Autoantibodies against chemokines post-SARS-CoV-2 infection correlate with disease course

Abstract

Infection with severe acute respiratory syndrome coronavirus 2 associates with diverse symptoms, which can persist for months. While antiviral antibodies are protective, those targeting interferons and other immune factors are associated with adverse coronavirus disease 2019 (COVID-19) outcomes. Here we discovered that antibodies against specific chemokines were omnipresent post-COVID-19, were associated with favorable disease outcome and negatively correlated with the development of long COVID at 1 yr post-infection. Chemokine antibodies were also present in HIV-1 infection and autoimmune disorders, but they targeted different chemokines compared with COVID-19. Monoclonal antibodies derived from COVID-19 convalescents that bound to the chemokine N-loop impaired cell migration. Given the role of chemokines in orchestrating immune cell trafficking, naturally arising chemokine antibodies may modulate the inflammatory response and thus bear therapeutic potential.

Fig. 1. Distinct patterns of chemokine antibodies in COVID-19 convalescents with different severity of acute disease.

a, Heatmap representing plasma IgG binding to 42 peptides comprising the N-loop of all 43 human chemokines, as determined by ELISA in healthy controls (Controls) and COVID-19 convalescents (COVID-19) of the Lugano cohort at month 6. Samples are ranked according to the level of SARS-CoV-2 RBD reactivity. Chemokine IgGs are ordered by unsupervised clustering analysis of ELISA signal. SARS-CoV-2 pseudovirus neutralizing activity (NT₅₀) and IgG binding to peptides corresponding to negative control, IFN-α2 and SARS-CoV-2 nucleocapsid protein (N) are shown. b, AUC of ELISA showing IgG antibodies to CCL19, CCL22 and CXCL17 (COVID-19 signature) in healthy controls and COVID-19 convalescents at month 6. Two-tailed Mann–Whitney U-test. c, Logistic regression analysis showing the assignment of COVID-19 convalescents and healthy controls based on CCL19, CCL22 and CXCL17 antibodies at month 6. d, AUC of ELISA showing CCL19, CCL22 and CXCL17 antibodies at months 6 and 12 in COVID-19 convalescents (n = 63). Wilcoxon two-tailed signed-rank test. e, Chemokine antibodies in previously hospitalized and outpatient COVID-19 convalescents at month 6, shown as ratio over healthy controls. Circle size indicates significance; colors show the log₂ fold-change increase (red) or decrease (blue), shown as ratio over healthy controls. Kruskal–Wallis test followed by Dunn’s multiple comparison test. f, Cumulative AUC of ELISA signal of the IgGs against the 42 chemokine N-loops in healthy controls and previously hospitalized and outpatient COVID-19 convalescents at month 6. Kruskal–Wallis test followed by Dunn’s multiple comparison test. g, t-SNE distribution of previously hospitalized and outpatient COVID-19 convalescents at month 6, as determined with the 42 datasets combined. h, Logistic regression analysis showing the assignment of previously hospitalized and outpatient COVID-19 convalescents based on CXCL5, CXCL8 and CCL25 antibodies (COVID-19 hospitalization signature) at month 6. In b and f, horizontal bars indicate median values. In a–h, AUC values are the average from two independent experiments. Healthy controls (n = 23) in a, b, c, e and f; COVID-19 convalescents (n = 71) in a, b and c, of which previously hospitalized (n = 50) and outpatient (n = 21) in e–h. Ab, antibody; AUC, area under the curve; FC, fold-change; ID, identity; m, months; NS, not significant. Source data

Fig. 2. Autoantibodies against specific chemokines in COVID-19 convalescents without persistent symptoms at month 12.

