Participants and ethical approval

The LPS-induced systemic inflammation (experimental human endotoxemia) study was approved by the Medical Ethical Committee Oost-Nederland (ref. no. NL61136.091.17). Eleven healthy male volunteers were recruited. All participants gave written informed consent and medical history, physical examination and laboratory tests and a 12-lead electrocardiogram did not reveal any abnormalities. Smoking, medication use, previous participation in experimental human endotoxemia studies or signs of acute illness within 3 weeks before the start of the study were exclusion criteria. For in vitro experiments, blood was drawn from healthy volunteers after obtaining written informed consent. Blood withdrawal for this purpose was approved by the Medical Ethical Committee Oost-Nederland (ref. no. NL84281.091.23). All study procedures were performed in accordance with the Declaration of Helsinki, including the latest revisions.

LPS-induced systemic inflammation study design and procedures

We performed a randomized, placebo-controlled observational study in which participants were allocated to receive either an intravenous LPS injection (n = 7) or a placebo injection with 0.9% NaCl (n = 4). All LPS-challenged participants received a second LPS challenge on d7 after the first LPS challenge, using identical procedures. Bone marrow and blood were collected at baseline (bone marrow: day −7 compared with first LPS challenge, blood: immediately before the first LPS challenge), 4 h after the first LPS challenge (4 h) and at d7. Bone marrow aspiration from the posterior iliac crest was performed by a skilled physician assistant of the Department of Hematology at Radboud university medical center (Radboudumc) in Nijmegen, the Netherlands. Bone marrow was collected in a sodium heparin solution (150 IU ml⁻¹, ratio 3:1). Blood was collected in tubes containing EDTA as an anticoagulant. Additional blood was collected before and at several time points after the LPS challenges.

During LPS or placebo challenge days, all participants underwent the same study procedures, except for administration of either LPS or placebo. Briefly, 24 h before hospitalization, participants needed to refrain from alcohol and caffeine and from 22:00 onward no food and drinks were allowed. Before the challenge, participants were admitted to the research ICU of Radboudumc. An intravenous cannula was placed in an antebrachial vein to administrate fluids and LPS or 0.9% NaCl. A radial artery catheter was inserted to withdraw blood and monitor blood pressure continuously. Prehydration (1.5 l of 2.5% glucose/0.45% NaCl) was administered intravenously in the hour before the challenge. Thereafter, a bolus of 2 ng kg⁻¹ of LPS (E. coli type O113, List Biological Laboratories, lot no. 94332B1) or saline (placebo) was administered intravenously and hydration fluid (2.5% glucose/0.45% NaCl) was continued at an infusion rate of 150 ml h⁻¹ for 8 h. During hospitalization (up to 8 h post-LPS administration), the heart rate was monitored using a four-lead electrocardiogram (M50 Monitor, Philips), core temperature was measured at 30-min intervals with a tympanic thermometer (FirstTemp Genius 2, Covidien) and LPS-induced symptoms (headache, nausea, cold shivers and muscle and back pain) were scored using a numerical six-point scale (0 = no symptoms, 5 = worst symptoms experienced ever) with the addition of 3 points in case of vomiting, resulting in a total symptom score ranging from 0 to 28.

Cell counts and plasma cytokine measurements

Blood cell counts were analyzed using a Sysmex XE-5000. For cytokine determination, blood was centrifuged directly after withdrawal (10 min, 2,000g, 4 °C) and plasma was stored at −80 °C until analysis. Concentrations of TNF, IL-6, CXCL8 (IL-8), IL-10, CCL3 (MIP-1α), CCL2 (MCP-1), CCL4 (MIP-1β) and IL-1RN (IL-1RA) were determined in one batch using a simultaneous Luminex assay (Milliplex, Millipore) on a MagPix instrument (Luminex).

