1-O-acylceramides are natural components of human and mouse epidermis.

The lipid-rich stratum corneum functions as a barrier against pathogens and desiccation inter alia by an unbroken meshwork of extracellular lipid lamellae. These lamellae are composed of cholesterol, fatty acids, and ceramides (Cers) in an equimolar ratio. The huge class of skin Cers consists of three groups: group I, "classical" long and very long chain Cers; group II, ultra-long chain Cers; and group III, ω-esterified ultra-long chain Cers, which are esterified either with linoleic acid or with cornified envelope proteins and are required for the water permeability barrier. Here, we describe 1-O-acylceramides as a new class of epidermal Cers in humans and mice. These Cers contain, in both the N- and 1-O-position, long to very long acyl chains. They derive from the group I of classical Cers and make up 5% of all esterified Cers. Considering their chemical structure and hydrophobicity, we presume 1-O-acylceramides to contribute to the water barrier homeostasis. Biosynthesis of 1-O-acylceramides is not dependent on lysosomal phospholipase A2. However, glucosylceramide synthase deficiency was followed by a 7-fold increase of 1-O-acylceramides, which then contributed 30% to all esterified Cers. Furthermore, loss of neutral glucosylceramidase resulted in decreased levels of a 1-O-acylceramide subgroup. Therefore, we propose 1-O-acylceramides to be synthesized at endoplasmic reticulum-related sites.

nGlcCerase-defi cient mice lack the neutral glucosylceramidase, which may lead to increased nonlysosomal GlcCer levels, and by that may potentially lead to decreased Cer levels as substrate for 1-O-acylation. LPLA 2 -defi cient mice lack LPLA 2 , which was reported to have 1-O-acylating activity on short chain Cers. Hence, it could have been a candidate for 1-O-acylation of Cers in epidermis. Lipid extracts from nGlcCerase-and LPLA 2 -mutants were obtained from the mouse lines published in ( 20,21 ). Mice with epidermal defi ciency of UDP-glucose ceramide glucosyltransferase ( Ugcg ) lack GlcCer-S in epidermal keratinocytes. Epidermal lipid extracts of Ugcg -keratin 14 (K14) mice were from ( 22 ). CerS3 is essential for the synthesis of Cers containing N -linked ultra-long chain fatty acids, which are abundantly expressed in epidermis. Lipid extracts used were from CerS3-mice, recently published ( 10 ), and lack -esterifi ed Cers.
Samples were extracted with solvents of per analysis grade. For UPLC-ESI-MS/MS analysis, dried aliquots were taken up in solvents used for chromatography which were of HPLC grade. All additives used were also of HPLC grade.
The adhesive tape (tesa® fi lm transparent, #57371) used for lipid isolation of human stratum corneum was composed of polypropylene with an acrylic adhesive.

Preparation of epidermis and isolation of epidermal lipids
Epidermis was separated from dermis and raw extracts of epidermal lipids were prepared according to ( 10 ). In brief, interscapular skin samples were rapidly dissected and snap-frozen. To separate epidermis from dermis, samples were treated with 500 mg/ml thermolysin buffer (containing 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 142 mM NaCl, 6.7 mM KCl, 0.43 mM NaOH, and 1 mM CaCl 2 at pH 7.4) for 2 h at 37°C. Isolated epidermis was cut into small pieces and lyophilized. Roughly 3 mg of dry weight epidermis was extracted with 2 ml of chloroform/methanol/water (30/60/8) by volume at 50°C for 15 min under sonication. Supernatant was taken after centrifugation and pellets were reextracted two additional times using the solvent mixtures of chloroform/methanol/distilled water (10/10/1) and chloroform/methanol (2/1) by volume. Supernatants were pooled and desalted using reversed-phase (RP)-18 columns.
Internal standards were added to aliquots of the lipid extracts prior to being analyzed by UPLC-ESI-MS/MS. Aliquots of lipid extracts from Ugcg-K14 mice were taken from the samples prepared for the publication of ( 22 ).

