A lyophilized open-source RT-LAMP assay for molecular diagnostics in resource-limited settings

Abstract

A critical bottleneck for equitable access to population-scale molecular diagnostics is the limited availability of rapid, inexpensive point-of-care tests, especially in low- and middle-income countries. Here, we developed an open-source reverse transcription loop-mediated isothermal amplification (RT-LAMP) molecular assay for pathogen detection. It is based on nonproprietary enzymes, namely, HIV-1 reverse transcriptase, Bst LF DNA polymerase, and UDG BMTU thermolabile uracil-DNA glycosylase. Formulated as liquid or lyophilized reaction mixtures, these reagents enable sensitive colorimetric detection of respiratory samples without the need for prior nucleic acid isolation. We evaluated our lyophilized RT-LAMP assay on clinical samples with suspected COVID-19 infection, demonstrating high sensitivity and 100% specificity compared with the gold-standard RT-qPCR. Reaction performance was unaffected by prolonged storage of lyophilized reagents at ambient or elevated temperatures. As a proof of concept, we evaluated the robustness and ease of use of lyophilized RT-LAMP reaction mixes through independent laboratory testing of COVID-19 samples in Ghana. Overall, our open-source RT-LAMP assay provides a flexible and scalable point-of-care test that can be adapted for rapid detection of various pathogens in resource-limited settings.

Figure S1.. Purification of open-source enzymes (related to Fig 1).

(A) Chromatograph of the ion exchange purification step of Bst LF. Ion exchange chromatography (IEC) was performed on Resource Q 6-ml ion exchange column. Absorbance measured at 280 nm in milliarbitrary units (mAU) is plotted against elution volume in milliliters. The dark blue box corresponds to the Bst LF protein fractions that were pooled for downstream assays and is visualized in the box of the same color in the SDS-PAGE gel image in (C). (B) Chromatograph of the size-exclusion purification step of HIV-1 RT. Size-exclusion chromatography (SEC) was performed on the HiLoad 16/600 Superdex 200 pg column. Absorbance measured at 280 nm in milliarbitrary units (mAU) is plotted against elution volume in milliliters. The dark red box corresponds to the protein fractions that were pooled for use in downstream assays and is visualized in the box of the same color in SDS–PAGE gel image in (D). (C) Coomassie-stained SDS–PAGE image of HIV-1 RT and Bst LF purification fractions. SDS–PAGE was run on the fractions of multistep purifications of HIV-1 RT and Bst LF. Dark blue box corresponds to the Bst LF protein fractions that were pooled from (A) and used in downstream assays. (D) Coomassie-stained SDS–PAGE image of HIV-1 RT size-exclusion chromatography purification fractions. Protein fractions collected during size-exclusion chromatography purification of HIV-1 RT were loaded onto SDS–PAGE. Note the heterodimeric nature of HIV-1 RT. Dark red box corresponds to the HIV-RT protein fractions that were pooled from (B) and used in downstream assays. (E) Chromatograph of the ion exchange purification step of BMTU UDG. Ion exchange chromatography was performed on the Resource Q 6-ml ion exchange column. Absorbance measured at 280 nm in milliarbitrary units (mAU) is plotted against elution volume in milliliters. (F) SDS–PAGE image of BMTU UDG purification fractions. Pooled protein–containing fractions from Ni-NTA and ion exchange chromatography are marked as “His-Trap” and “ResQ IonEx,” respectively. The final BMTU UDG lane contains the purified, concentrated protein after dialysis. (E) Dark green box corresponds to the BMTU UDG protein fractions that were pooled from (E) and used in downstream assays. (G) SDS–PAGE image of in-house–purified open-access RT-LAMP proteins. HIV-1 RT, Bst LF, and BMTU UDG enzymes were purified as described in this study. Approximately 1.25 μg of total protein was loaded per lane.

Figure 1.. Assessing the performance of open-source RT-LAMP enzymes.

