Sphingolipid serum profiling in vitamin D deficient and dyslipidemic obese dimorphic adults

Recent studies on Saudi Arabians indicate a prevalence of dyslipidemia and vitamin D deficiency (25(OH)D) in both normal weight and obese subjects. In the present study the sphingolipid pattern was investigated in 23 normolipidemic normal weight (NW), 46 vitamin D deficient dyslipidemic normal weight (-vitDNW) and 60 vitamin D deficient dyslipidemic obese (-vitDO) men and women by HPTLC-primuline profiling and LC-MS analyses. Results indicate higher levels of total ceramide (Cer) and dihydroceramide (dhCers C18–22) and lower levels of total sphingomyelins (SMs) and dihydrosphingomyelin (dhSM) not only in -vitDO subjects compared to NW, but also in –vitDNW individuals. A dependency on body mass index (BMI) was observed analyzing specific Cer acyl chains levels. Lower levels of C20 and 24 were observed in men and C24.2 in women, respectively. Furthermore, LC-MS analyses display dimorphic changes in NW, -vitDNW and –vitDO subjects. In conclusion, LC-MS data identify the independency of the axis high Cers, dhCers and SMs from obesity per se. Furthermore, it indicates that long chains Cers levels are specific target of weight gain and that circulating Cer and SM levels are linked to sexual dimorphism status and can contribute to predict obese related co-morbidities in men and women.

Obesity is a globally expanding pathology and up to 58% of the world's adult population is expected to be overweight or obese by 2030 1 . This condition is associated with vitamin D deficiency and is more common in women (11% of men and 15% of women were obese in 2014) [2][3][4] . Although a causal genetic association between obesity and vitamin D (25(OH) D) deficiency has been described 5 , the reasons for this association have not yet been clarified and could include a number of factors, such as differences in dietary intake or sun exposure, a decreased vitamin D bioavailability or altered vitamin D metabolism 6 . Furthermore, low plasma levels of vitamin D have been associated with obesity-related health complications, such as insulin resistance, type 2 diabetes and dyslipidemia [7][8][9] .
A recent differential proteomic study of sera of obese men and women from Saudi Arabia, in whom low vitamin D concentrations were present, identified a number of proteins that are differentially expressed in obese compared to lean weight people; these proteins belong to different pathways, including lipid metabolism, vitamin D function and immunity/inflammation 10 . Serum profiling by MALDI mass spectrometry on these same individuals indicated that increased inflammation and altered lipid metabolism were present in obese subjects 11 . Furthermore, polymorphisms of the synaptosomal-associated protein 25 (SNAP-25) gene associated with insulin resistance have been identified in obese individuals; in particular, the SNAP25 rs363050 (G) allele was shown to result in a reduced expression of SNAP25, associated with altered glycemic parameters 12 .

Results
Subjects' general characteristics and clinical parameters assessment. Serum Table 1. Gender and age composition were homogeneous across groups composed of normal weight and obese subjects (U-test p-value = 0.64, χ 2 p-value = 0.15 for age and gender respectively). -vitDNW subjects were characterized by dyslipidemia, with lower levels of HDL and higher levels of TRG and total cholesterol compared to NW (U-test p-value < 0.001 for HDL, TRG and total cholesterol), and with higher levels of TRG and lower levels of HDL compared to -vitDO (U-test p-value = 0.019 and 0.006 for HDL and TRG respectively). total sphingomyelin (SM) levels associated to dyslipidemia, vitamin D deficiency and obesity, HPTLC-primuline profiling was performed on SLs extracted from NW (n = 15), -vitDNW (n = 23) and -vitDO (n = 25) men and from NW (n = 8), -vitDNW (n = 23) and -vitDO (n = 35) women. After separation on HPTLC plates, SL bands were compared by a profiling approach based on HPTLC-densitometry and FDIC (fluorescence detection by intensity changes) emission after primuline staining (data shown in Fig. S1). From primuline stained HPTLC plates, bands corresponding to Cer and SM were identified by Rf comparison with standards, carrying a variation coefficient (CV) of 6-7%; bands with Rf = 0.923, Rf = 0.858, Rf = 0.176 and Rf = 0.152 respectively, were attributed to Cers (C20-C24), Cers (C14-C18), SMs (C20-C24) and SMs (C14-C18) (Fig. S2).
DiHexCer C18:0 levels were lower in NW men compared to women (p-value < 0.05) (Fig. 4A). In -vitDO subjects, higher levels of total HexCer and HexCer C16:0 were observed in women (Fig. 4B). Data from other sphingolipid classes in men and women are shown in Fig. S6.

