key: cord-0701299-fk0ri80v
authors: Sidarta-Oliveira, D.; Poblete Jara, C.; Ferruzzi, A. J.; Skaf, M. S.; Velander, W. H.; Araujo, E. P.; Velloso, L. A.
title: Assembly of an integrated human lung cell atlas reveals that SARS-CoV-2 receptor is co-expressed with key elements of the kinin-kallikrein, renin-angiotensin and coagulation systems in alveolar cells
date: 2020-06-04
journal: nan
DOI: 10.1101/2020.06.02.20120634
sha: 34d847dd92d06c3d409d1d392cc5fc52db35b807
doc_id: 701299
cord_uid: fk0ri80v

SARS-CoV-2, the pathogenic agent of COVID-19, employs angiotensin converting enzyme-2 (ACE2) as its cell entry receptor. Clinical data reveal that in severe COVID-19, SARS-CoV-2 infects the lung, leading to a frequently lethal triad of respiratory insufficiency, acute cardiovascular failure, and coagulopathy. Physiologically, ACE2 plays a role in the regulation of three systems that could potentially be involved in the pathogenesis of severe COVID-19: the kinin-kallikrein system, resulting in acute lung inflammatory edema; the renin-angiotensin system, promoting cardiovascular instability; and the coagulation system, leading to thromboembolism. Here we analyzed ~130,000 human lung single-cell transcriptomes and show that key elements of the kinin-kallikrein, renin-angiotensin and coagulation systems are co-expressed with ACE2 in alveolar cells, which could explain how changes in ACE2 promoted by SARS-CoV-2 cell entry result in the development of the three most severe clinical components of COVID-19.

COVID-19, the disease caused by SARS-CoV-2 infection, frequently opens with cough, fever, fatigue, and myalgia 1 , progressing to a severe illness in up to 20% of infected patients 2 . Severe COVID-19 is characterized by progressive dyspnea, which results from acute lung inflammatory edema leading to hypoxia 3 . In patients surviving the initial lung inflammatory burst, a number of other complications can sum to promote a rapid and frequently lethal deterioration of health [3] [4] [5] . Acute cardiovascular failure and coagulopathy are among the most frequent complications and could be placed alongside with acute respiratory failure 3-7 as components of a triad that leads to the highest death rates in COVID-19 (Fig. 1A) .

Angiotensin converting enzyme 2 (ACE2) is the receptor for SARS-CoV-2 8 . Binding occurs through viral spike protein 9 and depends on the serine protease TMPRSS2 for priming 8 . Physiologically, ACE2 is involved in the control of three independent but highly integrated systems, the kinin-kallikrein (KKS), renin-angiotensin (RAS), and coagulation (CS) systems (Fig. 1B) . Bradykinin is a potent inflammatory substance produced from high-molecular weight kininogen (HMWK) in a reaction catalyzed by the serine protease kallikrein 10 . Bradykinin can directly deliver its vasoactive and inflammatory actions through bradykinin receptor 2 or be further processed by carboxypeptidase N to form DR9-bradykinin that activates bradykinin receptor 1 to deliver inflammatory and pain signals 11 . ACE2 degrades DR9-bradykinin into inactive peptides and, together with angiotensin converting enzyme (ACE), which inactivates bradykinin, shuts down the KKS 12 . Angiotensin II (Ang II) is a pleotropic hormone involved in the regulation of blood pressure, blood volume, cardiac function, and electrolyte balance 13, 14 . It is produced from angiotensin I (Ang I) through the catalytic action of ACE, whereas ACE2 catalyzes its degradation into the inactive peptide angiotensin 1-7, thereby inactivating the RAS 15 .

The interaction of ACE2 with the CS is indirect and occurs via two mechanisms: 1) catalyzing the production of angiotensin 1-9, which reduces plasminogen activator and increases PAI-1, thus inhibiting fibrinolysis 16 and 2) modulating the activity of kallikrein, which in turn catalyzes the conversion of plasminogen into plasmin 17 .

Upon SARS-CoV-2 binding, ACE2 is internalized to endosomes, leading to a subcellular location shift that could alter its capacity to physiologically regulate the KKS, RAS and CS [18] [19] [20] [21] [22] . This could simultaneously impact the highly lethal COVID-19 triad: lung inflammation, cardiovascular failure, and coagulopathy. Despite the fact that ACE2 is expressed in several organs and tissues, both clinical and experimental evidence shows that SARS-CoV-2 promotes most of its pathological actions by initially infecting cells of the upper respiratory tract and, subsequently, alveolar cells in the lung [23] [24] [25] . It is currently unknown if lung cells expressing ACE2 are equipped with proteins that belong to the KKS, RAS and CS, which could potentially explain a cell-autonomous system that is disturbed by SARS-CoV-2 infection, leading to abnormal regulation of all three systems.

