key: cord-271032-imc6woht
authors: Schulte-Schrepping, Jonas; Reusch, Nico; Paclik, Daniela; Baßler, Kevin; Schlickeiser, Stephan; Zhang, Bowen; Krämer, Benjamin; Krammer, Tobias; Brumhard, Sophia; Bonaguro, Lorenzo; De Domenico, Elena; Wendisch, Daniel; Grasshoff, Martin; Kapellos, Theodore S.; Beckstette, Michael; Pecht, Tal; Saglam, Adem; Dietrich, Oliver; Mei, Henrik E.; Schulz, Axel R.; Conrad, Claudia; Kunkel, Désirée; Vafadarnejad, Ehsan; Xu, Cheng-Jian; Horne, Arik; Herbert, Miriam; Drews, Anna; Thibeault, Charlotte; Pfeiffer, Moritz; Hippenstiel, Stefan; Hocke, Andreas; Müller-Redetzky, Holger; Heim, Katrin-Moira; Machleidt, Felix; Uhrig, Alexander; Bosquillon de Jarcy, Laure; Jürgens, Linda; Stegemann, Miriam; Glösenkamp, Christoph R.; Volk, Hans-Dieter; Goffinet, Christine; Landthaler, Markus; Wyler, Emanuel; Georg, Philipp; Schneider, Maria; Dang-Heine, Chantip; Neuwinger, Nick; Kappert, Kai; Tauber, Rudolf; Corman, Victor; Raabe, Jan; Kaiser, Kim Melanie; Vinh, Michael To; Rieke, Gereon; Meisel, Christian; Ulas, Thomas; Becker, Matthias; Geffers, Robert; Witzenrath, Martin; Drosten, Christian; Suttorp, Norbert; von Kalle, Christof; Kurth, Florian; Händler, Kristian; Schultze, Joachim L.; Aschenbrenner, Anna C.; Li, Yang; Nattermann, Jacob; Sawitzki, Birgit; Saliba, Antoine-Emmanuel; Sander, Leif Erik
title: Severe COVID-19 is marked by a dysregulated myeloid cell compartment
date: 2020-08-05
journal: Cell
DOI: 10.1016/j.cell.2020.08.001
sha: 
doc_id: 271032
cord_uid: imc6woht

Summary Coronavirus Disease 2019 (COVID-19) is a mild to moderate respiratory tract infection, however, a subset of patients progresses to severe disease and respiratory failure. The mechanism of protective immunity in mild forms and the pathogenesis of severe COVID-19, associated with increased neutrophil counts and dysregulated immune responses, remains unclear. In a dual-center, two-cohort study, we combined single-cell RNA-sequencing and single-cell proteomics of whole blood and peripheral blood mononuclear cells to determine changes in immune cell composition and activation in mild vs. severe COVID-19 (242 samples from 109 individuals) over time. HLA-DRhiCD11chi inflammatory monocytes with an interferon-stimulated gene signature were elevated in mild COVID-19. Severe COVID-19 was marked by occurrence of neutrophil precursors, as evidence of emergency myelopoiesis, dysfunctional mature neutrophils, and HLA-DRlo monocytes. Our study provides detailed insights into the systemic immune response to SARS-CoV-2 infection and it reveals profound alterations in the myeloid cell compartment associated with severe COVID-19.

immune responses in blood samples in two independent cohorts of COVID-19 patients. 122

Activated HLA-DR hi CD11c hi CD14 + monocytes were increased in patients with mild COVID-123

19, similar to patients with SARS-CoV-2 negative flu-like illness ('FLI'). In contrast, 124 monocytes characterized by low expression of HLA-DR, and marker genes indicative of anti-125 inflammatory functions (e.g. CD163, PLAC8) appeared in patients with severe COVID-19. 126 The granulocyte compartment was profoundly altered in severe COVID-19, marked by the 127 appearance of neutrophil precursors due to emergency myelopoiesis, dysfunctional 128 neutrophils expressing PD-L1, and exhibiting an impaired oxidative burst response. 129

Collectively, our study links highly dysregulated myeloid cell responses to severe J o u r n a l P r e -p r o o f Results 131 Dual center cohort study to assess immunological alterations in COVID-19 patients 132 In order to probe the divergent immune responses in mild vs. severe COVID-19, we 133 analyzed blood samples collected from independent patient cohorts at two university medical 134 centers in Germany. Samples from the Berlin cohort (cohort 1) (Kurth et al., 2020) , were 135 analyzed by mass cytometry (CyTOF) and single-cell RNA-sequencing (scRNA-seq) using a 136 droplet-based single-cell platform (10x Chromium), while samples from the Bonn cohort 137 (cohort 2) were analyzed by multi-color flow cytometry (MCFC) and on a microwell-based 138 scRNA-seq system (BD Rhapsody). We analyzed a total of 24 million cells by their protein 139 markers and >328,000 cells by scRNA-seq in 242 samples from 53 COVID-19 patients and 140 56 controls, including 8 patients with FLI ( Fig. 1A+B , S1A, Table S1 ). 141

We first characterized alterations of the major leukocyte lineages by mass cytometry on 142 whole blood samples from 20 COVID-19 patients collected between day 4 and day 29 after 143 symptom onset, and compared them to 10 age-and gender-matched controls and 8 FLI 144

patients. We designed two antibody panels to specifically capture alterations in mononuclear 145 leukocytes (lymphocytes, monocytes and DCs, panel 1), and in granulocytes (panel 2, Table  146 S2). High-resolution SPADE analysis was performed with 400 target nodes and individual 147 nodes were aggregated into cell subsets based on lineage-specific markers, such as CD14 148

for monocytes and CD15 for neutrophils (Fig. S1B) . Uniform Manifold Approximation and 149

Projection (UMAP) analysis revealed distinct clustering of samples from COVID-19 patients, 150 FLI, and healthy controls, with marked changes of the monocyte and granulocyte 151 compartment (Fig. 1C) . Leukocyte lineages were compared in the earliest available samples 152

in COVID-19 patients (day 4 to 13), FLI, and controls (Table S1, Fig. 1D ). Since leukocyte 153 counts were not available for all control samples, we compared the control samples for 154

CyTOF ('ctrl CyTOF') to data from our recently published healthy control cohorts ('ctrl flow') 155 (Kverneland et al., 2016; Sawitzki et al., 2020) . The proportions of all major lineages were 156 highly similar, irrespective of the methodology (Fig. 1D) . Cell counts of the published cohort 157 could therefore be used as a reference to report absolute cell counts for leukocyte lineages 158

in COVID-19 samples. In line with recent reports Xintian et al., 2020) , 159 we observed elevated leukocytes and increased proportions of neutrophils in patients with 160

severe COVID-19 ( Fig. 1D) , whereas only proportional increases in neutrophils were evident 161 in FLI and mild COVID-19 patients (Fig. 1D) . Total lymphocytes and T cells were strongly 162 reduced in all COVID-19 and FLI patients, whereas non-classical monocytes were 163 specifically depleted in COVID-19 (Fig. 1D) . Increased neutrophils in severe COVID-19 and 164 loss of non-classical monocytes in both mild and severe disease, were validated in cohort 2 165 by MCFC (Fig. S1C, Table S1+3 ). Given the dramatic changes in various immune cell populations (Fig. 1C+D) , we next 171 assessed their composition and activation state by droplet-based scRNA-seq in 27 samples 172 from 18 COVID-19 patients (8 mild & 10 severe, cohort 1, Table S1 ) collected between day 173 3 and day 20 after symptom onset. A total of 48,266 single-cell transcriptomes of PBMC 174

