key: cord-0825499-40amg3d9
authors: Bennet, S. M. P.; Kaufmann, M.; Takami, K.; Sjaarda, C.; Douchant, K.; Moslinger, E.; Wong, H.; Reed, D. E.; Ellis, A. K.; Vanner, S.; Colautti, R. I.; Sheth, P. M.
title: Small-molecule metabolome identifies potential therapeutic targets against COVID-19
date: 2021-06-23
journal: nan
DOI: 10.1101/2021.06.18.21259150
sha: f42ed015c205fec6a549eed1186843dc0ad2dab4
doc_id: 825499
cord_uid: 40amg3d9

Background: Respiratory viruses are transmitted and acquired via the nasal mucosa, and thereby may influence the nasal metabolome composed of biochemical products produced by both host cells and microbes. Studies of the nasal metabolome demonstrate virus-specific changes that sometimes correlate with viral load and disease severity. Here, we evaluate the nasopharyngeal metabolome of COVID-19 infected individuals and report several small molecules that may be used as potential therapeutic targets. Specimens were tested by qRT-PCR with target primers for three viruses: Influenza A (INFA), respiratory syncytial virus (RSV), and SARS-CoV-2, along with asymptomatic controls. The nasopharyngeal metabolome was characterized using an LC-MS/MS-based small-molecule screening kit capable of quantifying 141 analytes. A machine learning model identified 28 discriminating analytes and correctly categorized patients with a viral infection with an accuracy of 96% (R2=0.771, Q2=0.72). A second model identified 5 analytes to differentiate COVID19-infected patients from those with INFA or RSV with an accuracy of 85% (R2=0.442, Q2=0.301). Specifically, LysoPCaC18:2 concentration was significantly increased in COVID19 patients (P< 0.0001), whereas beta-hydroxybutyric acid, Met SO, succinic acid, and carnosine concentrations were significantly decreased (P< 0.0001). This study demonstrates that COVID19 infection results in a unique NP metabolomic signature with carnosine and LysoPCaC18:2 as potential therapeutic targets.

COVID-19 represents one of the greatest public health challenges of the 21st century.

Unlike most respiratory viruses, SARS-CoV-2 has a longer incubation period and infected individuals present with a spectrum of symptoms ranging from asymptomatic to severe clinical disease requiring hospitalization. The majority of SARS-CoV-2 infections occur via the nasal mucosa (1). Understanding the host-pathogen interactions in the nasal mucosa may provide valuable insight into the identification of novel therapeutic targets. These targets may be used to interrupt the acquisition and limit disease progression of SARS-CoV-2. We thus examined if the nasal metabolomic profile for COVID-19 was distinct from those of other respiratory viruses, and whether examining . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Previous studies involving the nasal metabolome in individuals infected with respiratory viruses, including rhinovirus (RV) and respiratory syncytial virus (RSV), reported that the nasal metabolome was virus-specific, despite indistinguishable clinical presentations in infected individuals (2) . Furthermore, the concentrations of specific nasal metabolites positively correlated with viral load and disease severity and predicted the need for positive pressure ventilation in patients with a high degree of sensitivity and specificity (84% and 86%, respectively) (3) . The predominant changes in the nasal metabolome observed in response to respiratory viruses were identified to be hostderived, although some metabolite concentrations correlated with colonization with Haemophilus influenzae, Streptococcus pneumoniae and Moraxella catarrhalis (2) .

These studies suggest that evaluating metabolic signatures in the nasopharynx of COVID-19 patients compared to other respiratory viruses may provide insight into important host-mediated antiviral responses, further elucidate changes that may be occurring in the nasal microbial environment and potentially identify new therapeutic targets against COVID-19.

