A comparative NMR-based metabolomics study of lung parenchyma of severe COVID ‐ 19 patients

The COVID-19 pandemic was the leading cause of mortality due to a single infectious agent in 2020 and 2021. The fatality rate for individuals admitted to the ICU for this condition was over 60%. A metabolomic analysis based on nuclear magnetic resonance was performed on patients who died in the ICU due COVID-19 and other fatal respiratory diseases. Although there is information on metabolic signatures in COVID-19 patients' serum and plasma, little is known about the infection site in the lungs. We found statistically signicant differences between metabolites related to energy metabolism and inammatory processes, revealing a unique metabolic prole. Large-Scale Plasma Analysis Revealed New Mechanisms and Molecules Associated with the Host Response to SARS-CoV-2.


Introduction
The ongoing Coronavirus disease 2019 (COVID-19) could cause respiratory and systemic compromise such as pneumonias-like symptoms, acute respiratory distress syndrome (ARDS), and multiorgan failure 1 . The majority of patients diagnosed with COVID-19 attended the disease at home, 5-6% needed hospitalization in moderate care and between 1-2% required admission in the Intensive Care Unit (ICU).
In case series reported in the last year, hospital mortality was around 59% in patients admitted to the ICU, increasing up to 80% in those who required mechanical ventilation 2 .
Comorbidities including hypertension, diabetes, cardiovascular disease, and age may be potential risk factors among patients admitted in ICU 3 . On the other hand, viral load and also dysbiosis of respiratory and gut microbiota could in uence the clinical manifestations of COVID-19 4 . There is still no effective therapy for severe ill patients and a comprehensive understanding in metabolic changes and sepsis biomarkers are needed to address an improvement in the clinical management of patients.
Since it is now well accepted that COVID-19 is a systemic disease, several studies have employed metabolomic approaches to better understand the metabolic pathways involved in COVID-19 pathogenesis. Although there are many metabolomic studies in respiratory infections, the vast majority are focused on serum, plasma and bronchoalveolar lavage uid samples. Recently, metabolomic studies in serum of patients with ARDS due to COVID-19 revealed an altered amino acids metabolism, lipid metabolism, glycolysis and anaplerotic metabolism, suggesting an alteration in energy pathways, in ammatory response and oxidative stress 5 .
In this work we used an NMR-based non-targeted metabolomics approach to characterize the metabolome of lung parenchyma from fatal COVID-19.

Methods
The study was approved by the Ethics Committee from "Hospital Español -ASSE", a COVID-19 reference public hospital in Montevideo, Uruguay. Fragments of lung tissue were collected during clinical autopsy performed on patients with COVID-19 (n = 8) between November 2020 and February 2021 admitted in the ICU. Con rmation of infection by SARS-CoV-2 was made by RT-qPCR during ICU stay and informed consent was signed by direct family members before autopsy. All samples were obtained in the rst 2 h post-mortem and stored at -80°C.
On the other hand, lung fragments from non-COVID-19 deceased patients were collected between December 2016 and June 2018 also at Hospital Español -ASSE. This group includes microbiological and serological positive results for Klebsiella pneumoniae, Leptospira interrogans, and respiratory syncytial virus (n = 6). In all cases, lung tissues were obtained immediately after con rmation of death, and were stored immediately at -80°C for further studies.
The inclusion criteria for both groups were adults older than 18 years, admitted in the ICU with respiratory sepsis, septic shock or multi-organ failure, and all patients were mechanically ventilated.
An adaptation of previously published methods 6 was performed and lung fragment samples were homogenized and extracted with CHCl 3 /MeOH/H 2 O. Brie y, lung tissue was homogenized in water and methanol in a Bullet Blender (Next Advance, USA). Subsequently, chloroform was added for the nal ratio of 8:4:3 CHCl 3 :MeOH:H 2 O respectively, mixed in a vortex for 5 minutes, and centrifuged for 5 minutes at 5000g. Aqueous phase was lyophilized and resuspended in buffer phosphate in deuterium water (D 2 O) pH 7,4. Water-suppressed 1D-NOESY 1 H NMR spectra of aqueous tissue extracts were obtained at 500 MHz in a Bruker AVANCE III 500. A spectral width of 10 KHz, a data size of 32 K, and a total of 128 scans were employed to record each spectrum, using a relaxation delay of 4 s between scans. When required, gradient enhanced heteronuclear single quantum correlation (HSQC) protocol and 1D total correlated spectroscopy (TOCSY) spectra were acquired using standard pulse sequences provided with the spectrometer.

