Efficient Identification of High-Titer Anti–Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Antibody Plasma Samples by Pooling Method

We collected 2046 individual discarded/residual ethylenediaminetetraacetic acid whole blood specimens from blood donors from the Stanford Blood Center. These samples represent healthy volunteer blood donors as well as a small subset of CCP donors. Pooled samples were generated by combining the same 2046 individual plasma samples into pools of 6 in equal proportions, resulting in 341 pooled samples. An additional 10 218 samples were collected between April 2020 and July 2020. These 10 218 samples were tested in pools of 6 to determine the seropositivity rate in our blood donor population.

A subset of our donor samples (1413 individual samples) was tested individually and in parallel using the VG assay, a testing platform granted Emergency Use Authorization by the FDA²⁴  for the purpose of defining high-titer in CCP samples. Separate serum samples from the same collection were evaluated on the VG platform. Results were reported as either positive or negative for anti–SARS-CoV-2 IgG antibodies against the spike protein.³⁴  In addition to the qualitative antibody result, a signal-to-cutoff value for each sample was also provided for additional reference. In addition, 19 of 1413 individual samples were known CCP donors.

The ELISA procedure we used in this study was modified³⁵  from a protocol provided by Stadlbauer et al.³⁶  Plates were coated with 0.1 μg/well of SARS-CoV-2 RBD, which was expressed from a plasmid kindly provided by Florian Krammer, PhD, and incubated at 4°C overnight. Wells were washed with 300 μL of 0.5% tween in phosphate-buffered saline 3 times and blocked with 0.5% tween phosphate-buffered saline containing 3% non-fat milk powder for 2 hours. Plasma samples were diluted 1:100 in 0.5% tween phosphate-buffered saline containing 1% non-fat milk powder for individual plasma testing. Of pooled samples, 3 μL were diluted to 100 μL with 1% non-fat milk powder in 0.5% tween phosphate-buffered saline. Wells were washed, and 50 μL of the diluted plasma sample was added to the wells. Plates were sealed and incubated at 37°C for 1 hour. IgG detection antibodies (Invitrogen, Carlsbad, California) were diluted 1:6000, and 50 μL were added to each well following another wash. Plates were incubated at room temperature for 1 hour. Plates were washed again and 100 μL of 3,3′,5,5′-Tetramethylbenzidine was added to each well. Wells developed for 12 minutes and stop solution was added.

As negative control samples for our assay development, we used 96 serum samples collected from healthy volunteer blood donors in 2018 and tested each sample individually. The threshold for positivity for our individual sample ELISA (0.198) was defined as an odtical density (OD) 450 value 3 SDs above the highest OD450 in the sample set.

For pooled testing, the same 96 samples were pooled into 16 pools of 6 and tested using ELISA. Similar to the individual testing, the positivity threshold for P-RE (0.218) was defined as 3 SDs above the highest OD450 value in our negative control pooled sample set.

Two thousand forty-six blood donor samples were assayed individually by RBD ELISA (I-RE), and 57 samples were determined to be positive for anti-RBD IgG. These 2046 samples were assayed in parallel in 341 pools of 6 specimens by RBD ELISA (P-RE). Every positive pool was confimed by testing the individual samples it contained. Of 57 positive samples identified by I-RE, 32 were detected by P-RE, equating into a sensitivity of 56% (Figure 1, A and B). To confirm that the P-RE false negatives were indeed due to low-positive samples by I-RE that were diluted when pooled, we compared I-RE OD values of samples that were positive by both P-RE and I-RE against those that were positive by I-RE but negative by P-RE. Positive individual samples in positive P-RE pools had a significantly higher average OD value compared with positive individual samples in false-negative P-RE pools (0.77 versus 0.34; P < .001; Figure 1, A), confirming the lower sensitivity of P-RE testing. Of 284 individual plasma samples that tested negative, 2 came from positive pools resulting in a specificity of 99% (Figure 1, B). The 2 pooled samples that were negative by I-RE had relatively low P-RE absorbance values (0.27 and 0.33); we individually tested the 12 samples comprising these 2 pools and identified 1 individual sample in each pool with relatively high yet negative absorbances (0.11 and 0.12).

To further evaluate the performance characteristics of P-RE, we compared it against individual samples tested on the VG antibody platform, which has been recognized by the FDA to identify high-titer CCP donors. One thousand three hundred and twenty individual samples (including 12 CCP) tested by VG were pooled into 220 samples and assayed in parallel by P-RE. VG identified 24 positive individual samples, of which 20 were detected concordantly in pools by P-RE, giving a sensitivity of 83% (Figure 2, A). The P-RE OD450 values showed statistically significant correlation with the signal-to-cutoff values derived from VG (R² = 0.58, P < .001, Figure 2, B). In addition, we evaluated positive VG individual samples that were not detected in P-RE testing. Similar to the P-RE missing low-positive I-RE above, we found that P-RE also missed low-level positives by VG (Figure 2, C). Of note, for the 18 samples that were above the VG cutoff to be considered as high titer (signal-to-cutoff value > 9.5), all were detected on P-RE, indicating that pooling can accurately detect potential high-titer CCP donations. Of note, 1 sample was identified as positive by P-RE, and subsequently confirmed by I-RE, but negative by VG. This sample had a P-RE signal of 0.22.

