Standardization of enzyme-linked immunosorbent assays for serosurveys of the SARS-CoV-2 pandemic using clinical and at-home blood sampling

The extent of SARS-CoV-2 infection throughout the United States population is currently unknown. High quality serology is a key tool to understanding the spread of infection, immunity against the virus, and correlates of protection. Limited validation and testing of serology assays used for serosurveys can lead to unreliable or misleading data, and clinical testing using such unvalidated assays can lead to medically costly diagnostic errors and improperly informed public health decisions. Estimating prevalence and clinical decision making is highly dependent on specificity. Here, we present an optimized ELISA-based serology protocol from antigen production to data analysis. This protocol defines thresholds for IgG and IgM for determination of seropositivity with estimated specificity well above 99%. Validation was performed using both traditionally collected serum and dried blood on mail-in blood sampling kits, using archival (pre-2019) negative controls and known PCR-diagnosed positive patient controls. Minimal cross-reactivity was observed for the spike proteins of MERS, SARS1, OC43 and HKU1 viruses and no cross reactivity was observed with anti-influenza A H1N1 HAI titer during validation. This strategy is highly specific and is designed to provide good estimates of seroprevalence of SARS-CoV-2 seropositivity in a population, providing specific and reliable data from serosurveys and clinical testing which can be used to better evaluate and understand SARS-CoV-2 immunity and correlates of protection.

reference, well bottom is roughly 3.4 mm above the plate carrier. Once all plates have been washed with block added into the wells and returned to the output stack, all plates are then restacked into the original order into the input stack.
It is necessary to clean the syringe B dispense head and line soon after the block has been added to all plates to prevent clogging of the dispense tips. Each time a dispense happens throughout this protocol it will be assumed that the following cleaning protocol has been performed for the syringe head used. The bottle cap with the liquid sourcing end of line is fastened into an empty bottle and primed with 8 mL of air two times sequentially, next the same cap and line are secured into a bottle containing deionized water and primed with 8 mL five times, a clean empty bottle is then hooked up again and primed with 8 mL of air two times, next the cap and line are placed onto a bottle containing 70% ethanol and primed five consecutive times with 8 mL, followed up again with two 8 mL primes of air and a final deionized water rinse priming 8 mL five times. Before a new reagent is primed through the respective line two primes of 8 mL using a clean empty bottle is completed for separation of water and new reagent.
(5) After the 2 hour incubation (6) the plates are washed 3x with PBST to remove blocking solution and roughly 30 μL is left in all wells to prevent drying before sample addition. Here on out, it will be assumed that incubation steps happening at room temperature will be stacked in the BioTek input stacker with a lid, which will be removed before beginning next step, on the top plate. The height of the aspiration pins for the first two wash cycles is 3.68 mm above the carrier and the final aspiration happens at 4.70 mm above the plate carrier in order to retain some wash fluid as mentioned. (7) Currently samples have been added to the ELISA plates using a multichannel pipette; for the scaled-up process an Integra Biosciences VIAFLO electronic 96 channel pipette can be used to transfer 100 μL of the samples from a 96 deep-well block into multiple ELISA destination plates.
(8) Once the 1 hour incubation ends (9 + 10) all wells in the plates within the associated batch will be washed 3x with PBST followed by the dispensing 100 μL of the respective antibody into all wells of the 96-well plate using the BioTek EL406 washer/dispenser. The height of the aspiration pins for all three wash cycles is at 3.43 mm above the plate carrier and also include a secondary aspiration at the same height 2.06 mm to the left of well center. Each addition of PBST wash fluid also incorporates a 15 second soak prior to aspiration. This washing procedure is critical to decreasing background and fully removing the samples (primary antibody). Plates are restacked into the original order within input stack.
(11) After the 1 hour incubation, (12) all wells in the batch of plates are washed 2x with PBST and the final dispense of 300 μL is added using the 96-channel wash head. The height of the aspiration pins for these two cycles is 3.68 mm above the carrier, also includes a secondary aspiration at the same height 2.06 mm to the left of well center, and each addition of PBST wash fluid incorporates a 25 second soak prior to aspiration. The difference in soak time from the previous process of washing the samples out of the wells allows the timing to stay consistent for each assay plate. Plates are then restacked into the original order within the input stack with the remaining 300 μL of PBST.
Performing the initial two washes with PBST and leaving the last 300 μL in the wells enables the addition of substrate to happen quicker than if all three washes were to happen prior to substrate addition. With the critical nature of the 10 minute incubation for substrate and to increase throughput, it is desired to systematically increase the speed of substrate addition as in this protocol.
(13) The remaining PBST is removed from all wells of the plates using the BioTek 96-channel wash head.
A height of 3.43 mm above the plate carrier is set for the aspiration pins and a secondary aspirate at the same height 2.06 mm to the left of well center is performed. A volume of 100 μL of substrate is added to all wells using syringe A dispense head which is the designated line used for substrate only as it is a sensitive reagent. As plates are dispensed as loaded from the input stack they are returned to the output stack and after the last plate is finished all plates are restacked in the original order within the input stack. (14) Each plate is incubated with the substrate solution for 10 minutes.
(15) It is critical that the stop solution is added after the 10 minute incubation, therefore the plate batch size at this step is adjusted accordingly. As soon as the first plate has finished incubating with substrate the Biotek protocol for adding 100 μL of the stop solution using syringe B dispense head is started. An important note regarding the stop solution used in this ELISA protocol is that when 100 μL was set for the dispensing variable approximately 60 μL was actually added to the wells. In order to compensate for the characteristics of this liquid 175 μL was set as the dispense variable for the stop solution addition. It will be checked regularly for volume accuracy. Once all plates have received stop solution they are restacked into the original order into the input stack.
(16) Absorbance is measured at 450 nm and 650 nm. Data was initially collected using the BMG Labtech PHERAstar FSX, and the BioTek Epoch 2 has subsequently been used. The BioTek stacker containing the plates which have been dispensed with stop solution and returned to the original order is removed from the BioTek input stack position and placed into the input stack position for the Epoch 2. The read protocol is started, and plates are fed into the Epoch 2 for data collection. The BioTek EL406 96-channel wash head is cleaned at the end of each day to ensure no reagent buildup or clogging happens. This entails flushing 350 mL of deionized water followed by 350 mL of 70% ethanol and finally with 350 mL of deionized water. The 96-channel wash head dispense tips as well as the aspiration pins are cleaned in this process.
Based on thorough titering studies of both archival and PCR+ SARS-CoV-2 diagnosed patient sera (Supplementary Figure 7), we determined that the optimal titrations for the sample into ELISA were 1:400 to 1:1000 for IgG, and 1:400 for IgM and IgA (as IgM and IgA are at lower quantities in the blood when compared to IgG). Microsampler dilution was adjusted accordingly to 1:10 for IgM/IgA and 1:25 for IgG into PBS + 5.0% NFDM, or PBS + 0.03% Tween20, respectively, to ultimately result in 0.05% Tween20 in the final diluted sample for both 1:10 and 1:25 dilutions.