Isolation of human stratum corneum lipids
Isolation of human stratum corneum was performed by tape stripping in healthy male volunteers with ages ranging from 26 to 45 years. The volunteers were asked not to apply any cosmetic skin products on their ventral forearms within 16 h prior to sample collection. The human studies were approved by the ethics committee of the Medical Faculty, University of Heidelberg, Germany and informed consent of participants was obtained.
Both ventral forearms of each volunteer were stripped in four consecutive sites using ordinary adhesive tape in an area of 2 × lipid lamellae surrounding corneocytes of the stratum corneum. Here, special Cer structures with amide-bond ultra-long 2 chain -hydroxylated fatty acids exist that are formed by ceramide synthase 3 (CerS3) ( 10 ) ( Fig. 1 ). This -hydroxylation is a prerequisite for further esterifi cation of Cers with long chain fatty acids, predominantly linoleic acid ( ⌬ 9,12 -18:2), or esterifi cation with proteins of the cornifi ed envelope, as reviewed in (11)(12)(13)(14)(15). Although esterifi ed Cers gain hydrophobicity due to their ultra-long chain fatty acid and esterifi cation in the -position, they retain their polar head group structure with two hydroxy groups (three hydroxy groups in phyto-compounds).
An alternative metabolic pathway for Cers was suggested fi rst in the late seventies, when tissue or cells were shown to esterify labeled Cers in 1-O-position of the sphingoid base. Injection of 3 H-and 14 C-labeled Cers (d18:1;16:0 and d18:1;24:0) 3 directly into rat brains led, among other products, to 1-O-acylceramides ( 16 ). Subsequently, a group XV lysosomal phospholipase A 2 (LPLA 2 ) was discovered with the unique ability to transacylate in vitro radioactive short chain Cers in 1-O-position ( 17 ), reviewed in ( 18 ). Recently, a pathway leading to 1-O-acylceramides was described in yeast. It involves the diacylglycerol acyltransferases phospholipid:diacylglycerol acyltransferase (Lro1p) and diacylglycerol O-acyltransferase (Dga1p), which in contrast to the LPLA 2 also esterifi ed very long chain Cers in 1-Oposition ( 19 ). Here, we report the identifi cation and structural characterization of natural 1-O-acylceramides in mouse and human epidermis. Using ultra performance liquid chromatography coupled-electrospray ionization-tandem mass spectrometry (UPLC-ESI-MS/MS), we identifi ed and quantifi ed a group of almost 100 species in mice. Combinatorial calculations suggest the presence of more than 200 subspecies in mouse epidermis. To our knowledge, this study is the fi rst description of endogenously expressed 1-O-acylceramides in mammals.

