Pre-analytical considerations in the development of a prototype SARS-CoV-2 antigen ARCHITECT automated immunoassay

Clinical specimens

Remnant nasal and nasopharyngeal specimens used in the stability study were originally collected from March 30, 2020 to April 4, 2020 in a separate IRB-approved study at the Medical College of Wisconsin (Milwaukee, WI, n=18). Remnant specimens from the Abbott Rapid Diagnostics Division (ARDx, Scarborough, ME; n=25) were also used in the stability study; these specimens were originally collected under separate IRB-approved protocols at the University of Rochester (collected September 20, 2020 – May 7, 2022) and Beaumont Health System (Royal Oak, MI) and University of Florida College of Medicine (Gainesville, FL; collected September 23, 2020 – May 7, 2022). All patient specimens were de-identified prior to transfer and were processed by Abbott Diagnostics (Abbott Park, IL) for testing. Negative specimens for the stability study were pooled from the Abbott In-House Collections.

A total of 300 negative and 150 positive de-identified nasopharyngeal specimens used in the clinical performance study were obtained from Cerba Xpert (Cerba Research, Ghent, Belgium); these specimens were collected in the late spring and summer of 2021 at French national screening sites during the Delta variant surge under a separate IRB-approved protocol. FLOQSwabs Flexible mini-tip (Copan Diagnostics, Murrieta, CA) were used to collect nasopharyngeal specimens. After specimen collection, the swabs were swirled five times in Universal Transport Medium (UTM; Copan Diagnostics) without viral inactivating agent and incubated for 1 min. The swab was then pressed against the side of the tube to recover liquid and discarded. Specimens were stored at 2–8, −20, or −80 °C and transported to the central laboratory. The study was performed in accordance with the principles of Good Clinical Practice and the Declaration of Helsinki.

Assay development

Twenty-nine IgG antibodies and nine VHHs (the antigen binding domain of a camelid antibody heavy chain, also referred to as nanobodies) produced in house and five external antibodies were screened for use in the prototype SARS-CoV-2 antigen ARCHITECT immunoassay, for a total of approximately 350 antibody pairs. Initial screening was performed with the external antibodies as internal antibodies were not yet available. The initial assay format was selected based on sensitivity, specificity, and preliminary testing of positive and negative patient swab specimens in viral transport medium (VTM, CDC, Atlanta, GA) with known Ct values from PCR-based testing with the Abbott SARS-CoV-2 RealTime assay run on the m2000 instrument (Abbott Laboratories, Abbott Park, IL). A one-step assay format was found to give the best sensitivity: the specimen, capture, and detection reagent are added together, briefly incubated, then washed prior to signal generation. One of the external antibodies in the initial format was not manufacturable, so an internal antibody was substituted as the capture antibody. The final format included dual capture using two internal antibodies (Ab1 and Ab2) and dual detection using one internal antibody (Ab3) and one external antibody (Ab4) from the National Institute for Infectious Diseases in Japan found to be robust to SARS-CoV-2 variants. Figure 1 illustrates the final prototype SARS-CoV-2 antigen ARCHITECT immunoassay format [3], with chemiluminescence signals quantified as relative light units (RLUs). Extensive epitope mapping of the four antibodies predicted robust detection of various circulating SARS-CoV-2 variants, including Omicron BA.1 and BA.2 and other variants of concern, when used in the dual capture/dual detection format. A recent study also demonstrated antibody performance in detecting SARS-CoV-2 variants across the Abbott Binax, PanBio, ARCHITECT, and immunoblot assays [4].

Figure 1:

Prototype SARS-CoV-2 antigen ARCHITECT immunoassay. The assay is a dual antibody qualitative chemiluminescent microparticle immunoassay (CMIA). The assay uses a two-step reagent pipetting protocol in which the specimen, capture, and detection reagents are combined first and incubated for 18 min to allow antibody binding, followed by a 4 min soak/wash step. Two antibodies against the SARS-CoV-2 nucleocapsid protein are used for capture (Ab1 and Ab2) and detection (Ab3 and Ab4). The chemiluminescence signal is quantified as relative light units (RLUs). Nasal or nasopharyngeal swab specimens in universal transport medium (UTM) are inactivated with 0.2% Tergitol 15-S-9 based on a previous study [5]. Specimens are run at a final volume of 150 µL.

