Large-Scale SARS-CoV-2 Testing Utilizing Saliva and Transposition Sample Pooling

Summary

The protocol is intended to serve as a blueprint for universities and other organizations considering large-scale testing for SARS-CoV-2 or developing preparedness plans for future viral outbreaks.

Abstract

Identification and isolation of contagious individuals along with quarantine of close contacts, is critical for slowing the spread of COVID-19. Large-scale testing in a surveillance or screening capacity for asymptomatic carriers of COVID-19 provides both data on viral spread and the follow-up ability to rapidly test individuals during suspected outbreaks. The COVID-19 early detection program at Michigan State University has been utilizing large-scale testing in a surveillance or screening capacity since fall of 2020. The methods adapted here take advantage of the reliability, large sample volume, and self-collection benefits of saliva, paired with a cost-effective, reagent conserving two-dimensional pooling scheme. The process was designed to be adaptable to supply shortages, with many components of the kits and the assay easily substituted. The processes outlined for collecting and processing SARS-CoV-2 samples can be adapted to test for future viral pathogens reliably expressed in saliva. By providing this blueprint for universities or other organizations, preparedness plans for future viral outbreaks can be developed.

Introduction

The COVID-19 pandemic, caused by the SARS-CoV-2 virus, has caused the deaths of over 6.2 million people to date, with numbers rising every day¹. The gold standard of testing for SARS-CoV-2 is quantitative real-time (RT-q) PCR, with primers designed to target the viral genome, such as nucleocapsid, envelope, spike, and RNA-dependent RNA polymerase genes². At the beginning of the pandemic, sufficient capacity for SARS-CoV-2 testing was severely lacking. It arose from a lack of validated assays, testing components, clinical personnel, and an infrastructure unprepared to rapidly expand to accommodate pandemic-level, mass testing. Due to shortages, testing centers often required a physician's referral to be test eligible. These shortages resulted in delays for testing approval, long lines for uncomfortable nasopharyngeal sample collection, and lengthy wait times for results. Additionally, because of these constraints, testing efforts could not accommodate pre-symptomatic, mild, or asymptomatic carriers unknowingly spreading the SARS-CoV-2. The lack of easily accessible, widespread testing likely contributed to the uncontrolled spread of COVID-19.

Large-scale interval testing can be performed either as surveillance or screening. Both can be used to monitor local positivity rates in high-density or high-risk transmission areas and can be utilized to make public health decisions. Surveillance testing is intended to monitor the incidence and prevalence of disease in a population and is not used for individual diagnostics³. Surveillance results typically are de-identified and not returned to participants; laboratories conducting surveillance testing need not be clinically certified, nor are they required to use an FDA-authorized assay. Screening allows for results to be returned to individual participants, but screening laboratories in the United States must have a Clinical Laboratory Improvement Amendments (CLIA) certificate and meet all applicable CLIA requirements.

The Michigan State University (MSU) early detection program began in September 2020 and has processed over 350,000 samples. The program arose out of a research group's efforts to design a highly sensitive SARS-CoV-2 assay that did not require high demand testing supplies^4,5,6,7,8. The goals were to aid clinical labs to increase capacity and develop flexible processes to accommodate supply shortages while also developing a screening strategy to establish a return-to-work plan for the MSU College of Human Medicine. The initial efforts focused on alternative collection, extraction, and quantitation methods for SARS-CoV-2. High demand and subsequent shortages of nasopharyngeal swabs led to the evaluation of anterior nares samples collected with buccal swabs, and reagent shortages resulted in development of a sample extraction method adapted from early reports out of Wuhan, China⁹. To increase sensitivity for detecting SARS-CoV-2 in anterior nares samples, droplet digital PCR was substituted for RT-qPCR^6,7. Though droplet digital PCR is highly sensitive and can provide absolute values with an endpoint readout, it was determined that its use was not feasible for large-scale testing due to the lack of reliable high-throughput instrumentation for the technology. Additionally, self-collection of anterior nares samples based on levels of human RNase P was extremely variable, suggesting that it was not sufficiently reliable for mass testing.

An alternative to nasopharyngeal and anterior nares swabs is the collection of saliva. Respiratory viruses such as SARS-CoV, H1N1, and MERS were all historically detected in saliva^10,11,12,13. This was subsequently proven true for SARS-CoV-2^14,15,16,17. Direct comparison between saliva and nasopharyngeal samples showed saliva yields higher viral titers than nasopharyngeal swabs in matched samples, and that saliva is less variable with repeated sample collection¹⁴. Saliva has also been reported to be more sensitive in certain variants, such as Omicron, compared to Delta¹⁶. Added benefits to saliva collection are the relative ease of off-site self-collection without high-demand supplies, the ability to repeatedly retest the sample if needed, the elimination of on-site staffing requirements for sample collection, and the avoidance of participant queues which could increase the potential for viral transmission. The lab-assembled saliva kit was developed as a collaboration among lab assay developers, experts in the school of packaging, university branding experts, safety officers, and external manufacturing partners that produced the box and labeling system.

