key: cord-0745240-q14igi1k
authors: Liechti, Thomas; Iftikhar, Yaser; Mangino, Massimo; Beddall, Margaret; Goss, Charles W.; O’Halloran, Jane A.; Mudd, Philip; Roederer, Mario
title: Immune phenotypes that predict COVID-19 severity
date: 2022-03-10
journal: Res Sq
DOI: 10.21203/rs.3.rs-1378671/v1
sha: bc0dd15e57737960013df19830eb6a447e4c1139
doc_id: 745240
cord_uid: q14igi1k

Severe COVID-19 causes profound immune perturbations, but pre-infection immune signatures contributing to severe COVID-19 remain unknown. Genome-wide association studies (GWAS) identified strong associations between severe disease and several chemokine receptors and molecules from the type I interferon pathway. Here, we define immune signatures associated with severe COVID-19 using high-dimensional flow cytometry. We measured the peripheral immune system from individuals who recovered from mild, moderate, severe or critical COVID-19 and focused only on those immune signatures returning to steady-state. Individuals that suffered from severe COVID-19 showed reduced frequencies of T cell, MAIT cell and dendritic cell (DCs) subsets and altered chemokine receptor expression on several subsets, such as reduced levels of CCR1 and CCR2 on monocyte subsets. Furthermore, we found reduced frequencies of type I interferon-producing plasmacytoid DCs and altered IFNAR2 expression on several myeloid cells in individuals recovered from severe COVID-19. Thus, these data identify potential immune mechanisms contributing to severe COVID-19.

The recent COVID-19 pandemic caused an unprecedented global health crisis. 41

Demographic and socioeconomical factors affect disease severity and mortality (1). 42

Underlying health conditions such obesity and diabetes or gender with higher risk for 43 males have been associated with disease severity (1). Additionally, genetic predisposition 44 contributes to the development of severe COVID-19 (2, 3). GWAS identified several 45 genes encoding for pro-inflammatory chemokine receptors and molecules from the type 46 I interferon pathway, such as OAS1, DPP9, TYK2 and IFNAR2, that associate with the 47 development of severe COVID-19 (2, 3). Thus, tissue distribution of immune cells and the 48 responsiveness of innate immunity to infection may be key factors to prevent severe 49 outcome in COVID-19. While GWAS enable the identification of associations between 50 genetic variants and disease severity, such studies fall short of providing insights into the 51 mechanisms by which these genetic traits manifest disease susceptibility. Nearly all of 52 the SNPs identified in GWAS are regulatory and not coding in nature; the altered 53 regulation could be expressed on subsets of immune cells rather than organism-wide. 54

Thus, immunological studies such as immunophenotyping at the single cell level are 55 necessary to gain mechanistic understanding of how genetics affect immune responses 56

(4). 57 58 Chemokine receptors are crucial in regulating leukocyte trafficking and thereby 59 orchestrating immune responses (5, 6). Thus, chemokine receptors are critical in all 60 baseline levels. We first aimed to identify these persisting immune perturbations which 127 may contribute to long-lasting COVID19-related symptoms known as long COVID (19) . 128

To this end, we analyzed PBMC collected after recovery from mild, moderate, severe and 129 critical COVID-19 (Extended Data Fig. 1 and 2) . We focused on the moderate and severe 130 COVID-19 group as these groups showed the largest time range between symptom onset 131 and sample collection (Extended Data Fig. 1a ; Moderate, 24-129 days; Severe, 16-184 132 days). We applied two strategies to identify persistently affected immune traits. These 133 included i) linear regression of immune traits and time between symptom onset and 134 sample collection, and ii) comparison of samples collected before and after 60 days of 135 symptom onset using a Wilcoxon test. We opted to abstain from multiple testing correction 136 in order to avoid the inclusion of marginally significant true positive immune traits (i.e. 137

immune traits which truly change over time) in our analysis of stable immune traits. The 138 two strategies showed similar results (Fig. 1b) . We assessed the top hits from both 139 analyses (p<0.001 in at least one analysis, N = 24) to further delineate persistent immune 140 perturbations in COVID-19 (Fig. 2a) . 141