BIAcore binding kinetics

A BIAcore 4000 (GE Healthcare, Chicago, IL) was desorbed and sanitized per the manufacturer’s protocols, and a Series S CM4 Sensor Chip (Cytiva, Marlborough, MA) was conditioned with short, duplicate injections of 100 mM HCl (Sigma, St. Louis, MO), 50 mM NaOH (Sigma), 0.1% SDS (Cytiva), and 150 mM H₃PO₄ (Fisher Scientific, Hampton, NH). BIAnormalizing solution was used for hydrodynamic addressing and normalization of the conditioned chip in 1x HBS-N prior to successfully passing the instrument’s System Check with BIAtest solution (Cytiva). Mouse capture antibody reagent was immobilized onto the prepared CM4 chip using prepared Amine Coupling reagents per the manufacturer’s protocols for a Mouse antibody capture kit (Cytiva).

Purified mouse antibody (Abbott Laboratories) was diluted into Running Buffer (1x HBS-EP+ [Cytiva], 0.1% BSA [Sigma], 0.1% Carboxymethyl-Dextran [Sigma], plus an additional 1 M NaCl [Sigma]) to 7.5 μg/mL and captured onto the immobilized mouse antibody capture surface for 7 min at a flow rate of 10 μL/min. After all antibodies were captured, the flow rate was changed to 30 μL/min and the biosensor was equilibrated for 2 min at an increased flow rate. High-performance injections of purified SARS-CoV-2 nucleocapsid diluted into Running Buffer at 0.457–3,000 nM (3-fold dilution series) occurred for 3 min immediately followed by 6 min of Running Buffer alone over capture (captured antibody) and reference (no antibody) surfaces. All concentrations were injected randomly, and each concentration was tested twice with multiple blank injections (0 nM). The biosensor was regenerated with one 3-min injection of the Mouse Antibody Regeneration Solution (Cytiva).

After removing all warm-up/conditioning cycles, the double, reference-subtracted sensorgrams were fit to a 1:1 binding model accounting for mass transfer using the BIAcore 4000 evaluation software.

Specimen stability

Stability of native nasal and nasopharyngeal specimens (Medical College of Wisconsin and ARDx), individual and pooled patient specimens, native positive spiked, viral lysate, and recombinant nucleocapsid protein (rAg) in UTM (Copan Diagnostics), calibrator diluent, and negative patient pooled human nasal wash was evaluated. rAg and high positive samples (pooled from 18 nasal and nasopharyngeal specimens with an average Ct value of 23.11) were spiked into a negative nasal and nasopharyngeal specimen matrix. The pooled high positive sample yielded RLUs of >3 million and was diluted in negative matrix to reach the appropriate RLU target concentrations as a function of the signal-to-cutoff ratio (S/C), based on the 6.5 RLU value for calibrator diluent. rAg spiked in negative patient specimen matrix was stored at room temperature (RT) 15–30 °C for 6 and 20 h. Negative and high positive nasal and nasopharyngeal specimens were stored at 2–8 °C for 24, 48, 72 h, and 7 days and at room temperature (15–30 °C) for 1, 3, 6, and 24 h. To inactivate virus, specimens were pretreated for 10 min with a 50X Tergitol 15-S-9 solution in UTM diluted to a final concentration of 0.2% Tergitol 15-S-9 and 0.0016% sodium azide [5]. Specimens were run on ARCHITECT i2000SR or Beckman Coulter Access 2 (Access SARS-CoV-2 Antigen Assay) instruments after storage under various conditions. Clinical specimens stored at −20 or −80 °C were also assessed for stability and the impact of freeze/thaw on assay performance on the ARCHITECT i2000SR instrument. Stability was reported as % difference in the S/C or a positive shift in the S/C ratio. An S/C% difference of <20% was designated as stable for samples with an S/C of >1.00. An S/C shift of <0.20 for samples with an S/C of <1.00 was designated as stable.

