A Time-Series Metabolomic Analysis of SARS-CoV-2 Infection in a Ferret Model
Abstract
:1. Introduction
2. Materials and Methods
2.1. Ferret Challenge and Sample Collection
2.2. Metabolomics Analysis
2.3. Statistical Analysis and Data Integration
3. Results and Discussions
3.1. Viral Shedding following Challenge
3.2. Central Carbon Metabolism Variance in the Nasal Wash Samples
3.3. Chemical and Pathway Analysis of the Central Carbon Metabolism
3.4. Multivariate Analysis of the Central Carbon Metabolism and Discovery Metabolites
3.5. Time-Series Metabolomics Analysis of the Progression of SARS-CoV-2 Infection
4. Conclusions
Supplementary Materials
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- WHO. Update 95-SARS: Chronology of a Serial Killer. Available online: https://www.who.int/emergencies/disease-outbreak-news/item/2003_07_04-en (accessed on 26 September 2022).
- WHO. Summary of Probable SARS Cases with Onset of Illness from 1 November 2002 to 31 July 2003. Available online: https://www.who.int/publications/m/item/summary-of-probable-sars-cases-with-onset-of-illness-from-1-november-2002-to-31-july-2003 (accessed on 26 September 2022).
- WHO. Middle East Respiratory Syndrome Coronavirus (MERS-CoV). Available online: https://www.who.int/news-room/fact-sheets/detail/middle-east-respiratory-syndrome-coronavirus-(mers-cov) (accessed on 26 September 2022).
- WHO. Middle East Respiratory Syndrome Coronavirus (MERS-CoV)—Qatar. Available online: https://www.who.int/emergencies/disease-outbreak-news/item/2022-DON370 (accessed on 26 September 2022).
- WHO. Weekly Epidemiological Update on COVID-19—25 May 2022. Available online: https://www.who.int/publications/m/item/weekly-epidemiological-update-on-covid-19---25-may-2022 (accessed on 26 September 2022).
- Wang, H.; Paulson, K.R.; Pease, S.A.; Watson, S.; Comfort, H.; Zheng, P.; Aravkin, A.Y.; Bisignano, C.; Barber, R.M.; Alam, T.; et al. Estimating excess mortality due to the COVID-19 pandemic: A systematic analysis of COVID-19-related mortality, 2020–21. Lancet 2022, 399, 1513–1536. [Google Scholar] [CrossRef]
- The Pandemic’s True Death Toll. Available online: https://www.economist.com/graphic-detail/coronavirus-excess-deaths-estimates (accessed on 26 September 2022).
- Lei, R.; Huo, R.; Mohan, C. Current and emerging trends in point-of-care urinalysis tests. Expert Rev. Mol. Diagn. 2020, 20, 69–84. [Google Scholar] [CrossRef] [PubMed]
- Holmes, E.; Wilson, I.D.; Nicholson, J.K. Metabolic Phenotyping in Health and Disease. Cell 2008, 134, 714–717. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Wu, D.; Shu, T.; Yang, X.; Song, J.-X.; Zhang, M.; Yao, C.; Liu, W.; Huang, M.; Yu, Y.; Yang, Q.; et al. Plasma metabolomic and lipidomic alterations associated with COVID-19. Natl. Sci. Rev. 2020, 7, 1157–1168. [Google Scholar] [CrossRef]
