High plasma concentration of non-esterified polyunsaturated fatty acids is a specific feature of severe COVID-19 pneumonia

COVID-19 pneumonia has specific features and outcomes that suggests a unique immunopathogenesis. Severe forms of COVID-19 appear to be more frequent in obese patients, but an association with metabolic disorders is not established. Here, we focused on lipoprotein metabolism in patients hospitalized for severe pneumonia, depending on COVID-19 status. Thirty-four non-COVID-19 and 27 COVID-19 patients with severe pneumonia were enrolled. Most of them required intensive care. Plasma lipid levels, lipoprotein metabolism, and clinical and biological (including plasma cytokines) features were assessed. Despite similar initial metabolic comorbidities and respiratory severity, COVID-19 patients displayed a lower acute phase response but higher plasmatic concentrations of non-esterified fatty acids (NEFAs). NEFA profiling was characterised by higher level of polyunsaturated NEFAs (mainly linoleic and arachidonic acids) in COVID-19 patients. Multivariable analysis showed that among severe pneumonia, COVID-19-associated pneumonia was associated with higher NEFAs, lower apolipoprotein E and lower high-density lipoprotein cholesterol concentrations, independently of body mass index, sequential organ failure (SOFA) score, and C-reactive protein levels. NEFAs and PUFAs concentrations were negatively correlated with the number of ventilator-free days. Among hospitalized patients with severe pneumonia, COVID-19 is independently associated with higher NEFAs (mainly linoleic and arachidonic acids) and lower apolipoprotein E and HDL concentrations. These features might act as mediators in COVID-19 pathogenesis and emerge as new therapeutic targets. Further investigations are required to define the role of NEFAs in the pathogenesis and the dysregulated immune response associated with COVID-19. Trial registration: NCT04435223.


Principal component analysis (PCA) identifies two patterns linking lipid metabolism and outcomes that enables distinguishing between non-COVID-19 and COVID-19 patients. Using
PCA, four factors were retained for interpretation. These factors preserve 58.4% of the total information contained in the 25 original correlated variables ( Figure S1). The first component represents 26.6% of the data variability and was correlated to initial severity (SOFA score, pneumonia severity index (PSI)) and acute phase response (CRP). Levels of HDL cholesterol, Apolipoprotein A1 (ApoA1), HDL size, PLTP activity, LDL, ApoB, and LDL size were negatively correlated with this first component, while PLTP activity was positively correlated ( Table 2). The second component represents 13.8% of the data variability and was correlated with outcomes (duration of mechanical ventilation, ICU length of stay). NEFAs concentrations were correlated with this second dimension, as were ApoA1 and HDL diameter ( Table 2). The third component (10.1% of the data variability) represents the relationship between HDL, LDL and VLDL metabolism, but there was not a strong correlation with the outcome. The fourth component (7.8% of the data variability) is correlated with glucose metabolism and body mass index (BMI) ( Table 2). The projection of patient data on the plan defined by factors 1 and 2 made it possible to discriminate COVID-19 from non-COVID-19 patients (Fig. 2). By plotting patients according to pneumonia etiology, we noticed that non-COVID-19 patients with pneumonia of proven bacterial origin were represented by high factor 1 values, while COVID-19 patients were represented by high factor 2 values and moderate factor 1 values (Fig. 2). Mul-tiple linear regression showed that COVID-19 status was associated with high plasma NEFAs, low HDL cholesterol levels and low ApoE plasma concentrations, independently from body mass index (BMI), SOFA score and plasma CRP (Table 3). In COVID-19 patients, we estimated a proper excess of 0.12 ± 0.06 mg/dl for NEFAs (p = 0.04), and reductions of 0. 36  www.nature.com/scientificreports/ poprotein E (ApoE) concentrations (p = 0.01). LDL cholesterol, HDL diameter and PLTP activity were independently associated with CRP concentration but not with COVID-19 status (Table 3).
Fatty acid metabolism cascade exploration. As phospholipase A2 (PLA 2 ) mediates the release of fatty acids from membrane phospholipids, including linoleic and arachidonic acids, we measured both, secreted PLA 2 (sPLA 2 ) and lipoprotein-associated PLA 2 (lp-PLA 2 ) plasma concentrations. Plasma sPLA2 and lp-PLA 2 concentrations were not significantly different between the two groups (respectively p = 0.93 and p = 0.25) (Table S1). In addition, we did not observe any correlation between plasma concentration of sPLA 2 and lp-PLA 2 and PUFAs (Table S3).  www.nature.com/scientificreports/ In addition, two final products of the arachidonic cascade were measured, namely leukotriene B4 and prostaglandin E2 (PGE2). We observed a correlation between arachidonic acid and leukotriene B4 concentrations (r = 0.302, p = 0.02) but not with PGE2 (r = − 0.048, p = 0.9) (Fig. 4).

