Generation of a Large Repertoire of Monoclonal Antibodies against Dengue Virus NS1 Antigen and the Development of a Novel NS1 Detection Enzyme-Linked Immunosorbent Assay

Four serotypes of dengue virus (DENV-1–4) cause dengue disease in tropical and subtropical areas of the world (1). DENVs belong to the family Flaviviridae and genus flavivirus and are transmitted by the Aedes mosquito. This vector-borne febrile illness accounts for ∼390 million infections annually, of which ∼96 million infections are symptomatic (2). The four DENV serotypes share antigenic determinants with each other and with other flaviviruses (e.g., Zika virus [ZIKV], yellow fever virus [YFV], Japanese encephalitis virus [JEV], West Nile virus [WNV], and, to some extent, tick-borne encephalitis virus [TBEV]) (3, 4). Infection with one DENV serotype confers lifelong immunity to that particular serotype, and cross-protection with heterologous serotypes is not observed (5). Moreover, secondary DENV infections by heterologous serotypes are prevalent in the endemic regions and often result in more severe disease (6, 7).

DENV infection can only be confirmed by virus-specific laboratory tests, as the clinical symptoms of dengue are often indistinguishable from other febrile illnesses like chikungunya, Zika fever, yellow fever, malaria, etc. (8). The most common diagnostic methods for DENV include the detection of viral RNA, nonstructural-1 (NS1) Ag, and anti-DENV Abs.

NS1 is ∼50-kDa glycoprotein that is secreted in soluble hexameric form (∼300 kDa) from the virus-infected cells (9–11). Detection of NS1 Ag in the patient’s blood is the most practical direct detection method, as this does not require high resources like RNA detection (12). RNA detection requires highly skilled manpower, expensive temperature-sensitive reagents and instruments, and separate space for RNA extraction, reaction setup, and amplification/detection. Therefore, in most DENV-endemic areas, which are economically less advanced, RNA detection is a costlier and less practical solution compared with the Ag detection (13, 14). Moreover, the NS1 Ag can be detected in the patient blood from days 1 to 9 from symptom onset in primary DENV infection, which is potentially longer than the window for RNA detection (i.e., 3–5 d) (15). DENV infection can also be detected indirectly by detecting anti-DENV IgM and IgG Abs. However, Ab detection is insensitive in the early stage of infection and suffers from the problem of cross-reactivity with Abs induced by non-DENV flaviviruses (14).

Several NS1 Ag detection kits are commercially available in the ELISA and rapid lateral flow test formats (13, 16, 17). However, these kits often have poor sensitivity, particularly in secondary dengue infections (18–24). Moreover, the sensitivity of these kits differs for different DENV serotypes (16, 18, 19, 23, 25–27). It is believed that the Ab (IgG) response to NS1 from the previous infection interferes with the recognition of NS1 by the mAbs of the immunoassay (14, 28). To overcome these limitations, we made efforts to generate a large panel of mouse anti-DENV NS1 mAbs (n = 95) with broad specificities, as the properties of the Ag detection immunoassay depend on the target-specific mAbs used in the assay. Extensive characterization of this large mAb repertoire allowed us to identify a single pan–DENV-NS1 mAb (mAb recognizing all four serotypes) as a detection Ab and a mixture of four DENV subcomplex/serotypes-specific mAbs as capture for finalized pan–DENV-NS1 detection ELISA. All of the selected mAbs bind outside the wing domain of NS1, which helps in avoiding interference caused by IgGs from the previous infection, as the wing domain is a prime target of the Ab response (29–33). The developed NS1 Ag detection ELISA was evaluated using samples from febrile adults and pediatric subjects.

BSA fraction V, Trehalose, Tween-20, and other routine chemicals were procured from Merck/Sigma-Aldrich. The 96-well flat-bottom MaxiSorp polystyrene plates, EZ-linkNHS-PEG₄-Biotin, and poly-HRP–labeled streptavidin were from Thermo Fisher Scientific. StabilZyme SELECT, StabilZyme-HRP conjugate stabilizer, as well as MatrixGuard diluent were purchased from Surmodics (Eden Prairie, MN). Tetramethylbenzidine (TMB) substrate was procured from BD Biosciences. Precoated streptavidin microtiter plates were purchased from Kaivogen Oy (Turku, Finland). Hybridoma-SFM for production of mAbs was from Thermo Fisher Scientific, and the protein G-Sepharose for the purification of mAbs was from Cytiva. Normal human serum and normal mouse IgG were from Merck-Millipore. Sodium citrate normal human plasma and K2 EDTA normal human plasma were procured from SeraCare (Milford, MA). Goat anti-mouse IgG, HRP-conjugated goat anti-mouse IgG, HRP-conjugated streptavidin, and streptavidin were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA).

The clinical samples were collected after review and approval by the human ethics committee of the Translational Health Science and Technology Institute (THSTI), Faridabad, India, and the ethics committees of the respective hospitals (Christian Medical College, Vellore, India and All India Institute of Medical Sciences, New Delhi, India). Viral hepatitis (hepatitis B surface Ag [HBsAg]–positive/anti–hepatitis C virus [HCV] Ab–positive) and healthy controls (HIV-, HCV-, and hepatitis B virus–negative) were from commercially available panels from SeraCare. Sera from febrile pediatric patients were collected at All India Institute of Medical Sciences, and sera from febrile adult patients were collected at Christian Medical College.

Regardless of the clinical diagnosis and the tests performed in the respective hospital, all of the samples were subjected to the following tests at THSTI: Panbio DENV IgM Capture ELISA (Abbott, Abbott Park, IL), Panbio DENV IgG Capture ELISA (Abbott), and Scrub Typhus Detect IgM ELISA (InBios International, Seattle, WA). In addition, wherever enough sample volumes were available and clinical signs indicative of dengue, the FTD dengue differentiation (research use only) RT-PCR kit (Siemens Healthcare GmbH) was used to detect and differentiate DENV serotypes. For the evaluation of our developed ELISA, the following panels were made: DENV RT-PCR–positive (both pediatric and adult) (n = 84); DENV RT-PCR–negative but DENV Ab-positive (both pediatric and adult, clinically suspected for dengue and DENV IgM/IgG capture ELISA–positive) (n = 45); scrub typhus–positive (all febrile adults, clinically suspected of scrub typhus and scrub typhus IgM–positive) (n = 153); non-DENV (febrile adult and pediatric subjects, clinically not suspected for dengue, and negative in DENV IgM/IgG capture ELISA) (n = 148); and viral hepatitis (positive for HBsAg or anti-HCV Abs and negative in DENV IgM/IgG capture ELISA) (n = 41).

