key: cord-0897030-oygwojab
authors: Giroux, Nicholas S.; Ding, Shengli; McClain, Micah T.; Burke, Thomas W.; Petzold, Elizabeth; Chung, Hong A.; Rivera, Grecia O.; Wang, Ergang; Xi, Rui; Bose, Shree; Rotstein, Tomer; Nicholson, Bradly P.; Chen, Tianyi; Henao, Ricardo; Sempowski, Gregory D.; Denny, Thomas N.; De Ussel, Maria Iglesias; Satterwhite, Lisa L.; Ko, Emily R.; Ginsburg, Geoffrey S.; Kraft, Bryan D.; Tsalik, Ephraim L.; Shen, Xiling; Woods, Christopher
title: Differential chromatin accessibility in peripheral blood mononuclear cells underlies COVID-19 disease severity prior to seroconversion
date: 2022-04-07
journal: Res Sq
DOI: 10.21203/rs.3.rs-1479864/v1
sha: 6a95445a16628e1499e3b81c589cb8677193b069
doc_id: 897030
cord_uid: oygwojab

SARS-CoV-2 infection triggers profound and variable immune responses in human hosts. Chromatin remodeling has been observed in individuals severely ill or convalescing with COVID-19, but chromatin remodeling early in disease prior to anti-spike protein IgG seroconversion has not been defined. We performed the Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) and RNA-seq on peripheral blood mononuclear cells (PBMCs) from outpatients with mild or moderate symptom severity at different stages of clinical illness. Early in the disease course prior to IgG seroconversion, modifications in chromatin accessibility associate with mild or moderate symptoms are already robust and include severity-associated changes in accessibility of genes in interleukin signaling, regulation of cell differentiation and cell morphology. Furthermore, single-cell analyses revealed evolution of the chromatin accessibility landscape and transcription factor motif accessibility for individual PBMC cell types over time. The most extensive remodeling occurred in CD14+ monocytes, where sub-populations with distinct chromatin accessibility profiles were observed prior to seroconversion. Mild symptom severity is marked by upregulation classical antiviral pathways including those regulating IRF1 and IRF7, whereas in moderate disease these classical antiviral signals diminish suggesting dysregulated and less effective responses. Together, these observations offer novel insight into the epigenome of early mild SARS-CoV-2 infection and suggest that detection of chromatin remodeling in early disease may offer promise for a new class of diagnostic tools for COVID-19.

Coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), manifests with highly variable symptom severity 1 . Infected subjects demonstrate clinical trajectories that range from remaining asymptomatic to developing life-threatening illness. A growing body of evidence suggests that the range of clinical manifestations is the result of different immune responses, and that analysis of chromatin accessibility and gene expression in circulating leukocytes can de ne underlying molecular mechanisms 2 . Studies have shown suppressed immune responses in subjects with only mild symptoms, as indicated by de cient expression of Type I and III interferons 3 .

Subjects with more severe disease demonstrated up-regulation of pro-in ammatory factors, including IL-6 and TNF-alpha 4 . The number of monocytes and associated IL-6, CCL2, and CCL8 production in the peripheral blood also are elevated in subjects with severe COVID-19 5, 6 . However, regions of regulatory chromatin that contain transcription factor motifs become accessible prior to downstream gene expression and may offer the potential for even earlier detection of these evolving biological responses 7 .

Analyzing time-dependent changes at the epigenetic level is complicated by the variable time course of COVID-19 across patients, which even in the outpatient setting can range from a few days of asymptomatic viral shedding to prolonged febrile illness. Most studies have used clock time (days since illness onset), a measure that is broadly related to immune response stages in many subjects. Another option is to utilize direct measures of immune maturation to broadly characterize where during the host response to infection a given subject exists at a point in time. The development of speci c antibodies (IgG) against SARS-CoV-2 marks an in ection point in a COVID-19 patient's disease progression, indicating a transition from innate immunity to acquired immunity 8, 9 . IgG seroconversion typically occurs within two weeks of symptom onset and roughly coincides with the time that patients without critical illness will see clinical improvement 10, 11 . However, limited data are available examining the peripheral blood immune responses early in disease progression and prior to seroconversion. To test the hypothesis that the landscape of chromatin accessibility contains biomarkers that de ne early molecular mechanisms that underpin divergent immunologic responses in SARS-CoV-2 infection, we performed bulk and single-cell Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) and RNA-seq on longitudinal PBMC samples from patients before and after seroconversion.

