Ultrasensitive Measurement of Both SARS-CoV ‑ 2 RNA and Antibodies from Saliva

: Tests for COVID-19 generally measure SARS-CoV-2 viral RNA from nasal swabs or antibodies against the virus from blood. It has been shown, however, that both viral particles and antibodies against those particles are present in saliva, which is more accessible than both swabs and blood. We present methods for highly sensitive measurements of both viral RNA and antibodies from the same saliva sample. We developed an e ﬃ cient saliva RNA extraction method and combined it with an ultrasensitive antibody test based on single molecule array (Simoa) technology. We apply our test to the saliva of patients who presented to the hospital with COVID-19 symptoms, some of whom tested positive with a conventional RT-qPCR nasopharyngeal swab test. We demonstrate that combining viral RNA detection by RT-qPCR with antibody detection by Simoa identi ﬁ es more patients as infected than either method alone. Our results demonstrate the utility of combining viral RNA and antibody testing from saliva, a single easily accessible bio ﬂ uid.


■ INTRODUCTION
The two main tests for SARS-CoV-2 infection are molecular tests to detect the presence of the virus (RNA or antigen) and serological tests to detect the presence of antibodies against the virus. 1−7 Both tests have advantages and disadvantages.−10 There are several potential reasons for this: low viral load, variability in swabbing, or late swab collection relative to the time of infection. 2,5−19 Serology tests detect antibodies that develop against the virus during infection. 1,6−22 Combining RNA detection with antibody testing has the potential to increase the sensitivity of RT-qPCR alone 6,15,21,23−27 by reducing false negatives in RT-qPCR.
−30 However, studies have shown that both SARS-CoV-2 viral RNA 15,31−34 and antibodies 20,31,35 against the virus are present in saliva.Since saliva is easier to collect than either swabs or blood, its accessibility makes it an ideal biofluid for widespread diagnostic use.−33,36−38 This will depend largely on several factors such as patient selection, sample collection, and RNA extraction methodology.−33 Similarly, previous studies have shown a strong correlation between antibodies against SARS-CoV-2 in blood and saliva. 20,31,35lthough RT-qPCR is highly sensitive, there is great variability in the RNA extraction efficiency, which is generally performed using commercial kits.These kits have advantages in terms of ease but are prone to supply chain limitations. 39urthermore, using kits may lead to incompatibilities with upstream or downstream steps since the components are unknown to the user.Lastly, kits are not optimized for specific biofluids such as saliva, leading to potentially low RNA recovery.Antibody testing in saliva also presents challenges, namely, that concentrations of antibodies in saliva are much lower than they are in blood 20,31,35 and that different isotypes may be present in the different fluids.
We set out to develop a highly sensitive test to detect both viral RNA and antibodies against the virus from the same saliva sample.We developed an optimized RNA extraction protocol that is highly efficient for saliva.We also adapted a single molecule array 40 (Simoa)-based ultrasensitive test we previously developed for detecting SARS-CoV-2 antibodies in blood for use in saliva. 8,41Combining our RNA extraction protocol for RT-qPCR with this ultrasensitive antibody test, we are able to better classify COVID-19 patients as being actively or previously infected with SARS-CoV-2, demonstrating the utility of this approach for accurate diagnosis of COVID-19.

■ RESULTS AND DISCUSSION
In order to develop a highly sensitive test for both SARS-CoV-2 RNA and antibodies against the virus in saliva, we first optimized the RNA extraction.We developed two versions: a low-volume version (30 μL of saliva) that can be performed in a 96-well PCR plate and a high-volume version (300 μL of saliva) that can be performed in a deep-well plate.We started with a general method based on binding nucleic acids to carboxylated paramagnetic beads in the presence of a guanidinium thiocyanate lysis buffer 42,43 (Figure 1a) and optimized each step of the protocol using saliva samples.To evaluate our extraction method, we spiked in known amounts of synthetic SARS-CoV-2 RNA or heat-inactivated SARS-CoV-2 viral particles into PBS or human saliva and then quantified recovery after RNA extraction by RT-qPCR (Figure 1a, Supporting Information Figure S1).We tested a large number of parameters using this approach until we arrived at a protocol that consistently resulted in 50−100% recovery in 30 μL of saliva (Figure 1b,c).To maximize sensitivity, we opted to use more saliva (300 μL) in a deep-well plate format.However, as it is difficult to elute into a small volume using a deep well plate, we included a transfer step in our protocol where we transfer the beads (during a 70% ethanol wash step) from a deep-well plate to a PCR plate, allowing us to elute the RNA in only 10 μL (Figure 1a).
