Elucidating the chemical structure of native 1-deoxysphingosine

1-Deoxysphingolipids (1-deoxySL) are formed by an alternate substrate usage of the enzyme serine-palmitoyltransferase and devoid of the C1-OH-group present in canonical sphingolipids (SL). Pathologically elevated 1-deoxySL levels are associated with the rare inherited neuropathy HSAN1 and diabetes type 2 (T2DM) and might contribute to beta cell failure and the diabetic sensory neuropathy. In analogy to canonical SL it was assumed that also 1-deoxySL bear a (4 E ) double bond which is normally introduced by sphingolipid delta(4)-desaturase 1 (DES1). This, however, was never confirmed. We therefore supplemented HEK293 cells with isotope labeled D 3 -1-deoxy-sphinganine (1-deoxySA) and compared the downstream formed D 3 -1-deoxy-sphingosine (1-deoxySO) to a commercial synthetic SPH m18:1(4 E )(3OH) standard. Both compounds showed the same mass-to-charge ratio ( m/z ) but differed in their RPLC retention time and APCI in-source fragmentation, suggesting that the two compounds are structural isomers. Using dimethyl disulfide (DMDS) derivatization followed by MS 2 as well as differential mobility spectrometry combined with ozon-induced dissociation mass spectrometry, we identified the carbon-carbon DB in native 1-deoxySO to be located at (Δ14) position. Comparing the chromatographic behavior of native 1-deoxySO to chemically synthesized SPH m18:1(14 Z ) and (14 E ) stereoisomers assigned the native compound to be SPH m18:1(14 Z ). This indicates that 1-deoxySL are metabolized differently than canonical sphingolipids.


Introduction
Sphingolipids are typically formed by the condensation of serine and palmitoyl-CoA, a reaction catalyzed by the serine-palmitoyltransferase (SPT) enzyme. Besides these canonical substrates SPT can use other acyl-CoA's but also alanine or glycine as substrate which then forms a category of atypical 1-deoxysphingolipids (1-deoxySL) which lack the C1-OH group of canonical sphingolipids (1,2).
Several missense mutations in SPT which are associated with the rare inherited neuropathy HSAN1, induce a permanent shift in the substrate specificity of the enzyme resulting in increased 1-deoxySL formation. HSAN1 is a rare autosomal and dominantly inherited axonopathy and clinically characterized by a progressive loss of pain and temperature sensation often accompanied by neuropathic pain attacks and skin ulcers (3). 1-DeoxySL are toxic to primary sensory neurons in culture and lead to neurite retraction and the disruption of the neuronal cytoskeleton structure in a dose dependent manner (3,4). They also interfere with the survival and insulin secretory capacity of pancreatic beta cells and 1-deoxySL plasma levels have been found to be prospective biomarkers for the risk to develop T2DM (5-7).
The 1-deoxy-sphinganine (1-deoxySA) which is formed by SPT can be converted to 1-deoxyceramides (1-deoxyCer) but not to complex sphingolipids because of the missing 1-OH group. During catabolism 1-deoxyCer is degraded by ceramidase to form 1-deoxysphingosine (1-deoxySO) but not phosphorylated to form the catabolic intermediate sphingosine-1-P (S1P). This prevents its cleavage to hexadecenal by S1P-lyase meaning that 1-deoxySL cannot be degraded by the canonical catabolic pathway (2). Apart from that, it was assumed that 1-deoxySA is metabolized by the same set of enzymes as canonical sphingoid bases and that 1-deoxySO like SO bears a (4E) double bond which is introduced by the sphingolipid delta(4)-desaturase 1 (DES1).
However, we observed that natively formed 1-deoxySO showed a different RPLC retention time than a synthetic SPH m18:1(4E)(3OH) standard, though the m/z was identical for both compounds. This suggested that native 1-deoxySO and the synthetic SPH m18:1(4E)(3OH) are structural isomers probably differing in position and / or configuration of the carbon-carbon double bond. To further elucidate this difference we used a set of tandem mass spectrometry methods in combination with total synthesis to elucidate the real double bond position and configuration of native 1-deoxySO.

