LAMP-sequencing method
Sequencing libraries for detecting viral sequences in RT-LAMP products were prepared by a modified Anchor-Seq protocol (37, 40) using Tn5 transposase tagmentation instead of sonication for genomic DNA fragmentation (17). The relevant primers are summarized in table S4.
In detail, transposon adapters containing well-defining barcodes and unique molecular identifiers (UMIs) were annealed by mixing 25 μM oligos (P5-UMI-xi5001…5096-ME.fw, Tn5hY-Rd2-Wat-SC3) in 5 μM tris-HCl (pH 8), incubating at 99°C for 5 min, and slowly cooling down to 20°C within 15 min in a thermocycler. Transposons were assembled by mixing Tn5(E54K, L372P) transposase (100 ng/μl) [purified according to (41)] with 1.25 μM annealed adapters in 50 mM Tris-HCl (pH 7.5) and incubating the reaction for 1 hour at 23°C. Tagmentation was carried out by mixing 1.2 μl of the RT-LAMP product (~200 ng DNA) with 1.5 μl of loaded transposase in freshly prepared tagmentation buffer [10 mM [tris(hydroxymethyl)methylamino]propanesulfonic acid) (TAPS)] (pH 8.5), 5 mM MgCl2, and 10% (v/v) dimethylformamide] using a Liquidator 96 Manual Pipetting System (Mettler Toledo). The reactions were incubated at 55°C for 10 min. Reactions were stopped by adding SDS to a final concentration of 0.033%. Tagmented DNA of each plate was pooled and size-selected using a two-step AMPureXP bead (Beckman Coulter) purification to target for fragments between 300 and 600 bp. First, 50 μl of pooled reaction was mixed with 50 μl of water and bound to 55 μl of beads to remove large fragments. To further remove small fragments, the supernatant of this reaction was added to 25 μl of fresh beads and further purified using two washes with 80% ethanol before the samples were finally eluted in 10 μl of 5 mM tris-HCl (pH 8). One PCR per plate with 1 μl of the eluate and RT-LAMP–specific and Tn5-adapter–specific primers (P7nxt-GeneN-A-LBrc and P7-xi7001..7016, P5.fw) was performed using NEBNext Q5 HotStart polymerase (New England Biolabs) with two cycles at 62°C for annealing and 90 s elongation, followed by two cycles at 65°C for annealing and 90 s elongation, and 13 cycles at 72°C annealing and 90 s elongation. All PCR reactions were combined and 19% of this pool was size-selected for 400 to 550 bp using a 2% agarose/tris-acetate-EDTA gel and column purification (Macherey-Nagel). The final sequencing library was quantified by qPCR (New England Biolabs) and sequenced with a paired-end sequencing run on a NextSeq 550 machine (Illumina) with 20% phiX spike-in and 136 cycles for the first read, 11 cycles to read the 11-nt-long plate index (i7) and 20 cycles to read the 11-nt-long well index (i5) and the 9-nt-long UMI.
For trimming of the reads (i.e., removal of P7 Illumina adapter sequences), cutadapt (version 2.8) (42) was used. For validation of the origin of the sequence of the LAMP product (fig. S4A), 107 reads were randomly selected and used for the analysis. Reads were mapped to the SARS-CoV-2 reference genome (NC_045512.2) (43), using bwa-mem with default settings (version 0.7.17-r1188) (44). Virus genome coverage was determined with the samtools depth command (version 1.10) (45). Using bwa-mem, 80.6% of reads could be mapped to the virus genome (fig. S4, B and C). To analyze the remaining sequences, a k-mer analysis using a custom script was performed. Using 9-mers, this matched 93.5% of the nonmapped reads with a maximal Levenshtein distance of two to one of the LAMP primers or their reverse complement sequences (fig. S4D). This is explained by the fact that LAMP products can consist of complex sequence rearrangements.
For classification of samples by LAMP-sequencing, reads were assigned to wells and counted using custom scripts. A read was considered as a match to SARS-CoV-2 N gene if at least one of three short sequences (~13 nt, marked orange in fig. S4A) not covered by RT-LAMP primers was found in the read, otherwise it was counted as unmatched. Sequencing reads were grouped by UMI and by position of the matched sequence with the aim of removing PCR duplicates. A sample was considered if more than 200 total UMIs were observed and called positive if more than 10,000 virus-matching UMIs were observed.
There is a very wide gap in the number of virus-matching reads between positive and negative samples (fig. S5A): The count is either below 7000 UMIs or above 45,000 UMIs. This is why we placed the decision threshold for scoring a sample as LAMP-sequencing positive within this gap. The fact that also RT-qPCR–negative samples give rise to some UMI counts containing viral sequences is explained by template switching of unattached adapters that remain in the reaction after tagmentation, but no cause for concern due to the wide gap between negative and positive samples.
For a few samples, we saw so few reads (less than 200 UMIs) that we suspected that the multiplexing had failed and excluded them from the results. As most of these were in the same row of the same plate, we analyzed these samples after LAMP-sequencing by gel electrophoresis (fig. S5B) to check for DNA content after RT-LAMP. We found that the gel results agree with the RT-LAMP outcome, indicating that the failure likely was caused later, probably during multiplexing.