A Surrogate Enzyme-Linked Immunosorbent Assay to Select High-Titer Human Convalescent Plasma for Treating Immunocompromised Patients Infected With Severe Acute Respiratory Syndrome Coronavirus 2 Variants of Concern

Abstract

Background

The emergence of new severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants challenges the treatment of immunocompromised patients against coronavirus disease 2019 (COVID-19). High-titer COVID-19 convalescent plasma (CCP) remains one of the few available therapeutics for these patients. We have revisited the selection of CCP samples and evaluated their efficacy against the Omicron XBB.1.5 variant, the dominant strain in 2023.

Methods

A surrogate enzyme-linked immunoassay was reviewed to select CCP samples that ensure a protective level of neutralizing antibodies as the main correlate of protection. Antibody titers were analyzed in 500 serum samples from a population-based serosurvey at Mount Sinai Hospital in early 2023, and the results were validated with CCP samples (collected in 2020–2023) using an immunosuppressed mouse model.

Results

Using logistic regression modeling, we have redefined high-titer CCP against the new variant in the postpandemic era, where over 97% of the population has natural or vaccine-induced antibodies against earlier SARS-CoV-2 strains. Treatment of immunocompromised mice with two doses (100 μL/dose) of CCP plasma via intraperitoneal injection reduced lung viral titers by 46-fold 3 days post-XBB.1.5 infection.

Conclusions

These findings will guide future efforts in selecting high-titer CCP for emerging SARS-CoV-2 variants.

Graphical Abstract

Open in new tabDownload slide

Coronavirus disease 2019 (COVID-19) convalescent plasma (CCP) has been the subject of numerous clinical trials [1]. While several randomized controlled trials did not demonstrate efficacy in hospitalized patients [1, 2], studies have reported favorable clinical outcome when CCP is administered early in the disease course and contains high titers of anti–severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibodies [3, 4]. However, the introduction of monoclonal antibodies (mAbs), antivirals, and preventive measures such as vaccination rapidly diminished its use.

Not even 1 year after the pandemic was declared, SARS-CoV-2 variants of concern (VOCs) emerged and with the appearance of the Omicron subvariants, the health authorities withdrew virtually all mAb therapies. This was particularly problematic for individuals with various health conditions or those with immunodeficiencies, for whom antiviral drugs may not be compatible (Supplementary Table 1) or sufficient to provide complete protection against severe disease. Acknowledging these challenges, the United States Food and Drug Administration (FDA) issued emergency use authorization (EUA) for the high-titer CCP in patients with immunosuppressive conditions [5]. However, the therapeutic potential against current SARS-CoV-2 variants should be evaluated.

In this work we revisited CCP against SARS-CoV-2 VOCs, in the context of the XBB.1.5 variant. This variant was first identified in late 2022 and rapidly spread across regions including the United States and Europe due to its enhanced infectivity and immune evasion characteristics. To achieve this, we (1) investigated the effect of previous immunity against XBB.1.5; (2) validated a new surrogate enzyme-linked immunoassay (ELISA) to select high-titer CCP samples; and (3) assessed its therapeutic potential in an immunodeficient mouse model via intraperitoneal administration.

MATERIALS AND METHODS

Human Biospecimens

Serum and plasma samples were obtained from different cohorts: (1) 46 CCP samples with high immunoglobulin G (IgG) levels by Euroimmun AU test (median, 7.3 [interquartile range, 4.6–10.0]) collected 2020 to March 2021, from the Convalescent Plasma to Limit SARS-CoV-2 Associated Complications (CSSC-004) clinical trial (protocol number: IRB00247590); (2) 30 CCP samples from Blood Centers of America (collected January 2021, October 2021, and February 2022); (3) longitudinal serum samples from Protection Associated with Rapid Immunity to SARS-CoV-2 (PARIS) (study participants AS-01 to AS-22) and IRB-16-00791 (study participants AS-23 to AS-27) collected January–June 2020, July–December 2021, July–December 2022, and after infection in 2023 [6]; and (4) 500 uncharacterized serum samples from a serosurvey at Mount Sinai Hospital (New York) collected from February to March 2023 [7]. All participants provided informed consent for their use in the research study.