Materials and Methods
Unless stated differently all solvents and reagents were purchased from Sigma-Aldrich Chemie GmbH (Buchs, Switzerland) excluding methanol, which was purchased from Honeywell specialty chemicals Seelze GmBH, Germany.

Cell extract
HEK293 cells were fed with 1 µM deuterium labelled D 3 -1-deoxysphinganine (Avanti Polar Lipids, Alabaster, AL) or with the unlabeled 1-deoxysphinganine (Avanti Polar Lipids, Alabaster, AL). Cells were harvested after 24 h and the whole sphingolipid extract was hydrolyzed to get the free sphingoid bases as described previously with some modifications (7,8). The cell pellet was dissolved in 100 µL of PBS. 500 µL methanol including D 7 -sphingosine and D 7 -sphinganine (Avanti Polar Lipids, Alabaster, AL) as the internal standards were added. Lipids were extracted for one hour under constant agitation at 37°C. Samples were centrifuged to pellet precipitated proteins and the supernatant transferred into a new tube. Lipids were hydrolyzed by adding 75 μL of methanolic HCl (1 N HCl and 10 M H 2 O in methanol) and incubated for 16 h at 65°C. HCl was neutralized by adding 100 μL of KOH (10M). Then 625μL chloroform was added followed by 100 μL 2N ammonium hydroxide and 0.5 mL alkaline water to complete phase separation. The sample was vortexed, centrifuged at 16,000 g for 5 min, the upper phase discarded and the lower (organic) phase washed 3 times with alkaline water. The organic phase was finally dried under N 2 and stored at −20 o C until analysis.
Throughout this paper we refer to the extracted 1-deoxySO as native.

LC-MS Method
A commercial 1-deoxySO standard (SPH m18:1(4E)(3OH)) was purchased from Avanti Polar Lipids Inc. (Alabaster, AL). SPH m18:1(14Z)(3OH) was synthesized according to the method described below. The SPH m18:1(E)(3OH) standards (5E, 8E, 12E, 13E, 14E) were synthesized based on an unpublished method that will be issued elsewhere i . An LC-MS method described previously (8), was used to compare retention times and in source fragmentation. Sphingoid bases were separated by RPLC on a C18-column (Uptispere 120 Å, 5 μm, 125 × 2 mm; Interchim, Montluçon, France) and i Manuscript in preparation analyzed on a TSQ Quantum Ultra or a Q Exactive (Thermo, Reinach, BL, Switzerland) using an atmospheric pressure chemical ionization (APCI) interface. Mobile phases consisted of A: ultra-pure H 2 O / MeOH 1/1 v/v with 2.6mM ammonium acetate and B: MeOH. Gradient was set from 50% B to 100% B within 25 min followed by 5 min 100% B and 5 min of equilibration with a flow rate of 0.3 ml/min. For mass spectral detection, the following parameters were set on the APCI source: discharge current of 4 µA, vaporizer temperature of 450°C, sheath gas pressure 20 AU, aux gas 5 AU and capillary temperature of 200°C. For detection, the following parameters were set on the ESI source: spray voltage of 4.2 kV, vaporizer temperature of 30°C, sheath gas pressure 5 AU, aux gas 0 AU, capillary temperature of 320°C and for fragmentation in-source collision-induced dissociation (CID) was performed at 30 eV and higherenergy collision induced dissociation (HCD) of the ion at m/z 378.3 was performed with setting of 25.