Here, we evaluated the transcripts of 129,079 human lung cells previously submitted to single-cell RNA sequencing. We show that transcripts encoding for key elements of all three systems are highly co-expressed with ACE2 in alveolar cells.

ACE2 interacts with proteins belonging to the kinin-kallikrein, renin-angiotensin, and coagulation systems. A protein-protein interaction network (Fig. 1C ) revealed that ACE2 interacts closely with proteins that belong to the KKS; kininogen (KNG1), the substrate for bradykinin synthesis, and kallikrein (KLKB1), the enzyme that catalyzes this conversion. ACE2 also interacts with proteins of the RAS; ACE, that converts Ang I into Ang II, renin (REN), the enzyme that converts angiotensinogen (AGT) into Ang I, AGT itself and angiotensin receptor 1 (AGTR1). The interface of ACE2 with CS occurs mostly though KLKB1 that controls fibrinolysis; in addition, Factor II (F2, thrombin), was identified in the interactome.

Integrated analysis of 129,079 human single lung cells leads to the identification of cell types expressing ACE2. To investigate the lung cell types that are potentially targeted by SARS-Cov-2 due to ACE2 expression, we leveraged public single-cell RNA sequencing (scRNAseq) from three previously published datasets and a pre-print report [26] [27] [28] (https://doi.org/10.1101/742320). Despite active investigation of the cellular landscape of the human lung, the field lacks an integrated atlas in which a consensus can be established between various datasets. We addressed this issue by individually filtering and integrating each study control sample into a batch-corrected study-wise reference with Seurat v3 anchor-based integration ( Fig. 2A, Suppl. Fig. 1 , Suppl. Fig.   2A 2B) . In this embedding, each point represents a single-cell, and the position in the embedding represents its relative transcriptional identity when compared with other cells.

Grouping cells by study of origin shows that this approach is successful in removing experimental noise from data (Suppl. Fig. 2B ). Cells were clustered by applying the Louvain algorithm to the graph of each cell diffusion structure scoring . After removal of   singletons and doublets, 47 cell clusters were identified, comprising eight major groups:   alveolar cells, endothelial and lymphatic cells, monocytes/macrophages, T cells, B cells, fibroblasts, smooth muscle cells, and mast cells (Fig. 2B) . These clusters were annotated based on learned annotations published elsewhere (https://doi.org/10.1101/742320) (Suppl. Fig. 2C ) and on cluster marker gene expression (Suppl. Fig. 3 ). The clusters effectively correspond to biologically comprehensive cellular states that can be explored as a resource to COVID-19 studies. This potential is leveraged by dbMAP for the identification of rare cell types and cellular-state transitions in comparison against UMAP (Suppl. Fig. 2D ). An illustrative example is shown by B cell differentiation into plasma cells. In the dbMAP embedding, B and plasma cells form discrete populations connected directly by intermediate cells (Fig. 2B ), whereas UMAP represents these clusters as completely separated populations (Suppl. Fig. 2D ). We briefly explored cluster marker genes (Suppl. Figs. 4 and 5), albeit the exploration of this atlas exceeds the scope of this manuscript. An interactive browser with our data is available at https://humanlung.iqm.unicamp.br, allowing non-bioinformaticians to produce publication-level plots and tables in a community effort to further explore the human lung cellular landscape. We next investigated the expression of ACE2 We then investigated the expression of ACE2 and show that it is restricted to alveolar cells and fibroblasts, being practically absent from other cell clusters (Fig. 1C ).

TMPRSS2 is co-expressed with ACE2 in alveolar cells. The serine protease TMPRSS2 is required for SARS-CoV-2 spike protein priming and subsequent binding to ACE2 8 (Fig. 3A) . Following receptor binding, SARS-CoV-2 is internalized through endocytosis in a process that depends on the activity of phosphatidylinositol 3-phosphate 5-kinase (PIKFYVE) and its downstream effector, two-pore channel subtype 2 (TPCN2) 18 (Fig. 3A) . Moreover, the inhibition of cathepsin L (CTSL) dramatically reduces virus entry, suggesting a role for this lysosome protein in the process 18 (Fig. 3A) . Out of the four proteins currently described as players in SARS-CoV-2 cell entry, only TMPRSS2 gene expression selectively overlapped with ACE2 expression (Fig. 3B , 3C, and 3G). is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint 3E and 3G). CTSL is also expressed in virtually all lung-cell clusters. However, here there is a large predominance in macrophage clusters ( Fig. 3F and 3G ).

