Nonhydroxylated 1-O-acylceramides in vernix caseosa[S]

Vernix caseosa, the waxy substance that coats the skin of newborn babies, has an extremely complex lipid composition. We have explored these lipids and identified nonhydroxylated 1-O-acylceramides (1-O-ENSs) as a new class of lipids in vernix caseosa. These ceramides mostly contain saturated C11–C38 ester-linked (1-O) acyls, saturated C12–C39 amide-linked acyls, and C16–C24 sphingoid bases. Because their fatty acyl chains are frequently branched, numerous molecular species were separable and detectable by HPLC/MS: we found more than 2,300 molecular species, 972 of which were structurally characterized. The most abundant 1-O-ENSs contained straight-chain and branched fatty acyls with 20, 22, 24, or 26 carbons in the 1-O position, 24 or 26 carbons in the N position, and sphingosine. The 1-O-ENSs were isolated using multistep TLC and HPLC and they accounted for 1% of the total lipid extract. The molecular species of 1-O-ENSs were separated on a C18 HPLC column using an acetonitrile/propan-2-ol gradient and detected by APCI-MS, and the structures were elucidated by high-resolution and tandem MS. Medium-polarity 1-O-ENSs likely contribute to the cohesiveness and to the waterproofing and moisturizing properties of vernix caseosa.

Human skin is the largest complex organ in the human body with mainly protective function. The outermost layer of the skin is the stratum corneum, consisting of corneocytes filled with keratin filaments and surrounded by a continuous lipid matrix. The extracellular lipids form a being paid to polar lipids, mostly represented by ceramides (12,17,18).
In this work, we report on nonhydroxylated 1-O-acylceramides (1-O-ENSs) discovered during a search for unknown lipids in vernix caseosa. The 1-O-ENSs were isolated using adsorption chromatography on silica gel in several steps and their general structure was established using high-resolution and tandem MS, derivatization, and NMR. The molecular species within this class were comprehensively characterized using nonaqueous reversed-phase (NARP)-HPLC with APCI-MS.

Materials
Vernix caseosa (1-2 g) was collected from healthy newborn subjects delivered at full term (gestation weeks [39][40][41][42] immediately after the delivery. The samples were stored in amber glass vials at 80°C. The study was approved by the Ethics Committee of the General University Hospital, Prague (910/09 S-IV) and the samples were collected with written informed parental consent.

The isolation of total lipids
The isolation of total lipids was described earlier (16). Briefly, the lipids were extracted with methanol:chloroform (2:1 by volume) from 20 samples equally representing both genders. Each sample (300 mg) was processed separately. The chloroform extracts were dried over anhydrous magnesium sulfate, filtered through precleaned cotton wool, and combined. The solvent was removed using a rotary evaporator and a stream of nitrogen. In total, 6,000 g of vernix caseosa yielded 561.1 mg of total lipids. The lipids were reconstituted in chloroform:methanol (19:1 by volume) at the concentration of 30 mg/ml and stored at 20°C.

