key: cord-1038870-mdsjbfmx authors: Kuri-Cervantes, Leticia; Pampena, M. Betina; Meng, Wenzhao; Rosenfeld, Aaron M.; Ittner, Caroline A.G.; Weisman, Ariel R.; Agyekum, Roseline; Mathew, Divij; Baxter, Amy E.; Vella, Laura; Kuthuru, Oliva; Apostolidis, Sokratis; Bershaw, Luanne; Dougherty, Jeannete; Greenplate, Allison R.; Pattekar, Ajinkya; Kim, Justin; Han, Nicholas; Gouma, Sigrid; Weirick, Madison E.; Arevalo, Claudia P.; Bolton, Marcus J.; Goodwin, Eileen C.; Anderson, Elizabeth M.; Hensley, Scott E.; Jones, Tiffanie K.; Mangalmurti, Nilam S.; Luning Prak, Eline T.; Wherry, E. John; Meyer, Nuala J.; Betts, Michael R. title: Immunologic perturbations in severe COVID-19/SARS-CoV-2 infection date: 2020-05-18 journal: bioRxiv DOI: 10.1101/2020.05.18.101717 sha: 2ccf9cee81ab6b25002c6dfc392ea88276ce8225 doc_id: 1038870 cord_uid: mdsjbfmx Although critical illness has been associated with SARS-CoV-2-induced hyperinflammation, the immune correlates of severe COVID-19 remain unclear. Here, we comprehensively analyzed peripheral blood immune perturbations in 42 SARS-CoV-2 infected and recovered individuals. We identified broad changes in neutrophils, NK cells, and monocytes during severe COVID-19, suggesting excessive mobilization of innate lineages. We found marked activation within T and B cells, highly oligoclonal B cell populations, profound plasmablast expansion, and SARS-CoV-2-specific antibodies in many, but not all, severe COVID-19 cases. Despite this heterogeneity, we found selective clustering of severe COVID-19 cases through unbiased analysis of the aggregated immunological phenotypes. Our findings demonstrate broad immune perturbations spanning both innate and adaptive leukocytes that distinguish dysregulated host responses in severe SARS-CoV-2 infection and warrant therapeutic investigation. One Sentence Summary Broad immune perturbations in severe COVID-19 The coronavirus-19-disease pandemic caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has surpassed four million cases world-wide (4, 088, 842 as of 05/12/2020), causing more than 283,000 deaths in 215 countries (1). While asymptomatic in some, SARS-CoV-2 infection can cause viral pneumonia that progresses to acute respiratory distress syndrome (ARDS), and even multi-organ failure, in severe cases (2, 3) . Reports have shown that SARS-CoV-2 has the ability to productively infect lung epithelium, gut enterocytes and endothelium (4) (5) (6) . It is unclear whether disease severity is caused by the viral infection, the host response, or both, emphasizing the urgent need to understand the immune perturbations induced by SARS-CoV-2 (3) . Knowledge of the immunological signatures of severe COVID-19 is continually evolving. Although lymphopenia has been linked to disease severity, the majority of published studies are based on retrospective analyses of clinical data (3, (7) (8) (9) (10) (11) (12) (13) (14) . Immune profiling studies to date have been conducted as single case reports or focused only on moderate, severe or recovered COVID-19 with limited numbers of individuals (15) (16) (17) (18) , and have not necessarily reflected the range of comorbidities globally associated with severe COVID-19. Studies of peripheral blood mononuclear cells by mass cytometry or single cell RNA sequencing (scRNAseq) have provided valuable insights into possible immune perturbations in COVID-19 but have not assessed the contributions of granulocytic populations, or, in the case of scRNAseq, defined expression or modulation of cellular proteins (16) . In particular, modulation of granulocytic populations is suggested to be relevant during COVID-19 infection (12) . To address these issues, we conducted a comprehensive analysis of the overall immunologic state of 42 individuals with different trajectories of SARS-CoV-2 infection and COVID-19 (moderate, severe, and recovered), compared with 12 healthy donors using whole blood to capture the full breadth of immunological perturbations and activation occurring in circulating lymphocytes and major granulocyte populations. We further explored modulation of the B cell repertoire, its associations with the establishment of a SARS-CoV-2-specific humoral response, and activation of T cells relative to disease severity. Together our results reveal a potential platform for assessing disease trajectory, and identify distinct immune