A clinical cohort to study early COVID-19 with mild or moderate symptoms

To establish a cohort for the longitudinal study of early SARS-CoV-2 infection, subjects were selected from a large ongoing community-based prospective cohort, the Molecular Epidemiological Study of Suspected Infection (MESSI) in four experimental groups: 1) pre-pandemic healthy controls, 2) subjects exposed to SARS-CoV-2 as close contacts but negative for SARS-CoV-2 by qPCR and serology, 3) outpatients positive for SARS-CoV-2 by qPCR whose maximal illness was mild and 4) outpatients positive for SARS-CoV-2 by qPCR whose maximal illness was moderate. Demographic characteristics of study participants are shown in Table 1 . Symptoms associated with COVID-19 illness were self-reported to clinic personnel using a COVID-19 symptom survey developed by Duke University Medical Center consisting of 38 symptoms. Each subject reported a symptom and perceived severity of that symptom from 0 (none) to 4 (very severe). Subjects with mild symptoms exhibited a mean score of 12.9 ± 2.1 and 12.2 ± 2.8 for total PBMC assays (bulk) and assays of individual PBMCs (single-cell assays), respectively, which corresponded to the World Health Organization (WHO) Ordinal 8-point scale 1 (OS1, ambulatory and with no limitation of activities); subjects with moderate symptoms exhibited a mean score of 33.6 ± 2.4 and 36.0 ± 2.3 respectively which corresponded to the WHO Ordinal 8-point scale 2, (OS2, ambulatory and with limitation of activities). Interestingly, despite lacking any evidence of infection (negative SARS-CoV-2 qPCR and serological testing for at least two months after exposure) close household contacts who were exposed to SARS-CoV-2 but remained negative for infection showed broadly similar quantitative symptom scores to subjects with mild disease, with a mean score of 9.9 ± 2.9 and 3.3 ± 2.8 respectively ( Table 1 ). The sum of symptom severity across all symptoms reported was able to distinguish moderate subjects from mild or close contacts but did not distinguish mild from close contacts ( Fig. S1A-B) . However, the speci c symptoms experienced, and their severity, were different between all three groups ( Fig. S1C-D) . Loss of taste and smell, headache, malaise, and fatigue were most common in the moderate group. Coughing was most common in the mild group and runny nose in the close contacts.

