No link between type I interferon autoantibody positivity and adverse reactions to COVID-19 vaccines

February 23, 2024 steven

Population and clinical data

Inclusion criteria were: having received at least one dose of a COVID-19 vaccine that was being administered in Sweden during the study period (Comirnaty, Spikevax, Vaxzevria), being diagnosed with a condition or event that was attributed to the vaccine, being 18 years or older at time of recruitment, and having the capability to provide informed consent. Causality assessment for AEFIs were performed according to WHO criteria, as described previously¹¹. The vaccines are detailed as follows: Comirnaty is the proprietary name for the Pfizer-BioNTech mRNA vaccine, also known as BNT162b2. Spikevax is the proprietary name for the Moderna mRNA vaccine, also known as mRNA-1273. Vaxzevria is the proprietary name for the Oxford-AstraZeneca adenoviral-vector vaccine, also known as Covishield, ChAdOx1 nCoV-19, and AZD1222. Anonymous blood donor (BD) samples had been collected for research use before the COVID-19 pandemic at Uppsala University Hospital. APS1 patient samples were collected as part of an ongoing registry (Swedish Addison’s Disease Registry; ethics approval number: 2008/296-31/2).

Basic demographic data, elapsed time from vaccination until AEFI onset, and other clinical characteristics were collected from medical records and standardized questionnaires. Patients were classified according to AEFI diagnoses into the following groups: coagulation, neurological, allergic, cardiac, major adverse cardiac events (MACE), cytopenia, systemic disease, infection, vascular, and other. Exact diagnoses included in each group are detailed in Supplementary Table 1.

Sample collection

Venous blood samples of patients with AEFIs (n = 290) were collected into EDTA-containing tubes. All samples were centrifuged to obtain plasma (1500 × g, 10 min, 4 °C), aliquoted, and sent for storage at –70 °C. De-identified samples were received at or transferred to the Medical Biochemistry and Microbiology Department of Uppsala University for analyses.

Autoantibody screening via bead-based immunoassay

The screening of type I IFN autoantibodies was performed via an established bead-based anti-IgG assay that has been used previously with demonstration of reproducible results¹³. The large multiplex assay analysis plan was created to examine autoantibodies against 96 designated bead IDs (antigens, including technical controls) in a grand total of 2112 samples. Antibodies against IFN-α (IFNA1, 2, 4, 5, 6, 7, 8, 10, 14, 16, 17 and 21), IFN-β (IFNB), IFN-ɛ (IFNE), IFN-к (IFNK), and IFN-ω (IFNW) were measured in the study population for the specific purpose of the presented hypothesis. As part of the large multiplex assay, all samples underwent measurement of anti-human immunoglobulin G and antibodies against the primary proteins of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), including Spike (S protein), receptor binding domain (RBD), and nucleocapsid (N protein). Antibody levels against the Epstein-Barr virus nuclear antigen 1 (EBNA1) were also measured to assess the detection reliability and reproducibility of measurements.

Samples (1 ul) were diluted with a 2-step process: 1:25 in phosphate-buffered saline (PBS) and then a further 1:10 in a solution containing 0.05% Tween-20, 3% BSA and 5% non-fat milk in PBS. Magnetic beads (MagPlex®, Luminex Corp.) were coupled with commercially-available type I IFNs (and other proteins examined) at a concentration of 3 ug per 1.5 × 10^6 beads. For coupling, the AnteoTech Activation Kit for Multiplex Microspheres was used (Catalog code: A-LMPAKMM-10). The diluted samples (250 ul total volume) were incubated with 5 μl of the bead solution for two hours at room temperature with slight agitation achieved by a shaker set to 650 rpm. Following the incubation, beads were magnetized before washing with 0.05% Tween-20 in PBS (3X), and then resuspended in 50 microliters of 0.2% paraformaldehyde for 10 min. After another 3X wash process, a 30-min incubation with the secondary antibody (Invitrogen, H10104 lot#2384336) was performed. Measurement was carried out with a FlexMap 3D instrument (Luminex Corp) and results were recorded in AUs.

