Reverse transcription recombinase polymerase amplification-lateral flow assay for detection of pathogenic orthoflaviviruses in mosquito vectors

Ethics statement

The use of mosquito specimens in this study was approved by the Faculty of Science, Mahidol University Animal Care and Use Committee (SCMU-ACUC) under protocol number MUSC66-020-650. All experiments were conducted at the Center of Excellence for Vectors and Vector-Borne Diseases (CVVD), Faculty of Science, Mahidol University.

Recombinase polymerase amplification (RPA) primer design

Viral genome sequences of the genus Orthoflavivirus were retrieved from the International Committee on Taxonomy of Viruses Database (ICTV, https://ictv.global) and the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov). DNA sequences were obtained between June 2023 and June 2024. Detailed information of each sequence, including sample collection dates, is provided in Table S4. Sequence alignments were performed using Geneious Prime software (GraphPad Software LLC, Auckland, New Zealand) with the Clustal Omega algorithm under the “Group sequences by similarity” mode. Assignment of human pathogenic and non-pathogenic sequences was manually performed by reviewing relevant literature. Primers broadly targeting Orthoflavivirus were obtained from previous PCR studies (Table S3; (Bonnet, van Jaarsveldt & Burt, 2022; Bronzoni et al., 2005; Daidoji et al., 2021; Fulop et al., 1993; Grubaugh et al., 2013; Johnson et al., 2010; Kuno, 1998; Maher-Sturgess et al., 2008; Myhrvold et al., 2018; Patel et al., 2013; Pierre, Drouet & Deubel, 1994; Rice et al., 1985; Scaramozzino et al., 2001b; Tanaka, 1993; Vina-Rodriguez et al., 2017; Xue et al., 2021)) and assessed for binding ability to pathogenic Orthoflavivirus species, but not to non-pathogenic ones. Primers demonstrating the potential to bind across multiple pathogenic Orthoflavivirus species were shortlisted and served as the basis for designing new primers. These primers were further refined following the RPA design criteria outlined in the TwistAmp^® DNA Amplification Kits Assay Design Manual (part number: INASDM Revision 1, https://www.twistdx.co.uk). The final primer sequences were: Flavipath-F (5′-AAR GGH AGY MGN GCH ATH TGG TWY ATG TGG-3′) and Flavipath-R (5′- CCT TCH ACW CCD CYB HVD GAR TTY TYH CKV-3′). Primers were synthesized by Macrogen (Seoul, Republic of Korea) using phosphoramidite chemistry at a 50 nmole scale and purified by MOPC (cartridge purification). Probes were synthesized by Synbio Technologies (USA) at 100 nmole and 2 OD yield, with HPLC purification. For the dipstick test, primers were labeled with Digoxigenin (DIG) and fluorescein isothiocyanate (FITC) corresponding to the antibody binding sites of the Amplicon Detection 2T1C Dipstick (K-BIOSCIENCES, Pathumthani, Thailand). Detailed primer and probe sequences are provided in Table S1.

DNA and RNA sample preparation

Synthetic DNA targets were synthesized using phosphoramidite chemistry and ordered from Macrogen (Seoul, Republic of Korea) either as single-stranded oligonucleotides or as double-stranded oligo duplexes. All were prepared at a 50 nmole scale and purified by MOPC (Table S2). RNA targets were synthesized using the MAXIscript™ T7 Transcription kit (Invitrogen, Waltham, MA, USA), treated with TURBO DNase1 and 0.5 M EDTA, and precipitated with ammonium acetate/ethanol according to the manufacturer’s protocol. cDNA templates were synthesized using the SuperScript™ III One-Step RT-PCR System with Platinum™ Taq DNA Polymerase (Invitrogen), utilizing modified primers T7-DENV2 NS5-F and DENV-2 NS5-R (Patramool et al., 2013; Richardson et al., 2006) under the following thermal conditions: 55 °C for 30 min, 94 °C for 2 min, followed by 40 cycles of 94 °C for 15 s, 60 °C for 15 s, and 72 °C for 1 min.