Statistical Modeling
Let ! , " be the values from the census for proportion of people in each population. Let be the number of people in the sample. Let !# be the th sampled response and ! the number ofobservations from population 1 and "# be the th sampled response and " the number of observations from population 2.
The weights for an observation in population 1 or 2 are ! = $ ! % ! /% and " = $ " %"/% . The unadjusted estimate of the prevalence is the weighted estimate of the mean, which is where # are the observed proportions in each population. The variance of ) is Lang and Reiczigel 1 (2014) provide a prevalance estimate that adjusts for the sensitivity and specificity of the assay, taking into account the variability of the estimate of prevalence as well as the variability in the estimation of sensitivity and specificity. Thus, the Lang-Reiczigel method requires the sample sizes for validation of the assay, specifically, the number of true positives used to estimate the sensitivity, and the number of true negatives used to estimate the specificity. Although the method was developed for simple random samples, we make the following modification for other survey designs. Let ) and / ( )) be the prevalence and variance estimates under the design with perfect sensitivity and specificity. We wish to the know the sample size for the estimate of ) that would equal the variance of / ( )). Therefore set where the righthand side of the equation is modeled after the binomial variance from a simple random sample. In other words, the effective sample size is the simple random sample size that would give similar variance to the variance of the weighted design. Then we solve for )** so Then we input ) and )** into the Lang-Reiczigel method in place of the simple random sample prevalence estimate, , and its sample size, .
Here are the Lang-Reiczigel equations with those modifications 1 . The estimate of prevalence adjusted for sensitivity and specificity is with and being the estimates of specificity and sensitivity, and 0, and 0) being the number of observations used to estimate specificity and sensitivity, with being the upper quantile from the standard normal distribution with probability 1-/2, where = 1 − and is the confidence level for the confidence interval (e.g., 95%). Then The estimate of the confidence intervals are given by there is a typo in equation 4 of Lang and Reiczigel that defines / ( ), in our notation the typo is that the (1 − 5 ) " expression was mistakenly written as (1 + 5 ) " .)   8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30   31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58    Red line = threshold of lowest titer suggested to be used.  Archivals 101-200 run on automated setup