An immune-based biomarker signature is associated with mortality in COVID-19 patients

Immune and inflammatory responses to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) contribute to disease severity of coronavirus disease 2019 (COVID-19). However, the utility of specific immune-based biomarkers to predict clinical outcome remains elusive. Here, we analyzed levels of 66 soluble biomarkers in 175 Italian patients with COVID-19 ranging from mild/moderate to critical severity and assessed type I IFN–, type II IFN–, and NF-κB–dependent whole-blood transcriptional signatures. A broad inflammatory signature was observed, implicating activation of various immune and nonhematopoietic cell subsets. Discordance between IFN-α2a protein and IFNA2 transcript levels in blood suggests that type I IFNs during COVID-19 may be primarily produced by tissue-resident cells. Multivariable analysis of patients’ first samples revealed 12 biomarkers (CCL2, IL-15, soluble ST2 [sST2], NGAL, sTNFRSF1A, ferritin, IL-6, S100A9, MMP-9, IL-2, sVEGFR1, IL-10) that when increased were independently associated with mortality. Multivariate analyses of longitudinal biomarker trajectories identified 8 of the aforementioned biomarkers (IL-15, IL-2, NGAL, CCL2, MMP-9, sTNFRSF1A, sST2, IL-10) and 2 additional biomarkers (lactoferrin, CXCL9) that were substantially associated with mortality when increased, while IL-1α was associated with mortality when decreased. Among these, sST2, sTNFRSF1A, IL-10, and IL-15 were consistently higher throughout the hospitalization in patients who died versus those who recovered, suggesting that these biomarkers may provide an early warning of eventual disease outcome.

selectin/sCD62E, RAGE, sCD163, sVEGFR1/Flt-1, REG3A, S100A8, S100A9, MMP-9, lactoferrin, MPO, lipocalin-2/NGAL, LBP) were measured on customized, magnetic bead-based, multiplex assay (R&D Systems, Minneapolis, MN) according to the manufacturers specifications for standards and dilutions. The magnetic beads were analyzed on Bio-Plex 3D instrumentation (Bio-Rad, Hercules, CS). Standard curves were analyzed using nonlinear curve fitting and unknowns were calculated based on the derived equation. Samples that exceeded the highest standards were reanalyzed more dilute until the values fell within the range of the known standards. Two control plasma samples and a control sample spiked with a known quantity of each analyte were analyzed on each plate to assess the inter-plate variation and to determine the effect of the biological matrix on the measurement of each analyte. For most analytes, the control samples had <25% variation from plate to plate, and the recoveries were generally >70%.
Ferritin was measured in the clinical laboratories of the hospitals where patients were admitted. CXCL9 levels were measured using a Duoset ELISA kit (R&D Systems).
Total IL-18 and IL-18BP levels were measured as previously described (1). Briefly, serum was diluted 25-fold and assayed on a FLEXMAP 3D multiplex instrument per the manufacturer's instructions (Luminex). Recombinant IL-18 was used as standard (MBL International), and human IL-18BPa-Fc (R&D Systems) was run as a separate standard for IL-18BP. IL-18 and IL-18BPa beads were generated by conjugating capture antibody to magnetic beads per the manufacturer's instructions (Bio-Rad). All reagents used were derived from the same lots. Minimal variation between plates and runs was verified using bridging controls.
To measure pGSN, Nunc Maxisorp plates were coated with 100 μl of 5 μg rabbit anti-human pGSN pAb specific for the 24 amino acid extension unique to plasma gelsolin/ml 0.05M carbonate buffer pH 9.6 for 2 hr at room temperature. After washing 2x in wash buffer (150 mM NaCl, 25 mM Tris 7.4, 1 mM CaCl2, 0.05% Tween-20) plates were incubated in the wash buffer + 3% BSA for 10 minutes to block. Samples and recombinant standard were diluted in wash buffer + 1% BSA and incubated with detection buffer (wash buffer + 1% BSA with 0.4 μg mouse anti-gelsolin mAb clone GS-2C4 (Sigma) and 0.067 μg Goat anti-Mouse IgG (H+L) -HRPO (eBioScience/Invitrogen) for 30 min at room temperature with shaking. After washing 4x, 100 μl of TMB substrate (Invitrogen) was added per well and allowed to develop for 5 minutes before stopping with 100 μl 2 N H2SO4 and read at A450 and A650 prior to concentration determination using a 5-parameter logistic curve (GraphPad Prism 8).
Paired serum and EDTA plasma samples drawn concurrently from HV (n = 15) and patients with  were analyzed for all biomarkers. The paired data were plotted, and linear regression lines were determined. If the slope of the regression line for a particular biomarker fell between 0.6 and 1.4, the level in EDTA plasma and serum were considered equivalent and either serum or plasma was used for analysis.

