Regulation of Serum Sphingolipids in Andean Children Born and Living at High Altitude (3775 m)

Recent studies on Andean children indicate a prevalence of dyslipidemia and hypertension compared to dwellers at lower altitudes, suggesting that despite similar food intake and daily activities, they undergo different metabolic adaptations. In the present study, the sphingolipid pattern was investigated in serum of 7 underweight (UW), 30 normal weight (NW), 13 overweight (OW), and 9 obese (O) Andean children by liquid chromatography-mass spectrometry (LC-MS). Results indicate that levels of Ceramides (Cers) and sphingomyelins (SMs) correlate positively with biochemical parameters (except for Cers and Vitamin D, which correlate negatively), whereas sphingosine-1-phosphate (S1P) correlates negatively. Correlation results and LC-MS data identify the axis high density lipoprotein-cholesterol (HDL-C), Cers, and S1P as related to hypoxia adaptation. Specifically UW children are characterized by increased levels of S1P compared to O and lower levels of Cers compared to NW children. Furthermore, O children show lower levels of S1P and similar levels of Cers and SMs as NW. In conclusion, our results indicate that S1P is the primary target of hypoxia adaptation in Andean children, and its levels are associated with hypoxia tolerance. Furthermore, S1P can act as marker of increased risk of metabolic syndrome and cardiac dysfunction in young Andeans living at altitude.


Introduction
Several studies have addressed the comprehension of molecular mechanisms at the basis of a high-altitude hypoxia adaptation of Andeans and Tibetans, allowing them to permanently inhabit at 3500-4500 m of altitude [1][2][3]. These two ethnic groups are characterized by different traits associated with adaptive responses to hypobaric hypoxia, however the molecular mechanisms are still not completely clear. Advanced genome-wide scanning identified genes and genetic variants that contribute to human adaptation to altitude in Tibetans [4,5] and Andeans [6,7], confirming different adaptive mechanisms. Specifically, Andeans appear to develop metabolic syndrome [8], which is not described in Tibetans, and they have elevated hemoglobin concentrations compared to Tibetans,

Anthropometrics and Biochemical Differences
Data from comparison of biochemical parameters and anthropometric characteristics are summarized in Table 1.
HDL-C was higher in NW subjects compared to O (p-value < 0.01), whereas TC/HDL-C ratio was higher in O subjects compared to UW and NW (p-value < 0.05 and 0.01, respectively). Homeostatic Model assessment for insulin resistance (HOMA-IR) and insulin were higher in obese and overweight subjects compared to NW (p-value < 0.01 and 0.05, respectively, in both comparisons).

Sphingolipids and Biochemical Parameters Correlation
Scatter plots of total Cers, total SMs, and S1P levels with main biochemical parameters are summarized in Figure 1; a correlation table can be found in Supplementary Table S1. UW subjects showed a positive correlation between SMs levels and HDL-C (r = 0.854, p-value = 0.015).
In O subjects, a correlation was found comparing Cers and Glycaemia (r = 0.683, p-value = 0.043). summarized in Figure 1; a correlation table can be found in Supplementary Table 1. UW subjects showed a positive correlation between SMs levels and HDL-C (r = 0.854, p-value = 0.015).
In O subjects, a correlation was found comparing Cers and Glycaemia (r = 0.683, p-value = 0.043).