Proteins

Recombinant spike and receptor-binding domain (RBD) proteins from ancestral and XBB.1 were produced in Expi293F expression system [8]. At the time of the assay development, we did not have access to recombinant XBB.1.5 proteins; hence, XBB.1 was used as a proxy. XBB.1 and XBB.1.5 differ only in 2 mutations (G252V in the N-terminal domain [NTD] and F486P in the RBD). Recombinant SARS-CoV-2 NTD (catalog #40591-V49H), S2 subunit of spike (S2; catalog #40590-V08B), and ancestral nucleocapsid (N; catalog #40588-V07E) were purchased from Sino Biological (China).

Enzyme-Linked Immunosorbent Assay

Antigen-specific IgG titers were measured as previously described [8]. Antigens were coated at 2 μg/mL overnight at 4°C. Plates were washed and blocked for 1 hour at room temperature (RT). Serum samples were serially diluted 5-fold (initial dilution 1:50) in blocking solution followed by a 2-hour incubation at RT. ELISA plates were washed and anti-human IgG horseradish peroxidase (HRP) antibody (1:3000, Cytiva, catalog #10547065) was added for 1 hour. After washing, plates were developed using SigmaFast OPD (Sigma-Aldrich, catalog #P9187) for 10 minutes. Reactions were stopped by adding 3 M HCl and 492 nm absorbance was determined. Blank average absorbance plus 3 standard deviations was used as a cutoff to determine endpoint titers and area under the curve (AUC) using GraphPad Prism version 9.5.1. Results were standardized to binding antibody units per milliliter (BAU/mL) with the World Health Organization (WHO) standard National Institute for Biological Standards and Control (NIBSC) code 21/234 calibration.

Replication-Competent Vesicular Stomatitis Virus–Enhanced Green Fluorescent Protein SARS-CoV-2 Spike Neutralization Assay

Replication-competent vesicular stomatitis virus carrying an enhanced green fluorescent protein (GFP) reporter and expressing ancestral, XBB.1.5, or BQ.1.1 SARS-CoV-2 spike was produced as described previously [9]. Sera were heat-inactivated at 56°C for 30 minutes before use. Virus stocks were premixed with 10-fold serially diluted serum (initial dilution 1:10) and incubated for 15 minutes at RT. The virus–serum mix was subsequently transferred onto the BHK-ACE2 cells. Twelve hours postinfection, GFP counts were measured using a Celigo imaging cytometer (Nexcelom Biosciences, version 4.1.3.0) and the inhibitory dilution at which 50% neutralization is achieved (ID₅₀) was calculated [9]. Results were standardized to international units per milliliter (IU/mL) with the WHO standard NIBSC code 21/234 calibration.

Live Virus Neutralization Assay

Neutralizing antibodies against SARS-CoV-2 HP7 isolate were determined in Vero-E6-TMPRSS2 as previously described [10]. HP7 is an early isolate of SARS-CoV-2 (SARS-CoV-2/USA/DC-HP00007/2020, GISAID sequence EPI_ISL_434688) containing the spike D614G mutation (HP7) isolated from a COVID-19 patient at Johns Hopkins Hospital [11, 12]. Neutralizing antibody titers were calculated as the highest serum dilution that eliminated the cytopathic effect in 50% of the wells, and the AUC was calculated using GraphPad Prism 9.5.1 software.

Logistic Regression Model

The logistic regression models were calculated using GraphPad Prism 9.5.1 as follows:

where P(Y = 1) is the probability of the event Y happening, with CCP being protective (1);

X is the predictor variable, here it is the anti-RBD IgG titer (BAU/mL); β0 is the intercept; and β1 is the coefficient for the predictor variable from anti-RBD IgG titer (BAU/mL).

