Severity of COVID-19 Patients Predicted by Serum Sphingolipids Signature

The reason behind the high inter-individual variability in response to SARS-CoV-2 infection and patient’s outcome is poorly understood. The present study targets the sphingolipid profile of twenty-four healthy controls and fifty-nine COVID-19 patients with different disease severity. Sera were analyzed by untargeted and targeted mass spectrometry and ELISA. Results indicated a progressive increase in dihydrosphingosine, dihydroceramides, ceramides, sphingosine, and a decrease in sphingosine-1-phosphate. These changes are associated with a serine palmitoyltransferase long chain base subunit 1 (SPTLC1) increase in relation to COVID-19 severity. Severe patients showed a decrease in sphingomyelins and a high level of acid sphingomyelinase (aSMase) that influences monosialodihexosyl ganglioside (GM3) C16:0 levels. Critical patients are characterized by high levels of dihydrosphingosine and dihydroceramide but not of glycosphingolipids. In severe and critical patients, unbalanced lipid metabolism induces lipid raft remodeling, leads to cell apoptosis and immunoescape, suggesting active sphingolipid participation in viral infection. Furthermore, results indicated that the sphingolipid and glycosphingolipid metabolic rewiring promoted by aSMase and GM3 is age-dependent but also characteristic of severe and critical patients influencing prognosis and increasing viral load. AUCs calculated from ROC curves indicated ceramides C16:0, C18:0, C24:1, sphingosine and SPTLC1 as putative biomarkers of disease evolution.


Introduction
The 2019 coronavirus disease  is an ongoing pandemic that started in December 2019 in Wuhan (China), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). To date (August 2021), more than 4.4 million people worldwide have died from COVID-19 infection, and mortality risk increases in subjects with cardiovascular diseases, high blood pressure, diabetes, and metabolic syndrome. Furthermore, a high percentage of patients suffer from persistent symptoms and comorbidities associated with organ damage with long-term effects that remain unknown. The reason behind such a high inter-individual variability in response to SARS-CoV-2 infection is poorly understood, as well as the possible patient outcome. The majority of patients experience mild or moderate symptoms with a good prognosis [1]. However, about 20% of patients progress into severe or critical phase, suffering from severe respiratory failure, which requires mechanical ventilation and intensive care unit admission [2,3]. Disease severity is determined based on a set of clinical characteristics such as respiratory rate (≥30 breaths/min), mean oxygen saturation (≤94% in the resting state), or arterial blood oxygen partial pressure/oxygen concentration (≤300 mmHg). It is clear that prognostic markers of severity and possible predictors of patient outcomes are needed.
The virus infection occurs through the binding of viral Spike proteins (S-protein) to human cells utilizing a 2-step process that involves Angiotensin-Converting Enzyme-2 (ACE2) and Transmembrane Serine Protease (TMPRSS)-2. Once bound to ACE, the virus triggers the processing of ACE-2 through the ADAM-17/TNF-alpha-converting enzyme that induces the cleavage of ACE2's extracellular domain and internalization. Next, through the assistance of clathrin [4], viral particles and host cells fuse, and the intracellular structure of ACE2 aids viral transport from the cell membrane to the cytoplasm, developing SARS. Another way of SARS-CoV-2 entry is by a membrane serine protease TMPRSS2 that proteolytically cleaves and activates viral envelop glycoproteins. Recently a selective advantage in lung cells and primary airway epithelial cells has been attributed to TMPRSS2 that enables endosome-independent virus entry through the furin polybasic cleavage site without endosomal acidification, making chloroquine treatment ineffective [4][5][6][7][8].