Assay analytical performance

FDA defines the limit of detection (LoD) as the lowest concentration of SARS-CoV-2 at which 19 of 20 replicates per concentration generate a reactive test result (≥1.0 cutoff index [COI]), expressed as the immunoassay tissue culture infective dose (TCID50)/mL. The LoD for specimens in UTM, VTM, and sterile saline (0.9%) was determined for the prototype SARS-CoV-2 antigen ARCHITECT immunoassay in limiting dilution studies (two-fold serial dilution series) using an inactivated viral lysate from BEI Resources (USA-WA1/2020, Cat. No. NR-52281). We evaluated the within-laboratory imprecision of the prototype SARS-CoV-2 antigen ARCHITECT immunoassay using Clinical and Laboratory Standards Institute (CLSI, Malvern, PA) EP12-A2 as a guideline.

While the prototype SARS-CoV-2 antigen ARCHITECT immunoassay is a qualitative assay, we also assessed whether very high titer positive specimens have a high-dose hook (HDH) effect that could produce false negative results. An HDH effect occurs when a sample at an extremely high concentration falsely reads below the assay cutoff, while a more diluted sample shows higher results. We determined whether the prototype SARS-CoV-2 antigen ARCHITECT immunoassay exhibits and is critically susceptible to a HDH effect using CLSI EP34 as a guideline. HDH was evaluated with highest concentration of inactivated SARS-CoV-2 stock (Ct 13.0) available.

To determine the performance of the prototype SARS-CoV-2 antigen ARCHITECT immunoassay when a reagent kit remained on board the ARCHITECT i2000SR instrument for a minimum number of days, we conducted an in-use stability study using a classical, real-time testing protocol as defined by CLSI EP25-A. We also evaluated calibration curve stability.

Assay interference

Interference from endogenous substances was tested, including 0.5% bovine submaxillary gland mucin (Alfa Aesar, Ward Hill, MA; 5 mg/mL) and whole blood (4% v/v) negative for SARS-CoV-2 and SARS-CoV-2 antibodies (Boca Biolistics, Pompano Beach, FL). A swab containing 3 mL of a contrived positive specimen (+20% of cutoff by RLU) or 3 mL of a contrived negative specimen (−20% of cutoff by RLU) was inserted into 50 µL of each endogenous interferent solution, swirled, and incubated for 1 min. The swab was squeezed against the side of the tube and then discarded. 50X Tergitol pretreatment solution was added to each specimen at RT for a 15-min incubation prior to running on the prototype SARS-CoV-2 antigen ARCHITECT immunoassay [5].

To assess interference from drugs, a swab containing 3 mL of a contrived positive specimen (+20% cutoff by RLU) or 3 mL of a contrived negative specimen (−20% of cutoff by RLU) was inserted into 50 µL of each drug in solution, swirled, and incubated for 1 min. The swab was squeezed against the side of the tube and then discarded, and the specimen was run on the SARS-CoV-2 antigen ARCHITECT immunoassay. Swabs with 3 mL contrived positive specimen were mixed with 50 µL of each drug in solution and incubated for 1 min [6]. 50X Tergitol pretreatment solution was added to each specimen at RT for a 15-min incubation prior to running on the prototype SARS-CoV-2 antigen ARCHITECT immunoassay [5].

To test cross-reactivity, a swab was inserted into a tube containing 50 µL of each cross-reactant at its final concentration. The swab was then transferred into 3 mL of UTM, swirled five times, and incubated for 1 min at RT. Following the incubation, the swab was squeezed above the liquid line on the side of the tube and then discarded. The 50X Tergitol pretreatment solution was added to each specimen at RT for a 15 min incubation prior to running on the prototype SARS-CoV-2 antigen ARCHITECT immunoassay [5].

Clinical performance study

Specimens (Cerba Xpert) were tested in duplicate across two ARCHITECT i2000SR instruments using three calibrator lots. After arriving at the laboratory for testing, frozen specimens were thawed and all specimens were mixed by inversion 10 times. Three drops of 50X Tergitol pretreatment solution were added to 3 mL of each specimen in UTM [5], followed by gentle mixing by inversion 10 times. If present, visible precipitate was removed by centrifugation for 5 min. Specimens were incubated for 15 min at RT prior to running on the prototype SARS-CoV-2 antigen ARCHITECT immunoassay.

The sensitivity of the prototype assay run on the two ARCHITECT i2000SR instruments was determined using clinical specimens with known viral loads as reported by the vendor (Cerba Research; using the PerkinElmer^® SARS-CoV-2 Real-time RT-PCR Assay run on the BioRad CFX384 instrument). Sensitivity data for the prototype assay was compared to that of the Elecsys^® SARS-CoV-2 Ag immunoassay (Roche, Basel, Switzerland), as reported in the package insert [7].