Materials
Epidermal lipid extracts of glucosidase beta 2 (Gba2, nGlcCerase)-, LPLA 2 -, glucosylceramide synthase (GlcCer-S)-, and CerS3-defi cient mice were used for mass spectrometric analyses. 2 Long chain fatty acids or Cers are defi ned here as those with an acyl moiety ranging from 16 to 20 carbon atoms, very long chain as those with 22 to 26 carbon atoms, and ultra-long chain as those with 28 or more carbon atoms. 3 Cers are structurally described according to their sphingoid base and fatty acid moieties. In brackets, fi rst is described the type of sphingoid base. The initial letter refers to the number of hydroxy groups (m, mono; d, di; t, tri; or te, tetra), the following number corresponds to the amount of carbons of the linear chain and the number after the colon corresponds to the degree of unsaturation. After the semicolon, the fatty acid in 2-N -position is described fi rst with the number of carbons in the acyl chain and after the colon the number of unsaturations. If the fatty acid is hydroxylated, an "h" is placed in front of the number of carbons or " " when hydroxylation occurs at -position. Specifi cally for 1-O-acylceramides we included the description of the fatty acid attached in 1-O-position (including number of carbons and hydroxylation) prior to the specifi cation of their sphingoid base. Protonated 1-O-acylceramides were quantifi ed in single reaction monitoring (SRM) mode using 7.65 pmol of 1-O-oleoyl Cer(d18:1;17:0) as internal standard per aliquot of epidermal lipids corresponding either to 0.1 mg mouse tissue dry weight or to 5.1 cm 2 of tape used for stripping human stratum corneum. The capillary voltage was set at 2.5 kV, whereas the cone and the source offset were fi xed at 50 V. The source and desolvation temperatures were maintained at 90 and 300°C, respectively. The desolvation gas was delivered at 800 l/h, while the cone gas and the collision gas fl ow were fi xed at 150 l/h and 0.15 ml/min, respectively. 1-Oacylceramides were detected using a fi xed collision energy of 20 eV, while maintaining the dwell time at 8 ms. Instead classical Cers (d18:1;14:0-36:0) were measured using increasing collision energies ranging from 25 to 30 eV according to their molecular mass.
Signifi cant in-source decay of 1-O-acylceramides leading to water loss appeared. Therefore, each compound was detected by two transitions, which were added up for quantifi cation: i ) transition of the protonated acylceramide to fragment C (loss of a water molecule and the O-linked fatty acid); and ii ) transition of the mono-dehydrated and protonated acylceramide to fragment C. The decision basis for the incorporation of individual acylceramides into the fi nal multiple reaction monitoring list (supplementary Table II) was based on precursor ion scan m/z +264 mode by which all potential acylceramides with a C18-sphingosine (d18:1) were detected.
Samples were injected and processed using MassLynx, whereas mass spectrometric peaks were quantifi ed according to their peak area ratio with respect to the internal standard using TargetLynx (both v 4.1 SCN 843) both from Waters Corporation.

Validation of the method: linearity assay and matrix effects evaluation
Linearity of the method was analyzed keeping the amount of internal standard 1 constant at 7.65 pmol while diluting several times by 1:2 the amount of standard 2, 1-O-palmitoyl Cer(d18:1;16:0). Linearity of the method was obtained down to 8.8 fmol of 1-O-palmitoyl Cer(d18:1;16:0) with 10% deviation from the regression curve (supplementary Fig. IA). In a second experiment, the amount of murine epidermal lipid extract was diluted against a constant concentration of 7.65 pmol of internal standard, and the series of 1-O-acylceramides with the very long chain Cer 1.5 cm. For each stripping, tapes were pressed with a weight of 4.8 kg for 30 s prior to being peeled off with tweezers in a single sharp stroke. Tape strips were individually placed in glass vials and were extracted with 3.5 ml chloroform/methanol/water (30/60/8) for 15 min at 37°C with a subsequent centrifugation step for 10 min at 3,500 g . An aliquot (3 ml) of each of the four extracts containing the lipids of the consecutive tape strips of one forearm of each volunteer were pooled and dried. Thus, duplicates of three volunteers were processed. Extracts were then redissolved in methanol/ water (95/5, 2 ml) and centrifuged for 15 min at 21,000 g . Internal standards were added to a fi nal aliquot of the extracts (1 ml) prior to be quantifi ed by UPLC-ESI-MS/MS. The standards used were as follows: for 1-O-acylceramides, Cer(18:1;d18:1;17:0); for classical Cers, Cer(d18:1;14:0), Cer(d18:1;19:0), Cer(d18:1;25:0), and Cer(d18:1;31:0). To exclude contaminations or artifacts from the adhesive tape, the latter was used as a blank and extracted in parallel with the human probes. Aliquots were analyzed by UPLC-ESI-MS/MS for 1-O-acylceramides, as well as classical nonhydroxyfatty acid and sphingosine containing ceramides (NS-Cers), with acyl chains from 16 to 36 carbon atoms and a C18-sphingosine base. The amount of 1-O-acylceramides in fmol was normalized to the total amount of NS-Cers in pmol.

