Accelerated Development of a COVID-19 Lateral Flow Test in an Academic Setting: Lessons Learned

Conspectus

The COVID-19 pandemic further demonstrated the need for usable, reliable, and cost-effective point-of-care diagnostics that can be broadly deployed, ideally for self-testing at home. Antigen tests using more-detectable reporter labels (usually at the cost of reader complexity) achieve better diagnostic sensitivity, supporting the value of higher-analytical-sensitivity reporter technologies in lateral flow.

We developed a new approach to simple, inexpensive lateral flow assays (LFAs) of great sensitivity, based on the glow stick peroxyoxalate chemistry widely used in emergency settings and in children’s toys. At the peak of the COVID-19 pandemic, we had the opportunity to participate in the pandemic-driven NIH Rapid Acceleration of Diagnostics (RADx) initiative aiming to develop a deployable lateral flow diagnostic for SARS-CoV-2 nucleoprotein based on our novel glow stick-inspired light-emitting reporter technology. During this project, we screened more than 250 antibody pairs for analytical sensitivity and specificity directly in LFA format, using recombinant nucleoprotein and then gamma-irradiated virions spiked into negative nasal swab extracts. Membranes and other LFA materials and swabs and extraction reagent components also were screened and selected. Optimization of conjugate preparation and spraying as well as pretreatment/conditioning of the sample pad led to the final optimized LFA strip. Technology development also included optimization of excitation liquid enclosed in disposable droppers, design of a custom cartridge and smartphone-based reader, and app development, even a prototype reader usable with any mobile phone. Excellent preclinical performance was first demonstrated with contrived samples and then with leftover clinical samples. Moving beyond traditional academic focus areas, we were able to establish a quality management system (QMS), produce large numbers of customized LFA cassettes by contract injection molding, build in-house facilities to assemble and store thousands of complete tests for verification and validation and usability studies, and source kitting/packaging services and quality standard reagents and build partnerships for clinical translation, regulatory guidance, scale up, and market deployment. We were not able to bring this early stage technology to the point of commercialization within the limited time and resources available, but we did achieve strong proof-of-concept and advance translational aspects of the platform including initial high-performance LFAs, reading by the iPhone app using only a $2 plastic dark box with no lens, and convenient, usable excitation liquid packaging in droppers manufacturable in very large numbers.

In this Account, we aim to provide a concise overview of our 18-month sprint toward the practical development of a deployable antigen lateral flow assay under pandemic conditions and the challenges and successes experienced by our team. We highlight what it takes to coach a technically savvy but commercially inexperienced academic team through the accelerated translation of an early stage technology into a useful product. Finally, we provide a guided tutorial and workflow to empower others interested in the rapid development of translatable LFAs.

Key References

• Brosamer K.; Kourentzi K.; Willson R. C.; Vu B. V.. Glowstick-Inspired Smartphone-Readable Reporters for Sensitive, Multiplexed Lateral Flow Immunoassays. Commun. Eng. 2023, 2 ( (1), ), 31. 10.1038/s44172-023-00075-2 .¹Demonstrated a novel LFA platform using the shelf-stable, inexpensive chemistry of glow sticks to chemically excite standard fluorescent nanoparticle reporters. Emitted light is readable in color multiplex by an unmodified smartphone camera and achieves fluorescent-reporter levels of sensitivity.

• Paterson A. S.; Raja B.; Mandadi V.; Townsend B.; Lee M.; Buell A.; Vu B.; Brgoch J.; Willson R. C.. A Low-Cost Smartphone-Based Platform for Highly Sensitive Point-of-Care Testing with Persistent Luminescent Phosphors. Lab Chip 2017, 17 ( (6), ), 1051–1059 10.1039/C6LC01167E .²Demonstrated a simple, low-cost smartphone-based detection platform for highly sensitive luminescence imaging readout of point-of-care tests using persistent luminescent phosphors as reporters.

• Lei R.; Vu B.; Kourentzi K.; Soomro S.; Danthanarayana A. N.; Brgoch J.; Nadimpalli S.; Petri M.; Mohan C.; Willson R. C.. A Novel Technology for Home Monitoring of Lupus Nephritis That Tracks the Pathogenic Urine Biomarker ALCAM. Front. Immunol. 2022, 13, 1044743. 10.3389/fimmu.2022.1044743 .³Demonstrated a prototype device to noninvasively assess lupus nephritis activity at the point of care or even at home: a highly sensitive LFA coupled with smartphone-based imaging to detect lupus nephritis-specific biomarkers in urine.