(A) Comparison of DNA polymerase enzymes in RT-LAMP reactions. Detection rates of RT-LAMP reactions containing different DNA polymerases in addition to WarmStart RTx reverse transcriptase (NEB) and various concentrations of synthetic SARS-CoV-2 RNA. Results were obtained from four replicates per condition. (B) Comparison of reverse transcriptase enzymes in RT-LAMP reactions. Analogous to (A) but for different reverse transcriptase enzymes in combination with in-house Bst LF DNA polymerase. M-MuLV, Moloney murine leukemia virus reverse transcriptase; AMV, avian myeloblastosis virus reverse transcriptase; HIV-1, human immunodeficiency virus 1 reverse transcriptase. (C) Cross-contamination prevention in RT-LAMP via uracil-DNA glycosylase (UDG) enzymes. Diluted amounts of contaminating amplicons were added to RT-LAMP reactions containing different uracil-DNA glycosylase (UDG) enzymes and synthetic SARS-CoV-2 template. No UDG enzyme and nontemplate control reactions were included. Shown are time-to-threshold values from real-time fluorescence RT-LAMP reactions performed in duplicates. In-house BMTU UDG enzyme was tested against Antarctic Thermolabile UDG (NEB). (D) Thermostability test of BMTU UDG and commercial UDG enzymes. In-house BMTU UDG and commercial Antarctic Thermolabile UDG (NEB) were preincubated at different temperatures for 5 min. UDG enzyme was then added to qPCRs targeting either dTTP-containing DNA template or dUTP-containing DNA template. Cycle-to-threshold values (Ct) from duplicate reactions are shown for each condition. Crossed box = not determined. (E) Limit of detection of open-access and commercial RT-LAMP reactions. Reactions prepared from in-house Bst LF, HIV-1 RT, and BMTU UDG enzymes were compared with commercial reactions using 2X WarmStart LAMP Kit (NEB) containing engineered proprietary enzymes. Synthetic SARS-CoV-2 RNA at defined copy numbers was used as a template, and 20 replicates were performed per condition. Time-to-threshold values from real-time fluorescence RT-LAMP reactions are shown. (F) Specificity test of RT-LAMP reactions detecting different pathogens. RT-LAMP reactions were assembled with Bst LF and HIV-1 RT. Primers targeting SARS-CoV-2, Influenza A, or hRNase P were tested against SARS-CoV-2 RNA, Influenza A RNA (each individually spiked into HEK293-extracted RNA), HEK293-extracted RNA alone, and nuclease-free water as a no-template control. Reactions were performed in four replicates, and end-point relative fluorescent units are displayed. Source data are available for this figure.

Figure S2.. Benchmarking of open-source RT-LAMP reagents (related to Fig 1).

(A) Time to threshold for DNA polymerase enzyme comparison in RT-LAMP. Time-to-threshold data for the experiment shown in Fig 1A. Four replicates per condition were performed. (B) Time to threshold for reverse transcriptase enzyme comparison in RT-LAMP. Time-to-threshold data for the experiment shown in Fig 1B. Four replicates per condition were performed. (C) Time to threshold for sensitivity test of RT-LAMP reactions with UDG enzymes. Reactions were prepared with Bst LF and HIV-1 RT. In-house BMTU UDG, Antarctic thermolabile UDG, or no UDG enzymes were included. Four replicates per condition were performed. (D) Schematic representation of bead-LAMP workflow. Bead-LAMP was performed according to the protocol published previously (Kellner et al, 2022). In brief, 100 μl of sample (TCEP/betaine/proteinase K-inactivated sample) was mixed with 60 μl of magnetic bead slurry and left to incubate for 5 min to facilitate binding of nucleic acids to magnetic beads. The beads were separated on a magnet for 5 min or until the solution turned completely clear. The supernatant was discarded, and 200 μl of 85% ethanol was added. After 30 s, the ethanol was removed, and beads were left to dry out in the tube for a period of 3 min. RT-LAMP reaction mix was added directly on the top of the dried beads, and beads were resuspended in the reaction mix. The reaction vessel was capped/sealed and incubated for 35 min at 63°C. (E) Bead-LAMP increases sensitivity of open-source RT-LAMP reactions. Samples were prepared by diluting synthetic SARS-CoV-2 RNA in inactivated negative gargle sample in a serial dilution. The samples were tested by RT-LAMP and bead-LAMP in parallel to demonstrate the sensitivity boost provided by the simple bead enrichment protocol. (F) Example Bst LF titration for enzyme concentration optimization. RT-LAMP reactions were assembled with varying concentrations of Bst LF DNA polymerase and tested in five replicates on a dilution series of synthetic SARS-CoV-2 RNA near the detection limit. Reactions were allowed to run for 60 min to reveal possible spurious late-onset amplification. Six replicates were performed per condition. (G) Example HIV-1 RT titration for enzyme concentration optimization. RT-LAMP reactions were assembled with varying concentrations of HIV-1 RT and tested as in panel (F). (H) Example BMTU UDG titration for enzyme concentration optimization. RT-LAMP reactions were prepared for a forced contamination experiment. Note the lack of amplification for highest BMTU UDG concentrations, which reveals the inhibition or off-target activity. Two replicates were performed per condition. In (F, G), the empirically derived optimal concentration of the enzyme under our assay conditions is indicated in bold. Source data are available for this figure.

Figure 2.. Optimized, quick inactivation solution for direct-input RT-LAMP and RT–qPCR.