perturbation patterns in severe COVID-19 that merit consideration for therapeutic immunomodulation strategies to ameliorate disease severity and organ failure. We recruited 35 inpatients with active COVID-19, seven of whom had moderate and 28 with severe disease, seven recovered COVID-19+ donors, and 12 healthy donors (HD). All recovered donors reported mild disease, and did not receive inpatient care or COVID-19 directed therapy during the course of their illness. For inpatients, median follow up after enrollment was 27 days (range 20 -43) since blood draw. General demographics and clinical characteristics are shown in Table 1 . The median ages in the moderate and severe COVID-19+ groups were 59 and 68 years old, respectively, concordant with previous reports (8) , and were not significantly different (p=0.51). Both the HD and recovered groups were significantly younger than individuals with severe COVID-19+ (p<0.001 in both cases). In line with a recent publication (9) , the majority of the individuals in the severe and recovered groups were male (67.9% and 71.4%, respectively), while approximately 29% were male in the moderate disease group. The median number of days since onset of symptoms to disease progression in donors with severe COVID-19 was nine, similar to previous publications (3, 10) . Individuals with moderate disease also reported a median of nine days since onset of symptoms. In accordance with a recent report (19) , individuals with COVID-19 had high incidence of underlying pulmonary disease (11/35 including moderate and severe, 31.4%) and were current or former smokers (13/35 including moderate and severe, 42.7%, higher in individuals who developed severe disease). Hypertension and hyperlipidemia were the most frequent co-morbidities in moderate and severe COVID-19. The majority of individuals with severe COVID-19 presented with moderate and severe ARDS (20) , and hospital mortality was 14.3% within this group. Thromboembolic complications, metabolic, vascular and pulmonary disease were also observed more frequently among those with severe disease (Table 1) . As part of clinical care, D-dimer, procalcitonin, ferritin, lactate dehydrogenase, and C-reactive protein levels were measured in moderate and severe COVID-19 individuals. Median levels of D-dimer at the time of blood draw were 3.985 µg/ml in severe, and 0.62 µg/ml in moderate COVID-19 donors (severe n=20, moderate n=5; p=0.0022). We found higher levels of ferritin in the severe group compared to the moderate group (medians: 919.5 ng/ml in severe, n=20, and 162 ng/ml in moderate, n=5; p=0.007). Consistent with previous findings (13) , median procalcitonin values were relatively low, though higher in severe donors than in those with moderate disease (medians of 0.45 ng/ml, n=15, and 0.06 ng/ml, n=5, respectively; p=0.0014). Levels of lactate dehydrogenase and C-reactive protein were similar across groups. Bacterial co-infection was present in nine individuals with severe COVID-19, and in only one moderate donor. An extended list of clinical information of the analyzed individuals is shown in Table S1 . To assess the general landscape of immune responses and their perturbation during severe COVID- 19 , we performed extensive immunophenotyping to characterize the frequencies of circulating immune subsets in HD, or in moderate, severe and recovered COVID-19 individuals (Fig. 1, Fig. S1 ). We observed an expansion in the proportion of both neutrophil and eosinophil populations in severe COVID-19 donors compared to HD (median neutrophil frequencies within viable CD45+ cells: 79.9% in severe COVID-19 and 47.7% in HD; p<0.0001; and, median eosinophil frequencies within viable CD45+ cells: 0.68% eosinophils in severe COVID-19 and 0.17% in HD, p=0.0015; Fig. 1A -C). The neutrophil frequency also differed significantly between moderate vs. severe COVID-19 disease (p=0.0046, median frequency of 53% of viable CD45+ in moderate group), 8 but did not show increased activation or cycling (Fig. S2A) . Furthermore, we saw decreased expression of CD15 in neutrophils between HD and severe COVID-19 individuals (p=0.0095), but not in eosinophils (Fig. S2B) . We did not observe significant differences in the immature granulocyte frequencies between HD and COVID-19 individuals. However, the proportion of immature granulocytes in moderate and severe COVID-19 