Comprehensive proteomics and meta-analysis of COVID-19 host response

Abstract

COVID-19 is characterised by systemic immunological perturbations in the human body, which can lead to multi-organ damage. Many of these processes are considered to be mediated by the blood. Therefore, to better understand the systemic host response to SARS-CoV-2 infection, we performed systematic analyses of the circulating, soluble proteins in the blood through global proteomics by mass-spectrometry (MS) proteomics. Here, we show that a large part of the soluble blood proteome is altered in COVID-19, among them elevated levels of interferon-induced and proteasomal proteins. Some proteins that have alternating levels in human cells after a SARS-CoV-2 infection in vitro and in different organs of COVID-19 patients are deregulated in the blood, suggesting shared infection-related changes.The availability of different public proteomic resources on soluble blood proteome alterations leaves uncertainty about the change of a given protein during COVID-19. Hence, we performed a systematic review and meta-analysis of MS global proteomics studies of soluble blood proteomes, including up to 1706 individuals (1039 COVID-19 patients), to provide concluding estimates for the alteration of 1517 soluble blood proteins in COVID-19. Finally, based on the meta-analysis we developed CoViMAPP, an open-access resource for effect sizes of alterations and diagnostic potential of soluble blood proteins in COVID-19, which is publicly available for the research, clinical, and academic community.

Fig. 1. Study workflow.

a The serum was depleted of 14 high-abundant proteins, to provide identification of lower abundant proteins; b We infected the Calu-3 line with SARS-CoV-2 in vitro and collected the cells at 3 hours, 1 day, 3 days, and 7 days after infection. Non-infected Calu-3 cells and Calu-3 cells treated with ultraviolet light-inactivated SARS-CoV-2 were also cultured as controls; c In-depth LC-MS/MS proteomics workflow; d Serum proteome alterations were traced to proteome alterations of SARS-CoV-2-infected cells; e Combining twenty global MS proteomics datasets with our dataset, we performed a per-protein meta-analysis of 1517 proteins identified in at least two cohorts and in up to 1706 individuals.

Fig. 2. Protein identification.

a Venn diagram of all identified proteins (gene-centric) in HiRIEF LC-MS/MS and overlap with all proteins identified and analysed with affinity-based proteomics^, ; b Proteins ranked (x axis) based on their median precursor area intensity (y axis), coloured according to the method of identification. c Agreement in quantifying the clinical biomarkers of COVID-19 severity between HiRIEF LC-MS/MS and clinical chemistry assays: CRP, AST (gene name: GOT1), and ALT (gene name: GPT). The line represents the linear regression fit line and the surrounding shaded area 95% confidence intervals (CI). The p values for the Spearman correlation coefficients (r) were obtained with a two-sided t test.

Fig. 3. COVID-19 serum proteomics by HiRIEF LC-MS/MS.

a PCA based on proteins without missing values; b Differential alteration of serum protein levels in COVID-19. The shade of the points represent the log2 fold change (log2-FC) multiplied by -log10 (adjusted p value); c Heatmap of differentially altered serum proteins and relation with clinical parameters. The five most frequent REACTOME terms describing a protein function per cluster are annotated; d Agreement with PEA profiling of the COVID-19 soluble blood proteome—overlapping proteins that are significant at 5% FDR in both methods. The agreement is represented as proportion (in %) of proteins changing in the same direction out of the total number of overlapping proteins and with Spearman’s correlation coefficient (r). The PEA analysis is adjusted for age, sex, ethnicity, heart disease, diabetes, hypertension, hyperlipidaemia, pulmonary disease, kidney disease, and immuno-compromised status as covariates; e Agreement with SOMAscan profiling of the COVID-19 serum proteome—overlapping proteins that are significant at 5% FDR in both methods. The SOMAscan analysis is adjusted for age and sex as covariates.

