High Fat Diet Mediated Alterations in Serum Sphingolipid Profiles in An Experimental Mouse Model Measured by Matrix-Assisted Laser Desorption/Ionization-Time of Flight Mass Spectrometry

Abstract

Non-alcoholic fatty liver disease (NAFLD) is associated with hepatic steatosis, a benign condition caused by accumulation of lipids in hepatocytes, which may progress to steatohepatitis and cirrhosis. Recent studies suggest that sphingolipids are involved in the development and severity of NAFLD. The goal of this study is to identify the circulating sphingolipid species that are altered by chronic high fat diet (HFD) feeding and correlate these abnormalities with hepatic sphingolipids. We utilized a previously established experimental model of NAFLD generated by HFD feeding of 8-week-old male mice for 16 weeks. Lipids were extracted from serum samples by Folch method and analyzed with matrix assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) in the positive and negative ion modes. MALDI-TOF detected a total of 47 serum sphingolipids including sphingomyelins, sulfatides, ceramides, phosphosphingolipids, and glycosphingolipids within the mass range of 600–2000 Da. Principle component analysis demonstrated clear separation of hepatic sphingolipids from low fat diet (LFD) and HFD groups and partial overlap of serum sphingolipids with a variance of 53.5% and 15.1%, and 11.7% in PC1, PC2, and PC3, respectively. Chronic HFD feeding significantly increased expression of SM (40:0), SM(42:2), ST(42:2), Hex(6)-Cer (40:1), and Hex(4)-HexNAc (2)-Cer (34:1) in both serum and liver. In addition, HFD mediated percent changes in hepatic sphingolipids correlate linearly with the percent changes in serum sphingolipids as determined by Pearson correlation (P = 0.0002). Elevated levels of serum and hepatic sphingomyelins and glycoceramides are key factors mediating NAFLD development and may serve as peripheral markers of hepatic steatosis.

I. Introduction

Non-alcoholic fatty liver disease (NAFLD) has become the predominant cause of chronic liver disease worldwide with an estimated global prevalence rate of about 25% (Younossi, 2018) and a major cause of liver-related morbidity and mortality (Younossi, 2018; Mantovani et al., 2020). Non-alcoholic fatty liver (NAFL) can progress to steatohepatitis, fibrosis, and cirrhosis, which can result in liver failure and hepatocellular carcinoma (Polyzos et al., 2019; Stefan et al., 2019) Moreover, NAFLD is a risk factor for developing insulin resistance, dyslipidemia, and type 2 diabetes (Polyzos et al., 2019; Younossi, 2019). The co-occurrence of NAFLD and diabetes increase not only the progression of the disease to more advanced stages but also the risk for cardiovascular disease (Mantovani et al., 2020; Matsuzaka & Shimano, 2020). Therefore, NAFLD carries a serious economic burden and results in poor health-related quality of life.

Although a large body of clinical evidence suggests that NAFLD is associated with insulin resistance on hepatic and whole-body level, only a subset of individuals with NAFLD will develop insulin resistance, steatohepatitis, or other liver diseases (Sim et al., 2015; Brouwers et al., 2017). Accordingly, studies have focused on understanding the pathogenic factors that contribute to the progression of NAFLD. It has been shown that hepatic de novo lipogenesis is an important contributor of NAFLD and is associated with metabolic syndrome (Utzschneider & Kahn, 2006). The end product of de novo lipogenesis is saturated fatty acids, which are stored in triglycerides under healthy conditions. However, insulin resistance and endoplasmic reticulum stress increase de novo lipogenesis resulting in elevated levels of long chain saturated fatty acids in NASH (Lambert et al., 2014; Gluchowski et al., 2019). In addition, lipolysis has adverse effects on adipose tissue’s ability to store fat resulting in accumulation of fatty acids in the blood (Lambert et al., 2014; Cusi, 2009). Free fatty acids trigger formation of reactive oxygen species and lipid peroxidation thus exacerbating hepatic toxicity (Lambert et al., 2014; Gluchowski et al., 2019).