- Barberis, E.; Amede, E.; Tavecchia, M.; Marengo, E.; Cittone, M.G.; Rizzi, E.; Pedrinelli, A.R.; Tonello, S.; Minisini, R.; Pirisi, M.; et al. Understanding protection from SARS-CoV-2 using metabolomics. Sci. Rep. 2021, 11, 13796. [Google Scholar] [CrossRef] [PubMed]
- Shen, B.; Yi, X.; Sun, Y.; Bi, X.; Du, J.; Zhang, C.; Quan, S.; Zhang, F.; Sun, R.; Qian, L.; et al. Proteomic and Metabolomic Characterization of COVID-19 Patient Sera. Cell 2020, 182, 59.e15–72.e15. [Google Scholar] [CrossRef]
- Villoslada, P.; Alonso, C.; Agirrezabal, I.; Kotelnikova, E.; Zubizarreta, I.; Pulido-Valdeolivas, I.; Saiz, A.; Comabella, M.; Montalban, X.; Villar, L.; et al. Metabolomic signatures associated with disease severity in multiple sclerosis. Neurol. Neuroimmunol. Neuroinflammation 2017, 4, e321. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Jacyna, J.; Kordalewska, M.; Artymowicz, M.; Markuszewski, M.; Matuszewski, M.; Markuszewski, M.J. Pre- and Post-Resection Urine Metabolic Profiles of Bladder Cancer Patients: Results of Preliminary Studies on Time Series Metabolomics Analysis. Cancers 2022, 14, 1210. [Google Scholar] [CrossRef] [PubMed]
- Huang, X.; Zeng, J.; Zhou, L.; Hu, C.; Yin, P.; Lin, X. A New Strategy for Analyzing Time-Series Data Using Dynamic Networks: Identifying Prospective Biomarkers of Hepatocellular Carcinoma. Sci. Rep. 2016, 6, 32448. [Google Scholar] [CrossRef] [Green Version]
- Li, Y.; Xue, H.; Fang, S.; Wang, G.; Wang, Y.; Wang, T.; Shi, R.; Wu, J.; Ma, Y. Time-series metabolomics insights into the progressive characteristics of 3,5-diethoxycarbonyl-1,4-dihydrocollidine-induced cholestatic liver fibrosis in mice. J. Pharm. Biomed. Anal. 2021, 198, 113986. [Google Scholar] [CrossRef] [PubMed]
- Li, R.; He, H.; Fang, S.; Hua, Y.; Yang, X.; Yuan, Y.; Liang, S.; Liu, P.; Tian, Y.; Xu, F.; et al. Time Series Characteristics of Serum Branched-Chain Amino Acids for Early Diagnosis of Chronic Heart Failure. J. Proteome Res. 2019, 18, 2121–2128. [Google Scholar] [CrossRef]
- Beale, D.J.; Shah, R.; Karpe, A.V.; Hillyer, K.E.; McAuley, A.J.; Au, G.G.; Marsh, G.A.; Vasan, S.S. Metabolic Profiling from an Asymptomatic Ferret Model of SARS-CoV-2 Infection. Metabolites 2021, 11, 327. [Google Scholar] [CrossRef]
- Au, G.G.; Marsh, G.A.; McAuley, A.J.; Lowther, S.; Trinidad, L.; Edwards, S.; Todd, S.; Barr, J.; Bruce, M.P.; Poole, T.B.; et al. Characterisation and natural progression of SARS-CoV-2 infection in ferrets. Sci. Rep. 2022, 12, 5680. [Google Scholar] [CrossRef]
- Bauer, D.C.; Tay, A.P.; Wilson, L.O.W.; Reti, D.; Hosking, C.; McAuley, A.J.; Pharo, E.; Todd, S.; Stevens, V.; Neave, M.J.; et al. Supporting pandemic response using genomics and bioinformatics: A case study on the emergent SARS-CoV-2 outbreak. Transbound. Emerg. Dis. 2020, 67, 1453–1462. [Google Scholar] [CrossRef] [Green Version]