Flow cytometry of monocyte subsets

Blood from LPS-challenged healthy volunteers was phenotyped with antibodies against CD45-Cy5.5 (Beckman Coulter, cat. no. A62835), CD14-ECD (Beckman Coulter, cat. no. B92391), CD16-PE (BD Biosciences, cat. no. 332779), CD64-FITC (Beckman Coulter, cat. no. B49185), CD11b-PC7 (Beckman Coulter, cat. no. A54822), HLA-DR-APC (Beckman Coulter, cat. no. IM3635), DRAQ7 (Biostatus, cat. no. DR71000), CD192-BV421 (BD Biosciences, cat. no. 564067), CD15-KO (Beckman Coulter, cat. no. B01176) and dumpgates CD3-AA750 (Beckman Coulter, cat. no. A94680), CD19-APC A750 (Beckman Coulter, cat. no. A94681) and CD56-APC A750 (Beckman Coulter, cat. no. B46024) on a Navios flow cytometer (Beckman Coulter) at the Department of Hematology of Radboud UMC. Monocyte subtype populations were determined using the gating strategy depicted in Extended Data Fig. 9a. We used 1:100 dilution for all antibodies used in the present study.

In vitro, ex vivo and stimulation experiments

Ex vivo stimulation of monocytes for cytokine production

Classical monocytes (CD14⁺CD16⁻) were isolated from participants taking part in the experimental endotoxemia study at baseline (before LPS administration, d0) as well as at 4 h and 7 d after in vivo LPS administration. To this aim, peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-based density gradient separation (1,200g, 10 min, room temperature, with brake) in SepMate-50 tubes (STEMCELL Technologies). All samples were kept on ice in between procedures. PBMCs were subsequently depleted from neutrophils and intermediate and nonclassic monocytes using CD16 microbeads (Miltenyi Biotec) according to the manufacturer’s protocol. Hereafter, classic monocytes were positively selected using CD14 microbeads (Miltenyi Biotec). Cells were resuspended in culture medium (Roswell Park Memorial Institute (RPMI)-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate and 2 mM GlutaMAX), and 1 × 10⁵ cells were seeded in flat-bottomed 96-well plates. The cells were incubated with culture medium or various stimuli (Pam3Cys (10 µg ml⁻¹; InvivoGen), poly(I:C) (50 µg ml⁻¹; InvivoGen), E. coli LPS (10 ng ml⁻¹, Sigma-Aldrich, serotype 055:B5), flagellin (10 µg ml⁻¹, InvivoGen), resiquimod (R848, 0.35 µg ml⁻¹, InvivoGen), heat-killed E. coli (10⁷ per well, ATCC35218), S. aureus (10⁷ per well, American Type Culture Collection (ATCC), cat. no. 25923), P. aeruginosa (10⁷ per well, PA01), C. albicans (10⁶ per well, UC820) and Aspergillus fumigatus (10⁷ per well (with 10% human pooled serum), V05-27)) for 24 h. Hereafter, supernatants were collected and stored at −80 °C until analysis. The concentrations of TNF, IL-1β, IL-1RN (IL-1RA), IL-6, IL-10 and CCL4 (MIP-1β) in supernatants of ex vivo stimulated cell cultures were determined in one batch using a simultaneous Luminex assay on a MagPix instrument (Luminex).

Ex vivo stimulation of monocytes for analysis of BST2 expression

Pan-monocytes (containing all subsets) were isolated from participants (n = 3) in the experimental endotoxemia study at baseline (before LPS administration d0) as well as at 7 d after in vivo LPS administration. To this end, cryopreserved PBMCs (isolated as described above) were thawed, and monocytes were isolated using a human pan-monocyte isolation kit (negative selection, Miltenyi Biotec). Cells were resuspended in culture medium (RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate and 2 mM GlutaMAX). Subsequently, 1 × 10⁵ cells were seeded in round-bottomed, poly(propylene) 96-well plates and incubated in the presence and absence of E. coli LPS (10 ng ml⁻¹) and/or IFNβ (250 U ml⁻¹, R&D Systems) for 24 h, after which samples were stained and analyzed using flow cytometry (Cytoflex, Beckman Coulter). Flow cytometry antibodies used were mouse anti-human CD45-allophyocyanin (APC)/A750, mouse anti-human CD14-APC antibody (Beckman Coulter), mouse anti-human CD16-phycoerythrin (PE)/Cy7 antibody (BD Biosciences) and mouse anti-human BST2 (CD137/Tetherin) PE (Beckman Coulter).