RNA samples were extracted from the whole bodies of Aedes aegypti mosquitoes infected with dengue virus serotype 2 (Thai strain ThNR02/772) or Zika virus (Brazilian strain BE H 815744) without dissection. Viruses were propagated and mosquito infections performed according to Morales-Vargas et al. (2020) and Pompon et al. (2017). Mosquitoes were stored for approximately three years at −80 °C freezer before processing (n = 3 for each virus). Individual mosquito was homogenized in 1xPBS for 100 µL using a TissueLyser LT (Qiagen, Hilden, Germany) or pellet pestles, followed by centrifugation at 12,000 rpm for 5 min at 4 °C. The supernatant was mixed with one mL of Trizol LS Reagent (Invitrogen, USA), and RNA was extracted according to the manufacturer’s instructions, including the optional centrifugation step for high-fat samples. RNA pellets were resuspended in 25 µL DEPC-treated water (Invitrogen) and incubated at 55 °C for 10 min. Samples were stored at −80 °C until use. All molecular works was performed on clean benches decontaminated with 70% EtOH and RNase AWAY (Thermo Fisher Scientific, Waltham, MA, USA).

Recombinase Polymerase Amplification (RPA) reaction

The RPA reaction was performed using the TwistAmp basic kit (TwistDx™ Limited, UK) following the manufacturer’s instructions with modifications. Modifications included splitting the master mix for multiple reactions and adding a termination step at 65 °C for 10 min (Kumar et al., 2022; Patchsung et al., 2020). The master mix consisted of: 29.5 µL rehydration buffer, 4.0 µL Nuclease-Free water (Invitrogen), 2.5 µL Flavipath-F probe (10 µM), and 2.5 µL Flavipath-R probe (10 µM). Four 9.6 µL aliquots were prepared: each mixed with 1 µL of DNA template or nuclease-free water (for the non-template control, NTC). MgAcO solution (0.29 µL) was added immediately before incubation. Reactions were incubated at 37 °C for 30 min or different temperatures as specified, with manual mixing after 4 min, and then terminated at 65 °C for 10 min. Detection was performed using a dipstick (KB-AMP-2T1C-S; K-BIOSCIENCES) and three drops (approx. vol. 100 µL) of DNA running buffer (K-BIOSCIENCES), and developed at room temperature for 15 min. A visible control line indicated successful reaction conditions. Some samples were also analyzed via 2.0% (w/v) agarose gel electrophoresis stained with Red safe (Vivantis, Selangor, Malaysia) and in Tris-borate-EDTA buffer (TBE). All the reactions were performed in duplicate or triplicate sets.

The reverse transcriptase recombinase polymerase amplification (RT-RPA) protocol was modified from Bonnet, van Jaarsveldt & Burt (2022). The reaction mixture contained: 29.5 µL rehydration buffer, 6.8 µL Nuclease-Free water, 2.1 µL FlaviPath-F probe (10 µM), 2.1 µL FlaviPath-R Probe (10 µM), 1 µL M-MLV Reverse Transcriptase (Invitrogen™, USA), and 1 µL RNase OUT™ (Invitrogen). The reaction mixture was divided into four aliquots of 10.6 µL each. After addition of 1 µL RNA template or 1 µL NF for NTC control and 0.29 µL MgAcO solution, reactions were incubated at 37 °C for 20 min and terminated at 65 °C for 10 min. Products were visualized using the dipstick or gel electrophoresis.

