The ongoing severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic poses an enormous challenge to the world. SARS-CoV-2 has resulted in more than 100 million cases of coronavirus disease 2019 (COVID-19) worldwide, resulting in more than 2.3 million deaths as of 12 February 2020 (1
). Unlike many other viruses, SARS-CoV-2 displays high infectivity, a large proportion of asymptomatic carriers, and a long incubation time of up to 12 days, during which carriers are infectious (2
). As a result, transmission has been widespread, resulting in overwhelmed health care capacities across the globe (5
). Timely, reliable, and accurate diagnostic and surveillance tests are necessary to control the current outbreak and to prevent future spikes in transmission.
Reverse transcription polymerase chain reaction (RT-PCR), which detects viral nucleic acids, is the current gold standard for COVID-19 diagnosis (7
). Although RT-PCR is highly sensitive and specific (9
), it does not detect past infections—RNA is typically only present at high quantities during acute infection—and it does not provide insight into the host’s response to infection (11
). Serological assays, which detect antibodies induced by SARS-CoV-2, are a crucial supplement to nucleic acid testing for COVID-19 management (12
). Specifically, serological assays are important to track the body’s immune response (14
) and to potentially inform prognosis (15
) or immunity status (12
). Serological assays are also essential for use in epidemiological studies (16
) and are a critical enabling tool for vaccine development (17
SARS-CoV-2 is an enveloped RNA virus with four structural proteins: spike (S) protein, membrane (M) protein, enveloped (E) protein, and nucleocapsid (N) protein (18
). As the pandemic unfolded, several serological binding assays were developed including enzyme-linked immunosorbent assays (ELISAs) and lateral flow assays (LFAs). These assays measure either the level of total antibody or that of specific antibody isotypes that bind to viral proteins—normally S or N. Several studies have demonstrated promising clinical sensitivity and specificity for ELISA and some LFAs (19
). Furthermore, several ELISAs have been shown to correlate well with neutralizing antibody titers (21
) and thus may be useful clinically and in vaccine development (23
). However, both ELISA and LFAs have major disadvantages that limit their applicability for COVID-19 management. ELISA requires technical expertise, laboratory infrastructure, and multiple incubation and wash steps, limiting its applicability to settings outside of a centralized laboratory (24
). On the other hand, LFAs are portable, but they have lower sensitivity and provide qualitative results (25
), whereas a quantitative readout is preferred for clinical use, research studies, and surveillance applications. Collectively, these shortcomings of ELISAs and LFAs motivate the need for an easily deployable, point-of-care test (POCT) that can be manufactured in large volumes, has quantitative figures of merit equal to laboratory-based tests, and is as easy to use as an LFA.
To address the challenge of creating a user-friendly and widely deployable assay that can detect prior exposure to and immunological response against SARS-CoV-2, we developed a new multiplexed portable COVID-19 serological assay that is described herein. Our passive microfluidic platform provides sensitive and quantitative detection of antibodies against multiple SARS-CoV-2 viral antigens in 60 min with a single test from a single 60-μl drop of blood, plasma, or serum. We chose to quantify the antibody response against three different SARS-CoV-2 antigens because emerging studies have demonstrated that the primary antigenic target of the humoral immune response may inform disease progression and prognosis (14
). Thus, being able to differentiate the viral targets of antibodies—as we can with our platform—may be especially valuable. Furthermore, our portable test is completely automated and can function at the POC independent of a centralized laboratory using only an inexpensive handheld detector. We also show that our test can be easily modified to detect additional protein biomarkers, such as cytokines/chemokines, without compromising the performance of the serological assay, which may provide further clinical insight into disease severity and/or patient outcomes (2
). Collectively, these attributes suggest that our platform is a valuable tool for COVID-19 management both at the individual patient level (i.e., monitoring patients who may progress to severe disease) and for large-scale epidemiological studies at the population level. Furthermore, this platform is modular and can be easily modified to detect other pathogens or diagnostic markers simply by using a different set of biological reagents.