While saliva samples offer ample genetic starting material and RT-qPCR provides sensitive, reliable outcomes, the cost of reagents (primer/probes and master mix) made large-scale testing of individual samples a costly endeavor on an individual, per sample basis. Since the primer/probes and master mix are the most expensive components of the process, the goal was to seek solutions that would stretch their use and therefore decrease the per sample cost. Systemically optimizing sample pool size based upon incidence in the community and assay sensitivity has been proposed for SARS-CoV-2 testing¹⁸. However, when pools of any size indicate presence of SARS-CoV-2, all participants in the pool must be retested, resulting in lost time and increased opportunities for spread. To address these limitations, a two-dimensional pooling method was employed, like the process proposed by Zilinskas and others¹⁹ to conduct a first pass under the strictures of surveillance testing. In this process, 96 individual samples are placed in a 96-well plate consisting of 12 columns and eight rows. Each sample is included in a pool of eight and a pool of 12 on two different reaction plates. This results in every sample being uniquely represented with the two pools. Deconvolution of the pools based on the coordinates identifies potentially positive samples. Samples in pools where SARS-CoV-2 was not detected, do not move from surveillance testing to screening. Meanwhile, samples from individuals testing positive in the surveillance process are re-extracted through a CLIA-approved screening process. If confirmed positive, individuals are given their result, referred to the university physician's office, contact tracing is initiated, and the health department is notified. In total, an individual's sample is tested in three separate reactions before being declared positive, twice in surveillance pools and once as a single confirmatory screen, reducing the chances of a false positive. Sample pooling uses ~80% less reagents than running samples individually, resulting in a cost of ~$12 per sample.

Beyond the saliva kit, pooling strategy, and assay development process, the team also developed a logistics plan for distribution of kits, collection of samples, and reporting of results. Participants in the program pick up their kit, register the unique alphanumeric code on their tube, produce their sample and deposit it in one of the many drop-off bins where they are picked up daily and transported to the lab. The lab processes the samples, the technical supervisor reviews and uploads results, and participants are notified to check the results portal. This process has a turnaround time of 24-48 h from the time a sample is deposited. The collaboration from all parts of the institution were key for a successful large-scale implementation of this hybrid surveillance and screening process. The following procedures and descriptions of the testing program and infrastructure needed are intended as blueprints on how to scale-up testing for future surveillance and/or screening purposes.

Protocol

Studies performed to optimize the methods for the Early Detection Program were approved by the Michigan State University Institutional Review Board. All figures were reproduced with contrived samples and are representative of the observed human results. No data, information, or results shown in the manuscript are from any participant in the Michigan State University Early Detection Program.

1. Kit production

NOTE: During kit assembly, wear masks, gloves, eye protection, and lab coats at all times to prevent kit component contamination.

1. Preparation of kit components
   1. Using a permanent marker, draw a line denoting 1 mL on the 25 mL conical tube. Using a permanent marker, draw a line denoting 1 mL on the transfer pipet.
   2. Add four ceramic beads to the 5 mL sample tube. Pipet 1 mL of RNA stabilization solution into the sample tube and close the tube. Attach a barbell sticker consisting of a unique identifying code to sample tube with the data matrix barcode on the top and the identifier repeated in alphanumeric code on the side.
      NOTE: The ceramic beads are RNase/DNase-free and do not need to be further sterilized.
2. Place a 25 mL conical tube, 5 mL sample tube, transfer pipet, and small biohazard bag into the kit box (Supplementary Figure 1).
   ​NOTE: Conical tubes do not need to be DNase/RNase-free. Integrity of the sample is preserved by the RNA stabilization solution.

2. Kit uses for self-sample collection

1. As per the kit instructions, ask the participants not to eat or drink 30 min prior to providing a sample. Ask the participants to spit into the 25 mL conical tube until 1 mL of saliva is collected (marked line).
2. Using the pipet, transfer 1 mL of saliva (up to the marked line) into the 5 mL sample tube containing RNA stabilization solution and ceramic beads. Close and shake the sample tube for 15 s.
3. Ask the participants to register the alphanumeric code assigned to sample on the designated website. Once done, ask them to place the sample tube in small biohazard bag and deposit into a collection bin.

3. Sample intake and preparation for RNA isolation

NOTE: All steps take place in a biosafety cabinet or centrifuge bucket with a biocontainment lid.

1. Disinfect and visually inspect samples for quality control while still in the biohazard bag.
2. Reject the sample if (Figure 1): sample is of non-natural color (i.e., green, blue) or contains food particles or blood; sample is leaking or missing beads; sample has incorrect expected volume (<1.5 mL or >3 mL); sample does not have a barcode or alphanumeric code attached.
   1. If the sample is rejected, scan and record the sample into a rejection file with a note as to why the sample was rejected. If sample is rejected because it has no identifiers, log the sample as No Barcode in the rejection file.
3. If sample is not rejected, cut open the biohazard bag with scissors and remove the sample tube. Vortex the tube for 15 s then place the sample tube into a tube adaptor for a swing-bucket centrifuge.
4. Once tube adapters are full, secure the biocontainment lid over the centrifuge bucket and transfer to the centrifuge. Centrifuge samples at 4,100 x g for 2 min and return samples to the biosafety cabinet.
   NOTE: Centrifugation is used to pellet debris and force more viscous material in saliva to the bottom of the tube. This eliminates the need for proteinase K in the process. A full run consists of 768 samples (eight 96-well plates full of sample after RNA isolation).
5. Without disturbing the pelleted material, transfer sample tubes to a 96-slot tube rack pre-labeled with the Run ID which includes the rack number (1-8), date, and the corresponding group the samples will be pooled/assessed with (an identifier to denote the group, i.e., group A is the first run of the day).
6. When 96-slot tube racks are full or no samples remain, using a handheld barcode scanner, scan tubes into a plate map file (Supplementary File 1 for full pools; Supplementary File 2 for half size pools). In the event a barcode does not work, use the alphanumeric code on the tube to log the sample.
   NOTE: A mask is required to prevent the scanning of adjacent barcodes. A mask can be made by cutting a hole in opaque paper, then laminating it. Barcodes scanned into the plate map file will appear in the same orientation as in the 96-slot rack.
7. Spray the sample tubes in the rack with 70% ethanol and dab dry with a paper towel.
   NOTE: Wiping the tubes instead of dabbing can damage the barcodes.
8. Label a 96-deep-well plate of 2 mL well capacity with the Run ID. Transfer 200 µL of saliva sample from each sample tube to the corresponding well of the 96-deep-well plate. If samples are too viscous to transfer, vortex and centrifuge again. Reject sample if problem remains.
   NOTE: Saliva samples can be stringy and leave a snail-trail across the plate during transfer if the technician is not careful. This can lead to initial false positives and more samples that will eventually need to be re-run in the confirmation step (see section 6).
9. When all samples are transferred, cover the 96-deep-well plate with a silicone mat and store at room temperature. Save and store the 96-slot racks with remaining samples for confirmation of potentially positive samples.
   ​NOTE: The protocol can be paused here.