were analyzed together with 50,783 PBMC from publicly available control datasets (21 175 control donors, Table S1 ). UMAP and high-resolution cell type classification identified all cell 176 types expected in the mononuclear compartment of blood with a high granularity in the 177 monocytes, identifying five distinct clusters (cluster 0-4) (Fig. 2A+S2A, Table S4 ). 178 Monocytes in clusters 0-3 expressed CD14, cluster 4 comprised the non-classical 179 monocytes marked by FCGR3A (encoding CD16a) and low expression of CD14. Separate 180 visualization of cells in mild and severe cases revealed highly disease severity-specific 181 clusters (Fig. 2B) . A distinct subset of CD14 + monocytes (cluster 1)( Fig. 2A ) marked by high 182 expression of HLA-DRA, HLA-DRB1 and co-stimulatory molecule CD83 (Fig. S2D) , 183 engagement of which has been linked to prolonged expansion of antigen-specific T cells 184 (Hirano et al., 2006) , was selectively detected in mild COVID-19 (Fig. 2C ). In addition, we 185

identified another closely related CD14 + HLA-DR hi monocyte population (cluster 2), which 186 was characterized by high expression of IFN-stimulated genes (ISGs). However, upon closer 187 analysis, this cluster was found to originate from a single donor with mild COVID-19 (Fig .  188  2A-C, Fig. S2D ). Both cluster 1 and cluster 2 expressed high levels of ISGs IFI6 and ISG15 189 (Fig. S2D) . In patients with severe COVID-19, monocytes showed low expression of DR and high expression of alarmins S100A8/9/12 (cluster 3, Fig. 2A-C, Fig. S2D ). The most 191 prominent change in severe COVID-19 was the appearance of two distinct cell populations 192 (cluster 5+6), absent in PBMC of patients with mild COVID-19 and control donors ( Fig. 2A) . 193 Published markers (Kwok et al., 2020; Ng et al., 2019) identified cluster 5 and 6 as 194 neutrophils and immature neutrophils, respectively (Fig. 2A+B) . Immature neutrophils 195 (cluster 6) expressed CD24, PGLYRP1, DEFA3 and DEFA4, whereas neutrophil cluster 5 196 expressed FCGR3B (CD16b), CXCL8, and LCN2 (lipocalin 2) (Fig. 2C, Fig. S2A ). Their 197 migration within the PBMC fraction on a density gradient marked these cells as low-density 198 neutrophils (LDN). 199

In the second cohort, PBMC from 17 COVID-19 patients (8 mild, 9 severe, Table S1),  200  sampled between 2 and 25 days after symptom onset, and 13 controls, were collected for  201 scRNA-Seq on a microwell-based platform (BD Rhapsody). High-quality single-cell 202 transcriptomes for 139,848 PBMC were assessed and their population structure was 203 visualized using UMAP (Fig. 2D, Table S4 ). Data-driven cell type classification (Aran et al., 204 2019) and cluster-specific marker gene expression identified all cell types expected in the 205 PBMC compartment and revealed additional clusters and substructures (Fig. 2D+S2B) . 206 Similar to cohort 1, monocytes exhibited significant plasticity and were subclassified into 5 207 clusters (Fig. 2D , clusters 0-4). Disease severity-associated changes seen in cohort 1 were 208 validated in cohort 2 (Fig. 2E) . Immature and mature neutrophil clusters were detected in 209 both cohorts (clusters 5-6) and showed near identical marker gene expression (Fig. 2C) .

Similar to cohort 1, a prominent shift in subpopulation occupancy was observed in the 211 monocyte clusters (Fig. 2D+E) . 212

Based on the union of the top 50 genes for monocyte and neutrophil clusters, we found a 213 high correlation between the independently defined functional states within the monocyte 214 J o u r n a l P r e -p r o o f compartment, and mature and immature neutrophils in cohort 1 and cohort 2 (Fig. S2C) . 215 Violin plot representation of important marker genes illustrated distinct phenotypic states and 216 underscored the high similarity of the two cohorts (Fig. S2D) . 217 Disease-severity dependent alterations of the monocyte compartment and the appearance  218 of two LDN populations were detected in two cohorts of COVID-19 patients. 219

Predominance of HLA-DR hi CD11c hi inflammatory monocytes in mild and HLA-220 DR lo CD11c lo CD226 + CD69 + monocytes in severe COVID-19 221 The monocyte compartment is particularly affected by COVID-19, indicated by a loss of 222 CD14 lo CD16 hi non-classical monocytes (Fig. 1C+D) . Disease severity-dependent shifts in 223 monocyte activation were identified by scRNA-seq (Fig. 2) . We further explored the 224 phenotypic alterations of the monocyte compartment using mass cytometry ( Fig. 3A,C,D) . Increased levels of activated HLA-DR hi CD11c hi monocytes in mild COVID-19 249 patients was confirmed by MCFC in cohort 2 (Fig. 3E) . In severe COVID-19, we detected 250 increased expression of CD226 and CD69 (cluster 10) and/or decreased expression of HLA-251 DR, and total CD226 + CD69 + monocytes were elevated compared to controls. Cluster 10 252 expressed high levels of CD10, which is induced during macrophage differentiation (Huang 253 et al., 2020b). Thus, an alternative activation pattern of classical monocytes appeared to be 254 COVID-19 specific and was associated with severe disease. Besides activated lymphocytes, 255 also monocytes upregulate CD69 expression (Davison et al., 2017) , which promotes tissue 256 infiltration and retention (Cibrián and Sánchez-Madrid, 2017) . Similarly, CD226 expression 257 on alternatively activated monocytes might also promote diapedesis and tissue infiltration 258 J o u r n a l P r e -p r o o f (Reymond et al., 2004) . Together, this activation pattern may contribute to the reduction of 259 circulating monocytes in COVID-19. 260 261 HLA-DR lo monocytes persist in severe COVID-19 262 Next, we dissected COVID-19 associated phenotypic alterations of monocytes by scRNA-263 seq. Marker genes of the monocyte clusters derived from Fig. 2A showed that mild COVID-264 19 associated clusters 1 and 2 were characterized by an ISG-driven transcriptional program 265 (Fig. S3A) , and gene ontology enrichment analysis (GOEA) assigned these clusters to 'type 266 I interferon signaling pathway' (Fig. S3B) . A monocyte cluster marked by low expression of 267

HLA-DR and high expression of S100A12 and CXCL8 (cluster 3, HLA-DR lo S100A hi ) was 268 strongly associated with severe COVID-19 (Fig. S3A, 2B, S2D) . For further in-depth 269 analysis, we subclustered the monocyte compartment of the PBMC dataset of cohort 2 ( Fig.  270 2D, S3C, Table S1 ) resulting in 7 subclusters (Fig. 4a) . Cluster 1 was marked by high 271 expression of HLA-DRA and HLA-DRB1 and co-stimulatory molecule CD83 and was 272 therefore designated HLA-DR hi CD83 hi activated inflammatory monocytes (Fig. 4A+B , 273 S3D+E). We identified two major clusters (0, 2) and a smaller cluster 6 with low HLA-DR 274 expression, which were associated with severe COVID-19 (Fig. 4B, S3D+E) . Low HLA-DR 275 expression is an established surrogate marker of monocyte dysfunction (Venet et al., 2020) 276 which results in reduced responsiveness to microbial stimuli (Veglia et al., 2018) , suggesting 277 that cluster 0 and 6 are composed of dysfunctional monocytes. Genes of the S100A family 278

were expressed in both HLA-DR lo clusters (Fig. 4B ), albeit to a higher degree in cluster 0 279 (HLA-DR lo S100A hi , e.g. S100A12, Fig. S2D , S3E, well as pre-maturation markers like MPO and PLAC8 (Fig. 4B) , recently linked to immature 283 monocyte states in sepsis patients (Reyes et al., 2020) . In line with these findings, clusters 284 0, 2 and 6 were significantly enriched in a gene signature derived from sepsis-associated 285 monocytes ( Fig. 4C ) (Reyes et al., 2020) . Moreover, blood monocytes isolated from COVID-286 19 patients showed a blunted cytokine response to LPS stimulation, particularly monocytes 287 from patients with severe COVID-19 (Fig. 4D) . Accordingly, HLA-DR lo monocyte clusters (0, 288 2, 6) were detected almost exclusively in severe COVID-19 (Fig. 4E) . We next analyzed 289 time-dependent cluster occupancies per patient in cohort 2 (Fig. 4E+F) . Activated HLA-290 DR hi CD83 hi monocytes (cluster 1) were found in all cases of mild COVID-19, even at late 291 time points (Fig. 4E+F) . In contrast, HLA-DR lo CD163 hi monocytes (cluster 2) were present 292 mainly early in severe disease, while HLA-DR lo S100A hi monocytes (cluster 0) dominated the 293 late phase of disease (Fig. 4E+F) . Violin plots of ISG (Fig. S3D ) and visualization of marker 294 genes ( Fig. S3E) indicated differential expression patterns of IFN signature genes in 295 individual monocyte clusters. To reveal the kinetics of ISG expression, we plotted the 296 expression of ISG15 and IFI6 in the complete monocyte population for all patients that had 297 been sampled at least twice (Fig. 4G) . Expression levels were highest at early time points 298 and consistently decreased over time, clearly indicating that the IFN response in COVID-19  299 is inversely linked to disease severity and time (Fig. S3F+G) . In contrast, decreased 300 expression of HLA-DRA and HLA-DRB1 in severe COVID-19 is evident early on and 301 sustained over time. 302 J o u r n a l P r e -p r o o f Transcription factor prediction indicated a STAT signaling-driven gene expression program 303