1. Introduction

The COVID-19 pandemic emphasized the urgent need for user-centered, reliable, and cost-effective point-of-care diagnostics that can be broadly deployed, ideally for self-testing at home. The Willson laboratory at the University of Houston focuses on the development of sensitive, translatable analytical methods for point-of-care diagnostics, especially tests using novel reporter technologies readable by a smartphone, and had developed an early proof-of-principle of novel LFA technologies potentially applicable to this need.⁴⁻¹⁰

The COVID-19 pandemic led to the NIH Rapid Acceleration of Diagnostics (RADx) initiative to provide (milestone-based) incentives to technology developers to rapidly de-risk their technologies (which started at varied technology readiness levels¹¹) and move to translation and commercialization to increase U.S. testing capacity and more effectively manage the pandemic. A central element of this program was NIH-provided access to expert consultants to help address the scientific, technological, clinical, regulatory, and commercialization needs of each developer. Moreover, RADx served as a direct conduit to FDA, accelerating tests’ regulatory approvals. Clip Health, a previous Willson laboratory spinoff based on phosphorescent nanoparticles,⁴ was selected for the RADx program and achieved FDA EUA for their antigen LFA. The Willson UH laboratory was selected for the RADx program to de-risk a new approach to simple, inexpensive lateral flow assays (LFAs) of very great sensitivity, inspired by the glow sticks widely used in emergency settings and in children’s toys (as recently described¹).

In this Account, we provide a narrative of our 18-month sprint toward the practical development of a deployable antigen LFA under pandemic conditions and the challenges and successes experienced by our team. We provide a guided workflow to empower others interested in the rapid development of translatable diagnostics. We also discuss various translational specifics (usually considered only after typical academic assay development) that we, a technically savvy but commercially inexperienced academic team, had to navigate to translate our early stage technology to a useful product. Finally, we discuss the lessons learned. We were not able to bring this early stage technology to the point of commercialization within the limited time and resources available, but we did achieve strong proof-of-concept and advanced translational aspects of the platform, including initial high-performance Glow LFAs, read by an iPhone using only a $2 plastic dark box with no lens, and convenient excitation liquid packaging.

2. Glow LFA: From Inception to a Manufacturable Product

Antigen tests using more-detectable reporter labels (usually at the cost of reader complexity) achieve better clinical sensitivity, supporting the value of higher-analytical-sensitivity reporter technologies in lateral flow. Many of the most-sensitive diagnostic platforms use light-emitting labels. Unfortunately, most LFA formats of sufficient sensitivity are poorly suited for widespread home self-testing, including organic fluorophores,¹²⁻¹⁷ europium chelates,^18,19 up-converting phosphors,²⁰⁻²² and quantum dots²³⁻²⁷ that offer high analytical sensitivity but require a complex apparatus for excitation/detection^28,29 and/or are unstable, photobleachable, or require narrow excitation/emission wavelengths.³⁰ Enzyme-based chemiluminescent LFAs can provide ultrasensitive analysis⁵ but require often-unstable reporter substrates.

Inspired by the underlying technology of inexpensive, shelf-stable glow sticks,³¹ we have introduced chemi-excited “glow” LFA reporters—fluorescent nanoparticles chemically excited to emit light—that can be readily imaged using a smartphone camera.¹ When standard plastic glow sticks are bent, an inner glass ampule is broken, releasing 3% hydrogen peroxide + cosolvent that reacts with a diphenyl oxalate such as bis-2,4,6-(trichlorophenyl)-oxalate (TCPO) stored outside the ampule, producing the unstable 1,2-dioxetanedione intermediate (a strained cyclic dimer of CO₂; Figure 1).^32,33 When this highly reactive intermediate collides with certain fluorescent dyes (included in one of the liquids in a glow stick), energy is transferred to the dye and chemically excites it. The emission wavelength depends on the dye used, making it possible to excite glow sticks (and fluorescent LFA reporter particles; Figure 1) to emit in a variety of colors using the same activating chemicals for color-multiplex assays.

Figure 1.

(Top) Mechanism of the TCPO-H₂O₂ chemical excitation of glow sticks. (Bottom) Application of glow stick chemistry to the conventional lateral flow assay (LFA). LFA involves capillary wicking of a sample along a dry, porous membrane to a line bearing antibodies specific for a target analyte. In the presence of target analyte, fluorescent reporter particles bearing analyte-specific detector antibodies are sandwich-bridged to the membrane-bound capture antibodies and accumulate in lines to indicate a positive test. Application of glow stick chemiexcitation reagents makes the particles glow brightly with no excitation light or optical filters (adapted from ref (1)). This article is licensed under a Creative Commons Attribution 4.0 International License http://creativecommons.org/licenses/by/4.0/.