Lipid analysis by UPLC-ESI-MS/MS
UPLC-ESI-(triple quadrupole)MS/MS analysis was performed on a Xevo TQ-S tandem mass spectrometer coupled to an automated Aquity I class UPLC system using an ACQUITY UPLC® BEH C18 1.7 m column (length 50 mm, diameter 2.1 mm) all from Waters Corporation. The column was equilibrated in buffer A (95% methanol, 0.05% formic acid, and 1 mM ammonium formate) and lipids were eluted with increasing percent of buffer B (99% 2-propanol, 1% methanol, 0.05% formic acid, and 1 mM ammonium formate) at a fl ow rate of 0.45 ml/min as described in ( 23 ) with slight modifi cations as follows: The gradient went up to 90% B as listed in supplementary Table I. The total gradient time was fi xed at 6.5 min per run. As previously, samples were dissolved in 95% methanol and 5% water. However, care was taken that samples were treated in an ultrasound bath for 3 min at 40°C and then placed directly into the auto sampler, which was held at 20°C. The column was heated to 40°C and in general 10 l (100% of injection needle capacity) aliquots were injected.  1. General Cer metabolism. Cers reside at the epicenter of the sphingolipid metabolism. They can be converted into ubiquitously expressed SMs and GSLs. However, the glycan pattern of GSLs found is cell type specifi c, e.g., galactosylceramide synthesis is mainly restricted to myelin sheaths, renal tubular epithelial cells, and pancreatic islets, but is not present in the epidermis. Cers are also precursors of the bioactive lipids ceramide-1-phosphate and, via sphingosines, of sphingosine-1-phosphate. In the epidermis, Cers with ultra-longhydroxy acyl chains are further -esterifi ed, mainly with linoleic acid. These -esterifi ed Cers are imperative for adequate skin barrier functions. Reported here, epidermal Cers with N -linked long and very long chain fatty acids (C16-C24) may also be esterifi ed with saturated long and very long chain fatty acids (C14-C26) in 1-O-position leading to 1-O-acylceramides.
-esterifi ed Cers ( 10 )], and in samples of GlcCer-synthasedefi cient (keratinocyte-specifi c) mice, which do not contain epidermal GlcCers ( 22 ). Hence, the unidentifi ed peaks neither belonged to GlcCers nor to Cers with ultra-long chain Cer anchors. The lipids underlying these signals were sensitive to alkaline treatment and appeared enriched in lipid extracts of GSL-free epidermis, i.e., from keratinocytespecifi c GlcCer-synthase-defi cient mice. Simultaneously, alkaline treatment increased the amount of long and very long chain (C16-C24) Cers in wild-type but especially in GSL-free epidermis. Therefore, we hypothesized the unknown lipids to be alkaline-sensitive derivatives of long and very long chain Cers, e.g., esterifi ed Cers. Product ion spectra from the main peaks indeed revealed fragmentation patterns that included a parent Cer backbone: all fragmentation spectra contained three dominant product ions due to: i ) loss of a water molecule, ii ) loss of a water molecule and lignoceric acid, and iii ) a residual dehydrated d18:1-sphingosine base ( Fig. 2 , product ions in red small letters a, c, and e).

Subgroups of O-acylceramides identifi ed by specifi c product ion and retention time
In a next step, we performed precursor ion screens of the epidermal lipid extract using the c-type fragments, anchor Cer(d18:1;24:0) was quantifi ed. The obtained curve demonstrates the amount of lipid extract corresponding to 0.1 mg tissue dry weight within the linear range (supplementary Fig. IB).
To evaluate matrix effects from the adhesive tape in human epidermal extracts, a dilution of 1-O-palmitoyl Cer(d18:1;16:0) with a fi xed amount of 7.65 pmol internal standard was evaluated in the absence and presence of blank tape extracts obtained according to the method described for the isolation of human stratum corneum lipids (supplementary Fig. IC).