4. RNA isolation

1. Label a 96-well RNA isolation cartridge with the Run ID and place it on top of an empty collection plate.
   NOTE: Plates containing samples can be removed from the biosafety cabinet at this point.
2. Remove the silicone mat from the 96-deep-well plate. Transfer 400 µL of viral loading buffer into each well of the 96-deep-well plate.
3. Mix each sample by gently pipetting up and down two to four times and transfer the entire sample to the corresponding column in the 96-well RNA isolation cartridge.
   NOTE: Mixing large volumes will create bubbles that may lead to contamination of neighboring wells if performed haphazardly. Mixing in a deliberate fashion can lower this risk and protect the integrity of downstream processing.
4. Centrifuge the RNA isolation cartridge on top of the collection plate at 4,100 x g for 10 min to load samples onto the cartridge. Transfer the RNA isolation cartridge to a clean collection plate.
5. Add 500 µL of wash buffer to every well of the RNA isolation cartridge and centrifuge at 4,100 x g for 5 min. Transfer RNA isolation cartridge to a clean collection plate.
6. Add 500 µL of wash buffer to every well of the RNA isolation cartridge and centrifuge at 4,100 x g for 5 min. Transfer RNA isolation cartridge to a clean collection plate.
7. Add 500 µL of 100% ethanol to every well of the RNA isolation cartridge and centrifuge at 4,100 x g for 5 min. Transfer RNA isolation cartridge to a clean collection plate.
   NOTE: If the sample has not completely gone through the column at this point, the sample is too viscous or small pieces of debris in the sample are clogging the column. The sample needs to be removed completely from the cartridge as to not contaminate the downstream pooling process, and the sample number added to the rejection file.
8. Centrifuge the RNA isolation cartridge on top of the collection plate at 4,100 x g for 5 min to dry the columns. Transfer the RNA isolation cartridge onto a clean elution plate labeled with the Run ID and the word row, then place both on top of a clean collection plate.
9. Add 25 µL of DNase/RNase-free molecular grade water to each well of the RNA isolation cartridge. Centrifuge at 4,100 x g for 10 min to elute the RNA. If sample does not completely elute, re-centrifuge the sample for 5 min.
10. Transfer the elution plates onto pre-chilled cold plates and cover. Immediately proceed onto sample pooling and RT-qPCR.

5. Sample pooling and RT-qPCR

NOTE: The process is set up in a two-plate system (as seen in Figure 2). The elution RNA plate number corresponds to the produced column plate number and the row plate letter (A = 1, B = 2, C = 3, etc.). For deconvolution purposes, the column plate row letter will correspond to the row letter from the RNA elution plate, and the row plate column number will correspond to the RNA elution plate column number. For example, a sample in RNA plate #6 in position C4 will be found in the row reaction plate position F4, and in the column reaction plate position C6.

