Synergistic Role of NK Cells and Monocytes in Promoting Atherogenesis in Severe COVID-19 Patients

Patient characteristics

Patients were enrolled in the Ohio State University Intensive Care Unit Registry (BuckICU, IRB: 2020H0175) at the Ohio State University Wexner Medical Center and the James Cancer Hospital and Solove Research Institute between April 2020 and January 2021. BuckICU is a multi-disciplinary biorepository project that collects human biospecimens and associated clinical data from critically ill patients admitted to the ICU (severe) with respiratory failure, defined as any increase in respiratory support (supplemental oxygen or need for invasive or non-invasive mechanical ventilation) above baseline. For this study, we analyzed samples from thirty-one critically ill patients with severe COVID-19 disease (SARS-CoV-2 RNA+) (n=31), between 28 and 69 years of age. Patients with current malignant diseases, ongoing immunomodulatory treatment before hospitalization, and known diabetes mellitus were excluded. Seventeen healthy controls (Control) that were SARS-CoV-2 immunoglobulin G (IgG) seronegative and symptom-free at sampling were recruited as controls, with ages ranging from 25 to 65 years old. To minimize inter-experimental variability and batch effects between patients and controls, all samples (n=77) were acquired, processed, and analyzed fresh during the period between April 2020 and January 2021. Another group of individuals at an age range of 24 to 60 were recruited after three months of recovery from COVID-19 infection (n=29) (Recovered) at the peak of the COVID-19 pandemic in the United States (August 2020). Supplementary Table 1 and Materials and Methods both provide more specific patient characteristics. To study the innate cell response, we stained fresh whole blood with a 20-color innate cell–focused flow cytometry panel. All clinical laboratory analyses, including serology and polymerase chain reaction (PCR) for SARS-CoV-2, were performed using clinical routine assays in the clinical laboratories at The Ohio State University Wexner Medical Center, Columbus, Ohio. We obtained approval by the Research Ethics Committee in The Ohio State University under following IRB protocols (2020H0175, 2020H0228, 2014H0467, 2020H0339) and all patients gave written informed consent. For detailed clinical information, see Supplementary Table 1.

Flow cytometry staining and Data acquisition

Patients underwent a blood draw performed by a trained phlebotomist. Peripheral blood from healthy donors and patients was collected into sodium citrate tubes (BD, Franklin Lakes, NJ) and whole blood staining was performed. After collection, 120μL of whole blood was separated into polystyrene FACS (fluorescence-activated cell sorting) tubes and red blood cells were lysed using ACK buffer (Thermofisher, Waltham, Massachusetts) for 5 minutes followed by a wash with R10 media (RPMI-1640 supplemented with 10% FBS [fetal bovine serum], 2 mM L glutamine, 100 U/mL penicillin, and 100mg/mL streptomycin). After washing with phosphate-buffered saline (PBS), cells were pre-stained for viability exclusion using Zombie NIR Fixable Viability dye (BioLegend, San Diego, California) for 10 minutes, followed by a 20-minute incubation with a panel of directly conjugated monoclonal antibodies and TruStain diluted in equal parts of FACS buffer (PBS containing 0.1% sodium azide and 1% bovine serum albumin) and Brilliant Stain Buffer (BD Biosciences, San Jose, California). After staining, cells were fixed with 2% paraformaldehyde and were filtered by using filter caps for polystyrene FACS tubes before being acquired on a Cytek Aurora, a multispectral flow cytometry system with, 405-, 488-, and 640-nm lasers. Monoclonal antibodies used in the panel can be accessed in Supplementary Table 2.

Flow cytometry data analysis

FCS files were exported from spectralflo software and imported into FlowJo v.10.6.2 for subsequent analysis with the use of the following plug-ins to visualize the data: DownSample (3.2), and t-distributed stochastic neighbor embedding (tSNE). First, the compensation matrix for the 20-color flow cytometry panel was generated to correct for fluorescence spillover emissions and applied to files. Dataset as such was used for the downstream analysis in both manual gating (Supplementary Fig.1) and automated analysis. For the automated analysis, events were first down-sampled from the CD45⁺ CD3^- CD19^- gate across all samples using Down Sample. Clinical parameter categorical values for each sample were added to downsampled populations as metadata to enable identification of these groups, and these were then concatenated for analysis. tSNE was run using all parameters from the panel except Zombie NIR (Live/Dead), BV711(CD19), FITC(CD45) and APC-Cy7 (CD3).

