Proof of working of LAMP
Fluorescent LAMP in a tube
LAMP was performed with a primary LAMP protocol for N-gene cDNA detection. Twenty-five microliters of LAMP reaction volume constituted of the ingredients mentioned in Electronic Supplementary Material Table S6. We used different primer concentrations: F3 and B3 (0.6 μM), LF and LB (0.4 μM), and FIP and BIP (1.6 μM). Bst 3.0 polymerase was added to the reaction mixture at a final concentration of 0.32 U/μL. LAMP program was a single cycle of 40 min (65 °C for 30 min, 80 °C for 5 min, and 4 °C for at least 5 min). After LAMP amplification, SYBR-Green was added to the product to look for fluorescence or colorimetric signals. A bright green-yellow coloration was observed for the DNA sample and a dull orange coloration for the non-template control (NTC). The result was also visible with the naked eye (Fig. 2A).
LAMP readout via LFA
LAMP was modified to be read on the lateral flow test strips. Initially, FITC-labeled LF primer and biotin-labeled LB primer were used. LAMP was performed as described in Electronic Supplementary Material S3. LFA was performed as described in “Conclusion and summary.” The test strip readout indicated a clear positive result for the cDNA sample by producing two lines on the LFA test strip, while the NTC produced only one line on the LFA test strip (Fig. 2C(a, b)). This confirmed the compatibility of LAMP and LFA. However, replication experiments showed that the reproducibility was low due to unspecific amplification. As also reported previously, NTC is known to show unspecific amplification , so modifications in the LAMP reaction were made.
Bst 2.0 polymerase vs Bst 3.0 polymerase
WarmStart (with Bst 2.0 polymerase) RT-LAMP master mix (New England, Biolabs) and Bst 3.0 polymerase were tested in parallel to check the difference in the two enzyme activities. We observed that the master mix was less robust and produced more unspecific amplification as compared to Bst 3.0 polymerase. The combined efficacy and robustness of Bst 3.0 polymerase was the reason why it was chosen for all further experiments. The reaction mixture and protocol were modified accordingly.
Incorporating biotin-11-dUTPs to increase specificity of LAMP
For a reproducible LAMP-LFA readout, it is necessary to efficiently incorporate FITC and biotin labels into the amplification products. Firstly, we tested a combination of FITC-dUTP with biotin-LB primer but observed that FITC-dUTP is not incorporated during amplification using Bst 3.0 polymerase, for a 10-min LAMP annealing time. This likely is due to the FITC label hampering the activity of Bst 3.0 or FITC-labeled dUTP takes a longer time to incorporate. Consequently, we tested a combination of biotin-dUTP and FITC-LF primer. Biotin-dUTP was investigated for 1%, 5%, 10%, and 20% of dNTP volume in the 25 μL reaction, for 10-min LAMP annealing time. We assessed that 5% B-dUTPs gave us the best test band on the LFA (Fig. 2C(c, d)). The NTC gave a clear negative control, and hence, the reaction mixture was established for further experiments. In the abovementioned work by Tan et al., the incorporation of FITC-labeled dUTP during LAMP amplification in 30 min was reported ; however, we were not able to reproduce these results with Bst 3.0 polymerase in 10 min.
The LFA test strips used are standardized and stabilized for use at room temperature. For the LAMP product readout compatibility with LFA, the LF primer was tagged with the FITC label, generating a FITC-flanked region in the amplified LAMP product. Biotin-dUTPs were incorporated during elongation and generated biotin flanking regions in the amplified LAMP product. FITC in the amplified product was captured by GNP-anti-FITC-Ab conjugate at the sample loading area on the test strip. The agglomeration of GNPs on the test band and control band aided in visual readout on the LFA (Fig. 2C(c, d)). This result was obtained reproducibly and confirmed with the gel electrophoresis results (Fig. 2B).
Enzyme mixture for increasing robustness and specificity of LAMP
Since Bst 3.0 already has reverse transcriptase activity, RT-LAMP may be performed with only a single enzyme. Although the enzyme is self-sufficient, we added reverse transcriptase (RTase, 1 μL) to enhance reverse transcriptase activity, to increase the robustness of RNA amplification via RT-LAMP. Also, the non-specificity of LAMP needed to be reduced. It was noted that the primers in LAMP possibly dimerize and amplify, therefore generating unspecific NTC or background signals. The generation of false positives was prevented by using helicase (1 μL) in the LAMP enzyme mixture. Helicase is an ATP-dependent enzyme; hence, 1 μL of ATP (10 mM) was added to the LAMP reaction mixture. The addition of RTase decreases RNA amplification by 5 min, which brought the overall RT-LAMP reaction time to 15 min. Besides reducing the amplification time, the enzyme mixture did not cause any adverse effect on the readout of LFA.