Alkaline sensitive Cers of unknown composition in mouse epidermis
Screening total epidermal lipid extracts for Cers containing a C18-sphingosine base (d18:1) by MS/MS, we detected low abundant signals for compounds with an even protonated nominal mass in the range of m/z 750-1,000 (supplementary Fig. II). These signals did not fi t to any previously known epidermal Cers and GlcCers or ultra-long chain -esterifi ed Cers. Furthermore, these signals were present not only in samples of wild-type mice, but also in samples of CerS3-defi cient mice [lacking ultra-long chain and Fig. 2. Tandem mass spectrometric fi ngerprint of selected unknown Cers. CerS3-defi cient epidermal lipid extracts were injected on a RP-18 column and eluting compounds were monitored in product ion mode set to the corresponding m/z values of unknown Cers. Product ion spectra (0.2 s/spectrum) corresponding to the complete peak width (5-6 s, i.e., 25-30 spectra) of the eluting unknown Cer species (see Fig. 3 ) were summed up into each spectrum plotted here. In the product ion mode, molecular ions (precursor ions, blue m/z ) of unknown Cers were fragmented within the mass spectrometer by collision-induced dissociation to generate characteristic product ions (red small letters). Main fragments are annotated in bold letters and the corresponding acylceramide composition is written in larger letters on top. Alternative fragments and corresponding minor acylceramides coeluting within the same UPLC-peak are annotated in small letters with primes. Small red letters denote specifi c types of fragments described in the scheme of Fig. 4 . differentiated by the SRMs used ( Fig. 3B , horizontal line); e.g., 1-O-lignoceroyl-Cer(d18:1;16:0) ( Fig. 3A , series 16, peak I) and 1-O-palmitoyl Cer(d18:1;24:0) ( Fig. 3A , series 24, peak C) eluted simultaneously, but differed in their SRM by the transition to m/z 502 and 614, respectively.
We then compared the retention times of 1-O-acylceramides with those of linoleic acid esterifi ed -hydroxyc eramides (EOS-Cers). Whereas 1-O-acylated long chain Cers eluted earlier than EOS-Cers, 1-O-acylated very long chain Cers had a similar or even higher retention time than EOS-Cers on the RP-18 column. The additional free hydroxy group in 1-position and the two double bonds of the linoleic acid reduced the interaction with the RP-18 column so that EOS-Cers eluted much earlier than the corresponding 1-O-acylceramides with the same total amount of carbon atoms. which might correspond to doubly dehydrated Cer backbones. To exclude an involvement of known -esterifi ed ultra-long chain Cers, we used an extract of CerS3-defi cient mice, that lack Cers with fatty acids greater than lignoceric acid ( 10 ). For each c-type fragment scan, we identifi ed a series of lipids fi tting to the esterifi cation of the corresponding Cer backbone with long to very long chain saturated fatty acids (supplementary Fig. II).
Using the information of the precursor ion scans, we established a SRM-based RP-18 chromatography for a more specifi c and sensitive detection of 1-O-acylceramides ( Fig. 3 ). Using the transition of the molecular ions to their c-type fragment (see scheme of Fig. 4A ), which is specifi c for the parent Cer backbone, characteristic elution times of individual acylceramides were obtained that depended on the length of the acyl chain in ester linkage. Likewise, by keeping the 1-O-acyl chain constant the retention times increased with increasing size of the Cer backbone ( Fig. 3B vertical lines).
The characteristic retention time curves ( Fig. 3B ) allowed predictions with regard to the O-acyl chain length of 1-Oacylceramides. Interestingly, structurally isomeric 1-Oacylceramides eluted simultaneously and could only be for the 1-O-acyl chain were obtained. The O-acyl chain rather gave rise to a neutral loss ( Fig. 4A ). The product ion spectra of 1-O-palmitoyl Cer(d18:1;16:0) and the corresponding endogenous compound were not only identical at a collision energy of 20 eV ( Fig. 4A, B ), but also at 10, 30, and 45 eV with respect to the relative intensities of their fragment ions to each other ( Fig. 4B ). Furthermore, the synthesized compound eluted identical to the respective endogenous compound (t R = 3.70 min, Fig. 4C ; see also Fig. 3 , series 16). Thus, we concluded that the detected acylceramides of mouse epidermis represent 1-O-acylceramides.

