key: cord-0769595-6199zzce
authors: Bolouri, Hamid; Speake, Cate; Skibinski, David; Long, S. Alice; Hocking, Anne M.; Campbell, Daniel J.; Hamerman, Jessica A.; Malhotra, Uma; Buckner, Jane H.
title: The COVID-19 immune landscape is dynamically and reversibly correlated with disease severity
date: 2020-09-21
journal: bioRxiv
DOI: 10.1101/2020.09.18.303420
sha: ae05e2edadc215b6b7f0e1edc5aeb51936af459e
doc_id: 769595
cord_uid: 6199zzce

Despite a rapidly growing body of literature on COVID-19, our understanding of the immune correlates of disease severity, course and outcome remains poor. Using mass cytometry, we assessed the immune landscape in longitudinal whole blood specimens from 59 patients presenting with acute COVID-19, and classified based on maximal disease severity. Hospitalized patients negative for SARS-CoV-2 were used as controls. We found that the immune landscape in COVID-19 forms three dominant clusters, which correlate with disease severity. Longitudinal analysis identified a pattern of productive innate and adaptive immune responses in individuals who have a moderate disease course, whereas those with severe disease have features suggestive of a protracted and dysregulated immune response. Further, we identified coordinate immune alterations accompanying clinical improvement and decline that were also seen in patients who received IL-6 pathway blockade. The hospitalized COVID-19 negative cohort allowed us to identify immune alterations that were shared between severe COVID-19 and other critically ill patients. Collectively, our findings indicate that selection of immune interventions should be based in part on disease presentation and early disease trajectory due to the profound differences in the immune response in those with mild to moderate disease and those with the most severe disease.

The coronavirus-19-disease (COVID-19) pandemic has brought a worldwide focus not only on the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), but also on how immunity to this virus both promotes viral clearance and contributes to morbidity and mortality in infected individuals. There is a wide range of disease severity in SARS-CoV-2 infected individuals, ranging from asymptomatic infection to severe COVID-19 requiring mechanical ventilation, and in some cases, to death. Some factors have been identified that are associated with increased disease severity and poor outcome during COVID-19, including age, race, obesity, hypertension, and type 2 diabetes (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) . However, we still do not understand the biologic factors that contribute to disease severity and outcome. It is becoming clear that not only does the severity of disease vary amongst SARS-CoV-2 infected individuals, but the immune response can also vary widely leading to differing immune landscapes between patients. Therefore, it is important to understand how the immune landscape contributes to COVID-19 severity and outcome. Another important gap in our knowledge is how the immune landscape in COVID-19 resembles or is distinct from that seen in critically ill patients hospitalized for other reasons, since the immune landscape may change in the context of critical illness regardless of its etiology. In particular, it is important to determine if the early immune landscape can be used to inform which COVID-19 patients will have a severe disease course, and would benefit from early interventions.

Although we can learn about immunity to SARS-CoV-2 by assessing a snapshot of the immune response at one point in time, the immune response to infection is dynamic and is best studied over time. Early immune responses to viruses are dominated by the innate immune system, including neutrophils, monocytes, plasmacytoid dendritic cells (pDCs) and natural killer (NK) cells, while adaptive immune responses of T and B cells critical for viral clearance develop over days to weeks. Understanding how these populations change over time and relate to disease trajectory can give insight into the signature of a productive anti-SARS-CoV-2 immune response associated with clinical improvement, and whether immune dysregulation contributes to severe COVID-19. Additionally, early in the pandemic hospitalized patients were treated with a variety of experimental therapeutics, including the antiviral agent remdesivir, cytokine modulating therapies, and plasma from convalescent patients, all with varying efficacy in clinical studies and trials. However, how and if these treatments affect the immune landscape before and after therapeutic exposure has not been described. To address these outstanding and important questions regarding the immune response during COVID-19, we used mass cytometry integrated with detailed clinical data to examine how the immune landscape changes over time in severe and moderate disease through natural progression and recovery, and also in the context of immune intervention.