More recent studies provide evidence that besides fatty acids and triglycerides, other lipids such as sphingolipids are also involved in the development and progression of NAFLD (Leamy et al., 2013; Apostolopoulou et al., 2018). Increased levels of saturated fatty acids might have a direct effect on de novo sphingolipid synthesis because the long chain fatty acid palmitoyl is used as a substrate in the biosynthesis of sphingosine (Eisinger et al., 2014). Liver is the major organ for sphingolipid production, and it contains larger quantities of ceramides and sphingomyelins relative to other tissues (Sanyal & Pacana, 2015; Gault et al., 2010). Thus, liver is particularly susceptible to sphingolipid mediated toxicity. For example, high fat diet feeding caused NAFLD with hepatic insulin resistance and increased hepatic and serum ceramide expression (Kotronen et al., 2010; Régnier et al., 2019). In addition, sphingomyelins may also play a critical role in the transition between steatosis and NASH (Leamy et al., 2013; Longato et al., 2012). Increased concentrations of serum sphingomyelins and ceramides have been reported in obese children with NAFLD (Chocian et al., 2010; Gorden et al., 2015). Furthermore, sphingomyelinase, which synthesizes ceramide from sphingomyelin, might be involved in NAFLD. Sphingomyelinase activity increased by high fat diet induced NAFLD, while pharmacological inhibition of sphingomyelinase activity protected rodents from high fat diet induced hepatic steatosis, fibrosis, and liver damage (Kotronen et al., 2010; Wasilewska et al., 2018). These findings stress the importance for a more detailed characterization of sphingolipid composition in order to understand the mechanistic factors underlying hepatic insulin resistance and progression of NAFLD.

Unfortunately, only a few studies have determined hepatic lipid composition in humans due to the invasiveness of liver biopsy procedures. Although accumulating evidence implicates the importance of hepatic sphingolipids in the development and progression of NAFLD it is not well known whether circulating sphingolipids could serve as relevant biomarkers for NAFLD (Leamy et al., 2013; Gault et al., 2010; Longato et al., 2012; Hanamatsu et al., 2014). In this study, we utilize matrix assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry that enables detection and relative quantification of sphingolipids in serum samples in an adult experimental model of chronic high fat diet feeding. Our aim is to characterize serum sphingolipid profiles and test the hypothesis that high fat diet is associated with increased rates of circulating sphingolipids. In addition, we correlate serum sphingolipid alterations with those detected in the livers of paired samples to find out whether circulating sphingolipids can predict the development of NAFLD.

II. Materials and Methods

A. Materials

HPLC grade methanol and chloroform, 2,5-dihydroxybenzoic acid (DHB), and α-Cyano-4-hydroxycinnamic acid (HCCA) were purchased from Sigma Aldrich (St. Louis, MO). Acetonitrile and trifluoroacetic acid were purchased from Fisher Scientific (Waltham, MA). Peptide calibration standard mix was purchased from Bruker Daltonics (Billerica, MA). Male C57BL/6 mice were purchased from Jackson Laboratories. High fat diet (F3282) was purchased from BioServ (Frenchtown, NJ) and regular chow diet (P.2018.15) was purchased from Envigo (South Easton, MA).

B. Experimental Model

Male 8-week-old C57BL/6 mice were pair-fed with high-fat diet (HFD) or low-fat diet (LFD) for 16 weeks. The HFD provided 60% of the kcal in fat lard, 35.7% in carbohydrate, and 20.5% in protein, whereas LFD provided 18% of the kcal in fat, 58% in carbohydrate, and 24% in protein. The composition of fatty acids is listed in Supplementary Table IV. Mice were kept under standard humane conditions with food and water ad libitum and regular 12-hour light-dark cycle. Food intake was monitored daily and body weight was measured weekly. At the end of the diet feeding experiment, mice were sacrificed by inhalation of an overdose of isoflurane or cervical dislocation to collect their blood and internal organs. Procedures were approved by the Institutional Animal Care and Use Committee at Lifespan-Rhode Island Hospital and followed the guidelines established by the National Institutes of Health. Blood samples were collected in Eppendorf tubes and incubated at room temperature for 20 min to allow the blood to clot, and then centrifuged at 3,000 x g for 30 min to obtain the serum. Serum samples were stored at −80 °C immediately after collection.