Dipstick results were scanned using a Canon MF3010 Printer and Scanner, and analyzed with ImageJ software (v1.54 g, NIH, USA). The mean gray intensity of the test and control bands were normalized by subtracting background intensity (averaged of upper and lower positions), and the T/C ratio was calculated (Ahmed et al., 2022). A sample was considered positive if its T/C ratio was more than three times that of the NTC (cutoff value). Slight negative values from low signals were set to zero. Sensitivity was calculated as: true positives/(true positives + false negatives), and specificity as: true negatives/(true negatives + false positives). The RT-RPA-LFD assay was performed on mosquito RNA samples (30–100 ng/µL) or diluted dengue RNA samples (10⁰–10⁴ copies) to determine the detection limit.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)

RT-qPCR was performed using a Stratagene Mxpro3000P system (Agilent Technologies, USA) with the QuantiNova SYBR Green RT-PCR Kit (Qiagen, Hilden, Germany). The 10 µL reaction included: 5 µL 2x Master Mix, 2.9 µL RNase-Free Water, 0.1 µL 100x RT Mix, 0.5 µL 10 µM forward primer, 0.5 µL 10 µM reverse primer, and 1 µL RNA Template. Primers DENV-2 NS5-F and DENV-2 NS5-R (Richardson et al., 2006) were used for dengue virus quantification; Zika virus quantification used primers ZIKF and ZIKR (Han et al., 2018) (Table S1). Cycling conditions included reverse transcription at 50 °C for 30 min, denaturation at 94 °C for 2 min, followed by 40 cycles of 94 °C for 15 s, 60 °C for 1 min, and 68 °C for 1 min, with a subsequent melt-peak analysis. To compare the detection efficiency of the qPCR and RPA-LFD assays, the same thawed RNA aliquot was used for both tests and analyzed on the same day without blinding. Results were accepted if qPCR efficiency ranged from 80–120%, with correlation coefficients (r²) between 0.9–1. The limit of quantification (LOQ) was defined as the lowest concentration within the linear dynamic range. No-template controls (NTC) were included in all analyses. A Cq of NTC greater than 40 (no Cq) or no band was observed at the target location in gel electrophoresis was considered free from target contamination.

General PCR conditions used final volumes of 10–25 µL, with products visualized on 2.0% (w/v) agarose gels. The amplifications were performed according to the following parameters: 1 cycle of 2 min at 95 °C, 35 cycles of 30 s at 95 °C, 1 min at 55 °C, and 1 min at 72 °C, followed by 1 cycle of 5 min at 72 °C. DNA concentrations were measured using a NanoDrop One Microvolume UV-Vis Spectrophotometer (Thermo Fisher Scientific).

Statistical analysis

Statistical analysis was performed using R version 4.3.3 (2024-02-29) on platform: aarch64-apple-darwin20 (64-bit) (R Core Team, 2024). Data normality was tested using Shapiro–Wilk test. Non-parametric data were analyzed using the Kruskal-Wallis test with post hoc comparisons using the ‘dunn.test’ and FSA packages (Dinno, 2024; Ogle et al., 2023). Student’s t-test was applied for comparisons between two datasets. A p-value of less than 0.05 was considered statistically significant.

RPA primer design for pathogenic orthoflavivirus detection in mosquitoes

Primers targeting various orthoflaviviruses reported in previous studies were compiled (Table S3). Among the commonly targeted genes—Nonstructural Protein 5 (NS5), Nonstructural Protein 3 (NS3), and the Envelope protein (E) genes—NS5 was the most frequently used. As NS5 was the largest non-structural protein and conserved across different flavivirus species, it is a favorable target for molecular assays designed to detect members of this genus. Therefore, this study focused on broad-specificity primers targeting NS5.