As mentioned in the previous section, all of the commercial assays were run as per the manufacturer’s instructions. For the RT-PCR, total RNA was extracted from 140 µl of serum samples using the QIAmp viral RNA mini kit (Qiagen, Hilden, Germany). After an RT-PCR run, a threshold was set to determine the baseline for data analysis. The negative control should be below the threshold, whereas the positive control must show a positive amplification trace. All of the clinical samples with a threshold cycle value above the baseline and following an exponential amplification curve profile were considered positive for the presence of specific viral RNA. This is per the instructions from the manufacturer. Apart from the commercial assays used for the characterization of clinical samples, a commercial DENV NS1 ELISA (Platelia Dengue NS1 Ag from Bio-Rad Laboratories) was run in parallel with the developed ELISA, as per the manufacturer’s instructions.

Mammalian HEK-293 cells expressing recombinant secretory hexameric NS1 Ag of DENV-1–4, ZIKV, YFV, WNV, JEV, and TBEV were procured from The Native Antigen Company (London, U.K.). The oligomeric state of these Ags is confirmed by the manufacturer and a published study (34). DENV Ags were biotinylated using the EZ-linkNHS-PEG₄-Biotin kit as per protocol provided by the manufacturer. DENV-1–4 and ZIKV lysates, containing native NS1 Ag apart from the structural proteins, were procured from ZeptoMetrix Corporation (Buffalo, NY). We also expressed full-length NS1 and NS1 wing domain (26–173 aa) of DENV-1–4 and ZIKV recombinantly in the Escherichia coli, and these recombinant proteins were purified in denaturing conditions in the presence of 8 M urea. We have also produced in-house native viral Ags (cell culture supernatants) from DENV-1–4–infected mammalian Vero cells.

The animal experiments were performed after the review and approval by the institutional animal ethics committee of THSTI. Immunizations were performed on 6- to 8-wk-old female BALB/c mice. Three sets of mice were used for immunization. Preimmune sera were collected from each mouse separately for the control. Each set of four mice was immunized with: 1) sequential administration of recombinant secretory hexameric NS1 Ags produced in HEK-293 cells, 2) an equimolar mixture of DENV-1–4 recombinant secretory hexameric NS1, or 3) initial administration of a mixture of DENV-1–4 virus lysates followed by boosters with an equimolar mixture of DENV-1–4 recombinant secretory hexameric NS1 protein.

The mice were primed s.c. at two sites with the Ag-adjuvant emulsion comprising NS1 protein and CFA. Three sets of boosters were administered s.c. or i.p. with an emulsion of NS1 protein and IFA at the interval of 21 d each, and the final booster of Ag was administered i.v. without adjuvant. On the third day after the final booster, mice were sacrificed, and the spleen was collected for the fusion. Mouse spleen cells and Sp2/0 Ag-14 murine myeloma cells were fused to generate Ab-producing hybridoma cell lines as described earlier (35). The splenocytes and myeloma cells were used in a 1:1 ratio for the fusion step, and the fusion was mediated by polyethylene glycol. After the fusion step, cells were resuspended in hypoxanthine/aminopterin/thymidine media and plated in 96-well plates for 10 d at 37°C in a humidified atmosphere with 5% CO₂. After 10 d of undisturbed incubation in hypoxanthine/aminopterin/thymidine selection media, plates were examined for the presence of positive clones through ELISA. Culture supernatants were carefully aspirated from the master plates and screened for the presence of anti-NS1 Abs in an indirect ELISA using HEK-293–produced recombinant secretory hexameric NS1 proteins of four DENV serotypes.

The 96-well streptavidin-precoated plates were washed with PBS (pH 7.4) and blocked with 150 µl/well of BSA solution (1% BSA in PBS [pH 7.4]) for 1 h at room temperature (RT) (23 ± 2°C) with slow shaking. After the incubation, plates were washed twice with 500 µl/well of wash buffer (5 mM Tris, 0.9% NaCl, 0.02% NaN₃, and 0.05% Tween-20 [pH 7.5]) and incubated with 10 ng/50 µl/well biotinylated recombinant secretory hexameric NS1 Ag of all four DENV serotypes and unrelated Ags. After 1 h of incubation at RT, plates were washed two times with wash buffer and incubated with the hybridoma culture supernatant diluted 1:1 in assay buffer (50 mM Tris, 135 mM NaCl, 0.05% NaN₃, 0.5% BSA, and 0.01% Tween-40 [pH 7.75]) for 1 h at RT with slow shaking. Further, the plates were washed three times with wash buffer followed by the addition of 50 µl/well of 1:5000 diluted HRP-conjugated goat anti-mouse IgG for 45 min with slow shaking at RT. After incubation, plates were washed four times with wash buffer and developed with TMB substrate for 30 min. The reaction was stopped, and absorbance was measured as described above.

To determine the reactivity of hybridoma clones with E. coli–produced DENV NS1 and conserved wing domain, 96-well MaxiSorp plates were coated overnight at 4°C with 50 ng/50 µl/well of either full-length NS1 or NS1 wing domain of all four DENV serotypes and ZIKV diluted in carbonate-bicarbonate buffer (pH 9.6). The next day, plates were washed twice with 10 mM PBS (pH 7.4) and blocked with 1% BSA for 2 h at RT with slow shaking. The rest of the process was the same as described for recombinant secretory hexameric Ags.

Capture immunoassay was performed by coating 0.5 µg/ml of goat anti-mouse IgG in 50 mM carbonate-bicarbonate buffer (pH 9.6) on 96-well MaxiSorp plates. Plates were incubated overnight at 4°C. The following day, plates were washed twice with PBS (pH 7.4) and blocked with 150 µl/well of 1% BSA for 2 h at RT with slow shaking. After washing twice with wash buffer, the wells were incubated with the hybridoma cell culture supernatant diluted 1:1 in assay buffer for 3 h at RT with slow shaking. Next, the plates were washed twice with wash buffer, and 5 ng/50 µl biotinylated recombinant secretory hexameric NS1 Ag from four DENV serotypes was added to the wells. The Ags were incubated for 5 min. Next, the plates were washed three times with wash buffer, and 25 ng/50 µl/well of Europium-labeled streptavidin was added to the assay wells. Plates were then incubated for 45 min at RT with slow shaking, washed four times, followed by the addition of 100 µl/well of Europium enhancement solution. The enhancement solution facilitates the release of bound Europium from the immunocomplex, which is then measured with excitation at 340 nm and emission at 615 nm using the Envision multimode reader (PerkinElmer).