Our novel chemi-excited LFA format can achieve fluorescent-LFA sensitivity without the use of expensive fluorescence optics or any component which is slow or costly to manufacture. The Glow LFA workflow is identical to a gold nanoparticle LFA until the last step when (shelf-stable) excitation reagents (∼50 μL) are dispensed onto the LFA using a disposable dropper containing stable (crushable) glass ampules like those in a glow stick (Figure 1 or from a storage blister on the cassette). Early in the pandemic, when the need for widespread affordable testing became urgent, it became clear that the Glow technology potentially could be the basis of a clinically valuable diagnostic, for which the instructions for use are shown in Figure 2.

Figure 2.

Optimized, smartphone-read Glow LFA sample-to-answer workflow for the detection of SARS-CoV-2 nucleoprotein in nasal swab samples as shown in the instructions for use (complete instructions included in Supporting Information, SI3).

3. Development of the SARS-CoV-2 Glow LFA Prototype

To translate a proof-of-concept Glow LFA prototype to a readily deployable COVID-19 POC diagnostic under time pressure, we followed a parallel rather than a sequential approach, in which all parts were worked on simultaneously given the constraints of an academic environment (e.g., limited personnel, funds, instrumentation) along with those imposed by the pandemic (e.g., supply chain shortages, meeting restrictions, and laboratory operating constraints). We will discuss the major technical activities performed in the development of the translatable Glow LFA (Figure 3).

Figure 3.

Major technical components of the development of the Glow LFA toward a translatable, point-of-care SARS-CoV-2 rapid antigen test.

3.1. Antibody Screening

The performance of immunoassays depends critically upon the antibodies used, and thus, antibody screening is a key part of any immunoassay development. We screen antibodies in (dynamic) LFA rather than equilibrium-based ELISA (which often is not reliably predictive of LFA performance). Antibodies may compete for closely adjacent or even overlapping epitopes and may not be able to bind simultaneously to the analyte. Additionally, conformational changes induced in the analyte by detection antibodies may modify the affinity of (downstream) capture antibodies. Antibodies are immobilized either through passive adsorption on the nitrocellulose membrane (for capture antibodies on the test line) or via covalent conjugation onto reporter particles (for detection antibodies), targeting distinct surface functionalities. The specific immobilization method will differentially impact the conformation, orientation, and presentation of each antibody, ultimately affecting the preservation of active conformation and steric accessibility of antigen-binding sites. Thus, it is difficult to predict a priori the performance of various antibodies, complicating the rational selection of antibodies and forcing LFA developers to use a trial-and-error approach.

We acquired anti-SARS-CoV-2 nucleoprotein antibodies based on vendor reputation, consultant recommendations, and immediate availability. ELISA screening data and EC50 values were available for some of these, but no detailed binding³⁴ or epitope mapping data were available at that time. (A later study provided epitope mapping of 17 antibodies used in 11 EUA’d antigen tests as well as their susceptibility to mutational escape.³⁵) As new antibodies became commercially available, additional screening rounds were performed. We primarily screened monoclonal antibodies for their long-term reproducible supply. Overall, more than 250 antibody pairs were screened for binding to SARS-CoV-2 nucleoprotein (ACRO Biosystems, NUN-C5227; expressed in HEK293 cells and originally evaluated by Cate et al.³⁶). We used 3-mm-wide half-strip LFA strips encompassing only the nitrocellulose membrane and the absorbent pad. This format is frequently used for the initial LFA development. Others have developed a robotic-assisted platform for the efficient antibody screening on LFA strips.^36,37 The nucleoprotein is a highly conserved and abundantly expressed structural protein and the target of most FDA-cleared antigen tests. Antibodies were conjugated to commercial 200 nm blue fluorescent carboxylated particles (excitation 365 nm/emission 415 nm, ThermoFisher Fluorospheres F8805) using EDC/NHS conjugation chemistry¹ after we tested a variety of particles and confirmed their compatibility with Glow chemi-excitation. The same antibodies were also printed onto nitrocellulose (using the BioJet dispensers of a Biodot XYZ3060 system). Protein A (Arista, AGPTA-0101) was printed on the control lines. Candidate antibody pairs were first screened at 0 ng/mL and a high concentration (10 ng/mL) of the nucleoprotein. The strips were imaged with a CoolSNAP K4 CCD in a FluorChem SP gel cabinet (Alpha Innotech Corp., San Leandro CA) controlled by Micro-Manager software (https://micro-manager.org/). Antibodies were ranked based on the image brightness of the test line of the positive and negative tests (Figure 4). Antibody pairs were eliminated if high nonspecific binding (bright test line at 0 ng/mL) or low binding (no test line or a dim test line at 10 ng/mL) occurred. Top-performing pairs were tested again at 1, 0.5, and finally at 0.25 ng/mL of nucleoprotein. Initial screening was performed in buffer but later in a pooled nasal swab extract (a clinically relevant matrix). The top-performing antibody pair was finally chosen based on the presence of a bright test line signal at 0.25 ng/mL nucleoprotein and low nonspecific binding at 0 ng/mL nucleoprotein. The chosen pair was ExonBio, NP_3E6, detection antibody covalently conjugated via EDC-NHS chemistry to carboxylated particles (Thermo F8805) and ExonBio, NP_12E6, capture antibody immobilized on a strip.