Discussion
We investigated SL levels in three groups of Saudi individuals, normolipidemic with normal weight (NW), compared to vitamin D deficient and dyslipidemic subjects who were either obese (-vitDO) or had normal BMI (-vitDNW). Results herein show the presence of a general increase of total Cer in dyslipidemic (both -vitDNW and -vitDO) compared to normolipidemic NW subjects, linking total Cer abundance to dyslipidemia and vitamin D deficiency. Our results confirmed data showing higher levels of total Cer in obese (-vitDO) subjects compared to normolipidemic normal weight (NW) controls 20,26,27 . More interestingly, data herein for the first time highlighted that levels of total Cer are significantly higher not only in obese but also in dyslipidemic normal weight subjects. These results suggest that total Cer increase is associated with 25(OH)D deficiency and hyperlipidemia independently from obesity per se.
In countertrend with total Cer, total SM and total dhSM, were lower both in -vitDNW and in -vitDO compared to NW, suggesting also in this case a direct correlation of this class of molecules with vitamin D deficiency and hyperlipemia.
The strength of the present work is the investigation of sera from obese compared to dyslipemic NW and normolipidemic NW subjects that allows a better understanding of the significance of the fine regulation of Cers and SMs. In previous studies, sphingolipids abundance was compared in obese and lean healthy subjects 13,14,28 or lean normoglycemic 29 controls, without taking into account dyslipidemia. A recent study indicates the same relationship among 25(OH)D deficiency, total Cer increase and dyslipidemia in Andean children characterized by different BMIs 25 . It could be speculated that in Saudi Arabian subjects, total Cer increment in -vitDNW and -vitDO is related to palmitoyl-CoA /serine and fatty acids de novo biosynthetic route as suggested by the increment of total dhCer and of dhCer specific chains (Fig. S5) 30,31 . Ceramide generation, during HFD-induced obesity, is stimulated by the continuous and excessive supply of FFA from diet and adipose tissue providing substrate for serine palmitoyl transferase and Ceramide synthase CerS(1-6) isoforms 16,32 increasing plasmatic Cer levels. Notably, total SM and dhSM were decreased in -vitDNW and in -vitDO, suggesting a higher activity of sphingomyelinase in membranes 33 . In animal models, HFD administration stimulates the expression and the activity of both acidic sphingomyelinase (aSMase) and neutral sphingomyelinase (nSMAse) 34,35 . It has been observed that the pharmacological inhibition of aSMase by amitriptyline reduced HFD-stimulated ceramide release in blood 36 .
Little is known regarding SLs behavior during dyslipidemia and vitamin D deficiency. Thanks to their hydrophilicity, sphingolipids circulate in plasma bound to albumin and lipoproteins 37 . Particularly, Cer is present and equally distributed in HDL-C, LDL-C and VLDL-C 22 , while 70% of SM is carried by LDL/VLDL-C and 30% by HDL-C 22 . Interestingly, -vitDNW subjects showed higher levels of total Cer, dhCer and lower levels of total SM and dhSM, compared to normolipidemic NW, suggesting that -vitDNW subjects, independently from obesity, develop a sphingolipid plasma profile that is influenced by LDL-C, HDL-C and TRG levels. It can be speculated that normal-weight subjects, with vitamin D deficiency and dyslipidemia, develop altered plasma SL profiles because of the differential "transport" capacity of lipoprotein, whereas the alterations observed in obesity are related to changes in SLs biosynthesis (e.g., de-novo biosynthesis and sphingomyelinase pathway) (Fig. S5) 21 . Specific acyl chains of Cer and SM rather than total levels were found at variance in obesity supporting the role of different Cer synthase isoforms in the regulation of SLs production in obesity. Specifically in obese men, we found chains of Cer, namely Cers C20:0 and C24:2 at lower levels and higher levels of dhSM C18:0. Obese women had lower levels of Cer C24:1 and higher levels of SM C16:0. These results pinpoint the relevance of different Cers  www.nature.com/scientificreports www.nature.com/scientificreports/ synthase (1-6) isoforms in the regulation of single acyl chain Cer synthesis in obesity. Our results indicate that differentially abundant acyl-chains Cer are directly associated with increased SM and dhSM specific chains in obese subjects (SM C16:0 in women and dhSM C18:0 in men).
Although sexual dimorphism in body composition is evident in adulthood 38 , sex has been only sporadically considered in metabolomic and lipidomic studies in obesity 39 . Despite increased levels of circulating free fatty acids, women readily oxidize non-esterified plasma fatty acids, increasing ketone bodies production 40 or lipid re-esterification 41 and display improved insulin sensitivity.
Results from the present study indicate in men increased levels of Cer C20:0, C24:1 and C24:2 as a characteristic trait of normolipidemic normal-weight subjects confirming results from a larger cohort of Mexican Americans 42 , that attribute to cerS-2 higher levels of long-chain ceramides.
Also SMs characterized dimorphic changes being SM C18:0, C20:0, C22:1 and C24:1 higher in normolipidemic normal-weight men compared to women. These results are in contrast with other studies 43-45 that found higher levels of SM in women. However, once again, considering dyslipidemic normal-weight and obese women, SM were more abundant 44 , and particularly SM C22:0 and C14:1 were higher compared to men both dyslipidemic normal-weight and obese.
Concerning dhS1P, this molecule is present at higher levels in -vitDNW and -vitDO women compared to NW. It can be hypothesized that this molecule contributes to keep under control levels of dhCer thus subtracting dhCer to total Cer synthesis being levels of total Cer in -vitDNW and -vitDO similarly increased both in men and women.
It can be hypothesized that higher levels of hexosylceramide found in obese women could be related to overexpression of hexosylceramide synthase in adipocytes that suppresses insulin signaling 46 . Hexosylceramide has been described as a possible inducer of plaque inflammation and instability 47 . It must be stated that due to the similarity in the hydrophobic moiety and mass identity, hexosylceramides (glucosylceramides and galactosylceramides) are not easy to separate and quantify by reverse-phase chromatography 48 , therefore observed changes are related to both glucosylceramides and galactosylceramides and further studies will be required to precisely identify the contribution of single species to the total pool and the role of different glycosphingolipids.
Although the use of two approaches to study SLs in serum, the quantitative structure-specific measurements of all molecular species remains challenging.
The novelty of this study is that it demonstrated for the first time that Cer and SM levels correlate directly with dyslipidemia and vitD deficiency independently from BMI and that only specific acyl chains of Cer and SM are directly correlated with obesity per se. Furthermore, it highlights the importance of sexual dimorphism in determining circulating SLs levels independently from serum lipids and vitamin D status. Notably, results obtained in normal weight subjects indicated that sex influences serum concentrations of Cer and SM 42 . We confirmed those results and expanded them in the setting of obesity and vitamin D deficiency.
However, the inter-individual variability, the number of samples, the paucity of reference data could be seen as limiting factors in this study. These limitations notwithstanding, it is important to underline that results herein suggest that these putative biomarkers could be used to predict the risk for vitamin D deficient adult subjects to develop associated co-morbidity in obesity.
The possible effect of vitamin D and statins supplementation on Cer levels is currently being investigated in our cohorts to better clarify the role of drugs on SLs concentration.