Isotypes of 95 anti-NS1 mAbs were determined using a commercial mouse mAb isotyping kit (Sino Biological, Beijing, China). Briefly, isotype-specific rabbit anti-mouse IgG1, IgG2a, IgG2b, IgG3, and IgM were coated overnight on MaxiSorp plates. Similarly, to determine L chain, polyclonal rabbit anti-mouse Igκ and Igλ (Merck-Millipore) were coated on the MaxiSorp plates. Hybridoma culture supernatants were added to the coated plates, and detection was performed with HRP-conjugated anti-mouse IgG and TMB as substrate.

The sandwich immunoassay was performed with selected 10 anti-NS1 mAbs to identify suitable pairs. A total of 50 µl of 10 µg/ml purified anti-NS1 mAbs was passively coated on 96-well MaxiSorp plates. After overnight incubation at 4°C, plates were washed twice with PBS (pH 7.4) and blocked with 1% BSA. After 2 h of incubation, the wells were subjected to the addition of 25 µl/well of 0.5 µg/ml diluted biotin-labeled anti-NS1 mAbs (10 mAbs). This was immediately followed by the addition of 25 µl/well of 5 ng/ml of recombinant secretory hexameric DENV NS1 spiked in pooled normal human serum. After 1 h of incubation, plates were washed three times with wash buffer, and 50 µl of 25 ng/ml HRP-labeled streptavidin was added. Plates were then incubated for 30 min at RT with slow shaking. After incubation, plates were washed four times with wash buffer and developed with TMB substrate for 10 min. Finally, the reaction was stopped, and absorbance was measured as described above.

Affinity measurement of anti-NS1 mAbs was done using biolayer interferometry on the Octet K2 system. For the assay, anti-mouse IgG Fc capture biosensors were prehydrated with 10 mM PBS for 10 min. Posthydration, the biosensors were dipped into wells containing 200 µl of 10 µg/ml diluted anti-NS1 mAbs for 120 s. Biosensors were then transferred to fresh PBS-containing wells for 60 s to remove any nonspecific binding. Ag binding was then evaluated by dipping the mAb-coated biosensor into the buffer containing different dilutions of NS1 Ag (50, 25, 12.5, 6.25, and 3.125 nM) for 180 s, followed by a dissociation time of 600 s in another well with PBS. All of the sensorgrams were referenced, and data fitting was done with the 2:1 binding model through Octet analysis software.

The 96-well flat-bottom MaxiSorp plates were coated with 100 µl/well of 12.5 µg/ml of anti-NS1 mAbs mixture diluted in 50 mM carbonate-bicarbonate buffer (pH 9.6). After overnight incubation at 4°C, plates were washed two times with 500 µl/well of PBS and blocked with 200 µl/well of blocking cum stabilizing buffer (1% BSA, 3% trehalose, and 0.05% sodium azide in PBS [pH 7.4]) for 2 h at RT. The plates were tapped down gently to remove the blocking buffer and incubated again for 1 h at 37°C. Plates were then dried overnight in a desiccator. The next day, the plates were sealed in separate aluminum pouches with desiccant and stored at 4°C or RT.

Stabilization of biotinylated anti-NS1 mAb (FMN36), and poly-HRP–conjugated streptavidin in liquid form was done in StabilZyme SELECT and StabilZyme-HRP conjugate stabilizer, respectively. Biotinylated FMN36 mAb was diluted to 1 µg/ml in StabilZyme SELECT diluent supplemented with 0.5% Tween-20, whereas the poly–HRP-streptavidin was diluted to 0.5 µg/ml in StabilZyme-HRP stabilizer with 0.5% Tween-20. The diluted reagents were stored at 4°C or RT. The kit also contains assay diluent, namely MatrixGuard diluent supplemented with normal mouse IgG. The kit was also supplemented with 25× wash concentrate (125 mM Tris, 22.5% NaCl, 0.5% NaN₃, and 0.125% Tween-20 [pH 7.75]), ready-to-use TMB and stop solution (1 M H₂SO₄), assay calibrator (see below), positive control (recombinant secretory hexameric NS1 diluted in StabilZyme SELECT diluent), and negative control (pooled normal human serum).

Seventy-two sera from healthy controls were run on the developed NS1 ELISA kit. The cutoff was calculated using the formula: average absorbance of 72 negative samples + 7 ∗ SD of the absorbance of 72 negative samples. A cutoff calibrator was prepared by spiking 0.1 ng/ml DENV 1–4 NS1 mixture (0.025 ng/ml of each serotype) in pooled negative sera, giving similar absorbance as the cutoff value. The developed cutoff calibrator was run in triplicates on each assay plate.

All of the components of the developed kit, as described above, were used in the assay procedure. The dry plate was removed from the aluminum pouch, 50 µl of sample diluent, 50 µl of the sample (test samples/negative control/positive control/assay calibrator), and 50 µl of prediluted biotinylated anti-NS1 detection mAb were added to each well. The 150 µl mixture was incubated for 1 h at RT with slow shaking. Next, the plate was washed four times with 1× wash buffer (500 µl/well), and 100 µl/well of prediluted poly–HRP-streptavidin conjugate was added and incubated for 30 min at RT with slow shaking. Next, the plate was washed four times with 500 µl/well of wash buffer, and 100 µl/well of TMB was added and incubated for 10 min. The reaction was stopped by adding 100 µl/well of stop solution, and the absorbance was measured at 450 nm with 650 nm as reference wavelength using a microplate reader. The cutoff was calculated by taking the average absorbance of triplicate of the calibrator. The signal-to-cutoff (S/Co) ratio was calculated by the formula: S/Co ratio = absorbance of test sample/absorbance of calibrator. Test samples with S/Co value ≥1 were considered positive.