1. Label a new elution plate with the Run ID and the word row, and place on a pre-chilled cold plate. Transfer 10 µL of RNA from each well on the row elution plate to the corresponding row on the column elution plate, thus making a copy of the original elution plate (Figure 2).
   NOTE: Elution plates marked column will be pooled in the column direction and elution plates marked row will be pooled in the row direction.
2. For the column pooled plate, transfer the entire volume from each well across the plate into the column number that corresponds to the Run ID plate number. For the row pooled plate, transfer the entire volume from each well up or down the plate into the row number that corresponds to the Run ID plate number (1 = A, 2 = B, etc.; Figure 2).
3. Transfer 13 µL of each pooled sample into the corresponding column or row of a 96-well reaction plate (Figure 2).
4. For this protocol, use three probes against the nucleocapsid (N), ORF1ab, and spike (S) genes to detect SARS-CoV-2, and a fourth primer probe targeting MS2 phage as an internal control. In the pooled samples, spike an MS2 phage positive control into the master mix. In samples run individually, spike MS2 phage into the sample on the RNA isolation cartridge and use as an extraction control.
   NOTE: Master mix, primer probes, fluorescent dyes, and quenchers will vary based on the assay. Appropriate ratios and fluorescent channels should be determined by each lab.
5. Prepare the positive control by adding 2 µL of positive control solution and 11 µL of DNase/RNase-free molecular grade water to the positive control well on the reaction plate (Figure 2).
6. Prepare the no template control by adding 13 µL of DNase/RNase-free molecular grade water to the negative control well on the reaction plate.
7. Make master mix according to manufacturer's instructions and spike in MS2 phage positive control. Dispense 7 µL of master mix into each well of the 96-well reaction plates that contain samples or controls. Limit exposure to fluorescent light.
8. Cover reaction plates with optical film and vortex for 2 min to mix samples well. Centrifuge reaction plates at 650 x g for 5 min.
9. Transfer plates into the RT-PCR system and run with cycling parameters appropriate for the master mix. For master mix used here, the following cycles were run: 25 °C for 2 min, 53 °C for 10 min, 95 °C for 2 min, 40 cycles of 95 °C for 3 s, and 60 °C for 30s.
10. Once the samples are run, perform a deconvolution of the pools to identify potential positive samples manually (Figure 3) and via a R Script Shiny App.
    1. Export results from column and row reaction plates as spreadsheet files. Open the pool deconvolution R script Shiny app (source code available at: https://github.com/kochman1/JoVE-MSU-COVID-EDP).
    2. Drag and drop the plate map, column, and row files into the app and run the app. Save the list produced from the app.
       ​NOTE: The list produced from the app contains all of the barcodes entered into the plate map and whether they are likely positive or negative based on the pool deconvolution script. These samples that are flagged as potential positives are reprocessed individually.

6. Validation of positive samples

1. Label two 1.5 mL microcentrifuge tubes and an RNA extraction column for each sample that needs to be tested. Use one microcentrifuge tube for mixing sample, MS2 phage, and viral loading buffer, and the second microcentrifuge tube for the RNA elution step.
2. Add 5 µL of MS2 Phage RNA to the 1.5 mL microcentrifuge tube. Transfer 200 µL of the potentially positive sample to the tube and add 400 µL of viral loading buffer to the tube, mix by pipetting.
3. Transfer the entire contents to the pre-labeled RNA extraction column in a clean collection tube. Centrifuge at 10,000 x g for 2 min. Aspirate flow through from the collection tube.
4. Add 500 µL of wash buffer to the RNA extraction column and centrifuge at 10,000 x g for 2 min. Aspirate flow through from the collection tube.
5. Add 500 µL of wash buffer to the RNA extraction column and centrifuge at 10,000 x g for 2 min. Aspirate flow through from the collection tube.
6. Add 500 µL of 100% ethanol to the RNA extraction column and centrifuge at 10,000 x g for 1 min. Transfer the RNA extraction column to a clean collection tube and centrifuge at 10,000 x g for 1 min to dry the column.
7. Transfer the RNA extraction column to a pre-labeled 1.5 mL microcentrifuge tube for elution. Add 25 µL of DNase/RNase-free molecular grade water to the RNA extraction column. Centrifuge at 10,000 x g for 2 min to collect the RNA and transfer tube to wet ice.
8. Make master mix according to manufacturer's instructions. Dispense 7 µL of master mix into each well of a 96-well reaction plate that will contain the sample. Transfer 13 µL of each isolated RNA into the reaction plate.
   NOTE: The master mix is the same one used in the pooled sample protocol but does not include the MS2 phage RNA spike.
9. Prepare the positive control by adding 2 µL of positive control solution and 11 µL of DNase/RNase-free molecular grade water to the positive control well on the reaction plate. Prepare the no template control by adding 13 µL of DNase/RNase-free molecular grade water to the negative control well on the reaction plate.
10. Cover the reaction plates with optical film and vortex for 2 min to mix samples well. Centrifuge reaction plates at 650 x g for 5 min. Transfer plates into the RT-PCR system and run with cycling parameters appropriate for the master mix.
    1. For master mix stated here, use 25 °C for 2 min, 53 °C for 10 min, 95 °C for 2 min, 40 cycles of 95 °C for 3 s, and 60 °C for 30 s.

Representative Results

The vast majority of samples received by the lab to date have been accepted and passed the initial visual quality control step. The need to reject a sample is limited to reasons that can negatively influence sample processing and/or the overall results for the sample. Specifically, incorrect volume in the tube, consistency, or color not natural to saliva, an absence of ceramic beads used to aid in sample homogenization, and missing barcodes are all reasons to reject a sample (Figure 1).

Sample conditions were also examined to account for samples being exposed to outdoor elements while in the collection bins. The chosen RNA stabilization solution was tested against prolonged exposures to heat and freeze-thaw cycles. Contrived SARS-CoV-2 RNA samples in the absence of RNases were exposed to varying temperatures then processed for RT-qPCR. In the RNA stabilization solution, samples stored at 37 °C for 1 week compared to those stored at RT, 4 °C, or -20 °C, showed a Ct value higher by 1-3 units depending on which probe was examined (Supplementary Figure 2). In comparison, contrived RNA in DNase/RNase-free water could be up to 5 Ct higher than samples kept in water at RT, 4 °C, or -20 °C, and almost 8 Ct higher than samples in the RNA stabilization solution. To examine the effects of freeze-thawing on samples, contrived SARS-CoV-2 RNA samples were either processed immediately or subjected to up to three freeze-thaws, frozen at -80 °C, and then thawed to RT. Freeze-thawing samples in the RNA stabilization solution did not increase the Ct, however, the same samples in water were one to three cycles higher depending on the number of freeze-thaws (Supplementary Figure 3).

