Development of a Highly Multiplexed RT-LAMP Assay for Coverage of Genetic Sequence Diversity

Abstract

Nucleic acid amplification tests (NAATs) can achieve high accuracy for diagnosing infectious diseases by targeting conserved genetic sequences specific to the target organism. Isothermal NAATs, such as reverse-transcription loop mediated isothermal amplification (RT-LAMP), simplify instrumentation requirements facilitating point-of-care testing. However, sequence variation due to genetic variability can cause false negative results. Single-pot multiplex testing can improve sequence coverage, but RT-LAMP is complicated by requiring many primers for even a single assay, which can lead to non-specific amplification. We implemented a process that leveraged manual primer design to develop a highly multiplexed RT-LAMP assay (Chain LAMP) targeting 7 adjacent genomic target regions of HIV, one of the most diverse clinically relevant pathogens. This process departed from standards for RT-LAMP design, including omission of bumper primers whose activity was replaced by cooperative neighboring assays. The Chain LAMP is, to our knowledge, the highest order single-pot multiplexed RT-LAMP assay published. The assay has an analytical limit of detection of 25 copies of RNA/reaction without detectable non-specific amplification, translating to 1,000 copies of HIV/mL of plasma from a fingerstick sample, aligning with WHO standards for HIV viral load monitoring. When evaluated using 24 clinical RNA samples representative of global HIV diversity, the Chain LAMP demonstrated robust coverage of sequence diversity, amplifying all samples with minimal sensitivity variation. We performed mechanistic analysis with Nanopore sequencing, identifying liftoff of multiple assay regions for each sample indicating many initiation loci. The high level of multiplexing in the Chain LAMP effectively increases coverage of HIV sequence diversity.

Graphical Abstract

Introduction

Nucleic acid amplification tests (NAATs) play a crucial role in the diagnosis and treatment monitoring of infectious diseases. While reverse transcription-polymerase chain reaction (RT-PCR) is the gold standard NAAT, reverse-transcription loop mediated isothermal amplification (RT-LAMP) is an isothermal alternative that enables sensitive and specific detection, even at the point-of-care (POC) with no requirement for complex instrumentation, including a thermal cycler.^1–3 For NAATs, sequence variation due to genetic variability and mutations can cause mismatches between primers and target sequences, which impairs amplification and can produce false negative results.^4–9 Many human pathogens, most notably RNA viruses including HIV, HCV, SARS-CoV-2, and influenza A/B, can have high genetic sequence diversity.

Strategies to improve coverage of pathogen sequence diversity in NAATs include using proofreading polymerases to remove terminal mismatches however, this comes with increased test cost and complexity.^5,10 Another approach is the use of degenerate primer pools or nucleotides to cover specific variant bases.¹¹ This method is simple and inexpensive but often insufficient on its own to confer broad coverage. Single-pot multiplexed assays target multiple genomic regions in a single reaction, providing redundancy to mitigate mismatches in some assays and thus improve coverage of diverse sequences. This strategy has been employed effectively for several pathogens (Table S1).^12–16

Multiplexing also enables identification of multiple pathogens in a single sample and inclusion of internal controls. However, LAMP assays can be difficult to multiplex due in part to the large number of primers involved. Each LAMP assay requires 4–6 primers (compared to only 2 for PCR), which can interact to cause non-specific amplification and false positive results.^17,18 Multiplexed LAMP assays are rare, with a few examples developed using conventional primer design tools to target 2–3 sequence regions, and one assay that used manual primer design to create a multiplex for 6 pathogen targets (Table S1).¹²