Assay analytical performance

BIAcore was used to verify the binding kinetics of the capture and detection antibodies used in the prototype SARS-CoV-2 antigen ARCHITECT immunoassay (Table 1). TCID50/mL values were similar for specimens in UTM (5), VTM (3.75), and calibrator diluent (3.75). All 20 positive specimens with viral inactivation were positive on the prototype immunoassay (≥1.0 COI), as was inactivated viral lysate (USA-WA1/2020). Furthermore, the TDIC50 for the prototype SARS-CoV-2 antigen ARCHITECT immunoassay in UTM, VTM, and sterile saline were approximately 4-fold lower than those reported for the EUA Elecsys SARS-CoV-2 Ag immunoassay (Table 2). Within-laboratory imprecision over the 5-day study was ≤0.10 SD (<1.00 S/C) with a ≤10% CV (≥1.00 S/C). Assay reagents were stable when stored on board the instrument for 14 days, with a less than −20% shift from baseline (≥1.00 S/C). No HDH effect was observed with the highest concentration of SARS-CoV-2 stock available (Ct 13.0, RLU>1.0 × 10⁶). Endogenous interference with the assay is shown in Table 3; purified mucin (0.5%) and whole blood (4% v/v) did not interfere with assay performance (−0.06 and −0.01 S/CO, respectively). Drug interference and cross-reactivity results are shown in Supplementary Table 1 and Table 4, respectively. No interference was observed from any of the drugs evaluated, and no cross reactivity was observed.

Specimen stability

Loss of sample stability over time under various storage conditions was comparable for samples run on the prototype SARS-CoV-2 antigen ARCHITECT immunoassay or the Access SARS-CoV-2 Antigen Assay on the Beckman Coulter Access 2 instrument (Supplementary Figure 1). Positive patient specimens run under multiple conditions to simulate different sample handling and collection processes showed limited stability at RT with and without viral inactivation (Supplementary Figure 2A). Pre-inactivated specimens had a stability of <6 h at RT, and post-inactivation stability was <1 h. A single freeze/thaw cycle did not adversely impact specimen stability or antigen detection by the prototype assay, with all but one specimen remaining positive (S/C≥1.00) at RT (Supplementary Figure 2B). While purified mucin did not interfere with assay performance, the presence of mucus in positive clinical specimens was found to be a major driver of specimen instability, likely due to the activity of mucal proteases (Figure 2). Long-term storage (>30 days) at −20 °C did not lead to a significant loss of detected antigen (Supplementary Table 2).

Figure 2:

SARS-CoV-2 nucleocapsid instability. Recombinant nucleocapsid protein (rAg) was spiked into calibrator diluent, UTM alone, or a negative patient specimen panel in UTM and incubated at room temperature for the indicated times prior to running on the SARS-CoV-2 antigen ARCHITECT immunoassay. The presence of nasal mucus in the negative patient specimen panel significantly reduced nucleocapsid stability leading to a weaker assay signal (relative light units, RLUs) compared to rAg in diluent or UTM.

Assay clinical performance

Testing the positive and negative patient nasopharyngeal specimens in UTM, the prototype SARS-CoV-2 antigen immunoassay had a specificity of ≥98.5% and sensitivity of ≥89.5% across the two ARCHITECT instruments (Supplementary Figure 3). The assay cutoff signal ranged from ∼4,000 to 5,500 RLUs and was set to the mean calibrator response of ∼5,250 RLUs. All additional 150 negative specimens gave a response of <2,000 RLUs. The assay sensitivity was inversely correlated with Ct and was similar across the two instruments as well as with the Roche Elecsys immunoassay, based on values reported in the package insert [7] (Supplementary Table 3 and Figure 3).

Figure 3:

Clinical performance of the prototype SARS-CoV-2 antigen ARCHITECT immunoassay on nasopharyngeal specimens. Sensitivity across a range of Ct values was similar for the ARCHITECT immunoassay and the Roche Elecsys SARS-CoV-2 Ag immunoassay, based on the values provided in the package insert [7] (see also Supplementary Table 3).