The 500 samples obtained from an uncharacterized cohort [7] were utilized to develop the model [13]. XBB.1.5 ID₅₀ titers higher than the defined thresholds were assigned a value of “1” to denote treatment success, while values below were classified as “0,” indicating treatment failure.

Severe Combined Immunodeficiency Mouse Model

Mice experiments were performed in accordance with the Institutional Animal Care and Use Committee and Biosafety Level 3 (BSL-3) biocontainment facility of the Global Health and Emerging Pathogens Institute at Icahn School of Medicine at Mount Sinai.

Female CBySmn.Cg-Prkdc^SCID/J mice (BALB/c severe combined immunodeficiency [SCID], Jackson Laboratories, RRID:IMSR:JAX:001803) were used in these studies. Eight- to 10-week-old mice were anesthetized with a ketamine/xylazine cocktail and intranasally infected with 5 × 10⁴ plaque-forming units (PFU) of Alpha B.1.1.7 variant or a mouse-adapted XBB.1.5 [14]. Mice were treated with CCP samples given intraperitoneally at different time points. Lung viral titers 3 days postinfection were used as the readout for protection. The right lung lobes were formalin-fixed, paraffin-embedded (FFPE), and the left lung lobes were harvested and homogenized in 1 mL of sterile phosphate-buffered saline.

Immunohistochemistry

FFPE tissues were cut and stained by the Biorepository and Pathology Core using a Ventana Discovery Ultra instrument (Roche, Switzerland). SARS-CoV-2 was stained using a goat polyclonal anti-N antibody (1:750, Novus Biologicals, catalog #NBP312090).

SARS-CoV-2 Plaque Assay

SARS-CoV-2 plaque assays were performed in Vero-E6-TMPRSS2-T2A-ACE2 under BSL-3 conditions as previously described [9]. The plaques were immunostained with an anti-SARS-CoV-2 NP 1C7C7 (1:1000) mouse mAb and an HRP-conjugated goat antimouse secondary antibody (1:2000, Thermo Fisher Scientific, catalog #31430) and read using TrueBlue Peroxidase Substrate (SeraCare Life Sciences, catalog #5510-0030).

RESULTS

Immune responses against SARS-CoV-2 XBB.1.5 variant are cross-reactive, mainly dictated by previous exposure to the ancestral virus.

Immune imprinting describes how the first exposure to a virus shapes the immunological outcome of subsequent exposures to antigenically related strains [15]. We investigated longitudinal immune responses in 14 participants from the PARIS cohort [6, 16] from samples collected between 2020 and 2022. XBB.1.5 started circulating in late 2022; hence, the likelihood of significant exposure to that variant remains low in the cohort. Neutralizing antibody titers were measured at different time points using a pseudotyped virus assay expressing the ancestral and Omicron XBB.1.5 spike (Figure 1A and Supplementary Figure 1).

Neutralizing antibody titers were observed to remain high over time against the ancestral strain for all participants [16]. A similar trend but with low titers was seen against XBB.1.5 (Supplementary Figure 1). An average 32-fold reduction was observed between the 2 strains, independent of the exposure event and time point (Figure 1B–F). Overall, cross-reactive humoral immune responses against XBB.1.5 were detected in human serum samples caused by previous SARS-CoV-2 infections and vaccinations.

Development of a Surrogate ELISA for the Selection of High-Titer CCP Samples

Given the high level of preexisting immunity in the population, we next revisited the surrogate method to select high-titer CCP samples. The term “high titer” lacks a universal consensus but generally implies antibody levels associated with favorable clinical outcomes, which were studied during the first SARS-CoV-2 waves [3, 17, 18]. Under current EUA, CCP units must demonstrate high titers according to specified 2021 tests [13], which predominantly focus on quantifying IgG serum antibodies targeting the ancestral spike [19, 20]. Hence, they may not consistently capture neutralizing antibodies against new VOCs, like XBB.1.5 [21].