Dry plates coated with anti-NS1 mAb mixture as well as the diluted ready-to-use biotinylated anti-NS1 mAb and poly–HRP-streptavidin were stored at RT for 50 d. The performance of RT-stored components was assessed in comparison with the freshly prepared/diluted components (mAb mixture-coated plate, biotinylated detector mAb, and diluted poly–HRP-streptavidin conjugate). The assay procedure was the same as described above. A total of 50 ng/ml and 5 ng/ml of recombinant secretory hexameric NS1 spiked in pooled normal human serum was used as the sample for this comparison.

HEK-293–expressed recombinant secretory hexameric NS1 Ag from four DENV serotypes or DENV (serotypes 1–4)–infected Vero cell culture supernatants (native NS1) were diluted in three different matrices (namely pooled normal human serum, pooled normal human sodium citrate plasma, and pooled normal human EDTA plasma) and used as samples in the developed NS1 ELISA kit. Recombinant NS1 were 3-fold diluted from 20 ng/ml to 0.03 ng/ml, whereas native NS1 were diluted from 1:200 to 1:48,600 in the respective matrices. The assay procedure was the same as described above.

Platelia dengue NS1 Ag ELISA (Bio-Rad Laboratories) was carried out as per the protocol provided by the manufacturer. The Ag dilutions were only done in pooled normal human serum for analytical comparison of developed NS1 ELISA with Bio-Rad Laboratories ELISA. The dilutions were run in three replicates (three independent dilutions) on the developed NS1 ELISA kit and commercial Bio-Rad ELISA. The S/Co ratios were calculated for both the tests as the ratio of absorbance value of the test sample to the cutoff value. The highest dilution of each NS1 serotype with S/Co ratio ≥1 was considered the limit of NS1 detection for the particular serotype in that test.

Dilution curves were drawn using Prism 9.2.0 (GraphPad). Performance characteristics, including sensitivity and specificity with 95% confidence interval (CI) of the study ELISA, were calculated using the online version of MedCalc statistical software. For other analysis work, Microsoft Excel was used.

A repertoire of 95 anti-DENV NS1 mouse mAbs was generated using hybridoma technology. All of the mAbs bind to the DENV recombinant secretory NS1 (hexameric form), as this was the Ag used for immunization. The mAb repertoire was extensively characterized by determining the reactivity with E. coli–expressed monomeric NS1 and NS1 wing domain (aa 26–173) of DENV serotype 1–4 and ZIKV, as described in the Materials and Methods. Based on the reactivity analysis, Abs were divided into three primary groups: pan-DENV (mAbs binding to all of the four DENV serotypes; n = 31), DENV-subcomplex (mAbs binding to more than one DENV serotype but not to all four; n = 37), and DENV serotype-specific (mAbs binding to single DENV serotype; n = 27) (Table I). Considering the reactivity with secreted NS1, including ZIKV NS1, E. coli–expressed monomeric NS1, and NS1 wing domain, the generated repertoire can be put in 36 mAb classes (Table I, Supplemental Table I), indicating the repertoire’s breadth.

Out of 27 serotype-specific mAbs, 5 mAbs bind to DENV-1, 9 mAbs to DENV-2, 2 mAbs to DENV-3, and 11 mAbs to DENV-4 (Table I). Most of the pan-DENV NS1 mAbs were found to recognize the monomeric NS1 (28 of 31) and also the conserved wing domain of NS1 (24 of 31) (Table I). Fewer mAbs from other groups were found to react with the monomeric form of NS1 (15 of 37 from the DENV-subcomplex group and 12 of 27 from the serotype-specific group) and the wing domain (10 of 37 from the DENV subcomplex group and 14 of 27 from the serotype-specific group). Out of 14 ZIKV NS1 cross-reactive mAbs, 11 were from the pan-DENV NS1 group. Further, 8 of 14 ZIKV cross-reactive mAbs were found to recognize the conserved wing domain. One DENV-4–specific mAb and two subcomplex-specific mAbs recognize ZIKV NS1 (Table I). In addition, 55 mAbs were found to recognize both HEK-293 cell–expressed secretory NS1 and E. coli–expressed monomeric NS1, whereas 40 mAbs recognized secretory NS1 but not the E. coli–expressed monomeric NS1. More than half of the repertoire (48 of 95) binds to the wing domain (Table I). These results are similar to the information available about the dominance of anti-wing domain Abs in DENV-infected human subjects (29, 30, 36).

All of the 95 mAbs were also assessed for their capability to capture the NS1 Ag, during a very short interaction period (5 min), in a capture immunoassay format (data not shown). Ten mAbs that performed well in this stringent assay (high affinity) and do not bind to the wing domain were further evaluated in the Ab pairing study for the final Ag sandwich ELISA. Table I summarizes the characteristics of the shortlisted clones (indicated with d). All of these 10 mAbs have IgG1 isotype with κ L chain (Table I). These were scaled up for large production batches.

All 10 mAbs were tested as capture, and detection mAb and the HEK-293–expressed secreted hexameric NS1 Ags of four DENV serotypes were used as test Ags spiked in normal human serum. The detection Abs were biotin-labeled, and the Ag–Ab complexes were traced with HRP-conjugated streptavidin. A cross-tabulation analysis of all 10 × 10 combinations revealed the binding and blocking pairs, as presented in (Fig. 1 for the four DENV serotypes. A pair of capture and detection Abs providing a strong positive reaction for NS1 Ag was considered a binding pair. A pair with no reactivity with any NS1 protein was termed a blocking pair. Based on cross-tabulation screening and positive reaction for different DENV serotypes, we further shortlisted five mAbs for assay development (FMN91, FMN83, FMN6, FMN89, and FMN36). These mAbs are indicated with the e in Table I.

Different combinations of the shortlisted mAbs were formulated and explored as both capture and detection Ab (data not shown). After screening all of the possible combinations to detect NS1 from all four DENV serotypes efficiently, we deduced a mixture of four anti-NS1 mAbs, namely FMN91, FMN83, FMN6, and FMN89, from subcomplex and serotype-specific mAb groups as capture and a single pan–anti-NS1 mAb (FMN36) as a biotinylated detector mAb. The binding affinities (K_D) of these mAbs with DENV NS1 were determined using the biolayer interferometry and found to be in the subnanomolar range (Fig. 2). The selected mixture of four mAbs as capture paired with a single biotinylated pan–anti-NS1 mAb, detected with poly-HRP–conjugated streptavidin, were used in the final assay, as described in the Materials and Methods.