Here we describe a highly multiplexed RT-LAMP assay (Chain LAMP) that combines 7 assays (7-plex) in a single reaction to achieve high genetic coverage of HIV, validated through testing of extracted clinical RNA samples. HIV has the highest reported mutation rate of any biological entity¹⁹ and the greatest degree of genetic variability of any virus studied (up to 35% sequence diversity between Group M subtypes),^20–22 and thus provides one of the most difficult primer design challenges in terms of sequence diversity and coverage of variant subtypes. As commonly used software is insufficient to analyze primer interactions and sequence coverage for a multiplexed RT-LAMP assay, we established a design strategy to integrate these analyses into the primer design process. We designed the 7 assays in close proximity on the HIV genome, enabling the novel removal of bumper primers due to neighboring assay cooperativity to reduce non-specific amplification. This design approach can be used generally to improve coverage for RT-LAMP assays for other pathogens with high sequence diversity. We used Nanopore sequencing to understand the mechanisms underlying multiplexed RT-LAMP amplification. This assay represents the highest order single-pot multiplexed RT-LAMP assay to our knowledge published to date (as of Sept. 2024).

Materials and Methods

Coverage Analysis Software.

Chain LAMP and published primer sets were analyzed for HIV sequence coverage using a custom python script (available on GitHub).²³ Briefly, this script analyzes coverage between primer sets and sets of HIV sequences organized by subtype (Figure S9). Coverage is defined as (1) Perfect alignment of all primer regions (2) ROSALIND DxM Derived Coverage:^24–26 ROSALIND DxM derived severity score of ≤ 3 for all primers in a LAMP assay (Figure S10A). Derived ROSALIND scoring looks at individual sequences in lieu of an incident and prevalence approach in traditional ROSALIND DxM severity scoring.²⁵ In the context of multiplex assays, overall coverage is determined if at least one assay shows coverage. Finally, coverage is weighted based on global subtype prevalence.²⁷ Software can extract sequences from Los Alamos National Labs (LANL) alignments²⁸ or from GenBank ID. For development, coverage can be determined at the assay, primer and nucleotide level. Sequence alignments (LANL: Alignment ID-122FP8; 2022; M Group with CRFs; Filtered Web; Genome; DNA).²⁸

In-Silico Primer Analysis.

In-silico analysis of Chain LAMP assay mismatch severity with extracted clinical RNA samples was performed with a custom python script (available on GitHub).²³ Sequences were provided for samples from the External Quality Assurance Program Oversight Laboratory (EQAPOL).^29,30 Briefly, sequence information was obtained by GenBank ID and alignment score for each assay region was obtained from a ROSALIND DxM derived severity scoring (Figure S10A).²⁴

RT-LAMP Primer Design.

Our primers were manually designed against a 75% HIV consensus sequence, based on reference sequences of all global Group M subtypes (LANL: 2020; Alignment ID-120RP8; Subtype Reference; Genome; M group without recombinants (A-L); DNA).²⁸ Chain LAMP targets were designed in close proximity spanning the pol gene. RT-LAMP design omitted individual bumper primers (F3/B3) with a single set of flanking bumper primers and included degenerate bases within primers. Manual primer design enabled departure from some conventional RT-LAMP parameters (Supplementary Section S2). Initial primer design and targeted iterative redesign was performed using tools within Geneious with NUPACK,³¹ IDT OligoAnalyzer, and NCBI BLAST as supporting tools. See Supplementary Section S2 for more detailed information.

LAMP and RT-LAMP Amplification.

20 μL LAMP reactions contain 50 mM bicine at pH 8.3, 10 mM (NH₄)₂SO₄, 50 mM KCl, 8 mM MgSO₄, 0.1% (vol/vol) Triton X-100, 1X EvaGreen^® (Biotium, 31000), 1.4 mM dNTPs (NEB, N0447), 600 nM each FIP and BIP primers and 300 nm each LF and LB primers, 200 nM each flanking bumper primers, 200 nM Cy5-tagged UDP probe, 300 nM BHQ-1 tagged quencher probe, and 10 U (1.05 μg) TFpol polymerase (in-house polymerase; plasmid preparation, protein expression, and purification previously described).³ 2 μL of template or extracted sample was added to 18 μL reaction mixture. 20 μL RT-LAMP reactions contain additionally, 6 units of WarmStart^® RTx Reverse Transcriptase (NEB, M0380L) and 10 units RNAsin^® plus RNase inhibitor (Promega, N2611).