Kininogen is co-expressed with ACE2 in alveolar cells. Kininogen (KNG1), the precursor for bradykinin synthesis, is expressed in a large number of cells that also express ACE2 (Fig. 4A1, 4A6 , and 4B). KNG1 expression is restricted to alveolar type 1.1, type 1/2, and type 2.4, in addition to alveolar 2.2. and 2.3 clusters. Kallikrein (KLKB1) is a serine protease that catalyzes the conversion of kininogen into bradykinin. It is expressed predominantly in endothelial and lymph vessels cells ( Fig. 4A2 and 4B ). Bradykinin can either be converted into inactive kinins by the catalytic action of ACE or be converted into the active DR9-bradykinin. ACE is expressed predominantly in macrophages, monocytes, and endothelial cells ( Fig. 4A4 and 4B ). It is also expressed in scattered alveolar cells, T-lymphocytes, and fibroblasts ( Fig. 4A4 and 4B ). DR9bradykinin acts through bradykinin receptor 1 (BDRK1), which is predominantly expressed in fibroblasts, endothelial cells, and alveolar type 2.1, 2.2, and 2.3 cells ( Fig.   4A5 and 4B).

Angiotensinogen (AGT) is the precursor for Ang I; we show that most cells expressing AGT are fibroblasts and smooth muscle cells; however, a considerable number of type 2.4, 2.3, and 2.1 alveolar cells that express ACE2 also express AGT (Fig. 5A1 , 5A2, and 5B). Renin is the enzyme that converts AGT into Ang I; here, we show that in the lung it is predominantly expressed in alveolar cells and largely co-expressed with ACE2, particularly in alveolar cell types 1.2, 2.2, 2.3, 2.4, and NA 2 ( Fig. 5A1 , 5A3, and 5B).

Some macrophages and mast cells also express renin ( Fig. 5A3 and 5B ), which is virtually absent in the remaining clusters. Ang I is converted into active Ang II by ACE ( Fig. 5A4 and 5B) , which is largely expressed in macrophages and endothelial cells. Ang II exerts most of its cardiovascular effects by acting through angiotensin II receptor 1 (AGTR1). We show that most cells expressing AGTR1 are fibroblasts and muscle cells ( Fig. 5A5 and 5B ); in addition, a considerable number of alveolar type 1.1 and NA 2 cells that express ACE2 also express AGTR1 (Fig. 5A1 , 5A5, and 5B).

Fibrinogen gamma is co-expressed with ACE2 in alveolar cells. In addition to its action on the KKS, kallikrein (KLKB1) acts in the CS by converting plasminogen into the . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 4, 2020. . fibrinolytic substance plasmin; KLKB1 is expressed predominantly in endothelial and lymph vessel cells and in alveolar type 1.1, type 1/2, and type 2.4 cells ( Fig. 6A1 and   6B ). Another catalyst of plasmin production is tissue plasminogen activator (PLAT), which is inhibited by PAI-1 (SERPINE1). SERPINE1 is expressed predominantly by endothelial cells and secondarily by fibroblasts and muscle cells ( Fig. 6A2 and 6B ), whereas PLAT is expressed in endothelial and lymph vessel cells ( Fig. 6A3 and 6B ). 

Here, we present the largest single-cell analysis of the human lung cellular transcriptional landscape to date. This was achieved by assembling datasets provided by previous independent studies into an integrated human lung cell atlas containing 129,079 cells [26] [27] [28] . We further leveraged the power of this data by analyzing it with dbMAP, a novel dimensional reduction method that, together with the integrated dataset, are publicly available at https://github.com/davisidarta/humanlung and at https://github.com/davisidarta/dbMAP. An user-friendly interactive web-based platform is also available at https://humanlung.iqm.unicamp.br, in a powerful data exploration environment that holds potential to accelerate lung research. This data can also be used by future studies performing scRNAseq and data analysis of the human lung, so that others may have the option to add their study into this integrated atlas. Using this approach, we confirmed previous data that identified alveolar cells as those expressing highest levels of the SARS-CoV-2 receptor, ACE2, in the lung 23 . In addition, we identified cell types that express transcripts encoding for proteins potentially involved in SARS-CoV-2 cell entry. We demonstrated that expression of TMPRSS2, a serine protease that primes SARS-CoV-2 spike protein, largely overlaps with ACE2 expression in alveolar cells 23 , reinforcing the hypothesis of TMPRSS2 as a promising pharmacological target of COVID-19 29 . Moreover, we showed that transcripts encoding for PIKFYVE, TPCN2, and CTSL 18 are ubiquitously expressed in the lung and, even though they are expressed in alveolar cells with some degree of overlap with ACE2, their potential as therapeutic targets could be challenged due to lack of cell specificity.

