Development of a highly sensitive magneto-enzyme lateral flow immunoassay for dengue NS1 detection

Clinical samples and ethics statement

A panel of 120 sera was used to evaluate the clinical performance of the magneto-enzyme LFIA. The sera were obtained within 9 days post-onset of illness at Vietnam Military Medical University (Hanoi, Vietnam). All samples were tested for dengue RNA by RT-qPCR as previously described (Gurukumar et al., 2009) and also for dengue-specific IgM antibodies by Dengue IgG/IgM 3.0 Combo rapid test (CTK Biotech, Inc., Poway, CA, USA). Typing of positive samples was performed by nested RT-PCR (Lanciotti et al., 1992). Clinical samples positive for Hepatitis B virus (HBV) (viral load = 2.2 × 10⁵ IU ml⁻¹), Hepatitis C virus (HCV) (viral load = 4.3 × 10⁴ IU ml⁻¹), and Japanese encephalitis virus (positive with DRG^® JE IgM Antibody Capture ELISA) were also collected to determine the cross-reactivity of the assay. All samples were stored at −80 °C until use.

This study was approved by the Research Ethics Committee of Vietnam Military Medical University, Approval No. 18/QD-HDDD. Written informed consent was obtained from each participant or their legal guardians. Patients’ anonymity and confidentiality were guaranteed by the researchers involved in the study.

Preparation of biotinylated, antibody-conjugated magnetic nanoparticles

Conjugation of the Carboxyl-Adembeads (with a diameter of 100, 200, or 300 nm; Ademtech, Pessac, France) with monoclonal antibodies (10-2699 from Fitzgerald, North Acton, MA, USA or HM164 from EastCoast Bio, North Berwick, ME, USA; mAb) was performed as per manufacturer’s specifications with some modifications. To form mAb-Adembead complexes, 100 µl (three mg) of Carboxyl-Adembeads was first activated by incubating, on a Dynal Biotech rotary shaker, Thermo Fisher Scientific, Dynal Biotech, Waltham, MA, USA (15 rpm), with 240 µl of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) (Sigma, St. Louis, MO, USA) at the concentration of 10 mg ml⁻¹. After activation step, excess of EDC was removed and the nanoparticles were washed with one ml of Activation Buffer 1× (#10101; Ademtech, Pessac, France). Conjugation was then carried out by incubating the nanoparticles with 150 µg of detection antibody at room temperature for 2 h on a Dynal Biotech rotary shaker (15 rpm). Blocking of free carboxyl groups on magnetic nanoparticles was performed by incubating the immunocomplexes with 600 µl of bovine serum albumin (Sigma, St. Louis, MO, USA) (BSA, 0.5 mg ml⁻¹) at 37 °C for 30 min. After being washed with PBS 1×, primary amine groups of BSA and mAbs on the immunocomplexes were biotinylated using EZ-Link™ Sulfo-NHS-Biotin (Thermo Fisher Scientific, Waltham, MA, USA) as per manufacturer’s instructions. Specifically, the complexes were dissolved in 300 µl PBS 1× (pH = 7.4) before being mixed with 15 µl of 10 mM Sulfo-NHS-Biotin. The reaction mixture was incubated on a Dynal Biotech rotary shaker at 15 rpm for 2 h at room temperature before being washed with Storage buffer 1× (#10201; Ademtech, Pessac, France). The final product (biotinylated, mAb-conjugated magnetic beads) was stored in 1,500 µl of Storage buffer 1× (#10201; Ademtech, Pessac, France).