C. Lipid Extraction

Lipids were extracted from serum samples by the Folch method (Fucho et al., 2014). Briefly, 50 µl serum was mixed with 250 µl of chloroform: methanol (2:1, v/v) by vortexing vigorously for 30 seconds and incubating at 4 °C with an inverter mixer for 30 minutes to allow for extraction. The polar and apolar phases were separated by the addition of 250 μL of water. The mixture was vortexed for 30 seconds and centrifuged at 3,000 × rpm for 5 minutes. The lower phase containing the lipids was transferred to a glass vial, dried under the fume hood overnight, and stored at −80 °C.

D. MALDI-TOF MS Data Acquisition

The lipid pellets were dissolved in 100 µl HPLC grade methanol before mass spectrometry analysis. 5 µl of each lipid extract was mixed with the same amount of 2,5-dihydroxybenzoic acid (DHB, 75 mg/ml) dissolved in HPLC grade methanol. 1 µl aliquots of this mixture were spotted onto MALDI target plate (MTP AnchorChip 384 BC) in triplicate and air-dried. Mass spectra were acquired in the positive and negative ion modes using a MALDI-TOF/TOF (Ultraflextreme) mass spectrometer (Bruker Daltonics, Billerica, MA) equipped with the 1 kHz smartbeam-II Nd:YAG laser. Spectra were collected between 600–2000 Da mass range in the reflectron mode acquiring 1500 shots per well. For external calibration, we used peptide calibration standard II (Bruker Daltonics, Billerica, MA) and α-Cyano-4-hydroxycinnamic acid (HCCA) as matrix and obtained five or more calibration points over the mass range between 190 and 2465 Da in a cubic enhanced mode.

E. Data Analysis

Three mass spectra of each sample acquired between 600 and 2000 Da were imported into ClinProTools v3.0 (Bruker Daltonics, Billerica, MA) software for post-processing including baseline subtracting, peak defining, recalibrating, and calculating peak intensity. Lipids were tentatively assigned by comparing the precursor ion mass-to-charge (m/z) values with those previously identified in our lab, other publications, or Lipid Maps database (https://www.lipidmaps.org/resources/tools/bulk_structure_searches.php?database=LMSD) (Choi & Snider, 2015). Data bar plots were utilized to visualize the mean percent change in lipid ion intensities of HFD and LFD exposed samples. Statistical comparisons of mean intensity values were made by t-test analysis with a 5% false discovery (GraphPad Prism 8, San Diego, CA, USA). Principal Component Analysis (PCA) performed in ClustVis was used to reduce the dimensionality of high throughput data and compare sphingolipid ion expression patterns between HFD and LFD samples (Metsalu & Vilo, 2015). HFD mediated alterations of sphingolipids were further analyzed by Chi-square analysis with Yates’ correction (GraphPad Prism 8, San Diego, CA, USA).

III. Results

A. Serum Sphingolipid Profiles

The peak statistics report obtained from ClinProTools software identified 37 sphingolipid ions in the positive ion mode and 10 ions in the negative ion mode between the mass-to-charge ratio of 600 and 2000 Da. For putative peak assignments of the ions, we utilized previous MS/MS identifications performed in our laboratory, recent literature or Lipid Maps database MS search (http://www.lipidmaps.org/tools/ms/) (Dreisewerd et al., 2007; Hidaka et al., 2007; Summer, 2007; Meriaux et al., 2010; Colsch et al., 2011; Shanta et al., 2011; Gode & Volmer, 2013; Jackson et al., 2014; Soltwisch et al., 2015; Carter et al., 2016; Rao et al., 2016; Afsihnia et al., 2018). A list of sphingolipids annotated in this study is available in Supplementary Table III. The sphingolipids formed protonated ([M+H]⁺), sodiated ([M+Na]⁺), potassiated ([M+K]⁺), or deprotonated ([M-H]⁻) adducts. Sphingolipid subclasses include 1) ceramides (n=2; 4.3%), 2) sphingomyelins (n=21; 44.7%), 3) sulfatides (n=3; 6.4%), 4) glycosphingolipids (n=19; 40.4%), and 5) phosphosphingolipids (n=2; 4.3%) (Supplementary Table I).

