Distinct systemic and mucosal immune responses during acute SARS-CoV-2 infection

Abstract

Coordinated local mucosal and systemic immune responses following severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection either protect against coronavirus disease 2019 (COVID-19) pathologies or fail, leading to severe clinical outcomes. To understand this process, we performed an integrated analysis of SARS-CoV-2 spike-specific antibodies, cytokines, viral load and bacterial communities in paired nasopharyngeal swabs and plasma samples from a cohort of clinically distinct patients with COVID-19 during acute infection. Plasma viral load was associated with systemic inflammatory cytokines that were elevated in severe COVID-19, and also with spike-specific neutralizing antibodies. By contrast, nasopharyngeal viral load correlated with SARS-CoV-2 humoral responses but inversely with interferon responses, the latter associating with protective microbial communities. Potential pathogenic microorganisms, often implicated in secondary respiratory infections, were associated with mucosal inflammation and elevated in severe COVID-19. Our results demonstrate distinct tissue compartmentalization of SARS-CoV-2 immune responses and highlight a role for the nasopharyngeal microbiome in regulating local and systemic immunity that determines COVID-19 clinical outcomes.

Fig. 1. Systemic and mucosal antibody responses in patients with COVID-19.

a–i, Antibodies were measured in the plasma (a–c) of healthy controls (n = 12 donors) and patients with mild-to-moderate (n = 15), severe (n = 11) and critical (n = 23) disease or in the nasopharyngeal compartment (g–i) of healthy controls (n = 10 donors) and patients with mild-to-moderate (n = 10), severe (n = 10) and critical (n = 12 patients) disease using an ELISA-based approach with soluble CoV-2 spike protein (OD and AUC ‘ELISA’) and the ‘S-Flow’ FACS-based approach using cell lines stably expressing surface CoV-2 spike (‘S-Flow’). Heat map of statistically different (P < 0.05) antibody responses between healthy controls and patients with COVID-19 (moderate, severe and critical) in plasma (a) and nasopharyngeal compartment (g). In b and c, and h and i, individual antibodies responses by patient severity are shown. d–f, Pseudovirus neutralization in plasma samples from healthy controls (n = 12) and patients with mild-to-moderate (n = 15), severe (n = 11) and critical (n = 23) COVID-19. d, The percentage of pseudovirus neutralization against the SARS-CoV-2 spike protein was measured by analyzing luciferase-expressing pseudotypes. e, The percentage of patients with pseudotype neutralization above 50%. f, Correlation plots between the pseudotype neutralization (%) and presence of anti-spike IgA or IgG as measured by S-Flow. j–l, SARS-CoV-2 neutralization by nasopharyngeal samples from healthy controls (n = 10) and patients with mild-to-moderate (n = 10), severe (n = 10) and critical (n = 12) COVID-19. j, SARS-CoV-2 virus neutralization was measured using an S-Fuse assay that reads out productive infection by SARS-CoV-2. Neutralizing activity of each sample was expressed as the 50% effective dose (ED₅₀). k, Graph represents the percentage of patients with virus neutralization. l, Correlation plots between the neutralization (ED₅₀) and presence of anti-Spike IgA or IgG as measured by S-Flow. In a and g, P values were determined with a two-tailed Mann–Whitney test between healthy and infected individuals. In b–d and h–j, box-and-whisker plots show the minimum, maximum, interquartile range and the median. P values were determined with the one-sided Kruskal–Wallis test followed by Dunn’s post hoc test for multiple comparisons with Geisser–Greenhouse correction. In f and l, σ represents the Spearman coefficient; *P < 0.05; **P < 0.01; ***P < 0.001. In a and g, z-score scale is indicated, with upregulation shown in orange and downregulation shown in blue. AUC, area under the curve; MFI, mean fluorescence intensity; OD, optical density.

Fig. 2. Heterogeneous systemic and mucosal SARS-CoV-2 antibody responses.

a–f, IgA and IgG were assessed by S-Flow using cell lines stably expressing surface CoV-2 spike protein in plasma of healthy controls (n = 12) and patients with mild-to-moderate (n = 15), severe (n = 11) and critical (n = 23) COVID-19 or in the nasopharyngeal compartment of healthy controls (n = 10) and patients with mild-to-moderate (n = 10), severe (n = 10) and critical (n = 12) COVID-19. a, The percentage of IgA and IgG seroconversion in plasma and ‘naso-conversion’ (percentage of positive samples from the nasopharynx) versus disease severity. b, Correlation plots between the percentage of S-Flow anti-spike IgA binding and the percentage of S-Flow anti-Spike IgG binding in plasma (n = 61) and in nasopharynx (n = 42). c, Correlation plots between plasma and nasopharynx anti-spike antibody responses (n = 41). d, Correlation plot between the percentage of plasma pseudovirus neutralization and nasopharyngeal virus neutralization (ED₅₀; n = 41). e, Representation of antibody conversion among the patients; type A represents naso-positive and seropositive patients, type B represents naso-negative and seropositive patients, type C represents naso-positive and seronegative patients and type D represents naso-negative and seronegative patients. f, Representation of anti-spike responses in the different compartment for each patient type. In b, c and d, σ represents the Spearman coefficient.