a, Chemokine IgG (cumulative AUC of ELISA; left), RBD IgG (middle) and NT₅₀ (right) values in healthy controls (Controls) and COVID-19 convalescents (COVID-19) of the Lugano cohort at month 6 grouped as long COVID and no long COVID at month 12. Kruskal–Wallis test followed by Dunn’s multiple comparison test. b, AUC of ELISA showing CCL21, CXCL13 and CXCL16 antibodies (long COVID signature) at month 6 in COVID-19 convalescents defined as long COVID and no long COVID at month 12. Two-tailed Mann–Whitney U-tests. c, Logistic regression analysis showing the assignment of COVID-19 convalescents as long COVID and no long COVID at month 12, based on CCL21, CXCL13 and CXCL16 antibodies at month 6. d, ELISA showing CXCL16 antibodies binding to the CXCL16 N-loop peptide. Average of two independent experiments. e, Chemotaxis showing relative migration of the 300.19 preB cell line uniquely expressing CXCR6 in a CXCL16 gradient (1 nM). Mean + s.e.m. of three independent experiments. Paired, two-tailed Student’s t-test. f, ELISA showing CXCL13 antibodies binding to the CXCL13 N-loop peptide. Average of four independent experiments (mean + s.e.m.). g, Chemotaxis of primary CD19⁺ human B cells isolated from buffy coats in a CXCL13 gradient in the presence of the aCXCL13.001 antibody or isotype control. Mean ± s.e.m. of migrated cells in five high-power fields (HPFs). Average of three independent experiments with cells from different donors. Two-way repeated measures ANOVA followed by Šídák’s multiple comparisons test. In a and b, horizontal bars indicate median values and data are shown as average AUC from two independent experiments. Healthy controls (n = 23) in a; COVID-19 convalescents without (n = 22) or with (n = 41) long COVID at month 12 in a, b and c. OD, optical density. Source data

Fig. 3. Concentration of plasma chemokines during acute COVID-19 and in convalescence.

a, Plasma chemokine levels in the Milan (n = 44; acute and month 7) and Lugano (n = 12; acute, months 6 and 12) cohorts compared with healthy controls (Controls, n = 11). Horizontal bars indicate median values. Kruskal–Wallis test followed by Dunn’s multiple comparison test over healthy controls. b, Concentration of CXCL5, CXCL8 and CCL25 in healthy controls (n = 11) versus acute (n = 12, mild hospitalized; n = 26, severe hospitalized) and month 7 (n = 13, mild hospitalized; n = 31, severe hospitalized) COVID-19 convalescents in the Milan cohort. Horizontal bars indicate median values. Kruskal–Wallis test followed by Dunn’s multiple comparison test. c, Correlation between levels of chemokine (acute) and autoantibody (acute or month 7) by two-tailed Pearson correlation analysis in mild hospitalized (n = 12) and severe hospitalized (n = 26) COVID-19 convalescents in the Milan cohort. Source data

Fig. 4. Patterns of chemokine antibodies in COVID-19, HIV-1 and autoimmune diseases.

a, Autoantibodies against specific chemokines in COVID-19 convalescents (COVID-19, month 6, Lugano cohort) and in patients with HIV-1, AS, RA and SjS, shown as ratio over healthy controls (Control). Circle size indicates significance; colors show the log₂ fold-change increase (red) or decrease (blue), shown as ratio over healthy controls. Kruskal–Wallis test followed by Dunn’s multiple comparison test. b, AUC of ELISA showing IgG antibodies to CCL19, CCL4, CCL2, CXCL9 and CXCL12 in the same groups as in a. Horizontal bars indicate median values. Average AUC from two independent experiments. Kruskal–Wallis test followed by Dunn’s multiple comparison test over rank of the healthy control group. c, t-SNE distribution of COVID-19 convalescents (month 6) and patients with HIV-1, AS, RA and SjS, as determined with the 42 datasets combined. In a–c, healthy controls (n = 23), COVID-19 (n = 71), HIV-1 (n = 24), AS (n = 13), RA (n = 13) and SjS (n = 13). Source data

Extended Data Fig. 1. Chemokine N-loop antibodies in COVID-19 convalescents.

(a) Model showing the interaction between a chemokine and its receptor. Arrows point to the area of putative interaction between the N-terminus of the receptor and the chemokine N-loop (shown by spheres). Chemokine is magenta and chemokine receptor is cyan. (b) ELISA curves showing the levels of chemokine N-loop antibodies in healthy controls (Control, n = 23) and COVID-19 convalescents (COVID-19, n = 71) from the Lugano cohort at month 6. Average optical density (OD₄₅₀) measurements of two independent experiments. Source data

Extended Data Fig. 2. Analyses of chemokine antibodies in COVID-19 convalescents.