The fractionation of lipids
Approximately one-half of the total lipid extract was fractionated in three steps using semi-preparative TLC and HPLC. In the first step, lipids were separated on glass plates (60 × 76 mm) coated with silica gel (60 G for TLC; Merck Darmstadt, Germany) using the mobile phase of hexane:diethyl ether (93:7 by volume). The zones were visualized under UV light after spraying with rhodamine 6G (0.05% in ethanol). Silica gel with polar lipids (R F = 0-0.14) was scraped off the plates and the lipids were extracted with 10 ml of freshly distilled diethyl ether. The solvent was evaporated under a nitrogen stream. The procedure was used repeatedly (with approximately 3 mg of lipids separated in each step) and yielded 56.1 mg of polar lipids (F-1). In the second step, F-1 was reconstituted in chloroform:methanol (2:1 by volume) at a concentration of 30 mg/ml and separated on glass plates (90 × 120 mm) coated with silica gel. Chloroform:methanol (85:15 by volume) with 0.1% acetic acid was used as a mobile phase and the zones were visualized under UV light after spraying with 0.05% primulin in methanol:water (2:1 by volume). The silica-gel layer corresponding to R F = 0.86-0.96 was collected and extracted with chloroform:methanol (2:1 by volume). The procedure was used repeatedly to process 45.1 mg of F-1 (approximately 3 mg separated in each step) and yielded 20.0 mg of the fraction, F-2. The F-2 was dissolved in chloroform:methanol (2:1, v/v) at the concentration of 25 mg/ml and stored at 20°C until needed. In the final step, F-2 was separated on two Spherisorb HPLC columns connected in series (250 + 250 × 4.6 mm, 5 m particles) with a silica guard cartridge (4.6 × 10 mm), all from Waters, Milford, MA. The mobile phase flow (1 ml/min) delivered by a Rheos 2200 quaternary gradient pump (Flux Instruments, Reinach, Switzerland) was split postcolumn in a T-piece. A portion of the column effluent (100 l/min) was introduced into the MS detector, whereas the main portion of the flow (900 l/min) was directed via a switching divert valve either into a glass flask to collect the lipids (t R = 35-75 min) or into waste. The column was maintained at 30°C and the gradient was programmed from hexane (A) and hexane/propan-2-ol (96:4 by volume) (B) as follows: 0 min, 20% of B; 30 min, 29% of B; 60 min, 47% of B. The sample (10 l) was injected with an Accela autosampler (Thermo Fisher Scientific, San Jose, CA). The chromatographic separation was monitored using an LCQ Fleet spectrometer equipped with an APCI probe (Thermo Fisher Scientific). The APCI vaporizer and heated capillary temperatures were set to 270°C and 170°C, respectively. Nitrogen served as both the sheath and auxiliary gas at a flow rate of 15 and 17 arbitrary units, respectively. The MS spectra of the positively charged ions were recorded in the range of m/z 250-2,000. The whole HPLC/MS system was controlled by Xcalibur software (Thermo Fisher Scientific). The collected mobile phase with lipids was evaporated using a rotary evaporator (30°C, 170 mbar). The HPLC separation was repeated 22 times, providing 1.1 mg of F-3, later identified as 1-O-ENSs. It corresponded to approximately 1.0% of 1-O-ENSs in the total lipids of vernix caseosa. The whole isolation and fractionation procedure is depicted in Fig. 1.

Transesterification
Two transesterification procedures were used to investigate the structure of lipids in F-3: the first one for ester-linked FAs (50) and the second one for both ester-and amide-linked FAs (17). As regards the first method (50), a sealed glass ampoule with the sample (0.1 mg), solvent (chloroform:methanol, 2:3, v/v; 250 l), and acetyl chloride (30 l) was placed in a water bath at 70°C. After 60 min, the ampoule was opened and the reaction was terminated by adding silver carbonate (90 mg). In the second transesterification method (17), a sealed glass ampoule with the sample (0.1 mg) and a methanolic solution of HCl (1 mol/l; 100 l) was heated in a water bath at 75°C for 18 h. The reaction products were extracted with hexane.

GC/EI-MS and direct-probe EI-MS
FA methyl esters (FAMEs) were analyzed using a 6890N GC system (Agilent, Santa Clara, CA) coupled to a 5975B quadrupole mass spectrometer and equipped with an HP-5ms column (30 m × 250 m, 0.25 m; Agilent). The carrier gas was helium at 1 ml/min. The injector was held at 280°C and operated in the splitless mode (2 l of sample injected). The temperature program was as follows: 140°C (0 min), then 4°C/min to 320°C (10 min). The 70 eV EI mass spectra were recorded in the range of m/z 25-600; a solvent delay of 3.5 min was used. The temperatures of the transfer line, ion source, and quadrupole were 280°C, 230°C, and 150°C, respectively. For the EI-MS analysis without chromatographic separation, the sample was loaded on a direct insertion probe and inserted into the EI source of GCT Premier (Waters). The 70 eV EI mass spectra were recorded in the range of m/z 25-600.