As the COVID-19 pandemic unfolded, countries around the globe grappled with developing streamlined systems for diagnosis of acute infection using nucleic acid detection methods. Although there remains an urgent need for rapid and sensitive POCTs for acute diagnosis, developing accurate and reliable serological assays has been deemed an equally important endeavor to complement existing diagnostic strategies (12
). The challenge with developing an easy-to-use serology assay that can be broadly disseminated but that performs as well as centralized laboratory-based methods is highlighted by the large number of ELISA and LFA tests that have been developed. While LFAs are portable and easy to use and ELISAs are quantitative and highly sensitive, there remains a need for a technology that can merge the best attributes of each format.
The DA-D4 POCT is a promising platform to supplement existing diagnostic technologies to manage the COVID-19 pandemic because it marries the best attributes of LFAs and ELISAs—it is quantitative, easy to use, widely deployable, requires only a single 60-μl drop of blood, and can be performed with minimal user intervention. The SARS-CoV-2 DA-D4 assay can be used to measure antibody kinetics and seroconversion at the individual patient level directly from unprocessed blood or plasma. This test is highly sensitive and specific and is potentially suited for epidemiological surveillance at the population level using low-cost microfluidic cassettes that can be transported and stored for an extended period of time without a cold chain. Furthermore, it requires minimal user intervention to carry out the assay and provides a quantitative readout using a low-cost, handheld detector.
We show a strong correlation between the DA-D4 assay readout (for S1 and the RBD of S1) and neutralizing antibody titers, suggesting that this test may be useful in understanding efficacy and durability of natural or vaccine-induced humoral immunity and to potentially inform disease prognosis and population-level immunity. We also demonstrate that an additional prognostic biomarker can be easily incorporated into the test, which may be useful for monitoring disease severity and predict clinical outcomes. Combined, these attributes suggest that this platform may also be useful on the individual patient level to aid in clinical decision-making. While the results presented here mainly highlight the performance of the microfluidic chip, the open-format architecture with up to 24 individual assays per glass slide may be useful for scenarios where higher-throughput testing is demanded. The open format still has advantages compared to traditional ELISA because the open format only requires a single incubation step and one wash step, which reduces the hands-on time and equipment complexity required to complete the assay.
The DA-D4 has additional features that synergize to deliver a highly desirable serological assay. First, the DA sandwich format has advantages over other serological assay formats. Because total antibody is detected rather than a single antibody isotype or subclass, seroconversion in patients can be detected earlier, which reduces the chances of a false-negative result due to a test being administered too early in disease (39
). Furthermore, because the labeled reagent does not have species specificity, the single assay kit could be used in preclinical vaccine development studies to measure antibody responses in experimental animals (23
). The lack of species-specific detection antibodies also reduces the risk of high background signal caused by nonspecific antibodies binding to the surface and subsequently being labeled (46
). Last, the DA-D4 can be conducted in a single step to accomplish multiplex detection without the need for an intermediate wash step, which other assay formats require.
Second, all reagents needed to complete the assay are incorporated onto the nonfouling POEGMA brush that eliminates virtually all nonspecific protein adsorption and cellular adhesion, thereby enabling an extremely low LOD directly from undiluted samples (47
). Although many serological assays often dilute samples, the ability to test undiluted samples is advantageous, especially when combined with prognostic biomarker testing where dilution of low-concentration analytes can lead to an undetectable signal. Testing multiple dilutions can still be performed using our test when antibody levels become high, which could be used to calculate specific titers. POEGMA also acts as a stabilizing substrate for printed reagents, enabling long-term storage of chips without a cold chain (28
). In this study, results were generated over the course of 3 months from the same batch of tests stored in silica desiccated pouches at room temperature and ambient humidity.