Giving ACE2 and TMPRSS2 as specific targets of COVID-19, the plasma membrane and the lipid raft in which ACE2 and TMPRSS2 are embedded can actively participate in viral infection [9,10]. The lipid raft is enriched in sphingolipids. The latter is a bioactive class of amphipathic molecules involved in a number of biological functions such as inflammation, immune cell activation, and recognition of exogenous agents. Defects in sphingolipid metabolism are involved in the pathogenesis of pulmonary infections and several respiratory diseases (asthma, cystic fibrosis, chronic obstructive pulmonary disease) [11][12][13]. Recent papers described that viruses induce significant remodeling of lipid composition in host cellular membranes [14][15][16][17][18], and sphingolipid changes associated with viral infections have been reported in plasma or serum [19][20][21]. Furthermore, studies in freshly isolated human nasal epithelial cells of COVID-19 positive volunteers demonstrated that acid sphingomyelinase (aSMase) and ceramide C16:0 promote viral spreading and treatment with aSMase inhibitors such as ambroxol or amitriptyline almost completely inhibit the viral infection [22,23].
The present study aims to unravel, in serum, sphingolipids associated with membrane disturbances that can lead to viral load increase. Furthermore, we would like to get a better insight into the panel of sphingolipids characterizing SARS-CoV-2 affected patients undergoing severe illness with the final goal of identifying putative biomarkers able to predict COVID-19 disease evolution but also contributing to define which species are involved in viral load increase.
In this study, to overcome dynamic range limitations imposed by this class of molecules in serum [24], a combination of untargeted and targeted lipidomics was performed to investigate sphingolipid levels in a cohort of 83 subjects that included 24 healthy controls and 59 COVID-19 patients with different disease severity (i.e., mild, moderate, severe, and critical). The sphingolipid profile was correlated with clinical parameters. Results indicate a positive correlation for some sphingolipid species and no correlation for some others considered as protective. Furthermore, results indicated that the sphingolipid profile might be linked to disease severity, highlighting the relevance of monitoring this class of molecules to predict patient's outcome and proposing specific sphingolipid molecules to be targeted for treating COVID-19 patients.
Hydroxychloroquine was prescribed to 47 pts (80%), 22 of whom were treated with a Protease Inhibitor (Lopinavir/RTV, 9 pts, or Darunavir/COBI, 10 pts), in line with the standard of care at the time of the patient's hospitalization. Ten pts (17%) also received Tocilizumab in combination with one of the aforementioned treatments. Seven pts (12%) did not receive any therapy.
Death was reported in 4 pts (7%), all with moderate to critical disease severity, while the remaining patients were discharged after clinical and/or virological recovery.
Differences in serum sphingolipid levels were assessed for each sphingolipid acylchain towards HC and among COVID severity groups. To facilitate reading, results are described following the synthetic/degradation sphingolipid route ( Figure S1).  Figure 2A). Log2 fold changes were calculated to assess their levels in different groups compared to HC ( Figure 2B). Mild and moderate patients showed higher levels of DhSph that further increased in severe and critical groups. Mild patients are characterized by lower levels of DhCer C16:0 compared to HC. DhCer C18:0 and C22:0 levels followed the same trend, were lower in mild, moderate, and severe patients compared to HC, but not in critical patients. All patients showed higher levels of C22:1 and lower levels of C24:0 and C24:1 compared to HC. Box plots in Figure 2C display differences according to disease severity. DhSph levels increased in severe and critical patients vs. mild and moderate. DhCer C16:0 levels increased progressively in severe and critical groups vs. mild patients. In the critical group, DhCer C18:0 and C22:1 levels increased compared to moderate, whereas DhCer C22:0 increased compared to severe patients, indicating that critical patients are characterized by DhCers increments irrespectively from their chain length.  Box plots and whiskers represent the interquartile range with median value (central line) and the lowest and largest data point measured, respectively. Each measurement was performed in triplicate. Data were analyzed using Kruskal-Wallis ANOVA, followed by Dunn's post hoc test for multiple comparisons. * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.