The assay was converted in a kit format comprising a ready-to-use mAb mixture-coated dry-stable 96-well plate, the prediluted biotinylated detector mAb, and the poly-HRP–conjugated streptavidin in stabilized liquid form (see Materials and Methods). To study the stability of the components, the developed NS1 ELISA kit was incubated at RT (23 ± 2°C) for 50 d. The performance of the kit, subjected to RT incubation, was similar to freshly prepared components (Supplemental Fig. 1).

Seventy-two sera from the nonfebrile controls were run on the developed ELISA kit to determine assay cutoff. A cutoff absorbance of 0.120 was calculated, as explained in the Materials and Methods. To keep the consistency between different assay runs, a cutoff calibrator was prepared by spiking 0.1 ng/ml DENV 1–4 NS1 mixture (0.025 ng/ml of each serotype) in pooled normal human serum, giving similar absorbance (0.122) as the cutoff absorbance (0.120), determined using 72 negative samples. The prepared cutoff calibrator was run in triplicate on each assay plate to maintain consistency in the results (see Materials and Methods). The results of the test samples were determined by calculating the S/Co ratio (absorbance of test sample/absorbance of cutoff calibrator), and the ratio of ≥1 was counted as positive.

Next, we evaluated the compatibility of the developed ELISA kit with different blood-based sample matrices (i.e., serum, sodium citrate plasma, and EDTA plasma). As can be seen in Supplemental Fig. 2, no significant difference in the analytical sensitivity was observed among the different matrices.

We assessed the cross-reactivity of non-DENV flavivirus NS1 Ags in the developed DENV NS1 ELISA. Normal human serum spiked with HEK-293 expressed recombinant secretory hexameric NS1 Ag of ZIKV, WNV, YFV, JEV, and TBEV was run in the developed NS1 ELISA. The developed NS1 ELISA did not show any reactivity with non-DENV flavivirus NS1 Ags tested (Fig. 3A). We also assessed the developed NS1 ELISA’s specificity in the same experiment series by testing ZIKV lysate containing native NS1 (see Materials and Methods for details). Again, the developed NS1 ELISA did not show any cross-reactivity, even with the native form of ZIKV NS1 (Fig. 3B).

To find out the analytical sensitivity of the developed NS1 ELISA, we determined the limit of detection (LoD) for NS1 from each DENV serotype. This was done using recombinant secretory hexameric NS1 Ags and the native NS1 Ags (infected Vero cell culture supernatant). For comparison, the same samples were run on the commercial Platelia NS1 ELISA (Bio-Rad Laboratories). The head-on comparison of the S/Co ratio obtained for two tests indicated considerably higher sensitivity of the developed NS1 ELISA kit for all four DENV serotypes; the LoDs were 0.08 and 0.74 ng/ml (DENV-1), 0.24 and 6.66 ng/ml (DENV-2), 0.74 and 2.2 ng/ml (DENV-3), and 0.08 and 6.66 ng/ml (DENV-4) for the developed NS1 ELISA and Platelia ELISA, respectively (Fig. 4). Similar results were obtained for the native NS1 Ags (Fig. 4).

The developed ELISA kit was evaluated using a DENV RT-PCR–positive panel (panel 1; n = 84), DENV RT-PCR–negative but IgM/IgG-positive panel (panel 2; n = 45), and three DENV-negative panels (i.e., scrub typhus–positive [panel 3; n = 153], non–dengue fever [panel 4; n = 148], and viral hepatitis [panel 5; n = 41]). All of the panels contain only serum samples. The commercial Platelia NS1 ELISA (Bio-Rad Laboratories) was run in parallel for comparison. Compared to RT-PCR (panel 1), the developed ELISA showed 78.57% (95% CI 68.26–86.78) sensitivity, whereas Platelia NS1 ELISA showed a sensitivity of only 60.71% (95% CI 49.45–71.20) (Table II). The detection rate in panel 2 (RT-PCR–negative) was 42.22% (95% CI 27.66–57.85) for the developed NS1 ELISA and 17.78% (95% CI 8.00–32.05) for Platelia ELISA. All of the samples in panel 2 were suspected of dengue and positive for anti-DENV Abs (Table II). None of the dengue-negative samples from panels 3, 4, and 5 scored positive in the developed NS1 ELISA (specificity: 100% [95% CI 98.93–100.00]). The same was the case for Platelia NS1 ELISA (specificity: 100% [95% CI 98.07–100.00]), in which panels 4 and 5 were tested (Table II). The performance evaluation data in the scatter plot are presented in Supplemental Fig. 3.

DENV RT-PCR–positive samples (panel 1) were further categorized into primary and secondary dengue based on the reactivity in DENV IgG capture ELISA (Table III). Positivity in IgG capture ELISA indicates secondary DENV infection, as this test aims to detect elevated levels of DENV-specific IgG resulting from the anamnestic immune response (37, 38). In the secondary dengue, the developed NS1 ELISA showed a sensitivity of 70.18% (95% CI 56.60–81.57) compared with Platelia ELISA with a sensitivity of only 47.37% (95% CI 33.98–61.03) (Table III). Both of the ELISAs have high sensitivity for the primary dengue (Table III).

The performance of the developed NS1 ELISA was also assessed on DENV serotype level to determine any bias toward a particular serotype. The developed NS1 ELISA detected NS1 from all four serotypes equally (sensitivity 72.0–84.6%) (Table IV). However, poor performance was observed for Platelia ELISA for DENV-2, in which only 8 of 21 (38.10%) samples were detected compared with 17 of 21 (80.95%) in the developed NS1 ELISA (Table IV).

Dengue is endemic in almost all of the tropical and subtropical areas of the world. It does not have an efficient vaccine or antiviral treatment and demands early and accurate diagnosis to manage the patient’s condition. Most DENV infections are asymptomatic, but the symptomatic one can progress toward severe dengue if not diagnosed and managed early. NS1 is an ideal candidate for DENV detection, as it is a direct marker, unlike Ab detection, and is more straightforward than viral RNA detection. Moreover, NS1 is present in the blood of infected individuals from the first day of symptom onset to around day 9 in the primary infection (39), which is a potentially longer and clinically significant window compared with the presence of viral RNA. However, these detection windows are theoretical and vary on an individual level. Several NS1 detection ELISAs and lateral flow tests are available in the market and routinely used for DENV detection. However, multiple studies have shown reduced sensitivity of the NS1 Ag detection in secondary DENV infection, limiting this method’s utility (18–21, 40). The lack of adequate sensitivity in secondary infection is highly attributed to the presence of anti-NS1 Abs in blood from previous infections, as these Abs, due to the formation of Ag–Ab complexes, eventually mask the available epitopes on the NS1 Ag (14, 28, 41). Though immune-complex dissociation methods (e.g., heating) have proven to increase the sensitivity of NS1 detection tests in secondary infections (42–44), these methodologies are not easy to incorporate into an assay outside research laboratories. Another critical problem with the commercially available NS1 detection kits is that they do not detect four DENV serotypes equally and show serotype preference, leading to the differences in the sensitivity for different serotypes (16, 18, 19, 23, 25–27).