Flow cytometric studies in whole blood of COVID-19 patients
COVID-19 patient and HV whole blood samples were harvested in the morning and were processed for flow cytometry-based analyses within 3 hours of blood harvesting.

Analysis of peripheral blood smears
Routine peripheral blood smears were prepared using a Sysmex SP-10 automated slide maker stainer. Slides were scanned and cells were classified using a CellaVision DM96 instrument. CellaVision images were examined for monocyte and neutrophil morphology from seven COVID-19 patients hospitalized at the NIH Clinical Center. At least 100 monocytes and 100 neutrophils were evaluated per patient to determine the percent of each cell type that contained vacuoles.

Transcriptional analysis of whole blood from PAXgene tubes
Eighty-four COVID-19 patients had serial whole blood collections in PAXgene tubes; 11 (13%) had moderate disease, 8 (9%) had severe disease, 45 (54%) had critical disease and 19 (23%) ultimately succumbed to their disease. Total RNA was extracted from whole blood samples collected in PAXgene tubes (Qiagen, Germantown, MD) and subjected to transcriptional analysis of selected genes including IFNA2 was determined by NanoString (NanoString Technologies, Seattle, WA). A 28-gene type I IFN score and an 11-gene NF-κB score were calculated as previously described (3,4). An IFN-γ score was calculated based on 15 IFN-γ-regulated genes (5). Briefly, the 28-gene type I IFN score is the sum of the z-scores of 28 type I IFN response genes and the 11-gene NF-kB score is the sum of the z-scores of 11 NF-κB targets genes and the 15-gene IFN-γ score is the sum of the z-scores of 15 IFN-γ response genes. Individual gene z-scores were calculated using the mean and standard deviation of the NanoString counts from 22 HV (Supplementary Table 11).

Peripheral blood smear examination
Routine peripheral blood smears were prepared and wer examined for monocyte and neutrophil morphology from seven COVID-19 patients hospitalized at the NIH Clinical Center. At least 100 monocytes and 100 neutrophils were evaluated per patient to determine the percent of each cell type that contained vacuoles.

Statistical analyses
The effect of age (≥65 years), gender, and various medical comorbidities on biomarker    20

Supplementary Figure 11. CD8 + T cells of patients with COVID-19 exhibit enhanced IFN-γ production. (A)
Left panel depicts representative contour plots of IFN-γ production within CD8 + T cells in unstimulated whole blood of patients with COVID-19 (red) and healthy volunteers (black). Right panel depicts summary data of frequencies of IFN-γ producing cells within the indicated lymphoid cell subsets in patients with COVID-19 (n = 4) and healthy volunteers (n = 6). All quantitative data represent mean ± standard error of the mean. P value was calculated using an unpaired t-test with Welch's correction. *P<0.05. (B) Summary data showing no significant differences in the percent of the indicated IL-4 + or IL-17A + lymphoid cell subsets from COVID-19 patients compared to healthy volunteers.

Healthy volunteer COVID-19
Healthy volunteers COVID-19  Submitted as a separate Excel file.

Supplementary Table 4. Summary data of all biomarkers and their association with mortality in three mathematical models of univariable and multivariable analyses.
Submitted as a separate Excel file.

Supplementary Table 5. Evaluation of clinical factors that affect the concentration of immune-based biomarkers and laboratory tests.
Submitted as a separate Excel file.
Supplemental Table 6. Association between the longitudinal trajectory of biomarkers and the risk of death after COVID-19. Shown are the posterior median HR and 95% credible intervals (CrI), along with q-values for the posterior probability that the association was positive or negative. Biomarkers with q-values < 0.025 are highlighted in red. Survival submodel: proportional hazards model for risk of death versus expected biomarker concentration from the longitudinal submodel.

Longitudinal model
Data: All samples with missing biomarker and covariate data imputed.