Next, we determined the identities of lung cells expressing key components of the KKS, RAS, and CS pathways. Currently, there is neither experimental nor clinical evidence . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint suggesting that SARS-CoV-2 infection could lead to bradykinin-dependent lung inflammation and edema. However, in a rodent model of lung LPS-induced inflammation, the inhibition of ACE2 resulted in a significant increase lung inflammatory edema, and the simultaneous inhibition of bradykinin in this model dramatically reduced inflammatory damage 12 . In addition, in humans, severe lung edema can develop in a rare genetic disorder, hereditary angioedema, that results from bradykinin accumulation due to mutations of the C1 esterase/kallikrein inhibitor 30 . Clinical studies have shown that inhibitors of bradykinin can efficiently treat lung edema in these circumstances 31 .

Furthermore, the use of ACE inhibitors provides yet another line of clinical evidence for disturbed lung accumulation of bradykinin that can, on rare occasions, lead to severe outcomes 32, 33 . Here, we showed that kininogen, the precursor of bradykinin, presents considerable cellular expression overlap with ACE2 and that transcripts encoding for other components of the KKS are also expressed in lesser amounts in ACE2-expressing cells.

In contrast to KKS, abnormalities in the RAS have been widely reported in COVID-19 patients [34] [35] [36] . Obesity, hypertension, diabetes, and coronary insufficiency represent the greatest risk factors for severe COVID-19 5, 6 , and in all these diseases, there is an increased risk for abnormal regulation of the RAS 13, 37, 38 . In one of the largest series reporting COVID-19 patients published so far, the use of ACE inhibitors before infection was shown to reduce mortality by 33%, representing the most effective independent factor, among those analyzed, that could protect patients from a lethal outcome 6 . Here we showed that angiotensinogen, the precursor of Ang 1, and particularly renin, the enzyme catalyzing this conversion, are expressed in alveolar cells and that their expression overlaps that of ACE2. In addition, other key components of the RAS, ACE, and angiotensin receptor 1, are also co-expressed with ACE2 in a considerable number of cells.

COVID-19-associated coagulopathy can lead to a fulminant activation of coagulation, resulting in widespread thrombosis. Venous thromboembolism (VTE) is one of the leading causes of severe complications in COVID-19 patients 3, 39 . It was diagnosed in 20% of patients admitted to an intensive care unit (ICU) and its cumulative incidence increased progressively to 42% as patients remained in severe condition in the ICU 39 .

Developing VTE during the progression of severe COVID-19 increases the risk of death by 140% 39 . Moreover, D-dimer, a degradation product of fibrin, has been identified as an important predictive marker of severe COVID-19 and was directly correlated with poor prognosis 40, 41 . Because of the association between disturbed coagulation and severe disease progression, the use of anticoagulant treatment has been proposed as . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint potentially beneficial in COVID-19 22, 42 . In one therapeutic clinical trial, the use of lowmolecular-weight heparin resulted in reduced mortality in patients with high levels of Ddimer 43 ; however, in another study, preventive anticoagulation was associated with increased VTE 44 . These results suggest that prophylactic and therapeutic anticoagulation approaches have distinct outcomes in COVID-19; however, further studies are awaited in order to direct the establishment of optimal measures that could lessen the severity of COVID-19 coagulation abnormalities. Here, we showed that particularly FGG and KLKB1 are expressed in alveolar cells that also express ACE2.

Both FGG and KLKB1 play central roles in the control of the CS. FGG deficiency leads to mild or even severe bleeding 45 , whereas KLKB1 deficiency is associated with an increased risk of thrombosis 46 .

In conclusion, alveolar cells expressing ACE2, which are primary cellular targets for SARS-CoV-2 in the lung, also express transcripts encoding proteins that play pivotal roles in the regulation of the KKS, RAS, and CS. As all these systems are potentially affected during the progression of severe COVID-19, we propose that abnormal function of ACE2 as a result of SARS-CoV-2 infection could directly and cell-autonomously precipitate the development of acute lung inflammation, cardiovascular failure and thromboembolism, which are the hallmarks of severe COVID-19 46 .