In this study we hypothesized that SARS-CoV-2 induces a characteristic change to the nasal metabolome of patients and that could be used to both identify metabolites important in pathogenicity and potential therapeutic targets. We thus implemented a targeted LC-MS/MS metabolomics approach to (i) characterize small-molecule profiles in viral transport media from NP swabs of patients infected with INFA, RSV or COVID- 19 and asymptomatic controls; (ii) identify COVID-19 specific metabolite patterns; and . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint (iii) explore potential therapeutic pathways based on significant metabolites identified by a supervised machine learning model.

A total of 210 individuals were included in this study, comprising four patient groups: 44 asymptomatic controls (AC), 55 patients positive for SARS-CoV-2 (COV), 55 patients positive for INFA and 56 positive for RSV (Table 1) We studied the nasopharyngeal metabolome in patients who underwent standard-of-care or screen testing for respiratory infection. Viral transport media (VTM) from clinical samples were analyzed using a targeted, small-molecule screening kit (TMIC Prime) capable of quantifying 141-analytes over six chemical classes using a combination of LC-MS/MS and flow-injection analysis-MS/MS ( Supplementary Fig S1) .

Upon examination of individual analytes, 44 chemical species exhibited at least double the concentration in patient samples as compared with blank VTM, comprising amino acids (N=17), organic acids (N=4), biogenic amines (N=14) acylcarnitines (N=1) and lipids (N=8). Similarly, the feature selection step of our machine learning pipeline identified a subset of 28 metabolites that differed significantly between infected patients and asymptomatic controls, and 5 metabolites that distinguished SARS-CoV-2 from the other respiratory diseases. Combining these yielded 30 unique metabolites that we prioritized for multivariate analysis, including amino acids (N=15), organic acids (N=4), acylcarnitines (N=1), lipids (N=4), biogenic amines (N=5) and total hexoses (N=1) (Supplementary Table S1 ). Multivariate modelling of metabolite profiles scaled to blank . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint 6 VTM using partial-least discriminant analysis (PLS-DA) revealed separation of the four patient groups ( Figure 1A , B). A second model using orthogonal partial least squares discriminant analysis (OPLS-DA) focused on differences between asymptomatic patients and those with respiratory illness (INFA, RSV or COVID19) (Fig. 1C ). Using half of the data for training and the other half for testing, this model had an accuracy of 96%, a sensitivity of 98% and specificity of 86% (R 2 =0.771, Q 2 =0.72) in differentiating between groups (Fig. 1E) OPLS-DA loadings for the 30 unique metabolites that differed significantly among patient groups and controls are shown in Figure 2A . Twenty-eight analytes exhibited increased concentrations in patients with respiratory infections as compared with asymptomatic controls, including amino acids, lipids, organic acids, and biogenic amines. Most importantly, a smaller subset of analytes was observed to be specifically increased (LysoPCaC18:2) or decreased (MetSO, beta hydroxy-butyric acid, carnosine, and succinic acid) in COVID19 patients as compared with INFA or RSV patients ( Figure   2B ). Interestingly, carnosine and succinic acid were not found to be important factors in differentiating all respiratory patients from controls (Figure 2A and 2B). Despite their ability to distinguish among patient groups, none of these five metabolites correlated significantly with VL (qRT-PCR CT) for any of the three respiratory viruses ( Supplementary Fig. S2 ).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint Using targeted LC-MS/MS-based metabolomics, we identified unique metabolite profiles associated with the nasopharynx of patients with common respiratory infections. We observed striking differences in signatures that could be used to differentiate asymptomatic controls from patients with COVID19, INFA or RSV. Furthermore, we identified a COVID-19-specific signature that was characterized by altered concentrations of LysoPCaC18:2, beta-hydroxybutyric acid, Met SO, succinic acid, and carnosine, relative to INFA and RSV.