Figure 4.

LFA-based antibody screening was assessed by the brightness of the test line. The asymmetry of this results matrix around the diagonal illustrates the dependence of the performance of a given antibody on its environment - on polystyrene particles (carbodiimide-based covalent conjugation) or on the nitrocellulose membrane (adsorption via hydrophobic and electrostatic interactions³⁸). All anti-SARS-CoV-2 nucleoprotein antibodies tested are listed in the Supporting Information (SI1).

3.2. Transition to the Complete LFA Strip

The development of a complete LFA strip requires the optimization of sample and conjugate release pads and the development of conditions for spraying and drying the conjugated reporter particles onto the conjugate release pad. Multiple conjugate diluents were screened by spotting the particles onto a commonly used fiberglass conjugate release pad (Ahlstrom, #8964; also used in the final product) and identifying/optimizing components and concentrations to ensure good release and flow of the reporter particles, a strong positive signal, and low nonspecific binding using a basic LFA running buffer (also used in antibody screening; the composition was based on our previous experience and expert advice): 1× PBS (pH 7.4), 0.5% BSA (blocking agent to limit nonspecific binding between antibodies), 0.5% Tween-20 (surfactant to prevent particles from sticking), and 0.3% PEG3350 (polymer to prevent particle aggregation). The composition of the best-performing conjugate diluent was identified as 50 mM Tris, 1% casein (Sigma, C5679; blocking agent), 10% sucrose (Sigma, S9378), and 5% trehalose (Sigma, T9531) (the sugars are typically used to maintain protein stability and aid rehydration³⁹), pH 8.5. Particles were transferred to the diluent and sprayed at (an optimized concentration of) 0.15% solids onto conjugate pads using the AirJet dispenser of a Biodot XYZ3060 system and left to completely dry (overnight, room temperature, <20% RH; low humidity is important here) before LFA assembly.

The next step was testing with the appropriate sample matrix for intended use. While saliva was identified as the most desirable sample type in WHO’s early target product profiles for COVID-19 diagnostics,⁴⁰ based on the advice of our RADx team we early on decided to focus on the nasal swab extract to reduce complexity and build upon accumulated prior knowledge. The transfer of analyte from the nasal swab to the LFA requires the use of an extraction/running buffer that extracts the virions from the nasal swab, lyses them to liberate the nucleoprotein analyte, adjusts sample pH to optimize antibody binding, minimizes nonspecific binding, dilutes matrix interferents, and controls liquid flow. Using analyte spiked in nasal swab extract, we screened nitrocellulose membranes for maximum sensitivity while keeping the assay time below 15 min; the Sartorius CN140 membrane was chosen. Various sample pads were screened with spiked nasal swab extract for rapid absorption of the liquid sample and consistent flow distribution to the conjugate pad; cotton Ahlstrom 222 was chosen (cotton also achieves some filtration of the viscous nasal swab extract). All materials chosen are compatible with large-scale reel-to-reel manufacturing of LFA strips. As suggested by our translation consultants, some components of the LFA buffer were dried down on the sample pad to reduce extraction reagent complexity, improve reagent stability, and simplify manufacturing. In summary, prior to LFA assembly, the sample pad was pretreated with 100 mM sodium phosphate buffer, 100 mM NaCl, 1% BSA (Sigma A7906), 1% polyvinylpyrrolidone (Sigma, PVP-40) and 0.5% poly(acrylic acid) (Sigma, 323667; polymers to minimize particle aggregation), and 0.5% TERGITOL (Sigma, NP40S; nonionic surfactant used in immunoassays; may inactivate SARS CoV-2⁴¹), pH 7.2. Treated sample pads were dried in a forced air convection oven at 50 °C for at least 1 h and then stored at room temperature at RH < 20% at least overnight before LFA assembly.