Reactions were incubated at 65 °C for 75 minutes on a CFX96 (BioRad Laboratories, Hercules, California). DNA LAMP reactions included additional 95 °C heat spike for 5 seconds for dsDNA denaturation prior to amplification.³² Fluorescence measurement for FAM (dsDNA-binding dye) and Cy5 (UDP probe) signal were taken every 30 s. Analysis performed with Bio-Rad CFX Maestro 2.3 software (version 5.3.022.1030) with Cy5 channel baseline set as 10 to 60 cycles with a manual threshold of 50 RFU, and FAM channel baseline set as 10 to 50 cycles with a manual threshold of 50 RFU.

DNA Reference Template and Quantification.

Genomic DNA from 8E5 cell line (ATCC, CRL-3617) was used to prepare HIV DNA templates. The pol region (3kb) of HIV was amplified using flanking bumper primers (Table S2). The amplified PCR products were purified using Monarch PCR cleanup kit (NEB, T1030S) and length was verified using Agilent D5000 Screen Tape (Agilent, 5067– 5588). The extracted DNA template was quantified via Qubit High-Sensitivity dsDNA assay (ThermoFisher, Q32851). Samples were aliquoted for single use and stored at −20 °C until use.

RNA Reference Template and Quantification.

Inactivated HIV virus stock was a customized order from Zeptometrix (Zeptometrix, 0801032CFHI) by incubating the HIV-1 culture fluid of Subtype IIIB (ZeptoMetrix, 0801032CF) at 56 °C for 1 h. HIV RNA templates (Subtype B inactivate HIV reference isolate) were extracted from virus stock using QIAamp Viral RNA Mini Kit (QIAGEN, 52904) according to manufacturer instructions. Samples were aliquoted and stored at −80 °C until use. Extracted RNA was quantified using GeneProof HIV type 1 (HIV-1) D PCR Kit (GeneProof, HIV1D/GP/025). Experiments involving inactivated HIV virus were conducted following biosafety level 2 (BSL-2) procedures approved by Environmental Health and Safety (EH&S) at UW.

EQAPOL Samples Extraction and Quantification.

Panel of diverse HIV virus isolates was obtained from EQAPOL^29,30 that span HIV-1 M group major subtypes, circulating recombinant forms (CRFs), and unique recombinant forms (URFs), roughly mirroring global HIV subtype diversity (Table S4). The panel consisted of 24 culture supernatant virus stocks (A [3], B [4], C [5], D [2], F [1], G [3], CRF01_AE [2], CRF02_AG [3], and URF [1]) derived from HIV-positive plasma specimens and characterized, including providing genome sequences of each isolate.

Total RNA from 140 μL culture supernatant was extracted using QIAamp Viral RNA Mini Kit (QIAGEN, 52904) according to manufacturer instructions with the addition of 6 μL of GeneProof universal internal control and eluted into 60 μL of elution buffer. Extracted RNA was quantified using GeneProof HIV type 1 (HIV-1) D PCR Kit (GeneProof, HIV1D/GP/025). Samples were aliquoted and stored at −80 °C until use. Experiments involving cultured HIV virus were conducted following biosafety level 2 with biosafety level 3 practices (BSL-2+) procedures approved by UW EH&S.

EQAPOL Sample Testing.

Sample LOD was determined by a two-step testing process starting with 10-fold extracted RNA dilutions and further moving towards 2.15-fold dilutions based on results from Test 1. LOD was determined by duplicate liftoff of UDP probe signal within 70 minutes (Figure S16A).

Sequencing of LAMP products.

RT-LAMP amplicons were purified with AMPure XP Beads (Beckman Coulter, A63880) and quantified with Qubit dsDNA BR Assay Kit (ThermoFisher, Q32850) prior to library preparation and at intermediate steps. Libraries were prepared with 400 ng of purified DNA using the Native Barcoding Kit 24 V14 (Oxford Nanopore Technologies (ONT), SQK-NBD114.24) following the manufacturers’ instructions using NEBNext Ultra II End Repair/dA-Tailing Module (NEB, E7546S) for initial end repair, Blunt/TA Ligase Master Mix (NEB, M0367) for barcode ligation and the NEBNext Quick Ligation Reaction Module (NEB, E6056).