Different approaches have been reported in the literature to generate immunoassays for DENV NS1 detection. Initial studies have used anti-NS1 polyclonal Abs for assay development (11, 39). In these studies, the test performance was assessed with a limited number of clinical samples from patients infected with a single DENV serotype. The assays based on polyclonal Abs are also prone to batch-to-batch variation and are often not preferred.

mAbs, in contrast, ensure consistency in quality and supply. Researchers have extensively exploited the mAbs to develop NS1 Ag detection ELISAs. For example, Puttikhunt et al. (45) generated mAbs, using DNA immunization, that were employed to develop a pan-DENV NS1 ELISA (46). The analytical sensitivity of their ELISA was ∼50–160 ng/ml for different DENV serotypes, which is insufficient to provide a sensitive detection in blood.

Ding et al. (47) generated mouse mAbs using E. coli–expressed recombinant NS1 and developed a pan-DENV NS1 ELISA. However, the analytical sensitivity in terms of protein quantity was not determined, and the test was only evaluated using samples from patients infected with DENV-1 and DENV-3.

Gelanew et al. (48) generated mouse anti-NS1 mAbs using insect cells expressing recombinant NS1 and used the mAbs to develop the pan-NS1 detection ELISA. This ELISA employed heat treatment of the samples for immune-complex dissociation, which increased the absorbance value after heat treatment. However, the clinical evaluation was performed only using 15 DENV-positive and 20 DENV-negative patient samples. Moreover, the comparison with RT-PCR and the information on the infecting DENV serotype were missing. Furthermore, the procedure involving heat treatment of the sample is difficult to incorporate in assays used in clinical testing laboratories. Researchers have also developed NS1 ELISAs to detect single DENV serotypes (27, 43, 44). However, the utility of these ELISAs is uncertain in diagnosing DENV infection, as most of the DENV-affected areas have hyperendemicity with the circulation of more than one DENV serotype.

Additionally, researchers have tried developing NS1 immunoassays for DENV serotype differentiation. For example, Lebani et al. (49) isolated serotype-specific mAbs using a phage display Ab library and developed a serotype-specific DENV NS1 assay in which serotype-specific mAbs were immobilized on four different fluorescent microspheres. However, the analytical sensitivity of this test was 10–100 ng/ml for different serotypes. Moreover, the performance evaluation of this test was not shown with clinical samples. Additionally, this platform cannot be used for routine testing. Puttikhunt et al. (50) developed NS1 ELISAs, for DENV serotype differentiation, comprising a single capture mAb and four different serotype-specific detection mAbs in separate wells. However, in the clinical evaluation, DENV-3–specific ELISA showed cross-reactivity with DENV-1 samples. In another study, Röltgen et al. (51) developed four serotype-specific NS1 ELISAs by pairing four capture Abs with four different detector Abs. This should be noted that different wells or plates are required to detect different serotypes using this method. Four serotype-specific assays developed in this study were evaluated using 46 RT-PCR–positive samples with only 6 and 4 samples for DENV-3 and DENV-4, respectively. Samples used in this study were from travelers returning from endemic countries and likely to have primary DENV infection. The serotyping assays are valuable for epidemiological surveillance but not for the clinical diagnostic purposes.

Most experimental ELISAs, published in the literature, lack data concerning differences in the performance for primary and secondary DENV infection, do not show analysis on serotype-level bias, and lack the specificity analysis using samples from nondengue febrile patients. Our study has comprehensively addressed all of the above-mentioned limitations.

As the suitable mAbs are the most critical component in any Ag detection immunoassay, we generated mAbs with exceptional properties using hybridoma technology coupled with appropriate immunization strategies (see Materials and Methods), including: 1) sequential administration of recombinant secretory hexameric NS1 of four DENV serotypes in mice to ensure the recalling of response against conserved regions of the NS1 to obtain pan-DENV NS1 mAbs, 2) immunization of mice with the mixture of recombinant secretory hexameric NS1 of four DENV serotypes to obtain Abs of all specificities, and 3) immunization with crude native NS1 preparation followed by highly pure recombinant secretory hexameric NS1 Ag to obtain mAbs recognizing the native Ag. These approaches yielded a panel of 95 anti-DENV NS1 mAbs.

Considering the utility of the anti-NS1 mAbs for diagnostic applications and possible therapeutic applications (52), we have extensively characterized the generated mAb repertoire.

All of the generated mAbs recognized the native NS1 from DENV-infected Vero cell culture supernatant (data not shown), apart from recombinant secretory NS1 Ag, confirming the resemblance of Ag used for immunization (recombinant secretory hexameric NS1) with native NS1 Ag.

To understand the functional diversity of the generated mAb repertoire, we also checked the binding of our mAbs with E. coli–expressed monomeric NS1 and the conserved NS1 wing domain. The NS1 wing domain is highly conserved, a majorly disordered loop, and considered a hotspot NS1 domain targeted most during the human Ab response (29, 30, 36). As we wanted to make our test suitable for NS1 detection in both primary and secondary infection, it was prudent to avoid using mAbs recognizing wing domain to prevent epitope masking to some extent. We also screened the mAb repertoire for cross-reactivity with ZIKV NS1, as ZIKV is phylogenetically the closest flavivirus and cocirculates with DENV. In the immunoassay, cross-reactive mAbs on both sides (capture and detector) may lead to compromised performance. Therefore, we considered this as one of the criteria for the final mAb pair. Based on the binding of the mAbs with four DENV serotypes, ZIKV NS1, monomeric NS1 form, and NS1 wing domain, Abs were categorized in a total of 36 mAb classes indicating the success in generating anti-NS1 mAb repertoire with good functional diversity (Supplemental Table I). This diversity may even be higher, as we have not checked the reactivity with the dimeric form of NS1. It is possible that some of the mAbs may be binding to both hexamer and dimer due to shared epitopes and other mAbs binding exclusively to the hexameric form, creating further mAb classes.

