Prepandemic Metabolic Correlates of Coronavirus Disease 2019 (COVID-19) Severity and Long COVID Incidence in People Living With HIV

Abstract

Host metabolic dysregulation, especially in tryptophan metabolism, is intricately linked to coronavirus disease 2019 (COVID-19) severity and its postacute sequelae (long COVID). People living with human immunodeficiency virus (HIV; PLWH) experience similar metabolic dysregulation and face an increased risk of developing long COVID. However, whether preexisting HIV-associated metabolic dysregulations contribute in predisposing PLWH to severe COVID-19 outcomes remains underexplored. Analyzing prepandemic samples from PLWH with documented postinfection outcomes, we found specific metabolic alterations, including increased tryptophan catabolism, predicting an elevated risk of severe COVID-19 and the incidence of long COVID. These alterations warrant further investigation for their potential prognostic and mechanistic significance in determining COVID-19 complications.

Keywords: COVID-19; HIV; long COVID; metabolites; tryptophan.

Figure 1.

Prepandemic metabolic dysregulation linked to the severity of postpandemic coronavirus disease 2019 (COVID-19) in people living with human immunodeficiency virus (PLWH). A, Overview of the study design, illustrating the identification process of prepandemic plasma metabolite signatures predictive of severe COVID-19 and postacute sequelae of COVID-19 (long COVID). Abbreviation: MWCCS, MACS (Multicenter AIDS Cohort Study)–WIHS (Women’s Interagency HIV Study) Combined Cohort Study. B, Analysis of metabolites and metabolic ratios with prepandemic levels differing between PLWH who required hospitalization for severe acute respiratory syndrome coronavirus 2 infection and those who did not. Mean values and standard error of the mean are displayed, comparing each individual in the hospitalization group with average values from the control group. P values were calculated with the SelectKBest method from the scikit-learn library (version 1.2.2), using an analysis of variance–based approach to identify metabolites significantly correlated with the categorical variable. Abbreviations: HpODE, hydroperoxy-9Z,11E-octadecadienoic acid; K/T, L-kynurenine–tryptophan; Q/T, quinolinic acid–tryptophan. C, Metabolic enrichment analysis conducted using MetaboAnalyst 5.0 (accessible at http://www.metaboanalyst.ca/). Abbreviations: FDR, false discovery rate; tRNA, transfer RNA. D, Receiver operating characteristic (ROC) curves generated from a 5-fold cross-validation analysis of a classifier trained on an oversampled data set. Abbreviation: AUC, area under the ROC curve. E, Top, Depiction of predicted probability scores from the benchmarking data set using the classifier. E, Bottom, Illustration of the ROC curve from cross-validation analysis using the classifier trained on a balanced oversampled data set, with the original unbalanced data set used for cross-validation. Some of the graphics in A were adapted from BioRender .

Figure 2.

Prepandemic metabolic dysregulation linked to the postpandemic incidence of postacute sequelae of coronavirus disease 2019 (COVID-19), or long COVID, in people living with human immunodeficiency virus (PLWH). A, Analysis of metabolites and metabolic ratios whose prepandemic levels differ between PLWH who experienced long COVID symptoms for ≥3 months after their acute COVID-19 and those who did not. Mean values are shown with standard error of the mean, comparing the values of each individual in the long COVID group with average values from the control group. P values were calculated using the SelectKBest method from the scikit-learn library (version 1.2.2), using an analysis of variance–based approach to identify metabolites significantly correlated with the categorical variable. Abbreviation: Q/T, quinolinic acid–tryptophan. B, Metabolic enrichment analysis conducted using MetaboAnalyst 5.0 software (accessible at http://www.metaboanalyst.ca/). C, Receiver operating characteristic (ROC) curves generated from a 5-fold cross-validation analysis of a classifier trained on an oversampled data set. Abbreviation: AUC, area under the ROC characteristic curve. D, Top, Depiction of predicted probability scores from the benchmarking data set using the classifier. D, Bottom, Illustration of the ROC curve from cross-validation analysis using the classifier trained on a balanced oversampled data set, with the original unbalanced data set used for cross-validation.