The most prominent persisting perturbations occurred within switched (containing 143 memory B cells and plasmablasts) and memory CD20 + IgD -CD38 -/+ CD27 -/+ B cells (Fig.  144 2a). Switched and naïve B cells did not change in moderate COVID-19 over time but 145 significantly decreased and increased, respectively, in severe cases (Spearman's rank 146 correlation; Naïve: R 2 = 0.36, P = 0.002; Switched: R 2 = 0.44, P = 4*10 -4 ) to levels 147 observed in healthy individuals (Fig. 2b) . Both naïve and switched B cells did not differ 148 between study groups (Fig. 2b) . Similar dynamics occurred for CD38 + HLA-DRand CD38 -149 HLA-DR -CD4 naïve T cells which showed an increase and decrease, respectively, over 150 time in the severe COVID-19 group (Spearman's rank correlation; CD38 + HLA-DR -CD4 151 naïve: R 2 = 0.42, P = 5.8*10 -4 ; CD38 -HLA-DR -CD4 naïve: R 2 = 0.42, P < 6.5*10 -4 ) with 152 later timepoints reaching levels observed in healthy individuals (Fig. 2c ). In addition, 153 decreased CD38 + HLA-DRand increased CD38 -HLA-DR -CD4 naïve T cells occurred in 154 individuals recovered from severe and critical COVID-19 (Bonferroni-adjusted P-value 155 range 0.02 -1.46*10 -4 ) (Fig. 2c) . 156

Cross-presenting cDC1s induce potent CD8 T cell responses. Timepoints early after 158 onset of symptoms had reduced levels of cDC1s in severe COVID-19 cases, but these 159 increased later to levels observed in healthy individuals, suggesting perturbations of 160 severe and critical COVID-19 (Bonferroni-adjusted P-value range 0.01 -7.21*10 -6 ) ( Fig.  220 3d). In contrast, naïve CD8 + T cells and MAIT cells expressed similar levels between 221 study groups or elevated levels of TIGIT in individuals suffered from severe and critical 222 COVID-19, respectively (Bonferroni-adjusted P-value range 0.04 -0.004). 223

The frequency of MAIT cells was decreased in severe and critical COVID-19 (Extended 225 Data Fig. 4e ) (Bonferroni-adjusted P-value range 3.99*10 -4 -2.96*10 -6 ). Furthermore, 226 more individuals recovered from severe and critical COVID-19 showed reduced 227 frequencies of central memory (CM) CD4 + and CD8 + T cells (defined as CD45RA -228 CCR7 + CD27 + ) (Bonferroni-adjusted P-value range 0.02 -7.61*10 -6 ) (Extended Data Fig.  229 4f). In addition, individuals recovered from critical COVID-19 had elevated levels of 230 activated (defined as CD38 + HLA-DR + ) CD8 + effector and terminal memory T cells 231

(Bonferroni-adjusted P-value range 5.46*10 -4 -1.04*10 -6 ) (Fig. 3e) . DCs (pDCs) and CD14 + DC3s (Bonferroni-adjusted P-value range 0.00226 -3.71*10 -7 ) 246 ( Fig. 4a) . 247

The chemokine receptor profile on dendritic cells did not differ substantially between 249 individuals recovered from non-severe and severe COVID-19. We observed increased 250 expression of CX3CR1 on pDCs and cross-presenting cDC1s associated with disease 251 severity (Bonferroni-adjusted P-value range 0.00301 -1.83*10 -5 ) (Fig. 4b) . Frequency of 252 monocyte subsets did not differ between groups. However, classical and intermediate 253 monocytes from individuals recovered from severe COVID-19 had reduced expression of 254 pro-inflammatory chemokine receptors CCR1 and CCR2 (Bonferroni-adjusted P-value 255 range 0.04 -2.07*10 -9 ) (Fig. 4c) . In contrast, non-classical pro-inflammatory monocytes 256 showed no differences of CCR1 and CCR2 expression between COVID-19 severity 257 groups (Fig. 4c) . 258 259 Genome-wide association studies identified IFNAR2 as a risk factor for severe COVID-260 19 (2, 3). Furthermore, type I interferon response is critical for effective immune 261 responses against 11, 13, 14) . We measured expression of IFNAR2 on 262

and highest on pDCs and cDC1s, but expression could be detected on most subsets 264 including cDC2s, DC3s and monocyte subsets (Fig. 1a) . We found increased expression 265 of IFNAR2 on monocyte and dendritic cell subsets, except for cDC1s and pDCs, in (Bonferroni-adjusted P-value range 0.00564 -6.39*10 -8 ) and characterized by CCR9, 296 CXCR3 and TIGIT expression (Fig. 5d ). In contrast V 2V 9 T cell cluster 24 expressed 297 higher levels of CCR4 and CCR8 but lacked CXCR3 and TIGIT (Fig. 5d) . 298