In the Catabolic Pathway, Sphingosine Increased with Disease Severity Whereas S1P Decreased in Severe and Critical Patients
In the catabolic pathway ( Figure 4A), sphingosine (Sph) showed a progressive statistically significant increment according to disease severity. Furthermore, levels were higher in severe and critical groups vs. mild and in critical compared to moderate groups ( Figure 4B). In contrast, S1P levels decreased in severe and critical COVID-19 patients compared to controls ( Figure 4C), suggesting that in severe and critical patients, Sph is not further metabolized to S1P but contributes to the Cers pool level increase. of S1P serum levels. Each measurement was performed in triplicate. Data were analyzed using Kruskal-Wallis ANOVA, followed by Dunn's post hoc test for multiple comparisons. * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.

Glycosphingolipids and Disease Severity in COVID-19
Glycosphingolipids as hexosylceramides, lactosylceramides, and GM3s are involved in complex sphingolipid pathways ( Figure 6A). In respect to HC, a statistical analysis indicated the lowest levels of HexCers and GM3s in critical and in moderate patients (for HexCers and GM3s). In mild, the decrement was observed for GM3s ( Figure 6B). Mild patients showed higher levels of HexCer C20:0, C22, C24:0, C24:2, and GM3 C16:0, C20:0, C22:1, and C24:1 than the moderate group and lower levels of LacCer C16:0. Moderate patients displayed lower levels of HexCer C16:0, C20:0, C24:0, and GM3 C20, C22, and C24:1 compared to the severe group. Severe patients showed higher levels of HexCer 24:0, LacCer C18:0 compared to the critical group ( Figure 6C). However, driven by a previous study [25] that indicated GM3s increase in relation to COVID-19 in young patients, data were re-analyzed considering HC as two classes, young HC, and aged HC vs. patients. GM3 C16:0, the most abundant GM3 in human serum, increased in aged HC, and the level was higher than in mild, moderate, and critical patients (data not shown) as indicated in panel D.

Acid Sphingomyelinase Levels Increased with COVID-19 Severity
To further relate changes of Cer and SM specific chains in COVID-19 patients, levels of acid sphingomyelinase (aSMase) were assessed. This enzyme catalyzes the breakdown of SM to Cer and phosphorylcholine. Results from the ELISA assay showed a significant increase in acid SMase in severe compared to mild patients. ( Figure 7A). This increment paralleled the increase in Cer and decrease in SM levels in relation to disease severity. An increasing trend was also observed in critical patients. By analyzing HC according to age, young HC vs. aged HC, we observed that levels of aSMase were increased in aged HC, and the enzyme level was comparable to severe patients ( Figure 7B). Of note, three young subjects had aSMase levels comparable to aged subjects.  (A,B), serine palmitoyltransferase subunit 1 (SPTLC1) (C) and, caspase 3 (D) in serum of HC subjects, mild, moderate, severe, and critical patients, detected by ELISA assay. Each measurement was performed in duplicate. Data were analyzed using Kruskal-Wallis ANOVA, followed by Dunn's post hoc test for multiple comparisons. * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001.

Serine Palmitoyltransferase and Caspase 3 Levels Increased in Severe and Critical Patients
To further support our results in relation to DhSph dysregulation and decreased levels of S1P in severe and critical patients, levels of serine palmitoyltransferase, which catalyzes the condensation of serine and palmitoyl CoA in the first step of the de-novo synthesis of sphingolipids were assessed. Serum levels of subunit 1 (SPTLC1) were quantified by the ELISA assay. Enzyme levels increased in severe and critical COVID-19 patients compared to healthy subjects ( Figure 7C). To confirm cell death signaling in relation to an increase in pro-apoptotic species (ceramides and sphingosine) in COVID-19 patients, levels of Caspase 3 were assessed. Results showed an increase of Caspase 3 in severe and critical COVID-19 patients compared to healthy subjects and also in critical compared to moderate patients ( Figure 7D).