The libraries were sequenced with the MinION^™ nanopore sequencer (ONT) using a Flongle Flow Cell (R10.4.1) with adapter. Data acquisition was managed with ONT MinKNOW software (v24.02.8). Basecalling was performed using the ONT open-source software Dorado (v0.6.0), employing the super accurate model (v4.3.0). Sequence alignment and coverage frequency with Nextflow (v24.04) workflow in a docker container with samtools (v1.20)³³, chopper (v 0.8.0)³⁴, minimap2 (v2.28)³⁵ and python, with additional analysis of target coverage using python. Pipeline is available on Github.³⁶

Results and Discussion

Chain LAMP Concepts and Mechanisms.

Conventional LAMP assays contain 6 primers: inner primers (FIP and BIP), loop primers (LF and LB), and bumper primers (F3 and B3) (Figure 1B).^1,37 Bumper primers are required at initiation of amplification to displace an initial extension product, leading to a dumb-bell shaped seed for exponential amplification (Figure 1B).¹ Chain LAMP differs from conventional LAMP as the 7 assays are designed against targets regions in close proximity (Figure 1A). Bumper primers are omitted from each primer set since the initiation activity accomplished by bumper primers can be replaced by neighboring primer sets. When bumper primers are omitted from a single assay, no amplification occurs, however this activity is restored with primers from neighboring assay regions (Figure S1). Further, we found that performance of a single LAMP assay is improved by using multiple nested bumper primers (Figure S1B-D), and the multiple primers of each Chain LAMP assay (Figure S1E) may similarly serve to improve sensitivity by “bumping” adjacent amplicons. Here, we use only two bumper primers flanking all 7 Chain LAMP assays, and each assay is made up of 4 primers, reducing the total number of primers from 42 to 30 primers, decreasing the primer pairwise risk for cross reactivity (Figure 1A). Each of these redundant assay target regions can generate exponential amplification, reporting with a target-specific universal displacement probe (UDP) to a single fluorescence channel.¹³ Amplification of a single region is sufficient to create a detectable amplicon, which increases total coverage of genetic sequence diversity.

Figure 1. Multiplexed Chain LAMP design.

(A) Chain LAMP design with 7 redundant assay target regions. Individual target regions contain 4 primers, omitting individual bumper primers. A single set of bumper primers flank all assays, and neighboring assays can contribute strand displacement for LAMP initiation. Multi-target redundancy enables detection even if all, but one assay, fail to amplify. (B) Conventional LAMP design with 6 total primers, 2 inner primers (FIP and BIP), 2 loop primers (LF and LB), and 2 bumper primers (F3 and B3). (C) HIV sequence diversity. Group M HIV genomes from LANL²⁸ (Group M filtered web sequences; 2022; n=4777). Coordinates are based on HXB2 reference sequence. Plot shows the proportion of sequences, weighted by global prevalence²⁷, matching the consensus for 20 bp window regions and a 100 bp rolling average over these regions. The most conserved region of the genome, the HIV-1 pol gene serves as the target for our assay. Analysis is modelled after work by Ocwieja et al.³⁸ with updated sequence library and weighting to global prevalence.

The Chain LAMP was designed to achieve global coverage of HIV sequence diversity based on multi-target redundancy. To improve individual sequence coverage for each assay target, primer sets contain degenerate bases based on alignment to an HIV consensus sequence (Table S2).¹¹ We analyzed HIV Group M reference alignments (downloaded from Los Alamos National Laboratory (LANL))²⁸ representing global diversity of HIV to identify the design target. To accommodate the assay footprint of the Chain LAMP, we focused on the most conserved region of the genome, the pol gene, as our design target (Figure 1C).³⁸

Chain LAMP Assay Design and Development.

Most published LAMP assays are designed with software tools such as PrimerExplorer³⁹ (Table S1). These tools are largely derived from early empirical design rules and do not consider factors needed for highly multiplexed RT-LAMP development. Similar to other examples of highly multiplexed LAMP, we used manual primer design processes (Table S1).^12,13 In addition, we deviated from empirical design rules for primer melt temperatures (Figure S2A) and integrated coverage and primer interaction analysis into the primer design process. See Supplementary Section S2 for a detailed analysis of the RT-LAMP design process.