For all RT-qPCRs of both pooled and individual samples, the presence of the target gene is indicated by a Ct of ≤37. For a pool or individual sample to be considered positive for SARS-CoV-2, two-thirds of the viral targets (N, S, and ORF1ab) need to be present (Figure 4). The internal MS2 phage control must also be present in all wells to confirm that the absence of viral targets is not due to an unknown contaminant from the sample inhibiting the reaction (Figure 4). The absence of the MS2 phage control invalidates the assay for those wells and requires the samples to be reprocessed. By determining which pools on the row and column reaction plates do not show amplification of the MS2 phage control, manual pool deconvolution can be used to identify which sample is the cause of the assay inhibition in most cases. This can be confirmed by running the sample individually and a rejection notice reported to the participant.

Though sample pooling is an effective way to save time, reagents, and money, the pooling process does come with a caveat. As pooling samples effectively dilutes the original sample, it will produce a shift to a higher Ct compared to what the sample would have if run individually. A single positive sample in a pool of eight has a Ct shift of two to three cycles, whereas a single positive sample in a pool of 12 has a Ct shift of three to four cycles (Figure 5). Due to the Ct shift, weak positives can be missed by the script used to automate the process. If a technician observes a positive well in the pool of eight that does not have a corresponding positive well in the pool of 12, but does have a well where late amplification (Ct 37-39) occurred, manual deconvolution can be used to identify the potential positive from the pool. These samples mostly fit the positive criteria when run as a single sample, or in some cases the same individual was observed to be positive in the following weeks.

Figure 1: Sample quality control rejection criteria. Examples of saliva samples that would fail the initial quality control process. A sample that passes the quality control criteria is ~2 mL in volume, has the color and consistency of saliva, and the barcode and ceramic beads are present. Samples can be rejected for low volume (<0.5 mL), high volume (>2.5 mL), missing barcode, missing beads, discolored sample, or presence of food or other solid debris. Please click here to view a larger version of this figure.

Figure 2: Schematic for two-dimensional sample pooling (pools of eight and 12). The sample pooling process can accommodate up to eight, 96-well plates per run. (A) In preparation for sample pooling, each RNA elution plate is copied to an empty 96-well plate, with the original plate used to pool the rows and the copy plate used to pool the columns. (B) Entire contents of all 12 columns or all eight rows are pooled together into the correct column or row in the RNA plates. On RNA column plate 1, each row is pooled into the first column (same colors are pooled together). On RNA column plate 2, each row is pooled into the second column. This is repeated until all eight plates have been pooled, then columns are transferred into the column reaction plate in the respective order and a positive control (PC) and a no template control (NTC) are added to the plate. On RNA row plate 1, each column is pooled into the first row (different colors are pooled together). On RNA row plate 2, each column is pooled into the second row. This is repeated until all eight plates have been pooled, then rows are transferred into the row reaction plate in the respective order. (C) The final reaction plates are a compilation of up to 786 samples run in duplicate with no samples pooled with the same sample twice. Please click here to view a larger version of this figure.

Figure 3: Identification of a putative positive sample by pool deconvolution. Position of the positive well on the column and row reaction plates are used to identify the location of the original presumed positive sample. On the column plate, the column number represents the RNA elution plate number, and the row letter represents the row coordinate of the sample on the RNA elution plate. On the row plate, the row letter represents the RNA elution plate number (where A = 1, B = 2, etc.), and the column number represents the column coordinate of the sample on the RNA elution plate. In the example above, a positive well in the column reaction plate position D5 will be found on RNA elution plate 5 in row D, and a positive well in the row reaction plate position E4 will be found in column 4 on RNA elution plate 5; therefore, the presumed positive sample will be from RNA elution plate 5 in position D4. Please click here to view a larger version of this figure.

Figure 4: Amplification curves of positive and negative samples. Amplification plots from examples of contrived positive and negative samples. Assay looks at three SARS-CoV-2 targets, S, ORF1ab, and N genes, as well as an internal control spiked into the sample or master mix (MS2 phage). Amplification cycle cut off for detection of all targets is 37 cycles (denoted by the black vertical line). In the positive control, all targets have amplified before the cutoff. In the negative control, only the spiked in MS2 phage control has amplified, indicating there is no SARS-CoV-2 genetic material present in the sample. Please click here to view a larger version of this figure.

Figure 5: Higher Ct shift associated with sample pooling. Two different dilutions of an RNA SARS-CoV-2 positive control were prepared; dilutions (Dil) 1 and 2 were 1,000 and 10 copies/µL, respectively. RNA was isolated and eluted in water, and samples were run as either a single sample or pooled with seven (pool of eight) or 11 (pool of 12). Representative Ct for each of the three SARS-CoV-2 targets, N, ORF1ab, and S genes, are shown. Circles represent each technical replicate (n = 3), columns represent group means, error bars represent the standard error of the mean, and an asterisk denotes significant difference (p ≤ 0.05; one-way ANOVA). Please click here to view a larger version of this figure.

