Link between serum lipid signature and prognostic factors in COVID-19 patients

Although the serum lipidome is markedly affected by COVID-19, two unresolved issues remain: how the severity of the disease affects the level and the composition of serum lipids and whether

serum lipidome analysis may identify specific lipids impairment linked to the patients' outcome. Sera from 49 COVID-19 patients were analyzed by untargeted lipidomics. Patients were

clustered according to: inflammation (C-reactive protein), hypoxia (Horowitz Index), coagulation state (D-dimer), kidney function (creatinine) and age. COVID-19 patients exhibited remarkable

and distinctive dyslipidemia for each prognostic factor associated with reduced defense against oxidative stress. When patients were clustered by outcome (7 days), a peculiar lipidome

signature was detected with an overall increase of 29 lipid species, including—among others—four ceramide and three sulfatide species, univocally related to this analysis. Considering the

lipids that were affected by all the prognostic factors, we found one sphingomyelin related to inflammation and viral infection of the respiratory tract and two sphingomyelins, that are

independently related to patients' age, and they appear as candidate biomarkers to monitor disease progression and severity. Although preliminary and needing validation, this report pioneers

the translation of lipidome signatures to link the effects of five critical clinical prognostic factors with the patients' outcomes.

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), a positive-sense single-stranded RNA virus, is responsible for COVID-19 disease, which mainly affects the respiratory tract1.

COVID-19 results in a plethora of symptoms that contribute independently to the severity of the infection. Although most patients infected by SARS-CoV-2 experience very mild to moderate

illness, in some cases the infection leads to severe symptoms that eventually require hospitalization with a high mortality rate. The symptoms include respiratory failure, with a marked

decrease in the oxygen partial pressure/inspired oxygen fraction ratio (Horowitz index or P/F), and shock accompanied by multi-organ dysfunction2.

Laboratory findings in the sera obtained from severely ill COVID-19 patients often display high levels of biomarkers, e.g. C-reactive protein (CRP), D-dimer (DD) and excessive cytokines

release (also known as cytokine storm), linked to systemic hyper-inflammation, some of which are commonly found in patients with acute respiratory distress syndrome (ARDS)3. However, despite

some similarities, ARDS and severe COVID-19 represent separate clinical entities in terms of lung compliance and endothelial inflammation4. Nevertheless, several observations converge in

believing that COVID-19 targets multiple organs (lungs, liver, kidney, brain, testis, and intestine), impairing their function and leading to multi-organ injury3. Like many other viruses,

SARS-CoV-2 triggers several pathways linked to both the metabolome and the lipidome. Consequently, it is expected that small-sized metabolites become essential to support virus replication

by providing building blocks to assemble the viral nucleic acids, proteins, and membrane5,6. As for the lipidome signature, it has been shown that SARS-CoV-2 infection mobilizes the host

free fatty acids pool to support viral capsid-associated membrane formation7. Many viruses can profoundly alter the host lipidome and exploit the host energy resources to support their

replication, thereby causing marked changes in the plasma/serum lipidome8,9.

To fill the knowledge gap regarding the composition of the serum lipids set in COVID-19 patients, the main goal of the present study is to investigate the changes occurring within the

disease severity. The aim is two-fold: (1) to unravel how the lipidome is affected by progressive impairment of the most critical prognostic factors employed to assess the severity of

COVID-19 patients; and (2) to test whether the patients' outcome is marked by specific lipidome signatures independently of these prognostic factors.

At first, by employing an untargeted approach based on an up-to-date liquid chromatography tandem high-resolution mass spectrometry (LC-HR-MS) lipid screening technology, we identify the

most frequent lipidome alterations as a function of the severity of the prognostic factors. Then, we used the same opportunity to assess which lipidome signature is primarily occurring in

COVID-19 patients as a function of 7-day mortality. We reasoned that comparing the two lipidome signatures may give a unique insight into the identification of those lipids species that mark

not only the severity, but also the patients' outcome, thereby identifying those who most require aggressive therapy.

Table 1 reports the patients' main data at the time of blood sampling obtained within 48 h after hospitalization. Of these, twelve patients faced fatality within 7 days after

hospitalization. Here, we selected the following markers: (1) serum CRP, (2) the Horowitz Index (P/F), (3) serum creatinine (CR), and (4) serum DD, hallmarks of the prognostic factors

inflammation, hypoxia, kidney function and coagulation state, respectively. The fifth prognostic factor was age.

Untargeted lipidomics analysis on serum of COVID-19 patients identified about 1500 lipid species. The striking serum lipid profile differences that emerged when comparing COVID-19 patients

with healthy age-matched controls (n = 10), suggest that such differences are related to the disease state rather than the patients' age or the preceding lifestyle (Supplementary Figure S1).

For this analysis, COVID-19 patients were grouped in four quartiles according to CRP: CRP1 ( 118.0 mg/L, n = 10). The discriminant analysis (PLS-DA), used to discriminate the lipidomic

profiles among the four groups, showed a separation of 18.5% on principal component (PC1). The PC1 represents the new dimension in which the initial variables are compressed and represents

the maximum of the separation that can be reached within these clusters and variables (Fig. 1A). The VIP scores derived from PLS-DA were used for ranking the discriminating features, taking

a cut-off value > 1.5. From this cluster, it emerged that 240 lipid species, hereinafter called discriminant lipids, marked univocally the differences among the four groups as far as

inflammation was concerned. Univariate analysis was performed to validate this dataset and to test whether the trend in the discriminant lipids was consistent with the CRP level and with the

severity of the inflammation. Lipid classes that have a statistically significant modulation (Fig. 1B) include phospholipids, sphingomyelins (SM), CE, the antioxidant vitamin E (Vit. E) and

the anti-inflammatory bulk of lipids containing polyunsaturated fatty acids (PUFAs).

Discriminant analysis (score plot) of the lipidome as a function of inflammation (A) and hypoxia (C). Lipid classes with a statistical significance are shown in boxplots in function of

increasing inflammation (B) and hypoxia (D). Boxes: 25th–75th percentiles; lines: 10th–90th percentiles; crossing lines: median values; separate points: outliers. Statistical tests were

performed by one-way ANOVA and the Bonferroni post hoc test.

For this analysis, COVID-19 patients were grouped in four quartiles according to P/F as a marker of the respiratory failure severity: very mild (VM, P/F > 400, n = 11), mild (MI, P/F

200–300, n = 9), moderate (MO, P/F 100–200, n = 14), and severe (S, P/F  1.5). Pearson correlation analysis was performed to investigate the linear relationship between lipids and some

continuous clinical variables (i.e. CRP, P/F, DD and CR) and associated p values were subsequently used. p values