After generating the tSNE figures, manually gated immune cell subsets were overlaid to the figures to represent subsets from each cohort. (e.g., 17 “healthy controls,” 31“severe,” and 29 “recovered” patients). Computational analysis of flow cytometry data for monocytes were visualized by uniform manifold approximation and projection (UMAP) embedding. UMAP dimensionality reduction was done on the dataset using all parameters from the panel except Zombie NIR (Live/Dead), BV711(CD19) FITC(CD45) and APC-Cy7 (CD3). The data were projected in 2-dimensions using UMAP embeddings and clusters of major monocyte subsets were identified based on expression of lineage defining markers. We used R (version 4.2.1) with packages ggpubr, ggplot, psych, gridExtra, corrplot^8,9 to generate figures.

Measurements of inflammatory biomarkers, chemokines, and cytokines

The plasma collected from blood samples was stored at −20°C and batched until processing without a prior thaw. Using enzyme-linked immunosorbent assay (ELISA), the following inflammatory markers were measured (Supplementary Table 3) including: Tissue inhibitor of metalloproteinases-1 (TIMP-1), Tissue inhibitor of metalloproteinases-2 (TIMP-2), Monocyte chemotactic protein 1 (MCP-1), Lipoprotein-Associated Phospholipase A2 (LpPLA2), oxidized low-density lipoprotein (Ox-LDL). To detect plasma cytokine and chemokine levels, plasma samples were processed using the multiplexed ELISA-based platform Quantibody® Human Inflammation Array 3 (RayBiotech QAH-INF-3) in accordance with the manufacturer’s protocol. Raw optical data were analyzed using the manufacturer’s analysis tool to construct standard curves and determine absolute cytokine concentration.

In-vitro Ox-LDL uptake experiment

An in-vitro experiment was conducted to investigate the role of activated NK cells in the uptake of Ox-LDL by macrophages. 5ml of venous blood samples were collected from severe COVID-19 patients (n=6) in sodium citrate tubes, and Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll gradient centrifugation. The isolation of NK cells was performed using the Miltenyi Biotec NK cell isolation kit (130-092-657). Subsequently, NK cells were activated in vitro through incubation with autologous plasma. Concurrently, PBMCs were allowed to rest and were incubated in Teflon-coated wells for up to six days. This incubation stimulated the development of monocyte-derived macrophages (MDMs) from peripheral blood monocytes¹⁰. After six days, these monocyte-derived macrophages were exposed to a medium containing Ox-LDL, which had been commercially manufactured and dye-tagged with Dil, to assess the absorption of Dil-Ox-LDL (10 ug/mL). This assessment was carried out after a four-hour incubation period in the presence or absence of plasma-activated NK cells. The cells were subsequently washed twice with PBS, and the intensity of DiI staining was measured using flow cytometry. To identify macrophages, markers such as CD45, CD3, CD19, and CD14 were utilized. For fluorescence microscopy analysis, the cells were seeded into 12-well plates and incubated at 37°C for an additional 24 hours. Images were acquired using an EVOS microscope, and the presence of Dil was detectable using an RFP filter (absorption peak: ∼554 nm, emission peak: ∼571 nm).

