Direct detection of SARS-CoV-2 using non-commercial RT-LAMP reagents on heat-inactivated samples

RT-LAMP detection of SARS-CoV-2 has been shown to be a valuable approach to scale up COVID-19 diagnostics and thus contribute to limiting the spread of the disease. Here we present the optimization of highly cost-effective in-house produced enzymes, and we benchmark their performance against commercial alternatives. We explore the compatibility between multiple DNA polymerases with high strand-displacement activity and thermostable reverse transcriptases required for RT-LAMP. We optimize reaction conditions and demonstrate their applicability using both synthetic RNA and clinical patient samples. Finally, we validate the optimized RT-LAMP assay for the detection of SARS-CoV-2 in unextracted heat-inactivated nasopharyngeal samples from 184 patients. We anticipate that optimized and affordable reagents for RT-LAMP will facilitate the expansion of SARS-CoV-2 testing globally, especially in sites and settings where the need for large scale testing cannot be met by commercial alternatives.

The ongoing SARS-CoV-2 pandemic has had a tremendous impact on society, surpassing one million casualties worldwide and exposing the vulnerability of our globalised world to the spread of infectious disease. Extensive testing and isolation of confirmed cases has been proposed as a viable strategy to limit the spread of SARS-CoV-2 1-3 . Consequently, many methods for SARS-CoV-2 ribonucleic acid detection have been developed (reviewed in 4 ). The gold standard for RNA detection is the reverse transcription quantitative PCR (RT-qPCR), requiring specialized equipment of limited availability in many contexts (e.g., real time thermocyclers). Due to their intrinsic simplicity and high sensitivity, isothermal detection methods are increasingly appreciated as well suited for point-of-care (POC) testing 5 , and among them LAMP (loop-mediated isothermal amplification) is an attractive approach 6,7 . Early on during the COVID-19 pandemic many researchers, including ourselves, worked to adapt and optimize RT-LAMP detection methods for SARS-CoV-2 diagnostics [8][9][10] . This initial work quickly expanded with optimized oligonucleotide designs and simplified purification systems [11][12][13] . The specificity of RT-LAMP has been improved by applying CRISPR based detection 14,15 or high-throughput sequencing to confirm amplicon sequences and to facilitate mass usage 16,17 . Simplified isothermal approaches have the potential to facilitate SARS-CoV-2 diagnosis and thus contribute to the efforts against the COVID-19 pandemic. This is especially important when considering that frequent testing, even at a lower sensitivity, can contribute to a decrease in spread of the disease 1-3 .
Scientific Reports | (2021) 11:1820 | https://doi.org/10.1038/s41598-020-80352-8 www.nature.com/scientificreports/ However, access to reliable and affordable molecular biology reagents remains a problem. Even though the production and distribution have significantly increased since the start of the crisis, the need for testing to control disease spread and allow reopening of society keeps raising the demand. This is especially problematic in countries with limited distribution channels and a lack of economic resources to afford massive testing. To make molecular testing more accessible and allow its widespread implementation, multiple approaches have been developed. For example, it has recently been shown how a simplified RT-qPCR reaction for SARS-CoV-2 detection is feasible using only one enzyme 18 , or even how it is possible to use crude enzymes from lyophilized bacteria for LAMP 19 . Here we aim to contribute to the global effort to generate affordable SARS-CoV-2 tests based on in-house produced enzymes and thus democratize their use. We established protocols for simple production of DNA polymerases with high strand-displacement activity 20 and optimized their reaction conditions for RT-LAMP. We explore their compatibility with thermostable reverse transcriptases and provide a complete in-house reagent mix for fluorescent or colorimetric detection of SARS-CoV-2 RNA. We then benchmarked the optimized reaction mix with multiple commercial alternatives and RT-LAMP primer sets. Finally, we tested the ability of the produced reagents to correctly detect SARS-CoV-2 presence directly in non-purified clinical nasopharyngeal samples from 184 patients.