Direct infusion MS and HPLC/APCI-MS
The structure of intact lipids in F-3 was also elucidated by APCI-MS performed on an LTQ Orbitrap XL hybrid FT mass spectrometer having either an APCI or ESI probe installed (Thermo Fisher Scientific). The spectrometer was coupled to an HPLC system consisting of a Rheos 2200 quaternary gradient pump (Flux Instruments) and a PAL HTS autosampler (CTC Analytics, Zwingen, Switzerland). The system was controlled by Xcalibur software (Thermo Fisher Scientific).
The sample for direct-infusion APCI-MS was dissolved in chloroform (250 g/ml) and delivered by a syringe pump (3 l/min) to a low-dead-volume T-piece, where it was mixed with acetonitrile, flowing at 200 l/min into the APCI source. The vaporizer and capillary temperatures were set to 320°C and 180°C, respectively, and nitrogen was used as the sheath and auxiliary gas at a flow rate of 65 and 10 arbitrary units, respectively. The spectra of positively charged ions were recorded. As regards direct ESI-MS in the negative ion mode, the transesterification products were infused into the ion source at the flow rate of 3 l/min. The ion source, capillary, and tube lens voltages were set to 4 kV, 16 V, and 120 V, respectively, and the capillary temperature was 275°C. The sheath gas and auxiliary gas were operated at 50 and 10 arbitrary units, respectively.
The molecular species of 1-O-ENSs in F-3 were separated using a Nova-Pak C18 stainless-steel column (300 × 3.9 mm, particle size 4 m; Waters) at 30°C. The autosampler injected 10 l of the sample and the injection system was washed with chloroform: acetonitrile  . The peak areas were integrated and expressed in relative values. In the second step, the relative peak areas were divided using the relative proportions of FA loss fragments in the MS 2 step and the relative proportions of sphingoid base loss fragments (i.e., the loss of alkadiene) in the MS 3 step.

NMR
The NMR spectra of samples in CDCl 3 were measured on a Bruker AVANCE-III-HD 600 instrument ( 1 H at 600.13 MHz and 13 C at 150.9 MHz) equipped with a cryoprobe at 25°C. The spectra were referenced to the TMS ( 1 H) or solvent peak [ 13 C, using  C (CDCl 3 ) = 77.0 ppm]. The partial structural assignment of proton and carbon signals was achieved through a combination of 1D-1 H and 1D-APT-13 C-spectra with homonuclear 2D-H,H-COSY, heteronuclear 2D-H,C-HSQC, and 2D-H,C-HMBC spectra. Fig. 1. The scheme of the isolation procedure.