Discussion
In patients with severe pneumonia, we report here an independent association between lipid metabolism disorders and COVID-19. Multivariate analysis showed that high NEFAs, low ApoE and low HDL cholesterol concentrations are significantly associated with COVID-19 status, independently of body mass index, baseline severity (SOFA score) and acute phase response (CRP level). The characterization of NEFAs highlighted an increase in polyunsaturated fatty acids (mainly linoleic and arachidonic acids). The concentrations of NEFAs, and particularly linoleic acids, were correlated with the duration of the mechanical ventilation.
COVID-19 pneumonia has unique features when compared with pneumonia from other origins, and it displays a specific inflammatory pathogenesis [17][18][19] . The beneficial effect of dexamethasone in the most severe forms of COVID-19 argues in favor of a dysregulated immune response that mediates lung injury and outcome 20 . We recently identified a unique cytokine response in COVID-19 patients with severe pneumonia with higher plasma GM-CSF and CXCL10 (a Th-1 chemokine) in COVID-19 patients, that were independently associated with the longer duration of mechanical ventilation 21,22 . Obesity has been described as a factor of poor prognosis in COVID-19 patients and associated with disease severity 1,2 . In various medical settings, obese patients display immune dysregulation (e.g. Th1 hyperpolarization) and metabolic disorders, and there is an obvious relationship between the two 23 . We assumed that metabolic disorders may contribute to the pathophysiology of COVID-19. www.nature.com/scientificreports/ In the present study, although obesity was highly frequent in COVID-19 patients (52% of patients), BMI, diabetes, dyslipidaemia and plasma fructosamine levels did not significantly differ between non-COVID-19 and COVID-19 patients. Therefore, obesity has a similar frequency in severe pneumonia regardless of the origin and was not likely to acts as a confounding factor between the 2 groups of patients. Sepsis is associated with profound modifications in HDL (dramatic reduction in HDL-cholesterol, extensive remodelling of HDL particles) and LDL metabolism (lower LDL cholesterol concentrations) 24 . However, the mechanisms underlying this decrease are poorly described. PLTP activity is known to increase during systemic inflammation 25 . HDL and LDL were found to be protective in patients with sepsis 26 . The lower PLTP activity we observed in COVID-19 patients is consistent with the lower proportion of extra-pulmonary organ dysfunction (septic shock, lactatemia, RRT and creatininemia) and the lower acute phase response (CRP levels). With multivariate analyses, negative correlations between CRP and both HDL and LDL cholesterol levels were also consistent with these clinical settings. By contrast, low HDL and ApoE concentrations were independently associated with COVID-19 status. Interestingly, HDL particles are endowed with immune-modulatory properties partly through interaction with lipid rafts 27 and HDL-scavenger receptor B type 1 (SR-B1) has been demonstrated to  www.nature.com/scientificreports/ facilitate SARS-CoV-2 entry into ACE2-expressing cells 14 . Consequently, we assume that the more pronounced drop in HDL associated with COVID-19 origin could be the consequence of a higher consumption of HDL particles and/or the upregulation of SR-B1 expression. However, as in the ApoCOVID study, we did not find any association between HDL concentrations and outcomes 11 . Then, we depict here higher plasma PUFAs concentrations (higher C18:2 n-6 and C20:4 n-6) and a higher proportion of PUFAs among NEFAs in COVID-19 patients as compared to non-COVID-19 patients. High plasma NEFAs concentrations (including PUFAs) were previously reported in COVID-19 patients as compared to healthy patients, independently of the acute phase response 28 . However, to our knowledge, there has not yet been a comparison between patients with severe pneumonia according to COVID-19 status. This absolute and relative increase in PUFAs supports a specific alteration of the fatty-acid metabolism of COVID-19 patients. PUFAs are precursors of eicosanoids and platelet activating factor. Such metabolite production could contribute to the dysregulated immune response and coagulopathy that characterize COVID-19. For example, MAPK activation has previously been suggested to activate cytosolic PLA 2 , which in turn increases thromboxane and platelet hyperactivity (31). These recent data support the hypothesis that PUFAs and the PLA 2 pathway are potential key players in the pathogenesis of COVID-19. In our study, we did not find any correlation between the two plasmatic forms of PLA 2 (namely s-PLA 2 and lp-PLA 2 ) and PUFAs, but we do not rule out an increase in cytosolic PLA 2 expression and activity in COVID-19 patients, that we were unable to measure. However, we found a marginal positive correlation between NEFAs (and linoleic acid) and the three cytokines previously identified in higher concentrations and associated with poorer outcomes in COVID-19 patients 21 . That suggests the possible involvement of NEFAs in the specific immune response associated with COVID-19. Interestingly, SARS-CoV-2 comprises a high-affinity pocket that specifically accretes linoleic acids, modifying ACE 2 interaction in-vitro 15 . The interaction between linoleic acids and SARS-CoV-2 represents an emerging appealing therapeutic target that requires further investigation. The correlations between NEFAs concentrations (and linoleic acids proportion) and the longer duration of mechanical ventilation we observed in COVID-19 strengthens these hypotheses.
To date, only corticosteroids have been proved to be effective in reducing mortality in severe COVID-19 20 , strengthening the host-immune response as one of the most promising therapeutic targets. Since corticosteroids have many side effects, targeted therapies likely to dampen the dysregulated immune response in COVID-19 www.nature.com/scientificreports/ are urgently needed. Our data highlight NEFAs as a potential driver of this dysregulated immune response that could drive poorer outcome. Restoring lipid homeostasis could represent a promising approach likely to dampen some features of the dysregulated immune response in COVID-19, and as a consequence outcomes. For this purpose, PLA-2 inhibitor is a potential candidate. In unselected patients with severe sepsis, PLA-2 inhibitors was shown to be ineffective in reducing mortality 29 . However, as for sepsis, personalized medicine with targeted immunomodulatory therapy, guided by immune-related biomarkers is mandatory to deal with the heterogeneity of the host-immune response 30 . Modulating lipid metabolism, using lipid-associated biomarkers (e.g. NEFAs concentrations) for decision making could represent a promising therapeutic strategy that deserve to be investigated in COVID-19.
There are some limitations in our work. Because this is an observational study, only association and not causality can be inferred. Statistical analysis may suffer from a lack of power since only 61 patients were included, but this was compensated, at least in part, by the strong observed differences and the use of several statistical methods including PCA and multivariate analysis. However, given this small sample size, our results need to be confirmed in larger cohorts. At this stage, the comparison of fatty acids metabolism and immune response between non-COVID-19 and COVID-19 patients with severe pneumonia is still scarce in the literature. We used CRP as a surrogate for acute phase response. However, CRP is also a known acute phase reactant 31 . Finally, we used plasma collection, and cytosolic PLA 2 could not be measured to ascertain of the origin of plasma NEFAs.
In conclusion, COVID-19 was independently associated with high concentrations of polyunsaturated NEFAs and low concentrations of apolipoprotein E and HDL. These parameters may be key mediators of the COVID-19 pathogenesis and potential therapeutic targets. Further investigations are required to explore the role of cytosolic phospholipase A2 and PUFAs in the dysregulated immune response and the pathogenicity of COVID-19.