Identifi cation as 1-O-acylceramides
To identify the epidermal acylceramides as 1-O-acylceramides, we compared the retention times and fragmentation patterns of the endogenous compounds with synthetic 1-O-acylceramide standards. Comparing the fragmentation behavior of the two standard compounds, 1-O-palmitoyl Cer(d18:1;16:0) and 1-O-oleoyl Cer(d18:1;17:0), we could attribute the individual product ions to the Cer backbone (fragments b and c, Fig. 4A ), the sphingoid base (fragment e, Fig. 4A ), and to ions containing the amide bond acyl chain (fragments d and f, Fig. 4A ). No specifi c product ions  Fig. 3 ) from epidermal lipid extract of CerS3defi cient mice. Average product ion spectra were recorded from the chromatographic peaks as described in the Fig. 2 legend. The scheme on the right illustrates the structure of 1-O-acylceramides and the product ions (derived by collision-induced dissociation) annotated with small red letters within the spectra. Letters with a prime mark fragments from other minor acylceramides of the same nominal mass coeluting with endogenous Cp.16-C, i.e., Cer(16:0;d18:1;16:0). B: Comparison of product ion peaks in standard 2 [Cer(16:0;d18:1;16:0)] and acylceramide Cp.16-C from CerS3-defi cient epidermis recorded at different collision energies. Chromatograms were recorded in product ion mode. Product ion spectra of the complete peak at 3.7 min (see Fig. 4C ) were summed up to produce an average spectrum for evaluation. Note the identical product ion ratios from both samples at all collision energies investigated for all fragments (82.0-758.7). C: Bottom spectrum: mixture of the synthetic standards, Cer(16:0;d18:1;16:0) and Cer(18:1;d18:0;17:0); middle spectrum: raw lipid extract from mouse epidermis; upper spectrum: mixture of epidermal sample as seen in middle spectrum with standard Cer(16:0;d18:1;16:0). Note, the mixture contains a single peak at 3.7 min supporting further the identity of the synthetic standard and the endogenous compound. 4-day-old mutant mice ( Fig. 5 ). While Cers with a 1-O-very long chain fatty acid ester, e.g., 1-O-lignoceroyl Cers, increased only 3-to 4-fold, 1-O-palmitoyl and 1-O-stearoyl Cers increased 60-and 80-fold, respectively (supplementary Fig.  IV). In the epidermis of GlcCer-S-defi cient mice, 1-O-acylceramides increased to 31% of all esterifi ed Cers (31% of all esterifi ed sphingolipids) or to 16% of all sphingolipids, when comparing data obtained here with the previously published data ( 22 ) on the basis of C18-sphingosine-containing sphingolipids. Hence, Cers not transformed into GlcCer may further on be converted into 1-O-acylceramides.
We compared the quantities of 1-O-acylceramides with those of the corresponding unmodifi ed parent Cers obtained before and after mild saponifi cation, as the increase of parent Cers after saponifi cation should refl ect mild alkaline-sensitive Cer esters. These data had been obtained previously with a MS/MS instrument only ( 22 ). Indeed, the total amount of these parent Cers increased signifi cantly after saponifi cation in both control and GlcCer-S-defi cient epidermis. Furthermore, the difference in amounts refl ecting mild alkaline-sensitive derivates of these Cers correlated quite well the amounts of 1-O-acylceramides determined now by UPLC-MS/MS (supplementary Fig. V). While in wild-type epidermis one out of six parent Cers was found to be 1-O-esterifi ed, in GlcCer-S-defi cient epidermis every second parent Cer was 1-O-acylated.