We collected peripheral blood from 59 patients with COVID-19 (52 hospitalized patients and 7 ambulatory outpatients) at the Virginia Mason Medical Center, Seattle, Washington during the months of April and May 2020. Notably, we performed deep longitudinal sampling over the course of disease with an average of 4 time points per subject (Range: 1-18; Figure 1 ) allowing for detailed immune trajectories of recovery. Patients were classified based on maximum disease severity using a 7-point ordinal scale (OS) representing the following outcomes: 1, not hospitalized with resumption of normal activities; 2, not hospitalized, but unable to resume normal activities; 3, hospitalized, not requiring supplemental oxygen; 4, hospitalized, requiring supplemental oxygen; 5, hospitalized, requiring nasal high-flow oxygen therapy, noninvasive mechanical ventilation, or both; 6, hospitalized, invasive mechanical ventilation; and 7, death (12) .Of the hospitalized patients, 24 were classified as having severe disease on the basis of requiring management in a critical care unit (CCU); all required mechanical ventilation (maximal OS≥6), except one who was on high flow oxygen (maximal OS=5). The remaining 28 hospitalized patients were not in the CCU and were classified as having moderate COVID-19, with all requiring supplemental oxygen at some point in their hospital course (maximal OS=3-5).

The 7 ambulatory patients had mild disease (OS=2) and did not require hospitalization. For a control group, we also collected blood from 17 hospitalized patients who tested negative for SARS-CoV-2; four of these patients were admitted to the CCU and the remainder to the floor.

These patients were age and sex-matched to the hospitalized COVID-19 groups, and were admitted for a variety of conditions including respiratory (n=4), cardiac (n=4), gastrointestinal (n=3), neurologic (n=3) and miscellaneous conditions (n=3).

The demographic and clinical characteristics of all the patient groups are summarized in Table 1 . There was no significant difference in age or sex composition between severe, moderate and mild COVID-19 groups. Regarding racial distribution, there was an overrepresentation in the severe COVID-19 group of African American (16.7%) and Hispanic (37.5%) individuals based on the Washington state population, which is 78.5% white, 4.4% African American and 13% Hispanic (13). Duration of symptoms at time of presentation was longer in the severe disease group (median 9 days, range 3-22) compared to both the moderate (median 4 days, range 0-27) and mild (median 5 days, range 2-14) groups (p value=0.01). Duration of hospitalization was also significantly longer in the severe disease group (median 19 days, range 4-65) compared to the moderate disease group (median 6 days, range 2-28) (p value<0.01), although discharge was delayed for some patients due to restrictions placed on transfers to skilled nursing facility pending viral clearance from nasopharyngeal swabs.

Chronic medical conditions such as diabetes, hypertension and cancer were common in the hospitalized COVID-19 cohorts. Diabetes was present in 45.8% of the severe group, 28.6% of the moderate group and 28.6% of the mild group. Hypertension was present in 50% of the severe group and 67.9% of the moderate group but absent in the mild group. Cancer was present in 4.2% of the severe COVID-19 group, 21.4% of the moderate group and absent in the mild group.

Obesity was also more prevalent in the hospitalized COVID-19 cohort with a median BMI > 29 in both severe and moderate disease groups compared to a median BMI ~25 in the mild COVID-19 (p value = 0.08) and the hospitalized SARS-CoV-2 negative groups.

Because this cohort was from the early stage of the pandemic in the USA, hospitalized patients received a variety of experimental treatments, including hydroxychloroquine, remdesivir, tocilizumab and convalescent plasma (Supplemental Figure 1 ). Notably many patients received more than one type of experimental treatment. In the severe COVID-19 group, 7 patients (29.2%) received hydroxychloroquine, 17 (70.8%) received remdesivir, 8 (33.3%) received tocilizumab and 15 (62.5%) received convalescent plasma. Among the moderately ill, 2 (7.1%) received hydroxychloroquine, 11 (39.3%) received remdesivir, and 4 (14.3%) received convalescent plasma. The mild disease group did not receive any of these COVID-19 therapies.

We assessed the immune landscape by combining clinical data with mass cytometry (CyTOF) performed on whole blood samples recovered from the clinical laboratory. The CyTOF panel was designed to assess the composition of the innate and lymphocyte compartments and determine the maturation, lineage and activation status of these cell populations (Supplemental Table 1 , Supplemental Figures 2-4) . To better understand the impact of disease, we performed correlation analysis on the first sample collected for each patient in the COVID-19 cohort (n=59; Figure 2 and Supplemental Figure 5 ). The heatmap in Figure 2A shows all significant correlations between clinical data (disease severity ordinal score, age, BMI and CBC) and

CyTOF immune cell percentages of the total CD45+ (pan-leucocyte marker) cell compartment, whereas the correlation network in Figure 2B focuses only on correlations among major leukocyte populations identified by CyTOF. We found correlations consistent with the current literature. For example, white blood cell (WBC) counts and neutrophil counts were significantly correlated (Figure 2A) , not surprisingly given that neutrophils comprise a large proportion of WBC, and both are elevated in severe COVID-19 (14, 15) . Neutrophils in both the CBC and Together these findings for pDCs and basophils are consistent with recent studies reporting depletion of these cell types in acute COVID-19 (19, 20) . Notably, unlike other T cell populations the percentage of T follicular helper (Tfh) cells in the memory CD4+ compartment also showed a positive correlation with neutrophils, although this did not reach statistical significant ( Figure 3F ). Taken together these observations indicate that coordinate and counteracting changes in neutrophils, lymphocytes, pDCs and basophils drive the immune signature of COVID-19.