Computational Environment. In silico analyses were performed on two different machines. Locally, a ThinkPad P52 Workstation with 128GB RAM and a six-core Intel Xeon processor was used for data exploration. A high-performance computing cluster (HPCC) was used for data integration and computation of results. Analysis was performed in R version 3.6.2. A docker environment with a pre-installed RStudio image and loaded with all required packages is available at https://github.com/davisidarta/humanlung. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint containing the balanced expression of the union of each sample's expressed genes, resulting in three study-wise balanced datasets.

performed the union of the gene expression of available human lung peer-reviewed studies [26] [27] [28] . Prior to anchoring, 3,000 features were selected by ranking top features identified as highly variable in multiple studies. We proceeded similarly to the individual study level, where anchors were built between multiple samples, and searched for anchor correspondences between each study's own manifold by using Seurat v3

Integration. As a first step, dimensional reduction with CCA is performed. CCA effectively captures gene modules that are correlated and shared in pair-wise dataset comparisons, representing signals that define a shared biological state. Afterwards, mutual nearest neighbors (MNNpairs of cells, each from one dataset) that share a corresponding state are weighted in an 'anchor', which is scored so as to exclude spurious connections between unrelated cell types. The integration of these anchors into a single manifold was performed with the IntegrateData function, and the top 20,000 genes with higher dispersion were included in the final result. The data of Travaglini et al. were not included in the integrated atlas due to its lack of peer review.

However, we did use these data annotations for label transfer due to its high-quality cell-type annotation. This was performed by employing the FindTransferAnchor and TransferData functions, in which a classification matrix is transposed by multiplication by a weighting matrix to return prediction scores for each class for every cell in the atlas. Those labels were further used as guidance in the process of cell-type annotation, as well as the expression of cell-type marker genes.

Projection (dbMAP). Visualizing single-cell data is a challenging task in which comprehensive two-or three-dimensional embedding needs to be generated from an exceptionally large number of samples and observations. Previous approaches mainly relied on performing prior dimensionality reduction with PCA to denoise data and make it computationally easier to compute non-linear dimensional reduction methods such as t-stochastic neighborhood embedding (t-SNE), UMAP, and potential of heat diffusion for affinity-based trajectory embedding (PHATE). We have recently proposed diffusionbased Manifold Approximation and Projection (dbMAP), which excels at identifying rare cell populations and describing lineage dynamics as trajectories progress, branch, and cycle (http://dx.doi.org/10.2139/ssrn.3582067). Briefly, dbMAP encodes and denoises data by dissecting its diffusion structure, in which cell-cell similarity information is adaptive regarding each cell's individual neighborhood density. This information is then . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint propagated through a series of random walks, which are scaled to normalize the relative progression of diffusion during each walk. This approach effectively adapts diffusion maps to approximate the Laplace-Beltrami operator, which represents data intrinsic structure. The resulting eigenvectors are diffusion components which can be used for downstream clustering, layout visualization and pseudotime estimation, being a potential substitution for PCA in single-cell processing. These diffusion components address UMAP's assumptions on its input, thus optimizing its potential for visualization, a feature we explored with dbMAP. dbMAP is computationally scalable and robust to Importantly, visualization parametrization can be fine-tuned by the user for its specific dataset due to the fast UMAP layout computation of the structure components, although results overall are robust to small changes in these parameters. Parameters used for dbMAP embedding for individual studies and the integrated atlas, as well as those used for UMAP embedding of the atlas, are listed in Supplementary Table 3 .

Clustering. Clustering was performed by generating a k-nearest-neighbors graph from the structure components learned in dbMAP first step. For this, we applied the FindNeighbors function in Seurat with default parameters on the structure components. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint Doublet identification and removal. We identified five clusters that presented without specific marker genes and in spurious trajectories as doublets and removed them.

Three clusters scattered the white space between alveolar cells, monocytes/macrophages, and lymphocytes in pairwise trajectories between these three cell types, effectively corresponding to doublets. The two remaining doublet clusters were scattered between mast cells and monocytes. We provide code for the identification and removal of these doublet clusters. Table 4 ).

Code Availability. All code used for analysis of single-cell RNA sequencing data is available at https://github.com/davisidarta/humanlung. dbMAP is available at https://github.com/davisidarta/dbMAP as a python library, which can also be easily used within R with the reticulate package. A docker image containing all necessary packages for analysis with an RStudio interface is also available.