B. Differential Sphingolipid Expression in Serum of HFD- and LFD-Exposed Mice

For optimal experimental conditions to compare serum sphingolipid ion expression of HFD and LFD samples, lipids were extracted from 6 serum samples per group under identical conditions, mixed with DHB, spotted onto MALDI steel target plate in triplicates, and analyzed in the negative and positive ion modes for simultaneous data acquisition. The composition of serum sphingolipids was mostly similar in HFD and LFD groups with 29 ions expressed in both groups. However, 14 sphingolipids were detected only in the serum of HFD fed mice and 4 sphingolipids were only detected in the serum of LFD fed mice (Supplementary Table II). HFD resulted in differential expression of 1 ceramide, 1 ceramide phosphoinositol, 2 sphingomyelins, 1 sulfatide, and 9 glycosphingolipids that were not present in LFD serum. LFD resulted in differential expression of 1 ceramide, 1 lactosyl ceramide, and 2 sphingomyelins that were not present in HFD serum. The Chi-square test and Yates’ correction determined differential expression of sphingolipids (X2 = 12.68, 1 df; P = 0.0004) was statistically significant. These findings indicate altered serum sphingolipid composition due to chronic HFD feeding in mice.

C. HFD Effects on Serum Sphingolipid Profiles Determined by Principle Component Analysis (PCA)

PCA was used to reduce the dimensionality of high throughput MALDI MS data set by visualizing two of the components on the scatterplot (Fucho et al., 2014). PCA contains all serum sphingolipids including ceramides, sphingomyelins, sulfatides, glycosphingolipids, and phosphosphingolipids detected in the positive and negative ion modes within the mass range of 600 and 2000 Da (Fig. 1). Two dimensional PCA plots demonstrated separation of HFD and LFD serum samples with a variance of 38.2% and 21.3%, and 17.8% in PC1, PC2, and PC3, respectively. The first two components (PC1XPC2) (Fig. 1A) showed the most distinct two-dimensional projection followed by PC1xPC3 and PC2xPC3 (Fig. 1B and 1C) indicating differential effects of HFD on serum lipid ion profiles.

Fig. 1.

Principal Component Analysis (PCA) of serum sphingolipids detected by MALDI-TOF in the positive and negative ionization modes. Sphingolipids detected between 600 and 2000 Da mass range were compared between HFD (red) and LFD (blue) exposed mouse serum samples. Two dimensional PCA plots visualizing (A) PC1xPC2, (B) PC2xPC3, and (C) PC1xPC3 were generated in ClustVis software. X and Y axes show PC1, PC2, and PC3 that correspond to 38.2% and 21.3%, and 17.8% of total variance, respectively. HFD: high fat diet, LFD: low fat diet, PC: principal component.

D. HFD Effects on Serum Sphingolipid Expression

Comparative sphingolipid analysis of serum samples from HFD- and LFD-fed mice (n=6/group) revealed relative effects of HFD for sphingolipids detected between 600–2000 Da mass range. The mean peak intensity (or lipid abundance) of 47 sphingolipids was compared by t-test analysis with a 5% false discovery rate correction.

HFD increased expression of 17 (36.2%) sphingolipid ions, reduced expression of 1 (2.1%) ion, and had no effect on 29 (61.7%) ions, relative to LFD controls. (Supplementary Table I). Sphingomyelins (SMs) were mainly detected in the positive ion mode and ionized as [M+H]⁺, [M+Na]⁺, [M+K]⁺, or [M-H]⁻ adducts. HFD increased expression of 10 SM (47.6%), reduced 10 SMs (47.6%), and had no effect on 1 SM (4.8%). Among 3 sulfatides (STs) detected as deprotonated adducts, 2 STs were increased by HFD and 1 ST remained unchanged. 2 ceramide (Cer) and 2 ceramide phosphoinositol (PI-Cer) species were detected in the serum samples as protonated or deprotonated adducts. HFD increased expression of 1 PI-Cer, while the expression other PI-Cer and Cers did not change. Among 19 glycosphingolipids detected in the protonated form, expression of 4 (21.1%) glycosphingolipids was increased and 15 (78.9%) had not changed by HFD relative to control serum samples.