Methods
Design. This ancillary study of the ongoing prospective Lymphonie study (ClinicalTrials.gov NCT03505281) was registered in ClinicalTrials.gov (NCT04435223) 21 . The study fulfilled legal and ethics requirements: Approval was obtained from the CPP (Full name: Comité de Protection des Personnes SUD MEDITERRANEE V (reference 17.092); Registration number: 2017-A03404-49) for the original study and an amendment was obtained to include supplementary patients with severe COVID-19. All subjects (or their legal representatives) received written information and provided their consent to participate. The present report was drafted in line with the STROBE statement 32 .
Patients. Patients were enrolled in this prospective monocentric cohort if they fulfilled the criteria for the Lymphonie study (previously presented on the original report 21 ) : severe community-acquired pneumonia www.nature.com/scientificreports/ (CAP): (1) pneumonia (≥ 2 acute signs including cough, purulent sputum, dyspnea, chest pain, temperature < 35 °C or ≥ 38.5 °C, and novel radiological pulmonary infiltrate); (2) at least two criteria of poor prognosis according to the quick-SOFA score (systolic blood pressure ≤ 100 mm Hg, respiratory rate ≥ 22, Glasgow score < 15) and/or the need for mechanical ventilation (MV) and/or vasopressors; and (3) diagnosed within 48 h following admission. Non-inclusion criteria were: age < 18 years, pregnancy, person subject to a measure of legal protection, decision to limit care, known immune deficiency, chronic disorder known to cause deep lymphopenia (Cirrhosis, lympho or myeloproliferative syndrome, solid cancer or active systemic lupus), hospitalization for sepsis in the past 3 months. Non-COVID-19 CAP patients were included until February 20, 2020, one month before the COVID-19 pandemic started in Burgundy, France. COVID-19 patients were eligible if they tested positive for SARS-CoV-2 by reverse transcriptase-polymerase chain reaction on one respiratory sample. Patient with no available plasma sample were excluded from this analysis.