Establishing LAMP-LFA with viral cDNA
Standardization of the LAMP protocol
To establish the LAMP-LFA detection method, N-gene cDNA was used as template for LAMP amplification. The target in LAMP was a 200-bp section out of the 466-bp N-gene. LAMP was performed as described in “LAMP reaction mixture” and “LAMP program.” LAMP performance was analyzed at various amplification annealing time intervals, starting with 5 min up to 25 min (in steps of 5 min). For each sample, a LFA was tested along with confirmation of amplification with 2% agarose-gel electrophoresis. Different primer concentrations for each set of primers were also analyzed and then fixed for all future experiments, as described in Electronic Supplementary Material S3. The report by Tan et al. reports a working LAMP using a reduced 4 primer set (F3-B3 and FIP-BIP) . We also tried to use a primer-reduced set, but the system does not perform reliably with our LAMP system. The 6-primer system worked specifically for N-gene target in our work (Electronic Supplementary Material S2).
LFA readout for cDNA LAMP
cDNA detection was analyzed with two distinct tests and control lines on the LFA test strip. The test line is formed only when there is amplification of the template cDNA, and the control line is always seen as the reference to the validity of test strips. Non-template controls (NTCs) were observed to produce only one control line on the test strip, implying no amplification (Fig. 3A). Each experiment was performed with NTC for verifying the specificity of the cDNA reactions. A LAMP amplification time of 10 min was sufficient for an unambiguous LFA result for cDNA as template and was therefore fixed for consecutive experiments.
Statistical inference of cDNA-LAMP-LFA
To determine reproducibility of the assay, LAMP experiments were carried out for a total of 161 samples, out of which 28 were NTC and 133 were cDNA. The assay was 95.49% sensitive (Clopper-Pearson 95% CI of 90.44 to 98.33%) and 96.43% specific (with Clopper-Pearson 95% CI of 81.65 to 99.91%) (Fig. 3B). By using the McNemar statistical test for a paired nominal dataset, we confirmed statistically that the sensitivity of LAMP-LFA results corresponds significantly to the specificity of the assay (p-value = 0.064). Therefore, we determined that our LAMP protocol could be applied for viral RNA detection.
Proof of viral RNA detection via RT-LAMP-LFA
Target specificity confirmation
Eighty clinical RNA samples from patients that tested positive for SARS-CoV-2 were obtained with different CT values (confirmed by real-time qRT-PCR). The extracted RNA was obtained directly from human swab samples, so it contained a mixture of viral and human RNA. A human RNA control from 293 T HEK cells was tested with the LAMP primers to confirm no unspecific amplification of human RNA. The human RNA from 293 T HEK cells was used as negative controls. The RNA samples (obtained from RKI) were eluted in molecular-grade water for preservation and further experimental use. The elution water and some other elution buffers were tested as negative controls too.
Statistical inference of RT-LAMP-LFA
A wide range of CT values was tested, corresponding to 5.6 × 106 RNA copies/ml (CT 22) and 3.9 × 103 RNA copies/ml (CT 33) (Fig. 4A). RT-LAMP was performed for each sample in duplicates and for some in triplicates. It was observed that RNA was detected with RT-LAMP-LFA for all the RNA CT values tested, with 77.27% sensitivity (Clopper-Pearson 95% CI of 69.17 to 84.11%) and 97.30% specificity (Clopper-Pearson 95% CI of 85.84 to 99.93%) (Fig. 4C).
RT-LAMP-LFA was performed in a triplicate for CT-33 (3.9 × 103 RNA copies/ml) and the system could detect the CT-33 RNA with 100% sensitivity (Clopper-Pearson 95% confidence interval of 29.24 to 100.00%) and 100% specificity (Clopper-Pearson 95% confidence interval of 2.50 to 100.00%).
Variants of concern investigated
The RNA samples covered different variants of SARS-CoV-2, mainly wild type and VoCs alpha and delta. RT-LAMP could amplify and LFA could detect each of these variants, irrespective of the mutations, using the N-gene specific LAMP primers (Fig. 4B). The efficiency of the RT-LAMP-LFA was comparable with that of real-time qRT-PCR but will be improved with further experimentation.
Semiquantitative digital analysis of the LFA test strip
The quantification of LFA test and control lines was performed by using a smartphone-based in vitro diagnostics (IVD) device. The device has two main parts: (a) smartphone connected to a (b) test strip holder. The test strip is placed in the test strip slot and the result is visualized and saved on the smartphone. The results of the IVD device are presented in Table 1. Each row presents the analysis of one LFA. The readout from the device is relative intensity values. The relativity measure is to the test strip non-colored space next to the test and control lines.