Isolation and ex vivo stimulation of monocytes for bulk RNA-seq

PBMCs were isolated from participants in the experimental endotoxemia study at baseline (before LPS administration, 0 h) and at 4 h, 8 h, 24 h and 7 d and 7 d + 4 h after in vivo LPS administration. PBMCs were isolated using Ficoll-based density gradient separation (1,200g, 10 min, room temperature, with brake) in SepMate-50 tubes. Cells were washed with cold phosphate-buffered saline (PBS) (1,700 rpm, 10 min, 4 °C) and CD14⁺ monocytes were isolated by positive selection using CD14 microbeads (Miltenyi Biotec). Unstimulated cells were lysed using TRIzol (Invitrogen) and stored at −80 °C until preparation for bulk RNA-seq analysis (described separately below). For ex vivo stimulation, cells were resuspended in culture medium (RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate, 2 mM GlutaMAX and 10% human serum) and seeded in flat-bottomed, 96-well plates (2 × 10⁵ cells per well). Cells were left to attach for 1 h, after which culture medium was refreshed and monocytes were (re)stimulated with 10 ng ml⁻¹ of LPS (Sigma-Aldrich) for 4 h. After stimulation, monocytes were lysed using TRIzol and stored at −80 °C until preparation for bulk RNA-seq analysis.

Co-culture experiments with monocytes and T cells

PBMCs of nine healthy donors were isolated using Ficoll-based density gradient separation (1,200g, 10 min, room temperatures, with brake) in SepMate-TM-50 tubes. Cells were washed with cold PBS (1,700 rpm, 10 min, 4 °C), resuspended in culture medium (RPMI-1640, Dutch modification, Gibco, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate and 2 mM GlutaMAX) and counted on a Sysmex XN-450 (Sysmex Nederland). To obtain a pure lymphocyte fraction, PBMCs were depleted of monocytes using EasySep Human CD14 Positive Selection kit II (STEMCELL Technologies) in a 96-well, non-TC-treated, round-bottomed plate (Corning) in combination with the EasyPlate EasySep Magnet (STEMCELL Technologies). Lymphocytes were counted on a Sysmex XN-450 and brought to a concentration of 1 × 10⁶ ml⁻¹ in PBS for carboxyfluorescein succinimidyl ester (CFSE) labeling using the CellTrace CFSE Cell Proliferation kit (Thermo Fisher Scientific). Lymphocytes were labeled with 500 nM CFSE for 5 min at 37 °C with gentle agitation in the dark. After incubation, ice-cold fetal bovine serum (FBS) Xtra (Capricorn Scientific) was added in a 1:1 ratio to quench any remaining dye. CFSE-labeled lymphocytes were washed once with RPMI-1640, Dutch modification, supplemented with 10% FBS Xtra. After washing, lymphocytes were resuspended in 1 ml of RPMI-1640, Dutch modification containing 10% bovine calf serum (BCS, Capricorn Scientific) and counted on a Sysmex XN-450. Lymphocytes were subsequently kept on ice until cryopreserved classic monocytes (CD14⁺CD16⁻) obtained from participants in the experimental endotoxemia study were thawed. These classic monocytes were obtained from PBMCs isolated at baseline (before LPS administration, 0 h) and at 4 h after in vivo LPS administration, as described above. For co-culture experiments, cells were thawed and washed using pre-warmed (37 °C) complete medium (RPMI-1640, Dutch modification, supplemented with 20% FBS Xtra and 100 µg ml⁻¹ of DNase 1). Monocytes were then resuspended in ice-cold RPMI-1640, Dutch modification, containing 10% BCS and counted on Sysmex XN-450. Subsequently, 5 × 10⁴ CFSE-labeled lymphocytes were cultured in the presence and absence of 1 × 10⁵ monocytes (1:2 lymphocyte:monocyte ratio) in a 96-well, TC-treated, round-bottomed plate (Sarstedt). To activate lymphocyte proliferation and cytokine production, 5 × 10⁴ CD3/CD28 Dynabeads were added to the wells (Thermo Fisher Scientific) and lymphocyte or monocyte co-cultures were left to incubate for 3 d. Thereafter, supernatants were stored at −20 °C for analysis of TNF and IFNγ production using ELISA (R&D Systems), whereas T cell proliferation was analyzed using flow cytometry (Cytoflex, Beckman Coulter) after staining of cells with a mouse anti-human CD3-PeC7 antibody (Beckman Coulter).