Study procedures.
Ethylenediaminetetraacetic acid blood was obtained after inclusion of the patient and within 48 h of hospital admission. Plasma was collected in the biological resource center (CRB Ferdinand Cabanne; NF S96-900 certification), centrifuged at 2000 ×g for 10 min at 4 °C and stored at − 80 °C.

Variables of interest, clinical outcomes, and data collection. Clinical and biological parameters,
severity scores (Sequential Organ Failure Assessment (SOFA) (9) and the Pneumonia Severity Index (PSI) (11)) were calculated at the time of inclusion. Obesity was defined as body mass index (BMI) higher than 30. ARDS was defined according to the Berlin definition 33 . Septic shock was defined according to the third international consensus definitions for sepsis and septic shock 34 . Clinical outcomes were recorded up to 30 days after the hospital admission: 30-day mortality, number of hospital-, ICU-, and ventilator-free days. The 'ventilator-free days' outcome was defined as the number of days alive from day 1 of severe pneumonia to day 30 during which the patient was breathing without MV. Acute phase response is the systemic response to tissue damage 35 . We used CRP concentrations as a surrogate for APR because it is recognized as the prototype for acute phase reactant 31 .

Measurement of lipid parameters.
Total cholesterol concentration, total triglyceride concentration, high density lipoprotein cholesterol concentration (HDL-c) and low-density cholesterol concentration (LDL-c) were determined by colorimetric enzymatic methods. LDL-c and HDL-c were determined by direct methods. Apolipoprotein A1 (ApoA1) and B (ApoB) were measured by immunonephelemetry. All these parameters were determined on Dimension Vista analyzers using dedicated kits (SIEMENS). Apolipoprotein E (ApoE) was measured by enzyme-linked immunosorbent assay (ELISA) (THERMO FISHER SCIENTIFIC INC., Massachusetts). Serum amyloid A was measured by immunoassays using the pro-Quantum method (THERMO FISHER SCIENTIFIC INC., Massachusetts). Lipid peroxidation was measured by fluorescence using the Tbars method (BIOASSAY SYSTEMS, Hayward, USA).