Third, this platform can be easily multiplexed, which can be used to capture a more detailed picture of the host immune response to SARS-CoV-2 infection by quantifying the antibody level induced to multiple viral antigens—in this case, N, S1, and RBD—from a single sample without sacrificing ease of use. This is because each viral antigen is deposited at a spatially discrete location, which allows for a single fluorescent tag to be used during fluorescence imaging of the chip, thereby simplifying assay readout compared to other multiplexing technologies such as Simoa or Luminex assays, which rely on multiple different reporter molecules and a more complex readout (14
). This method also allows us to simultaneously measure the concentration of potential prognostic biomarkers directly from plasma (26
) without compromising the performance of the multiplexed serological assay. To the best of our knowledge, there are currently no tests on the market that can probe for antibodies against multiple viral antigens and prognostic protein biomarkers simultaneously.
Fourth, this platform is designed for POC deployment because it requires a single drop of blood that is readily obtained from a finger stick. This droplet is injected into the sample port of a gravity-driven microfluidic chip that requires no further user intervention beyond the concurrent addition of a few drops of wash buffer into a separate port. The assay runs by itself under the action of gravity and capillary action until all the fluid is drained from the microfluidic path by the absorbent pad at the bottom of the cassette, which fully absorbs and contains all liquid. This design eliminates the need for pumps, valves, or actuators and reduces the complexity and cost of the assay. Furthermore, it can be read out at the point of sample collection using the D4Scope, a highly sensitive and inexpensive handheld detector developed to work with the microfluidic chip. The D4Scope images a chip and provides a quantitative readout in less than 5 s, does not require an external power source or laboratory infrastructure, and can wirelessly transmit the results to a remote server over Wi-Fi. While smartphone-based diagnostics are becoming more popular, a benefit of this platform is that it does not rely on smartphone hardware and software, which change rapidly.
Where might this POC assay for COVID-19 serology and prognosis be useful? Serial quantification of antibody response and prognostic biomarkers would be most useful to monitor symptomatic and severe cases where use of available therapeutics, such as antiviral or monoclonal therapies, are indicated. Furthermore, it could be used to screen for patients with poor antibody responses who may benefit from convalescent plasma or monoclonal antibody therapy. We believe that this platform has potential utility in POC settings such as ICUs, urgent care clinics, and at the point of use—at locations where periodic surveillance of health care workers and other essential workers in close proximity to others for extended periods of time such as assembly-line manufacturing or food processing plants is desirable to assist in tracking clusters of disease and epidemiological studies. This platform could also be used as an inexpensive tool to study the longitudinal dynamics of antibody levels to inform reinfection potential, as coronavirus immunity often lasts only ~6 months (50
). Similarly, it could be used to monitor vaccine-induced humoral immunity, which could help determine if boosters are needed in certain vaccinated individuals. This technology is suitable for low-resource settings across the globe, where eliminating the need for sample storage and transport to a centralized testing facility, and the attendant cold chain, is desirable and where access to expensive, high-throughput clinical analyzers that process large volumes of serology and other sandwich immunoassays is limited. Similarly, remote and austere settings—such as the field-forward position of the military or other remote locations where pandemics often emerge—can also benefit from this platform, as the testing is carried out with a disposable cassette and a low-cost, lightweight, and handheld detector whose production can easily be scaled up to enable widespread and dispersed deployment.