To simplify delivery of the glow excitation liquid, we incorporated into the extraction/running buffer the hydrogen peroxide (H₂O₂, one of the components of the glow excitation liquid, see the next section). Formulations were tested using pooled, postdiagnostic PCR, leftover positive nasal swab extracts donated by local clinics and clinical laboratories. The optimized LFA buffer (500 μL in extraction tubes with dropper tips in the final test kit) was 100 mM sodium phosphate buffer, 100 mM NaCl, 0.5% IGEPAL CA-630 (Sigma, I8896; nonionic surfactant that enables virus lysis), 0.25% TERGITOL, 0.03% EMPIGEN BB (lauryl betaine; Sigma, 30326; amphoteric surfactant that enables virus lysis), 2.5% v/v H₂O₂, pH 7.4. There were careful discussions of the potential toxicity of IGEPAL if a child were to drink the extraction buffer in a home test setting; the amount used was deemed to be safe. Given the accelerated timeline, we performed a focused/limited screening and optimization of specific components and concentrations in order to achieve design freeze within the available time rather than an exhaustive exploration of all available materials.

3.3. Glow Excitation Liquid

Glow technology (aka the peroxyoxalate chemistry of common glow sticks) requires that separately stored oxalate dissolved in an organic solvent and hydrogen peroxide in a cosolvent be mixed to form the activating intermediate dioxetane that chemically excites nearby fluorophores, leading to light emission. Thus, optimization of the formulation, packaging, and application of the glow liquid was key in the development of a deployable Glow LFA. As we recently reported,¹ we screened various oxalate solvents and cosolvents to minimize toxicity and odor (http://www.thegoodscentscompany.com) and maximize glow signal strength and stability. In the initial prototype, the glow excitation liquids (component A: 15 mM TCPO dissolved in 33.3% butyl benzoate and 66.6% tributyl citrate; component B: 45 mL of tert-butanol, 5 mL of 30% H₂O₂ (3% final), and 1 mM tetrabutylammonium hydrogen sulfate) are stored in two glass ampules in a plastic dropper and activated by bending to break the ampules like a glow stick (Figure 7A,B). After the ampules are broken and mixed, the resulting excitation reagent is most active for only ca. 10 min, which is inconvenient for a practical multiple-testing workflow. We achieved stability (>8 h) of open droppers by formulating peroxide into the LFA buffer (see below) and storing only the oxalate solution in an ampule in the dropper (Figure 7C). The dropper tip size was iteratively optimized for consistent application of excitation liquid to the strip (ensuring a uniform liquid layer and avoiding any air pockets), and we demonstrated assay tolerance of substantial variation in the number of drops added (9–14 work equivalently).

Figure 7.

Packaging of Glow reagents. (A) Dropper prototype for Glow reagent storage, mixing, and delivery to LFA strips. (B) Commercial dropper tip assembly with crushable glass ampules to be filled with custom Glow reagents (James-Alexander Corp.). (C) Glow reagent packaged in a single-ampule glowstick form factor with a modified tip (Cyalume). (D) Concept of a cassette with built-in blister pouches sealed by breakable pinch valves. (E) Future design of the Glow cassette with onboard storage of Glow reagents in breakable blisters allowing one-push delivery to the LFA strip, limiting user exposure to reagents and simplifying the assay workflow.

3.4. Glow LFA Performance

During prototype development, we evaluated the analytical sensitivity of the Glow LFA prototype with commercial SARS-CoV-2 recombinant nucleoprotein (preliminary LoD was estimated at 62 pg/mL; Figure 5A) and gamma-irradiated SARS-CoV-2 virus (BEI, NR-52287) spiked in nasal swab extracts self-collected by laboratory members (asymptomatic and negative by EUA’d LFA for COVID-19; UH IRB: STUDY00002547). The preliminary LoD was estimated at 280 TCID50/mL of the gamma-irradiated SARS-CoV-2 virus in the pooled nasal swab extract (Figure 5B). The analyte concentration of 280 TCID50/mL was further tested in five replicates to confirm the LoD (Figure 5C). We adapted the Yale SalivaDirect PCR assay,⁵ which circumvents RNA extraction, and performed RT-PCR in house directly on nasal swab extracts to benchmark serially diluted pooled samples during LFA optimization and to check occasional fresh nasal specimens. There is no general consensus on the required analytical sensitivity to ensure clinical utility, but our Glow LFA was described by RADx as being (at that time) “the second most sensitive LFA” in the program, with the most sensitive being a commercial test based on the glow-in-the-dark nanophosphors previously developed in the Willson laboratory, with a reported LoD of 82 TCID50/mL.^42,43

Figure 5.