- Marsh, G.A.; McAuley, A.J.; Brown, S.; Pharo, E.A.; Crameri, S.; Au, G.G.; Baker, M.L.; Barr, J.A.; Bergfeld, J.; Bruce, M.P.; et al. In vitro characterisation of SARS-CoV-2 and susceptibility of domestic ferrets (Mustela putorius furo). Transbound. Emerg. Dis. 2022, 69, 297–307. [Google Scholar] [CrossRef]
- Muñoz-Fontela, C.; Dowling, W.E.; Funnell, S.G.P.; Gsell, P.-S.; Riveros-Balta, A.X.; Albrecht, R.A.; Andersen, H.; Baric, R.S.; Carroll, M.W.; Cavaleri, M.; et al. Animal models for COVID-19. Nature 2020, 586, 509–515. [Google Scholar] [CrossRef]
- Muñoz-Fontela, C.; Widerspick, L.; Albrecht, R.A.; Beer, M.; Carroll, M.W.; de Wit, E.; Diamond, M.S.; Dowling, W.E.; Funnell, S.G.P.; García-Sastre, A.; et al. Advances and gaps in SARS-CoV-2 infection models. PLoS Pathog. 2022, 18, e1010161. [Google Scholar] [CrossRef]
- Pallister, J.; Middleton, D.; Crameri, G.; Yamada, M.; Klein, R.; Hancock, T.J.; Foord, A.; Shiell, B.; Michalski, W.; Broder, C.C. Chloroquine administration does not prevent Nipah virus infection and disease in ferrets. J. Virol. 2009, 83, 11979–11982. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Caly, L.; Druce, J.; Roberts, J.; Bond, K.; Tran, T.; Kostecki, R.; Yoga, Y.; Naughton, W.; Taiaroa, G.; Seemann, T. Isolation and rapid sharing of the 2019 novel coronavirus (SARS-CoV-2) from the first patient diagnosed with COVID-19 in Australia. Med. J. Aust. 2020, 212, 459–462. [Google Scholar] [CrossRef] [Green Version]
- Pang, Z.; Chong, J.; Zhou, G.; de Lima Morais, D.A.; Chang, L.; Barrette, M.; Gauthier, C.; Jacques, P.-É.; Li, S.; Xia, J. MetaboAnalyst 5.0: Narrowing the gap between raw spectra and functional insights. Nucleic Acids Res. 2021, 49, W388–W396. [Google Scholar] [CrossRef]
- Karpe, A.V.; Beale, D.J.; Morrison, P.D.; Harding, I.H.; Palombo, E.A. Untargeted metabolic profiling of Vitis vinifera during fungal degradation. FEMS Microbiol. Lett. 2015, 362, fnv060. [Google Scholar] [CrossRef] [PubMed]
- Barupal, D.K.; Fiehn, O. Chemical Similarity Enrichment Analysis (ChemRICH) as alternative to biochemical pathway mapping for metabolomic datasets. Sci. Rep. 2017, 7, 14567. [Google Scholar] [CrossRef] [Green Version]
- Cevik, M.; Kuppalli, K.; Kindrachuk, J.; Peiris, M. Virology, transmission, and pathogenesis of SARS-CoV-2. BMJ 2020, 371, m3862. [Google Scholar] [CrossRef] [PubMed]
- Hu, X.; Xing, Y.; Jia, J.; Ni, W.; Liang, J.; Zhao, D.; Song, X.; Gao, R.; Jiang, F. Factors associated with negative conversion of viral RNA in patients hospitalized with COVID-19. Sci. Total Environ. 2020, 728, 138812. [Google Scholar] [CrossRef] [PubMed]
- Cogliati Dezza, F.; Oliva, A.; Cancelli, F.; Savelloni, G.; Valeri, S.; Mauro, V.; Calabretto, M.; Russo, G.; Venditti, M.; Turriziani, O.; et al. Determinants of prolonged viral RNA shedding in hospitalized patients with SARS-CoV-2 infection. Diagn. Microbiol. Infect. Dis. 2021, 100, 115347. [Google Scholar] [CrossRef]