Within myeloid cells, CD123 + CD5 -pDCs (CR1: 24, CR2: 28) and CD123 + CD5 + pre-DCs 300 (CR1: 28, CR2: 29) were significantly reduced (Bonferroni-adjusted P-value range 0.04 -301 4.11*10 -6 ) in individuals recovered from severe and critical COVID-19. These cells were 302 characterized by expression of CD38, CCR5 and high levels of CXCR3. CCR1, CCR2 303 and IFNAR2 were expressed at higher levels on pDCs while co-stimulatory CD86 was 304 lower and CD40 expression was lacking on both. CD14 -DC3s (CR1: 22; CR2: 21) and 305 CD14 + DC3s (CR1: 11) differ between individuals recovered from non-severe and severe 306 COVID-19 (Figs. 6a and b) in agreement with our manual analysis ( Fig. 4a and Extended 307 Data Fig. 3) . 308

We further examined myeloid cells from mild and severe COVID-19 cases using tSNE. On the contrary, we observed increased expression of IFNAR2 on basophils and myeloid 382 cells but not on B cells and pDCs in individuals recovered from severe COVID-19 (Fig.  383 5d). This is in contradiction with inferences from a recent study which combined GWAS 384 and bulk transcriptomics and identified reduced expression of IFNAR2 in lung and whole 385 blood as a risk factor for severe COVID-19 (2). In contrast to bulk transcriptomics, we 386

show at the single-cell level that IFNAR2 is only affected on certain blood immune cell 387 populations in individuals recovered from severe COVID-19. Notably, we measured 388 IFNAR2 only on B cells, basophils and myeloid cells and can therefore not determine 389 whether its expression is downregulated in other blood cell types. The dichotomy between 390 reduced pDC frequencies and elevated IFNAR2 expression on myeloid cells is puzzling. 391

However, interaction between type I interferon and its receptor results in endocytosis (42) 392 and it is therefore possible that constitutively expressed type I interferon might regulate 393 study, the majority of immune traits were at baseline in recovered patients (Fig. 1b) . 400

Nonetheless, we identified several immune traits which did not fully return to baseline 401 even weeks after symptom onset (Figs. 1b and 2). Most of these long-term perturbations 402 occurred in severe COVID-19, likely due to increased immune activation (24), and 403 affected mainly B and T cells (Fig 2) . The half-life of peripheral lymphocytes is longer 404 Table 3) . 

from each severity group were concatenated prior to tSNE analysis (perplexity = 30, theta 507 = 0.5, 5000 iterations) in order to maintain priority for tSNE computation equal among 508 patient groups. Fewer cells were used for TSNE in the case of innate-like T cells due to 509 limited numbers of cells in the critical COVID-19 group (total 27583 cells from all patients). 510

We expected lower diversity of NK cell subsets and therefore used 25000 cells per study 511 group for tSNE. 512 

We distinguished immune traits which were affected by long-term immune perturbations 529 or at steady-state within moderate and severe COVID-19 group. We focused on these 530 two study groups because they span across the longest period between symptom onset 531 and sample collection enabling the most precise analysis of long-term immune 532 trajectories after symptom onset (Extended Data Fig. 1a ). Of note, age correlated with 533 hospitalization length in severe but not critical cases and was significantly shorter in 534 severe COVID-19 cases (Extended Data Figs. 1b and c) . We used linear regression 535 between rank-normalized immune traits derived from both unsupervised clustering and 536 manual analysis and length of time in days between symptom onset and sample 537

collection. In addition, we compared immune traits in samples with less or more than 60 538 days between symptom onset and sample collection using Wilcoxon signed-rank test. 539

Long-term perturbated traits were defined as manually defined immune traits with 540 unadjusted P < 0.001 in at least one of the analyses (N = 24) . 541 542

Immune traits and FlowSOM clusters with unadjusted P > 0.05 in both analyses described 544 above (linear regression and Wilcoxon signed-rank test) were defined as stable immune 545 traits at steady-state (1365 manually defined immune traits and 291 FlowSOM clusters) 546

and were used to predict immune signatures associated with the development of severe 547 COVID-19. We rank-normalized the data and used logistic regression between 548 mild/moderate (group non-severe) and severe/critical (group severe) cases and corrected 549 for age and experiment (batch). P-values were adjusted using Benjamini-Hochberg false 550 discovery rate (56) and adjusted P-values < 0.05 were considered statistically significant. Residuals from linear regression between immune trait and age were used to calculate 971 statistics on age-corrected data. ANOVA with subsequent Wilcoxon test and Bonferroni 972 correction on residuals was performed for statistics highlighted in boxplots. * P < 0.05, ** 973 P < 0.01, *** P < 0.001, **** P < 0.0001 974 