We also compared the performance charactertistics between I-RE and VG. One thousand three hundred and ninety-four individual samples were assayed using I-RE and VG (Figure 3, A). We observed a R²  value of 0.82 when examining correlation between the values measured for individual samples in the 2 assays (Figure 3, B). Furthermore, 10 positives were identified through VG. Of these, 9 were also positive with I-RE testing, as well as an additional 12 positive specimens (Figure 3, A, C, and D).

The recent FDA guidance further outlined the requirement for high-titer CCP in COVID-19 patients.²⁴  Because P-RE misses low-level positive samples but can detect high-titer plasma samples from healthy blood donors, we next evaluated whether P-RE could serve as an effective strategy to screen for CCP donors. At Stanford Blood Center, individuals self-identify and register as possible CCP donors based on a previous laboratory-confirmed diagnosis of COVID-19 and having been asymptomatic for at least 2 weeks. Using VG, 19 CCP samples were assayed and 18 positives were identified, indicating a sensitivity of 95%. We tested 31 known CCP samples by I-RE and 32 CCP samples in pools by P-RE and were able to detect 94% (29 positives) and 88% (28 positives) of the CCP samples, respectively (Figure 4, A). In addition, 28 CCP samples were tested using both I-RE and P-RE to directly compare their abilities in detecting CCP. I-RE proved to be more sensitive by detecting 2 more CCP than P-RE (Figure 4, B).

We applied the P-RE methodology to 1703 pools collected from 10 218 healthy blood donors from April 19, 2020, to July 17, 2020. Positivity rate was tracked over time (Figure 5, A) and compared with virology testing data collected by Santa Clara County (SCC). Within the timeframe of our study, SCC observed weekly positivity rates reach as high as 10.81%; our study, on the other hand, had weekly positivity rates that peaked at 0.92% (Figure 5, A). Although the magnitude of the positivity rates differed, the peaks and troughs between the county data and ours coincided temporally. Overall, 43 of 1703 non-CCP pools tested positive which equates to an individual sample positivity rate of 0.4%. Furthermore, we observed a R² value of 0.21 when examining correlation between the positivity rates of the 2 populations (Figure 5, B).

COVID-19 continues to spread widely and cause morbidity and mortality worldwide, with therapeutics only showing modest effectiveness in a significant amount of cases.^6–8  One promising form of treatment is the use of high-titer CCP for passive immunity in hospitalized COVID-19 patients. Even with effective vaccines beginning to be approved and distributed, manufacturing and resource constraints mean that it will be months to years before vaccines are available worldwide, and sizeable fractions of the population in some countries may decline to be vaccinated.⁸  CCP will continue to be a treatment option in many parts of the world for the foreseeable future. Here, we have validated an efficient pooling and ELISA testing methodology to screen large numbers of samples for high titers of neutralizing antibodies to identify potential CCP donors.

In testing of 2046 individual samples in 341 pools, we found that only 56% of I-RE–positive individuals were also identified by P-RE, showing the expected decreased sensitivity of pool testing. The ELISA OD values of positive I-RE samples that were missed by P-RE were lower than those of samples that were positive by both methods. P-RE performed well, with 100% sensitivity in detecting individuals with high titers to RBD, which is the clinically relevant goal in screening for eligible CCP donors.

Furthermore, we compared both in-house assays with the VG, which has FDA Emergency Use Authorization status for identifying high-titer CCP donors.²⁴  Overall, we found that I-RE was more sensitive than VG when testing individual samples, while the sensitivity of P-RE was lower. This indicates that the decreased sensitivity of P-RE when compared with VG was due mainly to dilutional effects from pooling, and not likely from other technological differences between the 2 platforms. Despite the lower sensitivity of P-RE overall, we next explored its potential to be an efficient method for screening high-titer CCP samples.

We assayed CCP samples using both in-house methods in comparison to VG. Although P-RE demonstrated overall lower sensitivity than individual sample testing either by I-RE or Vitros assays, it nonetheless demonstrated high accuracy for use in CCP detection; all CCP samples detected individually by VG were successfully identified in pools by P-RE. Furthermore, P-RE's potential to increase efficiency 6-fold outweighs its reduction in sensitivity for low-positive samples. Our proposed use of P-RE favors specificity above sensitivity in screening for high-titer CCP for use in passive immunity. We believe it can be used to efficiently distinguish potential CCP donors from a large sample population. In addition, a set of CCP samples were separately pooled with non-CCP samples and was succesfully detected using P-RE, further supporting that P-RE is capable of detecting CCP efficiently.

We tested an additional 1722 samples by P-RE to track SARS-CoV-2 exposure over time and compared our results with our local county's public health COVID-19 testing data. Within the timeframe of our study, we compared the rate of viral nucleic acid test positivity rate in SCC with our own serologically positive blood donor population by P-RE. Although the serologic positivity rate of the blood donor population was significantly lower than the viral positivity rate of SCC, both rates trended similarly through time, suggesting that pooled testing of the blood donor population is a potential method to monitor trends in COVID infectivity rates in the general local population.

Our findings indicate that P-RE can be used to efficiently screen large healthy populations for high-titer SARS-CoV-2 antibody CCP donor identification, saving time and resources. Use of high-titer CCP in passive immune therapy is likely to remain a viable treatment option, particularly in lower resource regions.³⁷