Supplementary Figure 1: Image of a testing kit and associated components. Kit production begins with the preparation of all testing materials. Each kit is comprised of the following items: (A) kit box, (B) biohazard bag, (C) transfer pipet, (D) 25 mL conical tube, and (E) 5 mL sample tube. The kit box, specifically designed to hold all kit components, is labeled with instructions on how to use the kit and log the samples and has starburst punchouts to hold the 25 mL conical tube and 5 mL sample tube to make pipetting easier for participants. Participants spit into the 25 mL conical tube to the drawn black line (representing ~1 mL), then transfer 1 mL of saliva with the transfer pipet (black line denotes 1 mL mark) to the 5 mL sample tube. The sample tube is prefilled with RNA stabilization buffer, ceramic beads, and has a unique barcode/alphanumeric code sticker. The alphanumeric code is entered into the registration site by the participant to link the sample to the participant. The sample tube is sealed in the biohazard bag and deposited at one of the drop-off bins at sample collection sites. Please click here to download this File.

Supplementary Figure 2: Effects of temperatures on sample integrity. Two different dilutions of an RNA SARS-CoV-2 positive control were prepared in either molecular grade water or RNA stability solution (RSS) and aliquoted. Dilutions (Dil) 1 and 2 were 1,000 and 10 copies/µL, respectively. Aliquots were incubated at -20 °C (dark blue), 4 °C (light blue), room temperature (RT; dark green), or 37 °C (red) for 1 week. RNA was isolated and run as single samples. Representative Ct for each of the three SARS-CoV-2 targets, N, ORF1ab, and S genes, are shown. Circles represent each technical replicate (n = 3), columns represent group means, error bars represent the standard error of the mean, and an asterisk denotes significant difference (p ≤ 0.05; two-way ANOVA). Please click here to download this File.

Supplementary Figure 3: Effects of freeze-thaw cycles on sample integrity. Three different dilutions of an RNA SARS-CoV-2 positive control were prepared in either molecular grade water or RNA stability solution (RSS) and aliquoted. Dilutions (Dil) 1, 2, and 3 were 1,000, 10, and 0.5 copies/µL, respectively. Aliquots were either processed immediately (denoted as 0), or exposed to one, two, or three rounds of freeze-thaws. RNA was isolated and run as single samples. Representative Ct for each of the three SARS-CoV-2 targets, N, ORF1ab, and S genes, are shown. Circles represent each technical replicate (n = 3), columns represent group means, error bars represent the standard error of the mean, and an asterisk denotes significant difference (p ≤ 0.05; two-way ANOVA). Please click here to download this File.

Supplementary Figure 4: Schematic for two-dimensional sample pooling (pools of four and six). The sample pooling process can accommodate up to four 96-well plates per run. (A) In preparation for sample pooling, each RNA elution plate is copied to an empty 96-well plate, with the original plate used to pool the rows and the copy plate used to pool the columns. (B) The entire contents of columns or rows are pooled together into the correct column or row in the RNA plates. On RNA column plate 1, samples in columns 1-6 are pooled into the first column, and samples in columns 7-12 are pooled into the seventh column (same colors are pooled together). This is repeated until all four plates have been pooled, then columns are transferred into the column reaction plate in the respective order, and a positive control (PC) and a no template control (NTC) are added to the plate. On RNA row plate 1, rows A-D are pooled into row A, and rows E-H are pooled into row E (different colors are pooled together). This is repeated until all four plates have been pooled, then rows are transferred into the row reaction plate in the respective order. (C) The final reaction plates are a compilation of potentially 384 samples run in duplicate with no samples pooled with the same sample twice. Please click here to download this File.

Supplementary File 1: Plate map file. File containing the plate map template for pools of eight and 12. Barcodes are scanned into the file, populating the plate map. The plate map is a record of the location of each sample in the tube rack, the plate throughout RNA isolation, and serves as a key to identify the wells containing each sample in the row and column pooled sample plates. This file as well as the RT-qPCR results files from the row and column pooled sample plates are required for the automated deconvolution script. Please click here to download this File.

Supplementary File 2: Half-pool plate map file. File containing the plate map template for pools of four and six. This is a smaller pool format that can be used when there is a high prevalence of positive individuals. Barcodes are scanned into the file, populating the plate map. The plate map is a record of the location of each sample in the tube rack, the plate throughout RNA isolation, and serves as a key to identify the wells containing each sample in the row and column pooled sample plates. This file as well as the RT-qPCR results files from the row and column pooled sample plates are required for the automated deconvolution script. Please click here to download this File.

Discussion

During sample processing, there are steps requiring careful attention. The initial quality control step which looks at the sample volume, consistency, color, and presence of added beads is critical to the overall success of the process. Tubes with samples that do not contain the correct amount of saliva could produce a false negative, as too little saliva would result in not enough genetic material; conversely, too much saliva would not be in the correct ratio with the RNA buffer and RNA degradation could occur. In rare cases, liquid would be present in the individual biohazard bags, indicating the sample tube was improperly closed and the sample leaked. These samples were subject to the rejection process to eliminate potential cross contamination resulting from handling the tube. Atypical consistency and color of the sample can also be predictive of problems. Chunks of food and non-natural colored samples suggest that food or drink may have been consumed prior to the sample being collected. This could produce a false negative due to low signal coming from saliva after the mouth was flushed, or introduction of materials that could degrade RNA or make RNA extraction an issue. Viscous samples may not flow through the columns used for RNA extraction, possibly resulting in a false negative. Blood, tissue, and mucus such as what could come from a sputum sample could also cause issues with RNA extraction, preventing the sample from flowing through the column. An additional issue that can be associated with viscous samples is during sample transfer. During the transfer to deep well plates, a viscous sample could partially attach to the pipet tip and drag across the plate, leaving a snail-trail, that results in contamination of other wells on the plate. To prevent this, centrifugation of the entire saliva sample was done to pellet small debris that could clog the columns and to force the more viscous material to the bottom of the tube. This aids in sample transfer, but also eliminates the need for the addition of proteinase K to the samples, which is time consuming, expensive, and at times in short supply. Samples must also have the same number of beads in the tube as when the kit was made. The beads function to homogenize the sample during vortexing and can also be used as supporting evidence that the RNA buffer is present, and the sample tube was not spilled by the participant during sample collection. The presence of the barcode sticker on the tube and the scanning of the tube into the correct plate position is critical to be able to trace the result in the pool back to the original tube. Likewise, careful and accurate pipetting is required for the entire process to function. In the RNA pooling portion of the protocol, automation of the process has the potential to expediate the process and reduce the potential for human error. However, the cost, supply availability, and time to implement the process must be considered in the decision to use robotics.