Fig. 3. Systemic and mucosal cytokine production in patients with COVID-19.

Cytokines were measured in the plasma (a and b) of healthy controls (n = 12 donors) and in patients with mild-to-moderate (n = 15), severe (n = 11) and critical (n = 23) disease or in the nasopharyngeal compartment (c and d) of healthy controls (n = 10 donors) and in patients with mild-to-moderate (n = 10), severe (n = 10) and critical (n = 12) disease using a bead-based multiplexed immunoassay system, Luminex or the digital Simoa ELISA (IFN-α, IFN-β, IFN-γ, IL-6, IL-17A, IL-10 and TNF). a,c, Heat maps of statistically different cytokines (P < 0.05) between healthy controls and patients with COVID-19 (moderate, severe and critical), ordered by hierarchical clustering. Upregulated cytokines are shown in orange and downregulated in blue. b,d, Individual cytokine concentration plots by patient severity. e, Correlation plots between CCL2 concentrations in plasma and nasopharyngeal paired samples; n = 42. σ represents the Spearman coefficient. f, Heat map of statistically different cytokines and antibodies (P < 0.05) in patients having nasopharyngeal spike-specific antibodies (type A and type C) as compared with those lacking these antibodies (type B and type D). In a, c and f, z-score scale is indicated, with upregulation shown in orange and downregulation shown in blue. P values were determined with a two-tailed Mann–Whitney test between healthy and infected individuals. In b and d, box plots show the median ± minimum to maximum values. P values were determined with the Kruskal–Wallis test followed by Dunn’s post hoc test for multiple comparisons. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Fig. 4. SARS-CoV-2 antiviral immune responses are distinct locally and systemically.

a, Plasma viral loads evaluated by digital PCR and in nasopharyngeal swabs estimated by RT–PCR and expressed as relative copies (cp) per ml; n = 61 (left) and n = 42 (right). b, Correlation plots between viral load the in the plasma versus nasopharyngeal compartment. c,e, MDS projection for plasma compartment (cytokines, antibodies and blood viral load; c) and nasopharyngeal compartment (cytokines, antibodies and nasal viral load; e). The dotted lines represent the most associated analytes. d,f, Individual correlation plots between viral load and cytokines or antibodies. In a, box plots show the median ± minimum to maximum values. P values were determined with the Kruskal–Wallis test followed with Dunn’s post hoc test for multiple comparisons. In b (n = 42), d (n = 61) and f (n = 42), σ represents the Spearman coefficient. *P < 0.05; **P < 0.01; ***P < 0.001.

Fig. 5. Perturbations of nasopharyngeal 16S rRNA profiles in patients with COVID-19.

a–g, Nasopharyngeal bacterial communities were measured in healthy controls (n = 10) and in patients with mild-to-moderate (n = 10), severe (n = 10) and critical (n = 12) COVID-19. a, The percentage of relative abundance at the genus level. b, Shannon and Simpson diversity indices by patient severity. Data are presented as box plots with median ± minimum to maximum. c, Principal-component analysis of 16S bacterial profiles. d, Heat map of statistically different (P < 0.05) genus abundance between healthy controls and patients with COVID-19 (moderate, severe and critical). e, Heat map of statistically different (P < 0.05) genus abundance between patients with COVID-19 depending on disease severity. P values were determined with a two-tailed Mann–Whitney test. f,g, Plots showing the percentage of individual genus abundance by disease severity. In b, f and g, box-and-whisker plots show the minimum and maximum values, interquartile range and the median. Corynebacterium (critical versus healthy, P = 6.9 × 10⁻³; critical versus moderate, P = 2.6 × 10⁻³), Acinetobacter (critical versus moderate, P = 3.2 × 10⁻²), Cutibacterium (critical versus moderate, P = 6.6 × 10⁻³), Staphylococcus (critical versus healthy, P = 9.0 × 10⁻³), Peptostreptococcus (critical versus healthy, P = 3.3 × 10⁻²; critical versus moderate, P = 3.3 × 10⁻²). P values were determined with the one-sided Kruskal–Wallis test followed by Dunn’s post hoc test for multiple comparisons with Geisser–Greenhouse correction; *P < 0.05; **P < 0.01; ***P < 0.001. In e, z-score scale is indicated, with upregulation shown in orange and downregulation shown in blue.