10,000 randomly sampled cells is available as a Cerebro 47 interactive webpage which can be easily explored by non-bioinformaticians at https://humanlung.iqm.unicamp.br .

In Cerebro, users can readily visualize gene expression with dbMAP embeddings, search for each cluster differentially expressed genes, obtain functional enrichment scores for clusters of interest from a wide range of biomedical databases and export publication-level plots and tables. Fully processed data is also available as a complete Seurat object (.Rds) and as a Cerebro (.crb) file. All further data is available from the authors upon request. Table with is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . https://doi.org/10.1101/2020.06.02.20120634 doi: medRxiv preprint authors contributed in the interpretation of results and provided suggestions for data analysis. All authors wrote the manuscript and approved the final submitted version.

The authors declare no conflicts of interest. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 4, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 4, 2020. Fig. 6 < 5

Severe Covid-19

Clinical course and mortality risk of severe COVID-19

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the

Clinical Characteristics of Coronavirus Disease 2019 in China

Cardiovascular Disease, Drug Therapy, and Mortality in Covid-19

Evaluation of coagulation function by rotation thromboelastometry in critically ill patients with severe COVID-19 pneumonia

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Structural basis of receptor recognition by SARS-CoV-2

Hyperfibrinolysis increases blood-brain barrier permeability by a plasmin-and bradykinin-dependent mechanism

Sensory and signaling mechanisms of bradykinin, eicosanoids, platelet-activating factor, and nitric oxide in peripheral nociceptors

Attenuation of pulmonary ACE2 activity impairs inactivation of des-Arg(9) bradykinin/BKB1R axis and facilitates LPS-induced neutrophil infiltration

Renin-angiotensin system and cardiovascular risk

Cross-talk between the insulin and angiotensin signaling systems

Angiotensin-converting enzyme 2--a new cardiac regulator

Angiotensin-(1-9) enhances stasis-induced venous thrombosis in the rat because of the impairment of fibrinolysis

Plasma kallikrein cleaves and inactivates apelin-17: Palmitoyl-and PEGextended apelin-17 analogs as metabolically stable blood pressure-lowering agents

Characterization of spike glycoprotein of SARS-CoV-2 on virus entry and its immune cross-reactivity with SARS-CoV

Cell entry mechanisms of SARS-CoV-2

The renin-angiotensin system -a therapeutic target in COVID-19

Kallikrein-kinin blockade in patients with COVID-19 to prevent acute respiratory distress syndrome

COVID-19 and its implications for thrombosis and anticoagulation

SARS-CoV-2 Receptor ACE2 Is an Interferon-Stimulated Gene in Human Airway Epithelial Cells and Is Detected in Specific Cell Subsets across Tissues

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

SARS-CoV-2 entry factors are highly expressed in nasal epithelial cells together with innate immune genes

Single-Cell Transcriptomic Analysis of Human Lung Provides Insights into the Pathobiology of Pulmonary Fibrosis

Construction of a human cell landscape at single-cell level

scRNA-seq assessment of the human lung, spleen, and esophagus tissue stability after cold preservation

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

Therapeutic approaches in hereditary angioedema

Angioedema from angiotensin-converting enzyme (ACE) inhibitor treated with complement 1 (C1) inhibitor concentrate

Drug-induced pulmonary disease

Renin-Angiotensin-Aldosterone System Blockers and the Risk of Covid-19

Renin-Angiotensin-Aldosterone System Inhibitors and Risk of Covid-19

Hypertension, renin-angiotensin-aldosterone system inhibition, and COVID-19

Diabetes and hypertension: the bad companions

The Link Between Adipose Tissue Renin-Angiotensin-Aldosterone System Signaling and Obesity-Associated Hypertension

Incidence of venous thromboembolism in hospitalized patients with COVID-19

Dynamic relationship between D-dimer and COVID-19 severity

Monoclonal antibodies with equal specificity to D-dimer and highmolecular-weight fibrin degradation products

Thromboembolic risk and anticoagulant therapy in COVID-19 patients: emerging evidence and call for action

Anticoagulant treatment is associated with decreased mortality in severe coronavirus disease 2019 patients with coagulopathy

High incidence of venous thromboembolic events in anticoagulated severe COVID-19 patients

Clinical and molecular characterisation of 21 patients affected by quantitative fibrinogen deficiency

Genetic variants associated with deep vein thrombosis: the F11 locus

STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets

Comprehensive Integration of Single-Cell Data

Cerebro: interactive visualization of scRNA-seq data