The automated deconvolution process employs an R script that utilizes information from three files: the plate map file and the RT-qPCR result files from the row and column pool plates. When racks of samples are scanned into the plate map file, the sample is assigned a plate and well location for the RNA extraction process. An algorithm in the plate map file takes the plate and well location of the scanned sample and determines which wells will contain the sample in the row and column pool reaction plates based on the pooling process. After RT-qPCR, both the plate map file and the result files from the row and column pool plates are uploaded to a Shiny App used to run the R script. The script assesses the RT-qPCR results on the row and column pool plates, marking a well as positive if two or more SARS-CoV-2 targets have a Ct of ≤37. The script then uses the plate map file as a key, assigning each sample a designator of positive if the sample was present in positive pools on both pool plates, or negative if the sample was present in a positive pool on only one plate or was not present in any positive pools. While suitable for the purposes of the Early Detection Program, the automated deconvolution process for the pooling also comes with caveats. For example, in an ideal situation, pools will be either be clearly positive (Ct ≤37) or clearly negative (no SARS-CoV-2 target amplification), but this is not always the case. In non-ideal test outcomes, dilution of the samples by pooling can result in a positive result in the pool of eight without a corresponding positive pool of 12, which would result in a false negative call. In other words, the script is not written to detect mismatched or non-conclusive results. As such, manual oversight to look for non-ideal test outcomes is still required. If there is ever a question about a sample's positive call based on these non-ideal test outcomes, it is always run as an individual starting at the RNA isolation step. All samples denoted by the automated process as "positive" are also run as individuals starting at the RNA isolation step. Running samples individually based off the initial pools allows for the confirmation of the positive status of the sample. This also reduces the potential of false positives, as the sample would need to have been located in two positive pools on different plates and would need to meet the positive sample criteria as an individual sample starting at the RNA isolation step. Despite these quality control steps, automated deconvolution with secondary manual oversight still has the potential to miss positive individuals with low levels of the virus. If missed the first week, these individuals will likely be positive in their next test. It has also always been the stance of the Early Detection Program that if the individual is presenting symptoms, they should isolate and find a testing site immediately.

The lab process is subject to a few technical limitations that can be difficult to avoid. The assay currently used has primer/probes targeted against the N, ORF1ab, and S genes, with amplification ≤37 cycles of two out of the three targets required for a pool or sample to be considered positive for SARS-CoV-2. Mutations in the viral genome can lead to delayed or absent amplification of specific targets. For example, amplification of the targeted area of the S gene was significantly delayed in samples positive for the B.1.1.7 (Alpha) and Omicron variants, sometimes not amplifying within 40 cycles in weak positive samples. Sanger sequencing of the S gene of positive samples where amplification was delayed or absent confirmed the identity of the variant in each case. The inclusion of the N and ORF1ab primer/probes in the assay allows for detection of the virus independent of the S gene and supports the need to diversify the target areas when developing a testing program or assay. From the samples run, it has been observed that S gene target delay is not an issue with the Delta variant, but could again be problematic with other variants, as it is with the Omicron variant. Another technical limitation occurs when there are multiple positive samples pooled together in row and or column groups. High positivity rates on a plate can occur when all samples on a plate come from a common source that is a hot spot, such as from teams or organizations. In an attempt to counter plate bias, samples need to be randomized on intake. High overall positivity rates in the testing population can also occur naturally, with more infectious variants or other factors driving up the number of local cases. If the positivity rate is ~10% or greater, numerous negative samples are flagged as potential positives during pool deconvolution, resulting in an abnormally high number of samples that need to be re-extracted and processed in the CLIA approved screen. When this occurs, additional time and reagents are needed to process samples, cutting into the cost benefits of sample pooling. To address this issue, smaller pools of four and six have been used to successfully reduce the number of potential positives that would require re-extraction and processing. The transition to smaller pools only requires changes in the plate map and the pooling pattern. In the smaller pool format, RNA is only pooled across half the plate, instead of the entire plate (Supplementary Figure 4). This process maintains the general workflow for sample pooling that has been established while increasing sensitivity and the ability to resolve positive samples when the overall positivity rate is high. Using smaller pool sizes increases the amount of reagents used during the pooling and RT-qPCR portion of the process and the cost per sample, but overall, the pooling is still more cost/time efficient than running all samples individually.