Association of Social and Demographic 611 Factors With COVID-19 Incidence and Death Rates in the US

Genetic mechanisms of critical illness in COVID-19

Mapping the human genetic architecture of COVID-19

The genetic architecture of the human immune system: 631 a bioresource for autoimmunity and disease pathogenesis

Understanding the mechanisms that facilitate specificity, not 634 redundancy, of chemokine-mediated leukocyte recruitment

Chemokines and chemokine receptors: 637 positioning cells for host defense and immunity

The chemokine system in innate immunity

The many roles of chemokines and chemokine 642 receptors in inflammation

Innate immunity to influenza virus infection

COVID-19 and the human innate immune 646 system

Pathological inflammation in patients with COVID-19: a key 648 role for monocytes and macrophages

Human IFNAR2 653 deficiency: Lessons for antiviral immunity

Type I and Type III Interferons -Induction, Signaling, Evasion, 655 and Application to Combat COVID-19

Impaired type I interferon activity 661 and inflammatory responses in severe COVID-19 patients

The interferon landscape along the respiratory tract impacts the severity 668 of COVID-19

Milieu 684 Interieur, V. C. C. Co

Autoantibodies against type I IFNs in patients with life-688 threatening COVID-19

Inborn errors of type I IFN immunity in patients with life-712 threatening COVID-19

Production of type I interferons: plasmacytoid 714 dendritic cells and beyond

Long-term perturbation of the peripheral immune system months 719 after SARS-CoV-2 infection. medRxiv

Longitudinal analysis reveals 728 that delayed bystander CD8+ T cell activation and early immune pathology 729 distinguish severe COVID-19 from mild disease

Control of human viral infections by natural killer cells

Natural killer cell immunotypes related to COVID-19 disease severity

740 23. A. Cerutti, M. Cols, I. Puga, Marginal zone B cells: virtues of innate-like antibody-741 producing lymphocytes

Longitudinal analyses reveal immunological misfiring in 749 severe COVID-19

Single-Cell Analysis of 755 Human Mononuclear Phagocytes Reveals Subset-Defining Markers and Identifies 756 Circulating Inflammatory Dendritic Cells

Comprehensive Phenotyping of Human Dendritic Cells and 758 Monocytes

COVID-19 severity correlates with 765 airway epithelium-immune cell interactions identified by single-cell analysis

Discriminating Mild from Critical COVID-19 by 772 Innate and Adaptive Immune Single-cell Profiling of Bronchoalveolar Lavages

Human immune system variation

777 Chemokine receptor CCR5 promotes leukocyte trafficking to the brain and survival 778 in West Nile virus infection

The chemokine receptor CCR5 plays a key role in the 781 early memory CD8+ T cell response to respiratory virus infections

CX3CR1 is required for airway inflammation by promoting T helper cell 786 survival and maintenance in inflamed lung

Supplementary Data 5: Gating of several differentiation stages within innate-like T

Supplementary Data 8: FlowSOM analysis for B cells

Only markers 86 included in clustering (columns) are sown. Bar on left shows coloring of each FlowSOM 87 cluster and FlowSOM clusters were clustered based on similarity of MFI values using 88 hierarchical clustering (indicated by dendrogram and gap between rows). Bar graph in the middle shows the frequency of each cluster and bar graph on the right the 90 composition of each cluster based on manual gating annotation

CR1 and b) CR2 are shown. c) tSNE plots for CR1 (top) and CR2 (bottom) panel are 92 shown delineated based on COVID-19 severity group

Each tSNE plot contains 50'000 randomly subsampled 94 cells and not equally distributed across each individual sample

Residuals from linear regression between immune trait and age were used to calculate 1004 statistics on age-corrected data. ANOVA with subsequent Wilcoxon test and Bonferroni 1005 correction on residuals was performed for statistics highlighted in boxplots. * P < 0.05, ** 1006 

This is a list of supplementary les associated with this preprint. Click to download.220218Supplementarytables.pdf