(A) Scheme illustrating quick sample inactivation. Patient-derived (self- or professionally collected) respiratory sample is inactivated by the addition of an inactivation solution and heating for 5 min at 95°C. (B) Sample stability over time with different inactivation reagents. A mock sample in viral transfer medium was inactivated by adding inactivation reagents and heating for 5 min at 95°C. Samples were tested with one-step direct-input RT–qPCR at 0, 4, and 16 h of storage at RT (left), 4°C (middle), and −20°C (right). (C) Comparison of two commonly used respiratory tract samples. Sample stability after inactivation with different solutions was determined for independent swab samples in VTM or gargle lavage in HBSS. Ct values of open-source inactivation solutions were compared with QuickExtract solution. (D) Image demonstrating turbidity generated upon heat inactivation of VTM samples with different inactivation solutions. Smartphone image of mock nasopharyngeal sample in VTM taken after the addition of inactivation solutions and 5-min incubation at 95°C. (E) Colorimetry images of RT-LAMP reactions prepared after inactivation of samples with QuickExtract or in-house inactivation solution. Inactivated nasopharyngeal swab and gargle samples were tested by open-access RT-LAMP to assess the HNB colorimetric readout. (F) Comparison of time-to-threshold values for RT-LAMP and cycle-to-threshold values for RT–qPCR using direct-input samples. Time to threshold in minutes is reported for RT-LAMP, whereas cycle to threshold is reported for RT–qPCR. As evident from the poor correlation, quantitative measurement of target copy number is not possible via RT-LAMP. Source data are available for this figure.

Figure 3.. Open-source lyophilized RT-LAMP mixture.

(A) Schematic representation of lyophilization options for freeze-drying RT-LAMP enzymes and reagents. Glycerol-free enzymes can be lyophilized either as a concentrated enzyme mix or as a reagent mixture with dNTPs and primers. Trehalose is used in both options as a cryoprotectant. Mixtures are flash-frozen in liquid nitrogen, lyophilized using a freeze-dryer, and stored in dry conditions at various temperatures for transport. Before use, lyophilized mixtures are reconstituted. (B) Performance of reactions assembled with reconstituted RT-LAMP enzyme mix. RT-LAMP enzyme mix was lyophilized and stored at 22°C, 4°C, and −20°C. After 30 d, lyophilized enzymes were reconstituted and used to assemble RT-LAMP reactions; control RT-LAMP reactions were prepared with nonlyophilized enzymes stored at −20°C. Reactions were tested on a fourfold, five-step dilution series of a positive sample in four replicates by RT-LAMP, recording the time to threshold for each reaction. In addition, the sample dilutions were measured by direct-input RT–qPCR, and Ct values are displayed under the columns. (C) Time to threshold for reactions prepared using a reconstituted RT-LAMP reagent mix. RT-LAMP reagent mix was lyophilized and stored at −20°C, 4°C, 22°C, and 37°C. After 40 d, RT-LAMP reagents were reconstituted using the reconstitution buffer and reactions were tested on a fourfold, three-step dilution series of a positive sample in duplicates by RT-LAMP, recording the time to threshold for each reaction. In addition, the sample dilutions were measured by direct-input RT–qPCR, and Ct values are displayed under the columns. (D) Sensitivity comparison of RT-LAMP reactions assembled from lyophilized and nonlyophilized RT-LAMP reagents. Lyophilized and freshly prepared reactions were compared in 20 replicates using a dilution of SARS-CoV-2 synthetic RNA (Twist Biosciences) to assess the detection limit. Time-to-threshold values are shown. (E) Forced contamination experiment using lyophilized RT-LAMP reagents. In-house BMTU UDG enzyme, commercially available Antarctic Thermolabile UDG (New England Biolabs), and no UDG were included in three respective RT-LAMP reagent mixes for lyophilization. After freeze-drying, the reagent mixes were reconstituted into RT-LAMP master mixes and compared with freshly prepared reactions in a forced contamination experiment as described earlier (Fig 1C). (F) HNB colorimetric readout is compatible with lyophilized RT-LAMP reagent mixture. Reactions were prepared in parallel from cold-stored enzymes (left) and lyophilized reaction mix (right) to compare colorimetric readout. Smartphone images were taken after lyophilization/preparation of master mix (top row), after the addition of sample (middle row) and after a 35-min incubation at 63°C (bottom row) to showcase the color change. (G) Proof-of-concept multipathogen respiratory virus RT-LAMP test. An eight-well PCR strip was filled with singe-reaction aliquots of RT-LAMP reagent mix for different target RNAs and freeze-dried. A multipathogen test strip targeting SARS-CoV-2, human coronavirus NL63 (HCoV-NL63), influenza A, respiratory syncytial virus A (RSV A), and human RNase P (PC, positive control) was prepared from lyophilized reagents. Shown are the HNB RT-LAMP colorimetric results after reconstitution and addition of mock respiratory sample containing the respective pathogen RNA. A light blue color indicates a positive result. Source data are available for this figure.

Figure S3.. Assessment of performance of lyophilized RT-LAMP reaction mixes (related to Fig 3).