Discussion
In our study, sphingolipids were extracted from serum which contains extracellular vesicles, including exosomes and large plasma membrane-derived microvesicles released in the bloodstream; these vesicles transport proteins, lipids, and nucleic acids specific to the cell of origin. This untargeted/targeted sphingolipid analysis aimed to find changes in the sphingolipid composition, relate them with disease severity, and reveal new putative biomarkers to predict disease evolution toward severe conditions. Thus, the released or unreleased species from extracellular vesicles can provide a clue on the status of the plasma membrane and lipid raft according to disease severity and can provide hints or targets to counteract disease worsening and hypothesize new treatments.
Results from the present study refer to 24 HC compared to 59 COVID-19 patients divided into four groups. Considering changes in respect to HC, the increment of DhSph, Sph, and of specific Cer chains and the progressive decrement of SMs, with lowest levels in critical patients, were observed. More specifically, results indicated that DhSph, Sph, Cers C16:0, C18:0, and C20:0 increased in all patients. Furthermore, Cer C18:1, Cer C24:1, increased in moderate and severe/critical patients, which, besides that, also showed a Cer C24:2 increase. Alongside the increase in Sph and Cers, a decrease in sphingosine-1phosphate (S1P) was observed in severe and critical patients suggesting that the "sphingolipid rheostat" determining the cell fate is active [26]. In cells, ceramide and sphingosine are known to mediate apoptosis, cell cycle arrest, and differentiation, whereas S1P promotes proliferation and survival [27]. Our results collectively confirm a shift towards an apoptotic phenotype in severe and critical COVID-19 patients, as indicated by caspase 3 levels ( Figure 3C), with a loss of the protective effect of S1P [28]. Van Echten-Deckert et al. described that the conversion of Sph to S1P can inhibit serine palmitoyltransferase activity [29]. Our results indicated that in severe and critical patients, increased levels of DhSph are related to the activation of serine palmitoyltransferase, confirmed by ELISA. The increase in palmitoyltransferase occurs in severe and critical patients and parallel the decrease in S1P, characteristic of these two groups of patients, leaving them without S1P protective signaling. The inhibition of this enzyme is a potential target to sustain S1P levels in line with the observation of Asao Katsume et al. [30]. Further studies at cellular levels are ongoing to better clarify our findings.
Results from this detailed lipidomic analysis suggest that disease severity is associated with a profound metabolic rewiring of the sphingolipid machinery that involves several species with different chain lengths. Of note, ELISA of aSMase, previously described in cellular studies [22,23], confirmed our results. Furthermore, together with aSMase, the study introduced for the first time the involvement of serine palmitoyltransferase associated with DhSph and S1P dysregulation. The increase in Caspase 3 confirmed the proapoptotic role of Cers species and Sph, promoting the negative evolution of the disease.
Results on Cer C24:0 indicated a decrease in moderate and severe/critical patients. It has been described that the ratio Cer C16:0/C24:0 (but also C18:0/C24:0 and C24:1/C24:0) is a predictor of cardiometabolic risk, whereas Cer C24:0 had negative or no correlation with adverse cardiac effects [31]. Thus, these results from previous studies can suggest a negative impact of the Cer C16:0/C24:0 ratio on the plasma membrane and lipid microdomains that can act as platforms for signal transduction affecting their organization and function [32].
The progressive increase in acid SMase affects Cers and GM3s levels, and the increase/decrease in specific chains appear to be associated with disease severity. aSMase is positively associated with Cer C16:0, Cer C18:0, total GM3, GM3 C16:0, GM3 C20:0, and LacCer (C18:0, C20:0) and negatively associated with Cer 24:0 suggesting that changed levels of these species can be correlated with disease evolution and possibly with an increased viral load [18]. Considering the age of patients, aSMase and GM3s were positively associated with severity and were age-dependent. These species were, in younger patients, not significantly changed in mild compared to aged-matched HC, while a progressive increase was observed in aged and severely affected patients. Furthermore, three young HC showed aSMase enzyme levels comparable to aged HC, suggesting that high levels of this enzyme can predict the subject's frailty. Further quantitative studies have to be performed in a larger cohort to support our hypothesis. GM3 accumulation induces apoptosis through Caspase 3 activation, increases ROS production, and promotes cell senescence [33]. In melanoma cancer cells, GM3 increase is induced by aSMase, which is responsible for GM3 mediated apoptosis and an increase in gangliosides promotes immunoescape in these tumor cells [34]. In our patients, the increment of GM3 species directly correlates with increased levels of aSMase observed in severe patients. These results indicate that ganglioside levels and aSMase levels may trigger the negative evolution of disease prognosis, particularly in the elderly [35].