In order to understand whether the immune signature in COVID-19 differed by disease severity we determined the correlation between cell frequency and ordinal score at the time of sampling.

Increasing neutrophil frequency was positively correlated with increasing disease severity (Pearson correlation ~ 0.46, FDR-adjusted p < 0.01), while T cells, NK, pDCs and basophils were lower in severe disease (all FDR-adjusted p-values < 0.005) ( Figure 4A ). To determine if the immune landscape early in disease distinguishes severe from mild disease, we next performed a cross-sectional analysis of our population categorized based on an individual's highest disease score during the course of their illness using data from the first sample collected for each patient (Figures 4B-D, Supplemental Figure 5 ). The CBC data showed the greatest difference with disease severity in white blood cell counts with an increase in the absolute neutrophils and monocyte counts and low absolute lymphocyte counts ( Figure 4B ). However, these CBC results frequently fell within the normal range and notably, the hospitalized COVID-19 negative population showed very similar changes to those seen with severe COVID-19, suggesting that these findings are not unique to COVID-19 but are instead reflective of critical illness.

In contrast, the cross-sectional analysis of the CyTOF dataset identified two different patterns of immune alterations in the COVID-19 cohort: those that were also present in the hospitalized COVID-19 negative cohort, and those that were unique to severe COVID-19.

Immune cell populations that were similar between severe COVID-19 and hospitalized COVID-19 negative patients correlated with COVID19 disease score at all time points, as shown in Figure 4A . Specifically, there was an increase in neutrophils and HLA-DR lo monocytes with a decrease in T cells, NK, basophils and pDCs in severe disease ( Figure 4C ). Notably, for each of these cell types the changes seen in severe COVID-19 subjects were similar to the hospitalized COVID-19-negative cohort, suggesting that these changes are features of critical illness and not unique to severe COVID-19. Immune alterations unique to severe COVID-19 in this crosssectional analysis included increases in CD38+ CD8 T cells (FDR-adjusted p = 0.02), Tfh cells (FDR-adjusted p = 0.03) and plasmablasts (FDR-adjusted p = 0.00007) ( Figure 4D , Supplemental Figure 5 ). There were also increases in CD4 central memory T cells and HLA DR+ CD8 T cells although these were not statistical significant after adjusting for multiple testing ( Figure 4D ).

Unsupervised hierarchical clustering of the CyTOF data for each subject's initial sample identified three major clusters of patients ( Figure 4E ): a T cell predominant cluster with a relative decrease in neutrophils (cluster A), a cluster with mixed features including a predominance of monocyte, DC and NK cells (cluster B), and a third cluster (cluster C) whose patients had high levels of neutrophils and a relative paucity of other cell types. These clusters generally differentiated individuals based on their disease severity, with more moderate disease courses and good outcome associated with clusters A and B, while those with the most severe disease and death were associated with cluster C. These findings indicate that there is not one single immune signature in COVID-19, but that the immune response differs in individuals based on the ultimate disease severity.

To better understand the kinetics and coordinated changes in immune signatures, we tracked immune cell types in the blood over time based on date of admittance to the hospital. We focused on exploring differences in longitudinal analysis of moderate and severe patients based on distinct clustering between these groups as shown in Figure 4E ). However, it should be noted that this was not the case for all innate cells examined. For example, HLA-DR lo monocytes, which we and others found to be increased in severe COVID-19 ( Figure 4C ) (21) and are known to be increased in severe inflammatory syndromes such as sepsis (22) (23) (24) , were more dynamic in the severe COVID-19 cohort than the moderate COVID-19 cohort. DR lo monocytes in severe COVID-19 subjects Figure 5D ), likely in response to ongoing inflammation due to viral persistence. Consistent with this idea, the percentage of CD8 T cells expressing HLA-DR, a marker of activation, also increased over time in the severe COVID-19 cohort ( Figure 5D ), as did CD8 T cells expressing CD38 and PD-1 (Supplemental Figure 7B ) while total memory CD8 T cells increases were similar between moderate and severe patients ( Figure 5D ). Overall, our longitudinal analysis revealed that the immune trajectory differs between moderate and severe patients during the first two weeks after initial hospitalization. Patients with moderate disease showed signatures of a productive anti-viral response that resolved within the 2 weeks of the study time, whereas patients with severe disease showed signs of an aberrant response after hospital admittance that persisted for at least the first two weeks in hospital.