donors correlated inversely with the time since onset of symptoms (Fig. S2C) . In contrast to previous work (21), the total proportion of monocytes (CD14+ HLA-DR+), as well as monocyte subsets (defined by CD14 and CD16), was similar across groups (data not shown). Donors with severe COVID-19 had lower proportions of dendritic cells (DC) compared to moderate disease (p=0.003) and HD (p=0.0374; median percentage in viable CD45+ cells: 0.42% in severe, 0.64% in moderate and 0.49 in HD, Fig. 1A ), but not with recovered individuals. Consistent with previous reports (7, 8, (22) (23) (24) , we observed a relative decrease in the percentages of all lymphocyte subsets (Fig. 1A, B, D) . Severe COVID-19 individuals had significantly lower relative proportions of T cells (median frequency within CD45+ cells: 4.5% in severe COVID-19+ and 30.6% in HD; p<0.0001), CD161+ CD8+ T cells (median frequency of CD45+ cells: 0.002% in severe COVID-19 and 1.3% in HD; p<0.0001), innate lymphoid cells (ILCs, median frequency of CD45+ cells: 0.005% in severe COVID-19 and 0.03% in HD; p<0.0001) and natural killer (NK) cells than HD (median frequency of CD45+ cells: 0.95% in severe COVID-19 and 4.5% in HD; p<0.0001). We did not find significant differences in the frequencies of these cell subsets between HDs and moderate or recovered COVID-19 individuals. Within the NK cell lineage, we observed a drastic decrease in the frequencies of both CD56brightCD16-and CD56dimCD16+ NK cells in severe COVID-19 vs. HD (Fig. S2D ). In the recovered group, the proportions of T cells, CD161+ CD8+ T cells, ILCs and NK cells were higher than in donors with severe COVID-19 but similar to HDs (median frequencies within viable CD45+ cells: 22% of T cells, 0.1% of CD161+ CD8+ T cells, 0.014% of ILCs, 3.5% of NK cells). The proportions of regulatory CD4+ T cells and circulatory follicular CD4+ T cells were similar across studied groups (Fig. S3A, B) . Although we did not observe differences in CD4+ and CD8+ memory T cell subsets between groups (data not shown), we did find a negative correlation with the frequency of central memory T cells (TCM) and days since the onset of symptoms (Spearman r= -0.41 p=0.02 for CD4+ TCM; Spearman r= -0.61 p=0.0002 for CD8+ TCM, Fig. S3C ). Given that the neutrophil-to-lymphocyte ratio may be an independent risk factor for severe disease (25, 26), we examined the neutrophil:T cell ratio (based on their frequencies within viable CD45+ cells). Individuals with severe COVID-19 had a ratio of 15, while all other studied groups had ratios of less than 2.5. Furthermore, using logistic regression analyses, we did not find any associations between the reported frequencies and comorbidities (pooled together as vascular/metabolic disorders, underlying lung disease and bacterial infections, Table S1 ). Altogether, these data reveal multiple immunophenotypic abnormalities in severe COVID-19, which are not found in donors with moderate or recovered disease. Although we observed only marginal differences in the proportions of total B cells between the studied groups ( Fig. 1) , B cell plasmablasts were significantly expanded in severe COVID-19 donors compared to HD (Fig. 1D, Fig. 2A ; median frequency within B cells of: 9.7% in severe COVID-19 and 0.48% in HD, p<0.0001). These cells characteristically displayed high levels of Ki-67 and low levels of CXCR5 expression (Fig. S4A ). Similar to observations in the immune atlas of recovered COVID-19 donors (16), expanded plasmablasts were not found in this group (median frequency with B cells of 0.3% in recovered, p<0.0001 vs. severe donors). The frequency of plasmablasts in individuals with severe COVID-19 did not correlate with age, days since onset of symptoms or the presence of co-morbidities (data not shown), similar to one report based on scRNASeq analyses (16) . In the non-plasmablast B cell population, we observed a decrease in the percentage of CD21+CD27+ in moderate and severe groups compared to HD (median frequency of nonplasmablasts of: 24% in HD, 10.8% in moderate disease and 6.7% in severe disease). These proportions were highly significant by nonparametric test of trend (p=0.0008), but only the severe COVID-19 group reached statistical significance vs. HD (p=0.0061, Fig. 2B ). Recovered COVID-19 donors had similar levels of CD21+CD27+ non-plasmablasts as the HD group (median