Transcriptional pro les of PBMCs distinguish mild or moderate COVID-19 from healthy controls To identify underlying molecular mechanisms of disease severity in mild or moderate COVID-19 both prior and after seroconversion, PBMCs were isolated and pro led as total PBMCs (bulk assays) or as single nuclei isolated from individual PBMCs (single cell assays). Longitudinal sample collection from each subject spanned the transition from seronegative (IgG-) to seropositive (IgG+) for SARS-CoV-2 spike protein in the mild and moderate symptom groups (Supplementary Table S1 and S2). Differential gene expression analysis of bulk RNA-seq data from mild IgG-COVID-19 subjects compared to healthy controls identi ed a transcriptional signature of 45 genes with p < = 0.05; moderate IgG-COVID-19 subjects compared to healthy controls identi ed 71 genes with p < = 0.05 (Fig. 1A) . Similar comparisons using RNA-seq from mild IgG + subjects compared to healthy controls or moderate IgG + subjects compared to healthy controls identi ed 250 and 43 genes, respectively, with p < = 0.05 (Fig. 1C) . The gene symbols, gene names, fold change and p values adjusted for multiple hypothesis testing for the four comparisons are found in Supplemental Table S3 . Using principal component analysis (PCA), differential gene expression between healthy controls and subjects with either mild symptoms (Fig. 1B) or moderate symptoms (Fig. 1D) were su cient to distinguish the groups. However, both mild and moderate subjects clustered together and could not be easily distinguished. Considering the level of gene expression (log fold change, LFC) in mild and moderate subjects compared to healthy controls, differentially expressed genes were identi ed in only mild, only moderate or both symptom cohorts (Fig. 1E ). Comparing IgGsubjects, we identi ed 30 genes observed in both mild and moderate cohorts. In contrast, we did not identify any genes solely associated with moderate IgG + subjects and identi ed a further 43 genes observed in both IgG + mild and moderate cohorts. Healthy controls exhibited higher expression of the interleukin-8 precursor CXCL8 compared to other groups. Conversely, the interleukin-8 receptor, CXCR1, had higher expression in mild and moderate IgG-subjects. Lower expression of chemokine CCL3 and cytokine IL1B 12 , and higher expression of the myeloid cell plasticity regulator KLF6 13 was observed in both IgG-and IgG + COVID-19 subjects.

Chromatin remodeling prior to seroconversion can distinguish mild from moderate COVID-19

To test whether chromatin accessibility can distinguish COVID-19 disease severity in individuals early in their disease course (prior to seroconversion) an Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) 14 was used to quantify differential chromatin accessibility in three groups: subjects with mild symptoms, subjects with moderate symptoms and pre-pandemic healthy controls. A set of 455 differentially accessible chromatin regions (DARs) was found to distinguish IgG-subjects with mild symptoms from those with moderate symptoms at p < = 0.05 ( Fig. 2A) . Of these markers, 73 regions shared differential accessibility in mild and moderate IgG-subjects, demonstrating both severityassociated evidence of chromatin remodeling and a smaller conserved epigenetic response. Additionally, a set of 375 regions of differentially accessible chromatin was identi ed in IgG + subjects (Fig. 2B) . DARs speci c to the comparisons in 2A-B as heat map depictions ( Fig. 2C-D) for prior and post seroconversion shows overlap between mild and moderate symptom subjects compared to healthy controls. Functional enrichment was then performed for each set of differentially accessible regions to identify an association with transcription factor motifs and downstream pathways. In IgG-subjects with mild symptoms, we found an enrichment of transcription factors and pathways related to viral infection (Fig. 2E ). Fewer enriched pathways were observed in the IgG-subjects with moderate symptoms. Transcription factors found to be enriched in accessible chromatin from IgG + subjects with mild symptoms were associated with dysregulation of myeloid development via HOXA3, C/EBP-B and C/EBP-D 15 (Fig. 2F ). CD14 + monocytes are the primary cell type differentially activated in IgG-subjects

The chromatin accessibility signature detected in the bulk ATAC-seq datasets identi ed candidate epigenetic biomarkers associated with COVID-19 severity and showed that PBMCs undergo extensive chromatin remodeling in response to SARS-CoV-2 infection. To understand how each cell type contributed to this signature and track the evolution of the chromatin landscape of each cell type during seroconversion, we performed single-cell (sc)ATAC-seq pro ling of the PBMCs isolated from subjects with mild or moderate symptoms and healthy controls (Table 1 and Supplemental Table 2 ). The relative abundance of cell types collected from each subject cohort were similar between IgG-and IgG + timepoints (Fig. 3A) . Analysis of transcription factor motif accessibility was applied to PBMCs from IgGmild or moderate subjects to identify shared regulatory mechanisms. Transcription factors with highly accessible motifs and low gene expression were identi ed, consistent with epigenetically primed chromatin remodeling without corresponding transcriptional activity (Fig. 3B ). These motifs, including those of the CREB and ETS protein families, are characterized by increased binding activity as measured by footprint depth and relatively low average per-cell gene expression. In ammatory regulators including JUN/FOS, proteins from the KLF family, and IRF1 showed elevated gene expression. A complete list of transcription factors associated with accessible regions and low gene expression are found in Supplemental Table S4 .