(A) Performance of RT-LAMP reactions assembled with reconstituted RT-LAMP enzyme mix stored at different temperature conditions for 10 d. RT-LAMP enzyme mix was lyophilized and stored at 22°C, 4°C, and −20°C. After 10 d, enzymes mixes were reconstituted in nuclease-free water and used to assemble RT-LAMP reactions, with a control of cold-stored enzymes. Reactions were tested on a fourfold, five-step dilution series of inactivated positive sample in four replicates. Direct-input RT–qPCR Ct values are displayed under the columns. (B) Performance of RT-LAMP reactions assembled with reconstituted RT-LAMP enzyme mix 0 d after lyophilization. RT-LAMP enzyme mix was lyophilized and tested immediately after lyophilization and compared with cold-stored enzymes. Reactions were tested on a fourfold, five-step dilution series of inactivated positive sample. Direct-input RT–qPCR Ct values are displayed under the columns. (C) Sensitivity comparison of lyophilized versus nonlyophilized RT-LAMP reactions on a dilution series of synthetic SARS-CoV-2 RNA (top). Colorimetric readout of reactions presented in Fig 3D (bottom). Smartphone images were taken after a 35-min incubation of RT-LAMP reactions at 63°C. Color-stretched and color-converted images were generated using colorimetry.net. A notable shift in color is observed with lyophilized reagents, but reactions can still be robustly readout based on colorimetry alone. Source data are available for this figure.

Figure 4.. Performance evaluation of lyophilized RT-LAMP reagent mix on clinical specimens in Austria and Ghana.

(A) Performance of direct-input open-source RT-LAMP on clinical samples in Vienna. 192 samples with suspected COVID-19 infection were inactivated and tested in parallel with RT–qPCR targeting the N-gene, or HNB colorimetric RT-LAMP from lyophilized reagents. Shown are the time-to-threshold values obtained from real-time RT-LAMP reactions in comparison with RT–qPCR Ct values. Each dot represents an individual sample. The color of each dot indicates the colorimetric results from HNB RT-LAMP reactions. (B) Colorimetric HNB RT-LAMP results from samples shown in (A). Raw (top) or color-converted (bottom) images are shown. (C) Sensitivity of lyophilized RT-LAMP reactions. Each dot represents a single sample. A simple logistic regression analysis was performed for samples containing different amounts of viral RNA (measured via RT–qPCR) and plotted as a red line. 95% confidence intervals and 50% detection limits are indicated as dashed lines and added as a textbox. (D) Specificity of lyophilized RT-LAMP reactions. Each dot represents a single sample. RT-LAMP detection rates for true-negative samples determined by RT–qPCR are shown. (E) Clinical sample validation performed in Vienna, Austria, and Accra, Ghana. Lyophilized reagents were prepared in Vienna and shipped to Accra. Each site performed an independent validation study. (F) Performance of direct-input open-source RT-LAMP on clinical samples in Accra, Ghana. RNA from 192 samples with suspected COVID-19 infection was extracted and tested in parallel with RT–qPCR targeting the N-gene, or HNB colorimetric RT-LAMP from lyophilized reagents. Shown are the time-to-threshold values obtained from real-time RT-LAMP reactions in comparison with RT–qPCR Ct values. Each dot represents an individual sample. The color of each dot indicates the colorimetric results from HNB RT-LAMP reactions. (G) Colorimetric HNB RT-LAMP results from samples shown in (F). Raw (top) or color-converted (bottom) images are shown. (H) Sensitivity of lyophilized RT-LAMP reactions. Each dot represents a single sample. A simple logistic regression analysis was performed for samples containing different amounts of viral RNA (measured via RT–qPCR) and plotted as a red line. 95% confidence intervals and 50% detection limits are indicated as dashed lines and added as a textbox. (I) Specificity of lyophilized RT-LAMP reactions. Each dot represents a single sample. RT-LAMP detection rates for true-negative samples determined by RT–qPCR are shown. Source data are available for this figure.

Figure S4.. Agreement between RT-LAMP and RT–qPCR on clinical samples (related to Fig 4).

(A) Dilution series of synthetic SARS-CoV-2 RNA in TCEP/betaine/proteinase K–inactivated saline used for converting Ct values obtained on clinical samples into viral load in copies/μl. Synthetic SARS-CoV-2 dilutions were prepared in TCEP/betaine/proteinase K–inactivated negative saline as the dilution matrix. Defined dilutions were tested in replicates, and mean values of these measurements were used to generate an equation translating Ct value to copies/μl. (B) Correspondence of RT-LAMP and RT–qPCR. Shown are RT–qPCR Ct values with corresponding RT-LAMP time-to-threshold values and HNB RT-LAMP colorimetric results for samples shown in Fig 4B. Source data are available for this figure.