of 23.8%). Of note, the frequency of CD21+CD27+ non-plasmablasts was directly correlated with the age of the donors among moderate and severe COVID-19 (Spearman r=0.35, p=0.4, Fig. S4B ). In contrast, we observed a significant increase in the proportion of CD21-CD27-non-plasmablasts in moderate (median of 16.6%) and severe (median of 10.4%) COVID-19 individuals compared to HD (median of 2.3%; p=0.0182 and p=0.004, respectively). We next assessed the expression of Ki-67 and CD11c, to determine if any of these subsets were a potential source for the expanded plasmablast population (27) (Fig. 2C ). We did not observe a larger proportion of cycling Ki-67+ CD21-CD27-B cells in moderate or severe COVID-19 individuals when compared with HD. We also found a reduction in the frequency of CD11c+ cells within CD21-CD27-B cells in donors with moderate COVID-19 compared to HD that was specific to this group (medians of: 6.9% in moderate and 49% in HD; p=0.0162). Previous work has suggested that the SARS-CoV-2 IgG levels could be associated with disease severity (12, 28) . With this in mind, and due to the changes observed in B cell subsets, particularly the expansion of plasmablasts in severe COVID-19, we explored the humoral responses in these donors. The levels of total IgG in plasma and serum were equivalent across the groups (Fig. S4C) . We then quantified IgM and IgG specific for the spike receptor binding domain (RBD) of the SARS-CoV-2. The levels of both antibodies were significantly higher in the severe and recovered COVID-19 individuals (Fig. 2D) . While the frequency of plasmablasts did not correlate with the levels of spike RBD-specific IgM or IgG, there was a positive association between the levels of spike RBD-specific IgM and IgG and time since onset of symptoms ( Fig. 2E ) in the moderate and severe groups. Together these data indicate an exacerbated plasmablast response in severe COVID-19, as well as the development of a strong SARS-CoV-2-specific humoral response. Having observed the expansion of plasmablasts in severe COVID-19 donors, we sought to determine whether this expansion in severe-COVID-19 resulted from non-specific stimulation. Therefore, we examined the antibody repertoire within samples from randomly selected HD (n=3), moderate COVID-19 (n=3) and severe COVID-19 (n=7) individuals. To sequence antibody heavy chain libraries, we amplified genomic DNA was amplified using primers spanning across nearly the full-length variable (VH) gene sequence and the entire third complementarity determining region (CDR3). After quality control and filtering, the processed antibody heavy chain rearrangements were grouped together into a data set comprising 76 sequencing libraries and 109,590 clones across all 13 individuals (Table S2 and GenBank/SRA PRJNA630455). To evaluate the clonal landscape, we ranked the proportion of clones within the top ten (1-10), next 90 (11-100), next 900 (100-1,000), and most diverse clones with ranks above 1,000 (1,000+) (Fig. 3A) . Donors with severe COVID-19 had an unusually high proportion of large clones comprising the majority of their circulating antibody repertoire, with the fraction occupied by the top 20 ranked clones (D20 measure) the highest compared to the healthy and moderate SARS-CoV-2 infected patients (Fig. 3B, Fig. S5 ) The D20 rank measure in moderate and severe disease also correlated positively with the plasmablast fraction (Fig. 3C) . In many severe COVID-19 individuals we observed very large top copy clones, exceeding the diagnostic thresholds for clinically significant monoclonal B cell lymphocytosis (29). These large clones were readily sampled across multiple independently amplified and sequenced libraries ( Focusing on the most frequently used VH genes, VH genes from different families were used more often in severe COVID-19 donors compared to HD, including VH6-1 (7-fold), VH3-48 and VH3-15 (~6-fold) and several others (Fig. S6C) . We also looked for skewing in VH family usage, which revealed a modest relative increase in the proportion of VH3 family members among COVID-19 individuals compared to HD (Fig. S6D) . However, there was considerable inter-individual variation in the usage of VH3 vs. other family members, with some individuals (such as S25) exhibiting substantial skewing towards particular VH families (data not shown). Given the absence of