Distinct cellular immune profiles in the airways and blood of critically ill patients with COVID-19

Abstract

Background: Knowledge of the pathophysiology of COVID-19 is almost exclusively derived from studies that examined the immune response in blood. We here aimed to analyse the pulmonary immune response during severe COVID-19 and to compare this with blood responses.

Methods: This was an observational study in patients with COVID-19 admitted to the intensive care unit (ICU). Mononuclear cells were purified from bronchoalveolar lavage fluid (BALF) and blood, and analysed by spectral flow cytometry; inflammatory mediators were measured in BALF and plasma.

Findings: Paired blood and BALF samples were obtained from 17 patients, four of whom died in the ICU. Macrophages and T cells were the most abundant cells in BALF, with a high percentage of T cells expressing the ƴδ T cell receptor. In the lungs, both CD4 and CD8 T cells were predominantly effector memory cells (87·3% and 83·8%, respectively), and these cells expressed higher levels of the exhaustion marker programmad death-1 than in peripheral blood. Prolonged ICU stay (>14 days) was associated with a reduced proportion of activated T cells in peripheral blood and even more so in BALF. T cell activation in blood, but not in BALF, was higher in fatal COVID-19 cases. Increased levels of inflammatory mediators were more pronounced in BALF than in plasma.

Interpretation: The bronchoalveolar immune response in COVID-19 has a unique local profile that strongly differs from the immune profile in peripheral blood. Fully elucidating COVID-19 pathophysiology will require investigation of the pulmonary immune response.

Keywords: COVID-19; pneumonia; respiratory infection; viral infection.

Figure 1

BALF predominantly comprises T cells and monocyte-derived and alveolar macrophages. PBMCs and BALFMCs isolated from patients with COVID-19 admitted to the ICU were measured using spectral flow cytometry (n=17) with the dotted line separating data obtained from PBMCs and BALFMCs. (A, B) Unsupervised analysis using OMIQ is presented in optSNE plots wherein colours are applied to clusters after manual gating (see online supplemental figures 1,2). CD3+ and CD3− negative cells are depicted in separate optSNE plots. (C, D) Expression of lineage and exhaustion markers are depicted for all cell subsets using heat maps. (E–J) Quantification of general immune cell populations (E), T cell subsets (F) and monocyte/macrophages subsets (G) as well as their HLA-DR expression levels (H) in PBMCs and BALFMCs. HLA-DR expression on antigen presenting cells in BALFMCs (I). Statistical significance was tested after multiple testing correction for differences in cell populations or HLA-DR expression using Kruskal-Wallis (control vs patient PBMCs) or Friedmans test (patient PBMC vs BALFMC) and cytokine levels with Mann-Whitney U test (healthy vs COVID-19) or Wilcoxon signed-rank test (COVID-19 plasma vs BALF). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. αβ, αβ T cell receptor; ƴδ, ƴδ T cell receptor; Alt, alternative; Alv, alveolar; BALFMCs, bronchoalveolar lavage fluid mononuclear cells; cDC, conventional DC; Clas, classical; CM, central memory T cells; DC, dendritic cell; DN, double negative; EM, effector memory T cells; HLA-DR, Human Leukocyte Antigen–-DR isotype; ICU, intensive care unit; ILC, innate lymphoid cells; Int, intermediate; Mo/MQ, monocyte-like macrophage; MQ, macrophage; NK, natural killer cells; PBMCs, peripheral blood mononuclear cells; pDC, plasmacytoid DC; SCM, stem cell-like memory T cells; TEMRA, RA+ effector memory T cells; Treg, regulatory T cells; Trm, tissue-resident memory T cell.

Figure 2

BALF T cells comprise predominantly effector memory CD4 and CD8 T cells and CD8 Trm. PBMCs and BALFMCs isolated from patients with COVID-19 admitted to the ICU were measured using spectral flow cytometry (n=17). T cells were phenotyped (see online supplemental table 3 for all subsets) using CD27, CD28, CCR7 and CD45RA and CD95 in naïve (CD45RA+CD27+CD28+CD95−), stem cell-like memory (SCM; CD45RA+CD27+CD28+CD95+), effector memory-1 (EM1) (CD45RA−CD27+CD28+CCR7−), EM2 (CD45RA−CD27−CD28+CCR7+), EM3 (CD45RA−CD27−CD28+CCR7−), EM4 (CD45RA−CD27−CD28−CCR7−), effector memory Ra⁺ (TEMRA; CD45RA−CD27+CD28+CCR7−), central memory (CM; CD45RA−CD27+CD28+CCR7+), tissue-resident memory (Trm; CD103+CD28−; only for BALFMCs) and regulatory T cells (Treg; CD25+CD127−; only for CD4 T cells) (A, F). Activation (ie, HLA-DR⁺CD38⁺) is presented for different CD4 and CD8 T cell subsets (only for populations with >250 events) (B, F). Representation of PD-1 expression on different T cells subsets (C, G) in PBMC and BALFMC with concomitant quantification of total PD-1 expression (D, H,). Levels of IL-4 (I), IL17-a (J), granzyme B (K), IL-2 (L), IL-7 (M), IL-10 (N) and soluble PD-L1 (O) are presented in plasma and BALF. Box plots represent median±IQR. anti-N, anti-nucleacapsid; anti-RBD, anti-receptor binding domain of spike protein; BALF, bronchoalveolar lavage fluid; DN, double negative; ICU, intensive care unit; PBMC, peripheral blood mononuclear cells; PD-1, programmad death-1; PD-L1, programmed death-ligand 1. Statistical significance was tested with multiple testing correction using Kruskal-Wallis (A, B, E, F: control vs patient PBMC), Friedmans test (A, B, E, F: patient PBMC vs BALFMC), Wilcoxon signed-rank test (D, H, I–O: COVID-19 plasma vs BALF) or Mann-Whitney U test (I–O: healthy vs COVID-19). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Figure 3