Materials and methods
Expression and purification of strand displacing polymerases.. Displacement polymerases assayed for RT-LAMP polymerase activity included the Geobacillus stearothermophilus exonuclease-deficient family-A polymerase large fragment (Bst-LF), and its 'v5.9′ and 'v7.16′ derivatives previously reported by Ellington and colleagues to incorporate functional properties of the related Klentaq polymerase from Thermus aquaticus 20 . Prior to functional characterization, hexahistidine (H6)-tagged variants were screened for expression efficiency in various vectors, including pTetA (constructs kindly provided by Andrew Ellington), pET-16b and pET-28a (synthesized for this work). Constructs in pET-16b contained an ASRGS-H6 followed by the polymerase sequence, inserted between NcoI and BamHI sites in the multiple cloning region. Similarly, constructs in pET-28a were engineered by inserting the polymerase between NdeI and XhoI, 3′ to the incorporated GSS-H6-SSG and thrombin cleavage sequences.
Expression and Purification of thermostable reverse transcriptases.. The MashUp-RT plasmid (pET backbone) (kindly provided by https ://pipet tejoc key.com/) was transformed into BL21 (DE3) T1R pRARE2 expression strain and plated on L-Broth (LB) agar (Formedium, Norfolk, UK) plate containing 50 µg/ mL kanamycin. Overnight cultures were inoculated with fresh transformants and grown at 37 °C at 175 RPM in LB medium (Formedium, Norfolk, UK) containing kanamycin (50 µg/mL). Subsequently, the overnight culture was diluted into 1.5 L fresh LB medium supplemented with kanamycin and catabolite repression buffer (25% glycerol, 25% glucose, 1 mM MgSO 4 and 0.1 mM MnCl 2 ) and the cultures were grown at 37 °C in the LEX system (LEX-48, Epiphyte Three Inc., Toronto, Canada). At OD 600 0.9, the cultures were down-tempered to 18 °C before protein expression was induced with 0.5 mM IPTG and the culture was grown additionally for 24 h at 18 °C. The cells were harvested by centrifugation at 4500×g for 10 min at 4 °C and resuspended in MashUp-RT IMAC lysis buffer (25 mM Tris-HCl, 300 mM NaCl, 10% glycerol, 40 mM imidazole, 0.5% Triton X-100, pH 8) supplemented with 1 mg/mL lysozyme and one tablet Complete EDTA-free protease inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany) per 1.5 L culture. The resuspended cell pellets were stored at − 80 °C.
RT-LAMP primer design and positive control. We used either the iLACO 8 or As1e 12 primers as previ- All oligonucleotides were purchased from Thermo Fisher Scientific (Waltham, MA, USA) with standard desalting and dissolved in nuclease free water upon arrival.
We generated synthetic fragments of SARS-CoV-2 RNA (200-300 bp) by in vitro transcription of PCR fragments containing part of SARS-CoV-2 sequence and a T7 promoter. For iLACO we amplified the PCR product from the viral genome using T7_FW_iLACO (TAA TAC GAC   Reactions containing Bst 3.0 DNA Polymerase (New England Biolabs, Ipswich, MA, USA) and Saphir Bst2.0 DNA polymerase (Jena Bioscience, Jena, Germany) were largely performed as described for the in-house reagents. Isothermal amplification buffer II was used with Bst3.0 and Saphir Bst2.0 Turbo Buffer (Jena Bioscience, Jena, Germany) was used with Saphir Bst2.0, as recommended by manufacturers. Enzyme amounts were used as recommended by manufacturers. When using WarmStart Colorimetric RT-LAMP 2X Master Mix (New England Biolabs, Ipswich, MA, USA), 5 µl mix was used together with 1 µl 10X LAMP primer mix, 0.5 µl Eva Green dye, 1 µl of sample, and nuclease-free water to a final volume of 10 µl (double the volumes for 20 µl reactions). All reactions were run in a Step One Plus Real time PCR system (Applied Biosystems).
To test the RT activity of RTX and RTX exo-, we performed RT-qPCR according to the following protocol. 10 µl RT reactions were set up for each condition containing: either ThermoPol buffer, Isothermal Amplification buffer, or Isothermal Amplification buffer II; 1.4 mM each dNTP; 6 mM MgSO 4 (where appropriate); 0.2 µM iLACO-qPCR-rv primer (AGC ATG TTA GGC ATG GCT CT); specified amount of either RTX or RTX exo-, or 4 U/µl Superscript IV (SSIV), or no enzyme; 83 million copies of iLACO synthetic RNA template; and nucleasefree water up to 10 µl. RT was performed at 65 °C for 15 min. 1 µl of each RT reaction was used as template in a 10 µl qPCR reaction using 5 µl Fast SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) and 0.5 µM each of iLACO_qPCR_fw (TAT GGT GGT TGG CAC AAC AT) and iLACO_qPCR_rv primers. qPCR was performed according to the standard protocol.