in monocytes in COVID-19 (Fig. 4H) neutrophils and the remaining clusters as mature neutrophils (Fig. S4A) . Accordingly, pro-322 and pre-neutrophils were enriched for transcriptional signatures of neutrophil progenitors 323 derived from published single-cell data ( and pro-neutrophils in cluster 4 and 6 showed the highest proportion of cells with a 325 proliferative signature (Fig. S4B) . Clusters 0, 1, 2 (originally in cluster 4 in Fig. 2A ) 326 expressed mature neutrophil markers FCGR3B (CD16) and MME (CD10) (Fig. S4A) . including CD24, OLFM4, LCN2, and BPI, previously associated with poor outcome in sepsis 341 (Fig. 5B, S4A ) (Kangelaris et al., 2015) . 342

All LDNs also expressed high levels of alarmins S100A8 and S100A9 (Fig. 5D) , whereas 343 other S100 genes (e.g. S100A4, S100A12) were strongly induced in selected neutrophil 344 Alterations of the neutrophil compartment were further interrogated by mass cytometry of 362 whole blood samples of COVID-19 patients (n=8 mild + 9 severe, cohort 1), FLI patients 363 (n=8), and age-and gender-matched controls (n=9) (Table S1), using a panel designed to 364 detect myeloid cell maturation and activation states as well as markers of 365 immunosuppression or dysfunction (Table S2) . Unsupervised clustering analysis of all 366 neutrophils in all samples revealed 10 major clusters (Fig. 6A ) of immature (cluster 2, 5, 6, 367 7), mature (cluster 1, 3, 4) and remaining clusters of low abundancy (cluster 8, 9, 10). Based 368 on their differential expression of CD11b, CD16, CD24, CD34 and CD38, clusters 5 and 6 369 were identified as pro-neutrophils and cluster 2 as pre-neutrophils (Kwok et al., 2020; Ng et 370 al., 2019). The fourth immature cell cluster (7) showed very low expression of CD11b and 371 CD16, reminiscent of pro-neutrophils, but lacking CD34, CD38 and CD24 (Fig. 6A) , 372

suggesting a hitherto unappreciated pro-neutrophil-like population. The mature neutrophils 373 segregated into non-activated (cluster 1), partially activated (cluster 3) and highly activated 374 cells (cluster 4), based on the loss of CD62L and upregulation of CD64, as well as signs of 375 proliferative activity (Ki67 + ) (Fig. 6A) . 376 Neutrophils from COVID-19 patients clearly separated from those of controls and also FLI 377 patients in UMAP analysis (Fig. 6B) , and neutrophils in patients with severe COVID-19 were 378 distinct from those of patients with mild disease (Fig. 6B) . Cells from control donors 379 accumulated in areas enriched for mature non-activated cells (cluster 1) and immature pre-380 neutrophil-like cells (cluster 2). In contrast, neutrophils from FLI patients were mainly mature 381 non-activated (cluster 1) and mature highly activated (cluster 4) cells. Neutrophils from 382 COVID-19, particularly from patients with severe disease primarily occupied immature pre-383

and pro-neutrophil-like clusters. Plotting cell cluster-specific surface marker expression onto 384

the UMAPs (Fig. 6C ) as well as statistical analyses of cell cluster distribution and surface 385 marker expression among different patient groups supported these observations (Fig.  386  6D+E) . Samples from FLI patients contain a high proportion of highly activated mature 387 neutrophils, but barely any immature neutrophils. In contrast, severe COVID-19 is 388 associated with the appearance of immature pre-and pro-neutrophils (Fig. 6D+6E) . 389 Interestingly, immature cell clusters in severe COVID-19 showed signs of recent activation 390 like upregulation of CD64 (Mortaz et al., 2018) , RANK and RANKL (Riegel et al., 2012) , as 391 well as reduced CD62L expression (Mortaz et al., 2018) . In addition to loss of CD62L, 392 immature and mature neutrophils from severe COVID-19 showed elevated PD-L1 393 expression compared to control samples (Fig. 6E) We next assessed the dynamics of the changes within the myeloid cell compartment over 405 time. We grouped samples according to collection time as 'early' (within the first 10 days) or 406 late (during the following 20 days) after onset of symptoms. In both cohorts, we observed a 407 tendency towards (cohort 1) or significantly higher (cohort 2) proportions of granulocytes in 408 severe vs. mild COVID-19 patients, both at early and late time points (Fig. S5A) . We 409 observed a persistent release of immature neutrophils (e.g. cluster 6) in severe COVID-19 410 (Fig. S5B) showing high expression of CD64 and PD-L1, but downregulation of CD62L as a 411 sign of activation, dysfunction and immunosuppression (Fig. S5C ). In addition, severe 412 COVID-19 patients show further increased frequencies of mature, partially activated 413 neutrophils (cluster 3) at later time periods (Fig. S5B) . Thus, the neutrophil compartment of 414 severe COVID-19 patients is characterized by a combination of persistent signs of 415 inflammation and immunosuppression, which is reminiscent of long-term post-traumatic 416

complications (Hesselink et al., 2019) . 417

We also analyzed time-dependent phenotypic changes in the monocyte compartment by 418 mass cytometry. Non-classical monocytes started to recover in COVID-19 patients during 419 the later stages of the disease (Fig. S5A) . HLA-DR hi CD11c hi monocyte cell clusters also 420 declined at later time points in mild COVID-19 ( Fig. S5D,E,F) , which correlates well with the 421 longitudinal changes of IFI6 and ISG15 as well as HLA-DRA, and HLA-DRB1 expression 422 profiles (Fig. 4G+S3F) . In contrast, overall proportions of HLA-DR hi CD11c hi monocytes in 423 severe COVID-19 remained low throughout the course of the disease. Proportions of CD10 hi 424 macrophage-like cluster 10 and CD226 + CD69 + monocytes were generally higher at later 425 stages in severe COVID-19, which resembled the kinetics of HLA-DR lo S100A hi monocytes 426 identified by scRNA-seq (Fig. 4F ). This indicates a prolonged alternative activation of 427 monocytes in severe COVID-19 (Fig. S5E) . Table S1 ). Integrated visualization of 435 all samples of cohort 2 (fresh/frozen PBMC, fresh whole blood, 229,731 cells, Fig. S6A ) 436 revealed the expected blood leukocyte distribution, including granulocytes ( Fig. 7A, S6A , 437 Table S4 ). Cell type distribution identified by scRNA-seq profiles (Fig. S6B ) strongly 438 correlated with MCFC characterization of the same samples (Fig. S6C) . For further analysis 439 of the granulocyte compartment, we first combined the whole blood samples with the fresh 440 PBMC to guide the clustering of all major immune cells resulting in a total of 122,954 cells 441 (Fig 7A) . From these samples, we identified all neutrophil clusters and extracted the cells 442 derived from whole blood for subsampling, which revealed a structure of 9 clusters 443 (n=58,383 cells, Fig. 7B+C ). 444