Detection of (A) the SARS-CoV-2 nucleoprotein (ACRO Biosystems, NUN-C5227) and (B) the gamma-irradiated SARS-CoV-2 virus (BEI, NR-52287; lot 70039067) spiked in a nasal swab extract from subjects presumed negative for COVID-19 infection. The Glow LFA strips were imaged inside a 3D-printed dark box with a Samsung Note 8 camera (8 s exposure). Intensity profiles along the length of the LFA strips were extracted, and values for the integrated area under the curve for each peak (TL and CL) are shown in the tables below the LFA strips (inverted blue channel images; bright area at the bottom of the strips is the overlap between the conjugate pad and membrane with accumulated nonreleased reporter particles). (C) Confirmation of preliminary LoD by testing replicates of negative and low positive analyte (280 TCID50/mL) spiked in a pooled nasal swab extract (average (TL/CL) ± 1 SD, n = 5; negative 0.01 ± 0.006 and positive 0.07 ± 0.007). (D) Evaluation of clinical sensitivity by testing 40 PCR-positive leftover samples from LabCorp; the inset shows samples with 25 < Ct < 30. The cutoff was determined with 10 PCR-negative samples (see SI2).

Clinical sensitivity was further evaluated by testing 55 leftover, deidentified clinical samples (Supporting Information, SI2; 40 positive and 15 negative) purchased from LabCorp (collected as part of routine clinical care during 2020 and 2021 and provided along with cycle threshold value (Ct) values but with no information on infection status), the only samples readily available without establishing a de novo prospective collection protocol. These samples were supplied in liquid form, requiring modification of the dry swab extraction-based Glow LFA testing protocol. We developed a modified extraction protocol for these samples, mixing a 100 μL liquid sample with 25 μL of a 5× concentrated LFA extraction/running buffer. The Glow LFA correctly identified 100% of the samples with high viral loads (Ct < 25; Ct is inversely related to the viral load) and overall 91% of the positive samples with Ct < 30, the latter of which is considered to be the threshold for infectivity,²⁸ above which patients are considered noninfectious because there is such a low viral load in the sample (Figure 5D). All negative clinical samples were correctly identified (see SI2). Cross-reactivity studies were not required at this stage. To formally establish the cutoff, receiver operating characteristic (ROC) curve analysis needs to be performed. ROC curve analysis portrays the trade-off between specificity (false positive calls; type I error) and sensitivity (false negative calls; type II error). The 2021 FDA EUA guideline required antigen tests to demonstrate a minimum sensitivity of ≥80%, and previous studies in symptomatic people reported LFA sensitivities of around 85%. High specificity (>90%) has been reported for rapid COVID-19 antigen tests.

As COVID-19 became common in our area, we regretted not having established a general consent-and-test-upon-diagnosis protocol. However, we collaborated with a local hospital for specimen collection from hospitalized patients by a healthcare provider under an IRB protocol. Despite our promising results, we later realized that samples taken from inpatients who required nasal cannulas for oxygenation (drying the inside of the nose) were not representative of nasal swab samples typically collected in clinics or drive-through testing locations, which are the intended settings for our test. The two sample types showed significant differences in flow that ultimately affected the test performance.

3.5. Development of a Custom Glow LFA Cassette and Smartphone Reader

The initial development of Glow LFA was based on imaging multiple strips with a Samsung Note 8 and a 3D-printed dark box (coordinates available at https://www.thingiverse.com/thing:6462440). Images were taken using the built-in Samsung camera Pro mode and analyzed using in-house-developed software¹ (written in Python 3.6; deposited at https://github.com/willsonlab/LFA-Image-Analysis). We eventually developed an inexpensive (materials cost $0.96) 3D-printed, single-test LFA reader attachment (coordinates deposited at https://www.thingiverse.com/willsonlab/designs) for the iPhone XR to position the Glow LFA cassette and provide a light-tight imaging environment for reading without interference from ambient light. Initially, we incorporated a plano-convex macrolens to improve LFA imaging by allowing close-up focus, more efficient light collection, and sharper imaging. Although macrolenses are generally inexpensive ($10 each in quantities of 50), time pressure and the cost limitations of the intended use case suggested removal of the lens. We confirmed that the analysis app could work with the no-lens Glow images without sacrificing the sensitivity of the test. The shape of the first version of the attachment (Figure 6A) proved to be not very user-friendly, as it required the user to apply extra pressure to insert the phone into the attachment without being able to easily assess whether the phone was fully inserted into the reader. The second optimized version of the attachment (Figure 6B) requires only placing the phone onto it, and it holds the phone level. The reader attachment can easily be customized for a variety of phones, and the iPhone XR reader can be molded in large numbers for less than $1 each.