Results are summarized in Figure 10. Few studies assessed the sphingolipid plasma/serum profile in viral diseases. It has been investigated mainly in HCV and HIV infections. Results indicated that in the serum of acute and chronic HCV, a downregulation of Cer C24:0 and a simultaneous increase in the serum levels of Sph and DhSph were not revealed in HBV patients [15,19]. However, in HBV, the dysregulation of sphingolipids in serum can predict disease evolution toward acute-on chronic liver failure [21], indicating that levels of DhCer C24:0 can anticipate the negative patient's outcome. In HIV patients, elevated plasma levels of C16:0 and C24:1 ceramides correlated with immune activation and inflammation and were associated with antiretroviral therapy and progression of carotid artery atherosclerosis [20]. In line with our results, changes in specific chains abundance appear to be a common trait of viral infections and of their outcome. The risk of death from COVID-19 is about 10 times higher in countries where most of the population is overweight (COVID-19 and Obesity: The 2021 Atlas) and previous studies from our group highlighted the role of acyl chain composition in bodily fluids in obesity and obesity-related comorbidities [36][37][38]. In obese subjects, high levels of total Cers, lower levels of total SM, and of DhSM compared to normolipidemic normal weight controls were found. Interestingly the same pattern also characterizes dyslipidemic normal weight subjects. Furthermore, a detailed study by A. M. Poss, based on machine learning and sphingolipid profile in sera, identified sphingosine, DhCer C16:0, DhCer C18:1, and Cer C18:0 as dysregulated sphingolipids chains that can contribute to predicting the evolution of CVD [39]. Based on our results, these profiles could also represent a predictive signature for a negative evolution of COVID disease, confirming the susceptibility of obese and CVD patients toward a bad prognosis in COVID infection. Another point to be considered is age, in which accumulation of specific chains of ceramide and decrease in SM have been described in plasma, and results on aged and AD subjects indicated that dysregulation of sphingolipid metabolism is associated with frailty [40,41]. Aged subjects with a background of dysregulated sphingolipid metabolism are more prone to develop a severe phenotype, as CVD patients and obese. In aged HC, our results, although in a restricted cohort, indicated higher levels of aSMase and GM3s that may represent a risk factor for developing a severe illness. Data from SLs serum profiling suggested that viral infections, aging, and obesity are characterized by lipid raft remodeling that impacts membrane structural properties and cell signaling. An increase in Cer and DhCer levels and decrease in Cer C24:0 can cause stiffness and change in cell membrane curvature, making cells more permeable to small and large size molecules [42]. In addition, Cers can promote transmembrane lipid motion [43]. Our patients are characterized by increased acid sphingomyelinase levels that, besides promoting an increase in GM3 mediated apoptosis, also promote the synthesis of long acyl chains ceramides at the expense of SM. This would generate unbalance on the structure of the membrane with an accumulation of cholesterol in the inner leaflet, as suggested by studies by Slotte and Bierman [44]. It has also been demonstrated that imbalance in sphingolipids composition in the surface area between two leaflets leads to spontaneous endocytosis [45], which could contribute to the negative outcome observed in severe and critical patients in which unbalance of ceramide long chains and SM occurs. Furthermore, progressive increase in LacCer, Cer(d18:1/18:1), DhCer (d18:0/16:0), Sph, DhSph, GM3 (d18:1/16:0), GM3 (d18:1/20:0), GM3 (d18:1/22:0), GM3 (d18:1/24:1) in severe patients compared to mild, may promote structural remodeling, apoptosis, and immunoescape. Collectively these results indicate an active participation of sphingolipids in viral infection. Treatment with aSMase inhibitors appears to decrease viral proliferation [46]. However, further studies of the sphingolipid profile after treatment with aSMase inhibitors would be required to precisely reveal which changes are induced in the serum sphingolipid profile and if treatments at the cell level can also promote a structural membrane lipid raft remodeling inhibiting virus entry in frail aged subjects.