Using marker-and data-driven approaches as applied to LDN (Fig. 5D, S4A) , we identified 445 FUT4(CD15) + CD63 + CD66b + pro-neutrophils, ITGAM(CD11b) + CD101 + pre-neutrophils, along 446 with 7 mature neutrophil clusters ( Fig. 7B -D, S6D, Table S4 ). Heterogeneous expression of 447 various markers involved in mature neutrophil function including CXCR2, FCGR2A (CD32), 448 FCGR1A (CD64) or MME (CD10), pointed towards distinct functionalities within the 449 neutrophil compartment (Fig. 7E, S6D+E) . Seven of the nine 9 neutrophil clusters identified 450

in whole blood in cohort 2, could also be mapped to the fresh PBMC transcriptomes in 451 cohort 1 (Fig. S6F) , indicating that scRNA-seq of fresh PBMC in COVID-19 patients reveals 452 relevant parts of the neutrophil space. The transcriptional phenotype of pro-and pre-453 neutrophils (cluster 8+9) was corroborated in cohort 2 ( Fig. 7B-D, S6D) . 454

Heatmap and UMAP visualization of the cell type distribution identified pro-and pre-455 neutrophils mainly at late time points in severe COVID-19 ( Fig. 7F-G) . Furthermore, mature 456 neutrophils with a high IFN-signature (cluster 1) were associated with severe COVID-19 457 (Fig. 7E, S6G ). This cluster was also enriched for markers identified by CyTOF as 458 differentially expressed in patients with severe COVID-19 ( Fig. 6) , such as elevated 459 expression of CD274 (PD-L1) and FCGR1A (CD64) (Fig. 7H ). In addition to CD274, cells in 460 cluster 1 expressed genes indicative of a potentially suppressive or anti-inflammatory state, 461

including ZC3H12A (Fig. 7E) , which is known to suppress hepatitis C virus replication and 462

virus-induced pro-inflammatory cytokine production (Lin et al., 2014) . Cluster 2 was also 463 enriched for cells from COVID-19 patients, mainly from severe but also mild cases (Fig. 7F is mainly driven by E2F family members and pre-neutrophils mainly depend on ETS TFs 475 (Fig. S6H) . 476

Pseudotime analysis strongly supported the differentiation trajectory from pro-neutrophils 477 (cluster 8) via pre-neutrophils (cluster 6) to mature neutrophils in cluster 2 and 1 ( Fig. S6I-J) . 478 Particularly CD274 (PD-1L) was enriched in cluster 1 compared to cluster 2, supporting the 479 potential of neutrophils to progress towards a suppressive phenotype in severe COVID-19 480 (Fig. S6J) . Interestingly, CD177 is expressed in pre-neutrophils and persisting in cluster 1 481 further highlighting the newly emerging character of this cluster (Volkmann et al., 2020) . 482

Finally, we studied whether the persistent emergence of immature, potentially dysfunctional 483 neutrophils in severe COVID-19 patients can be captured under routine diagnostic 484

conditions. Therefore, samples of 32 COVID-19 patients ( Table S1 , cohort 1) were 485 characterized by routine hematology analyses using a clinical flow cytometry system 486 (Sysmex analyzer). Indeed, the assumption of rescue myelopoiesis in severe COVID-19 was 487 supported by significantly higher counts in the population of immature granulocytes (IG,  488 representing promyelocytes, myelocytes, and metamyelocytes) in this patient group ( Fig.  489 7K). We also found significant differences in the neutrophil compartment, when analyzing 490 the width of dispersion with respect to granularity, activity, and cell volume defined as NE-491 WX, NE-WY and NE-WZ, respectively. As compared to patients with mild course, severely ill 492 patients displayed increases in width of dispersion of activity and cell volume as surrogates 493

for increased cellular heterogeneity, immaturity and dysregulation in severe COVID-19 ( Fig.  494 7K), resembling previously described alterations in sepsis patients (Stiel et al., 2016) . 495

Furthermore, neutrophils of severe COVID-19 patients were partially dysfunctional as their 496

oxidative burst upon stimulation with standardized stimuli (E.coli or PMA) was strongly 497 impaired in comparison to control and mild COVID-19 neutrophils, whereas phagocytic 498 activity was preserved (Fig. 7L , Table S1 ). 499

Collectively, the neutrophil compartment in peripheral blood of severe COVID-19 patients is 500 characterized by the appearance of LDN, FUT4(CD15) + CD63 + CD66b + pro-neutrophils, and 501 ITGAM(CD11b) + CD101 + pre-neutrophils, reminiscent of emergency myelopoiesis, as well as 502 CD274(PD-L1) + ZC3H12A + mature neutrophils reminiscent of gMDSC-like cells, which might 503 exert suppressive or anti-inflammatory functions. 504 dysfunctional phenotype, PLAC8 was recently shown to suppress production of IL-1β and IL-559

18 (Segawa et al., 2018) . In fact, we observed that inflammatory cytokine production, 560

including IL-1β release, was impaired in monocytes from patients with severe COVID-19 561 (Fig. 4) . PBMC fractions in severe COVID-19 contained immature neutrophils, including pro-and pre-574 neutrophils, which was not observed in mild cases (Fig. 5) . These immature LDN showed a 575 surface marker and gene expression profile reminiscent of granulocytic MDSCs including 576 genes such as S100A12, S100A9, MMP8, ARG1 (Uhel et al., 2017) , and OLFM4, which has 577 been recently associated with immunopathogenesis in sepsis (Alder et al., 2017) . 578

Emergence of pro-neutrophils in severe COVID-19 was also detected by single-cell 579 proteomics on whole blood samples. Strikingly, both immature and the mature neutrophils 580

showed increased expression of CD64 and PD-L1 (Fig. 6+S5 ), similar to recently described 581 alterations in sepsis (Meghraoui-Kheddar et al., 2020). In addition to the altered phenotype, 582

we also observed an altered functionality. Neutrophils from patients with severe COVID-19 583

showed an impaired oxidative burst response, while their phagocytic capacity was preserved 584 (Fig. 7) . 585

Single-cell transcriptomics of whole blood samples revealed mature activated neutrophils in 586 both mild and severe COVID-19 (Fig. 7B, cluster 2) , however, expression of CD274 (PD-L1) 587

was only found in severe COVID-19 (cluster 1), and it increased in later stages of the 588 disease. Expression of PD-L1 on neutrophils has been associated with T cell suppression 589 (Bowers et Methodology: JS-S, DP, TK, SB, LB, EDD, MG, DW, MB, TSK, AS, OD, HM, ARS, CC, DK, EV,  664  CJX, AD, CT, SH, CLG, ML, EW, TU, MB, RG, (Table S4) . (Table S4) . within the monocyte space of cohort 1 (related to Fig. 2, Table S4 ). cluster ranked by adjusted p-values. 894 C, Back-mapping of monocyte clusters of cohort 2 (Fig. 4C) onto the PBMC UMAP of cohort 895 2 (Fig. 2D) . The legend shows the association of the colors to the clusters together with the 896 labeling of the clusters based on expressed marker genes (according to Fig. 2 and Fig.  897 S3D-F). 898 D, Violin plots of marker gene expression in the monocyte clusters identified in the complete 899 PBMC space of cohort 2 (Fig. 2C,D ) 900 E, Dot plot of the top 10 marker genes sorted by average log fold change calculated for the 901 monocyte clusters (Fig. 4C) . severe COVID-19 patients ( Figure 1A+B, Table S1 ). Information on age, sex, medication, 1020 and co-morbidities is listed in COVID-19 patients ( Figure 1A+B , Table S1 ) were included in the study. In patients who 1030

were not able to consent at the time of study enrollment, consent was obtained after 1031 recovery. Information on age, sex, medication, and co-morbidities are listed in Table S1 . 1032

COVID After one wash in DPBS cells were directly processed for scRNA-seq (BD Rhapsody) or 1058 multi-color flow cytometry (MCFC). Frozen PBMC were recovered by rapidly thawing frozen 1059 cell suspensions in a 37°C water bath followed by immediate dilution in pre-warmed RPMI-1060 1640+10% FBS (Gibco) and centrifugation at 300g for 5min. After centrifugation, the cells 1061

were resuspended in RPMI-1640+10% FBS and processed for scRNA-seq. Antibody 1062 cocktails were cryopreserved as described before (Schulz et al., 2019) . 1063

