key: cord-0944695-roao96p7
authors: Eddins, Devon J.; Yang, Junkai; Kosters, Astrid; Giacalone, Vincent D.; Pechuan, Ximo; Chandler, Joshua D.; Eum, Jinyoung; Babcock, Benjamin R.; Dobosh, Brian S.; Hernández, Mindy R.; Abdulkhader, Fathma; Collins, Genoah L.; Ramonell, Richard P.; Moussion, Christine; Orlova, Darya Y.; Sanz, Ignacio; Lee, F. Eun-Hyung; Tirouvanziam, Rabindra M.; Ghosn, Eliver E.B.
title: Pathogenic neutrophilia drives acute respiratory distress syndrome in severe COVID-19 patients
date: 2021-09-10
journal: bioRxiv
DOI: 10.1101/2021.06.02.446468
sha: b73c1c2fb8183f2306f751be78ae8fcadcece7b6
doc_id: 944695
cord_uid: roao96p7

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and the ensuing COVID-19 pandemic have caused ∼40 million cases and over 648,000 deaths in the United States alone. Troubling disparities in COVID-19-associated mortality emerged early, with nearly 70% of deaths confined to Black/African-American (AA) patients in some areas, yet targeted studies within this demographic are scant. Multi-omics single-cell analyses of immune profiles from airways and matching blood samples of Black/AA patients revealed low viral load, yet pronounced and persistent pulmonary neutrophilia with advanced features of cytokine release syndrome and acute respiratory distress syndrome (ARDS), including exacerbated production of IL-8, IL-1β, IL-6, and CCL3/4 along with elevated levels of neutrophil elastase and myeloperoxidase. Circulating S100A12+/IFITM2+ mature neutrophils are recruited via the IL-8/CXCR2 axis, which emerges as a potential therapeutic target to reduce pathogenic neutrophilia and constrain ARDS in severe COVID-19. Graphical Abstract The lung pathology due to severe COVID-19 is marked by a perpetual pathogenic neutrophilia, leading to acute respiratory distress syndrome (ARDS) even in the absence of viral burden. Circulating mature neutrophils are recruited to the airways via IL-8 (CXCL8)/CXCR2 chemotaxis. Recently migrated neutrophils further differentiate into a transcriptionally active and hyperinflammatory state, with an exacerbated expression of IL-8 (CXCL8), IL-1β (IL1B), CCL3, CCL4, neutrophil elastase (NE), and myeloperoxidase (MPO) activity. Airway neutrophils and recruited inflammatory monocytes further increase their production of IL-8 (CXCL8), perpetuating lung neutrophilia in a feedforward loop. MdCs and T cells produce IL-1β and TNF, driving neutrophils reprogramming and survival.

CXCL12 was not detected in the airways of patients, and there were no notable differences in plasma levels across patient groups (Extended Data Fig. 2 ). epithelial/stromal cells) with CXCR2-expressing neutrophils from the blood (blood cluster 2; Fig.   219 5a). The CXCL8 (IL-8)/CXCR2 pathway was identified as the primary recruitment axis for 220 circulating CXCR2 + neutrophils (Fig. 5b) . These data are in line with our finding that IL-8 was 221 increased in neutrophils by flow cytometry (Fig. 2 ) and secreted at very high levels in the 222 airways (Fig. 3) and suggests that the CXCL8/CXCR2 signaling axis is important for neutrophil 223 recruitment to the lungs during COVID-19 pathogenesis. Virtually all lung neutrophils 224 (particularly cluster 5), non-immune cells, and monocytes show the potential to recruit 225 circulating neutrophils (Fig. 5a,d) , indicating a robust and redundant mechanism of neutrophil 226 recruitment to the airways via the CXCL8/CXCR2 axis. Surprisingly, the immature neutrophils 227 (blood cluster 3) lacked CXCR2 (Fig. 5c) , suggesting that immature neutrophils are unlikely to 228 infiltrate the lung via IL-8. In contrast, a defined subset of mature neutrophils in blood 229 expressing high levels of CXCR2, along with interferon-induced IFITM2 and S100A11/12, 230 identify blood cluster 2 as the putative neutrophil subset that can infiltrate the lung via 231 4d and 5c-d). It is, therefore, probable that recent lung emigrants would still express detectable 232 levels of CXCR2, as well as IFITM2 and S100A11/12 (Fig. 5c ).

