RAPID POINT-OF-CARE DETECTION OF SARS-COV-2 BY ALVEO BE.WELL PLATFORM: ANALYTICAL AND CLINICAL EVALUATIONS

ETHICAL STATEMENT

A total of 253 clinical samples (mid-turbinate swabs) were collected in USA and United Kingdom during the pandemic seasons in 2020, two clinical sites in the US (Florida and Texas) and six sites in the United Kingdom in both Hampshire, including Basingstoke and North Hampshire Hospital, Microbiology Laboratory, Royal Hampshire County Hospital, and Basingstoke Covid-19 Community Hub, as well as in London, which includes King’s College Hospital in Denmark Hill and the South London Specialist Virology Centre of KCH NHS Foundation Trust. The protocol used for collecting and detecting the patient samples was approved by relevant local Institutional Review Boards (IRBs).

CHOICE OF SAMPLE

We designed our system for use with mid-turbinate nasal swabs (COPAN Diagnostics Inc.), as these can be collected by a healthcare provider or self-collected by patients. In analytical studies, all reactions contain pooled human nasal fluid (Lee Biosolutions) at 1% (v/v).

PRIMER DESIGN

We downloaded a reference sequence for SARS-CoV-2’s nucleocapsid (N) gene from GenBank (accession number NC_045512.2, bp 28274-29533). Using MAFFT 7 [13] and Jalview v2.11 [14], we aligned the reference N gene to full genomic sequences downloaded from the GISAID Epi-CoV database [15] and parsed all unique N gene sequences from the alignment.

Based on the N gene alignment, we used proprietary in-house software written in Python 3 (Python Software Foundation) to generate primer sets for RT-LAMP. Sequences for all LAMP primer domains (F3/B3, F2/B2, LF/LB, F1/B1) were chosen based on their ability to maintain an optimal average melting temperature (Tm) against every SARS-CoV-2 strain while never falling below a minimum effective Tm for any given strain. Our software also screened each primer set in silico to minimize unwanted secondary structures such as primer dimers (strongest allowable dimer = −9.5 kcal/mol) and hairpins (maximum allowable hairpin Tm = 50 °C).

OLIGONUCLEOTIDE AND TEMPLATE SEQUENCES

All primers were purchased from Integrated DNA Systems (IDT), as was a 700-bp double-stranded DNA target containing a fragment of the N gene sequence from GenBank Accession No. NC_045512.2. We used a MEGAscript T7 Kit (Thermo Fisher Scientific) to generate a 660-bp RNA transcript from the synthetic target. Heat-inactivated virus and gamma-inactivated virus (both SARS-CoV-2 isolate USA-WA1/2020) were obtained from BEI Resources. Genomic RNA (Hong Kong/VM20001061/2020) was obtained from American Type Culture Collection (ATCC).

AMPLIFICATION PARAMETERS FOR PROTYPE ASSAY DEVELOPMENT

In the development of the protype SARS-CoV-2 RT-LAMP assay, liquid reactions were utilized and performed on a real-time PCR instrument. We used WarmStart LAMP Master Mix from New England Biolabs (NEB). Reactions contained 50 mM Na+, 8 mM Mg++, 5.6 mM total dNTPs, F3 and B3 primers at 0.2 µM each, LF and LB primers at 1 µM each, FIP and BIP primers at 1.6 µM each, 50 U murine RNase inhibitor (New England Biolabs), 4% Ficoll-400k (Sigma-Aldrich), and 1% human nasal fluid (Lee Biosystems). Reactions were run isothermally at 65 °C for up to 60 minutes. Initial primer screening reactions and cross-reactivity studies were performed on a QuantStudio 5 thermocycler (Thermo Fisher Scientific) using EvaGreen intercalating dye (Biotium) as a reporter. This enabled rapid throughput and, based on bridging studies performed in-house, is predictive of sample reactivity on our cartridge.

CARTRIDGE STUDIES

A mixture of the reaction components, such as enzymes, dNTPs, and target-specific primers were dried directly within reaction wells on the cartridge. In the analytical limit of detection (LoD) studies, the equivalent of one swab (10 µL) of heat-inactivated SARS-CoV-2 (BEI) + nasal fluid (crude sample study) or purified genomic RNA (ATCC) + purified human DNA (Promega) were spiked into 500 µL Assay Buffer in 0.5-mL microcentrifuge vials. The mixed preparations were then pipetted into our cartridges and run in our analyzers using either 40 min at 65 °C (preliminary LoD) or 10 min at 50 °C followed by 40 min @ 65 °C (enhanced LoD study). Results were determined using either Internal Research and Investigative Software (in-house) or our be.well iOS app.

CROSS-REACTIVITY STUDIES

In accordance with FDA recommendations, viral samples were tested at reaction concentrations of ≥100k PFU/mL, while bacterial samples were tested at ≥1 million CFU/mL. If viral samples were insufficiently concentrated, extraction + concentration procedures were performed with a PureLink Total RNA/DNA kit (Thermo Fisher Scientific) to ensure that sufficient quantities of the pathogen could be added to the reaction.