The main goals of the design process were to ensure global HIV coverage while limiting non-specific amplification resulting from primer interactions.^17,18 The critical termini of primer regions, the 3’ end of F2, B2, LF, and LB and the 5’ end of F1c and B1c (Figure S3A) were vital to the design process. Although LAMP is relatively tolerant to mismatches, mismatches near the critical termini can impair both assay speed and sensitivity, with critical termini mismatches often causing complete assay failure due to inefficient polymerase binding and inability to initiate extension.^4,7–9 Non-specific amplification primarily results from interactions at these critical termini regions, as shown by a study of LAMP primer designs indicating that non-specific amplification results most strongly from a low ΔG of homo and hetero dimer at the critical termini.⁴⁰ During primer design and screening, we calculated ΔG of all primer interactions to identify and redesign primers with low ΔG interactions involving the critical termini (Figure S3-S4).³¹ With targeted iterative manual redesign we significantly improved individual assay performance and reduced interactions between assays to enable higher level multiplexing (Figure S3).

To enable high level multiplexing, we used an in-house, strand displacing TFv1 chimeric DNA polymerase (TFPol) previously shown to maintain rapid amplification and specificity in multiplexed reactions.³ For fluorescence detection we used Universal Displacement Probes (UDPs),¹³ a specific probe system that enabled specific detection even in the presence of non-specific amplification detected by double stranded DNA (dsDNA)-binding dye (Figure S6).

We were able to develop a 7-plex assay (Table S2) with no non-specific amplification detected by UDP probes within 85 minutes of reaction time (Figure S6).

Analytical Performance of Chain LAMP.

We verified functionality of each redundant target region in the Chain LAMP, with each of the 7 individual target regions generating detectable amplification with 1,000 copies of extracted RNA/reaction (Figure S7). For the multiplexed Chain LAMP assay, we showed an analytical limit of detection (LOD) of 25 copies of RNA/reaction in 35 minutes through assessments of serial dilutions of extracted RNA from a Subtype B inactivated HIV reference isolate (Figure 2A-B, Figure S8). Despite the high level of multiplexing, this LOD aligns with the sensitivity of the most sensitive HIV RT-LAMP assays (Figure 3A) and is near the LOD for the most sensitive published RT-LAMP assays (5–10 copies of RNA/reaction).⁴¹ Our LOD of 25 copies/reaction, when combined with appropriate sample preparation from a 50 μL fingerstick sample,⁴² translates to 1,000 copies of HIV RNA/mL of plasma, which corresponds to the World Health Organization (WHO)-recommended sensitivity to ensure viral suppression and negligible transmission risk.^43,44 As such, the analytical sensitivity of the Chain LAMP aligns with clinical guidelines for POC HIV viral load (VL) monitoring.

Figure 2. Chain LAMP analytical performance.

(A) Real-time UDP probe amplification signal of RT-LAMP reactions containing extracted reference HIV RNA. Reactions (n=2) detecting NTC, 10, 25, 100, 1000 copies of RNA/reaction. *Only one of two replicates of 10 copy reaction was amplified. (B) Evaluation of RT-LAMP analytical sensitivity at 25 copies per reaction of extracted reference HIV RNA (n=20). Time to result of individual reactions is plotted along with median and interquartile range.

Figure 3. In-silico RT-LAMP coverage analysis.