Concerning the diagnostic potentiality of identified molecules, the present study suggests possible candidates as putative biomarkers of disease evolution based on ROC/AUC curves. Ceramides C16:0, C18:0, C24:1, sphingosine, and SPTLC1 can be considered as putative markers with ROC/AUC outstanding values comparing severe and critical patient vs. HC and excellent values in the comparison of mild to severe vs. critical patients. aSMase and GM3s do not, unfortunately, have such excellent or outstanding discriminatory power being age-dependent. A validation study on a larger independent cohort is needed, and it is currently ongoing.
We believe that sphingolipid lipid composition has to be considered as a structural modifier that contributes to mechano-translate signals from lipid raft to cell membranes of endoplasmic reticulum, Golgi, lysosomes, and mitochondria, activating signaling processes. The identification of structural changes will represent a step forward not only in COVID 19 but also in other infectious or noninfectious disorders in which lipid transport is involved.
In the past few months, other groups investigated the COVID-19 sphingolipid serum profile; however, the results were not always uniform. Our data on Sph and Cers levels are in agreement with studies based on virus inactivation by chloroform/methanol [19,21,25]. In other studies adopting viral heat inactivation results were at variance [49], emphasizing the importance of an extensive validation with standardized procedures for sample collection, sample pre-treatment, and lipid extraction to define a robust protocol that can cope with the interindividual variability avoiding confounder factors.
The major limitation of this study is the restricted number of subjects and the absence of a verification study through targeted lipidomics, which will require an independent group of subjects from a larger longitudinal cohort. Considering correlation results, in which a greater sample size is usually required, caution should be used. Finally, further mechanistic validations are needed and are currently ongoing to confirm conformational changes in plasma membranes.
The present study conforms to the principles of the Helsinki Declaration and was approved by the Ethics Committee of Policlinic "Riuniti" of Foggia (protocol number 49/C.E./2021). All enrolled subjects provided their full informed consent.
General characteristics and laboratory findings of enrolled subjects are summarized in Table 1 and in Table S1.

Lipid Extraction
Sera were managed in a Class II biosafety cabinet, and sphingolipids were extracted according to a previous study, with minor modification [38]. Briefly, 0.1 mL of serum was mixed with 0.1 mL of ultrapure water and 1.5 mL of methanol/chloroform 2:1, and fortified with internal standards 200 pmol: sphingosine (d17:1), sphingosine-1-phosphate (d17:1), ceramide (d18:1/12:0), sphingomyelin (d18:1/12:0), and glucosyl (β)ceramide (d18:1/12:0). Samples were briefly sonicated and heated at 48 • C overnight. Then, 0.15 mL of potassium hydroxide (KOH) 1 M in methanol was added to every sample, and after 2-h incubation at 37 • C, the solution was neutralized with 0.15 mL of acetic acid 1 M and dried with Speedvac. Samples were then resuspended in 0.5 mL of methanol and transferred to a clean Eppendorf tube. Samples were dried, resuspended in 0.15 mL of methanol, and centrifuged for 3 min at 10,000× g. Liquid phases were collected in UPLC glass vials and stored at −80 • C.