Differential Mobility Spectrometry combined with Ozone-Induced Dissociation
A SelexION TM differential mobility spectrometer (DMS) was employed with a QTRAP ® 5500 triple quadrupole ion-trap mass spectrometer (SCIEX, Ontario, Canada). The instrument has been modified for ozone-induced dissociation (OzID) as previously described (10). Solutions for analysis were Sample solutions were subjected to ESI in a TurboV TM (SCIEX) and passed through the mobility spectrometer. Compensation voltage (CV) applied across the electrodes was scanned while the separation voltage (SV) was held constant at 4100 V. Ionograms were obtained from scanning CV and represent the sum of five mass spectra at each voltage point and have been smoothed using PeakView ® (SCIEX). The ESI source and DMS cell temperatures were 100 and 150°C, respectively; the ESI voltage was set to 5500 V. Nitrogen was set to 20 psi for each of the following: resolving gas in the DMS cell, the nebulizing ESI gas, and the MS curtain gas. Ions exiting the DMS cell into the mass spectrometer were mass selected in first quadrupole prior to isolation in the collision cell with ozone present. Ozone was produced by an external generator (Titan, Absolute Ozone, Alberta, Canada) operating at 220 g/Nm 3 (10.3% v/v ozone in oxygen) from which a small portion was mixed into the nitrogen collision gas input to the mass spectrometer through a variable leak valve (Nenion, Lustenau, Austria). The isolation time for ionized lipids in the collision cell was optimized between 100 ms and 15 s depending on the ozone-reactivity of a given ionized lipid (11,12). Following ozonolysis, ions were transferred to the third quadrupole region where mass analysis was performed using a trap-scan at 1000 Th/s. OzID spectra obtained at discrete CV values were averaged between 2 to 5 mins.

3-ol)
Eleven intermediate compounds en route to the target SPH m18:1 (14Z)(3OH) were synthesized as described in the supplementary info. A stirred solution of compound 11 (57 mg, 3 mmol) in 1,4dioxane (1 mL) at 0°C was treated with a solution of 4M HCl-dioxane (2 mL) over a period of 10 min.
The resulting reaction mixture was allowed to stir for 1 h at the same conditions, gradually warmed to ambient temperature, and followed by TLC analysis. After being stirred for additional 2 h (as monitored by TLC, Pet. ether/EtOAc 4:1; R f(adduct) = 0.5; R f(product) =0.0; visualized with KMnO 4 solution), the reaction mixture was concentrated under reduced pressure. The resultant residue was dissolved in with CH 2 Cl 2 (50 mL) and sequentially washed with saturated NaHCO 3 solution (40 mL), water (30 mL), and brine solution (40 mL). The organic layer was subsequently dried over anhydrous Na 2 SO 4 , filtered and concentrated in vacuo to afford a pale yellow oily residue. Flash column chromatography of the obtained crude amine over silica gel using ethylacetate and isopropanol as eluents (from 0-10% isopropanol in ethylacetate) provided the final compound 12 as colorless oil.