(A) Sensitivity and variation in sensitivity for published HIV RT-LAMP assays.^{11, 38,45–49} For assays that performed testing across subtypes, variation in sensitivity exceeds 3-log. All assays are designed in two highly conserved regions in the integrase and gag genes within close proximity. (B) Coverage analysis for Chain LAMP and published assays categorized by subtype. Coverage shown for HIV genomes from LANL (filtered web sequences; 2022; n=4777).²⁸ Coverage is defined by a ROSALIND DxM derived severity scoring.²⁴ *indicates subtypes for assays in which specific subtypes were implemented in design or testing processes (e.g. AE and B due to prevalence in China^46,48, B due to prevalence in the US and China^45,47). (C) Global coverage analysis for Chain LAMP and published assays. Coverage is defined as % of sequences for which there is a perfect match across all primer regions. Secondary coverage follows a ROSALIND DxM derived severity scoring.²⁴ Coverage shown for HIV genomes weighted by global prevalence of subtypes.²⁷

In-Silico Primer Coverage and Primer Mismatch Analysis.

Published HIV LAMP assays are often designed against a single HIV subtype^45–48 and have been shown to have variable amplification performance against multiple HIV subtypes samples (Figure 3A).^{11, 38,47} Using a custom script, we analyzed the Chain LAMP as well as published assays against a diverse alignment of HIV genomes from the LANL database,²⁸ importantly weighted to represent global sequence diversity and prevalence²⁷ (Figure S9).²³ Alignment was defined using two different methods: 1) as binary determination based on perfect match of all primers within each assay, and 2) a scoring methodology derived from ROSALIND DxM severity scoring that accounts for location of mismatches within primer regions, among other factors expected to affect severity of the mismatch (Figure S10A).²⁴ ROSALIND DxM severity scoring was developed in connection with the US FDA and NIH for a variety of amplification methodologies including RT-LAMP.^24–26 While the ROSALIND DxM derived scoring cannot reliably predict amplification for all samples, we found a statistically significant correlation (p<0.05) between scores and in vitro single assay threshold detection (Figure S12A). For this reason, we chose to use this scoring methodology for our in-silico analysis for relative comparison of coverage between assays. We defined coverage as alignment of at least one assay target within a multiplex. See Supplementary Section S4 for a detailed analysis of RT-LAMP coverage analysis.

HIV-1 Group M is made up of 9 subtypes and various recombinant forms that are increasing in prevalence.^22,27 Genetic sequence diversity within subtypes is up to 17%, while between subtypes can be as high as 35%, and even within a single infection some genes can mutate by up to 10%.^21,22 Using a ROSALIND DxM derived severity scoring, the Chain LAMP is able to maintain coverage across the predominant Group M subtypes (including A, B, C, D, F, G, CRF01_AE, and CRF02_AG) despite this diversity, with coverage above 93% for each individual subtype, while other assays showed highly variable coverage across subtypes and significant bias when one subtype was targeted in the design, as would be expected (Figure 3B). For HIV sequence alignments representing global diversity and prevalence, Chain LAMP achieved total global coverage of HIV sequence diversity above 95% while all other published HIV RT-LAMP assays had coverage below 54% (Figure 3C). The use of multiple redundant assays and degenerate bases in the Chain LAMP allows for a vastly increased in-silico coverage of HIV. Further, as HIV diversity increases over time,²⁷ coverage analysis can identify specific mismatches in an assay that can be corrected to maintain coverage (Figure S14).

Performance Against Clinical RNA Samples with Variant Sequences.

To characterize coverage of HIV diversity, we evaluated the Chain LAMP using a custom panel of 24 extracted clinical HIV RNA samples from EQAPOL isolates of diverse HIV subtypes designed to span global HIV Group M sequence diversity (including A, B, C, D, F, G, CRF01_AE, CRF02_AG, and URF) (Table S4).³⁰

The Chain LAMP amplified all clinical RNA samples. We initially tested clinical RNA with 1-log dilutions and all but one (95.8%) of the samples were detected at a threshold of 1,000 copies of RNA/reaction (Figure 4A-B). The single sample that was not detected by UDP was detectable with dsDNA-binding dye and had mismatches in the UDP probe-associated loop primer in the major amplified assay region, suggesting successful amplification but failure in probe integration (Figure S15). To achieve better LOD resolution, we performed more incremental dilutions (1/3-log) based on the results from the first test (Figure S16A) determining an LOD for the Chain LAMP with each sample (Figure 4C). Excluding the sample with identified probe reactivity issues, 22/23 (95.6%) were detected at 500 copies/reaction while 65.2% of samples were detected at 100 copies/reaction (Figure 4C-D).