Figure 1.
Total Ceramide (A), total sphingomyelin (B), and total S1P (C) scatter plots. Data for cers, SMs, and S1P are expressed as pmol/mL in log scale; data for Vitamin D are in ng/mL and data for insulin are in µU/mL; TG, TC, and LDL-C are in mg/dL; p-values from two tailed test, r value from Pearson's correlation.
SMs acyl chains followed the same trend highlighted for total SMs. In particular, SM (d18:1/18:0) was higher in OW subjects compared to both NW (p-value < 0.001) and O (p < 0.01); this increment was not observed in UW. SM (d18:1/20:0) was also higher in OW, comparing NW (p-value < 0.001) and O (p-value < 0.001) groups. SM (18:1/22:0) showed the same behavior of SM (d18:1/20:0) but significant differences were observed also in UW vs. NW, where higher levels were seen in the UW group (p-value < 0.001) and in UW vs. OW, where UW subjects were characterized by lower levels of SM (p-value < 0.001). Conversely, total SMs ( Figure 3) were higher in OW subjects compared to O (p-value < 0.01) and NW (p-value < 0.001) subjects, and were also higher in UW compared to NW (p-value < 0.01).
SMs acyl chains followed the same trend highlighted for total SMs. In particular, SM (d18:1/18:0) was higher in OW subjects compared to both NW (p-value < 0.001) and O (p < 0.01); this increment was not observed in UW. SM (d18:1/20:0) was also higher in OW, comparing NW (p-value < 0.001) and O (p-value < 0.001) groups. SM (18:1/22:0) showed the same behavior of SM (d18:1/20:0) but significant differences were observed also in UW vs. NW, where higher levels were seen in the UW group (p-value < 0.001) and in UW vs. OW, where UW subjects were characterized by lower levels of SM (p-value < 0.001).
Regarding total dhCers (Figure 4), the same trend of Cers was identified-NW and O subjects showed higher levels of serum dhCers compared to the OW group (p-value < 0.001 in both comparisons).
LC-MS analysis for total glycosphingolipids HexCer, diHexCer, and GM3 ( Figure 4) identified a characteristic pattern in all groups-NW patients have higher levels of the above-mentioned SLs species compared to UW (p-value < 0.001). For total HexCer, NW subjects showed higher levels also compared to OW and O (p-value < 0.05 and 0.001, respectively). For total diHexCer, O show lower levels compared to NW (p-value < 0.001), whereas UW subjects show lower levels compared to OW (p-value < 0.05). Total GM3 was higher in NW compared to OW (p-value < 0.01) and to O (p-value < 0.05) subjects. GM3 was higher in OW and O subjects compared to the UW (p-value < 0.05 in both comparisons) group. Regarding total dhCers (Figure 4), the same trend of Cers was identified-NW and O subjects showed higher levels of serum dhCers compared to the OW group (p-value < 0.001 in both comparisons).
LC-MS analysis for total glycosphingolipids HexCer, diHexCer, and GM3 ( Figure 4) identified a characteristic pattern in all groups-NW patients have higher levels of the above-mentioned SLs species compared to UW (p-value < 0.001). For total HexCer, NW subjects showed higher levels also compared to OW and O (p-value < 0.05 and 0.001, respectively). For total diHexCer, O show lower levels compared to NW (p-value < 0.001), whereas UW subjects show lower levels compared to OW (p-value < 0.05). Total GM3 was higher in NW compared to OW (p-value < 0.01) and to O (p-value < 0.05) subjects. GM3 was higher in OW and O subjects compared to the UW (p-value < 0.05 in both comparisons) group.
A clear trend for decrement was found across different BMI percentiles in S1P levels, and statistically significant differences were observed comparing UW vs. O (p-value < 0.05) and NW vs. O (p-value < 0.05). A clear trend for decrement was found across different BMI percentiles in S1P levels, and statistically significant differences were observed comparing UW vs. O (p-value < 0.05) and NW vs. O (p-value < 0.05).
To Summarize:  To Summarize:

Discussion
This is the first study considering a sphingolipid pattern in relation to high-altitude hypoxia