Like the dominance of the wing domain–targeting Abs in humans (29–33), we have found that more than half of our mAbs, from mice, bind to the wing domain (Table I), confirming the immunodominance of this conserved region. As expected, most of the pan-DENV mAbs (binding to all four DENV serotypes) bind to the wing domain (Table I). Interestingly, we also obtained 40 mAbs that only bind to the secretory NS1, showing no binding to the E. coli–expressed monomeric Ag. As explained before, the mAbs that bind exclusively to secretory NS1 may be targeted to epitopes that are either present only on a hexameric form or both on hexameric and dimeric form, as we have not checked the reactivity with dimeric form of NS1. Moreover, as the monomeric Ag was expressed in E. coli and lacked glycosylation, the nonbinding of mAbs to this Ag may also be due to this difference. Though most ZIKV cross-reactive mAbs were from the pan-DENV group, we found one DENV-4–specific mAb (not binding with DENV-1, -2, and -3) cross-reacting with ZIKV NS1, indicating that there can be common B cell epitope between a DENV serotype and ZIKV without a presence in the remaining three DENV serotypes.

From a broader perspective, we did not observe an explicit pattern for the specificity of generated mAbs based on the immunization strategy. However, it appears that different immunization approaches increased the functional diversity of the repertoire. Based on individual mAbs, there appears to be some pattern, e.g., the truly pan anti-NS1 mAb (FMN36, used as a detection Ab in the finalized assay) is derived from the mice immunized using sequential administration of NS1 protein of four DENV serotypes, which supports the idea of recalling of response against the conserved region of NS1, thereby generating a cross-reactive Ab. Moreover, the finalized assay’s capture Abs (FMN91, FMN83, FMN6, and FMN89), which were either subcomplex- or serotype-specific, were obtained from the immunization strategy involving the mixture of four DENV NS1s, indicating that immunization of animals with a mixture of NS1 proteins helps to generate Abs of all specificities. We further believe that an intelligent immunization strategy alone cannot guarantee the desired Ab specificity, and a robust screening component is equally critical.

The final assay configuration using a mixture of four mAbs as capture and a single pan-DENV NS1 mAb as detection Ab was established based on the following criteria: 1) appropriate pairing of high-affinity Abs to efficiently detect all four DENV serotypes, 2) removal of wing domain–recognizing mAbs to prevent epitope masking issue in secondary DENV infection, and 3) avoidance of a pair capable of detecting ZIKV NS1. All of the five final mAbs used in the assay had subnanomolar affinities, as determined using the biolayer interferometry-based kinetics experiment (Fig. 2). This indicates the success of the strategy used to select high-affinity mAbs by an immunoassay with a very short time (5 min) for interaction between soluble NS1 Ag and the coated capture mAbs. In addition to the performance of the mAb pair, the assay sensitivity also depends on the reporter system’s efficiency for signal generation. Accordingly, in the developed ELISA, we have used the biotinylated pan–DENV-NS1 mAb FMN36 as a detection Ab in conjunction with the poly–HRP-streptavidin conjugate instead of direct enzyme-conjugated mAb as a detector.

The developed NS1 ELISA was converted into a stable kit format in which the assay plate was dried stabilized (no need to coat the wells each time), and the detector mAb and poly–HRP-streptavidin were diluted in stabilizing solution (ready to use, no dilution required by the end user). The stability of the test components was assessed at RT (23 ± 2°C) for 50 d with no deterioration in the performance compared with freshly prepared components (Supplemental Fig. 1), indicating suitability for tropical areas where temperature can be high and intermittent power cuts are expected. Due to the ready-to-use components of the developed NS1 ELISA, a total assay run time of 100 min was achieved, which is essential for quick reporting of the results to the patient and health care provider.

We used 72 nondengue serum samples (not part of the negative panel used for test evaluation) and determined the assay’s cutoff absorbance. However, it is impossible to run 72 negative samples each time with test samples to calculate the S/Co ratio, which is why the commercial kits have a cutoff calibrator or control to calculate the result and avoid inconsistency between the assay runs. For this reason, we generated a cutoff calibrator based on the cutoff absorbance obtained with 72 negative samples. This calibrator was run in triplicates in all of the assay plates to calculate the plate-specific cutoff value. Despite its importance, this type of calibrator is often missing from the experimental ELISAs reported in the literature. Another important feature of an immunoassay should be the compatibility with different sample matrices. This helps streamline the testing workflow with other ordered tests (e.g., for febrile illness); several tests are ordered together, targeting different pathogens and host markers. Therefore, we checked the compatibility of the developed NS1 ELISA with serum, EDTA plasma, and citrate plasma and found it to have a similar performance with all three matrices (Supplemental Fig. 2).

Dengue shares multiple symptoms with illness caused by other flaviviruses like ZIKV, YFV, JEV, WNV, and TBEV (3, 4). Moreover, DENV also cocirculates with ZIKV, YFV, and JEV in different parts of the world (8), which demands cross-reactivity assessment of the DENV NS1 test with non-DENV flavivirus NS1 Ags. In our assessment, we found no cross-reactivity of NS1 from ZIKV, YFV, JEV, WNV, and TBEV in the developed NS1 ELISA (Fig. 3A). For this experiment, we used a very high concentration (500 ng/ml) of the recombinant secretory NS1 of ZIKV, YFV, JEV, WNV, and TBEV to ensure that cross-reactivity is not an issue even at high concentrations. These commercially procured Ags were produced in HEK-293 cells and are in native-like oligomeric confirmation, as indicated earlier (34). As ZIKV is most closely related to DENV, we also checked the ZIKV native NS1 Ag in the developed NS1 ELISA and found no cross-reactivity (Fig. 3B). These results indicate that the developed NS1 ELISA only detects NS1 from four DENV serotypes and not the NS1 from other flaviviruses. It should be noted that the detection Ab (FMN36) in the developed NS1 ELISA cross-reacts with ZIKV NS1. However, due to the choice of non–cross-reactive capture mAbs, the developed NS1 ELISA only recognizes DENV NS1 and not the ZIKV NS1.

During the analytical and clinical performance evaluation of the developed NS1 ELISA kit, we parallelly performed the testing on commercial dengue NS1 ELISA (Platelia from Bio-Rad Laboratories) to truly understand the performance of the developed NS1 ELISA. The choice of Platelia (Bio-Rad Laboratories) ELISA for comparison relies on the fact that this is the most extensively used NS1 ELISA globally, with several published reports on the performance of this test (18, 19, 28, 53–55).