After identifying the blood neutrophil cluster 2 as the putative source of lung-recruited progressive increase in IFI30, CCL3/4, along with abundant de novo CXCL8 and CXCR4 mRNA transcripts, reveal a transcriptionally active state in neutrophils that is poised to sustain a 264 hyperinflammatory milieu in the lung of severe COVID-19 patients.

Viral load in the respiratory tract does not correlate with disease severity

We then investigated whether the recruited neutrophils or other cell types (i.e., myeloid, transcriptome in the multi-omics single-cell analysis, we were able to assess viral mRNA 272 (vRNA) transcripts at a single-cell level (Fig. 7a) . Notably, we did not detect vRNA in any cell 273 types in the blood or airways of severe patients (Fig. 7b,c) . Further, SARS-CoV-2-specific RT-274 qPCR revealed that viral load was decreased in the upper airways of severe patients admitted 275 to the ICU compared to mild-acute patients (Fig. 7d,e) . However, we did note a gene signature 

Pulmonary TNF and IL-1β promote neutrophil reprogramming in the lungs

To ascertain which ligand-receptor pairs are potentially responsible for the transcriptional 284 reprogramming of neutrophils in the lung, we first performed differential gene expression (DGE) 285 analysis between blood neutrophil cluster 2 and lung neutrophils to identify genes that are 286 upregulated in lung-recruited neutrophils (Fig. 6a ). Next, we used the computational method 287 NicheNet 38 to identify the potential (prioritized) ligands that could induce the upregulation of the 288 genes identified by the DGE analysis, indicating their potential role in neutrophil reprogramming 289 in the lung. TNF, IL1B, and APOE were the highest prioritized ligands with high ligand activity 290 whose signaling axes have the regulatory potential to drive gene expression profiles observed in 291 lung-recruited neutrophils (Fig. 6b) . Importantly, NicheNet analysis revealed TNF as the ligand 292 predicted to increase BCL2A1 expression in pulmonary neutrophils, as well as CCL4 (MIP-1β) 293 and CXCL16. Furthermore, both TNF and IL1B are the ligands predicted to induce NFKBIA and 294 CXCL8 (IL-8), and IL1B shows the greatest potential to induce CCL3 (MIP-1α) expression in 295 recruited neutrophils. APOE is predicted to upregulate the expression of FCER1G in neutrophils 296 in the lung. Notably, TNF, IL1B, and HMGB1 are the ligands that have the widest range of 297 regulatory potential, inducing neutrophil reprogramming by upregulating most genes that we 298 identified as differentially expressed (Fig. 6a ) in lung-recruited neutrophils (Fig 6b) .

Pulmonary T cells represent the primary cell population expressing TNF in the airways, where 301 MdCs have the highest IL1B and APOE transcripts (Fig. 6c,e) . From the transcriptome data, 302 ligand-receptor pair analyses further identified putative signaling mediators for the top 15 303 prioritized ligands (Fig. 6d) . TNF is predicted to signal through a receptor that has a CALM1 304 association (Fig. 6d) . Intriguingly, it has been previously reported that CALM1 can bind other 305 transmembrane proteins, including ACE-2 39 , and regulate their cell surface expression 40 . IL1B is 306 predicted to signal through the canonical IL1B/IL1R2 pathway, and APOE through the recruitment to the airways (Fig. 6e) .

Collectively, we show here that TNF-and IL-1β-mediated transcriptional reprogramming of lung 313 infiltrating neutrophils leads to induction of proinflammatory NFKBIA, CCL3 (MIP-1α), and CCL4 314 (MIP-1β) along with TNFAIP3/6 and CXCL8 (Figs. 5 and 6). Of note, neutrophil-derived CCL3 315 and CCL4 can attract inflammatory monocytes from the circulation to the airways. Interestingly, 316 lung-infiltrating monocytes also produce elevated levels of CXCL8, potentiating the recruitment 317 of circulating neutrophils via the CXCL8/CXCR2 axis (Fig. 5a,d) . Hence, both neutrophils and 318 inflammatory monocytes in the airways exacerbate and potentiate pathogenic neutrophilia in the 319 lungs of severe patients in a positive feedback loop.