scRNA-seq data analysis

scRNA-seq data were obtained from a published dataset¹¹ The data file (in ‘rds’ format) downloaded from the cellxgene Data Portal (https://cellxgene.Cziscience.com/datasets) were read into R (version 4.2.1) session. Quality control was performed using Seurat (v3.1.5) to identify and filter out low-quality cells based on mitochondrial content and performing data preprocessing to prepare the data for downstream analysis. Cells with a percentage of mitochondrial genes greater than 10% were considered low quality and were filtered out. Prior to downstream analysis, the dataset was scaled and centered using “NormalizeData” and “ScaleData” functions. As the focus was on NK cells and monocytes and comparison between severe and control groups, specific subsets of scRNA-seq data set were separated and merged based on their metadata. The reanalysis included the standard Seurat workflow with “RunPCA“ and the number of Principle component analysis (PCA) dimensions used was 30, with a resolution parameter of 0.5. An elbow plot was generated to determine the optimal number of principal components for downstream analysis. UMAP plots were generated to visualize the cell clusters and identify potential subpopulations. After the cell clusters had been obtained, the Differentially expressed genes (DEG) in each cluster was computed with ‘FindAllMarkers’ (Supplementary Table 4)^12,13. Marker genes were determined based on their average log-fold change (log₂FC) and statistical significance (p-value and q-value).

Gene Ontology (GO) enrichment analysis was performed to gain insights into the biological processes associated with a set of genes of interest. The analysis was carried out using the clusterProfiler package in R (version 4.2.1)¹³. The significance threshold for enriched GO terms was set at a p-value cutoff of 0.05, and the Benjamini-Hochberg method was applied for multiple testing correction using GO terms with a q-value (adjusted p-value) below 0.05 were considered statistically significant. Cytoscape software (version 3.10.0) was used to visualize significant GO terms and integrate the GO terms to create a GO pathway network. For the DEGs, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment was performed for the top 200 DEGs. Enrichment analysis of the converted Entrez IDs was conducted to determine the significant KEGG pathways associated with the DEGs. The enrichKEGG function from the clusterProfiler package was utilized, specifying the gene IDs and the organism of interest (Homo sapiens, organism code ‘hsa’)¹⁴. This function identified overrepresented pathways based on a hypergeometric test. We employed the cnetplot function, to show the hub genes and pathways enriched.

Statistical Analysis

Graphs and heatmaps were prepared using GraphPad Prism, version 9. Data were statistically analyzed used their bundled software. One-way ANOVA¹⁵ was used to compare three or more groups followed by Dunn’s multiple comparisons test for pairwise comparisons. Comparisons between two groups were performed using the Wilcoxon rank-sum test. To analyze the relationship between immune signatures and plasma biomarkers, Pearson correlation coefficient was calculated and plotted on a correlogram using R 4.2.1. Scatter plots showing linear relationships between data was created for significant correlations. Data visualization was achieved using functions from the Seurat, ggplot2, and pheatmap packages.

Study design and cohort demographics

In this cross-sectional study, we assessed the immune profile in fresh whole blood samples from patients across three groups: 1) healthy donors: seventeen healthy age/sex-matched adults (n=17) who were SARS-CoV-2 immunoglobulin G (IgG) seronegative and asymptomatic at sampling as controls, 2) patients with severe COVID-19: thirty-one patients (n=31) with sampling on day one of intensive care unit admission for severe COVID-19 and respiratory failure, and 3) recovered patients: twenty-nine individuals with complete clinical recovery (n = 29) sampled three months after severe COVID-19 infection. Supplementary Table 1 and Materials and Methods both provide more specific patient characteristics. To minimize inter-experimental variability between patients and controls, all samples (n=77) were acquired, processed, and analyzed fresh between April 2020 and December 2020. Samples were analyzed using a 20-color spectral flow cytometer (Supplementary Table 1, gating strategy in Supplementary Fig.1).