Fourteen primer pairs from the literature were evaluated (Table S3). A set of 61 representative orthoflavivirus genome sequences—including 45 vertebrate-pathogenic and 16 non-vertebrate-pathogenic species—was retrieved from the ICTV and NCBI databases (Table S4). The binding potential of the 14 primer pairs to pathogenic and non-pathogenic orthoflaviviruses was compared (Table 1). Primer pairs that bound to >70% of pathogenic viruses were selected which included F8276d-F (Flav100F), pathogens 77.8% and non-pathogens 6.3%: F9063d-R (Flav200R), pathogens 100% and non-pathogens 87.5%: VD8, pathogens 84.4% and non-pathogens 12.5%: XF-F1, pathogens 88.9% and non-pathogens 25.0%: XF-F2, pathogens 75.6% and non-pathogens 37.5%: XF-R, pathogens 75.6% and non-pathogens 25.0%: YF1, pathogens 84.4% and non-pathogens 12.5%. However, as shown in Table S4, the genome sequences used for primer design in this study were derived from virus isolates collected between 1927 and 2016. Given the high mutation rates of RNA viruses, the effectiveness of the primers may be reduced against more recently emerging strains.

Although F9063d-R (Flav200R) showed complete binding to the selected pathogenic group, its high cross-reactivity (87.5%) with non-pathogens led to its exclusion from further consideration. Considering the optimal amplicon length for the TwistAmp^® RPA kit (80–500 bp) (TwistDx Limited, 2023), only three primer pairs—XF-F1/XF-R (264 bp), XF-F2/XF-R (215 bp), and YF1/XF-R (326 bp)—met this requirement.

RPA primer design criteria included a length of 30–36 bp, GC content of 20–70%, melting temperature (Tm) of 50–100 °C, and fewer than five mononucleotide repeats. Among the forward primers (Table S5), YF1 had a relatively low Tm (50.3 °C), and XF-F1 had higher base repetition (4) and lower GC content (39.9%) than XF-F2 which could affect the binding affinity with nucleotide targets. Thus, XF-F2 and XF-R were selected for further development.

To optimize the assay in a cost-effective manner, a synthesized double-stranded DNA oligonucleotide was used as the RPA template instead of RNA or in vitro-transcribed RNA. Given the 130 bp maximum length for custom oligonucleotide synthesis and the ideal RPA amplicon range of 100–200 bp (TwistDx Limited, 2023), a new primer was designed to complement XF-F2. XF-F2 had a better GC content (50.8%) than XF-R (60.3%) and was selected for modification. The newly designed forward primer, FlaviPath-F, was derived from XF-F2 by incorporating degenerative bases into a highly similar region across several sequences in order to broaden its detection scope in pathogenic orthoflaviviruses, and then the primer length extended from 22 to 30 bp. The reverse primer, FlaviPath-R, was designed approximately 130 bp downstream from the FlaviPath-F, targeting conserved regions among major pathogenic orthoflaviviruses.

Binding ability of the FlaviPath primers is shown in Fig. 1 and Table 1. FlaviPath-F matched 93.3% of pathogenic and 31.3% of non-pathogenic orthoflaviviruses. FlaviPath-R matched 40.0% of pathogens and none of the non-pathogens. While XF-R had higher coverage, FlaviPath-R was more specific to key mosquito-borne pathogens, including DENV, ZIKV, YFV, JEV, WNV, Wesselsbron virus (WSLV), and Murray Valley encephalitis virus (MVEV).

Establishment of RPA-LFD detection method

The RPA-LFD assay (Figs. 2A1–2A2) was first tested using synthetic single- and double-stranded DNA oligonucleotides of dengue virus serotype 2 from Thailand (Den2Th, Table S2, Fig. S1), following the manufacturer’s protocol (39 °C for 20 min). Clear test lines were observed for double-stranded DNA samples (normalized intensity: 0.20–1.1, Table S6), whereas only weak signals were seen with single-stranded templates (normalized intensity 0.02−0.08). The difference from the no-template control (NTC) was statistically significant for double-stranded DNA (p = 0.01). Therefore, double-stranded oligonucleotides were used for subsequent testing.

