Case report: Immune profiling links neutrophil and plasmablast dysregulation to microvascular damage in post-COVID-19 Multisystem Inflammatory Syndrome in Adults (MIS-A)

Despite surviving a SARS-CoV-2 infection, some individuals experience an intense post-infectious Multisystem Inflammatory Syndrome (MIS) of uncertain etiology. Children with this syndrome (MIS-C) can experience a Kawasaki-like disease, but mechanisms in adults (MIS-A) are not clearly defined. Here we utilize a deep phenotyping approach to examine immunologic responses in an individual with MIS-A. Results are contextualized to healthy, convalescent, and acute COVID-19 patients. The findings reveal systemic inflammatory changes involving novel neutrophil and B-cell subsets, autoantibodies, complement, and hypercoagulability that are linked to systemic vascular dysfunction. This deep patient profiling generates new mechanistic insight into this rare clinical entity and provides potential insight into other post-infectious syndromes.

Serology and auto-antibody assays. Serology and auto-antibody panels including xMAP SARS-CoV2 IgG (Luminex Corp., Austin, TX: FDA, EUA)) testing antibodies to S1 spike, receptor binding domain (RBD) of S1 and Nucleoprotein, with a normal reference range of 0 -700 median fluorescence intensity (MFI)); COVID ELISA IgA, G Ratio (Euroimmun AG, Luebeck, Germany: FDA EUA)), anti-cytokine (Luminex Corp.) , normal reference range 0-500 MFI), were all used according to the manufacturers protocols with fresh frozen EDTA plasma as described above. Auto-antibody scores were generated by ranking autoreactivity of each cytokine target (autoantigen) amongst the total patient population then summing the ranks for each patient to generate a cumulative autoimmunity score which was expressed as 1.04^score to generate values that would fully utilize the color scale of graph in Figure 6C.
Anti-nuclear antibodies (ANA) were detected by an indirect immunofluorescence assay (IFA) on HEp-2 substrates (NOVA Lite HEp-2, Inova Diagnostics Inc. San Diego, CA, USA) at a dilution of 1:80 and read on an automated instrument (NOVA View, Inova Diagnostics) which interpolates fluorescence intensity to an end point titer (3). IFA staining patterns were classified according to the International Consensus on Autoantibody Patterns (ICAP: https://anapatterns.org/index.php) (4). If multiple patterns were noted in individual sera, they were recorded as separate ICAP patterns and accompanying titers.
CellaVision Hematopathology. Peripheral smears were digitally scanned using the CellaVision DC-1 digital morphology analyzer (CellaVision AB, Lund, Sweden), with settings for a 500 cell differential count as previously described (2). Occasional specimens with fewer than 500 detectable cells were not specifically excluded. Automated cell identification and pre-classification was then performed using the CellaVision Peripheral Blood Application Software, followed by manual verification. This latter step was undertaken independently by local hematopathologists, with validity assessed by Bland-Altman analysis of independent reviews by external hematopathologists using the CellaVision Remote Review 8 Software.
Immunohistochemistry. Cells of interest identified in the peripheral smears were marked-off with a tungsten-carbide pen. Peripheral smears were then destained using a series of ethanol washes. Automated immunohistochemistry was then performed using the Leica Bond-III automated IHC system. Monoclonal CD3 (clone LN10), CD20 (clone L26), CD79a (11D10) and CD138 (clone MI15) mouse anti-human primary antibodies were used, with polymer-based Fast Red detection.
Leukocyte and lymphocyte preparation. Leukocytes and lymphocytes were isolated as previously published (1). In brief, for lymphocyte preparation, whole blood was spun (15min, 3000rpm, Room Temperature [RT]) and plasma was removed. Isolation Cocktail and Rapid Spheres (Easy Sep TM Direct Human Total Lymphocytes Isolation Kit: 19655, StemCell Technologies) were added to remaining whole blood. Samples were mixed and incubated for 5min incubation at RT, the sample volumes topped up to 5mL with D-PBS+2%FBS + 1mM EDTA. The diluted sample was incubated in the magnet for 5min, at RT. This last step was repeated twice before cell resuspension in 5mL of PBS+0.04% BSA. After 2 washes 7500 lymphocytes were resuspended in 25µL of PBS+0.04% BSA.
For leukocytes preparation, whole blood was collected in heparin containing vacutubes and mixed with 0.5M EDTA with PBS+2% FBS and EasySep RBC Depletion spheres (EasySep TM RBC Depletion Reagent: 18170, Stem Cell Technologies). After 5 min of magnet incubation, at RT, tubes were inverted and poured into a new tube and RBC depletion was repeated. After 2 washes, cells were resuspended in 25µL of PBS+0.04% BSA.
Single-cell RNA-Seq library construction, alignment, and quality control. A total of 15,000 single cells (containing an equal proportion of leukocytes and lymphocytes) were loaded for partitioning using 10X Genomics NextGEM Gel Bead emulsions. All samples were processed as per manufacturer's protocol (with both PCR amplification steps run 12X). Quality control and preloading cDNA quantification was performed using TapeStation D1000 ScreenTape assay. Sequencing was performed using Illumina NovaSeq S2 and SP 100 cycle dual lane flow cells over multiple rounds to ensure each sample received approximately 32,000 reads per cell. Sequencing reads were aligned using CellRanger 3.1.0 pipeline (5) to the standard pre-built GRCh38 reference genome. Samples that passed alignment QC were aggregated into single datasets using CellRanger aggr with between-sample normalization to ensure each sample received an equal number of mapped reads per cell. Aggregated healthy (n = 3) and recovered COVID-19 (n = 3) samples recovered 30,514 cells that were sequenced to 17,157 post-normalization reads per cell.