Comparison of PBMCs collected from subjects with mild or moderate symptoms identi ed speci c peaks enriched on the day of study enrollment (IgG-) versus a timepoint 2-4 weeks later (IgG+) for all cell types (Fig. 3C) . A total of 5988 peaks were identi ed as uniquely accessible at one of the three longitudinal (IgG-, mixed serology, and IgG+) timepoints. A similar experimental approach was applied to each major cell type individually to determine the extent of chromatin remodeling prior to seroconversion. The cells with enrichment of signi cant peaks at the early, IgG-timepoint were primarily CD14 + monocytes (Fig. 3D ).

Domains of regulatory chromatin (DORCs) were identi ed in all cell types collected from IgG-subjects with mild or moderate symptoms by counting the number of peak-to-gene linkages for each gene in the scATAC-seq datasets 7 . A total of 1109 genes with greater than 10 peak-to-gene linkages were de ned as DORC genes, and those known to be regulated by a super-enhancer were labeled (Fig. 4A) . These DORCs showed cell type-speci c pro les and appeared to have higher accessibility than gene expression in these cells, suggesting that these genes are being primed for activation (Fig. 4B) . Accessibility at the DORCs regulated by a super-enhancer was enriched in CD14 + monocytes and other myeloid cells, including dendritic cells and CD16 + monocytes, indicated by the presence of proximal gene regulatory elements with correlated accessibility to DORC gene loci (Fig., 4C) . These DORC genes regulated by superenhancers play a role in priming active chromatin states in CD14 + monocytes, consistent with the cell fate decisions identi ed that distinguished subjects with variable symptom severity.

Epigenetic pro les of CD14 + monocyte subpopulations differentiate mild from moderate IgG-COVID-19

We next identi ed transcriptomic and epigenetic pro les of sub-populations of CD14 + monocytes that distinguished IgG-subjects with mild from moderate symptoms. Single-cell gene expression data for CD14 + monocytes were processed using the uniform manifold and projection (UMAP) algorithm. Clusters 2 and 4 identi ed a unique sub-population of CD14 + monocytes in subjects with mild symptoms and cluster 7 identi ed a sub-population of CD14 + monocytes in subjects with moderate symptoms (Fig. 5A) . Transcription factors and downstream targets were identi ed for each cluster. CD14 + monocyte subpopulations from subjects with mild symptoms exhibited signi cant up-regulation of the regions that regulate IRF7, IRF1, and STAT1 transcription factors, whereas the CD14 + monocyte subpopulations from subjects with moderate symptoms were characterized by up-regulation of the CEBPB and KLF3 regulatory regions (Fig. 5B ).

Single-cell ATAC-seq data for CD14 + monocytes were processed using a similar approach to identify clusters with correlated gene expression regulatory activity (Fig. 5C ). Activity at the transcription factor motifs was estimated, and the correlations between scRNA-seq and scATAC-seq clusters were calculated in a pairwise manner (Fig. 5D ). Differential gene expression between the mild and moderate subpopulations identi ed interferon-stimulated gene activation as a distinguishing feature of symptom severity. The cell clusters from scATAC-seq with high correlations with clusters 2, 4, and 7 from scRNAseq were highlighted. We observed that scATAC-seq clusters 4 and 5 were correlated with scRNA-seq clusters 2 and 4, which identi ed similar populations of CD14 + monocytes from mild subjects (Fig. 5E ). We also observed that scATAC-seq cluster 5 was highly correlated with scRNA-seq cluster 7, identifying similar populations of CD14 + monocytes from moderate subjects. In addition to these transcriptomic markers, functional enrichment analysis was then performed for the transcription factors with differential activity and their downstream target genes (Fig. 5F ). Pathways identi ed centered on those of the adaptive immune response. Differential regulation of pathways related to the transition from innate to adaptive immunity, marked by dectin-1 and the toll-like receptor cascade, suggested that activation of T and B cell receptor signaling were observed in these sub-populations 16-18 .