We evaluated the level of neutralizing antibodies against the ancestral and XBB.1.5 SARS-CoV-2, and studied their correlations to binding antibody titers with a panel of SARS-CoV-2 antigens by ELISA (Figure 2, Table 1, Supplementary Figures 2 and 3). We quantified antibody titers targeting the N protein and the spike protein or its subunits (RBD, NTD, or S2 subunit) of ancestral or XBB.1 strains. We used blood serum samples collected from February to March 2023 from a population-based serosurvey at Mount Sinai Hospital [7] (Figure 2A).

Significant Spearman correlation coefficients were obtained for the spike protein or its subunits (Table 1). Lower correlation coefficients were found between XBB.1.5 neutralizing titers and the different ELISAs compared to the ancestral correlations (Table 1). The use of 500 samples and the high preexisting immunity against previous variants might explain these low values. For both ancestral and XBB.1.5 neutralizing antibody titers, the strongest correlations were observed against its homologous anti-spike and anti-RBD antigens (Table 1, ie, correlation coefficient of r = 0.92 for the ancestral and r = 0.61 for Omicron XBB.1.5 using the anti-RBD proxy antigen).

Next, we employed a logistic regression model to define an ELISA threshold to select high-titer CCP plasma against Omicron XBB.1.5. We used 60 IU/mL of anti-XBB.1.5 neutralizing antibodies as our correlate of protection from hospitalization similar to neutralizing antibody titer of 30–60 IU/mL reported in a previous work, which demonstrated to be protective when transfused early in the outpatient setting [3, 23] (Figure 2 and Supplementary Figure 4). These data were reported in the supplementary files of those articles. Using this model, an anti-RBD IgG titer against XBB lineage of 124 BAU/mL (95% confidence interval [CI], 52–295 BAU/mL) and a titer of 2030 BAU/mL (95% CI, 852–4837 BAU/mL) predicted a 50% and 90% probability of a potential favorable clinical outcome, respectively (Figure 2B). We also generated a similar model using the other antigens tested. If the ELISA is not performed against the VOC but against the ancestral virus, higher antibody titers are required for a favorable clinical outcome (616 BAU/mL [95% CI, 259–1384 BAU/mL] and 23 478 BAU/mL [95% CI, 10 458–52 711 BAU/mL] for a 50% and 90% probability, respectively; Figure 2C).

Different estimates of protective neutralizing antibody titers have been described in the literature, varying from 27 to 1444 IU/mL [3, 24, 25]. We also defined other logistic regression models using those thresholds (Supplementary Table 1). Clinicians can refer to these tables to identify the anti-RBD IgG titers associated with those ID₅₀ thresholds.

Validation of the Surrogate ELISA for High-Titer CCP Selection

To validate our methods, we obtained convalescent plasma from different cohorts (Figure 3A) [3, 6]. We measured ancestral and XBB lineage–specific binding (Figure 3B) and neutralizing titers (Figure 3C). Of note, pseudotyped virus neutralization titer was correlated to live virus neutralization for those samples [3] (Supplementary Figure 5).

A slight increase in anti-ancestral IgG levels was observed among the cohorts with time (Figure 3B). In contrast, anti-XBB IgG titers increased progressively over time, suggesting a relationship between the phylogenetic proximity of the circulating variant at the time of the CCP collection and the recent SARS-CoV-2 variant, like the XBB.1.5.

Neutralizing titers against the ancestral virus were very high in the first cohort analyzed, due to the selection of only high-titer CCP units [3] (Figure 3C). Titers observed in the subsequent cohorts presented lower titers, with an average of 1 × 10³ IU/mL. Notably, plasma collected prior to 2022 displayed negligible or very low neutralizing activity against the XBB.1.5 variant. This finding highlights the importance of cross-reactivity considerations for CCP therapy.