Fig. 6. Nasal microbiome influences local mucosal and systemic immune responses in patients with COVID-19.

a,c, DS projection for the nasopharyngeal compartment (cytokines, antibodies, neutralization, nasal viral load and nasal microbiome; a) and the plasma compartment (cytokines, antibodies, pseudoneutralization, blood viral load and nasal microbiome; c). The dotted lines represent the most associated analytes. b,d, Plots show individual correlations between the percentage of genus abundance and cytokines, antibodies or viral load. In b (n = 42) and d (n = 61), σ represents the Spearman coefficient. *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 1. Systemic and mucosal antibody responses in patients with COVID-19.

(a, b) Gating strategy and representative results for the S-Flow assay. (c) Correlation plots between the anti-Spike IgA or IgG OD ELISA and the S-Flow anti-Spike IgA or IgG binding (%) in plasma. (d) Total IgM, IgG and IgA and IgG1/2/3/4 were measured in plasma using a bead-based multiplexed immunoassay system Luminex. (e) Correlation plots between the anti-Spike IgA or IgG OD ELISA and the S-Flow anti-Spike IgA or IgG binding (%) in the nasopharyngeal compartment. (f) Total IgM, IgG and IgA and IgG1/2/3/4 were measured in nasopharyngeal compartment. (a-f) Antibodies were measured in the plasma of healthy controls (n = 12 donors), mild to moderate (n = 15 patients), severe (n = 11 patients) and critical (n = 23 patients) or in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). Nasopharyngeal IgA (Critical vs Healthy, p = 2.9 × 10-2; Critical vs Moderate, p = 4.0 × 10-2). In (c) and (e), σ represents Spearman coefficient and p the p value. In (d) and (f), box-and-whisker plots showing the minimum, maximum, interquartile range and the median. P values were determined with one-sided Kruskal-Wallis test followed by with Dunn’s post-test for multiple group comparisons with Geisser-Greenhouse correction; For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 2. Compartmentalized spike-specific antibody responses in COVID-19.

(a) Correlation plots between the anti-Spike IgA OD ELISA and anti-Spike IgG OD ELISA in plasma and nasopharynx. (b) Correlation plots between the anti-Spike IgA or IgG D.O. ELISA in plasma versus nasopharynx. (c) Correlation plots between plasma S-Flow anti-Spike IgG binding (%) versus anti-Spike IgG OD ELISA and nasopharyngeal S-Flow anti-Spike IgA binding (%) versus anti-Spike IgA OD ELISA. (d) Correlation plots between plasma S-Flow anti-Spike IgA binding (%) versus anti-Spike IgA OD ELISA and nasopharyngeal S-Flow anti-Spike IgG binding (%) versus anti-Spike IgG OD ELISA. (e) Heatmap representation of all antibodies measured in plasma and nasopharyngeal compartment. (a-d) Antibodies were measured in the plasma of healthy controls (n = 12 donors), mild to moderate (n = 15 patients), severe (n = 11 patients) and critical (n = 23 patients) or in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). In (a, b, c, d), σ represents Spearman coefficient and p the p value. For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 3. Systemic and mucosal cytokines production in COVID-19 patients.

(a) total protein (μg) content in nasopharyngeal samples. (b) MUC5AC content in nasopharyngeal samples. (c) Heatmap representation of statistically different (P < 0.05) plasma cytokines between critical COVID-19 patients and mild to moderate and severe COVID-19 patients. (d) Plasma cytokine concentration plots by patient severity. (e) Heatmap representation of statistically different (P < 0.05) nasopharyngeal cytokines between critical COVID-19 patients and mild to moderate and severe COVID-19 patients. (f) Nasopharyngeal cytokine concentration plots by patient severity. (a-f) Cytokines were measured in the plasma of healthy controls (n = 12 donors), mild to moderate (n = 15 patients), severe (n = 11 patients) and critical (n = 23 patients) or in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). In (c) and (e), P values were determined with the Mann-Whitney test. In (a), (b), (d) and (f), box plots with median ± minimum to maximum. P values were determined with the one-sided Kruskal-Wallis test followed by with Dunn’s post-test for multiple group comparisons with Geisser-Greenhouse correction. For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 4. SARS-CoV-2 anti-viral immune responses are distinct locally and systemically.

Correlation matrices between the (a) systemic compartment (plasma cytokines, plasma antibodies, serum virus neutralization, blood viral load) and (b) nasopharyngeal compartment (nasal cytokines, nasal antibodies, nasal virus neutralization, nasal viral load).

Extended Data Fig. 5. Impact of healthy donors on analysis of integrated anti-viral immune responses.