Temporal dynamics of innate immune signatures and dysregulations in severe COVID-19

We applied conventional flow cytometry and tSNE dimensionality reduction plots gated on CD45⁺CD3^-CD19^- cells, which allowed for the identification and visualization of 10 innate immune cell populations (Fig.1A). The clustering algorithm allowed for an unbiased representation of these clusters during each stage of the disease. Among the dominant innate cell subsets that were focused on this study, we observed a significant increase in the frequencies of eosinophils (p<0.05) and neutrophils (p<0.05) among severe covid-19 patients (Fig.1B). While no differences were observed in the frequency of total monocytes between the control and severe groups, an overall increase in their frequencies was observed following recovery (p<0.05). Total NK cell frequencies were significantly decreased during severe infection and returned to basal levels after recovery, similar to previously published data¹⁶. We next analyzed the connectivity of cellular profiles and cytokine levels (detailed in Fig.2) in each study group. Heatmaps depicting correlation between 48 plasma cytokines/chemokines and 10 major innate immune cell subsets for the control, severe and recovered groups constructed based on Pearson correlation coefficient (Fig.1C). The correlation coefficients between those cell subsets were used to construct the density plot for each group. The density plot for Pearson correlation coefficients between three groups (control, severe, and recovered) comparing immune cell and cytokine data, provides a visual representation of the distribution of correlation coefficients for various pairs of variables between these groups (Fig.1D). Each line on the density plot represents one of the groups, and the plot can help reveal central tendency, spread and variability. A higher peak indicates a stronger central tendency toward certain correlation values. Visually comparing the three density lines, you can assess differences or similarities in the correlations between the groups. Comparing the three group’s density lines, we observe a greater overlap between severe and recovered group, compared to controls. Overall, the correlograms and density plots imply that the recovered donor group has a more similar distribution to severe patients than to the controls. PCA also revealed that three groups have differential cellular and cytokine profiles (Fig.1E). The volcano plot (Fig.1F) visualizes a sustained dysregulation of immune and biochemical parameters despite recovery from the disease. Overall, our data suggest that severe COVID-19 patients’ innate cell responses are dysregulated, with some immune perturbations persisting even after assumed viral clearance and recovery (up to 90 days post-infection).

Differential inflammatory and cardiovascular biomarker profiles following severe COVID-19

Next, we analyzed 48 cytokines, chemokines, and plasma biomarkers related to cardiovascular risk, including those associated with the ‘cytokine storm’ and post-acute sequelae of SARS-CoV-2 infection, in plasma samples (Fig.2A). Interestingly, seven of 33 increased cytokine responses in the severe patients were also increased in the recovered population. Nine cytokines were decreased in severe patients and the recovered donors compared to levels in the healthy donors (Fig.2A). For a broader comparative analysis of proinflammatory protein profiles that exhibited differential expression among our study cohorts, we executed PCA on all the proteins identified in the plasma (Fig.2B). The control, severe, and recovered groups displayed distinct clustering, signifying significant dysregulation of inflammatory proteins both during and after severe COVID-19 infection. Interestingly, our findings reveal that metalloproteinase inhibitors, specifically TIMP-1 and TIMP-2, which are recognized for their pivotal roles in atherosclerosis, plaque development, and platelet aggregation¹⁷, were elevated in severe patients (Fig.2C and Fig.2D) and persisted even three months after recovery from severe COVID-19. We also found increased levels of MCP-1 (Fig.2E), which is known to mediate migration and infiltration of monocytes to tissues, in severe-COVID-19 patient’s plasma. We also found an increase in Interferon-gamma inducible Protein 10kDa (IP-10) (Supplementary Fig.2A), and decreased levels of CCL5/RANTES (Regulated on activation, normal T cell expressed and secreted) (Supplementary Fig.2B) and eotaxin (Supplementary Fig.2D) in severe patients compared to levels in the other groups. Levels of cardiometabolic and inflammatory biomarkers, including soluble CD14 (sCD14), lipopolysaccharide-binding protein (LBP), fatty acid binding protein 4 (FABP4), D-Dimer, C reactive protein (CRP), and tissue factor (Supplementary Fig.2), were elevated during severe disease, but levels in healthy and recovered participants were similar (Supplementary Fig.2).

Notably, CRP and D-dimer levels, indicative of COVID-19 severity, may correlate with complications like acute respiratory distress syndrome (ARDS) and cardiovascular events. Unexpectedly, levels of Ox-LDL and LpPLA2 were decreased in severe and recovered patients, compared to levels in healthy controls (Fig.2F-G). This surprising finding highlights a persistent disruption in mechanisms involving modified lipoprotein homeostasis during and after severe COVID-19 infection.