In vitro reversal of immunosuppression using IFNβ

Classic monocytes (CD14⁺CD16⁻) of eight healthy donors were isolated as described above. Cells were resuspended in culture medium (RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate and 2 mM GlutaMAX), and 1 × 10⁵ cells per well were seeded in flat-bottomed, 96-well plates. Cells were left to attach for 1 h, after which culture medium was refreshed with culture medium supplemented with 10% human serum. Cells were then incubated for 24 h in the presence or absence of 1 ng ml⁻¹ of LPS. After 24 h, cells were washed with warm PBS. Thereafter, cells were incubated with culture medium supplemented with 10% human serum for 5 d and this culture medium was refreshed on d3. On d6, culture medium was removed and cells were incubated for 24 h in culture medium with 10 ng ml⁻¹ of LPS in the presence and absence of IFNβ (100, 250 and 500 U ml⁻¹, R&D Systems) for 24 h. Afterwards, supernatants were collected and stored at −80 °C until determination of TNF and IL-6 using ELISA (R&D Systems).

Ex vivo reversal of immunosuppression using IFNβ

Cryopreserved classic monocytes (CD14⁺CD16⁻, isolated as described above) obtained from participants in the experimental endotoxemia study at baseline (before LPS administration, d0) and 7 d after in vivo LPS administration were thawed, washed (see above) and resuspended in culture medium (RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate, 2 mM GlutaMAX and 10% human serum). Subsequently, 3 × 10⁵ monocytes were stimulated with 10 ng ml⁻¹ of LPS in the presence or absence of IFNβ (100, 250 and 500 U ml⁻¹) for 4 and 24 h in a 96-well, TC-treated, flat-bottomed plate. After 4 h, cells were stored in RLT buffer containing 40 mM dithiothreitol (QIAGEN) for future RNA isolation and, after 24 h, supernatants were stored at −80 °C until determination of TNF and IL-6 using ELISA.

In vitro monocyte maturation using IFNβ and responsiveness across monocyte subsets

Maturation experiments: classic monocytes (CD14⁺CD16⁻) of six healthy donors were isolated as described above. Monocytes were subsequently incubated in RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate, 2 mM GlutaMAX and 50% human serum for 30 min on ice to block unwanted antibody FC receptor binding. Thereafter, 2.5 × 10⁵ monocytes were incubated for 1 h in 5-ml Falcon, poly(propylene), round-bottomed tubes (Corning) with culture medium (RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate, 2 mM GlutaMAX and 10% human serum) in the presence or absence of 10 µg ml⁻¹ of mouse anti-human IFNα/-β receptor chain 2 antibody (Merck Life Science). Next, 100 U ml⁻¹ of IFNβ (R&D Systems) or culture medium was added and cells were incubated for 3 d, after which they were stained and analyzed using flow cytometry (Cytoflex, Beckman Coulter). Flow cytometry antibodies used were mouse anti-human CD45-APC/A750, mouse anti-human CD14-APC antibody (Beckman Coulter) and mouse anti-human CD16-PE/Cy7 antibody (BD Biosciences).

Responsiveness to IFNβ across subsets: pan-monocytes (containing all subsets) of four healthy donors were isolated as described above. Cells were resuspended in culture medium (RPMI-1640, Dutch modification, supplemented with 50 µg ml⁻¹ of gentamicin, 1 mM sodium pyruvate, 2 mM GlutaMAX and and 50% human serum) was added for 30 min on ice to block unwanted antibody FC receptor binding. Subsequently, 1 × 10⁵ cells were seeded in round-bottomed, poly(propylene), 96-well plates and incubated in the presence and absence of 10 µg ml⁻¹ of mouse anti-human IFNα/-β receptor chain 2 antibody for 1 h. Subsequently cells were incubated with or without of E. coli LPS (10 ng ml⁻¹, Sigma-Aldrich, serotype 055:B5) and IFNβ (100 U ml⁻¹, R&D Systems) for 24 h, after which samples were stained and analyzed using flow cytometry. The flow cytometry antibodies used are described above.

Bulk RNA-seq

Total RNA extraction and cDNA synthesis

Total RNA was extracted from monocytes using the RNeasy RNA extraction kit (QIAGEN), incorporating on-column DNase I (QIAGEN) DNA digestion. Afterwards, ribosomal RNA was removed using riboZero rRNA removal kit (Illumina). The efficiency of rRNA removal was confirmed using a reverse transcription quantitative PCR (RT–qPCR) with primers for glyceraldehyde 3-phosphate dehydrogenase (as internal control) and 18S and 28S rRNA. RNA molecules fragmented into ~200-bp fragments by incubating in fragmentation buffer (200 mM Tris acetate, 500 mM potassium acetate and 150 mM magnesium acetate, pH 8.2) for 7.5 min at 95 °C. First-strand complementary DNA from fragmented RNA was synthesized using SuperScript III reverse transcriptase enzyme (Life Technologies) according to the manufacturer’s protocol and followed by second-strand cDNA synthesis.