IN SILICO ANALYSIS

We used in-house Python 3 scripts for all in silico analyses. For exclusivity studies, tables of k-mers for each primer length were generated from all target sequences, and the percentage of matching nucleotides (% homology) between each primer and equal-length k-mer were calculated. Reverse complements of each primer were also analyzed to account for homology in double-stranded templates. Any primer/k-mer combination exceeding 80% homology was flagged. The locations and sequences of high-homology k-mers were tracked to inspect potential amplicons.

SOFTWARE AND DATABASES

Thermocycler data were analyzed using QuantStudio Design & Analysis Software v1.5.1 (Thermo Fisher), and cartridge run results were analyzed using in-house software written in C#. Data plots and tables were produced with JMP 15 (SAS Software, Inc.), Python 3, and Microsoft Excel.

PRIMER DESIGN

The SARS-CoV-2 assay targets a segment of the nucleocapsid (N) gene in accordance with WHO detection protocols. Primers were designed with a proprietary, internally developed RT-LAMP primer design software and an alignment of all available SARS-CoV-2 sequences from the GISAID EpiCoV database [15] as of March 17, 2020. Over 100,000 candidate primer sets were initially generated, from which eight were chosen for empirical testing in a thermal cycler. Purified genomic RNA of SARS-CoV-2 was used to determine an approximate limit of detection for the primer set in thermal cycler reactions. The best primers were chosen based on amplification speed (<20 minutes for most replicates), sensitivity, and lack of false-positive amplifications.

PRIMER SPECIFICITY

To evaluate specificity, we screened our primer set in silico against a known panel of respiratory pathogens (Table 1), including common coronavirus strains (HKU1, NL63, OC43, 229E) and those most associated with severe outcomes (MERS and SARS). All instances of a primer showing ≥80% homology against any subsequence within a pathogen were flagged in accordance with FDA guidelines for in silico reactivity. While six bacterial species showed reactive homology against a single primer, this indicates a low risk for false-positive amplification due to the inability of a lone primer to amplify a target exponentially. Only SARS-CoV-1, a close genetic relative of SARS-CoV-2 (∼88% homology in the N gene), showed reactive homology against more than 1 primer.

Next, we tested a panel of these pathogens in wet reactions against our primer sets. In accordance with FDA guidelines, viral and bacterial samples were tested in the reactions at ≥100k PFU/mL and ≥1 million CFU/mL, respectively. No amplifications were seen within the 40-minute cutoff window for any of the pathogens, except for one S. salivarius replicate (∼35 min time to result, TTR) and the SARS-CoV-1 N gene plasmid. Due to the lack of homology between S. salivarius and our primers, we expect the lone amplification is the result of components in the bacterial broth. Mild cross-reactivity with the high-concentration (40 million C/mL) SARS-CoV-1 plasmid is likely the result of homology between SARS-CoV-1 and SARS-CoV-2 but we regard this as presenting no risk, as SARS-CoV-1 has not been identified in humans since 2004 [16]. All samples included 1% (v/v) human nasal fluid, which contains human DNA as well as genetic material from a diverse panel of respiratory microbes. The lack of amplification suggests that our assay will not cross-react with genetic material normally present in respiratory samples.

INCLUSIVITY OF PRIMER SETS

To predict inclusivity, we used computer modeling of our primer set against a multiple sequence alignment (MSA) of SARS-CoV-2 genomes from the GISAID database [15] on December 30, 2023. We aligned each primer to each genome and calculated the respective Tm (melting temperature) using nearest-neighbor thermodynamic data from the SantaLucia lab [17]. A genome was considered detectable so long as all primers met predefined Tm thresholds. Of the >1.8 million sequences in the GISAID MSA, all met or exceeded the thresholds, indicating that our assay can detect all strains of SARS-CoV-2 currently in circulation. This includes highly infectious variants of concern such as Alpha, Delta, and Omicron, as their hallmark mutations are primarily in a region (the spike protein) that our assay does not target [18,19].

LIMIT OF DETECTION STUDIES

Single-Step Heating LoD

We used simulated crude samples to conduct two limit of detection (LoD) studies with the Alveo be.well COVID-19 test. LoD is defined here as the lowest sample viral load that will be detected 95% of the time. In the first study, cartridge runs were performed for 40 min at 65°C, though positive results were determined in <25 min in most cases. To simulate unpurified samples assayed directly from patient swabs, varying concentrations of heat-inactivated SARS-CoV-2 virions were added to assay buffer containing 1% (v/v) human nasal fluid. These contrived samples were then loaded into cartridges and assayed. An initial 3-fold dilution series was performed with 3 replicates at each virus concentration. The highest concentration in the series without a 100% detection rate was 1,300 copies/sample (C/sample). We used the next highest concentration in the series (4,000 C/sample) as a starting point for a series of increasingly concentrated dilutions at up to 20 replicates each. Beginning our increasing series at 4,000 C/sample, detection rates at each concentration were found to be between 70% and 86% until we reached a 100% detection rate at 17,000 C/sample (Figure 4).