Correlations between cell populations in PBMCs and BALFMCs. The bronchoalveolar and systemic immune response were compared by correlating αβ-T cells and CD4/CD8 in PBMC and BALFMC (A) and comparing plasma and BALF cytokine levels (B). Cytokines are ordered in ‘implicated in COVID-19 cytokine storm’ (IL-6, CXCL10/IP-10 and CCL2/MCP-1) and ‘anti-viral responses’ (IFN-α, IFN-β, anti-RBD IgG and anti-N IgG) with box plots displaying median±IQR. All PBMC populations obtained using manual gating were correlated to BALFMC populations using spearman correlation. All significant correlations with a Rho >0.6 are depicted in a circus plot wherein red depicts a positive correlation and blue depicts a negative correlation and line thickness resembles the goodness-of-fit (ie, Rho) (C). The four populations with the strongest correlation are presented in dot plots (D). αβ, αβ T cell receptor; ƴδ, ƴδ T cell receptor; Alt, alternative; Alv, alveolar; BALFMC, bronchoalveolar lavage fluid mononuclear cells; cDC, conventional DC; Clas, classical; CM, central memory T cells; DC, dendritic cell; DN, double negative; EM, effector memory T cells; ILC, innate lymphoid cells; Int, intermediate; Mo/MQ, monocyte-like macrophage; MQ, macrophage; NK, natural killer cells; PBMC, peripheral blood mononuclear cells; pDC, plasmacytoid DC; SCM, stem cell-like memory T cells; TEMRA, RA+ effector memory T cells; Treg, regulatory T cells; Trm, tissue-resident memory T cell.

Figure 4

Reduced T cell activation in both BALF and peripheral blood of patients with an ICU stay of >14 days. samples were stratified based on moment of sampling in ≤14 days (n=8) and >14 days (n=9). Only one sample per patient was included per group. Immune cell population from PBMCs (A) and BALFMCs (C) were clustered using omiq unsupervised clustering and presented as optSNE plots. The ten populations with biggest relative differences (sorted from left to right) are depicted for PBMC (B) and BALFMC (D). cytokine levels in plasma and BALF, as measured using Luminex, were compared ≤14 days and >14 days of ICU stay wherein box plots represent median ±interquartile range (E). Alv, alveolar; BALFMC, bronchoalveolar lavage fluid mononuclear cells; cDC, conventional DC; CM, central memory T cells; DC, dendritic cell; EM, effector memory T cells; ICU, intensive care unit; ILC, innate lymphoid cells; MN, monocyte; Mo/MQ, monocyte-like macrophage; NK, natural killer cells; PBMC, peripheral blood mononuclear cells; pDC, plasmacytoid DC; TEMRA, RA+ effoctor memory T cells; Treg, regulatory T cells; Trm, tissue-resident memory T cell. Statistical significance was tested with multiple testing correction using two-way analysis of variance (ANOVA) (B, D) or Mann-Whitney U test (E). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Figure 5

Contrasting patterns of T cell activation in BALF and peripheral blood between survivors and non-survivors. Samples were stratified based on ICU mortality (n=4 non-survivors (blue) vs n=13 survivors (red)). Only one samples per patient was included in each group N. (A, C) Immune cell population from PBMCs (A) and BALFMCs (C) were clustered using omiq unsupervised clustering and presented as optSNE plots. (B, D) The 10 populations with biggest relative differences (sorted from left to right) are depicted for PBMC (B) and BALFMC (D). Cytokine and antibody levels in plasma and BALF, as measured using Luminex and ELISA, respectively, were compared in fatal and non-fatal COVID-19 cases and presented in box plots as median±IQR (E) BALFMC, bronchoalveolar lavage fluid mononuclear cells; cDC, conventional DC; CM, central memory T cells; DC, dendritic cell; EM, effector memory T cells; ICU, intensive care unit; ILC, innate lymphoid cells; MN, monocyte; Mo/MQ, monocyte-like macrophage; NK, natural killer cells; PBMC, peripheral blood mononuclear cells; pDC, plasmacytoid DC; TEMRA, RA+ effoctor memory T cells; Treg, regulatory T cells; Trm, tissue-resident memory T cell. Statistical significance was tested with multiple testing correction using two-way analysis of variance (ANOVA) (B, D) or Mann-Whitney U test (E). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.