Bulk RNA-seq library preparation and sequencing

Gene expression libraries were prepared using KAPA HyperPrep kit (KAPA Biosystems) according to the manufacturer’s protocol. In brief, synthesized double-stranded cDNA was incubated with end repair and A-tailing buffer and enzyme initially for 30 min at 20 °C and then for 30 min at 65 °C. Library-specific adapters were ligated to tailed DNA molecules using DNA ligase enzyme by incubating for 15 min at 15 °C. Ligation reaction was cleaned up using Agencourt AMPure XP reagent (Beckman Coulter) and subsequently amplified using ten cycles of PCR. Finally, 300-bp fragments were selected using a 2% E-gel selection system (Invitrogen). Size selection was validated with a 2100 BioAnalyzer system (Agilent). Prepared libraries were sequenced utilizing NextSeq 500 machine (Illumina) with a paired-end sequencing setup.

Bulk RNA-seq data analysis

Raw RNA-seq reads were aligned to the hg38 reference genome and gene expression profiles were quantified using STAR aligner³⁸. Genes with <50 mapped reads on condition average were excluded from the analysis. For each comparison of RNA-seq profiles, gene expression data of corresponding samples was normalized and differentially expressed genes (DEGs) were identified using the DESeq2 (ref. ³⁹) analysis package utilizing fold-change >2 and q value (Benjamini–Hochberg-adjusted P (P_adj) value) <0.05 as statistical significance cutoffs.

GO and GSEA of bulk RNA-seq data

To infer significantly enriched gene ontologies (GOs) for identified gene sets of interest, such as DEGs, we used the clusterProfiler⁴⁰ analysis package. Gene set enrichment analysis (GSEA) for the comparison of d0 and d7 monocyte gene expression profiles was done using the fgsea⁴¹ package.

ScRNA-seq sample preparation and sequencing

For scRNA-seq, mononuclear cells (MNCs) were isolated from peripheral blood and bone marrow samples using Ficoll-based density gradient separation (1,200g, 10 min, room temperature, with brake) in SepMate-50 tubes, as described above, and cryopreserved at −80 °C until further processing. No depletion or selection of a specific myeloid MNC was performed and the profiling was unbiased. Single Cell Gene Expression 3ʹ v.3 (10x Genomics) was utilized following the manufacturer’s protocol. In brief, approximately 10,000–15,000 single cells were loaded into a channel of chromium chip and the loaded chip was inserted into the chromium controller. After generation of single-cell gel bead-in-emulsion and reverse transcription of RNA, cDNA amplification, fragmentation and adapter ligation were done. The quality of prepared sequencing libraries was assessed using 2100 Bioanalyzer (Agilent). The scRNA-seq libraries were sequenced on a NextSeq 500 (Illumina) or NovaSeq (Illumina).

ScRNA-seq data preprocessing and analysis

Demultiplexing the raw BCL files was done using Cell Ranger mkfastq (v.3.1.0) software and the resulting fastq files were mapped to the human GRCh38 reference genome using Cell Ranger count software with default parameters. The output count matrix was imported to R analysis software and further analyzed using Seurat (v.4.0.4)⁴². Low-quality cells with mitochondrial percentage >15% or with <200 genes and <40,000 UMI (unique molecular modifier) counts were excluded from the analysis. Cells expressing multi-canonical lineage markers at the same time, such as T and B cell-specific markers, were identified as potential doublets and removed from the analysis. Afterwards, gene expression profiles were normalized to sequencing depth and scaled to 10,000 counts and log(transformed). Batch correction for interdonor differences was done utilizing the RunFastMNN function from the SeuratWrappers package, which is the R implementation of the MNN⁴³ batch correction method. Cells were embedded on a two-dimensional view using UMAP¹⁹ and clustered using the FindClusters function from the Seurat package with a resolution of 1. Cluster- or cell-type-specific marker genes were identified using FindMarkers function from the Seurat package with default parameters. Based on well-known cell-type markers, we annotated each cluster and visualized violin plots of gene expression for several cell-type-specific markers using the stacked_violin function of the scanpy (1.9.1)⁴⁴ analysis package in Python. Cell density plots were generated using the embedding_density function of the scanpy package. For lineage-specific analysis, the same steps were done by initial extract of corresponding lineage cells from the whole bone marrow dataset and further embedding and clustering. Unless specified, all figures were generated using the ggplot2 visualization package and all heatmaps were generated using pheatmap package in R.