• Download figure
• Open in new tab

Figure 4.

Result times for each cartridge in the single-step heating LoD study. Each point represents the time to result from one cartridge. Empty circles at 40 min are placeholders for cartridges that failed to amplify. 17,000 copies/sample is the lowest concentration at which ≥95% of cartridges amplified. Average time to result was 14.1 min.

Multi-Step Heating LoD

The second LoD study was designed to improve rehydration of dried reagents and increase the efficiency of the reverse transcription step before proceeding to exponential DNA amplification. In this study, cartridges were held for 10 min at ambient conditions (approx. 21 °C) and then for 10 minutes at 50 °C before increasing the temperature to 65 °C for LAMP. Prior to testing positive samples, 20 blank cartridges were run to determine the appropriate cutoff time for the assay. Using 5 cartridges per concentration, we ran a twofold dilution series beginning with 8,000 C/sample down to 250 C/sample. We achieved 100% detection (5 out of 5 cartridges) at all concentrations except for 250 C/sample, where 3 out of 5 cartridges detected SARS-CoV-2 virus. To determine the actual LoD, we then ran 500 C/sample and 2,000 C/sample at 20+ cartridges per concentration. Using an assay cutoff time of 30 min at 65 °C, we achieved a hit rate of 85% at 500 copies/sample (equivalent to 1,000 C/mL) and 95% at 2,000 copies/sample (equivalent to 4,000 C/mL) (Figure 5). If our 65 °C cutoff time were 20 min instead of 30 min, we would still detect 95% of cartridges at 2,000 C/sample using the same assay run time as in the initial, one-step LoD study. This value corresponds to 4,000 C/mL in the buffer tube and implies that our system can detect approximately 100 copies in each 25 µL reaction well (see Figure 5) 95% of the time. This LoD is on the same order as that of the CDC’s EUA-authorized RT-PCR kit (LoD = 1,000 C/mL) [20] and without requiring sample extraction or a thermocycler.

• Download figure
• Open in new tab

Figure 5.

Result times for each cartridge in the multi-step heating LoD study. Each point represents the time to result from one cartridge. Empty circles at 50 min are placeholders for cartridges that failed to amplify. 2,000 copies/sample is the lowest concentration at which ≥95% of cartridges amplified. Average time to result was 31.4 min. These are clinically significant concentrations owing to patients shedding below approximately 100,000 copies/swab generally being noninfectious [2,21].

CLINICAL EVALUATION

To evaluate the clinical performance of the be.well COVID-19 assay, two IRB-approved prospective clinical trials were conducted during September 2020 through December 2020, with one trial taking place in the United States and the other in the United Kingdom. For the US clinical evaluation study, patient samples were collected from two clinical sites (Florida and Texas) using the Copan mid-turbinate swabs for both Alveo be.well COVID-19 test and comparator PCR test authorized by FDA. Both samples were shipped to a testing site in California after collection and tested with the be.well COVID-19 test and the EUA authorized Abbott RealTime SARS-CoV-2 Assay performed on the Abbott m2000 System. The be.well COVID-19 test results were compared to results from the Abbott RealTime SARS-CoV-2 Assay to determine the clinical sensitivity and specificity and the 95% confidence interval (CI). The positive percent agreement (sensitivity) was calculated to be 94% (95% CI, 79.2% to 99.2%) with a negative percent agreement (specificity) of 97% (95% CI, 92.7% to 99.2%) as shown in Table 2.

The prospective clinical performance evaluation was also carried out in multiple sites in Hampshire and London of the United Kingdom. Testing was spread across four sites in Hampshire (Basingstoke and North Hampshire Hospitals, Microbiology Laboratory, Basingstoke and North Hampshire Hospitals) and 2 sites in London (King’s College Hospital and Denmark Hill London). Two samples were collected from all the patients. The first sample was taken from the nostril of the patient using Copan pediatric flocked swabs for the Alveo be.well Test according to the instructions of the assay. The second sample was collected from the nose and throat using a nasopharyngeal adult swab and preserved in viral transport media. The first sample was then tested immediately in the be.well COVID-19 test on site, while the second sample was taken to the local laboratory for PCR testing. The Altona RealStar® SARS-CoV-2 RT-PCR Kit 1.0 with an LOD of 100 copies/mL was used as the comparator instruments in London, while the CERTEST VIASURE SARS-CoV-2 Real Time PCR Detection Kit with an LOD of 20 copies/reaction was used in Hampshire. After the study, the PCR results were combined from both London and Hampshire sites for comparison to the be.well COVID-19 results. Ct values for positive specimens in the Abbott assay ranged from 16-45. The combined clinical sensitivity (positive percent agreement) and specificity (a negative percent agreement) was demonstrated to be 94% (95% CI, 84.6% - 97.5%) and 95% (95% CI, 93.2%-98.4%), respectively (Table 2).