All anti-human antibodies pre-conjugated to metal isotopes were obtained from Fluidigm 1066

Corporation (San Francisco, US). All remaining antibodies were obtained from the indicated 1067 companies as purified antibodies and in-house conjugation was done using the MaxPar X8 1068 labeling kit (Fluidigm). TLRpure; Innaxon, UK). After stimulation, cell-free supernatants were collected and tested 1160

for IL-1β, IFNγ, and TNFα, respectively, using the cytokine bead assay Legend-Plex 1161

Mix&Match test was used to report differences in IG count, whereas mixed-effect-analysis and Sidak's 1182 multiple comparison test was applied to report statistical differences of NE-WX, NE-WY and 1183 NE-WZ between mild and severe COVID-19 patients. 1184 1185 10x Genomics Chromium single-cell RNA-seq 1186 PBMC were isolated and prepared as described above. Afterwards, patient samples were 1187 hashtagged with TotalSeq-A antibodies (Biolegend) according to the manufacturer's protocol 1188 for TotalSeq TM -A antibodies and cell hashing with 10x Single Cell 3' Reagent Kit v3.1. 50µL 1189 cell suspension with 1x10 6 cells were resuspended in staining buffer (2% BSA, Jackson 1190 Immuno Research; 0.01% Tween-20, Sigma-Aldrich; 1x DPBS, Gibco) and 5 µL Human 1191

TruStain FcX TM FcBlocking (Biolegend) reagent were added. The blocking was performed 1192 for 10min at 4°C. In the next step 1µg unique TotalSeq-A antibody was added to each 1193 sample and incubated for 30min at 4°C. After the incubation time 1.5mL staining buffer were 1194 added and centrifuged for 5min at 350g and 4°C. Washing was repeated for a total of 3 1195

washes. Subsequently, the cells were resuspended in an appropriate volume of 1x DPBS 1196 (Gibco), passed through a 40µm mesh (Flowmi TM Cell Strainer, Merck) and counted, using a 1197

Neubauer Hemocytometer (Marienfeld). Cell counts were adjusted and hashtagged cells 1198

were pooled equally. The cell suspension was super-loaded, with 50,000 cells, in the 1199 Chromium TM Controller for partitioning single cells into nanoliter-scale Gel Bead-In-1200

Emulsions (GEMs). Single Cell 3' reagent kit v3.1 was used for reverse transcription, cDNA 1201 amplification and library construction of the gene expression libraries (10x Genomics) 1202

following the detailed protocol provided by 10x Genomics. Hashtag libraries were prepared 1203 according to the cell hashing protocol for 10x Single Cell 3' Reagent Kit v3.1 provided by 1204

Biolegend, including primer sequences and reagent specifications. Biometra Trio Thermal 1205

Cycler was used for amplification and incubation steps (Analytik Jena). Libraries were 1206 quantified by Qubit TM 2.0 Fluorometer (ThermoFisher) and quality was checked using 2100 1207

Bioanalyzer with High Sensitivity DNA kit (Agilent). Sequencing was performed in paired-end 1208 mode with a S1 and S2 flow cell (2× 50 cycles) using NovaSeq 6000 sequencer (Illumina). the Qubit dsDNA HS Kit (ThermoFisher) and the size-distribution was measured using the 1229

Agilent high sensitivity D5000 assay on a TapeStation 4200 system (Agilent technologies). 1230

Sequencing was performed in paired-end mode (2*75 cycles) on a NovaSeq 6000 and 1231

NextSeq 500 System (Illumina) with NovaSeq 6000 S2 Reagent Kit (200 cycles) and 1232

NextSeq 500/550 High Output Kit v2.5 (150 Cycles) chemistry, respectively. 1233

Data pre-processing of 10x Genomics Chromium scRNA-seq data 1236 CellRanger v3.1.0 (10x Genomics) was used to process scRNA-seq. To generate a digital 1237 gene expression (DGE) matrix for each sample, we mapped their reads to a combined 1238 reference of GRCh38 genome and SARS-CoV-2 genome and recorded the number of UMIs 1239

for each gene in each cell. 1240 1241 Data pre-processing of BD Rhapsody scRNA-seq data 1242

After demultiplexing of bcl files using Bcl2fastq2 V2.20 from Illumina and quality control, 1243

paired-end scRNA-seq reads were filtered for valid cell barcodes using the barcode whitelist 1244 provided by BD. Cutadapt 1.16 was then used to trim NexteraPE-PE adapter sequences 1245

where needed and to filter reads for a PHRED score of 20 or above (Martin, 2011) . Then, 1246 STAR 2.6.1b was used for alignment against the Gencode v27 reference genome (Dobin et  1247 al implemented in Seurat. 1258

We excluded cells based on the following quality criteria: more than 25% mitochondrial 1260 reads, more than 25% HBA/HBB gene reads, less than 250 expressed genes or more than 1261 5,000 expressed genes and less than 500 detected transcripts. We further excluded genes 1262 that were expressed in less than five cells. In addition, mitochondrial genes have been 1263 excluded from further analysis. 1264

LogNormalization (Seurat function) was applied before downstream analysis. The original 1266

gene counts for each cell were normalized by total UMI counts, multiplied by 10,000 (TP10K) 1267 and then log transformed by log10(TP10k+1). 1268

After normalization, the count data was scaled regressing for total UMI counts and principal 1270 component analysis (PCA) was performed based on the 2,000 most variable features 1271 identified using the vst method implemented in Seurat. Subsequently, the scRNA-seq data 1272 from cohort 1 was integrated with publicly available 10x scRNAseq data of healthy controls 1273 using the 'harmony' algorithm (Korsunsky et benchmarking data from healthy controls and 10x v3.1 scRNA-seq data from cohort 1). We 1277 downloaded the count matrices for the publicly available scRNA-seq data and filtered the 1278 cells using the above-mentioned quality criteria prior to data integration. For two-dimensional 1279 data visualization we performed UMAP based on the first 20 dimensions of the 'harmony' 1280 data reduction. The cells were clustered using the Louvain algorithm based on the first 20 1281

'harmony" dimensions with a resolution of 0.4. 1282

Differential expression (DE) tests were performed using FindMarkers/FindAllMarkers 1284 functions in Seurat with Wilcoxon Rank Sum test. Genes with >0.25 log-fold changes, at 1285 least 25% expressed in tested groups, and Bonferroni-corrected p-values<0.05 were 1286 regarded as significantly differentially expressed genes (DEGs). Cluster marker genes were 1287 identified by applying the DE tests for upregulated genes between cells in one cluster to all 1288 other clusters in the dataset. Top ranked genes (by log-fold changes) from each cluster of 1289 interest were extracted for further illustration. The exact number and definition of samples 1290 used in the analysis are specified in the legend of Fig. 2A and summarized in Table S1 . 1291

Clusters were annotated based on a double-checking strategy: 1) by comparing cluster 1293 marker genes with public sources, and 2) by directly visualizing the expression pattern of 1294

CyTOF marker genes. 1295

Significant DEGs between each monocyte cluster and the rest of monocyte subpopulations 1297

were identified by FindMarkers function from the Seurat package using Wilcoxon Rank Sum 1298 test statistics for genes expressed in at least 25% of all monocyte clusters. P-values were 1299 corrected for multiple testing using Bonferroni correction and genes with corrected p-values 1300 lower or equal 0.05 have been taken as significant DEGs for GO enrichment test by R 1301 package/ClusterProfiler v.3.10.1 (Yu et al., 2012) . 1302

To systematically compare the similarity of marker gene expression in the identified 1305 monocyte/neutrophils subpopulations between the two cohorts, the Spearman correlation 1306 coefficients were calculated based on the union of the top 50 marker genes of each cluster 1307 sorted by fold change in the two cohorts, based on their average expression of all cells in the 1308 specific subpopulation. The pairwise comparisons were performed, and the correlation 1309 coefficients were displayed using a heatmap. 1310