This study demonstrates that the epigenomes of critical subsets of PBMCs are remodeled extensively early in SARS-CoV-2 infection and re ect disease severity in IgG-subjects with mild and moderate symptoms. Speci cally, differential activity of transcription factors and chromatin accessibility prior to anti-spike IgG seroconversion were observed to distinguish disease severity, preceding the later transcriptional response. The fact that these signals are robust even in a small cohort of subjects with subtle differences in disease severity (WHO OS 1 vs 2) suggests the potential power of epigenetic approaches to classify subsets of SARS-CoV-2 infection. This study complements previous work that identi ed signatures of accessible chromatin in severe or convalescing COVID-19 [19] [20] [21] [22] The data show that transcription factor motifs enriched in peaks accessible prior to seroconversion had the highest occupancy in cells from the myeloid lineage, speci cally CD14+ monocytes, dendritic cells, and CD16+ monocytes. KLF and CREB transcription factor families, which are known to regulate monocyte-macrophage polarization, have elevated motif accessibility in subjects with mild or moderate symptoms, suggesting an epigenetic priming mechanism controlling cell fate 23, 24 . This distinction is important as not only are CD14+ monocytes generally more activated in early COVID-19, as has been described, but here they are shown to differentially activated across subjects with variable disease severity. These differences include modulation of IFN-γ, as reported previously. Speci cally, mild symptom severity (or effective control) is marked by upregulation of classical antiviral pathways including those regulating IRF1 and IRF7. With increased severity, these antiviral signals diminish, suggesting dysregulated and less effective responses underlie moderate disease. These early epigenetic changes occur prior to transcriptional manifestations and in this cohort were also more stable and predictive of disease severity than gene expression alone. These ndings suggest that epigenetic approaches focusing on chromatin accessibility may offer even greater promise than other host-based molecular diagnostics and predictive tools, as has been suggested for DNA methylation-based pro les 25 . Our work is consistent with previous studies that identi ed DNA-methylation signatures, which included viral response and interferon signaling, that will predict SARS-CoV-2 infection and clinical outcome 25 and with genome-wide DNA methylation signatures associated with severe COVID-19 that highlight hypermethylation in IFN-related genes, hypomethylation in in ammatory genes and increased epigenetic age 26 .

Interestingly, CD14+ monocytes underwent the most extensive chromatin remodeling over time and exhibited epigenetic pro les at early seronegative times that distinguish disease severity. In contrast, after seroconversion motifs enriched in accessible peaks were not seen in monocytes, instead showing the highest activity in B cells and plasmablasts, consistent with the transition from in ammatory signaling to adaptive immune development in this phase. Our work is consistent with epigenetic pro les of immune cells of individuals convalescing from COVID-19 that show establishment of immunological memory 22 .

There are limitations to this study, including the relatively small sample size, nonstandard enrollment and sampling times that vary across individuals, and a lack of subjects with critical illness to examine how these epigenetic changes manifest in more severe disease. However, development of a host-response assay that leverages the highly sensitive epigenetic biomarkers established early during infection has the potential to ll a clear unmet clinical need in the care of patients with COVID-19 27, 28 . In summary, the evolution of the chromatin landscape in circulating leukocytes during COVID-19 primes host immunological responses at early times, is mediated primarily by CD14+ monocytes and correlates with an observed divergence in disease severity. These changes temporally precede transcriptional manifestations of pathways related to adaptive immunity. Together, these observations offer novel insight into severity-associated variations in host responses to SARS-CoV-2 infection and suggest that detection of critical components of chromatin remodeling in early disease may offer promise for a new class of diagnostic tools for COVID-19.