Autoantibody confirmation via ELISA

The initial multiplex screening (bead-based assay, Luminex) results were confirmed via optimized ELISA methods for several IFNs that were selected due to the presence of at least one sample with elevated response in the bead-based screening. That is, ELISA re-analysis was performed for a certain IFN if at least one ‘elevated response’ had been observed in either the AEFI or the BD group for said IFN (defined as >1500 AUs, based on Bastard et al.¹⁴). According to this criterion, we performed confirmatory testing for IFNA2 (number of samples with elevated response = 1), IFNA6 (n = 7), IFNA8 (n = 2), and IFNK (n = 1). For each antigen, the highest 8 samples (including those with elevated response) were included in the analyses (total n = 32). Starting sample dilution was 1:10 and was increased based on optimization goals described in the “Supplementary Methods” (1:20, 1:40, 1:80, 1:100, 1:160, 1:320, 1:1000, 1:2000, 1:5000, 1:10000, 1:20000, 1:25000, 1:50000, and 1:100000). In addition to the tested samples, we included three patients with APS1 as positive controls, one sample known to have high cross-reactivity (or non-specific binding), and three known-negative BDs during the course of each ELISA optimization (Supplementary Fig. 1).

Neutralization analysis

The neutralization properties of equivocal responses (n = 4) detected in the multiplex autoantibody assay were analyzed via cell culture experiments –modified from previously-reported methods²³. The experimental design involved (i) cell plating, (ii) co-transfection with Firefly (type I IFN-stimulable) and Renilla luciferase (constitutive expression) genes, (iii) stimulation with IFNA2 & addition of samples, and (iv) detection via a dual luciferase reporter assay. On day 1, HEK293T cells were seeded at 45000 cells/well in a 96-well plate (clear, flat-bottom, cell culture) with a final volume of 90 ul growth media per well (Gibco DMEM GlutaMAX + 10% fetal bovine serum + 100 units of penicillin-streptomycin). Transfection was performed with the Firefly pGL4.45[luc2P/ISRE/Hygro] and Renilla pRL-SV40 internal control luciferase vectors (Promega; #E414A and #E2231, respectively). The transfection solution was created in OptiMEM media with a 3:1 (ul:ug) ratio between the X-tremeGENE9 transfection reagent (Sigma-Aldrich; 6365787001) and total DNA (inter-vector ratio: 2:1 between Firefly and Renilla). The solution was incubated for 15 min and added to the wells (10 μl). On day 2, following overnight incubation, stimulation was performed with a final concentration of 10 ng/ml IFNA2 in wells (MedChemExpress; HY-P7022), except for non-stimulation controls. Immediately after stimulation, plasma samples were added into the wells to create a final plasma dilution of 1:10 in media, except for non-plasma controls. The plasma samples tested for neutralization included APS1 samples, AEFI samples with equivocal positivity (n = 4), and BD samples. On day 3, following 24 h of incubation, the Dual-Luciferase Reporter Assay System (Promega; #E1960) was used for analysis as described by the manufacturer (cell lysis, transfer of lysates to white opaque plates, and measurement with sequential addition of substrate and inhibitory/activating solutions). To perform quantification, we employed a plate reader that had luminescence quantification capabilities with magenta (Firefly) and green (Renilla) filters (Tecan, Magellan). The Firefly:Renilla ratio was used to assess neutralization. Technical controls confirmed experimental success, APS1 samples showed strong neutralization (ratios of <0.050), and BDs showed similar results to non-plasma controls–indicating non-neutralization (Supplementary Fig. 1).

Statistics

To obtain descriptive data and perform statistical analyses, we utilized the SPSS software (version 25.0; IBM, NY, USA). Continuous data were summarized in the form of mean ± standard deviation. Categorical data were summarized with absolute (n) and relative frequency (%). Normality of distribution in continuous variables was checked via evaluation of Q-Q plots or histograms. When required, the lack of normal distribution was confirmed via the Shapiro-Wilk or the Kolmogorov-Smirnov (Lilliefors correction) tests. The Kruskal-Wallis test was used to compare continuous variables among diagnosis subgroups (and the BD group), and post hoc adjustments were performed with the Bonferroni correction. In the comparison of groups formed according to the presence/absence of ‘elevated response’ (>1500 AUs), analyses for continuous data were performed with the Mann–Whitney U test and we used appropriate Chi-square tests for categorical data. For data visualization in the form of scatterplots and the heatmap, we respectively used the “ggplot2” and “pheatmap” packages installed on RStudio software (“Cherry Blossom” release, 2023.03.1-Build 446)²⁴. All code used to analyze data are available upon reasonable request from the corresponding authors. Dimensionality reduction was performed via principal component analysis (PCA) with use of the “prcomp” and “factoextra” packages in RStudio. The APS1 group was excluded from PCA. All type I IFN values were standardized with the calculation of Z-scores. Antibody levels for IgG, EBNA1 and SARS-CoV-2-related proteins were not included in the PCA in order to be able to detect the potential effects of IFNs –since smaller effects (and subgroups) could have been masked by variables with far greater impact (Supplementary Fig. 3).