obvious or uniform VH restriction among COVID-19 individuals, we next analyzed the CDR3 sequences for shared characteristics in the COVID-19 donors. In individuals with severe disease, CDR3 sequences exhibited greater variation in length (Fig. 3G ), and were significantly longer among the top copy sequences (Fig. 3H ). To determine if the antibody heavy chain sequences from COVID-19 individuals are generated commonly or infrequently, we searched the Adaptive Biotechnologies public database, which consists of 37 million antibody heavy chain sequences (31), revealing 3995 matches to the CDR3 amino acid sequences in our dataset. Among the 50 most frequent clones in the COVID-19 individuals, the CDR3 lengths of the matching or "public" clones were shorter than the CDR3 lengths of the nonshared or "private" clones ( Fig. 3I) , indicating that the top copy clones in COVID-19 with long CDR3 sequences are mostly private. Finally, to determine if there were any collections of clones that harbored similar CDR3 amino acid sequences, we computed the edit distances of all of the amino acid sequences in the top 50 clones of each of the individuals. If there were sequence convergence, we would have expected to find clusters of sequences separated by 3 or fewer amino acids. We found no evidence of co-clustering of CDR3 sequences; rather, over 99% of the edit distances for the severe COVID-19 individuals' top copy clone pairs were more than 3 amino acids apart (Fig. 3J ). Consistent with this finding, alignment of top copy clone CDR3 amino acid sequences from severe COVID-19 individuals revealed highly variable amino acid sequences (Fig. S6E ). Taken together, these data show that severe COVID-19 is associated with large, oligoclonal B cell expansions with antibodies enriched for long and divergent CDR3 sequences. Acknowledging the characteristic differences in innate cell subset frequencies in severe COVID-19 individuals (Fig. 1) , we further assessed the phenotype of innate immune cells. CD161 has been reported to be a marker of inflammatory monocytes and NK cells (32-34). Despite having observed a decreased frequency of CD161+ CD8 T cells (Fig. 1A, D) , the frequencies of CD161+ monocytes and CD38+CD161+ NK cells were similar across study groups (Fig. S2E ). We next assessed the frequency and expression of CD16 by neutrophils, monocytes, NK cells and immature granulocytes. While the proportions of CD16+ monocytes and immature granulocytes were consistent between groups, severe COVID-19+ individuals had significantly lower circulating CD16+ NK cells in compared with HDs (median percentages of 68% in severe COVID-19 and 85.5% in HD; p=0.0023; Fig. 4A ; also observed when analyzing NK cell subsets in Fig. S2D ). . We did not, however, find significant associations between the frequency or expression of CD16 and IgG levels (Fig. S2F) . Although we found a decrease in the frequency of CD16+ monocytes in some severe COVID-19 individuals, this was not consistent amongst the whole cohort (Fig. 1A ). The monocyte CD16 expression level tended to decrease with disease severity Altogether, these findings indicate a substantial perturbation of the innate immune system in severe COVID-19. Whether this dysregulation is consequence or contributing factor towards COVID-19 severity remains to be defined. T cell activation has been reported in acute respiratory and non-respiratory viral infections (37-39). Consistent with recent case reports (15, 40, 41) , we observed increased activation of both memory CD4+ and CD8+ T cells in severe COVID-19 individuals compared to other study groups ( Fig. 5A and B). However, unlike the plasmablast response, heightened T cell activation was not observed in every severe COVID-19 individual and instead demonstrated significant heterogeneity. While overall the frequencies of CD38+ and HLA-DR+ CD38+ memory CD4+ and CD8+ T cells in severe COVID-19 were elevated compared to HD (CD4+, 7.6%, 2.2% vs 2.7%, 0.2%, p=0.009 and p<0.0001, respectively; CD8+, 9.2%, 3.9% vs. 0.6%, 0.09%; p<0.0001 for both cases), we did not find statistically higher Ki-67+ CD4+ or CD8+ T cells in COVID-19 individuals compared to HD. However, a subset of severe COVID-19 donors clearly had increased levels of Ki-67+ CD4+ and CD8+ T cells, reaching as high as ~25% in some individuals. The frequency of PD-1+ memory CD4+ T cells (44.3% in