Numeric raw data for the amplification curves is provided in Table S1.
Clinical samples. We obtained 184 nasopharyngeal samples from Karolinska University Hospital, Huddinge (Stockholm) collected between May 20th and June 1st 2020. We used pseudo-anonymized surplus material previously collected for clinical diagnostics of SARS-CoV-2. This is in accordance with the Swedish Act concerning the Ethical Review of Research Involving Humans, which allows development and improvement of diagnostic assays using patient samples which were collected to perform the testing in question. Thus, according to Swedish law neither specific ethical approval nor informed consent was required. Pseudo-anonymized means that data was process in such a way that it can no longer be attributed to a specific subject without the use of additional information, and the additional information is separate from the authors, making it virtually impossible to identify any individual. We picked samples collected in either Ethics statement. We used pseudo-anonymized surplus material previously collected for clinical diagnostics of SARS-CoV-2. This is in accordance with the Swedish Act concerning the Ethical Review of Research Involving Humans, which allows development and improvement of diagnostic assays using patient samples which were collected to perform the testing in question. Pseudo-anonymized means that data was process in such a way that it can no longer be attributed to a specific subject without the use of additional information, and the additional information is separate from the authors making virtually impossible to identify any individual. Additional ethical approval for RT-LAMP diagnosis was obtained by the appropriate Swedish Authority (Dnr 2020-01945, ethics committee: Etikprövningsnämnden).

Results
Purification and optimization of alternative LAMP enzymes. As a starting point to optimize alternative reagents for LAMP testing, we used the thermophilic strand-displacing polymerases developed by the Ellington Lab 20 . We used both Bst LF from Geobacillus stearothermophilus as well as the chimeras v5.9 and v7.16 of Bst LF and Klentaq (Thermus aquaticus) which displayed higher thermal stability 20 . We performed simple enzyme expression and purification to enable effortless production of good quality reagents for hundreds of thousands of tests (see methods). To validate the performance of the synthesized enzymes in SARS-CoV-2 RT-LAMP detection, we first benchmarked their ability to amplify in vitro produced SARS-CoV-2 RNA fragments using the iLACO primer set that we previously developed 8 ( Fig. 1 and Table S1).
We tested the intrinsic reverse transcriptase (RT) activity of all enzymes, and compared their ability to efficiently amplify and detect the synthetic RNA fragment with or without supplementing the reaction with a commercial thermostable RT enzyme (Superscript IV, SSIV). Addition of a thermostable RT was essential to enhance the performance of all tested enzymes (not shown), thus, we carried out further optimizations in the presence of SSIV. To facilitate quantitative comparison between samples, we performed all the optimizations using continuous fluorescence detection. However, to increase throughput and avoid the use of expensive equipment, end-point measurement of fluorescence or colorimetric detection can also be applied 8 www.nature.com/scientificreports/ We optimized the composition of the reaction buffer and the amount of strand displacing DNA polymerase used (Fig. 1A,B). For a fixed template amount we minimized the time to detection (color in Fig. 1B) and maximized the time between true positive and false positive (spurious amplification, number in Fig. 1B; see Fig. S1 for detail). Non-specific spurious LAMP amplification is a known phenomenon that can lead to false positive detection and is templated by the primers themselves 22 . In some cases, the amplification of spurious non-specific sequences can be detected by performing a melting curve analysis (Fig. S2) 23 . However, to avoid false positives, reaction conditions and components should be optimized to maximize sensitivity while minimizing spurious amplification. Variation of any component of the reaction (e.g., oligos, enzyme, buffer composition…) as well as the physical reaction conditions (e.g., temperature, time…) can affect the spurious amplification. For example, while increased enzyme concentration decreases time to detection, it also increases false positive amplification (Fig. 1B,C). We observed that the optimal concentration of KCl varied between the enzymes tested. 150 mM KCl was optimal for Bst LF, 50 mM for v7. 16  www.nature.com/scientificreports/ capacity of betaine 24 , guanidine 25 and DMSO 26 to increase sensitivity and reduce nonspecific amplification. Unfortunately, in our hands none of these strategies improved sensitivity or specificity. It is important to note that even the dye used for fluorescent detection can alter the measured RT-LAMP sensitivity and specificity and thus should also be accounted for in optimization. In our conditions, Eva Green dye provided a higher dynamic range than SYBR Green I, when measuring difference in fluorescence before and after RT-LAMP reaction. We compared the three polymerases based on the speed of amplification, the specificity, and the range of conditions where the former two parameters are acceptable. Specifically, we selected the minimal amount of enzyme capable of amplifying 10 5 copies of RNA template before 20 min with no measurable amplification of the non-template control before 50 min. We found that v7.16 outperformed v5.9, and both outperformed Bst LF. Amplification curves for LAMP at optimal conditions with v5.9 and v7.16 are shown in Figs. 1C,D (raw data in Table S2). We proceeded with both v5.9 and v7.16 in the following experiments.