While the results presented here are promising, there are several issues identified during this study that require further investigation before its deployment. First, our cohort of individuals with SARS-CoV-2 infection consisted of adults with clinically severe disease, which is not representative of the entire spectrum of COVID-19 disease severity. These samples were chosen to demonstrate proof of concept of the DA-D4 assay and because these samples were locally available through an existing biobank. We recognize that a larger sample size that spans the disease severity spectrum is required to develop a more robust measure of sensitivity and specificity of the DA-D4 serology test for SARS-CoV-2. Similarly, we were not able to match demographics in our negative control group, which may have introduced confounding variables in our analyses. Because of limitations in the volume available from archived samples, we were not able to directly compare the performance of our test to ELISA or LFAs. These studies will be conducted on additional samples in future studies that are designed to address this precise issue and will allow us to assess the concordance in the clinical performance metrics between the different analytical methods. Furthermore, several of the samples we tested saturated the readout of our assay, which limits the dynamics we can measure once high antibody titers are achieved. This limitation could be addressed by testing individual samples on separate microfluidic chips at various dilutions, which would effectively increase the dynamic range of our assay and yield more precise quantitative titer. In addition, because of the DA design of our assay, we are also not able to discriminate between specific antibody subclasses or isotypes, which has been shown to be important for other diseases. This assay format also required that we use a truncated form of the N protein—expressed in Escherichia coli
—as the detection reagent to avoid high signal at low antibody concentrations due to dimerization of full-length N. This may have caused our assay to underestimate the titer of anti-N antibodies for two reasons: (i) the bacterial expression system we used does not perform glycosylation, which could negatively impact antibody recognition, and (ii) the truncated form does not allow us to detect antibodies that are developed against the C-terminal domain, which also contains immunogenic epitopes (51
). This limitation is compensated for by the fact that we can easily multiplex using the DA-D4 format and thus detect antibodies directed against different antigens to maintain high sensitivity and specificity. Despite these limitations, we believe that our assay is poised well to complement existing diagnostic solutions once additional validation studies encompassing larger patient cohorts are completed. We are actively developing an improved version of the test that requires less sample volume and has a shorter run time to better match the time to results and volume requirements of existing LFAs.
In summary, we have developed a COVID-19 serological assay that merges the benefits of LFAs and ELISAs. We used this test to simultaneously measure the antibody levels for multiple viral antigens and a potential prognostic biomarker directly from plasma and whole blood. For COVID-19 management, our platform may be useful to better understand patient antibody responses, provide actionable intelligence to physicians to guide interventions for hospitalized patients at the POC, to assess vaccine efficacy, and to perform epidemiological studies. Furthermore, our platform is broadly applicable to other diseases where sensitive and quantitative antibody and/or protein detection is desirable in settings without access to a centralized laboratory. Overall, we believe that our platform is a promising approach to democratize access to laboratory quality tests, by enabling rapid and decentralized testing with minimal user intervention to locations outside the hospital.
MATERIALS AND METHODS
The DA-D4 assay is based on the design of the D4 immunoassay, reported elsewhere (28
). Briefly, a polymer brush composed of POEGMA was “grafted from” a glass slide by surface-initiated atom transfer radical polymerization (48
). Recombinant SARS-CoV-2 proteins were then printed onto POEGMA-coated slides as capture and detection spots. Capture spots of the following proteins were printed as ~170-μm-diameter spots using a Scienion S11 sciFLEXARRAYER (Scienion AG) inkjet printer: spike S1 (Sino Biological, catalog #40591-V05H1), spike RBD (Sino Biological, catalog #40592-V02H), and nucleocapsid protein (Leinco, catalog #S854). Each protein was printed as a row/column of five identical spots. Next, 12 excipient pads of trehalose with 1.6-mm spacing were printed from a 10% (w/v) trehalose solution in deionized water around the periphery of the capture antigen array using a BioDot AD1520 printer (BioDot Inc.). To print the detection reagents, S1 (Sino Biological, catalog #40591-V08H) and N-NTD (produced in-house) were first conjugated to Alexa Fluor 647 (per the manufacturer’s instructions) and then detection spots of the fluorescent protein conjugates of these proteins were printed on top of the excipient pads as 12 1-mm-diameter spots. A schematic of the chip that shows the spatial address and dimensions of the capture spots, trehalose pad, and detection spots is shown in fig. S1. After printing and final assembly, D4 chips were stored with desiccant until use. The amount of reagent deposited for the open format and microfluidic format was identical, with the only difference being the relative spot placement (fig. S1, A and B). For DA-D4 assays that also detected IP-10, an additional column of five spots of capture antibody (R&D Systems, catalog #MAB266) was included and anti–IP-10 detection antibody (R&D Systems, catalog #AF-266) was included in the detection cocktail for the open-format chips.