General structure elucidation
The mass spectra of unknown lipids infused into the APCI source showed protonated molecules accompanied by more abundant water-loss fragments in the range of m/z 750-1,200 (Fig. 2). The exact masses of the protonated molecules were consistent with the elemental compositions C n H 2n-x O 4 N + (x = 0, 2, 4, or 6). The presence of nitrogen pointed to structures related to ceramides. In the next step, the lipids were transesterified using conditions strong enough to cleave both ester-and amide-linked FAs. Infusion experiments using an ESI source operated in the negative ion mode revealed signals consistent with sphinganine (chloride adduct at m/z 336.3), sphingosine (chloride adduct at m/z 334.3), and other sphingoid bases (Fig. 3A). High-temperature GC/MS and direct-probe EI-MS showed a number of mostly saturated FAMEs with 8-32 carbons having straight and methyl-branched chains, but no hydroxyl group (supplemental Fig. S1a; supplemental Table  S1). The masses of the intact lipids were too high to be explained by a sphingoid base linked to one FA. Therefore, two fatty acyls in the structures were assumed. The absence of hydroxy FAs led us to a structure with the two fatty acyls linked directly to the sphingoid base by means of one amide and one ester bond. To confirm the hypothesis, transesterification was performed once again, but this time using milder conditions preserving amide bonds. Products having a sphingoid base linked to a fatty acyl by an amide bond were sought in the reaction mixture. Indeed, APCI-MS showed ions corresponding to C n H 2n-x O 3 N + (x = 2, 4, 6, or 8) and their water-loss products consistent with Cers (NS) and Cers (NdS) (Fig. 3B). These Cers have free hydroxyl groups on C1 and C3 of the sphingoid chain. Although both of them can be esterified, the C1 hydroxy group was considered more likely to be involved in the ester bond formation. Many ceramides, e.g., sphingomyelins and glycosphingolipids, are functionalized on C1, whereas the C3 hydroxy group remains unmodified. Therefore, we hypothesized that the unknown lipids were, in fact, 1-O-ENSs (Fig. 4). To test the hypothesis, we compared the MS and NMR data of the lipids isolated from vernix caseosa with a commercial standard (17:0 ceramide-1-O-18:1). The mass spectra of the standard (Fig. 5) were consistent with the spectra recorded for the unknown lipids. NMR spectra were taken for the isolated lipid class, i.e., for a rich mixture of molecular species. Although no conclusions could be made regarding the side chains, the chemical shifts for the central part of the molecules observed in 1 H and 13 C NMR spectra were almost identical to those obtained from the standard (supplemental Fig. S2). Hence, the hypothesis was confirmed and the lipids isolated from vernix caseosa were identified as 1-O-ENSs. Later, it was confirmed  , and protonated amides corresponding to the amide-linked FA. In the case of the most common sphingoid base, sphingosine, the eliminated alkadiene corresponded to hexadecadiene. In addition, protonated doubly dehydrated sphingoid base ions (m/z 264.3 for sphingosine) were formed. These low-mass ions were not always detected because of the ion trap cut-off effect. The MS 2 spectra were utilized to identify ester-linked FAs (Fig. 5A). Further fragmentation of the ions formed by the elimination of FA (MS 3 spectra) led to the neutral loss of alkadiene ([M + H  H 2 O  FA  C n H 2n-2 ] + ) and protonated doubly dehydrated sphingoid base ions. Fragments of protonated amides corresponding to amide-linked FAs were detected as well. The MS 3 spectra made it possible to identify amide-linked FAs and sphingoid bases (Fig. 5B). The APCI mass spectra closely resembled the ESI spectra of 1-O-ENSs published previously (52).

The 1-O-ENSs in vernix caseosa
The NARP-HPLC data revealed an enormous complexity of 1-O-ENSs in vernix caseosa. The chromatogram contained many coeluting peaks (Fig. 6A), which could be partially resolved using reconstructed chromatograms. For instance, reconstructed chromatograms for selected monounsaturated 1-O-ENSs provided several broad peaks. Interestingly, the pattern was very similar for monounsaturated species differing in the total CN (Fig. 6B-D). The relatively large peak widths could be explained by the coelution of species with the same total CN, but various lengths of   individual chains or positions of the double bonds. The chromatographic separation of 1-O-ENSs with the same total CN was most likely caused by methyl branching of the aliphatic chains. We had already shown that the methyl branching of lipids significantly decreased their retention in RP-HPLC (53). The MS 2 spectra taken in the separated peaks looked similarly, indicating almost the same distribution of the fatty acyl chain length (supplemental Fig. S3). Hence, various degrees of methyl branching seemed to be a reasonable explanation for the peak separations. The peak patterns looked differently in the reconstructed chromatograms for 1-O-ENSs with a different DB; nevertheless, similarly to monounsaturated 1-O-ENSs, the profiles were alike within each group (supplemental Fig. S4). The complexity of 1-O-ENSs in vernix caseosa can be further illustrated using chromatograms reconstructed for the neutral loss of a specific ester-linked FA observed in MS 2 . As for the monounsaturated 1-O-ENSs, the profiles consisted of dozens of peaks and looked almost the same regardless of the FA (Fig. 7A-C). The reconstructed chromatograms, however, changed dramatically when the neutral loss of the same FA was plotted for 1-O-ENSs with a different total DB (Fig. 7C-E).
Representative mass spectra of 1-O-ENSs from vernix caseosa are given in   reversed-phase chromatography, the retention of lipids depends on the number of carbon atoms and double bonds in acyl chains. Therefore, the retention times of lipids are usually in good agreement with their ECN values calculated as the total CN minus twice the DB, i.e., ECN = CN  2DB (14,16,54,55). This is also the case of 1-O-ENSs, which makes the validation possible (Fig. 9).
The most abundant 1-O-ENSs identified in vernix caseosa are summarized in Table 1. The list of 30 molcular species represents almost one-third of the total intensity of lipids within this class. For a complete list of 1-O-ENSs see supplemental Table S2. Ester-linked fatty acyls were mostly saturated; they consisted of 11-38 carbons, where evennumber chains predominated (Fig. 10A). Although amidelinked fatty acyls spanned a similar range of chain lengths (12-39 carbons), their profile was significantly different (Fig. 10B). The amide-linked fatty acyls were almost exclusively saturated, dominated by chains with 24 and 26 carbons. The fatty acyl profiles calculated from LC/MS data were in good agreement with the GC/MS data recorded for transesterified samples (supplemental Fig. S1). Some additional peaks in GC/MS traces (particularly FAME 16:0 and FAME 18:0 in supplemental Fig. S1a) indicated minor impurities of other lipids in the sample (fraction F-3). Concerning sphingoid bases, they appeared to contain 16-24 carbons and up to two double bonds. Sphingosine (d18:1) was the most abundant, but sphinganine (d18:0) and d19:1 were also detected in significant amounts (Fig. 10C). The most abundant 1-O-ENS in vernix caseosa proved to be 24:0;d18:1;26:0, which formed 2.3% of these lipids.

