Early peripheral blood MCEMP1 and HLA-DRA expression predicts COVID-19 prognosis

Abstract

Background: Mass vaccination has dramatically reduced the incidence of severe COVID-19, with most cases now presenting as self-limiting upper respiratory tract infections. However, those with co-morbidities, the elderly and immunocompromised, as well as the unvaccinated, remain disproportionately vulnerable to severe COVID-19 and its sequelae. Furthermore, as the effectiveness of vaccination wanes with time, immune escape SARS-CoV-2 variants could emerge to cause severe COVID-19. Reliable prognostic biomarkers for severe disease could be used as early indicator of re-emergence of severe COVID-19 as well as for triaging of patients for antiviral therapy.

Methods: We performed a systematic review and re-analysis of 7 publicly available datasets, analysing a total of 140 severe and 181 mild COVID-19 patients, to determine the most consistent differentially regulated genes in peripheral blood of severe COVID-19 patients. In addition, we included an independent cohort where blood transcriptomics of COVID-19 patients were prospectively and longitudinally monitored previously, to track the time in which these gene expression changes occur before nadir of respiratory function. Single cell RNA-sequencing of peripheral blood mononuclear cells from publicly available datasets was then used to determine the immune cell subsets involved.

Findings: The most consistent differentially regulated genes in peripheral blood of severe COVID-19 patients were MCEMP1, HLA-DRA and ETS1 across the 7 transcriptomics datasets. Moreover, we found significantly heightened MCEMP1 and reduced HLA-DRA expression as early as four days before the nadir of respiratory function, and the differential expression of MCEMP1 and HLA-DRA occurred predominantly in CD14+ cells. The online platform which we developed is publicly available at https://kuanrongchan-covid19-severity-app-t7l38g.streamlitapp.com/, for users to query gene expression differences between severe and mild COVID-19 patients in these datasets.

Interpretation: Elevated MCEMP1 and reduced HLA-DRA gene expression in CD14+ cells during the early phase of disease are prognostic of severe COVID-19.

Funding: K.R.C is funded by the National Medical Research Council (NMRC) of Singapore under the Open Fund Individual Research Grant (MOH-000610). E.E.O. is funded by the NMRC Senior Clinician-Scientist Award (MOH-000135-00). J.G.H.L. is funded by the NMRC under the Clinician-Scientist Award (NMRC/CSAINV/013/2016-01). S.K. is funded by the NMRC under the Transition Award. This study was sponsored in part by a generous gift from The Hour Glass.

Keywords: Biomarkers; COVID-19; Gene expression; HLA-DRA; MCEMP1; Pathogenesis; RNAseq; Single cell sequencing; Systematic review; Transcriptomics.

Fig. 1

Comparative analysis of transcriptomics datasets allows discovery of genes that are predictive of severe COVID-19. a, Schematic of the multi-cohort analysis workflow, validation studies and single cell analysis that eventually identified MCEMP1 and HLA-DRA gene expression as the most consistent prognostic biomarker for severe COVID-19. b, Heatmap showing the number of datasets that the genes are differentially expressed (Fold-change > 1.3, p < 0.05) in severe compared to mild COVID-19 patients. Transcripts that were differentially regulated in 4 or more datasets are displayed. Orange indicates upregulated in the acute samples taken from severe compared to mild COVID-19 patients, blue indicates downregulated in severe compared to mild COVID-19 patients and black indicates no change. c–e, Forest plots of MCEMP-1, ETS1 and HLA-DRA transcript expression in severe COVID-19 patients compared to mild COVID-19 patients based on 7 transcriptomic datasets. x-axis represents the log2FC between severe and mild COVID-19 patients in the respective datasets. Whiskers represent the 95% CI. Mean statistics are obtained by first determining the average log2FC values for each dataset, followed by calculating the mean log2FC and 95% CI of these log2FC values. Red indicates upregulation in severe COVID-19 patients whereas blue indicates downregulation as compared to mild COVID-19 patients. Sample size for each dataset is indicated in Table 1.

Fig. 2

MCEMP1 and HLA-DRA transcript expression levels can discriminate severe from mild COVID-19 patients as early as 4 days before nadir of respiratory function. a–f, Scatterplots showing correlation between MCEMP1, HLA-DRA and ETS1 gene expression levels in whole blood that are associated with a–c, C-reactive protein (CRP), d–f, IFN score or g–i, lymphocyte counts. The respective Pearson's correlation (r) and p-values are shown. Patient samples are based on the study by Ong et al., 2021. j, Normalised expression levels of MCEMP1, HLA-DRA and ETS1 expression in severe and moderately severe COVID-19 patients (n = 6) over time relative to nadir of respiratory function (day 0), based on a study by Ong et al., 2021. k, Combination of MCEMP1 and HLA-DRA gene expression levels in mild COVID-19 (n = 4), and severe COVID-19 (n = 6) subjects as measured by real-time PCR. Combined scores are determined by subtracting individual Z-scores for MCEMP1 to Z-scores of HLA-DRA. Samples are taken on the first day of hospital admission for the severe and mild COVID-19 patients. Relative expression of each gene is normalised to B2M, which is the housekeeping gene. Normality was determined by the D′ Agostino & Pearson test (p > 0.05), and error bars indicate standard deviation of the data. ∗p < 0.05 (Unpaired t-test).

Fig. 3

Sensitivity of MCEMP1 and HLA-DRA across 7 transcriptomic datasets. a, Box plots showing mean AUC and 95% CI of MCEMP1, HLA-DRA and combined scores (obtained by subtracting individual Z-scores for MCEMP1 to Z-scores of HLA-DRA) based on 7 severe vs mild COVID-19 transcriptomic datasets detailed in Table 1. AUC values were first obtained for each study, followed by calculation of the mean AUC and 95% CI of the AUC values. b, ROC curve showing suitability of using MCEMP1 and HLA-DRA combined scores in discriminating severe COVID-19 and mild COVID-19 patients in 7 datasets. c, Table showing AUC values and 95% CI for each dataset when both MCEMP1 and HLA-DRA expression levels were used for discriminating severe and mild COVID-19. d–f, ROC curve for comparing severe and mild d, influenza, e, RSV and f, dengue patients in 3 independent datasets based on combined scores of MCEMP1 and HLA-DRA expression levels. Sample size for the datasets are indicated in Table 1 and Supplementary Table S5.

Fig. 4

Single-cell PBMCs transcriptomes highlight that differential expression of MCEMP1 and HLA-DRA occurs predominantly in CD14+ monocytes. a and b, UMAP of 48,583 PBMCs based on GSE149689. Charts are coloured to show the a, disease severity status and b, cell type annotations. c and d, Violin plots showing the expression of c, MCEMP1 and d, HLA-DRA in different immune cell subsets in 4 healthy, 4 mild and 6 severe COVID-19 subjects. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 (Wilcoxon rank sum test).