Multidimensional scaling (MDS) projection for (a) systemic compartment (plasma cytokines, plasma antibodies, serum virus neutralization, blood viral load) and (c) nasopharyngeal compartment (nasal cytokines, nasal antibodies, nasal virus neutralization, nasal viral load). (b) and (d) Cytokines and viral load were measured in the plasma of healthy controls (n = 12 donors), mild to moderate (n = 15 patients), severe (n = 11 patients) and critical (n = 23 patients) or in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). (b) and (d) show individual correlation plots between viral load and cytokines or antibodies. In (b) and (d), σ represents Spearman coefficient and p the p value. For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 6. Perturbations of nasopharyngeal 16 S rRNA profiles in COVID-19 patients.

PCA and PERMANOVA test analysis of 16 S bacterial profiles of healthy versus critical COVID-19 patients. PCA analysis of 16 S bacterial profiles color coded by smoking status (b) and sex (c). (D) Non-metric multidimensional scaling (MDS) of 16 S bacterial profiles. (e) Partial least squares-discriminant analysis of 16 S bacterial profiles (f) Bacterial load (16 S rRNA) plotted by patient severity. Box plots with median ± minimum to maximum. (g) Individual correlation plots between Genus abundance (%) for ‘cornerstone’ and ‘pathobionts’. (f) and (g) Bacterial communities were measured in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). In (g), σ represents Spearman coefficient and p the p value. In (F), P values were determined with the one-sided Kruskal-Wallis test followed by with Dunn’s post-test for multiple group comparisons with Geisser-Greenhouse correction. For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 7. 16 S rRNA microbiome analysis in COVID-19 patients using DADA2 pipeline.

Nasopharyngeal bacterial communities were measured in healthy controls (n = 10), mild to moderate (n = 10), severe (n = 10) and critical (n = 12) COVID-19 patients. (a) Relative abundance (%) at the Genus level. (b) Shannon and Simpson diversity indices by patient severity. Data are presented as mean values + /- SEM. Simpson diversity index (Severe vs Moderate, p = 2.1 × 10-2), Shannon diversity index (Severe vs Moderate, p = 4.1 × 10-2), (c) PCA analysis of 16 S bacterial profiles (d) Heatmap representation of statistically different (P < 0.05) Genus abundance between healthy controls and COVID-19 patients (moderate, severe, critical). P values were determined with the Mann-Whitney test. (e) and (f) Individual Genus abundance (%) plots by disease severity. For all plots, bacterial communities were measured in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). In (b), (e) and (f), box-and-whisker plots showing the minimum, maximum, interquartile range and the median. Staphylococcus (Critical vs Healthy, p = 6.5 × 10-3), Prevotella (Critical vs Moderate, p = 2.7 × 10-2; Critical vs Severe, p = 2.5 × 10-2). P values were determined with the one-sided Kruskal-Wallis test followed by with Dunn’s post-test for multiple group comparisons with Geisser-Greenhouse correction; For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 8. Microbiome regulates mucosal cytokines and antibody responses.

Correlation matrices between the (a) nasopharyngeal compartment (nasal cytokines, nasal antibodies, nasal virus neutralization, nasal viral load and nasal microbiome) and (b) systemic compartment (plasma cytokines, plasma antibodies, serum virus neutralization, blood viral load and nasal microbiome).

Extended Data Fig. 9. Impact of healthy donors on microbiome analysis of integrated SARS-CoV-2 immune responses.

Multidimensional scaling (MDS) projection for (a) nasopharyngeal compartment (nasal cytokines, nasal antibodies, nasal virus neutralization, nasal viral load and nasal microbiome) and (c) systemic compartment (plasma cytokines, plasma antibodies, serum virus neutralization, blood viral load and nasal microbiome). The dotted lines represent the most associated measures. (b) and (d) show individual correlation plots between Genus abundance (%) and cytokines and specific antibodies. (b) and (d) Cytokines and viral load were measured in the plasma of healthy controls (n = 12 donors), mild to moderate (n = 15 patients), severe (n = 11 patients) and critical (n = 23 patients) or in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients). In (b) and (d), σ represents Spearman coefficient and p the p value. For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.

Extended Data Fig. 10. Influence of age on integrated SARS-CoV-2 immune responses.

(a) Age distribution of the cohort based on disease severity. Box plots with median ± minimum to maximum. (b) Correlation plots between age and nasal S-Flow anti-Spike IgG or IgA binding (%). (c) Correlation plots between age and nasopharyngeal microbiota Shannon and Simpson diversity and Staphylococcus genus abundance (%). For all plots, analytes were measured in the plasma of healthy controls (n = 12 donors), mild to moderate (n = 15 patients), severe (n = 11 patients) and critical (n = 23 patients) or in the nasopharyngeal compartment of healthy controls (n = 10 donors), mild to moderate (n = 10 patients), severe (n = 10 patients) and critical (n = 12 patients).In (b, c), σ represents Spearman coefficient and p the p value. For all panels: *P < 0.05; **P < 0.01; ***P < 0.001.