DISCUSSION
Ceramides are essential lipids for the skin barrier function, protecting the body against desiccation and microbial infections. Together with FFAs and cholesterol, they form lamellar membranous structures surrounding corneocytes in the stratum corneum. They are also important in vernix caseosa, where barrier lipids typically do not form lamellar structures. The level of barrier lipids is significantly lower than in the stratum corneum, and the dominant components of vernix caseosa are nonpolar lipids. Such composition ensures the protection of the fetal skin from direct contact with the amniotic fluid and facilitates the maturation of epidermis in utero (12,56).
The very first mention of 1-O-ENSs comes from 1977, when they were tentatively identified in rat brain after the injection of precursor ceramides (58). As regards endogenously expressed lipids, 1-O-ENSs together with 1-O-EASs were identified in humans and mice in 2013 as new classes of epidermal ceramides (31). Their masses were in the range of 750-1,000 Da, and they contained sphingosine and long to very long acyl chains with an even CN. They were found to comprise 5% of all esterified ceramides in the epidermis of mice. In humans, their total amount accounted for approximately 30% of the corresponding unmodified parent ceramides (NS). The most abundant 1-O-ENSs contained 24:0 or 16:0 chains in ester-linked FAs, whereas amide-linked acyls mostly consisted of hydroxy-16:0 or 16:0 chains. The biosynthesis of these ceramides was suggested to take place at endoplasmic reticulum-related sites and involve acylation by DGAT2 (31). Mouse epidermal 1-O-ENSs have recently been investigated using direct infusion MS (52). Approximately 710 molcular species composed mainly of saturated long-chain acyls and sphingosine were found. The high number of isobaric species illustrated the complexity of this lipid class. The double bond in unsaturated 1-O-acyl and N-acyl chains appeared to be located at the position n-9. The 1-O-ENSs analyzed in this work were entirely nonhydroxylated species (1-O-ENSs). They were isolated using adsorption chromatography on silica gel, which ensured efficient separation from other lipids of different polarity, including 1-O-EASs, if they were present. In total, 1.1 mg of 1-O-ENSs were isolated, which corresponds to approximately 1% of lipids in vernix caseosa. As ceramides comprise about 50% of stratum corneum lipids and 1-O-ENSs form <2.0% of total ceramides (23), we can conclude that the content of 1-O-ENSs in the lipids of vernix caseosa and stratum corneum is approximately the same.
Like mouse and human epidermal 1-O-acylceramides (31, 52), 1-O-ENSs from vernix caseosa were found to be a highly complex mixture of molecular species. The complexity stems from the high variability of fatty acyls differing in their chain length, methyl branching, and DB. Although 1-O-ENSs exhibit a preference for certain acyls (Fig. 10A, B), the overall incorporation of fatty chains from these "pools" is highly random. Consequently, there are at least 2,300 molcular species of 1-O-ENSs in vernix caseosa, but the real number is very likely even higher. When compared with previous works (31,52), the higher number of the molecular species identified in this work can be attributed to the analytical method, i.e., MS coupled to thoroughly optimized chromatography. HPLC can efficiently distinguish isobaric 1-O-ENSs differing in chain branching, which is its principal advantage over direct-infusion MSbased approaches. Methyl-branched fatty acyls are known to occur in vernix caseosa ceramides, particularly in acylceramides (17). Methyl-branched 1-O-ENSs have been found to exist in vernix caseosa, which is obvious both from HPLC and GC data. The GC/MS of the transesterified 1-O-ENS sample revealed three series of FAMEs. Based on their retention behavior (59), we speculate that they corresponded