The neutrophil space was investigated by subsetting the PBMC dataset to those clusters 1312 identified as neutrophils and immature neutrophils (cluster 5 and 6). Within those subsets, 1313

we selected top 2,000 variable genes and repeated the clustering using the SNN-graph 1314

based Louvain algorithm mentioned above with a resolution of 0.6. The dimensionality of the 1315 data was then reduced to 10 PCs, which served as input for the UMAP calculation. To 1316 categorize the observed neutrophil clusters into the respective cell cycle states, we applied 1317

the CellCycleScoring function of Seurat and visualized the results as pie charts. 1318

A gene signature enrichment analysis using the 'AUCell' method (Aibar et al., 2017) was 1319 applied to link observed neutrophil clusters to existing studies and neutrophils of cohort 2. 1320

We set the threshold for the calculation of the area under the curve (AUC) to marker genes 1321 from collected publications and top 30 of the ranked maker genes from each of neutrophil 1322

clusters from cohort 2. The resulting AUC values were normalized the maximum possible 1323 AUC to 1 and subsequently visualized in violin plots or UMAP plots. 1324 1325

ScRNA-seq UMI count matrices were imported to R 3.6.2 and gene expression data 1328 analysis was performed using the R/Seurat package 3.1.2 (Butler et al., 2018) . 1329

Demultiplexing of cells was performed using the HTODemux function implemented in 1330

Seurat. After identification of singlets, cells with more than 25% mitochondrial reads, less 1331 than 250 expressed genes or more than 5,000 expressed genes and less than 500 detected 1332 transcripts were excluded from the analysis and only those genes present in more than 5 1333 cells were considered for downstream analysis. The following normalization, scaling and 1334 dimensionality reduction steps were performed independently for each of the data subsets 1335 used for the different analyses as indicated respectively. In general, gene expression values 1336

were normalized by total UMI counts per cell, multiplied by 10,000 (TP10K) and then log 1337 transformed by log10(TP10k+1). Subsequently, the data was scaled, centered and 1338 regressed against the number of detected transcripts per cell to correct for heterogeneity 1339 associated with differences in sequencing depth. For dimensionality reduction, PCA was 1340 performed on the top 2,000 variable genes identified using the vst method implemented in 1341

Seurat. Subsequently, UMAP was used for two-dimensional representation of the data 1342 structure. Cell type annotation was based on the respective clustering results combined with 1343 data-driven cell type classification algorithms based on reference transcriptome data (Aran 1344 et al., 2019) and expression of known marker genes. 1345

ScRNA-seq count data of 229,731 cells derived from fresh and frozen PBMC samples 1348 purified by density gradient centrifugation and whole blood after erythrocyte lysis of cohort 2 1349 (Bonn, BD Rhapsody) were combined, normalized and scaled as described above (see Fig.  1350 S6A). After variable gene selection and PCA, UMAP was performed based on the first 20 1351 principal components (PCs). No batch correction or data integration strategies were applied 1352 to the data. Visualization of the cells (Fig. S6A) showed overlay of cells of the same type 1353 (e.g. T cells clustered within the same cluster, irrespective of cell isolation procedure). In 1354 other words, cell type distribution was unaffected by the technical differences in sample 1355

handling. Data quality and information content was visualized as violin plots showing the 1356 number of detected genes, transcripts (UMIs) and genic reads per sample handling strategy 1357 split by PBMC and granulocyte fraction. 1358

ScRNA-seq count data of 139,848 cells derived from fresh and frozen PBMC samples of 1360 cohort 2 (Bonn, BD Rhapsody) purified by density gradient centrifugation were normalized 1361 and scaled as described above. After variable gene selection and PCA, UMAP was 1362 performed and the cells were clustered using the Louvain algorithm based on the first 20 1363

PCs and a resolution of 0.4. Cluster identities were determined by reference-based cell 1364 classification and inference of cluster-specific marker genes using the Wilcoxon rank sum 1365 test using the following cutoffs: genes have to be expressed in more than 20% of the cells of 1366 the respective cluster, exceed a logarithmic fold change cutoff to at least 0.2, and exhibited a 1367 difference of >10% in the detection between two clusters. The exact number and definition of 1368 samples used in the analysis are specified in the legend of Fig. 2D and summarized in 1369 Table S1 . 1370

To compare shifts in the monocyte and neutrophil populations in the PBMC compartment of 1373 COVID-19 patients, the percentages of the cellular subsets -as identified by clustering and 1374 cluster annotation explained above for the two independent scRNA-seq data sets (cohort 1 1375 and cohort 2) -of the total number of PBMC in each data set were quantified per sample and 1376 visualized together in box plots. To determine the statistical significance of differences in cell 1377

proportions between the different conditions, a Dirichlet regression model was used, due to 1378 the fact that the proportions are not independent of one another. The R/RDirichletReg 1379 (Maier, 2014) package was used. The p-values were corrected for multiple testing using the 1380

Benjamini-Hochberg procedure. 1381

The monocyte space was investigated by subsetting the PBMC dataset to those clusters 1383 identified as monocytes (cluster 0-4), removing cells with strong multi-lineage marker 1384 expressions, and repeating the variable gene selection (top 2,000 variable genes), 1385 regression for the number of UMIs and scaling as described above. The dimensionality of 1386 the data was then reduced to 8 PCs, which served as input for the UMAP calculation. The 1387 SNN-graph based Louvain clustering of the monocytes was performed using a resolution of 1388 0.2. Marker genes per cluster were calculated using the Wilcoxon rank sum test using the 1389 following cutoffs: genes have to be expressed in >20% of the cells, exceed a logarithmic fold 1390 change cutoff to at least 0.25, and exhibited a difference of >10% in the detection between 1391 two clusters. The exact number and definition of samples used in the analysis are specified 1392 in the legend of Fig. 4A and summarized in Table S1 . 1393 J o u r n a l P r e -p r o o f

For each patient and time point of sample collection, the proportional occupancy of the 1395 monocyte clusters was calculated, and the relative proportions were subsequently visualized 1396 as a function of time. 1397

ScRNA-seq count data derived from fresh PBMC samples purified by density gradient 1399 centrifugation and whole blood after erythrocyte lysis of cohort 2 (BD Rhapsody) were 1400 normalized, scaled, and regressed for the number of UMI per cell as described above. After 1401 PCA based on the top 2,000 variable genes, UMAP was performed using the first 30 PCs. 1402

Cell clusters were determined using Louvain clustering implemented in Seurat based on the 1403 first 30 principle components and a resolution of 0.8. Cluster identities were assigned as 1404 detailed above using reference-based classification and marker gene expression. 1405

Subsequently, the dataset was subsetted for whole blood samples after erythrocyte lysis and 1406 clusters identified as neutrophils and immature neutrophils, and re-scaled and regressed. 1407

After PCA on the top 2,000 variable genes, the neutrophil subset data was further processed 1408

using the data integration approach implemented in Seurat (Stuart et al., 2019) based on the 1409 first 30 PCs removing potential technical biases of separate experimental runs. UMAP and 1410 clustering were performed as described above on the top 12 PCs using a resolution of 0.3. 1411

Differentially expressed genes between clusters were defined using a Wilcoxon rank sum 1412 test for differential gene expression implemented in Seurat. Genes had to be expressed in 1413 >10% of the cells of a cluster, exceed a logarithmic threshold >0.1. The exact number and 1414 definition of samples used in the analysis are specified in the legend of Fig. 7A and  1415 summarized in Table S1 . 1416

After cell type classification of the combined scRNA-seq data set of fresh PBMC and whole 1419 blood samples of cohort 2 described above, 89,883 cells derived from whole blood samples 1420 after erythrocyte lysis were subsetted. Percentages of cell subsets in those whole blood 1421 samples of the total number of cells were quantified per sample and visualized in box plots 1422 separated by disease stage and group. 1423 

To categorize the cells within the neutrophil clusters into the respective cell cycle states, we 1434 applied the CellCycleScoring function of Seurat and visualized the results as pie charts. 1435

Trajectory analysis was performed using the destiny algorithm v3.0.1 (Angerer et al., 2016) .