Figure 4. Chain LAMP amplification of EQAPOL extracted clinical RNA samples.

(A) Example real-time UDP probe amplification signal of Chain LAMP with extracted RNA from Sample A1.3. From first round of testing, sensitivity was determined at 1,000 copies of RNA based on liftoff of duplicate samples. Second round of testing began testing at 1,000 copies to determine final LOD. (B) Liftoff time for duplicate extracted viral isolate testing in initial 1-log dilution testing. (C) Sensitivity of Chain LAMP with each viral isolate determined by sequential sensitivity testing, organized by subtype. LOD for one sample (C.3) is shown based on dsDNA-binding dye and not probe due to identified problems with probe reactivity. (D) LOD of Chain LAMP at various thresholds. One sample (C.3) excluded from analysis due to identified problems with probe reactivity.

Sample LOD varied between 21.6–1,000 copies of RNA/reaction, less than a 2-log variation (1.67 log) (Figure 4C). In a different context, for quantification in HIV VL monitoring, the FDA allows a total permissible variation of 1-log,⁵⁰ with independent testing showing up to a 2-log variation for clinical samples.⁵¹ To date, only a few HIV RT-LAMP assays have been evaluated for their ability to accommodate sequence variability. These methods use degenerate primers¹¹ or target highly conserved regions, specifically a small section of the integrase gene.^38,47 However, these approaches still exhibit significant sensitivity variations across samples and HIV subtypes, with differences in LOD exceeding a 3-log variation (Figure 3A). Further, FDA standards assess analytical inclusivity at 5X LOD.⁵² In our case with an analytical LOD of 25 copies/reaction, testing at 5X LOD (125 copies/reaction) resulted in detection of 65.2% of clinical RNA samples. This remains somewhat below the 95% detection rate that would be required by FDA but demonstrates increased coverage compared to conventional LAMP assays.

We compared clinical RNA sample coverage between the Chain LAMP and a single assay (Assay 2) that has been identified as the strongest individual assay (Figure S7). At a 1,000 RNA copy/reaction threshold, the single assay had a 62.5% coverage compared with 95.8% coverage with the Chain LAMP (Figure S17A-B). This means that the Chain LAMP multiplex provides a direct coverage benefit over a single assay.

Multiplexed LAMP Mechanisms.

As highly multiplexed RT-LAMP assays have not been extensively studied, we sought to understand some of the underlying mechanisms. We began by creating a model system, leveraging the UDP probe system to individually tag four separate assay regions within a 4-plex RT-LAMP. This revealed amplification of subtype B reference DNA for two of the target regions (Assay 2 and Assay 5) as a major and minor product (Figure S18A). These assays were also the fastest amplification regions in individual assay region testing (Figure S7), suggesting that faster assays in a multiplex will outcompete others if they are able to react. Further, when these two primer sets were omitted from the 7-plex assay, there was only a minor reduction in assay sensitivity (Figure S18B). This indicates that the multiplex is robust to omission of even the strongest target regions, validating the concept of multi-target redundancy for the Chain LAMP.

For all EQAPOL clinical RNA samples, amplicons from a reaction containing 10⁴ copies of RNA were sequenced by Oxford Nanopore sequencing (Figure 5A). Samples showed amplicons corresponding to multiple target regions with up to 5 targets present in amplicons from a single sample. This indicated multiple RT-LAMP initiation loci within each reaction. Further, in agreement with the model system (Figure S18A) many tests had a major amplicon product with a single target region dominating in sequencing frequency. In 66% of samples, Assay 2 emerged as the dominant assay, as observed in model system testing (Figure S18A), likely as it exhibited the fastest amplification when tested individually (Figure S7). In contrast, Assay 4, which exhibited the slowest amplification when tested individually (Figure S7), was the only target region not represented in any of the sample amplicons.

Figure 5. Nanopore sequencing and ROSALIND DxM derived analysis of EQAPOL extracted clinical RNA samples.