Preparation of the immunochromatography strips

Test strips were manufactured according to Posthuma-Trumpie, Korf & Van Amerongen (2008). Briefly, a Linomat V (Camag, Muttenz, Switzerland) was used to dispense antibodies onto a nitrocellulose membrane of 2.5 cm wide. For the control line, goat anti-mouse IgG (M5899; Sigma, St. Louis, MO, USA) was dispensed at a dose of 2.0 µg cm⁻¹ at the position two cm away from the dipping point. For the test line, capture antibody (10-2698 from Fitzgerald or HM026 from EastCoast Bio) was dispensed at a dose of 0.5–2.0 µg cm⁻¹ at the position 1.5 cm away from the dipping point. The membrane was then dried for 2 h at 37 °C. An absorption pad (Extra Thick Blot Paper, BIO-RAD, Hercules, CA, USA) was applied to the dried membranes, which were then cut to a width of four mm by an Autokun cutter (Hangzhou Autokun Technology, Hangzhou, China). Finally, test strips were sealed in aluminum packages with a desiccation pad and stored at 4 °C until use. Three nitrocellulose membrane types were tested, namely CNPC-SS12, 10 µm with wicking time of 140 ± 28 s/40 mm (MDI Technologies, Ambala Cantt, India), UniSart^® CN140 with wicking time of 95–155 s/40 mm, and UniSart^® CN 95 with wicking time of 65–115 s/ 40 mm (Sartorius, Goettingen, Germany).

Assay procedure

Briefly, 10 µl of sample was mixed with 90 µl of running buffer (PBS 1×; 0.5% BSA; 1% Tween^®-20; (Sigma, St. Louis, MO, USA) pH = 7.4) and one to five µl of biotinylated, mAb-coated magnetic nanoparticles in a well of a low-binding 96-well plate (Corning, Corning, NY, USA) before being flowed vertically onto a LFIA strip. After 25 min, if brown signals are observed at both test line and control line, a positive result could be concluded. If the signal is only observed at the control line, a signal amplification step is carried out by dipping the test strip into a well containing 30 µl of Streptavidin-PolyHRP80 conjugate at the concentration of 1–10 ng ml⁻¹ (Fitzgerald, Acton, MA, USA). The excess of Streptavidin-PolyHRP80 conjugate was washed by dipping the test strip into a new well containing 50 µl of the running buffer, followed by absorbent pad removal and application of 200 µl of 1-Step™ Ultra TMB-Blotting Solution (Thermo Fisher Scientific, Waltham, MA, USA) on the test strip. The enzyme-substrate reaction was maintained for 10 min at room temperature before reading the results.

Analytical sensitivity and cross-reactivity

Analytical sensitivity of the magneto-enzyme LFIA was determined using recombinant NS1 of four dengue serotypes (The Native Antigen Company, Kidlington, UK). These recombinant antigens were produced in mammalian HEK293 cells (purity greater than 95%). Serial dilutions of each NS1 serotype were spiked into negative sera (confirmed by both RT-qPCR and Dengue IgG/IgM 3.0 Combo rapid test as described) and were subjected to analysis using the same procedure as above.

The cross-reactivity of the developed assay was tested using recombinant NS1 from Zika virus at the concentration of 100 ng ml⁻¹; an HBV-positive serum, an HCV-positive serum, and a Japanese encephalitis virus-positive serum. The analytical procedure was carried out the same as specified above.

Data analysis

Data were analyzed according to Jacinto et al. (2018). For optimization experiments, test strips were captured by a Perfection V600 scanner (Epson, Suwa, Japan). Optical densities of test lines and control lines were digitalized to obtain signal values using ImageJ software (ver.1.47; HIH, Bethesda, MD, USA) (Schneider, Rasband & Eliceiri, 2012). The images were converted to 32-bit grayscale to acquire gray level intensities. A 40 × 140 pixels sq. region of interest was used to assess the signal intensities, which were then averaged by taking three or eight replications. GraphPad Prism 6.0 (GraphPad Software Inc., San Diego, CA, USA) was used to statistically analyze and graph the data. The LOD for each dengue serotype was determined as mean + 3SD of blank sample. Unpaired, two-tailed t-tests were performed to determine statistical significance. For testing clinical performance, results were read with the naked eye.