Discussion
This is the first study considering a sphingolipid pattern in relation to high-altitude hypoxia exposure. In previous reports, a discrepancy between the lipid and metabolic profile of Koya children, born and living at 3775 m a.s.l., compared with age and sex matched lowlanders [15,16], was described. This group was characterized by slightly high glucose compared to lowlanders, lower insulin level, lower HDL level, whereas TC/HDL-C, LDL-C/HDL-C, and triglycerides/HDL-C were higher in high-altitude children, and particularly in overweight and obese [8,[15][16][17]37] children. An independent study performed in the San Pedro de Cajas district, located in the Central Andes of Peru at 4100 m a.s.l., showed similar results, with a high prevalence of hypertriglyceridemia (53.9%) and low HDL-C (45.3%) in indigenous descendants of the Amerindian populations, identifying these traits as characteristic of this population living at altitude [38]. In the present study, metabolic parameters show that mean levels of HDL-C are lower than the safety cut-off (45 mg/dL) for UW and O subjects, while triglycerides range from borderline (79-99 mg/dL) in UW, NW, and OW subjects to high (>110 mg/dL) in O subjects, confirming a particular dyslipidemic pattern in this group. More importantly, results from the present study identified the axis HDL-C, ceramide qualitative and quantitative variations, and S1P levels as related to hypoxia adaptation and body weight in Koya children, and indicated that total ceramide levels correlate directly with several biochemical parameters, whereas S1P correlate indirectly. Figure 5 summarizes the levels of sphingolipids or glycosphingolipids in the four groups and indicates that UW and O subjects are characterized by similar glycosphingolipids levels, with S1P values varying, being high in UW subjects and low in O subjects.
Andes of Peru at 4100 m a.s.l., showed similar results, with a high prevalence of hypertriglyceridemia (53.9%) and low HDL-C (45.3%) in indigenous descendants of the Amerindian populations, identifying these traits as characteristic of this population living at altitude [38]. In the present study, metabolic parameters show that mean levels of HDL-C are lower than the safety cut-off (45 mg/dL) for UW and O subjects, while triglycerides range from borderline (79-99 mg/dL) in UW, NW, and OW subjects to high (>110 mg/dL) in O subjects, confirming a particular dyslipidemic pattern in this group. More importantly, results from the present study identified the axis HDL-C, ceramide qualitative and quantitative variations, and S1P levels as related to hypoxia adaptation and body weight in Koya children, and indicated that total ceramide levels correlate directly with several biochemical parameters, whereas S1P correlate indirectly. Figure 5 summarizes the levels of sphingolipids or glycosphingolipids in the four groups and indicates that UW and O subjects are characterized by similar glycosphingolipids levels, with S1P values varying, being high in UW subjects and low in O subjects. Figure 5. Summary of the altered ceramide-biosynthesis pathways. Different size arrows indicate the different involvement of Cers biosynthetic pathways in UW and O. LC-MS and LC-MS/MS results for total Cers, total dhCers, total SMs, total HexCer, total dihexCer, and total GM3. Data are expressed as pmol/mL in Log scale; p values are expressed as * p-value < 0.05, ** p-value < 0.01, and *** p-value < 0.001.
In UW children, Cer levels seem to arise from de novo biosynthesis (being dhCers levels comparable to NW as S1P). Conversely, O subjects are characterized by similar levels of ceramides as NW, whereas S1P is significantly decreased, making this molecule a candidate as a central node in the metabolic adaptation to hypoxia of Koya children.
A previous important study demonstrated that increment of S1P is associated with increased capacity of oxygen delivery from erythrocytes, by increasing 2,3-diphosphoglycerate and activating Figure 5. Summary of the altered ceramide-biosynthesis pathways. Different size arrows indicate the different involvement of Cers biosynthetic pathways in UW and O. LC-MS and LC-MS/MS results for total Cers, total dhCers, total SMs, total HexCer, total dihexCer, and total GM3. Data are expressed as pmol/mL in Log scale; p values are expressed as * p-value < 0.05, ** p-value < 0.01, and *** p-value < 0.001.
In UW children, Cer levels seem to arise from de novo biosynthesis (being dhCers levels comparable to NW as S1P). Conversely, O subjects are characterized by similar levels of ceramides as NW, whereas S1P is significantly decreased, making this molecule a candidate as a central node in the metabolic adaptation to hypoxia of Koya children.
A previous important study demonstrated that increment of S1P is associated with increased capacity of oxygen delivery from erythrocytes, by increasing 2,3-diphosphoglycerate and activating enzymes involved in glycolysis [39]. In the circulation, 2/3 of the S1P is bound to HDL, carried out by the Apolipoprotein M (apoM), and 1/3 is bound to albumin [40,41]. S1P bound to apoM promotes vasorelaxation by phosphorylating endothelial nitric oxide synthase (eNOS) [42]. The eNOS is responsible for the generation of nitric oxide (NO) in the vascular endothelium and plays a crucial role in vascular tone [43] and homeostasis through NO. Further, eNOS uncoupling generates endothelial dysfunction due to superoxide anions and hydrogen peroxide overproduction. A previous study from our group demonstrated that the NO releasing donors causes ceramide accumulation in tumor cell lines, inducing a switching of the Akt hyperactivated signal leading to UPR (unfolded protein response) stress response [44]. There is a possibility that ceramide accumulation is also, in this case, associated with UPR activation. Further studies are required to elucidate the mechanism at the basis of ceramide levels of high-altitude living subjects.
Another study by our group on muscle tissues of Tibetans living at altitude and Tibetan lowlanders indicates an increment of e-and nitric oxide synthase (n-NOS) in muscle tissue and of 2,3-diphosphoglycerate (2,3-BPG). The latter is an erythrocyte-specific glycolytic intermediate, which facilitates O 2 release found to be increased in animal models overexpressing S1P [39]. In animal models, intracellular S1P promoted increased production of 2,3-bisphosphoglycerate. In light of the present results, the authors can postulate that UW children utilizing S1P can activate the same biochemical pathway promoting oxygen delivery to maintain energy production and homeostasis, thereby keeping levels of ceramides as low as possible. Conversely, O children, characterized by lower levels of S1P, are unable to reduce ceramide levels; even if the salvage pathway is decreased, they may lose the capacity to adapt to a lack of oxygen, therefore activating the lipogenic and lipotoxic pathways [45][46][47][48].
Another point that characterizes UW and O children is the decreased levels of glycosphingolipids generated in the salvage pathway. The activation of Neuraminidase 3 (NEU3) sialidase in transgenic mice expressing NEU3 sialidase caused reduced tissue and circulating levels of GM3 [49]. This molecule was found increased in our studies on muscle tissue of animal models exposed to prolonged severe hypoxia [50]. In the present study, it can be speculated that decreased levels of GM3, Hexosylceramides, and dihexosylceramides can be associated with the activation of NEU 3 sialidase. The latter has been recently described as a target of HIF 2 alpha, and thus regulated by hypoxia and associated with obesity [51].
Concerning levels of different ceramide acyl-chains, it can be hypothesized that the elevated C16:0-ceramide levels observed in O subjects can contribute to insulin resistance [23,52]. Ceramides substantially upregulate fatty acid uptake and synthesis through direct or indirect mechanisms and could be central modulators of obesity and hepatic steatosis [53]. Furthermore, specific ceramide synthases (in this case, CerS6) have been associated with weight gain and glucose intolerance [23]. Several studies addressed the issue of the role of S1P in inflammation. In this context, O are characterized by low levels of S1P compared to UW. It has been described that S1P improves energy homeostasis in obesity, particularly in the early phase, acting as protective molecule to control overweight and inflammation [54]. This observation is in agreement with our results, since in OW levels of S1P are high compared to O. Furthermore, this molecule, in conjunction with its receptor, decreases endothelial inflammation and promotes cardioprotection mediated by HDL-Apo-M-S1P binding and release [55][56][57]. Further studies in this direction are required.
The major limitations of this study are the restricted number of subjects and the absence of a verification study, which will require an independent group of subjects. Further data based on 2,3-BPG, pH, partial pressure of carbon dioxide (PCO 2 ), of oxygen (PO 2 ), and haemoglobin-O 2 saturation (SO 2 ) will be necessary to determine the abnormal O 2 affinity or altered metabolism of RBC [58]. This data, in conjunction with lipemia, could be supportive of our results. Another important aspect that has not been addressed in this study for unavailability of fresh serum samples is the study of RNAs from peripheral blood mononuclear cells (PBMCs) for the quantitative assessment of transcripts of enzymes involved in ceramide pathways and of circulating cytokines or adipokines. Studies in this direction are ongoing and are aimed at opening new avenues in the complex field of hypoxia tolerance in high-altitude dwellers.
In conclusion, ceramides levels and S1P are the primary targets of hypoxia adaptation in populations born and living at high altitude. An important point highlighted in this study is that S1P can be a circulating biomarker of hypoxia tolerance and a valuable marker of increased risk of metabolic syndrome and cardiac dysfunction, being directly quantifiable in blood by selected reaction monitoring (SRM) mass spectrometry.
The study was carried out following the rules of the Declaration of Helsinki and was approved by the Human Rights Committee of the Salta Health Ministry (project number: 4518; date of approval: 8 November 2010). Each caregiver and child gave written informed consent after an explanation of the study and before its initiation.