To identify key immune cell populations that are associated with either clinical improvement or decline, we focused our analysis on samples taken from individuals before and after a change in ordinal score, reflective of disease severity. We assessed changes in the absolute abundance of immune cell populations by CBC or in the frequency of immune cell subsets in our CyTOF analyses across these key clinical times. We identified subjects that had samples drawn across a score improvement of ≥ 2, or a score decline of ≥ 1 ( Figure 6A ). This analysis identified several populations whose abundance or frequency was significantly altered upon changes in ordinal score ( Figure 6B ). Consistent with the lymphopenia observed in severe COVID-19, we found that absolute lymphocytes decreased with clinical decline whereas an increase in the absolute number of lymphocytes was associated with clinical improvement. The increase in lymphocytes was mediated by a general increase in the frequency of naïve and memory CD4+ and CD8+ T cells as well as NK cells, but not B cells. The frequency of pDCs also increased in subjects Figure 6C ). This analysis demonstrates that the immune landscape is dynamic in COVID-19, and that resolution of key features of severe disease resolve co-incident with improvement in clinical status.

Early immune signatures of tocilizumab, but not convalescent plasma, treatment in severe

To determine if there were immune signatures of tocilizumab or convalescent plasma treatment, we identified 7 patients treated with tocilizumab and 7 patients treated with convalescent plasma in our cohort who had CyTOF samples both before and after treatment (Supplemental Table 2 ).

Notably, these patients all had severe disease and there were stringent criteria for the use of tocilizumab including rapidly escalating oxygen needs combined with an IL-6 level > 20x upper limit of normal (ULN); and CRP >125 mg/dl (ULN, 7). Marked elevations in ferritin, LDH and D-dimer were also weighted in the decision making process. All patients were also treated with remdesivir, with the exception of one patient in the convalescent plasma group. Additionally, 6/7 patients in the tocilizumab group analyzed were treated with convalescent plasma prior to tocilizumab treatment (1-6 days pre tocilizumab). None of the 7 patients in the convalescent plasma group were treated with tocilizumab during the time points analyzed.

We first assessed serum C-reactive protein (CRP) levels in these two groups as a measure of the effectiveness of tocilizumab treatment, which should reduce this marker of systemic inflammation. Indeed, treatment with tocilizumab swiftly reduced serum CRP in all patients ( Figure 6A ). Serum ferritin was also reduced mainly in those patients with very high concentrations pre-treatment (Supplemental Figure 10A ). In contrast, convalescent plasma treatment had no consistent effect on CRP levels ( Figure 7A ). Therefore, tocilizumab treatment showed an acute clinical signature of reduced inflammation in patients with severe COVID-19, whereas convalescent plasma did not consistently affect these measures.

We then compared acute changes in immune populations in the blood before and after tocilizumab or convalescent plasma treatment by assessing the closest CyTOF sample before day of treatment (range from day -4 to day 0) with the first CyTOF sample available after treatment (Range: days 2 to 9 post-treatment). In the tocilizumab group there were several populations of immune cells that differed significantly before and after treatment ( Figure 7B ). In contrast, there were no significant changes after convalescent plasma treatment in the immune cell populations analyzed by CyTOF ( Figure 7C , Supplemental Figures 10B-C) . The significant changes in response to tocilizumab treatment included a reduction in the percent of neutrophils and an increase in the percent total T cells, eosinophils, basophils, and DCs among CD45+ cells ( Figures 7B, 7D, 7E ). There were also increases in several CD4 and CD8 T cell subpopulations, and no changes in any B cell populations after tocilizumab ( Figure 7B ). Moreover, our findings for T cells, B cells, neutrophils and basophils were consistent with our signature of clinical improvement shown in Figure 5 . However, there was not complete overlap between the tocilizumab signature and the clinical improvement signature as NK cells and pDCs were not significantly changed by tocilizumab ( Figure 7B , data not shown) but were increased in improving patients ( Figure 6B ). In the tocilizumab group, we also identified increased populations associated with T cell activation, including HLA-DR+ and CD38+ CD4 and CD8 T cells ( Figures 7B, 7D, 7E) . In summary, we observed a clear acute signature of tocilizumab treatment that shares some but not all features of the immunologic changes seen with clinical improvement, whereas there is no acute change in the immune landscape with convalescent plasma treatment, in patients with severe COVID-19.