Another potential source of limitations to the lab is the lack of supply availability. Though the goal was to use supplies that were not in high demand by clinical labs or hospitals, there are common materials that have been in short supply. Specifically, the centrifuge tubes used for sample collection have been difficult to obtain at times. In these cases, alternative products needed to be substituted. For example, the tube that the participant spits in were originally a 25 mL conical tube, however, this could be substituted with a 50 mL conical or any other wide-mouthed tube to collect saliva. Pipet tip shortages were another area of supply concern. When the lab was first designed, the choice to not use robotics was decided for several reasons: 1) to invest in the futures of workers and create jobs; 2) saliva samples can be difficult to transfer due to viscosity issues; and 3) robotics require specific types of pipet tips that were in short supply. By choosing the manual route with single- and multi-channel pipets, there were many more sources where pipet tips could be purchased. Additionally, a pipet tip washer was purchased to offset the scarcity of pipet tips. The pipet tip washer was confirmed to remove all traces of DNA/RNA from the tips, allowing the reuse of non-filtered tips in the lab and cutting down on the overall cost of tips and plastic waste. Though not a limitation caused by the lack of supplies, storage space is another concern that must be considered in the lab. Large shipments of tubes, tips, and reagents all require space and appropriate storage conditions (i.e., 80, 4 L bottles of ethanol require ample flammable cabinet space).

Although the focus is on the methods, there is a critical infrastructure element with testing that must also be addressed. The process uses a homemade kit comprised of readily available components and reagents packaged in a box designed by the university's School of Packaging. The box has all the instructions on how to register and submit a sample, as well as starburst cut-outs that turn the box into a tube holder. Production of the kits requires time and labor, as well as a distribution strategy. The process utilizes numerous kit pickup sites and secured sample drop-off bins around heavily trafficked areas on campus and select off-campus sites. In addition to lab personnel, a team is required to collect samples and deliver them to the lab. Another critical element to the process is technical support and design of a sample registration and reporting system. Unique 2D barcodes paired with an alphanumeric code were attached to each tube. The registrant logs into the program website, links their barcode to their account, and can monitor the progress of the sample (when it came into the lab, when it is processed, and the final result). Furthermore, a communication team is needed to recruit participants, alert participants to any changes in the program, and answer questions participants may have. Since the start of the program, a partnership has also formed with the university's genomics core and the State of Michigan Bureau of Laboratories to track variants. Positive samples that fit the Ct cutoff criteria are sent for Sanger sequencing of the portion of the S gene where mutations have been reported. This allows the program to know when variants appear and where. Positive cases identified through the Early Detection Program are generally representative of the asymptomatic population on campus. The data is combined with clinical testing data from symptomatic cases, and self-reported data within the residential population on campus to generate outbreak curves. These trends in cases are reported along with wastewater surveillance results on a weekly basis so that appropriate actions can be suggested to leaders.

Though the primer/probe sets used in the methods are specific to SARS-CoV-2, the overall process is not limited to a single virus. SARS-CoV, H1N1, MERS, RSV, influenza, and other viruses have all been shown to be detectable in saliva^10,11,12,13. The same process used for SARS-CoV-2 could be used for any of these other viruses by switching the primer/probe targets and adjusting pool size based on viral load in saliva. The ability and blueprints to quickly scale-up the lab for future pandemics speaks to the larger implication of the methods presented. Early detection and constant monitoring when paired with isolation are effective tools to prevent spread from isolated cases. This large-scale process can also be effective in testing entire populations of cohabitants when an outbreak is suspected, as can be indicated through surveillance/screening testing of individuals, or wastewater surveillance. There are many university labs worldwide that are able to perform these tests in a surveillance or screening capacity, if not function as a diagnostic lab. The potential to make a difference already exists, but labs just need to be allowed to perform the testing and officials need to be willing to take appropriate action based on the results.

Materials

Name	Company	Catalog Number	Comments
1 Step MM, no ROX	Thermo Fisher	A28523
1.2 mlDeep well Plates	Fisher	AB0564
100 mL reagent reservoirs	Corning	4872
2.8 mm Ceramic Beads	OMNI	19-646
25 ml conical w/screw cap	VWR	76338-496
50mL V bottom reservoirs	Costar	4870
5430-High-Speed Centrifuge	Eppendorf	22620601
5ml Eppendorf Tube	Fisher	14282300
8 strip tubes for QuantStudio	life technologies	4316567
Beta Mercaptoethanol	Fisher	AC125472500
Ethanol 200 Proof, Molecular Biology Grade	Fisher	BP28184
Microamp Endura Optical 96-well fast clear reaction plate with barcode	life technologies	4483485
Microamp Fast Optical 96 well plate	Fisher	4346906
Mini Microcentrifuge	Corning Medical	6770
optical caps for strip tubes	life technologies	AB-1820
Optical Film	Thermo Fisher	4311971
PCR plate sealing film Non-optical	Fisher	AB-0558
PCR Plate semi-skirted	Fisher	14230244
QuantStudio 3 Real-Time PCR System, 96-well, 0.1 mL	Thermo Fisher	A28136
Quick RNA Viral Kit confirmation	Zymo	R1035
Reagent Reservoir, 100ml	DOT	229298
RNA Shield	Zymo	R1200-1L
Small Biohazard Bags	Fisher	180000
Taqpath RTPCR COVID19 kit	Thermo Fisher	A47814
Thermo Scientific Sorvall ST4R Plus Centrifuge	Thermo Fisher	75009525
Transfer Pipet	Fisher	22170404
Viral 96 Kit	Zymo	R1041
Vortex Mixer	Fisher	2215414