Cohort recruitment, biological sample collection, and initial phenotyping Protection of human subjects was in accordance with research protocols approved by the Duke University Institutional Review Board, consistent with the Declaration of Helsinki. Written informed consent was obtained from all research subjects or their legally authorized representatives. Subjects with con rmed or suspected SARS-CoV-2 infection or their close contacts were identi ed in the outpatient setting and enrolled into the Molecular and Epidemiological Study of Suspected Infection protocol (MESSI, IRB Pro00100241). All close contacts and subjects with mild or moderate COVID-19 were longitudinally sampled from enrollment to convalescent phase. Biological samples and demographic information were collected prospectively at rst visit (Day 0) and at weekly intervals on Day 7 and Day 14. At each visit, infection with SARS-CoV-2 was con rmed using quantitative polymerase chain reaction (qPCR) of nasopharyngeal (NP) swab samples, and serology testing was performed for IgG against the SARS-CoV-2 spike domain. All subjects with mild or moderate COVID-19 progressed from seronegative (IgG-) to seropositive (IgG+). Close contacts were qPCR negative and IgG-at all time points; healthy controls were enrolled pre-pandemic and were not tested for SARS-CoV-2 or spike protein IgG. Self-reported symptom surveys were performed at each visit. To quantify symptom severity, the sum of 38 de ned symptoms, each scored 0-4 (0-none, 1-mild, 2-moderate, 3-severe, 4-very severe), was determined from symptom onset through each longitudinal sample collection. SARS-CoV-2 q-PCR tests used virus RNA extracted from NP samples in 140 uL of viral transport medium (VTM) according to manufacturer's instructions (QiaAmp Viral RNA minikit). SARS-CoV-2 nucleocapsid (N1) and human RNase P (RPP30) RNA copies were determined using 5 µL of isolated RNA in the CDC-designed kit (CDC-006-00019, Revision: 03, Integrated DNA Technologies 2019-nCoV kit). SARS-CoV-2 IgG ELISA tests for antibody response were performed using the anti-SARS-CoV-2 spike S1 domain IgG ELISA assay (EUROIMMUN Medizinische Labordiagnostika AG, Lübeck, Germany).

PBMCs were puri ed using the Ficoll-Hypaque density gradient method according to manufacturer's instructions. Brie y, whole blood was collected in ACD Vacutainer tubes (BD) and processed within 8 hours by dilution 1:2 in PBS, layered onto the Ficoll-Hypaque (Sigma Aldrich) in 50 ml conical tubes, and centrifuged at 420 x g for 25 minutes. Buffy coat was collected, washed twice in D-PBS by centrifugation at 400 x g for 10 minutes to isolate peripheral blood mononuclear cells (PBMCs) which were assessed for viability and cell count using a Vi-Cell automated cell counter (Beckman-Coulter). PBMCs were adjusted to 10 x10 6 cells/ml in cryopreservation media (90% FBS, 10% DMSO), frozen at -80 o C using CoolCell LX (BioCision) for 12-24 hours and stored in liquid nitrogen vapor phase.

RNA extraction, total RNA-seq, and analysis RNA was extracted from 300K cells using the Zymo Direct-zol miniprep kit (Cat# R2051) and RNA quality assessed using the Agilent DNA tape screen assay. The RNA Integrity Number (RIN) scores for all samples were > 7.0. Total RNA libraries were generated using the NuGEN Ovation® SoLo RNA-Seq Library Preparation Kit (Cat# 0500-96). Libraries were sequenced using Illumina NovaSeq 6000 instrument with S4 ow cell and 150 base pair paired-end reads.FASTQ les were generated from the NovaSeq BCL outputs and quality was assessed with FASTQC 29 . Differentially expressed genes were identi ed between subjects with different disease severity using the limma package and voom to model variance 30 .