Principle of the magneto-enzyme lateral flow immunoassay

The concept of the magneto-enzyme LFIA developed in this study is to use biotinylated, mAb-conjugated magnetic nanoparticles as the detection complex. Biotin molecules on this complex allow subsequent binding with Streptavidin-polyHRP conjugates, generating amplified signals when TMB substrate is added.

The assay consists of two main steps: (i) sample analysis on the lateral flow strip and (ii) signal amplification (Fig. 1). Specifically, sera of suspected dengue patients are diluted 10-fold in running buffer and mixed with biotinylated, mAb-coated magnetic nanoparticles in a well of a low-binding 96-well plate. The mixture was then allowed to flow onto the test strip by capillary force. When NS1-mAb-magnetic nanoparticle complexes reach the test line position, antigen-antibody sandwiches will be formed, consisting of capture antibodies, NS1, and the biotinylated magnetic nanoparticles. If dengue NS1 exists in high concentration in the specimens, test line signal could be observed directly (within 25 min). In case of low NS1 concentration (faint or no test line observed), a signal amplification step could be performed by flowing Streptavidin-polyHRP conjugates onto the test strip, washing off all the non-bound HRP conjugates and color development by TMB addition (total assay time of 90 min). Goat anti-mouse IgG, immobilized at the control line position, will always generate a signal of brown color, independent of the presence of dengue NS1 in the specimens.

Optimization of the magneto-enzyme lateral flow immunoassay

The performance of magneto-enzyme LFIA strongly depends on (1) antibodies used, (2) magnetic nanoparticles and nitrocellulose membranes, (3) capture antibody application, (4) amounts of detection complexes. Therefore, the effects of these parameters on the test performance were determined.

Firstly, two antibody pairs from Fitzgerald (10-2698: capture antibody and 10-2699: detection antibody; pair #1) and EastCoast Bio (HM026: capture antibody and HM164: detection antibody; pair #2) were tested. At the same concentration of 20 ng ml⁻¹, all the four DENV serotypes were detected using the two antibody pairs. However, pair #1 produced higher signal intensities, especially when detecting DENV-1 and 2 (Fig. 2; Fig. S1). Therefore, 10-2698 and 10-2699 (Fitzgerald) were chosen to be immobilized at the test line position and conjugated to magnetic nanoparticles, respectively. Since DENV-4 generated the lowest signal intensities, further optimizations were carried out using DENV-4 NS1.

Three different sizes (100, 200, and 300 nm) of Adembeads super-paramagnetic nanoparticle were tested. Immunocomplexes prepared from 300 nm Carboxyl-Adembeads were not able to migrate properly along the strip and caused aggregation at the dipping zone, which affected the readout (Fig. S2). Therefore, they were excluded from further analysis. Figure 3 clearly showed that super-paramagnetic nanoparticles with diameter of 200 nm displayed the best performance.

A total of three nitrocellulose membrane types were also evaluated. The highest signals were obtained with the membrane type CNPC-SS12, 10 µm (Fig. 4; Fig. S3). Therefore, “CNPC-SS12, 10 µm” from MDI technologies was chosen for subsequent experiments.

In order to further increase the intensity of test line signals, the optimal amount of capture antibody to be immobilized at the test line was determined. Results showed that the intensity increased proportionally with the amount of antibody used, and reached the saturation when the capture antibody was immobilized at 1.5 µg cm⁻¹ (Fig. 5; Fig. S4). The difference in signal intensity between immobilization of capture antibody at 1.5 µg cm⁻¹ and that at 2.0 µg cm⁻¹ was not statistically significant (P = 0.909).

Similarly, amounts of mAb-magnetic bead complexes and Streptavidin-polyHRP conjugate were optimized. Increasing the amount of mAb-magnetic bead complexes could enhance signal intensities at the test line (Fig. 6; Fig. S5). From the results presented, five µl of biotinylated mAb-magnetic bead complexes were used for detection of NS1 antigen in the next experiments. For the Streptavidin-polyHRP conjugate, various concentrations ranging from one to 10 ng ml⁻¹ were tested. High streptavidin-polyHRP concentration could enhance signal intensities at the test line (Fig. 7; Fig. S6). However, false-positive results were observed at the highest concentration tested (10 ng ml⁻¹). As a result, the optimal concentration of HRP conjugates was five ng ml⁻¹.