(A) Nanopore sequencing results show normalized sequencing frequency of Chain LAMP target regions (assays 1–7) in EQAPOL extracted clinical sample amplicons (10⁴ input copies of RNA/reaction). Frequency is normalized within each sample by the target region with highest individual frequency. (B) ROSALIND DxM derived analysis representing derived mismatch severity scoring for Chain LAMP target regions with each EQAPOL sample sequence. Each sample’s “Best Assay” column identifies the lowest mismatch severity across all assays, all samples showing “low” or “medium” severity indicative of strong in-silico coverage for the multiplex.

We performed in silico ROSALIND DxM^24–26 derived severity scoring analysis of all clinical RNA sample sequences with Chain LAMP primers. This analysis showed variable severity scores across assay target regions for each sample. However, for all samples there was at least one region with only “low” to “medium” ROSALIND mismatch severity (Scores: 0–3), indicating strong in silico coverage of the samples (Figure 5B). We assessed the correlation between ROSALIND DxM derived severity score and presence of amplicons in sequencing results. We found a significant correlation (p<0.05) only in the case of Assay 1 and 2 (Figure S12B), which were two of the strongest individual assays and the dominant assays determined by sequencing in 87% of samples. In the highly multiplexed system, severity of primer-target mismatches has a significant effect on target region amplification, however only for the dominant individual assays driving amplification.

Within the Chain LAMP, performance of individual assay targets was driven by both individual assay speed as well as the severity of mismatches with the target. The value of the highly multiplexed RT-LAMP assay goes beyond redundancy with many assay target regions showing initiation and amplification in the same reactions. Sequence analysis and nanopore sequencing are effective tools that can be used to understand assay behavior in the context of sequence variability and identify future potential improvements in the assay (e.g., redesigning Assay 4, which showed no representation in clinical sample testing).

Conclusions

We developed a highly multiplexed RT-LAMP assay, the Chain LAMP, to address sequence diversity for detection across HIV subtypes. The Chain LAMP had a strong analytical LOD with extracted RNA of 25 copies of RNA/reaction, aligning with the WHO clinical standard for HIV VL monitoring of 1,000 copies of HIV/mL of plasma, and no non-specific amplification. The Chain LAMP demonstrated increased in-silico coverage across Group M HIV subtypes compared to published RT-LAMP assays. We tested the assay against extracted HIV samples representing global sequence diversity and demonstrated minimal variation in sensitivity between samples with 95.6% of samples detected at a sensitivity of 500 copies of RNA/reaction. Despite strong coverage of genetic sequence diversity there is still variability in sensitivity between samples, bringing the LOD of some clinical RNA samples above clinically relevant thresholds.

Nanopore sequencing, in connection with mismatch severity analysis, enabled understanding of mechanisms underlying a highly multiplexed RT-LAMP assay. Both individual assay speed and mismatch severity affected the performance of assay regions within a multiplex, and multiple initiation loci were present in all samples. Future work will involve further mechanistic analysis with sequencing to identify improvements in assay design.

The design process made clear many of the difficulties behind designing highly multiplexed RT-LAMP assays. Commonly used software is not capable of analyzing multiplex primer interactions and cannot account for sequence diversity during primer design. We integrated analysis of primer interactions and coverage analysis into the primer design process to enable development of a highly multiplexed RT-LAMP assay. Further, we departed from standard design rules for melt temperatures and reduced the total number of primers through omission of bumper primers based on cooperativity of neighboring primer regions. This represents, to our knowledge, the highest order published single-pot multiplexed RT-LAMP assay (as of Sept. 2024). As HIV represents the most challenging case of sequence diversity for assay design for infectious diseases, this approach can be used for development of assays for other pathogens with high sequence variability such as HCV, SARS-CoV-2, and Influenza.

Supplementary Material

Supporting information:

Supporting information, consisting of supporting figures and discussion is categorized into five sections: assessment of published RT-LAMP assays and Chain LAMP primer sequences; RT-LAMP primer design considerations; Detailed Chain LAMP analytical testing; Assessment of HIV coverage considerations; Detailed Chain LAMP sample testing (PDF).