We used our previously defined logistic regression model (Figure 2) to analyze the probability of achieving a favorable clinical outcome with these CCP samples based on ELISA results. As shown in Figure 3D, regardless of whether the ELISA was directed against the ancestral virus or the Omicron XBB variant, similar probabilities were observed, with no significant difference between the geometric means (P = .4206, nonsignificant with Mann-Whitney test). The probability of a favorable clinical outcome increased over time, ranging from 11.7% at the beginning of the pandemic to 78.4% for plasma collected in 2023.

Focusing on 2023 breakthrough infections, we evaluated the ID₅₀ titers against BQ.1.1, a distinct sublineage of the Omicron variant that also emerged in 2023 (Figure 3A). All samples contained very high levels of neutralizing antibodies against the ancestral virus. However, only 67% and 75% of the samples had neutralizing antibodies above the protective threshold against XBB.1.5 and BQ.1.1, respectively. Interestingly, the samples that were protective against XBB.1.5 were also protective against BQ.1.1 (Figure 3E, different samples identified by numbers). Hence, in the absence of a validated test to select CCP samples against BQ.1.1 or any other current Omicron variant, our surrogate XBB.1.5 ELISA method proves to be better than current 2021 FDA-specific tests [19].

An SCID Mouse Model to Evaluate CCP Therapeutic Efficacy

To assess the therapeutic effect of CCP in immunocompromised individuals, we established a novel animal model following the recommendations for CCP use [13]. We selected SCID mice, characterized by a BALB/c background with a mutation inducing severe combined immunodeficiency (Prkdc^SCID).

Mice were intranasally infected with the Alpha B.1.1.7 variant with 5 × 10⁴ PFU. Mice were treated with 100 µL of clinically validated CCP samples [3] (Figure 3A) collected during the Alpha era (Figure 4B). CCP was administered at 1 and 2 days postinfection as a therapeutic countermeasure, which is equivalent to the 2 units of high-titer CCP recommended for human treatment [13]. SARS-CoV-2 viral titers in the lungs were measured 3 days postinfection as described in another study [27]. For comparison, 200 µL of CCP was injected 2 hours prior to infection as a prophylactic control group (Figure 4A).

Mice did not experience significant weight loss and survived the infection (Figure 4C). Treated groups showed no viral titers (95% CI, 50–50) compared to 7.3 × 10⁵ PFU/mL (95% CI, 1.4 ×10⁴ to 3.7 ×10⁷) observed in the untreated group (Figure 4A). These results correlate with the presence/reduction of SARS-CoV-2 N protein in the lungs (Figure 4E).

CCP Protection Against Omicron XBB.1.5 Infection in the SCID Mouse Model

In the case of an Omicron XBB.1.5 infection, we used a mouse-adapted version of XBB.1.5 [14]. The CCP samples used were collected in 2023 from the PARIS cohort at 10, 47, and 60 days postinfection, with a probability of favorable clinical outcome of 90% and 80% against XBB.1.5 infection when using the XBB.1 and ancestral models, respectively (Figure 5B). These samples were selected as the ones that provided the highest probabilities of protection in that cohort. All mice survived and did not exhibit any significant weight loss (Figure 5C). A lung viral titer reduction of 46-fold and 256-fold was observed for the therapeutic and for the prophylactic group, respectively, compared to the nontreatment control (Figure 5D). These outcomes were consistent with immunohistochemistry results (Figure 5E).

DISCUSSION

The use of convalescent plasma from patients with COVID-19 remains an important approach for treating immunocompromised individuals. Among the different therapeutic activities presented by this immunoglobulin-based therapeutic, the primary correlate of protection from hospitalization is its neutralizing activity [28] although in the future, other correlates of protection might complement these results, such as the contribution of Fc-receptor functions or anti-inflammatory cytokines [29, 30]. In this work, we have used the neutralizing capacity against the homologous Omicron XBB.1.5 as our correlate of protection to revisit the definition of high-titer CCP against VOCs.