Epidermal 1-O-acylceramide synthesis occurs in the absence of LPLA 2
LPLA 2 contains transacylation activity and has been shown to produce 1-O-acylceramides, especially using short Quantifi cation reveals 1-O-acylceramides are a minor group of esterifi ed Cers in wild-type mouse epidermis We investigated epidermal lipid extracts from wild-type and mutant mice of a mouse strain with a constitutive cellspecifi c defi ciency of GlcCer-S (gene, Ugcg ) in K14 expressing cells (basal keratinocytes) 4 days after birth. From this strain we had previously published the Cer, glucosylceramide, and SM levels in epidermis at this age. The mutant mice do not synthesize GlcCers in epidermal keratinocytes anymore and have a defi cient water permeability barrier ( 22 ).

Transacylation and 1-O-glucosylation seem to compete for Cer as substrate
Acylation of Cers in 1-O-position makes it impossible for Cer to be transformed into GSLs through either 1-O-galactosylation or 1-O-glucosylation, or into SM. To investigate whether these pathways compete with each other, we analyzed the epidermis of mice defi cient in GlcCer synthesis. In fact, the sum of 1-O-acylceramides increased about 7-fold in

DISCUSSION
The formation of 1-O-acylceramides was fi rst observed in rat brain using radioactively labeled substrates ( 16 ). In the 1990s, an enzyme with Cer 1-O-acyltransferase activity was identifi ed in MDCK cells and mouse brain, spleen, liver, and kidneys ( 17 ). This enzyme (LPLA 2 ) predominantly acts as phospholipase A 2 . In vitro, LPLA 2 has been shown to be most active using short chain Cer (C2-Cer) for 1-O-acylation ( 17,18 ). Recently, a novel Cer 1-O-acylation pathway was discovered in yeast. Here, very long chain Cers (C26-Cer) were 1-O-acylated and the reaction was catalyzed by the diacylglycerol acyltransferases Lro1p and Dga1p ( 19 ). Lecithin:cholesterol acyltransferase (LCAT) and LPLA 2 are human homologs of the yeast Lro1p. Whereas LCAT and LPLA 2 are localized either outside of the cell or within lysosomes, Lro1p acts in the lumen of the ER ( 19,31 ). Also the activity of Dga1p, which is a homolog to the mammalian diacylglycerol O-acyltransferase 2 (DGAT2), is localized in the ER, but it is most prominent in lipid droplets (32)(33)(34). Interestingly, mice defi cient for DGAT2 exhibit not only lipopenia, but also skin barrier abnormalities ( 35 ).
Whereas the previous studies shed light on the biosynthetic pathways of Cer 1-O-acylation in yeast and mammalian cells, an in-depth analysis of the composition of endogenous 1-O-acylceramides and their natural concentrations is missing. Here, we report the complete endogenous structures of almost a hundred 1-O-acylceramides as natural components of murine and human epidermis/ stratum corneum. However, combinatorial calculations for mouse epidermis hint to even more (>200) components eventually present in trace amounts.
Epidermis is one of the richest sources for Cers and displays a high Cer structural diversity ( 8,9 ). Van Smeden et al. ( 9 ) and Wertz and Downing ( 36 ) pointed out the presence of a very hydrophobic series of unknown esterifi ed chain Cers ( 17 ). We investigated the involvement of LPLA 2 in epidermal 1-O-acylceramide synthesis by comparing 1-Oacylceramide levels from extracts of controls and LPLA 2 mutants. At the 5% level, no differences were observed at birth or 4 days after birth, excluding a major role of LPLA 2 in epidermal 1-O-acylceramide synthesis ( Fig. 5 and supplementary Fig. IV).