generate UMAP representations all events of a given population of interest were down-1473 sampled to 70,000 cells and then embedded using the tumap function (R uwot package,  1474 https://CRAN.R-project.org/package=uwot) parameterized by local neighborhood 50, 1475 learning rate 0.5, and using the indicated markers ( 3  8  11  14  22  24  3  8  14  6  10  12  13  17  13  22  7  11  16  8  13  13  18  15  20  6  7  6  8  11  7  8  12  5  8  11  5  7  11  19  9  12  16  17  19  23  9  16  9  16  8  15  13 STAT3   FKBP5  LGALS9  IFITM3  IFIT2  ISG15   IFI27  MX2  IFI6  IFIT1   HERC5  OASL  MX1  IFIH1  IFI44   IFI44L  OAS2   SERPINB1   IL1R2   SERPING1  CD163  RNASE1   IFI16   OAS3   ADAR LGALS3BP SPI1  DEFA3  DEFA4  HSP90AA1  MPO  ELANE  PRTN3  CD24  BPI  CD63  CLEC5A  FUT4  HEXA  PDE4D  C1QBP  CEACAM8  ANXA1  GSN  CLEC12A  NLRC4  OLFM4  CYBB  LCN2  LGALS3  LTF  MMP8  HP  CD101  CAMP  S100A8  S100A12  CD177  TSPO  RAB27A  S100P  ITGAM  S100A9  S100A6  IFI6  ISG15  LY6E  IFI16  GBP1  CCR1  C3AR1  IFIH1  DDX58  FCGR1A  AIM2  ZC3H12A  TLR2  ABCA1  ICAM1  INPP4B  FBXL17  SLC38A1  CLEC2D  ITGA4  SELL  S100A11  CXCR2  TLR5  CLEC4E  FCGR2A  ADAM8  SLC11A1  NLRP12  TLR4  C5AR1  CORO1A  CMTM6  TNFRSF1A  TNFRSF1B  RAC1  NLRP3  PTPRC  PTGS2  SIRPA  NCOA4 MME S100A4 % exp. 

Transcriptome meta-analysis deciphers a 1524 dysregulation in immune response-associated gene signatures during sepsis

SCENIC: Single-cell 1528 regulatory network inference and clustering

Is a Candidate Marker for a Pathogenic Neutrophil Subset in Septic Shock

Destiny: Diffusion maps for large-scale single-cell data in R

Reference-based analysis of lung single-cell 1536 sequencing reveals a transitional profibrotic macrophage

Gene ontology: Tool for the unification of 1539 biology

The pathogenicity of SARS-CoV-2 in hACE2 transgenic mice

Targeting potential 1545 drivers of COVID-19: Neutrophil extracellular traps

Myeloid-derived suppressor cell subsets drive glioblastoma growth 1548 in a sex-specific manner

Human intestinal pro-inflammatory CD11chighCCR2+CX3CR1+ 1552 macrophages, but not their tolerogenic CD11c-CCR2-CX3CR1-counterparts, are expanded 1553 in inflammatory bowel disease article

Immune Suppression by Neutrophils in HIV-1 Infection: Role of PD-L1/PD-1 1556 Pathway

Presence of SARS-CoV-2 reactive T 1559 cells in COVID-19 patients and healthy donors

L-arginine 1561 metabolism in myeloid cells controls T-lymphocyte functions

Recommendations for 1564 myeloid-derived suppressor cell nomenclature and characterization standards

Integrating single-1567 cell transcriptomic data across different conditions, technologies, and species

The Gene Ontology Resource: 20 years and 1571 still GOing strong

Deciphering myeloid-derived suppressor cells: 1574 isolation and markers in humans, mice and non-human primates

Neutrophils which migrate to lymph nodes modulate CD4+ T cell response by a PD-L1 1578 dependent mechanism

Clinical and immunologic features in severe and moderate Coronavirus 1581 Disease

COVID-19 severity correlates with airway 1584 epithelium-immune cell interactions identified by single-cell analysis

CCL2 Promotes Colorectal Carcinogenesis by Enhancing 1588

Polymorphonuclear Myeloid-Derived Suppressor Cell Population and Function

CD69: from activation marker to metabolic 1591 gatekeeper

From mice to monkeys, animals studied for coronavirus answers

MAFB Determines Human Macrophage Anti-Inflammatory Polarization: Relevance for the 1597

Pathogenic Mechanisms Operating in Multicentric Carpotarsal Osteolysis

Neutrophils with myeloid derived 1601 suppressor function deplete arginine and constrain T cell function in septic shock patients

Platelet, monocyte and neutrophil activation and glucose tolerance in 1605

Favorable Anakinra Responses in Severe Covid-19 Patients with Secondary 1609

STAR: ultrafast universal RNA-seq aligner

Genomewide Association Study of Severe 1615

Covid-19 with Respiratory Failure

CD163 1618 expression defines specific, IRF8-dependent, immune-modulatory macrophages in the bone 1619 marrow

Complex Immune Dysregulation in COVID-19 Patients with Severe Respiratory Failure

Targets of T cell responses to 1626 SARS-CoV-2 coronavirus in humans with COVID-19 disease and unexposed individuals

Complex heatmaps reveal patterns and 1629 correlations in multidimensional genomic data

Impaired type I interferon activity and 1632 exacerbated inflammatory responses in severe Covid-19 patients

Normalization and variance stabilization of single-cell 1635 RNA-seq data using regularized negative binomial regression

Neutrophil heterogeneity and its role in infectious 1639 complications after severe trauma

Engagement of CD83 ligand induces prolonged 1643 expansion of CD8+ T cells and preferential enrichment for antigen specificity

Stroke-induced 1647 immunodepression and dysphagia independently predict stroke-associated pneumonia -1648 The PREDICT study

Dexamethasone in Hospitalized Patients with Covid-1651 19 -Preliminary Report

Simultaneous inference in general parametric 1653 models

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Induced CD10 expression during monocyte-to-macrophage differentiation 1659 identifies a unique subset of macrophages in pancreatic ductal adenocarcinoma

Should we stimulate or suppress immune responses in COVID-19? 1663 Cytokine and anti-cytokine interventions

Gene List to a Gene Regulatory Network Using Large Motif and Track Collections

Heterogeneity among septic shock patients in a 1670 set of immunoregulatory markers

Human suppressive neutrophils CD16 bright /CD62L dim exhibit decreased 1673 adhesion

Toward understanding the origin and evolution of cellular organisms

Increased expression of neutrophil-related genes 1678 in patients with early sepsis-induced ARDS

Suppress Lymphocyte Proliferation through Expression of PD-L1

Incidence of thrombotic complications in critically ill ICU patients with COVID-19

Fast, sensitive and accurate 1689 integration of single-cell data with Harmony

Web-based analysis and publication of 1691 flow cytometry experiments

Immunologic 1695 perturbations in severe COVID-19/SARS-CoV-2 infection

Studying the 1698 pathophysiology of coronavirus disease 2019: a protocol for the Berlin prospective COVID-1699 19 patient cohort (Pa-COVID-19)

Age and gender leucocytes 1702 variances and references values generated using the standardized ONE-Study protocol

Combinatorial Single-Cell Analyses of Granulocyte-Monocyte 1706

Progenitor Heterogeneity Reveals an Early Uni-potent Neutrophil Progenitor

Spleen-derived IFN-γ induces generation of PD-L1 + -suppressive neutrophils during 1710 endotoxemia

Immunophenotyping of COVID-19 and influenza highlights the role 1713 of type I interferons in development of severe COVID-19

Least-squares means: The R package lsmeans

PAD4 mediated histone hypercitrullination induces heterochromatin decondensation and 1718 chromatin unfolding to form neutrophil extracellular trap-like structures

Arginine deficiency is involved in thrombocytopenia and 1722 immunosuppression in severe fever with thrombocytopenia syndrome

Single-cell landscape of bronchoalveolar immune cells in patients with COVID-1726 19

The Molecular Signatures Database Hallmark Gene Set Collection

MCPIP1 Suppresses Hepatitis C Virus 1732 Replication and Negatively Regulates Virus-Induced Proinflammatory Cytokine Responses

Dysregulated myelopoiesis and 1735 hematopoietic function following acute physiologic insult

Antibody responses to SARS-CoV-2 in patients with COVID-1738 19

Longitudinal analyses reveal immunological misfiring 1741 in severe COVID-19

Ebola Virus Disease Is 1744 Characterized by Poor Activation and Reduced Levels of Circulating CD16+ Monocytes

Single cell RNA sequencing of human liver 1748 reveals distinct intrahepatic macrophage populations

DirichletReg: Dirichlet Regression for Compositional Data in R

Cutadapt removes adapter sequences from high-throughput sequencing 1752 reads

Validation of diagnostic gene sets to identify critically ill patients with sepsis

Deep immune profiling of 1758 COVID-19 patients reveals patient heterogeneity and distinct immunotypes with implications 1759 for therapeutic interventions

The Innate Immune System: Fighting on the Front 1761 Lines or Fanning the Flames of COVID-19?