Statistical differential abundance analysis

To perform differential abundance analysis and find out significantly reduced (or induced) populations we used the miloR package⁴⁵. This package uses the k-nearest neighbor graph base differential abundance analysis, taking into account the metadata from different donors for each condition. We used the suggested workflow of the miloR package for the analysis, setting the k parameter to 30 and the prop parameter to 0.1 for LPS systemic inflammation and late sepsis datasets and 0.05 for the COVID-19 dataset. In datasets with significantly different cell numbers between conditions, cell numbers were downsampled to the same number. Cells were visualized using the ggplot2 package and rasterized for visualization purposes.

Single-sample GSEA and non-negative matric factorization

To identify signaling pathways responsible for variations observed between cell types and time points, we performed single-sample GSEA (ssGSEA) using the VISION⁴⁶ analysis package. We utilized hallmark, GO and Reactome gene sets from molecular Signature Database (MSigDB)⁴⁷. After calculation of the enrichment of each gene set for each single cell, to cluster similar gene sets into one meta-gene set we measured Pearson’s correlation of different gene sets and clustered highly similar gene sets into one meta-gene set, which we defined as the signature set. Using terms in each signature set we annotated each of them. We performed the same analysis for each of the three lineages studied in this article (HSCs and myeloid cells, B cells and pDCs, and T and NK cell lineages).

To identify the underlying gene program responsible for the generation of infMonos and T_inf cells at the 4 h time point of LPS systemic inflammation experiments, we performed non-negative mature factorization (NMF) using the RunNMF function of STutility⁴⁸ package with the default parameters obtaining 40 factors per gene programs. Afterwards, highly similar gene programs were clustered together, generating a meta-gene program. The identified meta-gene program highly enriched in infMonos and T_inf cells was visualized using ggplot2. The top genes contributing to each of gene programs are listed in Supplementary Table 5. The same analysis procedure was done to determine gene programs that are highly active in the intermediate and nonclassic monocyte region. Top genes contributing to NCM-enriched and IFN-I gene programs are listed in Supplementary Table 5.

Analysis of early and late sepsis patient data

For the early sepsis dataset, we downloaded publicly available scRNA-seq data from the Broad Institute Single Cell Portal with the accession no. SCP548 and extracted monocytic and T cell compartments of the data using cell annotations from the corresponding dataset. We performed batch correction for interindividual differences using RunFastMNN function from SeuratWrappers R package and visualized cells using UMAP.

For the late sepsis dataset, we downloaded publicly available scRNA-seq data from the Gene Expression Omnibus (GEO) database with accession no. GSE175453 and extracted the monocytic compartment for the dataset using the cell annotations from the corresponding dataset. We performed batch correction for interdonor differences using the RPCA function from the Seurat package and visualized cells using UMAP. We generated cell density plots for the healthy and late sepsis samples using the embedding_density function of scanpy package.

Analysis of early COVID-19 or sepsis and late COVID-19 data

For the early mixed COVID-19 or sepsis dataset, we downloaded publicly available scRNA-seq data from the CELLxGENE database (https://cellxgene.cziscience.com/e/ebc2e1ff-c8f9-466a-acf4-9d291afaf8b3.cxg). We extracted monocytic and T cell compartments of the data using cell annotations from the corresponding dataset. We performed batch correction for interdonor differences using the RunFastMNN function from the SeuratWrappers R package.

For the COVID-19 convalescent dataset, we downloaded publicly available scRNA-seq data from the GEO database with accession no. GSE158055. We extracted monocytic compartment from the dataset using the cell annotations from the corresponding dataset. We performed batch correction for interdonor differences using the RunFastMNN function from the SeuratWrappers R package and visualized cells using UMAP.

Analysis of neutrophil data

For the neutrophil dataset, we downloaded healthy blood and bone marrow data from ArrayExpress under accession no. E-MTAB-11188. We performed batch correction for interindividual differences using the RunFastMNN function from the SeuratWrappers R package and visualized cells using UMAP.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.