Differential-mobility spectrometry (DMS) combined with Ozone-Induced Dissociation
DMS has previously been deployed successfully for lipid isomer separation prior to mass spectrometric analysis (10,13). Here it was combined with OzID which exploits the reaction between mass-selected lipid ions and gaseous ozone inside a mass spectrometer to drive fragmentation diagnostic of the position(s) of carbon-carbon double bonds (12,14). In these experiments the D 3labeled hydrolyzed extract from HEK cells was spiked with a 1:1 mixture of the unlabeled commercial SPH m18:1(4E)(3OH) and synthesized SPH m18:1(14E)(3OH) structural isomers and the mixture subjected to electrospray ionization. Figure Table 1). Of this pair of ions, the aldehyde is mass-shifted and appears as expected at m/z 107 for the D 3 -isotopologue while the m/z 123 is absent in this spectrum: possibly due to the lower abundance of the CV = 20 V feature in the native extract. Interestingly, an abundant water loss was observed in the spectra in Figure 3C, which is much diminished in the spectra in Figure 3B providing further evidence of the structural difference between the two lipids.
DMS-OzID analysis was also performed on non-hydrolyzed lipid extract from HEK cells.  Table 1for details). Thus, tert-butyllithium was added to 1-pentyne at -78°C, and the in situ generated 1-lithio-1-pentyne was sequentially reacted with DMPU and bromo substrate 4 to afford compound 5 in a satisfactory yield. Subsequent stereoselective hydrogenation of alkyne 5 was accomplished with Lindlar catalyst after optimization of reaction conditions to afford the desired Zconfiguration of the alkene 6 as a sole isomer (see supporting information Table 2 for more details). It is noteworthy, that initial attempts to furnish alkene 6 led to a product that was contaminated (10-15%) with inseparable by-products. Careful analysis of 1 H-NMR spectra of the crude mixture identified these by-products as the E-configured alkene and the over reduced-product 1-bromopentadecane.
Cleavage of the THP-protecting group in 6 under acid-catalyzed ethanol treatment proceeded smoothly to provide the hydroxyl alkene 7 which was subsequently subjected to Appel reaction with CBr 4 /PPh 3 to furnish the Z-bromo-alkene 8.
Next, the obtained bromo-alkenyl chain 8 was converted into the corresponding Grignard reagent 9 and subsequently coupled to the Weinreb amide 2 (prepared from L-alanine in two steps, see the supporting information for details). Toward this end, Weinreb amide 2 was reacted with 0.9 equivalent of methylmagnesium bromide (as sacrificial base), followed by addition the Grignard reagent 9 to afford ketone 10 in 78% isolated yield. The subsequent diastereoselective reduction of the carbonyl group in 10 with lithium tri-(tert-butoxy)-aluminum hydride (TBLAH) (1.7 equiv.) in absolut ethanol at -78°C was accomplished without reduction of the double bond to furnish the desired anti-amino alcohol 11 as a mere stereoisomer. Acidic hydrolysis of the tert-butyl carbamate protecting group of 11 with 4M HCl-dioxane solution yieled the corresponding hydrochloric salt of SPH m18:1(14Z)(3OH) (.HCl). The desired product 12 was finally obtained after neutralization workup in 71% yield.

Retention-time correlation of native lipids with synthetic standards
The RPLC retention times of the synthetic SPH m18:1(E)(3OH) standards with double bonds in positions 5E, 8E, 12E and14E, were compared. We observed an inverse logarithmic correlation (R 2 =0.96) between retention times and the double bond position ( Figure 5A). The more the DB was positioned towards the omega end the earlier the molecule eluted from the column. The closest match in retention time between native 1-deoxySO and the synthetic standards was seen again for SPH m18:1(14E)(3OH) although the elution time between the two compounds still differed by about 30 seconds ( Figure 5B).
To further elucidate the DB configuration, we compared the retention times of native 1-deoxySO with the synthetic standard in (14Z) configuration ( Figure 5B and supplementary Figure 3). The SPH m18:1(14Z) eluted at 13.0 min and was therefore coinciding with the elution time of native 1-doxySO. From these results we conclude that native 1-deoxySO bears a (14Z) double bond.

Discussion
1-deoxySO is an atypical sphingolipid, lacking the 1-hydroxylgroup of canonical sphingosine. It is a downstream metabolite of 1-deoxySA which is formed by SPT due to its alternative activity with alanine. Because of the missing C1-hydroxyl group, 1-deoxySO cannot be phosphorylated to sphingosine 1-phosphate and therefore also not degraded by S1P-lyase (2).
Pathologically elevated 1-deoxySL play an important role in the inherited neuropathy HSAN I but were also found in other conditions like the metabolic syndrome and type 2 diabetes (3, 7). However, the metabolism of 1-deoxySLs and their molecular structures have not been investigated in detail yet.
Comparing native 1-deoxySO to a synthetic SPH m18:1(4E) standard by RPLC we observed a In conclusion we provide strong evidences from multiple, orthogonal analytical techniques to unambiguously assign native 1-deoxySO as it is formed in HEK293 cells to be a SPH m18:1(14Z)(3OH) structure. This furthermore implies that 1-deoxy-ceramides are metabolized distinct from canonical sphingolipids. However, more detailed studies are needed to further investigate the enzymatic pathways and metabolic steps involved in the conversion of these lipids.