Lipid Extraction
Sphingolipids were extracted from sera according to a previous study [60], with minor modification. 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: Sphinganine (d17:0), sphinganine-1-phosphate (d17:0), C12 Ceramide (d18:1/12:0), C12 Sphingomyelin (d18:1/12:0), and Glucosyl (β) C12 Ceramide. Samples were briefly sonicated and heated at 48 • C overnight. Then, 0.15 mL of 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.

UPLC-MS for Sphingolipids
Liquid chromatography-mass spectrometer configuration included a Waters Acquity UPLC system linked to a Waters LCT Premier Orthogonal Accelerated Time of Flight Mass Spectrometer (Waters, Millford, MA, USA). The instrument operated in positive or negative electrospray ionization mode. Full scans were obtained in a window, ranging from 50 to 1500 Da. Accuracy and reproducibility were maintained employing an independent reference spray via LockSpray. First, 10 µL of sphingolipid extract were injected and separated on an analytical column, kept at 30 • C, 100 mm × 2.1 mm id, 1.7 µm C8 Acquity UPLC BEH (Waters), 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. Sphingolipids quantification was carried out using the ion chromatogram obtained for each compound using 50 mDa windows. The linear dynamic range was determined by 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™ 4.1 Software (Waters, Millford, MA, USA), and lipids were annotated as lipid subclasses as follows (sphingosine backbone/number of carbon atoms of the fatty acid: number of unsaturations of the fatty acid).

Statistical Analysis
Participants were grouped according to their BMI percentiles and their characteristics were described using median and interquartile range (if continuous) or counts and percentages (if categorical). Comparison of age, weight, height, BMI, TC, HDL-C, CT/HDL, LDL-C, Vitamin D, glycaemia, HOMA-IR, insulin, and TG were investigated between groups using One way ANOVA with Bonferroni's correction if data were normally distributed, otherwise the Kruskal-Wallis' with Dunn's correction was adopted. Correlation tables for sphingolipids and biochemical parameters were obtained using Pearson's or Spearman's correlation, when data were parametrically or not distributed, respectively. Differences in Cers, dhCers, SMs, dhSMs, HexCers, diHexCers, GM3, and S1P from LC-MS and LC-MS/MS quantitative data were assessed among groups.
Statistical analyses were performed using SigmaPlot software version 12.0 and Prism software version 7.0.