While several metabolomics studies related to COVID-19 have emerged, the use of both targeted and untargeted approaches applied to a range of biosamples makes comparing results among studies challenging (4) . The current study is unique as it employed a targeted approach to profile VTM acquired from standard-of-care swab kits from a diverse cohort of patients with qRT-PCR-confirmed COVID-19, IFNA or RSV, as well as asymptomatic controls. Although two previous studies have analyzed NP swabs, one study assessed VTM using matrix-assisted laser desorption/ionization mass spectrometry (MALDI-TOF MS) (5) . The other study analyzed fresh swabs directly by ambient ionization methods including DESI and LD-REIMS and focused on lipid profiling (6) . Both studies revealed diagnostic accuracies of >80%. At least three studies investigating the serum metabolome of COVID-19 patients identified changes in the tryptophan-kynurenine pathway associated with regulation of inflammation (7) . We also observed an increase in kynurenine concentration in our respiratory model, but this metabolite was not COVID-19 specific. Our results are consistent with Bo Shen et al. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 23, 2021.

8 interleukin-6 stratified COVID-19 patients compared to controls purportedly due to renal dysfunctional and marked alterations in nitrogen metabolism (7) . In contrast, we saw an increase in amino acids in respiratory virus infection compared to control, yet this may reflect a difference in systemic (serum) vs local (nasopharynx) compartments sampled.

Of the three studies involving serum metabolome analysis, Blasco et al. (9) calculated a diagnostic accuracy for COVID-19 as 74%.

A brief literature review of analytes in the COVID-19 metabolomic profile yields potential insight into the infection pathway (Fig 3) . In particular, carnosine and LYSOC18:2 had strong loadings in the OPLS-DA model, with the former decreasing, and the latter increasing, in COVID-19 patients relative to patients with INFA or RSV.

Carnosine, a naturally occurring dipeptide, has a wide range of protective effects in humans, which are largely attributed to its powerful antioxidant actions (10) . Several mechanisms could explain the depleted levels of carnosine in COVID-19 patients. First, decreased carnosine levels may signify decreased production of the dipeptide by the host. The olfactory system is among the richest sources of carnosine in humans (11, 12) , and carnosine present in the nasal swabs likely originated from the olfactory epithelium at the roof of the nasal cavity. The downregulation of carnosine could reflect decreased biosynthesis/secretion by olfactory sensory nerves or progressive loss of these neurons. Second, reduced carnosine levels may be the result of increased dipeptide degradation. Carnosine is largely metabolized by carnosinase-1 (CN1) (10) .

Although CN1 is expressed by the human olfactory epithelium (13) , the nasal cavity is not typically considered a site of high carnosinase activity, such that intranasal administration of carnosine has been employed in a preclinical model of Parkinson . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint disease as a means to avoid degradation by carnosinase (14) . A third, and more probable explanation, is that the diminished carnosine levels indicate depleted carnosine stores. Our data show that even basal levels of the dipeptide are completely exhausted in COVID-19 patients. Saadah et al. (15) predicted that COVID-19-induced oxidative stress would result in carnosine depletion. The protective effects of carnosine are widely attributed to its antioxidant, anti-glycation, and anti-inflammatory properties (10) . For instance, reduced circulating levels of low-density lipoprotein (LDL) has been associated with increased risk of acute kidney injury in COVID-19 patients (16), and carnosine, known to block lipid peroxynitrite-mediated modification of human LDL at physiological levels (17), may protect against LDL degradation. Intriguingly, recent papers suggest that carnosine may also protect against SARS-CoV-2 infection through more specific mechanisms. Molecular docking and modelling studies identified carnosine as the most promising drug candidate to prevent the binding of SARS-CoV-2 to the ACE2 receptor (15) . Sustentacular cells of the olfactory system co-express the ACE2 receptor as well as TMPRSS2, a protease that facilitates viral entry, making these cells highly susceptible to SARS-CoV-2 (18) . Infection of these cells has been implicated in anosmia 9 , a recognized symptom of COVID-19 (19) . Given that the olfactory epithelium is a major producer of carnosine and this dipeptide's vital neuroprotective role in this system (20) , loss of carnosine may lead to olfactory nerve damage resulting in anosmia in COVID-19.