(a) t-SNE distribution of healthy controls (Control, n = 23) and COVID-19 convalescents from the Lugano cohort at month 6 (COVID-19, n = 71), as determined with the 42 datasets combined. (b) Correlations of antibodies to the N-loop and C-terminal peptides of the same chemokine by two-tailed Pearson correlation analysis. ELISA was performed on a subset of samples (healthy controls, n = 5; COVID-19 convalescents from the Lugano cohort at month 6, n = 31). Average of two independent experiments. (c) Chemokine antibodies in COVID-19 convalescents from the Lugano cohort at month 6 (n = 71), shown as ratio over healthy controls (n = 23). Circle size indicates significance; colors show the Log₂ fold-change increase (red) or decrease (blue), shown as ratio over healthy controls. Two-tailed Mann–Whitney U-tests. (d) Unsupervised hierarchical clustering analysis with CCL19, CCL22 and CXCL17 antibodies showing the distribution of COVID-19 convalescents (month 6, Lugano cohort) and healthy controls in two separate clusters. Two-tailed Fisher’s exact test. (e) Left, AUC of ELISA showing CCL19, CCL22 and CXCL17 antibodies in healthy controls (n = 23) and COVID-19 Milan cohort during acute disease (n = 40) and at month 7 (n = 44). Kruskal-Wallis test followed by Dunn’s multiple comparison test. Right, logistic regression analysis assignment of the COVID-19 Milan cohort and healthy controls based on CCL19, CCL22 and CXCL17 antibodies during acute disease and at month 7. (f) Same as in (e) but for the Zurich cohort at month 13. Healthy controls (n = 23), and COVID-19 convalescents (n = 104). Two-tailed Mann–Whitney U-tests. Source data

Extended Data Fig. 3. Correlation analyses of chemokine antibodies.

(a) ELISA showing RBD IgG antibodies as OD₄₅₀ of serial plasma dilutions (top panel) and as AUC (bottom panel) in healthy controls (Control, n = 23) and COVID-19 convalescents (COVID-19, n = 71) from the Lugano cohort at month 6. Average of two independent experiments. Horizontal bars indicate median values. Two-tailed Mann–Whitney U-tests. (b) Neutralizing activity against SARS-CoV-2 pseudovirus shown as relative luciferase units (RLU) normalized to no plasma control (top panel) and half-maximal neutralizing titers (NT_50, bottom panel) in healthy controls (n = 9) and in COVID-19 convalescents (n = 71) from the Lugano cohort at month 6. Average of two independent experiments. Horizontal bars indicate median values. Two-tailed Mann–Whitney U-tests. (c) Two-tailed Pearson correlations of RBD IgG and NT₅₀ values to each other and with age. Average of two independent experiments. (d) Two-tailed Pearson correlations of CCL19, CCL22, CXCL17 antibodies and of the cumulative signal of the 42 chemokine antibodies with RBD IgG, NT₅₀ values and age. (e) AUC of ELISA showing CCL19, CCL22 and CXCL17 antibodies in healthy controls (n = 23) and COVID-19 convalescents (n = 71) from the Lugano cohort at month 6 grouped by gender (n = 33, females; n = 38, males). Data are shown as average of two independent experiments. Horizontal bars indicate median values. Kruskal-Wallis test followed by Dunn’s multiple comparison test. Source data

Extended Data Fig. 4. Chemokine antibodies in COVID-19 convalescents over time and upon COVID-19 vaccination.

(a) Diagram of the time points of blood collection after onset of COVID-19 symptoms in the Lugano cohort. (b) AUC of ELISA showing RBD IgG antibodies at month 6 and 12 in vaccinated (n = 19) and non-vaccinated (n = 40) COVID-19 convalescents from the Lugano cohort. Average from two independent experiments. Two-tailed Wilcoxon signed-rank test. (c) AUC of ELISA showing chemokine antibodies in COVID-19 convalescents from the Lugano cohort at month 6 and 12 (n = 63). Two independent experiments. Two-tailed Wilcoxon signed-rank test. (d) Diagram of the time points of blood collection after onset of COVID-19 symptoms in a subset of previously hospitalized COVID-19 convalescents from the Lugano cohort. (e) ELISA showing CCL19, CCL22 and CXCL17 antibodies in healthy controls (Control, n = 10) and COVID-19 convalescents (COVID-19) from the Lugano cohort at day 15 (acute, n = 12), and at month 6 (n = 12), 12 (n = 11) and 18 (n = 7). Average OD₄₅₀ values from two independent experiments. One-way ANOVA test followed by Tukey’s multiple comparison test. Data are shown as median±range. (f) AUC of ELISA showing chemokine antibodies in SARS-CoV-2 naïve individuals (n = 16) before and at month 4 on average after COVID-19 mRNA vaccination. Two independent experiments. Pink lines represent the signal of a positive control plasma sample with broad reactivity (CLM70). RBD IgG is shown alongside as control (right panel). Two-tailed Wilcoxon signed-rank test with false discovery rate (FDR) approach. Source data

Extended Data Fig. 5. Chemokine antibodies in previously hospitalized and outpatient COVID-19 convalescents.