HFD mediated changes were statistically significant for 6 sphingolipids and trend effects (0.05 < P < 0.1) were observed for 6 sphingolipids, whereas the remaining 35 sphingolipids failed to reach significance due to low levels of difference in mean peak intensity of serum lipids between HFD and LFD groups (Table I). HFD significantly increased 3 sphingomyelin species including SM(40:0) + Na⁺ (P =0.005), SM(42:2) + Na⁺ (P = 0.02), and SM(42:2) + H⁺ (P = 0.05), 1 sulfatide ST(42:2) – H⁻ (P = 0.005) and 2 glycosphingolipids including Hex(6)-Cer(40:1) + H⁺ (P = 0.001) and Hex(4)-HexNAc(2)-Cer(34:1) + H⁺) (P = 0.003) relative to control serum samples. HFD had trend increase effect (0.05 < P < 0.1) on the expression of SM(38:0) + Na⁺, SM(40:0) + H⁺, HexCer(d32:1) + H⁺, Hex(3)-HexNAc(2)-Cer 44:2;O₂ + H⁺, Hex(3)-HexNAc-NeuGc-Cer 38:1;O₂ + H⁺, and PI-Cer(44:0) – H⁻ ions in serum samples. HFD had no significant effect on the remaining 35 sphingolipids tentatively identified as detected as SM(26:1) + H⁺, SM(32:2) + Na⁺, SM(34:1) + H⁺, SM(35:0) + H⁺, SM(38:1) + H⁺, SM(38:4) + H⁺, SM(38:1) + Na⁺, SM(40:2) + H⁺, SM(40:1) + H⁺, SM(40:2) + Na⁺, SM(40:1) + Na⁺, SM(42:1) + H⁺, SM(43:1) + H⁺, SM(42:3) + Na⁺, SM(42:1) + K⁺, SM(42:1) – H⁻, ST(42:3) – H⁻, ST(42:1) – H⁻, Cer(42:1) + H⁺, Cer(42:0(2OH)) – H⁻, PI-Cer(t46:0) – H⁻, LacCer(d38:3) – H⁻, LacCer(d38:2) – H⁻, GD1(36:1) – 2H²⁻, Hex(6)-Cer 38:1 + H⁺, Hex(3)-HexNAc-NeuAc-Cer 38:1 + H⁺, Hex(3)-HexNAc(2)-Fuc-Cer 34:1 + H⁺, Hex(3)-HexNAc-KDN-Cer 42:1 + H⁺, Hex(4)-HexNAc-Fuc-Cer 38:1 + H⁺, Hex(3)-HexNAc-KDN-Cer 44:2 + H⁺, Hex(3)-HexNAc-NeuGc-Cer 40:1 + H⁺, Hex(4)-HexNAc(2)-Cer 36:1 + H⁺, Hex(6)-Cer 42:1 + H⁺, Hex(3)-HexNAc-Fuc-KDN-Cer 34:1 + H⁺, and Hex(3)-HexNAc-NeuGc-Cer 42:2 + H⁺.

TABLE I:

EFFECTS OF HFD ON SERUM SPHINGOLIPID EXPRESSION IN HFD- AND LFD-FED MOUSE AS DETECTED BY MALDI-TOF IN THE POSITIVE AND NEGATIVE IONIZATION MODES

Sphingolipid Ions	Ion Intensity (mean ± SD)		P value
	LFD	HFD
SM(40:0) + Na+	12.7 ± 3.7	31.2 ± 10.2	0.005
SM(42:2) + Na+	6.4 ± 1.2	10.3 ± 2.4	0.02
SM(42:2) + H+	6.1 ± 1.1	10.9 ± 2.9	0.05
ST(42:2) - H−	10.1 ± 2.5	23.9 ± 6.2	0.005
Hex(6)-Cer(40:1) + H+	7.3 ± 1.6	15.4 ± 2.9	0.001
Hex(4)-HexNAc(2)-Cer(34:1) + H+	8.1 ± 1.9	16.0 ± 3.3	0.003

The mean peak intensity (or lipid ion abundance) of serum sphingolipids obtained from HFD and LFD exposed mice were compared by t-test analysis with a 5% false discovery rate correction. HFD: high fat diet, LFD: low fat diet, SM: sphingomyelin, Cer: ceramide, Hex: hexose, HexNAc: N-acetylhexosamine, SD: standard deviation.