Optimization of alternative RT enzymes compatible with LAMP.
Once we optimized the conditions for efficient strand-displacement DNA activity required for LAMP, we decided to explore alternatives to the commercial thermostable RT enzymes. First, we explored the use of a thermostable synthetic reverse transcriptase, RTX, developed by the Ellington lab 27 , which can also be used for direct RT-qPCR of SARS-CoV-2 18 . However, supplementing either v5.9 or v7.16 with different concentrations of RTX did not improve their ability to detect SARS-CoV-2 RNA ( Fig. 2A and Table S1). Addition of high concentrations of RTX lead to abnormal amplification curves, while lower concentrations did not improve the RT-LAMP activity of v5.9 and v7. 16. Similar results were obtained using both the RTX versions containing or excluding the proofreading domain ( Fig. S3A-D). Importantly, the absence of improvement in the LAMP reaction was not due to lack of RT activity of these enzymes, as demonstrated by standard RT-qPCR (Fig. S3E). Thus, we concluded that RTX is not compatible with RT-LAMP. In addition to RTX, we aimed to explore other non-commercial alternatives. MashUp-RT is based on FeLV-RT, which is intrinsically more accurate than MMLV-RT 28  www.nature.com/scientificreports/ thermostability and fidelity (https ://pipet tejoc key.com/). We expressed MashUp-RT and tested its ability to enhance RT-LAMP reaction in combination with v5.9 and v7.16. The addition of 0.025 μg/µl reaction was enough for a performance identical to, or even better than, that of SSIV (Fig. 2B,C). This shows that MashUp-RT is a thermostable RT enzyme compatible with LAMP that is capable of enhancing RT-LAMP detection sensitivity for SARS-CoV-2 RNA.

Benchmarking of alternative RT-LAMP enzyme mix and oligo combinations.
Once we optimized non-commercial enzyme RT-LAMP conditions, we benchmarked their ability to detect SARS-CoV-2 against commercial alternatives. After testing multiple oligo primer sets optimized for SARS-CoV-2 8,10,12 , we decided to focus on the use of iLACO 8 and As1e 12 targeting ORF1ab, which worked better in our hands. We compared our optimized enzyme mix with commercial alternatives from multiple providers (i.e., WarmStart Colorimetric master mix (NEB), Bst3.0 (NEB) and Saphir Bst2.0 Turbo (Jena Biosciences)) using synthetic RNA ( Fig. 3 and Table S1). Although all enzymes were able to detect SARS-CoV-2 RNA, we identified clear differences in both sensitivity and background level. In particular, the performance of the RT-LAMP reaction using v7.16 enzyme and As1e primers was comparable to the best commercial alternatives, with a detection speed similar to Bst3.0 and lower spurious amplification. Unexpectedly, we found that Bst3.0 needed to be supplemented with RT activity to detect SARS-CoV-2 RNA (Fig. S4). The performance of primer sets was also affected by the different enzymes used. For example, v5.9 and Saphir Bst2.0 perform better in combination with iLACO primers, while other enzymes perform better with As1e or similarly with both. Thus, we concluded that in-house produced v5.9 or v7.16 in combination with MashUp-RT and iLACO or As1e primers were competitive alternatives for RT-LAMP detection of purified RNA.