Monocyte subset dynamics in COVID-19: Implications for immune dysregulation

Next, we examined monocyte subsets in our study populations. We observed elevated levels of CD14⁺⁺CD16⁻ “classical” monocytes in the recovered donors’ whole blood compared to the other two groups (Fig.3A and B). In contrast, intermediate monocytes (CD14⁺⁺CD16⁺) were increased in severe patients (Fig.3B), while the frequency of CD14⁻CD16⁺ “non-classical” monocytes was reduced in severe COVID-19 patients. Monocytes were visualized in a two-dimensional space using UMAP in Flowjo 10.9.0 (Fig.3C), which showed unique sub-clusters depending on the disease condition. Control and recovered groups had similar clustering patterns, distinct from the severe group, and some island clusters exhibited differences, suggesting varying features between these cohorts. No statistically significant changes were observed in CD27, CXCR3, and PD-1 receptor expression in classical and intermediate monocytes. However, the trend for decreased expression in PD-1 and increased CXCR3 expression in non-classical monocytes in the severe COVID-19 group suggests a more proinflammatory phenotype displayed by this subset (Fig.3D and Supplementary Fig.3C). We noted significant increases in CXCR5-expressing intermediate monocytes among our study populations compared to controls, with elevated frequencies persisting from severe disease into the recovery phase (Fig.3D and E). Similar monocyte/macrophage subsets have previously been reported in atherosclerotic lesions¹⁸. In contrast, severe COVID-19 patients exhibited low levels of intermediate and classical monocytes expressing tissue residency markers (e.g., CD69) in their peripheral circulation, and this reduction was also seen in the recovered group (p<0.05) (Fig.3D, and F). Our results indicate persistent dysregulation of circulating monocytes post-COVID-19 recovery, involving alterations in frequencies, subsets, and functional markers.

Dysregulated NK cell phenotypes associated with COVID-19 infection

We and others have observed that COVID-19 infection alters total NK cell frequencies in peripheral blood^16,19. Using a previously described gating strategy, CD56 and CD16 receptor expression were used to identify five distinct NK cell subsets²⁰ (Fig. 4A and B). NK cell subsets clustered differently among the three groups. Overall, patients with severe illness had considerably lower frequencies of all five NK cell subsets (Fig.4D). Furthermore, these subsets showed noticeable differences between groups in terms of expression of activation and tissue homing markers (Supplementary Fig.4). In all NK subsets, CD69, a marker of activation and tissue residency²¹ was increased in individuals with severe COVID-19. Levels of CD69 on NK cell subsets was similar between healthy and recovered donors (Fig.4F). We also noted an increasing trend in HLA-DR expression during severe infection, suggesting NK cell activation during severe disease (Fig. 4G). Additionally, NKG2A expression (indicating an inhibitory NK phenotype)²² was decreased among NK subsets in severe COVID-19 patients (Supplementary Fig.4B), confirming NK cell activation during severe infection. We also noted significantly lower expression of CD27 in three of the NK subsets except CD56^dimCD16^- and CD56^-CD16^dim subsets in severe COVID-19 patients and recovered donors compared with controls (Fig.4E). CD27⁺ NK cells are particularly important in mediating NK cell /T cell interaction via the CD27-CD70 pathway²³.

Implications of activated NK cells and monocytes on Oxidized LDL levels and potential atherogenesis in severe COVID-19