We determined the analytical sensitivity by testing the serial dilution of DENV recombinant secretory NS1 and the native NS1 in pooled normal human serum. The same spiked samples were also run parallelly on the Platelia ELISA (Bio-Rad Laboratories). We observed a comparable detection limit for DENV-3 NS1 for both tests. However, for other serotypes, the developed NS1 ELISA kit has higher sensitivity (up to 81-fold) than Platelia ELISA (Fig. 4). In addition, the higher analytical sensitivity of the developed NS1 ELISA kit was consistent for both the recombinant and native NS1.

Detection of DENV using RT-PCR is often considered as the gold-standard method in the early stage of infection (56–59). To assess the performance of the developed NS1 ELISA, we used human serum samples that were positive for DENV by RT-PCR (panel 1). The same panel was also run on the commercial Platelia NS1 ELISA (Bio-Rad Laboratories). Like the results with purified NS1, the developed NS1 ELISA was found to have higher sensitivity with the DENV RT-PCR–positive clinical samples as compared with commercial ELISA (Table II, Supplemental Fig. 3). One of the major limitations of NS1 detection methods is the poor sensitivity in the secondary dengue, and, as secondary dengue is prevalent in hyperendemic areas and likely to have more severe disease, we decided to check the performance of our ELISA with samples from this important group. To this end, we divided the RT-PCR–positive samples in primary and secondary dengue based on the negative or positive results, respectively, in the DENV IgG capture ELISA. In the secondary DENV infection, due to the anamnestic response, a high amount of anti-DENV IgGs are produced in a very short time, leading to a positive result in IgG capture ELISA (37, 56, 60). In the RT-PCR–positive secondary dengue cases, the developed NS1 ELISA showed a sensitivity of 70.18% (95% CI 56.60–81.57), whereas the Platelia ELISA showed a sensitivity of only 47.37% (95% CI 33.98–61.03) (Table III). These results indicate the success of our strategy in which we have used mAbs, in the developed NS1 ELISA, that bind outside the wing domain and likely avoid the masking effect due to the Abs from primary infection. Our observation of poor sensitivity of Platelia ELISA in the secondary DENV infection is consistent with the two meta-analysis studies (13, 18) reporting that the sensitivity of Platelia NS1 Ag ELISA reduced from 60% (95% CI 46–75) to 42% (95% CI 34–50) and from 88% (95% CI 85.8–89.9) to 60.8% (95% CI 57.8–63.8) during secondary infection.

In the clinically suspected but RT-PCR–negative samples (n = 45), the developed ELISA detected NS1 in 42.22% (95% CI 27.66–57.85) samples, whereas Platelia ELISA detected NS1 in only 17.78% (95% CI 8.00–32.05) samples. The NS1 detection rate in this group is lower than in the RT-PCR–positive group. This difference could be due to the following reasons: 1) the samples negative for NS1 may be truly DENV-negative, as these samples were also negative in RT-PCR. The presence of anti-DENV Abs cannot be considered a definitive diagnosis for dengue, and the fever may also be due to some unknown infection. 2) NS1 may have been cleared from circulation due to the late stage of sampling, as the Abs have already appeared.

The higher sensitivity of the developed NS1 ELISA indicates the success of our approach in the selection of high-affinity mAbs. Even for panel 2, with samples from clinically suspected dengue with positive serology, our ELISA was found to have higher sensitivity than the commercial ELISA (Table II, Supplemental Fig. 3).

As we have the information on the infecting DENV serotype in the RT-PCR–positive samples (panel 1), we looked for any serotype-specific bias in the developed NS1 ELISA. This is very important as multiple studies have shown that the commercial NS1 assays do not perform equally for four DENV serotypes (16, 18, 19, 23, 25, 27). No serotype-level bias was observed in the developed NS1 ELISA (Table IV). However, we found poor sensitivity for DENV-2 in commercial Platelia NS1 ELISA (Table IV). This is consistent with the poor analytical sensitivity of Platelia ELISA for DENV-2 observed by us (Fig. 4) and the published literature (16, 18, 61). The unbiased performance of the developed NS1 ELISA for all four DENV serotypes results from the systematic selection of mAbs for our test.

As the clinical signs of dengue are indistinguishable from other febrile illnesses, it is crucial to determine the specificity of an NS1 ELISA in the samples from febrile subjects who are negative for recent DENV infection. Our developed ELISA showed 100% (95% CI 98.93–100.00) specificity with challenging panels containing serum samples from scrub typhus fever (panel 3; n = 153), other non–dengue fever (panel 4; n = 148), and viral hepatitis (panel 5; n = 41) patients (Table II, Supplemental Fig. 3). In south and southeast Asia, dengue and scrub typhus are the leading causes of acute febrile illness (62–64), and that is why it was important for us to include the samples from scrub typhus–positive febrile individuals and patients with other causes of fever.

This study is unique as we have not only created a very large anti-DENV NS1 mAb repertoire and characterized the mAbs in great detail but also made the ELISA kit that is stable at RT for a reasonable duration and has an assay calibrator to remove the variation between the assay runs. These features are often missing in the experimental assays reported in the literature. Moreover, we have used well-characterized clinical samples for clinical evaluation of the developed NS1 ELISA and compared the performance with RT-PCR as well as with a very well-established commercial NS1 ELISA. We have also looked at the performance of NS1 ELISA in primary and secondary dengue as well as the differences at the serotype level. Furthermore, we have used large clinical panels from subjects with other febrile illnesses for specificity analysis. Finally, we have also tested cross-reactivity potential with NS1 from ZIKV, YFV, JEV, WNV, and TBEV. Our work also has some limitations. For example, despite no cross-reactivity with the secretory recombinant ZIKV NS1 and native ZIKV NS1, cross-reactivity cannot be completely ruled out as we have not run ZIKV-infected patient samples in the developed NS1 ELISA. In addition, the subgroup analysis for individual serotypes has a limited number of samples (12 to 26 for each serotype).

Conclusively, we have generated a large repertoire of anti-DENV NS1 mAbs with wide specificities and developed a highly specific and sensitive NS1 ELISA with improved performance for DENV NS1 detection in primary and secondary DENV infection without any serotype bias.