Two new 1764 immature and dysfunctional neutrophil cell subsets define a predictive signature of sepsis 1765 useable in clinical practice

Barcoding of

Live Human Peripheral Blood Mononuclear Cells for Multiplexed Mass Cytometry

Platinum-conjugated antibodies for 1770 application in mass cytometry

The CD14+HLA-DrlO/NEG 1772 monocyte: An immunosuppressive phenotype that restrains responses to cancer 1773 immunotherapy

Pathological inflammation in patients with COVID-19: a 1775 key role for monocytes and macrophages

Ultra-high-throughput 1778 clinical proteomics reveals classifiers of COVID-19 infection

Frequencies of circulating MDSC 1782 correlate with clinical outcome of melanoma patients treated with ipilimumab

Neutrophil Extracellular Traps 1786 (NETs) Contribute to Immunothrombosis in COVID-19 Acute Respiratory Distress 1787 Syndrome

Persisting low monocyte human leukocyte 1790 antigen-DR expression predicts mortality in septic shock

Update on 1793 neutrophil function in severe inflammation

Different phenotypes of non-classical monocytes associated with 1797 systemic inflammation, endothelial alteration and hepatic compromise in patients with 1798 dengue

Heterogeneity of neutrophils

Detection of SARS-CoV-2-Specific Humoral and Cellular Immunity in COVID-1803 19 Convalescent Individuals

CyTOF workflow: differential discovery in high-1806 throughput high-dimensional cytometry datasets

A comprehensive single cell transcriptional landscape 1809 of human hematopoietic progenitors

Immunopathogenesis of coronavirus infections: 1811 Implications for SARS

Biological basis and pathological 1813 relevance of microvascular thrombosis

A subset of neutrophils in human 1816 systemic inflammation inhibits T cell responses through Mac-1

Decoding human fetal liver 1819 haematopoiesis

Extracting a cellular hierarchy from high-1822 dimensional cytometry data with SPADE

Mortality rates of patients with COVID-19 in 1824 the intensive care unit: A systematic review of the emerging literature

Immunotherapies for 1827 COVID-19: lessons learned from sepsis

An immune-cell signature of 1830 bacterial sepsis

DNAM-1 and PVR regulate monocyte 1833 migration through endothelial junctions

Human polymorphonuclear neutrophils express RANK and are activated by its 1836 ligand, RANKL

Immunosuppression for hyperinflammation in 1838 COVID-19: a double-edged sword?

Convergent antibody responses to 1841 SARS-CoV-2 in convalescent individuals

Differential Redistribution of Activated Monocyte and Dendritic Cell Subsets to the 1845 Lung Associates with Severity of COVID-19

Hepatic acute-phase proteins control innate immune 1849 responses during infection by promoting myeloid-derived suppressor cell function

Invariant NKT cells reduce the 1853 immunosuppressive activity of influenza A virus-induced myeloid-derived suppressor cells in 1854 mice and humans

Regulatory cell therapy in kidney 1857 transplantation (The ONE Study): a harmonised design and analysis of seven non-1858 randomised, single-arm, phase 1/2A trials

Human neutrophils in the 1860 saga of cellular heterogeneity: insights and open questions

Myeloperoxidase can differentiate between sepsis and non-infectious SIRS and predicts 1863 mortality in intensive care patients with SIRS

Emerging Principles in Myelopoiesis at 1866 Homeostasis and during Infection and Inflammation

Surface barcoding of live PBMC for multiplexed mass 1868 cytometry

Minimizing Batch Effects in Mass 1871

Production through Regulation of Autophagy and Is Associated with Adult Still Disease

Neutrophil 1877 Diversity in Health and Disease

Neutrophil Fluorescence: A New Indicator of Cell Activation 1880 during Septic Shock-Induced Disseminated Intravascular Coagulation

Neutrophil Activation during Septic Shock

Comprehensive Integration of Single-Cell 1886 Data

Human CD62Ldim neutrophils 1889 identified as a separate subset by proteome profiling and in vivo pulse-chase labeling

Interleukin-3 receptor in acute leukemia

Leukocyte Protease Binding to Nucleic Acids Promotes Nuclear 1895

Localization and Cleavage of Nucleic Acid Binding Proteins

Type I IFN immunoprofiling in COVID-1898 19 patients

Early expansion of circulating granulocytic 1901 myeloid-derived suppressor cells predicts development of nosocomial infections in patients 1902 with sepsis

Myeloid-derived suppressor cells coming 1904 of age review-article

Myeloid cells in sepsis-1906 acquired immunodeficiency

Expression of DNAM-1 1908 (CD226) on inflammatory monocytes

Kidney injury enhances renal G-1911 CSF expression and modulates granulopoiesis and human neutrophil CD177 in vivo

Clinical Characteristics of 138 Hospitalized Patients with

Coronavirus-Infected Pneumonia in Wuhan, China

Dysregulation of the immune response 1917 affects the outcome of critical COVID-19 patients

A single-cell atlas of the peripheral 1920 immune response in patients with severe COVID-19

Pathological findings of COVID-19 associated with acute respiratory distress 1926 syndrome

Increased formation of neutrophil extracellular traps is associated with gut 1929 leakage in patients with type 1 but not type 2 diabetes

Myeloid-derived suppressor cells: Their 1931 role in the pathophysiology of hematologic malignancies and potential as therapeutic targets

ClusterProfiler: An R package for 1934 comparing biological themes among gene clusters

Clinical course and risk factors for mortality of adult inpatients with COVID

China: a retrospective cohort study

Overly Exuberant Innate Immune Response to SARS-CoV-2 Infection. 1940 SSRN Electron

A dynamic immune response shapes COVID-19 1943 progression

Neutrophil extracellular traps in COVID-19

CoV-2 infection induces profound alterations of the myeloid compartment • Mild COVID-19 is marked by inflammatory HLA-DR hi CD11c hi CD14 + monocytes • Dysfunctional HLA-DR lo CD163 hi and HLA-DR lo S100A hi CD14 + monocytes in severe

• Emergency myelopoiesis with immature and dysfunctional neutrophils in severe COVID-19

Analysis of patients with with mild and severe COVID-19 reveals the presence of dysfunctional neutrophils in the latter that is linked to emergency myelopoiesis

In brief, the neutrophil space was subsetted to only severe patients (early and late) and only 1438 the most prominent clusters of the latter (clusters 1,2,6,8). The normalized data were scaled 1439 and regressed for UMIs and a diffusion map was calculated based on the top 2,000 variable 1440 genes with a sum of at least 10 counts over all cells. Based on the diffusion map, a diffusion 1441 pseudo time was calculated to infer a transition probability between the different cell states 1442 of the neutrophils. Subsequently, the density of the clusters along the pseudotime and 1443 marker gene expression for each cluster were visualized. 1444 Enrichment of gene sets was performed using the 'AUCell' method (Aibar et al., 2017 (Aibar et al., ) 1445 implemented in the package (version 1.4.1) in R. We set the threshold for the calculation of 1446 the AUC to the top 3% of the ranked genes and normalized the maximum possible AUC to 1. 1447The resulting AUC values were subsequently visualized in violin plots or UMAP plots.