Previous studies have reported increased neutrophilia in severe COVID-19, particularly in the 323 circulation 14, 44, 45, 46 . In contrast, reports on exacerbated airway neutrophilia and the implication of lung neutrophils as the main cell type driving ARDS in severe patients have yielded 325 inconclusive results 7, 8, 17, 47 . Here, we present the first comprehensive study of the lung immune highlighted some of the socio-economic and behavioral inequalities that may have contributed 329 to troubling disparities in COVID-19-associated morbidity and mortality 48 , with almost 70% of 330 deaths being Black/AA patients in some areas 1, 2 . Although the socio-economic and behavioral 331 differences indeed contribute to health disparities among demographics, a systematic 332 investigation to determine the immunological features that characterize disease severity within 333 Black/AA patients is lacking. Our systems biology approach addresses this knowledge gap and 334 reveals new therapeutic targets to inhibit neutrophil migration, retention, and/or survival in the 335 lung as potential effective interventions for individuals with severe disease that have been 336 disproportionally affected by COVID-19.

Although previous studies on severe COVID-19 in humans have not been conclusive with 

The heightened numbers of neutrophils in the lung are likely to induce and sustain inflammatory 350 signatures by an autocrine/paracrine feedback loop among neutrophils, and paracrine signaling 351 to other cell types that can potentiate disease severity. Notably, NE is a potent serine protease 352 that we show is released from degranulating neutrophils in the lungs (see Fig. 2 ), and has 353 potential to stimulate production of TNF, IL-1β, and IL-8 52, 53 , and also abrogate protective 354 effector functions of T cells and MdCs in the lung by cleaving cell surface receptors such as part, the reduction of FcγRII (CD32) expression on pulmonary neutrophils (see Extended Data 358 Fig. 5a ). Of note, the reduction in CD32 expression may prevent IgG-mediated suppression of 359 ISG induction 14 in pulmonary neutrophils. Indeed, we demonstrate a pronounced signature for 360 the ISG IFI30 (see Fig. 5f ). Similarly, we show by intracellular staining that pulmonary 361 neutrophils are producing exacerbated levels of IL-1β protein that is likely released upon progression. For example, we and others 16 have shown that infiltrating neutrophils sustain local production of calgranulins (S100A8/9/11/12), which can signal through and activate myeloid cells via TLRs and RAGE receptors, compounding the already hyperinflammatory lung milieu. pathogenesis in severe patients. Non-immune cells (e.g., stromal and epithelial cells) are known 373 targets of SARS-CoV-2 in the lung 56 . Therefore, these cells are also posited to influence 

Interestingly, CXCR4 is upregulated in recruited neutrophils in severe patients (Figs. 4d,i and 390 5f), which may promote neutrophil survival and retention during pneumonitis and/or further 391 influence inflammatory neutrophil phenotype. Accordingly, CXCR4 was shown to promote 392 transcriptional reprogramming of neutrophils in pulmonary tissues 64 . Although we do not find 393 detectable levels of the canonical CXCR4 ligand (i.e., CXCL12) in the airways, we did observe 394 transcripts for HMGB1 (Fig. 6) , which is an alternative ligand for CXCR4. Since the IL-8/CXCR2 395 pathway is most notably increased and likely the primary neutrophil recruitment axis to the 396 airways, we speculate that CXCR4 signaling instead may promote neutrophil survival and 397 retention at the site of inflammation, which has been previously reported 65 . Additionally, CXCR4 398 signaling can stimulate de novo CXCL8/IL-8 production 66, 67 , and has been shown to promote 399 neutrophil extracellular trap release during malaria disease progression 68 . Alternatively, CXCR4 400 is also associated with neutrophil aging and senescence 69, 70, 71 . As such, elevated CXCR4 may 401 be associated with prolonged neutrophil survival in COVID-19 pathogenesis. In a prior study 72 , 

We also contend that recognizing the lung pathology in severe COVID-19 to be a neutrophilic 431 and hyperinflammatory disease is paramount to achieve better outcomes in next-generation 432 therapies. Although the lung pathology in COVID-19 is initiated by a viral infection, severe 433 patients in the ICU no longer show signs of uncontrolled viral replication. In fact, not only did we 434 not detect viral transcripts by scRNA-seq within the cells in the airways, but we also noted 435 decreased viral burden in severe patients in the ICU versus mild-acute patients seen in the 436 outpatient clinic (Fig. 7) . This is further supported by our previous study where we performed 437 plaque assays on the respiratory secretions from severe patients and revealed significantly 438 diminished, if any, viral plaques from the endotracheal aspirates 76 . This may explain, in part, known to date to employ integrated multi-omics single-cell investigation of immunity in Tables 1 and 2 

Splicing-aware aligner STAR 80 was implemented to align FASTQ inputs to the reference 558 genome, and the resulting files are automatically filtered by CellRanger to include only cell 559 barcodes representing real cells. This determination is based on the distribution of UMI counts.