Single-cell RNA-Seq computational analyses and workflows.
Filtered feature-barcode HDF5 matrices from aggregated datasets were imported into the R package Seurat v.3.9 for normalization, scaling, integration, multi-modal reference mapping, louvain clustering, dimensionality reduction, differential expression analysis, and visualization (6). Briefly, cells with abnormal transcriptional complexity (fewer than 500 UMIs, greater than 25,000 UMIs, or greater than 15% of mitochondrial reads) were considered artifacts and were removed from subsequent analysis. Cell identity was classified by mapping single cell profiles to the recently published PBMC single-cell joint RNA/CITE-Seq multi-omic AZIMUTH reference (7). Since no published reference automates granulocyte annotations, neutrophil clusters were manually annotated by querying known markers (i.e. CSF3R, S100A8, S100A9, MMP8, MMP9, ELANE, MPO) (8). A cell state-specific 'perturbation score' was calculated to reflect the magnitude of response elicited by factoring in number and cumulative FC of consensus DEGs. Perturbation scores were visualized using Nebulosa-generated density plots (9). Cells with high mitochondrial read were subsetted to enable further characterization of these cells, as there were high mitochondrial reads specifically in the first MIS-A sample.
3D human microvascular assays. Human umbilical vein endothelial cells (HUVEC, 10-donor pooled, Lonza) were labeled with EGFP via lentiviral transfection and used at P5-7 using a previously published protocol (Nature Methods). In brief, HUVEC-GFP were mixed with NHLF (Lonza) to a final ratio of 10x10^6/ml HUVEC to 2x10^6/mL NHLF with 2.5mg/mL final human plasmin-depleted fibrinogen (EMD-Millipore) activated by 4U/mL thrombin (Sigma) to encapsulate cells within a fibrin hydrogel. After 7 days of daily media changes, media was removed and cryopreserved human plasma diluted 1:20 with HBSS and spiked with 7.5µg/mL Fibrinogen-Alexa546 (Invitrogen) was perfused through microvessels for 10 minutes. Fibrinogen deposition was imaged in real-time using an Sp8 resonant scanning microscope at 1-2 minute intervals.