to iso-, anteiso-, and straight-chain esters (supplemental Table S1). Odd-carbon FAMEs were of all three types (iso-, anteiso-, and straight-chain), whereas even-carbon FAMEs lacked anteiso-isomers, likely because anteiso-chains are biosynthesized from the odd-carbon precursor, isoleucine (60). Iso-FAMEs predominated in odd-carbon series, where the most abundant even-carbon FAMEs were of the anteiso type. A very similar profile of branched fatty acyl chains was observed previously in acylated ceramides (17). Besides the methyl branching, fatty acyls in vernix caseosa 1-O-ENSs are characterized by considerable chain lengths. Saturated fatty acyls contained up to almost 40 carbons (38 in esterlinked, 39 in amide-linked), and the most typical chains consisted of 24 carbon atoms. Unsaturated fatty acyls were not common; they were virtually absent from amide-linked chains (about 0.1% of unsaturated chains) and formed approximately 10% of O-acyls. The low content of unsaturated chains is favorable for the resistance to oxidative damage on exposure to air after delivery. Saturated and monounsaturated very-long-chain fatty acyls are commonly found in epidermal ceramides, where they contribute to the barrier function of the skin (23,30). Odd-carbon chains formed 20% of N-linked acyls and 24% of O-linked acyls in vernix. Sphingosine (d18:1) was by far the most abundant sphingoid base in the 1-O-ENSs of vernix caseosa, like in human and mouse epidermis (31,52). In addition, the research revealed sphinganine (d18:0), five sphingosine homologs, seven sphinganine homologs, and two sphingadienines, which are known constituents of ceramides (26). A direct comparison of 1-O-ENS molecular species from vernix caseosa, human skin, and mouse skin is difficult because of the different analytical methodology used. However, it seems that vernix caseosa and epidermal samples share numerous similarities. For instance, the major structures of the most abundant 1-O-ENSs in mouse epidermis are reported to be 24:0;d18:1;24:0 and 24:0;d18:1;26:0 (52), which are species richly represented in vernix caseosa as well. The overall distribution of 1-O-ENS masses is also similar for the discussed samples (31,52). As discussed above, HPLC/APCI-MS 2 has made it possible to disclose the fascinating diversity of the 1-O-ENSs of vernix caseosa. There are various combinations of straight and branched chains, which for given chain lengths results in two or more 1-O-ENS isomers. A significant portion of 1-O-ENS molecular species thus exists in several forms differing only in chain branching. To illustrate this fact, fully characterized 1-O-ENSs listed in supplemental Table S2 were searched for duplicities (lipids with the same abbreviation). Only 594 out of 972 records were unique, which means that 39% of all identified 1-O-ENSs were branched variants. The use of APCI for MS detection was found to be convenient for the 1-O-ENSs. Because the initial water loss already takes place during ionization, only two fragmentation steps (MS 2 and MS 3 ) are required for complete structural characterization, without the need for instruments with MS 4 capabilities. As far as lipid quantification is concerned, it is important to note that the response factors of individual molecular species depend to some extent on the nature of aliphatic chains. The response is mainly affected by the overall DB: the larger their number, the higher the detection sensitivity (53). Consequently, the quantitative data reported in this work might be slightly biased.
The biological function of 1-O-ENSs in vernix caseosa has yet to be clarified. The moderate polarity and the free hydroxyl group of these ceramides might contribute to the cohesiveness of vernix caseosa composed of nonpolar sebaceous lipids and polar barrier lipids. In this way, 1-O-ENSs could contribute to the waterproofing and moisturizing properties of vernix caseosa.