Nuclei puri cation, ATAC-seq and analysis Nuclei were extracted from frozen PBMCs. Brie y, 100K cells were spun down at 300 x g for 5 minutes at 4 o C. The supernatant was removed, and cells were mixed with 100 uL of lysis buffer (10mM NaCl, 3mM MgCl2, 10mM Tris-HCl pH7.4, 0.1% Tween-20, 0.1% NonidetTM P40) and lysed on ice for 4 minutes. Wash buffer (1 mL; 10mM NaCl, 3mM MgCl2, 10mM Tris-HCl pH7.4, 0.1% Tween20) was added before centrifuging at 500 x g for 5 minutes at 4 o C. ATAC-seq libraries were generated as described 14 . Brie y, transposition mix (25 μL 2× TD buffer, 2.5 μL transposase (Tn5, 100 nM nal), 22.5 μL water) (Illumina Cat# 20031198) was added to the nuclear pellets, incubated at 37°C for 30 minutes, and samples puri ed using the Qiagen MinElute PCR Puri cation Kit (Qiagen Cat#28004). DNA fragments were PCR ampli ed for a total of 10-11 cycles and resulting libraries puri ed using the Qiagen MinElute PCR Puri cation Kit. The libraries were sequenced with an Illumina Novaseq 6000 S4 ow cell using 100 bp paired-end reads. FASTQ les were generated from the NovaSeq BCL outputs and used as input to the ENCODE ATAC-seq pipeline (https://github.com/ENCODE-DCC/atac-seq-pipeline) using the MACS2 peak-caller with all default parameters. Differential accessibility was calculated between groups of subjects with different disease severity using the csaw package 31 .

Single-cell (sc)RNA-seq and analysis Frozen PBMCs were thawed, and count and cell viability were measured by Countess II. The cell viability exceeded 80% for all samples except PBMC samples from CC subjects, which had viability between 70-80%. For single cell (sc) RNA-seq, 200K cells were aliquoted, spun down, resuspended in 30 uL PBS+0.04%BSA+0.2U / uL RNase inhibitor, and counted using Countess II. GEM generation, post GEMRT cleanup, cDNA ampli cation, and library construction were performed following 10X Genomics Single Cell 5' v1 chemistry and quality was assessed using Agilent DNA tape screen assay. Libraries were then pooled and sequenced using Illumina NovaSeq 6000 platform with the goal of reaching saturation or 20,000 unique reads per cell on average. Sequencing data were used as input to the 10x Genomics Cell Ranger pipeline to demultiplex BCL les, generate FASTQs, and generate feature counts for each library. Dimensionality reduction and cell type annotation was accomplished using gene-barcode matrices generated using CellRanger count were analyzed using Seurat 3 with the default parameters unless otherwise speci ed 32 . Regulatory network inference was accomplished by converting the scRNA-seq Seurat object into a SingleCellExperiment and used as input to analysis with the SCENIC package 33 .

PBMCs were thawed and nuclei were extracted as for ATAC-seq. The single-cell suspensions of scATACseq samples were converted to barcoded scATAC-seq libraries using the Chromium Single Cell 5′ Library, Gel Bead and Multiplex Kit, and Chip Kit (10x Genomics). The Chromium Single Cell 5′ v2 Reagent (10x Genomics, 120237) kit was used to prepare single-cell ATAC libraries according to the manufacturer's instructions. Quality was assessed using Agilent DNA tape screen assay. Libraries were then pooled and sequenced using Illumina NovaSeq platform with the goal of reaching saturation or 25,000 unique reads per nuclei on average. Sequencing data were used as input to the 10x Genomics Cell Ranger ATAC pipeline to demultiplex BCL les, generate FASTQs, and generate feature counts for each library.scRNAseq and scATAC-seq were integrated using fragment le outputs generated using CellRanger ATAC count were analyzed using ArchR following the standard work ow and with default parameters unless otherwise speci ed 34 . Feature and motif enrichment analysis (peak calling) was performed using MACS2 via the addReproduciblePeakSet() method in ArchR which uses pseudo-bulk replicates of cells grouped on a speci c design variable. The correlations between chromVAR transcription factor deviation scores and scRNA-seq derived gene expression data were calculated using the ArchR method correlateMatrices() to identify activators and repressors. Topic-based clustering was performed for the CD14+ monocytes from the mild or moderate subject cohorts using the R package cisTopic 35 .