Application of optimized RT-LAMP to clinical nasopharyngeal samples. Since we and others
have already shown the utility of RT-LAMP for the detection of SARS-CoV-2 from patient purified RNA [8][9][10][11]17 , we now wanted to explore the ability of our in-house enzyme mix to detect SARS-CoV-2 also in non-purified samples 11,17 . We first explored its compatibility with different virus transport media commonly used for the collection of nasopharyngeal swabs: Virocult (MWE), Sigma Transwab (MWE), eSwab (Copan) and Beaver (BEAVER biomedical) (Fig. 4A and Table S1). Three of the four tested media (Virocult, Transwab and eSwab) were compatible with both v7.16 and v5.9 reactions when adding 10% of the reaction volume. However, the Beaver media required additional dilution (three-fold) to avoid inhibition and was compatible only with v5.9. This limits the total amount of patient sample that could be used per reaction and thus decreases the overall sensitivity.
Having established the reaction conditions, we used 184 nasopharyngeal patient samples collected in Virocult, TransSwab and eSwab transport media that have been previously analyzed using GeneXpert SARS-CoV-2 detection (Cepheid) at the Karolinska University Hospital (Sweden). We selected 142 positive samples across all GeneXpert Ct ranges, and 42 samples determined to be negative for SARS-CoV-2. We used our developed v7.16 + MashUp-RT reaction mix in combination with either iLACO or As1e primer sets (Fig. 4B,D and Table S2). While both primer sets were successful in detecting samples with high viral load, the As1e primers showed better overall sensitivity. Performing technical duplicates demonstrated that the sensitivity was further increased if a sample was called as positive when at least one replicate had amplified (Fig. 4C). Unfortunately, in all cases patient samples with low-to-medium content of SARS-CoV-2 RNA (higher GeneXpert Cts) were often misclassified as false negatives. Finally, we compare the results obtained with our in-house reagents with state-of-the-art

Discussion
Here we have optimized and validated the production and application of alternative non-commercial reagents for simplified RT-LAMP based SARS-CoV-2 detection. We demonstrate their ability to identify COVID-19 positive patients using unextracted nasopharyngeal swabs in common virus transport media. And finally, we show that even omitting RNA purification, RT-LAMP can easily detect samples with high to medium SARS-CoV-2 RNA content. This is especially important when considering that frequent testing, even at a lower sensitivity, can help to effectively decrease disease spread [1][2][3] . During this optimization process we obtained some general recommendations that can facilitate the application of RT-LAMP in different settings. In addition to the initial benchmarking with alternative RT-qPCR approaches, it is important to routinely control for spurious amplification and cross contamination. For example, it is common to observe variation in the quality and specificity of different oligonucleotide batches (even from the same manufacturer). In some circumstances, oligonucleotide batches can contain a trace amount of target regions from the SARS-CoV-2 genome due to oligo cross contamination during oligonucleotide synthesis or handling. Thus, reactions with negative water controls should always be performed in parallel to ensure the specificity of the test.
The activity of each new batch of enzymes and reagents should first be calibrated with a known serial dilution of target RNA. In particular, we recommend assessing multiple concentrations of the newly produced protein (v5.9 or v7.16) in a range from 0.01 to 1 µg/µl in reaction. We propose to estimate batch activity by assaying the minimal amount of enzyme capable of amplifying 10 5 copies of RNA template before 20 min with no measurable amplification of the non-template control before 50 min. If end-point measurement is used instead of real-time tracking, the time at which the reaction is stopped should also be optimized. These parameters should be estimated whenever a new batch of enzyme is produced, which is usually enough for tens of thousands of reactions. However, we suggest to re-assay if loss of activity is suspected.
Although direct RT-LAMP in unextracted samples can detect most cases with medium to high SARS-CoV-2 viral load, to bring it to a RT-qPCR detection sensitivity level it is still necessary to perform at least some simple sample purification and RNA concentration [9][10][11]15 . In addition, even if the use of viral transport media can be tolerated by RT-LAMP, less complex media, or no media at all, might be more advantageous 29 . Our work also highlights the importance of controlling for the presence of salt and other components that may affect RT-LAMP sensitivity and specificity. In summary, we hope that this work will facilitate the implementation of affordable RT-LAMP SARS-CoV-2 diagnostics in settings where the use of commercial reagents is not possible. This could decrease the total cost per test in order to facilitate routine testing of large portions of the population. www.nature.com/scientificreports/