive results when diluted 10 or 20 times (as shown in Fig. 4). The possible reason is that the antibodies against the RBD may not be equal to NAbs against SARS-CoV-2, not only neutralizing antibodies, but there are also non-neutralizing antibodies against RBD in the samples that combine with both detection and capture RBD to cause false positive results. We regard the cVNT titre 1:4 as the cutoff value and dilute the samples 40 times before testing. At this time, the results can accurately determine whether the sample is negative or positive. The samples with titre <1:4 are negative and the samples with titre ≥1:4 are positive.

Based on the results, we tested more clinical samples (n = 554). Each sample was diluted by 40 times and then applied to the LBs-LFIA for observing whether a red band formed on the T-line after 20 min. As shown in Table 4, Table 6 cases negative samples yielded a false positive, and 5 cases positive samples yielded a false-negative. The LBs-LFIA demonstrated a good sensitivity (98.29%) (95%CI: 95.84%–99.37%) and specificity (97.70%) (95%CI: 94.82%–99.06%) compared with cVNT (Table 5). The Cohen's kappa coefficient was 0.960 (Table S6), indicating a significant correlation between the LBs-LFIA and cVNT. Moreover, in terms of sensitivity and specificity, LBs-LFIA is comparable to FBs-LFIA, while it has an advantage of instrument-free detection than FBs-LFIA; hence, it's more suitable for rapid detection of NAbs in POCT.

Table 4.

Concordance analysis between LBs-LFIA and cVNT.

Samples		cVNT
		Negative	Positive	Total
LBs-LFIA	Negative	255	5	260
	Positive	6	288	294
	Total	261	293	554
	Concordance	97.70%	98.29%	98.01%

Since centrifuges and other instruments are required for the separation of serum or plasma, this greatly limits application of the FBs-LFIA in POCT. Being able to directly test whole blood samples will be beneficial to increase the applicable scenarios of LFIA. In this study, whole blood was collected from vaccinated people by blood collection vessels without or with different anticoagulants and centrifuged to obtain serum and plasma, and the NAbs titre was 1:32 determined by cVNT. Each sample was applied to LBs-LFIA after diluted 40-fold, and the results indicated that serum or plasma detection was more sensitive than whole blood detection (Fig. 5 ). The possible reason is that whole blood contains a large number of blood cells, resulting in lower NAbs levels than plasma and serum in the same volume. Meanwhile, the anticoagulant had no obvious effect on the test results. These results further illustrate that LBs-LFIA has a good prospect for POCT applications, and it's suitable for home self-test and community hospital.

Fig. 5.

The LBs-LFIA is suitable for multiple types of samples. a) Photos of different type of samples detected by LBs-LFIA. b) Bar graph of detection results (n = 3). The x-axis is the sample number, the y-axis is T-line gray value. The No.1 is plasma collected by blood collection tubes with sodium citrate; The No.2 is plasma collected by blood collection tubes with heparin sodium; The No.3 is plasma collected by blood collection tubes with EDTA-2K; The No.4 is serum; The No.5 is whole blood collected by blood collection tubes with sodium citrate; The No.6 is whole blood collected by blood collection tubes with heparin sodium; The No.7 is whole blood collected by blood collection tubes with EDTA-2K.

3.6. Comparison of different LFIAs for detection of NAbs

Recently, a series of LFIAs for detection NAbs have been reported, and we compared the two LFIAs established in this work with previously reported LFIAs in several aspects (Table 6). In contrast, most of the LFIAs are based on blocking strategy that use the RBD-ACE₂, so the signal output is reverse. In addition, our LFIAs are validated by the gold standard (cVNT), while most LFIAs previous were compared with commercial ELISA kit (cPass SARS-CoV-2 Neutralization Antibody Detection Kit, GenScript USA Inc.). We also used Genscript ELISA kit (sVNT) to detect 197 sera samples, and it showed a 100% sensitivity but only 69.07% specificity, compared with cVNT (Table S7). Therefore, our LFIAs has good accuracy and reliability in clinical application. Importantly, LBs-LFIA is a potential tool for the quantitative detection of NAbs based on the correlation between sample dilution fold and NAbs titre. The drawback of the LFIAs is that it cannot distinguish between IgG and IgM of NAbs, so the next major research goal is to explore how to realize the detection of NAbs types.