Detailed methods for qPCR and antibody tests, and epigenetic and genomic pro ling, including RNA extraction, sequencing, differential expression analysis and differential chromatin accessibility analysis for both bulk and single cell analytic approaches can be found in Supplemental Methods. 

Imbalanced Host Response to SARS-CoV-2 Drives Development of COVID-19

Altered transcriptional and chromatin responses to rhinovirus in bronchial epithelial cells from adults with asthma

Type I and Type III Interferons -Induction, Signaling, Evasion, and Application to Combat COVID-19

An in ammatory cytokine signature predicts COVID-19 severity and survival

Severe COVID-19 Is Marked by a Dysregulated Myeloid Cell Compartment

Heightened Innate Immune Responses in the Respiratory Tract of COVID-19 Patients

Chromatin Potential Identi ed by Shared Single-Cell Pro ling of RNA and Chromatin

The role of IgG Fc receptors in antibody-dependent enhancement

Rapid Decay of Anti-SARS-CoV-2 Antibodies in Persons with Mild Covid-19

Antibody responses to SARS-CoV-2 in patients with COVID-19

Antibody Responses to SARS-CoV-2 in Patients With Novel Coronavirus Disease

Chemokines (CCL3, CCL4, and CCL5) Inhibit ATP-Induced Release of IL-1beta by Monocytic Cells

KLF6 contributes to myeloid cell plasticity in the pathogenesis of intestinal in ammation

ATAC-seq: A Method for Assaying Chromatin Accessibility Genome-Wide

Dysregulation of macrophage development and phenotype in diabetic human macrophages can be rescued by Hoxa3 protein transduction

TLRs and innate immunity

Dectin-1: a signalling non-TLR pattern-recognition receptor

Longitudinal Analysis of T and B Cell Receptor Repertoire Transcripts Reveal Dynamic Immune Response in COVID-19 Patients

A blood atlas of COVID-19 de nes hallmarks of disease severity and speci city

Epigenetic Landscapes of Single-Cell Chromatin Accessibility and Transcriptomic Immune Pro les of T Cells in COVID-19 Patients

Multi-omic pro ling reveals widespread dysregulation of innate immunity and hematopoiesis in COVID-19

Single-cell epigenomic landscape of peripheral immune cells reveals establishment of trained immunity in individuals convalescing from COVID-19

Transcription factors STAT6 and KLF4 implement macrophage polarization via the dual catalytic powers of MCPIP

Kruppel-Like Factors in Vascular In ammation: Mechanistic Insights and Therapeutic Potential

Host methylation predicts SARS-CoV-2 infection and clinical outcome

Genome-wide DNA methylation pro ling of peripheral blood reveals an epigenetic signature associated with severe COVID-19

A novel severity score to predict inpatient mortality in COVID-19 patients

Early prediction of the risk of severe coronavirus disease 2019: A key step in therapeutic decision making

A Quality Control Tool for High Throughput Sequence Data

Precision weights unlock linear model analysis tools for RNA-seq read counts

csaw: a Bioconductor package for differential binding analysis of ChIP-seq data using sliding windows

Comprehensive Integration of Single-Cell Data

We thank Eric van Gieson and Thomas Thomou for their advice throughout the project, and Andrew Macintyer, Thomas Oguin, and their teams at the Duke Regional Biocontainment Laboratory for COVID IgG serology and PCR viral load testing. We thank Raul Louzao, Christina Nix, Carolyne Whiting, Tyffany Locklear, Pam Isner, Anna Mazur, Jack Anderson, Maria Miggs, Aleah Bowie, Julie Steinbrink, Robert Rolfe, Jorge Prado for technical expertise. We wish to express our deep appreciation to the participants and their families for their generosity in collaborating with us in this work. 

This is a list of supplementary les associated with this preprint. Click to download.