severe and 25.7% in HD, respectively; p=0.0084), but not CD8+ T cells, was also higher in the severe COVID-19 group compared to the HD group. For all measures, CD4+ and CD8+ T cell activation in recovered donors was equivalent to the HD group. Of note, the proportion of PD-1+ memory CD4+ T cells, but not of PD-1+ CD8+ T cells, in moderate or severe COVID-19 correlated with donor age (Fig. S3D ). In addition, the Due to limited samples, we did not include the moderate or recovered COVID-19 groups for this analysis. We found a significantly higher proportion of cytotoxic CD8+ T cells in severe COVID-19 than in HD (median frequency within memory CD8+ T cells of 48.7% and 27.2%, respectively; p=0.048). The frequencies of T-bet+ cells, as well as the levels of expression (measured by median fluorescence intensity) of perforin+ and granzyme B+ cells within the cytotoxic memory CD8+ T cell subset were similar between groups ( Fig. S3E -F). Cytotoxic CD8+ T cells from severe COVID-19 donors also had an increased proportion of cells expressing CD38 or co-expressing PD-1 and CD38 compared to HD (medians of 8.2% and 1.8%, respectively; p=0.0082; Fig. 5D and Fig. S3G ). These data indicate a heightened status of immune activation and frequency of cytotoxic CD8+ T cells during severe COVID-19, not observed in moderate or recovered disease. Finally, we performed an unbiased analysis to determine if the immune cells in severe COVID-19 disease cohort could be differentiated from the healthy, moderate, and recovered cohorts. We included all analyzed immune phenotype parameters described thus far, including the expression of activation markers within specific CD4+ and CD8+ T cell memory subsets (data not shown). We scaled all flow cytometry generated data using z-score, and performed hierarchical clustering ( Fig. 6A) . From this analysis, the data from 21/28 of the severe COVID-19 patients co-localized to a distinct cluster within the hierarchical tree. We further analyzed these data by principal component analysis, where we again found selective clustering of individuals with severe COVID-19 ( Fig. 6B ). The top parameters driving the clustering of the severe COVID-19 were associated with T cell activation in CD4+ and CD8+ T cell memory subsets, frequency of plasmablasts and frequency of neutrophils (Table S3) , also evidenced in the heat map shown in Fig. 6A . Independent analyses of the severe COVID-19 group did not produce separate clustering, likely due to reduced sample number. However, it is clear from the heatmap analysis that distinct patterns within the severe COVID-19 disease cohort may be present that further subdivide these individuals into different subgroups. Taken as a whole, our analysis reveals a characteristic immune phenotype in severe COVID-19, distinct not only from HD but also from other COVID-19 individuals with moderate or recovered disease. immunopathology. In line with a recent report (57), we did not observe clear sequence convergence of VH genes amongst all the severe COVID-19 individuals, but VH3 family members were enriched in some individuals. CDR3 sequences from individuals with severe COVID-19 had higher edit distances than individuals with mild disease or HD. While their size, somatic mutation status and association with the plasmablast fraction are suggestive of active participation in the immune response to SARS-CoV-2, it is unknown if these clones can recognize the virus, confer protection, or contribute to immunopathology. Future comparisons of our data to antibodies of known specificity may provide important insights into the dynamics of antibody responses in different phases of the illness and may reveal important differences between antibodies produced in the context of moderate vs. severe disease. T cell activation is typically observed during acute viral infections (58-60), and as expected (15, 18) we observed increased activation of both CD4+ and CD8+ T cells in severe COVID-19 that correlated with the plasmablast frequency. However, T cell activation was very heterogeneous across the severe COVID-19 patients, being equivalent to baseline in some while reaching up to ~25% of memory CD8+ T cells in others. This heterogeneity is relatively unusual compared to the symptomatic phase in other acute infections in humans, such as HIV, EBV, HCMV, HBV, and Ebola, where activation is uniformly