Figure 6.

(A) Early working prototype of the smartphone reader and (B) final, optimized, user-friendly smartphone reader. (C) CAD drawing of the custom LFA cassette. (D) Blowout model of the custom, injection-molded black plastic Glow LFA cassette housing the LFA strip.

In parallel, we designed a custom, black, Glow-centric LFA cassette (Figure 6C,D), an activity not usually included in an academic project, with the expert help of Symbient. The cassette includes many engineered features (e.g., offset pressure bars centered between overlapped sections of the strip to hold strip components together, round-on-hex interlocks to prevent assembly problems, sample well seals, and a raised support surface under the nitrocellulose membrane) and not only provides housing for the LFA strip but also enhances test performance by regulating sample application and flow, imaging, and result interpretation. While there are many off-the-shelf, inexpensive cassettes available, LFA performance can be substantially improved by optimizing cassette design for the specific strip, especially its thickness at critical overlap points. This was an iterative process, with SLA 3D-printed reusable prototypes used for functional testing with actual strips. Points of interest included sample containment or leakage, strip flooding, consistent flow, and even development of test and control lines. We also customized the design of the outer surface of the cassette for slip-resistance and enhanced grip. Finally, the cassette was injection molded by Kanani Biologicals in thousands with a custom-made 8-cavity mold using black high-impact polystyrene (HIPS). To ensure proper closure and avoid under- or overcompression, cassettes were closed using a commercial automatic assembly roller (KinBio).

We also developed an iOS software app for LFA image analysis and result call-out in native Swift to access the AVFoundation framework, allowing full control of camera settings (focus, ISO, and exposure). Images are captured in RAW format, and the signal is averaged across the strip into a 1-D profile to compensate for any hot or dead pixels and line nonuniformity. Ten images are taken, and profiles are then averaged to reduce acquisition noise. The total analysis time is <20 s. The app can detect excessive background, defective activation (by the summed intensity of test and control lines), and outdated or incompatible materials by QR code reading, light leakage background by intensities at the two ends of the window, and a mispositioned/inverted cassette by image analysis. Most modern digital phone cameras have a similar spectral response, but the sensitivity varies considerably among models. We have successfully tested our assay with Samsung and iPhone cameras, which account for more than 75% of the US market (https://www.statista.com/statistics/620805/smartphone-sales-market-share-in-the-us-by-vendor/). These phones allow us to access RAW-format image data and lock down the exposure, ISO, and focus settings.

We also have begun to develop a free-standing agnostic reader with a low cost of parts and the ability to communicate with most smartphones by Bluetooth and/or Wi-Fi to obviate any effects of OS updates by phone manufacturers and to reduce the cost of equipment exposed to potentially hazardous samples.

4. Other Translational Issues

For real-world impact, a technology needs to reach a readiness and maturity level beyond that typically reached in an academic laboratory to ensure manufacturability at scale, extraordinarily high reproducibility, and the potential for successful commercialization and widespread adoption. Point of care tests and especially home tests must be developed under tight economic constraints with considerable attention to manufacturing cost and scalability. These considerations must be addressed very early in the development process, as small details can greatly affect manufacturability.

Narrowing the intended use case and understanding the intended end users are essential for successful test deployment. We engaged external marketing consultants to estimate the potential addressable market opportunity and target market segmentation to assess the competitive landscape to direct final product configuration based on the voice of the customer (VOC) and to develop a commercial plan including identifying possible test distributors.

We identified potential commercial partners for test manufacture, kitting/packaging, and regulatory-compliant reagent preparation. In parallel, we initiated accelerated stability studies for the preliminary determination of the shelf life of the various test components (reagents and LFA strip). To assess the robustness of the test, we performed initial flex studies to assess test performance in the presence of common potential use errors (e.g., varied number of drops of sample applied or improper application of the Glow reagent). Nevertheless, to evaluate the reliability and user-friendliness of a point-of-care (POC) test, formal, thorough flex studies should be conducted under conditions simulating potential user-error sources and environmental stresses.