We developed a surrogate ELISA for the selection of CCP samples that provides a practical method to estimate the level of neutralizing antibodies against the circulating VOCs. ELISA is an easy, accessible, and cost-effective method that could be rapidly implemented in many clinical settings. Although the correlation between ELISA and neutralization titers is low, the logistic regression model is significant, and provides valuable insights, particularly in resource-limited settings. We tested several SARS-CoV-2 antigens and showed that RBD provided the highest correlation of binding and neutralizing antibodies. We determined a threshold of 124 BAU/mL and 616 BAU/mL when using the XBB-lineage RBD and the ancestral RBD, respectively. There is a 5-fold difference in the IgG threshold when using the ancestral antigen compared to the homologous variant. This highlights the increased immune escape of new VOCs and the need for updated high-titer CCP selection thresholds.

Due to the limited clinical data on high-titer CCP treatment in the postpandemic era, we implemented a preclinical model to evaluate the protective effect of CCP in a research setting. We demonstrated that CCP treatment resulted in a reduced SARS-CoV-2 viral loads in an infected immunocompromised mouse model. This efficacy of CCP was observed for Alpha infection, with clinically validated samples, and for XBB.1.5 infection with convalescent plasma samples collected in 2023, when XBB.1.5 was dominant.

The present work has focused on XBB.1.5 variant as a model virus. However, the results are relevant for all SARS-CoV-2 VOCs. We believe the present findings have significant implications for the selection of high-titer CCP in the postpandemic era and may be applicable to phylogenetically related variants as shown for BQ.1.1. However, further validation will be necessary to confirm these applications. One of the main limitations is the need to perform these correlations empirically since SARS-CoV-2 preexisting immunity in the population and immune evasion of emerging variants is continuously changing. Considering recent evidence suggesting the potential of an antibody-based therapy to treat persistent infections [1], high-titer CCP could be also an important asset to improve long-COVID treatment.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The graphical abstract was created by Jill Gregory (Medical Illustrator, Mount Sinai). Figures were created with BioRender.com (https://BioRender.com/d72t319). We thank all of the study participants for their generosity and the study team members of the Personalized Virology Initiative for recruiting and following participants, collecting data and biospecimen, and processing and archiving biospecimen. We kindly appreciate the work of Dr Randy Albrecht, Carlos Franco, and the Center for Comparative Medicine and Surgery team overseeing the Conventional Biosafety Laboratory Level 3 (C-BSL-3) and Animal BSL-3 facilities (Icahn School of Medicine at Mount Sinai [ISMMS], New York). Immunohistochemical analyses were performed with the help of Dr Monica Garcia-Barrios in the Biorepository and Pathology Core (ISMMS). We also thank Dr Ralph Baric (University of North Carolina, Chapel Hill) for providing the variant mouse-adapted XBB.1.5 strain. BHK-ACE2 cells and initial pseudotyped virus neutralization assay were provided by Dr Benhur Lee (ISMMS). The help of Shreyas Kowdle with the Celigo instrument is also greatly appreciated.

Author contributions. Conceptualization and design: V. D., D. S., P. P., and I. G.-D. Pseudotyped virus rescue and characterization: S. S., N. L., W. S., and I. G.-D. Animal experiments: V. D., A. A., T. Y. L., and I. G.-D. Protein and virus reagents: G. S., J. M. C., A. A., and F. K. Cohorts: K. S., V. S., D. S., and F. K. Live virus neutralization assay: J. S., A. P., and D. S. Data analysis: V. D., P. P., and I. G.-D. First draft of the manuscript: V. D., P. P., and I. G.-D. All authors reviewed and approved the manuscript.

Financial support. This work was partially funded by the Centers of Excellence for Influenza Research and Response (contract number 75N93021C00014 to P. P.) and the Collaborative Influenza Vaccine Innovation Centers (contract number 75N93019C00051 to P. P.).

Data-sharing statement. Data will be available in Immport:SDY2918.

References

Author notes