detectable but to varying, and sometimes much higher, degrees (61-64). However, given the degree of lymphopenia observed in the severe COVID-19 patients, it is possible that activated T cells are migrating to, or sequestered in, the lung in response to the virus (23, 65-68), making it unclear if T cell activation is found in other sites as suggested by case study reports (6, 69) . We also observed a marked reduction in the frequency of CD161+ Gated on viable CD45+ cells Lines on the graphs indicate the median of the group. Differences between groups were calculated using Mann-Whitney rank-sum test. **** p<0.0001, ***p<0.001, *p<0.05. Lines on the graphs indicate the median of the group. Differences between groups were calculated using Kruskal-Wallis test with Dunn's multiple comparison post-test and Mann-Whitney ranksum test. **** p<0.0001, ***p<0.001, **p<0.01, *p<0.05. H12 R6 H4 H7 H3 R5 R2 M2 M4 M5 S1 R1 H9 S3 M7 S7 H5 H8 R3 M1 S2 H11 H1 R4 H2 H10 S20 S16 S5 S4 S13 S22 S21 S12 S6 S15 S26 S25 M3 S23 S24 S14 S8 S27 S9 S11 S17 S10 APACHE III scoring was based on data collected in the first 24 hours of ICU admission or the first 24 hours of hospital admission for subjects who remained on an inpatient unit. Clinical laboratory data was collected from the date closest to the date of research blood collection. Peripheral blood samples processed within 3 hours of collection. After plasma separation, 1 ml of whole blood was separated for staining and the remaining volume was used for PBMC isolation using SepMate tubes (StemCell Technologies, Vancouver, Canada) following manufacturer's instructions. Flow cytometry experiments were performed on whole blood or freshly isolated PBMC. For whole blood stains, leukocytes were obtained after lysis of red blood cells using ACK buffer (Thermofisher, Waltham, MA) during 5 minutes followed by a wash with R10 media (RPMI-1640 supplemented with 10% FBS, 2 mM L-glutamine,100 U/ml penicillin, and 100 mg/ml streptomycin). After washing with phosphate-buffered saline (PBS), cells (whole blood derived leukocytes or PBMC) were prestained for the chemokine receptor CCR7 for 10 min at 37°C 5% Antibody heavy chain sequencing DNA was extracted from blood using Gentra Puregene Blood Kit (Qiagen). Immunoglobulin heavy-chain family-specific PCRs were performed on genomic DNA samples using primers in FR1 and JH as described previously (5 (6), using a similar protocol described previously (7) . DNA was chosen for this analysis because it provided a parsimonious means of evaluating the B cell repertoire, with one template per cell, and because replicate sequencing libraries could be used to provide rigorous clone size estimates (7) . Briefly, paired reads were assembled using default parameters, sequences that had an average quality score less than 30 were excluded, ends of each read which had an average quality score less than 30 within a window of 20 bases were trimmed, sequences shorter than 100 nucleotides were excluded, and bases with a quality score less than 30 were masked with an N. Sequences with ten or more Ns were then discarded. Sequences were annotated with IgBLAST, (8) and imported into ImmuneDB v0.29.9 (9) for further processing and data visualization. To group related sequences together into clones, ImmuneDB hierarchically clusters sequences with the same VH gene, same JH gene, same CDR3 length, and 85% identity at the amino acid level within the CDR3 sequence (5 Specific color coding was assigned per individual for cross comparison across graphs and Figs. Lines on the graphs indicate the median of the group. Differences between groups were calculated using Kruskal-Wallis test with Dunn's multiple comparison post-test, or Mann-Whitney rank sum test. *p<0.05, ns, not significant. and recovered (n=7) COVID-19+ quantified using a cytometric bead array assay. Specific color coding was assigned per individual for cross comparison across graphs and Figs. Lines on the graphs indicate the median of the group. Differences between groups were calculated using Kruskal-Wallis test with Dun's multiple comparison post-test. ns, not significant. Days Sx Start, days since onset of symptoms accounted from the time of blood draw. Hypoxia Severity: NC, nasal cannula; NIV / HFNC, non-invasive ventilation and/or high flow nasal cannula; ARDS, acute respiratory distress syndrome; ECMO, extracorporeal