Loss of neutral glucosylceramidase activity does infl uence individual 1-O-acylceramide levels
GlcCer is synthesized at the cytosolic side of the cis -Golgi ( 24,25 ). Then, it is either used to generate more complex GSLs on the luminal side or it may be degraded back to Cer by neutral glucosylceramidase at the cytosolic side of the Golgi, endoplasmic reticulum (ER), or plasma membrane ( 26,27 ). If neutral glucosylceramidase is located at sites en route to 1-O-acylation of Cers, it could enhance local concentrations of Cers used for 1-O-acylation. Correspondingly, loss of neutral glucosylceramidase then would lead to decreased substrate availability and concomitantly lower 1-O-acylceramide levels. Previously, elevated steady-state GlcCer levels were reported for liver and testis of neutral glucosylceramidase-defi cient mice ( 20,28,29 ), but were also observed in the spleen, kidney, and brain of these mice (unpublished observations). In epidermis, we analyzed GlcCers with a Cer anchor also present in 1-O-acylceramides and observed a 2-to 3-fold increase in mutants. Correspondingly, levels of the more abundant Cers with long (C16-) acyl chains, but not of Cers with the very long (C24-) acyl chains, decreased down to 52% (16:0) and 61% (h16:0) (supplementary Fig. VI). These Cers may be used as substrates for 1-O-acylation. And indeed signifi cantly lower levels of 1-O-acylceramides with N -linked C16-acyl chains were observed in neutral glucosylceramidase-defi cient epidermis ( Fig. 5B ). As longer 1-O-acylceramides were not altered, the overall decrease of 1-O-acylceramides (80% as compared with controls) was not signifi cant at the 5% level ( Fig. 5A ).

1-O-acylceramides are present in human stratum corneum
To investigate whether 1-O-acylceramides also occur in human stratum corneum and/or the surface of human skin, we isolated total lipids from the ventral forearms of three male volunteers by tape stripping. UPLC-coupled tandem mass spectrometric analyses revealed that 1-Oacylceramides are present at 280 ± 70 fmol per pmol of NS-Cers. The latter represent about 7.4% of all Cers in human stratum corneum ( 30 ). In humans, 1-O-lignoceroyl Cer(d18:1; h16:0) was the most prominent peak, followed by 1-O-palmitoyl Cer(d18:1;16:0) and 1-O-palmitoyl Cer(d18:1;h16:0) ( Fig. 6 ). However, if all parent Cers were summed up, palmitoylation was the most prominent Cer modifi cation in the 1-O-position followed by myristoylation and lignoceroylation (supplementary Fig. VII). The total amount of 1-O-acylceramides accounted for 31 ± 6% of the corresponding unmodifi ed parent Cers. might be a lamellar body-lipid droplet contact site. Lamellar bodies fi nally transport -esterifi ed Cers and GlcCers to the interface of stratum granulosum and stratum corneum where they are released to form extracellular lipid lamellae and the cornifi ed lipid envelope ( 11 ). However, the cellular origin of epidermal 1-O-acylceramides still has to be established; and sebaceous glands, a rich source of unpolar lipids ( 39 ), may not be excluded.
Interestingly, defi ciency of the cytosolically oriented neutral glucosylceramidase, which is tightly attached to membranes of the Golgi and the ER ( 26 ) or the plasma membrane ( 27 ), was accompanied by a decrease of those 1-O-acylceramides for which we also observed a decrease of the corresponding precursor (parent) Cers, suggesting neutral glucosylceramidase activity to be topologically related and upstream of 1-O-acylceramide production.
As pointed out earlier ( 19 ), 1-O-acylceramides should not support lipid bilayer formation. However, 1-O-acylceramides are also more polar than triglycerides, wax esters, or cholesterol esters. 1-O-acylceramides contain a free hydroxyl group, which allows them to undergo oriented hydrogen bonding interactions . Due to their predicted physicochemical properties, 1-O-acylceramides might constitute building blocks of extracellular lipid lamellae and may contribute to their stability and correct lamella phase organization. Lamella phase organization has been shown to be disturbed in patients with atopic eczema ( 40 ). In spite of their relatively low abundance (5% of all esterifi ed Cers, or 2.3% of all Cers, which account roughly for about half of the total stratum corneum lipid species by weight) ( 41 ). 1-O-acylceramides could contribute to a functional water permeability barrier as they are among the most hydrophobic epidermal Cers . However, this has to be clarifi ed in future investigations.