E. Comparison of HFD Mediated Alterations in Serum versus Hepatic Sphingolipid Profiles

Comparative analysis of serum and liver samples from HFD and LFD fed mice revealed a similarity between the effect of HFD on serum and hepatic sphingolipids. Hepatic lipid analysis was obtained from our previous study that focused on HFD mediated hepatic sphingolipid alterations including 6 paired samples per group (Dreisewerd et al., 2007). Data presentation was simplified by including only the lipids detected in both LFD and HFD samples. The percent change differences in mean lipid ion abundance were graphed in data bar plots to visualize relative effects of HFD on serum and hepatic sphingolipid ion expression. Data bar plots were separated by sphingolipid subclass and sphingolipids were detailed in the ascending order based on the total number of carbon atoms in the structure.

HFD-associated increases in sphingolipid ion expression were represented by the red bars to the right, whereas HFD-associated reductions in sphingolipid expression were indicated by blue bars to the left (Fig. 2A). Side by side comparison of the data bar plots for serum and liver sphingolipid changes revealed that the directional effects of HFD on relative sphingolipid intensity were quite similar. Although SM (32:2) + Na⁺ and SM (42:1)-H⁻ were not present in the liver samples, the remaining sphingomyelins and sulfatides were expressed in both serum and liver. The abundance of 11 sphingomyelins and 2 sulfatides was increased in both serum and liver samples of HFD-fed mice whereas only 3 sphingomyelins were reduced in HFD-fed mice livers and sera relative to controls. T-test analysis revealed that the alterations in 6 sphingolipids expressed in serum and 14 sphingolipids expressed in the liver samples were statistically significant or had trend effects (Fig. 2A). PCA plot demonstrated clear separation of LFD and HFD groups in the liver and partial overlap in the serum samples with a variance of 53.5% and 15.1%, and 11.7% in PC1, PC2, and PC3, respectively (Fig. 2B). Correlation analysis revealed that HFD mediated percent changes in hepatic sphingolipids correlate linearly with the percent changes in serum sphingolipids as determined by Pearson correlation coefficient (r = 0.7872, P = 0.0002) (Fig. 2C).

Fig. 2.

Comparison of HFD mediated alterations in serum versus hepatic sphingolipids. (A) Data bar plots demonstrate effects of HFD on serum and hepatic sphingolipid ion expression as detected by MALDI-TOF in the positive and negative ion modes. Plots depict percent differences in mean peak intensity levels of sphingomyelins and sulfatides observed in the liver and serum samples of HFD and LFD exposed mice. The scale bars shown under the plots display the range of HFD responses from −300 to 300 relative to controls. Sphingolipids are organized with respect to increasing number of carbon atoms in the structure. Red bars to the right represent HFD-associated increases in sphingolipid expression, whereas blue bars to the left represent HFD-associated reductions in sphingolipid expression. (B) PCA of serum and hepatic sphingolipids detected by MALDI-TOF in the positive and negative ion modes. Sphingomyelins and sulfatides detected by MALDI-TOF between the mass range of 600 and 2000 Da were compared between HFD-serum (blue), LFD-serum (purple), HFD-liver (red), and LFD-liver (green) samples. Two dimensional PCA plot visualizing PC1xPC2 was generated in ClustVis software. X and Y axes show PC1 and PC2 that explain 53.3% and 15.1% of total variance, respectively. (C) Correlation of HFD-associated percent differences in hepatic versus serum sphingolipid ion expression. Correlation analysis was performed by Pearson correlation using GraphPad Prism 8. P < 0.05 was considered significant. HFD: high fat diet, LFD: low fat diet, SM: sphingomyelin, ST: sulfatide, NS: not significant, PC: principal component.