Fig. 4. Comparison of serum protein alterations to proteome alterations in Calu-3 cells infected with SARS-CoV-2.

a Day 3 after infection. The agreement is represented as proportion (in %) of proteins changing in the same direction out of the total number of overlapping proteins and with Spearman’s correlation coefficient (r); b Day 7 after infection; c Boxplots for selected proteins consistently upregulated at day 3 and day 7 after infection, and in serum. The boxplots show protein levels in SARS-CoV-2-infected Calu-3 cells at different time points, compared to non-infected cells and cells treated with UV-inactivated SARS-CoV-2. All cells in each condition were cultured as biological replicates (n = 3 each). In addition, boxplots of serum levels of the respective protein in COVID-19 patients compared to healthy controls are presented. The box centre represents the median, the lower and upper box limits the 25^th and 75^th percentile, respectively, and whiskers’ limits the minimum and maximum values of the data after removing outliers. ns = non-significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.005, **** = p < 0.001. The p values were determined with a two-sided t test and adjusted for multiple testing with the FDR.

Fig. 5. Gene set enrichment analyses of organ-associated protein sets and MsigDb gene sets in serum of COVID-19 patients.

a Enriched organ-associated protein sets at 5% FDR, permutation test. The sets consist of proteins deregulated in a specific organ obtained from COVID-19 patients, regardless if the protein was identified in another organ-associated protein set. We tested 33 protein sets (17 upregulated and 16 downregulated) from 10 organs: adrenal glands, blood vessels, brain, heart, kidney, liver, lungs, lymph nodes, spleen (white pulp), and thyroid; b Enriched organ-specific protein set that was downregulated in the white pulp of the spleen, filtered for any protein that was downregulated in another organ in the same dataset; c Enriched hallmark gene sets at 5% FDR; d Enriched KEGG pathways at 5% FDR. The rank (x axis) of the protein belonging in the gene set is based on log2-FC comparing mean serum levels between COVID-19 and PCR-negative healthy controls. The enrichment score is plotted on the y axis.

Fig. 6. Protein identification in studies included in the meta-analysis.

a Number of participants in relation to protein identification—as reported in the publication. The line was fitted with local polynomial regression; b Proteins identified compared to proteins included in the meta-analysis, i.e., proteins identified in at least three COVID-19 patients and at least three SARS-CoV-2 PCR-negative controls; c Proportion of proteins identified in the studies. Only identifications that had a unique match to a protein with a gene name were included—those that had two gene names per protein were excluded; d Map depicting the overlap of proteins identified in a given study.

Fig. 7. Meta-analysis summary and selected proteins.

a Volcano plot showing the SMD on the x axis and -log10 p values on the y axis, random-effects model. Proteins above the dashed line (p < 0.05) were statistically significant. The shade of the points represents SMD multiplied by -log10 (p value); b Proteins identified in ≥ 18 cohorts that had a statistically significant SMD. The error bars represent 95% CI of the SMD estimates; c Agreement in percentages and Spearman’s correlation coefficient (r) between log2-FC by HiRIEF LC-MS/MS (p < 0.05, 5% FDR) and SMD estimates in the meta-analysis (p < 0.05); d SROC summary estimates of sensitivity and specificity of all the proteins identified in ≥ 3 studies. Proteins with 95% CI of either sensitivity or specificity including 0.5 (the chance dashed line) were statistically non-significant. FPR false positive rate; e Top 11 proteins based on AUC that were identified in all cohorts. The SROC curves are based on the bivariate model, along with a 95% CI tolerance ellipsoid per protein; f Heterogeneity of the underlying ROC curves per protein per study (n = 7606) used in estimating the SROC curves (n = 971). The estimated α shape parameter for lowest heterogeneity (Q) is plotted on the x axis; α of 1 = no preference, α > 1 indicates ROC preference for sensitivity and α < 1 indicates ROC preference for specificity. The log ratio of mean sensitivity and mean specificity is presented on the y axis; 0 = no preference (horizontal double-dashed line), values ≻ 0 = model prefers sensitivity, and values ≺ 0 = model prefers specificity. The ROC curves for a protein in a study were labelled as having a preference of sensitivity and specificity if they had values above the 80^th quantile (qtl) and below the 20^th quantile of α, respectively. g Inverse relationship between mean specificity (x axis) and mean sensitivity (y axis). h Relationship between the number of cohorts included in the per-protein estimates and absolute difference between specificity and sensitivity (left) and AUC (right).