Fabrication and analytical testing of open-format DA-D4
Open-format slides were prepared by adhering acrylic wells to each slide, which splits one slide into 24 independent arrays (see fig. S1A for a schematic and Fig. 1B
for an image). To validate the analytical performance of the test, dose-response curves were generated using antibodies targeting SARS-CoV-2 antigens (Sino Biological, catalog #40143-MM05, 40150-D001, and 40150-D004) spiked into undiluted pooled human serum. Open-format chips were incubated with a 13-point dilution series (run in triplicate) for 30 min, briefly rinsed in a 0.1% Tween 20/phosphate-buffered saline (PBS) wash buffer and then dried. Arrays were imaged on an Axon Genepix 4400 tabletop scanner (Molecular Devices LLC).
Fabrication and analytical testing of microfluidic DA-D4
The microfluidic chip fabrication process is described in detail in the Supplementary Materials. Briefly, the microfluidic chip was fabricated by adhering complementary layers of precision laser-cut acrylic and adhesive sheets onto a POEGMA substrate that had been functionalized with the relevant capture and detection reagents. The resulting assembly features a reaction chamber, timing channel, sample inlet, wash buffer reservoir, and wicking pad that automates the sample incubation, sample removal, wash, and drying steps. Simulated doses were prepared using antibodies targeting SARS-CoV-2 antigens (Sino Biological, catalog #40143-MM05, 40150-D001, and 40150-D004) spiked into undiluted pooled human serum. Six doses (including a blank) were tested on the microfluidic DA-D4 in the following way: (i) The user dispenses 60 μl of sample into the sample inlet using a pipette. (ii) The user dispenses 135 μl of wash buffer into the wash reservoir of the cassette using a pipette. (iii) The user waits 60 min for the cassette to run to completion. During this time, (a) fluorescently labeled antigens dissolve and form sandwiches with the antibodies of interest and the immobilized capture antigen in the reaction chamber. (b) A small volume of sample traverses the timing channel, which governs the incubation time. (c) The sample reaches an absorbent pad situated at the end of the timing channel that rapidly wicks away all sample from the reaction chamber, ending incubation. (d) As the sample clears, wash buffer enters the reaction chamber, removing residual sample and unbound reagent before it is also wicked away, leaving a cleaned and dry imaging surface. We observed a less than ±10% variation in the designed 23-min incubation time for the data presented in Fig. 1F
. The remaining difference in time accounts for washing and drying time. (iv) The cassette is ready for analysis on the D4Scope. The vertical orientation of the cassette works in conjunction with the POEGMA brush to maintain low background fluorescence. Cellular and other sample debris can collect on the brush surface owing to gravitational forces, even if no binding is occurring. The vertical orientation ensures that these debris fall harmlessly toward the timing channel during the wash step. This proved especially important when testing with undiluted human whole blood samples.
De-identified heat-inactivated EDTA plasma samples (57°C for 30 min) were accessed from the Duke COVID-19 ICU biorepository (Pro00101196, PI Bryan Kraft) approved by the Duke Health Institutional Review Board (IRB) via an exempted protocol approved by the Duke Health IRB (Pro00105331, PI Ashutosh Chilkoti). Briefly, eligible patients included in the repository were men and women ages 18 years and above who were admitted to an adult ICU at Duke University Hospital with SARS-CoV-2 infection confirmed by PCR testing and who gave informed consent. Samples were collected on study days 1, 3, 7, 14, and 21. In addition to biological samples, clinical data on these patients were also collected including demographics, laboratory data, and clinical course. This study was performed in collaboration with the biorepository team and we have complied with all relevant ethical regulations.
Ten negative control plasma samples were collected under a normal blood donor protocol (Pro00009459, PI Tony Moody) and were collected from 2014 to 2019 (before the COVID-19 outbreak). All patient information, including demographics, is unknown to the investigator team. An additional 11 negative control samples were purchased commercially (Lee Biosolutions Inc.). Last, 20 negative control samples and 18 samples from patients infected with coronavirus 229E (n = 2), HKU1 (n = 4), NL63 (n = 2), and OC43 (n = 10) were collected under Pro00001698. All samples were accessed via an exempted protocol approved by the Duke Health IRB (Pro00105331, PI Ashutosh Chilkoti). Blood was either purchased commercially (Innovative Research Inc.) or accessed from the ICU biorepository (Pro00101196, PI Bryan Kraft) in EDTA-collection tubes and was tested within 48 hours of sample collection.