A growing literature indicates that the immune landscape is profoundly altered by and differs between individuals dependent on disease severity (18, 20, 25, 26) . Whether the immune landscape is a reflection of disease severity, a source of severe disease or a combination of the two is still not fully understood. Here, we utilized recovered samples from the clinical laboratory to rapidly assess peripheral blood cell populations by CyTOF and this data was analyzed in conjunction with clinical laboratories and disease severity scores. Importantly, we were able to collect samples longitudinally among the hospitalized individuals, allowing us to examine the evolution of immune responses through natural progression and recovery, and in the context of immune intervention. Novel aspects of our study included: 1) deep longitudinal sampling allowing for detailed immune trajectories of recovery, 2) a control cohort of moderate and severely ill hospitalized COVID-19 negative patients, and 3) analysis of immune signatures associated with tocilizumab and convalescent plasma treatments.

Notably, we found that at the time of initial sampling the immune landscape in COVID-19

forms three dominant clusters that relate to disease severity. When we examined individual cell populations based on disease severity, we found, as others have, that the neutrophil to lymphocyte ratio is increased in individuals with severe COVID-19 (15) (16) (17) (18) . Furthermore, this inverse relationship with neutrophils applies to basophils, DC, NK and monocytes, and only modestly with B cells, and is most pronounced among T lymphocytes with the exception of Tfh, which are positively correlated to neutrophil numbers, a finding also consistent with the current literature (18) . Interestingly, many changes seen in severe COVID-19 compared to mild and moderate disease were also seen in our hospitalized COVID-19 negative control cohort.

Including this unique control group allowed us to identify differences between critically ill Caveats for our tocilizumab analysis include the small cohort size, and that all but one of these patients were treated with convalescent plasma prior to tocilizumab treatment. Therefore, it is possible that convalescent plasma acts synergistically with tocilizumab to cause the immune signature we identified. Interestingly and in contrast to tocilizumab, we saw no clear immune signature of convalescent plasma within 7 days, suggesting either our cohort was too small to see changes, the immune populations change after the times we analyzed, or convalescent plasma does not act at the level of blood leukocyte populations. It is clear that further investigation is needed to determine if tocilizumab has a therapeutic role in COVID- 19 , and in what patient population it would be useful and this may be determined in part by the character and trajectory of the immune landscape of the patient.

The demographics of our COVID-19 patients were consistent with published case reports.

African Americans and Hispanics were overrepresented in the severe COVID-19 group to the population of Washington State, which is consistent with reports from other states in the USA (10, 11) . We also found that type 2 diabetes was more common in those with severe disease compared to moderate or mild disease. Notably, all groups have higher diabetes prevalence than the US or Washington rates (28); the highest prevalence in Washington state is among 65-74

year olds at 21.5%, which is more than doubled in the cohort with severe disease described here.

reduced T-cell function and chronic inflammation have been postulated as potential mechanisms driving this increased risk (29) . In addition, some glucose-lowering agents used in diabetes are known to impact the immune system (reviewed in (30) . Full analysis of the differential impact of diabetes and its treatment on our immune signatures is beyond the scope of this work but merits further analysis.

There are limitations to this study. Due to the urgency of the pandemic, we chose to use recovered clinical samples for our study and thus the collection schedule and sample availability was dictated by the treatment needs of the patient. This meant that we did not have the same time points for every patient, and that we could not match between groups the medications that individuals were already taking due to pre-existing comorbidities, some of which may impact the immune responses seen here. In addition, the differences between the mild, moderate and severe COVID-19 groups may reflect the time from disease onset, which significantly varied between these groups, and/or differences in viral burden, which we could not assess. The subjects in our hospitalized control group were not matched to the SARS-CoV-2 positive groups by race, although they are well matched by age.

In summary, we have identified unique features of the immune landscape in moderate versus severe COVID-19 along with features that are common to moderate and severe non-COVID illness. Importantly, our findings indicate that selection of immune interventions should be based in part on disease presentation and early disease trajectory due to the profound differences in the immune response in those with mild to moderate disease and those with the most severe disease. Finally, our characterization of the variety of immune signatures in COVID-19 provides insight into the types of immune interventions that may be beneficial in the treatment of severe disease.