In order to determine whether the changes we observed in innate cells were related to cardiometabolic biomarker levels, we performed Pearson correlational analysis to assess associations between plasma biomarker data acquired through ELISA and flow cytometry immune cell frequencies. As shown in Fig.2F and G, while most biomarkers returned to basal levels after recovery, Ox-LDL and LpPLA2 (which co-traffics Ox-LDL into tissue) levels remained decreased significantly even after recovery. Ox-LDL is recognized as a key player in the development of atherosclerosis (atherogenesis)²⁴. We identified multiple correlations between immune cell data and plasma Ox-LDL (Fig.5A and B). We discovered a strong negative correlation between CD69⁺ NK cells and Ox-LDL levels (Fig.5C-D and Supplementary Fig.5A-D) as well as a negative correlation between Ox-LDL and activated (CD69⁺) total monocytes (Fig.5F), activated (CD69⁺) classical monocytes (Fig.5G) and intermediate monocytes (Fig.5H) among our severe COVID-19 patients. Monocytes migrate into the arterial wall and differentiate into macrophages, where these cells can internalize Ox-LDL and transform into foam cells²⁵. Therefore, our findings suggest a novel mechanism underlying the reduction of Ox-LDL levels in severe COVID-19 patients, potentially driven by activated NK cells contributing to increased consumption of Ox-LDL from the circulation by MDMs, leading to increased foam cell formation and subsequent atherogenesis in COVID-19 patients.

Transcriptional profiling and functional differentiation of monocytes and NK cells in severe COVID-19

We validated the phenotypic changes observed in flow cytometry and plasma biomarker assays using a publicly available single-cell RNA sequencing (scRNA-seq) dataset from Schulte-Schrepping et al.¹¹ For both NK cells and monocytes, we identified 2307 and 1815 DEGs (adjusted p-value < 0.05 & positive log₂FC) (Supplementary Fig.6A), selecting the top 200 DEGs for GO enrichment analysis (Supplementary Table 4 and 5). UMAP figures in Supplementary Fig.6B showcased immune cell annotations and clustering based on disease condition (control vs. severe COVID-19), with DEG analysis separately conducted for NK cells and monocytes. Heatmaps in Supplementary Fig.6C (monocytes) and Fig.6D (NK cells) provided an overview of transcriptional profile differences. Our analysis, incorporating the top 200 DEGs, revealed that transcriptional profiles phenotypically and functionally distinguished severe COVID-19 patients from uninfected counterparts. Volcano plots (Supplementary Fig.6D and E) displayed upregulated (red) and downregulated (green) genes. Our flowcytometry analysis revealed that severe patients and recovered donors both exhibited an increased presence of intermediate monocytes (as shown in Fig.2B). This difference was found to be statistically significant (p<0.05). Incidentally, our transcriptome analysis revealed an upregulation of CD36, a gene known to be more expressed in intermediate monocytes. CD36 encodes a receptor for Ox-LDL, and it plays a role in binding and internalizing Ox-LDL. Additionally, our analysis predicted multiple epigenetic modifications occurring as a result of severe infection. This prediction was corroborated by a high number of DEGs associated with histone modification and DNA methylation. Examples of these genes include H1-4, H1-3, H1-10, H1-2, and SETD2 (Supplementary Table 4). In monocytes, we observed significant upregulation in expression of CXCL2, IL1b and NLRP3 other notable genes suggestive of a proinflammatory phenotype in monocytes. (Fig.6A & Supplementary Table 4). Similarly, in NK cells, we detected upregulation of genes indicative of migration, inflammation, and activation, including ICAM3, TNFAIP3, CD69, and IL12RB1 (Fig.6B & Supplementary Table 4). Some genes like ICAM3, TNFSF12, TNFSF1214, NLRP3 and BRD2 were upregulated in both cell types (Supplementary Table 4).

Pathway analysis reveals intriguing interplay between viral response and lipid storage regulation in macrophages, highlighting persistent changes in innate cells

Furthermore, we conducted GO enrichment analysis to explore the functional enrichment of the DEGs, and the KEGG database was used to group these differentially expressed genes into gene pathways. We utilized the GO database and considered GO terms with a p-value < 0.05 significantly enriched (Fig6C-D & Supplementary Table 5). Our analysis revealed significant enrichment in GO terms associated with lasting cellular epigenetic changes. This pattern was consistent in both NK cells and monocytes during severe infection. Both cell types showed enriched pathways related to dysregulated unfolded protein response, particularly regulation of endoplasmic reticulum (ER) stress. In monocytes, we found significant enrichment in GO terms like “leukocyte chemotaxis”, “myeloid leukocyte migration”, and “phagocytosis engulfment”, suggesting potential activation and differentiation of migratory, phagocytic macrophages. In NK cells, GO pathway analysis highlighted the importance of “defense response to virus” and “viral-induced Pattern recognition receptor (PRR) pathway signaling” in combating SARS-CoV-2 infection.