Testing of prepandemic healthy controls, specificity panel, and ICU samples on the microfluidic DA-D4
The plasma samples (prepandemic healthy controls, specificity panel, and ICU biorepository) were thawed from −80°C storage and allowed to reach room temperature before testing. Blood samples were tested at room temperature. The same procedure used to test the simulated samples as described in “Fabrication and analytical testing of microfluidic DA-D4” was used for testing of all clinical samples. The only exception was that a modified microfluidic flow cell described in the Supplementary Materials that required the use of 200 μl of wash buffer was used for testing whole blood.
D4Scope fabrication and operation
The D4Scope design, fabrication, and assembly are described in detail in the Supplementary Materials. Briefly, the D4Scope’s optical elements—the laser, band-pass filter, lens, and camera—and processing elements—the Raspberry Pi 4, touchscreen, and cabling—are mounted in a custom 3D-printed chassis. Fully assembled, it weighs ~5 pounds. The D4Scope can be powered through either a portable battery pack or wall power. Once connected to the power source, the D4Scope automatically runs our custom imaging Python program. The user removes the light protection cover from the cassette loading port and slides the microfluidic cassette with the glass side toward the detector. The light protection cover is then replaced enclosing the cassette. The user is then prompted to enter the sample ID # and chip ID # using either the touchscreen or optional attached keyboard and mouse.
The D4Scope has two fine adjustment knobs on the cassette loading port that allow for precise vertically and horizontally movement of the cassette relative to the laser source to ensure that the DA-D4 array is perfectly centered with the excitation source. Each array has coprinted two control spots that will always be uniformly bright across all tested samples and align with two superimposed alignment cross hairs on the live video feed of the D4Scope. Using the “toggle video” function on the user interface activates the laser and camera to provide a live view of the imaging area for this alignment. Once aligned, the toggle video function can be pressed again to end the live view, and the “capture image” function can be used to collect and save the resulting image onto the on-board hard drive and, optionally, to a cloud-based server defined by the end user. The live-view feature should be used sparingly to prevent photobleaching of the sample. For this study, we manually analyzed the resulting fluorescence intensity using Genepix Analysis software. However, we have developed an algorithm for automatic analysis of spot intensity and instantaneous results on our open-format platform, which will be reported elsewhere.
Live SARS-CoV-2 MN assay
The SARS-CoV-2 virus (Isolate USA-WA1/2020, NR-52281) was deposited by the Centers for Disease Control and Prevention and obtained through BEI Resources, National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH). SARS-CoV-2 MN assays were adapted from a previous study (52
). In short, plasma samples are diluted twofold and incubated with 100 TCID50
(median tissue culture infectious dose) virus for 1 hour. These dilutions are transferred to a 96-well plate containing 2 × 104
Vero E6 cells per well. Following a 96-hour incubation, cells were fixed with 10% formalin and cytopathic effect (CPE) was determined after staining with 0.1% crystal violet. Each batch of MN includes a known neutralizing control antibody (clone D001; SINO, catalog #40150-D001). Data are reported as the inverse of the last dilution of plasma that protected from CPE, log10
Open-format DA-D4 slides were fabricated as described above using all reagents needed for antibody detection and IP-10 detection. Citrated plasma samples from 10 patients were procured from the ICU biorepository. Sixty microliters of each sample was added to two separate DA-D4 chips and incubated for 30 min, and the chips were then rinsed using 0.1% Tween 20 in 1× PBS. All slides were scanned with the Genepix tabletop scanner.