IV. Discussion

Long-term high fat diet feeding causes changes in the hepatic and peripheral sphingolipidome, as demonstrated by several studies (Leamy et al., 2013; Postic & Girard, 2008; Apostolopoulou et al., 2018; Longato et al., 2012; Dreiseward et al., 2007; Hidaka et al., 2007; Summer, 2007; Meriaux et al., 2010). Herein, we provide information about the most abundant and commonly detected sphingolipids as well as new sphingolipid species that have not been published before. Most importantly, we report several sphingolipid subclasses that increased significantly in the serum and liver after 16 weeks of high fat diet feeding and could play a significant role in the development of NAFLD. This study describes how circulating sphingolipid profiles are altered with high fat diet in an experimental model of NAFLD, correlates the changes in serum and hepatic sphingolipids mediated by chronic high fat diet feeding, and highlights the potential implications of these changes on NAFLD progression.

The composition of serum sphingolipids detected by MALDI-TOF was similar to previous studies (Gorden et al., 2015; Meriaux et al., 2010; Colsch et al., 2011) consisting of three major sphingolipid subclasses: ceramides, sphingomyelins, and sulfatides. In addition, we detected several glycosphingolipids within 1500 and 1700 Da mass range, which have lower abundance in the serum relative to other sphingolipids, and thus they are less studied sphingolipid species in NAFLD. MALDI mass spectra also included several phospholipid and glycerolipid subclasses; however, these lipids are beyond the scope of this study.

Chronic high fat diet feeding caused substantial alterations in hepatic lipid composition and profiles as demonstrated by excellent separation of high fat diet and control groups in the PCA plot. In contrast, for serum, PCA revealed a partial overlap between these two groups suggesting that the effects of chronic high fat diet feeding on circulating sphingolipid profiles are not as uniform and severe as observed with hepatic sphingolipids.

The data bar plots demonstrate the magnitude of change and direction of responses to high fat diet feeding on sphingolipid ion expression. Among 47 serum sphingolipids, chronic high fat diet feeding had statistically significant stimulatory effects (P < 0.05) on 6 sphingolipids and trend effects (0.05 < P < 0.1) on 6 sphingolipids in the serum samples. On the other hand, high fat diet feeding had a striking effect on hepatic sphingolipids with 14 sphingolipid species significantly increased to a much higher level of expression relative to serum sphingolipids. HFD-associated increases in hepatic and serum sphingolipids could indicate increased hepatic and peripheral insulin resistance as observed in previous models of high fat diet induced NAFLD in response to accumulation of toxic sphingolipids (Postic & Girard, 2008; Kotronen et al., 2010; Meriaux et al., 2010). In addition, it is important to emphasize that although high fat diet had a more striking stimulatory effect on liver, the course of action of serum and hepatic sphingolipids remained the same suggesting that circulating sphingolipids might be informative for hepatic changes during the development of NAFLD.

Previous studies provided evidence for altered sphingolipid metabolism in NAFLD, implicating an important pathophysiological role for this lipid class (Leamy et al., 2013; Gault et al., 2010; Hanamatsu et al., 2014). Sphingolipids contribute to the development of NAFLD via various pathways mainly including insulin resistance, inflammation, and oxidative stress (Leamy et al., 2013; Kotronen et al., 2010; Hanamatsu et al., 2014; Shanta et al., 2011). The contribution of ceramides to these pathophysiological processes of NAFLD has been extensively studied yet less is known about the role of other sphingolipids such as sphingomyelins, sulfatides, and especially glycosphingolipids (Kotronen et al., 2010; Gode & Volmer, 2013). Sphingolipid metabolism is tightly regulated, and a slight change in the level of an intermediate or activity of an enzyme involved in sphingolipid metabolism could have substantial effects on the sphingolipidome with consequences for cell structure, metabolism, and signaling (Gault et al., 2010). Sphingomyelin is the most abundant sphingolipid in the mammalian cells, mainly localized in the plasma membrane (Eisinger et al., 2014). Sphingomyelins can be hydrolyzed into ceramide through the sphingomyelinase pathway, therefore a slight change in sphingomyelin concentration can significantly influence ceramide abundance and cause toxic and degenerative effects (Gault et al., 2010). Our study shows that chronic high fat diet feeding causes upregulation of hepatic and serum sphingomyelin expression, which is in agreement with previous reports (Apostolopoulou et al., 2018; Summer, 2007; Meriaux et al., 2010). Imaging mass spectrometry studies revealed that sphingomyelins are associated with the central veins of the liver, where de novo lipogenesis occurs however, they become increasingly delocalized with the progression of NAFLD (Shanta et al., 2011; Gode & Volmer, 2013). Furthermore, areas of fibrosis were characterized with increased sphingomyelins including SM(40:1) in human livers diagnosed with NASH (Gode & Volmer, 2013). Another study identified SM(41:1) as one of the potential predictors of NAFLD (Jackson et al., 2014). Herein, we report increased intensity of circulating SM(40:0) and SM(42:2), which are likely to contribute to NAFLD progression due to high structural similarity with SM(40:1) and SM(41:1) as well as accumulation in HFD exposed mouse serum and liver samples. Furthermore, sphingomyelin species marked in the sera and livers of high fat diet-fed mice were significantly increased in other experimental models of hepatic steatosis induced by chronic alcohol or tobacco nitrosamine NNK exposures providing a strong rationale for further investigation of these sphingolipid markers in the development of NAFLD (Rao et al., 2016).