Using our newly developed 33-parameter CyTOF panel, we characterized the immune response Correlation graph: The correlation graph in Fig. 2A was built from the matrix of Pearson correlations in Fig. 2B using the R iGraph package (31) Heatmap: The heatmap in Figure 4E was generated using Euclidean distance and the clustering method Ward.D2.

Smoothed time-course graphs: Time-series data from each patient were organized in terms of the relative number of days from the date of the first sample (hereon denoted pseudo-time), and then aligned by first sample. To reduce the potential effects of outlier samples, median values were calculated for each severity category and each day for the samples available. If no samples were available at a given pseudo-time day, we inferred a value using linear interpolation between the before and after pseudo-time points. The vertical bars at each pseudo-time point are equal to one standard deviation from the indicated median value. Plot point with no error bars are those with only one sample or represent an inferred value. Loess smoothing was performed on the median values for each disease severity class using the geom_smooth function in the R ggplot library (32) UMAP: The UMAP plots in Fig. 5A were generated directly from the CyTOF signal intensities following archsinh transformation with a co-factor value of 5. To ensure against batch and other potential confounding effects, we specifically selected samples collected and stained in a highly uniform fashion from a single donor and z-score normalized probe intensities for each sample prior to UMAP projection to 2D. 

Factors associated with COVID-19-related death using OpenSAFELY

Associations of type 1 and type 2 diabetes with COVID-19-related mortality in England: a whole-population study. The lancet Diabetes & endocrinology

Risk factors for COVID-19-related mortality in people with type 1 and type 2 diabetes in England: a population-based cohort study. The lancet Diabetes & endocrinology

Clinical Characteristics and Risk Factors for Mortality of COVID-19 Patients With Diabetes in Wuhan, China: A Two-Center

Obesity and COVID-19 Severity in a Designated Hospital in Shenzhen

Clinical Characteristics and Outcomes of Patients With Diabetes and COVID-19 in Association With Glucose-Lowering Medication

Diabetes in COVID-19: Prevalence, pathophysiology, prognosis and practical considerations

Risk Factors for Intensive Care Unit Admission and In-hospital Mortality among Hospitalized Adults Identified through the U.S. Coronavirus Disease 2019 (COVID-19)-Associated Hospitalization Surveillance Network (COVID-NET)

Obesity Is a Risk Factor for Greater COVID-19 Severity

Hospitalization and Mortality among Black Patients and White Patients with Covid-19

Assessment of COVID-19 Hospitalizations by Race/Ethnicity in 12 States

A Trial of Lopinavir-Ritonavir in Adults Hospitalized with Severe Covid-19

Clinical and immunologic features in severe and moderate Coronavirus Disease

Dysregulation of immune response in patients with COVID-19 in Wuhan, China

Neutrophil-to-lymphocyte ratio predicts critical illness patients with 2019 coronavirus disease in the early stage

Immune Phenotyping Based on the Neutrophil-to-Lymphocyte Ratio and IgG Level Predicts Disease Severity and Outcome for Patients With COVID-19. Frontiers in molecular biosciences

Comprehensive mapping of immune perturbations associated with severe COVID-19

A dynamic COVID-19 immune signature includes associations with poor prognosis

Immunomonitoring from Acute to Recovery Phase of Severe COVID-19

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

Monocytic HLA-DR expression in intensive care patients: interest for prognosis and secondary infection prediction

Genomic landscape of the individual host response and outcomes in sepsis: a prospective cohort study

Downregulation of Blood Monocyte HLA-DR in ICU Patients Is Also Present in Bone Marrow Cells

Longitudinal analyses reveal immunological misfiring in severe COVID-19

Deep immune profiling of COVID-19 patients reveals distinct immunotypes with therapeutic implications

Tocilizumab in hospitalized patients with COVID-19 pneumonia. medRxiv preprint. 2020. 28

COVID-19 pandemic, coronaviruses, and diabetes mellitus

Coronavirus Infections and Type 2 Diabetes-Shared Pathways with Therapeutic Implications

The igraph software package for complex network research

Elegant Graphics for Data Analysis

We would like to acknowledge the Benaroya Family Foundation, the Leonard and Norma Klorfine Foundation, and Glenn and Mary Lynn Mounger for their funding of this project. We also acknowledge the Allen Institute for Immunology for their funding support for the development of the CyTOF panel used in this study. We also would like to thank Carmen Mikacenic for help with obtaining IRB approval, and Henry T. Bahnson for help with statistical analysis. See Supplemental Acknowledgements for BRI COVID-19 Research Team details.