Results
As indicated in Table 1, all the patients in this study had acute respiratory sepsis and had high APACHE-II scores upon admission (mean 20.6 ± 8.4 points). Patients required mechanical ventilation and were on vasopressor support in all cases. Multiorgan failure occurred in all of the COVID-19 patients, with an average ICU stay of 17.6 ± 4.9 days. When compared to non-COVID pneumonias, they also had a higher percentage of comorbidities on admission (diabetes, hypertension, COPD, or obesity) and a lower PAFI (PaO 2 /FiO 2 ) score. The latter group spent an average of 11.2 ± 8.3 days in the ICU and had a slightly lower rate of multi-organ failure (83%). a Standard deviation, b Chronic obstructive pulmonary disease, c The arterial oxygen pressure/inspired fraction of oxygen (PaO2/FiO2 or PAFI) is a gas exchange indicator that has been widely used to assess the severity of lung damage in critically ill patients. d Acute respiratory distress syndrome As shown in Table 2, the analysis indicates different metabolomic pro les between both cohorts, showing an increase or decrease of some speci c metabolites in the tissue samples.
We compared the NMR spectra of eight lung tissue extracts from COVID-19 autopsies, and six non-COVID-19 autopsies. Although the low number of samples of each condition, principal component analysis (PCA) showed good discrimination between groups [see Additional le 1]. In addition, statistical differences between groups were observed. Orthogonal projections to latent structures discriminant analysis and statistical total correlation spectroscopy analyses of the resulting data, in combination with classical NMR dereplication experiments, allowed us to identify 21 metabolites among the two cohorts, 11 of them were differentially expressed ( Table 2). The amino acids valine, alanine, methionine, glycine, tryptophane, phenylalanine, tyrosine, asparagine and the metabolic intermediate fumarate, were signi cantly increased in samples from COVID-19 patients. On the other hand, choline and α-glucose were signi cantly decreased among these samples.

Discussion
The aim of this study was to compare the metabolome of lung parenchyma in fatal COVID-19 from other severe infections and whether we could discriminate a distinctive metabolic pro le using NMR-based non-targeted metabolomics technique. We found good separation in PCA analysis which indicates a distinct metabolic footprint in lung parenchyma of COVID-19 infection.
In rst place, lactate was the most widely expressed metabolite across cohorts with no statistically signi cant differences between them. This nding is consistent with the known fact that high plasma lactate concentrations are a marker of poor prognosis and an indicative of metabolic acidosis in critically ill patients and were expected to be higher in both groups 7 . In fatal COVID-19, we also found a signi cant increase in several tissue amino acids, such as phenylalanine, alanine, valine, methionine, and asparagine. In patients with COVID-19, amino acid metabolism may be altered as a compensatory mechanism to restore NAD reducing power levels in the context of lung injury and hypoxia 8 . In ammatory markers like tryptophan 9 , on the other hand, are higher in COVID-19 than in other acute pneumonias. Lung injury in COVID-19 leading to severe refractory hypoxemia was observed in recently published series, and is one of the main causes of mortality in this study.
Choline levels were also found to be signi cantly lower in COVID-19 samples. This has been reported in previous studies in the serum of severe COVID-19 patients, where an increase in choline consumption caused by activation of macrophage TLRs receptors was linked to extracellular cytokine secretion 11 . The presence of pro-in ammatory components in bronchoalveolar lavage uid is elevated even in severe COVID-19 patients treated with glucocorticoids, suggesting that slowing down the cytokine storm is a critical strategy for disease control 12 .
One of the main limitations of our study is the low number of cases, but we were able to nd statistical differences in metabolite concentrations between groups. We also excluded healthy controls since we decided to compare COVID-19 biomarkers with severe non COVID-19 pneumonias.
In conclusion, distinct metabolic signatures including energy metabolism and in ammatory pathways have been identi ed that distinguish COVID-19 from other deadly pneumonias induced by multiple respiratory infections.

Declarations
Ethics approval and consent to participate The study was approved by the hospital's ethical committee (Bioethics and Research Ethics Committee -Hospital Español) and carried out in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki's guiding principles. Direct family members or the patient's legally authorized representative provided informed consent if the patient was unable to consent for any reason.

Consent of publication
Not applicable