Untargeted Lipidomics
Liquid chromatography-mass spectrometer configuration included Waters Acquity UPLC system linked to a Q-TOF Synapt G2-Si (Waters, Milford, MA, USA). Ten µL of sphingolipid extract were injected and separated on a C8 Acquity UPLC BEH (Waters), 100 mm × 2.1 mm id, 1.7 µm, kept at 30 • C, using the following linear gradient: 0.0 min: 80% B; 3 min: 90% B; 6 min: 90% B; 15 min: 99% B; 18 min: 99% B; 20 min: 80% B, at 0.3 mL/min flow rate. Phase B consisted of 1 mM ammonium formate in methanol, 0.05 mM formic acid, while phase A was 2 mM ammonium formate in H 2 O, with 0.05 mM formic acid. The ESI ionization source was operated in positive ion mode with 30 V of sample cone voltage and 3.0 kV of capillary voltage. The desolvation temperature was set to 150 • C, and the ion source was 120 • C. The desolvation gas flow was set to 600 L/h. Data were captured under centroid mode, and the range of mass scan was set to 50-1500 Da. Accuracy and reproducibility were maintained employing an independent reference spray via LockSpray. Sphingolipids' quantification was carried out using the ion chromatogram obtained for each compound using 50 mDa windows. The linear dynamic range was determined by the injection of standard mixtures. Positive identification of compounds was based on the accurate mass measurement, with an error <5 ppm and its retention time, compared to that of a standard (±2%). Mass spectra were analyzed by MassLynx™ 4.2 Software (Waters, Milford, MA, USA), and lipids were annotated as lipid subclasses as follows (sphingosine backbone/number of carbon atoms of the fatty acid: number of unsaturation of the fatty acid). MS/MS spectra were acquired, and the assignment of species was based on precursor ions and product ions m/z 264.268 and m/z 266.286, which correspond to sphingosine backbone (d18:1) and dihydrosphingosine backbone (d18:0), respectively.

Targeted Lipidomics
Sphingosine, dihydrosphingosine, S1P, and dihydroS1P were quantified using a Xevo TQ-S micro mass spectrometer (Waters, Milford, MA, USA). Sphingolipid extracts were injected and separated on a C8 Acquity UPLC BEH 100 mm × 2.1 mm id, 1.7 µm (Waters) kept at 30 • C, with the following gradient: 0.0 min: 80% B; 3 min: 90% B; 6 min: 90% B; 9 min: 99% B; 12 min: 99% B; 14 min: 80% B, at 0.3 mL/min flow rate. Phase A and phase B were the same as for untargeted lipidomics. An electrospray interface operating in positive ion mode was employed to obtain MS/MS spectra by acquiring MRM transitions of:  4, cone voltage 20 V, collision energy 16 eV. The capillary voltage was set at 3.5 kV. The source temperature was set to 150 • C. The desolvation gas flow was set to 1000, and the desolvation temperature was set to 350 • C. Data were acquired by MassLynx™ 4.2 software and quantified by TargetLynx software.

Enzyme-Linked Immunosorbent Assays
Concentrations of acid sphingomyelinase, serine palmitoyltransferase (subunit 1), and caspase 3 in serum of patients were measured using Human Acid sphingomyelinase ELISA Kit (SMPD1) (Abcam, ab277075), SPTLC1 ELISA kit (AssayGenie, Dublin, Ireland), and Human Caspase 3 Instant ELISA kit (Thermo Fisher Scientific, Waltham, MA, USA), respectively, according to the manufacturer's instructions. The absorbance was measured on a microplate reader (BioRad) using the wavelength of 450 nm.

Statistical Analysis
Spearman correlations between clinical data and sphingolipid levels were performed with data from COVID-19 patients. Samples with missing clinical data were excluded from the calculations. Correlations between age and sphingolipid levels were performed for all the subjects. Correlation plots for all variables grouped by sphingolipid class were presented. Only correlations with r > 0.3 and p-value < 0.05 were indicated and further discussed. ROC curve analyses were performed by Metaboanalyst 5.0 to determine the diagnostic value of serum sphingolipids and enzymes in COVID-19 patients, considering the area under the ROC curve (AUC) with 95% confidence interval (CI), sensitivity, specificity, and cut-off value.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available in Supplementary Tables. Study data from this human study other than those published in this work are under privacy regulations but can be obtained on a case-to-case basis upon reasonable request from the corresponding author.