(a) AUC of ELISA showing chemokine antibodies in healthy controls (Control, n = 23) and COVID-19 convalescents from the Lugano at month 6 that were previously hospitalized (n = 50) or outpatient (n = 21). Average of two independent experiments. Horizontal bars indicate median values. Kruskal-Wallis test followed by Dunn’s multiple comparison test. (b) Chemokine antibodies in previously hospitalized (n = 50), shown as ratio over outpatient (n = 21) COVID-19 convalescents of the Lugano cohort at month 6. Circle size indicates significance; colors show the Log₂ fold-change increase (red) or decrease (blue), shown as ratio over outpatient COVID-19 convalescents. Kruskal-Wallis test followed by Dunn’s multiple comparison test. (c) Left, AUC of ELISA showing CXCL5, CXCL8 and CCL25 antibodies in healthy controls (n = 23) and in mild (n = 13, acute; n = 27, month 7) versus severe (n = 27, acute; n = 31, month 7) hospitalized COVID-19 from the Milan cohort during acute disease and at month 7. Kruskal-Wallis test followed by Dunn’s multiple comparison test. Right, logistic regression analysis assignment of mild and severe hospitalized COVID-19 convalescents from the Milan cohort based on CXCL5, CXCL8 and CCL25 antibodies during acute disease and at month 7. (d) Same as in (c) but for the Zurich cohort at month 13. Healthy controls (n = 23) and COVID-19 convalescents that were previously hospitalized (n = 38) or outpatient (n = 66). Kruskal-Wallis test followed by Dunn’s multiple comparison test. Source data

Extended Data Fig. 6. Correlation analyses of CXCL5, CXCL8 and CCL25 antibodies in COVID-19 convalescents.

(a) Two-tailed Pearson correlations of CXCL5, CXCL8 and CCL25 antibodies with RBD IgG, NT₅₀ values and age for COVID-19 convalescents from the Lugano cohort at month 6 that were previously hospitalized (n = 50) or outpatient (n = 21). Average of two independent experiments. (b) AUC of ELISA showing CXCL5, CXCL8 and CCL25 antibodies in healthy controls (Control, n = 23) and in COVID-19 convalescents (n = 71) from the Lugano cohort at month 6 grouped by gender (n = 33, females; n = 38, males). Average of two independent experiments. Horizontal bars indicate median values. Kruskal-Wallis test followed by Dunn’s multiple comparison test. (c) AUC of ELISA showing NT₅₀ and RBD IgG values in COVID-19 convalescents (n = 71) from the Lugano cohort at month 6 grouped by disease severity (n = 50 hospitalized; n = 21, outpatient) and by gender (n = 33, females; n = 38, males). Average of two independent experiments. Horizontal bars indicate median values. Two-tailed Mann–Whitney U-tests. Source data

Extended Data Fig. 7. Chemokine antibodies and long-term COVID-19 symptoms.

(a) Incidence of symptoms in patients with long COVID from the Lugano cohort at month 12 (n = 32, previously hospitalized; n = 9 outpatient). (b) Analysis of age (left), gender distribution (middle) and time from COVID-19 onset to month 12 sample collection (right) in COVID-19 convalescents in the Lugano cohort without (n = 22) and with long COVID (n = 41). Horizontal bars indicate median values. Two-tailed Mann–Whitney U-tests. (c,d) Cumulative AUC of ELISA showing chemokine IgG antibodies in COVID-19 convalescents without and with long COVID at month 12 in the Lugano cohort grouped by disease severity (c; n = 10, outpatient no long COVID; n = 9, outpatient long COVID; n = 12, hospitalized no long COVID; n = 32, hospitalized long COVID)) or by gender (d; n = 10, females no long COVID; n = 19, females long COVID; n = 12, males no long COVID; n = 22, males long COVID). Average of two independent experiments. Horizontal bars indicate median values. Two-tailed Mann–Whitney U-tests. (e) Two-tailed Pearson correlation of the cumulative signal of the antibodies against the 42 chemokines in COVID-19 convalescents from the Lugano cohort at month 6 (n = 63) and the number of their self-reported symptoms at month 12. Average of two independent experiments. (f) AUC of ELISA showing chemokine antibodies at month 6 in no long COVID (n = 22) and long COVID (n = 41) groups from the Lugano cohort. Data are shown as average AUC of two independent experiments. Horizontal bars indicate median values. Two-tailed Mann–Whitney U-tests. (g) Left, AUC of ELISA showing CCL21, CXCL13 and CXCL16 antibodies in no long COVID (n = 69) and long COVID (n = 35) groups from the Zurich cohort at month 13. Two-tailed Mann–Whitney U-tests. Right, logistic regression analysis assignment of no long COVID and long COVID groups from the Zurich cohort based on CCL21, CXCL13 and CXCL16 antibodies at month 13. Source data

Extended Data Fig. 8. Human monoclonal antibodies that impede chemotaxis.