It was essential to initiate a quality management system (QMS; we used the cloud-based software Greenlight Guru) as early as possible, formally documenting processes, procedures, and staff responsibilities to ensure that the test device consistently met the user and regulatory (FDA) requirements and performance criteria. We had weekly meetings with quality consultants (Rook Quality Systems) who guided us in the development of design inputs that addressed the intended use and user need, systematic identification of possible risks, failure modes and effects analysis (FMEA) and associated severity scoring, and implementation of the appropriate design controls and risk mitigation strategies. A traceability matrix tabulated the user needs, design inputs, design outputs, design verification, and design validation processes for both the Glow device and the Glow app. Detailed work instructions (WI) for all reagents and the LFA strip were drafted, and all critical suppliers/vendors were surveyed as part of risk management. A design history file (DHF) was finally initiated.

We faced significant pandemic-induced delays and supply chain issues, most notably for swabs, extraction tubes, and nitrocellulose membranes. We evaluated a variety of swabs on the basis of availability, interacted with a variety of international vendors with complex shipping logistics to source the appropriate extraction tube with a filter dropper tip for the sample, and placed large-volume open orders for nitrocellulose membranes to ensure future supply.

Usability^11,44 is of paramount importance in the intended use and was assessed throughout the project, initially by naive users (undergraduate and graduate students in our academic laboratory) and later by a group of medical students at the UH Medical School experienced with COVID-19 rapid tests. We provided participants with detailed test instructions and surveys to assess the usability of the Glow app and the overall workflow. Each group provided useful feedback, and improvements were integrated into the next prototype. Finally, high-level feedback was provided by the Georgia Tech HomeLab (RADx sponsored). Application of the Glow reagent using the dropper was the most error-prone step. We are currently working on more user-friendly Glow reagent packaging (e.g., new blister-based cassette design, Figure 7D,E).

Table 1. Lessons Learned.

1	Access to high-affinity antibodies is crucial for the development of high-performance tests.
2	Access to a variety of LFA materials early on allows for efficient material screening.
3	Test early with dried reagents, in a cassette, and with well-characterized relevant clinical samples.
4	LFA cassette design is important, has many subtleties, and should be considered very early.
5	Maintaining formal and organized documentation is essential to the success of the project and the eventual initiation of formal quality management.
6	Keep the specific use case in mind from early on and understand specific user requirements, capabilities, and needs.
7	A technically excellent academic team is not sufficient for successful practical translation. Recruit translational consultants and external experts as early as possible.
8	Usability is often overestimated by the developers of a technology. Conduct formal usability testing with naive users, and consider engaging consultants.
9	Continuously assess the available data as fairly as possible and pivot if necessary.
10	Be coachable and open to feedback from specialists in areas where you are not experienced.

5. Conclusions and Future Outlook

We were not able to bring our early stage Glow technology to the point of commercialization within the limited time and resources available, but we achieved strong proof-of-concept and advanced translational aspects of the Glow platform including initial high-performance Glow LFAs, reading by iPhone using only a $2 plastic dark box with no lens, convenient excitation liquid packaging, user friendliness, and manufacturability. The current lull before the seemingly inevitable emergence of pathogen X is an appropriate time to assess and share our experience, and we hope to help empower others interested in the rapid development of translatable LFAs and other diagnostics.

Biographies

Katerina Kourentzi is a Research Associate Professor of Chemical and Biomolecular Engineering at the University of Houston. Her research is focused on the translation of molecular recognition fundamentals into the development of point-of-care diagnostics.

Kristen Brosamer received her Ph.D. in Biomedical Engineering from the University of Houston in December 2022. She was then a postdoctoral associate in the Department of Chemical and Biomolecular Engineering at the University of Houston. She is currently a Clinical Research Associate at Meridian Bioscience.

Binh Vu is a Research Assistant Professor of Chemical and Biomolecular Engineering at the University of Houston. His research is centered on bringing low-cost diagnostic technologies to underserved communities without sacrificing the sensitivity of laboratory-based assays.

Richard C. Willson is the Huffington-Woestemeyer Professor of Chemical and Biomolecular Engineering and Biology and Biochemistry at the University of Houston. His work is focused on the development and translation of novel diagnostics, PAT, and purification technologies.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.accounts.4c00075.

• List of all anti-SARS-CoV-2 nucleoprotein antibodies tested; LFA testing of clinical samples; and Glow Rapid Antigen Test—Instructions for Use (PDF)

The authors declare the following competing financial interest(s): B.V. and R.C.W. are named inventors of IP, which could relate to the subject of this Account.