4. Conclusions

Quantitative detection of SARS-CoV-2 NAb titres is crucial to screen therapeutic antibodies from convalescent patients, predict humoral protection, evaluate vaccine efficacy, and optimize immunization strategies. The traditional method used to detect NAb titres is cVNT, but it requires culture viruses and cells, is time-consuming, and poses an infection risk to the operator. In order to rapidly and accurately detect the NAbs level in the immunized population on a large scale, we developed two LFIAs based on double antigen sandwich that enables the detection of SARS-CoV-2 NAb in clinical sera samples.

First, we used fluorescent beads containing lanthanide europium as tracer to establish a sensitive LFIA (FBs-LFIA). Unlike the LFIA based on ACE₂-RBD, the FBs-LFIA requires only the RBD recombinant protein that is safe and readily available and provided a forward readout result in 15 min. After optimized the parameters, FBs-LFIA showed good analytical performance, the LoD was 1.13 ng mL⁻¹, and the linear range was 0–2000 ng mL⁻¹ using four-parameter fitting (R² = 0.9954). The detection results of clinical samples showed that the FBs-LFIA reached 97.44% (95% CI: 93.15%–99.18%) sensitivity and 98.28% (95% CI: 95.37%–99.45%) specificity compared with the cVNT (Table 3), Cohen's kappa coefficient (0.957) indicates that there is a significant correlation between the two NAb detection methods (Table S5). In addition, the FBs-LFIA was no cross-reactive with sera infected by other pathogenic bacteria. These results show that FBs-LFIA is a fast, safe, sensitive, accurate and robust NAbs detection tool.

Secondly, a visual LFIA (LBs-LFIA) based on carboxyl modified latex beads was developed to further simplify the testing process, reduce the dependence on instruments, expand the applicable scenarios for testing, and make testing applicable to home self-test. The LoD of LBs-LFIA is 7.11 ng mL⁻¹, and linear range from 0 ng mL⁻¹ to 2500 ng mL⁻¹ (R² = 0.9844). While it has a wide clinical application because it's suitable for whole blood samples and provides the possibility of detecting NAbs at home by collecting fingertip blood with a miniature needle. Moreover, in the clinical test, the LBs-LFIA achieves a sensitivity of 98.29% (95%CI: 95.84%–99.37%) and a specificity of 97.70% (95%CI: 94.82%–99.06%) compared with cVNT (Table 5), and the Cohen's kappa coefficient (0.96) indicated a significant correlation between both NAb detection methods (Table S6). In short, the LBs-LFIA has good POCT application prospects in evaluating vaccine efficacy and guiding vaccination.

Of course, LFIAs based on double antigen sandwich strategy have certain limitations that both LFIAs produce false negatives and false positives. For this, the possible reason is that the antibodies against the RBD may not be equal to NAbs against SARS-CoV-2. Liu et al. reported that 61 out of 121 monoclonal antibodies which bind to the S trimer can neutralize SARS-CoV-2 [19], and Ju et al. reported that 7 out of 16 RBD-binding antibodies possessed neutralizing activity [20]. McCallum et al. reported that 15 out of 41 monoclonal antibodies against NTD showed neutralizing function [18]. Therefore, not all monoclonal antibodies that bind to the S protein or RBD protein possess neutralizing ability and not all NAbs bind to the RBD protein. Another reason for the false positive result could be that LFIAs is more sensitive than cVNT because all the false positive samples were obtained from the vaccinated group that may contain low NAb levels even if the titre is < 1:4. These low NAb levels cannot block the RBD–ACE₂ interaction, but they can be detected by LFIAs.

In conclusion, two rapid, simple, cheap, stable and user-friendly LFIAs we developed were highly consistent with the gold standard, cVNT, and no risk of infection because without any live virus and pseudo-virus. The FBs-LFIA and LBs-LFIA can quickly evaluate the NAbs level of vaccinated people, which have good application prospects in evaluating vaccine efficacy and guiding vaccination. In the next step, we intend to evaluate the difference between the NAb level in vaccinated people and that in clinical patients, and thus explore the relationship between the neutralizing antibody level and the protective effect.

Credit authorship statement

Ying Zhang: Conceptualization, Methodology, Writing –original draft. Yixiao Chen: Validation, Formal analysis, Writing –original draft. Yong He: Validation, Data Curation, Writing –original draft. Yizhe Li: Validation, Resources. Xiaoli Zhang: Conceptualization, Methodology, Resources. Jiajie Liang: Conceptualization, Methodology. Jinyong He: Resources, Data Curation. Shaofang Lu: Resources, Data Curation. Zhixing Gao: Validation, Resources. Jianhua Xu: Resources, Project administration, Funding acquisition. Yong Tang: Project administration, Supervision, Funding acquisition, all authors have read and agreed to the published version of the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.