Interestingly, we observed an overlap between DEGs and enriched GO terms in NK cells and monocytes. Genes like STAT, TNFAIP3, and NFKBIA were differentially expressed and enriched in multiple GO terms. We used Cytoscape v3.10.0 to create network visualizations of enriched GO pathways for monocytes (Fig.6E) and NK cells (Fig.6F). The monocyte network has 57 nodes, while the NK cell network has 39 nodes, each representing a specific GO pathway. The edges between nodes indicate potential relationships between pathways based on shared genes. Lighter node colors indicate greater statistical significance for pathway enrichment. Connections between pathways in the network imply shared genes, suggesting possible crosstalk or cooperation in biological processes. Node color signifies pathway enrichment significance, and lighter shades indicate stronger significance. Edges represent shared genes, hinting at potential functional links, like crosstalk or cooperation between pathways due to shared genes in biological processes. In monocytes, a significant network emerges with interconnected pathways such as macrophage activation, phagocytosis, ER stress, acute inflammatory response, and lipid storage regulation. This network has the potential to fuel atherogenesis through immune system activation. A prominent connection arises in NK cells between defense response to viruses, PRR signaling, and stem cell differentiation regulation. This suggests that NK cells, when encountering viruses, adopt an activated migratory phenotype, potentially becoming dominant in circulation. Furthermore, the KEGG pathway analysis (Supplementary.Fig.6G-H) adds another layer of evidence to the notion of enduring alterations in innate cells and their potential impact on the post-acute consequences of SARS-CoV-2 infection. The amalgamation of differential gene expression and GO enrichment analysis offers a holistic understanding of the molecular terrain associated with severe COVID-19 and the enduring modifications in innate cells. This insight can potentially reveal valuable biomarkers and therapeutic targets for further exploration.

Enhanced Ox-LDL uptake by monocyte-derived macrophages in the presence of autologous NK cells treated with plasma from severe COVID-19 patients

Based on our observations of innate immune responses and plasma biomarkers in vivo and analysis of published single-cell data, we have identified a significant correlation between multiple activated NK cell subsets and plasma Ox-LDL. Our findings strongly imply that these activated NK cells play a pivotal role in enhancing the uptake of circulating Ox-LDL. To further investigate the contribution of activated NK cells to atherogenesis and substantiate our hypothesis, we conducted an in vitro experiment to explore the impact of activated NK cells on Ox-LDL uptake by macrophages (Fig.7A).

We employed a method as outlined in one of our earlier publications. Frozen PBMCs from severe COVID-19 patients (n = 6) were used to isolate NK cells. These NK cells were subsequently incubated with autologous plasma. Concurrently, autologous MDMs were generated from the same patient’s PBMCs by allowing them to rest in Teflon-coated wells for up to six days. Following a six-day incubation period, we incubated these MDMs for four hours with commercially manufactured Dil-labeled Ox-LDL, both in the presence or absence of activated autologous NK cells. Interestingly, we observed significantly higher Dil-Ox-LDL uptake in the presence of activated NK cells (Wilcoxon signed rank-test; p<0.05 for all) (Fig.7B and C). The levels of Dil-Ox-LDL were quantified using flow cytometry, and this was further confirmed through EVOS microscope images (Fig.6D). Microscopic images were acquired to detect cells with internalized Dil-Ox-LDL using an RFP/TRITC filter set (absorption peak: ∼554 nm, emission peak: ∼571 nm). In summary, our findings collectively suggest that activated NK cells, in conjunction with other plasma factors, may be linked to increased Ox-LDL uptake and increased cardiovascular risk in individuals with severe COVID-19.