- Cevik, M.; Tate, M.; Lloyd, O.; Maraolo, A.E.; Schafers, J.; Ho, A. SARS-CoV-2, SARS-CoV, and MERS-CoV viral load dynamics, duration of viral shedding, and infectiousness: A systematic review and meta-analysis. Lancet Microbe 2021, 2, e13–e22. [Google Scholar] [CrossRef]
- Agarwal, V.; Venkatakrishnan, A.J.; Puranik, A.; Lopez-Marquez, A.; Challener, D.W.; Horo, J.C.; Badley, A.D.; Halamka, J.D.; Morice, W.G.; Soundararajan, V. Quantifying the prevalence of SARS-CoV-2 long-term shedding among non-hospitalized COVID-19 patients. MedRxiv 2020. [Google Scholar] [CrossRef]
- Gombar, S.; Chang, M.; Hogan, C.A.; Zehnder, J.; Boyd, S.; Pinsky, B.A.; Shah, N.H. Persistent detection of SARS-CoV-2 RNA in patients and healthcare workers with COVID-19. J. Clin. Virol. 2020, 129, 104477. [Google Scholar] [CrossRef]
- Bojkova, D.; Costa, R.; Reus, P.; Bechtel, M.; Jaboreck, M.-C.; Olmer, R.; Martin, U.; Ciesek, S.; Michaelis, M.; Cinatl, J. Targeting the Pentose Phosphate Pathway for SARS-CoV-2 Therapy. Metabolites 2021, 11, 699. [Google Scholar] [CrossRef] [PubMed]
- Mesri, E.A.; Lampidis, T.J. 2-Deoxy-d-glucose exploits increased glucose metabolism in cancer and viral-infected cells: Relevance to its use in India against SARS-CoV-2. IUBMB Life 2021, 73, 1198–1204. [Google Scholar] [CrossRef]
- Bai, X.; Narayanan, A.; Skagerberg, M.; Ceña-Diez, R.; Giske, C.G.; Strålin, K.; Sönnerborg, A. Characterization of the Upper Respiratory Bacterial Microbiome in Critically Ill COVID-19 Patients. Biomedicines 2022, 10, 982. [Google Scholar] [CrossRef] [PubMed]
- Forst, C.V.; Zeng, L.; Wang, Q.; Zhou, X.; Vatansever, S.; Tu, Z.; Zhang, B. Tissue Specific Age Dependence of the Cell Receptors Involved in the SARS-CoV-2 Infection. bioRxiv 2021. [Google Scholar] [CrossRef]
- Wang, M.-Y.; Zhao, R.; Gao, L.-J.; Gao, X.-F.; Wang, D.-P.; Cao, J.-M. SARS-CoV-2: Structure, Biology, and Structure-Based Therapeutics Development. Front. Cell. Infect. Microbiol. 2020, 10. [Google Scholar] [CrossRef] [PubMed]
- Sun, N.; Fernandez, I.E.; Wei, M.; Witting, M.; Aichler, M.; Feuchtinger, A.; Burgstaller, G.; Verleden, S.E.; Schmitt-Kopplin, P.; Eickelberg, O.; et al. Pharmacometabolic response to pirfenidone in pulmonary fibrosis detected by MALDI-FTICR-MSI. Eur. Respir. J. 2018, 52, 1702314. [Google Scholar] [CrossRef] [PubMed]
- Cao, W.; Feng, Q.; Wang, X. Computational analysis of TMPRSS2 expression in normal and SARS-CoV-2-infected human tissues. Chem.-Biol. Interact. 2021, 346, 109583. [Google Scholar] [CrossRef]
- Lodge, S.; Nitschke, P.; Kimhofer, T.; Wist, J.; Bong, S.-H.; Loo, R.L.; Masuda, R.; Begum, S.; Richards, T.; Lindon, J.C.; et al. Diffusion and Relaxation Edited Proton NMR Spectroscopy of Plasma Reveals a High-Fidelity Supramolecular Biomarker Signature of SARS-CoV-2 Infection. Anal. Chem. 2021, 93, 3976–3986. [Google Scholar] [CrossRef]