(a) Gating strategy for sorting CCL8 N-loop specific B cells by flow cytometry. (b,c) Representative flow cytometry plots showing human B cells binding to the CXCL16 (b) or CXCL13 (c) N-loop peptide (gate). The frequency of antigen-specific B cells is shown. (d) AUC of ELISA showing antibodies to the CCL8 N-loop in healthy controls (Control, n = 23) and COVID-19 convalescents from the Lugano cohort at month 6 (COVID-19, n = 71), and identification of individuals with high antibody reactivity. Average of two independent experiments. Horizontal bars indicate median values. (e) Representative flow cytometry plots showing human B cells binding to the CCL8 N-loop peptide (gate). The frequency of antigen-specific B cells is shown. (f) ELISA showing CCL8 monoclonal antibodies binding to the CCL8 N-loop. Average of two independent experiments. (g) Chemotaxis showing migration of human monocytes in a CCL8 gradient (n = 4) in the presence of aCCL8.001 (n = 2), aCCL8.005 (n = 4) or isotype control antibody (n = 4). Mean±SEM of migrated cells in 5 high-power fields (HPF). Up-pointing triangle is antibody alone, and down-pointing triangle is buffer control. Two-way RM ANOVA followed by Šídák’s multiple comparisons test. (h) AUC of ELISA showing antibodies to the CCL20 N-loop in healthy controls (n = 23) and COVID-19 convalescents from the Lugano cohort at month 6 (n = 71), and identification of individuals with high antibody reactivity. Average of two independent experiments. Horizontal bars indicate median values. (i) ELISA showing CCL20 monoclonal antibodies binding to the CCL20 N-loop. Average of two independent experiments. (j) Chemotaxis showing relative cell migration of the 300.19 preB cell line uniquely expressing CCR6 in a CCL20 gradient (1 nM) in the presence of aCCL20.001 or isotype control antibody. Mean+SEM of 3 independent experiments. Two-tailed Mann–Whitney U-tests. (k) Chemotaxis showing cell migration of preB 300.19 cells expressing CCR2 toward a CCL7 (100 nM) or CCL8 (100 nM) gradient, or of preB 300.19 cells expressing CXCR1 towards a CXCL8 gradient (1 nM), in the presence of plasma IgGs from a subset of COVID-19 convalescents from the Lugano cohort at month 6 (n = 24 for CCL7 and CCL8; n = 16 for CXCL8) or healthy controls (n = 8). Technical triplicates (Mean±SEM) of migrated cells in 5 high-power fields (HPF). Two-tailed Mann–Whitney U-tests. Source data

Extended Data Fig. 9. Chemokine antibodies in HIV-1 infection and in AS, RA and SjS.

(a) AUC of ELISA showing chemokine antibodies in healthy controls (Control, n = 23), COVID-19 convalescents (COVID-19, n = 71; Lugano cohort at month 6), HIV-1 (n = 24) AS (n = 13), RA (n = 13), and SjS (n = 13). Average of two independent experiments. Horizontal bars indicate median values. Statistical significance was determined using Kruskal-Wallis test followed by Dunn’s multiple comparison test over rank of healthy controls. (b) Venn diagram showing the chemokines targeted by autoantibodies across the autoimmune disorders AS, RA and SjS. Red and blue colors indicate either an increase or decrease compared to healthy controls with P < 10⁻⁴. Kruskal-Wallis test followed by Dunn’s multiple comparison test over rank of healthy controls as in (a). Source data

Extended Data Fig. 10. Chemokine antibodies in Lyme disease (Borrelia infection) and clustering of COVID-19, HIV-1 and autoimmune diseases based on chemokine antibodies.

(a) AUC of ELISA showing chemokine antibodies in healthy controls (Control, n = 30) and in Lyme (Erythema migrans) during acute disease (n = 26) and at month 6 post-infection (n = 23). Average of two independent experiments. Horizontal bars indicate median values. Statistical significance was determined using Kruskal-Wallis test followed by Dunn’s multiple comparison test. (b) Heatmap representing the unsupervised hierarchical clustering analysis of COVID-19 convalescents (n = 71; Lugano cohort at month 6), HIV-1 (n = 24), AS (n = 13), RA (n = 13) and SjS (n = 13), based on normalized AUC of ELISA values for plasma IgG binding to 42 peptides comprising the N-loop of all 43 human chemokines. The distribution of the groups within each cluster is shown. Source data