IP-10 levels were measured using the LEGENDplex Human Proinflammatory Chemokine Panel (13-plex) and LEGENDplex Human Anti-Virus Response Panel (13-plex) obtained from BioLegend. Assays were performed with patient serum per the manufacturer’s instructions. The assay was performed using a Beckman Coulter CytoFLEX flow cytometer, and data processing was performed using BioLegend’s Bio-Bits cloud-based software platform. Each sample was tested in triplicate, and the results are reported as the mean of these triplicates.
Statistical analysis was performed using GraphPad Prism version 8.4.1 (GraphPad Software Inc.). All data were log-transformed for analysis. To establish statistical significance between negative and positive cohorts (Fig. 2, B to D
), unpaired t
tests were used. Pearson’s r
correlation was used to assess the degree of correlation between measurements and was calculated using GraphPad Prism.
We thank D. Montefiori for providing laboratory space to complete the clinical validation studies. We also thank R. Sahm for completing live SARS-CoV-2 microneutralization assays, which were performed in the Virology Unit of the Duke Regional Biocontainment Laboratory, which received partial support for construction from the NIH/NIAD (UC6AI058607; GDS). We thank the nurses in the ICUs of Duke University Hospital for collecting the blood samples used for this study and thank P. Lee for supporting the ICU Biorepository. We thank T. Moody for access to the prepandemic negative control samples used in this study, T. Denny for providing access to BSL2+ laboratory facilities to run the prepandemic negative samples, and H. Register for assistance with testing the negative samples on the DA-D4 POCT. Funding: A.C. acknowledges the support of the National Science Foundation (grant no. CBET2029361); the National Cancer Institute through grants P30-CA014236, R01-CA248491, and UH3-CA211232; Department of Defense United States Special Operations Command (grant no. W81XWH-16-C-0219); Defence Academy of the United Kingdom (grant no. ACC6010469); and the Combat Casualty Care Research Program (JPC-6) (grant no. W81XWH-17-2-0045). B.D.K. receives funding from NHLBI (K08HL130557). Author contributions: J.T.H. and D.S.K. are co–lead authors, who equally participated in experimental design, data collection, data analysis, manuscript drafting, figure creation, and manuscript revision. L.B.O. participated in experimental design and collection and analysis of data for clinical validation. J.L. developed the D4Scope used throughout the study and drafted text and figures related to the D4Scope. D.S.K., C.M.F., and A.M.H. developed the microfluidic cassette. G.K. cloned, expressed, and purified the N-NTD. S.A.W., S.O., and Z.Q. participated in data collection and analysis. C.M.F., D.Y.J., and A.M.H. were responsible for conceptualization, investigation, and manuscript revision. C.P. oversaw statistical analysis and assisted in study design. J.G.A. and T.W.B. assisted in procurement and testing of negative control samples. T.O. designed and ran live virus microneutralization assays, analyzed data, and participated in writing the manuscript. G.D.S. participated in writing the manuscript. B.D.K., C.W.W., L.C., L.G.Q., S.K.N., B.A.S., I.A.N., and L.B.O. contributed to the development of the Duke COVID-19 biorepositories and oversaw clinical data acquisition. T.W.B., B.D.K., and C.W.W. participated in manuscript revision. A.C. is the principal investigator who directed the studies, helped plan experiments, analyzed data, and participated in writing and editing the manuscript. All authors read and approved the manuscript. Competing interests: A.C., D.S.K., C.P., A.M.H., J.L., and J.T.H. are inventors on two provisional patents related to this work filed by Duke University [no. 63/068,432, filed (21 August 2020), not yet published, and no. 63/116,511, filed (20 November 2020), not yet published]. Both are entitled “Microfluidic assay device” and both describe innovations used for the D4 microfluidic cassette described in this work. A.C. and J.L. are inventors on a patent related to this work filed by Duke University [no. WO/2020/223713, filed (2 May 2020), published (5 May 2020)]. The patent is entitled “Devices and methods for imaging microarray chips” and describes innovations used for the D4Scope described in this work. Immucor Inc. has acquired the rights to the D4 assay on POEGMA brushes for in vitro diagnostics from Sentilus Inc. (cofounded by A.C. and A.M.H.). All other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.