- Krishnan, S.; Nordqvist, H.; Ambikan, A.T.; Gupta, S.; Sperk, M.; Svensson-Akusjärvi, S.; Mikaeloff, F.; Benfeitas, R.; Saccon, E.; Ponnan, S.M.; et al. Metabolic Perturbation Associated With COVID-19 Disease Severity and SARS-CoV-2 Replication. Mol. Cell. Proteom. 2021, 20, 100159. [Google Scholar] [CrossRef]
- Lawler, N.G.; Gray, N.; Kimhofer, T.; Boughton, B.; Gay, M.; Yang, R.; Morillon, A.-C.; Chin, S.-T.; Ryan, M.; Begum, S.; et al. Systemic Perturbations in Amine and Kynurenine Metabolism Associated with Acute SARS-CoV-2 Infection and Inflammatory Cytokine Responses. J. Proteome Res. 2021, 20, 2796–2811. [Google Scholar] [CrossRef]
- Li, W.; Hulswit, R.J.; Widjaja, I.; Raj, V.S.; McBride, R.; Peng, W.; Widagdo, W.; Tortorici, M.A.; Van Dieren, B.; Lang, Y. Identification of sialic acid-binding function for the Middle East respiratory syndrome coronavirus spike glycoprotein. Proc. Natl. Acad. Sci. USA 2017, 114, E8508–E8517. [Google Scholar] [CrossRef] [Green Version]
- Tortorici, M.A.; Walls, A.C.; Lang, Y.; Wang, C.; Li, Z.; Koerhuis, D.; Boons, G.-J.; Bosch, B.-J.; Rey, F.A.; de Groot, R.J. Structural basis for human coronavirus attachment to sialic acid receptors. Nat. Struct. Mol. Biol. 2019, 26, 481–489. [Google Scholar] [CrossRef]
- Barberis, E.; Timo, S.; Amede, E.; Vanella, V.V.; Puricelli, C.; Cappellano, G.; Raineri, D.; Cittone, M.G.; Rizzi, E.; Pedrinelli, A.R. Large-Scale Plasma Analysis Revealed New Mechanisms and Molecules Associated with the Host Response to SARS-CoV-2. Int. J. Mol. Sci. 2020, 21, 8623. [Google Scholar] [CrossRef] [PubMed]
- Bruzzone, C.; Bizkarguenaga, M.; Gil-Redondo, R.; Diercks, T.; Arana, E.; García de Vicuña, A.; Seco, M.; Bosch, A.; Palazón, A.; San Juan, I.; et al. SARS-CoV-2 Infection Dysregulates the Metabolomic and Lipidomic Profiles of Serum. iScience 2020, 23, 101645. [Google Scholar] [CrossRef] [PubMed]
- Li, G.; Wu, X.; Zhou, C.-l.; Wang, Y.-m.; Song, B.; Cheng, X.-b.; Dong, Q.-f.; Wang, L.-l.; You, S.-s.; Ba, Y.-m. Uric acid as a prognostic factor and critical marker of COVID-19. Sci. Rep. 2021, 11, 17791. [Google Scholar] [CrossRef] [PubMed]
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Karpe, A.V.; Nguyen, T.V.; Shah, R.M.; Au, G.G.; McAuley, A.J.; Marsh, G.A.; Riddell, S.; Vasan, S.S.; Beale, D.J. A Time-Series Metabolomic Analysis of SARS-CoV-2 Infection in a Ferret Model. Metabolites 2022, 12, 1151. https://doi.org/10.3390/metabo12111151
Karpe AV, Nguyen TV, Shah RM, Au GG, McAuley AJ, Marsh GA, Riddell S, Vasan SS, Beale DJ. A Time-Series Metabolomic Analysis of SARS-CoV-2 Infection in a Ferret Model. Metabolites. 2022; 12(11):1151. https://doi.org/10.3390/metabo12111151
Chicago/Turabian StyleKarpe, Avinash V., Thao V. Nguyen, Rohan M. Shah, Gough G. Au, Alexander J. McAuley, Glenn A. Marsh, Sarah Riddell, Seshadri S. Vasan, and David J. Beale. 2022. "A Time-Series Metabolomic Analysis of SARS-CoV-2 Infection in a Ferret Model" Metabolites 12, no. 11: 1151. https://doi.org/10.3390/metabo12111151