Analytical sensitivity

Different concentrations of NS1 antigen were spiked into negative sera to determine the analytical sensitivity of the magneto-enzyme LFIA. The assay could detect directly (without the signal amplification step) NS1 at 1.0 ng ml⁻¹ for DENV-2; 2.5 ng ml⁻¹ for DENV-1 and DENV-3; and 10 ng ml⁻¹ for DENV-4 when observed with the naked eye (Figs. S7 and S8). With the additional signal amplification step, sensitivities were increased by 10-fold for all four serotypes: 0.1 ng ml⁻¹ for DENV-2; 0.25 ng ml⁻¹ for DENV-1 and DENV-3; and one ng ml⁻¹ for DENV-4 (Fig. S9). A calibration curve of log(NS1 concentration) vs. test line signal was also plotted for each dengue serotype and the theoretical LOD was defined as mean of the blank + 3SD. Figure 8 showed that the LODs of our assay (without the signal amplification step) for NS1 from DENV-1, DENV-2, DENV-3, and DENV-4 were approximately 1.4, 0.7, 1.4, and 6.6 ng ml⁻¹ respectively. Of note, LOD of the gold nanoparticle-based LFIA using the same antibody pair was 7.9 ng ml⁻¹ for DENV-2 NS1 (Fig. S10). These results revealed that the use of 200 nm Carboxyl-Adembeads enhanced the analytical sensitivity of LFIA. To our knowledge, the assay developed in this research is the most sensitive NS1 test to date. Before this study, dengue NS1 (serotype 1) could be identified at the lowest concentration of 4.9 ng ml⁻¹ (Kumar et al., 2018).

Cross-reactivity tests

The specificity of our assay was assessed using recombinant NS1 from Zika virus and clinical samples with high viral loads of HBV, HCV, or Japanese encephalitis virus. Results showed an absence of cross-reactivity, even with NS1 from Zika virus (Fig. 9; Fig. S11).

Clinical performance

Among 120 clinical sera, six samples were found positive for both RT-qPCR and Dengue IgG/IgM 3.0 Combo rapid test, 64 samples were only found positive for RT-qPCR, and 50 samples were found negative by both tests (Table S1). Typing of positive samples by nested RT-PCR (Lanciotti et al., 1992) showed that 11 (15.7%) were positive with DENV-1, 55 (78.6%) were positive with DENV-2, and 4 (5.7%) were positive with DENV-4 (Fig. S12). This panel was then used to compare the analytical performance of the developed magneto-enzyme LFIA to that of Dengue Ag rapid test CE kit (CTK Biotech, Inc., Poway, CA, USA). The commercial assay detected only 66 out of 70 positive samples. By contrast, the magneto-enzyme LFIA developed in this study was able to detect all 70 positive sera from dengue-infected patients (Fig. S13). Of note, signal amplification by polyHRP-TMB reaction allowed the detection of four positive specimens that were found negative by the commercial assay (Fig. S13B). Moreover, these four samples were negative for IgM and IgG by Dengue IgG/IgM 3.0 Combo rapid test (CTK Biotech, Inc., Poway, CA, USA). Therefore, our assay is likely to be able to diagnose primary dengue infection at the very first days of the disease. Besides, no false positive results were obtained by both the commercial assay and the magneto-enzyme LFIA (Tables 1 and 2). To summarize, our magneto-enzyme assay showed higher sensitivity (100% vs. 94%) and accuracy (100% vs. 97%) than Dengue Ag rapid test CE (CTK Biotech, Poway, CA, USA); while the specificities of the two assays were equivalent (100%) (Table 3). These results also indicated that the developed assay could accurately detect DENV-1, -2, -4 NS1 in clinical samples.