ADT reads were aligned to a feature reference file containing the antibody-specific barcode 561 sequences. To recover neutrophils, we applied our SuPERR-seq pipeline as previously 562 described 30 . Briefly, we recovered neutrophils from CellRanger unfiltered count matrices by by GEX to confirm viable neutrophil identity. A threshold for mitochondrial content per barcode was determined for each sample independently and applied as a cutoff to remove dead or dying 567 cells (Extended Data Table 7 ). Most samples show high cell viability with a minimal proportion of 568 dead cells.

The UMI counts of the GEX data were log-normalized by the "NormalizeData" function in Seurat 571 before downstream analysis, following the optimal workflow we previously described for sample 572 normalization and data integration 82 . Center log-ratio (CLR) transform in Seurat was performed 573 on ADT UMIs when recovering neutrophils from the unfiltered matrices. For surface protein 574 visualization to classify major lineages using our SuPERR-seq workflow 30 , ADT UMIs were 575 normalized using the R package Denoised and Scaled by Background 83 (DSB) to remove 576 ambient UMI counts (i.e., background) prior to manual sequential gating by surface expression 577 (Extended Data Fig. 4) in SeqGeq v1.7 (FlowJo, LLC). DSB uses empty droplets to calculate 578 background expression, which was manually selected according to the distribution of total ADT 579 per cell in the raw count matrices (Extended Data Table 8 ). To minimize the influence from non-580 informative empty droplets, we removed cell barcodes with less than 100 total ADT UMIs before 581 plotting the ADT distribution.

Before integrating the multiple datasets, we first classified major lineages in individual samples 584 based on a combination of gene transcript and surface protein markers (SuPERR-seq 585 workflow 30 ) as in Fig. 4 for samples where the ADT library was of sufficient quality to allow 586 manual gating (Extended Data Fig. 4) . Cell barcodes within each major lineage that co-587 expressed markers exclusive to other major lineages were considered cell doublets and 588 removed (Extended Data Fig. 4 ). In addition, we removed cell barcodes with extremely high 589 total ADT UMIs, which we considered to be aggregated cells. To efficiently integrate replicate 590 samples, we concatenated major lineages derived from the same tissue in different donors. To 591 minimize batch effects and optimize data integration, we followed the data normalization and 592 merging strategies described previously 82 . Briefly, samples were first treated individually, and 593 log-normalized count matrices were scaled/Z-transformed, and the "vst" method of the Seurat genes was determined using Seurat::FindMarkers(min.pct = 0.1), keeping only those genes with extracted from the respiratory secretions of COVID-19 patients using the Quick-RNA™ Viral Kit using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems™) per the 657 manufacturer's instructions, then diluted 1:5 in nuclease-free water. 10 µL diluted cDNA was

Formal Analysis. Mindy R. Hernández: Resources. Fathma Abdulkhader: Investigation.

School Fellowship, and RPR was supported by the NIH T32-HL116271-07 Fellowship. The 722 graphical abstract and Fig. 7d were generated in part using BioRender. We thank Keivan Zandi, 

Racial and ethnic disparities in COVID-19 incidence by age, sex, 16. Silvin, A. et al. Elevated Calprotectin and Abnormal Myeloid Cell Subsets Discriminate 790 Severe from Mild COVID-19

Endotracheal aspirates contain a limited number of lower respiratory tract Concentration (pg/mL) of 15 analytes interrogated by Mesoscale analyses in plasma (gray 1069 circles) and respiratory supernatant (Resp. SNT; green squares) from healthy control (HC), 1070 mild-acute (MA), and severe COVID-19 patients. (b, e, h) Representative flow cytometric 1071 intracellular staining for IL-8 (CXCL8), IL-1β, and IL-6, including the full stain and fluorescence 1072 minus one (FMO) controls in both blood and ETA neutrophils (CD66b + ). (c, f, i) UMAP 1073 visualizations of CXCL8 (IL-8), IL1B, and IL6, which were also measured by intracellular flow 1074 cytometry staining

1076 black dotted line = median lower limit of detection (LLOD) for assays (see Extended Data Table Extended Data Figure 4. Gene signature for immature neutrophils is lacking in healthy 1222 donors and lungs of COVID-19 patients. UMAP visualizations of genes that identify immature 1223 neutrophils in the blood healthy individuals (a) and severe COVID-19 patients (b)