To see how our protocol compares to a commonly used commercial kit, we performed a head-to-head comparison with the ThermoFisher Scientific MagMAX Viral/Pathogen Nucleic Acid Isolation kit.We first extracted RNA from equal volumes of saliva spiked with viral particles and eluted the captured RNA into equal volumes (Figure 1d).In this comparison, we found that our custom protocol had more than twice the recovery of MagMAX.We also compared our protocol to MagMAX using the maximum input volume and minimum elution volumes for both protocols using saliva with the same concentration of viral particles.In this case, our protocol recovered 16× more RNA (Figure 1e).We then compared our protocol to three recently developed protocols for measuring RNA from saliva without purification.These protocols, SalivaDirect, 44 TCEP inactivation, 45 and 95 °C heating, 46 all use a small volume of saliva (<10 μL) since the final volume of RT-qPCR reactions is usually 20 μL.We compared our highvolume RNA extraction to these three protocols using saliva containing the same concentration of viral particles.We found that we detected significantly more RNA in extraction-based protocols, largely due to the ability to use larger volumes of saliva (Figure 1e).
After validation of our RNA extraction protocol, we turned to combining RNA detection with antibody detection in clinical saliva samples from COVID-19 patients.We have previously developed single molecule array (Simoa)-based ultrasensitive profiling of IgG, IgM, and IgA antibodies against SARS-CoV-2 nucleocapsid (N), spike (S), S1, and RBD protein targets in blood. 8Our assay employs a bead-based, digital ELISA for high-throughput, automated ultrasensitive detection of antibodies in small volumes.Using only 40 μL of saliva per sample (10 μL for antibodies and 30 μL for RNA), we characterized 12 antibody interactions and quantified SARS-CoV-2 RNA across 18 saliva samples.We tested two prepandemic saliva samples from healthy individuals and 16 samples from symptomatic individuals who visited the MGH Respiratory Infection Clinic during the pandemic (see the Supporting Information Table S1 for a clinical characteristics summary).The patients were all tested by RT-qPCR from NP swab samples upon arrival to the clinic, and saliva was collected on the same day.
Combined measurement of SARS-CoV-2 RNA and IgG, IgM, and IgA levels against S1 (Figure 2, Supporting Information Figure S2) revealed that five patients were positive for either RNA, antibodies, or both, in saliva compared to three patients by NP RT-qPCR alone.Two patient samples were positive for antibodies in saliva but negative for SARS-CoV-2 RNA by RT-qPCR (in both NP swabs and saliva).Since these patients displayed severe respiratory illness upon presentation to the hospital, their RT-qPCR results were likely false negatives.RNA was detected in saliva for two of the three patients with positive NP swabs.For the patient with RNA detected in the NP swab but not saliva, antibody levels were above the threshold for IgA and IgM against S1.Since our previous work showed that IgA-S1 displayed the best separation between positives and controls, this sample would be classified as positive in our antibody assay.We also measured saliva from a subset of the patients at six late time points (>9 days after the first positive PCR).We found no RNA was detected but antibody levels were high for most of the immunoglobulin subtypes we measured, as expected (Supporting Information Figure S3).Finally, to see if we could achieve multiplexed detection of viral RNA and antibodies on the same platform from one sample, we developed a Simoa assay for the direct detection of SARS-CoV-2 RNA.Although we found this assay to be less sensitive than RT-qPCR, the Simoa assay detects RNA without amplification.We combined this assay with our antibody assay to detect RNA and antibodies spiked into saliva samples (Supporting Information Figure S4, Table S2), demonstrating multiplexed detection of RNA and antibodies on the Simoa platform.