Severe cases of COVID-19 are associated with multi-organ damage arising from oxidative stress (21) , raising the possibility of global depletion of carnosine in these patients and underscoring its therapeutic potential (22) . Recently, Kulikova and . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. whereby LysoPCaC18:2 was higher in COVID19 samples compared to healthy controls (28) . Additionally, a study inhibiting Cytosolic Phospholipase A2α (cPLA2α) which produces lysophospholipids, had significant effects on lowering coronavirus RNA and protein accumulation due to the importance of phospholipids in the creation of replicative organelles, emphasizing the therapeutic potential of lipid metabolism pathways. Moreover, the same study found that inhibition of cPLA2α had no impact on the replication of influenza A thus, consistent with the COVID19-specific pattern of LysoPCaC18:2 concentration observed (29) .

A limitation of our study was that extensive validation of the TMIC Prime kit for use with VTM was not conducted, including evaluation of matrix effects arising from VTM. While recovery of most synthetic metabolites was demonstrated at a single level . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 23, 2021. in VTM, accuracy may have been affected by matrix interferences. For the purposes of this pilot study, we used the TMIC Prime kit as a rapid screening method to evaluate broad metabolite profiles in patients with respiratory diseases that will form the basis of more refined assays for individual metabolites in future studies. While internal standards corrected for recovery of analytes from VTM, we could not account for differences in yields of NP swabs. Furthermore, assay performance metrics for certain lipids such as LysoPCaC18:2 were not determined as we lacked a synthetic standard, and the internal standard used for this lipid was non-specific.

In conclusion, we demonstrated that the metabolome of the nasopharynx can be measured from clinical nasal swabs, and that metabolite profiles identified using machine learning methods can differentiate patients with COVID-19 from other respiratory virus infections (e.g. INFA/RSV). Our study identified key metabolites specifically altered in COVID19 such as carnosine and LysoPCaC18:2 that have previously been implicated in viral replication and symptom generation. This enables us to propose mechanisms contributing to viral infection and propagation as well as potential targets for COVID-19 therapy.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint

All experimental protocols were approved by and conducted in accordance with the Queen's University Health Sciences and Affiliated Teaching Hospitals Research Ethics Board (HSREB Files 6029794, 6029811) . For assay of amino acids, biogenic amines, and lipids, a second aliquot of VTM were prepared as described above. Samples were aliquoted onto a filter paper disc in each well of the 96-well filter plate. Samples were dried for 30 minutes on an N2 evaporator. For derivatization of amino acids and biogenic amines, a 5% solution of . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint phenyl-isothiocyanate (PITC) was prepared in equal parts ethanol/pyridine/water. 50 µL of the 5% PITC solution was added to each filter paper. The plate was covered and incubated at room temperature for 20 minutes. The plate was dried for 90 min to remove excess liquid. 300 µL of 5 mM ammonium acetate in methanol was added to each well, and the plate was shaken at room temperature for 30 minutes to extract analytes from the filter paper. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint MS/MS were quantified using a relative quantification approach using a single representative internal standard for each analyte class. Supplementary Table 2 presents assay performance parameters for prioritized metabolites measured by LC-MS/MS. Four quality control samples based on solution standards (QC 1-3) and a lowlevel spiked VTM sample were measured 3 times on each of 3 assay days. 89% of measurements in solution standards were within 20% of target values, and all but one analyte exhibited total CVs of <20%. Total CVs for spiked VTM was more variable, with only 52% of analyte measurements exhibiting CVs of <20%. Mean % differences from target concentration was +28% (range: -14.3-112%). Poorer assay performance metrics in VTM are likely due to the presence of matrix interferences that we were unable to evaluate in the current study.