Lipoprotein size determination.
The different fractions of plasma were separated according to their density by ultracentrifugation (TLA100.2 rotor, BECKMAN, Palo Alto, CA). Densities were adjusted by adding appropriate volumes of potassium bromide solution. Three consecutive ultra-centrifugations were conducted at the appropriate densities in order to separate the plasma into 4 fractions: the triglyceride-rich fraction (d < 1.006), the LDL fraction (1.006 < d < 1.063), the HDL fraction (1.063 < d < 1.21) and the lipoprotein-free fraction (d > 1.21).
Triglycerides, cholesterol (free and total) and phospholipids in lipoproteins fractions isolated by ultracentrifugation were measured enzymatically by colorimetry with commercially available kits (DIASYS DIAGNOS-TIC SYSTEMS GMBH, Holtzheim, Germany) according to manufacturer's instructions. Lipoprotein size in isolated fractions was calculated based on the lipid composition as follows 36 : lipoprotein diameter (Å) = 20 * (radius of surface + radius of core); radius of surface being equal to 2.05 nm; radius of core being equal to 3 * PLTP and CETP activity measurement. Phospholipid transfer protein (PLTP) activity was measured using a commercially available fluorescence activity assay (ROAR BIOMEDICAL INC., New York) according to the manufacturer's instructions. Fluorescence measurements were performed over time on a Vicor 2 multilabel counter (PERKIN ELMER). Phospholipid transfer activity was calculated from the slope of fluorescence increase between 1 and 30 min.
Cholesteryl ester transfer protein (CETP) activity was measured using a commercially available fluorescence activity assay (ROAR BIOMEDICAL INC., New York) according to the manufacturer's instructions. Fluorescence measurements were performed over time on a Vicor 2 multilabel counter. Cholesteryl ester transfer activity was calculated from the slope of fluorescence increase between 1 min and 2 h.

Measurement of glucose metabolism parameters.
Fructosamine and glucose concentrations were measured on total plasma by colorimetry using commercially available kits (CLINISCIENCES, Nanterre, France, for fructosamine and glucose). Insulin concentrations were measured by ELISA (MERCODIA AB, Sweden).  (13 variables) and the following 12 clinical and biological parameters: BMI, PSI, SOFA, Lactate, ICU and hospital length of stay, ventilation duration and free days at day 30, CRP, PCT, creatinine and NT-ProBNP. PCA identifies factors, called principal components, that induce the most variation in the overall data. These factors are uncorrelated and can be expressed as a linear combination of the correlated original variables (OV). We can inverse these formulas to express each OV as a linear combination of factors. Coefficients defining these linear combinations are interpreted as correlation coefficients. A positive (or negative) coefficient means that the OV is positively (or negatively) associated with the factor. An absolute value approaching 1 indicates that the original variable has a strong influence on the value of the factor, while an absolute value below 0.3 is weak. Each factor describes a percent variation of the OVs. In order to determine the number of components to retain, we used the scree plot test: the number of factors to retain corresponds to the break point between factors with large eigenvalues and those with small eigenvalues on the scree plot 37 and the clinical interpretability of these factors 38 . The labeling was primarily descriptive and based on our interpretation of the factors structure for the OVs strongly associated with the factor. In addition, OVs data can be projected and plotted on the plans defined by the retained factors, which reduces the dimensionality of a large amount of potentially interrelated OVs into a smaller set of factors. This makes it possible to observe variations and strong patterns among patients in a new, two-dimensional space without losing important information 39 .
Spearman's rank correlations were computed between clinical characteristics and pertinent lipoprotein metabolism parameters associated with COVID-19 status in univariate analysis comparison tests. Multivariable linear regression was then performed between the selected lipoprotein metabolism parameters as independent variables and the COVID-19 status as a principal explicative covariate. To account for potential confounders, we constructed models adjusted for BMI, CRP and SOFA score. Absence of autocorrelation and heteroscedasticity were assessed using the DW statistic and the White test 40 , respectively. Model variability explicative power was quantified using the R 2 coefficient. Association results were expressed as mean differences ± SE. A p-value below 0.05 was considered statistically significant. Analyses were performed using SAS software (version 9.4, SAS Institute, Inc., Cary, NC, USA).
Ethics approval and consent to participate. This ancillary study of the ongoing prospective Lymphonie study (ClinicalTrials.gov NCT03505281) was registered in ClinicalTrials.gov (NCT04435223). The study fulfilled legal and ethics requirements: Approval was obtained from the CPP (Comité de Protection des Personnes SUD MEDITERRANEE V; 2017-A03404-49) for the original study and an amendment was obtained to include supplementary patients with severe COVID-19. All subjects (or their legal representatives) received written information and provided their informed consent to participate.