Several studies have shown that saliva is suitable for both SARS-CoV-2 viral RNA detection and antibody measurements, but these measurements have not been made together using the same sample.In this study, we combined RT-qPCR with antibody measurements on a small number of symptomatic COVID-19 patient saliva samples.We demonstrate that combining these measurements identifies SARS-CoV-2 infection in patients that RT-qPCR alone misses, providing information that can help guide clinical decision making.This is particularly important in patients who present to the hospital with COVID-19 symptoms but have negative RT-qPCR results despite having been infected with SARS-CoV-2 (for example, because the immune system has already cleared the virus).
To maximize sensitivity of RNA detection, we developed and optimized a high-efficiency saliva RNA extraction protocol without the use of kits.We found our protocol yields much higher levels of RNA from saliva than a widely used commercial isolation kit when performed manually.We also directly compared our RNA extraction method to several recently published no-extraction methods for SARS-CoV-2 detection by RT-qPCR in saliva.Although these methods greatly increase ease and throughput of RT-qPCR testing, this comes at a cost of sensitivity.Our high-volume RNA extraction protocol allows for a 30-fold concentration of the saliva sample in addition to the inactivation and removal of enzymatic inhibitors present in saliva.In a comparison utilizing saliva spiked with virus, our high-volume RNA extraction protocol led to the detection of at least 56 times more viral RNA than the use of inactivated saliva as input.
To maximize the sensitivity of our antibody measurements, we employed assays that we recently developed using single array (Simoa) technology. 8The high sensitivity of these assays is particularly advantageous in saliva, where antibody levels are known to be low. 20,31,35−49 We envision that the sensitivity of the RNA assay can be further improved, for example, by adding Cas13a-based detection. 50Future studies may also incorporate additional protein biomarkers in saliva to measure inflammation (cytokines, etc.) or other aspects of host response to increase the utility of multiplexed saliva diagnostics for COVID-19.

Figure 1 .
Figure 1.Custom RNA extraction protocol recovers viral RNA with high efficiency.(A) Schematic representation showing our bead-based extraction protocol in low-volume and high-volume versions.(B) High-efficiency viral RNA recovery from PBS and saliva using custom protocol.PBS or saliva (30 μL for low volume or 300 μL for high volume) were spiked with 10 000 copies of synthetic SARS-CoV-2 RNA.Our custom RNA extraction protocol was performed, and extracted RNA was compared to the input amount using RT-qPCR.(C) Recovery of RNA is calculated as the RNA level of extracted SARS-CoV-2 RNA relative to the spike-in amount.(D) Direct comparison of custom protocol with commercial kit using the same input and elution volumes.RNA was extracted from 200 μL of saliva with 50 particles/μL of heat-inactivated virus and eluted into 50 μL using either our custom protocol or the MagMAX kit and quantified by RT-qPCR.(E) Comparison of extraction and extraction-free methods.For all methods, saliva spiked with 50 viral particles/μL was used, and the same volume of inactivated saliva or eluted RNA was compared by RT-qPCR.For comparison of RNA extraction methods, maximum input volumes (400 μL for MagMAX vs 300 μL for custom) and minimal elution volumes (50 μL for MagMAX vs 10 μL for custom) were used.Relative RNA levels (log scale) represent each condition relative to spike-in aliquot on the same RT-qPCR plate.Fold-differences relative to custom protocol are indicated above each condition For the saliva control (noinactivation), 8/12 replicates had no detectable levels of RNA and were not plotted.Four replicates were repeated across three different days for each condition (different colors represent different days).Error bars for all figures indicate the standard deviation.