Our fully open and reproducible analysis pipeline, written in R (v 4.3), is available online where optimal separation of patient groups was observed. Orthogonal partial least squares discriminant analysis was used to plot analyte profiles among patient groups. In panel C, all patients with a respiratory illness were grouped into a single category and compared to asymptomatic subjects. In panel D, COVID19 patients were compared to all other patients with influenza A and RSV were tr into a single category. The 95% confidence region is circled for each category. Confusion matrices based on test/train cohorts using 50% of the data are shown in panels E and F from which the accuracy, sensitivity and specificity of identifying patients with a respiratory infection in general (E) or patients with COVID19 among patients with respiratory illness (F) was determined. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted June 23, 2021. ; https://doi.org/10.1101/2021.06.18.21259150 doi: medRxiv preprint

The central role of the nasal microenvironment in the transmission, modulation, and clinical progression of SARS-CoV-2 infection

Respiratory Syncytial Virus and Rhinovirus Bronchiolitis Are Associated With Distinct Metabolic Pathways

Associations of Nasopharyngeal Metabolome and Microbiome with Severity among Infants with Bronchiolitis. A Multiomic Analysis

Mass Spectrometry Techniques in Emerging Pathogens Studies: COVID-19 Perspectives

Detection of SARS-CoV-2 in nasal swabs using MALDI-MS

Rapid Detection of SARS-CoV2 by Ambient Mass Spectrometry Techniques

COVID-19 infection alters kynurenine and fatty acid metabolism, correlating with IL-6 levels and renal status

Proteomic and Metabolomic Characterization of COVID-19

The specific metabolome profiling of patients infected by SARS-COV-2 supports the key role of tryptophan-nicotinamide pathway and cytosine metabolism

Physiology and pathophysiology of carnosine

Carnosine-related dipeptides in neurons and glia

Carnosine in the primary olfactory pathway

Cellular localization of carnosinase in the human nasal mucosa

Evaluation of Carnosine Intervention in the Thy1-aSyn Mouse Model of Parkinson's Disease

Carnosine to Combat Novel Coronavirus (nCoV): Molecular Docking and Modeling to Cocrystallized Host Angiotensin-Converting Enzyme 2 (ACE2) and Viral Spike Protein

Alterations in the lipid profile associate with a dysregulated inflammatory, prothrombotic, anti-fibrinolytic state and development of severe acute kidney injury in coronavirus disease 2019 (COVID-19): A study from

Prevention of peroxynitrite-dependent damage by carnosine and related sulphonamido pseudodipeptides

SARS-CoV-2 Receptors and Entry Genes Are Expressed in the Human Olfactory Neuroepithelium and Brain

COVID-19 and anosmia: A review based on up-to-date knowledge

Neuroprotective effect of carnosine in the olfactory bulb after vanadium inhalation in a mouse model

COVID-19 and cytokine storm syndrome: can what we know about interleukin-6 in ovarian cancer be applied?

COVID-19 and Senotherapeutics: Any Role for the Naturally-occurring Dipeptide Carnosine?

A new derivative of acetylsalicylic acid and carnosine: synthesis, physical and chemical properties, biological activity

An overview of the pathogenic mechanisms involved in severe cases of COVID-19 infection, and the proposal of salicyl-carnosine as a potential drug for its treatment

Characterization of the Lipidomic Profile of Human Coronavirus-Infected Cells: Implications for Lipid Metabolism Remodeling upon Coronavirus Replication

Lipid-based therapies against SARS-CoV-2 infection. Reviews in Medical Virology

Omics-Driven Systems Interrogation of Metabolic Dysregulation in COVID-19 Pathogenesis

Large-Scale Plasma Analysis Revealed New Mechanisms and Molecules Associated with the Host Response to SARS-CoV-2

Inhibition of Cytosolic Phospholipase A(2)α Impairs an Early Step of Coronavirus Replication in Cell Culture

Chemical Composition of Commercial Cow's Milk

Analysis of the Human Adult Urinary Metabolome Variations with Age, Body Mass Index, and Gender by Implementing a Comprehensive Workflow for Univariate and OPLS Statistical Analyses

The authors would like to thank KHSC and the Clinical