Materials and Methods
Participants and sample collection. Adult Table 1. Vitamin D deficiency was defined as circulating serum 25(OH)D < 50 nmol/l (<20 ng/ml) 50 . www.nature.com/scientificreports www.nature.com/scientificreports/ Sphingolipid analysis. For sphingolipid analysis two different strategies were considered: High performance thin layer chromatography (HPTLC)-primuline profiling were first conducted to analyze all serum samples (from 23 NW, 46 -vitDNW and -vitDO subjects), whereas LC-MS analysis were carried out on 4 sub-pools per group, both for men and women (24 total sub-pools).

Reagents and Chemicals. Propan
The use of primuline as a fluorophore for quantitative purposes was first described by Domínguez et al. 53 as FDIC, i.e. fluorescence detection by intensity changes. FDIC detects fluorescent emission produced by non-specific and electrostatic interactions between the primuline and hydrocarbon chains in the ceramide backbones of sphingolipids [54][55][56][57] . In our work, primuline was used for semi-quantitative purposes, to detect changes in ceramides and sphingomyelins levels among NW, -vitDNW and -vitDO men and women.

Sphingolipids by LC-MS.
For LC-MS analysis, sera were randomly sub-pooled into 4 groups of men and 4 groups of women, homogeneous for biometric parameters and HPTLC quantitative profiles. SL extracts were analyzed in the presence of internal standard 52 , prepared as described by Merrill et al. 59 , including an alkaline hydrolysis step to remove phospholipids, and analyzed by Waters Aquity Ultra Performance Liquid Chromatography (UPLC) system connected to a Waters LCT Premier orthogonal accelerated time of flight mass spectrometer (Waters, Millford, MA), operating in positive electrospray ionization mode. Spectra were acquired according to 52 . Positive identification of compounds was based on the accurate mass measurement with an error <5 ppm and its LC retention time, compared to that of a standard (±2%). Mass spectra were analyzed by MassLynx ™ 4.1 Software.
LC-MS/MS analyses of sphingosine, S1P and dhS1P were carried out on an Acquity UPLC system coupled with a Xevo TQ-MS triple quadrupole mass spectrometer. The mass spectrometer was operated in the positive ESI mode, and analytes were quantified by multiple reaction monitoring (MRM Statistical analysis. Participants were grouped according to sex, obesity and dyslipidemia status and their characteristics were described using median and interquartile range, if continuous, and percentages, if categorical ( Table 1). Comparison of serum cholesterol and triglycerides were performed using Mann-Whitney U-test. Circulating levels of Cer and SM were transformed as logarithm and were compared by sex, weight group and dyslipidemia status using a generalized linear model (GLM) adjusted for BMI and age. Retro transformed least square means from the GLM models were reported along with their standard errors and comparisons were performed using the Tukey adjustment to control for inequalities of groups sizes 60 . For LC-MS analysis, sera were randomly sub-pooled into 4 matched groups of men and women considering biometric parameter and HPTLC quantitative profiles. The pooling was adopted as a method to reduce the variance among biological groups increasing the power to detect changes when few samples are available and the variance is high 61,62 . All statistical analyses were performed using the SAS software version 9.4. Statistical tests were two tailed and type-I error rate was set at 5% (α = 0.05).

Data availability
The data generated during and/or analysed during the current study are included in this published article (and its Supplementary Information files). When needed, further information are available from the corresponding author on reasonable request.