Sulfatides are acidic glycosphingolipids also known as 3-O-sulfogalactosylceramides, derived from galactosyl ceramides via esterification of a sulfate group to 3-hydroxyl of the galactose moiety (Afshinnia et al., 2018). Previous studies demonstrated that serum sulfatide levels were strongly correlated with hepatic and that increased sulfatide levels were also associated with upregulation of hepatic expression of cerebroside sulfotransferase, a key enzyme in sulfatide synthesis in a PPAR-α dependent manner (Metsalu & Vilo, 2015). PPAR-α has a protective role in several liver diseases. For example, it is downregulated in alcoholic liver disease through attenuation of increases in oxidative stress (Yalcin et al., 2020). PPAR-α is also shown to be downregulated following liver transplantation in mice (Puri et al., 2009). Constitutive activation of PPAR-α ameliorates hepatic steatosis and inflammation in mice (Kahle et al., 2015). Conceivably, HFD-associated increases in hepatic and serum sulfatide may reflect compensatory responses to overcome hepatic steatosis and inflammation.

We also observed significantly increased levels of two neutral glycoceramides, Hex(6)-Cer(40:1) and Hex(4)-HexNAc(2)-Cer(34:1) in the sera of high fat diet-fed mice. This finding suggests that HFD-mediated increase in ceramides levels are not via scavenge salvage pathway which leads to production of ceramides through the catabolism of hexosylceramides. Glycosphingolipids have been proposed as potential markers of hepatic steatosis by previous studies (Montefusco et al., 2018). Hence, inhibition of glycosphingolipid synthesis by pharmacologic modulation of glucosylceramide synthase has been shown to improve insulin sensitivity and reverse hepatic steatosis in mice (Montefusco et al., 2018; Puri et al., 2007). Future mechanistic studies are needed to further investigate the role of specific glycosphingolipid species in the progression of NAFLD.

Strengths of this study are the simultaneous analysis of serum samples obtained from a mouse model of chronic HFD-induced NAFLD and comparison of serum sphingolipid alterations with hepatic sphingolipids from paired liver samples. Herein, we provide a sphingolipidomic signature to elucidate molecular mechanisms in relation to lipid dysregulation in NAFLD. Many of the changes reported in this paper would not be possible with sole measurement of total lipid or fatty acid content. The main limitation of this study is that our data reveals associations but not causal relationships. Therefore, we don’t know yet whether there is a strong correlation between the identified sphingolipids and hepatic insulin resistance or inflammation. Although we observed a linear correlation between percent changes in serum and hepatic sphingolipids, it is unknown whether serum sphingolipid alterations are mediated by hepatic sphingolipids. Future studies will aim to incorporate in vitro mechanistic studies. It is also important to note that only male mice have been included in this study.

V. Conclusion

This study comprehensively analyzed sphingolipid levels in the sera of high fat diet-fed mice and compared alterations in circulating sphingolipids with hepatic sphingolipids. This allowed us to identify specific sphingomyelin and sulfatide species in serum that reflect the respective changes in the liver. Furthermore, these lipid metabolites relate to hepatic insulin resistance, oxidative stress, and inflammation suggesting that they contribute to progression of simple steatosis to non-alcoholic steatohepatitis. These findings further highlight focusing on specific sphingolipid species for future mechanistic studies to decipher their role for hepatic steatosis and insulin resistance.