membrane oxygenation. APACHE, acute physiology and chronic health evaluation. One patient required mechanical ventilation for encephalopathy but did not fulfill ARDS radiographic criteria ("Ventilated non-ARDS"). a Vascular and metabolic disorder category included any of the following: obesity, cardiovascular disease, hypertension, diabetes mellitus and hyperlipidemia. b Underlying pulmonary disorder category included asthma, sarcoidosis, chronic obstructive pulmonary disease or interstitial lung disease. Y, yes. N, no. c Excluded from all reported analyses. Clinical Characteristics of Coronavirus Disease 2019 in China Clinical features of patients infected with 2019 novel coronavirus in Wuhan SARS-CoV-2 productively infects human gut enterocytes Endothelial cell infection and endotheliitis in COVID-19 Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study Clinical predictors of mortality due to COVID-19 based on an analysis of data of 150 patients from Wuhan, China Clinical course and outcomes of critically ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective, observational study Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan The clinical course and its correlated immune status in COVID-19 pneumonia Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan Clinical and immunological features of severe and moderate coronavirus disease 2019 Virological assessment of hospitalized patients with COVID-2019 Breadth of concomitant immune responses prior to patient recovery: a case report of non-severe COVID-19 A single-cell atlas of the peripheral immune response to severe COVID-19. medRxiv High-dimensional immune profiling by mass cytometry revealed immunosuppression and dysfunction of immunity in COVID-19 patients Elevated exhaustion levels and reduced functional diversity of T cells in peripheral blood may predict severe progression in COVID-19 patients Respiratory Pathophysiology of Mechanically Ventilated Patients with COVID-19: A Cohort Study Acute respiratory distress syndrome: the Berlin Definition CD16 BV711 (clone 3G8) and Perforin APC (clone B-D48) from Biolegend 5 (clone N418) from eBioscience; CXCR5 BV750 (clone RF8B2 HLA-DR BV605 (clone G46-6), CD123 PE (clone 9F5) The Live/Dead Fixable Aqua Dead Cell Stain Kit (Invitrogen) was used for viability exclusion, and Human Trustain FcX (Biolegend) was used to prevent unspecific binding Total IgG was measured using a Hu Total IgG CBA Flex Set Bead (BD Biosciences) on plasma or serum samples following manufacturer's protocol ELISA plates (Immulon 4 HBX, Thermo Scientific) were coated with 50 µL per well of recombinant protein diluted in PBS to a final concentration of 2µg/mL and plates were incubated overnight at 4°C. The next day, ELISA plates were washed 3 times with PBS containing 0.1% Tween-20 (PBS-T) and monocytes or NK cells (gated in total CD56+ NK cells), defined by the single expression of CD161+ or co-expression of CD161 and CD38, respectively. F) Spearman correlation of the percentage of CD16+ and expression (MFI of CD16+ cells) and plasma/serum RBD-specific IgG levels in moderate (orange triangles) and severe COVID-19+ individuals (dark red circles). G) MFI of Specific color coding was assigned per individual for cross comparison across graphs and Figs. Lines on the graphs indicate the median of the group. Differences between groups were calculated using Kruskal-Wallis test with Dunn's multiple comparison post-test Acute respiratory distress syndrome: the Berlin Definition SARS-CoV-2 Seroprevalence Among Parturient Women A serological assay to detect SARS-CoV-2 seroconversion in humans. medRxiv SARS-CoV-2 Seroconversion in Humans: A Detailed Protocol for a Serological Assay, Antigen Production, and Test Setup An atlas of B-cell clonal distribution in the human body pRESTO: a toolkit for processing high-throughput sequencing raw reads of lymphocyte receptor repertoires Computational Evaluation of B-Cell Clone Sizes in Bulk Populations IgBLAST: an immunoglobulin variable domain sequence analysis tool ImmuneDB, a Novel Tool for the Analysis, Storage, and Dissemination of Immune Repertoire Sequencing Data MiXCR: software for comprehensive adaptive immunity profiling Binary codes capable of correcting deletions, insertions and reversals or more identity at the amino acid sequence level in the CDR3 (see Materials and Methods).