Metabolomics Profiling of Critically Ill Coronavirus Disease 2019 Patients: Identification of Diagnostic and Prognostic Biomarkers

Abstract

Objectives: Coronavirus disease 2019 continues to spread rapidly with high mortality. We performed metabolomics profiling of critically ill coronavirus disease 2019 patients to understand better the underlying pathologic processes and pathways, and to identify potential diagnostic/prognostic biomarkers.

Design: Blood was collected at predetermined ICU days to measure the plasma concentrations of 162 metabolites using both direct injection-liquid chromatography-tandem mass spectrometry and proton nuclear magnetic resonance.

Setting: Tertiary-care ICU and academic laboratory.

Subjects: Patients admitted to the ICU suspected of being infected with severe acute respiratory syndrome coronavirus 2, using standardized hospital screening methodologies, had blood samples collected until either testing was confirmed negative on ICU day 3 (coronavirus disease 2019 negative) or until ICU day 10 if the patient tested positive (coronavirus disease 2019 positive).

Interventions: None.

Measurements and main results: Age- and sex-matched healthy controls and ICU patients that were either coronavirus disease 2019 positive or coronavirus disease 2019 negative were enrolled. Cohorts were well balanced with the exception that coronavirus disease 2019 positive patients suffered bilateral pneumonia more frequently than coronavirus disease 2019 negative patients. Mortality rate for coronavirus disease 2019 positive ICU patients was 40%. Feature selection identified the top-performing metabolites for identifying coronavirus disease 2019 positive patients from healthy control subjects and was dominated by increased kynurenine and decreased arginine, sarcosine, and lysophosphatidylcholines. Arginine/kynurenine ratio alone provided 100% classification accuracy between coronavirus disease 2019 positive patients and healthy control subjects (p = 0.0002). When comparing the metabolomes between coronavirus disease 2019 positive and coronavirus disease 2019 negative patients, kynurenine was the dominant metabolite and the arginine/kynurenine ratio provided 98% classification accuracy (p = 0.005). Feature selection identified creatinine as the top metabolite for predicting coronavirus disease 2019-associated mortality on both ICU days 1 and 3, and both creatinine and creatinine/arginine ratio accurately predicted coronavirus disease 2019-associated death with 100% accuracy (p = 0.01).

Conclusions: Metabolomics profiling with feature classification easily distinguished both healthy control subjects and coronavirus disease 2019 negative patients from coronavirus disease 2019 positive patients. Arginine/kynurenine ratio accurately identified coronavirus disease 2019 status, whereas creatinine/arginine ratio accurately predicted coronavirus disease 2019-associated death. Administration of tryptophan (kynurenine precursor), arginine, sarcosine, and/or lysophosphatidylcholines may be considered as potential adjunctive therapies.

Keywords: biomarker; coronavirus disease 2019; diagnoses; intensive care unit; metabolomics; prognoses.

Figure 1.

A, Subjects plotted in two dimensions following dimensionality reduction in their respective metabolites by stochastic neighbor embedding. Green dots represent healthy control subjects, whereas orange dots represent age- and sex-matched coronavirus disease 2019 positive (COVID19+) ICU patients (ICU day 1 plasma). The dimensionality reduction shows that based on the plasma metabolites, the two cohorts are distinct and easily separable. The axes are dimensionless. B, Feature classification, demonstrating the top eight plasma metabolites that classify COVID19+ status versus healthy control subjects with their % association. C, Receiver operating characteristic analysis of healthy control subjects versus COVID19+ patients, using an arginine/kynurenine ratio, demonstrates an area-under-the-curve (AUC) of 1.00 (p = 0.0002). The cutoff value is 15.6. The diagonal broken blue line represents chance (AUC 0.50).

Figure 2.

A, Feature classification demonstrating the top eight plasma metabolites that classify coronavirus disease 2019 positive (COVID19+) status versus healthy control subjects with their % association. B, Receiver operating characteristic analysis of COVID19+ versus coronavirus disease 2019 negative (COVID19–) ICU patients, using the arginine/kynurenine ratio, demonstrates an area-under-the-curve (AUC) of 0.98 (p = 0.005). The diagonal broken blue line represents chance (AUC = 0.50). C, A time plot, demonstrating the Arginine/Kynurenine ratio for both COVID19+ (orange dots) and COVID19– (blue dots) patients over 10 ICU days. The two cohorts are significantly different on ICU days 1 and 3 (***p = 0.005). Healthy control range values are represented by green shading.

Figure 3.

A, Coronavirus disease 2019 positive (COVID19+) ICU patients plotted in two dimensions following dimensionality reduction of their respective metabolites by stochastic neighbor embedding. Blue dots represent COVID19+ ICU patients that survived their ICU stay, whereas orange dots represent COVID19+ ICU patients that died (ICU day 1 plasma). The dimensionality reduction shows that based on the plasma metabolites, the two cohorts are distinct and easily separable. The axes are dimensionless. B, Feature classification, demonstrating the top eight plasma metabolites that classify COVID19+ ICU patient outcome as alive or dead with their % association. Plasma creatinine was the leading outcome predictor metabolite. C, A time plot, demonstrating the creatinine/arginine ratio for COVID19+ ICU patients over 10 ICU days that either survived (blue dots) or died (orange dots). The two cohorts are significantly different on ICU days 1 and 3 (**p = 0.01). Healthy control range values are represented by green shading.