To optimize reaction time, the RPA-LFD assay was performed for 5, 10, 15, 20, and 25 min with the Den2Th double-stranded DNA oligonucleotide (Figs. 2B1–2B2). Amplification was observed at all time points (averaged normalized intensity 0.06–0.13), with the best signal at 20 min. Though the differences were not statistically significant across time points (except compared to NTC, p = 0.04), 20 min was chosen for future reactions.

The effect of incubation temperature was also evaluated (Figs. 2C1–2C2). Positive results were seen across all tested temperatures between 20 and 37 °C (normalized intensity 0.09–0.44). Statistically significant differences from the NTC were observed at all tested temperatures within the range (p ≤ 0.04). The strongest signal (average normalized intensity: 0.30) occurred at 37 °C, which was selected as the optimal reaction temperature.

Cross-reactivity evaluation

To assess cross-reactivity, synthetic double-stranded DNA oligonucleotides representing DENV, ZIKV, JEV, WNV, and three insect-only orthoflaviviruses (AEFV, NAKV, and LAMV) were synthesized (Table S2) and tested in the RPA-LFD assay (Figs. 3A1–3A2). Positive test results were observed for DENV, JEV, ZIKV, and WNV (normalized intensity 0.18–0.69), while AEFV, NAKV, and LAMV showed no signal above the cutoff threshold of 0.03 (normalized intensity 0.00–0.02). These results matched primer predictions (Table 1).

RT-RPA was then conducted on RNA extracted DENV serotype 2- and ZIKV-infected mosquitoes (Figs. 3B1–3B2). Infection status was confirmed by RT-qPCR (Figs. S2 and S3). All virus-infected mosquito RNA samples (A–D) produced a clear RT-RPA signals using FlaviPath-F and FlaviPath-R (DENV: 0.23–0.60, ZIKV: 0.12–1.11). In contrast, uninfected laboratory colony samples (Neg 1 and 2) and NTCs showed negligible signal (≤ 0.01). These results confirmed the effectiveness of the RT-RPA-LFD assay in distinguishing infected from uninfected mosquito samples.

Detection limit determination

The detection limit of the RPA-LFD test for orthoflaviviruses was evaluated using a serial dilution technique with the RNA extracted from dengue-infected mosquito samples. These samples (A, B, and C) contained dengue virus at densities of 10⁵, 10⁴, and 10⁶ RNA copies, respectively (Figs. 4A and 4B), as quantified by RT-qPCR using an absolute quantification method with in vitro transcribed dengue serotype 2 RNA as a standard (Fig. S2). Sample A was serially diluted tenfold down to a single gene copy.

Positive RT-RPA-LFD results were obtained across all tested dilutions (10¹–10⁴ RNA copies), though with varying detection rates (60–100%). Mosquito samples B and C also yielded positive RPA results in the RT-RPA-LFD assay (Figs. 4A and 4B). In comparison, RT-qPCR consistently detected dengue virus only at concentrations above 10² RNA copies (Figs. 4C and 4D). While the RT-RPA-LFD achieved 88.6% sensitivity (5 false negatives out of 44 samples, Figs. 3–4) and 89% specificity (2 false positives out of 24 samples, Figs. 2–4) in the range of 10¹–10⁴ RNA copies, RT-qPCR demonstrated 100% sensitivity and specificity within the tested range of 10²–10⁴ RNA copies. The agreement between RT-RPA-LFD and RT-qPCR, measured by Cohen’s Kappa, was 0.6 (standard error: 0.17), indicating substantial agreement (Landis & Koch, 1977).

Although RT-RPA-LFD demonstrated lower reliability compared to RT-qPCR analysis, it provided faster processing and greater practical utility, particularly in field settings. These findings suggest that RT-RPA-LFD is suitable for detecting low viral loads or pooled samples, however, caution is advised due to the risk of false negative results at concentrations below 10⁴ RNA copies. The estimated limit of detection for the RT-RPA-LFD assay was 10⁴ RNA copies, as not all positive mosquito samples were consistently detected below this threshold.